Supercapacitors
Scientific Fundamentals and
Technological Applications
Electrochemical
Supercapacitors
Scientific Fundamentals and
Technological Applications
8. E. Conway
Fellow of the Royal Society of Canada
University of Ottawa
Ottawa, Ontario, Canada
Conway, B. E.
ElectrochemIcal supercapacitors , scientific fundamentals and
technological applications I B.E. Conway.
cm.
p.
Includes bibliographical references and index.
1. Storage batteries. 2. Electrolytic capacitors.
double layer.
1. Title.
TK2941.C66 1999
621.312424--dc21
3. Electric
98-48209
CIP
ISBN 978-1-4757-3060-9
ISBN 978-1-4757-3058-6 (eBook)
DOI 10.1007/978-1-4757-3058-6
10987654321
A C.I.P. record for this book is available from the Library of Congress.
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted in
any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
Foreword
The first model for the distribution of ions near the surface of a metal electrode
was devised by Helmholtz in 1874. He envisaged two parallel sheets of charges
of opposite sign located one on the metal surface and the other on the solution
side, a few nanometers away, exactly as in the case of a parallel plate capacitor.
The rigidity of such a model was allowed for by Gouy and Chapman independently, by considering that ions in solution are subject to thermal motion so
that their distribution from the metal surface turns out diffuse. Stern recognized
that ions in solution do not behave as point charges as in the Gouy-Chapman
treatment, and let the center of the ion charges reside at some distance from the
metal surface while the distribution was still governed by the Gouy-Chapman
view. Finally, in 1947, D. C. Grahame transferred the knowledge of the structure of electrolyte solutions into the model of a metal/solution interface, by envisaging different planes of closest approach to the electrode surface depending
on whether an ion is solvated or interacts directly with the solid wall. Thus, the
Gouy-Chapman-Stern-Grahame model of the so-called electrical double
layer was born, a model that is still qualitatively accepted, although theoreticians have introduced a number of new parameters of which people were not
aware 50 years ago.
Irrespective of the structural details, it has long been accepted that a double
layer exists at the electrode/electrolyte solution boundary, which governs adsorption phenomena and influences charge transfer reaction rates, and where
electrostatic energy is stored as in a capacitor a few molecular diameters thick.
Nevertheless, the existence of a double layer has always been inferred from indirect observations of related properties and quantities, but never directly
vii
viii
Foreword
probed, so much that it was compared to the Arab Phoenix: "Everybody says it
exists, nobody knows where it is." This until recently, when it was realized that
the energy stored per unit surface area of an electrode is noticeable per se and
becomes technologically very interesting with the introduction of new materials
with an exceptionally extended active surface: especially treated carbons, some
transition metal oxides, electrosynthesized conducting polymers. The interfacial capacity is further increased if the purely capacitive charge is supplemented by a Faradaic charge related to bidimensional redox reactions or
tridimensional intercalation processes.
"Supercapacitors" are devices that store electrical energy on the basis of
the above phenomena and that can be discharged at a much higher rate than conventional batteries. They have aroused interest for various applications, including electric vehicles, in particular cars as well as trains. I should say that in spite
of our awareness of the principles, supercapacitors have appeared on the scientific scene rather suddenly, or at least this has been the impression ofthose who
have realized that something was happening at the technological level. Of the
many examples we can produce of innovations developed in technology first
and then "discovered" from a fundamental point of view, supercapacitors furnish an authoritative example ofthe reverse: a technological innovation pushed
by fundamental knOWledge.
The situation is now that fundamental researchers know everything of the
electrical double layer but ignore its application to supercapacitors, while engineers know of supercapacitors but may ignore the fundamentals of their operation. This monograph comes at an opportune time to fill this gap, with a
balanced presentation of fundamentals aimed at applications, and applications
related to fundamental principles. B. E. Conway has worked for more than 50
years in almost all areas of electrochemistry, particularly interfacial electrochemistry. He is therefore a "veteran" in the field, being the first to realize the
potentialities of some materials for their double layer energy.
This volume offers what cannot be found in any other work for its comprehensiveness, exhaustiveness, and focus. For the first time a highly theoretical
topic, the electrical double layer at electrodes, is shown to manifest itself in
highly technological applications. It is with a real sense of pride that electrochemists in the near future will press the accelerator in their electric car knowing
that certain performances are possible only thanks to the discharge of the "socalled" (but is it indeed there?) electrical double layer of which technologists
have long maintained "electrochemistry can do without it."
The content of this book is useful both for scientists working in fundamental research and technologists, in particular those interested in electrochemical
energy conversion and chemistry and physics of electrified interfaces, as well
as for engineers working in the field of electrochemical power sources and electrical energy storing devices. They will find the book an invaluable source of in-
Foreword
ix
formation and inspiration. For the way the topics are presented, people working
in the area of materials chemistry and physics will find this book of great general
interest in view of the typical dependence of the performance of supercapacitors
on the structure of materials.
Milan, Italy
Sergio Trasatti
Preface
Systems for electrochemical energy production originated with Volta's discovery in 1800 of "voltaic electricity" and were developed in various forms during
the nineteenth century. Toward the end of that period, reversibly chargeable batteries for electrical energy storage and utilization became a major development
in applied electrochemistry and during the present century have been improved
to a high state of the art. They also represent a large fraction of the economic
activity in industrial electrochemistry.
In relatively recent years, but originating with Becker's patent in 1957, a
new type of electrochemically reversible energy storage system has been developed that uses the capacitance associated with charging and discharging of the
double layer at electrode interfaces or, complementarily, the pseudocapacitance
associated with electrosorption processes or surface redox reactions. In the first
case, large interfacial capacities of many tens of farads per gram of active electrode material can be achieved at high-area carbon powders, fibers, or felts,
while, in the second case, large pseudocapacitances can be developed at certain
high-area oxides or conducting polymers where extents of Faradaic charge (Q)
transfer are functionally related to the potential of the electrode (V), giving rise
to a derivative corresponding to a capacitance dQ/dV.
These large specific-value capacitors, especially of the double-layer type,
are perceived as electrical energy storage systems that can offer high power-density in discharge and recharge, and cycle lives on the order of 1 to 106, many times
those of conventional batteries. A variety of uses of such electrochemical or socalled "supercapacitors" are now recognized and a new direction of power-source
development, complementary to that of batteries, is well established.
as
xi
xii
Preface
An important aspect of this monograph is that it gives a comprehensi ve account of the electrochemical science and technology of these capacitor systems.
An attempt is made to present a self-contained and unified treatment of the field,
including essential details of the background science (e.g. of double-layer capacitance and the origins of pseudocapacitance, the electrolyte solutions used in
electrochemical capacitors) as well as basic concepts of electrode kinetics and
interfacial electrochemistry, dielectric polarization theory, porous electrodes,
and conducting polymer materials that give rise to large specific capacitances.
In this way, understanding and study of the material presented in this volume
will not require frequent reference to other textbooks of physical. chemistry or
electrochemistry .
The text contains many illustrative diagrams and cross-references between
chapters, and includes many literature references. For the convenience of the
reader, three or four diagrams have been duplicated from one chapter or another
to avoid the necessity of seeking earlier or later pages in the volume where
cross-referenced material is cited.
The author's work in this field originated with a research contract between
Continental Group Inc. and the University of Ottawa's Electrochemistry Group.
We would like to acknowledge here the work carried out by Drs. H. AngersteinKozlowska, V. Birss, J. Wojtowicz, and Visiting Professor S. Hadzi-lordanov
(University of Skopje) with Mr. Dwight Craig (electrical engineer) of Continental Group in the period 1975 to 1981. More recently, new work in this field is
being carried out at the University of Ottawa and is supported by the Natural
Sciences and Engineering Research Council of Canada. For this work, acknowledgment is made to Dr. W. 1. Pell and Mr. T. C. Liu.
Special thanks are due to Dr. B. V. Tilak of Occidental Chemical Corp.,
N.Y., for his critical reading of the manuscript before its submission for publication, and for his suggestions for additions and revisions. Appreciation is expressed to Dr. Tilak and Dr. S. Sarangapani (ICET Inc., Norwood, Mass.) for
their detailed examination of Chapter 20 on technology development, and in
particular for their suggestions for the best systematic organization of the manifold aspects of the subject treated in that chapter. Thanks are also due to Drs. S.
GoUesfeld (Los Alamos National Laboratory) and J. Miller (1. M. Inc., Shaker
Heights, Ohio) for reading the chapters on conducting-polymer capacitors and
ac impedance, respectively. We are grateful to Dr. Miller for permission to reproduce some of his computer-generated graphs and data on ac impedance
evaluation of capacitors.
The author is also most grateful to Drs. S. P. Wolsky and N. Marincic for
their permission to draw on various diagrams and tables from the proceedings
of papers presented at the seminars on electrochemical capacitors held at Deerfield Beach and Boca Raton, Fla, over the period 1991 to 1997, under the auspices of Florida Educational Seminars Inc. (abbreviated as FES in the text).
Preface
xiii
Finally, special thanks are due to Denise Angel, who typed, with great efficiency and accuracy, all the chapters of this volume in several drafts, exercising literacy and care that would be difficult to match. Grateful thanks are also
due to Eva Szabo for drafting most of the diagrams.
Ottawa, Canada
B. E. Conway
Contents
Chapter 1
Chapter 2
1
8
9
11
11
12
13
14
14
15
15
17
18
18
19
20
xvi
Contents
Chapter 4
22
25
28
29
31
31
33
34
37
41
41
45
46
48
50
51
53
59
59
61
63
64
64
67
68
Contents
xvii
4.2.1.
4.3.
4.4.
4.5.
4.6.
4.7.
4.8.
4.9.
4.10.
4.11.
4.12.
Chapter 5
68
68
70
71
72
73
74
74
75
76
77
78
78
79
79
79
80
81
81
81
82
83
83
86
86
87
88
91
92
Contents
xviii
5.5.
5.6.
5.7.
5.8.
5.9.
Chapter 6
Chapter 7
105
108
114
116
117
119
121
123
124
125
127
129
133
133
135
136
136
137
Contents
xix
7.5.3.
Chapter 9
138
139
141
141
143
148
149
151
152
156
156
157
160
161
165
168
169
170
176
180
180
xx
Contents
186
193
196
198
199
203
203
208
209
212
213
217
219
220
221
224
224
224
236
236
239
241
243
246
Contents
xxi
248
253
255
255
256
257
259
264
265
267
270
276
277
277
279
282
284
285
286
290
292
293
xxii
Contents
299
304
312
314
320
322
327
331
332
334
335
337
338
343
344
345
345
347
350
351
360
Contents
xxiii
361
362
365
366
366
366
366
373
374
375
377
380
383
383
383
403
405
406
408
411
415
416
417
421
421
425
427
433
xxiv
Contents
436
440
443
443
444
448
448
452
456
459
462
463
464
468
474
475
479
486
486
489
491
491
496
497
497
501
502
Contents
xxv
502
503
504
505
506
506
509
510
513
513
513
515
518
524
525
528
528
530
532
538
547
549
555
556
xxvi
Chapter 18
Contents
557
557
559
561
562
567
568
569
573
574
575
578
579
582
582
583
584
586
586
587
591
592
595
595
Contents
xxvii
597
598
599
600
600
602
602
602
603
605
605
605
606
606
609
610
610
612
613
613
615
615
618
618
618
620
620
622
623
625
627
Contents
xxviii
20.6.
20.7.
20.8.
20.9.
20.10.
20.11.
20.12.
20.13.
20.14.
20.15.
20.16.
20.17.
20.18.
Chapter 21
627
634
635
641
643
644
644
648
649
649
651
655
658
663
666
666
667
667
668
670
671
673
674
Electrochemical
Supercapacitors
Scientific Fundamentals and
Technological Applications
Chapter 1
Chapter 1
Although the experimental phenomenology of electrical charging of surfaces, including those of the Leyden jar condenser, was well understood in the
mid- and later part of the eighteenth century, the full physical significance of
charging or discharging processes at the plates of capacitors could not have been
at all fully understood until the atomic nature of electricity, the electron, was
characterized and its properties directly determined. 1,3,4 Similarly, the charging
or discharging of capacitors by a flow of electrical charges in wires could not
have been understood until the electron theory of metals had been developed
and the mechanism of current flow in metal conductors was understood. Nor for
that matter could the physical and chemical significance of the charging of amber be understood until satisfactory ideas about the ionization of molecules and
macromolecules (in that case, through frictional or triboelectrical effects) had
been formulated, partly through the results on spectroscopic ionization limits of
electron energy states in molecules or atoms. Thus, the mechanisms of electrical
charge storage in capacitors remained poorly understood at the atomic physical
level until some 140 years after the development of the Leyden jar capacitor and
related contemporary electrical machines. Nevertheless, it was Faraday who
had some of the first (in principle) correct ideas about polarization in dielectrics
and the significance of dielectric strain and lines of (electrical) force in the dielectric materials of charged condensers.
At this point it should be stated that the charging of metallic plates of a capacitor involves production of an excess (negative plate) or deficiency (positive
plate) of the density of the delocalized electron plasma of the metal over a short
distance (ca. 0.1-0.2 nm, the Thomas-Fermi screening distance) from the formal outer surface of the metal plate. However, each plate has its own electric
potential (an equipotential) throughout its material, except very near its surface.
Hence, local charge density variation within the so-called "Thomas-Fermi
screening distance" has to arise according to the Poisson relation that expresses
the field gradient in terms of the local space charge density and the Gauss relation that expresses the field as a function of surface charge density (see Chapter
4).
At an insulator such as amber, the excess charge density that arises upon
charging has a different origin that is associated with localized molecular ionization (localized oxidation) of the insulator material at its surface, or in some
cases is due to negative ionization by localized uptake of electrons at electron
acceptor sites on the surface (localized reduction). These latter phenomena are
the subject of "triboelectricity."
The principle that electrical energy can be stored in a charged capacitor
was known since 1745; at a voltage difference, V, established between the plates
accommodating charges +q and -q, the stored energy, G, is 112 CV 2 or 112 qV,
G being a Gibbs (free) energy which increases as the square ofV.
Chapter 1
~I
FIGURE 1.1. Photographs of various designs and arrangements of condensers (capacitor devices), including first the Leyden jar. (Reproduced from
Encyclopaedia Britannica, 1929 edition, with permission). (1) A Leyden jar, the simplest type of condenser. (2) A battery of Moscicki glass condenser
tubes-an improvement on the Leyden jar, introduced about 1904 by Moscicki. (3) The transmitter at one of the stations of the British Broadcasting
Corp., showing two air dielectric condensers in the lower compartment (center and right-hand side). (4) Ship's radio transmitter, showing mica dielectric
condenser on right side on shelf over generator. (5) Mica dielectric condenser of special construction for shortwave radio stations. (6) Condenser gallery
of the Rugby radio station of the British post office. (7) Interior of the bell box used with British office telephone instrument, showing paper dielectric
condenser (bottom of box). (8) Top row: several patterns of fixed condensers. Bottom row: several types of variable condensers commonly used as tuning
capacitors in old radios.
"i.
\r ~.!~
,~~
en
CD
Ic'
'tI
e!.
C'j'
~...
:I:
a.
III
:J
o:J
c:
!l
5"
Chapter 1
The utilization of this principle to store electrical energy for practical purposes, as in a cell or battery of cells seems to have been first proposed and
claimed as an original development in the patent granted to Becker in 1957.8
The patent described electrical energy storage by means of the charge held in
the interfacial double layer at a porous carbon material perfused with an aqueous
electrolyte. The principle involved was charging of the capacitance, Cd!> of the
double layer, which arises at all solid/electrolyte interfaces, such as metal, semiconductor, and colloid surfaces (and also at the phase boundary between two
immiscible electrolyte solutions9 ).
Carbon is an element almost uniquely suited for fabrication of electrochemical capacitors of the double-layer type. It exists in several, well-known allotropic forms-diamond, the fullerenes, and graphite; the latter and glassy
carbon can be generated in the form of high-area fibers or felts. Amorphous carbons and carbon black are available as high specific-area powders. The fiber or
felt materials are particularly convenient for formation of electrode structures
having good mechanical integrity, while the high-area powders are more difficult to handle. However, glassy carbon, graphite, and carbon black materials are
convenient for forming high-area electrode structures, often on a support matrix. From an electrochemical point of view, carbon is relatively, though not entirely, unreactive and thus has a potential voltage range of almost ideal
polarizability (Chapters 6 and 7), approaching 1.0 V in aqueous solution and
possibly up to 3.5 V in nonaqueous media.
After Becker, the Sohio Corporation in Cleveland, Ohio, also utilized the
double-layer capacitance of high-area carbon materials, but in a nonaqueous
solvent containing a dissolved tetraalkylarnmonium salt electrolyte. Such systems provide higher operating voltages (3.4-4.0 V) owing to the greater decomposition voltage of nonaqueous electrolytes than those for aqueous ones. Thus
they can accommodate higher charge densities and provide larger specific energy storage since the storable energy increases with the square of the voltage
attainable on charge.
A different principle, originating from ref. 14, was utilized and developed
in 1975 onward to 1981 by ConwaylO in Ottawa, under contract with Continental Group. This was based on the concept of D. Craig that was developed at
Hooker Corp. Here, in one type of system, the large, so-called "pseudocapacitance," CIP' associated with the potential dependence of extents of electrochemical adsorption of H or monolayer levels of electrodeposition of some base
metals (Pb, Bi, Cu) at Pt or Au was used ll as a basis for an energy-storing capacitor. In another type of system, the pseudocapacitance associated with solid
oxide redox systems was used, especially that developed over some 1.4 V (practical range 1.2 V) in aqueous H 2S04 at RU02 filmsy-15 This system approaches
almost ideal capacitative behavior, with a large degree of reversibility between
charge and discharge, and multiple cyclability over some lOS cycles (see Chapter 11).
Work on the latter type of system has been continued by Pinnacle Research
Corp., Cupertino (now at Los Gatos, California), in the former laboratories of
Continental Group. Useful military applications have been developed, but the
Ru materials required are too expensive for the development of a large-scale capacitor, e.g., for use in hybrid systems with batteries for electric-vehicle motive power.
Pseudocapacitance arises whenever, for thermodynamic reasons, there is
some continuous dependence of a charge, q, passed Faradaically in oxidation or
reduction, upon the electrode potential, V. Thus, a derivative dq/dV can arise
that corresponds to a pseudocapacitance that is directly measurable, or
utilizable, as a capacitance (Chapter 10).
The large capacitances (on the order of several or more farads per gram)
that can be developed with the RuOz film system and also with the carbon
double-layer-type capacitors led to the terms "supercapacitor" or "ultracapacitor" being coined, respectively, for these two types of high specific capacitance devices. Recently it has been suggested that the more general term
"electrochemical capacitors" be used to refer to these systems. However,
this name should not be confused with "electrolytic capacitor," the latter
term referring to the well-known, moderately high-capacitance device (on the
order of tens of millifarads) that is based on a thin-film oxide dielectric formed
electrolytically with a gel electrolyte on such metals as Ta, Zr, Ti, or AI.
The key practical factor that allows very high capacitances, on the order of
farads or tens of farads rather than millifarads, to be achieved in a small volume,
say 1 cm3, is the utilization of high-area materials such as activated porous carbons for which real areas are up to 1000 to 2000 m 2 g-l. Similarly, with the
Ru02 pseudocapacitance system, the material, which is a hydrous oxide, has a
quasi-3-dimensional, electronically conducting structure, giving accessibility to
protons and electrons 15 that are involved in two or three (Chapter 11) successive, reversible oxidation or reduction steps in charge or discharge, respectiVely.
The continuous dependence of the extents of oxidation or reduction on electrode
potential (over 1.4 V), with corresponding passage of charge, leads to the high
specific redox pseudocapacitance of this material,16 which is usually coupled
with an appreciable double-layer capacitance component (see Chapter 11).
The use of high-area carbon or oxide redox systems has led to the commercial production of practical high-capacitance electrochemical capacitor devices
such as that developed (the Gold Capacitor) by Matsushita Electric Industrial
Co. (Osaka, Japan) and by Pinnacle Research; the latter has been developed
mainly for military applications. The commercial products are designed to provide standby power for random access memory (RAM) devices or telephone
equipment, as power sources for operating activators, and as elements for long
time-constant circuits, etc. An attractive technology employing Ru02 in a thin
film applied to a Nafion membrane, or a powder treated with Nafion, has been
developed by Giner, Inc. (Waltham, Massachusetts) and gives high specific ca-
Chapter 1
1~.
SCOPEOFTHEMONOGRAPH
REFERENCES
1. 1. J. Thomson, Phil. Mag., 5,346 (1903); see also J. J. Thomson, The Electron in Chemistry,
Franklin Institute Lectures, Chapman and Hall, London (1923).
2. W. Crookes (1879), quoted by S. Glasstone in Textbook of Physical Chemistry, Van Nostrand,
New York (1940).
3. R. A. Millikan, Phys. Rev., 2, 143 (1913).
4. 1. S. Townsend, Electricity in Gases (1879), quoted by Glasstone as in Ref. 2.
5. G. Johnstone Stoney, Phil. Mag.,ll, 381 (1881); Sci. Trans. Roy. Soc. Dublin, 4,583 (1891).
6. H. von Helmholtz, 1. Chern. Soc., Lond., 39, 277 (1881).
7. B. E. Conway, Ionic Hydration in Chemistry and Biophysics, Elsevier, Amsterdam (1981).
8. H. E. Becker, U.S. patent 2,800,616 (to General Electric Co.) (1957).
9. Z. Samec, 1. Electroanal. Chern., 103, I (1979).
10. B. E. Conway, 1. Electrochem. Soc., 138, 1539 (1991).
11. B. E. Conway and H. A. Kozlowska, Acct. Chern. Res., 14,49 (1981).
12. S. Hadzi-Jordanov, H. A. Kozlowska, M. Vukovic, and B. E. Conway, 1. Electrochem. Soc.,
125, 1471 (1978).
13. S. Hadzi-lordanov, B. E. Conway, and H. A. Kozlowska, 1. Electroanal. Chern., 60, 359
(1975).
14. S. Trasatti and G. Buzzanca, 1. Electroanal. Chern., 29, App. 1 (1971).
15. R. Galizzioli, F. Tantardini, and S. Trasatti, 1. Appl. Electrochem., 4, 57 (1974).
16. B. E. Conway, in Proc. Symp. on Electrochemical Capacitors, F. M. Delnick and M.
Tomkiewicz, eds., vol. 95-29, 15, Electrochem. Society, Pennington, N.J. (1996).
Chapter 2
2.1. INTRODUCTION
A modern technological society demands the use and storage of energy on a major scale, employing large and small systems for that purpose. Energy stored as
potential energy is involved in hydroelectric systems through the hydrostatic
"head" of water behind dams; it is also stored in a potential sense in fuels (e.g.,
coal, oil, and cryogenic hydrogen) and becomes available, albeit with rather
poor efficiency, through combustion utilizing steam-piston, steam-turbine, and
internal combustion engines of various kinds as energy transduction devices.
Energy may also be stored as rotational kinetic energy in flywheels.
Electrical energy can be stored in two fundamentally different ways: (1) indirectly in batteries as potentially available chemical energy requiring Faradaic
oxidation and reduction (see Sections 2.4.2 and 2.5) ofthe electrochemically active reagents to release charges that can perform electrical work when they flow
between two electrodes having different electrode potentials (i.e., across the
voltage difference between the poles of battery cells); and (2) directly, in an
electrostatic way, as negative and positive electric charges on the plates of a capacitor, a process known as non-Faradaic electrical energy storage. The efficiency of these two modes of storing electrical energy is usually substantially
11
12
Chapter 2
larger than that of fuel combustion systems, which are limited by thermodynamic Carnot cycle considerations while electrochemical systems usually involve more reversible processes, with direct conversion of potentially available
chemical energy to free (or Gibbs) energy, tJ.G.
2.1.2. Modes of Electrical Energy Storage by Capacitors and Batteries
13
ter). This is a major and characteristic difference between battery and capacitor
electrical energy storage systems.
It must be emphasized at the outset that there has never been an aim or projection of a possible substitution of batteries by supercapacitors; rather, opportunities arise for complementary operation of electrochemical capacitors that
are electrically coupled in discharge and recharge with batteries, while other
kinds of applications especially favor capacitor-type behavior, e.g., for backup
power systems. Also, there are stand-alone opportunities for using multiply rechargeable electrochemical capacitors in a variety of independent functions, as
briefly indicated in Chapter 1. General aspects of the electrochemistry and technology of batteries are to be found, for example, in Refs. 1 and 2.
In the early stages of the development of electrochemical capacitor technology and related fundamental work, there was some confusion in the electrochemical engineering field and at symposia about the differences between the
properties and operation of a battery and a supercapacitor, and what advantages
one might have over the other. This confusion may have been assisted by some
groups calling the capacitor devices "ultracapacitors" and others, "supercapacitors," the latter as originated by the Ottawa group in 1975. The present preferred
name, proposed by Burke, as referred to in Chapter 1, is now more scientific and
generic, namely "electrochemical capacitors."
This chapter identifies and explains some of the similarities and differences between electrochemical capacitors and batteries in relation to the electrochemical processes that are involved in their discharge and recharge cycling,
and in their potential uses as electrical energy storage devices. In particular, the
fundamentally different mechanisms of charge storage that are normally involved will be emphasized, along with the consequent, usually different, relations between the extents of charge accommodated at the electrodes and the
electric potential differences (cell voltage) between pairs of electrodes having
conjugate, , polarities.
One of the main and kinetically significant differences between capacitors
and batteries is that the electrodes of the latter usually undergo substantial phase
changes during discharge and recharge (minimally though for the intercalation
systems), which lead to kinetic and thermodynamic irreversibility. On the other
hand, capacitors of the double-layer type require only electrostatic charge accommodation with virtually no phase change, though some small but significant
reversible electrostriction of the electrolyte can arise upon charging.
14
Chapter 2
(2.1)
Such a reaction at the surface of an electrode M usually gives rise to a potentialdependent pseudocapacitance (see again Chapter 10) and the quantity Oe is related to the so-called "electrosorption valence" treated in Refs. 3, 4, and 5.
To summarize, the important differences in the charge storage processes
are as follows:
2.2.1. Non-Faradaic
The charge storage is achieved by an electron transfer that produces chemicalor oxidation state changes in the electroactive materials according to Faraday's laws (hence the term) related to electrode potential. Pseudocapacitance
can arise in some cases. The energy storage is indirect and is analogous to that
in a battery.
In a battery cell, every electron charge is Faradaically neutralized by
charge transfer, resulting in a change of oxidation stage of some redoxelectroactive reagent, e.g.,
NP+02-0H- + e + H+ ~ NF+20H-
(2.2)
15
Table 2.1 contains a summary of the types of capacitors and their mode of
energy storage: electrostatic or Faradaic, the latter in the case when pseudocapacitance arises (Chapter 10). Types d and f, which are printed in bold, are the
principal kinds treated in this volume.
Normally, capacitors function as elements of electronic circuits or communications equipment, or as ballast for starting electric motors or electric discharge
TABLE 2.1. Types of Capacitors and Mode of Energy Storage
Type
(a) Vacuum
(b) Dielectric
(c) Oxide electrolytic (thin film)
(d) Double-layer
Examples
Mica, Mylar, paper
Ta205, Al 20 3
C preparations,
powders, fibers
Undeveloped
polyaniline,
polythiophenes
Fe(CN)~--Fe(CN)~-,
V 2+N 3+N0 2+
16
Chapter 2
tubes (fluorescent lights). As has been explained earlier, devices of very large
capacitance are now available for storing electric energy in various applications.
Table 2.2 summarizes the types of batteries currently extant. 1,2 These are
generally classified as primary (nonrechargeable) or secondary (multiply rechargeable). The discharge or recharge mechanism is mainly Faradaic, although
all electrode interfaces exhibit a double-layer capacitance that is reversibly
chargeable. For batteries the latter mechanism accommodates about 2-5% of
the total charge accepted.
In a different class from the battery systems listed in Table 2.2 are fuel cells
in which the anode and cathode (02 or air) reactants are supplied on a continuous basis from external reservoirs, and the electrode surfaces provide an interface for either electrocatalytic oxidation or reduction of the reagents supplied.
The primary metal-air cells are operated as semi-fuel cells, but the "fuel" is an
easily oxidizable base metal and a gas-diffusion catalyzed air or O 2 cathode is
employed. Such cells using Al are not rechargeable except by mechanical replacement of the metal anodes. However, if Zn is used, electrochemical recharging is possible, but requires a bifunctional catalyzed counter electrode capable
of evolving H2 on recharge or reducing O2 (air) on discharge.
Faradaic
Faradaic
Faradaic (exhibiting
intercalative pseudocapacitance)
17
Electrochemical capacitors
Carbon (double-layer capacitors), H on Pt
(UPD)
Underpotential-deposited Pb on Au
Ru02
lr0 2
Modified cheaper transition metal oxides,
metal nitrides
H in Pd, LaNi s(?)
Conducting polymer electrodes, e.g.,
polyanilines, polythiophenes
Hg-HgO
Li-SOCI 2
Li-S02
Li-CFx
Intermediate transitional systems:
Li-TiS2; Li-MoS2;
Li-Mo02; Li-Mn02;
Li-C002; Li-C
As discussed and illustrated in Chapters 16 and 17, an electrochemical capacitor, like a battery cell, requires two electrodes, one of which is charged
negatively with respect to the other, the charge storage and separation being
electrostatic. At each ofthe two electrodes, double-layer electron and ion charge
separations are established across the electrode interfaces. The macroequivalent
circuit (in the absence of self-discharge processes) is represented by two capacitances linked in series with ohmic resistances representing the resistivities of the
solution and the separator, as in a battery. In the usual case where the electrodes
are high-area, porous matrices, a further microequivalent circuit is needed (see
Chapter 2
18
Chapters 14 and 16) to represent the electrical behavior of the internal surfaces
and the electrolyte-containing interstices.
As with batteries, bipolar electrode configurations can be fabricated for
higher voltage series combinations; these diminish the internal resistance of the
device, but require that the edges of the electrodes be carefully sealed to the case
in order to avoid shunt, i.e., leakage currents. Such systems optimize power density.
The equivalent circuit for most battery-type charge storage systems involves simply a Faradaic resistance (R F ) that represents the potential dependence of the reciprocal of the rate of the oxidation and reduction charge transfer
process; it is in parallel with (an always significant) double-layer capacitance,
Cdl Under some high-rate discharge or recharge conditions, some diffusion
control may arise, in which case RF is in series with a so-called Warburg C-R
impedance element written as W. In addition, with some rechargeable battery
systems (e.g., of the intercalation type), a pseudocapacitance element may also
be required to represent the impedance behavior of the Faradaic process (Chapters 10 and 17).
For both electrochemical capacitors and battery cells, a solution resistance
element, Rs' in series with the Faradaic impedance, ZF, is usually necessary in
order to fully represent the charging or discharging behavior. In fact, Rs is usually very important in the evaluation and performance of capacitors and batteries for high-rate discharge applications, and is an important influence on the ac
impedance spectrum (see Chapters 16 and 17) of the device.
In the absence of self-discharge processes, or any parallel pseudocapacitance, the macroequivalent circuit of an electrochemical capacitor involves only
a solution resistance (Rs) and a double-layer capacitance Cd1 However, Rs and
Cd1 have distributed components in the microequivalent circuit in the case of
high-area, porous electrode materials, as treated in Chapter 14. The same applies
to battery electrodes fabricated in porous material configurations, a procedure
and design that are commonly adopted in battery technology; then complex microequivalent circuits also apply, but with the inclusion of the essential RF components.
19
face, it is easily seen that the charge accommodated per atom will be 3011 0 15 f..lc.
This is equivalent to 30 x (10- 61105) x 6 x 1023110 15 electrons per atom where
105 and 6 x 1023 are taken approximately as the Faraday constant and
Avogadro's number, respectively. The above quotient works out to be 0.18
electrons per atom stored as a double-layer charge at 1 V for the example
chosen, but this is the average delocalized charge distribution associated
with conduction-band electrons.
By contrast, in most battery processes, redox reactions involving usually
one or two valence electron charges per atom (sometimes three for Al or Bi) or
molecule of electroactive reactant are involved. Thus, electrochemical capacitor
charge storage involves, per atom of active available surface area, only about 20
or 10% (respectively) of that involved (indirectly) or available with battery redox materials. Hence the energy densities available with capacitors are usually
substantially smaller than those with batteries. This is a well-recognized limitation, but it is usually compensated by much larger cyclability (cycle life) and
often better power density attainable from electrochemical capacitors. Of
course, because the energy density of a capacitor increases with the square of
the voltage on charge, a substantial improvement in energy density is attained
by using nonaqueous electrolytes instead of aqueous ones, in which the decomposition voltage can be increased to 3 or 3.5 V or more per cell unit.
Elsewhere in this volume (Chapter 14) it will be shown that with large-area
carbon materials (e.g., 1000 m2 g-I), very large specific capacitances of 300 F
g-I can (theoretically) arise. It is then of interest to compare energy densities attainable at double-layer capacitors with those realized in an Ni-Cd battery.
For a capacitor electrode of 1000 m2 g-1 operating at 1 V and having a specific double-layer capacitance of 30 f..lF cm- 2 (e.g., for carbon blacks), the total
capacitance is 300 F g-l, as mentioned earlier. At 1 V, the energy, G, stored is
G = (112) CV2 = 112 x 300 x 12 = 150 Wsec g-1 or J g-1 or 150 kJ kg-I, theoretically. Equivalently, in watt-hours (1 Wh =3600 J), the above energy density
is about 42 Wh kg-I. (In practice, it will usually be substantially less than this
figure owing to the inaccessibility of the electrolyte solution to the fine pores of
the porous electrode structure in the case of carbon powders or felts, and the
weight of the packaging structure and electrolyte.)
A comparison with an Ni-Cd battery can be made as follows: The molecular weight of NiOOH is 92 and the equivalent weight of Cd is 112/2, i.e., 56.
The electron number is 2 for Cd (1 for the equivalent weight of 56) and approximately 1 for Ni OOH being reduced to Ni(OH)2. (The oxidation state of Ni in
20
Chapter 2
charged nickel oxide electrodes can approach higher values at low temperatures
and in strong aqueous KOH.)
For a working voltage of 1.2, the ideal energy density (~G = -zFE) will be
1.2 x 1 x 105/148 J g-I == 8.1 X 105 J kg-I = 810 kJ kg-I where 105 is approximately F, the Faraday constant in coulombs mole-I. With I Wh = 3600 J, the
(theoretical) energy density for Ni-Cd (electroactive materials alone) is then
225 Wh kg-I. Thus an electrochemical double-layer capacitor electrode
(charged to I V working potential) would have about 20% of the energy density
of an Ni-Cd battery electrode, both figures being based on ideal, theoretical performance.
A double-layer capacitor based on nonaqueous electrolyte technology
would have a substantially larger energy density than that for I-V aqueous solution charging. In an actual two-electrode capacitor device, with one electrode
worked against the other, the energy density will be about (112)2 smaller,
namely, about 10 Wh kg-I in aqueous electrolyte, since each electrode can be
discharged down to only about half its initial voltage following charging.
"
21
-----~-------------------.-----,
....... - .. __ ..... --
Recharge
Discharge
'.
'\: IDEAL
\1
BATIERY
..J
IZ
__ IDEAL CAPACITOR
I-
oa.
,.
I
I
: State of
: charge indication
I
I
-41---
FIGURE 2.1. Difference of discharge and recharge relationships for a capacitor and a battery: potential as a function of state of charge, Q.
rather than Qll2 M (or 1I2Y). This difference can be illustrated by the discharge curves shown schematically in Fig. 2.1 where the voltage on the capacitor declines linearly (for potential-independent Cdl ) with the extent of charge,
while that for an ideal battery remains constant* as long as two phases remain
and are in equilibrium (upper curves in Fig. 2.1).
The decline of the capacitor voltage arises formally and phenomenologically since C = QIVor V = QIC; therefore dVldQ = lie. The ideal battery cell
voltages on discharge and recharge, as a function of state of charge, are shown
as parallel lines of zero slope4 in the upper parts of the diagram. These two lines
differ as a result of any cathodic and anodic polarization (including so-called
ohmic IR potential drop due to internal or solution resistance) arising in the discharge or recharge half-cycles. In the sloping discharge and recharge line for the
capacitor (Fig. 2.1), there will also be significant IR drop, depending on the discharge or recharge rates, that is, the discharge line will actually be somewhat
separated from the recharge line by a voltage difference equal to 2IR.
The significance of the difference between the energy density for the capacitor and a battery cell charged to the same potential can be illustrated by ref-
*In practice, there is usually some decline ofvoJtage in a battery with increasing extent of discharge.
In certain battery systems e.g., LirriSz , Li/CoOz, and metal hydride anodes, this decline is larger
and occurs for fundamental thermodynamic reasons connected with the form of the sorption
isotherms.
Chapter 2
22
erence to the working diagram shown in Fig. 2.2. Here the voltage vs. charge
relations for a capacitor and an ideal (constant voltage) battery are compared. In
actuality, most batteries, with the exception of Li-SOCI2, do exhibit some decline of voltage with decreasing state of charge, but this is for reasons quite different from that in the case of capacitor discharge.
The line VB is the voltage across the battery cell as a function of charge Q
held during a constant rate charge, Vc is the plot of voltage across a capacitor
being charged up to the same final voltage, VB' with accommodation of the same
accumulated charge, Q. However, since the line Vc is for capacitor charging, it
has a slope of lie (see Fig. 2.1).
The integrals VQ under the two lines of the working diagram correspond
to the energy of charging. Thus it is seen that VC'dQ is half VBdQ, i.e., the
energy of charging the capacitor to a terminal voltage VB with charge Q is half
that for charging the battery with the same charge also to VB' The superimposed
hatched and slashed areas in Fig. 2.2 correspond to this difference of energy
stored. As shown mathematically in Chapter 4, the factor of half in the charging
energy of a capacitor arises for the same reason that it enters into the Born equation for the self-energy of charging of spherical ions in vacuo or in a dielectric
(see Chapter 4).
23
..... dQ~
Full charge
~Battery
~ Capacitor
f
FIGURE 2.2. Working diagram illustrating storage of energy E = VdQ for a battery at ideally
constant charging voltage V = VB, and for a capacitor with progressively changing voltage, V =
Vc, leading to EB =QVB or Ec =(t)QVc.
Chapter 2
24
a) RU02 film
u;
..0
:::J
6.25
VI
IE
u
O~-+--+--r~--~-+--+--+--~-r~--;--+--r----
<t
E
........
c ~ 6.25
~ 0
'- 0
:::J J:
UI-
'"
POTENTIAL/V, RHE
0.2
1.0
0.6
1.4
b) Pb/PbCI 2
I.f)
o
33.2
<t
........
r
z
w
a::
a::
-0.717
o~~
-0.217
__~~~L-~~-=~~L-____-0.317
- L____~~
POTENTIAL/V, RHE
::::>
U
0
33.2
----+PbCI 2
66 .4
:I
FIGURE 2.3. (a) Cyclic voltammogram for RU02 in 1 M aqueous H2S04 (298 K) showing mirror-image symmetry and responses to successive switching along the scanned potential range. (b)
Cyclic voltammogram for Pb-PbCIz battery electrode showing typical irreversibility arising with
3-dimensional materials undergoing chemical phase changes.
25
age of that for negative, discharging, currents and switching the direction of
sweep during the potential scan produces an almost immediate reversal of direction of current. This behavior is also characteristic of a reversible, capacitative charge and discharge process (compare the theoretically calculated curves
in Chapter 10, Fig. 10.1).
Another characteristic feature of the behavior of this Ru02 example is that
over a wide range of sweep rates, s, the current response, I, at any potential
across the scanned range is linear in s. Since I = Cs, this means that C is independent of sweep rate and virtually pure capacitative behavior is exhibited; it is
actually mainly pseudocapacitative (see Chapter lIon RU02) when the RU02
has been formed in a hydrous state, e.g., by an electrochemical procedure.?
Figure 2.3(b) shows, by contrast, the behavior of oxidation and reduction
at Pb-PbClz, a battery-type electrode (Pb + 2cr ~ PbC12 + 2e). The voltammogram indicates a completely irreversible process: the potential range for formation of PbCl 2 from Pb is widely different from that for the reverse process,
reduction of PbCl2 back to Pb. Also, switching the direction of sweep during the
voltammetric trace does not immediately reverse the sign of current at that
switching potential, contrary to behavior observed with a capacitance.
In this system, which is typical for a battery-type reaction (Pb-PbS0 4 is
similar), overvoltages are required in both directions of the reaction in order to
pass significant currents. Hence the current response profiles for each of the directions of reaction lie wide apart, on each side of the reversible potential, along
the electrode potential axis of Fig. 2.3(b). The reversible potential corresponds
approximately to the potential at which the current profile I(E) crosses the potential axis, i.e., where I(E) = O.
Returning to the situation of a reversible process, it should be mentioned
here that if such a reaction is diffusion controlled, then the cyclic voltammogram is again not of the mirror-image type, the potential for the anodic peak current being more positive than that for the cathodic one. This is because, even if
the reaction is reversible, different potentials are required on the anodic
sweep from those on the cathodic one to establish given concentration
gradients in the boundary diffusion layer that lead to the same (equal but
opposite) diffusion-controlled currents. If such a reversible reaction is
conducted in a thin-layer cell where the solution thickness is less than the
width of the boundary diffusion layer, then the voltammogram tends to acquire
mirror-image symmetry like that for a reversible surface film process.
26
Chapter 2
host cathode materials, e.g., MoS 2, TiS 2, V60 13 , and C00 2 . 89 Normally these
cathode materials, coupled with an Li anode or a Li-C anode, would be regarded
as battery cathode materials. 1.2 However, the forms of the charge and discharge
curves and associated pseudocapacitance, and even the cyclic voltammetry profiles, are similar to those for 2-dimensional electrosorption (underpotential
deposition, UPD) (see Chapter 10). Hence this class of materials exhibits prop-
...
--- .......
a)
...........
2.2
>
.....
I.L.
.......
.......... ~
Li/TiS 2
.....
,~
w
1.6
b)
.... -,
>
.....
xl>
<3<1
I
2
I
I
I
I
Li/TiS 2
!/V
I.
, ....
- _....
.......
FIGURE 2.4. (a) Discharge curve for Li+ intercalation into TiS2 as a function of extent of discharge. X. expressed as a stoichiometric factor in LixTiS2 (solid line, calculated relation 8 ). (b) Differential coefficient of the curve for discharge of Li+ into TiS2 as in Fig. 3(a). (From Haering and
MacKinnon 8 .)
27
erties intermediate between those of bulk-phase battery reagents 1 and quasi-2dimensional pseudocapacitor electrodes.
The thermodynamics of sorption of Li into layer-lattice intercalation host materials have been worked out by Haering and MacKinnon8 in terms of 3-dimensional
lattice statistics which lead to 3-dimensional sorption isotherms in site fraction X, having forms similar to that of the relation for 2-dimensional adsorption into lattice arrays.9 Thus, such systems exemplify the transition between supercapacitor and battery
behavior.
Figure 2.4(a) shows an example of the sorption (discharging) isotherm for
a 3-dimensional intercalation process, e.g., for Li into TiS z.8 Since a Faradaic
charge is required for deposition ofLi into an intercalation host, X depends continuously on the charge passed. The relation of the potential V to X can be differentiated, so that formally a pseudocapacitance arises, as indicated by the
differential quantity shown in Fig. 2.4(b) as a function of X. It is seen that it has
a form like that of a cyclic voltammogram for a process exhibiting pseudocapacitance. This is because a cyclic voltammogram for an electrosorption process
gives directly, as the current response, the differential coefficient of the charging curve or the sorption isotherm (see Chapter 10).
The essential and common thermodynamic feature of systems that can give
rise to pseudocapacitative behavior in electrochemical capacitors is thus the relation of free energy to the log of the configurational activity terms, i.e., ()/l-()
for 2-dimensional adsorption or Xll-X for 3-dimensional sorption and
[Ox]/[Red]/(l-[Ox]/[Red]) for redox processes (see Table 2.4).
While double-layer capacitance behavior on carbon materials has usually
been distinguished from pseudocapacitance exhibited by RU02 and other systems, it has now become recognized that various high-area carbon preparations
TABLE 2.4. Correlation of Types of Systems Giving Rise to Pseudocapacitance
with Application to Development of Electrochemical Capacitors
System type
Essential relations
:R
Ltinto MA2
MZ+ + S + ze ~ S.M
E = EO + (RTlzF) In X/(1 - X)
X =occupancy fraction of layer lattice sites
(e.g., for U+ in TiS 2)
E = EO + (RTlzF) In {}I(l - (})
Note: (b) and (c) can be regarded as mixing of occupied (X or 0) sites with unoccupied sites (1 - 0) or (I - X).
Systems (a), (b), and (c) are dealt with in detail in Chapter 10.
28
Chapter 2
also exhibit a small but significant pseudocapacitance. This appears to arise because of electrochemically active redox functionalities (Chapter 9) that exist on
the surfaces of high-area carbon powders and felts, depending on the conditions
of their pretreatment.
Battery electrodes often become overcharged either adventitiously or because of degradation and aging of their plates. Overcharge processes occur
when the voltage between the plates exceeds the thermodynamic or, in practice,
the kinetic limits for decomposition of the electrolyte solution. Also, at high imposed charging rates, some current component will often pass in parallel with
the Faradaic charging process owing to decomposition of the battery's electrolyte, usually to form H2 and 02. This is caused by overvoltage effects.
In the case of an electrochemical capacitor being charged, there is no particular thermodynamic cell voltage of the kind to which a particular battery
should be charged. Thus, after the thermodynamic potential (1.23 V at 298 K)
for decomposition of water (the normal "voltage window" for the solution) is
reached, further charging to higher potential differences between the capacitor
electrodes results in increasing rates of solvent decomposition, which is similar
to that in overcharging of a battery. In both cases the double-layer capacitance
continues to be increasingly charged in proportion to the prevailing capacitance
at those potentials, but in addition, with carbons, some electrochemical oxidation of the active material can take place. The current is, however, divided into
a charging current, ic' and a Faradaic leakage current, iF, that normally will increase exponentially with the potential beyond the solution's decomposition
limit. The situation is thus:
(2.3)
or, correspondingly, in terms of respective current densities, i.
The leakage current component iF increases as the electrode potential increases, usually exponentially, and at sufficiently high potential, where decomposition of the solution takes place, iF ic and i ~ iF. Further details are
presented in Chapter 3. This situation is the opposite of self-discharge, while for
the ideally polarizable electrode or capacitor, i = icFor optimum efficiency in charging a capacitor, iF should be a minimum
over the whole range of potentials corresponding to the charging process over
the practical operating potential range of the capacitor: ca. 1.3 V for aqueous
electrolytes and 3.5-4.0 V for nonaqueous, aprotic solvent solutions. If iF is ap-
29
preciable toward the end of the charging half-cycle of a capacitor, there will be
a corresponding appreciable Faradaic self-discharge current following termination of charging when i is cut to zero, until the generated overvoltage has decayed. In practice, self-discharge down to lower voltages often continues (see
the discussion in Chapter 18). Once overcharge sets in, the efficiency of further
charging to higher interelectrode potentials across the capacitor rapidly diminishes owing to the "Tafel" character of the potential dependence of the iF component (Chapter 3).
Disadvantages
Limited energy density
Poor volume energy density
Low working voltages (compared with electrolytics; satisfactory compared with batteries)
Aq. voltage range 0 - 1.4 V; nonaq. to 4.5 V. In practice, 3.5 V.
Nonaq. embodiments require pure, H20-free materials; more expensive.
Requires stacking for high potential operation (electric vehicles)
Hence, good matching of cell units is necessary
30
Chapter 2
Electrochemical capacitor
Supercapacitor
DoG == constant
Has corresponding single-valued potential on
discharge
Usually not reversible, in the sense that
recharge curve is not mirror image of
discharge curve, e.g., in cyclic voltammetry
31
Battery
REFERENCES
1.
2.
3.
4.
5.
6.
7.
D. Linden, ed., Handbook of Batteries, 2nd ed., McGraw-Hill, New York (1995).
H. A. Kiehne, Battery Technology Handbook, Marcel Dekker, New York (1989).
W. Lorenz and G. Salie, Zeit. Phys. Chern., N.F. 29,390,408 (196\).
J. W. Schultze and F. D. Koppitz, Electrochim. Acta, 21, 327, 337 (1976).
B. E. Conway, H. A. Kozlowska, and W. B. A. Sharpe, Zeit. Phys. Chem., N.F., 98, 61 (1975).
S. Sarangapani, B. V. Tilak, and C.-P. Chen, J. Electrochem. Soc., 143, 3791 (1996).
S. Hadzi-Jordanov, H. Angerstein-Kozlowska, and B. E. Conway, 1. Electrochem Soc., 125,
1473 (1978).
8. R. R. Haering and R. MacKinnon, in Modem Aspects of Electrochemistry, B. E. Conway,
lO'M. Bockris, and R. White, eds., vol. 15, Chapter 4, Plenum, New York (1981).
9. B. E. Conway, Electrochim. Acta, 38, 1249 (1993).
Chapter 3
3.1. INTRODUCTION
On the basis of the assumption that this volume will be read not only by specialists in electrochemistry but also by electrochemical engineers and technologists,
as well as materials scientists, it has been considered desirable to include a brief
chapter on the basic aspects of charge transfer at electrode interfaces, especially
their energetics and kinetics. Some notes on the requirements for experimental
measurements on capacitor and battery electrodes are also included.
Ideally, the behavior of capacitative charge or energy storage devices, unlike the situation with batteries, would not involve any excursions into, or applications of electrode kinetics. This statement is predicated on the assumption
that the charging or discharging of a capacitor device involves only electrostatic
processes and not any electrode processes of the Faradaic kind, the kinetics and
energetics of which, if significant, would lead to deviations from pure capacitative behavior.
However, electrode kinetic effects do come into play whenever a capacitor
electrode ceases (practically) to be any longer ideally polarizable (Chapters 6
and 7). This can happen on overcharge when solution decomposition sets in
and/or when Faradaic self-discharge processes occur owing to the presence of
oxidizable or reducible impurities or reactive functional groups at carbon surfaces. Thus, in the case of electrochemical capacitors, some indirect or, in some
cases, direct involvement of Faradaic electrode processes can occur as follows:
33
34
Chapter 3
1. in adventitious overcharge or overdischarge of double-layer-type capacitors when decomposition of the electrolyte occurs;
2. in charge or discharge of the small but significant pseudocapacitative component of the capacitance that occurs with most carbonbased double-layer capacitor devices;
3. in the self-discharge processes that most electrochemical capacitors undergo on open circuit, following charging; and
4. in the basic Faradaic mechanisms of charging or discharging pseudocapacitors of the oxide or conducting polymer redox type, or of the adsorption type (see Chapters 10, 11, and 12) where the kinetics and
energetics of electrode processes are directly involved.
The electrode processes that are the subject of this chapter are of the socalled "Faradaic" type involving electrolytic charge transfer of electrons between electrode interfaces and ions or molecules in solution; hence the origin of
the term, which is named after Michael Faraday's work, especially that on electrolytic decomposition of solutions. However, it should be mentioned that at the
time (from 1834 onward) of this work, the details of such charge transfer processes involving electricity and chemistry I.2 were far from being understood. It
remained for Crookes and for J. J. Thomson (1897) to demonstrate much later
the common involvement of electrons in matter and their characterization in
terms of their charge/mass ratio and charge (by Millikan in 1909 and 1913). As
noted in Chapter 1, Faraday's laws implied an atomic unit of electricity, now
called the electron. However, it still took many years for it to be understood how
electrons became involved in electrochemical charge transfer processes. A
fuller understanding had to await: (1) knowledge about the electronic states and
their energy levels in atoms and molecules, a topic dependent on interpretation
of spectroscopic and ionization phenomena by Bohr and Sommerfeld and (2)
knowledge of the state(s) of free and bound electrons, their number density in
metals and semiconductors (by Fermi and Bloch), and their thermal (Richardson) and photoemission (Einstein) characteristics in relation to the so-called
"electronic work function" (t/J) for emission of electrons from metals.
The important conclusion from this work was that electrons are ubiquitously present in all metals, with relatively high concentrations being free, mobile, and delocalized-about one per atom; they are thus available for exchange
with ions or molecules across the metal's interface. Later, the somewhat different states of electrons, (or deficiencies thereof, i.e., holes) at much lower concentrations in semiconductors became understood.
35
At the same time, the energetic states of ions (or molecules) in solutions in
polar solvents had to be understood since the elementary act of an electrode
process is the transfer of an electron to (cathodic) or from (anodic) a reactant in
solution and from or to the electrode.
A basic factor in such electron transfer processes began to be understood
in a fundamental way in the 1930s and in more detail in the 1940s and 1950s;
it was that electrochemical electron transfer processes occur in a so-called "radiationless" manner, i.e., with no photochemical absorption or emission of light
(radiation) quanta. This means that there must be selective adjustment of electron energy levels in the electrode, relative to those in a reducible or oxidizable
reactant, so that the transfer of electrons takes place between levels that are
equal in energy. The solvation energy of the reactant ions or molecules in such
electron transfer processes is a major factor 3- 6 in determining the condition for
radiationless electron transfer and the rates of such reactions.
The general basis of this concept originated in papers by Gurney,7 Gurney
and Fowler,8 and Butler,9 was followed in more detail by the important paper
by Weiss,3 and later developed in a series of papers by Marcus 45 and Bockris
and Parsons. 6 The principal concepts are as follows:
1. The electrons involved in a Faradaic redox process are transferred from
(in a cathodic process) or to (in an anodic process) the Fermi level of the delocalized conduction-band electrons of the metal electrode. The electrons in the
metal obey the so-called "Fermi-Dirac statistics," and the upper level of halfoccupancy (the probability of the occupancy of that level being 0.5 at finite temperatures) is designated the "Fermi level." Electrons at this level differ in energy
from those in a vacuum with zero kinetic energy by a quantity, CP, the electronic
work function of the metal. <l> varies from about 2 eV (ca. 46 kcal mol-lor 184
kJ mol-I) for the alkali metals to about 5 e V or more for Pt and other noble metals. The value of cP for a given metal also depends appreciably on the orientation
of the crystal face at which the electron is exchanged. cP can be regarded as the
negative of the electron affinity ofthe bulk metal through its surface and is thus
related in a general way to the electronegativity of the metal forming the electrode material, and to its surface structure and ionization potential.
The energy required to remove an electron in a cathodic process is compensated by the energy gained when it has been transferred to an electron acceptor (oxidant, "Ox," reactant) and vice versa for an anodic process, from a
reductant, "Red."
2. The reactants that are electron acceptors, "Ox," or electron donors,
"Red," in solution are "solvated" by the electrolyte solvent (hydrated in the case
when the solvent is water). The transfer of an electron to or from such an Ox or
Red ion in solution is associated with a major change in the solvation energy of
the ion and also involves the energy of ionization 10 (see later discussion). When
molecular reactants are involved, the electron transfer usually results in the for-
36
Chapter 3
mation of an ion, or an ion radical, or a radical, with a corresponding large sol vation energy or with coupled generation or consumption of a proton in order to
achieve an overall charge balance in the reaction.
The overall energy changes in a simple electrode reaction can be usefully
represented by means ofan energy cycle (a Born-Haber cycle) based on the first
law of thermodynamics, which is related to the principle of conservation of energy. For example, for electron transfer involved in the reduction of a ferric to
a ferrous ion in aqueous solution at a metal electrode M (a single electrode reaction), we can write the energy cycle as:
e (M)
------+~
!l.G
Fe 2+
aq
from which the overalll1G for this half-cell electrode redox process is
(3.1)
Here Gs quantities are the Gibbs energies of solvation (hydration) of the ions
Fe3+ and Fe2+, and both are large negative energy quantities; J is the ionization
energy of the Fe 2+ ion in the gas phase to form Fe3+ in the gas phase, and IPM is
the electronic work function of the electrode metal M. The I1G quantities are
preferably expressed as the standard-state values, I1Go or GO, for a unit activity
or fugacity of the species in solution and in the gas phase.
The G, J, and IPM quantities in Eq. (3.1) are each relatively large, having
magnitudes of ca. tens or a few hundreds of kilocalories or kilojoules per mole.
However, their values with different signs add up algebraically to relatively
small numbers of 0 to lOO kJ mol-I. A similar situation arises for the I1Go values and corresponding acidity constant (pKa) values for acid ionization processes in aqueous solutions where solvation and bond energies determine the
2.3 RT(pKa) or I1Go values.
In the example treated earlier it is seen that the work function of the metal
M enters directly into the determination of the I1G or I1Go value for the half-cell
process and hence into its hypothetical electrode potential E = -I1G/zF. However, since a single half-cell potential can never be measured, only the I1E (i.e.,
the potential difference between two single electrodes) is experimentally accessible. In this case the work functions of the respective two electrode metals do
not enter into the experimentally measured emf of the cell since their difference
37
Chapter 3
38
Vac--r---.
Redox reoction
0+ e(M)+ R
(non - matching energy levels)
(a)
Vac --.--,,-
0+ e(M) ..... R
at electrode potential V (matching energy levels)
FIGURE 3.1. Conditions for energy mismatching (a) and energy matching (b) at the electrode
Fermi level for facile electron transfer to occur in a radiationless process (based on Ref. 3 and later
representations by Gerischer).
for the redox process Ox + e ~ Red; flG s in Eq. (3.2) is positive for a decrease
of net charge.
When the result of electron transfer is the production of an intermediate
chemisorbed with energy A (A being negative), condition (3.1) becomes9
l/> - eV - 1- flG s + A == 0
(3.3)
39
8
N
------=-
Ti
.Nb
IE
Cu
__;'w
.Mo
,TO
.Au
AQ,W-eF
AQ e"~
Ni
CI'
Bi
Mo
<t
....'"
AI
-e-.--..--iII.~Ir_
Rh
'd
0
3.5
':'
_~
4.0
4.5
5.0
PI
5.5
40
Chapter 3
-10
Langmuir adsorption of H
-3
+160
-160
FIGURE 3.3. Theoretical relation of Parsons for dependence oflog io for the HER on the standard
Gibbs energy, LlC:ds.H. for chemisorption of H at the electrode metal. (From Parsons. 12 Copyright
1958 Royal Society of Chemistry.)
11; this must be the energy of adsorption (AH) of H at the metal. The apparent
relation to (j) arises because AH usually depends on (j),13 e.g., for the initial heats
of chemisorption at low coverage.
Parsons 12 derived a theoretical relation for the dependence of log io on the
standard Gibbs energy (LlG~) for chemisorption of H at the metal, and its form
is as shown in Fig. 3.3. A maximum arises in log io when LlG~ is zero, because
log io is dependent on the coverage product, ~(1 - BID1-a, which is determined
by LlG~. This function has an obvious maximum when BH 0.5 and a 0.5, the
transfer coefficient, and gives rise to the so-called "volcano relation" illustrated
in Fig. 3.3. Thus, volcano relations can occur in electrocatalysis for reasons
~imilar to those for regular heterogeneous reactions, as was discussed in classical catalytic work e.g., by Balandin. 14 An experimental plot of log io for the
HER vs. heat of adsorption of H based on critically selected data was given by
Trasatti 15 and is replotted in Fig. 3.4.
Only for those metals that have a strong affinity for some electro sorbed
species (e.g., H at Pt, Rh, Ru, Ir or Pb or Bi atoms at Au or Ag) does UPD occur
or alternatively, in the OPD potential range, when a desorption step is rate
determining, does appreciable potential-dependent surface coverage of the
electrosorbed species occur. Then, for such conditions, adsorption pseudocapacitance (Chapter 10) arises. Thus, the energetics and kinetics of charge transfer processes that lead to an adsorbed, discharged (or partially discharged)
species (see Ref. 16) are of major importance in determining which and what
cr:
41
ill
....
.E
E
u
--
0)
.Q
30
50
70
90
First it is useful to note that the rates of electrode processes are directly
measurable as currents, J, or as current densities, i, expressed per square centimeter of available accessible area (A) of the electrode surface, which is important for electrode reactions at interfaces of high-area, porous electrode
materials. Thus,
i=zFv
(3.4)
where v is the reaction velocity in mole s-1 cm- 2 and z is the number of faradays
passed per mole of reactants. The current density i is related to the current J passing at the available electrode area A by
42
Chapter 3
=I1A
(3.5)
Equation (3.4) applies to a given electrode process only if its current efficiency is 100%; otherwise a current efficiency factor <1 must be applied. For
example, in the cathodic electrodeposition of Zn from Zn 2+ or ZnO~- ions in
aqueous solution, significant co-deposition of H, with H2 evolution, takes place,
so Eq. (3.4) applies properly only to either the partial current density for Zn
deposition or to the partial current density for H2 evolution. The relative proportions must be analytically evaluated.
The primary feature of the kinetics of electrode processes, distinguishing
them from kinetics of regular chemical reactions, is their dependence on the potential, V, of the electrode, which is usually referred to that of some reference
electrode in contact with the electrolyte solution.
This dependence on V arises energetically, as explained earlier, because of
the dependence of the electron work function (/J on V: (/Jy = (/Jy=o e V. Since
the electron is a reactant in the electrode process, changes of its energy in the
metal relative to that in vacuum modify the activation energy for the charge
transfer process and hence the rate constant of the electrode reaction.
Thus if the energy of the initial state of the reactant ion or molecule, plus
the electron in the metal, is changed by eV or VF per mole of electrons
(96,500 coulombs) then the activation energy is normally changed by some
fraction, p, of VF. p is commonly around 0.5 0.1 for many electron charge
transfer processes, and is called the "barrier symmetry factor." It is analogous!?
to the Brj/lnsted coefficient for a series of related homogeneous proton transfer
processes which proportionates their Gibbs energies of activation to the standard overall Gibbs energies of the respective protonation reactions.
The significance of the effect of change of VF in relation to p can be illustrated in the schematic energy diagram of Fig. 3.5, which shows the effect on
the activation energy AE* (or ilC* for the Gibbs energy) that is involved in the
rate equation. It is found that
(3.6)
over an appreciable range of change of V. P= 0.5 when the energy curves for
the course of the electrode reaction (e.g., discharge of H from an H30+ ion or
from water) along its reaction coordinate, d (Fig. 3.5) cross one another symmetrically, i.e. (numerically) with the same slopes dG/dd at and near the crossing point corresponding to the transition state (activated state) of the electron
transfer reaction. For the discharge of H30+ or solvated metal (M) ions, the immediate product of the electron transfer step is an adsorbed atom (H) or adion
of M at the electrode surface.
43
I
I
, ,,
\
____________ J
INITIAL STATE
REACTION COORDINATE, d
FIGURE 3.5. Significance of effect of change of electrode potential, as VF, on energy of activation, till*, of an electron transfer reaction in relation to the barrier symmetry factor, p. (Equivalent
to an electrochemical Brnsted relation. 17)
The basic kinetic equation for such processes is as follows (Ref. 11 and
General Reading refs.), taking the H deposition case as the example:
i = zFk cH:s (1 - 0H)exp - [fJVFIRT]
(3.7)
44
Chapter 3
tl = V - Vrev or V = tl
+ Vrev
(3.8)
(3.11)
The reciprocal of this derivative is called the "Tafel slope," and is denoted by b,
so that
b = -2.3 RTI[JF
(3.12)
with In i expressed as 2.3 log i; then b is written empirically in the Tafel equation:
tl
= a + b log i
(3.13)
after Tafel's work (1905) on the HER, which arose originally from his research
on the electrochemical reduction of strychnine. Various Tafel slope values (b)
can arise for the HER, depending on the rate-controlling mechanism that obtains
or the H coverage conditions (eH 1 and potential dependent, or eH ~ 1 and
thus independent of tl).
More generally Eq. (3.7) can be written in the form
~
ilzF =KCH:s(1-
(3.14)
(3.15)
45
where the two terms on the right-hand side (rhs) ofEq. (3.14) or (3.15) represent
the forward and backward rates of the reaction, both of which are significant
near (within 10-15 m V) the reversible potential. Equations (3.14) or (3.15) are
called the "Butler-Volmer equation."
When the forward direction is predominant at appreciable values of 11,
i = io exp[-pl1FIRT],
(3.16)
(3.17)
A comparison with Eq. (3.13), the empirical Tafel relation, reveals that
11
2.3RT I . 2.3RT I .
=PF
og lo-PF og
1
(3.18)
so that
Another aspect of the properties of Eq. (3.15) for near-equilibrium conditions (small 11 ) is that the two exponential terms can be linearized in a Taylor
series expansion to the approximation:
(3.20)
=-iorJ FIRT
(3.21)
46
Chapter 3
(3.23)
Note that in Eq. (3.15) a negative sign is taken for the cathodic tT in the HER.
This latter relation (3.23) is useful for characterizing the dynamics of the
electrode reaction in terms of:
1. the so-called "microscopic polarization" (low tT) of the electrode reaction where i is linear in tT; and
2. the impedance of the reaction under alternating voltage modulation
where the Faradaic resistance, RF , can be determined as well as the
double-layer capacitance with which RF is in parallel (Randles'
equivalent circuit) sometimes together with a diffusional, so-called
Warburg impedance, W (see Chapters 16 and 17).
The Faradaic resistance has special significance in work on the frequency
response (impedance analysis) of electrochemical capacitors in cases where
either some Faradaic self-discharge process takes place or when pseudocapacitance is involved in a primary way in the behavior of the device. In these cases,
an RF component, which is usually dependent on the dc-level potential to which
the device is polarized, must be included in the equivalent circuit that represents
the frequency response of the capacitor system. This matter is treated in detail
in Chapters 16 and 17 on impedance spectroscopy.
47
"~
Log I
----
OVERPOTENTIAL, Tj,+VE
Log I
~I \"~
Log io
TAFEL
FIGURE 3.6. The anodic and cathodic current components of the potential dependence of an
electrode reaction rate near its reversible potential: graphical representation of the Butler-Volmer
relation (Eq. 3.15) illustrating the dependence of the net current density, i =ia - ie, on overpotential
around the reversible potential (linear micropolarization relation) and the significance of the exchange current density, io.
The second significant point is that both curves intercept the axis of i values at +io for the anodic curve and -io for the cathodic curve. The distance between the intercepts in terms of current density is two times i o .
With increasing negative '7, the ic becomes progressively larger than ia until ic ia and Tafel behavior applies, i.e., ic then increases exponentially with
-'7. On the other side of the curve, for positive polarization, ia behaves similarly
with positively increasing '7. On logarithmic Tafel plots (i.e., for log ic or log
ia plotted against appreciable negative or positive '7 values), the intercepts oflog
ic or log ia on the '7 axis give the log of the exchange current density. This is one
of the standard procedures for experimental determination of log io or i o.
If the net current density, i = ic - ia (see the Butler-Volmer relation in the
form ofEq. 3.20), is plotted along with the individual component's partial ic and
48
Chapter 3
For so-called "activation-controlled" electrode processes, " increases linearly with log i in the absence of any resistive (ohmic) potential drops in the solution. However, for high current densities and/or large io values, concentration
polarization may set in owing to the kinetic limitations associated with finite
rates of diffusion of the reactant ions or molecules to the electrode surface.
Then, toward high i values, the experimental Tafel plot of I] vs. log i can curve
upward with an approach toward a diffusion-controlled limiting current as illustrated in Fig. 3.7. Such diffusion-limited currents can, however, be increased by
means of solution stirring or forced flow, or by rotation of the electrode. Then
the diffusion-limited currents can be increased in proportion to the square root
of the electrode rotation rate (Levich relation) in the case of an electrode having
49
LOG [i 0]
z
o
!;i
~
Cb
W
U
8
I-
z
~
u
ct
w
0::
the geometry of a disk with the rotation axis normal to its surface. Such experiments can be conducted quite easily with modern rotated-disk electrode equipment so that the role of diffusion limitation can be quantitatively characterized
and, for appropriate conditions, the diffusion constant of the reactive species determined. However, this procedure is inapplicable for porous electrodes.
50
Chapter 3
Equations (3.7) or (3.14) have been written for a simple discharge reaction
exemplified by the first step of the hydrogen evolution reaction (HER) or by the
forward direction, or both directions of the H underpotential deposition (UPD)
process. It should be mentioned that other steps in the HER may be rate determining, such as electrochemical atom-ion desorption of H forming H 2, or the
chemical desorption of two adsorbed H atoms by catalytic recombination to H 2.
Such steps are represented by electrochemical rate equations different from
(3.7) or (3.14). However, the details are not relevant to the material treated in
this volume but may be found worked out at length in the General Reading Refs.
3,4, and 5. The case actually examined here illustrates the general principles and
energetics of electrode process kinetics involving electron transfer and overvoltage.
When a step subsequent to that of the initial electron transfer in the
cathodic direction of the HER, Tafel slope values (b) can be different from those
usually observed, namely, 2.3RTlPF with p:::: 0.5. They then depend on (1) the
type of desorption step involved, i.e., whether it is of the electrochemical kind
(with a further electron transfer event) or of the chemical kind, e.g., discharged
radical recombination or dissociation (as for H2 formation in the HER or involving CH3COO- and CH3 in the Kolbe reaction that produces C 2H 6 from discharge of acetate); and (2) on the extent of coverage by the electrosorbed
intermediate, i.e., whether the fractional coverage (J is 1 and then potential
dependent or whether at elevated overvoltages (J--+ 1 and is then virtually independent of potential.
The Tafel slopes that arise are then generally of the form 2.3RTI(n + P)F
or 2.3RTlmF where n or m are integers; the former arises when desorption of an
adsorbed intermediate is rate controlling via a further electron transfer and the
coverage by the intermediate is potential dependent according to some quasiequilibrium electrosorption isotherm (Fig. 3.8); the latter arises for a nonelectrochemical desorption step involving, for example, recombination of adsorbed
radicals (m :::: 2) (Fig. 3.8) or dissociation of a radical (m :::: 1), e.g., CH3COO".
Values of nor m are usually the integers 1 or 2, but for more complex sequences
(e.g., in anodic O 2 evolution), n can be 2 or m 4. 19
When the coverage, by an adsorbed intermediate (e.g., H) as a function
of potential is known or can be assumed, the Tafel slope, b, takes the useful general form:
e,
(3.24)
51
C)
o
>
a:
o>
where the first term on the rhs ofEq. (3.24) is the logarithmic derivative ofthe
adsorption isotherm, O(V) (e.g., for H in the H2 evolution reaction), and the second term arises from the potential dependence of the electron charge transfer
rate through the barrier symmetry factor (fJ =:: 0.5) or electrochemical Brjijnsted
factor. 16 Special cases arise when d In OldV =FIRT or d In OldV =0 (0 ~ 1).
In the case of underpotential deposition (e.g. of H), no continuous Faradaic
currents pass (i.e., fJFIRT = 0 in Eq. 3.24), and adsorption pseudocapacitance
arises because 0 (e.g., 0H) is a thermodynamic function of V over some defined
potential range, about 0.05 to 0.35 V, depending on the substrate metal, which
is usually a noble one.
52
Chapter 3
(3.25)
where IfIl is the potential at the inner limit of the diffuse double layer (i.e., at the
Helmholtz compact layer) relative to the potential in the bulk of the solution in
the absence of net ionic currents that could lead to an ohmic iR drop. This is the
potential that is derived from the theory of the diffuse double layer according to
Gouy and to Chapman (Chapter 6, references 7 and 9, respectively). (In the case
where the ions are significantly chemisorbed [usually anions at positively
charged electrode surfaces or near the potential of zero charge, pzc, at negatively charged surfaces as well], the potentiallfll is appreciably modified.)
In the kinetic equation (Eq. 3.9), CH,+s must be substituted according to Eq.
(3.25). In addition, in Eq. 3.7, when double-layer effects are to be properly included in the rate equation for i, account must be taken of the fact that it is the
local interfacial potential difference V - 1fIl> rather than V (or 11), that has to be
written in the rate equation. When these two effects are included (in the absence
of specific chemisorption of the reacting ions), the full electrode kinetic rate
equation for the example of proton discharge becomes from Eq. (3.7):
(3.26)
In dilute solutions, the theory for ion distribution in the region near and up to
the compact Helmholtz layer of the double layer formally gives an approximate
solution for the ionic concentration dependence of IfIl for dilute solutions (ionic
strength < ca. 0.1 M) as:
(3.27)
Note that for a cathodic process (i.e., one involving discharge of cations), IfIl is
a negative potential (relative to the bulk solution potential) that becomes numerically smaller with increasing ionic strength. The effective thickness of the
diffuse double layer (the Debye length) becomes smaller in proportion, approximately, to the log of the ionic strength at relatively low ionic concentrations.
The basis of double-layer effects in electrode kinetics was thoroughly
worked out by Frumkin and his school in the 1930s and by Gierst in more specialized ways in the 1950s and 1960s (see Refs. 11 and 20). For the case of specific (chemisorbed) ions in the double layer, usually polarizable and
electron-donative anions, the double-layer effects cannot be treated in a generalized way, i.e., without individual evaluations of their adsorption behavior and
their effects on the local potential difference across the compact Helmholtz
53
(3.28)
*The term interphase is distinguished from interface in order to recognize the 3-dimensional nature
of the former as a region of finite thickness residing on the 2-dimensional interface defined by the
boundary of the electrode surface with the solution.
Chapter 3
54
and
Aq
c=
=f idt
f i dtlAV= i AtlAV
(3.29)
(3.30)
over some time interval At. Equation (3.30), as written, applies if the capacitance, C, is constant with potential. Often experimentally it is not, so AV deviates from a linear dependence on time at constant current. C is thus obtained as
the reciprocal of the slope of the relation between AV and At (Fig. 3.9), the time
elapsed to some point on the AV vs. At curve or differentially, at some point on
it, as (dAV I dtr l when C is not constant with AV. The measurable dependence
of AVon time at a constant current (i.e., on charge passed) is commonly referred
to as the "charging curve."
This applies to a so-called "ideally polarizable" electrode (see Chapter 6)
where the current i simply passes charge into the interface without any Faradaic
processes of transfer of charge across the double layer taking place, leading to
chemical change at the electrode surface. As long as that condition obtains, the
electrode interface remains ideally polarizable and i is purely a double-layer
charging current, i dl .
However, in practice, the charging current may be maintained in a voltage
range across the interface where Faradaic decomposition of the solution begins
to take place at that electrode (and then also at the counter electrode). The dou-
lIe
a
TIME (AT CONSTANT i)
FIGURE 3.9. Constant current charging curve for an electrode interface having a double-layer
capacitance C.
55
ble layer continues to be increasingly charged with the rise of Ll V, but the current
i then becomes divided into two components, i d1 and iF, where iF is the Faradaic
current, which increases exponentially with Ll V when Ll V exceeds a value corresponding to a thermodynamic reversible potential for solution decomposition
(e.g., H2 evolution) at that electrode. This increase of iF follows a Tafel relation
in the overpotential, Yf, involved, as discussed earlier and represented by Eq.
(3.13). Then the charging situation can be written as
(3.31)
i = C(dVldt) + iF
where C(dVldt) = i d1 and iF has the form i F = io exp [aYfFIR1J, which is the Tafel
equation in exponential form (Eq. 3.7), with a the transfer coefficient.
The charging curve then has the form shown in Fig. 3.9 (for constant C)
and as Ll V increases beyond its value corresponding to solution decomposition,
the Ll V vs. Lll relation increasingly deviates from linearity as the iF component
of i becomes progressively larger, i.e., a greater leakage current across the double layer passes until the Faradaic process is the main charge-carrying component, iF i d1 Then the electrode is behaving in a nonideally polarizable way.
The situation of transition from double-layer charging to mainly Faradaic decomposition of the electrolyte as the potential is raised is illustrated in Fig. 3.9.
Region ab is for pure double-layer charging; region bc includes an increasing
component of the passage of charge in a parallel Faradaic reaction. ab is the region of ideal polarizability of the interface, prior to decomposition of the solution over region bc. Figure 3.10 shows schematically the relative values of the
LL
UJ
Z
oc...
o==
o
onset of non-ideal
polarizability, iF 0
---:....-
!Z
UJ
Constant i
= idl+ iF
a::
a::
;:)
POTENTIAL
FIGURE 3.10. Components of the constant charging current of Fig. 3.9 at an electrode interface
when a Faradaic partial current, iF, passes in parallel with the double-layer charging current, idl.
56
Chapter 3
57
Potential- step
w
(J)
z
oD..
V2
..J
tzw
oD..
(J)
a::
a::
a::
;:)
I:!!
v,
"" ,
,,
iF>O
"-
',iF"O
......
FIGURE 3.11. Current-response transient for charging an electrode interface under the influence
of a step of potential from VI to V2.
arly with time in a three-electrode cell and the resulting time-dependent response current (Fig. 3.12) is registered by an analog (X-Y recorder or oscilloscope) or digital (transient recorder, computer, or digital oscilloscope)
recording instrument.
In such a voltage sweep experiment at a sweep rate s = dVldt (= constant),
the capacitative charging current, i, is
i
= C (dVldt) = Cs
(3.32)
or
(3.33)
C=ils
In a sweep-reversal experiment (cyclic voltammetry) at constant is, the current
response profile is ideally a rectangle along the time-potential axis when C is
constant (Fig. 3.12, line a) or if it is not, then the differential profile of C is generated, often with a peaked structure (line b in Fig. 3.12).
4. In a self-discharge experiment through a load resistance, R, the time
dependence of potential is (see Chapter 18)
Vet) = Va exp [-tIRC]
(3.34)
Z'= lIjwC
(3.35)
58
Chapter 3
VI
.!!..
:.0
0
t-"
Z
c
ct
IIexs)
~
I
i
I
I
+
i
a::
a::
::J
U
I
I
a::
__
I
I
a..
VI
CJ)
CJ)
:c0
.&;.
"0
--
IV2
-......
--
.-_
....
Sweep - signal
vO) _ -
-!
oil(
< .................
..
.~
....
ct
VI
U
II
'C
POTENTIAL,V
I time)
I
>1
I
I
1.::EILl
l~
'0
1%
,0
t~
,ILl
jl3
,0
tg:
a::
a::
=>
U
V>
POTENTIAL, V
( time)
a..
V>
a::
:c0
.&;.
FIGURE 3.12. Cyclic voltammogram for an interface having a potential-independent capacitance (C = ils) (curve a) and potential-dependent capacitance (curve b),
and ideally the real component Z' is infinity for an ideally polarizable electrode,
Most capacitor systems, especially those of the electrochemical type, have
a complex R-C equivalent circuit of distributed Rand C due to porosity (see
Chapter 14) and/or complex electrochemical reactions, and hence do not have
a simple impedance behavior as a function of frequency, In some cases, the im-
59
pedance behaves like that of a transmission line having a 45 phase angle over
a wide frequency range instead of 90 for ideal capacitance. Finite or significant
R limits power output performance as discussed in Chapter 15.
Angerstein-Kozlowska22 has given a detailed account of various experimental procedures for electrochemical experimentation, including the design of
cells and preparation of solutions, especially techniques for measurements in
very pure solutions and on clean electrode surfaces. Almost all packaged electrochemical capacitors or batteries are sealed two-electrode (or a stacked series
of two-electrode-) systems. However, electrochemical test and evaluation information is often required on individual electrodes (the positive or the negative
one) of the device. In this case, a third, unpolarized, electrode is required as a
reference probe. Provision for such measurements must then be made by using
a three-electrode or three-compartment cell of the type shown in Fig. 3.13. Alternatively, a third (reference) electrode probe can be inserted between two
working plate-type electrodes in a cell of the type shown in Fig. 3.14. Then, during tests, the potentials of each working electrode on charge, discharge, or open
circuit (for self-discharge measurements) can be separately recorded against the
potential of an appropriately compatible reference electrode.
Some commonly available reference electrodes are as follows:
The Hz-H+-H20 or Hz-OH--H20 electrode at platinized Pt.
The PdH electrode at Pd containing sorbed H.
The dynamic hydrogen electrode where H2 is generated in situ at low current density at a small platinized Pt electrode.
Hg-Hg2Clz-CI-, reversible to Cl- ion.
Hg-Hg2S04-S0~-, reversible to SO~- ion.
Hg-HgO-OH-, reversible to OW ion at pH > 9.
Ag-AgX-X-, reversible to halide, X-, ions.
Ag-AgOH film-OH-, reversible to OW ion.
Fe-FeF2 film or Cu-CuF2 film for P- ion solution.
Ag-Ag+-AgCI0 4 in nonaqueous solvents.
Glass electrode, sensitive to pH in aq. solutions.
Ion-selective electrode (Orion Co.) sensitive to various specific ions in solution.
60
Chapter 3
61
Ref.
iii
I.
1:1
I-
Plate-type cell
in plastic frame
(Schematic)
I
I
I
J)
I
I
Ii
"I I
~
~
1111
~separator
FIGURE 3.14. Plate-type test cell with arrangement allowing insertion of a reference electrode
probe.
When a reference electrode is used, potential measurements or other connections must be made into or through a high input impedance (ca. 106 ohms)
recording device in order to minimize current drain from the reference electrode
to below several microamperes. This ensures that the reference electrode maintains its proper reference potential to within a millivolt, i.e., it does not itself become polarized on account of its inclusion in the measurement circuit. A typical
three-electrode measurement circuit is shown in Fig. 3.15.
3.10.2. Instruments
The various electrochemical instrumental procedures available for studying the behavior of capacitor devices and batteries, and electrode processes in
general, are well treated in the monographs by Delahay (General Reading Ref.
1) and by Bard and Faulkner (General Reading Ref. 6).
Basically, for the evaluation of electrical response functions of electrochemical capacitors, the following equipment is required:
1. A potentiostat [e.g., EG and G (PAR), Hokuto or Wenking] that is capable of addressing and controlling cells up to 5 V. Some higher voltage options
62
Chapter 3
FUNCTION
- GENERATOR (5)
POTENTIOSTAT
R
C~
;;;;,CELl
I
-'V"",.N'y
\ C I
',/
'\ \
I
x- y
1-0 RECORDER
'-9
\., //--'"
/
"
{, W
,
,,- .....
'R'
.I
POTENTIOMETER r--
63
Most packaged electrochemical capacitor devices are two-terminal systems, so no reference electrode can be used as a third potential probe without
making an opening in the device's case. Testing is therefore conducted mainly
by recording charge and discharge curves at controlled currents at various tem-
64
Chapter 3
peratures and by potentiometric recording of open-circuit potentials or float currents (Chapter 18) when self-discharge rates are being evaluated. Under these
conditions, of course, an overall evaluation of the two-electrode system is obtained so that information on the behavior of each electrode interface is not
available from such measurements. However, in practice, useful performance
information on the charge and discharge capacity, energy density, and power
density of capacitor devices is easily obtainable by means of two-terminal measurements, but is less informative in a fundamental direction than data obtained
at individual electrodes in three-electrode cells.
REFERENCES
1. L. Pearce Williams, Michnel Faraday: A Biography, Chapman and Hall, London (1965).
2. B. E. Conway, in Electrochemistry, Past and Present, Chapter II, ACS Symposium Series
390, American Chemical Society, Washington, D.C. (1989).
3. 1. Weiss, Proc. Roy. Soc., London A222, 128 (1954).
4. R. A. Marcus, 1. Chem. Phys., 24, 966 (1956).
5. R. A. Marcus, Ann. Rev. Phys. Chem., 15, 15 (1964).
6. R. Parsons, and 1. O'M. Bockris, Trans. Faraday Soc., 47, 914 (1951).
7. R. W. Gurney, Proc. Roy. Soc., London, AI34, 137 (1931); A138, 378 (1932).
8. F. P. Bowden, Trans. Faraday Soc., 28, 368 (1932).
9. 1. A. V. Butler, Proc. Roy. Soc., London, A157, 423 (1936).
10. B. E. Conway, Ionic Hydration in Chemistry and Biophysics, Chapter 13, p. 260, Elsevier,
Amsterdam (1981).
11. B. E. Conway, Theory and Principles ofElectrode Processes, Ronald Press, New York (1964).
12. R. Parsons, Trans. Faraday Soc., 341053 (1958).
13. B. E. Conway and 1. O'M. Bockris, 1. Chem. Phys., 26, 532 (1957).
14. A. A. Balandin, Z. Phys. Chem., B2, 289 (1929).
15. S. Trasatti, 1. Electroanal. Chem., 39, 183 (1977).
16. 1. W. Schultze and F. D. Koppitz, Electrochim. Acta, 21, 327, 337 (1976).
17. B. E. Conway, in Progress in Reaction Kinetics, G. Porter, ed., vol. 4, Chapter 10, Pergamon,
Oxford (1967).
18. B. Losev, quoted by L. Antropov, Theoretical Electrochemistry, p. 395, Mir Pub!., Moscow,
1972.
19. K. 1. Vetter, Electrochemical Kinetics, Springer-Verlag, Berlin (1965).
20. L. Gierst, in Electrochemical Society Symposium on Electrode Processes, E. Yeager and P.
Delahay, eds., Electrochemical Society, Pennington, N.J. (1961).
21. D. C. Grahame, Chem. Rev., 41, 441 (1947).
22. H. Angerstein-Kozlowska, in Comprehensive Treatise of Electrochemistry, E. Yeager, J.
O'M. Bockris, B. E. Conway, and S. Sarangapani, eds., vol. 9, Chapter 2, Plenum, New York
(1984).
65
Chapter 4
4.1. INTRODUCTION
The electrochemistry of double layers, and the ions and solvent molecules constituting them, involves the electrostatic energies and molecular or ionic distributions of these species in high interphasial fields. At charged electrode
interphases in double layers, the electric fields can become as high as 107 V
cm- 1, and similarly in the solvation shells of ions. The electrostatics of such
interphases are concerned with:
1. the energies of individual charges and molecular electric dipoles, and
of their interactions;
2. the motions of individual charges (ions) in fields;
3. the configuration of groups of charges on ionized complex molecules
(e.g., ionic centers arising at conducting polymers) currently being developed as pseudocapacitors;
4. the orientational movement of electric dipoles in homogeneous fields;
5. the translational motion of dipoles in inhomogeneous fields;
6. the interaction of solvent dipoles with electric fields, and quadrupoles
with field gradients; and
7. the effective local dielectric coefficient in the double-layer interphase.
67
68
Chapter 4
Magnetic interactions are much weaker and are only important in spectroscopy,
in some solids, and in transition metal ion properties.
When a charged electrode interface is in contact with an ionic solution,
there is an accumulation of ions of one sign or the other, forming a double layer
(Chapter 6) and causing an orientation of dipoles of the solvent. Since ions in
the double layer are normally situated in a liquid solvent in contact with the electrodes, the electric polarization of the medium will also be involved in determining the properties of the charged particles at or near an electrode interface. In
order to provide a basis for discussing the details of such questions in the theory
of the double layer and electrochemical capacitors, we present in this chapter
some elementary principles of electrostatics.
Perhaps the two most basic laws in physics and chemistry that determine
the short-range properties of substances are Coulomb's law of electrostatic
force, F, between two charges q1 and qz, distant r from one another, and the corresponding law of interaction that determines the force between two magnetic
dipoles. A third basic relation is the Pauli exclusion principle, which places restrictions on quantized energy states in assemblies of electrons in atoms or
molecules and is hence fundamental in defining interatomic forces arising from
electronic interactions in molecules. Analogous effects occur in metals, giving
rise to their band structure.
Coulomb's law has been experimentally verified as an inverse square law
like that for gravitation. The general relation for the force F between two
charges in a vacuum, is given by
(4.1)
where F is an attractive force when q1 and qz are of opposite sign and a repulsive
one when they are of like sign. The constant k depends on the units in which the
charges and distance are expressed and the force F is also dependent on the electrical properties of the medium in which the charges q1 and qz are situated when
it is not a vacuum.
4.2.1.1. Units
69
k = 1I4nKo
(4.2)
where KO is the permittivity of the vacuum. The rationalized system is introduced so that certain fundamental relations in electrostatics, such as Gauss's and
Poisson's equations, can be written without the 4n factor which otherwise enters
into them. The value of KO is
KO = 8.85435
(4.3)
In some texts the symbol eo is used for the permittivity of free space, but
the symbol Ko is preferred here in order to avoid confusion with the static (zero
frequency) dielectric constant of a medium, also usually denoted by eo or eO. KO
is also sometimes referred to as the diabattivity of free space.
Electrostatic calculations in the rationalized units system thus depend on
forces calculated from Fr=QIQzI4nKoe,z or Ur=QIQ2/4nKoBr from corresponding energies. The charges are written QI, Qz here in coulombs to distinguish them from q], q2 in Eq. (4.1), where the units are to be taken in esu.1t is
often, however, simpler to perform such calculations in the esu system with e
then taken simply as the dielectric constant (see later discussion) of the medium;
then forces appear as dynes and energies as ergs per particle. Conversion to
joules mol- I is easily achieved by multiplying by NAil 07 , where NA is
Avogadro's number, noting that 107 ergs = 1 joule. Itis obvious that the two systems must give the same results with the appropriate interconversion of units.
Thus a unit charge in coulombs is 3 x 109 times that in esu, or 1 C = 3 X 109 esu.
The conversion factor 3 x 109 (actually more exactly 2.998 x J09) originates
from the speed oflight in a vacuum, 2.998 x 1010 cm S-I, and is 1110 of this figure, corresponding to the coulomb being a practical unit that is 1110 of the value
of the absolute electromagnetic unit of charge, which is itself defined in terms
of the magnetic field produced in a circular coil of one turn having a radius of
1 cm, when a unit charge per second (current) is flowing.
70
Chapter 4
'1
= q = 113 X
2
Therefore, equating the result in joules from the two systems, it is seen that
Ur=1 em
(4.6)
or
KO
10-16
(4.7)
71
qx1
(4.8)
-F(x)=-eX2
00
V(X)
(4.9)
=q / ed
or, in general,
J(x)
=q / eX
(4.10)
The corresponding electric field E is defined as the gradient of the potential. For the above case,
E = dJ/dx = -q/ex2
(4.11)
Hence field varies as the inverse square of a distance x from a charge while potential varies as the inverse first power of this distance (Fig. 4.1). From the form
of Coulomb's law, it is evident that the force on a charge q' in the field of q is
given by
F=_lxq'
(4.12)
eX2
i.e., force is the product of the charge multiplied by the field due to another (or
other) charge(s). This formula provides the basis for the definition of the unit
charge given earlier.
72
Chapter 4
Potential
cp
Distance x
FIGURE 4.1. Relation between electric potential and electric field as functions of distance, x,
from an ionic charge (schematic).
73
4>A
(0 )
-qEd 3
(b)
q2
Field Resultant
FIGURE 4.2. (a) Scalar summation of potentials due to a distribution of charges at various distances dl. d2. d3 from point A. (b) Vectorial summation of fields to derive resultant field at a given
point due to the distribution of charges shown in (a).
Field intensity gives a measure of the so-called "force" associated with the
action of the field on a unit charge. One "line" per square centimeter is defined
as existing in classical electrostatics for each dyne per esu of field strength.
Lines of forces correspond to the tangential force experienced by a charge in the
field at the particular point concerned. Density of lines of force corresponds to
the flux and electrification. The concept of lines of force originated with Faraday.
The force at ron a unit charge due to q is q/r2 (8 = 1), i.e., the field intensity.
This field can be represented by drawing q/r2lines of force per square centime-
74
Chapter 4
ter at r. The total of such lines at a distance is the line density multiplied by the
total cross-section at r, i.e.,
(4.13)
Ae
(4.14)
75
FIGURE 4.3. Model for discussion of the basis of Gauss's theorem in terms of normal induction
across an element of sectional surface "S through a surface Be at distance r from a charge q.
fIn' ds =4nq
(4.15)
Let q be the enclosed charge and the flux of induction will be In . ds, and this
will be I cos eds and be equal to the field (q/?-) X (the shaded circle area, BM).
However ,the area BM/?- measures the solid angle Je. Then the flux of induction across the area Be is qde. Hence the total flux is
f qde =q f de =-q . 4n
(4.16)
(4.17)
which is a form of Gauss's theorem. If the plate has two surfaces (i.e., it is not
a bulk conductor on one side), the result is E = -2nq/e.
76
Chapter 4
the field components dJldx, dJldy, or dJldz, increase or decrease in the various
orthogonal directions of the Cartesian coordinate system. In terms of p and the
dielectric constant I'. of the medium in which the space charge resides, Poisson's
equation is written
div
(I'.
grad J) = -4np
(4.18)
In terms of the x, y, Z coordinate system, and for the case where I'. is not dependent on the field components dJldx, dJldy or dJldz, the above equation is written
in terms of the partial field derivatives
(4.19)
Usually the series of partial second derivatives is represented by the symbol V2, "del squared" (the Laplace operator), so that
(4.20)
The mathematical operations required to obtain the Poisson equation can be
complex, depending on the degree of rigor involved in its derivation, but a simple treatment gives the required result by application of Gauss's theorem (Eq.
4.17) (General Ref. 1).
The idea that a charge is associated with an energy is less familiar than the
concept of field or potential due to a charge. As with all energy quantities, it is
necessary to specify with respect to what reference state the energy is measured.
In the case of a charge, the energy originates from the work associated with
building up that charge (e.g., q) by adding elements of charge Jq brought from
infinity. Hence the "energy of the charge" is the work done in the building up
of that charge from an initial value of zero to its final value q, with increments
of charge, Jq, transported from reference distance 00 where the potential due to
the charge built up at any stage in the charging process is zero.
Let Aq be the charge established at a given stage of the charging of the initially uncharged particle, where A is a fraction 0 < A < 1. The electrical work
done, oWe, in bringing up the next element of charge oq is (AqJq)la if a is the
radius of the particle being charged. The total work done, We' for transfer of all
the charges Jq (i.e., from A =0 to A = 1) is given by
77
q
W =f Aqdq
e
a
o
(4.21)
The integral is evaluated by regarding the fraction A. as being progressively increased from 0 to 1, so that the integration is made with respect to A. rather than
q, i.e., because,jq can be written as q.,j A.. Then
1
AqqdA. n 2 f
We= f -a-=~ AdA. =q2/2a
o
0
(4.22)
The same result may also be obtained by considering the charging energy
(1I2)C(11J)2 of a spherical particle treated as a conductor having a capacitance
C equal to its radius, a.
If the charging is carried out in a medium of dielectric constant, e.
(4.23)
A similar concept of "self-energy" applies to a capacitor which has an energy
of 1I2CV2 or 1I2C(I1Jf
W'=('8~)Ad
(4.24)
where E is the field 11J/d. Ad is the volume of dielectric in the capacitor so that
eE2/Sn: must be the electric pressure or really the internal tension associated
with the field E, which follows from considering the units of the left- and righthand sides of Eq. (4.24). Equation (4.24) also provides a way of evaluating the
self-energy of an ion by integrating the PdV work experienced by the dielectric
from the ion's periphery at a to 00. Thus
eW
=f. 4n:r2 . dr =q2/2ea
Sn:
~
(4.25)
78
Chapter 4
= a.,E
(4.26)
Induced dipole
moment fL i-et
Atom in field E
FIGURE 4.4. Electric polarization of an atom or molecular particle in a field leading to an induced dipole moment, j1.i.
79
This work, Wi' is done on the atom or molecule; hence the positive sign for
Wi' The induced dipole, fli, however, interacts with the field E that produced it
with an energy -fl iE or -aeE2. Hence the total energy of induced polarization is
-aeE212.
e,
e,
In a nonuniform field, a dipole not only tends to become aligned with the
field but to be translated down the field gradient to minimize its energy (i.e., to
maximize the numerical value of flE). The movement of the dipole continues
as E increases (e.g., near an ion where the field varies as ze/B?, or near an electrode surface in its double layer) until the dipole experiences some sufficient
contact repulsive force. This is the reason for the internal compressional force
near ions in a polar medium or in a double layer. This leads to "electrostriction"
or a decrease in volume of the dielectric medium near ions.
Chapter 4
80
Field direction
------------------------.~
(t)
ILl
1-'=
cose
a
Force -eE
Field E
Force ~eE
--b
FIGURE 4.5. Interaction of a permanent dipole of moment, fl., with an electric field E: (a) Interaction energy with the field, fl.E cos O. (b) Torque resulting in orientation.
Field direction
81
Ii)
(b)
e
Field
Field direction E9
Field direction 6)
Field E
(0 )
Enercn +p.E
FIGURE 4.6.
Energy - jl.E
!lR.
4.10.1. Dielectrics
82
Chapter 4
I~C"
fJ-/1/9~
:
Ct
Cf
ce
(4.28a)
flCHP
(4.28b)
and
The Sp3 geometry of each molecule is such that 3flcCl cos t/J =flcc! and 3flcH
cos t/J = flCH. Rence the molecules should have the same overall dipole moment.
This is not the case: flCHC! 3 = 1.1 D; flCH 3C! = 1.85 D [lD (debye) = 10- 18 esu].
The discrepancy could arise because the tetrahedral angle is not exactly maintained in chloroform due to repulsion of the CI atoms and/or because there is
some depolarization among the C-CI bonds, making them have lower bond moments than the single C-CI bond in CRCI 3; the latter effect is the main one.
In flexible polyatomic molecules, such as conducting polymers, the overall
dipole moment will depend on the conformation the molecule may adopt. For
83
example, dioxane may exist in a boat (I) or chair (II) conformation. Structure I
has the higher dipole moment and will be preferred in the high fields near ions
or in double layers at electrode surfaces; it may also be preferred due to intermolecular H bonding by water.
84
Chapter 4
Aq
dG=-dq
C
(4.29)
G = f dG = f
~ . qdA
(4.30)
=~fA.dA
(4.31)
1
1
=_q2/C=- CV 2
2
2
(4.32)
recalling that always C = q/V. C contains the dielectric constant, e (>1) of the
dielectric medium between the plates; e = 1 for vacuum, hence CCe) > CCe = 1).
This implies that a greater energy is stored in a capacitor constructed with a dielectric having e > 1 than in one in a vacuum by the factor e. The extra energy
over that in a vacuum when e > 1 and geometrical dimensions of the capacitor
are the same is stored within the dielectric medium as an energy ofpolarization
of the dielectric, as explained in Section 4.4.7 concerning polarization of solvent
by an ionic charge, and especially in Chapter 5.
It is interesting that the calculation of electrical energy stored in a charged
capacitor can be conducted (as above) in a way closely similar to the derivation
of the "self-energy" of the charge on a chemical ion, i, due to Born l (1920). This
self-energy (a Gibbs electrostatic energy) is given (Eq. 4.23) by
(4.33)
(4.34)
where ri is the radius of the ion in vacuum, rie is its effective radius in the dielectric solvent medium, and Zie is its charge. Note that these G i quantities depend on half the square of the ionic charge, Zie, like the result in q2/2C for energy
in the case of capacitor charging. Then, by analogy between these two cases, it
is seen that C is equivalent to ri for a spherical ion. That is, the capacity of a conducting sphere is equal to its radius, a principle in classical electrostatics.
Another aspect of electrochemical and physicochemical interest is that
since the stored energy of a capacitor is a Gibbs energy, it has an entropy com-
85
ponent S == -dGldT, the temperature derivative of that energy. Other things remaining constant, the only quantity involving temperature dependence is the dielectric constant of the dielectric medium determining C in the relation C =
Au.old (see Eq. 4.14 and the text following that equation). Hence the entropy of
charging is given (for B > 1) by
(4.35)
(4.36)
(4.37)
Since de/dT is normally a negative quantity for most if not all dielectric materials, it is seen that the entropy of charging a capacitor having B > 1 is negative.
This is due to the electric polarization of the dielectric that is established in the
field between the plates of the capacitor. It is analogous to the entropy of magnetic polarization in magnetized paramagnetic or ferromagnetic metals. Analogously also, if one could switch off the electric polarization instantaneously,
then an adiabatic dielectric depolarization cooling effect would arise which
might have practical technological value. However, in a double-layer electrochemical type of capacitor, such an effect would tend to be reversed owing to
the internal joule heating effect associated with the discharging current required
to transport accumulated ionic charges away from the two electrode interfaces
into solution when the two double layers were discharged, i.e., when V and q ~
O. In a double-layer capacitor, this entropy is associated mainly with solvent dipole orientation in the double layer and coupled ion accumulation at the capacitor plate interfaces.
The distribution of charge on conducting capacitor plates is another matter
of interest: the distribution tends to be uniform except for edge effects. The electronic charge () on the conducting plates is a delocalized charge of conductionband electrons: an excess or deficiency of surface charge density compared with
the situation at two uncharged plates.
At the surfaces of an electrochemical double-layer type of capacitor, the
situation can be somewhat different because the charges in the solution, complementary to the plus or minus electronic charge densities, are composed of
discretely charged ions of finite sizes; also, the distance of closest approach of
86
Chapter 4
Chapter 5
5.1. INTRODUCTION
88
Chapter 5
(5.1)
(5.2)
or
where Ko is the permittivity of free space (see later discussion), sometimes symbolized by BO' In the presence of a dielectric medium between the plates, C is
increased (cf. Eq. 4.14) to
C=Asl4nd
(5.3)
89
or
C=AK/d
(5.4)
C =AK/d =A Kof./d
(5.6)
CIA = Koe!d
(5.7)
90
ChapterS
proximately equal to the square of the refractive index, arising from electronic
and atomic polarization. For polar but non-H-bonded liquids, it has values
around 6 to 40, while for H-bonded liquids, it can attain much higher values,
e.g., 78 for water or 113 for HCN at ordinary temperatures. Some oxide materials can have relatively large t: values of several hundred. This is important for
fabrication of so-called "electrolytic" capacitors in which a thin oxide film is the
dielectric.
It is seen that C scales with t:, other conditions being constant. This is related to the involvement ofthe dielectric constant in Coulomb's law, where the
repulsive or attractive force between two charges, ql and q2, in a vacuum (Chapter 4),
(5.8)
or
(5.11)
i.e., they are attenuated by the factor lit: if they are situated in a dielectric medium having t: > 1. Similarly, capacitance of a given geometrical configuration
of plate electrodes is enhanced by the factor t: (> 1) over that for a vacuum (t: =
1) and correspondingly more (free) energy can be stored for a given applied
voltage, Ll V, when t: > 1, the energy being 112 CCLl V)2 (Chapter 2).
For two capacitors having identical geometries, the difference of stored
free energy for t: > 1 (real dielectric) and t: = 1 (vacuum) is
(5.12)
91
92
Chapter 5
inverse "log" law in time usually applies or a "parabolic" law for some conditions; see Ref. 3.)
The situation with double-layer capacitors is different since at each of the
capacitor electrodes a double layer exists across which there is a drop in potential (see Fig. 6.3, Chapter 6) between the electrode plate and the electrolyte
solution; the dielectric of this double molecular capacitor is composed of only
one or two molecular layers of solvent, including solvated anions or cations in
that solvent layer, and it is polarized by the field associated with the differences
in potential between the metal and solution, especially that part which falls
across the first 0.3 to 0.6 nm (Helmholtz layer) of the solution near the electrode
surface. Interfacial dielectric properties of this very thin film have to be formulated (see Chapter 7) to account for the magnitude of the double-layer capacitance (16 to 50 JiF cm- 2) and its dependence on electrode potential, the sign of
the charge of the electrode plate, and the nature of the solvent and the electrolyte
solute.
Some of the principal aspects of the molecular and ionic modeling of the
double layer and its effective average dielectric properties are given in Chapters
6 and 7. In order to provide the background for the discussion in those Chapters,
we return now to the more general theory of bulk dielectrics in order to explain
the concepts of polarization and polarizability more completely, and their relation to experimentally measurable dielectric constants and molar polarizabilities. These are fundamental matters that lie at the bases of capacitance behavior
in electrostatics, and in the phenomenology of double-layer capacitance and
electrical behavior of supercapacitors.
Apart from the invention and use of the Leyden jar (Chapter 1), Faraday
was the first to consider the state of dielectrics in charged capacitors. In fact it
was in correspondence with the polymath, Whewell of Cambridge, that the term
"dielectric" was suggested (December 1836) and it has remained in use ever
since. Faraday! initiated ideas of "polarization" due to electric fields and recognized the induction of electric charges,with polarization being induced by one
polarized or polar particle into or onto another. He also recognized concepts
such as specific inductive capacity (8), lines of inductive force, and the strain introduced in an electrified dielectric medium by an imposed field. Thus Faraday
recognized some of the properties of dielectrics and the nature of electric inductive polarization before the middle of the nineteenth century. Of course, it was
many years later before these ideas became more exactly and mathematically
formulated by workers such as Mosotti, Clausius, Coulomb, Frohlich, Debye,
and Kirkwood in terms of molecular polar models. Unfortunately, as is well
93
known, 1 despite his depth of scientific perception and his inventiveness, Faraday never developed his concepts and interpretations of physical phenomena in
mathematical terms.
The electrostatic behavior of two identical configurations of capacitor
plates is considered here. One capacitor [Fig. 5.1(a)] has a vacuum dielectric,
the other [Fig. 5.1 (b)] a real dielectric medium having 8 > 1. An external field
Eo is applied and is operative between the plates generating or arising from plus
and minus charge densities on the plates, equally and oppositely. For a given
voltage difference, ~ V, between the plates or a corresponding field ~ Vld, charge
densities 0'1 qtfA or 0'2 q21A appear, respectively, on the plates of the two
capacitors. 0'1 and 0'2 differ on account of the difference in 8 for the two cases;
81 = 1; 82 > 1. This difference corresponds to an induced charge density p,
which appears at the plates, changing the original a value [Fig. 5.1(b)]. When
the dielectric medium is present [Fig. 5.1(b)], the field between the plates is reduced from the vacuum value, Eo, to a diminished value E.
(0)
Field Eo +
.-
Field Eo
Field E~
+
+
Vacuum
...
+CT
-CT
Dielectric
(b)
/ ..........
....---
/--
External field
'......--"......,, Field EE o)
,,..---"',/" ""
-"""-,
--,- -,
Eo
1/--
. . . . . /,/"
+p
,
I
Eo
-p
FIGURE 5.1. Fields E between charged capacitor plates: (a) in vacuum; (b) in the presence of a
dielectrically polarizable medium. Original charge density a; induced charge density p.
ChapterS
94
+
+
+
+
+
+
+
b
+
+
+
1 cm2
+ - -.....~E---...
FIGURE 5.2. Schematic diagram of induction of polarity in a dielectric medium by an externally
applied field. (a) Induced charges or orientation of polar molecules by an external field Eo in capacitor. (b) Resulting polarization vector, P, corresponding to induction of polarity, -, +, on the
faces of a centimeter cube.
95
The origins of the polarization thus induced in the dielectric are fourfold:
1. The first is a redistribution of electronic charge density (Fig. 5.3),
mainly of valence electrons, within the individual atoms or molecules
of the dielectric material; it is characterized by the so-called "electronic
polarizability," denoted by a, of the molecules. This a depends on the
number of polarizable electrons and thus on the volume of the molecule
(a has the units of cm3 mol-I, on the order of 10-24 ) and on the type(s)
of orbitals involved in the molecules, e.g., (J or 7r. a for 7r orbitals is
greater than a for (J orbitals; for example, with benzonitrile C6H5 . C=N,
the polarizability is anisotropic, along rather than across the molecule.
2. There is a slight change in the atom positions or bond lengths of the
molecule, and hence its polarity. This is called "atom polarization."
3. Third, for intrinsically polar molecules such as H 20, HCI, HF, and
CH3CN, a larger degree of polarization per unit volume is set up within
the dielectric due to an orientational response (Fig. 5.4) of the already
existing dipoles of the molecules of the dielectric medium. This is a statistical effect, evaluated in important work by Debye and by Langevin,4
and is the main source of dielectric polarization in fluids of intrinsically
polar molecules. Since for ordinary fields of e.g., 10-100 V cm- 1 there
is a small but significant net average orientation of dipoles, this has an
influence on the induced electronic polarization in the molecules of the
dielectric, depending in an average way on their momentary orientations in the field disturbed by thermal fluctuations in the bulk fluid. This
effect does not of course arise with atomic fluids or 3-dimensionally
symmetric molecules (e.g., CH4 , CCI 4 ) except to a small extent through
dipole induction by the experimental field. For intrinsically polar molecules, polarization of the dipole orientation is normally substantially
larger than the polarizations of types (1) and (2).
Field E
fL~
8+-8-
.....
.
..
..
fJ-i = a E
FIGURE 5.3. Redistribution of electronic charge density in a polarizable molecule in an electric
field (induced polarity).
96
ChapterS
Average moment.. ~
,,/ \
-- -
-----
, . Field
-,/.--
-IfI'
-..
(0)
~E<kT
(Normal polarization
experiment)
-..
E
---..
-----
Field
(b)
~E.
kT
( Dielectric saturation
e.o. at ions or
electrode interfaces)
FIGURE 5.4. (a) Orientational polarization of intrinsically polar molecules in a small electric
field. (Effect is additional to the induced electronic polarization in Fig. 5.3.) (b) Approach to full
orientation of solvent dipoles in the very strong field (p.FJkT 1) of the double layer (dielectric
saturation condition).
4. Finally, when there is substantial intermolecular electrostatic dipoledipole interaction between molecules of a dielectric material (e.g.,
CH3N0 2 , (CH3hCO), and especially when the molecules are H-bonded
as in liquid HF, H 20, HCOOH, CH3 COOH, H-CONHz, esters, propylene carbonate, etc., the process of orientational polarization becomes a
cooperative one between molecules in small assemblies, giving rise to
much larger polarizations per unit volume for a given field. These are
the media that exhibit high dielectric constants in the range 50 to 113 at
ordinary temperatures, e.g., CH3OH, H 20, liquid HCN, liquid HF, and
HCOOH.
From the point of view of choice of dielectric solvents for double-layer capacitors (see Chapters 8 and 13), the use of H-bonded solvents, though giving
high e values, has the disadvantage of relatively low decomposition voltages in
the range 1.23-1.5 V and rather high melting points. Higher operating voltages,
per cell, for double-layer capacitors therefore require so-called "aprotic"
nonaqueous solvents (Chapter 8) for their electrolytes (e.g., tetraalkylammonium salts) since such solvents, which do not have electrolytically active
(i.e., electrolytically depositable) H or other atoms, have substantially higher
decomposition voltages, e.g., 3.4 to 4.0 V. Useful practical examples are acetonitrile, benzonitrile, N,N-dimethylformamide, propylene carbonate, dimethoxyethane, and the tetrahydrofurans. Such solvents are also preferred
media for construction of nonaqueous lithium battery systems operating at 33.S V per cell. Here the requirement is of course that the very base metal, Li not
discharge H2 from any "active" H atoms in functional groups of the solvent
97
molecule. "Active" H atoms in a solvent are usually those that are acidic and
participate in intermolecular hydrogen bonding. However, as is shown in Table
8.1, Chapter 8, a number of aprotic solvents do have quite high dielectric constants (20-30), but since they are not H-bonded do not exhibit the strong cooperative polarization behavior that is found for such associated liquids as H 20,
CH30H, CH3 COOH, etc. Hence they have lower dielectric constants.
In terms of the diagrams in Figs. 5.2 and 5.4, a vector quantity P is defined
as the dielectric polarization, which is equivalent to a dipole moment per unit
volume [Fig. 5.2(b)]. P is thus equivalent to a charge p (per square centimeter)
induced at the plates of the capacitor multiplied by a unit length of 1 cm, i.e., a
dipole moment per unit (1 cm3 ) volume. p is sometimes written as ai, the induced a diminishing the original a value.
In the absence of a polarizable medium (8 = 1), the field due to a surface
charge density a per square centimeter is [Fig. 5.1(a)] Eo =4na (Gauss's relation), which is also referred to as the dielectric displacement, D. Owing to the
presence of the dielectric, the field Eo, which in the absence of a dielectric is due
to a, becoming reduced (see Coulomb's law, E = -qI8?-) by 4np [which is related to the induced charge density p, Fig. 5.1(b)], so that
E=4n(a-p)
or
E=Eo-4np
(5.13)
(5.14)
so that
8
(5.15)
98
ChapterS
The induced dipole moment, Ili' is proportional to the field for low or moderate fields (i.e., Ili = aE where a is the distortional polarizability), mainly owing to redistribution of electron density in the molecules. It has the dimensions
of volume and magnitude on the order of 10- 24, i.e., - (10- 8)3 or angstroms
cubed (nm3 x 10- 3).
For a dilute gas, the effective polarizing field is the external field, Eo. If M
is its molecular weight and p its density, MIp is the molar volume (here p is the
density) or MINAP is the molecular volume (NA =Avogadro's or Loschmidt's
number) and the induced polarization, Pi is
p. =
I -
molecules
dipole moment
x ---'~---unit volume
molecule
NAP
=-(a~)
(5.16)
(5.17)
Then from Eq. (5.18) defining Ili' and using Eq. (5.15)
(e - 1) Mlp = 4nNA a
(5.19)
If the dielectric is not a gas but a condensed phase in which molecular interactions are considerable and short range, the influence of the molecular surroundings on a given dielectric molecule has to be taken into account: a given
dipole in a hypothetical molecular-sized cavity in the medium induces charge
density on the walls of that cavity which then generates an opposing (induced)
local field, the reaction field. Then the effective field Ef becomes related to E
and P in a relatively complex way, leading to
(5.20)
and the induced moment becomes
,ui=a(E+
~n p)
99
(5.21)
~ = coth (,uE)_
,u
l kT
kT
,uE
(5.23)
Ii =(,u2/3kT) E
(5.24)
100
Chapter 5
(5.25)
a relation derived by Debye with the aid of the Langevin function for "ji/f1 above
(Eq.5.23).
In the case of dielectric liquids composed of strongly interacting dipoles (e.g., CH3 N0 2 , H 20, CH3 0H, amides, and amino acids in zwitterion
form), high dielectric constants are manifested. Water has an B value of 78
at 298 K and HCN 113 at that temperature; these are hydrogen-bonded or socalled "associated" liquids where the orientation of a given molecular dipole
drags neighboring dipoles in the same direction, especially with hydrogenbonded solvents. Such solvents also have relatively high viscosities and long
dielectric relaxation times, on the order of 1O- 11 s, in comparison with 10- 12
- 5 X 10- 13 S for weakly interacting solvent molecules. Similar effects occur
with other strongly dipolar but not hydrogen-bonded solvents, but the resulting B values are rather lower, e.g., 35.9 for CH3 N0 2 , and 38.0 for CH 3 CN
and N,N-dimethylformamide.
This cooperative orientation effect enhances the polarization P for a given
field, giving rise to large B values. Some solid ceramic materials also have large
B values, but for different reasons.
Several attempts have been made, especially by Kirkwood and co-workers7,8 to develop a theory for high dielectric constant solvents based on a treatment of cooperative dipole orientation polarization. The effect is crudely
analogous to coupled orientation of iron filings when a magnet is placed below them when they are spread on a sheet of thin paper. However, such a
model lacks the important factor of thermal disorientation, which applies to
molecular interactions and orientations in any liquid at finite temperature
[see the Debye treatment utilizing the Langevin function (Eq. 5.23) referred
to earlier].
The problem is again to express the polarization resulting from the applied
field as a function of that field, but taking into account the cooperative nature
of interactive orientations of dipoles. In the case of strongly hydrogen-bonded
fluids such as water or hydrogen fluoride, the )O-H- - - 0( or F-H---F- hydrogen
bonds remain strongly interactive for displacements of the hydrogen bond from
linearity of up to ca. 35. In the case of water, the interactions are 3-dimensional
(fourfold) or, for HF, they are mainly quasi-linear (twofold); hence the substantially
101
g = 1 + z < cos
(Jij>
(5.27)
where ()ij depends only on the coupled orientations of the i th andjth molecules,
each with z nearest neighbors.
Another relation is the Kirkwood-Frohlich equation:
(5.28)
where B~ is a dielectric constant of a supposed continuum in which the permanent dipoles (f.ld) are embedded; Vm is the molar volume, M/p; and p denotes
density.
The case of water as a dielectric has, of course, special significance for
many electrochemical systems, including the double layer at carbon and metaUwater electrode interfaces. For water it is found experimentally that the Kirkwood g factor decreases with increasing temperature from a value of 2.75 at ooe
(in liquid water) to 2.49 at 83C. This follows the general properties of water,
such as viscosity and X-ray diffraction structure, which both decrease with temperature. Theoretical values were calculated in a paper by Kirkwood and Oster. 8
Assuming free rotation about the H bond in water (not a realistic condition!), the previous equation (5.27) leads to
g = I + z cos 2(J12
(5.29)
102
Chapter 5
where z is the average number of neighbors in the first coordination shell and e
is the HOH bond angle in water, about 105. The value of z can be derived from
X-ray diffraction data. Other general and more detailed aspects of dielectric polarization are treated in Refs. 9, 10, and 11, the monograph volumes by Frohlich,
Smyth, and Bottcher.
The correlation factor will also be important in the orientation polarization and its contribution to the local dielectric constant of the interphasial
double layer where solvent dipole orientation occurs in response to that part
of the metal-solution potential difference at an electrode/solution interphase
across the compact Helmholtz region. The interaction energy factor in the
treatment of interphasial solvent dipole orientation is of major importance in
defining the approach to dielectric saturation in the double layer at electrodes as
treated in a paper by Bockris, Devanathan, and Miiller I2 and in several recent
papers by Marshall and Conway.13,14 In particular, some minimum strength of
dipolar interaction is required to avoid a "catastrophe"J3,I4 in the values calculated for the contribution of dipole orientation to the double-layer capacitance
arising from the reciprocal dependence of the orientation surface-potential contribution I2 on the electrode surface charge density at a polarized electrode. This
difficulty arises in the two-state orientation model (Ising-type model in magnetic polarization) first considered for the double layer. 12 In more sophisticated
three-state or cluster models for interphasial water polarization, these difficulties are avoided (see Chapter 7).
103
Using the treatment by Booth l5 (see Chapter 7), Conwayl6, Bockris and
Ammar 17, and Grahame l8 made the first calculations of the local dielectric coefficient of water in the diffuse and compact parts of the double layer at electrodes in aqueous media. Much more is now known about the state of the water
dielectric in the double layer, e.g., see Chapter 7 and Ref. 19. The calculations
indicate that the e for water is lowered appreciably from its normal bulk (zerofield) value of78 at 298 K in the diffuse layer and more so in the compact Helmholtz layer, where its value can descend to about 10-20, depending on the
electrode/solution potential difference. More recent estimates 10 indicate a minimum value of around 6. The various treatments of high-field effects on the dielectric constant of water, including experimental attempts to detect a lowering
at high fields (but much less than those in double layers), are summarized by
Smyth in Ref. 9, Chapter 3, p. 88.
The types of calculation referred to here are really based on bulk models
of the water dielectric. These cannot, however, be expected to reliably represent
the behavior of a lattice of water molecules only two or three molecular diameters wide at the electrode interface, not to mention the local solvational influence of the ions that also occupy an appreciable volume fraction of the compact
layer comprising this thin film of interphasial solvent molecules. In fact, depending on the electrolyte and the electrode/solution potential difference, some
30-40% of the interphasial water molecules can be electrostatically influenced
by the high radial fields of neighboring cations or anions. 2o Under such conditions, the bulk dielectric saturation theories of high-field polarization must be
Onset of dielectric
saturation
!z
C3
ii:
u..
8
o
cr
~
w
---
~~~t~nt.-rr-
------------------------5
'
ChapterS
104
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
L. Pearce Williams, Michael Faraday: A Biography, Chapman and Hall, London (1965).
B. E. Conway, Electrochernical Data, Elsevier, Amsterdam (1952).
N. F. Mott and B. Cabrera, Rept. Prog. Physics, 12, 163 (1949).
P. Debye, Phys. Zeit., 13, 97 (1912); see also P. Langevin, 1. Phys., 4 (4), 678 (1905).
O. F. Mosotti, Mern. di Mathern. e Fisica, Modena, 24 (II) 49 (1850).
R. Clausius, Die rneclwnische Wiinne theorie /I, p. 62, Braunschweig (1879).
J. G. Kirkwood, 1. Chern. Phys., 7, 911 (1936).
J. G. Kirkwood and G. Oster, 1. Chern. Phys., 11, 175 (1943).
C. P. Smyth, Dielectric Behavior and Structure, McGraw-Hill, New York (1938).
H. Frohlich, Theory of Dielectrics, Oxford University Press, Oxford (1949).
C. J. F. Bottcher, The Theory of Electric Polarization, Elsevier, Amsterdam (1952).
1. O'M. Bockris, M. A. V. Devanathan, and K. Miiller, Proc. Roy. Soc., Lond., A274, 55
(1963).
S. Marshall and B. E. Conway, 1. Electroanal. Chern., 337, 1 (1992).
S. Marshall and B. E. Conway, 1. Electroanal. Chern., 337,19 (1992).
F. Booth, 1. Chern. Phys., 19, 391, 327, 1451 (1951).
B. E. Conway, Ph.D. thesis, University of London (1949).
B. E. Conway, J. O'M. Bockris, and I. A. Ammar, Trans. Faraday Soc., 47, 756 (1951).
D. C. Grahame, 1. Chern. Phys., 18, 903 (1950).
Symposium on the Structure of the Electrified Interface, D. J. Henderson and O. R. Melroy,
eds., Electrochirn. Acta, 36, 1659-1894 (1991).
B. E. Conway, 1. Electroanal. Chern., 123, 81 (1981).
G. M. Torrie and G. N. Patey, Electrochirn. Acta, 36, 1677 (1991).
A. M. Brodsky, M. Watanabe, and W. P. Reinhardt, Electrochirn. Acta, 36, 1695 (1991).
Chapter 6
6.1. INTRODUCTION
106
ChapterS
negatively) with respect to the electrolyte solution, with a separator (as in a battery cell) usually being arranged between the two electrodes. The two-electrode,
two-interface system in a single capacitor cell is illustrated in Fig. 6.1(a).
While the capacitative behavior of individual electrode interfaces can be
studied experimentally, the electrical circuitry must always include a second
(counter) electrode and also preferably a separate reference electrode with respect to which the potential measurements can be independently scaled and controlled. Similarly, a double-layer capacitor device must always involve (at least)
two electrodes. This is an unavoidable requirement of all electrode interface
measurements and all constructions of electrochemical devices, e.g., capacitors
and battery cells. If the electrochemical behavior of each of the pair of electrodes
is to be examined, then a third, reference, electrode is required.
In the charged condition there are two interphasial drops in potential across
the capacitor cell, one across each double layer [Fig. 6. 1(a)]. Upon discharge,
there is also a current-dependent, ohmic IR potential drop within the solution,
and the opposite on recharge [Fig. 6.1(b)].
Since one of the main technological directions of electrochemical capacitor development is in the field of the double-layer type of capacitor, this Chapter
is devoted to a somewhat detailed description of the structure of the double layer
and the origin of the capacitance associated with it. A detailed understanding of
the properties and structures of double layers is essential for a full understanding
of the operation of double-layer capacitors and the capacitance values that are
achievable per square centimeter, or per cubic centimeter of active porous material in various systems. This will also require the mathematical representation
of the properties of the double layer related to the nature of the electrolyte and
solvent on the solution side of the electrode-materiaVsolution interface and to
the properties of the electrode itself. The theoretical treatment and modeling of
the double layer are discussed in Chapter 7.
Determinations of the double-layer capacitance and its potential dependence provide detailed information for examining the behavior and structure of
electrified interfaces. This is mainly because the dependence of the charge, q,
held on each side of the electrode interface (electron density on the metal and
ion density on the solution side) on the measured electrode potential, E (which
is equivalent to the Fermi level of electron energy states in the metal) is directly
determinable as a differential quantity, Cd!> dq/dE; such types of quantities always give more resolved information than corresponding integral relations, e.g.,
total charge q plotted vs. the electrode potential E at which it has been accumulated, or q divided by E at a given electrode potential. That is why cyclic voltammetry and ac impedance measurements are especially valuable and
preferred in studies of the double layer and the resulting electrochemical capacitor devices; they both give differential information. A typical relation between
107
Electrolyte
-~ ++ +-+
-~ +- +-+
-~ ++ +-+
+ -+ - +-~ + -+ - +-~ + -+ - +-
-:/
Porous
electrode
Separator
Porous
electrode
Configuration of an electrochemical
capacitor requiring two electrodes
and thus two double-layers
b
Electrolyte
,/
~
+
+
+
+
+
+
+
+
+
Separator
108
Chapter 6
f( y)
J,/f(dY/dX)
\
\
\
"....
FIGURE 6.2. The relation between arbitrary integral signal curve x,y, (e.g., for a double-layer
charge as a function of potential) and its differential coefficient, dyldx (e.g., for differential capacitance).
an arbitrary integral curve and its differential, illustrating this point, is shown
schematically in Fig. 6.2.
It is appropriate first to describe the structure of the double layer, i.e., of
the interphasial region of the metal/solution boundary. The term interphasial region! is deliberately chosen over the word "interfacial" (see Chapter 3, section
3.8) since the boundary region is really a 3-dimensional one, albeit very thin,
some 0.3 to 0.5 nm in thickness, rather than the 2-dimensional one implied if
the term "interfacial" were used. This means that the concepts of 3-dimensional
phases can be used to describe the double layer, such as surface concentration
or surface excesses, e.g., of ions of the electrolyte and solvent molecules.!,2
It is logical to describe these models in the order in which they were historically proposed in the literature, which is also the order of their progressive
approach toward a correct description of the structure of the interphase at electrode surfaces.
The concept of a double layer corresponds to a model consisting of two array layers of opposite charges, separated by a small distance having atomic dimensions, and facing each other, as on the plates of a two-plate hardware
capacitor. This model was adopted by von Helmholtz3 to describe his perception
of the distribution of opposite charges, quasi-2-dimensionally, first at the interface of colloidal particles. It is illustrated in Fig. 6.3(a), which shows its compact
structure, and is referred to as the Helmholtz double-layer mode1. 3,4
..
..
..
109
+ _.+ -+
+
+
"'S
IT..
(0)
(b)
+ -
fs
01 FFUSE LAYER
0--
=8-
HELMHOLTZ LAYER
(c)
FIGURE 6.3. Models of the double layer: (a) Helmholtz model, (b) Gouy point-charge model
(specific charges a per unit area as indicated for anions and cations as an example), (c) Stem model
for finite ion size with thermal distribution, combining Helmholtz and Gouy models.
In the original model for colloid interfaces, the charges on the surface side
of the double layer arise either from acid-base ionization, as with proteins or
polyelectrolytes, or on account of the adsorption of ions, as at lyophobic colloids. On the solution side of the double layer, counterions of opposite sign of
charge accumulate to balance the charge on the colloid, forming a double-layer
array of positive and negative charges.
The Helmholtz model was later adapted to the case of electrode interfaces
where, on the metal side, a controllable surface density of excess negative or
positive charge can arise that corresponds to an excess or deficiency of electron
charges of the delocalized electron plasma of the metal. Owing to the high free
electron (e) density in the metal (approximately 1 e per atom), any net charge
density of electrons at the surface is strongly screened, so the gradient of electron density at a charged metal interface is highly localized over a distance of
only 0.05 to 0.2 nm, the so-called Thomas-Fermi screening distance (Fig. 6.4).
Because the wave function amplitudes of the conduction-band electrons retain
significant but diminishing magnitudes outside the formal electrode surface
plane, there is significant spillover of electron density into the double layer on
the solution side of the interface5,6 and the effect is potential dependent.
In the case of p- or n-semiconductors, the charge-carrier (hole or electron)
densities are, however, very much smaller than in metals by a factor varying
from about 10-4 to 10-15. As a consequence, there is a distribution of the charge
carriers away from the interface (Fig. 6.5) but extending into the bulk of the
semiconductor over a relatively large distance that is inversely related to the
charge-carrier density (cf. the situation for ion distribution in an electrolyte in
the solution side of the interface, Section 6.3). This distribution of charge carriers within semiconductors, near their interfaces, follows mathematically exactly
Chapter 6
110
Thomas - Fermi
screening distance
..
Appro)!.. 1e
per atom in
bulk
---
SOLUTION
METAL
\,/Electran overspill
.... - - - - to zero
~Nominol surface
plane
t
Locus of
atomic
nuclei
0
~
1
cr
-----
~9>
n space - charge
/
region
SEMICONDUCTOR
e
IONIC SOLUTION
FIGURE 6.5. A space charge distribution of charge carriers within an n-type semiconductor near
its surface that is like ion distribution in the diffuse layer in solution.
111
the same form as for ion distribution in the diffuse layer on the solution side
(Gouy-Chapman theory for dilute ionic solutions).
Some time after von Helmholtz's model was proposed, it became realized
that ions on the solution side of the double layer would not remain static in a
compact array as in Fig. 6.3(a) but would be subject to the effects of thermal
fluctuation 7 according to the Boltzmann principle. 8 This latter effect would depend on the extent to which the electrostatic energy Ue (together with any
chemisorption energy Uc ) of the ions' interactions with the charged metal surface exceeded, or were exceeded by, the average thermal energy, kT, at temperature, T, K, i.e., the ratio (Ue + UJlkT.
Gouy7 introduced this thermal fluctuation factor into a modified representation of the double layer in which the counterions conjugate to the metal
surface's electron charge were envisaged as a 3-dimensional diffusely distributed population of cations and anions [Fig. 6.3(b)] of the electrolyte having a
net charge density equal and opposite to the virtually 2-dimensional electron excess or deficit charge on the metal surface.
In this model, the ions were assumed to be point charges. Historically, this
was an important restriction since it led to a failure of Gouy' s model on account
of (1) an incorrect potential profile and local field near the electrode surface and
(2) consequently a too-large capacitance being predicted, that quantity being defined as the rate of change of net ionic charge on the solution side with the
change of metal-solution potential difference across the interphase. The interphasial capacitance associated with this model is commonly referred to as the
"diffuse" double-layer capacitance.
A full mathematical treatment of the Gouy diffuse-layer model was given
in some detail by Chapman in 1913 9 (see Section 6.3), based on the combined
application of Boltzmann's energy distribution equationS and Poisson's equation lO for the relation between ionic space charge density in the interphasial region to the second derivative of electric potential, IJI, with respect to distance
from the electrode surface. It is interesting to note that the mathematics and principles used by Chapman anticipated the approach taken by Debye and Huckel
in 1923 11 in determining ion distribution in three dimensions around a given ion
in their treatment of activity coefficients and conductance of electrolytes. Later
it was used by Onsager,12 in an improved treatment of the conductivity of electrolytes.
In both Chapman's and Debye and Huckel' s treatments of ionic charge distribution, the key equation resulting from the combination of Boltzmann's energy distribution function and Poisson's electrostatic equation lO has been
referred to as the "Poisson-Boltzmann" equation. It is also utilized in the treatment of band profiles and space charge effects in semiconductors. 13
The serious problem with the Gouy-Chapman treatment, overestimation
of the double-layer capacitance, was overcome by Stern in 1924 14 in the next
112
Chapter 6
stage of development of the theory of double layers. In his model and calculations it was recognized that the inner region of the ion distribution could be
treated in terms of an adsorption process according to Langmuir's adsorption
isotherm, and the region beyond this inner layer, into the solution, could be validly treated in terms of a diffuse region of distributed ionic charge [Fig. 6.3(b)]
as treated by Gouy 7 and by Chapman. 9 In addition, if the ions were recognized
as having finite size, including the annular thickness of their hydration shells
(their so-called Gurney cosphere radii 15 ), it was easy to define a geometrical
limit to the compact region of adsorption of ions at the electrode surface [Fig.
6.3(c)]. This is taken to correspond to a Helmholtz type of compact double layer
having a capacitance CH , while the remaining ionic charge density beyond this
compact ion array is referred to as the "diffuse" region of the double layer, having a capacitance Cdiff . Cdiff and CH are conjugate components of the overall
double-layer capacitance, Cd], related by the equation
1
Cd!
1
CH
1
Cdiff
-=-+--
(6.1)
-----II I
I 1-1- -
On account of the reciprocal form of the terms of Eq. (6.1), it will be seen
that the Cd! will be determined by the smaller of the two components, CH and
Cdiff . This is of considerable importance in determining the properties of the
double layer and its capacitance as a function of electrode potential and ionic
concentration of the solution.
The original paper by Stern!4 in Zeitschrift for Elektrochemie (1924) is
somewhat obscurely written, but Parsons, in an important article in 1954,4 presented a much clearer version of this treatment in which the limit for a distinction between the Helmholtz compact layer and the diffuse layer beyond it [Fig.
6.3(c)] can be understood in terms of the distance of closest approach of counter
anions or counter cations to the metal electrode surface.
By introducing a distance of closest approach of finite-sized ions and thus
geometrically defining a compact Helmholtz inner region of the double layer,
the problem of a far too high capacitance that arises in the Gouy-Chapman
treatment is automatically avoided. This difficulty arises since the capacitance
of two separated arrays of charges increases inversely as their separation distance, so very large capacitance values would arise in the limit of infinitesimally
small (point charge) ions very closely approaching the electrode surface.
113
The Stern theory of the double layer remained a good basis for general interpretations of electrode interface phenomena, including double-layer effects
in electrode kinetics 16 until the detailed work of Grahame in the 1940s on the
double layer capacitance at the mercury electrode in aqueous electrolyte solutions reported in various papers, particularly in the seminal source review paper
in Chemical Reviews in 1947.17 Its semicentenary was celebrated at a special anniversary Electrochemical Society symposium in May 1997.
Grahame's work emphasized the great significance of the specificity of
double-layer capacitance behavior at Hg to the nature of the cations and anions
of the electrolyte, particularly the size, polarizability, and electron-pair donor
properties of the anions of the electrolyte.
This led Grahame to make an important distinction between an inner and
an outer Helmholtz layer in the interphase which correspond to the different distances of closest approach that can arise for anions vis a vis cations at the electrode surface. This difference of distance of closest approach is mainly caused
by the fact that most common cations are smaller than common anions (Table
6.1) and retain solvation shells due to strong ion-solvent dipole interaction.1O18.23 Thus, the Gfahame model (Fig. 6.6) consists of three distinguishable
regions: the inner Helmholtz layer. the outer Helmholtz layer, and always a diffuse ion distribution region. At extremes of polarization (i.e., for high positive
or high negative charge densities), one or the other of the Helmholtz layer regions dominates, with a population of anions or cations, corresponding to such
polarizations. Because anion distances of closest approach are usually smaller
than hydrated cation distances of closest approach, the inner layer capacitance
at positively charged electrode surfaces is usually about twice that at a corresponding negatively charged surface, (16-25 f.1,F cm- 2), though this depends on
the metal and the ions of the electrolyte, and the solvent. These aspects of double-layer capacitance behavior are of great significance for understanding the
properties of double-layer supercapacitors and the magnitude of capacitance
that can be achieved per square centimeter over various ranges of potential and
at various electrode materials.
TABLE 6.1.
Li+
Na+
K+
Rb+
Cs+
NH!
H30+
F
ClBr[-
0.136
0.181
0.195
0.216
114
Chapter 6
GOUY-CHAPMAN
I
I
1.
NEUTRAL
MOLECULE
I1//~
1 c..OUTER'HELMHOLTZ LAYER
t"-----lNNER ....
:2.
FIGURE 6.6. General representation of the structure of the double layer showing different regions for adsorption of hydrated cations and less hydrated anions (Grahame model!?), together with
solvent molecules and an adsorbed neutral molecule.
30 X 10-6
20 X 10-6
115
These figures are shown just to give an idea of the extent to which electron deficiency or electron excess can arise on the electrode surface of the double layer.
These charge densities depend on the electrode potential, E, and hence can be
controlled or modulated by direct, alternating, or pulsed voltage.
At capacitor plates, as also at the interfaces of two-electrode double-layer
electrochemical capacitors, the surface charge densities (plus and minus
charges) arise on account of the application of a potential difference, till, between the two electrodes. In response to the applied till, electrons from one surface are driven through the external circuit containing the polarizing device (a
power supply, a battery, or a regenerative braking dynamo in an electric vehicle
hybrid system) to the other surface, establishing a difference of sign of charge
density between the two plates. In an electrochemical capacitor, the respective
plus and minus charge densities on the two plates are matched by net equal and
opposite accumulations of respective negative (anion) and positive (cation)
charge densities in the interphasial regions of the solution [Fig. 6.3(c)] over distances from ca. 1 to 100 nm into the bulk solutior., depending on total ionic concentration.
Note also that when an interphasial potential difference is generated
through a Nernstian thermodynamic equilibrium (as for the electrode potentials
of a pair of battery cell electrodes on open circuit), a double layer is also spontaneously set up at each electrode interface but is not generated by charge flow
from or to an external source of electric charge; the electrode equilibration processes generate their own double layers. Upon charge or discharge of the battery,
the two double-layers will become more, or less, charged, depending on the direction of current flow and the changes (if any) of electrode potentials resulting
from charging or discharging.
Their dependence on the electrode potential E corresponds to the development and manifestation of the double-layer capacitance, Cdl = dqldE, or integrally Cdl = q/E or !l.qltill. It is of interest chemically that the above extents of
controllable variation of surface electron excess or deficiency are comparable
with the local changes of electronic charge density that can arise chemically on
various conjugated aromatic ring structures such as benzene or naphthalene,
owing to the presence of substituents such as -OH, -S03H-, -CH3, -NHz, and
-COOH, where charge density changes arise on account of electronic inductive
and resonance effects.
116
Chapter 6
117
relatively concentrated aqueous acid (H 2S04) or base (KOH) solutions that are
used as electrolytes for a number of electrochemical capacitor embodiments, it
is clear that there must be very little free water in the interphase that is not influenced by the ions either in the bulk electrolyte or in the Helmholtz layer. 19
This is an important but little realized aspect of double~layer modeling and has
significant practical consequences, in that there may be local changes of solvent
activity and balance during charging or discharging processes.
Chapter 6
118
Doublelayer
I"
.\
Bulk
electron density
~tozero
METAL
tt
SOLUTION
Nom;nol metal
surface plane
LOCus of centers
of metal surface
atoms
FIGURE 6.7. Schematic profiles of emergent electron density at a metal at three potentials: one
positive (1), one negative (3) to the potential-of-zero charge, and one (2) at the potential-of-zero
charge. (After Lang and Kohn ref. 5).
vent molecules in the compact region of the double layer. This model has been
envisaged and treated by Lang and Kohn,5 and more recently by Amokrane and
Badiali. 6 More details are discussed in Chapter 7.
The variation of the locus of the electron density overspill with changing
electrode potential is important for modeling such processes as solvent-dipole
adsorption and orientation, and the chemisorption of ions, especially anions,
each of which factors determines, among others, the capacitance of conductor/solution interfaces as in a double-layer supercapacitor.
In energy terms, this modification of surface-region electron density by a
changing potential corresponds to effective changes of the electronic work function, cP, of the metal (when it is an electrode) or its electron affinity (-cP), which
is one of the bases of the effects of changing potential on the kinetics of electrode processes l8 (Chapter 3) and on the chemisorption of ions.
The overall structure of the electrode/solution interphase is seen to be quite
complex. Its construction in terms of surface-chemical, ion hydration, and double-layer aspects is illustrated schematically in Fig. 6.8.
119
METAL/ELECTROLYTE INTERFACE
_ _ _ _ .--------OUTER HELMHOLTZ LAYE.R
DIFFUSE LAV'ER
FIGURE 6.8. Overall construction of the electrode/solution interphase illustrating its surfacechemical and double-layer aspects. XM and Xd are the intrinsic surface potentials of the metal and
adsorbed dipoles, respectively.
10-8 V cm- 1
(6.2)
for a double-layer thickness of 3.8 A (0.38 nm). The latter figure corresponds
to the distance of closest approach of simple hydrated cations, e.g., Na+. The actual value of the thickness of the double layer will depend on the crystal ionic
radius of the ion and the thickness of its time-average hydration or solvation
shell.
120
Chapter 6
From this it is seen that E has a very high value-some 2.9 x 107 V cm- I .
Sometimes nonelectrochemists are puzzled as to how such enormous fields can
exist without electrical breakdown. Thus, in an ordinary capacitor, e.g., with a
polystyrene dielectric, fields cannot be sustained beyond ca. 5000 Vern-I. However, in the double layer, the behavior is quite different since there is no bulk
dielectric in the normal sense associated with dielectrics in regular capacitors.
Only the water of hydration of the ions and the monolayer film of adsorbed solvent water at the electrode interface (Fig. 6.6) constitute the dielectric of the
double-layer capacitance.
The situation is analogous to that on the atomic scale in, e.g., a molecular
dipole such as HCl or H 20 where the internal, interatomic field is also on the
order of 107 Vern-I. Similarly, the interion () local fields in an ionic crystal
are of about the same magnitude. No charge transfer breakdown by passage of
charges between the ions can occur since they are very stable in their regular
ionic states, Na+ and Cl-, e.g., in a rock-salt crystal. In the double layer, leakage
currents across it can arise only when thermodynamically and kinetically, electron charge transfer processes are allowed beyond certain critical potentials and
corresponding interphasial fields, as outlined earlier.
The interphasial field, E, across the double layer can also be calculated in
a different way by employing the electrostatic equation of Gauss for the field
generated normal to a metal plate charged to a density of q charges per square
centimeter.
E=-4nq/e
(6.3)
where e is the dielectric constant (coefficient) ofthe medium in which the field
is established by the charge density q. In the double layer, e is probably on the
order 6 but varies with the field.
Earlier we showed that the maximum charge density sustainable at an Hg
electrode in aqueous electrolyte is about 0.17 e/atom of the surface. The electron
charge is 4.8 x 10- 10 electrostatic units (esu, see Chapter 4). Then we find
E=
4n x 4.8 x 10-10
6
(6.4)
where the factor 3 x 1015 is approximately the number of metal atoms per square
centimeter in a close-packed Hg surface. Thus E =5 x 105 esu cm-I. In practical
units of V cm- 1, E is then found to be 5 x 105 X 300 V cm- 1, the factor 300 being
from the definition 1 V = 1 esu/300. The resulting figure for E, depending on
the value to be assigned for e, is seen to be of the same magnitude as that derived
above by dividing a typical potential difference value across the double layer by
an appropriate thickness value d. A very high value for E again results and of
121
course has to be consistent with the first approximate way of estimating the
field. The agreement or disagreement obviously depends on the value chosen
for e in Eq. (6.4). More sophisticated calculations ofE would take into account
the microscopic distribution of charges of ions and associated solvent dipoles,
and the electron overspill from the metal toward the solution side of the interface in the whole interphasial region (Fig. 6.8) constituting the double layer.
122
Chapter 6
(6.5)
At potentials approaching 0.23 V or higher Hg becomes electrochemically
oxidized according to the reaction (e.g., in KCI solution)
2Hg + 2CI- ~ Hg 2Cl 2 + 2e
(6.6)
(6.7)
HgH + OH-
(6.8)
followed by
HgH + H 20 + e
Hg + H2 + OH-
(6.9)
is very poor (exchange current density, io = 10- 13 A cm-2) so that significant currents for Faradaic charge transfer across the double layer (corresponding to a
leakage current across the interphase) do not begin to pass until potentials more
negative (vs. RHE or SHE) than ca. -0.9 to -1.0 V are attained. Then between
+0.23 V and ca. -0.9 V, only double-layer charging currents effectively pass so
that within that range the interphase approaches ideal polarizability with an almost ideal nonleaky double-layer capacitance being then manifested which can
be experimentally characterized by means of transient charging curves or ac impedance. The electrode interphase is then capacitorlike in its electrical behavior,
so that charging energy can be stored.
Gold is another metal that exhibits almost ideal polarizability over a certain
range of potentials: in this case between -0.2 V and + 1.30 V in aqueous H 2S04
solution. Beyond + 1.30 V, surface oxidation begins, eventually leading to O 2
evolution on a gold-oxide film while below ca. -0.2 V, H2 evolution leakage
currents become significant (> 10-5 A cm- 2). Within the above limits, the electrode interphase behaves like an almost ideal capacitor but with significantly potential-dependent capacitance, as is also the case at Hg. 4 ,17
It should be noted that with excursions of potential, positively or negatively, beyond the potential limits for solvent-electrolyte decomposition, further double-layer charging (or discharging) still occurs so that the total current
density passing is then the sum of a double-layer charging component, idb and
123
Cdl
The behaviors of equivalent circuits (a) and (b) are easily distinguished by the
difference between their impedance spectra with variation of frequency; thus (a)
is purely capacitative while (b) has a maximum capacitative impedance for a
given value of Cdl and RF (for a particular potential) at a certain frequency.
124
ChapterS
The Faradaic leakage resistance, RFo in circuit b is very important as the basis of self-discharge (see Chapter 18) in electrochemical capacitors and in battery cells. Its role will be analyzed in Chapter 18.
In the case of electrodes that are base metals, nonideal polarizability usually occurs because of the possibility of anodic corrosion or oxide film formation at potentials already near the H2 reversible potential; this leads to Faradaic
leakage currents in parallel with double-layer charging. In addition, such metals
usually have larger exchange current density for H2 evolution from water, so
they cannot be polarized very far cathodically to the RHE or SHE potentials
without appreciable H2 evolution currents arising.
REFERENCES
1. E. A. Guggenheim, J. Chern. Phys., 4, 689 (1936).
2. N. K. Adam, The Physics and Chemistry of Surfaces, 3rd ed., Chapter 3, p. 107, Oxford
University Press, Oxford (1941).
3. H. von Helmholtz, Ann. Phys. (Leipzig), 89, 211 (1853).
4. R. Parsons, Chapter 4 in Modem Aspects of Electrochemistry, J. O'M. Bockris and B. E.
Conway, eds., vol. 1, Chapter 4, Butterworths, London (1954).
5. N. D. Lang and W. Kohn, Phys. Rev., Bl, 4555 (1970); B3, 1215 (1971).
6. S. Amokrane and J. P. Badiali, in Modem Aspects of Electrochemistry, J. O'M. Bockris, B. E.
Conway, and R. White, eds., vol. 22, Chapter 1, Plenum, New York (1992).
7. G. Gouy, Ann. Phys., Paris, 7, 129 (1917); J. de Phys., 9, 457 (1910).
8. R. H. Fowler and E. A. Guggenheim, Statistical Thermodynamics, p. 77, Cambridge
University Press (1939).
9. D. L. Chapman, Phil. Mag., 25, 475 (1913).
10. B. E. Conway, Ionic Hydration in Chemistry and Biophysics, Chapter 10, Elsevier,
Amsterdam (1981).
11. P. Debye and E. Hiickel, Phys. Zeit., 24, 185 (1923).
12. L. Onsager, Phys. Zeit., 27, 388 (1926); 28, 277 (1928).
13. M. Green, in Modem Aspects of Electrochemistry, J. O'M. Bockris, ed., vol. 2, Chapter 5, p.
343, Butterworths, London (1961).
14. O. Stem, Zeit. Elektrochem., 30, 508 (1924).
15. R. W. Gurney, Ionic Processes in Solution, Dover, New York (1940).
16. A. N. Frumkin, Zeit. Phys. Chem., AI64, 121 (1933); Acta Physicochim., USSR, 6, 502 (1937).
17. D. C. Grahame, Chem. Rev., 41, 441 (1947).
18. J. A. V. Butler, Proc. Roy. Soc. Lond.,A 157, 423 (1936).
19. B. E. Conway, J. Electroanal. Chem., 123, 81 (1981)
20. B. E. Conway, Theory and Principles of Electrode Processes, Ronald Press, New York (1964).
21. F. P. Bowden and E. K. Ridea1, Proc. Roy. Soc., Lond., AI07, 486 (1925); A114, 103 (1927);
A119, 680; 686 (1928).
22. J. A. V. Butler, Electrocapillarity, Methuen, London (1940).
23. J. D. Bernal and R. H. Fowler, J. Chem. Phys., 1, 515 (1933).
Chapter 7
In order to provide a fundamental basis for understanding the properties and behavior of double-layer types of capacitor devices, this chapter gives a broad account of the theoretical treatments of the structure and capacitance of the double
layer at electrode interfaces. This topic has been one of major activity and interest in electrochemistry for about a hundred years, and has now found substantial
technological applications. In 1997, the Electrochemical Society Sponsored a
major symposium on the double-layer to recognize the 50th anniversary of Grahame's seminal paper! in Chemical Reviews (1947).
Electrostatic and thermodynamic treatments of the double layer are based
on a model in which the interphasial region between an electrode and an ionic
solution is ideally polarizable (Chapter 6), i.e., a potential difference can be established between a metal electrode surface and the inner boundary of an electrolyte solution without Faradaic charge transfer processes taking place. The
interphase is then ideally capacitative and the electrode is referred to! as an
ideally polarizable one. The interphase then has a pure double-layer capacitance that ideally is frequency independent in an ac evaluation of that capacitance. This is the ideal requirement for an electrochemical capacitor device.
In practice, some frequency dependence is commonly observed, i.e., the
phase angle for the double-layer capacitance may not have the ideal value of 90
at all frequencies, and potential-dependent dc leakage can also occur (Chapter
18). Deviations from ideal capacitative behavior can arise when there is some
125
126
Chapter 7
dielectric loss associated with the solvent orientation polarization in the doublelayer dielectric (at very high frequencies) and/or when there are some slow anionic chemisorption processes that lead to lower frequency losses. In either case,
there is energy dissipation in the charging and discharging cycles at an incompletely polarizable electrode interface. Another source of nonideal behavior is
distribution of the double-layer capacitance over a porous electrode surface
(Chapter 14). In a practical electrochemical capacitor device, frequency dependence of the overall capacitance is generally observed and is due, in addition to
the "porous electrode" effect, to coupling with other equivalent series resistance
(esr) components.
As emphasized by Grahame, l the double layer at an ideally polarizable
electrode is in a state of electrostatic equilibrium, in contrast to other types of
electrode interphases across which free charge transfer takes place (nonpolarizable interphases). This corresponds to an electrochemical thermodynamic
equilibrium, ideally of a Nernstian kind.
Theoretical treatments of the structure (Chapter 6) and properties of the
double layer can be conveniently considered in a hierarchy of four sections:
1. The first is the diffuse part of the double layer [Fig. 6.3(b), Chapter 6],
beyond (toward the solution) the inner contact layer of ions (solvated or
otherwise) that defines the so-called "Helmholtz layer" [Fig. 6.3(a),
Chapter 6]. Historically, this was the first theoretical examination of
double-layer properties in work by Gouy and by Chapman. 2,3
2. The second is the compact or Helmholtz layer,4 which became treated
in terms of adsorption of ions as part of Stem's combination5 of the
Helmholtz and Gouy models, taking into account the finite sizes of the
ions of the electrolyte, which determine their distances of closest approach to the electrode surface [Fig. 6.3(c), Chapter 6].
Later, Grahame 1 treated the compact region in terms of two layers,
one for anions located at their distances of closest approach to the electrode metal, and the other for hydrated cations located at their somewhat
longer distances of closest approach, which are determined by their hydration (or solvation) radii, as shown in Chapter 6. The two regions distinguished by Grahame are referred to as the "inner Helmholtz layer"
(or plane) and the "outer Helmholtz layer," (or plane), respectively.
3. Work on the two classically distinguished regions of the double layer
in (1) and (2), was followed by important treatments of the role of adsorbed solvent dipoles and their potential (field)-dependent orientation
in the compact regions of the double layer6- 9 in determining the capacitance.
4. Finally, in relatively recent years, attention has been paid to the potential-dependent spillover of electron density from the formal surface of
127
the metal toward (or away from) the solution and compact layers, depending on the sign of charge of the electrode metal, and its influence
on the interphasial capacitance of the electrode. 10-12
The above factors in double-layer behavior are interactive and determine
the overall capacitance of the electrode interphase and its dependence on electrode potential, the types of ions in the electrolyte solute, and the solvent in
which they are dissolved. Such an interphasial system is obviously complex and
has been modeled according to various levels of recognition of this complexity.
These are treated in the following sections.
For completeness and for historical reasons, in discussing the models of the
double layer, we first outline the electrostatic statistical treatment of the diffuse
part of the double layer. In practice, however, the behavior of the diffuse-layer
capacitance is of less significance for the properties of electrochemical capacitors than is that of the inner layer since the latter dominates the overall interfacial properties at the high concentrations of electrolyte commonly employed
in capacitors to minimize equivalent series resistance or internal pore resistance.
However, treatments of the diffuse part of the double layer have formed a cornerstone in our understanding of double-layer properties, and an accurate theory
of diffuse-layer capacitance is an essential requirement for evaluating compact
layer capacitance. This latter mainly determines the capacitance of doublelayer-type devices, through application of the reciprocal sum relationship (Eqs.
7.11 and 7.12) to experimental data on overall capacity. Direct theoretical treatment of the inner layer itself, although included in Stern's paper,5 has been slow
to develop in reliable form, owing to the complexity of the modeling involved
(see later discussion).
The treatment of the diffuse-layer ion distribution [Fig. 6.3(b), Chapter 6]
and its dependence on the surface charge density, qM, on the metal, which is related to the latter's electrode potential relative to the potential at which the electrode bears a zero charge (the potential-of-zero charge), proceeds in the
following way: Poisson's electrostatic equation is used in conjunction with the
Boltzmann distribution law4 and applied to the electrostatic energy experienced
by both cations and anions owing to the interaction of their charges with the potential that results from the electrode surface charge or, more particularly, to the
gradient of the electric field near the electrode.
If IfIr is the potential at a distance r from the electrode surface, the local concentration of cations and anions, c+ and c, at r will be4
128
Chapter 7
(7.1)
where cO terms refer to the bulk average stoichiometric concentrations of cations
and anions related by c+z+ = IczJ, and z+ and z_ are the charge numbers of the
cations and anions, respectively. The space charge density pat r is then
Pr - 4 ed?-
(7.2)
if only electrostatic interactions determine the ion distribution. As in the Debye-Hiickel theory, a further relation is available between /fir and Pr by use of
Poisson's equation relating field gradient to space charge density, P, but expressed for a one-dimensional electrostatic distribution of ions in the form
(7.3)
since Br may be a function of r in the double layer as first treated by Conway,
Bockris, and Ammar. 13 ,14 Substitution of Pr from Eq. (7.2) into Eq. (7.3) gives
(7.4)
Using the identity
(7.5)
in Eq. (7.4), and integrating with Br ='ii, a mean constant value of the dielectric
constant B in the double layer (since B is not very sensitive to variation of r except very close to the electrode surface), gives
(7.6)
for the boundary conditions d/flr1dr -7 0, r -7 00; and /fir -7 /fIs, the potential in
the bulk of the solution where Pr =O. The term -1 in the braces in Eq. (7.6)
arises from the integration constant, i.e., when r -7 00 , exp[(/fIr - /fIs)lkT] = 1 for
d/flr1dr=0.
129
At this point in the treatment of the diffuse double layer it is assumed that
Gauss's relation
(7.7)
can be applied near the electrode surface bearing the charge density, qM, at the
distance of closest approach, a, of the ions to the electrode surface. Sa is the local
value of the dielectric constant, s, in that region. Sa is taken as a mean value
though Sa may depend on potential (field 13 ) and will probably have a different
mean value on each side of the pzc.
Also, solvent dipole orientation (cf. dielectric saturation effects) at the
electrode surface6- 9 leads to a lower value of S in the region r::e a than further
away from the electrode. From Egs. (7.6) and (7.7), qM is obtained with the
above assumptions as
e,
(7.8)
C (=
qM-
(7.9)
If the ions are point charges as considered by Gouy, If/a becomes identical
with <PM, the potential of the metal with respect to the solution, and distribution
ofthe ions continues right up to the metal surface, a = o. By Eg. (7.9), aguantity
dqMlo( If/a - If/s) can be evaluated as
dqM
=(z2e2ce]112 cosh [ze(lf/a - If/s)]
o(lf/a - If/s)
2nkT
2kT
(7.10)
130
Chapter 7
which may be identified with the differential double-layer capacity, C, that is associated with the ionic atmosphere charge distribution dependent on afllrlar and
hence on qM' This capacitance will be referred to as Cdiff.
A comparison of values of the capacity calculated from Eq. (7.10) with
those measured experimentally from the dependence of surface charge (qM), at
the metal on electrode potential E, indicates, however, a major discrepancy with
regard to the rate of change of C with JM - fils and the absolute magnitude of C
at values of JM - fils> ca. 0.1 V. Only near the potential at which qM = 0 (the
pzc) is the experimental behavior numerically comparable with that predicted
by Eq. (7.10).
The principal reason for the failure of this relation lies in the assumption
that the ions are point charges so that the ionic charge distribution can be continuous up to the electrode surface. This allows a large space charge to arise very
near the electrode, and hence an anomalously large capacitance is calculated. If,
however, the ions have a finite size (e.g., as determined by their crystallographic
or, more appropriately, their primary hydration radii, a,), the continuous charge
distribution must be cut off at r =a, and the potential-distance relation near the
electrode is then discontinuous as in Fig. 6.1 (c) in Chapter 6.
The importance of the distance of closest approach a for ions of finite size,
referred to earlier, was discussed by Grahame l but was recognized by Stern5 and
was mentioned by Gouy in regard to an effective thickness of the double layer.
The overall potential difference, JM - fils, between the metal and the solution can then be regarded as made up of two parts JM - fill and fill - fils where
fill is the mean potential at r = a, i.e., fila == fill' fill is referred to as a mean potential since the potential in the double layer will fluctuate laterally across the interphasial region because of the discreteness of the ionic charges involved. The
symbol fill is used to maintain conformity with the earlier Russian literature, but
J2 has been used as the symbol for this quantity in some later publications. Both
these contributions can be dependent on qM so that
(7.11)
The first term is the reciprocal of the overall capacity C mentioned earlier, and
the other terms have a similar nature, so that Eq. (7.11) can be written
1
Cl
C2
-=-+-
(7.12)
where C l and C2 are obviously defined by comparison with the two terms on the
right-hand side ofEq. (7.11). Equation (7.12) also implies that C l and C2 are additive in a series combination. C l is the Helmholtz compact layer capacity and
131
For small deviations of potential around the pzc, this function for C is
mainly determined by C2 and increases approximately exponentially with '1/1 'l/s on each side of the pzc.
Itis seen that when C 1 C2> C= C2 (fromEq. 7.10 with 'l/a= '1/1), or when
C2 Cj, C = C 1; the first inequality is valid when '1/1 - 'l/s is quite small, and
the second when it is large. It is important to note that it is the value of the
smaller capacity contribution that mainly determines the overall value of C
since the two contributions are in a series relationship (Eq. 7.12), as discussed
in Chapter 6 (Eq. 6.1).
Apart from giving a quite accurate representation of the diffuse-layer capacitance, the Gouy-Chapman theory is important because it provides the
means of evaluating the inner-layer capacitance, C 1, which is of greater interest,
from the overall C measured, through Eq. (7.12). In fact, at the rather high concentrations of electrolytes employed in aqueous double-layer capacitors, it is the
compact or Helmholtz layer capacitance that mainly determines the capacitance
of a double-layer electrochemical capacitor device.
Equation (7.13) represents the data for capacitance at the mercury/solution
interface (e.g., for a KCI solution!), if C 1 = 18 f1F cm-2 is assumed when cations
preferentially popUlate the interface region (l/JM - 'l/s negative) and C, = 38 f1F
cm- 2 when the anions dominate in the double layer (l/JM - 'l/s positive). Typical
capacity-potential profiles for a mercury electrode are shown in Fig. 7.1 (from
Ref. 1). The capacity contribution C 1 is referred to as the capacity of the Helmholtz layer and C2 is that of the diffuse layer or ionic atmosphere region. Further
divisions of C! into contributions from the inner or the outer Helmholtz layers,'
referred to earlier, will be discussed later.
In Eq. (7.13) it will not generally be validl,15 to assume that the same potential '1/1 applies to the position of closest approach for anions and for cations.
Owing to the more polarizable nature of anions and their specific affinity as adsorbed surface ligands for some metal surfaces, anions may approach more
closely than hydrated cations at a given potential, so that '1/1 will not have quite
the same significance for anions as for cations. It is convenient to call this potential'l/2 for the region of closer approach and specific adsorption, i.e., chemisorption associated with anions. This matter is now examined in more detail
according to the theory developed by Stern5 and reviewed by Parsons. 15
132
Chapter 7
NaGI
NoF
46
0.1 M
Hump
28
20
0.001 M
12
~Influence of
Ie
Cditt.
LL.
::i...
.......
W
U
0.8
<t
0.4
-0.4
-0.8
-1.2
-1.6
-20
l-
e:::;
a:
O.IM
<t
a::
w
80
Nol
42
NoNO:;
1M
1M
.-l
34
W 60
.-l
CD
::>
0
0
26
40
O.IM
18
20
10
0.4
-0.4
-0.8
-1.2
-1.6
ELECTRODE POT.:NTIAL
10
0.8
0.4
~ELATIVE
-0.4 -0.8
-1.2
-1.6
TO p.z.c.lV
FIGURE 7.1. Differential capacitance of the Hglaqueous solution interface at 298 K in foure1ectrolytes: NaF, NaCl, NaI, and NaND3. Concentrations indicated. Reprinted with permission from
Chem. Rev., 41, 441 (1947). Copyright American Chemical Society.
133
We referred earlier to the fact that the diffuse layer, as a model for the double layer, is applicable only to dilute electrolyte solutions over a small potential
range near the pzc. Stern5 recognized that a satisfactory theory of the double
layer must take into account both the finite size of the ions adsorbed and any
specific chemisorption interactions they may experience with the metal surface.
The charge qs on the solution side of the double layer may thus be regarded as
being made up of two contributions, qz and ql> arising, respectively, from the
space charge associated with the ionic atmosphere and from the charge associated with specifically adsorbed ions. Then, since qM =-qs,
(7.14)
The charge qz will be given by Eq. (7.8) with tJlz as the relevant potential, i.e.,
tJla = tJl2' The charge ql was obtained by Stern in terms of site-fraction statistics
for solution and surface phases by applying a form of the Langmuir isotherm
with an electrochemical Gibbs energy of adsorption determining the extent of
adsorption as a function of solution concentration of the ions.
Stern's derivation for the quantity tJlz was unnecessarily involved, and Parsons l5 has given a useful clarification of the calculation which need not be repeated here. The result of applying the Langmuir isotherm is
ql
= eQ
[1 +
e-~-p-[L1-V-~-lk-T-]1
(7.15)
where L1C; are the standard electrochemical Gibbs energies of adsorption of the
indicated cations or anions, i.e., the L1C; terms normally contain both electrical
and chemical Gibbs energy contributions, the latter defined for some standard
state concentration or mole fraction Xs of salt in the bulk solution. The term Q
is defined as the number of sites per square centimeter available for adsorption
at the electrode interface. It is assumed that this number is the same for cations
and anions, and that the adsorbed ion can freely replace solvent molecules,
which are also necessarily adsorbed at the surface, with a relative fractional coverage depending on coverage by chemisorbed anions.
Reference has been made earlier to the fact that the L1C; terms in Eq. (7.15)
are electrochemical Gibbs energies of adsorption; following Stern's assumptions, they may be written for convenience in the form
Chapter 7
134
(7.16)
where ~G~ terms are the chemical standard Gibbs energies of adsorption of the
indicated ions (), and 1f/2 is the mean potential in the plane of specifically adsorbed ions. Experimentally, however, it will not usually be possible to separate
the ~G~ terms into their component chemical and electrical contributions.
An explicit mathematical expression for the inner-layer capacitance, C1, in
Eq. (7.12) corresponding to aqM/a(M-lf/d in Eq. (7.11) or dq1td(ifJM - If/l) in
Eq. (7.15), is not easy to obtain directly, though the electric potential is implicit
in the ~G~ terms. Hence, in practice, experimental data for the overall C are
split, reciprocally, into C I and C2 components by calculating the theoretical C2
diffuse-layer capacitance for a given concentration and electrode potential, and
combining it reciprocally with the experimentally measured C: lIC I = lIC -
lIC2
In most cases only one type of ion is specifically adsorbed (the anion, except in cases of organic electrolytes such as tetraalkylammonium salts), so that
Eq. (7.15) could be written for anions as
(7.17)
since usually for dilute solutions, (lIXs) expLl.G;lkT 1. It is clear that qltz-e
is the number nl of ions in the adsorbed layer and that nltQ is the relative cov.
15
f hese IOns,
erage lot
I.e.,
(7.18)
135
form of the isotherm written as Eq. (7.18) may in fact be physically the more
reasonable one.
Stern expressed the integral capacity Kl of the double layer as
(7.19)
and assumed that !fI2 = !fib which enables values of!fll to be calculated as a function of M, the potential of the metal, for known Kl and solution concentrations
in the absence of specific adsorption. (Usually this evaluation is most conveniently performed numerically by choosing !fIl values and obtaining M as the dependent variable, since !fIl enters the arguments of exponentials in Q2') The term
Kl is the integral capacity of the region between the metal surface and the plane
of centers of the adsorbed ions. The assumptions !fI2 = !fIl and AC; = 0 are
equivalent to taking ql =0, i.e., when specific adsorption is absent. Stern did not
distinguish, however, between !fI2 and !fIl for the case of specific adsorption, and
this distinction constituted an important element of Grahame's treatment l referred to earlier.
As mentioned several times before, the inner-layer capacitance cannot be
properly interpreted, especially with regard to its substantially differing values
on the positive and negative sides of the pzc at Hg (Fig. 7.1) or Au, withoutrecognizing different values of the closest approach distance, a, for positively and
negatively charged electrode surfaces. This situation corresponds to different
effective radii of anions and cations, owing to their usually differing hydration
and crystallographic radii, and because anions tend to be more strongly chemisorbed (owing to their greater lone-pair electron donicities) than corresponding
isoelectronic cations, e.g., Na+-P-, K+-CI-, Rb+-Br-. Other factors (such as hydrophobicity) are involved in the adsorption of large organic cations (e.g.,
Et4N+, n-Bu4N+), which have major effects on the interphasial capacitance of
electrodes. These ions are important in nonaqueous solution capacitors, where
they often form the electrolyte with co-anions such as PF6, BF.;, CI04, and are
adsorbed at carbon powder or fiber surfaces.
The recognition of the importance of different distances of clos~st approach for cations and anions, and the corresponding distinction between inner
Helmholtz and outer Helmholtz planes for the description of the compact layer,
was one of Grahame's major contributions,l complementing his detailed treatment of the thermodynamics of the interphase at the ideal polarized electrode.
In the case of anions, their adsorption at electrodes can be usefully regarded as a quasi-chemical process involving partial charge transfer. The extent,
0, of this charge transfer is related to the so-called "electrosorption valence." It
Chapter 7
136
7.5.1. General
137
In parallel with this theoretical work were new (in the 1960s) experimental
studies by Hills and Payne20- 23 on the properties of the double layer at Hg in
nonaqueous solvents, the electric polarization properties of which are substantially different from those ofliquid water. 24
Also to be noted is the fact that ions (cations or anions, depending on the
electrode potential relative to the pzc) populating the compact (Helmholtz) region of the double layer are usually substantially hydrated (especially cations 24 )
and hence locally influence the orientation and state of polarization of the water
molecules that also always populate the double layer. 6- 8 Thus the structure of
the double layer at the molecular level is quite complex.
The energies of solvation of ions in nonaqueous solvents are usually substantially less than in water (hence the lower solubilities of many salts in such
solvents), so that the tendency for ions, especially anions, to become chemisorbed at electrodes is greater in aprotic media than in water. This leads to an indirect effect of the solvent type on double-layer capacitance.
Solvents are of three main types: (1) associated ones such as water and
D 20, methanol and other alcohols, hydrogen fluoride, formamide, and carboxylic acids in which strong interactions and correlated orientations of the molecules (Chapter 5) occur through hydrogen bonding; (2) polar ones where there
are still quite strong electrostatic dipole-dipole interactions but without the directional specificity and strength of H bonding; examples are acetonitrile, nitromethane, acetone, dimethyl sulfoxide, dimethoxyethane, propylene
carbonate, ethyl and methyl carbonates, dimethylformamide, methylene dichloride, and esters; and (3) finally nonpolar ones such as hydrocarbons and fluorocarbons.
The use of nonaqueous aprotic solvents for electrolytes in carbon capacitors has become very attractive because substantially higher voltages on charge
(up to 3.5-4.0 V) can be obtained owing to the higher solution decomposition
voltages that apply (Chapter 2) with such systems. Hence, higher energy and
power densities can be achieved. This preferred use of nonaqueous solutions for
electrochemical capacitors in order to maximize energy density has enhanced
interest in the specific dependence of double-layer structure and capacitance on
the type of solvent in recent years. Thus the fundamental treatments of the role
of solvent polarity and structure in determining double-layer capacitance behavior have come into new focus in recent years. This matter is sufficiently important for a separate chapter (8) to be devoted to it, in relation to the development
of nonaqueous electrochemical capacitors.
There are several indirect aspects of the use of nonaqueous solvents for
higher voltage capacitors. (1) They are usually more difficult to purify than
138
Chapter 7
water and hence can contain small but significant quantities of impurities that
can be either electrochemically reduced or oxidized, leading to self-discharge.
(2) Because of their usually smaller solvating power for ions, ion adsorption at
electrode interfaces can be stronger, modifying the double-layer capacitance behavior. (3) Because their dielectric constants are usually lower than that of
water, specific double-layer capacitances tend to be lower than in aqueous solutions. Also, (4) because of these lower dielectric constants, the specific resistivities of their solutions of electrolytes are usually less. This tends to raise
equivalent series resistance values in nonaqueous solution capacitors, leading to
diminished power densities (see Chapters 15 and 17).
For aprotic solvents, the polarity is often measured by their bulk dielectric
constants. However, in evaluating local intermolecular interactions, and interactions with charged surfaces and ions, the individual dipole moment and/or the
donicity of unshared electron pairs are usually found to be the more significant
quantities. In fact, as Table 13.1 in Chapter 13 shows, the dielectric constant is
little related to the lone-pair doni city numbers (based on heats of complexation
of the solvent to a Lewis acid such as SbF5). The behavior of the metal/solution
double layer depends very much on these electrical polarity properties of the
solvent of the electrolyte solution in contact with the electrode metal surface.
The electrical state of the solvent in interphasial double layers can be usefully determined from evaluations of the effective dielectric constant, 8, of the
solvent medium in the double layer, which is derived from capacitance measurements. Under certain conditions, complementary information can also be obtained by applying Fourier transform infrared spectroscopy to thin solution
layers at electrode surfaces. The latter procedure provides information on the intra- and intermolecular vibrations and hindered rotations (librations) of the solvent molecules. The former procedure provides information on the overall
electrical polarizability ofthe solvent molecular system in the narrow interphasial region of the electrode metal boundary with the solution where molecular
orientation6- S can be influenced by the electrical field across the double layer.
The effective dielectric constant, 8, of the double layer at Hg at potentials
negative to the pzc in an electrolyte of aqueous NaP, the ions of which are not
chemisorbed (specifically adsorbed), can be calculated from the relation referred to earlier, namely (in the system of rationalized electrostatic units,)
C=A8KOld
(7.20)
where Ko = 8.85 X 10- 12 F m- I is the permittivity of free space. C for the above
system has a value of about 17 flF cm-2 negative to the pzc. Taking the thickness
139
140
Chapter 7
(7.21)
141
include the effects of electrostriction (see Ref. 17), while lateral interaction effects in the inner layer of oriented adsorbed water molecules are only crudely
treated. For example, the angular dependence of H bonding in water-water interactions 25 .27 ,28 is not considered and cooperative effects in H bonding have received little attention in treatments of water in the double layer, except
qualitatively in terms of semiempirical cluster models, and in lattice treatments
by Parsons and Reeves 29 and Guidelli,3o referred to in Section 7.5.6.
Ni
Ni +NJ-
(7.23)
where U is the net energy of the oriented dipole in the field. When lateral interactions are significant, 8 U has the form
U = flE + Wn NJ-IN - Wn NilN
(7.24)
where N = Ni + NJ-, W is the pairwise lateral interaction energy between oriented dipoles and 2n is the coordination number in the interphase. The direction
of fl in relation to that of the field vector E determines the sign of flE and hence
that of the energy U.
Since
142
Chapter 7
Ni
NJ,
-=e2UlkT
(7.25)
9\ =
Ni -NJ,
=tanh UlkT,
(7.26)
[~: - 9\WnlkT]
(7.27)
N +N-I-
i.e.
9\ =tanh
fl(9\ - J) - 4naqM
1 + aneld
(7.29)
where <5 = tan2012 for a mean angle 0 of preferential orientation of the H20 dipole toward the electrode, e.g., due to asymmetric dipole image interactions
and/or to specific chemical affinity of one end of the dipole, such as H 20 for the
electrode metal. fl(9\ - <5) is the mean moment per site on the surface in the array
143
of oriented dipoles and a here is the electronic polarizability of the dipoles determining the magnitude of induced moments.
The relation between Ll V, d, and the other quantities shown can be used to
obtain an empirical expression for the effective dielectric constant Be by writing
LlVld =mean field E and E= -4nqM/Be by means of Gauss's law.
Then Be is seen to be given9 by
(7.30)
and its variation with field toward a saturation value depends on the variation
of the dipole orientation function 9\ with field; 9\0 is any residual (specific) orientation when qM = 0 at the potential-of-zero charge referred to earlier; it is indicated experimentally.
The main weakness of the treatments by Watts-Tobin and Mott,6 and by
Levine, Bell, and Smith9 is the neglect ofH-bond structural effects2s.27-30 in the
interphase, i.e., correlated orientation effects. Such effects are not adequately
taken into account by the lateral interaction term wn9t, since this has no general
angular dependence on orientation,2s.27 only extreme orientations being con sidered. 8.9
The limiting two-state situation is undoubtedly a serious oversimplification since molecules of the water dielectric in the interphase will be oriented at
various angles and some average polarization orientation angle will be a function of field and Wn e , and will also be determined by the angular dependence of
H-bond energy between neighboring molecules,2s.28 The problem then becomes
a 2-dimensional version of the Kirkwood theory for associated dielectrics so
that the simple, elegant, but perhaps oversimplified features of the Watts-Tobin
and Mott model 7.8 would become lost. The problem of some overestimate32 of
the dipole orientation effect in the two-state modelS was solved by Fawcett33 by
introducing the idea of a three-state orientation distribution, with the "third"
state lying flat along the electrode surface as in Reeve's mode1. 29
Experimental studies at Hg indicate that some specific direction of orientation of water dipoles tends to occur34- 36 at Hg, Ga, and other metals in the absence of an opposing Coulombic field due to excess charge. Hence it is
necessary to take into account both orientation effects due to chemisorption and
those due to changing electrode charge. This was recognized by Levine, Bell,
and Smith,9 while Bockris, Devanathan, and MulIerS noted an equivalent preferential orientation effect which they considered would occur because of interactions of the noncentral dipole in water with its image charges in the metal (Hg)
144
Chapter 7
surface. This is, of course, a different effect from that associated with specific
interaction of the lone-pair orbitals on 0 in H 20 with the atoms in the Hg surface, but both are connected with the electron charge distribution in the water
molecule and its donor and acceptor properties at metal interfaces. At molecules
containing an S atom (e.g., thiourea), the specific affinity of the S-end of the molecular dipole for Hg is very much larger.
Frumkin and Damaskin37 assumed that the contribution of the surface dipole potential (Xd) to the total surface potential could be written
(7.31)
145
10
:3
4
60
...
I
u
lL
40
-~
30
20
10
-~
-10
__- L____L - _
-1~
-20
'i",/ ~Ccm-2
FIGURE 7.2. Inner-layer capacitance as a function of surface charge qM calculated according to
Damaskin and Frumkin,37 for the parameter values indicated. Points: data for Hg.
From the data required to empirically fit the experimental results, an estimate of the area of the clusters of adsorbed water was given as 0.24 nm2 The
mean dipole moment of the clusters was found to be ca. 0.75 D. The effective
value of B was also derived asf(HM)'
It should be noted that a capacity hump arises in curves 2 and 3 of Fig. 7.2,
but displaced from the pzc. This is an important feature of most of the experimental capacitance results at the Hg/aqueous solution interface, which has been
the subject of much previous discussion,S,41 including the idea that H 20 dipole
reorientation was involved. The hump is, however, usually sensitive to the type
of anion of the electrolyte, which led Bockris, Devanathan, and MulIerS to offer
an explanation in terms of changing anion adsorption and repulsion in the double layer with increasing positive qM' Probably both oriented chemisorbed water
146
Chapter 7
and specifically adsorbed anions 36 playa role since the state of orientation of
chemisorbed water will almost certainly depend on the presence of anions in the
compact layer.
Curve 4 in Fig. 7.2, for K2 = 50 Ji.F cm- 2, corresponds to the experimental
behavior of Ga where the capacity rises sharply at the pzc with no hump.
Stronger chemisorption of water dipoles corresponds to a lowering of K2 (from
200 to 50).
Following Damaskin and Frumkin's paper,37 Parsons 42 presented a more
detailed treatment of the state of a solvent at an electrode interface in terms of a
primitive four-state model. A similar multistate treatment was subsequently
given by Bockris and Habib,43 but apparently without comparison with the previous 42 work.
Parsons 42 took account more quantitatively of the number of molecules per
square centimeter in the double layer and avoided some of the arbitrary assignments of parameters, discussed earlier, that were involved in the DamaskinFrumkin theory.37 Solvent molecules were assumed to exist in two states of
aggregation at the surface of the electrode: free molecules and small clusters.
The free molecules could adopt either of two extreme "up" or "down" orientations, as in the Watts-Tobin model,6,7 with a dipole moment Ji.. The clusters
could also adopt one of two extreme orientations, depending on the orientation
of their resultant perpendicular dipole moment, Ji.c, with relation to the metal
surface. Ji.c was expressed per molecule of the cluster and a parameter p was defined as the ratio /1cl/1 to characterize the state of polarization of the cluster. The
solvent polarization was expressed in terms of Boltzmann orientation distribution functions, A+ and A_, for clusters and other terms for free dipoles.
Inner-layer capacities were calculated as a function of a reduced charge parameter S and are shown in Figs. 7.3(a) and 7.3(b) for the two values of the reduced dipole capacity K with A_ taken as 10-8 A series of curves was given for
various A+ values, varying from 10-2 to 10-6 or 10-7. With A+ taken as 10-2 and
A_ as 10-4, the general features of the experimental capacitance curve for aq.
NaF at 0C 41 are quite closely reproduced. The calculated capacities C which
are plotted in Figs. 7.3(a) and 7.3(b) are expressed in terms of the ratio
C =C/Ko where Ci is the inner-layer capacitance derivable from analysis of experimental results and Ko is an inner-layer capacity for a fixed orientation of dipoles.
The physical significance of A+ must be mentioned because it determines
the general shape of the capacitance vs. S curves [Figs. 7.3(a) and 7 .3(b)], e.g.,
in changing from a form similar to that for Hg [A+ = 10-4 - 10-5 in Fig. 7.3(a)
and 7.3(b)] to that for Ga (A+ = 10- 2). A+ is of course analogous to Damaskin
and Frumkin's parameter K2; here it measures the energy associated (among
other things42) with binding of the dipole to the metal. A large value of Ji.b.+ (0end of H 20 dipole adsorbed at the metal) displaces the sharply rising part of the
.---.-----.-----,--
-7
147
-6
-5
-4
-3
-2
(0 )
( b)
C.
-6 -4 -3-2
10
-10
-10
capacity curve to more positive S values where the field effect on orientation becomes dominant. In this work, the inner-layer capacity behavior is quite well accounted for by H 20 dipole and cluster orientation.
With solvents other than water, different types of inner-layer capacitance
behavior can be distinguished, as shown schematically in Fig. 7.4. It is possible
that type (ii) (Fig. 7.4) would manifest behavior like that of water if a wider
range of positive qM could be explored. Type (iii) is, however, different and corresponds to weak association and binding to the surface. The transition from
type (i) to type (iii) is qualitatively reproduced by the theoretical calculations as
larger values of log A+ in Figs. 7.3(a) and 7 .3(b) are taken. In certain solvents,
two humps appear, one on either side of the pzc.
In such cases, it seems that ion interaction in the double layer must be invoked 8 to explain at least one of the humps. Also, in water, the development of
the hump depends very much on the nature of the electrolyte anion (Fig. 7.1),
so the capacitance behavior cannot be treated only in terms of water dipole or
cluster orientation without some allowance for specific solvent orientation effects caused by the adsorbed anions in addition to that caused by excess surface
charge, qM' Indeed, from the known 24 specific influence of ions on the local state
of bulk water, it would be very surprising if similar effects were not of major
importance in the solvent layer in the interphase.
Chapter 7
148
l i)
l i il
( iii)
qM
FIGURE 7.4. Schematic representation of three principal types of inner-layer capacitance profiles as a function of qM for aqueous and nonaqueous solvents: (i) water type; (ii) ring-compound
type, e.g., proPllene carbonate, sulfolane; and (iii) humpless type, e.g., CH30H, HCOOH, NH3.
(After Parsons. )
149
Electrocapillary curves at Ga in aqueous solution show much more asymmetry than at Hg. This is, however, not due to stronger anion adsorption. Anions, such as 1-, are in fact less strongly adsorbed. 49 Thus, the extent of increase
of capacitance on Ga as potential changes to more positive values (Fig. 7.2) cannot be accounted for quantitatively by increasing adsorption of anions. Stronger
specific adsorption and orientation of water dipoles than at Hg is the explanation
of the different properties of the double layer at Ga in comparison with Hg.37 It
is clear that if the specific adsorption of water at metals depends on the chemical
identity of the metal; then image interactions alone are not the only factor in
preferential orientation of water at Hg and other metals. Kemba1l 5o has studied
the related question of entropies of ad layers of water and methanol at the vapor/Hg interface and given a detailed statistical-mechanical interpretation of the
results in tenns of mobilities of the adsorbate molecules.
The general difference of adsorbability and specific orientation of H 20 dipoles in aqueous solution at various metals was noted by Trasatti34- 36 in relation
to the dependence of the potentials-of-zero charge of the metals on their elec-
Chapter 7
150
FIGURE 7.5. Laterally associated H20 dipole arrays in the solvent dipole layer at an electrode
surface. (After Parsons and Reeves. 29)
~/~'cy--T~
If
9
T
1
fJ
i. ~~~-~ - ~
~"'1--<~~
-------
~---electrode-surface
Plane
FIGURE 7.6. As in Fig. 7.5, for an element of the W bonded interphasial H20 lattice at an electrode surface. (After Guidelli. 30)
151
The contribution of potential-dependent solvent orientation to the interfacial capacitance of solid metal electrodes is well demonstrated from work on
well-ordered, single-crystal surfaces. Unlike polycrystalline solid metal interfaces, single-crystal surfaces exhibit a well-defined single pzc and hence welldefinable surface charge densities and interphasial fields as a function of
displacement of electrode potentials from the pzc.
A good example of double-layer studies on single-crystal metal surfaces is
that by Valette41 on Ag(1lO) in aqueous NaF, NaCI0 4, and KPF6 electrolytes
for which the order of relative strengths of specific adsorption is F- >
CIO:! > BF4 =:: PF6 ~ O. This order is approximately the inverse of that at Hg
electrodes and is attributed to local adsorption on surface defects.
Figure 7.7 shows the curve for the inner-layer differential capacitance for
Ag(llO) in 0.04 M aqueous KPF6. Data for three other concentrations down to
0.005 M were recorded at five frequencies. Minima in the curves of the overall
100
IE
u
u...
50
::I..
...
......
O~----~-2~0~-----~1~O------~O~----~1~0------J20~--
qM/,.,.C cm- 2
FIGURE 7.7. Double-layer capacitance behavior of single-crystal Ag(llO) as a function of potential around the pzc in aqueous 0.04 mol dm- 3 KPF6 at 298 K. Reprinted from G. Valette, 1. Electroanal Chern., 122, 285 (1981), with permission from Elsevier Science.
152
Chapter 7
capacitance arise from the (reciprocal) contribution of the diffuse-layer capacitance in the usual way (Eq. 6.1 in Chapter 6). The PF6 anion is minimally adsorbed.
When the diffuse-layer capacitance is factored out from the total series interphasial capacitance combination, the Helmholtz, compact-layer capacitance
can be evaluated in the usual way and gives rise to the interesting curve of Fig.
7.7. The inner layer capacity, thus decoupled from the diffuse-layer capacity,
shows a very well-defined maximum that is almost independent of (KPF6) concentration and is located close to the zero charge (qM = 0) condition (Fig. 7.7).
At substantial positive or negative surface charge densities, the capacity tends
to values of about 25 flF cm- 2, which is rather larger than that for Hg, namely,
about 16 flF cm-2, which corresponds to an interphasial B of the water layer of
about 5.7. Valette considered the structure of the Ag( 11 0) surface in terms of
the channeled arrangement of surface-Ag atoms in relation to possible structured arrangements of adsorbed water molecules. His analysis gave a good account of the observed capacity values at substantial () values of the surface
charge density in terms of the model of Ref. 42.
Appreciable self-orientation of water dipoles caused by the discontinuity
of the water structure at the electrode wall and to the role of dipole image interactions with the metal is possible; in fact, some statistical-mechanical calculations 5 ! do indicate a degree of icelike structuring of water molecules at a wall
interface52 with the solvent water. The whole topic of states of gases and liquids
in confined systems, such as fine pores, is attracting much current interest (e.g.,
see Refs. 53-60) and is relevant to double-layer behavior in high-area porous
supercapacitor electrodes.
This section is concluded with a summary (Table 7.1) of the inner-layer
differential capacitance, Ci , of eight solid metals and two liquid alloys at their
respective potentials-of-zero charge. Values of the respective inner-layer relative perrnittivities, Bi, are given in the third column and are based on an assumed
effective thickness of the inner layer of 0.31 nm. It is seen that both Ci and Bi are
appreciably dependent on the chemical nature of the metal. This reflects the
specificity of solvent adsorption at the metal coupled with the metal electron
component lO-!2 (see Section 7.6) of the double-layer capacitance and the free
electron density in the metal.!!
7.5.9. Recent Modeling Calculations
153
TJ(Ga)C
Bi
Sn
Pb
Cd
In(Gar
Zn
Ga
Cj I tlF cm-2a
23
28
34
36
39
40
52
60
100
135
8.0
9.8
11.6
12.6
13.7
14.0
18.2
21.0
35.0
47.3
Source: Reprinted from S. Trasatti, 1. Electroanal Chern., 123, 121 (1981), with permission from Elsevier
Science.
arnner-Iayer differential capacity at the potential-of-zero charge.
brnner-layer permittivity derived from C; with the inner-layer thickness taken as 0.31 nm.
CLiquid alloys.
Hg-aqueous solutions, albeit on a semiempirical basis. Some more recent papers have approached this problem by means of advanced statistical-mechanical
and computational simulation procedures. Several of these papers are summarized below.
A-
L7
FIGURE 7.8. Icelike ordertcd H-honded water structure near an electrode surface. The two inclinations of the dipoles in the contact layer with respect to the surface normal ~ are distinguished by
the horizontal and vertical shadings. For an uncharged surface they are energetically equivalent.
(From Torrie and Patey.51)
154
Chapter 7
(a)
0.0
FIGURE 7.9. Probability density p(8,r) for molecular orientations as a function of distance rfrom
contact with a neutral surface for a fluid of water molecules with tetrahedral symmetry. All distances are in units of the solvent diameter, ds 2.8 A. (a) 80H, OH-bond orientation; (b) 81" dipolru
orientation. (From Torrie and Patey.51)
155
(a)
\0.'>
'\
180
p(e~, r)
~
(b)
~
FIGURE 7.10. As in Fig. 7.9 but for a surface charge density of -13 j1C cm- 2 Note the more
intense orientations. (From Torrie and Patey.51)
156
Chapter 7
It has been realized for some time that at the metal/vacuum interface there
is a spillover of electron density just outside the nominal atomic surface of a
metal. This arises because the wave functions of the delocalized electrons of the
157
metal (approximately one per atom) allow a finite probability of finding electron density outside the metal to an extent that diminishes exponentially outward from the metal surface. Qualitatively, the electron spillover effect was
recognized a long time ago (e.g., Fig. 2 in Ref. 65) in relation to the intrinsic surface potential, X, of metals.
In the case of an electrode, this spillover effect depends on the electrochemical polarization of the metal so that at potentials positive to the pzc, this
spillover electron density is drawn in toward the metal while at negative potentials it is expelled outward to a corresponding extent. The variation may be over
1-2 A (0.1-0.2 nm) or so. This region over which there is an interfacial gradient
of electron density is referred to as the Thomas-Fermi screening distance (see
Fig. 4 in Chapter 6).
Development of this concept of a spillover electron density was initially
due to Lang and Kohn,lO,66 but some earlier treatment is attributed to Rice. 67
More recently, the spillover effect has been examined in more detail by Badiali
and co_workers 12,68,69 and by Schmickler,11,70 especially as it may influence
double-layer properties and the capacitance. The treatments employ the jellium
model for the behavior of the delocalized electron plasma in or near the metal
surface, taking into account the solid atom nuclei array as a diffuse positive
background, a "jellium."
As there is a variation of spillover electron density with electrode potential
around the pzc, a corresponding capacitance contribution, CM , formally arises,
which is referred to as the "metal contribution" to the double-layer capacitance.
It adds as a series contribution to the capacitance, being reciprocally additive to the
other contributions (compact layer, diffuse layer, and solvent polarization). Hence,
only when it is small does it lead to a significant contribution to the overall C.
Insofar as this metal contribution was not included in the classical conventional treatments of the double layer, taking account of it will introduce some
modification to the evaluation of the other components. For example, conventionally, the compact (Helmholtz) layer capacitance contribution, CH , is usually
evaluated from the overall Cd1 value (at a given potential) (actually as l/Cd1 ) by
subtracting a calculated value of the reciprocal diffuse-layer capacitance, since
the diffuse-layer capacitance theory is regarded as reliable at least for dilute solutions up to about 0.1 M. If an additionalllCM term should be included, obviously then some changes in lICH will result, depending how small CM is. In fact
CM is really part of CH' This complication has been pointed out and emphasized
by Badiali. 69
7.6.2. Profile of Electron Density at Electrode Surfaces
Chapter 7
158
Double-
Bulk
electron density
- t o zero
SOLUTION
METAL
metal
t ~Nominal
surface plone
L - ofLOCuS
of centers
metal surface
L---
atoms
FIGURE 7.11. Profiles of electron density at the surface of an electrode at three potentials corresponding to: (1) positive surface charge density, (2) zero charge density at the pzc, and (3) negative
charge density. (Schematic; repeated from Chapter 6.)
electron spillover toward the solution and adsorbed ions in the Helmholtz layer
region. This can influence (1) chemisorption of these ions, especially anions;
and (2) the orientation, and orientation distribution, of solvent molecules, e.g.,
Nonideal metal
Ideal metal
-cr
-cr
+cr
Za=o
FIGURE 7.12. Jellium model of charge distribution at a metal interface: left, microscopic view
of the metal interface; right, representation of the metal interface by a charged plane located at Zo.
D is the position of an ideal charged plane. (After Amorkrane and Badiali. 69)
159
H 20 dipoles (or quadrupoles). For example, on the negative side, electron spillover can influence shielding of the protons of oriented H 20 molecules at the
metal surface. Both these effects can have an influence on the interphasial capacitance in a rather complex, interactive way, and can also affect the specific
adsorption of ions, especially for anions, as noted earlier.
In the treatment of the surface region of metals from the point of view of
the electron-density distribution, the so-called "jellium model" is commonly
used. In this model, the discrete nature of the semi-infinite array of ion cores of
the crystal structure of the metal near its interface is ignored and is replaced by
a uniform, structureless background of positive charge, ending abruptly at the
metal's nominal surface (Fig. 7.12). This charge is equal and opposite to that of
the free-electron gas (approximately one valence electron per atom). Owing to
the structureless nature of the jellium, the properties of the electron-jellium sys-
-- -- ..........",
a)
....
" "
........\ ..
"
\ ".
\
'.
\
"....
\
'..
\.0.5
-10
z/eu
-5
z/eu
FIGURE 7.13. Calculated electron-density profiles n(z)/n for (a) rs = 6 and (b) rs = 2 in the jellium
model. (-) from Smith,?l (---) from Lang and Kohn. lO In (a), the dotted line shows the effect of
second gradient terms. (After Amorkrane and Badiali. 69 )
160
Chapter 7
tern are invariant parallel to the metal interface and vary only normal to it, making a substantial mathematical and modeling simplification.
A treatment of this model in an approximate way was first given by
Smith71 in 1969 and a more exact one was published in a now well-known paper
by Lang and Kohn lO in 1970. In Fig. 7.13, we show the electron-density profile
for the jellium model n(z)/n (along the normal coordinate z) for rs = 6 and for
rs = 2 according to the calculations of Smith71 and Lang and Kohn. lO The parameter rs is defined by rs = (3/4nn)I/3 where n is the bulk free-electron density,
which varies from metal to metal; in fact, rs is the radius of the sphere containing
on the average one electron. The significance of the surface-region variation of
electron density for the structure of the electrode/solution interphase, in particular that component associated with the surface region of the metal itself, has
been explained in the referenced papers and in works by Amokrane and Badiali,
as reviewed in Ref. 69.
161
responding to the inner limit of the diffuse layer for the case of finite-sized ions
(Stern-Grahame model 1,5).
The overall result of a rather complex calculation73 is that '1/ as a function
of distance r into the solution and of concentration is given by the complex expression:
COSh[U'l/I/2] - 1
2Kr=ln--~~~~--
cOSh[U'l/I/2] + 1
In cosh[u'l//2] - 1
cosh[u'l//2] + 1
(7.32)
where u = ze/kT and '1/1 is again the potential of the inner limit of the diffuse
layer at the Helmholtz plane.
The interesting point about this relation is that the final function for 'I/(r),
which appears as the variable '1//2, is equal to the product 2(Kr), i.e., K and r
scale together in determining 'I/(r), so that values of r (proportional to the square
root of concentration) are larger for smaller values of K (nll2) for a given value
of '1/, relative to '1/. The first term on the right-hand side ofEq. (7.32) for 2Kr is
a constant for given electrode polarization and solution conditions.
The above calculation also allows calculation of the zeta potential (~) associated with the distance over which the diffuse layer is mobile with respect to
the Helmholtz plane in the case of a moving electrode surface or a colloidal particle subject to electrophoresis. The result is also significant for porous doublelayer capacitor electrodes in relation to electro-osmotic flow of electrolyte upon
discharge or recharge. Several years after Eversole and Lahr's paper, Grahame,
in his review, 1 gave an approximate expression for the '1/ potential as a function
of distance across the double layer, a result that is useful for restricted conditions.
162
Chapter 7
the electrolyte concentration or ionic strength (cf. Ref. 73 and Section 7.7). In
the cases of ions in the electrolyte itself, the ionic atmosphere charge distribution, conjugate to a central reference ion's charge, extends effectively to
a 3-dimensional distance that is reciprocal in the square root of the ionic
strength according to the Debye-Htickel theory.74.75 This relation applies to dilute solutions, but for strong solutions (ca. 5 M or more), such as those employed
in capacitors, the concept of an ionic atmosphere breaks down, as it does also
for the diffuse-layer ion distribution at electrode interfaces. Nevertheless, a
rough guide is that the diffuse layer at electrodes effectively extends about 100
nm into the solution, i.e., for the If! potential to fall to within a few millivolts of
the solution potential, Js' for a 0.001 M solution of a uni-univalent electrolyte,
about 10 nm for 0.01 M and 1 nm approximately for a 0.1 M solution. For
stronger solutions (> 1 M), a rather different model consisting of a quasi-lattice
of anions, cations, and solvent molecules near the electrode interface over distances ~l nm has to be adopted 15 and a diffuse layer, in the original sense, ceases
to exist.
At an electrode interface distributed within a microporous structure, the
configuration of the distribution of electrode potential into the solution phase is
much more complex than at an infinite planar electrode. 72 The problem arises
on account ofthe thickness,76 i.e., the Debye length, (K- i ) of the diffuse part of
the double layer in relation to the width or space available for the ion distribution to be set up within the pores of the electrode.
In the original Debye-Hiickel theory,74,75 the Poisson-Boltzmann equation relating the second-derivative operator, V21f!, to the ionic space charge density became transformed to a second-order differential equation, V21f! = K21f!,
where K2 involves the ionic strength I of the solution,16 i.e., K is square-root in
I. It was also shown that K- i had the dimensions of distance, namely, the effective radius of the ionic atmosphere distribution around any given ion. A similar
significance applies to the effective thickness of the diffuse-layer ionic atmosphere in double-layer theory, which is inversely related to a function of the ionic
strength of the electrolyte.
From this it can be appreciated that in very fine pores of a porous electrode
matrix, say 2-10 nm in extent (diameter), the diffuse region of the double layer
in dilute electrolyte solutions could extend all the way across the pore diameter
and overlap72 with the diffuse layer ion distribution at the other side of the pore
(Fig. 7.14). This situation then becomes similar to that envisaged and mathematically treated by Verwey and Overbeeck in their monograph (see general
reading reference 1) on the theory of the stability oflyophobic colloids, and the
fall of potential with distance in the diffuse layer73 between the inner boundaries
of the double layer is then less than for free, noninterfering interfaces. The ion
distributions then overlap, so that the charge density and potential equations become much more complex 72 than those given in the Gouy-Chapman theory.
163
..J
<{
to-
5a.
-----/,
'-
.....
---
.,- ./
-'><-
RADIAL
DISTANCE
FIGURE 7.14. Overlap of diffuse double-layer potential profiles (and corresponding ion distributions) in an element of a pore in a porous electrode matrix (schematic).
In a real electrochemical capacitor, however, operating with acidic or alkaline electrolytes at several molar or higher concentrations, this situation of
overlapping diffuse layers will rarely arise. The net charge density on the pore
surfaces (e.g., of carbon particles) will be already almost fully screened by the
compact layer counterion charges and residually by a substantially collapsed
diffuse layer only -0.5 nm in thickness. Then the overlap situation that would
arise in dilute electrolytes will no longer apply.
Another situation of a different kind may arise, however, on charging of
high-area double-layer capacitors: depending on the electrolyte concentration
and its bulk amount relative to the weight and area of the carbon powder matrix
electrode, significant differential depletion of the electrolyte concentration can
occur because of the migration and electrostatic accumulation of cations at the
interface of the negatively charged electrode and of anions at the conjugate positively charged electrode. Then the bulk cation plus anion electrolyte concentration can be appreciably lowered; for sufficiently diminished bulk region
concentration, the diffuse layer will expand its penetration into the bulk and
some overlap effect might then become established between the sides of pores.
Similarly, the bulk conductivity of the electrolyte could become significantly diminished on charge, leading to enhancement of the internal resistance
of the capacitor device, a matter of practical significance and concern for the
maintenance of optimum operating power. Such problems can be of greater significance in the case of nonaqueous electrolyte solutions in higher voltage capacitors, and have been experimentally observed.
The situation in a double-layer capacitor is rather different from that arising at porous electrocatalytic electrodes (e.g., Raney Ni); there, in fine pores the
164
Chapter 7
165
350r----------------------------,
A-2225
BETl786, 36 (144)F/g
300
PA-1500
BET 1532,31 (124)F/g
250
A-2200
BET 1556, 37(148)F/g
C'3
2::! 150
v
o
......
0..
E 100
PF-36
BET 551, 58 (232) Fig
::l
50
o~~~~~~~~~~~~~~~
10
20
30
40
50
60 o 70
80
90 100
FIGURE 7.15. Cumulative pore surface area in a porous electrode matrix as a function of pore
width. Annotations refer to proprietory porous carbon preparations. (After Horovitz. 8l )
depth and detail than is required here to provide the basis for understanding the
origins of double-layer capacitance of supercapacitor devices.
REFERENCES
I. D. C. Grahame, Chern. Rev., 41, 441 (1947).
166
Chapter 7
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
167
168
Chapter 7
83. S. K. Rangarajan, Specialist Periodical Reports (Chemical Society, London), 7,203 (1980);
see also 1. Electroanal. Chern., 82, 93 (1977).
Chapter 8
8.1. INTRODUCTION
170
Chapter 8
2.
3.
4.
5.
Some of the nonaqueous electrolyte solution aspects of this field are dealt
with in Chapter 13, which is specifically on that topic.
Usually the solvents favored for electrochemical capacitor applications (as
for lithium batteries) are aprotic ones, i.e., those not containing electrochemically active H atoms, such as on -NHb -OH, -COOH functionalities in the molecular structure. This is an obvious requirement for higher voltage operation in
order to avoid discharge ofH2 from active protons in the structure of the solvent,
as with alcohols or carboxylic acids. The most commonly employed and practically satisfactory nonaqueous solvent for higher voltage electrochemical capacitors is propylene carbonate, which is sometimes mixed with other solvents
such as dimethoxyethane to improve the electrolytic conductivity of dissolved
R4N+ salts.
In practice, not only is the decomposition voltage of the nonaqueous solution important for capacitor design and operation, but favorable dissociating
ability for dissolved electrolytes is also required (see Chapter 13). This is also
indirectly related to the solubility of a particular electrolyte in a given solvent.
Both factors determine the conductance of the solutions as a function of concentration, an important matter for optimizing power density. These factors are determined mainly by the Gibbs energy of solvation of the ions of the electrolyte
in the particular solvent under consideration 1,2 (see Chapter 13).
Substantially less work has been reported in the literature on this topic than
on electrode interfaces in aqueous media. 3 Practically speaking, there are usually experimental problems in purifying and drying solvents, which can add
costs in producing nonaqueous solution capacitors. Various activities have been
carried on commercially in this field at carbon powder and fiber electrodes, but
they often involve proprietary knowledge and techniques. On the fundamental
research side, most work has been directed to studies of the Hg electrode interface in nonaqueous solutions as a basis for comparison with the very extensive
171
results that have been obtained for this metal in an aqueous medium; e.g., see
review Refs. 3,4, and 5.
The types of study include:
1. comparative evaluation of the double-layer capacitance of electrode
surfaces in various nonaqueous solvent media in relation to their behavior in water;
2. examination of the behavior of various electrolytes containing different
anions and cations, especially those of various tetraalkylammonium
salts and some low-melting point, pure organic salts having alkylpyridinium and alkylimidazolium structures;
3. theoretical examination of double-layer capacitance behavior in relation to solvent type and molecular structure;
4. the relation of double-layer capacitance and solution properties to energies of ion solvation 1 in various nonaqueous solvent media; and
5. adsorption behavior of various solvent molecules at electrode interfaces.
Some of these topics, however, have not been worked on in either the depth or
detail that has characterized studies of corresponding aspects of aqueous solution behavior. 3- s
Whether the solvent is structured or relatively unstructured is of major importance in the capacitance behavior of the interphase. It is also important in determining the state of solvation of ions and especially their partial molar
entropies in solution. 1,6-8 Structure in solvents is usually the result of hydrogen
bonding, although some highly polar aprotic solvents such as nitromethane and
acetonitrile can also exhibit structure due to strong dipole-dipole interactions.
Gases are unstructured, solids are highly structured; over small but significant ranges of distance, the extent of structure in liquids is experimentally recognizable (e.g., by X-ray or neutron diffraction), but only over short ranges of
intermolecular distance, two to three molecules in range. The degree of structure
in a liquid depends on the intermolecular potential energy and its orientational
dependence in the case of dipolar solvents and especially H-bonded solvents.
The extent of structure is usually related to the dielectric constant and viscosity.
Intermolecular structure in liquids can be determined by evaluating the radial distribution/unction (rdf) and the orientational distribution function around
a given central reference molecule or center. These functions are calculated
from observations of X-ray and neutron diffraction. 9 Some of the earliest work
of the X-ray diffraction type was done on water, 1,10-12 the structure of which is
related to that of ice. The diminution of short-range structure with increasing
temperature was also demonstrated l 1,l3 (Fig. 8.1). Theoretical calculations of
ChapterS
172
0.2
0.3
0.4
R Inm
0.5
0.6
(b) 83C
0.2
0.3
0.4
R Inm
0.5
0.6
FIGURE S.l. Radial distribution functions for the structure of water at 1.5 and 83C. First, second, and third nearest-neighbor regions indicated by 1, 2, and 3. Solid line, experimental curves;
dashed line as fitted by calculations of Pople and Lennard-Jones. 13 (Based on Refs. 10, II, and 13).
rdfs have also been made,12-14 with considerable success (Fig. 8.1). However,
most of this work has focused on water and ice, with a little also on methanol. 15
The structural aspects of solvent properties are important in double-layer
capacitance behavior for reasons similar to those for ionic solvation: thus, it is
either the interfacial field at a charged electrode surface or the radial field and
field gradient beyond the periphery of an ion, into solution, that tends to bring
about dipole orientation along the field lines (depending on the dipole-field energy, ,uE, compared with kD. This orientation is opposed by structural interactions, e.g., intermolecular hydrogen bonding and/or dipole-dipole interactions.
173
Adsorbed ion
co-sphere overlap
Adsorbed
j~s..- ...
Diffuselayer ions
Co-plane for
adsorbed H20 layer
FIGURE 8.2. Overlap of regions of solvent in the double layer at an electrode that are influenced
both by the double-layer field and by the field of ions in the Helmholtz layer: (a) overlap between
co-plane and ion-cosphere solvent and (b) overlap between ion solvation cospheres.
Chapter 8
174
Intermediate
Nitromethane
Acetonitrile
Propylene and ethylene
carbonate
N,N-dimethyl formamide
Dimethyl sulfoxide
Dimethyl sulfone
Sulfolane
Esters of carboxylic acids
Hexamethylphosphoramide
Dimethoxyethane
Dioxane and other cyclic
ethers
Ketones
Lactones
Pyridine
Unstructured
Hexane
Benzene
Oils
2-Methyl tetrahydrofuran
Ethers with large
R groups
Fluorocarbons
well as, in a general way, by their freezing points and viscosities. Structured solvents have high dielectric constants ca. 30 to 115 (HCN)I, and relatively long
relaxation times ca. 1O- 11 S for water. Intermediate solvents have dielectric constants from 10 to ca. 20 and relaxation times of 10- 12 to 5 x 10-13S. Unstructured
solvents have low dielectric constants, 3-8.
The bulk dielectric constants in the above ranges might be expected to be
mirrored in the capacitances and the dielectric constants of double layers containing the respective solvents, but exact values are not directly known. They
can be estimated roughly from measured capacitances and assumptions about
double-layer thicknesses in various solvent media.
However, some of the earliest work on the double layer at electrodes in
nonaqueous electrolytes by Frurnkin 16 showed that the capacitance at Hg did not
correlate with the bulk dielectric constants of the solvents employed. Perhaps
this is not surprising since the behavior of solvent molecules in the field of the
interphase and adjacent to a metal surface depends more on microscopic molecular properties, such as the dipole moments of individual solvent molecules,
or of their polar functional group(s) (e.g., OH, NHz, CN) and their lone-pair
donicities, rather than the behavior of an interacting bulk assembly, the overall
properties of which determine the dielectric constant.
The origin of this absence of systematic correlation is the lack of a direct
dependence of the dielectric constant of solvents on their individual molecular
dipole moments, especially for associated liquids. In the case of vapors, their dipole moments can be derived directly from the temperature dependence of the
175
respective measured dielectric constants, based on the theory of Debye and the
Clausius-Mosotti equation (Chapter 5). However, for liquid-phase dielectrics
and for vapors at relatively elevated pressures, the relationship between dipole
moment and dielectric constant is more complex and its interpretation depends
on models of orientational correlation associated with short and longer range
molecular interaction, especially in the case of the associated solvents.
A further complication arises at charge densities where significant lowering of the dielectric constant values can take place due to dielectric saturation
effects 17- 19 at high interphasial fields resulting from the high surface charges.
Similar dielectric saturation effects occur at solvated ions in the solvation cospheres and have been treated theoreticallyp,20,21 In fact, the dielectrically polarized layer of solvent next to a charged electrode interface behaves quite
similarly to that existing around and near an ion in its solvation coordination
shell (see Fig. 8.3). Dielectric saturation effects tend to diminish the capacitance
of an electrode interface. There are also compressional effects (electrostriction 22,23) that increase the molecular density of solvent in the double layer,
which tends to change the capacitance in the opposite direction. 23
For the same reasons that protic solvents cannot be used for Li batteries,
because of potential electrochemical discharge of H atoms bound at electronegative atom centers, similar restrictions apply to the choice of nonaqueous
solvents and their electrolyte solutes for high operating (cathode) voltage electrochemical capacitors. Also, solvent oxidation on the positive plate of a
charged capacitor can introduce some other limitations. These two effects are
determined by the electrocatalytic properties of carbon surfaces for reduction or
oxidation if potentials adventitiously exceed those for solvent or solute discharge on accidental overcharge of the capacitor device. Therefore, in practice,
there are only a few nonaqueous solvents, or mixtures thereof, that can be reliably used for high-voltage capacitors. Usually the preferred choices are propyl-
-'
I
I
\
-+_1
~\
\;'
",
---- -
a) Ion
:--/
-- --r-~ , ,
+
+
+
+
~
+ ...+
--I'
--
b) Charged electrode
FIGURE 8.3. Models of approach to dielectric saturation at (a) an ion (in its primary solvation
shell) and (b) in the charged double-layer interphase (Helmholtz-layer ions not shown).
176
Chapter 8
One of the earliest works on double-layer capacitance in a nonaqueous medium was that by Frumkin l6 on mercury. As mentioned earlier, Frumkin made
the significant observation that double-layer capacitances of the Hgisolution interface for various nonaqueous solvents did not parallel the order of the respective bulk dielectric constants of those solvents. Ocrent24 recorded
electrocapillary curves for alcohol-water mixtures and Payne25 ,26 has published
a useful review on double-layer work in nonaqueous solutions up to 1975. Another review of the whole field, with an extensive bibliography larger than it is
appropriate to give here, was published by Parsons 2 in 1976.
Damaskin and Frumkin 27 and later Parsons 28 ,29 provided an explanation of
the effect of the nature of the electrode metal on the form of capacity curves (in
this case for water) by assuming that the solvent dipole orientation treatment of
the Watts-Tobin model (Chapter 7) could be applied to small clusters of water
molecules together with small numbers of monomeric water molecules adsorbed at the interface. The model also gave a good account of the temperature
dependence of the double-layer capacitance of the Hg interface in aqueous solutions. These treatments were thus based realistically on the associated nature
of the water solvent and the temperature dependence of its structure. l
The treatment by Damaskin and Frumkin27 and in particular its improvement by Parsons,28 takes into account four states of the solvent (water) in the
double layer: free H 20 dipoles and small, H-bonded clusters, each in two supposed orientations; this was outlined in Chapter 7.
177
Parsons 2 classified the qualitative behaviors of the double-layer capacitance behavior of nonaqueous solutions at Hg as a function of surface charge
density, q, (compared with water) into three classes, as shown in Chapter 7:
1. Waterlike capacity curves: these have a hump near the pzc and minima
either side, that on the positive branch being at higher C than that on the
negative branch. In addition to water, this group includes N-monosubstituted formamides and related amide solvents, as well as dimethyl sulfoxide.
2. Singly humped capacity curves: these are related to the first group since
a hump is observed. The minima are absent, probably because of experimental limitations to the ranges of potential that could be covered. The
solvents representing this case, however, have a common feature in that
they are all cyclic, e.g., ethylene and propylene carbonates, sulfolane,
4-butyrolactone, and 4-valerolactone.
3. Water-unlike capacity curves. These show only a single, parabolalike
minimum, and include ammonia, alcohols, N,N-dimethylformamide
and N,N-dimethylacetamide, acetonitrile, acetone, ethyleneglycol and
pyridine-a series of solvents that show hardly any chemical molecular
similarities!
As shown in Chapter 7, the form of the aqueous solution capacity curves
for Hg can be fitted quite well by the four-state model of Darnaskin and
Frumkin 27 and better by the improved treatment by Parsons. 28 Therefore it was
of interest2,3o to examine how well this type of treatment fitted data for other solvents ofthe group (1) class. In fact, the fits are quite successfueo except that for
the solvent formamide (hydrogen bonded); although the general form of the capacity vs. potential profile is accounted for (Fig. 8.4), the potential ofthe hump
is substantially shifted from the pzc of Hg in this solvent. The experimental data
are those of Dutkiewicz, Garnish, and Parsons. 3!
For group (2) behavior, a similar comparison between experimental data
and the fitted curve is shown in Fig. 8.4 for the inner-layer capacitance at Hg in
ethylene carbonate from the results of Fawcett and Mackey.3o The curve has a
broad maximum (hump) and a long tail on the negative branch. These data are
best fitted with a W atts-Tobin two-state model,32 as improved by Levine, Bell
and Smith. 33
Group (3) behavior can be represented by that of methanol using Graharne's data for that solvent at Hg?5 The experimental and fitted behavior is
shown in Fig. 8.5. The general shape of the curve with a deep minimum negative
to the pzc is well fitted but, in detail, at larger positive and negative surface
charges, not so accurately. It is significant that curves of this type cannot be ac-
178
Chapter 8
?:'
\,
'g 20
'l!
D.
a.
c
u
GI
.2 15
"."-.
"..,..
'."'0
GI
.......
.5
+10
q1fL C
cm- 2
-10
FIGURE 8.4. Inner-layer capacity of an Hg electrode in ethylene carbonate (circles indicate the
results of Fawcett and Macke/o) fitted by a theoretical model based on the treatment by Levine,
Bell, and Smith. 33
34
...
cu
>-
CI.I
j5 16
::l
Cl
10
FIGURE 8.S. Differential capacity of an Hg electrode for 0.1 M solution of NH4F in methanol
(results of Grahame 35 ). Note difference of shape from aqueous solution curves.
179
CH -CH-CH
3
0,
'c"""
1\
1. Propylenecarbonate
II. y-Butyrolactone
counted for by other theories except by introduction of some charge (field)dependent electrostrictive effect. 22.23
Of the various nonaqueous solvents in which the double-layer capacitance
ofHg has been determined (see review in Ref. 25), Payne's results 34 for propylene carbonate (see diagram I) and y-butyrolactone (see diagram II) are probably
of most interest for the development of electrochemical capacitors. Propylene
carbonate (I) is the cyclic carbonic (di-)ester of 1, 2-propanediol, while ybutyrolactone (II) is the internal (cyclic) ester of y-hydroxy-n-butyric acid
(i.e., a y-lactone). They have obvious chemical similarities and solvent properties, as well as relatively high dielectric constants (65 and 39, respectively).
In Payne's experiments,34 0.1 mol dm- 3 KPF6 was used as electrolyte.
Both solvents approach a maximum in capacitance of the Hg electrode at about
+0.4 V vs. the potential of the normal calomel electrode (in water) (Fig. 8.6).
The capacitance hump lies positive to the potential-of-zero charge. The profiles
30
lI..
E
o
20
a.
o
...
CLI
>-
30
~~.................
BL ,,-' ....
::t.
>.
1\",
_. . .,J-._
.... \ ..
EC '
,/
VL
\\.
\\
20
\ ,\
CLI
.
,'.
' ....~~-.- ........ _
.2 10
:0
,,
-BL
'-. -'-'-EC-
',......... --VL-
10
180
Chapter 8
of capacitance vs. potential from +0.6 V to -1.5 V, vs. the above calomel reference potential, follow approximately the same shapes. Toward the more negative potentials, the capacitance falls to 8 or 13 f1F cm-2, which is not too
different from the values for water at Hg under similar conditions(16 f1F cm-2).
Work on the amide solvents has been more extensive. 25
Capacitance humps
are observed on the negative side of the pzc, but the profile for N,N-dimethyl
formamide is more featureless.
The amide solvents have, however, been little used in preparing electrolyte
solutions for electrochemical capacitors and their stabilities are less than that of
propylene carbonate, which is the favored solvent, sometimes in mixtures with
dimethoxyethane (DME). Such solutions, using R4N+ salts as electrolytes, have
relatively high decomposition potentials in the anhydrous state and thus enable
electrochemical double-layer capacitor devices to be operated up to a plate voltage difference of about 3.5 V per cell, or a swing voltage of 1.75 V.
In conclusion, a useful review on the behavior of non-aqueous solvents at
metal interfaces, by Trasatti,36 should be mentioned.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
181
Chapter 9
9.1. INTRODUCTION
9.1.1 Historical
A substantial fraction of the whole body of electrochemical capacitor development has depended, and continues to do so, on an understanding of the
properties of carbon, especially in its more dispersed and conducting forms, including materials having fibrous or foamlike morphologies. It has been known
for many years that there are two or really three natural allotropic forms of carbon: diamond, graphite, and the fullerenes, as noted in an earlier chapter. The
graphite modification is crystalline and a moderately conducting, anisotropic,
electronic conductor. Diamond is a highly ordered and insulating material. In
addition, amorphous forms of carbon are well known, and transitions between
amorphous and graphitic structures are possible at elevated temperatures, but
are possible for diamond only at extremes of conditions. In more recent years,
other modifications of carbon have been prepared or become known, namely,
the fullerenes (which may exist in interstellar space), glassy carbon, and exotic species such as C 2, which was also detected in interstellar space. Carbon "blacks" such
as Shawinigan black or acetylene blacks (soots) are well known and have relatively
high specific surface areas in the range of 30-1000 m2per gram. However, they are
not usually suitable for fabrication of capacitor electrodes.
Carbon has a unique position and electronic structure (ls2, 2s2, 2p2) in the
Periodic Table, being midway along the first eight-element period, so it has in183
184
Chapter 9
termediate electronegativity and fourfold bonding possibilities in saturated valence interactions, as in diamond and the lowest hydride, CH4, in which both
structures are built with four equivalent covalent bonds hybridized between the
two 2s and the two 2p orbitals. As a result of its special position in the Periodic
Table, carbon, like silicon but less so, has some semiconducting properties and
the band structures of its various forms have been treated in a number of publications, e.g., Refs. 1,2, and 3.
For electrochemical double-layer capacitors, an activation pretreatment of
existing carbon materials made initially from thermal carbonization of coal,
pitch, wood, coconut shells, or polymers such as polyacrylonitrile, is the preferred procedure. This is necessary to achieve the best real areas and optimum pore
accessibility and hence capacitance per gram, the best stability against degradation
with time under cycling, and minimal self-discharge rates after charging.
The pretreatment or activation procedures are often proprietary so that
their details are usually difficult to specify. However, methods include heat
treatment up to 2000-2800 C in an inert (Nz) or an Hz atmosphere, or in vacuum. At such temperatures, micro graphitization can occur and some surface or
edge functional groups are dissociatively distilled or pyrolyzed off. Alternatively, treatment in COz or hydrothermal treatment in steam at an elevated temperature (-1000 0c) is also effective in changing or removing surface oxygen
functional groups and opening up microporous structures, e.g., in fibrous carbon
materials. Under these conditions, reactions between the carbon and HzO or
CO 2 are possible, together with changes of unsaturation of ring structures in
graphite crystallites. Other oxidizing treatments (e.g., utilizing chromic or nitric
acids) are also sometimes used in some applications.
Two principal morphologies of carbon for capacitor electrode fabrication
are usually employed: high specific-area material in powder form or high-area
carbon fibers, which give rise to good mechanical integrity of fabricated electrode structures and good electrical conductivity. Recently, Hyperion Catalyst
Corp. has reported the production of microtubular carbon fibers (related to the
fullerenes) that have high specific area and lead to excellent porous structures
for electrochemical capacitors (see Chapter 19).
Extensive exploratory research has been done in recent years on the electrochemistry of fullerenes, but they are unlikely to become materials for capacitor electrodes owing to the volumes that would be required and their relatively
high cost. Work on fullerenes has been reported in a series of well-supported
symposia at Electrochemical Society meetings over the past 6 or 7 years.
Apart from the patent in 1957,4 it appears that the first activity in the use
of high-area carbon for a double-layer energy storage device was that by Sohio
in the early 1960s. This has been referred to in a paper by Currie5 in which the
author records some of the early work by Sohio on fuel cells fabricated from
electrodes made from high-area carbon particles bearing a catalyst loading. The
185
capacitor development originated directly from the experience gained in the fuel
cell electrode field under the direction of Rightmire, who realized that a 400
m 2g- 1 porous carbon material would give rise to a double-layer capacitance of
80F g-l, assuming a specific surface capacitance of 20 f.1F cm- 2. An energy storage device for electric vehicles, employing large capacitors, was already envisaged in the early 1960s.
A brochure published by Sohio Corp., dated April 25, 1969, described a
so-called "electrokinetic capacitor" having the capability of substantial energy
storage. Twenty 100-f.1s discharges at 20 A or 66% of the decline of initial voltage could be achieved in 20 f.1s.
Sohio Brochure for Double Layer Capacitor
GOHI~
The choice of name, "electrokinetic capacitor," for the device was perhaps
inappropriate however: the charging or discharging processes in a double-layer
electrochemical capacitor involve virtually no electrode kinetics, Faradaic processes are not usually involved to any substantial extent in such devices except
on overcharge. The actual term "electrokinetic" derives from colloid science
and is connected with the movement of colloidal particles in an electric field-a
very different phenomenon from double-layer charging. The front page of Sohio's announcement brochure (shown in a publication by Argade from Technochern Corp.) is reproduced above, with permission.
Compared with the extent of research on the interfacial capacitance of mercury (Chapter 6) and gold and platinum, work on the double layer at carbon has
been substantially less both in extent and depth of interpretation, except in work
by Yeager et a1. 1,2,6 Nevertheless, important results relevant to the use of carbon
186
Chapter 9
materials in electrochemical double-layer capacitors are to be found in the literature, while the monograph by Kim Kinoshita on carbon3 is an important
source book for information on the electrochemical and physicochemical properties of carbon in its various forms and the commercial sources of modifications of carbon. Some of the latter are important in electrochemical capacitors
or primary batteries. See also General Reading Refs. 2, 3, and 4 in this chapter.
Carbon materials for electrochemical capacitors are far from being electrochemically inert and this reactivity of carbon has been extensively studied (see
review material in Kinoshita3). Oxidation or reduction of the redox functionalities that are initially present in various carbon formulations gives rise to one
kind of electrochemical reactivity while more extended voltage excursions (e.g.,
in cyclic voltammetry or in overcharging or overdischarging carbon capacitor devices) can cause additional introduction of O-containing functional groups or electrochemical hydrogenations in the opposite (cathodic) direction of polarization.
Unlike the case of most metals, the element carbon exists in several allotropic forms with very different properties, as referred to earlier. With the exceptions of diamond and the fullerenes, the other modifications of carbon are
variously used in construction of electrochemical capacitors and of course in
batteries of the Leclanche or alkaline types involving Mn02> and in some Li primary batteries, as well as support materials for fuel cell catalysts and conducting
polymer capacitor (Chapter 12) matrices.
For electrochemical capacitors, the carbon for the double-layer type must
have (1) high specific real areas, on the order of 1000 m 2 g-l; (2) good intra- and
interparticle conductivity in porous matrices; and (3) good electrolyte accessibility to intrapore surface area (Chapter 14). In addition, the surface conditioning of powdered or fibrous carbon materials for capacitor fabrication is of
substantial importance for achieving the best performance, namely, good specific capacitance and conductivity and, most important, minimum self-discharge rates. This means the preferred carbon materials should be free from
impurities (e.g., Fe species, peroxides, O 2 , quinones) and surface quinonoid
structures that can set up self-discharge processes that must be minimized
(Chapter 18).
Because of residual surface "valencies," most carbons that have been exposed to air (which is the usual situation) contain adsorbed (mostly chemisorbed) oxygen, which can lead to various O-based surface functionalities 3 ,7
(ketonic, peroxyl, hydroxyl, quinonoid). It is these types of surface state that can
be modified by high-temperature treatment.
187
188
Chapter 9
FIGURE 9.1. Crystal structure of the benzenoid layers of graphite and an element of their conjugated bonding.
edges of the 2d-structure planes in the case of graphitic materials), their relative
proportion in the material tends to scale with the specific surface areas (cm2 or
m2 g-1) of the material. For such redox reactivity, a significant redox pseudocapacitance, CtfJ (see Chapter 10) arises as well as the desired predominant doublelayer capacity, Cdl In fact, with some carbon materials, it appears that CtfJ can
amount to some 5-10% of the total realizable capacitance, and scales approximately with the Cdl value. However, the distinction between the C and Cdl contributions (in principle possible by means of impedance analysis), becomes
difficult when the interfaces are part of a porous matrix system. The extent of
this scaling depends very much on the nature and conditions of high-temperature pretreatment the carbon material may have undergone prior to fabrication
189
ketone
phenolic
carboxylic
carbinol
o-hydroquinoid
lactone
into a capacitor device and its prehistory of exposure to atmospheric air, leading
to oxygen chemisorption.
In addition to electrochemical (redox) activity, various carbon materials
exhibit titratable acidic and/or basic functionalities related to some of the surface structures 7 illustrated in Table 9.1. Such surface structures influence the
wettability of carbon materials, a matter of considerable importance in maximizing the access of electrolyte to the surface of carbon modifications in the
preparation of electrochemical capacitor matrices. Note that the requirements
190
Chapter 9
for capacitors in this respect are the opposite of those preferred for gas-diffusion
fuel cell electrodes, in which hydrophobic additives are usually incorporated to
maintain a three-phase gas/liquid/solid interface in the electrode matrix.
Edge quinonoid groups that can exhibit redox C activity are also titratable
as acidic functions in their reduced form (=0
-OH ~ -0- + H+). Acidic
phenolic and carbonxylic surface groups probably have only acid or base activity.
The presence of various polar surface functionalities on graphitic materials
thus not only affects the potential-of-zero charge (Chapters 6 and 7) but also the
contact angle with the electrolyte under conditions of incomplete wetting of
pores in capacitor-type, high specific-area electrodes. A correlation between the
contact angle of water and carbon, and the quantity of oxygen and hydrogen associated with carbon blacks was indicated in work by Studebaker and Snow. 9
As might be expected, electrochemical oxidation of carbons increases their wettability due to an increase in 0 content associated with hydrophilic surface
groups. Electrochemical reduction has an opposite effect. These results were
obtained in work by Kinoshita and Bett. lo Related to the wettability is the heat
of immersion of carbon black ll in a slurry, which is substantially larger at low
(2-3) than at neutral pH, indicating the role of basic oxygen groups. An important point for electrochemical capacitors is that the accessible porosity, which is
determined by wettability, will influence the double-layer capacitance that can
be obtained, especially in fine-pore, high-area materials.
A voltammogram for the processes occurring on a Union Carbide ZTA
grade of graphite, studied by Blurton,12 is shown in Fig. 9.2. Figure 9.3 shows
comparative behavior at a heat-treated Vulcan carbon. An interesting related series of results on a carbon fiber electrochemically oxidized to various potentials
in dilute aqueous HN0 3 were given (Fig. 9.4) by Kozlowski and Sherwood,13
who examined the CIs XPS spectra. As potential was increased, component species having lower electron binding energies progressively increased; increase in
carbonyl- and quinone-type functionalities with increasing potential (up to 3.0
V) corresponds to the changes observed in the XPS spectra. This interpretation
is not entirely supported, however, since in other XPS work,14 increasing electrochemical oxidation from 0.5 through to 3.0 V in H 2S04 indicates the dominance of )C=O groups at low potentials but an increase in C-O- (ether)
groups at more positive potentials, perhaps a chemically surprising result.
The identification of various surface functionalities on carbon materials
has also been conducted by means of infrared and Raman spectroscopy. These
techniques provide rather detailed information on surface groupings that are
characteristic of various types of carbons (blacks, C fibers, graphitic materials,
activated carbons) and their pretreatments. A detailed account of this work, together with examples of the characteristic spectra observed, is to be found in
Chapter 3 of ref. 3.
191
250
200
150
100
-E
0
=l
---c
Q)
t)
50
0
-50
-100
-150
-200
0.25
0.65
1.05
Potential V
1.45
FIGURE 9.2. Voltammogram for ZTA grade Union Carbide Corp. graphite in 0.5 mol dm- 3
aqueous H2S04 at 303 K. The sweep rate is 10 mV S-I. Reprinted from K. F. Blurton, Electrochim.
Acta, 18, 869 (1973), with permission from Elsevier Science.
1.2r----.-----.----.-----~--~r_--~----~
U
"C
0.8
0.4
<t
E
.....
w
a::
a::
-0.4
....Z
::>
.~
"C
-0.8
..c
-1.2 :----='::------::-'-:---~L:----_::l::---...L-----.JL---l
0.2
0.6
O.S
1.0
1.2
1.4
POTENTIAL/V
FIGURE 9.3. Cyclic voltammograms for the oxidation of heat-treated (2700 DC) Vulcan XC-72
carbon at + 1.2 V in 95% H3P04 at 160 DC. Increasing current responses correspond to increases in
anodic oxidation time: 0,5,60, and 250 min. The sweep rate is 14 mV s-l. Reprinted from K. Kinoshita and 1. A. S. Belt., Carbon, 13,403 (1975), with permission from Elsevier Science.
Chapter 9
192
Cd)
(a)
2.0 V
0.5 V
l!l
'c
:::l
(b)
1.0 V
?:
c
'(jj
(C)
(f)
3.0 V
1.5 V
290
285
290
Binding energy I eV
285
FIGURE 9.4. XPS CIs spectra of carbon fiber electrochemically oxidized at various potentials
vs. SCE for 20 min. in 0.0027 mol dm-3 aqueous RN03. (From Kozlowski and Sherwood. 13 Copyright 1984 Royal Society of Chemistry.)
TABLE 9.2.
193
Intercalate
Li, Na, K, Rb, Cs
Limiting composition
C 6Li, C1zLi, C 18Li, C64Na, CgM,
C z4 M, C 36M, C 48 M, C 60M
(M =K, Rb, Cs)
(CF)n' (C 2F)n' C 4F, CgBr, C 8Cl, CsICI
C 446(CuBrz)
C 6.g(Cr0 3), C18S0(Sb204)
Note: Compositions from K. Kinoshita, Carbon: Electrochemical and Physiocochemical Properties, Wiley,
New York (1988).
A substantial amount of work on the double-layer capacity of graphite basal planes 1- 3 edge sections, and powder materials has been reported in the literature (see Ref. 1 for a detailed tabulation). Unfortunately, values of C are
194
Chapter 9
rather discrepant, varying from 16 to 60 f1F cm- 2 for the basal plane (polished)
of ordinary pyrolytic graphite in 0.9 M aqueous NaF to 60 f1F cm-2 on a similar
surface in 0.5 M aqueous KCl. However, on a cleaved basal plane (which is
likely to be cleaner), a low value of 12 is reported for 0.5 M aqueous KCl. For
stress-annealed pyrolytic graphite, values of C of around 20 are reported! for a
peeled-off basal plane film in aqueous NaF, NaOH, and H 2S04 , while for polished edge-plane sections, higher values from 50 to 70 f1F cm-2 are found. 2
Carbon blacks have C values from ca. 4.5 to 10,11,18 "active" carbons
around!9 10 to 15 f1F cm- 2, while graphite powders exhibit C values of 20 to 35
f1F cm- 2. On the other hand, "graphite cloth" TGM 285 exhibits!810w valuesbetween 1 and 3 f1F cm- 2. This is a surprising result and may be a consequence
of uncertain evaluation of true real areas of test samples or of surface contamination by nonpolar organics in the preparation of the cloth. The specific capacitance values for powder materials are given here per real square centimeter, but
this is usually based on some not always reliable "real surface area" measurement per gram. Thus the derived apparent capacitances per "real square centimeter" may not be entirely reliable and can depend on pore accessibility and
pretreatmen t.
Another problem is the necessity for specifying the potential range of
measurements (around the pzc). Differences in C by a factor of 2 between potentials that are substantially positive and substantially negative to the pzc can
arise at well-defined surfaces, as is known for Hg.20
Generally, the quality of experimentation and control of conditions in experiments 4 at carbon have not been at all as good or rigorous as work at Hg or
at single crystals of Au, e.g., in work by Hamelin.2! Other problems arise because carbon surfaces are not as ideally polarizable (see Ref. 20) as Hg or Au
because of surface oxidation reactions and the possibility of intercalation processes taking place, especially upon cathodic polarization. Thus the electrochemical characterization of carbon/solution interfaces is substantially more
complex than that for various metals, especially the noble ones and mercury.
Despite the reservations expressed here, a series of interesting papers was
published by Soffer et a1. 22- 3 ! on porous carbon materials, some of the earliest
in the field, about the time at which Sohio began to develop carbon double-layer
capacitors. Soffer and Folman22 realized that at high-area porous carbon electrodes, the uptake of ions from solution as a carbon interface was being charged
could be directly measured, in their case by determining the changes of conductivity of the neighboring solution. Thus, using 2 g of a carbon electrode having
500 m2 g-! real area and an integral double-layer capacitance of 10 f1F cm- 2,
they showed that the electrode would adsorb 1 mmol of ions for a I-V change
in potential. (They later developed this phenomenon for selective removal of
ions from bulk solutions.)
195
In another part of this work in 1971 they demonstrated how an electrocapillary curve (surface free energy vs. potential) for a solid porous carbon electrode
(product FC-12, St. Mary's, Pennsylvania) could be determined by directly
measuring dimensional changes of the electrode. The carbon material used had
an electrocapillary maximum potential (pzc) of 0.080 0.01 V on the SCE
scale. Of course, such pzc values will depend on the source of the carbon, its pretreatment, and the chemical nature of surface functionalities. 7 ,32
Among the papers 22- 31 by Soffer et al. was an important one,24 relevant to
supercapacitors, on double-layer capacitance and rates of charging of ultramicroporous carbon fiber-type electrodes. Using linear sweep voltammetry, they
observed adsorption of ions from solution into pores as small as 0.37 nm. The
conductivity of the solution in pores less than 0.7 nm in diameter was found to
be several orders of magnitude lower than in a supporting 0.1 M NaCl freely in
contact with the outer interface of a porous carbon electrode. Highly oxidized,
ultramicroporous carbon materials show substantially lower double-layer capacitance at positive polarizations than nonporous carbon surfaces. It was concluded that ions can penetrate the smallest pores after depletion of their
hydration shells, as do solvated Li+ ions when they become intercalated into
TiS 2 or C hosts in hi-ion anodes. Thus, the double-layer charging rate in the microporous carbon is much lower if the NaCI is replaced by 0.1 M LiCl. The surface dipoles of chemisorbed oxygen were supposed to lead to repulsive
interactions with the electrolyte anions.
Significant semiconducting properties are exhibited in certain carbon materials having relatively small electronic conductivity.l,2 In such cases, there is
a distribution of charge carrier concentration (charge density) within the bulk of
the carbon material, giving rise to a space charge inside the carbon/solution
boundary. As with regular semiconductors, this leads to a drop in potential
within the material and because of the electrode potential dependence of the
space charge density, a space charge capacitance, Csc> is developed. Its behavior
is exactly the same as that of the ionic diffuse-layer capacitance outside the carbon/solution interface, except for the dielectric constant of the medium in which
the charges reside. The same Gouy-Chapman mathematics (Chapter 7) apply
to the treatment of this space charge region, employing the Poisson-Boltzmann
equation. Further details are described in Section 9.6.
Since Csc is in a series relation with the other components of capacitance,
it adds to them in a reciprocal manner and is hence only significant in the overall
C when it (Csc ) is small; thus:
1
1
CH
Cdiff
Csc
-=-+--+C
(9.1)
196
Chapter 9
such materials to behave in an almost metallic way (metalloid behavior). However, there are some experimental indications 1,2 for the basal-plane from double-layer capacity measurements that semiconductor properties are in fact
exhibited.
Because of the asymmetric structure of crystalline graphite in three dimensions, there will be corresponding differences in the electronic work function of
the basal plane and edge sections. According to general principles of interfacial
electrochemical science,33 it follows that there must be related differences of potential-of-zero charge. This then leads to specific differences in anion adsorption as a function of graphite-crystal electrode potential, on the basal plane
relative to edge sections. Of course, in powdered graphite preparations, these
differences in properties will be unevenly distributed, causing heterogeneity in
surface properties and interfacial electrochemistry.
A very large amount of work has been done on oxidation of carbon including of course the ultimate oxidation processes, combustion to CO and CO 2,
Electrochemically, oxidation is much more restricted, especially at ordinary
temperatures, but fuel cells employing a coal slurry have been investigated in a
number of works and applications.
For carbon-based double-layer capacitors, the state of superficial oxidation
especially at edge planes, is of most significance because this can determine the
charge accommodation per gram and the corresponding capacitance. It can also
influence the self-discharge characteristics (Chapter 18) of carbon capacitors
and how carbon materials can be conditioned by heating them to elevated temperatures in a vacuum, nitrogen, or hydrogen, and sometimes in water vapor.
One of the earliest conferences devoted to the state of carbon was the Faraday Discussion in 1937 on chemical reactions involving solids, published in the
Transactions of the Faraday Society. This volume included a classic paper by
Meyer34 on surface reactions of graphite with oxygen, carbon dioxide, and
water vapor; and other papers on the characterization of surface groups,? (e.g.,
ketonic, ketenic, and quinonoid) on carbon, including the possibility of formation of surface peroxide groups due to direct combination with dioxygen molecules, either bridged or end-on.? At elevated temperatures (1450 DC), some
reaction to form surface groups [e.g., a-diketo functions (ortho-quinone)] occurs with release of H 2, though some rupture points at edges lead to decomposition of such groups.
Under conditions of anodic electrolysis, the a-diketo grouping on graphite,
which corresponds to an ortho-quinone structure, is capable of being oxidized. 32
At low current density in H 2S04, the maximum volume of CO generated is
197
equal to that of CO 2, This process occurs at rupture points on the edges of the
graphite planes, with breaking of the outer C-C bonds; a new keto group is
formed in the graphite structure while the outer keto group is oxidized to a carboxyl group.
Sihvonen proposed further complex surface-edge chemical reactions?2
Figure 9.5 is a schematic diagram of various reaction product groupings at the
surface of an element of the graphite structure. As indicated earlier, these structures also determine the electrochemical interfacial state of graphitic carbon materials and their double-layer properties, including pzc values and specific
adsorption of ions. It must be emphasized, however, that a variety of high specific-area carbon preparations used in fabricating electrochemical double-layer
capacitors are not graphitic in structure, though some of the surface groupings
referred to here may be present.
In other work, the role played by surface oxides in the oxidation of carbon
was treated by Strickland-Constable,? who emphasized the significance of
:::;C=O groups on the surface, like those in the structure of acetone. He concluded that the oxidized state of carbon surfaces consists of covalently bound
surface compound species. This conclusion was in agreement with Schilow,35
who showed that when outgassed charcoal was treated with molecular 2, surface compound species developed that had both acidic and basic properties, but
such species corresponded to only a small fraction of the bound oxygen. In this
work it was indicated that amorphous carbon was still mainly graphitic, but con-
FIGURE 9.5. Surface-edge chemical-oxidation structures and related reaction products developed at graphite. After Sihvonen;32 in the original paper, Roman numerals identified various reactions and functionalities at the surface that are referred to in that paper.
198
Chapter 9
199
Usually in descriptions of the properties of double-layer capacitor electrodes, average values of specific capacitance, about 20 to 25 f.lF cm- 2, are
quoted. Kinoshita tabulated 3 a series of values for various carbon materials. A
detailed and informative experimental study of the capacitance vs. potential for
the basal plane and edge orientation of stress-annealed pyrolytic graphite was
carried out by Randin and Yeager 1.2 and also reported by Yeager, Molla, and
Gupta. 6
The results, which are based on many experimental points over a potential
range of 0 to 0.5 V (vs. NHE), demonstrate symmetrical V-shaped curves with
minima at 0.0 V vs. NHE and a virtual independence of the capacitance curves
on the chemical identity of the electrolyte (0.5 M aqueous H 2S04 , 1.0 M NaOH,
0.9 M NaF, 0.8 M NaF + 0.1 M NaI, and 9.0 M NaF + 0.01 M NaI). Selected
results 6 are shown in Figs. 9.6 and 9.7. Minimum values of capacitance (near
0.0 V, vs. normal H2 electrode) of only ca. 3.1 (Fig. 9.7) or 2.9 (Fig. 9.6) f.lF cm-2
also arise. It is to be presumed that this potential of the minimum is the potentialof-zero charge.
In comparison with behavior at Hg20 or Ag (Chapter 6), these results are
completely different. The capacitance in solutions of moderate or high ionic
strength exhibits a minimum rather than a maximum20 (near the potential-ofzero charge) and the curves vs. potential are insensitive to the type of electrolyte,
especially to type of anion.
Boronated (0.3%) graphite exhibits more structure in the capacitance profiles 6 and an appreciable dependence on the type of electrolyte (0.5 M aqueous
H 2S04 relative to 1.0 M NaOH, Fig. 9.7). The capacitance values are also somewhat higher, 4.9 to 6.5 f.lC cm- 2, but still substantially lower than at Hg. Cyclic
Chapter 9
200
4.0f-
3.0f-
+0.5
E/V
VS
-0.5
NHE
FIGURE 9.6. Capacitance vs. potential (NHE reference) curves for basal plane of stress-annealed
pyrolytic graphite in NaF + Nal solutions (pH 6) at 298 K and 1000 Hz. x = 0.9 mol dm-3 NaF; /}.
= 0.8 NaF + 0.1 NaI; 0 = 0.9 NaF + 0.01 Nal. (From Yeager, Moller, and Gupta. 6 Reproduced by
permission of The Electrochemical Society, Inc.)
voltammograms taken at 0.1 V s-1 at these surfaces are also shown in the figures 6 ; they are rather featureless.
On the edge orientation, however, the capacitance curves show much more
structure, as a function of potential; also, there is a major dependence on the type
of electrolyte (aqueous 0.05 M H2S04 and 1.0 M NaOH) and, in addition, the
capacitance values are much higher, being in the range 70 to 100 flF cm-2. The
behavior in 1.0 M NaOH exhibits a maximum around -0.4 V (vs. NHE), which
is reminiscent of behavior at Hg.2o The symmetry of the curves for the basal
plane, the low minimum of the capacitance values, and the non dependence of
the behavior on the presence of adsorbable anions clearly indicate that it is not
a graphite/solution double-layer capacitance that is being probed but rather a
space charge on the inside of the graphite basal plane interface.
201
7.----,-----------,-----------.----------,
BORONATED (0.3 %)
6
C\I
Ie
u
IJ...
:t
......
,,
,,
,,
' I.
"
........
_--/
//
//
WITHOUT BORON
,/1 M NAOH 10.5 M H2S04
3~----L-----------~----------~--------~
0.5
-0.5
E V vs NHE
FIGURE 9.7. Capacitance vs. potential curves (NHE reference) and current vs. potential curves
below, for basal plane of boronated and nonboronated stress-annealed pyrolytic graphite at 298 K
and 1000 Hz. Compare with Fig. 9.6. (From Yeager, Moiler, and Gupta. 6 Reproduced by permission
of The Electrochemical Society, Inc.}
(9.3)
where IfI is the potential at the surface and Ce is the electronic charge carrier concentration taken as 6 x 1018 cm- 3 and a dielectric constant, B ::: 3.0.
It is useful to note here that the space charge capacitance Csc behaves like
the ionic diffuse-layer capacitance20 in metal-electrolyte solution behavior at
202
Chapter 9
metals (see Chapter 6); it has rapidly rising values on each side of the minimum
at the potential-of-zero charge. Also, since it is combined in series with a Helmholtz layer capacitance, CH, the overall C is determined here by the smaller
Csc (cf. Eq. 9.1):
(9.4)
with CH being taken as about 20 J.lF cm-2 .
At the edge exposure (Fig. 9.8), the behavior is like that of a normal metaVsolution interface where the capacitance is determined by the series combination of an ionic/electronic Helmholtz compact-layer capacitance (CH ) with an
ionic diffuse-layer capacitance, Cdiff' Hence the much larger overall capacitance
values. Note that the edge regions may contain surface oxygen functionalities 3.7
(Table 9.1) that may enhance ionic adsorption.
The behavior ofboronated graphite (Fig. 9.7) is again different. This is due
to both p-doping and probably also to introduction of a new surface-state distribution that changes the potential distribution at the graphite/solution interfaces
across the Helmholtz layer.
One conclusion from this work, that is relevant to fabrication of practical
double-layer capacitor devices is that use of microcrystalline (graphitic) carbon
powder preparations may give lower available capacitances per gram or per cu1'Or----.------.-------r------.------~
1M NoOH
(\/
IE
90
()
l.L.
::l
.......
80
O.05M H2SO4
70
FIGURE 9.8. Capacitance vs. potential curves (NHE reference) for edge-oriented pyrolytic
graphite as per conditions for Figs. 9.6 and 9.7. Curves for 1.0 mol dm- 3 NaOH (B) and 0.05 mol
dm- 3 H2S04 (A). (From Yeager, Moller, and Gupta. 6 Reproduced by permission of The Electrochemical Society, Inc.)
203
bie centimeter than other types of carbon, depending on the number of basal
plane exposures relative to that of edges. Also to be noted is that high-temperature pretreatments that enhance graphitic structures will have the effect of increasing the amount of semiconductor behavior of the thermally conditioned
carbon material.
Electrochemical double-layer capacitors employing high-area carbon materials 37 usually require carbon powders, fibers, felts, or so-called "aerogels"
that have been heat treated at elevated temperatures in N2 , 02> or water vapor.
This has the effect of modifying surface functionalities, opening up or changing
pore structures, distilling out impurities, or (at very high temperatures, >2300
0c) increasing the degree of micro-graphitization, usually with some diminution of specific surface area or change of porosity. Chu and Kinoshita38 have
summarized the effects of pretreatments (Table 9.3). Modification of surface
area and change of bound oxygen content upon heat treatment of a Vulcan carbon XC-72 at various temperatures for 2 h in an inert atmosphere is shown in
Fig. 9.9 as a function of temperature over the range 0 to 2700 0c. 38 Beyond ca.
1200 DC, surface oxygen functionalities are mainly removed (as CO or CO 2)
when an inert atmosphere or vacuum is used. The presence of oxygen-containing functionalities is undesirable because of their involvement in cycle-life stability or in self-discharge processes (see Chapter 18) that probably take place
through Faradaic redox reactions in the matrix, although the details of such effects are sparse.
TABLE 9.3. Modification of Carbonaceous Materials
Method
Liquid-phase oxidation
(e.g., oxidizing acids)
Gas-phase oxidation
(e.g., O 2, H20)
Plasma treatment
(e.g., atomic oxygen)
Result
Increase in surface area and porosity, lower density, increase
in concentration of surface functional groups
Increase in surface area and porosity, lower density, increase
in concentration of surface functional groups
Increase in surface area and porosity, improvement in
wettability, increase in concentration of surface functional
groups, decrease in electrical resistance, increase in
electrocatalytic activity for oxygen reduction
Decrease in surface area and porosity, higher density, more
graphitic surface structure, decrease in concentration of
surface functional groups
204
Chapter 9
4oor----.----.-----r----.----~--~2.0
1.5
l(1l
N
-0~
.......
.......
1.0
c:
CI.l
c:
<t
c:
CI.l
CI.l
....
(1l
>-
;:)
)(
(/)
100
0~--~----1~0~0~0~~~--2-0~0-0----~--3-0~006
Temperature /
FIGURE 9.9. Surface area modification and bound oxygen change due to heat treatment of a Vulcan-type XC 72 carbon. (From Chu and Kinoshita. 38 Reproduced by permission of The Electrochemical Society. Inc.)
The type of heat treatment and the nature of the precursor of the carbon material are major factors in choice and optimization of high-area carbons for double-layer capacitors, as they are also for Li-ion host materials for Li battery
anodes. For example, activated carbon fibers (ACFs), having ca. 1600 m 2 g-l,
can be prepared for capacitor devices by carbonization of phenolic resins followed by activation in steam at 1000 dc. The relation of types of carbon materials to their precursors is illustrated in Fig. 9.10 (Chu and Kinoshita38 ).
The graphitization effect of high-temperature treatments is usually greatest
with "soft" carbons, which show the largest changes of microstructure. Hard
Precursor
Carbon Material
for Double-Layer Capacitors
Hydrocarbon gas
Carbon black
Polymeric material
Carbon fiber
Petroleum
205
Microbeads
Coal
Carbon aerogel
Renewable fuel
Carbon powder
FIGURE 9.10. Relation of types of carbons to their precursors. (From Chu and Kinoshita. 38 Reproduced by permission of The Electrochemical Society, Inc.)
carbons derived from carbonization of cellulose or coconut shells do not graphitize so easily, even at temperatures as high as 2800 0c.
Attainable specific double-layer capacitances in f.lF cm- 2 for various carbon materials in different electrolytes are listed in Table 9.4 and depend quite
substantially on the pretreatment history and overall morphological characteristics. For example, it seems uncertain at present whether high-area fibrous
materials or powder-type matrices are preferable for optimum capacitor performance. The latter is determined by a variety of factors such as average pore
size, pore-size distribution, accessibility of pore structures to electrolyte, wettability (related to residual oxygen functionalities), and distributed internal ohmic
resistance, which determines the microequivalent circuit behavior related to the
ac or pulse response impedance.
Chu and Kinoshita investigated various other pretreatments of a fibrous
carbon material (Spectracorp A-lO and A_20).38 Cyclic voltammetry showed
the changes that can be induced by heat treatments in air or oxygen, as exemplified by the current responses at 5 mV S-1 shown in Fig. 9.11 for activated
carbon fibers. Liquid-phase pretreatments in NH40H; a superacid; in H 20 2 ; and
with l,4-naphthoquinone gave various degrees of improvement in the specific
capacitance, F g-l, as shown in Fig. 9.12.
The comparative effects of surface oxidation and etching of three fiber
preparations (structures illustrated schematically)-rayon, polyacrylonitrile
(PAN), and a mesophase-are shown in Fig. 9.13; it is seen that quite substan-
206
Chapter 9
Carbonaceous material
Activated carbon
Carbon black
Electrolyte
10% NaCI
Double-layer
capacitance
(,uF cm- 2)" Remarks
19
1 M H 2SO4
31 wt.% KOH
10
edge plane
Reference
Surface area 80
to 230 m 2 g- I
-I
6
7
Graphite powder
10% NaCI
35
Surface area 4 m 2 g- I
Graphite cloth
0.168 N NaCI
10.7
Glassy carbon
0.9 NNaF
-13
Solid
Carbon aerogel
4MKOH
23
"Values based on estimates of real areas but subject to uncertainties discussed in the text (Section 9.3).
References
1. I. Tanahashi, A. Yoshida, and A. Nishino, 1. Electrochem. Soc., 137,3052 (1990).
2. A. Soffer and M. Folman, 1. Electroanal. Chem., 38
3.1. Randin and E. Yeager, 1. Electroanal. Chem., 36, 257 (1972).
4.1. Randin and E. Yeager, 1. Electrochem. Soc., 118, 711 (1971).
5. S. Evans, 1. Electrochem. Soc., 113, 165 (1966).
6. K. Kinoshita and J. Belt, Carbon, 11,403 (1973).
7. E. Gagnon, 1. Electrochem. Soc., 122
8. K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, pp. 294-295, Wiley, New York
(1988).
tial increases in specific capacitance can be obtained using the indicated pretreatments.
Sullivan et a1. 39 described another example of an advantageous treatment
of a carbon material for use in an electrochemical capacitor. By means of chemicalor electrochemical oxidation of glassy carbon, these authors showed that a
double-layer capacitor material results that is particularly suitable for high
power-density applications due to its low internal resistance. An active surface
region is generated that is porous and permeable to electrolyte. A high specific
capacitance of ca. 250 F cm-3 over a I-V potential step charging and discharging
program was obtained. 39
207
FIGURE 9.11. Effect of heat and gas-phase treatment on carbon fibers as revealed by cyclic voltammetry at 5 mV s-I. (From Chu and Kinoshita. 38 Reproduced by permission of The Electrochemcial Society, Inc.)
208
Chapter 9
500r-------------------------------------~
1,4-NophothoQu i none
400
Superocid
0> 300
LL
.......
Initiol
Q)
(.)
c:
200
( .)
0-
100
209
__
0.6
10'l
Initial
LL
-.........
>- 0.4
' (3
o
0..
o
0.2
Rayon
PAN
Mesophase
FIGURE 9.13. Comparative effects of surface etching and oxidation on specific capacitance of
three fibrous carbon materials. (From Chu and Kinoshita. 38 Reproduced by permission of The Electrochemical Society, Inc.)
210
Chapter 9
time-dependent degradation of high-area carbons and/or be involved in selfdischarge processes (Chapter 18) following charging of electrochemical double-layer-type capacitors.
Singer40 reported electron spin resonance studies on carbons at that symposium and much prior and subsequent literature exists on this subject (see the
bibliography in Ref. 40). Leading early work was that by Ingram and Bennett,41
Ueberfeld, Etienne, and Combrison,42 and Winslow, Baker, and Yager,43 as
well as that by Castle44 and Collins, Bell, and Kraus45 somewhat later. All these
authors found ESR signals in carbonized materials, and at graphites and carbon
blacks. Interesting effects of molecular O 2 on the ESR behavior (due to chemisorption because the O2 molecule itself has unpaired electrons) were detected
that are relevant to the effects of O2 on capacitor carbon materials and high-temperature pretreatment effects. The number and complexity of the ESR effects indicate the useful sensitivity of ESR spectroscopy to the microscopic molecular
and electronic structure of carbon materials.
Most high-area carbon materials have appreciable surface and inner-surface densities of free radicals depending on their pretreatment. This free-radical
character has been examined and quantitatively evaluated by means of electron
spin resonance spectroscopy. Since electrochemical engineers and technologists may not be familiar with the technique of electron spin resonance spectroscopy and the significance of the results obtained from it for carbon materials, a
brief outline of its principles is given below.
Electron spin resonance spectroscopy, sometimes referred to as electron
paramagnetic resonance (EPR) spectroscopy, is based on changes in the electron spin energy states of molecules having an odd number of valence shell electrons in an applied magnetic field. An electron has a magnetic moment, Ite'
associated with its spin state, Ite = -gfJS where hS/2n is the spin angular momentum vector of the electron and ms is the spin (S) quantum number, which can take
values of 112 and lead to different energy states in a magnetic field. g is a dimensionless constant called the "electron g-factor" and fJ is the electronic Bohr
magneton, ehl2nm ec. me is the mass of the electron and c the velocity of light.
The energy state of the electron in the magnetic field, H, is Jlfi and is dependent on the direction of orientation of the magnetic dipole in the field H.
In the absence of a magnetic field, the energy of an electron with spin quantum number ms = + 112 is indistinguishable from that for ms = -112 (so-called degeneracy). However, in the presence of a field H, the degeneracy is lifted and
differing energies of the plus and minus spin states arise, leading to a difference
of energy levels by hv =gfJH where hv is the quantum of energy corresponding
to a characteristic frequency v (depending on H), and is equal to gfJH. Transitions from thermally excited spin states (ms = + 112) to the lower energy spin
state (ms = -112) can occur and the corresponding frequency v =gfJHIh can be
evaluated.
211
Chapter 9
212
Parallel to layers
Perpendicular to layers
50
0.06
1.3 X 10-6
27 X 10-6
1.06 X lO"
3.6 X lO"
Source: Reprinted from P. A. Thrower. 1. Electoanal Chern., 36,401 (1972), with permission from Elsevier
Science.
and surface species was clearly demonstrated in the ESR results of Harbour and
Walzak. 46
A process that can lead to generation of free-radical centers is the formation of peroxides by reaction of unsaturated organic structures with molecular
oxygen. This can occur at the (edge) double bonds of graphitic structures. The
subsequent dissociation of bridged or linear -O-OH groups can lead to -0' or
-0-0' free radicals on the surface of carbon materials. In the presence of dissolved 02> on cathode electrode charging in an electrochemical capacitor,
HO or H 20 2 can be formed. In fact, carbon is a well-known electrode surface
for electrochemical generation ofH20 2. H 20 2 can then give rise to free radicals
by well-known surface or bulk reactions, especially in the presence of traces of
Fe species (Haber-Weiss reaction), and is one of the species involved in selfdischarge processes (Chapter 18) at charged capacitors.
A paper by Thrower47 is relevant to the variety ofESR signals arising from
carbon materials. He discusses the great multiplicity of carbons, their physical
properties (Table 9.5) (depending on their source), and preparation and pretreatment procedures. The effects of heating at elevated temperature are discussed,
including X-ray diffraction evidence for modification of crystallite size and
mean spacing between the carbon layer planes. Procedures for preparation and
orientation of carbon fibers from rayon, polyacrylonitrile, and pitch are also
considered in this paper, including characterization methods based on X-ray diffraction, scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Such carbon-fiber materials are of major technological
interest for fabrication of electrochemical capacitors having good mechanical
integrity and low equivalent series resistance values (see Chapter 17).
213
Related to chemical and ESR investigations on the nature of carbon surfaces is the determination of their electronic work functions, rjJ. rjJ measures the
energy required to excite an electron into a vacuum, limitingly with zero kinetic
energy, from the Fermi level of the electron energy distribution in the bulk material (the level having a probability of half-occupancy at finite temperature and
214
Chapter 9
150
-e-e- OXYGEN
AS C02
C02' CO. H20
IE 130
u
e-m
.....
Z 110
iii
a::
~ 90
:2
o
o
I
40
60
80
100
150
200
thus the energy state to or from which, on average, electrons are withdrawn or
donated in cathodic or anodic reactions). However, information on the suiface
state of conducting materials is given through lP measurements 50 since lP always contains a surface dipole contribution, X, often amounting to 20-30% of
lP. This contribution occurs because the experimentally determined lP value always includes the electrical work of transfer of an electron out from the metal
surface across the metal interface at which a surface potential contribution, ilX,
arises due to any molecular surface dipole structures. Thus changes in X caused
by the presence of such dipoles can be determined and information on the electrical aspects of surface functionalities derived .
where n is the number of dipoles per square centimeters of surface having a surface dipole moment component, fl, normal to the surface. As n approaches its
215
o
>Q)
---"
'0
C'
-eI
c
0
.0
0
...
"S-
...........
0.1
O~
_ __ L_ _
4
~~
__
__
10
pH
FIGURE 9.15. Relation of work-function difference between various carbons and a gold plate,
and pH of slurries of the respective carbons. Reprinted from T. J. Fabish and D. E. Schliefer, Carbon, 22, 19 (1984), with permission from Elsevier Science.
216
Chapter 9
300~--~----.----.-----r----~--~----~
o
250
C\I
IE 200
u
01
\-
Q)
.......
c
'in
\0
Q)
E
E
150
'0
.....
0
Q)
100
50~--~----~----~--~----~--~~--~
pH
FIGURE 9.16. Dependence of heats of immersion of a high-area carbon black on pH in an aqueous slurry. Reprinted from T. Murata and Y. Matsuda, Electrochim. Acta, 27, 795 (1982), with permission from Elsevier Science.
ference between the carbons and a gold reference plate is plotted against the
slurry pH. 52 Corresponding data were obtained for seven water-extracted carbon blacks subjected to various surface treatments.
The pH relations show characteristic minima at an isoelectric pH around
6.3. The carbon black work function values are assumed to be directly related
to base uptake33 and are therefore determined by acidic polar functional groups
217
A factor that has to be taken into account when using graphitic materials
for capacitor electrodes is the possibility of intercalation of species dissolved in
the electrolyte, especially anions at the anode of a capacitor electrode (on
charge) or possibly also of some cations at the negative electrode. Intercalation
into the layer structure of graphite has been known for many years,32 but more
recently its practical implications have been recognized.
Beck and Krohn 53 studied the reversible intercalation and deintercalation
of anions at highly crystalline graphite electrodes. Interfacial charge transfer
processes coupled with diffusion can be involved. Riidorff and Hofmann 54 were
the first to perform an electrochemically driven intercalative sorption of strong
sulfuric acid (cf. Ref. 19) into graphite. The overall reaction was written 54 as
The intercalation of cations was regarded by Beck and Krohn 53 as involving first a step of electrochemical oxidation of the graphite lattice to form a
macroradical cation. This process is analogous to oxidation of a macrocyclic unsaturated hydrocarbon such as perylene, which first forms a radical cation. The
second, probably coupled, step involves transfer of an anion through the carbon
electrolyte phase boundary:
(9.7)
where n solv. represents the number of solvating solvent (water) molecules associated with the anion A-in solution. The Gibbs-energy driving force for this
process will be determined by the difference of accommodation energy of the
guest anion in the C~ host and the (change in) energy of solvation of the anion.
In addition, there will be an electrical Gibbs energy component that is determined by the electrode potential of C x relative to that of the solution in which
A - initially resides. The energy of A-in the lattice will also depend on the extent, if any, to which the intercalated anion undergoes charge transfer with the
lattice instead of forming the ion pair, C~A-.
Examples (positive and negative) of reversible electrochemical intercalation of anions (which can be significant at positive capacitor electrodes of graphitic powder) into a composite of natural graphite crystal flakes with 20%
polypropylene (CPP) are listed in Table 9.6. The examples cited include acids
commonly employed as electrolytes for aqueous solution electrochemical ca-
218
Chapter 9
Aqueous acids
Positive examples
Sulfuric acid
Perchloric acid
Tetrafluorboric acid
Hydrofluoric acid
Trifluoroacetic acid
Methanesulfonic acid
Hexafluorophosphoric acid
Negative examples
Nitric acid
Phosphoric acid
Arsenic acid
Amidosulfonic acid
Hexafluorsilicic acid
Telluric acid
Chloroacetic acid
Dichloroacetic acid
Trichloracetic acid
Benzenesulfonic acid
p- Toluenesulfonic acid
H2F2
CF3COOH
CH 3S0 3H
HPF6
0.1-18
0.1-12
0.1-8
1-50
1.5-10
2-8
2-8
HN0 3
H3P0 4
H3AS04
NH T S0 3H
H 2SiF6
H 2Te04
CICH 2COOH
Cl 2CHCOOH
Cl 3CCOOH
C 6H 5 S0 3H
CH r C6 H4-S0 3H
4-7
4-48
80%
0.75-1.5
1-2.8
1
1-5
90% (slight deintercal.)
1.3-9
1-4
6
H 2SO4
HCI04
HBF4
Source: From Beck and Krohn. 53 Reproduced by permission of The Electrochemical Society, Inc.
pacitors. The intercalation processes were studied by means of slow (mV s-l)
cyclic voltammetry in the cases of 4 M concentrations of H 2S04 , HCI04 and
HBF4 solutions.
Ebert55 reviewed the general electrochemistry of intercalation compounds
of graphite. The general basis of lattice oxidation, forming a delocalized positive charge on the benzenoid framework of the larger crystal structure with intercalation of anions as counter charges, was emphasized. The sorbed anions
were regarded as being screened by co-sorbed solvent molecules, i.e., the anions
retained some degree of solvation. Thus the mechanism is a phase-transfer of the
solvated anions coupled with electron transfer from the lattice (anodic intercalation). The case of sorption of cations (Li+) into carbon host materials now constitutes the major field of lithium ion anodes in the lithium ion battery field.
The whole question of the electronic and lattice-mode structure of graphite
is related to the topic of intercalation. This subject was considered in a review
by Dresselhaus,56 including the 2-dimensional band structure of graphite. 3,22
219
REFERENCES
\.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
220
Chapter 9
Chapter 10
222
Chapter 10
modynamic reasons between the extent of charge acceptance (IJ.q) and the
change of potential (IJ.V), so that a derivative d(IJ.q)/d(IJ.V) or dq/dV, which is
equivalent to a capacitance, can be formulated and experimentally measured by
dc, ac, or transient techniques (Chapters 16 and 17).
The capacitance exhibited by such systems is referred to as pseudo-capacitance since it originates in a quite different way from that corresponding to classical electrostatic capacitance of the type exhibited (mainly) by double-layer
capacitors. The term "pseudocapacity" was also used by Grahame in his 1947
review (Chapters 6 and 7), but to refer to the reversible discharge of ions (e.g.,
ofNa+) at high negative potentials at Hg. This usage is, however, different from
that employed here where extents of Faradaically delivered charge are a function of potential but no continuous currents pass in time.
It is now known that double-layer carbon capacitors exhibit perhaps 1-5%
of their capacitance as pseudocapacitance due to the Faradaic reactivity of surface (edge) oxygen-functionalities (depending on the conditions of preparation
or pretreatment of the carbon material). On the other hand, pseudocapacitor devices, like batteries, always exhibit some electrostatic double-layer capacitance
component proportional to their electrochemically accessible interfacial areas-probably about 5 to 10%.
At this point, it is useful to compare the potentialities for charge storage in
a double-layer capacitor with that in a pseudocapacitor device and a battery
electrode as shown in Table 10.1. From a thermodynamic point of view, pseudocapacitance originates whenever some property, y, proportional to charge
passed, is related to potential by an equation of the form
y/(l- y) = K exp(VF/RT)
(10.1)
Battery; 1 to 3 electrons per atom or molecule of bulk phase, depending on oxidation states of
reagents and products.
223
cies (or vice versa) in a redox system in solution [e.g., aq. Fe(CN):-lFe(CN)~-]
or one in a hydrous oxide, e.g., Ru02, Ir02' C030 4 (Chapter 11).
Each of the above types of examples corresponds in a formal general way
to an electron transfer process between an oxidized species Ox (e.g., H+ or H 20,
Li+, metal ion redox reagent) and a reduced species, Red (e.g., adsorbed H, Li+
in a host cathode lattice, metal ion in the reduced state in the redox reagent); thus
Ox+ze~Red
d IdV='-.
Kexp(VFIRT)
y
RT [1 + K exp(VFIRT)]2
(10.2)
224
Chapter 10
amined under ac modulation (Chapters 16 and 17); it varies with the square root
of frequency. This pseudocapacitance has no value for energy storage purposes
and is significant only in the frequency response of electrochemical reactions involving diffusion (Randle's equivalent circuit treatment, Chapter 7).
In Section 10.1, the electrochemical and physical significance of pseudocapacitance and its difference from double-layer capacitance were outlined.
Here we present the mathematical theory of pseudocapacitance based first on
the example of the potential (V)-dependence of coverage, 0, of an electrode surface by a Faradaically deposited or desorbed electroactive species such as H.
Formally similar equations arise for an electrochemical redox reaction (see Section 10.5). The pseudocapacitance (eq,) arises as a quantity proportional to
dOldV in the case of electrosorption processes. 1-3
Two kinds of treatment of pseudocapacitance have been given that are
complementary. One is of an equilibrium thermodynamic kind, given in the papers of Eucken and Weblus 4 and of Conway and Gileadi;l the other is based on
a kinetic approach for processes that exhibit pseudocapacitance but under conditions displaced from equilibrium as treated in papers by Gileadi and Conway,5
Gileadi and Srinivasan,6 and Kozlowska, Klinger, and Conway,? the latter for
two-step processes involving a chemisorbed species. These latter nonequilibrium treatments include the equilibrium condition as a special case and lead
also to a representation of the cyclic voltammetry behavior of experimentally
observable, electrochemical 2-dimensional surface processes.
The treatments are based on writing kinetic equations for 2-dimensional
surface processes, taking account of (1) fractional surface coverage 0; (2) fraction of surface sites unoccupied, 1-(), (3) the exponential effects of potential on
the rates of the charge transfer processes that are involved in the formation or
removal of the 2-dimensional film of electrochemisorbed species, and (4) lateral
interaction effects (g factor, not to be confused with the esr g factor, p. 210).
10.2.2. Electrosorption Isotherm Treatment of Pseudocapacitance: A
Thermodynamic Approach
1-8
MHads+H20
8
(I)
225
(10.3)
taking the argument of the exponent as negative together with negative V for the
cathodic reaction I. This type of process is well known at the catalytic noble
metals such as Pt, Rh, Ru, Ir, and also for some base metal adatom deposition
processes at Au or Ag. For H, it is observed at potentials positive to the reversible hydrogen electrode potential, i.e., before cathodic H2 evolution takes place.
For this reason, the process is referred to as "underpotential deposition" to distinguish it from the related process of deposition of H during H2 formation at
cathodic overpotentials when continuous net currents for H2 formation pass.
Then process I is formally also the first step of the H2 evolution process, but
takes place under different conditions of H coverage and electrode potential
from those for UPO, in particular on a surface fully covered by UPO H (fJ 2= 1)
in the case of polycrystalline Pt, Rh, Ir, and Ru. At Pd and Ru, H also passes into
the metal lattice by a sorption process after discharge of the H+ ion or from
water.
Processes like that for UPO of H on Pt also occur with electrodeposition
of base metals (e.g., Cu, Bi, Pb) on noble metals such as Au, Ag, or Pt. Thus
monolayers of Cu, Bi, or Pb can be electrosorbed from the corresponding cations in solution at potentials positive to the respective reversible potentials for
deposition of the regular crystalline bulk metals.
Thermodynamically, the condition required for UPD of 2-dimensional
monolayers prior to bulk phase formation is that the Gibbs energy of the H-to-M
or base metal-to-M bonds be lower than that for H-H or for the intermetallic
bonds in the bulk phase of the deposited base metal. Thus, e.g.
(10.4)
or similarly for base metal to substrate M bonding in the UPO of the base metal
on substrate M in relation to G for bond strength in the lattice of the bulk base
metal. The above relation for the Gs corresponds to the actual numerical value
of the bond energy of H to M being greater than that for H to H in 112 H 2, or
the metal atom substrate bond energy being numerically greater than the metalto-metal interatomic bond energy in the bulk metal crystal, taking into account
the coordination number of the metal crystal structure. It is in systems that behave thermodynamically in the above way that adsorption pseudocapacitance
can arise if the kinetics of the processes are sufficiently facile, which is usually
the case.
226
Chapter 10
If the charge for formation ofa monolayer ofH on M in process I is ql, then
a pseudocapacitance C~ can be derived from Eq. (10.3) as
(10.5)
qlF
= RT
KCH+ exp(VFIRT)
[1
+ KCH+exp(VFIRT)]2
(10.6)
0=
KCH+ exp(VFIRT)
(10.7)
1 + KCH+ exp(VFIRT)
For H on polycrystalline Pt, ql = 210 pC cm-2 ; at the principal index faces
of single-crystal Pt, somewhat different values of ql apply. The UPD of Pb on
Au involves a 2e discharge per atom and the resulting ql is about 380 pC cm- 2,
allowing for the relative atomic diameters of Pb and Au.
An alternative and useful expression for C~ can also be obtained l in terms
of 0 and I - 0, using Eq. (10.7).* Thus
(l0.8)
227
same "configurational" Gibbs energy term RTln 81( 1 - 8). Then the electrosorptlon isotherm becomes of the Frumkin type, namely,
--8 =Kexp(-g8)
1-
(10.9)
. cH+exp(VFIR7)
Here K exp (-g8) can be regarded as a coverage-dependent adsorption equilibrium constant differing from the Langmuir one, K, by the exponential factor in
the argument, -g8. Thus K for 8 = 0 (Langmuir case) is diminished with increasing 8 for positive g values corresponding to pairwise adsorbate repulsions in the
adlayer. For further discussion, see Hill,9 Thomas,lO and Rangarajan,l1 where
implicit assumptions in relations having the form ofEq. (10.9) are examined, as
well as problems with underpotential deposition of adatoms where questions of
the significance of partial charge transfer in the electrosorption valence 12,13
arise.
Evaluation of the Crp corresponding to Eq. (10.9) by differentiating Eq.
(10.9) in the form 8 exp(g8)/(1 - 8) w.r.t. V leads to the relation
8(1 - 8)
qlF
Crp - RT . 1 + gee I -
(10.10)
e)
228
Chapter 10
..--...
<pI>
(g>O)
"0
"0
r::T
<l>
'--'"
"
"9-
<t
a::
u
z
<t
~0.5
<t
!l.
<t
u
O~~------~~--L---
________
-=~
______
ELECTRODE POTENTIAL
FIGURE 10.1. Coverage, 8, isotherms and Ctp vs. electrode potential profiles for an electrosorption process involving a zero and a positive g value, i.e., for lateral repulsion (Eq. 10.10) (schematic,
cf. Refs. 5 and 6).
Cp (= i/s)
(farad cm -2)
~VII2
8p
dV/dlog slk J
(V)
2.14 x 10 3
0.091
0.50
1.57 x 10-3
0.126
0.63
(V)
0.118
(P = 0.5)
229
Reversible
-'-'-'- Irreversible
0.8
0.6
>
..........
0.4
+-
Q)
+-
Langmuir
case
I
t
0... 0.2
.....
-4
,.
/.---
~.
'
12
16
20
9
FIGURE 10.2. Widths (AV1I2) at half-height of C,p vs. V profiles as a function of the lateral interaction parameter, g. (From Conway and Gileadi 1 and Kozlowska et al? Copyright 1962 and 1977
Royal Society of Chemistry.)
therm equation will contain two exponential terms, exp -[fOIRT] and
exp[VFIRT] (i.e., in combination, exp[(VF - jfJ)IRTJ. This illustrates how VF
must be increased as/is increased positively in order to maintain a given value
of (VF - /O)IRT, which then determines the value of the configurational term9
01(1 - 0) of the isotherm function (Eq. 10.9).
If the g value is negative, up to g = -4, the opposite conclusion applies; the
electrosorption is easier with increasing coverage due to attractive interactions
in the adlayer, so that a relatively lower potential V is required to attain a given
coverage, and the effective potential range required to reach almost full coverage (0 --7 1) is then diminished. Because the required charge (ql) is still the same
for monolayer formation, the resulting linear-sweep voltammogram must be
sharper than for the Langmuir case, g or/ =0 (see Fig. 10.3). Note that the integral under the voltammogram at a sweep rate s = dVldt gives the charge
passed:
V2
V2
VI
VI
(10.11)
230
Chapter 10
0 .0
1.0
0.8
9 -8
0.6
0.4
0 .2
0.0
-0.2
0.0
0.2
OA
0..6
Potential/Volts
FIGURE 10.3. (a) Ctp vs. electrode potential profiles for an electrosorption process involving
negative g values, i.e., lateral attractions. (b) Related isotherms for coverage, B, as a function of potential for positive and negative g values.
231
for a sweep between potential limits V2 and VI. Then since i(V)/s is a pseudocapacitance CJ' the integral is equivalent to charging the CJ from potential VI
to V2. The charge passed is then CJ (V2 - VI) or f~2 CJ . dV since CJ is rarely constant with changing potential, especially for elec'trosorption processes. I - 7
Experimentally, the pseudocapacitance behavior represented by the above
equations for CJ can be directly recorded by means of cyclic voltammetry since
for a sweep rate, dV/dt,
CJ dVldt= i
(10.12)
or
CJ =iI(dVldt)
(10.13)
where i is the recorded response current density for a system exhibiting pseudocapacitance, addressed by a potential that varies linearly in time t. Thus linearsweep or cyclic voltammetry is the most convenient technique for accurately
recording CJ behavior, including situations commonly encountered where CJ is
far from constant and the CV may exhibit more than one current maximum in
the swept potential range. This situation is common on polycrystalline surfaces
where several maxima in CJ vs. V profiles are observable in UPD studies (Fig.
lOA); it is interesting that even for well-prepared single-crystal surfaces [e.g.,
ofPt(lOO)], more than one peak in the CJ profile can arise as for H [Fig. 10.4(a)].
Similarly, multiple peaks arise in the UPD ofPb adatoms on single-crystal surfaces of Au [Fig. lO.4(b)].
In evaluation of pseudocapacitance by means of cyclic voltammetry, it
must be remembered that there will always arise components of response currents owing to the ubiquitous presence of the double-layer capacitance. This
will lead to a double-layer charging current, Cdl (dVldt), that is proportional to
the sweep rate; this current is normally 5 to 10% of the (maximum) pseudocapacitance charging currents when CJ is appreciable. The double-layer charging
component can usually be distinguished from the pseudocapacitance component by its different frequency dependence; thus, the double-layer capacitance
response is usually maintained at up to 105_106 Hz in a well-designed measurement system while pseudocapacitances are often dispersed in the range 103-104
Hz due to electrode kinetic rate limitations.
Writing the isotherm function, Eq. (10.9), logarithmically, i.e.
In 0/(1- 0) + gO =In KCH+ + VFIRT
(10.14)
232
Chapter 10
....
I
E
u
ci
:1.
::::::
100
a
FIGURE 10.4. (a) Cyclic voltammetry profiles for UPO of H adatoms on clean Pt single-crystal
surfaces: I, (111); 2, (tOO); and 3, (110) surface. (b) Cyclic voltammetry profiles for UPO of Pb
adatorns at indicated single-crystal surfaces of Au. [(a) From Clavilier et aI., 1. Electroanal. Chern.,
295, 333 (1990); (b) from Engelsman et al. 18]
it is seen that differentiation W.r.t. V will give rise l to two reciprocal capacitance
components 2 :
RT
- . d 1n[8/(1- 8)]
qlF
(lO.15a)
and
RT
qlF' g
(1O.15b)
cal reasons lO different from those envisaged by Temkin,14 namely, surface site
heterogeneity. Since the terms ofEqs. (1O.15a) and (1 0.15b) arise from addition
in Eq. (10.14), they can be regarded as corresponding to two capacitance contributions in series; that is, a configurational9 capacitance contribution from the
Oil - 0 term ofthe isotherm and an interaction contribution from the exp(-gO)
term ofEq. (10.9). Obviously, as g ~ 0, the adsorption behavior reverts to that
represented by a Langmuir isotherm.
For intermediate values of 0 (e.g., 0.3 < 0 < 0.7) and appreciable values of
g (e.g. ~ 10), 0 becomes approximately linear in V, a result also corresponding
formally to Temkin's isotherm. This corresponds to a potential range over
which C is approximately constant with changing potential, a condition that is
desirable if pseudocapacitative electrochemical capacitor devices are to be practically useful over a reasonable charging voltage range.
Later in this chapter it is shown that a Nernst-type redox equation has the
same form as the electrochemical Langmuir isotherm equation. 2 Hence, upon
234
Chapter 10
+ 2e ~ 2S 2-
Cathod ic
1.389
..
(10.16)
10
(bl
.....
............. ,
<.>
'~
<;
u
.""
.........
(0)
..'....,
'
/ 7',
.... _-/" ... . . .,
" ....""",
./
"
1.041
",-
"
... ,
,/
0 ,694
0 .347
N
IE
<{
0,347
,,j
0,694
'j
1.041
<t
1.389
Anodic
-0.1
0 .5
Potential (Volt, E H )
FIGURE 10.5. Almost-reversible cyclic voltammogram for reduction and reoxidation of the
surface region of FeS2 (pyrite). (From Conway and Ku. 15 )
235
The process takes place in the surface region ofFeS2, producing two sulfide ions
that remain in the interface of the lattice. In order to maintain charge balance
with respect to Fe2+ ions, presumably S2- becomes protonated to SH- in the
course of the reduction.
Using cyclic voltammetry, an almost reversible mirror-image cathodic and
anodic cyclic voltammogram is generated, as shown in Fig. 10.5. The system
can be cycled thousands of times with little degradation. For comparison, FeS
(the salt of S2- ions) behaves irreversibly on reduction, as is seen in the cyclic
voltammogram illustrated in Fig. 10.6. This proves indirectly that it is the disulfide-sulfide redox reaction that gives rise to the pseudocapacitance exhibited by
4 .40
Cathodic
3 .52
..
u
'0
.r::. 2 .64
'0
1.76
0 .88
N
IE
u
0 .00
<l:
...
0 .88
1.76
.':::' 2 .64
'0
3 .52
FIGURE 10.6.
10.5.)
Anodic
Irreversible cyclic voltammogram for FeS. (From Conway and Ku 15; contrast Fig.
236
Chapter 10
FeS2' The reversibility arises presumably because there are no phase changes involved and the process is restricted to the surface region. The disulfide system,
FeS2> behaves in a way typical of other disulfides, e.g., the organic molecule
cystine, which is reversibly reducible to cysteine.
e,
237
500~------------------------------------------'
400
N
IE
u 300
u..
:t.
":>-
200
5'00
FIGURE 10.7. Kinetically calculated pseudocapacitance profiles for a one-electron surface process displaced progressively from equilibrium, depending on relative values, KNE, of the rate constants for electrochemical desorption. (From Gileadi and Conway.5 Reprinted with permission from
1. Chern. Phys., 31, 716. Copyright 1964 American Institute of Physics.)
depends on which is the principal desorption mechanism in the overall overcharge process, here H2 evolution.
A different type of calculation of the disturbance to equilibrium pseudocapacitance was made by Srinivasan and Gileadi 6 for the case of cyclic voltammetry behavior. Here the effect of increasing sweep rate, s, in relation to the rate
500
400
N
IE 300
u
u..
....:t.
:>- 200
r
U
<1
100
L)
0.1
0.2
0.3
0.4
0.5
0.6
0.7
POTENTIALIV
FIGURE 10.8. As in Fig. 10.7 but for a recombination controlled reaction. (From Gileadi and
Conway.5 Reprinted with permission from 1. Chern. Phys., 31, 716. Copyright 1964 American Institute of Physics.)
238
Chapter 10
constants kl and k-l for the forward and reverse directions of the electrosorption
reaction (e.g., process I) was examined in terms of electrochemical kinetic equations for each direction of the surface process. Analogous equations (but not involving surface coverages) could be written for a Nernstian redox process
similar to those relations for polargraphic behavior of a redox reaction displaced
from equilibrium.
The treatment by these authors 6 is based on introduction of a linearly timevariant potential into the rate equations for an electrosorption process, corresponding to the conditions of cyclic voltammetry at some constant sweep rate,
s = dV/dt. The discontinuous change of sign of s arises at the two limits of the
potential span applied to the electrode in cyclic voltammetry. There is a corresponding change of sign (direction) of the response current at those limits if capacitative behavior is exhibited by the system.
The kinetic treatment of pseudocapacitance given below closely follows
that developed by Srinivasen and Gileadi6 for a process such as reaction I, following the more thermodynamic treatment given by Conway and Gileadi 2 that
is based on adsorption isotherms for electrosorbed species outlined earlier in
this chapter. Using the example of reaction I involving nondiffusion-controlled
electrosorption of H, the Faradaic current density, i, is written in the form of a
Butler-Volmer equation (see Chapter 3 on electrode kinetics)
z~ = kJcH+(l- 0) exp(jJVFIRT) -
where kl and k-l are again the rate constants for the forward and reverse directions of process I for which z = 1, and Pis the barrier symmetry factor for the
charge transfer process (P =0.5). At equilibrium i =0, then a function for O(V)
can be written as
B
KlcH+ exp(VFIRT)
(10.18)
1 + K1cH+ exp(VFIRT)
which is recognized as the Langmuir type of electrosorption isotherm (Eq. 10.7)
earlier with K KI =ktfk_ l ) involving an exponentially dependent adsorption
equilibrium constant, KI exp(VF/RT). In Eq. (10.17) the sign (ve) of the argument of the exponent has been combined for convenience with the negative sign
of V for a cathodic process so that i increases with V in the usual way (as with
Eq. 10.3), with the opposite for the reverse direction of the reaction.
The net Faradaic current density is also qJdB/dt where ql has been defined
earlier as the charge for monolayer formation or desorption, ca. 210 flC cm- 2 .
From the relation for O(V), dB/dt is obtained6 as
dBldt=
K1cH+
. RFT exp[V(t)FIRT] s
K1cH+ + exp - [V(t)F IRI1
(10.19)
239
(10.21)
240
Chapter 10
=zFkJ(l -
(}) exp[,BVFlR7J
(10.25)
i.e., a Tafel-type equation, which after differentiation for linear sweep behavior,
gives
dUdt = zFkJ exp[pV(t)FfRT] . [(1 - B),BsFfRT - d{}fdt]
(10.26)
(10.27)
241
The latter relation can be recognized as a Tafel-type equation giving the displacement of Vp as a logarithmic function of sweep rate, s, with a Tafel slope of
RT/PF noted by Gileadi and Srinivasan 6 and treated in some detail by Kozlowska and Conway, 16 corresponding to a logarithmic function of i dependent
onpVF/RT.
The peak current is again found to be proportional to s, but with an apparent maximum pseudocapacitance for the nonreversible behavior of process I,
given by
(l0.30a)
(jJ = 0.5)
(l0.30b)
(10.31)
For the reversible case,ep = 0.5.
Note that for an electrochemical (UPD) adsorption or surface redox process, continuous Faradaic currents cannot arise as they can for regular Faradaic
processes. At low potentials, the response current in a potential sweep is small
because the potential is relatively small; as the potential is increased, the current
initially increases exponentially but the 1 - eterm decreases until it eventually
becomes zero, so the current is cut off. A maximum in the current therefore
arises and it is observed (in the reversible case) when =1 - =0.5. Analogous
effects arise in a surface redox reaction where the fractions of active materials
[Ox]/([Ox] + [Red]) and [Red]/([Ox] + [Red]) (analogous to e and 1 - e) are
changing throughout the potential sweep. For bulk solution redox reactions, the
limit of current density at a sufficient limit of the potential sweep becomes determined by the diffusion-limited current, as in polarography.
242
Chapter 10
(10.32)
and
k1cH+ exp[[JVFIRT]
B=--------~~~~----~-------
- ilzF[k1cH+ exp[[JVFIRT]
(10.33)
from which also dBldt and dildt for Vet) = V(t = 0) + st can be derived. These
differential equations cannot be solved analytically. but Srinivasan and Gileadi 6
gave a series of numerical solutions plotted as linear-sweep voltammograms in
terms of Ct/J asfiV) for various s values and the Vp values as anfilog s) for various
kl and k-l values (Figs. 10.9 and 10.10).
Kozlowska, Klinger, and Conway7 gave similar plots, but for generalized
kinetic parameters defined as reduced sweep rates slk 1 Then reversible conditions correspond to small slk! values and irreversibility to large slk!. Kozlowska
and Conway16 also demonstrated the transition to Tafel behavior in plots of Vp
vs. log s from quasi-equilibrium behavior for s < So (the limit of increasing
-3 -2
2.0
N
'E
u
(/)
"Cl
~
...... 1.0
o<"l
Q
u"
/.
,....
.'.-..
:-
. \.
i
.
\\. .
.\ .:
:'\
:'
\
\
243
0.4
0.3
>
"-
Q.
>
c
0.2
:.=
Q)
0.1
a..
0.0
electrochem'lcall
electrochemical
reaction
reaction
reversible
I
irreve sible
0.05L---------~--------~0~--------2~------~4
-4
-2
log ( :1)
FIGURE 10.10. Peak potential values for linear-sweep voltammograms for various slkl values
(based on Gileadi and Srinivasan6 and Kozlowska, Klinger, and Conway\
sweep rate up to which complete reversibility is maintained) to irreversible behavior when S > So or s So.
The latter authors also considered the effect of the introduction of interaction effects, g > 0 or g < 0, on the shapes of the voltammetry (C) curves (Fig.
10.11). They showed the behavior of cyclic voltammograms when sweep reversals were conducted from various potentials in the sweeps corresponding to incomplete coverages, e < 1, and the resulting shapes of the current response
curves for the opposite directions of sweep after successive sweep reversals
from those potentials. Examples are shown in Figs. 10.12 and 10.13.
244
Chapter 10
0.6
....
Ie
()
C/I
ec
u..
..,
g=10
1- =10
0.4
0.2
""C
0.0
0.4
g
M
Potential (Volts)
0.2
0.4
0.6
0.8
1.0
FIGURE 10.11. Effect of interaction effects on cyclic yoltammograms for a process displaced
from equilibrium. s > so. (From Conway and Klinger, unpublished.)
exceed the exchange current density for the reversible process, i.e., for conditions where either the forward or the reverse terms of Eq. (10.17) become dominant on charge or discharge. Such changes of conditions are exactly the same,
in principle, as those that arise in the polarization of regular, continuous
Faradaic reactions as the overvoltage is progressively increased from zero to appreciable values.
Thus, the behavior of surface processes involving underpotential deposition is closely analogous to that of regular Faradaic reactions except that the finite availability of surface sites (1 - ~ 0 as ~ 1) and a corresponding finite
value of ql restricts the possibility of continuously (and exponentially) increasing currents as the potential is raised since the processes take place 2-dimensionally on a finitely limited density of surface sites, about 1015 cm- 2
Note also that a surface (UPD) electrochemical reaction that is proceeding
under equilibrium or quasi-equilibrium conditions has a current that is determined at all potentials in the sweep, not by an overvoltage, but by a potential
defined by the equilibrium adsorption isotherm, e.g., Eq. (10.7). Thus the latter
is a form of the Nernst equation for relative extents of vacancy (1 - e) and oc-
t=)O-3
245
: =10- 2
1.6
0.8
Ie
-0.'
0.2
0.4
-0.1
./.
f\'
/0.2
0.4
0.8
(,)
f/)
1.6
~
........
2.4
"0
'0.
(b)
(0)
.1..-10
k -
t=10-1
1.6
0.8
L.z
-0.'
'Y
0.5
,0/I
0.8
1.6
jJ)
:r,(00.1 0.'
Vp.v'o
(c)
W
I
I
(d)
POTENTIALS/V
FIGURE 10.12. Anodic and cathodic CV profiles for various initial cathodic values and for
various degrees of reversibility of reaction (I). The values of slk are (a) 10-3, (b) 10-2, (c) 10- 1, and
(d) 10; g = 0 for all cases. Reprinted from H. A. Kozlowska, 1. Klinger, and B. E. Conway, 1. Electroanal. Chern., 75, 45 (1977), with permission from Elsevier Science.
246
Chapter 10
1.0
~ =10
g=2
0.5
IE
III
"C
...
I.L.
..........
r<l
0.5
1.0
FIGURE 10.13. As in Fig. 10.12 but with a finite g value. Reprinted from H. A. Kozlowska, J.
Klinger, and B. E. Conway, J. Electroanal. Chern., 75, 45 (1977), with permission from Elsevier
Science.
reaction) can take place according to the kinetics of the surface process or of the
redox reaction will a true overvoltage arise. This occurs when,16 in a linear potential sweep modulation process, the sweep rate s is greater than so, or the
Faradaic current density i is significantly greater than io. The relation between
current responses to linear potential sweep modulation and those to alternating
voltage modulation is discussed in Chapter 16.
From the theoretical treatments of adsorption pseudocapacitance, it is evident that appreciable values of C!p for a single-state adsorption process arise only
over a narrow potential range of about 120 mV when g::: O. Only for large values
of g is the range of appreciable C!p extended to 0.5 to 1.0 V, but the values are
247
then much lower than the maximum C~ when g = O. However, the range is still
usually substantially larger than the double-layer capacity of the same area of
electrode.
When multiple-state adsorption pseudocapacitance arises, as for UPD of
Pb on polycrystalline AU,IS appreciable C~ arises over about 0.7 V, but five
peaks in the C~ vs. V profile arise, several even on single-crystal surfaces of Au.
Only for the redox pseudocapacitance developed at Ru02 in aqueous H 2S04
(Chapter 11) is a wider potential range (ca. 1.4 V) realized, over which C~ is
within 10% constant. At Ir0 2, similar behavior is observed, but only over a
smaller potential range (0.9 V) and with greater dependence of C~ on potential.
The 1.4-V range attainable at Ru02 is near the maximum possible in aqueous
solutions owing to the limitation of the decomposition voltage of water. Simulation analysis of the redox pseudocapacitance by hydrous RU02 indicates that
the broad range of almost constant capacitance can only be accounted for by
several (three) oxide redox processes overlapping in their potential ranges for
charge accommodation, as illustrated in Fig. 11.16 in Chapter 11.
In the case of the electro active conducting polymers such as polyaniline,
polypyrrole, and poly thiophene (Chapter 12), the potential range of appreciable
C~ (depending on the thickness of the electrochemically developed films) is
about 0.7-0.8 V. However, Rudge et al. I9 have developed some derivatized
poly thiophene materials that show two regions of pseudocapacitative activity
corresponding to p and n doping well separated by a substantial voltage difference of about 3.0 V. These materials are of special current interest; see Chapter
12.
In aqueous solutions, the range of operation of any electrochemical capacitor is determined theoretically by the decomposition voltage of the solvent,
water, i.e., 1.23 V at 298 K and standard pressures ofH2 and O2, In practice, this
range may extend somewhat further, depending on the overvoltages for H2 and
O 2 evolution at the negative and positive electrodes as a function of current density (the Tafel parameters). However, no significant quantities of H2 and O 2
should be allowed to build up since then serious self-discharge (Chapter 18) can
set in.
In the case of oxide redox capacitor systems, the potential ofthe lowest redox couple should therefore be a little positive to the hydrogen reversible potential and that of the highest should be just below the oxygen potential. This
will achieve the maximum range of redox pseudocapacitance, as found with
RU02' which (as explained elsewhere) must involve several redox processes
overlapping in their potential ranges. If the onset of reversible oxidation and reduction behavior is for thermodynamic reasons substantially more positive than
the hydrogen reversible potential, as with Co and Ir thick-film oxides, then the
achievable operating range for the pseudocapacitance-based electrochemical
capacitor will be too short for practical use.
248
Chapter 10
Other oxide-film systems, e.g., the thick films that can be generated at Pt,
Rh, Au, or Ni (in NiOOH), can only be generated anodically at high potentials
in the Oz-evolution potential region and their reduction is highly irreversible.
They hence have no value as pseudocapacitor materials.
It is uncertain if the operating potential range of RU02 could be substantially extended in nonaqueous electrolytes. Protons have to be available, so
some acidic solute in a nonaqueous solvent would be required, e.g., CF3S03H
in a polar nonaqueous medium. Use of an Li salt as electrolyte would in principle be possible, but then the system would become equivalent to an Li+ ion intercalation battery system with Li+ insertion and deinsertion instead of the
corresponding protonic processes that take place in aqueous acidic solutions at
RU02' Such systems are already well known with cathode materials such as LiCo0 2, Mn02, or V60 13 . As explained in the following section, Li+ intercalation
battery systems behave as a transitional class of electrochemical power sources
since they exhibit pseudocapacitance because their state of charge is dependent
on potential for fundamental thermodynamic reasons, but they are commonly
regarded as nonaqueous electrolyte "batteries."
The generality of Eqs. (10.1) and (10.2) was emphasized earlier in this
chapter insofar as they apply equivalently to a 2-dimensional electro sorption
process (UPD), a 3-dimensional sorption process with charge transfer (electrochemical intercalation), or a redox reaction at an electrode. The detailed development in previous sections was addressed to adsorption pseudocapacitance,
based on Eqs. (10.6), (10.7), (10.9), and (10.10).
Capacitors based on inorganic redox reagents also have a capacitance that
is electrochemical rather than electrostatic in origin. Consider the following example:
(10.34)
for the ferri-ferrocyanide couple. This can be written in an abbreviated general
form for the (any) oxidized (Ox) and reduced (Red) species:
Ox+ e~Red
(10.35)
The Nernst equation for the equilibrium redox potential E for such a 1 e
system is written
E = EO + (RT/F) In[Ox]/[Red]
(10.36)
249
where [ ] represents the concentration or activity of the indicated species. Equation (10.36) can be rewritten, for a given molar quantity, Q, of reagent (i.e., Q
total [Ox] + [Red] available for reaction) in terms of the fractions:
E = EO + (RT/F) In[OxlQ]/[RedlQ]
(10.37)
(10.38)
(lOAO)
[I +exp(Ll FfRT)]2
250
Chapter 10
0.6
I
BOUND
,'-,/REDOX
I
, IE
0.4
pc
DIFFUSION
/REDOX
0.2
,,
c:
.2
U
-c:
,- /
,
I
(10.41)
~~
::l
c:
....
....
Q)
::l
-0.4
I
I
I
I
-0.6
I
I
I
I
I
I
I
;""
'-'
Potential IV
FIGURE 10.14. Cyclic voltammograms for the Fe(CN)~-lFe(CN)~- redox couple giving rise to
pseudocapacitance: (a) free in solution (diffusion controlled) compared with surface bound; (b) experimental example of ferrocyanide bound on poly(vinylpyridine). (From Conway and Duic, unpublished. )
251
-.l
1 mA
A (lnitiol)
I \
,
," ,
,,, ,
I
"
....
_--
",'----
,
,,
I
,
,
\
"
I
I
lOOmV
--I
I--
where X is the 3-dimensional site fraction occupancy by the sorbed guest atom
or ion, e.g., Li.1t is seen that this equation has the same form as Eq. (10.1) (after
conversion to its logarithmic form) and hence will give a capacitance relation
like Eq. (10.2), or the corresponding equations for 2-dimensiona1 electrosorption or for the redox case. It is because of Eq. (10.41) that Li ion intercalation
batteries can be regarded in some sense as pseudocapacitors since their extents
of charge are a continuous function of their voltage for fundamental reasons.
However, their response times in a capacitative sense are longer than with double-layer capacitors owing to the necessity for interlayer diffusion in the solid
host lattice.
The similarity to adsorption pseudocapacitance arises directly because Li+
ion accommodation is really a quasi-2-dimensional phenomenon, occurring at
the layer surfaces of the van der Waals gaps in the layer-lattice structures of
e.g., TiS 2 or MoS 2. It is interesting that the quasi-2-dimensional intercalation
processes involving Li+ ions take place in distinguishable progressive stages, as
revealed by cyclic voltammetry.2o Examples are shown in Figs. 10.15 and
Chapter 10
252
-200
o
DISCHARGE
200
-1.2
-1.0
-1.4
-1.8
-1.6
POTENTIAL/V
FIGURE 10.15. Slow cyclic voltammogram (sweep rate 9 /iV
ciated with U+ intercalation into M002. (From Sacken. 21 )
S-I)
0.03
0.02
0.01
<
0
-0.01
..............-. ..._..-'-'" B
-0.02
-0.03
1
1.5
2.5
FIGURE 10.16. Differential electrochemical sorption and desorption curves showing three-state
accommodation of Li in a fullerene, C60 (compare Fig. 10.15). (From Chabre?2)
253
10.16, where two or three current response peaks are resolved in slow cyclic
voltammetry for Li+ intercalation and deintercalation at Mo0 2 (Fig. 10.15)21
and remarkably at a fullerene (Fig. 10.16).22 It is seen that this behavior is
closely similar to that observed [e.g., as in Figs. 10.4(a) and 1O.4(b)] for 2-dimensional electrosorption (Section 10.2.2) of H onto single-crystal Pt surfaces
or of Pb onto gold surfaces. In both types of cases it must be assumed that successive lattice array structures are set Up20,23 having distinguishably characteristic Gibbs energies as the lattice occupancy fraction (0 for 2-dimensional
adsorption or X for 3-dimensional intercalation) progressively increases or decreases. 20 In both these cases, the asymmetry between the sorption and desorption peaks is attributable to the slowness of diffusion. The three cases
(2-dimensional underpotential deposition, redox systems, and quasi-2-dimensional Li+ intercalation processes) and their corresponding Nernst-type equations are summarized in Table 10.3.
System Type
(a)
Redox system:
Ox+ze~Red
(b)
Intercalation system:
RT
=0+ zF InX/(I-X)
(c)
Underpotential deposition
MZ+ + S + ze~ SM
(S
Note: (b) and (c) can be regarded as mixing of occupied (X or IJ) sites with unoccupied sites. (1 -IJ) or (1 - X).
Also IJ + (1 -IJ) 1 or X + (1 - X) 1.
Another source of pseudocapacitance exists when so-called "specific adsorption" of ions, usually anions, takes place at electrodes. Specific adsorption
(i.e., chemisorption of ions) was discussed in Chapter 6 on the double layer. The
254
Chapter 10
effect occurs when the adsorption is not only electrostatic in origin (i.e., dependent on the charge on the ion and the surface charge density at the electrode) but
also depends on some "specific" electronic interaction between the ion's valence electrons and the electrode's surface orbitals.
Such effects are usually accompanied by a partial transfer of charge, e.g.,
as in the following quasi-chemical equation for adsorption of an anion A-at an
electrode M:
M
+ A- ~ MJA(l-oj- + 0 e(in M)
(II)
where 0 is some fraction of an electron charge. Process (II) has the nature of a
Lewis acid-base donor-acceptor interaction, the anion being the donor. Adsorption takes place more strongly when M is positively charged (Lewis-acid
character) but at the potential-of-zero charge for many anions. The quantity 0
(or 1 - 0) is a component of the so-called "electrosorption valence" of the anion,
which characterizes its chemisorption behavior at electrodes. 12,13
The charge passed in process (II) can be determined microcoulometrically
in a thin-layer cell or by means of cyclic voltarnmetry: for example, Fig. 10.17
shows the current response for HS0 4ion chemisorption at Au in a single potential-sweep cycle; the charge under the anodic or cathodic sides of the voltammogram gives the extent of partial charge passed, plus the charge
accommodated in the double layer over the same potential range. These two
charge components are not easily separated except under favorable conditions
by means of ac frequency-response analysis (Chapters 16 and 17). The pseudocapacitance associated with processes like II is not normally appropriate for
5
Au (111)
02
-5
FIGURE 10.17. Current-response profile for specific adsorption of sulfate ion at polycrystalline
Au recorded by single-cycle voltarnmetry.
255
In recent years, during the course of development of double-layer capacitors based on high-area carbon powder or fiber materials, it has come to be recognized that some fraction of the capacitance exhibited is Faradaic rather than
electrostatic in origin; that is, there is some pseudocapacitance component. It is
believed to arise from the electrochemical reactivity of surface groups (see
Chapter 9), e.g., on the edge sites of graphitic planes of the carbon particles.
Quinonoid or aldehydic groups, which could provide redox activity, may be responsible for this pseudocapacitance component of the overall capacitance of
these electrode materials.
The extent of this Faradaic capacitance probably lies in the range 1 to 5%,
the latter figure being a maximum, of the total capacitance measurable at low
frequencies. Surface functionalities present on high-area carbon materials depend very much on the origin and pretreatment of the carbon, as discussed in
Chapter 9 and in the monograph by Kinoshita. 24
256
Chapter 10
In the case of a porous electrode (Chapter 14), however, the distinction between Cdl and CtP , when the latter is significant, is much more difficult owing to
the distributed nature of both CtP and Cdl throughout the porous matrix, with a
progressively increasing electrolyte resistance down pores (Chapter 17) in a series-parallel arrangement with the C components. For such reasons, it is usually
difficult to distinguish a CtP component in carbon double-layer capacitors, although it is probably significant owing to reactive surface functional groups
(Chapter 9).
The discussion given here concerning the distinction between double-layer
and pseudocapacitance in experimental measurements has implicitly assumed
that the processes involved in double-layer charging and Faradaic passage of a
charge in charging the pseudocapacitance at the same electrode are independent.
This, however, is not necessarily true and there can in fact be some coupling between double-layer charging and simultaneous Faradaic processes. Although it
is a somewhat arcane topic, this matter has been thoroughly discussed and
treated in a review by Parsons.z5 It can be a significant factor when quasiFaradaic processes such as anion-specific adsorption with partial charge transfer12,13 are involved. However, more detailed examination of this problem is
beyond the scope of this chapter, so readers are referred to Ref. 25 and the literature cited there.
REFERENCES
1. B. E. Conway and E. Gileadi, Trans. Faraday Soc., 58, 2493 (1962).
2. E. Gileadi and B. E. Conway, in Modem Aspects of Electrochemistry, J. O'M. Bockris and B.
E. Conway, eds., vol. 3, Chapter 2, Butterworths, London (1965).
3. E. Gileadi, Electrosorption, Plenum, New York (1967).
4. A. Eucken and B. Weblus, Zeit. Elektrochem., 55, 114 (1951).
5. E. Gileadi and B. E. Conway, 1. Chem. Phys., 31, 716 (1964).
6. E. Gileadi and S. Srinivasan, Electrochim. Acta, 11, 321 (1966).
7. H. A. Kozlowska, 1. Klinger, and B. E. Conway, 1. Electroanal. Chem., 75, 45 (1977).
8. M. Boudart, 1. Amer. Chem. Soc., 72, 3566 (1952).
9. T. L. Hill, Introduction to Statistical Thermodynamics, Addison-Wesley, Reading, Mass.
(1960).
10. 1. G. N. Thomas, Trans. Faraday Soc., 57, 1603 (1961).
11. S. K. Rangarajan, Specialist Periodical Reports (Chemical Society, London), vol. 7, p. 203
(1980).
12. W. Lorenz and G. Salie, Zeit. Phys. Chem., N.F., 29, 390,408 (1961).
13. 1. W. Schultze and F. D. Koppitz, Electrochim. Acta, 21, 327, 337 (1976).
14. M. I. Temkin, Zhur. Fiz. Khim., 15, 296 (1941).
15. B. E. Conway and J. Ku, 1. Coli. Interfacial Sci., 52, 542 (1973).
16. H. A. Kozlowska and B. E. Conway, 1. Electroanal. Chem., 95,1 (1979).
17. 1. O'M. Bockris and H. Kita, 1. Electrochem. Soc., 108, 676 (1961).
18. K. Engelsman, W. J. Lorenz, and E. Schmidt, 1. Electroanal. Chem., 114, 1 (1980).
19. A. Rudge, 1. Davey, I. Raistrick, and S. Gottesfeld, 1. Power Sources, 47, 89 (1994).
257
Chapter 11
260
Chapter 11
equation with regard to its electrode potential, can give rise to a pseudocapacitance over a range of potentials. This is the main basis of the pseudocapacitance
developed at RU02 in anodic and cathodic cycling; it also originates with a simple redox couple (e.g., Fe(CN)t-Fe(CN)~-) in a thin-layer cell or anchored on
a polymer as shown in Chapter 10. However, the latter type of system gives appreciable pseudocapacitance only over about a 120-mV potential range,
whereas RU02 gives relatively constant and appreciable capacitance over a 1.4V range. 5,6 Later in this chapter we suggest how this important difference may
originate in terms of several overlapping redox processes within the l.4-V potential range.
Ruthenium oxide films can also be formed at Ru metal electrodes by potential cycling over about a 1.4-V range above the hydrogen reversible potential,
e.g., in aqueous sulfuric acid. Initially monolayer oxide formation and reduction
take place (Fig. 11.1), like that at Pt. However, upon continuous cycling, the oxide film progressively grows (Fig. 11.2) since it is not reduced to Ru metal at
the least positive potential of the sweeps (+0.05 V, RHE), as is the initial monolayer at Ru or Pt.
Ruthenium dioxide films thus formed give a cyclic voltammogram having
the shape shown in Fig. 11.2 and a corresponding charge and discharge diagram
closely similar to that for a capacitor. 5- 7 It was on this basis that Trasatti and
Buzzanca6 recognized the interesting pseudocapacitative behavior of this Ru02
film material, since an almost rectangular, mirror-image cyclic voltammogram
(Fig. 11.2) is characteristic of a capacitance. A useful but brief review on RU02
as a electrochemical capacitor material for electrical energy storage has recently
been published by Trasatti (see general reading ref. 1). The Ru02 material
formed thermochemically on Ti also gives a voltammogram similar to that in
Fig. 11.2.
Especially with the RU02 films thermally formed on Ti with Ti0 2, or with
Ta205, the cycle life in capacitor charge and discharge between 0.05 and 1.2 V,
or even to 1.40 V (RHE) under some conditions is remarkable, allowing cycling
over 105 times with little degradation. What small degradation may take place
arises due to RuO~- formation at the high potential ends of the anodic voltage
excurSIOns.
By integration ofthe I vs. E voltammogram curves, the charges, q, for supposed oxidation and reduction of Ru atoms in the thermally prepared Ru02
films were related4 to the fraction (%) of Ru atoms involved in solid-phase redox reactions (Table 11.1). This fraction is actually quite small and tended to
decrease with increasing temperature of calcination between 350 and 450C.
The q values in Table 11.1 cannot be related only to surface redox processes;
some bulk-phase reactions must also take place and some double-layer charging
must of course also be appreciable. This conclusion has been confirmed by
___- r
---
./"""
261
---
"
0
cu
0
cu
~ u
;;; .5
J
j
'"
";
..z
Irreversible
region
E
u
<t
c.
~
:;
u
.5
....I:
u
2.5
--Reversible profile
/'"
FIGURE 11.1. Cyclic voltammograms for Ru metal in aqueous I mol dm-3 H2S04 at 298 K
showing behavior of initial cycles where monolayer oxide formation and reduction, like that at Pt,
arises. Reprinted from S. Hadzi-Jordanov, H. A. Kozlowska, and B. E. Conway, 1. Electroanal.
Chern., 60, 359 (1975), with permission from Elsevier Science.
262
Chapter 11
3.0
'0
o
c:
ANODIC-_-_ _ _ _
-:::::=====:::.
2.0
<{
c::t
1.0
......
I-
a::
a::
-1.0
:>
:0 -2.0
.s::.
0
(j
-3.0
FIGURE 11.2. Progressive increase of oxide film growth with potential cycling at Ru, exhibiting
qlmC
cm- 2
17
35
70
50
18
67
57
42
Fraction (%)
6.7
6.9
6.9
4.9
1.8
6.6
5.6
4.1
Source: Reprinted from S. Trasatti and G. Buzzanca, J. Eleclroanal. Chern., 29, Appl. 1 (1971 l, with permission
from Elsevier Science.
263
possible mechanisms of the oxide redox processes. Nonstoichiometry of the oxide material was concluded to be significant and is no doubt in part responsible
for the good electronic conductivity of this oxide material.
Soon after this work by Trasatti, Hadzi-Iordanov et at7 8 demonstrated
closely related pseudocapacitative behavior of RU02 films generated anodically
at Ru metal by potential cycling (at ca. 25 mV s-') between 0.05 and 1.40 V
(RHE) in aqueous H 2S04 , In fact, the voltammetric behavior was even more reversible (mirror-image, anodic and cathodic, voltammograms) than that at thermally formed RU02 on Ti. Reversibly behaving Ru02 films could also be
generated thermally on the surface of a glass tube7 provided with an internal-toexternal Pt-wire contact. The anodic and cathodic charge values derived from
integrating the l/E voltammetry plots progressively increased with cycling and
the so-called "charge enhancement factor" (CEF)" was linear with increasing
number of cycles up to a CEF corresponding to about 200 equivalent monolayer
charges. Similar behavior was found at Ir (Fig. 11.3). It is interesting that enhancement of electrocatalysis for anodic Cl 2 evolution at these oxide films followed the CEF linearly,9 suggesting that the latter reflected a progressive
increase of accessible oxide surface area.
In the work by Hadzi-Iordanov, Kozlowska, and Conway8 on oxide film
formation and cycling at Ru metal, the very initial stages of submonolayer formation and reduction of oxide species at the metal/aqueous H 2S04 interface
were distinguished (see Fig. 11.1) from the subsequent, much more reversible,
oxidation and reduction processes that arise after progressive film thickening
has occurred (Fig. 11.2), which takes place at Ru upon extended
potentiodynamic cycling between 0.05 and 1.40 V (RHE). The initial current
response profile at a clean, oxide-free Ru surface is similar to that at Pt.
0.5
0.25
<t
.....
I-
o~--~~~--------------------~
c::
~ -0.25
U
-O.5L-________~--------~~------~
0.5
POTENTIAL, EH IV
1.0
1.5
FIGURE 11.3. Cyclic voltammograms for (almost) reversibly formed and reduced oxide films of
lr02 on Ir. Annotations indicate distinguishable, almost-reversible, states on the curve. State M is
underlying Ir metal with adsorbed H. (From Mozota and Conway, unpublished.)
264
Chapter 11
Following the first publication by Hadzi-Iordanov, Kozlowska, and ConwaySon the behavior of RU02 films and their processes of formation, in 1975
the Ottawa group were the first to undertake research and development on RU02
materials for commercial electrochemical capacitor development and production. This work was done in collaboration with Continental Group Inc., through
their laboratories at Cupertino, California. During a 6-year project, various embodiments of the Ru02 pseudocapacitor device based on mixed oxides (RuOz +
TiO z or TazOs) were developed and tested, including bipolar-stacked electrodes
giving 12-V devices. Unfortunately, much ofthis work remains in confidential
reports. This operation was taken over around 1981 by Pinnacle Research Corp.,
who have successfully continued development of this technology to the present,
mainly with military funding. Further details are discussed in Chapter 20 on
Technology Development.
A related reversible oxide formation and reduction process that exhibits
substantial adsorption pseudocapacitance was discovered and examined by
Kozlowska, Conway, and Sharpl2; it arises in the initial (submonolayer)
stages of oxide film formation and reduction at Pt and Au in aqueous electrolyte solutions and was quantitatively characterized, again by means of cyclic
voltammetry. Between 0.85 and 1.0 V (RHE) at Pt, a range of reversible electrochemical deposition and desorption processes can be identified, especially at
low temperatures, before irreversible place-exchange surface reactions take
place. An "envelope" of such reversible, almost mirror-image current responses
can be recorded. However, the potential range, ca. only 0.15 V, over which these
processes are reversible and hence exhibit pseudocapacitance (see Chapter to),
is too short to be of practical interest, even for specialized purposes for which
the use of Pt might be justified.
11.2. INTRODUCTION
It was explained in Chapter 1 how there are essentially two types of electrochemical capacitors: one based on mainly electrostatic energy storage in the
charged double-layer capacitance associated with the interfaces of solutions and
large specific-area, carbon-based materials (e.g., carbon powders, fibers,
and felts); the other kind originating from the pseudocapacitance associated with Faradaic adsorption (electrosorption) or redox processes in
which potential-dependent extents of passage of Faradaic charge can
arise; hence a capacitance can be formally measured (Chapter 10).
Two principal types of material in this second category have been discovered and studied: (1) electroactive oxide or hydrous oxide films of transition
metals,I-4 e.g., RuOz and IrOz, and some other oxides such as C0 30 4, Mo0 3 ,
W0 3 , and molybdenum nitride of uncertain composition; and (2) electrogener-
265
266
Chapter 11
The device was designed for possible use in a microminature electrical neural
stimulator control module, but the achieved capacitance of 300 flF or a to-cell
series unit much exceeds the 0.7 flF required for this particular application, so
opportunity for further minaturization is available.
The RU02 used was prepared by thermal decomposition of RuCl 3 in an
NaN0 3 flux. Two approaches were employed for forming the Ru02-ionomer
267
268
Chapter 11
269
dc polarization at some appropriate positive potential (e.g., > 1.0 V at Pt), the
rate of oxide film extension is directly dependent on the log of the time of polarization (direct log law) for thin-film formation 23 - 25 while for longer times the
inverse log law in the sense of Mott and Cabrera24 applies.
These two principal growth laws for oxide film formation correspond
to different mechanisms of oxide extension and to different ionic and electronic conductivities of the resulting films. In some cases (e.g., at Pt), the
mechanism changes after some initial reconstructed oxide film is formed,
e.g., to an extent corresponding formally to "pt0 2,,25 which is actually two
layers of PtO in a compact film state. Further growth beyond this film occurs
in a hydrous oxide state that is independently reducible from the compact film
component. However, neither component of the overall film exhibits reversible reduction and reoxidation like Ru02 or Ir0 2. The formation of Ru02 films
upon cycling also appears (like Pt) to take place in two stages, distinguished
in this case by two regions of potential dependence, as shown in work by
Hadzi-lordanov et al. 7 ,12 and illustrated in Fig. 11.4.
E
.24
.,
.
...;
~3
.II:.
c:
.2
Ci
'02
'j(
l:
'0
~1
03
0 .5
0 .7
0 .8
1.0
1.2
1.4
POTENTIAL, V IE H )
FIGURE 11.4. Two regions of potential dependence of oxidation charge for Ru in aqueous
H2S04 at 298 K. Reprinted from S. Hadzi-Jordanov, H. A. Kozlowska, and B. E. Conway, 1. Electroanal. Chern., 60, 359 (1975), with permission from Elsevier Science.
270
Chapter 11
(11.1)
RuOz + H+ + e ~ RuOOH
01.2)
may be a preferred representation for the singular rest potential. A similar type
of reaction has been proposed for Ni0 2 at low temperatures in aqueous KOH in
the Ni-Cd battery. However, the broad range of oxidation and reduction in the
voltammograms must correspond to several overlapping stages of that process.
The standard potential for the first reaction has been reported as 0.94 V
(RHE) (Pourbaix Atlas of Electrochemical Equilibria). Galizzioli, Tantardini,
and Trasatti5 measured potentiostatically the quasi-stationary potentials set up
at RuOz films at small anodic polarizations, 10-5_10- 3 A cm-2. However, on
open circuit, time-dependent self-discharge and recovery potentials are observed26 ,27 (see later discussion), which eventually approach ca. 0.75 V in acid
solutions. 26
Data for the Gibbs energy of formation of Ru02 enabled a value of 0.43 V
to be derived for the standard potential of the process
RuOz + 4H+ + 4e ~ Ru + 2H zO
(11.3)
but this is not the stationary potential observed in acid media. However, under
cyclic voltammetry conditions, Ru02 is almost certainly, like Ir02, not reduced
271
back to the metallic state in the negative-going sweep to 0.05 V (see Ref. 28),
so the process in Eq. (11.3) is not relevant to the experimental situation. (Ellipsometry work by Gottesfeld et a1. 22 showed that an iridium oxide film remained
when the underlying surface of Ir became electrodeposited with H.)
In potential-step experiments at Ru02 from ca. 1.4 to 0.05 V, Galizzioli,
Tantardini, and Trasatti5 observed that the electrochemical behavior became
like that of a capacitor. This was one of the early indications of capacitative behavior (as a pseudocapacitance) of Ru02, as mentioned earlier. Another indication was the observation of mirror-image cyclic voltammograms for RU02
reported by Trasatti and Bucanzza6 and by Hadzi-Jordanov et al? However, this
type of observation could be alternatively interpreted as a double-layer current
response at a large distributed surface area of Ru02> as will be seen later (see
Refs. 16 and 17).
Returning to quasi-thermodynamic potentials, Pell, Liu, and Conway26 recently found that either a thermochemically or electrochemically formed (hence
a hydrous film) film of RU02 decayed in potential (on open circuit; see Chapter
18 and Refs. 29 and 30) after anodic polarization to 1.4 V, down to 0.75 V,
while, after discharge to 0.1 V, it recovered in potential to almost that value,
which remained stable for many hours. Thus 0.75 V seems to correspond to a
well-defined redox potential; it happens to correspond (at pH -0) to the line in
the Pourbaix pH potential diagram for Ru for equilibrium between Ru and
RU203, which is an interesting result but one inconsistent with the difficulty of
reducing Ru oxides back to the metallic Ru state, the problem referred to in Eq.
(11.3).
Although superficially (e.g., in the cyclic voltammograms) the thermally
and electrochemically grown Ru02 films behave similarly, Pell, Liu, and Conway26 found that the pH dependence of their stationary potentials was reproducibly quite different, being -30 mV for the former film and -60 m V for the latter.
By contrast, Daghetti, Lodi, and Trasatti31 found also a value of -60 mV for the
thermally formed oxide film. The two values found in the work of Ref. 26 indicate
a qualitative difference between the thermodynamically potential-determining
processes that are involved at the initially anhydrous and the electrochemically
formed hydrous oxide films. Thus the -60-mV slope (corrected for the pH dependence of the RHE) requires a process involving an equal number of electrons and
protons, while the -30-mV slope requires a ratio of elH+ of 2.
A similar recovery behavior (see Ref. 26) at thermally formed RU02 had
been observed earlier by Arikado et aI.,27 but the potential approached was ca.
0.5 V rather than 0.75. In this work, a role for proton diffusion in the recovery
process was proposed.
In the voltammogram for RU02 in 1 M aqueous HCI04 at 60 m V s-l, two
conjugate (anodic and cathodic) pairs of peaks are observed (but not well resolved) at ca. 1.14 and 0.6 V (RHE) which were attributed5 to the RU20rRu02
272
Chapter 11
and RU20rRu couples. However, as remarked earlier, Ru02 or other lower oxides are not returned to the Ruo state in the cathodic sweep except when only a
monolayer of Ru surface oxide has been formed 7 as at Pt. 12 In fact, it is precisely
the incomplete reduction of RU02 back to Ru on each sweep that is believed to
be the origin of the phenomenon of electrochemical growth of RU02 films on
potential cycling from 0.05 to 1.4 V. The increase in thickness of the RU02 film
is up to several microns, as is visible under the electron microscope or even under the optical interference microscope. 13
Under potentiodynamic conditions, the initial stages of oxide film formation are typical of an irreversible, slow anodic process, as also found at pt. 12
Some coverage by surface oxide can be established in an anodic sweep up to a
certain potential, say, 1.0 V. Upon arrest of the sweep, growth continues at constant potential; then when the anodic sweep is restarted, the original i-V voltammogram is only picked up at some 50 to 100 mV higher potentials. This
means that the oxide that is formed at one (higher) potential (V2) in a sweep can
also be formed upon holding in time at a lower potential VI' < V2 .
Galizzioli, Tantardini, and Trasatti5 concluded that not all the Ru atoms in
the Ru02lattice were redox active, giving rise to pseudocapacitance; that is, in
a potential range of 1.35 to 0.3 V, the formal process Ru4+ + e -4 Ru 3+ (in the
oxide) was considered to be restricted to a number of atoms corresponding to
the degree of nonstoichiometry. Coupled with this change of oxidation state
(transfer of electrons) is a corresponding injection or withdrawal of protons to
maintain a charge balance through conversion of 0 2- ions to OH-. Possibly
these coupled redox and proton-exchange processes take place only in the nearsuiface region 26 of RU02, especially when it has been thermochemically prepared. Electrochemically prepared RU02 films are hydrous in nature, so that a
more facile proton penetration, coupled with redox changes of oxidation state,
can take place, as suggested earlier.
The three-dimensional and near-surface (see Refs. 32 and 33) chemical
constitution of thermally formed RU02 was investigated34 using depth profiling
by means of sputtering. Oxygen species are rich in the surface region but decrease in concentration toward the bulk, where Ru also decreases (Fig. 11.5).
Cl- ion is adventitiously present and increases somewhat toward the bulk; it
arises from the RuC1 3 used to make the thermochemically generated RU02 film.
A new charge storage mechanism involving protonation ofRu02 (presumably coupled with reduction) was recently claimed by Zheng and Jow. 16 However, it does not seem any different from the basis of operation of the redox
pseudocapacitance of RU02 recognized in various, much earlier papers 4- 7,27 and
more recently by Sarangapani et al.,9-21 as well as with oxides of Ni. 35 Good
specific capacitance was obtained with an RU02 . xH 20 material made by a solgel precipitation process, followed by heating between 25 and 400C. Values of
capacitance were up to 350 F g-I while, with the hydrous form, up to 720 F g-I
273
0
...J
W
>=
1.2
>-
0:
W .0
(/)
\
I/)
<l
0 c
Z ::J
0 ..: 0.8
~
..... 0
'.---.....-.......-.
16 0-
.'.....
U
w
a...
(/)
0.4
10
15
20
were obtained. The X-ray diffraction patterns of the materials formed at various
temperatures were recorded as shown in Fig. 11.6. Higher annealing temperatures promoted the formation of more crystalline oxide structures (Fig. 11.6),
which are less active for charge accommodation, presumably owing to their
more anhydrous character.
Further details of the work in Ref. 16 were given in a following extended
paper (Ref. 17) in which, among other things, cyclic voltammograms, which
give the capacitance (C = i/s; see Chapter 10), were comparatively shown for
preparation temperatures of Ru02 varied over the range 25 to 400C. The capacitative response currents diminished with increasing temperature of the heat
treatment. The charge and discharge behavior was represented by almost linear
relations between charge and potential, giving almost constant capacitance over
a potential range of 1.0 V, as had been found in other earlier work. 5 ,7,8 The capacitance as a function of cycle number remained virtually constant after an initial fall by ca. 6% during the first 700 cycles. The specific capacitance (F g-l)
remained constant for various preparations with specific area (m2 g-l) over a
range of 40 to 96 m2 g-l. From this result it was concluded 18 that the capacitative
current response was due, not only to the surface of the oxide, but arose appreciably on account of redox pseudocapacitance (Chapter 10) taking place within
the bulk of the hydrous oxide material, confirming earlier indications.
274
Chapter 11
7000 ,-----:-:--:-:-.---:--------:-::-::-::-:-----,
(110)
(101)
(211)
6000
(200)
(111 )
O+---~--~--~--_r--_r~~--_,--~
20
30
40
50
60
2 Theta (degrees)
FIGURE 11.6. X-ray diffraction plots for RU02 thermally formed at various temperatures. Indices identify anhydrous RU02 diffraction peaks. (From Zheng and Jow. 16 Reproduced by permission
of The Electrochemical Society, Inc.)
FIGURE 11.7. Thick, anodically grown RU02 films on Ru metal showing spalling off. (From
Birss, Myers, and Conway.13 Reproduced by permission of The Electrochemical Society, Inc.)
275
FIGURE 11.8. Columnar structures of RU02 developed under anodic cycling at Ru metal. (From
Birss, Myers, and Conway. i3 Reproduced by permission of The Electrochemical Society, Inc.)
276
Chapter 11
ration. This determines the microsurface structure and thus the real-to-apparent
surface area ratio.
The macroscopic features of electrochemically prepared RU02 (i.e., films
formed by cycling Ru metal between 0.05 and 1.4 V) were compared with those
of the oxide thermochemically generated on Ti in a paper by Birss, Myers, and
ConwayY Both scanning electron microscopy (SEM) and interference optical
microscopy procedures were used.
Electrochemically grown films were up to several microns in thickness and
spalling off detached sheets of the oxide could be seen (Fig. 11.7) in some photos. Under the SEM, remarkable columnar structures could be observed (Fig.
11.8), which could account for the large specific capacitance found with this
material. In constrast, the material thermochemically formed on Ti had the mudcrack appearance well known in the DSA technology field.
We conclude this section by remarking again that capacitative behavior
quite similar to that of RU02 is observed with Ir0 2 in acid solution. However,
reversible charging arises only at potentials positive to ca. 0.6 V. This is due to
the effect of potential, below 0.6 V (RHE), on the electronic conductivity of
Ir0 2 as the result of a change in oxidation state. This behavior is also paralleled
by an electrochromic effect.
with
(11.7)
leading to "Ru(OH)2," can take place. Relatively free mobility ofthe protons is
essential, as in the basis of the proton-jumping mobility of protons in aqueous
acid or alkaline solutions. The H+ can be looked upon as an intercalated species,
but unlike Li+ in Mn02 or CoOl> it has localized states (e.g., on 0 2- as OH-) or
in any water molecule, as H30+ at acid pH. A similar proton migration process
may also occur at the thermally formed RU02, but probably only in or near its
surface,5,16,17,37 as studied by Arikado, Iwakura, and Tamura,27 and it depends
on the temperature of the prior heat treatment. 16,17
The above processes may go on more easily in the near-surface region of
RU02 material than in the bulk and will be sensitive to structural and nonstoichiometric 5 imperfections in the oxide. In parallel with the above redox charge and discharge mechanism, a significant quantity of charge
will also be accommodated through double-layer charging since the
RU02 materials usually have a relatively large specific area of 26 to 90
m 2 g-l, depending on formation temperature, as noted earlier. Further details of the possible redox steps are given in Section 11.7.
278
Chapter 11
Since the conversion of hydrated RuO to hydrated RU02 requires 2e, J in the Eq.
(11.9) would be about 1. It is important that in their paper, and in J ow's, 16,17 a
hydrated form of RU02 or other oxides is postulated. This is significant for the
mechanism of charging and discharging where facile proton insertion and deinsertion37,38, coupled with electron transfer through the electronically conducting
oxide, is supposedly involved. Such a mechanism has precedents in the mechanisms of electrochemical reduction that have been established for NiOOH to
Ni(OHh in the Ni-Cd battery and for the so-called "homogeneous mechanism"
for the first-electron reduction of Mn02.37,38
Based on X-ray emission analysis and SEM examination, Mitchell, Rand,
and WOOdS 14 deduced that ruthenium oxide films formed after 14,000 cycles
were about 200 nm thick (see Ref. 13 on related SEM work). They concluded
that such films were predominantly ofRuO stoichiometry, with a considerable
degree of nonstoichiometry (see Ref. 5), which would account for their good
electronic conductivity. However, a dissolution and redeposition mechanism,
possibly involving RU04, was suggested.
Charge and oxide thickness evaluations by Mitchell, Rand, and Woods 14
gave a value of 96 mC cm- 2passed between 0.06 and 1.5 V for an electrode covered by a 200-nm film. Taking the density of hydrated RU02 as 25-40% of that
of rutile RU02, ca. 65 wt. % Ru would be calculated. Thus the surface oxide
should contain between 20 and 40 f1g cm-2 of Ru and require a charge of 40-70
mC cm- 2 for oxidation of (supposed) RuO to RU02' These figures are of the
same order of magnitude as the obse:ved charge passed in the cyclic voltammograms. Because there is a high surface area of the cycled material, appreciable
charge is also involved in double-layer charging and discharging, as noted earlier.
A comparison with the charge and discharge behavior of thermochemically generated RU02 is important and was also considered by Mitchell, Rand,
and WOOdS. 14 The correlations between the experimental and calculated redox
charges led to the conclusion that each Ru atom in the electrochemically formed
oxide film is involved in the one-electron redox reaction, written as Eq. (11.2),
a conclusion also reached by Arikado, Iwakura, and Tamura. 27 However, this is
not the case for the thermally formed film, where the voltammetric charge (for
otherwise similar conditions) is only 3 mC cm- 2 even though the film is visible 13
and appears substantial. The above results 14 support the conclusion by Trasatti
and Buzzanca6 that only 2-7% of the atoms in thermally produced RU02 are
electrochemically involved in a redox electron-transfer process (Table 11.1).
Thus it was supposed that such electron transfers were localized to sites where
water molecules remained trapped in the lattice during its formation. In fact,
Kozawa37 had reported that hydration of RU02 is essential for discharge capacity to be exhibited, which is a reasonable requirement for proton transport to occur.
279
With ruthenium oxide electrodes having real areas of 60-80 times the apparent areas, Arikado, Iwakura, and Tamura27 deduced that approximately one
electron per surface atom of Ru was involved in the electron-transfer redox
process (see Ref. 14 and Eq. 11.2). However, for films prepared at 450 to 550o e,
the charge values were too large for it to be concluded that only surface-atom
reactions were involved, so that some solid-phase, 3-dimensional redox processes must also be taking place, but not in the whole material. Note that for some
DSA RU02 electrode preparations, the real-to-apparent area ratios are around
400.
Another significant result obtained in work by Mitchell et a1. 14 was proof
of the involvement of proton diffusion (presumably via a jumping process from
OH- to 0 2- etc. in the hydrated oxide). This was found by some clever experiments using Ru02 supported on a Pd membrane from or to which protons could
move. RU02 reacts with an H from the Pd (in an H+ + e process) involving the
reduction step RU02 ~ RuOOH (Eq. 11.2), as with NiOOH or Mn02?8,39
In addition to cyclic voltammetry work on Ru02, giving highly reversible
cyclic voltammograms, Raistrick 15 has carried out electrochemical quartz microbalance studies on this material and BET area measurements. The capacitance was found to be proportional to the surface area, and a figure of 380 F g-1
was found for material of 130 m2 g-l. This corresponds (for e.g., a I-V operating
span at a single electrode) to 380 x 106/130 x 104 JiF cm- 2, i.e., ca. 300 JiF cm- 2.
This is of course at least ten times the specific double-layer capacitance expected at a metal, or oxide film/solution interface and confirms the involvement
of an appreciable Faradaic (redox) pseudocapacitance. However, it has been
suggested by Zheng and JOW 16,17 that this redox capacitance arises principally
at the surface sites on the oxide film. By means of the quartz microbalance technique, Raistrick et al. I5 ,40-44 showed that the redox process involves only proton
exchange (presumably at vicinal OH--0 2- ions) in the surface.
280
Chapter 11
2~
.
70
N
IE
u
E 60
"cr
o~~
.0
50
0
0.05
0 .10
S"2/ lmV
0 .15
5-'),,2
FIGURE 11.9. Voltammetric charge, q, for redox processes at RU02 as a function ofreciprocal
square root of sweep rate, s. (I) aq. HCI04 and (2) aq. KOH. Reprinted from S. Ardizzone, G.
Fregonara, and S. Trasatti, Electrochim Acta, 35, 263 (1990), with permission from Elsevier Science.
01.10)
281
In the careful chronocoulometric work by Doblhofer et al.,34 further evidence for a distinction between surface and bulk processes at inner and outer regions in the charging of RU02 was obtained. Oxidation states between 2 and up
to 6 in the surface region were indicated. Also, the related role of diffusion processes in the charging was recognized.
A typical dependence of the voltammetric charge, q, on sweep rate for
RU02 in aqueous HCI04 and KOH, found by Ardizzone, Fregonara, and
Trasatti45 , is shown in Fig. 11.9. These data plot out linearly in a reciprocal
square-root of the sweep rate, i.e., a relation to the square-root of time. This
strongly confirms the role of the diffusion of protons coupled with the Faradaic
oxidation-state charge as potential is linearly scanned in time.
The diffusion-controlled proton transport is assumed to be along hydrated
surface structures containing OH groups in the porous oxide structure. The
mechanism of proton diffusion is assumed to be of the Grotthus proton hopping
kind between OH- and 0 2- groups, or between such centers and structural water
in the hydrous surface regions. The Ru02 grows homogeneously porous with
the loading employed and this seems also to be the case with electrolytically
grown films. It should be noted that in a high-area, rather porous material such
as RU02, there is more likely to be a continuous range of accessibilities of regions to proton transfer and redox oxidation-state changes in a way analogous
to the dispersion of double-layer charging in a porous carbon double-layer capacitor owing to a progressive voltage drop down pores.
Another interesting result in this work by Trasatti et al. was the observation
of a greater dependence of voltammetric charge on Ru02 film loading (on a Ti
substrate electrode) when the charge was evaluated in aqueous KOH than in
aqueous HCI0 4. This was attributed to higher oxidation states (from RuII to RuVII) being accessible in alkali compared with only Run to Ru VI in acids. Note that
this supposition is contrary to that assumed in other work discussed earlier,
where only Ru II to Ru III or possibly Ru lV redox transitions were envisaged, but
Ru VI was indicated in the work in Ref. 34.
In terms of a facile proton-transfer process being involved in a way coupled with oxidation-state changes at RuZ+ sites, it seems that the following step,
indicated earlier, has to be involved:
(11.7)
with each of the ions occupying lattice sites coordinating Ru z+ sites. In addition,
intralattice ionization of hydrate water (hence the importance of the ruthenium
oxide being hydrous in its bulk and/or at its surface regions) could provide additional OH- (or 0 2-) sites through such steps as
H 20
H+(lattice) + OH-(lattice)
(11.11a)
282
Chapter 11
(l1.11b)
if oxidation states of Ru higher than +IV were to be generated at elevated electrode potentials. In such cases, however, less reversibility might be expected
since new coordination arrangements, due to the requirements of an increased
coordination number charge balance around Ru z+, would be expected to arise
with accompanying phase changes involving heavy (OH- or 0 2-) particles (i.e.,
not just H+). Such phase-change, coordination rearrangements are usually far
from being reversible.
The extra charge accommodation in voltammetry in alkaline solution over
that in acid was about 25%. The charge accommodation was also sensitive to
temperature of calcination, being maximal at about 100 mC cm-2 at a temperature of ca. 320C. The sensitivity of the properties of RU02 to firing temperature
has been known for some time and was also systematically studied in earlier
work by Trasatti and more recently by Zheng and Jow 1617 in their preparation
of Ru02xH20 hydrous materials for electrochemical capacitors based on redox
charge storage.
283
284
Chapter 11
----
0.5
<t
E
......
0.1
-0.1
J:
......
<I
0.2
0.4
0.6
0.8
POTENTIAL / V vs SeE
FIGURE 11.10. Quartz crystal microbalance results for RU02 in 0.5 M aq. H2S04: upper curves,
the cyclic voltammogram at 50 mV s-1; lower curves, simultaneous frequency shift and corresponding weight change. Total capacitance 6.8 mF. (From Atanososki. 32 )
285
The figures of Raistrick for the real area per gram (130 m 2) and the capacitance of 380 F g-1 translate to a specific capacitance of 3 x 1061130 X 104 flF
cm- 2, i.e., about 300. This is about 10 times a probable double-layer capacitance
value, confirming the pseudocapacitative nature of the charging process involving surface redox processes plus a usual 10% double-layer contribution. For a
summary of thermal decomposition processes, see S. Trasatti in Electrodes of
Conductive Metal Oxides (Part B), S. Trasatti, ed. Elsevier North Holland, New
York (1981).
On dc discharging, the response of the RU02 system is typical of a true capacitance. Figure 11.11 shows a normalized plot of dc discharge behavior of
Ru02 in 1 M H 2S04 into parallel load resistances, R, having values from 100
kQ to 63 Q, i.e., for various rates of discharge. The experimental points in Fig.
11.11 fall on a common curve, indicating constancy of capacitance under these
100r-----------------------------------------,
R= 100 Kn (lI)
= 20Kn (0)
= 10 Kft (x)
= 5 Kit (0)
=250 n (0)
63
n (+)
>
".....
0::
10
10
100
tiRe
FIGURE 11.11. dc discharge kinetics of an RU02 film capacitance through six values of load resistance, R, as indicated, expressed as a plot of RI vs tiRe. (From Wojtowicz and Conway, unpublished.)
286
Chapter 11
..
.~16F
'
. ..
3.76 F
.-..
.'
.1
-1
I..L.
'.
-2
"-
<.!)
-3
..
....J
-4
.-11.7p.F
-5
-6
-4
-3
-2
-1
LOG f(Hz)
FIGURE 11.12. Dispersion of capacitance C (as log C) of a small RU02 electrode with frequency
in ac modulation over the range 10-3.4 Hz to 10 kHz. (From Wojtowicz and Conway, unpublished.)
dc conditions. The form of the plot derives from the discharge equation, namely
Vet) =exp[-t(RC)] + constant and V(t) =I(t) R, the time-dependent potential difference across C as it discharges at current I(t) into R. tlRC is a reduced time,
scaled by the time constant, RC.
The behavior under ac modulation exhibits a major dispersion of capacitance (Fig. 11.12), which is typical of porous electrode behavior. RU02 can also
be dispersed as high-area films, thermally formed on glass, glass particles, or
carbon. Their ac impedance behavior is very similar to that of the anodic (hydrous) film on metallic Ru (Fig. 11.2). It is seen that the accessible capacitance
of this particular small RU02 electrode falls from 16 F at 10-3.4 Hz to 11.7 f.1.F
(the double-layer capacitance) at 10 kHz.
287
0.60
0.30
<{
......
.....
QO.2
0.2
-0.30
-0.60
FIGURE 11.13. Cyclic voltammogram for an electrochemically formed and cycled oxide film on
W in aqueous 1 mol dm-3 H2S04. (From Conway and Birss, unpublished.)
288
Chapter 11
1 . 0 0 r - - - - - - - - - - - - - -- - - - - - - - - - ,
0 .50
<l
E
......
IZ
w
a::
a:: 0.00
:J
-0.50~---~---~------~---~---~
0.5
0 .75
1. 25
1. 5
1.75
POTENTIAL / V
FIGURE 11.14. Cyclic voltammograms for an electrochemically formed C0304 oxide film on Co
in aqueous KOH. (From Simpraga and Conway, unpublished.)
behavior arises on Mo. However, the operating ranges are only about 0.8 V, substantially less than that for RU02'
Other cheap and easily prepared materials that exhibit large specific
pseudocapacitance values are the family of conducting, electro active polymers
such as polyaniline, polypyrrole, poly thiophene, and their derivatives. Their behavior can be regarded as analogous to that of redox-type pseudocapacitors (see
Chapter 12). They represent a special class of materials and will be dealt with
separately in Chapter 12.
This chapter is concluded by reiterating the important and fundamental difference between pseudocapacitative, reversible discharging and recharging behavior [Fig. 11.15(a)] of RU02 and other oxides, and the irreversible behavior
in the cyclic voltammograms for a battery electrode system (e.g., Pb-PbCI 2), as
illustrated in Fig. 11.15(b), where the electrochemical half-cell reaction is
Pb + 2CI- ~ PbCl 2 + 2e. In the first type of system, the anodic and cathodic
voltammograms are almost mirror images of each other; in the second case
of a battery-type material, the anodic and cathodic voltammograms are
widely separated and involve substantial anodic and cathodic overvoltages
around a reversible potential at the crossover potential [Fig. 11.15(b)] for a
change of sign of currents from anodic to cathodic, or vice versa. The irreversibility in the case of the battery-type systems such as Pb-PbCI 2, Pb-PbS04 ,
and Mn02-Mn(OHh is due primarily to irreversibility in the phase changes involved and to the widely different Gibbs energies of the oxidized and reduced
forms of the couples involved.
289
u 12.5
B
~ ~
v
v;
c:(
6 .25
<1'1
(\J
'E
~-+--~~~~--+--+--~--~~--t--+--i--i--;----
c::(
......
::J
6.25
U
UB
o
J:
I-
12.5
<t
U
0 .2
0 .6
1.0
1.4
~ 66"r
Pb/PbCI 2
:;
~H
"II
..
or.,
POTENTIAL/V, RHE
,0)"
-0.2"
FIGURE 11.15, Contrast between cyclic voltammogram for (a) reversible redox processes at an
RU02 film electrode in 1 M H2S04 and (b) the totally irreversible processes of formation and reduction of PbCh on Pb where major changes in phase and chemical substance take place. (Diagram
repeated from Chapter 2.)
290
Chapter 11
linear region
Wo
,/
""
,.---
-- --
---------- Q
.g...
- - - L Crt/=::: constant ~
Z
~/
(,!)
~
C5:
<:{
u
c:t
:::c
u
POTENTIAL ( - +ve)
FIGURE 11.16. Illustrating overlap of three redox capacitance (C) responses required to give
almost constant net capacitance over an appreciable potential range, as observed with RU02. Q is
the accumulated charge (schematic). (B. V. Tilak, unpublished).
The capacitance behavior of Ru02 is remarkable in that the cyclic voltammogram approaches that for a capacitor having almost constant capacitance,
C, over a 1A-V range, albeit with some small but discernable maxima [Fig.
1 1. 15(a)]. For the electrochemically formed material, the capacitance is substantially a redox pseudocapacitance (see Chapter 10), so the almost constant
C must arise from the overlap of several redox processes on the potential scale,
as illustrated in Fig. 11.16. It would be difficult to account for the form of the
experimental current response profile in cyclic voltammetry through a singlestate redox reaction, even with a substantial interaction factor (see Chapter 10,
Fig. 10.1) analogous to that for an adsorption pseudocapacitance. Of course, in
parallel with the C contributions, probably up to 10% of the accumulated
charge would be due to double-layer charging on the extended hydrous oxide
surface, as discussed elsewhere in this chapter.
With Ru02, three successive redox stages can be recognized that involve
oxidation states Run, Rull , Ru lv , and probably also Ru VI. The voltammogram,
unlike those for oxidation and reduction processes at Pt or Au, does not involve
any stage of reduction to Ru o, the bare metal surface.
291
paper by Atanasoski. 32 Two topics were addressed: (1) the surface chemistry of
RU02, with emphasis on the oxidation state of Ru species and (2) the distribution of Ru at the pseudocapacitor material interface. From a combination of
electrochemical, photoelectrochemical, and ultrahigh vacuum (URV) surfaceanalysis studies, it was concluded that the electrochemically significant interaction between the oxides of Ru and Ti was confined to the interfacial region.
Ti0 2, as a rutile structure, functions as a support and stabilizing agent for the
RU02, which otherwise tends to disappear from the electrode structure when polarized (anodic corrosion) as RU02 and volatile RU04 (cf. OS04). This is the basis of the high stability of DSA RuOT Ti0 2 composite electrodes.
A significant aspect of this work was that single-crystal Ru02-Ti02 was
examined in comparison with thermally formed (as in the DSA electrode formation process) RUOT Ti0 2 films. The single-crystal work allowed the reconstructed surface of Ru02 to be compared with the unreconstructed (1 x 1)
surface and the hexagonal structure, in relation to surface content of atoms.
The surface content and Ti-Ru and O-(Ti + Ru) ratios were determined by
means of Auger spectra taken at the same spot on the surface; the data are shown
in Table 11.2.
Three possibilities have been proposed for the incorporation and distribution of RU02 thermally generated on an anodic surface oxide film of Ti0 2.32
Further literature on this direction of work on RU02 is found in Refs. 33 and 48
to 53.
The coordination and electronic band structure of RU02 (rutile geometry)
were reviewed and illustrated in Ref. 54 in connection with the use of RU02 as
the DSA anode electrocatalyst material for commercial Cl 2 production. Some
of the same factors (e.g., electronic conductivity) that make this oxide an excellent anode electrocatalyst also make it a very attractive redox pseudocapacitor,
e.g., through participation of several overlapping redox stages which lead to the
electronic conductivity. It is interesting that PbOb which is also an electronic
TABLE 11.2.
Temperature of
measurement (0C)
300
250
120
Ambient
Source: From Atanasoski 32
Ti
Ti
Ru
Ru+Ti
Ru
Ru+Ti
0.22
0.21
0.23
0.22
0.32
0.29
0.18
0.16
0.82
0.81
0.80
0.83
0.66
0.60
0.48
0.45
292
Chapter 11
conductor, does not exhibit redox pseudocapacitor behavior, presumably because a continuous mixture of oxidation states allowing electron and proton
hopping (in the hydrous oxide structure) is not possible with Pb0 2, which is directly reduced to a phase of a different structure, PbO or PbS04 .
The use of mixed oxides, that is with additions of Ti0 2, Ta205, and La203,
has formed an important aspect55 of the developing technology of RuOrbased
capacitor devices. In a recent paper by Tilak, Birss, and Chen,56 the electrochemical impedance behavior of such mixed oxides was examined over a range
of RU02 contents in RuOr Ti0 2 mixtures. The effects are dramatic, as illustrated
in Fig. 11.17, where the phase angle, q'J, vs. log [frequency] plots are shown for
Ru02 content that increases from 5 to 40%. The experiments were conducted in
5 M aqueous NaCl at 298 K at the open-circuit potential.
90r-----.-----.-----.-----.-----r---~
,
,,
,
,
60
,,
",
~--,
,/'" \
,/
,__
,/
,~---------
/'
/'
',50/0
"
.,., '10/, \
'"
\
15%' I
\
\\
\\
"
\1II
II
II
-----
\I
1\
"
~I
II
30
1\
"",,
o~----~----~----~----~----~--~
10-1
10 2
104
FREQUENCY/Hz
FIGURE 11.17. Frequency dependence of phase angle I/J in impedance spectra of fresh RU02 +
Ti02 composite electrodes having six different Ru contents by indicated percent at 298 K in 5.0 M
NaCI at the open-circuit potential. (From Tilak, Birss, and Chen, in press.)
293
At low Ru02 percentages (5, 10, and 15%), the </J vs. log [frequency] behavior corresponds substantially to the high-frequency double-layer response.
However, as more Ru02 is present (20 to 40%), the </J curves correspond to a
much lower frequency-response characteristic of the (redox) pseudocapacitance, with the </J dispersion curve rather suddenly moving up toward the pure
capacitance limit of 90 at 0.1 Hz. The complex-plane plots contain related information but are more complicated to discuss. These results of Tilak et al. show
that the added oxide components do not act just as diluents of the RU02, but intimately influence its pseudocapacitative behavior.
In conclusion, it is useful to compare and contrast the oxide film formation (or film oxidation) and reduction processes that take place at some of
the transition metals. This is conveniently shown by the series of comparative, semischematic, cyclic voltammograms displayed in Fig. 11.18 for Pt,
Rh, Ir, Ru, Au, Co, W, Mo, and Ni, some with oxide films either in thick- (multilayer) or thin-film (near monolayer) states. Attention is directed to the contrast
between the complete irreversibility of the oxidation and reduction processes
observable at Pt, Rh, Au, and Ni (even in thick-film states), and the excellent
reversibility observed with thick films on Ru, Ir, Co, W, and Mo. This latter behavior is manifested most clearly in the mirror-image oxidation and reduction
voltammograms, mUltiply recordable at such oxide film-bearing metals. As was
indicated in Chapter 10, reversible processes leading to pseudocapacitance require (1) chemically and structurally reversible oxidation and reduction reactions; (2) good electronic conductivity of the oxide film, and (3) facile electron
and proton hopping in hydrous oxide lattice structures, allowing proton intercalation and deintercalation. Here analogies exist to Li+ intercalation into transition metal oxide and sulfide electrode materials.
294
Chapter 11
...
....
----)
2.2 ~
+
.~~------------------------~
FIGURE 11.18. Comparative features of voltammograms for oxide formation, or oxidation, and
reduction for several transition metals with thin or thick oxide films; note cases of reversible vis
vis irreversible behavior (semischematic).
295
Nj (thick film)
.~~--------------------~~~
FIGURE 11.1S.
Ru (thick film)
.~~-----------------------4.
Continued
REFERENCES
1. 1. B. Goodenough, in Progress in Solid-State Chemistry, N. Reiss, ed., vol. 5, Pergamon, New
York (1971).
2. W. D. Ryden, A. W. Lawson, and C. C. Sartain, Phys. Letters, 26A, 209 (1968).
3. H. Schafer, G. Schneidereit, and W. Gerhardt, Zeit. Anorg. AUg. Chemie, 319, 372 (1963).
4. S. Trasatti and G. Lodi, Conductive Metal Oxides, S. Trasatti, ed., vol. A, p. 338, Elsevier,
Amsterdam (1980).
5. D. Galizzioli, F. Tantardini, and S. Trasatti, 1. Appl. Electrochemi., 4, 57 (1974).
6. S. Trasatti and G. Buzzanca, 1. Electroanal. Chem., 29, Appl. 1 (1971).
7. S. Hadzi-lordanov, H. A. Kozlowska, M. Vukovic, and B. E. Conway, 1. Electrochem. Soc.,
125, 1473 (1978).
296
Chapter 11
297
40. I. D. Raistrick, in Proc. First Inti. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational Seminars, Boca
Raton, Fla. (1991).
41. C. Ho, I. D. Raistrick, and R. A. Huggins, 1. Electrochem. Soc., 127, 343 (1980).
42. I. D. Raistrick and R. A. Huggins, Solid State Ionics, 7, 213 (1982).
43. I. D. Raistrick, Solid State Ionics, 9/10, 425 (1983).
44. I. D. Raistrick, Rev. Chim. Min., 21, 456 (1984).
45. S. Ardizzone, G. Fregonara, and S. Trasatti, Electrochim. Acta, 35, 263 (1990).
46. B. E. Conway, V. Birss, and J. Wojkowicz, 1. Power Sources, 66, 1 (1997).
47. S. L. Roberson, R. F. Davis, and D. Finello, in Proc. Sixth Inti. Seminar on Double-Layer
Capacitors and Similar Energy Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida
Educational Seminars, Boca Raton, Fla. (1996).
48. L. J. Atanasoska, W. E. O'Grady, R. T. Atanososki, and F. Pollak, Surface Sci., 202, 142
(1988).
49. L. J. Atanasoska, R. T. Atanasoski, F. Pollak, and W. E. O'Grady, Surface Sci., 2309, 95
(1990).
50. M. R. Kozlowski, P. S. Tyler, W. H. Smyrl, and R. T. Atanoasoski, Surface Sci., 194,505
(1988).
51. P. S. Tyler, M. R. Kozlowski, W. H. Smyrl, and R. T. Atanasoski, 1. Electroanal. Chem, 237,
295 (1987).
52. M. R. Kozlowski, P. S. Tyler, W. H. Smyrl, and R. T. Atanasoski, 1. Electrochem. Soc., 136,
442 (1989).
53. F. Hine, M. Yasuda, T. Noda, T. Yoshida, and J. Okuda, 1. Electrochem. Soc., 126, 1439
(1979).
54. B. E. Conway and B. V. Tilak, in Adv. in Catalysis, 38,122-135 (1992).
55. D. Craig, Canadian Pat. No. 1,196,683 (1985).
56. B. V. Tilak, V. Birss, and c.-P. Chen, Proc. Electrochemical Society (Grahame) Symposium,
97-17, 372 (1997), c. Korzeniewski and B. E. Conway, The Electrochemical Society,
Pennington, N.J. (1997).
57. L. S. Roblee, M. J. Mangaudia, E. D. Lapinsky, A. G. Kimball, and S. B. Brummer, Mat. Res.
Soc., Symp. Proc., 55, 287 (1986).
58. R. A-Blageni, A. G. Kimball, L. S. Roblee, G. Kahanda, and M. Tomkiewicz, 1. Electrochem.
Soc., 134, 2637 (1987).
59. X. Beebe and T. L. Rose, IEEE Trans. Biomed. Eng. BMK-35, 494 (1988).
60. X. Beebe, R. D. Raub, and T. L. Rose, Electrochim. Acta, 34, 795 (1988).
61. J. D. Klein, S. L. Clauson, and S. F. Cogan, 1. Vac. Sci. Technol., A7, 3043 (1989).
62. B. A-Blageni, M. M. Boucher, A. G. Kimball, and L. S. Roblee, 1. Mat. Res., 4, 440 (1989).
63. 1. D. Klein, S. L. Clauson, and S. F. Cogan,l. Mat. Res., 4,1505 (1989).
64. F. M. Kalliber and T. L. Rose, 1. Electrochem. Soc., 136, 1765 (1989).
Chapter 12
300
Chapter 12
mers has been likened to that by impurity dopants added to intrinsic semiconductors such as Si or Ge and the terms "p-doping" and "n-doping" are used to
describe the results of electrochemical oxidation or reduction, respectively.
These two terms do not seem, however, altogether appropriate because they do
not correspond to their respective significances in semiconductor science where
the doping arises from the actual addition of chemical electron-donor or electron-acceptor impurities which give rise to a new conduction band or valence
band of states and a diminished band gap. The electrochemical processes at the
conducting polymers are electrochemical redox reactions associated with sequential Lewis acid- or Lewis base-producing steps so that the polymer molecules are converted to multiply charged structures through electrochemical
Lewis-type reactions involving electron withdrawal or electron donation.
It is the possibility of introducing positive or negative charge centers on
electronically conjugated polymer chains by means of electrochemical oxidation or reduction reactions, the extents of which depend on electrode potential,
that has given rise to the substantial current and earlier interest lO- 14 in these materials for supercapacitor devices. The capacitance developed is substantial and
can be thought of as of the redox pseudocapacitance type (see Chapter 10) that
is associated with Faradaic redox processes. A coupled double-layer capacitance, probably about 2-5% of the specific pseudocapacitance, would be expected (Chapter 6) also to be observed, a point to be discussed further later. The
utilization of this pseudocapacitance associated with the redox properties of
such polymers in electrochemical capacitor devices is reviewed in more detail
below.
Conducting-polymer systems offer opportunities for storing electrochemical energy largely through their redox capacitance. Therefore, like Li intercalation systems or RU02, they store energy largely by Faradaic processes that
correspond formally to "battery-type" behavior but exhibit electrical characteristics which are those of a capacitor, i.e., there is a functional relation between
the charge (q) accommodated and the potential (V) of the electrode, giving a derivative dqldVthat corresponds to (pseudo) capacitance (Chapter 10). However,
as explained later in this Chapter, because of the metal-like properties of these
polymers in their conducting states, it can be argued formally that the capacitance developed is ofthe double-layer type. Because the charging and discharging reactions do not involve any appreciable structural alterations, such as phase
changes, the processes are highly reversible, as for proton intercalation into
RU02 or Li ions into some oxides and sulfides.
Such polymer materials are much cheaper than RU02 or Ir02 and can generate comparably large specific capacitances, F g-l. Being organic chemicals in
nature and subject to periodic oxidations (on charge) or reductions (on discharge), at least in most cases they do not have the very long-term stabilities or
cycle lives ofthe mixed-oxide type of electrochemical capacitor. However, very
301
substantial cycle lives have been achieved with not too much degradation, and
such performance is adequate for many applications. Also, work in recent years
(see later discussion) is revealing new polymer electrode materials that can be
cycled over substantially wider potential ranges than polyaniline or polypyrrole.
The possibility of employing electroactive conducting polymers such as
polypyrrole or polyaniline as electrochemical (pseudo)capacitor materials was
suggested in papers by Gottesfeld 11-14 and by Conway.1O These proposals were
predicated on the excellent reversibility of electrochemical charge and discharge which these materials exhibit over a potential range of 0.8 to 0.9 V, and
the associated mirror-image cyclic voltammograms which they also display and
which are an essential feature for capacitor-type electrical energy storage
(Chapter 10). Such materials are of course much cheaper than RU02, which also
gives mirror-image behavior, but are not as stable over very long-term cycling
(10 5_106 cycle life).
The essential features of the electrochemical behavior of conducting polymers that make them suitable as electrochemical capacitor materials are (1) the
continuous range of states of oxidation that arise with increasing electrode potential and (2) the reversibility of the Faradaic processes corresponding to
charge withdrawal and reinjection. The processes involved thus give rise in cyclic voltamrnetry to almost mirror-image voltamrnograms that are characteristic
of capacitor discharge and recharge behavior, and also to response currents that
increase with sweep rate.
The progressive dependence of charge on potential is analogous to that
originating in the titration of a polymeric acid or base [e.g., poly(methacrylic
acid)] as pH is increased (Br~nsted-type ionization due to transfer of protons),
giving rise to poly anions from polyacids or polycations from polybases. However, such polymers do not become electronically conducting on account of the
Br~nsted ionizations (in that case) induced by pH change. The monomer pKa is
about 4, but as the polyacid is progressively titrated with a base, the pKa continuously increases to about 10-11 (Fig. 12.1), i.e., the standard Gibbs energy
of the deprotonation of -COOH groups to give -COO- centers progressively
changes owing to interaction effects between the ionic charges on (side groups
of) the polymer chain, namely, -COO- groups. Analogously, with the conducting polymers, Lewis ionization by oxidation (transfer of electrons) causes a progressive change of standard redox potential (increase of Gibbs energy) owing
to lateral interactions between the generated positive (and delocalized) charge
centers, which makes further oxidation more difficult (higher electrode potential required) the more a given degree of oxidation is established.
In the case of polyelectrolytes derived from potential polyelectrolytes
[e.g., poly(methacrylic acid) or poly(vinylpyridine)], the ionization arises by a
Br~nsted deprotonation or protonation, respectively, i.e., by acid/base ionization processes involving protons. In strong bases or acids, the degree of ioniza-
302
Chapter 12
14
pH
Monomer
pK o
0.5
tion can be up to about one ionic center on the chain per monomer unit. With
the conducting polymers, charges are introduced on the polymer chains by a Lewis acid or base type of process formally involving oxidation or reduction with
electron transfer, the electrode surface behaving as the conjugate Lewis base or
acid. In both B[fmsted and Lewis types of ionization at polymer chains, high linear charge densities can be generated. However, such high charge densities cannot usually be sustained without some coupled charge-shielding process taking
place. In both cases this is achieved by counterion association with the chain
charges, leading to a Helmholtz type of double-layer configuration, as illustrated in Fig. 12.2.
In effect, both with Lewis ionization and Brpnsted protonic ionization, a
quasi-linear, one-dimensional cylindrical Helmholtzlike double layer of
charges becomes developed with the involvement of counterions of the electrolyte. In the case of the conducting polymer materials, it becomes somewhat difficult to distinguish whether the charging process is to be regarded as of the
I-dimensional double-layer (see Fig. 12.2) type or I-dimensional redox pseudocapacitance charging, but it is not just a semantic matter. Insofar as the charge
development can be regarded as having some of the characteristics of a Faradaic
reaction, the latter description of polymer capacitance behavior seems formally
preferable. An alternative view is that the distribution of redox potentials is related to a range of electrochemical Gibbs energies in the polymer for sites that
are candidates for being charged. However, the impedance spectroscopy of such
conducting, electroactive polymer matrices has features similar to those of the
303
Polymer
Backbone
(a)
ID
POOPING
..
POEOOPING
Current
Collector Neutral
Plate Conducting
Polymer
FUm
Current
Collector POoped
Plate Conducting
Polymer
Film
Solution
<tl Cations
9 Anions
9
Solution
(b)
NOOPING
..
N-DEDOPING
Currenl
Collector Neutral
Plale Conducting
Polymer
Film
Solution
@Callons
9Anlons
<tl
Current
Collector NOoped
Plate Conducting
Polymer
Film
Solution
304
Chapter 12
The discovery of conducting polymers, some of the earliest being polyacetylenes, 1-4 opened up a new interdisciplinary area of chemistry that lies between
electrochemistry and polymer science. The possibilities of various applications
were soon recognized5,14,16,17 in the fields of electronics, lithium batteries, and
recently for development of redox-pseudocapacitor devices in the electrochemical capacitor field (see later discussion). Recently, the properties and applications of polyaniline were discussed and usefully reviewed by Genies et a1. 9
A wide variety of physicochemical investigations on several conducting
polymers and their derivatives have been reported in recent years. Various spectroelectrochemical procedures have been used to investigate the properties of
305
polyaniline, including UV-visible spectroscopy,5,9,18-21 (FTIR) spectroscopy,22-24 and ellipsometry12,14; and direct electrochemical procedures such as
cyclic voltammetry,6 and ac impedance spectroscopylO with complementary
electrochemical quartz microbalance techniques, as recently employed in work
by Gottesfeld. 25
Pickup and co-workers have made extensive26- 32 physicochemical investigations coupled with electrochemical studies on conducting copolymers such
as conjugated bithiazole and bis-(thienyl) bithiazoles. Properties such as redox
potential, conductivity, solubility, and morphology were studied. An interesting
examination of the possibility of controlling the morphology of electroformed
poly(3-methylthiophene) at a rotating Pt disk electrode has been described. 29 At
a stationary electrode, precipitation dominates at long times, giving rise to porous, uneven films, but dense compact films are generated at relatively high
electrode rotation rates (-1600 rpm). This procedure may be useful for forming
electrochemical capacitor materials from such polymers.
In earlier work by Hillman and co-workers 33 it had been concluded that for
poly thiophene a monolayer film of polythiophene was first formed from the
monomer and on that film instantaneous nucleation and 3-dimensional growth
then occurred. The large number of overlapping and growing nuclei formed during the early stages of deposition (cf. Avrami's models) lead to subsequent 1dimensional growth perpendicular to the substrate electrode surface. On the
other hand, with a substituted thiophene (thiophene-3-acetic acid), polymer
films were formed by a 2-dimensional nucleation and growth mechanism, as
found in work by Li and Albery.34,35
In work by Zhao and Pickup,29 using a rotated disk electrode, impedance
plots in the complex plane showed behavior typical of a porous electrode (finite
transmission line). The extrapolated low-frequency resistance intercepts increased at higher rotation rates. Electron micrographs showed major differences
between the morphologies of films grown at a stationary electrode and those
grown at the rotated one; in the former case, fibrous (perhaps fractal) deposits
are easily distinguished. In Gottesfeld's electron micrography work,l1 photos
taken over a range of magnifications of derivatized poly thiophene films on a fibrous support showed clear evidence for self-similarity of structures, indicating
fractal growth.
In work with Mao,26,28 Pickup investigated the cyclic voltammetry redox
behavior of ferrocyanide electrostatically bound in polypyrrole-based, anionexchange, conducting polymers. Very reversible cyclic voltammograms for ferrocyanide oxidation and re-reduction were observed as first found by
Conway36,37 using ferrocyanide bound in adsorbed poly(vinylpyridinium) cation films at Pt and Au. These films gave rise to good redox pseudocapacitor behavior, but only over a 0.2-V range, as found also by Mao and Pickup.26,28 The
binding in the film converts the VOltammetry behavior from one characterized
306
Chapter 12
6
+
Note that it is the location of the radical (.) site that varies among the resonating
forms of the molecule; the positive charge is located on the aniline N center. The
generation of the free-radical center engenders polymerization.
Following that polymerization, and sometimes coupled with it, various
stages of deprotonation can arise, giving rise (Scheme 1) to recognizable and
characterized intermediates, known in classical organic chemistry; the names
are usually trivial ones, related to the colors of the intermediates. Corresponding
acid or base protonation or deprotonation processes can occur.
In the oxidation and reduction processes, transitions between insulating
and conducting states of the polymers can occur (see the reaction schemes on
following pages). A complex variety of molecular structures can be involved,
as shown in the formulas that follow. Some of the intermediates are highly col-
307
LE UCOEMERAl..D1NE
HN~~~
H
HNcu.o~
H
EMERAlJ)1NJ!
~"U.O~
H
ored, golden, green, etc., and the colors change during the progression ofpolymerization and oxidation. The electropolymerizations can be effected most conveniently in situ on an inert substrate electrode subjected to cyclic polarization
(cyclic voltammetry) over an appropriate potential range. The resulting polymer
films adhere to metal or carbon substrates and are reasonably stable.
It is outside the scope of this chapter to discuss in any detail the organic
chemistry of aniline polymerization and the associated series of complex oxidation (and reduction) steps that can take place. Also, the redox behavior is dependent on the protonation or deprotonation associated both with pH variation
and with oxidation or reduction, respectively. Anion binding (association) is
usually an important concomitant process accompanying oxidative polymerization and the oxidative half-cycle associated with charging, and leads to the formation of double layers.
However, we give in a few reaction schemes and chemical formulas some
of the essential changes that occur in aniline upon its polymerization and oxidation since they are important for understanding the processes of charge accommodation and the development of metallic conductivity. These are
308
Chapter 12
aqueous medium
strong acid
I:-@-@-IN+-H
-N+
X=H
H
NH2
$
X
aqueous medium
strong base
~CNlpyrldine
H
H
X-@-~-@-~
X-@-N .. N-@-X
SCHEME 3.
309
-@-~-@-~-
NH-NH
-@-~-@-a-
NH-~H
+
++
+
.
-@-~+-@-~-
NH2 -NH
-@-E+-@-~t-
NH-~H
-@-N==O=~-
==O=N==O=N-
N=N
N=N
SCHEME 4.
(.)
t-O-t:=o-iA-l
Salt Corm : S
AmiDe form: A
(b)
SahCcirm :s
Amine form : A
(cj
OxIdized form
Reduced form
SCHEMES.
310
Chapter 12
illustrated in the accompanying schematic formulas above and below. These reaction schemes have been discussed in much more detail in the original literature (see review Refs. 9 and 38).
It can be seen how a linear structure, capable of being successively oxidized (or reduced) down a polymer chain in one dimension, is analogous to the
2-, or 3-dimensional situation where successive oxidation or reduction stages
can occur in conducting hydrous oxides of transition metals (Ru02, Co02) or
progressively in quasi-2-dimensional intercalation processes involving Li, e.g.,
in TiS 2 , MoS 2, and Co02. In all these cases, large capacitances can arise as
pseudocapacitance, offering the potential for using such materials as supercapacitor electrodes.
The main points of interest, for the purposes of a supercapacitor material,
are (1) the possibility of a linear structure being developed with a sequence of
oxidation and re-reduction sites that could give rise to a redox process for which
charge accepted or delivered is a function of potential; and (2) the development
of a semimetallic conductivity along electronically conjugated double-bond
structures in a linear array in the oxidized state of the molecule (cf. Schemes 2
and 5).
Notwithstanding the complex series of states that the poly aniline structure
can exist in (see above), it is unclear at the moment why poly aniline gives
clearly three current-response peaks in cyclic voltammetry (Fig. 12.3) (formally
reminiscent of underpotential deposition of base metals on a noble metal where
appreciable pseudocapacitance arises but originates for different reasons) in the
oxidation and reduction cyclic voltammogram while o-anisidine (the o-OMesubstituted aniline) gives only one main peak (Fig. 12.4).
In practice, the electrochemical response of polyaniline is also quite dependent on the co-anion of the electrolyte, the behavior in HCI, H 2S04 , or
HCI0 4 being appreciably, indeed almost qualitatively, different. This can be
atttributed to different anion binding to the polymer system in its various oxidation states, which is analogous to the specificity of anion adsorption at electrodes of Au or Hg and their effects on H UPD at Pt, and also to counterion
binding at polyelectrolyte chains. Such anion binding at elementary units of the
poly aniline chain is illustrated in Scheme 6 and Fig. 12.2.
Scheme 1 shows five states of progressive oxidation (dehydrogenation at
the -NH- centers) of the so-called "leucoemeraldine" polymer (reduced form;
"leuco" = relatively uncolored). The completely dehydrogenated polymer (at
the Ns except for the end groups) is called "pernigraniline." Corresponding
changes in conductivity occur in the course of progressive oxidation. 39
A reaction mechanism for radical cation coupling of aniline, leading to the
formation of a linear polymer, is shown in Scheme 2 which indicates (cf. Refs.
16 and 39) the steps that are believed to occur by radical coupling, leading to
the formation of linear polymer structures. Other aromatic coupling steps are
311
AMINE FORM
'::fF
1 S INSULATING
NG
OX!t RED
2 A INSULATING
(TYPE I)
{N=O=N-aH+X-
H+
~ tN=O-~--c.
(TYPE II)
-'--N=O=N~)( ~ 1 N = O = N -~ \
"\
~n
\:
INSULATING IMINE
)(~n
+e-
'=O=N-o-'N
\
-
)
X
... ACID.
BASE
((==O=N-o-~=)\-t
2 S CONDUCTING
2 A INSULATING
ox!t
RED
((=O=N-o-N
3 A INSULATING
ox!t
RED
H~: ((=er-O-ry::"'l
3 S CONDUCTING
SCHEME 6.
312
Chapter 12
H
a)
Insulatlng-deprotonated
reduced tonn
nHw H+ solv
b)
w
n
Hsolv
H
Hsolv
(-o-J-o-J-t~ (-o-~+==O=~+-)n
protonated oxidized tonn
c:onctx:tor
first oxidation level
protonated tonn
InsUaior
Hsolv
c)
Hsolv
(--O-~'=O=~+-t~ (--Q-N:=O=N-r
2)(
d)
H+ solv
H+solv
H+solv
(-o-N ==O=N-r ~(-o-~+==O=~+-t
2)(
Hsolv
e)
H+solv
2)(
.2I
H H+solv H
(-o-~+==O=~'-)n~ (-o-~-o-~-t
2)(
.2T"
SCHEME 7.
Various chemical structures of poly aniline can occur (Scheme 4), including salt forms, as shown in Scheme 5 and 6.
The relationship between insulating and conducting stages of electroactive
conducting polymers is of great importance in redox supercapacitor devices.
Some examples for the polyaniline structural units were shown in Scheme 3 and
4. The role of pH (protonation, deprotonation) involvement and anion (A -) association in the transition between conducting and insulating states is shown in
Scheme 7.
The general behavior of the conducting polymers that have potential use
as capacitor materials is formally that of a redox pseudocapacitor (Chapter 10).
313
They exhibit the typical mirror-image type of cyclic voltammogram (Fig. 12.3)
that is a required aspect of reversible charging behavior over a range of potentials as exhibited by Ru02 and by regular double-layer-type capacitors.
At each potential along the voltammetric scan, a new equilibrium potential
is established that corresponds to some fraction (y) of the polymer chains having
been oxidized, leaving some remaining fraction (1 - y) unoxidized or in the reduced state (see Chapter 10). y represents a I-dimensional fractional extent of
oxidation of the polymer chain (or of full linear charge density) analogous
to '9\, the 3-dimensional state of oxidation in a bulk redox system, or to
the 2-dimensional state of occupancy of sites in an electrosorption process,
e.g., UPD of H on a metal surface lattice.
The electro-oxidation isotherm for charge accommodation as a function of
potential will, in the first analysis, be like a Langmuir isotherm or the exponential form of a Nernst equation. However, the substantial range of charge acceptance on the voltage polarization scale that arises (ca. 0.8-0.9 V or more)
indicates the strong role of intercharge repulsion or corresponding electronic interaction effects, so that potentials for charge formation on the polymer chains
depend on the charge density already resident on the chain, a behavior also characteristic of double-layer capacitance.
Most of the electroactive polymers can be generated in the conducting state
by chemical or electrochemical oxidation, which introduces positive charges
(so-called "p-doping") on the repeating units of the polymer chains. Conjugate
reduction returns the structures back to a less or nonconducting condition. Some
polymers, however (e.g., poly acetylenes, poly-o-phenylene, and polythiophenes), can also be rendered conducting by reduction, corresponding to socalled "n-doping."
In contact with a solution, the polymer matrix is perfused by the electrolyte
so that charge neutrality during charging or discharging is achieved by ion association with the charged chains, as with polyelectrolytes. Hence ions (anions
or cations) have to move in and out of the matrix upon charge or discharge. Correspondingly, changes of solvation and electrostriction around the polymer
chains must take place in concert with the charging or discharging processes, as
also occurs in polyelectrolyte ionization. Conformational changes of the polymer chain also take place to minimize its Gibbs energy, as with polyelectrolytes.
The ion distribution (ion association or ion binding), corresponding to the
requirement of electroneutrality was illustrated schematically in Fig. 12.2 in relation to polymer chain oxidation or reduction.
As with adsorption pseudocapacitance and other pseudocapacitance behavior, the key factors with the electroactive, conducting polymers are (1) that
the material itself has or acquires intrinsic electronic conductivity and (2) more
significantly, that the cyclic voltammograms for oxidation and re-reduction are
e,
314
Chapter 12
(almost) mirror images of each other over an appreciable range of electrode potentials, e.g., 0.8 V for PANI (polyaniline) in aqueous solutions.
In Chapter 10 it was explained that at each and any potential in the scanned
range, a reversible electron transfer (redox) process is established so that cycling
between the oxidized and the reduced states is facile and reversible. The capacitance charging current vs. potential profiles are differentials of the redox titration curve for the material and virtually no kinetic polarization (Chapter 3) is
involved except at high sweep rates (s > so; see Chapter 10) or at overcharge potentials.
As emphasized earlier, mirror-image cyclic voltammetry behavior is essential for a pseudocapacitor system to behave electrically like a regular capacitor or, more particularly, like a double-layer capacitance. This kind of
requirement is well demonstrated by the behavior of a variety of electroactive
conductive polymers, hence their potential value as cheap electrochemical capacitor materials, as recognized in the recent literature.l0.ll.13.14.36.40
315
ANILINE
IZ
a::
0:
POTENTIAL
:::>
FIGURE 12.3. Sequence of voltammograms for polyaniline oxidation and re-reduction in successive cycles as the film builds up on a gold electrode in aqueous H2S04. (From the author's laboratory with Gu Ping; cf. Gholamian. 6)
o-ANISIDINE
IZ
0: ~~----------------~~---------/~------0:
POTENTIAL
:::>
FIGURE 12.4. Voltammogram for a poly-o-anisidine film grown on a gold electrode in aqueous
H2S04. (From the author's laboratory, with Gu Ping.)
316
Chapter 12
Interesting results related to this question were found by D' Aprano and Leclerc,39 who showed voltammograms for poly(2-methylaniline) (PMA), poly(2methoxyaniline) (PMOA), poly(2-methoxy-5-methylaniline) (PMOMA), and
poly(2,5-dimethoxyaniline) (PDMOA) in 1 M aqueous ReI. These authors also
recorded the conductivity of films ofthese materials (Fig. 12.5) and found maxima at 0.42 V (vs. SeE) for PANI, PMA, and PABA; 0.35 V for PMOA and
PMOMA; and 0.28 V for PDMOA, the films being virtually nonconductive in
the fully reduced (leucoemeraldine) or fully oxidized (pernigraniline) states in
the case of PANI. They also examined the optical (electronic) spectra of these
PMOMA with the change in the oxidation state (Fig. 12.6).
In this work some attempt was made to explain the multiplicity of peaks
observed in the voltammetry of PANI; it was suggested that they may arise as
a result of some degradation products formed during the electrochemical polymerization and/or side-couplings (branching defects in the linear polymerization). The selectivity of head-to-tail (isosteric) couplings was diminished by
methoxy substitution on the ring. The range of appreciable anodic and cathodic
electroactivity ofthese polymers was about 0.5 to 0.7 V, which is similar to that
for PANI itself.
A different basis for explanation of the multiple current peaks observed
with PANI films was offered by Jiang, Zhang, and Xiang. 16 Peaks these authors
designated a and d at 0.13 V and -0.04 V (vs. SeE) may relate to protonation
and deprotonation processes (I, II) while they thought that peaks designated as
band cat 0.42 and 0.35 V may be associated with electron transfer coupled with
anion ion pairing with the charged chain (II, III). Some of the anodic and
cathodic separation between the respective peaks may be due to an iR drop in
the film and/or to some kinetic irreversibility.
Another and possibly more correct explanation is that the multiplicity of
the peaks for PANI arises simply from a stepwise oxidation to the first radical
cation and thence to the dication (bipolaron), based on electron-proton or on
electron-counterion coupling, depending on the acidity or the counterion' s mobility, i.e., its Stokes's law diameter. Alternatively, another probable reason is
that some component of artha coupling could lead to the minor peak(s) observed between the two main ones, as suggested by Genies et aI.9
An interesting and original variant was prepared and investigated by Gottesfeld et al. 14,42: poly-3-(4-fluorophenyl)-thiophene (PFPT) in an electrolyte of
tetramethylammonium trifluoromethanesulfonate (TMATFMS) in acetonitrile.
Structures of the PFPT (formula 3) and related nonfluorinated repeating (y)
groups (formulas 1 and 2) are shown below.
The cyclic voltammogram of PFPT (Fig. 12.7) is strikingly different from
that for PANI in that it exhibits two widely separated regions of almost reversible electroactivity that correspond to both positive and negative charge injection (coupled with ion association). The voltage range within each
317
0.006r-----------------~~----------_,
PMOMA
'E 0.004
u
If)
"~
0.002
PDMOA
0.3
IE
u 0.2
If)
"~
0.1
0.0
-0.2
0.6
FIGURE 12.5. Conductivities of PMOMA and PDOMA conducting polymer films as a function
of electrode potential. (From d' Aprano and Leclerc. 39 Reproduced by permission of The Electrochemical Society, Inc.)
1.0.-------...,-----.-----,------,------,
PMOMA
::J
......
w
u
.....
................
Z
<t
m
a::
o
(/)
<t
400
500
600
700
800
WAVELENGTH I nm
FIGURE 12.6. Changes in electronic s:wctra of a PMOMA conducting polymer film with oxidation state. (From d' Aprano and Leclerc. 9 Reproduced by permission of The Electrochemical Society, Inc.)
318
Chapter 12
electro activity region is about 0.4 V, but the two regions (peak-to-peak) are
separated by 2.6 V. In a two-electrode system, this allows one region to be
worked against the other with a wide redox pseudocapacitance charge exchange, i.e., with a capacitor operating over half of the 2.6-V range given above.
The current response behavior in cyclic voltarnmetry is better than that for
poly thiophene itself (formula 1) which also, however, exhibits two separate regions of positive or negative charge accommodation.
F
~
s
The forms of the voltammograms are reminiscent of those for the polycyclic hydrocarbon, rubrene, in acetonitrile, oxidized and reduced at a Pt electrode. However, no polymer film is formed in that case, but the neutral molecule
can be either reduced to a carbanion radical (-) or to a carbonium ion radical (+)
in a diffusion-controlled process. The anion and cation radicals recombine in the
diffusion layer with the emission of red light43 (electrochemiluminesence).
The behavior of the poly thiophene system is different from that of PAN!.
The latter can be oxidized to a p-type polycation (with associated co-anions) but
on re-reduction reverts to a relatively less conducting neutral state which limits
the power achievable with this material as a (pseudo )capacitor electrode. fn the
case especially of the fluorophenyl derivative, 3, of thiophene, the polymer exhibits both oxidizability and reducibility (n-doping) so there are two potential
ranges of electroactivity (Fig. 12.7). Thus in the fully charged state of a twoelectrode capacitor made with this material, a total voltage drop of about 3.0 V
across the cell can be achieved. This corresponds to maximum positive charge
injection into the positive electrode and maximum negative charge into the
negative plate. Corresponding co-anion and co-cation insertion adjustments into
the polymer interphases take place to achieve local charge balance as in the
charging of the double layer at a metal/electrolyte interface. Thus a
quasi-I-dimensional double layer tends to be set up locally along the
polymer chains, though the latter are not linear for en tropic reasons, as
with any polymer in solution except stretched rubber.
It should be mentioned that the discharge of such a capacitor device based
on polythiophenes means that the excess electron (and counterion) charge on
each plate is brought down to zero, but in this case a potential difference of about
2.0 V remains between the plates (Fig. 12.7) rather than zero V, as would be the
319
-,- -
-'-'
60
- - - Me4
NCF3 SO"3
- -- - - BU 4 NPF6
40
zo
<l:
'-
r-
0: -20
0:
::>
U -40
25 mV/s
-6.0
-z.O
-2 .5
-1 .5
- 1.0
-0.5
0 .5
1.0
POTENTIAL/V vs Ag/Ag+
FIGURE 12.7. Cyclic voltammograms forpoly-(3-fluorophenyl)thiophene in two electrolytes indicated showing two potential ranges of electro activity. Reprinted from Rudge et aI., 1. Power
Sources, 47,89 (1994), with permission from Elsevier Science; and Ren et a1. 42)
case with a PANI Type I capacitor (Fig. 12.8 and section 12.5). This is because,
with the poly thiophene material, high charge capacity with reversible charge injection is associated with only about 0.5 V at each of the extreme ends of the
negative and positive potentials of the overall 3.0-V range (Fig. 12.7) of the vol-
tammogram of the active material. This is an advantage in that all of the charge
(b)
(a)
~Vl-:
FIGURE 12.8. (a) Schematic representation of half-cycle charge and discharge operation of a
Type I combination of polymer electrodes making an electrochemical capacitor, (b) The decline of
potential, V, with time or charge, Q, on discharge. Reprinted from Rudge et aI., 1. Power Sources,
47, 89 (1994), with permission from Elsevier Science.
320
Chapter 12
Capacitor systems utilizing electroactive conducting polymers were classified into three types in a paper by Rudge et al. 14 : in a Type I electrochemical
capacitor, both electrodes are polymers of the p-dopable type, i.e., with oxidation leading to positively charged polymer chains. In the fully charged state, one
electrode will be in the fully p-doped (positive) state and the other in the uncharged state; a potential difference of 0.8-1.0 V will then normally become established between the electrodes, with half this range being available for
capacitor charge and discharge. Thus redistribution of charge can take place on
discharge down to half the initially established potential difference between the
two electrodes, as with a double-layer capacitor and most other embodiments. IO,40 Their behavior is illustrated schematically in Fig. 12.8.
In a Type II capacitor, two different p-dopable polymers are used that have
different ranges of potentials for oxidation and re-reduction. Their behavior in
a capacitor on discharge or recharge half-cycles is illustrated42 in Fig. 12.8.
Finally, in a Type III capacitor, the F-functionalized type of poly thiophene
polymer is used which can be both p- and n-doped in the same molecule, e.g.,
the poly-3-(fluorophenyl)-thiophene, the cyclic voltammogram for which 14 was
shown in Fig. 12.7. Then the discharging half-cycle of the p-doped electrode can
be worked against the discharge half-cycle of the n-doped electrode, but with
the substantial operating voltage difference Vo (Fig. 12.10) between the potential ranges for the two processes in the same molecule.
(b)
(a)
""'-I----V2- ---:
I
--'""'----::-t-+----;-.. V
,.I Q 2 ,,~-):
,,'
v
&t
L.----~--7--
01
0 &t
FIGURE 12.9. As in Fig. 12.8 but for the Type II capacitor electrodes. Reprinted from Rudge et
a!., 1. Power Sources, 47, 89 (1994), with permission from Elsevier Science.
321
(b)
(a)
' .~--------- V3 ----------~~:
~
:--VO~
i I
Va
3 ~ __I
\
\
Va
I
'--________________~i .. 0 & t
Oa
'''I .. '
FIGURE 12.10. As in Figs, 12,8 and 12,9, but for the two-region, Type III electroactivity shown
in Fig, 12.7. Reprinted from Rudge et aI., 1. Power Sources, 47,89 (1994), with permission from
Elsevier Science.
TABLE 12.1. Charge and Energy Density Obtained in Experiments on the Three
Schemes for the Application of Conducting Polymers to Electrochemical Capacitors
Capacitor
typea
II
III
Charge density
Energy density
Voltage (V)
(C/cm 2)b
(C/g)c
(J/cm2)b
(J/g)c
(Wh/kg)c
1.0
1.5
3,1
1.2
2.2
1.3
86
120
52
0,56
1.9
3.5
41
100
140
11
27
39
Source: Reprinted from Rudge et aI., J. Power Sources, 47, 89 (1994), with permission from Elsevier Science.
"Type I: polypyrrole; Type II: polypyrrole end poly thiophene; Type III: functionalized poly thiophene:
poly-3.(4-f1uorophenyl) thiophene.
bCharge and energy densities calculated per geometric square centimeter of the carbon paper electrodes,
<Charge and energy densities calculated per gram of active material on both electrodes in the capacitor
configuration. This does not therefore include the mass of the carbon paper support material.
322
Chapter 12
terials-design viewpoint and from the stored energy density that is achievable
with the higher operating voltage that is obtained. Some performance data obtained in this work for the three types of polymer capacitor systems are shown
in Table 12.1.
Most of the electroactive polymer systems exhibit a nonconstant pseudocapacitance, namely, a current-response curve in cyclic voltarnmetry that exhibits a maximum or several maxima. This behavior is in contrast to that of Ru02'
the voltammogram for which approaches the ideal rectangular shape expected
for a regular capacitor, i.e., it exhibits almost constant pseudocapacitance over
a 1.4-V range, as illustrated earlier (Chapter 11).
12.6. COMPLEMENTARY STUDIES USING OTHER PROCEDURES
The cyclic voltammetry studies of poly thiophene derivatives by Rudge et
al. 14 were complemented by the use of several other procedures: scanning electron micrography of polymer films grown on carbon-fiber substrates (Fig.
12.11), as shown in Fig. 12.12. In earlier studies of conducting polymer films,
FIGURE 12.11. Scanning electron micrograph of fibrous carbon paper matrix for growth of conducting polymer films. Reprinted from Rudge et ai., 1. Power Sources. 47.89 (1994). with permission from Elsevier Science.
323
FIGURE 12.12. (a) Scanning electron micrograph of polyfluorophenylthiophene grown on fibrous carbon paper (Fig. 12.11) to a total charge equivalent of 10 C cm-2. (From Rudge et a1. 14) (b)
As in (a) but for a sleeve of polymer grown around an individual fibril to a total charge equivalent
of 10 C cm- 2. Reprinted from Rudge et aI., 1. Power Sources, 47, 89 (1994), with permission from
Elsevier Science.
324
Chapter 12
~~
O.52V
_ _----~OV
-4~--~----~----~--------~--~
FIGURE 12.13. Simulated (upper curves) and experimental (lower curves) relations between the
and
for conversion of a polyaniline film on Pt in 2 M Hel at a 50ellipsometric parameters
mV s-I scan rate at 500 nm and a 60 incidence angle, between 0 and 0.52 V. The simulated curves
correspond to three conversion modes A, B. and C. (From Gottesfeld. Redondo. and Feldburg. 12
Reproduced by permission of The Electrochemical Society. Inc.)
".!l "'I'
325
tive index (n) and optical absorption coefficient (K) of the film, usually by
means of computer fitting or simulation procedures. Since the electrochemically formed polymer films are usually nonuniform, theoretical difficulties arise
in making a rigorous analysis of the optical data. However, the progressive conversion of the films on potential cycling can be easily followed, as was shown
in the detailed work by Bard et al.,47,48 represented in the plots of Fig. 12.14.
This investigation of propagation directions (during film conversion) by means
of ellipsometry was first described for conducting polymers in Ref. 12 (Fig.
12.13).
In the EQCM method, the resonant frequency of an electrically driven, longitudinally vibrating, piezoelectric quartz crystal (covered by a thin, conducting
metal film of Pt or Au) is measured, and the changes in frequency (lowering)
due to development of a film, or to adsorption of ions or molecules, are recorded. This system detects the nanogram sensitivities at weight changes caused
by film formation, so that growth of very thin films can be followed 14,25 in situ
by means of cyclic voltammetry.
The SEM pictures of the polymer films (Fig. 12.12) showed growth of the
polymer around individual fibers of a carbon fiber substrate, and microscopic
inspection at increasing levels of magnification revealed self-similar structures
of the film down to very small dimensions, i.e., the film growth was essentially
fractal, at least in a qualitative way. Because of this nonuniformity of the polymer deposits, theoretical difficulties can arise in interpreting the ellipsometric
parameters experimentally evaluated, as Dignam49 has shown for the effects of
roughness of surfaces.
In the case where the polymer films are of interest for electrochemical capacitor development, the use of EQCM and ellipsometry techniques by Rudge
et al. 14 provides an important in situ way of following the growth of the films
-OXIDIZED
v-REDUCEDINEUTRAL)
40
1Jt0
100
30
........... ,
20
0
10
0
60
80
100
120
f:j,0
140
160
180
FIGURE 12.14. Overlying ellipsometry plots for continuing growth of polyvinylferrocene films.
(From Carlin, Kepley, and Bard. 47 Reproduced by permission of The Electrochemical Society, Inc.)
326
Chapter 12
327
significantly over the potential range for which polypyrrole is highly conducting (0.44 to 0.84 V, RHE). The space charge capacity was constant from 0.74
to 0.54 V but became reduced as potential was lowered to 0.34 V. It appears that
as the potential is reduced toward the supposed redox potential (0.34 V, RHE)
of the film, the structure of the interphase begins to change from that of a mixed
metal semiconductor to that of a semiconductor.
Using the results for polypyrrole perchlorate, it was shown that a MottSchottky relation could be constructed by plotting the inverse square of the
space charge capacitance against the potential; this gave a flat-band potential of
-0.23 V (RHE). The slope of the plot was negative, as expected for a p-type
semiconductor.
The best fit of the impedance behavior to various equivalent circuits required a constant-phase element corresponding to a value of the a-exponent
[a = (D - It' where D is the fractal dimension of the roughness] equal to
0.53, which gave D = 2.89 for the fractal-represented surface. This corresponds to a material having an interface with a significantly 3-dimensionally
distributed surface, as concluded by Gottesfeld" from his SEM images (Fig.
12.12) of poly thiophene films on carbon fibers.
328
Chapter 12
propagation of the conversion process will be uniform within the film volume.
Taking r as a typical duration of electrochemical perturbation, d as a film thickness, and P as the dominant resistivity, the above conditions correspond to r >
pdC where C is the electrical capacitance per unit cross-sectional area of the
film associated with the charging process.
Mode B: This arises when p for electronic resistivity is much less than the
ionic resistivity. Then the film conversion process propagates as a front moving
from the film/electrolyte interface toward the substrate/film interface. The condition r < Pion dC then applies.
Mode C: This is for the condition when the ionic resistivity is negligible
compared with the electronic resistivity, i.e., for the condition r < Pelect de. Then
the propagation front moves in a direction opposite to that in mode B.
Mode D: A fourth mode arises for directional propagation of film conversion as in modes B or C, but with graded film oxidation or reduction along the
thickness direction as determined by diffusion from the leading interface into
the volume of the film.
Schematic representations of modes A, B, and D, as distinguished by Lee,
Kwok, and Bard,48 are shown in Fig. 12.15. Depending on which of modes A,
B, or D applies to the mechanism of conversion of films at electrodes, distinguishable types of ~ vs. If! plots of the ellipsometric parameters of the film will
arise, as shown by Clayton and De Smet. 52
Further details of the use of ellipsometry and of ECQM measurements on
electrogenerated polymer films are to be found in Refs. 12,25,44,46, and 48.
A detailed ellipsometric study of electrochemically formed polymer films
derived from bipyrazine, aniline, and vinylferrocene was described by Carlin,
Kepley, and Bard. 47 The in situ optical procedure allowed the growth of these
electroactive films to be followed in terms of If! or ~ plots, which are periodic
in the ellipsometry diagrams (Figs. 12.13 and 12.14) with repeated, almost overlying lines when the thickness d is some multiple of approximately half the
wavelength of the incident light. Difficulties arise,49 however, in the rigorous interpretation of the~, If! data, with porous, relatively thick layers that are nonuniform; however, see the last section of the article by Gottesfeld, Kim, and
Redondo. 46
Studies of the electrochemical behavior of conducting polymer films that
have potentialities for electrochemical capacitor devices of the redox pseudocapacitance type can be usefully complemented (see Section 12.6) by use of other
nonelectrochemical, in situ procedures. Rishpon et aI., with Gottesfeld25 have
utilized an original system for simultaneous ECQM and ellipsometry measurements as a means to obtain more detailed information on the physicochemical
and physical behavior of these interesting materials during their growth as films.
Ib in''" tJ
tsr
Bth
D
DO
329
arlier
part of
conversion
film thickness
iddle
port of
conversion
..
loter
port of
conversion
MODE A
MODEB
MODED
FIGURE 12.15. Schematic representation of modes 1, 2, and 4 for conducting polymer propagation (as defined by Lee, Kwok, and Bard48 ) during film growth.
A combined EQCM and ellipsometer setup was constructed25 which allowed simultaneous measurements of a weight change caused by the growth of
a polyaniline film and the changes in the ellipsometric parameters, Ll and'll, and
relative reflectivity that are associated with film growth. Some typical optical
and gravimetric records for film growth of PAN! over time are shown in Fig.
12.16.
Figure 12.17 shows experimental and simulated'll vs. Ll plots for galvanostatic growth of a PANI film at a current density of77 p,A cm- 2 (potential
0.50 V vs. SCE). The corresponding film growth, as measured by EQCM (in
terms of the decrease of oscillation frequency, Llf, of the quartz crystal of the
EQCM; Sowerby equation) was shown25 to scale linearly with the average optical thickness determined from the ellipsometry measurements, as seen in Fig.
12.18. The derived apparent film density was 1.45 g cm -3. Under potentiostatic
growth conditions, the apparent density was substantially less, 1.36 g cm- 3 . This
shows that the electrochemical conditions of film growth determine the physical
state of the grown films, probably reflecting different extents of solvent incorporation into the film during its growth.
Chapter 12
330
36
-200
31
::r:
<J
\fir
~r
'-
-400
110
26
21
tis
FIGURE 12.16. Simultaneously obtained optical (ellipsometric) and gravimetric (EQCM) results
for growth of poly aniline film. Reprinted from Rishpon et aI., J. Electroanal. Chern., 294, 73 (1990),
with permission from Elsevier Science.
~_ _ _'4..;;i,",,7o...-
44
SOLID CURVE:
A. 1.51 - 0.01 i
41
If
38
35
3~00
120
140
/)0
160
180
200
FIGURE 12.17. Experimental (points) and simulated (solid line) li vs. IjI plots for galvanostatic
film growth of poly aniline at a current-density of 77 flA cm-2 (potential = 0.50 V vs. SCE). Reprinted from Rishpon et aI., J. Electroanal. Chern., 294, 73 (1990), with permission from Elsevier
Science.
331
1Zoor------.-------r------,-------.---__~
./
960
720
Af/Hz
"0/
480
0~----~~~-----7.60~----~9~0------~12~O----~1~
OPTICAL THICKNESS; nm
FIGURE 12.18. EQCM results for polyaniline film growth scaling linearly with the ellipsometricall y determined film thickness values. Reprinted from Rishpon et aI., J. Electroanal. Chern., 294,
73 (1990), with permission from Elsevier Science.
332
Chapter 12
REFERENCES
1. R. B. Kaner and A. G. MacDiarmid, 1. Chem. Soc., Faraday Trans., 1,80,2109 (1984).
2. R. B. Kaner and A. G. MacDiarmid, Synth. Metals, 14, 3 (1986).
3. A. G. MacDiarmid, R. B. Kaner, R. 1. Mammone, and A. 1. Heeger, 1. Electrochem. Soc., 128,
1651 (1981).
4. K. Kaneto, M. Maxfield, D. P. Nairns, and A. G. MacDiarmid, 1. Chem Soc., Faraday Trans.,
1,78,3417 (1982).
5. T. Kobayashi, H. Yoneyama, and H. Tamura,l. Electroanal. Chem., 177,281 (1984); 177,
281 (1984) and 161, 419 (1984).
6. M. Ghoiamian, J. Sundaram, and A. Q. Contractor, Langmuir, 3, 741 (1987).
7. H. Letheby, 1. Chem. Soc . London, 15, 161 (1862).
8. J. Schmidt, Organic Chemistry, trans. by H. Gordon Rule, p. 782. Gurney and Jackson,
London (1943).
9. E. M. Genies, A. Boyle, M. Lapkowski, and C. Tsintavis, Synth. Metals, 36, 139 (1990); see
also 1. Electroanal. Chem., 195, 109 (1985).
10. B. E. Conway in Proc. 34th Intl. Power Sources Symposium, p. 319, Catalog no. 90CH 2863-9,
Institute of Electrical and Electronics Engineers. New York (1990).
11. S. Gottesfeld, in Proc. First Int!. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds. Florida Educational Seminars, Boca
Raton, Fla. (1991).
12. S. Gottesfeld. A. Redondo. and S. W. Feldberg. 1. Electrochem. Soc., 134, 271 (1987).
13. A. Rudge, I. Raistrick. S. Gottesfeld, and 1. P. Ferraris, Electrochem. Soc., 39, 273 (1994); and
Electrochim. Acta, 39, 273 (1994).
14. A. Rudge, J. Davey, I. Raistrick. and S. Gottesfeld, 1. Power Sources. 47,89 (1994).
15. J. O'M. Bockris and D. Miller, in Conducting Polymers: Special Applications, Proc.
Workshop, L. Alcacer, ed., D. Reidel, Boston, Mass. (1987).
16. Z. Jiang, X. Zhang, and Y. Xiang, 1. Electroanal. Chem., 351, 321 (1993).
17. Z. Jiang, Y. Xiang, and J. Wang, 1. Electroanal. Chem., 316,199 (1991).
18. D. E. Stilwell and S.-M. Park, 1. Electrochem. Soc., 135, 2491, 2497 (1988).
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
333
Y.-B. Shim, M.-S. Won, and S.-M. Park, 1. Electrochem. Soc., 137, 538 (1990).
C. D. Batich, H. A. Laitinen, and H. C. Zhou, 1. Electrochem. Soc., 137, 883 (1990).
H. Weiss, O. Pfefferkorn, G. Kotora, andB. D. Humphrey,!. Electrochem. Soc., 136,3711 (1989).
T. Ohsaka, Y. Ohnuki, N. Oyama, G. Katagiri, and K. Karnisako, 1. Electroanal. Chem., 161,
399 (1984).
M. A. Habib and S. P. Maheswari, 1. Electrochem. Soc., 136, 1050 (1989).
N. S. Sariciftci, H. Kuzmany, H. Neugebauer, and A. Neckel, 1. Chem. Phys., 92,4530
(1990).
J. Rishpon, A. Redondo, C. Derouin, and S. Gottesfeld, 1. Electroanal. Chem., 294, 73
(1990).
H. Mao and P. G. Pickup, 1. Am. Chem. Soc., 112,1776 (1990).
X. Ren and P. G. Pickup, 1. Phys. Chem., 97, 5356 (1993).
H. Mao and P. G. Pickup, 1. Chem. Soc., 1. Phys. Chem., 96, 5604 (1992).
Z. S. Zhao and P. G. Pickup, 1. Chem. Soc., Faraday Trans., 90, 3097 (1994).
X. Ren, S. K. Mandai, and P. G. Pickup, 1. Electroanal. Chem., 389, 115 (1995).
I. H. Jenkins and P. G. Pickup, Macromolecules, 26, 4450 (1993).
X. Ren and P. G. Pickup, 1. Electroanal. Chem., 396, 359 (1995).
A. R. Hillman and E. F. Mallen, 1. Electroanai. Chem., 220, 351 (1987).
F. B. Li and W. J. Albery, Langmuir, 8, 1645 (1992).
F. B. Li and W. 1. Albery, Electrochim. Acta, 37, 393, (1992).
B. E. Conway, in Proc. Third IntI. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., paper no. I, Florida Educational
Seminars, Boca Raton, Fla. (1993).
B. E. Conway and L. Duic, Report to Continental Group Inc., on supercapacitor research and
development, July (1978).
A. A. Syed and M. Dimesan, Talanta, 38, 815 (1991).
G. D' Aprano and M. Leclerc, 1. Electroanal. Chem., 351, 145 (1993).
B. E. Conway in Proc. Symposium on Electrochemical Capacitors, F. M. Delnick and
Tomkiewicz, eds., 95-29, 15, Electrochemical Society, Pennington, N.J. (1996).
Gu Ping and B. E. Conway, (in press) 1999.
X. Ren, J. Davey, A. Rudge, 1. Ferraris, and S. Gottesfeld, in Proc. Fourth IntI. Seminar on
Double-Layer Capacitors and Similar Energy Storage Devices, S. P. Wolsky and N. Marincic,
eds., Florida Educational Seminars, Boca Raton, Fla. (1994).
B. E. Conway and A. Pighin, 1. Electrochem. Soc., 122, 5 (1975).
B. E. Conway, in Techniques oj Electrochemistry, E. Yeager and A. 1. Salkind, eds., Chapter
5, Wiley, New York (1972).
M. Born and E. Wolf, Principles ojOptics, Macmillan, New York (1964).
S. Gottesfeld, Y-T. Kim, and A. Redondo, in Physical Electrochemistry, I. Rubinstein, ed.,
Chapter 9, Marcel Dekker, New York (1995).
C. M. Carlin, L. J. Kepley, and A. 1. Bard, 1. Electrochem. Soc., 132, 353 (1985).
48.
49.
50.
51.
52.
334
Chapter 12
54. D. Finello, D. Harrison, and R. K. Bunting, in Proc. Second Inti. Seminar on Double-Layer
Capacitors and Similar Energy Storage Devices, S. P. Wolsky and N. Matincic, eds., Florida
Educational Seminars, Boca Raton, Fla. (1992).
55. K. Sugimoto, as in Ref. 57.
Chapter 13
13.1. INTRODUCTION
336
Chapter 13
its power output capability. The second is through anion adsorption from the
electrolyte, determining in part the specific double-layer capacitance, especially
at potentials positive to the potential-of-zero charge of the carbon electrode material employed. The third is through the dielectric properties of the solvent,
which also determine the specific double-layer capacitance value and its dependence on electrode potential, as well as the extent of ionization or ion pairing
of the solute salt, which influences the conductance.
In Chapters 1 and 2 it was shown how two types of supercapacitors have
been developed in recent years: one based on the large double-layer capacitance
developed at high-area porous or powdered carbon matrices! invaded by a suitable electrolyte and the other based on utilization of the large pseudocapacitance2 ,3 associated with underpotential deposition, electrosorption processes
(e.g., H on Pt, Pb on Au), or with redox processes at high-area, porous conducting oxide materials [e.g., RuOz, Ir02 (in acid), or Co0 2 (in alkali)], as treated in
Chapter 10. Both types require high conductivity of their respective electrolytes
to deliver optimum power (see Chapter 15).
In order for highly reversible charge and discharge processes to take place
at high rates (giving rise to higher power operation), the internal electrolyte and
structural resistance of such porous, high-area materials must be minimized;
otherwise the principal desirable feature [expectation (2)] of capacitative fast
charging or discharging would be lost or seriously attenuated. Put in other terms
of impedance response, the phase angle of the impedance behavior should be as
near as possible to 90 0 at all frequencies-a rather idealized requirement in
practice, especially for porous electrode structures (see Chapter 14).
One of the principal factors determining the power spectrum (i.e., the
availability of power withdrawn over various time domains or in various frequency ranges) of porous electrode supercapacitors is therefore the resistivity of
the electrolyte, R e , in the pores and the extent of its physical contact with the
maximum accessible area of the pores of the matrix. In addition, minimum contact resistance, Reo between particles of the electrode structure is desirable. The
latter is a function of the materials used in the electrode. Diminution of contact
resistance can be achieved with compression of the matrix, but this usually diminishes the volume fraction of electrolyte, resulting in a relative increase of
Re. An optimized combination must be achieved.
In this chapter, the properties of solutions of electrolytes and their role in
the equivalent circuit of porous electrode supercapacitor systems is examined,
In addition to aqueous acid and alkaline solutions for low-voltage aqueous supercapacitors, some nonaqueous electrolyte systems that are available for capacitors operating at rather higher voltages (e.g., 3.5 to 4.0 V) will be
considered. For the latter purposes, some of the electrolyte solutions or solvents
available for Li battery technology are applicable, with substitution of Li salts
by tetraalkylarnmonium ones. We examine first some fundamental factors that
337
determine the conductance of electrolytes because this determines the esr of porous electrode systems used for capacitor devices.
The two principal factors involved in conductance are (1) the concentration of free charge carriers, cations, and anions of a gi ven salt or acid solute; and
(2) the ionic mobilities, or conductance contributions per ion, of the dissociated
ions of the electrolyte solute.
The factors that in turn determine (1) and (2) are (3) the solubility of the
salt in the solvent of choice; and (4) the degree of dissociation, a, of the dissolved salt "molecules" into free ions, or conversely, the extent of cation-anion
association or pairing of the ions of the dissolved salt or acid. Three stages of
ion pairing, depending on the extent of retention of solvation, are illustrated in
Figs. 13.l(a), 13.1(b), and 13.1(c).
(c)
FIGURE 13.1. Types of ion pairs in an electrolyte solution: (a) solvated-ion pairs, (b) solventshared ion pairs, and (c) contact ion pairs.
Chapter 13
338
MA
(1 - a)c
ac
ac
339
where 0 < a < 1 and activity coefficients have been taken as unity. For many
salts and acids in water, a ~ 1 (strong electrolytes) except toward the limit of
solubility of very soluble salts. However, it is important (in a negative way) that
for nonaqueous, high operating voltage (3.5-4.0 V) supercapacitors, the extents
of dissociation of many salts are substantially less than lowing to ion association (weak electrolytes). The association is often between partially solvated
ions, leading to a species referred to as "solvent-shared ion pairs" (see Fig.
13. 1(b).
A- (mS) M+ (nS)
(13.3)
where m and n can be less than the respective solvation numbers for the free individual solvated ions. 4 The total concentration of salt and dissociated ions depends on the solubility (see later discussion). When the species in Eq. (13.3) are
extensively formed, especially in low dielectric constant solvents, free ions are
withdrawn from their charge-conducting role, so both K and A decrease.
For weakly dissociated salts or acids in poor solvents, the A value is mainly
determined by K and a in Eq. (13.2) for process (13.1) or related equilibria. In
fact (according to Arrhenius), Ac is then determined almost entirely by the dissociation, and A/Ao = a where Ao (or A~) (> Ac) is the equivalent conductivity
at infinite dilution, lim(Klzc) as c ~ 0 (see Table 13.1 for aqueous Hel and acetic acids as examples).
Substitution of a in Eq. (13.2) by A/Ao leads to Ostwald's dilution law by
which the applicability of the Arrhenius relation for a in Eq. (13.2) can be tested
in various ways for weak electrolytes. Such plots give distinguishable relations
for weak and strong electrolytes (Fig. 13.2) where, for the latter, the Arrhenius
equation, a = A/Ao, does not apply.
For the solutions employed as electrolytes for nonaqueous solvent doublelayer capacitors, there is significant weak electrolyte behavior so that a is appreciably less than its value, which is near 1 in aqueous solutions. This usually
leads to larger esr values for nonaqueous solution devices than for aqueous ones
using the same electrode materials and cell geometries. Also the Ao values are
usually different.
In strongly dissociating solvents such as water, Kc is relatively large
and a ~ 1, so that the concentration of species (13.3) is small. However,
Ac is still found to progressively decrease with c in relation to Ao for infinite
dilution. This is due to the long-range electrostatic attractions between the
340
Chapter 13
Ao = 426.16
Ac
0.028408
0.081181
0.17743
0.31836
0.59146
0.75404
1.5768
1.8766
425.13
424.87
423.94
432.55
422.54
421.78
420.00
419.76
0.0116
0.02666
0.03355
0.05139
0.05995
0.07169
0.1059
0.1212
cI03
(mollliter)
0.028014
0.15321
1.02831
2.41400
5.91153
12.829
50.000
52.303
Ac
210.38
112.05
48.146
32.217
20.962
14.375
7.358
7.202
K(c) X
10
1.760
1.767
1.781
1.789
1.789
1.803
1.808
1.811
free,dissociated ions (it also occurs with the ions arising from weakly dissociated salts, but the effect on A is much less than that due to small K or Iowa),
which also depends on the dielectric constant, e, of the solvent. Up to moderate
concentrations, Ac decreases with the square root of c, an effect treated in the
well-known theory of Debye and Huckel (1923), and later by Onsager (1926).
The different effects of an increasing concentration of a dissolved salt or
acid on equivalent conductivity in the case of weakly dissociated (or ion paired)
electrolytes compared with strongly dissociated salts for which long-range interactions are more important are illustrated in Fig. 13.2 and are usually easily
distinguishable experimentally. Intermediate cases, of course, arise.
___ A
~""
,
, \,
,, '
,,
,
CD
FIGURE 13.2. Comparative conductivity behavior of a strongly and a weakly dissociated (or ionpaired) electrolyte as a function of concentration (schematic).
341
The electron-withdrawing CF3S02 functions bonded to OH, NH, or CH promote strong acidity of the Hs on those 0, N, or C groups. The latter two are referred to as imide- and methide-type ions. They have good conductivities as
salts in both aqueous and nonaqueous solvent media, and have also been employed as acid fuel cell electrolytes. In addition, low melting point alkyl pyridinium and alkyl imidazolium salts offer interesting possibilities as nonaqueous
electrolytes having good conductivities.
Some specific conductances at 22C of these and other salts of the imidazolium cations are listed in Table 13.2 together with their van der Waals volumes, Vc and Va' for the cations and anions. Data for aqueous 3.9 M H 2 S04 and
0.65 M tetraethylammonium-Bf4 in propylene carbonate are shown for comparison; they are mostly lower than the values for the latter two electrolytes. In
the organic molten salt series, 1m and Me represent the imide and methide coanions.
342
TABLE 13.2.
Chapter 13
Melt
EMI AICI4
EMIIm
EMIMe
DMPIAICI 4
DMPIIm
DMPIMe
3.9MH2S04
0.65 M TEA BF4fPC
15.0
7.0
1.3
7.1
2.5
0.5
575
10.6
118
118
118
152
152
152
113
144
206
113
114
206
TABLE 13.3.
Electrol yte
0.1 M KCI
3.0MH2S04
1.0 M TEA BFiPC
DMPI AICI4
EMllm
EMI 1m: C6H6 a
Source: From Koch et al. 5
a60:40 vol. %; 1m = (CF3 . S02):z~r.
2.65
2.50
2.57
2.35
3.43
3.43
15
10
7
5
12
12
24
21
15
3.4
26
22
343
1 M TEA BF4IPC
DMPI AIC14
EMllm
EMI 1m + C 6H 6 c
C dl a (int)C,uF cm-2)
Cdl bTheory
480
420
300
24
21
15
3.4
26
22
240
120
68
520
438
100
and/or associated poor wettability, i.e., unsuitable contact angles between the
molten electrolytes and elements of the carbon surface.
vIE = zeI6ny/ri(s)
(13.4 )
344
Chapter 13
345
(13.6)
When the (negative) energy U in the above equation is > or > > kT, the extent of ion pairing is substantial (a 1). Similarly, a usually increases with a
rise in temperature, leading to improved conductance. This of course is a common problem with supercapacitors and batteries, i.e., their internal resistance increases at lower temperatures and their power delivery capability (Chapter 15)
becomes substantially diminished.
In some cases (Fig. 13.1), when solvated ions are paired their solvation
shells become shared (a case of solvent-shared ion pairs), so that x <
r;,s,+ + r;,s,_. In poorly solvating solvents, the plus and minus ions come into a
contact situation [Fig. 13.1(c)] (contact ion pairs, Fuoss). Then, the maximum
plus/minus interaction arises electrostatically but is counterbalanced by a substantialloss of solvation energy of the two ions.
In these close-encounter situations, which lead to diminution of the conductance of an electrolyte, the normal value of B for the solvent becomes inapplicable and a lower local value due to dielectric saturation should be used for
the region between and near the ions. Thus ion-pairing equilibrium constants for
an electrolyte in various solvents do not follow simply a liB relation although
there is often a general qualitative trend. For example, three solvents having the
same nominal B (=20) (e.g., acetic anhydride, n-butyronitrile, and trimethylphosphate) can exhibit quite different ion-pairing effects for a given electrolyte
solute, with consequent substantial differences between the Ac vs. cor c I12 relations, the latter corresponding to the behavior of fully dissociated electrolytes
(Debye-Hiickel effect).
However, for the short-range interaction effects involved in ion pairing, it
is often the lone-pair electron donicity that is more significant than the B (Table
13.5). In the cases of the three solvents above, the respective donicity numbers
are quite different, namely, 10.5, 16.6, and 23 in a range of 0 (reference) to 38.8
for a variety of solvents (Table 13.5).
The ion pairing in various solvents can be quantitatively characterized by
infrared, Raman, and NMR spectroscopies, often in a specific way.4,II,I2 A summary of factors determining electrolyte conductance is given in Table 13.6.
Here the choice for supercapacitor electrolytes, for obvious reasons, is sulfuric acid or KOH for carbon-type double-layer capacitors using an aqueous
medium. However, the decomposition voltage limit is then theoretically 1.23 V
346
Chapter 13
TABLE 13.5.
Solvent
1,2-Dichloroethane (ref. solvent)
Sulfuryl chloride
Thionyl chloride
Acetyl chloride
Tetrachloroethylene carbonate
Benzyl chloride
Nitromethane (NM)
Dichloroethylene carbonate (DEC)
Nitrobenzene (NB)
Aceticanhydride (AA)
Phosphorus oxychloride
Benzonitrile (BN)
Selenium oxychloride
Acetonitrile (AN)
Sulfolane (TMS)
Propanediol-l,2-carbonate (PDe or PC)
Benzylcyanide (nitrile)
Ethylene sulfite (ES)
iso-Butyronitrile
Propionitrile
Ethylene carbonate (EC)
Phenylphosphonic difluoride
Methylacetate
y-Butryronitrile
Acetone
Ethyl acetate (ETAC)
Water
Phenylphosphonic dichloride
Diethyl ether
Tetrahydrofuran (THF)
Diphenylphosphonic chloride
Trimethylphosphate (TMP)
Tributylphosphate (TMP)
Dimethylformamide (DMF)
N,N-Dimethylacetamide (DMA)
Dimethylsulfoxide (DMSO)
N,N-Diethylformamide
N,N-Diethylacetamide
Pyridine (PY)
Hexamethylphosphoramide (HMPA)
Hydrazine
Ethylenediamine
Ethylamine
t-Butylamine
Ammonia
Trithylamine
Source: From Gutmann. lO Most e data are for 298 K.
DN
0.1
0.4
0.7
0.8
2.3
2.7
3.2
4.4
10.5
11.7
11.9
12.2
14.1
14.8
15.1
15.1
15.3
15.4
16.1
16.4
16.4
16.5
16.6
17.0
17.1
18.0
18.5
19.2
20.0
22.4
23.0
23.7
26.6
27.8
29.8
30.9
32.2
33.1
38.8
44
55
55.5
57.0
59.0
61.0
e
10.1
10.0
9.2
15.8
9.2
23.0
35.9
31.6
34.8
20.7
14.0
25.2
46.0
38.0
42.0
69.0
18.4
41.0
20.4
27.7
89.1
27.9
6.7
20.3
20.7
6.0
81.0
26.0
4.3
7.6
20.6
6.8
36.1
38.9
45.0
12.3
30.0
347
or practically, in kinetic terms, 1.3-1.4 V. As for battery systems (lead-acid, NiCd), relatively concentrated electrolytes are required to minimize the esr and
maximize power capability. However, strong solutions of acid are much more
corrosive than strong solutions ofKOH or NaOH, so the latter may be preferred
for some embodiments. These hydroxide electrolytes are very soluble in water
and, because of the OH- anion, have very good conductivities. Corrosion of the
hardware components is an important factor to consider in design and reliability, and with regard to the self-discharge characteristics of the capacitor device
(Chapter 18).
Both the acid and alkaline electrolytes have advantageously high equivalent conductivities in aqueous medium owing to the special mechanism of proton transport that determines their conductance (proton hopping).
13.6.2. Nonaqueous Media
348
Chapter 13
(a)
pore
Cdl
liquid
electrolyte
(b)
FIGURE 13.3. (a) Transmission-line model (Chapter 14) for a single pore. Re and Rs are the ohmic resistance of the electrolyte and the solid electrode material, respectively, is the electrode
Faradaic impedance and Cdl is the double-layer capacitance. (b) When RI = Rs = 0, the equivalent
circuit in (a) becomes simplified to that shown as (b). (See Chapters 14 and 16).
Z,
349
Materials
5.68 M HC1IH 2O
2.04 m MgCI21H2O
1.806 m MgC1 21H2O/
EtOH(20.3 wt.%)
1.547 m MgC1 21H2O/
EtOHl(39.8 wt.%)
1.362 m MgCl 2IEtOH
0.662 m LiCIOiPC
1.652 m LiCIOiPCI
DME(28.2 wt.%PC)
LiCVKCI eutectic
(mS/cm)
Remarks
849
160.3
69.3
Maximum
Maximum
Maximum
37.9
Maximum
23.0
5.420
14.38
Maximum
Maximum
Maximum
1615.0
(45rC)
EMIClAICl 3 X(A103) = .5
22.6
RbCu16I7CI13
RbAg4I5-single crystal
LiI
LiI(AI 20 3)
LiULi 2SIP2S5 (45/371l8)(wt.%)
LiCF3S03IPE0 6
LiCl4lPClPoly(vinylidene
fluoride)
ECIPClLiCI041PAN
(38/33/8/21 )(mol%)
I M LiCI0 4IPC Aerosil 200
340
260,270
0.01,0.042
10 to 0.1
I to 2
0.01
2
Cu+ conductor
Ag+ conductor
Li+ conductor
1.7
20C
5.8
22C
Li + conductor
60C
100C
350
Chapter 13
Electrolyte
Anodic limit
Cathodic limit
LiClO4
LiClO4
BU4NPF6
BU4NPF6
2.6 V
2.3 V
3.7V
3V
-3.5 V
-2.2V
-3.0
-3 V
Et4NPF6
-2.8
2.4 V
Working
electrode
Pt
Pt
Pt
Pt
Reference
Ag-AgCI0 4
Ag-Ag+
Ag-AgCI
Ag-AgCI
4. Intermediate dielectric constant aprotics such as acetonitrile, DMF, dimethyl acetamide (DMA) and butyrolactone (BL).
From various works it appears that the best nonaqueous electrolyte for
higher voltage electrochemical capacitors is Et4NBF4, 1 M, in propylene carbonate, possibly mixed with dimethoxyethane to improve (lower) its viscosity
and thus raise ionic mobilities. The above solution has a specific conductance
of 13 mS em-I.
13.6.3. Molten Electrolytes
The use of low melting-point organic salts as fused electrolytes for aluminum batteries using AICl4 or AI 2CI? salts of alkylimidazolium cations has
been explored in recent years. In an original development, Koch, Nandjundiah,
and Godlman l5 have employed these kinds of salts in fused electrolyte capacitors. With suitable anions, good anodic decomposition voltages have been
achieved, as was demonstrated comparatively in the anodic half-cycle voltammograms for Li salts of various anions (AsF6, PF6, Im-, Me-) in acetronitrile.
The voltammetry behavior of methylethylimidazolium (+) AICl 4 electrolyte in three melts (Lewis basic, Lewis acidic, and neutral mixtures) was studied
by Melton and Simpson. 16 Depending on composition, the operating ranges lie
between +2.4 and -2.2 V vs. an Al reference electrode in the melt. At a Pt electrode, AsF6 and PF6 imidazolium salts had anodic decomposition limits of ca.
5.1 and 5.2 V vs. Li.
A relatively new family of anions (also examined as the basis for Li battery
electrolytes and for aqueous solution electrolytes for fuel cells where O 2 reduction catalysis is improved), based on trifluoromethylsulfonic acid, has also been
explored for capacitor electrolytes. The family of structures is shown below.
The oxidation potential limits, vs. Li, for various complex anions and cations at glassy-carbon (gc), W, and Pt electrodes are listed in Table 13.8 below.
Trifluoromethylsulfonate
(triflate)
Perfluoromethylsulfonyl
imide
351
Perfluoromethylsulfonyl
methide
352
Chapter 13
1
+3
+2
+'
Aqueous
-,
-2
-3
\------111.11 H,SO,(Pt)
I - - - - - - - l i pH 7 Buffer IPII
p, {
~{
C{
rl- - - - - - - ( 1 1 . 1 1 KCllHgi
)--------lI'.M
HCIO,ICI
1
) - - - - - - - - - - 1 1 0,' .II KCIICI
Nonaqueous
MoeN
p,
1 - - - - - - - - - - - - - - - - - - - ; 1 0,' .II TBABF.
I - - - - - - - - - - - - - - - - - l i ~,~~l TBAP
\----------------11
Senzonitrile
O".If TBABF.
\ -_ _\---------------...,1
_ _ _ _ _ _ _ _ _ _ _ _---;1 PC
\--------------;1
1------------11
>-
"C
(J')
t:
W
Cl
I
I-
)-------------ll
-,
0,' AI TBAP
CH,CI,
O.lA.TEAP
0,',01 TBAP
~,~'.w KI
1
-3
-2
E V "', SCE
~----------.--~------------I
"F(1')
I
I
I
O~--------_+--_+------------Lr--------+_---I------
V-
::)
.,
+2
+3
~~'M TBAP
THF
....----- ,
i.(dl)-
~-----U
"C
.c
c;
1
:
V+ , / POTENTIAL
I
..
I FARADAIC
: DECOMPOSITION
~I------------~I
:
-FA-R-A-O--i.
....
I-C-- 1
OECOMPOSITION:
FIGURE 13.4. Estimated potential ranges (decomposition limits) for aqueous and nonaqueous
electrolyte solutions. (From A. 1. Bard and L. R. Faulkner, Electrochemical Methods, Wiley, New
York (1980). Reprinted by permission of John Wiley & Sons.) The lower curves illustrate decomposition limits, V+ and V-, in a cyclic yoltammetry experiment; i(dl) = double-layer charging currents and iF-Faradaic decomposition currents.
353
354
Chapter 13
Solvent/solution properties
Dielectric permittivity of the solvent
Donicity or electron acceptor numbers
Solvent viscosity
Solvent dielectric relaxation times
Mixture properties of solvents
Melting point of solvent or solution
Temperature coefficients of viscosity and dielectric
permittivity
Solvent vapor pressure, as a function of temperature
Electrochemical stability range
Liquid phase temperature range (difference between boiling
and freezing points)
li K
Zn Cd
355
AgCI Cu
- -+ +- - - - - - - - - - -1- -I -'It- -
Ag
"t ~ - 1" -
_~i~ ___________~~.sj~A~~
__ ~g_
__________
Li
K
--t- -t- -
~,n_~~u__~
-f!-
___ _
MCI
Ag
-------------t---+
AgCI
Ag
----------------~-~--+-2
-3
-1
POTENTIAL I V vs H2 /H+
DMSQ
SCHEME 1. Potential scales for several solvents, using the hydrogen electrode in the given solvent as the conventional reference electrode. The point of zero charge of mercury in the absence of
specific adsorption is marked as a cross. Reprinted from R. Parsons, Electrochim. Acta, 21, 681
(1976), with permission from Elsevier Science.
whatever solute ions are within it) is a quantity of prime importance because it
determines the theoretical and often practical voltage range over which the device can be operated. The practical operating range can usually exceed the thermodynamic decomposition limit if substantial overvoltages at the cathode (e.g.,
Hg) or anode (e.g., Pb0 2) arise in aqueous solutions, or for processes in
nonaqueous media.
Reduction potentials
-0.5
0.5
1.0
1.5
POTENTIAL I V vs H 2 /H+
SCHEME 2. Reduction potentials of redox systems in water versus nhe(aqueous) and in acetonitrile vs. normal H2 electrode (acetonitrile). Reprinted from A. 1. Parker, Electrochim. Acta, 21, 671
(1976), with permission from Elsevier Science.
356
Chapter 13
Ag Li
CI Rb K
Rb K
--+--t+
-?
CI
Rb K
Li Ag
-t----++ - - - - --++- CI
CI
-r- - - - - - - - - -+
Rb K
----1-1- -
CH 3CN
DMF
-1---
CI
HCONH 2
- -
Ag
Ag
-1-
DMSO
Li
_ -Sd____ _Zt_1~_
--+---H
Cd
Zn Cu
Cd
Zn
--+---+-+
---f- H
-1-----
---I----f--
H
- --1--
--I- -----I-I-
Cd
Zn Cu
Cd
Zn
Cu
--+----1----1-
--1---H
------1-
DMSO
----+-I
1\
12
a~ leV
18
20
22
al/eV
SCHEME 3. Real solvation energies of some common ions in various solvents. The lengths of
the arrows indicate schematically the probable contribution of the oriented solvent dipole at the surface to the real solvation energy of a monovalent cation ex potential effect). Reprinted from R.
Parsons, Electrochim. Acta, 21, 681 (1976), with permission from Elsevier Science.
357
tetraalkylammonium salts, often with PF6 or BF4 anions, are the preferred solutes for capacitors, while Li+ salts are often involved in high-energy rechargeable battery systems where deposited Li is the anode material, as metal or an
intercalate in carbon.
The donor properties of solvents as classified by Gutmann 10 (Table 13.5)
provide a basis for ordering of solvation energies, mainly of cations, among
nonaqueous solvents. The donicity scale also provides a measure of the strength
of interaction of various solvents with polar molecular solutes, e.g., CF3I where
the 19F NMR chemical shift is linearly related24 to the donicity number (DN)
(Fig. 13.5). An opposite trend is observed when the chemical shifts of 23Na+
(CI04") in various solvents are plotted against the DN (Fig. 13.6).
From an electrochemical direction, there are trends of polarographic halfwave potentials, E 1I2 , for Zn2+, Cd2+, and T1+ with solvent donicity as shown in Fig.
13.7. This plot reflects the dependence of the Gibbs energy of solvation of the cation
MZ+ in the solvents on the DN in the redox reaction MZ+ + ze + Hg ~ HgIM. Generally, the E1/2 values shown tend to increase with solvent donicity number. However, in this plot from the literature, it is unclear on what scale the E1I2 values have
been referenced. If a reference electrode had been used internally in the respective
solvents, then a solvent effect on the reference electrode potential would have been
15
DMF. DMA
12
TMPO
OMSO
a.
a.
"-
u..
a>
(,()
16
19 F
358
Chapter 13
12
......
8:
If)
flO
0
-3
FIGURE 13.6. Relation of Gutmann donor numbers lO to 23 Na+ (CI04) chemical shifts in various
solvents.
included. Alternatively, if an external reference had been used,25 then solventdependent, liquid-junction potentials would be included and their trends would
be difficult to predict.
In a reverse sense, a scale of acceptor numbers (AN) has also been proposed, as listed in Table 13.10. 26 Clear relationships are also found between
ANs and solute-solvent interaction behavior [e.g., with 13C=O chemical shifts
of acetone in various solvents and with the Gibbs energies of solvation of the
Cl- ion (relative to acetonitrile)] in a series of nonaqueous solvents and water.
The AN values are also related to values on other scales for Lewis acidbase interaction strengths, e.g., Kosower's Z-values 27 and the values of Dimroth
DMF
-0.3
>
......
DMSO
AC
SF
-0.1
0.9 L-__
____~__~____-L__~~__- L _
10
ON
20
30
FIGURE 13.7. Trends of polarographic half-wave potentials for Zn2+, Cd2+, and TI+ with solvent
donicity values.
359
ANa
Tetrahydrofurane (THF)
Diglyme
Dioxane
Dimethylacetamide (DMA)
Benzonitrile
Dimethy lformamide
Acetonitrile (AN)
Ethanol
Water
8.0
10.2
10.8
13.6
15.5
16.0
19.3
37.1
54.8
a AN
et a1. 28 The AN values, in relation to solvation effects in electrolytes for electrochemical capacitor applications, are relevant more to the states of anions in various solvents while the donor numbers apply more to the states of cations.
Generally, of course, a good donor number for cation solvation will be a poor
one for anions, and vice versa for acceptor numbers.
Herein lies the basis of optimization of solvent properties for maximization
of electrolyte solubility and dissociation, leading to good specific conductivity
and minimum esr in capacitors: mixtures of two or more solvents having respectively good AN and DN values, can be advantageously employed in developing
optimized electrolyte solutions for nonaqueous capacitor systems. Also, good
donor and good acceptor mixtures tend to provide widely miscible solutions in
binary and ternary solvent mixtures. A suitable choice of mixed solvents can
lead to maximized conductance of capacitor electrolytes and hence to desired
high-power performance that can be otherwise restricted by the internal resistance of the electrolyte.
The charge distribution in several dipolar aprotic solvents that determines
in part their DN or AN values has been illustrated by Parker23 as shown in the
diagram below, which indicates resonance canonical forms and resulting dipole
moments. The resonance aspect can be important in accounting for dipole moment values or magnitudes that are not always obvious from inspection of the
molecular formula. Thus, for example, CO might be expected superficially to
have a dipole moment similar to the )CO groups in ketones for which the dipole moment is quite large. However, the dipole moment of CO is very small
and it is an almost nonpolar gas. This is because two canonical forms,
0+
0-
0-
0+
C =OandC=O,
360
Chapter 13
)N
)N
) N-P
)N
++
)N-P-O
HMPA
)N
5.20
) S
++
)S-O:
-c
N:
++
-c
N:
C
)~
++
)N
)~
++
)N
OMSO
3.90
AN
3.40
C-S
OMF
3.90
SOMF
4.40
are comparably significant in the resonance hybrid, so a very low dipole moment results. In the diagram, the dipole moments are in Debyes, i.e., units of 10!8 esu (see Chapter 4), which is approximately the electronic charge, 4.80 x
10- 10 esu, multiplied by an effective charge separation distance of about 0.0205
nm (or 0.205 A).
Similarly, DN and AN values depend on resonance aspects of the electronic structure of the solvent molecule except in simple cases. Examples 23 are
illustrated below.
Here we return to the key practical matter; that is, the complex equivalent
circuit by which it is necessary to represent2930 the electrical response of porous
electrode supercapacitors and the associated distribution of Rand C elements.
Here the main concern is the distributed electrolytic resistance Re in the liquid
electrolyte, as shown in Fig. 13.3, which illustrates the equivalent circuit of a
lengthy pore (see Chapter 14) filled with a resistive electrolyte in contact with
an electronically conducting surface at which a double-layer capacitance Cdl or
a C + Cdl is exhibited. 3 ! As explained in Chapter 14, such a circuit has no single
or unique relaxation frequency w" or values of the time constant "ReC;" there
is a widely distributed range of Wr values or ReCs which corresponds to a power
spectrum on discharge or recharge.
361
This situation occurs because the inner regions of the pore (i.e., its surface
elements) are electrically accessed only through a progressively increasing Re
down the pore (ERe), i.e., there are progressively increasing IR drops. The double-layer capacitance is completely chargeable, or dischargeable, only under dc
or low-frequency time scales. Hence a dispersion of capacitance occurs. A detailed treatment of the electrical behavior of porous electrodes is given in Chapter 14.
An important but little-investigated effect in charging of high-area supercapacitor powder electrode materials occurs on account of the separation of cations and anions by their respective accumulation in the high-area double layers
at the cathode and the anode. The extents of separation and accumulation match
the charges developed on each side of the capacitor. The net charge accumulation is at least about 15 f.iC cm- 2, so that for a 1000 m 2g- 1 carbon matrix, the
net (univalent) ion accumulation within about 0.3 nm of the distributed surface
will be equivalent to 1000 x 104 x 15 X 1O-6 C g-l, i.e., 1.5 x 102/96,500 mol g-l
of carbon, "" 1.5 x 10-3 .
This is the quantity of electrolyte ions removed from the bulk electrolyte
invading the powder matrix upon completion of a charge. Depending on the volume fraction of solution in the matrix, this extent of removal can significantly
diminish the conductance of the electrolyte remaining in the pores of the electrode. For, say, a 2 M electrolyte (i.e., 2 x 10-3 mol cm-3) and a 66% porosity,
in the uncharged state there would be ca. 1.3 x 10-3 mol g-l of matrix. Thus the
charging process can appreciably deplete the concentration of free ions in the
bulk electrolyte in the pores, depending on the porosity ofthe matrix and the initial concentration of electrolyte in the discharged state of the electrodes. This
depletion then, of course, has a negative effect on the conductivity of the remaining electrolyte, so power performance is impaired and will depend on the
state of charge.
In charging electrochemical capacitors, there is not only a migration of cations and anions of the electrolyte to the surfaces of respective cathode and anode
matrices, which can deplete the bulk concentration of the internal electrolyte, as
explained above, but there can also be a significant electro-osmotic movement
of the whole solution on charge (depending on the electrode or cell design), or
in an opposite direction on discharge, that should be taken into account in assessing the solution factor in the performance of electrochemical double-layer
or other porous capacitor systems. The electro-osmotic movement depends on
(1) the sign and density of the surface charge on the porous matrix surface, (2)
362
Chapter 13
M;
~sub
II
Ai
(SUblimatio~t
+ A- (mS)
(solution)
M+(nS)
With acids, HA, there is a chemical proton transfer reaction with the solvent (here H 20):
HA + (m + n) H 20
363
Ways are available for separating the derived salt (M+, A -) values into individual ionic components,4,33 but they can have no fundamental thermodynamic basis. However, various empirical procedures are quite successful. 33
Calculations based on ion-solvent dipole electrostatic energies give values
in moderate agreement with the data derived from experiment. Approximately,
the ion-dipole energies gained when an ion enters a polar solvent balance the
energy that must be provided to break up the crystal lattice of strongly interacting ions in the solid phase and transfer them as free but solvated ions into the
solution bulk.
The equilibrium solubility of a salt is detennined by the standard Gibbs energy (~GO) of solution of the ions from the salt crystal. Negative ~GO values correspond to good solubility of a salt (i.e., to provision of a high concentration of
plus and minus ion charge carriers) and good conductance.
Thus the conductance of a salt, among other things, depends greatly on the
polarity or dielectric properties of the solvent. The latter also detennine what
fraction of the dissolved ions at a given concentration (below the solubility
limit) are plus or minus paired (factor 3), usually partially retaining their solvation shells of coordinated solvent molecules [Fig. 13.I(b)].
Table 13.11 gives some idea of ionic hydration energies for some simple
monovalent ions relative to the value for the proton. They are all large (negative)
values on the absolute scale. Di- or trivalent ions have much larger solvation energies, which are related to the square of the charge number. There have been
various attempts to make a priori calculations of individual ionic hydration (or
solvation) energies. The first procedure was carried out by Born in 192034 and
was based on the difference between the self-energy of charging (see Chapter
4) the ion in a solvent of dielectric constant s and that in a vacuum (B = 1). This
difference is a Gibbs energy given by
(13.7)
for an ion of charge Zie and radius rio Equation (13.7) gives the correct magnitude for I1G s for ions in water. Better estimates are obtained if calculations 35 ,36
are based on the electrostatic energy of coordination and orientation of solvent
dipoles around the ion, with some allowance for solvent structure modification4
in the region around the solvated ion. A critical examination of the various types
of ab initio calculations and corresponding models used is found in Ref. 14 and
in the relevant original papers.
Equation (13.7) relates the I1G s to the dielectric constant of the solvent, B,
in the term liB. From the nature of this relation it will be seen that it will not give
a good account of the specificity of ion-solvent interaction energies since for B
values above about 15-20 (i.e., for a variety of common solvents that are good
364
Chapter 13
ClBr-
r-
OW
CIO]
S2-
0.0
137.7
163.8
184.0
189.9
197.8
118.7
147.1
182.8
-366.3
-348.8
-340.7
330.0
-345.7
-427.6
-849.4
0.0
-2.4
5.1
13.6
16.5
17.2
-18.7
3.7
14.6
-63.1
-49.49
-45.79
40.32
0.0
138.4
162.3
179.9
185.0
192.7
124.3
146.0
178.5
-347.5
-334.0
-326.0
318.3
-83.1
824.6
Source: From Benjamin and Gold 21 Copyright 1954 Royal Society of Chemistry.
Notes: Absolute value of MIl for H+ aq. is -265 5 kcal mol- 1 and for!!S1 it is -5 I cal- I K- 1 mol-I. Absolute
values for other ions are then obtained by adding -265 to the above values for cations or subtracting it from the
values for anions.
media for dissolving salts), lie is a small quantity compared with 1, so that tl.G s
remains a relatively large negative quantity. In fact, it can be seen from Eq.
(13.7) that except for quite small values of e for nonpolar solvents, the magnitude of tl.G s is determined mainly by the loss of the ion's charging energy in a
vacuum (the Zie212ri term with e = 1) as the ion is transferred into the solvent
medium. The ion-dipole interaction mode1 3435 gives a better account of specificity for ion-solvent interactions through the individuality of solvent dipole
moments and donicities,1O as well as coordination numbers for solvent molecules around ions.
Since the solubility equilibrium constant (the solubility product Ksp) is
logarithmically related to AGO (AGo = -RTln Ksp ), relatively small differences
of AGO, associated with differences oflattice and solvation energies, make large
differences of Ksp; these can vary from 10- 13 to Wi for a variety of salts, and
also from one solvent to another. Hence, large differences of conductance can
arise, so that for battery or supercapacitor electrolytes either well-dissociated
acids or extensively soluble electrolytes must be used.
For nonaqueous solvents, solvation energies are usually smaller than for
water so solubilities and conductances are often less. Tetraalkylammonium salts
or salts of large anions such as Ph4B-, Ph4As-, PF6, AsF6, and BEl are often
reasonably soluble in nonaqueous solvents, although they are somewhat ion
365
paired. The differences between solvation energies of ions in nonaqueous solvents and in water are usually not well accounted for by Eq. (13.7) in terms of
their dielectric constants. Better comparisons can be obtained by using dipole
moments 35 36 and donor numbers, while taking into account steric factors and
the accessibility of the polar center to the ion, as noted earlier in this chapter.
366
Chapter 13
b.p.(C)
Acetonitrile
y-Butyrolactone
Dimethoxyethane
N,N-Dimethylformarnide
Hexamethyl-phosphorotriarnide
Propylene carbonate
Tetrahydrofuran
2-Methyltetra-hydrofuran
Dimethyl sulfoxide
Dimethyl sulfite
Sulfolane (tetra-methylenesulfone)
Nitromethane
Dioxolane
81.60
202
84
158
233
241
66
80
189
126
285
101.2
78
f.p.(C)
-45.7
-43
-58
-61
7.2
-49
-108.5
18.55
-141
28.86
-28.6
-95
d(g cm- 3)
0.771
1.125
0.859
0.944
1.02
1.19
0.88
0.855
1.096
1.207
1.262
1.131
1.06
,u(D)
37.5
39
7.20
36.7
29.75
64.4
7.58
7.3
46.6
22.5
43.3
35.94
7.13
3.44
4.12
3.86
5.38
4.94
1.75
367
Cl.
u
....
f:"
100%
PC
mOl %
100%
OME
..
I
e
Ne
u
en
.....
<;
100 "I.
PC
mol "I.
100 "I.
OME
368
Chapter 13
"0
DOL-THF
(/) 5
.....
-<
A3
3~--~----~----~--~
25
50
75
DOL concentration/mol %
100
CJ)
....
-<
2
'0
25
50
75
100
FIGURE 13.A4. As for LiBF4 but with 1 M LiCI04. (Data from Matsuda et al.7.8)
369
.(/)
5~-------------------'
DOL-DME
o 4 ~ _ _ ._. .....
E
E
:l
b
~
.JfI:.
(/)
t
o
A5
25
50
75
DOL concentration /mol
100
0/0
FIGURE 13.AS. Walden product (AI}) for DOL-DME and DOL-THF mixtures with 1 M LiBF4.
(Data from Matsuda et aI. 7,8)
15V\
.I
(5
to
Ne
0
(/)
......
<:
A6
0
25
50
75
100
DME or THF conelmol %
DMSO-DME;
370
Chapter 13
to~--------------~
/.~.
"J
; "j
'0
"'eu
VI
'-
<:
A7
DME concentrotion Imol %
DMSO-
80~--------------------'
100
371
15
'j
(5
10
t\I
(/)
.......
<:
A9
o-
40
20
60
r
FIGURE 13.A9. Correlation between A (LiCl04 I M) and Br of solvent: 0 PC-DME; t:, S-DME;
DMSO-DME. (Data from Matsuda et al.7,8)
1/1
0
Cl.
'0
E
.......
~
<:
A10
0
40
20
60
r
FIGURE 13.AIO. Correlation of Walden product AI} (l M LiCI04) with solvent er: 0 PC-DME;
DMSO-DME; t:, S-DME. (Data from Matsuda et aI. 7 ,8)
Chapter 13
372
3
0u
'-
All
OL----2~0~--~~--~----~--~1=OO
Moss % PC
GO
AsFs
CIO"..,
~.
Ace
BU4N+, BPh;
30
Li+
-x___..
xe~
CI04'
o AsF.-
BPh.Bu4~
10
...
A1:2.
50
"lope
o Lt
100
Pure PC
FIGURE 13.A12. Variation of mobility at infinite dilution, A~, S cm2 mol-I, for some cations
and anions plotted against % PC in PC-DMM mixtures.
373
THF
DMSO
DMS
SL
AN
DME
DMM
DOL
HMPA
PC
EC
NM
BL
DCM
13.12.4 Conductivities
Salt
298 K
273 K
248K
8.31
6.83
6.46
4.95
6.52
6.44
7.51
7.15
8.30
6.12
4.76
7.70
3.31
1.94
4.62
3.87
3.23
2.58
3.76
3.40
4.24
4.12
4.89
3.45
2.54
4.36
1.91
0.99
2.21
1.86
1.48
1.07
1.65
1.72
1.45
1.97
2.33
1.56
1.14
2.06
0.91
0.43
Source:
From Morimoto. 37
374
Chapter 13
REFERENCES
1. 1. Raistrick (review) in Electrochemistry of Semiconductors and Electronics, 1. McHardy and
F. Ludwig, eds., p. 297, Noyes Publ., N.1. (1993); see also 1. Currie in The Chemistry and
Physics of Composite Media, Electrochemical Society Proc., vol. 85-8, M. Tomkiewicz and
P. N. Sen, eds., Electrochemical Society, Pennington, N.J. (1985).
2. B. E. Conway, 1. Electrochem. Soc., 138, 1539 (1991).
3. B. E. Conway and E. Gileadi, Trans. Faraday Soc., 58, 2493 (1962).
4. B. E. Conway, Ionic Hydration in Chemistry and Biophysics, Elsevier, Amsterdam, (1981).
5. V. R. Koch, C. Nanjundiah, and 1. L. Godlman, in Proc. Fifth Inti. Seminar on Double-Layer
Capacitors and Similar Energy Storage Devices, S. Wolsky and N. Marincic, eds., Florida
Educational Seminars, Boca Raton, Fla. (19995).
6. V. N. Plakhotnik et al., Elektrokhimiya, 24, 964 (1988).
7. M. Morita, T. Fukumasa, M. Motoda and Y. Matsuda, J. Electrochem. Soc., 137, 3401 (1990).
8. M. Morita, M. Goto, and Y. Matsuda, Bull. Chem. Soc., Japan, 59,1967 (1986).
9. L. Werblan and A. Balkowska, J. Electroanal. Chem., 354, 25 (1993).
10. V. Gutmann, Chemistry in Britain, 7, (3), 102 (1971).
11. A. I. Popov, Pure Appl. Chem., (IUPAC), 41, 275 (1975).
12. B. W. Maxey and A. 1. Popov, J. Am. Chem. Soc., 89, 2230 (1967); 91, 20 (1969).
13. J. Barthel and H.-J. Gores, Pure Appl. Chem., (lUPAC), 67, 919 (1995).
14. A. 1. Bard and L. R. Faulkner, Electrochemical Methods, Wiley, New York (1980).
15. V. R. Koch and C. Nandjundiah, in Proc. Fourth Inti. Seminar on Double-Layer Capacitors
and Similar Energy Storage Devices, S. Wolsky and N. Marincic, eds., Florida Educational
Seminars, Boca Raton, Fla. (1994).
16. T.1. Melton and R. 1. Simpson, J. Electrochem. Soc., 137, 3865 (1990).
17. J. Robinson and R. Osteryoung, J. Am. Chem. Soc., 101, 323 (1979).
18. 1. Barthel and H.-I. Gores, in Chemistry of Nonaqueous Solutions: Current Progress, G.
Mamentov and A. 1. Popov, eds., Chapter 1, VCH Publ., New York (1994).
19. J. Barthel and H. Popp, J. Chem. Information Compo Sci., 31, 107 (1991).
20. B. E. Conway, in Ionic Hydration in Chemistry and Biophysics, Chapter 11, Elsevier,
Amsterdam (1981).
21. L. Benjamin and V. Gold, Trans. Faraday Soc., 50, 797 (1954).
22. R. Parsons, Electrochim. Acta, 21, 681 (1976).
23. A. 1. Parker, Electrochim. Acta, 21, 671 (1976).
24. R. H. Erlich, R. Roach, and A. 1. Popov, J. Am. Chem. Soc., 92, 4989 (1970).
25. D. P. Wilkinson and B. E. Conway, Electrochim. Acta, 38, 997 (1993).
26. U. Mayer, V. Gutmann, and W. Gerger, Mh. Chem., 103, 1360 (1972).
27. E. M. Kosower and E. P. Kleindinst, J. Am. Chem. Soc., 78, 3483; 5700 (1956) and 80,3253;
3267 (1958).
28. K. Dirnroth, C. Reichardt, T. Siepmann, and F. Bohlmann, Ann. Chem., 53, 661 (1963).
29. Third IntI. Seminar on Double-Layer Capacitors and Similar Energy Storage Devices, S.
Wolsky and N. Marincic, eds., Florida, Educational Seminars, Boca Raton, Fla (1993).
375
30. Second Inti. Seminar on Double-Layer Capacitors and Similar Energy Storage Devices, S.
Wolsky and N. Marincic, eds., Horida Educational Seminars, Boca Raton, Ha.
31. L. Bai, L. Gao, and B. E. Conway, 1. Chem. Soc., Faraday Trans., 89, 235, 243 (1993).
32. N. K. Adam, The Physics and Chemistry of Surfaces, Oxford University Press, Oxford (1941).
33. B. E. Conway, 1. Soln. Chem., 7, 721 (1978).
34. M. Born, Zeit. Physik, I. 45 (1920)
35. 1. D. Bernal and R. H. Fowler, 1. Chem. Phys., I, 515 (1933).
36. D. D. Eley and M. G. Evans, Trans. Faraday Soc., 34, 1093 (1938).
37. T. Morimoto, in Proc. First Inti. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, eds. S. Wolsky and N. Marincic, Horida Educational Seminars, Boca Raton,
Ha. (1991).
Chapter 14
14.1. INTRODUCTION
The point has often been made that electrochemistry is a two-dimensional science since its processes, including charging or discharging of double layers,
proceed on surfaces rather than in bulk. When the condition of2-dimensionality
is imposed on an electrode reaction (e.g., at a plane electrode interface), then a
coupled condition arises, namely, reagents must diffuse to and products diffuse
away from the electrode surface. Mechanical or hydrodynamic assistance to this
mass-transfer condition is often provided, (e.g., in electrode-synthetic processes), but the requirement of diffusion or migration to or from the electrode still
obtains. Even in charging the capacitance of the double layers at high-area electrodes, there is some reorganization of the distribution of cations and anions of
the solution, especially in a two-electrode double-layer capacitor where ions of
opposite charge must be differentially accumulated at the two respective electrode interfaces of each cell of the device. Hence there must be electrolytic and
diffusional migration of ions during charging or discharging.
The limitations of planar electrodes exhibiting 2-dimensional electrochemistry are largely attenuated if high-area, porous, or particulate electrodes
having a quasi-3-dimensional nature, are employed. Then diffusion limitations
are eliminated or largely attenuated since path lengths for transport of ions or
molecules in the solution are very much diminished relative to the situation at
plane electrodes (semi-infinite diffusion). The transport situation is then locally
like that in a thin-layer cell.I
377
378
Chapter 14
379
important are (1) the pore-size distribution and (2) the accessibility of electrolyte to the distributed pore area of the material. Coupled with the last factor is
wetting by the electrolyte and capillary action within the pore matrix. The latter
can be assisted by the addition of surfactants. The requirements for optimum
utilization of a high-area electrode material for a double-layer capacitor are the
opposite of those for a fuel cell, where a broadly distributed gasfsolid/liquid
"three-phase" interface is necessary to achieve maximum power. For a capacitor
electrode, zero residual gas occupancy of pores is the desideratum, but is not always achievable. While pore-size distributions can be determined on high-area
porous carbon materials (e.g., by employing the mercury penetration method up
to high injection pressures), such data usually do not reflect the actual distribution of electrolyte-occupied pores, which is a much more difficult factor to
evaluate reliably.
High-resolution scanning electron microscopy can provide visualization of
pore structures in the dry state, down to ca. lO-nm scales, while on very thinsectioned preparations, transmission electron microscopy provides information
almost down to atomic resolution, 0.5 - 1 nm. However, few detailed studies of
this kind are available.
From another direction, heats of wetting by an invading electrolyte can be
useful, but more for comparative purposes between two or more preparations
rather than for the absolute evaluation of the heat of wetting. Some results of this
kind are shown for carbon materials in Chapter 9.
The wetting of, and capillary action within, pores are determined by (1) the
local interfacial tension and (2) the contact angle, two not-unrelated quantities.
However, in very fine pores, the significance of the contact angle becomes
nebulous since a macroscopic meniscus is not formed in the regular way. Of
course, in addition, the occupancy of pores by the device's electrolyte is of
prime importance in determining the ohmic resistance of capacitor electrodes
and their equivalent series resistance.
In the following pages of this chapter a brief account is given of the nature
of porous electrodes and the models that have been proposed to represent them,
together with some analysis of their ac and dc behavior. The problems that occur
in the case of electrochemical capacitors originate from nonuniform primary
current distribution,2 (i.e., that determined by inhomogeneities of ohmic solution resistance along conducting pore channels and/or of the electrode matrix)
or finite resistance of the matrix structure itself. This leads to a continuous variation of electrical relaxation rates (reciprocals of RC time constants and consequent frequency response) throughout the matrix, inward from the outer
geometrical interface of the electrode structure in contact with bulk solution, toward the inside distribution of pore channels and pore surfaces, the latter providing the large-area interfaces for double-layer charging. Some model
diagrams of electrode pores will be seen later in this chapter.
380
Chapter 14
It must be noted that the actual micropore structures (so-called "nanostructures") and pore-size distributions depend greatly on the provenance of the higharea carbon materials and their pretreatments, e.g., whether they are derived
from carbonization of synthetic polymers (e.g., acrylonitrile) or of natural organic substances, such as, coconut shells. Comparisons of performances must
take this factor into account.
First we consider the charging behavior of a simple, model series RC element where charge flowing into the capacitance C is impeded by the necessity
of it having to flow through the resistance R. If a potential difference Ll V is applied to a series RC circuit, an initial surge of current I takes place, diminishing
in time until the potential across the capacitor component Cis LlV and I has decayed to zero. The potential difference across R ( = IR) at any time (t) is given
by
~f J.dt=constant
(14.1)
LlVC=IRC+ J.dt
(14.2)
LlV.C)
dl
d [- =O=RC-+I
dt
dt
(14.3)
LlV=IR+
or
Differentiating
then
(14.4)
or
(14.5)
so that
t
-lnl=-+r
RC
(14.6)
381
The integration constant r is found from the condition that at t =0, all of
the potential drop LlV is falling across R (/ is at its maximum value), i.e., 1=
LlVIR. Then
(14.7)
LlV
-In-=r
R
Hence the overall time dependence of I upon charging of C through R is
In LlVIR =_t_
I
RC
(14.8)
In l(t=O) =_t_
I
RC
(14.9)
or
Alternatively
-I- =exp[-tIRq
I(t =0)
(14.10)
That is, the charging current through R into C declines exponentially on a time
scale characterized by a reduced time, tIRC, where RC is the time constant of
the series RC combination. I~O as t~oo; then Ll V = the potential across C and
that across R is zero since I has become zero.
The above elementary example was worked through in order to illustrate
the situation with more complex equivalent circuits. (The equivalent circuit to
a real system is that arrangement of R, C and possibly inductive L components
in series and/or parallel connections that reproduces the observed frequency response of the experimental system, including the phase angle as a function of
frequency w or log wand the correct relations between the real and imaginary
impedance components Z' and Z' in the complex plane over a range of experimental frequencies. In more complex cases, the equivalent circuit may not actually represent the physical basis of the process occurring at the electrode
although for an interfacial electrode reaction, a real double-layer capacitance
component, Cdt> must always form a component of the equivalent circuit with
which often R, other C, or L components may be in series and/or parallel.
From the above, it can be seen how a complex network of parallel RC circuits (each a series RC combination) such as that illustrated in Fig. 14.1 will
have a distribution of characteristic RC time constants and therefore different
charging or discharging relaxation times, depending on the values of Rand C
for each circuit element in the network. A network such as that as shown in Fig.
382
Chapter 14
LR;-----------------
-z"
I I
I I
electrolyte
resistance R
within pore
electrolyte impedance Z
(double-layer capacitance
and faradaic impedance)
FIGURE 14.1. The equivalent circuit of a pore invaded by a conducting electrolyte and having a
distributed RC network (transmission-line model). Typical-Z' vs. Z response asf(w) also shown.
14.1 represents in a crude way the equivalent circuit of a porous high-area capacitor device. The overall frequency-response behavior approximates that of a
transmission line that exhibits a -45 0 phase angle over a wide range offrequencies, such as the Warburg impedance (see later discussion and Chapter 10) involved in a diffusion-controlled Faradaic reaction. In fact, as shown in the paper
by de Levie,3 the differential equations for the charging behavior of the capacitance in a transmission-line network have some formal mathematical similarity
to the equations for diffusion or heat transfer. 4 In addition to the series RC elements, there may be potential-dependent Faradaic or ohmic leakage resistance
elements locally in parallel across the double-layer capacitance elements which
determine the self-discharge behavior (Chapter 18) of the network.
It will be noted that the time scale for charging any RC series elements becomes
normalized in the ratio tiRe, a reduced time scale, i.e., common charging behavior occurs for that scale. The time-constant, Re, appears here as it does in other cases,
e.g., discharge through a resistor or the critical frequency, Wr = liT = liRe, for
the top point of the Z" vs. Z' semicircle in the complex-plane plot for a parallel
RC circuit or the admittance plot for a series RC circuit (see Chapters 16 and
17).
383
Chapter 14
384
t~~~~<
z+dz
./
FIGURE 14.2. Element of the equivalent circuit of a pore identifying the quantities involved in
the treatment of its impedance. (After de Levieh
down the pore, both having values expressed per unit length down the pore in
the direction z, z being measured from the orifice of the pore at Z =o. V and i are
the potential and current,* respectively. The current-conducting matrix phase is
assumed to have negligible resistance (the lower line in the equivalent circuit
diagram, Fig. 14.1).
The ideal pore behaves equivalently to a uniform RC transmission line.
This analogy occurs because an electric power transmission line has a continuously increasing resistance from one end down its length, and a self-capacitance
leakage pathway (to ac) to the atmosphere all the way down its length.
The differential equations for V and i as a function of Z down the pore and
as a function of time for charging processes are written 3 as:
av
az
dy=-dz =-zRdz so
ai
az
aV
az
-+iR=O
av
at
di=-dz=-C-dz
(14.11)
(14.12)
a2i -RC ai =0
az2
at
(14.13)
Similarly, we may get by differentiating Eq. (14.11) d/az and combining it with
Eq. (14.12)
*In order to keep consistency with the mathematical development in Ref. 3, the symbol i is used
here for current and the letter I is defined in Eg. (I4.20b).
385
(14.14)
Equations (14.13) and (14.14) were identified3 as the basic differential equations describing the system under investigation in terms of i and V, respectively.
In the following sections Eq. (14.14) will be solved with
V(z,O)
=V(oo,t) =0;
V(O,t)
=E
V(z,O)
=V(oo,t) =0;
V(O,t)
and
As a result of the mathematical analogy between Eqs. (14.13) and (14.14), the
solutions ofEq. (14.13) may be directly written 3 as:
i(z,O)
i(O,t)
=I
i(z,O)
i(O,t)
=I sin wt
in the galvanostatic
and
in the ac cases.
The solutions of these equations were obtained with the aid of a Laplace transformation.
For the case of potentiostatic conditions:
.
1
I(Z,t)="R
dV
dZ
E'rr-RC~
="R-Y
lm J exp[-rlt]
(14.15)
from which the current i(O,t) at the orifice of t e pore can be obtained as
i(O,t)
=~~ (~;)
(14.16)
The conditions under which the boundary condition V( oo,t) = is permissible were investigated. 3 The assumption of semi-infinite length is found to be
virtually correct when I'? 4(t1R C)1 12, f being the real length of a pore with one
closed end, or half the real length of a pore with two open ends, e.g., at opposite
sides of the macroscopic electrode surface, in which case each half counts as a
whole pore. r is a characteristic time constant, r =Z2RC I 4.
For the case of alternating voltage, V(O,t) = E sin wt, the steady but periodic
part of Vis
(14.17)
386
Chapter 14
(14.18)
and
i(O,t)=~-V(1I2WRC)
(14.19)
From the above equation for V(O,t) and Eq. (14.19), it can be concluded that the
phase angle between the potential imposed on a pore at its orifice z = and the
resulting charging current is 45, so the charging current of a pore behaves formally like the Warburg diffusion impedance of a flat electrode. The 45 phase
angle, independent of frequency w, is a characteristic feature of transmissionline behavior.
Finally, the condition V(oo,t) = will virtually be fulfilled if
t>
- -V(wRC)
with
i(z,O)
=i( oo,t) =
(14.20a)
and
i(O,t) =I
(14.20b)
reads
i=lerfCZ~ (~;)=lerfCf
In order to obtain V we use Eq. (14.11) from which
V(z,t)
=R f idz
(14.21)
387
(14.22)
so
V(Ot)
~ ~(n:cJ
04.23)
IR
VCz,t)
=R
(14.24)
and
VCO,t)
(14.25)
General conclusions from this model can be represented 3 in two diagrams, Figs.
14.3 and 14.4, in which the potential scaled as VIE or current scaled as ill is plotted as a function of reduced time tlr, or the current is plotted as a function of
time for different depths down the pore.
As concluded by de Levie,3 upon the imposition of potentiostatic or galvanostatic conditions (e.g., through a potential or current pulse) at the orifice of
..........
1..0..---------------------,
0.8
...
w
......
>
10
tIT
15
20
FIGURE 14.3. The potential (current) response, VIE or il/, of the pore as a function of reduced
time tlr, when a potential (current) step function with amplitude E(l) is applied at the orifice of the
pore. The characteristic time T depends on the coordinate z along the axis of the pore, on the solution
resistance R, and on the double-layer capacitance C, both per unit pore length, according to T =
iRc/4 (see p. 384). Reprinted from R. de Levie, Electrochim. Acta, 8, 751 (1963), with permission
from Elsevier Science.
388
Chapter 14
1.0
.....
"
I-
">
-1.0
t: ....
4 =(wRC 12)1/2
FIGURE 14.4. The potential (current) response VIE or ill, at different distances z from the orifice
of the pore as a function of time t, when a sinusoidal potential (current) with amplitude E(I) and
angular frequency w is maintained at Z =0; R is the solution resistance per unit pore length and C
the double-layer capacitance per unit pore length. Numbers below curves are values of the quantity
7J(wRCI2)!I2. Reprinted from R. de Levie, Electrochim. Acta, 8, 751 (1963), with permission from
Elsevier Science.
the pore in contact with an electrolytic solution that penetrates the pore(s), the
potential or the current within the pore changes more sluggishly the greater the
depth within the pore. In the potentioharrnonic case, however-which as a consequence of the neglect of transient terms, is in the Levie treatment equivalent
to the galvanoharrnonic case-the frequency maintained at the orifice is left unchanged throughout the pore, whereas its amplitude decreases with increasing
z down the pore.
Finally, it is very useful to define a quantity characteristic of the fraction
of the pore effectively playing a part in the charging process. The foregoing
equations suggest the use of z ..J(tlr) = ..J4tIRC or zl..J(1l2wRC) for such a
penetration depth in the dc pulse or ac cases, respectively. This penetration
depth increases with time in potentiostatic or galvanostatic measurements, and
decreases with frequency in the alternating voltage or alternating current techniques. Hence, after a very short time or at high frequencies, what is virtually
measured is the capacitance response of the more or less flat external electrode
surface, the pores being unable to follow the variations in potential or current
sufficiently quickly. Thus there is a major dispersion of capacitance. After a certain lapse of time or at lower frequencies, the influence of the pores, however,
becomes relatively more important because the inner electrode surface is usually very large in comparison with the outer or projected one. Consequently, if
the pores are adequately described as circular cyclindrical, but of nonuniform
diameter, the narrow pores, as it were, tune in after a longer time or at lower frequencies than the wide ones.
These conclusions are very important for recognition of the power limitations associated with porous electrode capacitor devices, especially those based
389
75
.....
-e50
40 2
10-
rZw
-ir-1
10
104
LOG [r~w/D]
FIGURE 14.5. The phase angle I/J, in degrees, between the applied alternating voltage and the resulting alternating current in cyc1indrical diffusion, plotted as a function of mean diffusion coefficient D, radius ro of the cylinder and angular frequency w. Reprinted from R. de Levie, Electrochim.
Acta, 8, 751 (1963), with permission from Elsevier Science.
390
Chapter 14
relevance to electrochemical capacitor behavior occurs in charging and discharging behavior where transport and concentration (or dilution on discharge)
of counterions in the Helmholtz layer (Chapter 6) are required to balance the
changes in surface charges on the interfaces of the conducting electrode particle.
The ratio of the current icy], in cylindrical diffusion, to the current ip , in planar diffusion, is plotted (for potentiostic conditions) in Fig. 14.6 from de Levie's
calculations as a function of diffusion coefficient, D, radius ro of the cylindrical
pore and time t (reduced time scale, tlT6ID). A corresponding plot of phase angle J between the applied av and the ac response current in cyclindrical diffusion is shown in Fig. 14.5 as a function offrequency expressed as wl(D1r5). Of
course the variables tl(DIT6) and w(DIT6) are to be regarded as conjugate quantities since inverse frequency corresponds to a response time (cf. RC time constants, T, and corresponding characteristic frequencies, 1/r).
Finally, the more complex and complete case of a combination of doublelayer capacitance, diffusion, and coupled Faradaic reaction was also treated by
de Levie (Ref. 3, parts I and II). The equivalent circuit element in this case involves a Faradaic leakage impedance ZF across the capacitative elements; ZF
may also include a series Warburg impedance element. The representation of
such conditions is shown in Fig. 14.7. The system now behaves as a uniform RC
transmission line but with leakage, the leakage and part of the capacitance, however, being frequency dependent.
The calculations for this case were based on a model similar to that in Fig.
14.7 with C and ZF combined as a total impedance of the electrode interface per
1.0
0.8
....
0.6
0.4
....""'-
Q.
>u
....
0.2
0
10 5
10
FIGURE 14.6. The ratio of the current icyl in cylindrical diffusion to the current ipl in plane diffusion, ceteris paribus, in potentiostatic measurements, plotted as a function of mean diffusion coefficient, D, radius, ro, of the cylinder, and time t. Reprinted from R. de Levie, Electrochim. Acta,
8,751 (1963), with permission from Elsevier Science.
391
l.:9.!
FIGURE 14.7. A small section dz of the equivalent circuit of a pore with uniformly distributed
electrolyte resistance, R, double-layer capacitance, C, and Faradaic impedance, ZF, all taken per unit
pore length. z is the distance from the orifice of the pore measured along the axis of the pore; V and
i are the potential and the current respectively. The electrode itself has no resistance (lower horizontalline).
unit length of pore. Diffusion, coupled to the Faradaic impedance, was considered only in the radial directions, not along the direction of the pore. This is justifiable as long as the penetration depth is much larger than the pore radius.
The general differential equations, analogous to those for the non-Faradaic
case are, for the infinitesimally small section, dz, of the pore, are
dV
dV=-dz=-Rdz so
dV .
-+lR=O
(14.26)
. ~
dl =- dz
-....!.... + -
d'
dZ
(14.27)
dZ
dZ
=- -ZV dz
dZ
so
V
Z
=0
and likewise differentiating Eq. (14.26) and W.r.t. z combining with Eq. (14.27),
giving
(Fe R
---e=O
dZ 2 Z
(14.29)
(14.30)
392
Chapter 14
in which the reciprocal of the real part of ..J(R/Z) can be recognized as a penetration depth like that for the non-Faradaic case.
Further analysis yielded
.( )
1 de
z Z =- R dZ
= ..J(ZR)
eX1J -Z~
'lZ )J sm. wt
(14.31)
and
i(O) = _bsinwt
"'/(ZR)
(14.32)
The phase angle J between potential and current at Z =0 is given by the ratio of
the imaginary to the real component of lI..J(ZR) and since the argument of the
root of a complex quantity is half the argument of the complex quantity itself,
it is found that
b
a
2J =arctan-
(14.33)
in which we denote the complex quantity lIZR by a + bj, a and b being real coefficients, j ==."J-l and the minus sign for J being omitted.
Other cases involving the radial diffusion effect were also treated by de
Levie but are outside the scope of this discussion of the behavior of porous electrodes. They are treated in some detail in Ref. 3.
In a second paper9 containing the fourth part of his seminal work on the
electrochemical behavior of porous electrodes,9 de Levie showed that the theoretical treatment of the first three (contiguous) parts of that work could fully account for the behavior of a model brush electrode designed to simulate the
behavior of a porous electrode. The brush electrode, made from parallel Pt wires
with the wire tightly wound around the brush arrangement, is shown in Fig.
14.8. The experimentally determined impedance behavior, plotted in the complex plane (Z" vs. Z') for this model electrode system, is shown in Fig. 14.9 for
three concentrations of KCI electrolyte at 25C: 0.1 M, 0.3 M, and 1.0 M. The
experimental data plot out exactly as 45 lines corresponding to the 45 phase
angle expected for a cyc1indrically porous electrode. The brush electrode was
comprosed of 2000 Pt wires 50 J1 in diameter, each about 10 mm long <f dz = 1
cm).
Geometrically such an electrode matrix does not provide strictly cylindrical pores; in fact, it has a conjugate space between the circular metal cylinders
in contact, having hexagonal symmetry. However, the simulation provides an
excellent match to the behavior of a cylindrically porous electrode. A sintered
Pt wire,
O.5mm
393
cp
Jeno gloss
platinum brush
electrode
FIGURE 14.8. Platinum wire brush electrode as model of porous electrode having parallel pores.
(After de Levie. 3)
silver powder electrode, made from spherical silver particles, also gave impedance behavior exhibiting a 45 phase angle independent of frequency.
A useful summary relation of the qualitative variation of phase angle on
log (frequency) can be given as shown in Fig. 14.10, which illustrates how the
160
20
30
120
40
....... 80
-N
40
50
150
250
tIn
FIGURE 14.9. Measured cell impedance of the model platinum brush electrode (Fig. 14.8) in 0.1
to 1.0 M aqueous KCl at 298 K plotted in the complex plane for a frequency range of 20 Hz to 1
kHz. Lines are drawn with a slope of 45, corresponding to transmission-line behavior. Reprinted
from R. de Levie, Electrochim. Acta, 8, 751 (1963), with permission from Elsevier Science.
Chapter 14
394
50r-------------------------------------------~
pore
double-layer
exhaustion
capacity
30
(capaci10r)
diffusion
......
.g...
10
OL---------------______
Log
~~
____
__________
tJJ
FIGURE 14.10. Schematic representation of the dependence of the phase angle of the apparent
electrode impedance of a pore on the angular frequency w. In each horizontal part of the graph, the
dominating response effect in that frequency region is indicated. (After de Levie. 3)
spectrum of relaxation frequencies characterizes various distinguishable response processes that can be taking place in a porous electrode. In a porous capacitor electrode system, ideally the electrode Faradaic reaction is absent but,
as indicated earlier, the role of electrolytic migration and diffusion of counterions can be significantly dependent on specific electrode area (cm2 g-l) and the
concentration of the electrolyte that is employed. This latter factor, of course,
also determines the solution resistance in the pores.
In practical impedance measurements on porous electrode devices, the resistance outside the matrix, in series with the impedance of the multipore system, must be allowed for and may usually be distinguished experimentally
through evaluation of data and plotting of Z" and Z' components of Z in the complex plane (see Chapters 16 and 17). A simplified approach is possible3 for
small-amplitude av when only the steady-state harmonic response is considered.
For such a condition 5 m V), Z is dependent only on the angular frequency w.
Then, for the small element, dz, of the pore equivalent circuit and denoting by
Z the impedance of the electrode/electrolyte interface per centimeter and similarly for R, the ohmic resistance per centimeter,
dV =-iR-dz
so
dV/ dz + iR =0
di=-dZ
so
dildz+=O
(14.34)
and
(14.35)
395
d2i _ Ri = 0
(14.36)
d2V_RV=O
(14.37)
dz 2
and
dz 2
These are the basic differential equations for the model under the above conditions. Then, for the semi-infinite pore:
V(oo) = 0;
and
V(z)
= E exp(-z -YRIZ)
(14.38)
and also
.
l(Z)
_~
=-1 . -dV =- E exp( -z'lRIZ)
R dz -YZR
(14.39)
with
i(O)=~
ill
(14.40)
Thus the pore behaves as an impedance, .,izR, which can be called the apparent pore impedance, Za, since Z, and correspondingly ZR, is a complex quantity that can be represented by real and imaginary components in the complex
plane. In this plane, Z is represented by a vector of length IZI and a phase angle
. The correspondin~ctor {Z has the length ..yjzJ and phase angle /2. Since
R is scalar,
and -vZR differ only in their vector magnitude. The calculations3
lead to the following conclusions under the assumption that the curvature of the
pore has no influence on Z:
rz
From the mathematical equations developed, it turns out that the reciprocal
of the real part of-VRIZ can be identified as a quantity characteristic of the frac-
396
Chapter 14
tion of the pore length that is effectively involved in the ac response measurements. It is thus a penetration depth, A, which is defined as
A ="IZlR
/ R cos(J/2)
(14.41)
Table 14.1 shows calculated (Ref. 3) maximum values of A for various values
of the radius (ro) and electrolyte specific resistance (P).
The behavior of a simple redox reaction in a semi-infinite pore was treated
in Bockris and Srinivasan's monograph6 on fuel cells, based on analyses published by Newman and Tobias lO and by Micka and Rousar. II - 13 Expressions for
the current and overvoltage distribution in the pore were derived, based on elliptic integral calculations.
These treatments are relevant more to Faradaic processes at fuel cell e1ectrodes l4 than to capacitors although, as stated earlier, significant relevance
arises when self-discharge Faradaic leakage currents are involved. We give below some of the principal equations leading to relative current distribution and
total current, It, and for the overvoltage distribution down a pore.
Two cases were considered6 : (1) for an overvoltage <100 mV and a transfer coefficient (Chapter 3) value a = 0.5, for which both backward and forward
currents of the Faradaic reaction near equilibrium (based on the Butler-Volmer
equation; see Chapter 3) have to be taken into account; and (2) for the simpler
case where overvoltages are> 100mV and then only the forward direction of the
reaction need be treated.
The differential rate dIz of the Faradaic reaction in the element dz along the
pore of radius ro is given in terms of the overvoltage, t7, at Z =z; the transfer coefficient, a; and the exchange c;:urrent density, i o, by
100
10
0.03
0.1
0.3
100
0.01
0.03
0.1
0.3
1000
0.003
0.01
0.03
0.1
Source:
1000
Reprinted from R. de Levie, Electrochim. Acta, 8, 751 (1963), with permission from Elsevier Science.
Notes: ro = Radius of the pore (j1); p = specific electrolyte resistance (ohm cm); C = double-layer capacitance,
taken to be 20 J1. cm- 2; w = angular frequency, taken to be 500 rad S-I (approx. 80 Hz).
(If other components of Z are present besides the double-layer capacitance,
above.)
The penetration depth deduced for the earlier general case was d
[a:;)-
exp [(a - 1)
~~]}
397
(14.42)
(14.43)
Knlo
where I z is the current generated in one pore from z = 0 to z = z and K is the specific conductivity of the electrolyte. Differentiating Eq. (14.43) and combining
the resulting equation with (14.42) leads to
d2lJ
dz2
(2). { l(al]RTF) -
= Kr2
exp
10
FJ }
I]
exp -(1 - a) RT
(14.44)
=1]0
at
z=O
(14.45)
and
2
(dl]) z=O
Iz=O =Knr 2 dz
atz=O
(14.46)
..
Iz=t= It
-;(dl])
=Knro l dz
z=O
(14.47)
tan( If/)x=x
it
tan( If/)x=l
(14.48)
Chapter 14
398
and
2RT (4a)112 . h Yo
. =Kn~aT
sm "2 tan ()
If! =1
It
(14.49)
where:
cos(lfI) =
sinh(yol2)
(14.50)
sinh(yl2)
kI(k,lf!)
=al12x
(14.51)
(14.52)
'II
f(k,lfI)
=fo (1 -
k2 d.1f!2
sm IfI
)112
2iof2F
a=--
(14.53)
KRTro
k=
x=l'
(14.54)
'IF
y= 2RT
(14.55)
cosh(Yol2)
= sin ()
(14.56)
f(k,lfI) is the elliptic integral of the first kind and tables of this integral are available as a function of () and 1fI.
Case 2 for 'I 100 mY.
For this case, the relative current-density distribution, the total current density, and the overpotential distribution were given 6 as follows:
iz
tan[(a1l212) exp(yol2)x]
(14.57)
it = Knr~
1
2RT
(YoJ~U
IF al12 exp (Yo)
("2 tan [(a
2 12) exp ("2
399
(14.58)
and
4RT
11z-11o=Flnsec
(14.59)
The latter equation is the expression for the ohmic potential drop in the solution
fromz = 0 to z = z. According to this equation and Eqs. (14.53) to (14.55), it follows that the ohmic potential drop in the solution depends on the parameters
i o, 11 0, ro, and tIn the general case, for which Eqs. (14.57) to (14.59) must be used, IzlIt is
initially linear with x (= zit), having a slope of 45 with the x axis, but as 110 increases, it varies rapidly with x (Fig. 14.11). The overpotential vs. current density plot in Fig. 14.12 shows two sections depending on the current density-at
lower current densities the normal Tafel slope is observed, whereas at higher
1.0r------------------lP
0.8
+,0.6
.......
"-N
.......
0.4
0.2
0.3
0.5
zll
0.7
0.9
FIGURE 14.11. Current-distribution relations, Izi I" for a case of activation and ohmic overpotential (elliptic integral calculation) along (vt) a two-phase porous electrode. Parameters: io = 10-6
A cm- 2 ; K =0.1 ohm- I em-I; 1'2 =5 X 10- 1 em; 0 -170 =0.243 V; ~ -170 =0.312 V. (From Bockris
and Srinivasan. 6 Reproduced with permission of the McGraw-Hili Companies.)
Chapter 14
400
1.0
>
0.8
:;: 0.6
CD
15
e- OA
CD
>
0 0 .2
O~------~~-----J--------~
10 -2
10-1
101
______~
current densities the Tafel slope is double the normal value. On the other hand,
for very low io 10-9 A cm-2) or large specific conductivity (K) values >10
ohm- 1 cm- 1, the current distribution is linear with distance down the pore (solid
lines, no points, in figure 14.11) and the predicted Tafel slope retains its normal
value.
The pioneering theoretical work by de Levie3,9 was taken further by
Keiser, Beccu and Gutjar, 15 who derived relations for the impedance as a function of frequency for various geometries of a single pore and applied the results
to the case of a porous electrode. By means of comparisons calculated for various electrical models in relation to the overall impedance curve, they showed
that the average pore structure of a porous electrode could be estimated. Again,
the mathematics are complex and laborious. Essentially, the transmission-line
equivalent circuit was used but various pore shapes, as illustrated in Fig. 14.13,
were investigated. They gave rise to a characteristic series of complex-plane
curves of Z' vs. Z' for selected parameters (Fig. 14.13) representing different
form-factors for the pores.
First, a cylindrical pore was considered oflength t and radius r, with a uniformly distributed double-layer capacitance at a smooth surface. The electrical
properties of the pore were regarded as made up of RC elements in a transmission-line arrangement along a length variable, x as in Ref. 3.
401
0.4
0.3
N
I
0.2
o. ,
0.1
0.2
0.3
0.4
0.5
Z
FIGURE 14.13. Pore shapes treated by Keiseret al. and corresponding complex-plane impedance
plots. Also shown by dashed lines. Reprinted from Keiser et aI., Electrochim. Acta, 21, 539 (1976),
with permission from Elsevier Science.
For the impedance change, dZ, along the pore length x (z in de Levie's notation), the complex resistance was written
Z - dZ = Rdx + Z/( 1 + jwZCdx)
(14.60)
(14.61)
Z' - jWCZ} + R = 0
(14.62)
z:
402
Chapter 14
defined asf(I1); for a cylindrical poref(l1) == 1 but has other values for differently
shaped pores (see Fig. 14.13). Then an integration of the differential equations
above gives an explicit result:
Z'
(14.63)
For other pore form factor values, only numerical integration can be performed.
The general curve for -Z'IRo vs. Z"fRo, i.e., a complex-plane plot, is shown in
Fig. 14.14 for a series of ifA values.
For a series of R, C combinations (R" C,; Rb C2 RN, CN) in a transmission-line network, Zb the impedance of N-k elements gives the recursion formula
(14.64)
Normalizing the form factorJ(rf) as
(14.65)
for noncylindrical pores, there results after further mathematics'S
Z'H
4 (i)' A+ IIZZ]
= IINJI + II [
(14.66)
403
0.6
2
0.5
(t)
0.4
0
_0::
...... 0.3
N
I
0.2
0.1
0
0.1
0.2
Z'IR o
0.3
0.4
0.5
FIGURE 14.14. Master complex-plane impedance curve for form-factor values, ttl., and
Ro = IJ7rKr2. l. = 0.5 (rKlwC)ll2; notation of Ref. 15. Reprinted from Keiser et aI., Electrochim. Acta, 21, 539 (1976), with permission from Elsevier Science.
Note that for pore-form numbers 3 and 4, there is a tendency for an initial
(high tf).. values) semicircle to become developed. It is interesting that this has
been observed experimentally with a compressed porous Ni electrode. 7
Chapter 14
404
CIC
0~-~8~~~-~4~~~0~~~4~~~8~
X
FIGURE 14.15. Average relative ionic concentration across the cell for various values of the y
parameter. The dashed line represents the ionic concentration profile before charging; Yl = 1.155,
Y2 = 10.40, Y3 = 22.49, Y4 = 46.20, and Y5 = 104. Reprinted from C. J. E. Farina and K. B. Oldham,
1. Electroanal. Chern., 81, 21 (1977), with permission from Elsevier Science.
12
10
8
><
"0
"
"0
-8
-4
FIGURE 14.16. Plots of the undimensionalized electric field d~/dX vs. x. y values as in Fig. 14.15.
Reprinted from C. 1. E. Farina and K. B. Oldham, J. Electroanal. Chern., 81, 21 (1977), with permission from Elsevier Science.
TABLE 14.2.
10-2
10-3
10-4
405
Half-width (run)
Minimum
Maximum
7
21
11
34
67
106
212
335
669
1060
Source: Reprinted from C. J. E. Farina and K. B. Oldham, J. Electroanal. Chem., 81, 21 (1977), with
permission from Elsevier Science.
Various interfaces (e.g., coastlines and crystals) have in recent years been
modeled in terms of so-called fractal geometries, i.e., structures that are selfsimilar on various scales. Similar applications have been made to porous electrodes. Mulder et al. 8 have examined Tafel currents and related admittance
spectra at fractal interface electrodes, e.g., those having a typical self-similar
406
Chapter 14
00 0
o
b)
eoeDeoe
oDe
a
FIGURE 14.17. (a) Elements of an electrode/solution interface having a fractal structure based
on hexagonal (60/120) elementary geometry. (b) Cross-section geometry of a square fractal pore
structure. (Sapoval, ref. 23).
407
nv
27
64
125
216
343
3x3x3
4x4x4
5x5x5
6x6x6
7x7x7
ns
26
56
98
152
218
n/n v
0.963
0.875
0.784
0.704
0.636
Note that for sufficiently large n, nslnv ~ 61n. For a sufficiently large spherical
particle (e.g., one having a diameter of 50 atomic diameters, that is about 15
nm), the number in the bulk compared with the number in the surface would
scale approximately as 4/3 n?/4n?, i.e., with rl3. In practice, some particles of
powder materials (e.g., soots) are probably fractal and so have greater surfaceto-volume atom ratios than in the case of regular cubic clusters. For those interested in mathematics, the differences between sequential ns values given above
for cubes of atoms with n > 2 form an arithmetic progression of numbers advancing by 12.
The electrochemical accommodated charge is proportional to ns and thus
the stored energy would be (qns) x V for a potential V where q is the excess
charge of the interface calculated per atom of the substrate based on ca. 1.5 x
10 15 atoms (real cm2rl. The weight of material required to accommodate this
charge is proportional to nv x A, so the energy density of the accommodated
charge will be
energy density = (qns) V/ Anv
408
Chapter 14
where A is the atomic weight of the substrate material, q the charge accommodated per atom, as above (e.g., 0.18 to about 0.3 e atom-I), and Vis the voltage
(relative to that of another electrode).
As an example take a carbon sample having 100 m2 g-1 (a low figure), i.e.,
106 cm2 g-l. This is 1.2 X 107 cm2 per mole (A = 12) of carbon; but 1 mol == 6 X
1023 atoms and 1 cm2 accommodates on the order of 1015 atoms; therefore
nslnv for such a sample of Cis (1.2 X 107 X 10 15 )/(6 X 1023 ) = 0.02, or 2%. From
the formula deduced above for n,lnv> 2, we find n =:: 300 for such carbon
particles. This illustrates the commonly perceived importance of utilizing
carbon materials that have the highest possible real area per gram but that
have materials properties (e.g., wettability and pore distribution) of such a
kind that maximum access of electrolyte to the particle surface (the ns factor)
is still maintained, allowing the best achievement of stored charge and hence
energy density. Working out a compromise between these two factors in
double-layer capacitor devices is usually a materials-science art in this field,
depending very much on the pretreatments of the carbon materials and their
biological or fossil origins (see Chapter 9). In fact, for various carbon materials having a range of particle sizes and corresponding specific areas per
gram, the materials giving maximum storable energy (energy density) may
not coincide with the specifications of another material that delivers maximum power density (see Chapter 15). This factor has to be taken into account in optimizing materials specifications in designing an electrochemical
capacitor for particular purposes.
The question of the role of pore size and size distribution in porous carbon
materials for electrochemical capacitors was usefully considered by Shpz who
made comparative double-layer capacitance and BET (N z) surface area determinations on various high-area carbon materials.
Various methods have been employed for real surface area measurements,
the best known one being due to Brunauer, Emmett, and Teller (BET) and variants based on Kr desorption into a chromatograph detector. The BET method
is, however, less satisfactory for microporous materials. Other procedures are
due to Dubinin and to Joyner, Barrett, and Halenda; and another is based on uptake of Zn 2+ ions by adsorption from liquid (aqueous) solution. The latter procedure may probably be assumed to give real areas relevant to electrolyte
solution/carbon interfaces since the samples are actually wetted by the electrolyte.
According to IUPAC classification, pore sizes can be grouped in three
classes: micropores, <2 nm; mesopores, 2 to 50 nm; and macropores, >50 nm.
409
These figures may be compared with the diameter of a hydrated sodium ion (ca.
0.65 nm) or of a water molecule (0.286 nm).
With Spectracorp microbeads of activated carbon, measured double-layer
capacitance values (200 to 2600 J1F cm2 external area) are very high, so much
of the microporous structure must apparently contribute to the capacitance. Surface-area distributions with respect to pore width determined on two Spectracorp carbon samples (M20 and M30) are shown in Fig. 14.18. A large
proportion of the real surface area arises from pores in the 0.5- to 5-nm range of
diameters. It was the conclusion of Soffer et aL 23 that it is these smallest pores
that give rise to most of the BET area that are wetted, consequently contributing
to the measured double-layer capacitance.
The dc capacitance distribution as a function of total surface area (m 2g- 1)
is shown in Fig. 14.19. The dc capacitance divided by external surface area,
plotted against micropore surface area and divided by external surface area, is
shown in Figs. 14.20 and 14.21 for activated microbe ads and activated carbon
fibers. These results illustrate the relative importance of the microstructure area
in obtaining the desired large double-layer capacitance values per gram. The
slopes of the lines in these two graphs are significantly different, being 0.07 for
the microbeads and 0.20 for the activated fibers.
400
-I
0>
N
I..
Miera
.I.~I.
_ _ _ _ _M_e_sa
______
.,.~-tl.~f-a_e_ra----l
. ...
300
"<l:
0::
200
<l:
U
~
MIa
M 30
100
0::
::>
Cf)
o
10 2
PORE WIDTH/A
FIGURE 14.18. Surface-area distribution plots vs. pore width (log scale) for two carbon preparations, MID and M30. Ranges of micro, meso, and macro classification are also indicated.
(Ashursht Technology Center, Inc., from Shi.22)
410
Chapter 14
4.5
'0>
M20B
ll...
......
4.0 I-
W
U
M20A
M20C
3.5 l-
<t
<t
0..
MISA
3.0 I-
"
""',
M30
M20E
Q
MIO
2.0
1000
MISC
2.5 l-
M,,,
M200
MISB
I
1250
I
1500
1750
,.-,
N
30
'--'
0
<I>
'-
<t
. 25
<I>
\ ..
20
en
"0
c
...
15
<I>
-;;:
UJ
......
QJ
u
c
10
.E
'u0
0.
30
60
90
120
,....,
411
3.0
2.5,
(1)
.E
5
(/)
2.0
1ii
1.5
g
";(
w
......
c:
1.0
8.
0.5
0.0
(3
/-
,....
_ _ _ _ .~ _ _ _ _ _ _L _ _ _ __ 1 . _
5
10
15
20
It must be stated, however, that although the very small micropores contribute substantially to the real area per gram, this fraction of real area is not necessarily the most useful or electrochemically accessible since it is accessed only
through an appreciable cumulative solution resistance (transmission-line
model). Also, whether such pores are wetted by the electrolyte and hence electrochemically chargeable is questionable. Certainly they will contribute only in
a small degree to the charge storage capacity under high-rate or short-duration
power-pulse discharge or recharge, contrary to the implications of the conclusions in Ref. 22. These matters have been the subject of controversial discusSIOn.
-2 .0
-60.0
-80.0
0.0
1.0
--0.5
CD
III
"l:l
0.0 . _I
"0
-20.0
g' -40.0
--
'"<D
i\l
'0.
C -1 .0
--
"E -1 .5
(i)
10
2 .5
3.0
103
log (w/rad 5-')
10'
102
Z(real)/O em'
2.0
. 3.
1.5
104
3.5
lOS
4.0
(b)
(a)
-2 .0
-60.0
-80.0
0.0
0 .0
--0.5
-1.0
0-
J:::.
<0
VI
10- '
v.v
-20.0
0>
III
0>
:?
--
"E -1.5
0
(i1)
10
0.5
10 2
10 3
log(w/rad s-')
10 '
1.5
2.0
Z(reall/O em'
1.0
:1
104
2.5
10 5
3 .0
(b)
(a)
-......
':s'
CD
"C
1\1
I\)
....
-1 .2
::~
0.0 -1
10
-20.0
10
0.3
0.9
cm 1
1.2
10 2
103
log(w/rad s - ')
10 '
Z(reall/O
0.6
104
1.5
105
1.8
(b)
(a)
FIGURE 14.22. Progression of impedance and phase angle responses [marked (b)] for a porous Pt model electrode with increasing concentration of KOH
electrolyte: (i) 0.5 M, (ii) 1.0 M, and (iii) 4.0 M. (From Bai and Conway.27 Copyright 1993 Royal Society of Chemistry.)
Q.
.c
'"
III
~ -40.0
-...
0
0,
0.0
0 .0
~ -0.3
'"E
01
C -0.6
C
-...
'"E -0.9
(iii)
-1>0
....
w
iil
6'
o
D)
1/1
:::I
o~
0'
"2-
~
>
"0
Co
1/1
ac:
"8
o'
...
:r
D)
<
CD
III
()
:r
CD
3
414
Chapter 14
1.0
C\I
IE
0.11-
1.L..
Ol
o 4
0.'
"0
0.4
.......
<>
-0.2 -
0
-l
-0.5
0.00
-0.05
<>
.0
o <>
-0.10
Overpotentiol / V
I-
- 1
-0.15
VS
-0.20
-0.25
RHE
FIGURE 14.23. Double-layer capacitance C(log) of the porous Pt model electrode as a function
of potential for 5 KOH concentrations: (I) 0.1, (2) 0.2, (3) 1.0, (4) 2.0, and (5) 4.0 M. (From Bai
and Conway?? Copyright 1993 Royal Society of Chemistry.)
415
(mol dm- 3)
0.1
0.2
1.0
2.0
4.0
0.37-0.61
0.65-0.85
2.01-2.50
3.92-4.66
8.18-8.43
ca.
ca.
ca.
ca.
ca.
13.4-22.1
23.6-30.8
72.8-90.6
142-169
296-305
method. 27 It is seen that there is a major favorable change (increase) of Cd! values with increasing KOH concentration and corresponding conductivity, so that
the evaluated roughness factor (based in this case on a Cdl value of 25 f.lF cm- 2)
depends similarly on the electrolyte concentration (Table 14.3). The effect is a
major one and shows that real areas of porous electrodes can be estimated from Cd!
measurements only with highly conducting invading electrolytes and at low frequencies. For an ac modulation of 10 mVat a frequency wHz, the maximum dVldt
would be equivalent to linear sweep voltarnmetry at 10-2 w (Chapter 16).
The porous electrode's electroactive surface becomes unfolded progressively with increasing conductance of the electrolyte. The key factor here is that
the penetration depth of the ac modulation signal (see Ref. 3) becomes larger as
the conductivity of the electrolyte is increased. When the conductivity of the
electrolyte has become sufficiently high, the ohmic resistance elements, R, of
the transmission-line model all become substantially diminished. In other terms,
it is also seen (see Fig. 14.22) that the role of a CPE in the equivalent circuit becomes diminished with increasing electrolyte conductance.
416
Chapter 14
stitution and its porosity. Such effects do not seem to have been examined in
studies on double-layer capacitor charging or discharging, or reliability testing,
but it appears that such evaluations are desirable.
REFERENCES
1. C. N. Reilley, Rev. Pure Appl. Chem., 18, 137 (1968).
2. J. Newman, Electrochemical Systems, Prentice-Hall, Englewood Cliffs, N.J. (1972).
3. R. de Levie, Electrochim. Acta, 8, 751 (1963) (in 3 parts); also see A. Winsel, Zeit.
Elektrochem., 66, 287 (1962).
4. J. Crank, Mathematics of Diffusion, Clarendon Press, Oxford (1975).
5. J. Miller in Proc. Second IntI. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, S. Wolsky and N. Marincic, eds., Rorida Educational Seminars, Boca Raton,
Ra (1992); also see Proc. Third Inti. Seminar (1993).
6. J. 0' M. Bockris and S. Srinivasan, Fuel-Cells, Their Electrochemistry, Chapter 5, McGraw
Hill, New York (1969).
7. T. Pajkossey, J. Electroanal. Chem., 300, I (1991).
8. W. H. Mulder, J. H. Sluyters, T. Pajkossey, and L. Nyikos, J. Electroanal. Chem., 285, 103
(1990).
9. R. de Levie, Electrochim. Acta, 9, 1231 (1964) (part 4 of the series).
10. J. S. Newman and C. W. Tobias, J. Electrochem. Soc., 109, 1183 (1962).
II. K. Micka, Adv. Chem. Series, 47, 73 (1965).
12. K. Micka and 1. Rousar, Electrochim. Acta, 21, 599 (1976).
13. K. Micka and 1. Rousar, Coll. Czeek. Chem. Comm., 40, 921 (1975).
14. S. Srinivasan and H. D. Hurwitz, Electroehim. Acta, 12,495 (1967).
15. H. Keiser, K. D. Beccu, and M. A. Gutjahr, Electrochim. Acta, 21, 539 (1976).
16. E. J. W. Verwey and Th. G. Overbeeck, Theory of the StabilityofLyophopic Colloids, Elsevier
(1948).
17. C. 1. E. Farina and K. B. Oldham, J. Electroanal. Chem., 81, 21 (1977).
18. A. T. Hubbard, J. Electroanal. Chem., 22,165 (1969).
19. A. T. Hubbard and F. C. Anson, J. Electroanal. Chem., 9, 163 (1965).
20. A. P. Brown, M. Reischmann, and D. Pletcher, J. Electroanal. Chem., 50, 65 (1974).
21. B. Sapoval, Solid State Ionics, 23, 253 (1987).
22. H. Shi in Proc. Fourth Inti. Seminar on Double-layer Capacitors and Related Devices, S.
Wolsky and N. Marincic, eds., Rorida Educational Seminars, Boca Raton, Ra (1994).
23. A. Soffer, D. Golub and Y. Oren, J. Electroanal. Chem., 260, 383 (1989).
24. R. Kotz, M. Sullivan, M. Bartsh, and D. Haas, in Proc. Fourth IntI. Seminar on Double-layer
Capacitors and Related Devices, S. Wolsky and N. Marincic, eds., Rorida Educational
Seminars, Boca Raton, Ra (1994).
25. F. W. Geurts and Z. Horovitz, in Proc. Fourth IntI. Seminar on Double-layer Capacitors and
Related Devices, S. Wolsky and N. Marincic, eds., Rorida Educational Seminars, Boca Raton,
Ra (1994).
26. C. Niu, E. K. Sichel, R. Hoch, D. Moy, and H. Tennent, in Proc. Sixth IntI. Seminar on
Double-layer Capacitors and Related Devices, S. Wolsky and N. Marincic, eds., Rorida
Educational Seminars, Boca Raton, Ra. (1996).
27. L. Bai and B. E. Conway, J. Chem. Soc., Faraday Trans., 89, 235, 243 (1993).
28. For example, see A. W. Adamson, Physical Chemistry of Surfaces, 4th ed. pp. 200-203,
Wiley, New York (1982).
Chapter 15
The relation between power and energy requirements in the operation of "prime
movers" has long been an important topic in the engine industry, starting with
the period of steam and improvements in the efficiency of the Newcomen atmospheric steam engine by Watt through his separate condenser, and the invention of indicator diagrams for recording efficiency.
The power and energy requirements at the wheels for typical urban and
suburban driving in a typical compact car that weighs 2000 lb. are listed in
Table 15.1 based on assessments made in 1967. 1,2 These figures relate to the
power-density (p.d.) and energy-density (e.d.) performance of batteries (or
battery-capacitor hybrids) required for contemporary electric vehicles.
Driving cycles (Federal Urban Driving Schedules, PUDS) in recent years
have been analyzed in relation to the use of electrochemical capacitors in hybrid
battery-capacitor energy sources where the capacitor component provides a
load-leveling function for smoothing and handling the high power drains involved in acceleration or uphill driving. Such studies have emphasized the importance of evaluating the dependence of the achieved energy density on the
operating power density of such systems.
During the past 30 years or so, specific power density (e.g., watts per kilogram) and energy density (watt-hours per kilogram) have been empirically plotted against each other, often on logarithmic scales in "so-called Ragone plots."
These types of plots originated in a conference paper presented by Ragone3 at
417
418
Chapter 15
TABLE 15.1. Power and Energy Requirements at the Wheels for Typical Urban
and Suburban Driving
(Compact Car, 2000 lb)
Power or energy
Maximum power (kilowatts)
Energy per mile (watt-hours per mile)
Maximum specific power- (watts/lb)
Specific energy for 50 milesa (watt-hours/lb)
Source:
Urban cycle
Suburban cycle
22.1
234
44.2
23.4
40.1
242
80
24.2
419
1000
AlI/Zn
CD
...J
Li/CIZ
::
.......
a::
LLI
~
a..
u
u..
10
[j
LLI
~ORGANIC
a..
(/)
ELECTROLYTE:
10
11 - -
laO
1000
INTERNAIL
COMBUSTION
ENGINE
NIICd
j'
CD
-.J
CD
...J
~
......
a::
w
.......
0:
LLI
a..
a..
ILl
10
(/)
a::
ii:
1:
!J5
1~
____________~~LUL------.------------~
10
100
1000
FIGURE 15.2. Ragone plots for batteries, fuel cells, and other motive-power sources in relation
to driving power and range requirements. (Reprinted with permission from SAE paper No. 680453
Copyright 1968 Society of Automotive Engineers, Inc.)
420
Chapter 15
batteries and advanced lead-acid cells together with Zn-air and AI-air batteries
of relatively high energy densities. To these systems must be added the electrochemical capacitors which, although of low energy density, can provide attractively high power densities.
A theoretical basis for the hooked shape of Ragone plots referred to above
was not provided in Ragone's paper,3 but the origin of this shape is easy to understand qualitatively in terms of increasing electrode polarization (hence diminution of cell voltage, which determines energy density) as higher power
demands are made on a battery or capacitor at higher rates of discharge (higher
current densities). The principles involved are as follows, noting that electrical
power density is cell voltage (V) multiplied by discharge current density,
i: p.d. = iV. With increasing i, the product iV (the power, P) will tend to
increase, but with further increase in i, V begins to fall due to polarization
losses. These effects are treated in Section 15.2.
The Idaho National Engineering and Environmental Laboratory (INEEL)
compared Ragone plots for advanced electrochemical capacitors and Ni-Cd and
lead-acid batteries (specifications not given) using 1995 data (quoted in
Miller4 .) This comparison is reproduced in Figure 15.3. It will be seen that quite
103
DOEnear-..,.. goal
IIU pow. P....cld
_/
102
...
101
Q)
UUwall24V org.l8III/-packagtd
Q.
100
Primary alkali..
100
101
102
WYnO,
103
421
high energy densities in the range 7-S Wh kg-I for 3-V nonaqueous solvent,
carbon double-layer-type capacitors and about 1-2 Wh kg-I for I-V aqueous
capacitors (Maxwell Corp.) can be attained. Pinnacle Corp. 2S-V capacitors
(probably of the RuOz mixed-oxide type) lie at about O.S Wh kg, with about 400
W kg-I p.d. rating. The 3-V organic electrolyte solution capacitors lie at about
200 to 3000 W kg-I. The Department of Energy' s (DOE) projected goal is about
1.7 kW kg-I. The currently quoted power density levels are all well above those
estimated for Ni-Cd and lead-acid systems, demonstrating the substantial advantage of electrochemical capacitors over appropriately chosen batteries when
high power density is the primary requirement. This behavior is observed notwithstanding the intrinsic problem with high-area double-layer capacitors of
distributed internal pore resistance within the electrode matrices (Chapters 14
and 17). The initial slopes of the Ragone plots for the capacitors (on the log-log
scale) are comparable to those for the battery system data plotted and indicate
the role of similar iR and i2R losses caused by internal resistance (these factors
are analyzed later in this chapter).
422
Chapter 15
until all the active material is consumed in (practice, an unusual situation), then
the energy density is
(15.1)
(15.2)
(15.3)
423
424
Chapter 15
cessible areas of active materials in batteries or capacitors are usually not accurately known or, in the latter case, depend effectively on the current drain I that
is due to the distributed nature of pore sizes and depths in porous carbon matrices (Chapter 14) or other high-area electrode materials. Thus, power densities
are expressed as W g-I or kW kg-I. Often single values of p.d. are given for particular designs of batteries or capacitors, but this is not a satisfactory basis for
specifying p.d. for a particular battery cell or capacitor device since the power
that is developed on discharge (or required on recharge) depends, as already
noted, on the rate factor i or I (see Eqs. 15.4 or 15.5) or, for capacitors, as the
particular voltage, Ll V, during discharge at which p.d. is to be expressed (Section
15.2.5).
The density factor in energy density is a more definitely definable quantity,
at least for the case of the maximum theoretical e.d. It derives from the charge,
Q, that is Faradaically available from the cell redox reactions in the case of battery e.d. and is expressed per mole of active material or, more usually, per gram
or kilogram of such material, according to Eq. (15.2), taking into account the
number, z, of equivalents per mole.
In the case of double-layer capacitor devices, of course, most of the charge,
Q, arises, not from Faradaic charge transfer processes involving changes of the
oxidation state of reagents (and often also changes of phase that lead to irreversibility), but from removal (on discharge) or addition (on recharge) of electrostatically accommodated charge (Chapters 2 and 6). In most
double-layer-type capacitor devices, only a small fraction of the total accommodated charge is associated with Faradaic processes (e.g., originating from surface or edge functionalities at graphitic C), as discussed in Chapter 9.
By contrast, in the case of redox-type pseudocapacitance systems [e.g.,
Ru02 or the electroactive conducting polymers (Chapter 12)], an opposite situation obtains: most of the (pseudo-)capacitance arises from redox Faradaic-type
charge transfer processes plus about 5-10% of the total storable charge from the
double-layer capacitance associated with the usually relatively high real areas
per gram of such materials. These two types of processes usually have different
characteristic relaxation times or equivalent frequencies in pulse or av modulation, and can thus be distinguished. The relative extents of capacitance charge
storage in these materials depend, however, very much on their method of
preparation and electrochemistries.
Returning to the question of the p.d. of double-layer type capacitors, it is
usually considered that no kinetic limitations restrict their p.d.s, as occurs in
Faradaic processes (Tafel-type polarization, Eq. 15.4), since most of the charges
are accommodated purely electrostatically and thus are not subject to kinetic
limitations in the rates of charge or discharge. Fundamentally, however, such kinetic limitations can arise to some degree, especially for anion adsorption and
desorption in the case of the more polarizable and donative anions (e.g., Cl-, Be
425
CNS-), where certain degrees of quasi-Faradaic, partial charge transfer arise (as
known at Hg). This is equivalent to the development of a pseudocapacitance
component in the measurable overall interfacial capacitance. Such behavior can
occur at the edges or faces of graphitic carbon materials (see Chapter 9).
15.2.2. Power Density
In the case of power density, it was emphasized that it is not possible to define or calculate a theoretical maximum value as it is with e.d.s, where Faraday's
laws and equilibrium reversible potentials are involved. The power delivered
from a device corresponds to a certain rate of discharge into some load resistance at a particular arbitrary current density, and being defined (as expressed
in words earlier) as
power = current (I) x voltage (V)
obviously depends on J for a particular Vor Ll V of a cell. However, Vor Ll V also
normally decrease with increasing J owing to two kinds of polarization losses
of voltage: (1) an ohmic JR drop caused by the internal resistance of the cell
electrolyte and electrodes and (2) kinetic polarization effects associated with
Tafel's relation between log (current density) and overvoltage tJ. This is written
as
In i = In io + atJ FIRT
(15.4)
=III Vrev -
JR - EtJ(l)]
PR - IEb In i + JEb In io
(15.5)
(15.6)
where bs are the Tafel slopes RTlaF of the respective anode and cathode reactions. III Vrev is the ideal power factor for increasing J if Ll Vrev were uninfluenced
by the magnitude of J, i.e., P would increase linearly with I. This is actually observed for sufficiently low J values that the ohmic and Tafel polarizations re-
426
Chapter 15
~Vrev -
= ~ Vrev
(15.7)
(15.8)
The rate of change of P with current given by the above differential equation has
several components: (1) an initial linear increase at low currents that is proportional to the voltage, ~ Vrev , provided by the electrochemical power source; (2)
with increasing current drain, a linear rate of decrease in power with current that
is proportional to 2IR arising from the PR term; (3) a rate of decrease in power
that is related to the log of the current (from the I[a + b log i] term); and (4) a
constant term equal to the sum of the Tafel parameters a and b also arising from
differentiating the I[a + b log i] term. In addition (see Section 15.3) there can be
a concentration polarization effect on p.d. Thus the rate of decrease of power
with increasing current (decreasing load resistance) exhibits complex behavior
and, as shown earlier, the p.d. plotted against current or current density exhibits
a maximum that is determined by the polarization parameters, especially R at
high current densities.
The p(max) arises at a current given by
2IR + Eb In IIA = ~ V rev + Eb In lolA - LAb
(15.9)
Note that p(max) is determined both by I and In I; thus with increasing I, it is seen
that the rate of decrease in P is actually mainly determined by the ohmic IR term,
which tends practically to be a more important factor than the kinetic polarization at high I unless R is relatively negligible. However, there is in addition the
constant factor, LAb, which is related to the sums of the Tafel slopes for the anodic and cathodic partial reactions of the cell. Also, as expected, better maximum power is attained when the reversible cell voltage, ~ Vrev' is larger, i.e., for
a bigger open-circuit emf or driving force of the cell. On the basis of the equa-
427
tions given above, some examples of P plotted against I for several values of the
parameters Rand b have been given in Ref. 5.
Note that at sufficiently high power drain (virtually a short circuit), Ll V will
be reduced to zero so that/eLl V) = Ll V =0, which occurs when (see Eg. 15.6)
(15.10)
(15.11)
= IR + 11(/)
(15.12)
Practically, this occurs when I tends to be sufficiently large that the IR drop virtually opposes the reversible cell voltage so that no power is available any more.
An additional factor can sometimes be the concentration polarization in
certain cells, e.g., with soluble redox reagents or fuel cells with soluble fuel,
such as CH3 0H. That is, at sufficiently high current drains, power becomes limited by mass transport of the reagent or by zero concentration of the reagent at
the reactive electrode surface. An analogous effect can sometimes occur in an
electrochemical capacitor when a high-area electrode material is combined with
a limited volume of electrolyte. Then, upon charging, significant gradients of
ion concentration can arise and most of the electrolyte ions become adsorbed at
the high-area interfaces of the electrodes (see Section 15.3 and Chapter 6).
15.2.3. Relation to Energy Density
The e.d. can be written simply as Q.LlV. Then, as above for P, Ll V is diminished from Ll Vrev by the polarization terms, i.e.
(15.13)
From this equation it is seen that the factors that diminish e.d. from its ideal
maximum (in this case zero I value) Q.Ll Vrev are similar to those that influence
P as a function of I. However, in this case, differentiating e.d. w.r.t. I gives simply
(15.14)
assuming Q does not decrease with I. It is interesting that this also has a maximum or minimum when
Eb/I=-R
(15.15)
Eb=-IR
(15.16)
or
428
Chapter 15
From the above relations for P and e.d., P can be plotted against e.d. for various
increasing values of I, giving theoretical Ragone plots for variously chosen values of the parameters (Fig. IS.4), as recently calculated in a paper by Pell and
Conway5, which quantitatively evaluates the relative roles of Tafel polarization
and "IR" drop.
In relation to the hook shape of most Ragone plots (Le., e.d. decreasing as
p.d. approaches elevated values), note that the eventual decline in p.d. with increasing I or I1A after its initial increase as I.!l.Vrev' given by Eq. (1S.5), is due
mainly to the term in I squared. That term increases with Ior IIA (depending on
the value of R) more rapidly than IEb In I1A or than (lIA)Eb In /lA, the transcendental term arising from Tafel-type polarization in Eq. (1S.6). See the analysis
given later in Eqs. (1S.31), (IS.32), and (IS.33). For the quasi-3-dimensional
materials constituting electrochemical capacitor or battery devices, the
current density factor, /lA, is more commonly and practically replaced by
the 3-dimensional current density, I per gram or per kilogram, of the active material, as noted earlier. Some further examples of Ragone plots for
practical systems are given in Chapter 20.
Power losses caused by polarization were considered in a general way in
Srinivasan and Bockris' monograph 6 on fuel cells, for which the power factor
is as important as it is for large electrochemical capacitors designed for battery
hybrid systems in electric vehicles having high power requirements. The general form of the power vs. Faradaic current drain relationship is shown in Fig.
IS.S for an electrochemical power source. A detailed analysis for a series of polarization conditions, including IR drops for 1 and 0.01 ohms (as exemplified in
the linearly scaled cell voltage vs. current plots) gave 6 the cell terminal voltage,
E, vs. current, I, profiles illustrated in Fig. IS.6. The kinetic and ohmic parameters used for these calculations are listed in Table IS.2. The exchange current
density, io, and the indicated (Table IS.2) anodic and cathodic transfer coefficients, aa and a c ' are the relevant kinetic parameters, together with the
diffusion-limited current densities, it. The terminal cell voltages are the
anode-to-cathode emfs remaining after subtraction of all polarization
voltage losses at each electrode. Similar considerations apply to a pair of
capacitor electrodes being discharged against each other.
A generalized set of polarization relations is shown in Fig. IS.7 where the
series of curves representing increasing contributions from the iR drop are plotted on reduced-variable scales: overpotential '1 act + iR expressed as '1 act +
(ioR) X (ilio) ('1 act = kinetic activation overvoltage); current density as ilio and
iR drop expressed as ioR. For zero R, the polarization is purely of the Tafel
type, so the relation is logarithmic in ilio, and curved on the linear ilio scale,
as shown in Fig. IS.7. Conversely, with increasing R, the overall polarization
becomes almost linear in ilio when the iR drop becomes dominant in '1. Arbitrary
'0'
\
\
\
\
.
.,00
11
10"'
... -.-.-.
b = 0.120, a' 0.60
b = 0.030, a =0.15
,
\
'.
"
\
,
\
n
1~
tOO
D..
10-'
,
\
.,
Energy'density x Ot (VI
\
\
.,
100
':"":"":"''-.
.,
.,
\
1~
"' D
\
\
............--
--~-...,..
b.0.030,a.0.~.
_._._'_"
....... -.-'
,
1~
b 0.120. a 0.60'.
1~
10-4'
:1;10-'
...
(d)
(b)
r.-;:-.....- .....---...............
-~to~'
10-'
~
~
l1t'
,0-<
Q.,01
~,o-1
<=
]!-
~to"
'w
...:
1,.
to'
larization according to Eq. (15.6) for a battery having .1.Vrev = 1 V, io = 10-5 A kg-I, for four sets of Tafel parameters as indicated on the curves. Reprinted from W. Pell and B. E. Conway,J. Power Sources, 63, 255 (1996),
with permission from Elsevier Science.
FIGURE 15.4. Examples of calculated Ragone plots showing effect of ohmic polarization for fixed Tafel po-
10'4'
10"1
t.,10'
itD-r
(c)
Ut1
Energy-density x Ot (V)
,,
b 0.120. a 0.60"
IQO~.----~----~----~~----~--~~--~~~
10-1
10-4'
,0->
-;;-'0-'
:J:
D..
10"1
1ito
..
<=
.....iP
1,.
(a)
lOr~-:--....-.....-........,.......,.---....---.......-
1/1
IX
o~
Ie
ID
iil
'<
cO
:J
ID
...~
[
!a-
1/1
f...
0.
II
:J
~
'<
:J
430
Chapter 15
1.0
,Ideal power-current /
relation without
/
'T] losses
/
:i:
"CV... 0.5
~
a..
/
,/
,- ,/
,/
,-
,-
,-
,-/
,,-
,-
,-
Moximum power
,- I
Power shutdown
due to diffusion
limitation
Cell current I A
FIGURE 15.5. General form of cell polarization as a function of current for an electrochemical
power source involving Faradaic, JR, and concentration polarization voltage losses. (From Bockris
and Srinivasan, 6 based on an analysis of fuel cell performance. Reproduced with permission of the
McGraw-Hill Companies.)
but reasonable values of the parameters io (10- 6 A cm- 2) and b, the Tafel slope,
dlJ/dlog i (= 0.1), were taken. Here, as before, io is the exchange current density.
Note that the behavior illustrated in Figs. 15.5 and 15.7 applies to an electrochemical power source involving a Faradaic process as in a battery, fuel cell,
or pseudocapacitance type of supercapacitor. For the double-layer type of capacitor device, the activation overvoltage (Tafel) component of polarization
will not normally be involved except on overcharge.
From the analysis by Bockris and Srinivasan 6 it again follows that power
losses depend on energy dissipation caused by ohmic loss (PR) and kinetic polarization ilJ where IJ = a + b In i generally, as noted earlier. The dependencies
of power production on current, corresponding to the polarization curves of Fig.
15.6 (conditions as in Table 15.2), are shown in Fig. 15.8. The differential variation of p.d. with current density will arise from the above two components of
polarization (see Eq. 15.6) as follows:
d(p.d.)/d i (ohmic) = -2iR
(15.17)
(15.18)
431
> 1.5-r---------------------.
.....
W
..J
<!
~ 1.0
b
a..
...J
...J
'."O- .a..
'x
" :c.............x
-,~~
0.5
-,
--G---o-n....
-u.
3 II
--..... ---1:.---6-___A- __t.-"t>n~
~
~-6
_A
W
U
1 ..........~
o
0.2
0.4
0.6
CURRENT/A
1.0
0.8
FIGURE 15.6. Plots of terminal cell potential E vs. current I showing effect of variation of Aio
at constant values of AiL and Ri (see Table 2 for assumed kinetic parameters). Solid lines are for Ri
=I ohm and broken lines are for Ri =0.01 ohm. (Based on calculations of Bockris and Srinivasan,
ref. 9)
1.0r----------:-----=----~-__:'-__,
"""0
......
-;; 0.8
0::
II
=0:: 0.6
0::
II
I='"
O~
__
10
40
60
80
100
ilio
FIGURE 15.7. Effect of combination of linear (ohmic) and exponential (Tafel) polarization contributions on generalized current density (i) vs. potential (17) relations for an electrode process.
Scales are in terms of reduced variables: overpotential scale 17 + iR (iR ioR x i/io) (see text) and
current density scale normalized to io as ilio. Tafel slope taken as b 0.118 V; i io exp [17lb]. For
ioR = a (lowest curve), the curve is a purely Tafel relation. linear on a log scale of ilio and curved
(as shown) on the linear scale. (Calculations by the author.)
432
Chapter 15
Aaio,a
10 3
>1
10-3
>1
10-3
(A)
2
3
4
Acio,c
(A)
AaiL,a
(A)
AciL,c
(A)
10 3
10-3
10-3
10-3
10-6
aa
ac
112
112
112
112
112
112
112
1/2
R;
(ohms)
1
1
0.01
0.01
1
aExamples taken from Bockris and Srinivasan, 6 based on analysis for a fuel cell power source. Subscripts a and
c indicate the quantities for anode and cathode reactions, respectively.
Ref. 6, p. 183.
where a and b are the parameters of the Tafel relation for the kinetics of the
power-producing electrochemical process.
It is seen that the rate of power loss due to the ohmic polarization effect (resistance of the electrolyte solution together with any intrinsic resistance of the
electrode matrix and/or current collectors) increases linearly with the current
density i while the kinetic polarization power loss increases at a rate that is a
constant, -(a + b), together with a logarithmic factor, -b In i, where b is the
0.8
0.6
I
I
.....
a::
W
3:
0.4
.0'
3:
oQ..
,..0-
~,c---c-~
I
I
I
I
I
I
0.2
I
I
0.2
0.4
0.6
CURRENT/A
0.8
1.0
FIGURE 15.8. Examples of power production vs. current curves in a cell subject to polarization
behavior according to the polarization curves of Fig. 15.6 and conditions listed in Table 15.2. Numbers on curves correspond to calculation numbers indicated in Table 15.2. (Based on calculations
by Bockris and Srinivasan. 6)
433
Tafel slope (smaller b values give a smaller rate of increase in power loss with
increasing i). With increasing i, the ohmic power loss will tend to increase more
rapidly than that caused by kinetic overvoltage effects, as was seen from Eq.
(15.6) earlier.
It can be seen that the ohmic polarization component of the rate of power
loss with increasing i is dominant when
2iR (a + b) + b In i or a + b (1 + In i)
(15.19)
(15.20)
iR = a + b In i
(15.21)
i -!?.In i = aiR
(15.22)
or
i.e.,
At small values of i, on the order of 10-4 _10- 6 A cm- 2 , the In term -(bIR)ln i
will be positive and dominate over the linear term in i, but at higher i values the
latter term becomes dominant. This corresponds to the form of Tafel plots at
high current densities (Fig. 15.7) when R is not insignificant. When R is small,
the (bIR) In i term is, of course, numerically dominant (i.e., kinetic polarization
principally determines the power loss), and vice versa when R is sufficiently
large. These conclusions are rather obvious.
15.2.4. Power and Energy Density Relationships for Capacitors
As has been emphasized earlier, the intrinsic energy density of electrochemical capacitor devices is substantially smaller than that of battery cells of
the same volume. Higher e.d. can be achieved by using nonaqueous electrolyte
double-layer capacitors in which the maximum voltage on charge can be up to
434
Chapter 15
3.5 to 4.0 V. Since the energy of a capacitor (see Chapter 2) charged to a plate
voltage difference of Ll V is 1I2CLlV2 , there is an obvious advantage in e.d. with
such nonaqueous embodiments. The power densities achievable are higher but
there are some disadvantages from the higher PR-drop factor in the power that
arises from the usually higher resistivity of nonaqueous electrolytes over that of
aqueous ones at comparable electrolyte concentrations.
Three factors distinguish the e.d. and p.d. characteristics of electrochemical capacitors from those of battery cells: (1) The capacitor voltage, Ll V, falls
continuously on discharge so the e.d. depends on the level of charge (i.e., the
corresponding Ll V) at which the e.d. is calculated; overall, the e.d. must be calculated by means of an integration of Q.d(Ll V) over the range of voltage on discharge, as with a battery cell having a declining voltage on discharge, e.g., with
Li-TiS2 or Li-MoS2 intercalation systems, or with Li ion cells. (2) Because of
the fall of Ll V with the extent of discharge of a capacitor device, the power density attainable from such a capacitor depends intrinsically on the fractional state
of discharge JQIQ since Ll V(JQ) = (oQIQ)/C. Thus, for a given controlled discharge current, I, the p.d. progressively declines during discharge. Usually the
initial power density at the beginning of discharge is referred to, i.e., when Ll V
has its highest value. With a capacitor system, constant current discharge is only
attainable with a progressively adjusted load resistance, diminishing in proportion to the falling Ll V of the capacitor device undergoing discharge.
In the above equations for e.d. and p.d., it is seen that the same polarization
terms, IR and Lb In I1A, enter into the relations for e.d. and p.d., diminishing the
electric potential difference, Ll V, from its maximum value, Ll Vrev , with increasing I. Hence, formally, a simple ratio, r, of p.d. (= I.Ll V) to e.d. (= Q.Ll V) can be
written (or equivalently with i) as
~=
r= I.LlVIQ.LlV= I1Q
e.d.
(15.23)
(15.24)
Thus the relation between p.d. and e.d. is, in the simplest analysis, just the rate
factor, lILlt for the discharge, as might be intuitively apparent; the shorter the
duration of discharge for a given Q, the greater the power delivered in that discharge. This ratio, it is seen, is actually independent of the polarization factors
that diminish both p.d. and e.d. individually but equivalently in the same way,
at least for the ideal simple conditions considered above.
The third factor that characterizes the behavior of capacitor energy storage
devices is that the internal (equivalent series) resistance is not a simple series re-
435
sistance but a distributed one, as discussed and illustrated in Chapter 14. Thus
power is not simply defined by I multiplied by the potential difference between
the plates (at a given level of discharge) minus the IR drop since the effective
value of the latter quantity depends on the rate of discharge. At high rates, especially in pulse regimes, only the outer regions of the porous matrix, bearing
the double-layer charge, are accessed, while at low rates most of the whole of
the double-layer charge, including that from inner surface elements of the matrix, can be accessed. Thus the quantification of the power density of carbontype double-layer capacitors is more complex than that for most battery
systems. However, some similarities exist since a distributed pore and matrix resistance also arises in a number of battery systems (e.g., the Zn-Mn02 primary
battery) and in Li-SOCI 2 cells, which employ porous electrode materials. Then
the distribution of ohmic resistance and double-layer capacitance is equivalent
to that for the porous electrode case treated by de Levie and co-workers 7 as outlined in Chapter 14. Similar considerations apply to most fuel cells. 6
A further point with double-layer capacitors is that the kinetic polarization
terms ILb In IIA do not appear in the p.d. or e.d. equations in most cases, leaving
only the IR drop term to diminish available voltage on load. However, as mentioned earlier, the IR drop is itself a complex quantity since the effective R depends on the time scale of discharge or of potential modulation in the case of
evaluation of such capacitor devices by impedance spectroscopy over a wide
range of frequencies.
Because in purely electrostatic charge accommodation there are ideally no
kinetic limitations to p.d., double-layer capacitor devices are perceived to provide high power densities, substantially larger than those available from most
battery-type systems. However, as discussed in Chapter 14, there are unavoidable power limitations of other kinds associated with the complex distributed
series resistance of pore structures and the electrolytic pathways within them.
These give rise to a power spectrum within which only a fraction of the charge
is accessible at high power levels, or with short-period pulses or high-frequency
av modulation.
A different kind of kinetic limitation (of a non-Faradaic kind) can, however, arise in microporous double-layer capacitors having large real areas of the
charge-accommodating material. When large distributed areas of porous electrodes are charged, there is the requirement that appreciable quantities of the
ions (anions and cations) of the electrolyte must become separated and migrate,
respectively, to the surfaces constituting the positively and the negatively polarized electrodes of the device. This can involve a kinetic limitation associated
with the ionic mobilities (equivalent conductances) of the anions and cations. A
net separation of anion and cation charges between the two capacitor electrodes
must take place since a double-layer capacitor is constituted of two double layers distributed over the large real areas of two separated high-area electrodes im-
436
Chapter 15
mersed in an electrolyte and divided by a separator interface (actually two interfaces). As a consequence of this situation, depending on the concentration of
the electrolyte in relation to the required maximum ionic charge accommodation (determined by c./). V) at the electrode matrix interfaces, the above effect
of ion separation can lead to a significant increase in the esr of the capacitor device upon charge, as was pointed out in a recent paper8 (see Section 15.3).
In the case of oxide-type redox pseudocapacitance systems (e.g., RU02 and
Mo0 3), the discharge processes involve (see Chapter 11) electron transfer to
Ru4+ or even some Ru 6+, through Ru 3+, probably to Ru 2+ (but not down to Ruo
metal, as in the reduction of oxides at Pt or Au), coupled with proton transfer in
the hydrous oxide matrix to 0 2-, giving OH- sites and producing 0 2- sites from
OH- sites at the originally negative electrode of the pair. Thus proton migration
from the negative electrode to the positive (RU02) one is required on discharge.
In the normally used aqueous H 2 S04 electrolyte, there can also be HS04 migration in the opposite direction, depending on the transference numbers of cation and anion in the system or whether an ion-exchange membrane separator is
employed. These ion migration processes in a hydrous oxide material can lead
to power limitation at high current drains into low resistance loads. Note
that the proton transfer involved will probably follow the nonclassical,
quantum-mechanical proton-jumping process and water-structure rearrangement treated in papers by Bernal and Fowler9 and in more detail by
Conway, Bockris, and Linton.lO In that mechanism, the proton transference number is about 0.8 to 0.85 in free aqueous H 2S0 4 electrolyte at
pH < 1.
(15.25)
437
for a given constant discharge current i (= dq/dt = C dV/dt), and, for C invariant
with potential, dV/dt = constant for a given fixed i. Here C is in units of farads
cm- 3 .
In another way, we can write the charge remaining on the capacitor at a
time t beyond some initial condition (i) when the charge was qi as
(15.26)
The corresponding potential Vt is q/C, i.e.
Vt = (qi -
J i.dt)/C
(15.27)
so that the power density at time t after the start of discharge at i A cm- 2 will be
t
(P.d)t = i (qi -
J i dt)/C
(15.28)
(p.d)t = Vi -
if
(15.29)
~ dt
= i (Vi -
(15.30)
J~ dt)
o
i
which eventually becomes zero when fb -c dt =q/C == Vi.
Alternatively, a capacitor device could be discharged under conditions of
constant power, iV = constant. For such a situation, as charge is being withdrawn and V declining, it would be necessary to arrange instrumentally that i
will be increasing in order to maintain the product iV constant, K. Then i =
K V-i so that the condition
di/dV = _K(V)-2
(15.31)
438
Chapter 15
(15.32)
(15.33)
i.e., i increases as V decreases, which is the basis ofthis case; see Section 15.4.3.
These relationships show that evaluation of power density for a capacitor is not
as simple as for an ideal battery.
In addition, as for a battery, the p.d. will decline (at a given degree of
charge, qt, held by the capacitor) with increasing discharge current density when
there are ohmic polarization losses. This will certainly be the case with porous,
high-area capacitor electrodes, especially when less conductive nonaqueous
electrolytes are employed, as in the higher operating voltage double-layer capacitors. In the case of electrochemical capacitors utilizing redox or adsorption
pseudocapacitance, additional Faradaic (Tafel-type) polarization losses will occur, as discussed earlier for the case of p.d. vs. e.d. relations (Ragone plots) for
batteries. As with battery systems, the rated power density of capacitors will depend on the current drain at which the rating is evaluated; it will also depend on
the state of charge in the capacitor case, as explained above, much more so than
for most battery systems except for Li intercalation types, where quasi- or
pseudocapacitor behavior can occur (Chapter 10).
An important difference between capacitor and battery power rating is the
relation between p.d. and e.d. for these two types of electrical energy storage
systems. We have seen from Eqs. (15.23) and (15.24) that there is a direct relation between p.d. (iV) and e.d. (qV) for a given initial cell voltage since for both
cases the factor by which the cell voltage is diminished by polarization effects
with increasing current density is at least theoretically the same. However, in the
case of capacitors, the density of stored energy (for a given capacitance, C)
scales with the square of the voltage attained at the end of charge (Chapter 2)
while the power density remains linear in that voltage, although for both properties the voltage can be controllably raised (unlike the thermodynamic voltage
of a battery cell) up to or just below the practical decomposition limits for the
potential in an electrolyte solution, i.e., ca. 1.3 V for aqueous solutions or ca. 3.5
V for nonaqueous aprotic solvent solutions. This statement implies that the ratio
of e.d. to p.d. for a given capacitor is a function of potential on charge, i.e., this
ratio increases toward completion of charging, which is not normally the case
for battery cells except for the Li intercalation type and to some extent for
NiOOH-Zn or NiOOH-Cd battery cells. Some examples of Ragone plots calculated5 for capacitors are shown in Fig. 15.9.
Power densities attainable with electrochemical capacitors vary considerably, depending on the constituent materials and electrode and current-collector
design, but state-of-the-art figures are now in the range of 5 to 15 kW kg-I,
439
104
10'
----"\
102
b
....
10'
Z!'
iii
c
100
~
CD
10"
~
CD
-=0
tI..
..........
R-a.t.O - - - -
R-'.D.O - - - _
R-fa.o.O ..
'.
10'2
C-a.UFkf'I
10"
10"
10'2
10"
100
10'
which is substantially larger than figures being quoted 4 or 5 years ago. Corresponding energy densities are in the range of 8 to 20 Wh kg-I. Some advanced mixed-oxide capacitors are claimed to provide p.d.s on the order of
1000 W cm- 3 ! However, this figure seems unrealistic.
Recent performance figures from Maxwell Corp. are 3-10 kW kg-I, although some other claims in the literature go as high as 15 kW kg-I. Using derivatized poly thiophene conducting polymers, the Los Alamos National
Laboratory groupll has achieved a p.d. of 35 kW kg-I. For electric vehicle applications in hybrid configurations with batteries (Chapter 19), requirements for
energy density projected by the automobile industry are currently 16 Wh kg-I.
It is interesting to examine this requirement in terms of absolute electron
density per atom, assuming a carbon-based capacitor; 16 Wh kg-I is equivalent
to 16 x 3600 C kg-I at 1 V operating potential. For a 1000-m2 g-I (= 1010 cm2
kg-I) porous carbon powder capacitor electrode, this energy density at 1 V is
equivalent to a charge storage of (5.76 x 104/105 ) x 6 X lOz3 electrons per 1010
cm2, or to (5.76 x 104/10 5) x 6 X 1023/10 10 electrons per cmz, i.e., about 0.035
electrons per atom, taking the real surface of carbon to contain about 1015 atoms
per cm- z.
440
Chapter 15
Since the specific maximum charge at, say, 1 V corresponding to a doublelayer capacity of about 30 flF cm- 2 is about 0.18 e atom- 1 (see Chapter 6), it is
seen that the above figure for an e.d. of 16 Wh kg- 1 corresponds to utilization
of only about 20% the surface. This, however, is not an unreasonable figure
since by no means all of the surface of a porous, high-area electrode with appreciable tortuosity is electrochemically accessible to the electrolyte solution unless a repeated vacuum-filling procedure is used. Even then, the total specific
real dry (BET) area is not accessible because of surface tension and contact angle factors.
Summarizing, an absolute evaluation of projected power density is not
practicable because it depends on the operating current density attained and
overvoltage vs. current density polarization relationships, as well as the depth
of discharge (dod) in the case of capacitor energy storage. In the latter case, the
dod factor arises for the fundamental reasons discussed earlier and distinguishes
p.d. relations for capacitors from those for most battery systems.
441
...J
<t
i=
z
lJJ
0..
0::
W
LOG [i 0)
I
I
LOG [CURRENT-DENSITY J
FIGURE 15.10. Onset of concentration polarization in a Tafel relation at high current densities
(schematic). Effect of stirring also shown (see also Chapter 3).
RT
Yf c = zF In(cz/c e)
(15.35)
iri
= 1 - iii/=- -.tl
(15.36)
This equation is an important basis for the development of equations for the
electroanalytical technique of polarography. Thus, as i ~ i l (the onset of full
442
Chapter 15
z
o
~
a::
I-
~
~
<t
a:
DISTANCE FROM ELECTRODE
SURFACE PLANE (order of fOO}L)
FIGURE 15.11. Profiles of concentration change across the diffusion layer at an electrode as concentration polarization sets in (schematic). Initially, Ce = Cb; later as i ~ it, Ce ~ 0 (see Eq. 15.30).
concentration polarization), cecb or Ce ~ O. Then 11 c can approach theoretically large values. Normally some other process (e.g., decomposition of the solvent) then takes over.
With regard to power consumption, the onset of concentration polarization, with i limited to a maximum value i /, leads to power shutdown. In cases
where solid oxide electrode reagents are involved (e.g., in nickel oxide batteries
or oxide pseudocapacitors), concentration gradients of protons can be set up and
lead to some analogous concentration polarization.
Concentration polarization of another type can arise in charging of doublelayer capacitors where anions and cations of the electrolyte have to be separated
from one another and concentrated, respectively, at the interfaces of the positive
and negative high-area, porous carbon electrodes. This can lead to increases of
internal resistance on charge, as discussed earlier in this chapter.
A different kind of concentration polarization can arise also in the Faradaic
charging or discharging of solid oxide electrodes, as with Ru02 pseudocapacitors. Here gradients of the oxidation state of the ions of the oxide material (coupled with proton diffusion) can develop upon fast charging and vice versa on
discharge. On returning the electrodes to an open-circuit state, relaxation of the
oxidation-state gradients takes place and further addition or removal of charge
(at a given potential) can then be achieved. Examples of this behavior are shown
in Chapter 18.
443
Formally, according to Ref. 12, the C-rate for discharge or recharge is defined by the expression 1= fx Cn where I is the current in amperes, C is the numerical value of the rated charge capacity of a cell or battery in Ampere-hours,
n = C-rate at which the system was rated, andfis a multiple or fraction of C. As
an example, the 0.1 Cor ClIO discharge rate for a battery rated at 5 Ah capacity
(18,000 C) is 0.5 A. Conversely, a 250-rnAh cell, discharged at a current of 50
rnA, would be discharging at a 0.2 C or CIS rate. Following the discussion in
Ref. 12, this is calculated according to the following definition:
A further example for clarification of this procedure for defining the rates
of discharge and the power of operation on discharge or recharge is as follows:
the designation for a CII 0 rate for a battery rated at 5 Ah charge capacity at a
CIS or 0.2 C rate is 0.1 CO.2. In this case, the ClIO rate is equivalent to 0.5 A or
500rnA.12
Another procedure for specifying the current is through the hourly rate.
This is defined as the current at which the battery will become completely discharged over a period of a specified number of hours. For a given selected number of hours for complete discharge, the current will obviously be larger, ideally
in proportion to the charge capacity, but lower for a larger number of hours chosen for discharge.
It must be noted generally that the practical or available ampere-hour capacity usually becomes smaller at higher discharge rates or power drains. The
decrease of capacity with increasing discharge rate is a complex phenomenon
dependent on the type of active materials and the construction of the battery. It
also usually depends on the cycle history of the battery and the temperature. The
effect occurs on account of materials isolation during high-rate discharge, development of internal gradients of electrolyte concentration, etc., and is difficult
to formulate quantitatively, unlike the theoretical open-circuit materials capacity; i.e., excluding weights of current collectors, cell casings, terminal blocks,
etc. If the discharge is conducted intermittently, the realizable total capacity is
usually larger than that attainable with continuous discharge at the same current.
For battery-powered systems, a constant power mode of discharge is becoming more commonly used. A method analogous to the C-rate is employed
to express the charge or discharge rate in terms of power. Thus (see Eq. 15.70)
P =M x En where P is the power in watts, E is the numerical value of the rated
444
Chapter 15
energy of the cell in W aU-hours, n is the E rate at which the cell was rated, and
M is again some fraction or multiple of E. Thus the power level at 0.5 EO.2 or
EO.212 for a cell rated as 1200 mWh of capacity at the 0.2 E or EI5 rate is 600
mW.
It was mentioned that the practical charge capacity is usually substantially
less than the theoretical capacity of the actual electroactive materials calculated
according to the open-circuit thermodynamic potential difference between the
anode and the cathode together with the number of gram-equivalents, i.e., the
number of faradays deliverable per mole on discharge. A histogram illustrating
this difference for a number of anode-cathode combinations is shown in Fig.
15.12.
445
1000
Level
Level
100
a.
.1&&!..8
FIGURE 15.12. Theoretical and practical charge-capacity values for battery systems shown as
histograms. 1, Leclanche dry cell; 2, alkaline-Mn02; 3, magnesium-cuprous chloride; 4, zinc-mercuric oxide; 5, Li-Mn02; 6, lithium-S02; 7, magnesium-cuprous chloride; 8, lead-acid; 9, nickelcadmium; 10, zinc-silver oxide; II, Li-Mn02. Level A, theoretical capacity (active materials only);
level B, theoretical capacity of a practical battery; level C, actual capacity. (From Linden. 12 Reproduced with permission of the McGraw-HilI Companies.)
This matter therefore revolves around how current density is to be expressed. If it is in terms of bulk quantity or volume of active material (as for a
carbon-type double-layer capacitor), then C-rates can be expressed as numbers
that are independent of the mass of material. If current densities normalized per
gram or per kilogram of active material are used, then C-rates are not dependent
on the masses or numbers of equivalents of materials that are involved. (This
seems to be the basis of the usual convention.) However, if the discharge proc-
446
Chapter 15
(15.37)
and
.
Q(AH)
1
.
= - h - x Tg = Q (A g-l), then p.d. = l x V = Q x V watts g-l, (15.38)
1=
1OQ(Ah)
h
= 10 Q (A), then p.d. = 10 Q x V watts;
(15.39)
and
.
IOQ
10 g
l=~Ah) x--= Q
i x V 10QV
(A g-l), thenP=--=--= Qx V watts g-l
10 g
10 g
(15.40)
447
These examples show again that on an absolute basis, different power values
arise for a given C-rate, depending on the quantity of electrochemical reactive
material. On a bulk per gram basis, the C-rate corresponds to a particular specific power, i.e., a power density related to a current density expressed on a bulk
per gram basis, I g-l, == i. This latter representation of power density in terms of
C-rate is preferred.
In practice, C (the rated capacity) is the charge determined at a specified
current (time to complete discharge down to a defined cut-off potential); unlike
an ideal situation, V will not normally remain constant and depends on the state
of charge and current density. Also, the practically available charge, Q, is usually less than the expected theoretical value according to Faraday's laws (Fig.
15.12). The cell voltage is normally a function of the state of the charge (as is
that of a capacitor for more fundamental reasons) and current density, and includes terms related to iR drop and other sources of polarization as discussed in
Section 15.1.
The activation overvoltage contribution is usually expressed in terms of
current density on a per square centimeter basis. For a 1-g sample of reagent,
nominally in a cubic form, and a density of 1 g cm- 3 , the surface area would be
cm2 ; ~or a corresponding lO-g sample, the surface area would be 27.8 cm 2
({10 x ill x 6). Then the external-surface current densities would be, respectively, for the
(15.41)
. IOQ(A)/1O g _
-2
lO-g sample. 27.8/10 g - 0.36 Q (A cm ).
(15.42)
448
Chapter 15
terials, however, the real surface area in m2g- 1 of the starting material(s) is usually known, at least approximately, so a real current density, expressed per
square centimeter, could ideally be stated for some defined operating conditions. Even then, though, a real current density evaluation is elusive since by no
means all of the pores of such porous matrices are accessible to the electrolyte
and thus to electric current (Chapter 14). The same limitation also applies to a
number of battery systems (e.g., Ni-metal hydride) in which powder-material technology is involved, and also to fuel cells where three-phase, gas/liquid/metal interfaces occur.6
The general situation is that the overall density of current, i, expressed on
a bulk per-gram basis for the active material, will not usually scale with the real
current density on a per square centimeter basis, especially with bulky, densely
packed electrode structures, since the fraction of available active area that is effective for passage of current will depend on current density in the case of porous electrodes. The conclusion here is then that the relation between rates of
discharge (or recharge) expressed as C-rates and power levels depends on how
the involved current densities are expressed, namely, per square centimeter of
reactive surface area (or that available for double-layer charging) of the reactive
material, or per gram or kilogram of the active material, or per cubic centimeter
or meter.
In an interesting paper on optimizing energy and power density requirements for a given system, Miller4 has discussed conditions that are met by using
the smallest and lightest device for a given constant power load. As an example
to illustrate the principles involved,13 he considers the nominal hypothetical
Ragone plot shown in Fig. 15.13, as follows: say that the smallest energy storage
device is required that can deliver 25 units of energy (along the axis of ordinates
in Fig. 15.14, a log scale) at a nominal power level of 25 on the axis of abscissae (also on a log scale; the units are arbitrary). Selecting point b in Fig.
15.14 (e.d. = 3, p.d. = 3) means that a device of size 8.33 mass units would
exactly meet the required energy and power needs. On the other hand, for a
system requiring 25 units of energy at a power level of 5 (e.d. = 5, p.d. = 1), the
optimum operating point is at a, where an energy storage unit of 5 mass units
meets the requirements. Note that this latter case is in a regime of the Ragone
plot where e.d. is almost constant with diminishing p.d., i.e., there is little polarization influencing the performance of the device. The opposite arises typi-
449
100 r - - - - - - - - - - - - - - - - - - - - ,
~
(,!)
a::
w
Z
BATTERY
10--r-----.:;-~-
HYBRID
- - - - - - - - - - - - - - __ _
"
'
\
I
CAPACITOR
~
(,!)
9 0.1
+------'---,.-----+------r----~
10
100
1000
10000
FIGURE 15.13. Hypothetical Ragone working diagram to illustrate the optimization of system
requirements for a capacitor-battery hybrid (schematic). Lower curve for capacitor. upper for battery component. (From Miller 4 )
10
,.--.
>-
<.!)
a::
w
z
w
u
G: , r
u
IJJ
Cl.
~
<.!)
0.1
0.1
I
1
r
10
450
Chapter 15
cally below point c where any further attempt to obtain higher power density
leads to a rapid roll-off of energy density.
When it is desired to determine the minimum size of an energy storage system to support a constant power load, it is necessary to locate on the Ragone
curve the point that has the required energy-to-power ratio (Eq. 15.17). The
Ragone plots in Fig. 15.15, based on data from Murphy and Davis, 14 give performance information on various types of capacitor and battery systems. One
proposed energy-to-power ratio for an electric vehicle has been given as 1.4
MJ/40 kW, i.e., a 36-s time response. Then the optimum operating points (see
Figs. 15.14 and 15.15) are found as the coordinates of the intersection points of
the 35-s time response discharge line with the respective Ragone plot for the desired capacitor system.
100
I
I
r---
Max power
NiCd
I
Ni~d
'"---- t--
Pb-acid -I"-
.-.../
IMax power
Pb-acid
10
odE
1/
"
--
I~
'
Maxwell
3Vorganic
-.;..
..... 1
Maxwell
24Vorganic
I Semi-Pac;aged
"I~
Maxwel~b
1/
~ear-te~m g~,
'/
r-,
1Vaqueous
r-
~
I
'"Pinnacle
28V aqueous /
.1
10
100
Wlkg
1000
iI
10000
FIGURE 15.15. Ragone plots for various electrochemical capacitor and two battery systems. A
time response of 36 s for electric vehicle operation is represented by the 45 straight line, intersections of which with the Ragone relations (see also Fig. 15.13) provide the respective points for optimum operation of each type of system. (From Miller,4 based on data of Murphy and Davis. 14)
451
Miller4 has also treated the problem of optimizing the hybridization of a capacitor with a battery, the essence of contemporary approaches for development
of electric vehicle power systems. 15
Figure 15.13 shows hypothetical Ragone plots for a battery and a capacitor. The plot for the battery represents good energy density, much larger than
that of the capacitor, but with the latter capable of delivering energy at much
larger power densities. The aim of combining the capacitor with the battery is
to extend the energy-power performance rating of the system to that represented
by the dashed line in Fig. 15.13. This has the effect of allowing the battery's energy to be delivered at the capacitor's (higher) rate.
The capacitor must, however, be adequately decoupled from the battery in
order to achieve the possible benefits of this combination. This requirement exists because the battery's voltage during discharge varies much less than that of
the capacitor, whose voltage must diminish almost linearly with the state of discharge (V =qlC), as discussed in Section 15.2.5. The required decoupling can
be achieved by dc-dc converters or resistors; in some cases just the esr of the battery itself suffices. 4
A significant improvement in battery systems other than those for electric
vehicles can be achieved, e.g., for improvement of the operating lifetime of battery power sources for a telephone. As an example, a 1.3-F capacitor connected
in parallel with an Li ion battery achieved a fivefold improvement in operating
lifetime.
A model calculation, also by MiIler,4 showed that a capacitor having a
volumetric energy density as small as 0.1 % of that of the coupled battery enabled the operating time of a phone (as load) to attain a maximum (Fig. 15.16)
when about 20% of the energy storage volume was used to accommodate the
capacitor component. With less capacitor volume, power deficiencies occur, so
operating time is diminished. The same trend occurs on the other side of the
maximum: when the capacitor volume is larger; then in the combined system
there is depletion of energy delivery capability. This approach by Miller provides a useful basis for optimization analyses of the behavior of capacitor-battery
hybrid systems that will have increasing importance in future technologies.
The effect of hybridizing rechargeable, high-performance batteries with
Polystar's "Aerocapacitor" was demonstrated by Juzkow, Gold, and Kraemer 17
in a series of Ragone plots. Such diagrams showed that for hybridization with
an Li ion, an Ni-MH, and an Ni-Cd battery, the inclusion of a coupled electrochemical capacitor has a clear and major influence in allowing high-power
drains to be achieved without the usual corresponding diminution of energy
density as a result of polarization effects. It provides a clear demonstration of
the value of hybridization of capacitors with batteries to level load and obtain
power performance much improved over individual battery systems.
FIGURE 15.16. Operating time for discharge from 4.1 to 3.0 V of an organic electrolyte capacitor
vs. fractional volume of capacitor as exemplified by use in a digital telephone unit. (From Miller. 4)
Miller4 has shown how conditions for maximum power delivery can be derived. A simple series RC circuit shown below is considered:
----~
1~lc-------
where Rs is a series resistance (e.g., an esr) and any parallel leakage resistance
is infinite. A discharge current of I will lower the initial potential Vi to a value
V= Vi-IRs
(15.43)
(15.44)
dPldl = 0 = Vi - 2lRs
(15.45)
453
i.e.
Imax.p
= V;l2Rs
(15.46)
VmaxP = V-iRs = Vi -
[2~s) Rs = V;l2
(15.47)
= V?14Rs
(15.48)
ImaxP
When the maximum deliverable power is Pmax, i.e., Vi 2/4Rs, then the potential is V;l2 and the current through Rs, by Ohm's law, is (V/2)IR s' The power
delivered to a load resistor by such a series RC circuit that is initially charged
to its working voltage is given4 by
P = 4Pmax/(1 + r)2 exp[-2r/(1 + r)]
(15.49)
where P max is as given above, r = tiRsC, i.e., time divided by the time constant,
RsC, of the circuit and r is the ratio RdRs of the load resistance to the esr, (when
RL = Rs, RL is referred to as a "matching load" but Rs is usually RL for a good
capacitor). For this condition, RL = Rs, the voltage across the capacitor would
immediately drop from Vi to V;l2 when discharge was commenced; the delivered energy would then be zero for any voltage limit lower than V;l2.
Some examples calculated by Miller4 are illustrated in Fig. 15.17, where
the normalized power P (through load RL)IP max is plotted against reduced time,
r. The end point of each line corresponds to the time to reach a voltage across
the plates of V/2; the discharge time for a matched load is zero, as indicated
above. From Fig. 15.17 it follows, e.g., that for RL = 5 x Rs, the peak power
would be 0.11 times the maximum power, Pmax' The capacitor voltage V/2
would be reached after time r = 3.1. Thus, the initial power is 11 % of the maximum power level. The final power when a voltage of V/2 has been attained is
4% of P max , and the discharge lasts 3.1 time constants, r.4
The delivered energy, EdeJ> at a specified (average) power level of discharge, P av for a series RC circuit can be derived. Figure 15.18, which is also
based on the calculations by Miller,4 shows the result for a constant-current
discharge, where energy and power are expressed as normalized quantities
Ede/Etotal and Pa/P max . The initial voltage is again taken as Vi and declines
during the constant current discharge to V/2. Figure 15.18 is, in fact, a generalized Ragone plot applicable to a capacitor of any type if its behavior can
be represented by that of a series RC circuit. The value of the delivered energy,
EdeJ> at low power levels is 0.75 EtotaJ> as referred to earlier, and is the energy
available from discharge over the potential range Vi to V;l2. At high power levels, the delivered energy approaches zero, a characteristic commonly illustrated
454
Chapter 15
r=
,.......,
S
E
0.1
0......
C>
ooJ
O.Ol;------,r-----.------,~--__;
.=t/RsC
FIGURE 15.17. Normalized power (PIPmax) plotted against the time constant T = tl RsC. The parameter r is the ratio of load resistance (RL) to series (esr) resistance Rs. (From Miller. 4)
(5
W
.......
Q)
0.1
"0
0.01~----------------~-----------------"
0.01
0.1
P ave/ P max
FIGURE 15.18. Delivered energy (relative to total energy) vs. normalized average power
(PavIPmax ) for a series RC circuit discharged at constant current from an initial voltage Vi to V;l2.
(From Miller. 4)
455
0.2
1.1
0.44
174Sl
1.5
--'J ~F
---L-_ _ _-'--_ _
0.09
0.8
"ii 0.6
"0
W
......
<II
"0
0.4
0.2
0.2
0.4
0.6
0.8
P ave/Pmax
FIGURE 15.19. Dissipated energy, relative to delivered energy, in a series RC circuit discharfed
at constant current as a function of normalized average power (as in Fig. 15.18). (From Miller. )
456
Chapter 15
------ - - - ----
--- -
.... "-
--
\
\
\
\
I
0.1
"-
Qj
"0
P ave/Pmax
FIGURE 15.20. Ragone-type plot for reduced delivered energy vs. reduced average power for the
five-unit distributed Rand C network shown above. The behavior of the series RC circuit is shown
by the dashed line. (From Miller. 4)
457
0.8
Qj
-0
0.6
......
II)
-0
0.4
0.2
0.2
0.4
0.6
0.8
P ave/ P max
FIGURE 15.21. Relative dissipated energy curve vs. reduced power for the five-unit distributed
Rand C equivalent circuit. The dashed line is the behavior of the series RC circuit (cf. Fig. 15.20).
(From Miller. 4)
3. Constant power: The current during the discharge increases as the battery or capacitor voltage decreases, thus discharging the system at a
constant power level (power = current x voltage) (see Section 15.4.3).
As illustrated usefully in the monograph by Linden,12 the discharge current
as a function of discharge time can be plotted and compared for the various discharge modes as shown in Fig. 15.22. For this figure it is assumed that at the
commencement of discharge, the discharge current has a common value for all
three modes. Figure 22(b) shows the variation of the cell voltage for the three
modes. As the cell voltage declines during discharge, the current in the case of
mode 1 corresponds to the drop in cell voltage as determined by Ohm's law:
I = VIR [Fig. 15.22(a)]. For mode 2, the discharge time or service life is
lower because the average current is larger. For mode 3, the current increases with decreasing voltage to maintain IV = constant according to I =
PIV. Figure 15.22(c) is a plot of the power level for each of the above modes
of discharge.
For each of the above conditions, the ampere-hour capacity (to a given cutoff voltage) delivered by the battery is the same. The current varies, depending
on the mode of discharge. The delivered service-hour performance, therefore,
is inversely proportional to the average current.
The same relationships can be plotted,12 but with the assumption that the
average current (to a given end voltage) is the same for all three modes of dis-
458
Chapter 15
a)
Constant power :
....C....onstant
curre~~'),..- .. )..;
~~~----~-~~~~~
0'------------"'Discharge t i m e -
t '.
b)
i . .~-~-:.~.:..::.::::.'-. .:>. :
'0 Conslonl currenl
>
Q)
c)
J.constont power
t. .
Conslont resislance
Discharge time_
....
'~....
onslonl resistance
I
I
Constant power I
~~--
,,-,,-,,-
-'~::-"-"I
I
I
Discharge timeFIGURE 15.22. Discharge profiles under different discharge modes for the same average current:
(a) current profile during discharge, (b) voltage profile, and (c) power profile. (Based on plots
shown by Linden 12 ; schematic.)
charge. Under such a condition, the service or discharge time is about the same
for each mode. From an application point of view, the most realistic case is the
assumption that the power under all three modes of operation is the same at the
end of the discharge.
On a scale of power density plotted against the logarithm of the current
density, the falloff of p.d. with increasing current density is very sharp. This
happens, of course, when the sum of the polarization terms becomes comparable
with, but still smaller than, the iV product for high i values. On a linear scale of
i values, the drop off of p.d. looks less abrupt. When the polarization terms become equal to iV, a power shutdown occurs since there is effectively no voltage
available in the system; a kind of short-circuit current still passes but at a voltage
V ~ 0; hence power formally becomes zero.
Note that for a battery cell containing two electrodes, the Tafel polarization
term, i(a + b log i), in the p.d. equation is composed of two terms, one for each
cell half-reaction; the a and b Tafel parameters are not normally identical for
such a combination, and one half-cell process may also be subject to concentra-
459
tion polarization more than the other, depending on the electroactive reagents,
their concentrations, and the electrode-process mechanisms that obtain.
15.5.4. Constant Power Discharge Regime for a Capacitor
Today, the common procedure for rating the power density of electrochemical sources is by means of the constant power discharge mode, referred to
earlier. It is often more useful in practical evaluations of energy storage devices,
especially when the capability to deliver power is an issue. Usually use of this
mode means that as the power source becomes drained of charge and its voltage
drops, the current must be increased to maintain IV constant. In the case of batteries, the decline of voltage is a secondary effect and may be caused by conversions of the electro active materials on discharge, accompanied by phase
changes. In the capacitor case, a similar voltage decline occurs on discharge but
for primary reasons, i.e., because C = !J.QI!J. V or !J. V = !J.QIC, as discussed in Section 15.2.5 and earlier.
With capacitors, however, any test procedure must be accompanied by a
decline in voltage across the device during whatever mode of discharge is the
basis of the test. In a corresponding test of a battery, a decline in voltage on discharge (although often significant) is not a necessary requirement; for an ideal
two-phase battery electrode system, such a decline in voltage is zero, except for
intercalation processes involved in (e.g., Li-TiSz, Li-Co0 2) rechargeable cells
where a voltage decline on discharge is an intrinsic aspect of their thermodynamic behavior (see Chapters 2 and 10), like that of a capacitor.
For constant power discharge of a capacitor, the voltage must decrease
with time t, but the power product I X V must remain constant, say, K; i.e.,
IV=K
(15.50)
1= CdVldt
(15.51)
and therefore must increase as V declines (dVldt negative) during the discharge.
This situation therefore differs from the constant power discharge of an ideal
battery. The power will be given by
I x V =-CVdVldt =K
(15.52)
-CVdV=Kdt
(15.53)
or
460
Chapter 15
Integrating
CV2
--=-Kt+c
2
(15.54)
CV[
(15.55)
c=-2
Then
(15.56)
so that
(Vi> V)
(15.57)
is the condition for the variation of potential in time for constant power discharge of a capacitor, the specific capacitance of which has been assumed to be
independent of V, an ideal case. It is seen from the above final result that for a
given constant IV power product (= K), the variation of (Vr - V2) with time
scales with 2KIC, i.e., at larger constant power drains, a shorter time scale for a
given C value will obviously be required to discharge the capacitor at that constant K, compared with a smaller K rate.
It is seen that Eq. (15.57) can be rewritten to represent the difference of energy stored initially at Vi and that at V, after some time t, divided by that time.
This difference is an overall power quantity (energy divided by time for delivery), represented by the selected K rate.
The variation of current with time must be the inverse of the variation of
potential for constant power discharge conditions, i.e.,
dJldt = -dVldt
(15.58)
and
d(JV)
= 0 = JdV + VdJ
(15.59)
461
of discharge through a load leakage resistance when V ---7 0 only as t ---7 00 (see
Chapter 18).
The cutoff time needed for the potential V across the capacitor to reach the
cutoff value of zero at a constant power drain of K watts is easily found from
Eq. (15.56). Thus, for V = 0:
(l/2)CVr = Kt (V = 0)
(15.60)
=0) =(l/2)CVr IK
(15.61)
i.e.,
t (V
=G/K
(15.62)
where Gi is the initial (electrostatic) Gibbs energy stored by the charged capacitor. As expected, t(V = 0) becomes smaller as K, the constant power drain, becomes larger.
In practice, a cut-off time on a test may be chosen for V to be above zero
but it should be some limit V less than Vi, namely, Vi > V> 0, commonly V =
0.5 Vi. The cut-off time for complete discharge is obviously the time corresponding to exhaustion of all the initially stored (l/2)CVr free energy, consumed at a constant power drain of IV watts (see Eqs. 15.52 and 15.61).
Related to the analysis given above for the power density of capacitors,
Rudge et al. ll made an empirical power-density evaluation of a poly-3-(4fluorophenyl) thiophene (PFPT) conducting polymer-based pseudocapacitor by
dividing the experimentally determined energy density of the system by the determined RC time constant (= 4 s) calculated from the results of charging and
discharging a pair of electrodes made from the above material. A figure of 35
kW kg-I resulted, which is substantially larger than current goals for electrochemical capacitor performance for electric (hybrid)-powered vehicles. I8 These
polymers are thus very promising capacitor materials (see Chapter 12), provided electro-organic degradation side reactions do not limit their cycle life,
which should be >105 . Equation (15.56) evidently represents the difference of
stored energy in the capacitor as V declines from Vi during a time interval, t,
down to a cut-off limit when V = 0, i.e., for all the stored energy to have been
consumed in the constant power discharge regime.
In the galvanostatic mode, the situation is simpler since I = constant and V
declines with time according to the charge remaining on the capacitor after the
charge l.dt has been consumed. The power associated with the discharge process is then [(constant) x Vet), which declines continuously according to the decline of Vet) with t given by
462
Chapter 15
(15.63)
Vet) = q(t)/C
I
= [q(initial) -
f I.dt)/C
(15.64)
= [q(initial) - q(t))/C
(15.65)
assuming C is independent of V. Obviously Vet) ~ 0 as q(t) ~ -q(initial), corresponding to completion of discharge. The power associated with discharge
will be I Vet), i.e.,
(15.66)
= I[q(initial) - It)/C
(15.67)
so that the power drain decreases, in this case linearly with time during discharge at constant I, as may be expected owing to the declining V with time.
A method (see Section 15.4.1) analogous to the C-rate procedure for rate
comparisons can also used to express the discharge or recharge rate in terms of
power, P:
P = M x En (in watts)
(15.68)
As defined previously, E is the numerical value of the rated energy of the energy
storage device in watt-hours which, in the case of a capacitor, depends on the
voltage reached on charge just before decomposition of the solution.
In the constant power discharge procedure for evaluating power density,
the test procedure is continued until some arbitrary or standard cut-off voltage
is reached, the current being continuously increased to maintain the constant
power delivery of the system during the test. On this basis a watts or kilowatt
per kilogram rating can be specified as a practical p.d. As mentioned earlier, an
absolute, theoretical p.d. cannot be specified since the power drain depends on
the rate or current density of discharge. However, since p.d. generally exhibits
a maximum with increasing current, this p.d. (max) could provide a useful, more
theoretically significant measure of power density. In the case of capacitor discharge (for the case of double-layer-type capacitors at least), the Tafel-type polarization term i(a + b log i) in the p.d. equation will normally be absent since
no Faradaic currents will be passing, at least ideally.
15.5.5. Effects of Temperature
Quite generally, the power density attained by both electrochemical capacitors and batteries improves with an increase in temperature. This is due to
(1) the usual decrease in ohmic resistance, especially ofthe electrolyte (with an
463
effective activation energy of about 16 - 20 kJ mol-I), with an increase in temperature; and (2) a decrease in activation polarization (Tafel polarization) with
temperature associated with an activation energy of 28 to 80 kJ mol-I, depending on the electrode reactions involved. Also, the percentage of rated capacity
that is available as a function of discharge current density normally improves at
elevated temperatures of operation. This effect is more complex and occurs for
a variety of complementary and indirect reasons, including those in (1)
and/or (2).
(15.69)
For a given value of Von charge, it is only C that can have a temperature dependence, i.e., a derivative with respect to T.
It was shown in Chapter 6 that the double-layer capacitance is a complex
quantity determined by the geometry of the compact Helmholtz layer, the size
and solvation of cations and anions, and the field-dependent orientation of solvent dipoles in the interfacial field, as well as electronic effects connected with
potential-dependent electron-density overspill (Chapter 7). The solvent dipole
orientation effect can be formally treated in terms of an interphasial dielectric
permittivity K which will be temperature dependent, usually diminishing with
increasing temperature. 19,20 This mainly reflects diminishing orientational dielectric polarization with increasing temperature in a given interphasial field.
Since C is represented formally as
(15.70)
per unit area, where the dielectric constant, B, is the relative permittivity KIKo,
with KO being the permittivity of free space referred to in Chapter 5. Therefore,
with B tending to decrease with increasing temperature and d, the effective
thickness of the compact layer, tending to increase due to thermal fluctuations
of the structure of the interphase, C will usually tend to decrease with increasing
temperature.
Therefore, at a given voltage on charge, the electrostatic energy of the double layer will tend to decrease with temperature so that dGldT will be negative.
464
Chapter 15
Hence, with Eq. (15.69), the electrostatic entropy of the double layer21 - 24 will
tend to increase with temperature. This can be interpreted in a qualitative way
as the result of higher temperature opposing ordered electrostatic orientation
and electrostriction2o,25,26 in the double layer. There will also be an entropy contribution from the series diffuse double layer capacitance, but for the strong
electrolyte solutions usually used in the operation of electrochemical capacitors,
the reciprocal contribution (Chapter 6) of this diffuse-layer capacitance will be
negligible in the overall interfacial capacitance.
Various theoretical treatments of the entropy of the double layer at electrode interfaces have been given in the literature 21 ,22 and are based on the solvent dielectric properties in the interphase and the related dipole orientation
behavior in the interphasial field. This area is, however, outside the scope of this
chapter. Suffice it to say that the entropy of the double layer is some 20 to 40 J
K- 1 mol- l more negative at appreciable charges than at the potential-of-zero
charge and depends on whether anions or cations are preferentially populating
the double layer.23,24 These figures correspond to T/)"S quantities in G of about
6 to 12 kJ mol-I, a not-insignificant contribution.
Some practical attention has, however, been given to the entropy of the
double layer as a function of extent of charge. By analogy with the phenomenon
of adiabatic demagnetization (the relaxation of magnetic dipole polarization)
employed to attain very low temperatures near zero K, it has been suggested that
a similar effect could be utilized for cooling (but not near zero K) an electrolyte
by allowing the interphasial dielectric polarization in large-area double-layer
capacitor devices to relax adiabatically with the expected absorption of heat
from the surroundings, determined by a T/)"S quantity. However, unlike magnetic dipole relaxation, the relaxation of electrostatic dipole orientation polarization in a double layer must be coupled with relaxation of accumulated ionic
and electronic surface charges at the interface. Forced discharge to achieve rapid
decay of interfacial polarization, leading to absorption of heat, will tend to be
opposed by inevitable [2R losses generating joule heating in opposition to the
cooling caused by dielectric depolarization.
465
Thus, in a series stack, say, often IF capacitor units,(1/C);= 10 x (1/1) so that the
overall C is only 0.1 F! However, for a I-V rating of each unit, a 10-V combination
results and the energy stored in each unit at 1 V is 112 x 1 F x (1 vi;= 0.5 J, independent of the series connection. Were the 10 units connected in parallel, the stored
energy would be 10 x 112 x 1 F x (1 V)2;= 5 J, the same as the total for the series-stacked combination. This is a requirement of the principle of conservation
of energy since energy cannot be lost or gained by switching a charged parallel
combination of capacitors to a series one.
In the case of electrolytic capacitors, the operating voltage can be up to 200
V per unit having a capacitance of, say, 100 flF. The stored energy at 200 V in
such a unit would be 100 X 10-6 F x (200 V)2 x 112;= 2 J, which is not too much
different from that for the example of the 1-F electrochemical capacitor charged
to the nominal potential difference of 1 V!
With a parallel combination of many such high-voltage electrolytic capacitors and each combination being coupled severalfold in a high-voltage series, a
good e.d. system could be constructed. The practical problem for EV application would probably be the volume required using this arrangement, compared
with that for multiply stacked electrochemical capacitors of much higher intrinsic specific capacitance but much lower voltage rating, Ll V. Any advantage for
the electrolytics, apart from disadvantageous volume requirements depends, in
order of magnitude, on how 10-6 (i.e., on the order of microfarad units) compares with (Ll V)2 based on Ll V being on the order of 100(V) at least. Of course
even 100-mF electrolytics are today not too large.
Another aspect of the electrical behavior of electrolytic capacitors is their
quite low equivalent series resistance (the esr discussed in Chapters 14 and 16
for porous capacitor electrodes). This means that they have a good high, as well
as low-frequency response behavior without the limitation of large RC time
466
Chapter 15
constants (Chapter 16) for inner surface area regions that are electrically, relatively isolated by appreciable electrolyte resistance in interparticle pores (see
Chapter 14). Then the power-density performance aspect of electrolytic capacitors will be expected to be superior to carbon double-layer embodiments. This
is another significant advantage over the latter type of capacitors, especially for
situations where short-pulse transient behavior is technologically desirable in
certain applications. Applications substituting electrochemical capacitors for
high-voltage operating requirements will, however, have to avoid the intrinsic
polarity associated with electrolytic capacitors, but that could be advantageous.
Further evaluations of these possible advantages are desirable.
Finally, in the context of this section, the development of the so-called
"Hybrid" capacitor by the Evans Corp. must be mentioned. This rather original
device is a hybrid combination (U.S. Pat. No. 5,369,547) of an anodic oxide film
electrolytic capacitance (anode electrode ofTa-Ta20s) together with a relatively
high electrochemical capacitance (the other pole) of the oxide film (RU02) type
as cathode pole. A high potential difference is sustained across the thin, anodically formed, barrier-layer oxide film and a low potential difference across the
other capacitative film so that no significant electrolyte decomposition can set
in. Good capacitance values (e.g., 0.68 F) in a small, thin disk device are
achieved with operating voltages of 100-150 V. With scaleup, these devices
may offer some opportunities in the EV power-train field.
The advantageous basis of the Evans "Hybrid" capacitor has to be considered in relation to the common technology of fabrication of Ta or Al electrolytic
capacitors. Such devices employ a thin, inSUlating, anodically formed oxide
film on the electrode metal, e.g., Ta or AI. Usually, these capacitors are asymmetric, employing dissimilar anode and cathode materials. In these the dielectric oxide film on the anode is the thickest and sets the working voltage of the
unit. In preparation of the oxide films, the support metal is anodized at a certain
voltage and the oxide film grows to a certain thickness at that voltage, until the
rate of further extension of the film becomes negligible. The growth follows one
of the logarithmic laws of oxide film development. 27 ,28 Typically, the anodic
oxide films at Al or Ta can sustain a potential difference of 100 - 200 V across
them in the final quasi-steady state.
The complementary cathode usually has a higher capacitance because of
its greater surface area, but the masses are commonly similar. Manufacturers use
the highest specific surface area materials for cathodes because these materials
are ,best for low voltage anodization.
The thinner dielectric film at the cathode and its substantial real surface
area cause the cathode capacitance (Cc ) to be higher than the anode capacitance
(Ca ). Then since the overall capacitance, C, is determined by the series equivalent circuit of the device, namely,
1.=_1 +~
C Ca Cc
467
(15.72)
and since Cc Ca , the overall C is determined by the Ca across which the high
voltage can be sustained. The device capacitance always approaches the smaller
value of the two component electrode film capacitances owing to the nature of
the relation (15.72).
Because the electrochemical capacitor cathode in the hybrid requires little
volume, available space can be used to double the size of the anode. The result,
in the Evans device, is a capacitor with four times the original energy density.
A Hybrid capacitor occupies about one-fourth the volume and needs only about
one-fourth the amount of raw material required for a production electrolytic capacitor.
The anode is a pressed-sintered disk made from high capacitance density,
high-purity, commercially available tantalum powder. Formation of the tantalum pentoxide dielectric was done electrochemically in the usual way. This
yields a high specific-surface-area porous electrode pellet with an integral dielectric oxide coating. The high-capacitance pseudocapacitor cathode was prepared on titanium foil according to a recipe developed to yield electrodes having
about 6 mF cm-2 of electrode area.
A complex-plane, impedance spectroscopy plot for this device is shown in
Fig. 15.23 and exhibits the typical frequency response of a transmission-line,
multiple time-constant (RC) system. The device becomes almost purely capaci-
.100
flHz
(/)
.200
.c.
2: ......
N
0400
0
.o:soo
' 1600
(,9.4 kHz
Z'lohms
FIGURE 15.23. Complex-plane impedance for the Hybrid electrolytic capacitor of the Evans
Corp.
468
Chapter 15
The basic importance of internal resistance, mainly of the contained electrolyte, on high-rate capacitor performance has been emphasized in Section
15.2.2 and was examined more practically by Reichman and Msotek. 29 The corresponding effects of electrical resistance on the dependence of realized capacitance on discharge rate (power-density factor) were also discussed. Modeling
calculations were carried out29 by taking selected values for the specific conductivity of the matrix, the electrolyte solution (strong aqueous sulfuric acid), a
carbon-plate substrate, and the conductivity of the separator. The electrode
thickness was taken as 0.076 cm, the substrate thickness as 0.064 cm, and the
separator thickness as 0.00254 cm. The total ac resistance of 1.0-V capacitor
cells is predicted to be 0.29 ohm cm2 by adding the resistances of the components based on their respective thicknesses. An arbitrary or material-specific parameter in this simulation was the specific conductance of the matrix, which
depends on the activation and compression procedures used. The role of polarization resistance, which is significant at high rates, was also considered in the
evaluation, but this would, it seems, occur only if the charging process became
partially Faradaic.
Some graphical representations of the performance data are shown in Figs.
15.24 and 15.25 for small and large carbon particle sizes and for small, medium,
and large amounts of electrolyte. The polarization curves for a low-electrolyte,
double-layer capacitor were also evaluated 29 for different states of charge from
38 to 100%.
469
90
0~
.......
CI)
(.)
0
+-
80
'u
0
'0.
0.
0
Large particles
70
OJ
N
E
....
0
'0..
60 -
"'0
500
FIGURE 15.24. Achieved relative capacitance vs. discharge current in a double-layer capacitor
fabricated from small D or large 0 particles. (From Reichman and Msotek?9)
Ragone-type plots were constructed for four resistance values ("0.0", 0.5,
and 1.0 ohm cm2) associated with a capacitance of 5.5 F cm-2 (F cm-3) (Fig.
15.26). The power range was ca. 70 to 1580 W kg-I. The Ragone characteristic
curves were derived29 according to the following equations:
(15.73)
P == (1I2)i(V - iR)
dt
(15.74)
with
i == [C dVldt]/[RCIdt + 1]
(15.75)
or
C
dt = --;-. dV - RC
I
(15.76)
470
Chapter 15
o~
.....
(c)
U
~
<t
U 75
>
i=
<[
-I
0::
60
FIGURE 15.25. As in Fig. 15.24 but for (a) low 0, (b) medium 0, and (c) high <> electrolyte concentrations (esr effect; concentrations unspecified). (From Reichmann and Msotek. 29 )
Note that their Eq. (15.73), with 112, does not seem correct as written. The relation for R = 0 is the vertical line in Fig. 15.26 since no diminution of operating
voltage with increasing current (power), i.e., no iR drop, occurs in this case.
Another practical aspect of the importance of internal resistance is as follows. As in battery construction, the maximization of volume energy density of
electrochemical capacitors is required, together with the minimization of esr.
This would normally be achieved by compaction of the matrix material under
mechanical pressure during some stage of the manufacture. However, while
compaction will help to minimize interparticle resistance in carbon-based
double-layer capacitors, compression can have the effect of squeezing out
the electrolyte, thus diminishing the active pore area accessible for interfacial charge accumulation. Hence careful optimization of fabrication conditions is required in the construction stages of capacitor devices.
Typical battery and capacitor e.d. and p.d. requirements for electric vehicle
applications have been examined by Haskins and Dzieciuch30 for the U.S. Advanced Battery Consortium and are summarized in Tables 15.3 and 15.4. Circuit
response equations were solved numerically for both constant-power pulses and
for the FUDS cycle dynamic-load profile. Figure 15.27 shows pulse power results plotted in a way that allows general relative sizing of capacitors for a range
of power requirements. The effects of finite esr are included in this figure. Below about a 2-to-l ratio of peak pulse to sustained power, esr does not strongly
471
1600
1200
Ie>
.><:
::......
800
....
Q)
a..
400
0L-~
0.0
0.8
1.2
1.6
2.0
affect the capacitance required for adequate hybrid-system performance. However, a key secondary parameter for capacitor sizing is the internal resistance, RB , of the battery which, if of appropriate magnitude, can provide the
necessary coupling between the battery and capacitor components. Rs can be estimated, according to Haskins and Dzieciuch,30 from the sustained power requirement on the basis that it is the battery that has to provide all of this power
at no less than the system's minimum voltage, Vrnin ; thus
65 kW
35 kW
40 kWh
330 V dc
250 to 385 V dc
-30 to +65C
600 cycles, 5 years
400 kg
300 liters
$6000
112% per day
472
Chapter 15
250 to 385 V dc
60 F
900 Wh
-30 to +65C
45 milliohm
25 rnA
300,000
60 kg
80 liters
$1000
w
)(
wI-
U W
ZIIl
0.1
<t-1
!Q(l.
I-=>
Ill'
W w
a:u
~~
wII-u
I-<t
<t(l.
1D<t
u 0.5
1.5
2.5
:3
FIGURE 15.27. General results for capacitor sizing for three esr capacitor-battery resistance
ratios, r, in a hybrid electric vehicle application. (From Haskins and Dzieciuch. 30 )
473
60
50
s::.
~40
"-
fJ 30
1JJ
Q.
en 20
10
:=
~
"-
a::
1JJ
:=
0f--"-t+-------"--4J++-It-.-F=-
160
200
240
280
320
TIME Isec
FIGURE 15.28. Typical response of a battery-capacitor hybrid energy delivery system to maximum overall power requirement in the FUDS cycle. (From Haskins and Dzieciuch. 30)
capacitor-battery hybrid power source. The watt-hour energy capacity was SOO.
The resulting advantageous behavior is clearly shown in [Fig. 1S.29(b)], with
the capacitor in comparison to the battery behavior without capacitor load leveling [Fig. IS.29(a)]. The power-drain spikes for acceleration or hill climbing
were up to ca. 50 kW, while with the coupled capacitor they did not exceed 14
kW [Fig. 1S.29(b)]. Capacitor power spikes reached about 30 to 4S kW. Pulsedpower systems are envisaged for such capacitor-battery hybrid coupling.
The advanced electric vehicle goals that have stimulated large, high-power
electrochemical capacitor design and construction (e.g., by Maxwell Laboratories at Auburn University), were as of 1991: energy density 100-200 Wh kg-I;
peak power density, lSO-300 W kg-I, distance life, SO,OOO-lOO,OOO miles. The
cost was estimated at $100-150 per kilowatt-hour. These requirements are now
revised upward, but with cost estimates lower.
Continuing, possibly similar, electric vehicle requirements in up to eleven
northeast states and the District of Columbia have led to a 1991 projection by
Automotive News for annual sales of electric vehicles to be over 100,000 starting
in 1998, rising to over SOO,OOO by the year 2003. However, in the year this was
written (1996), the first projection seems now to be either exaggerated or optimistic, and can depend on the contemporary political and economic climate.
More recent estimates (1998) by the Electric Vehicle Association of the Americas for sales in Japan and the US of electric vehicles and hybrid electric vehicles
474
Chapter 15
60
o} Without supercopacitor
Battery
Power/kW
Amps
Volts
100
Time or distance-
Battery
Power/kW
30
b} With supercapacitor
(Wh) cap = 500
Capacitor
Power/kW
Time or distance-FIGURE 15.29. Effect of hybridizing a large-value electrochemical capacitance with a 500-Whcapacity battery on the battery power drain profile in the FUDS cycle. (a) without capacitor and
(b) with capacitor. Also shown is the state of charge (SOC) of the capacitor component expressed
as % of Q, the maximum charge available (diagram b). (From Burke. 15 )
are 17,500 and, in Asia, greater than 25,000 electric motor scooters. In Europe
(France), electric vehicle sales in 1998 have been around 6000.
475
2.5r----------~1---r-------,.----------------------/~--------.
I
I
.......
~
0..
,I
LI
II
Rocketdyne
II /
IX/ / -,'
-' -'
. .:----
PIE = 50
/
-'
/
/
/
WOfid
Flywheel
/Consortium
______
--
P/.E =25-
-:;-::"-~-----
PIE -10-
United Technologies
-~ Research C~t~r __
- - - - -American
---
1/1/"' ... , - ~
M9t~r
:,,-,..--_~-=-I
nique-'
/ / . / Mobility
0.5
/
/
PIEI =100
Magn~t
/ /
,I
I " I I
1.0
a:
w
>fZ
P/E~ 2
I
University of Texas
Center for Electromechanics.
2.0
(fi
10
20
30
Flywheel Systems
40
50
60
REFERENCES
1. Report of the Panel on Electrically Powered Vehicles to the Commerce Technical Advisory
Board, U.S. Dept. of Commerce, October 1967. (Library of Congress Code: TO 886,5 U.S.
1967, V; I.)
476
Chapter 15
2. The Automobile and Air Pollution, Part II, Sub-panel Report to Panel on Electrically Powered
Vehicles, Commerce Technical Advisory Board, U.S. Department of Commerce, December
1967. (Library of Congress code: FC US 1 DC, 67A71.)
3. D. V. Ragone, in Proc. Soc. Automotive Engineers Conference on Rev. of Battery Systems for
Electrically Powered Vehicles, Society of Automotive Engineers, Warrendale, PA. (1968).
4. J. Miller in Proc. Fifth IntI. Seminar on Double-Layer Capacitators and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational Seminars, Boca
Raton, Fla. (1995).
5. W. Pell and B. E. Conway, J. Power Sources 63,255 (1996).
6. J. O'M. Bockris and S. Srinivasan, Fuel-cells: Their Electrochemistry, McGraw-Hill, New
York (1969); see also S. Srinivasan and H. D. Hurwitz, Electrochim Acta, 12,495 (1967) .
7. R. de Levie, Electrochim. Acta. 8,751 (1963).
8. B. E. Conway in Proc. Fifth IntI. Seminar on Electrochemical Capacitors and Related Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Education Seminars, Boca
Raton, Fla. (1995).
9. 1. D. Bernal and R. H. Fowler, J. Chem. Phys., I, 515 (1933).
10. B. E. Conway, 1. O'M. Brockris, and H. Linton, J. Chem. Phys. 24, 834 (1956).
11. A. Rudge,l. Raistrick, S. Gottesfield, and J. P. Ferraris, Electrochim. Acta 39,273 (1994).
12. D. Linden, ed., Handbook of Batteries, 2nd. ed. McGraw-Hill, New York (1995).
13. D. V. Ragone, B. M. Barnett, and 1. M. Bentler in Proc. Twelfth Inti. Seminar on Primary and
Secondary Batteries and Applications, Florida Educational Seminars, Boca Raton Fla. (1995).
14. T. C. Murphy and P. B. Davis, Extended Abstracts, Battery Division, 188th Electrochemical
Society Meeting, pp. 112-113, Electrochemical Society, Pennington, N.J. (1995).
15. A. F. Burke, in Proc. Fourth IntI. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational Seminars, Boca
Raton, Fla. (1994).
16. 1. R. Miller, Extended Abstracts, Battery Division, 188th Electrochemical Society Meetings,
pp. 106-107, Electrochemical Society, Pennington, N.J. (1995).
17. M. Juzkow, S. Gold, and B. Kraemer, in Proc. Sixth IntI. Seminar on Electrochemical
Capacitors and Related Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational
Seminars, Boca Raton, Fla. (1996).
18. A. F. Burke, in Proc. Second IntI. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational Seminars, Boca
Raton, Fla. (1992).
19. 1. B. Hasted, Aqueous Dielectrics, Chapman and Hall, London (1973).
20. B. E. Conway, Ionic Hydration in Chemistry and Biophysics, Elsevier, Amsterdam (1981).
21. B. E. Conway and L. G. M. Gordon, J. Phys. Chem., 73, 3609 (1969).
22. H. D. Hurwitz, S. Sanfeld, and A. Steinchen-Sanfeld, Electrochim. Acta, 9,929 (1964).
23. G.1. Hills and S. Hsieh, J. ElectroanaI. Chem, 58, 289 (1975).
24. G. J. Hills and R. Payne, Trans. Faraday Soc., 61, 326 (1965).
25. B. E. Conway, 1. E. Desnoyers, and R. E. Verrall, J. Chem. Phys., 43, 243 (1965).
26. 1. Padova, J. Chem. Phys. 40,691 (1964).
27. N. F. Mott and H. Cabrera, Rept. Prog. Phys., 12, 163 (1948).
28. L. Young, Anodic Oxide Films, Academic Press, New York (1961).
29. B. Reichman and R. Msotek, in Proc. First IntI. Seminar on Electrochemical Capacitors and
Similar Energy Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational
Seminars, Boca Raton, Fla. (1991).
30. H. 1. Haskins and M. A. Dzieciuch, in Proc. First IntI. Seminar on Electrochemical Capacitors
and Similar Energy Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational
Seminars, Boca Raton, Fla. (1991).
477
31.. A. Burke, in Proc. First Inti. Seminar on Electrochemical Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational Seminars, Boca
Raton, Fla. (1991).
32. M. Belager in Proc. Sixth Inti. Seminar on Double-Layer Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida Educational Seminars, Boca
Raton, Fla. (1996).
Chapter 16
AC Impedance Behavior of
Electrochemical Capacitors and Other
Electrochemical Systems
16.1. INTRODUCTION
480
Chapter 16
and any potential-dependent Faradaic resistance (Eq. 16.5) the system may exhibit, to be separately evaluated. The evaluation of esr for various electrode
structures and choices of electrolyte (Chapter 13) is especially important for device design and optimization of power density. In addition, for favorable conditions, the pseudocapacitance of an electrode can be distinguished from
double-layer capacitance over a range of frequencies, which is sometimes a matter of practical significance.
Since some applications of electrochemical capacitors are for low-frequency
electrical filtering and for other uses in electronics, ac impedance measurements
give direct information on device performance for such purposes. In the regime of
response behavior to time-dependent potentials, the method of so-called "alternating current" (ac) impedance spectroscopy has become a principal procedure for examining fundamental and applied aspects of electrochemical capacitor
behavior. Experimentally, however, it is the potential that is modulated (typically by i> 5 mV) over a wide frequency range. The term "spectroscopy" has become used in this field because the electrical and kinetic response of
electrochemical systems is determined over a wide range of frequencies that are
applied to modulate the system, and various kinetic or associated electrical "relaxation processes" can be characterized as in other types of spectroscopy. For
example, there are close analogies to dielectric absorption spectroscopy and dielectric relaxation, as in the work by Cole and Cole. l Basically the magnitude and
phase relation of an ac current generated in response to an applied low-amplitude
(5 mV) alternating voltage (av) is evaluated and recorded by some phase-sensitive
recording device or other kind of apparatus. Two commercially available instruments currently used for this purpose are the Solartron Frequency Response Analyser system, made in the United Kingdom or the EG and G system, available
in North America. Various other systems are produced in Germany or the
United Kingdom. The principle of operation of the Solartron system differs
from that of phase-sensitive, "lock-in" instruments.
The analysis and interpretation of the ac response behavior of an electrode/solution interface, i.e., the "impedance spectroscopy" of the system, as in
an electrochemical capacitor, is usually carried out by one or another of the following procedures:
1. The observed frequency-response behavior is fitted to that expected according to equivalent-circuit models of various complexities, supposedly simulating the behavior of the actual system, and evaluating the so-called "transfer
function." Three equivalent circuit elements are commonly employed: ohmic
resistors, R; capacitances, C, or pseudocapacitances, Crp; and inductances, L, or
pseudoinductances, Lrp. Their real or imaginary frequency response in the impedance Z is summarized below.
481
Z=
2'=
ZR
Zc
ZL
Z"
-j/OJC
jwL
I I
482
Chapter 16
recognized on the basis that: (1) an ohmic resistance, R, component has a frequency-independent impedance, Z' = R, with a phase angle of zero while (2) a
pure capacitance component, C, has a frequency-dependent imaginary impedance Z" = l/jOJC == -j/OJC with a phase angle of 90 at all frequencies with zero
impedance as OJ -t 00 and infinite impedance as OJ -t O. Combinations of Rand
C components have a complex dependence of their Z on OJ and exhibit a frequency-dependent phase angle.
A useful summary of the frequency-response behavior of various equivalent circuits has been given by Gabrielli in a publication of the Solartron Instrumentation group titled "Identification of Electrochemical Processes by
Frequency Response Analysis" (1980). *
Electrode-solution systems, especially those involving porous, high-area
materials, as in electrochemical capacitors or batteries, have complex combinations of C and R components and also some pseudoinductive (L!) behavior. For
electrode-solution systems, the interfacial impedance is associated with (1) a
double-layer capacitance, Cdl (ubiquitously present-see Chapter 6); (2) a solution resistance, R s ' in series with the electrode interfacial impedance; and (3)
a Faradaic impedance, ZF, associated with any potential-dependent Faradaic
charge-transfer resistance, RF (see later discussion) and pseudocapacitance, C!.
This latter can arise when the electrode/solution interface is not ideally polarizable so upon polarization, some potential-dependent Faradaic current
passes, leading in some cases (e.g., for cathodic H2 evolution) to a pseudocapacitance associated with potential-dependent coverage by adsorbed species
(H) involved in the Faradaic reaction. The distinction between ideally polarizable, purely capacitative electrode interfaces and Faradaically leaky interfaces (also exhibiting double-layer capacitance) was explained in Chapter 6.
An electrochemical double-layer capacitor would have an ideally capacitative (i.e., ideally polarizable) double-layer capacitance. However, in practice
this is rarely the case, so real systems have some ohmic solution resistance,
some ohmic interparticle contact resistance, and often (at high potentials on
charge) some Faradaic leakage resistance due to overcharge or to Faradaic redox
reactions caused by impurities (see Chapter 18).
The impedance associated with a Faradaic reaction arises as follows. The
current density, i, as a function of overvoltage " is given by the Butler-Volmer
equation (Chapter 3)
i =io[(exp[fl"FIRT] - exp - (1 -
P)" FIRT])
(16.1)
"The author's address at that time was CNRS Laboratoire GR4 Physique des Liquides et
Electrochimie, Universite Pierrre et Marie Curie, 4 Place lussieu T22, 75230 Paria Cedex 05,
France. The addrews of Solartron, the publishers of this useful short monograph, is The Solartron
Electronic Group Ltd., Famhorough, Hampshire, GU 147 PW, England.
483
where io is the exchange current density passing in both directions of the reaction at equilibrium (zero overpotential '7) and fJ is a barrier symmetry factor usually near 0.5 for simple electron transfer processes. (More generally, fJ may be
replaced by the transfer coefficient, a, for more complex reaction mechanisms.)
....
The tw~ terms in the brackets, multiplied by i o, are the forward (i) and backward (i ) components of the net current density, i:
~
<-
i=i-i
(16.2)
Sufficiently near equilibrium (i.e., for small overvoltages of 5 m V), the exponentials ofEq. (16.1) can be linearized to the first term of a Taylor series expanSIOn, glVlng [for fJ'7F1RT or (l-fJ)'7F1RT 1) (cf. Eqs. 3.20 to 3.23 in
Chapter 3)]
(16.3)
i = io'7F1RT
(16.4)
i.e.
d'7ldi =RT1ioF= RF
(16.5)
Rs
484
Chapter 16
more complex equivalent circuits (considered later in this chapter and in Chapter 17 for porous electrodes), and for an electrode process or a Faradaically leaky
interfacial capacitance (double-layer capacitance as in a double-layer type of
electrochemical capacitor).
RF as written above is the Faradaic resistance at the reversible potential.
However, the rate ofthe Faradaic leakage, according to Eq. (16.3), is dependent
on the overvoltage, ", so RF has other potential-dependent (lower) values at finite
This can also be verified through impedance experiments conducted at
various finite, potentiostatically controlled, dc-level overpotentials where the
electrode process is no longer at equilibrium. When such an experiment is conducted, the diameters of the semicircles of the Z" vs. Z' complex-plane plots for
a range of frequencies decrease with increasing" since RF is decreasing. This
change corresponds (when logs are taken) to a Tafel plot for the electrode process. An example is shown schematically in Fig. 16.1.
Far from equilibrium (" FlRT 1), only the first term in the brackets of
Eq. (16.1) is significant (or for opposite polarization, the second); then the Tafel
equation in exponential form is recovered. Thus,
"s.
i"
= ioexpp1] F IRT
(16.6)
av modulation can still be applied around some imposed 1] value that is associated with the current density i'l. Then, for a small modulation of /:l.1] to 1] + /:l.1],
say, and for non-small" values,
i,,+!l>}
(16.7)
i
FIGURE 16.1. Complex-plane plots of -Z" vs. Z for a parallel RFC circuit (with series resistance,
Rs) where RF changes with potential for a Tafel-type Faradaic process; schematic.
485
(16.8)
For sufficiently small D.17, linearization of the exponential term in D.17 gives
(16.9)
ilJ+!l1J can be represented by ilJ + D.i, the latter being the response current-density
increment due to D.11, i.e.
ilJ +!l1J
= ilJ +D.i
(16.10)
Then
ilJ + D.i = ilJ + i'1(jJD.17 F IR1)
(16.11)
(16.12)
so that
Since i" is greater than io by the Tafel exponent, RF,IJ < RF , i.e., for '1
librium. In fact, since ilJ = io exp P'1FIRT for appreciable '1 values,
=at equi(16.14)
(16.15)
This is why diameters of semicircles in complex-plane plots diminish with increasing '1. Hence impedance spectroscopy determinations of RF,IJ at various
overpotentials can lead to information on the Tafel parameters for the electrode
process involved.
Note that, in the above discussion, the modulation in av experiments will
be D.'1 around some potentiostatically adjusted '1 value. Unlike the situation for
modulation around the reversible potential, when '1 = 0, no backreaction current
component arises for behavior at appreciable '1 values. This is the reason, of
course, why the P factor arises in Eq. (16.15) for RF,IJ and not in RF for 11 = 0.
486
Chapter 16
Vet) = Vosin WI
(16.16)
FIGURE 16.2. Rotating vector (2) representation of time dependence of an alternating voltage
signal at frequency f rad s-I, =wl2n Hz.
487
V(t) Vo.
I=--=-smwt
(16.17)
(16.18)
Hence
I(t)
=losin wt
(16.19)
so that I(t) is in phase with V(t), i.e., I and V are related in the same way as for
a dc circuit.
Figure 16.3 shows the relations of I and V(t) as functions of t. The corresponding rotating vector diagram for this situation for a resistance is shown in
Fig. 16.3(c). Note that V(t) and I lie on the same vector in wt.
In the case of a capacitance C (Fig. 16.4), such as a pure, ideally polarizable
electrochemical double-layer capacitance, the instantaneous charge q on the capacitor is
q = CV(t)
=CVosin wt
(16.20)
The corresponding charging current, i.e., the rate of passage of charge q with
time t, will be
1= dq I dt = wCVocos wt
(16.21)
Alternatively,
l.w
lo)
-------- t
l c)
FIGURE 16.3. Relation of I and V to time in an alternating voltage signal across a resistance R:
(a) circuit, (b) signals, and (c) rotating vector diagram.
488
Chapter 16
I.w
1 ------
(a)
(b)
(c)
FIGURE 16.4. Relations of I and V to time in an alternating voltage signal across a capacitance
c: a) circuit; b) signals; c) rotating vector diagram (lout of phase with Vby -90).
I(t) = CdV(t)/dt
=wCVocos wt
(16.22)
and the maximum current Imax (or current amplitude) is evidently then
Imax = wCVo (for cos wt = 1)
(16.23)
= Vo/(lIwC)
(16.24)
and define an impedance Zc as lIwC. Because the current I is out of phase with
the voltage V, we write the impedance Zc as IljwC where j =-{1 or
(16.25)
which is an imaginary quantity; the impedance of a resistor R is simply ZR = R,
a real quantity.
The analysis of more complex equivalent circuits involving C, R, and possibly L components (see Chapter 17) involves the mathematical separation of
489
IZI
w
FIGURE 16.5. Relation of ZR. Zc. and ZL to frequency compared.
the real and imaginary components of Z (designated Z' [real] and Z" [imaginary D. This is usually performed by a so-called "rationalizing" operation on the
overall expression for impedance as a function of w (see later example). The relationships among Zc, ZL, and ZR(= R) are illustrated schematically in Fig. 16.5.
In summary, it can be stated that the instantaneous potential differences in
a series circuit addressed by an av add algebraically, as for a dc circuit, but the
voltage amplitudes add vectorially. Formally, for any RCL circuit, its impedance can be defined as the ratio of the voltage amplitude to the current amplitude, corresponding to Ohm's law: R = VII.
16.2.2. Root-mean-square and Average Currents in ac Studies
fay
=_1_
(16.26)
For a sinusoidally varying current I(t) = Isin wt, the average value of current
over a half-cycle (0 to n) from t = 0 to t =n/w will be
490
Chapter 16
_L..J-
1ov _.2.l~~~~
- 7r """
FIGURE 16.6. Average value of a sinusoidal current over a half-cycle (211n, Eq. 16.27).
n/w
lay = w
(16.27)
as illustrated in Fig. 16.6; thus, the average current is 21n (about two-thirds) of
the maximum current and the area under the rectangle in Fig. 16.6 (area =
2l1n x n/wis equal to the area (i.e., a charge) under one loop (half-cycle) of the
sme curve.
In the case ofrms current or voltage, Fig. 16.7 is used to represent the situation. If let) =lsin wt
[I(t)]2
12
=12sin2 wt =-(1
2
,.,
I
"
(16.8)
cos2wt)
,-","
FIGURE 16.7. Average value of the square of a sinusoidally varying current over any number of
cycles is PI2 (root-mean-square value = 112).
12
12
=-2 - -cos
2wt
2
491
(16.29)
The average value of P, or the mean square current, is equal to the constant term
12/2 since the average value of cos wt over any number of cycles is zero; thus
(16.30)
The root-mean-square current is the square root of this average squared current,
i.e.,
I rms = ,f(l2)av = I/{i
(16.31)
where I is the maximum current, written earlier as Imax. Similarly, the rms value
of a sinusoidal av is
(16.32)
For a meter reading on a rectifier ac ammeter of, say, 1 A, the corresponding rms
value would be 1.11 A. The average value of the rectified current, in any number
of cycles, is the same as the average current in the first half-cycle in Fig. 16.7
or it is 2/n times Imax. Thus, if an ac ammeter deflects full scale with a current
10 through it, the same deflection will occur when the average value of the rectified (ac) current, 21/n, is equal to 10 . The current amplitude, I, for such fullscale deflection is then
1= n1012
(16.33)
492
Chapter 16
In the complex-plane representation of impedance behavior of an electrochemical system, the imaginary component (usually the capacitative one), Z" of
Z, is plotted against the real (ohmic) one over the range of frequencies studied.
For simple circuits combining C and R components, such plots usually consist
of one or more semicircles in the complex plane, sometimes with the center of
a semicircle depressed below the Z' axis.
The self-consistency of impedance analysis of a given equivalent-circuit or
kinetic-mechanistic representation (see later discussion) in terms of frequency
response can sometimes be usefully tested by applying or examining the socalled "Kramers-Kronig relations".4 These have their origins in analysis of optical and dielectric absorption behavior, but the impedance relations for
electrochemical systems as a function of frequency are formal analogs.
In order to demonstrate how a (semi) circular plot of Z" vs. Z' arises, we
consider the equivalent-circuit originally treated by Sluyters and Sluyters3 :
c
Circuit 1
R
The overall impedance Z is a function of frequency w. The components of the
impedance are:
Z(C) = lIjwC =- -j/wC (= Z")
Z(R)
= R (=- Z')
(16.34)
(16.35)
Combining the Zs for the two components of the circuit with parallel, i.e., reciprocal addition,
lIZ=jwC+ lIR
= (jwRC + l)/R
(16.36)
(16.37)
(16.38)
493
w2R2C2+ 1
R(jwRC+ 1)
(16.39)
Therefore
Z=
jwR2C - R _
jwR2C
- (wZR2C2 + 1) (wZR2C2 + 1)
(16.40)
The first term on the rhs of Eq. (16.40) is identified as the imaginary component
of Z (to be denoted by Z") since it contains j, while the second term is the real
component (to be denoted by Z'). It is the two terms Z' and Z" representing the
real and imaginary components of Z that are plotted against each other in a complex-plane (Nyquist) diagram for various values of w.
In most systems of practical significance, an additional series resistance,
Rs, representing the ohmic resistance of the solution and separator, and any internal resistance of the electrode materials, must be included. The latter component is called the equivalent series resistance and is discussed in more detail in
Chapter 17.
We can now proceed to demonstrate the geometrical form of such a diagram.
For w in Z' we can write
(16.41)
i.e.,
(16.42)
Also
"2
wZR4C2
(Z ) = (wZR2C2 _ 1)2
(16.43)
=R -
Z .R2-(RIZ't2
(16.44)
=(R -
Z')Z
(16.45)
RZ + (Z')2 =0
(16.46)
Therefore
(16.47)
494
Chapter 16
It is seen that this is the equation for a circle when Z" is plotted against Z' (complex-plane plot), with the circle having a radius RI2, and two intercepts on the
Z' axis for zero Z" having values of zero and R. The impedance behavior for this
case is therefore represented by a semicircle with positive Z' and negative Z"
values (Fig. 16.8).
If the circuit is
.------1
I-C - - ,
Circuit 2
R
i.e., including a resistance Rs (representing solution resistance or esr), the semicircle becomes displaced along the Z' axis by Rs. There are now two finite intercepts on the Z" axis for Z" = 0, i.e., when Z' = Rs and Z' = Rs + R.
Two limiting cases for circuit 1 are easily recognized. In the equation for
Z as/(w):
1. When w ~ 00, Z ~ Z" =-j/we, i.e., the Z" is small and Z is determined
by the e, which provides the main ac current pathway (-jlwe R)
since the circuit is a parallel combination of Rand C.
2. When w ~ 0, Z" ~ 00, so the only (then dc) current pathway is through
R andZ ~ Z' = R.
(()
((),=1/RC
":-'
.
,,
(()
'R/2
FIGURE 16.8. Complex-plane impedance plot for the simple parallel RC circuit 1 showing toppoint characteristic frequency COr = 1IRe.
495
The complex-plane plot for the parallel Rp-CdJ circuit 1 is shown in Fig.
16.8. A similar analysis can be performed for a series RC circuit
-----~~--~llrc-------
Circuit 3
for which
(16.48)
Z=R+ lIjwC
=R-j/wC
This requires no rationalization so that ZIt is -j/wC. Limiting cases are obviously
w~oo,Z~R;
w ~ 0,
RuJC2 +
jwC
1 + R 2uJC2 1 + R 2uJC2
(16.49)
Y"
Y'
in which real (Y') and imaginary (Y") components are immediately recognized.
(0
~
z'
FIGURE 16.9. Complex-plane impedance plot for the series RC circuit 3.
496
Chapter 16
t
:
1/R
:{
y'
FIGURE 16.10. Admittance (Y) plane plot for the series RC circuit 3. (Compare Fig. 16.9 for the
parallel circuit Z plane plot.)
Denoting
(16.50)
(16.51)
B2 + G - (RI2)2 = (RI2)Z
with the circle (semicircle) centered at B = 0 with G = RI2 on the G axis with
radius RI2. Thus the series RC circuit gives a semicircular plot of Y' vs. Yin
the admittance plane over a range of frequencies, w.
16.3.2. Time Constant and Characteristic Frequency OJr
The top point of the semicircle arising from the complex-plane plot (Fig.
16.8) is a characteristic point corresponding to a frequency Wr =11RC where the
product RC of the magnitudes of Rand C represents a time constant of the circuit. The characteristic top point of the semicircle of the admittance plot is obtained by evaluating the condition for its maximum, i.e., in B vs. G (cf. Eq.
16.51):
(16.52)
Then
dB
dG
1 + (wRC)Z
2wRCZ
(16.53)
=0
l.e.,
(16.54)
1 - [c0R 2CZ] =0
for the characteristic frequency w =Wr the top point. The last equation obviously gives
497
OJ,RC = 1
or
OJ, = lIRC
(16.55)
lIRC has the units of reciprocal time (s-l) so that RC itself is a characteristic
time constant of the circuit for the response to av modulation, as for the Z case
(see Section 16.4).
RC is represented by r = 1/OJ" as before. OJ, is also called the "characteristic frequency" of the circuit. Similar conditions apply to the frequency response of the parallel RC circuit considered earlier in this chapter.
Since for a well-defined semicircular complex-plane plot, the top point can
be identified as the point of intersection of a perpendicular erected from the Z'
axis at RFI2 on the Z" vs. Z' curve, then OJo can be evaluated. Since the RF at this
point is also known, the top-point identification leads to a convenient way of
evaluating C, provided it is not frequency dependent.
Chapter 16
498
= 0; AVR =0 as t ~
AVe = Vas t
00
and
(16.56)
00
=0 at t ~
00.
d(AVJ
d(AVc)
=AVR IR = C dt
- - or AV = R C - R
dt
(16.57)
(16.58)
(16.59)
since V = constant. Then, integrating
t
When t
+ k =-RCln(A V)R
(16.60)
499
so that ~ VRIV = exp( -tIRC), i.e., in a time scale of t reduced by division by the
time constant RC.
The time-dependent current let) through R and into C cannot be explicitly
written; thus, in terms of charge acquired by C,
let)
(16.62)
(16.63)
or
let)
V
=Rexp(-tIRC)
(16.64)
This meets the boundary condition requirements since let) = VIR when t = 0 and
l(t) ~ 0 as t ~ 00. The response function for the charging current variation in
time is seen to be scaled in tIRC, as expected for problems of this kind. These
relations are to be regarded as having general significance for RC circuits.
An alternative treatment of the problem in terms of accumulated charge, q,
on C, is as follows:
from the condition V =lR + ~ VC'
(16.65)
Rearranging,
RC
dq
=dt
(CV- q)
(16.66)
and integrating
-RCln(CV - q) =t + k'
(16.67)
(16.68)
500
Chapter 16
Hence
(16.69)
q = CV(1 - exp[-tlRCD
(16.70)
so that
I =Ii exp[-tIRC]
(16.71)
In this case, the discharge current from C (= -CdVJdt) is equal to the current
through R, namely, VJR. Thus,
-CdVJdt= VJR
(16.72)
so that
1
V
dVcldt=
c
-lIRC
(16.73)
Integration gives
10gV= -(1IRC)t + k
When t
(16.74)
501
giving
V/Vi = exp[-tIRC]
(16.75)
= V(t)IR = (V/R)exp[-tIRC]
(16.76)
again an exponential function in reduced time. As before, the boundary conditions for this case are fulfilled since I(t) = ViR at t = 0 because no charge has
yet been lost from the capacitor; and as t ~ 00, I(t) ~ 0 as the discharge of Cis
completed.
In both cases it is seen that the time-dependence of the electrical response
is scaled exponentially in the reduced time tIRC, i.e., the RC product determines
the effective duration of the charging or discharging processes for some initial
applied voltage or initial voltage on the capacitor. This is a general type of conclusion that applies to RC circuits and in a more complex way to electrochemical capacitors that have a distributed R-C network due to the porosity of the
material (Chapter 14) at which double-layers are established.
A similar characteristic RC parameter also arises when the ac impedance
behavior of resistive and capacitative circuits is evaluated. For example, for the
case of the simple parallel R-C circuit, the maximum value of the capacitative
(imaginary) component of the impedance (Z') arises at a critical frequency we =
lIRC = lite> as was shown earlier, where te is the RC time constant of the network. For such a parallel RC circuit, as W ~ 00, Zc ~ 0, and R is bypassed;
hence Z = O. On the other hand, when W ~ 0, no current passes through C and
Z = R. For intermediate frequencies
1 1
1
-=-+Z R Zc
(16.77)
for the parallel addition of impedance; that is, introducing win Zc,
l/Z=.l+ jWC
R
(16.78)
as discussed earlier.
16.4.2. Formal Significance of the RC Product as a Time Constant
502
Chapter 16
value, the capacitance will be charged from a given applied voltage, Ll V, at a rate
inversely related to R; similarly, the larger the C, the longer will be the time required for the capacitor to acquire the applied voltage difference, Ll V, between
its plates with the charge being admitted through a given R.
Formally, the significance of RC as a characteristic time can be seen by reference to the units oftheRCproduct. Thus R == [voltage] [currentr l (Ohm's law)
or [voltage] [charger I [time]. Correspondingly, C == [charge] [voltagerl. Hence
RC == [time] or reciprocal frequency. The same result arises if Rand C are expressed in the fundamental units of mass, length, and time (mIt); then it is found
thatRC== t, a time quantity. Thus RC == (Ke} t-It) x (Kot) == t where KO is the permitivity of free space discussed in Chapter 4.
16.5.1. AC Bridges
Generator
FIGURE 16.11. Elementary four-arm bridge balancing circuit for resistance (R) or impedance
(Z) measurements. Test cell is labelled as C (not necessarily a capacitance).
503
I I
IZI =AA'IBB',
(16.81)
sin tP =DD'IAA'
(16.82)
z =IZlexp(-jtP)
(16.83)
and
504
Chapter 16
AA'
IZI = 88'
.
Sln
DO'
= AA'
Z =IZI exp(- j)
Lissajous figures
e(t) =6E Sinwt
i(t)
= ~z~
X of osci Iloscope
Sin (w +)t -
II
"
FIGURE 16.12. Lissajous figure generated between the y and x-axes of an oscilloscope or X-Y
recorder in relation to the impedance vector magnitude z and the phase angle, I/! (schematic).
I I
This method is specially applicable to low frequencies, say <1 Hz, but today is
not commonly used although it is a useful procedure for visually demonstrating
(on an oscilloscope) the phase relation between voltage and current in RC circuits.
In this approach, a phase-sensitive amplifier is used which enables inphase and out-of-phase components of a response signal, generated as a result
of the application of an applied modulation signal, to be recorded in an analog
or digital manner. Essentially, the system separates the real and imaginary components of the impedance. For analog phase-sensitive detection, the low limit
of the perturbing frequency is between 0.1 and 1 Hz.
The term "lock-in," applied to such instruments, refers to the situation
where the instrumental response is locked in in-phase or 90 out of phase with
the frequency modulating the response of the experimental system. Such sys-
505
tems are also used in other ways, e.g., in detecting reflectance or optical absorption changes at electrodes brought about by sinusoidal or square-wave, low-amplitude modulation of electrode potential. This arrangement allows major
improvements in signal-to-noise ratios to be achieved.
16.5.4. Digital Frequency-Response Analyzers (Solartron and Other
Instruments)
Here signal components Set) cos wt and Set) sin wt, corresponding to the
imaginary (Z") and real (Z') components of the investigated impedance, are subjected to an integration procedure in comparison with signals from the test system. A simplified block diagram is shown in Fig. 16.13.
The Z' (real) and Z" (imaginary) components are generated from
T
(16.84)
(16.85)
r--------------------------,
I
I
I
I
I
I
Generator
1---'---
1m
I
I
IL
7T" /
~=-'--'
o ~L---:-,-_+---I
I---'--'--Re
I
I
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ...JI
x(t)
S(t)
System
under
test
FIGURE 16.13. Simplified block diagram of the Solartron Frequency Response Analyser system.
506
Chapter 16
where Set) is a cell response signal that may include some noise, and T is a signal
response integration time. Longer integration times, analogous to a signal-averaging procedure, lead to better noise rejection, but they also can lead to problems of changes of state in electrode surfaces during prolonged integration
periods. Total noise rejection is achievable in principle if T ~ 00, but this is not
a practically attainable condition. In the case of the Solartron model 1250, a frequency range of 10-4 to 6.5 x 104 Hz can be covered in a frequency scan of an
electrode process or capacitor behavior.
Normally, the frequency scan must be conducted at an electrode or a capacitor polarized at a potentiostatically-controlled potential or held at a controlled potential difference, respectively. The frequency scan should be performed
over a range of various controlled potentials or potential differences, which may
be a quite time-consuming process. This is why care must be exercised in interpreting results at capacitors or electrodes at which some irreversible changes of
state might take place over periods of an hour or more. A comparison between
scans of frequency response in the upward and downward directions of frequency change can be useful in evaluating such time-dependent effects, e.g.,
due to impurity adsorption or electrode deactivation.
The use of scans at various potentials within the operating range of electrochemical capacitors is very desirable in establishing a full evaluation of capacitor performance. This is because the specific interfacial capacitance of
many metal electrode materials, and of carbons of various types (see Chapter 9),
is often appreciably dependent on potential (relative to some reference electrode
probe), as is also Hg, the classic case for metals (Chapter 6).
We indicated in an earlier section of this chapter how the impedance behavior of an electrochemical system can be interpreted in terms of simulations
based on equivalent circuits and equivalent-circuit modeling. This approach,
however, is somewhat arbitrary, intuitive, and empirical, not to mention that two
or more equivalent circuits having different parameter values for the components may sometimes equally well represent the experimentally observed frequency-response behavior of the process. Only in the case of rather simple
charge-transfer processes not involving several steps with adsorbed intermediates or film formation can the equivalent-circuit approach represent, with some
good relation to reality, the frequency-response behavior of a Faradaically active electrode interface.
507
508
Chapter 16
'\
,,
FIGURE 16.14. Complex-plane impedance plot for a Faradaic process involving transition to diffusion control at low frequency, w, associated with the presence ofthe Warburg impedance element,
W, exhibiting a 45 phase angle.
with W in parallel with the double-layer capacitance is called the "Randles circuit" after Randles' paper in Discussions of the Faraday Society, 1947. 2
ell
Rs
Randles
circuit
In that discussion, Erschler8 presented a closely related paper. One ofthe earliest
papers on ac determination of double-layer capacitance was that by Proskurnin
and Frurnkin9 in 1937, while ac applications in polarography (ac polarography)
were made in an important paper by Breyer and Gutmann, 10 with later developments by Smith. II ,12
The superimposition of an av modulation on the slow sweep of potential is
used in polarography experiments to generate a response signal that is the differential coefficient of the polarographic wave, giving a maximum at the halfwave potential, EI/2, of the polarographic wave where the concentrations of
reduced and oxidized species are equal at the electrode interface; this allows a
maximum in the redox electron exchange rate to occur. This has relevance to the
evaluation of double-layer capacitor behavior where impurities in redox systems may lead to attenuation of ideal polarizability and the appearance of leakage currents, causing self-discharge (Chapter 18).
When diffusion control is a limiting factor in the kinetics of an electrode
process, the Warburg impedance, W, representing the diffusion behavior under
509
av modulation, gives rise to a 4SO straight line in the complex plane plot of Z"
vs. Z'. A transition from the semicircular Z" vs. Z' plot to a 45 line is often observed in the scan of OJ toward low frequencies, as illustrated in Fig. 16.14. The
Warburg impedance exhibits a frequency dependence on square-root of OJ-I
both in Z" and Z'; hence the 45 straight line. The square-root of the OJ relation
arises for quite general reasons connected with the dynamics of diffusion where
the time factor in semi-infinite diffusion leads to Wbehavior in one form or another (OJ-1 12 in ac impedance, t l12 in time dependence of concentrations, squareroot of rotation rate in rotating-disk electrode studies). The t ll2 relations arise for
fundamental reasons originating from the Stokes-Einstein equation for the
mean-square displacement, :x 2, of particles diffusing in a concentration gradient:
:x 2 = 2Dt or (:x2)112 =...J2Dt where D is the diffusion coefficient of the diffusing
species. Diffusion problems are dealt with through Fick' s two laws (differential
equations) representing the dynamics of diffusion:
flux equation: dN / dt =DAde/dx
(16.86)
(16.87)
where dNldt is the number N of moles of a solute diffusing per second (flux)
across a laminar of area A normal to the concentration (or chemical potential)
gradient, dc/dx, along the x direction, the vector of diffusion in the solution.
It should be noted here that a phase angle of 45, independent of frequency,
can also arise for quite different, nondiffusional reasons when a distributed RC
network arises in a porous electrode (Chapter 14). The frequency response is
then like that of a transmission line that also has a linear Z' vs. Z' plot with a
45 phase angle. This is experimentally observed in some high-area, porous,
carbon-based double-layer capacitors as was discussed in Chapter 14.
=Fk exp(aFE/RT)
(16.88)
.- .
I - Iss
l,d 2i (E - E)2
+ !Ji
dE (E - E)
ss + 2 d2
ss + ..... .
(16.89)
Chapter 16
510
If a modulation (E - Ess)
so that
I - Iss
di It
(16.90)
= dE
Therefore
(16.91)
(16.92)
H 20 + M + e- +-7
MHads + OH-
(step 1)
(16.93)
k_1
k,
2 MHads ~ Hz + 2M
(step 2)
(16.94)
(step 3)
(16.95)
=k l (1- (})exp(jJFEIR7) -
Vz
=kz(}exp(jJFEIR7)
k_l(}exp[(1 - P)FEIR7J
(16.96)
(16.97)
and
(16.98)
In terms of current density i
ifF = vI
511
+ V2 =To(E,e)
(16.99)
(16.100)
(16.101)
de
-=jw7)
dt
(16.102)
and
with
Tl = f(E,e)
Hence,
(ar,)
+ de
(a,,)
TO,ss
+ dE
Tl = Tl,ss
+ dE
TO
11 (E - Ess)
Oss)
(16.103)
and
..
TO - TO,ss
11 (E - Ess)
1'0
[ar,]
ae
'E(0
- Oss)
= T/ F and E - Ess = E
(16.104)
(16.105)
(16.106)
Then we can write
512
Chapter 16
(16.107)
IIF= -
'E
with
- (arllaE E+ (aro)ae
'E e
(qllF)jwe =
(16.108)
(16.110)
-(~)o
(~;l
-(~J, [(~l,
!IF
-YJW]
(~;1
(16.111)
r(~i) -~Jwl
E
and
(16.112)
in the usual way.
513
Note that Oss can be calculated by setting Eq. (16.100) equal to zero, i.e.,
(qtfF)(d8!dt) = VI - V2 - 2V3 = 0, the usual steady-state condition.
16.6.4. Relation to Linear-Sweep Modulation and Cyclic Voltammetry
16.6.4.1. Methodology
s =dVldt
(16.113)
Addressed to a capacitor of capacitance C, the response current l(t) to a linear voltage sweep is
Chapter 16
514
let)
= C (~~)= Cs
(16.114)
l(t)/s =C
(16.115)
or
The direction of current will depend on the sign of dVldt, which is periodically
reversed in cyclic voltammetry at the ends of the voltage excursion, Ll V, i.e., the
amplitude.
If C is constant with potential, cyclic voltammetry generates a rectangular
current response record, as illustrated in Chapter 3, which is symmetrical around
the zero-current line if the same s is applied in each direction of the voltage
sweep. For many systems, however, C is not constant with V in the sweep and
may not be constant with sweep rate when kinetic or diffusion limitations to currents arise, e.g., when C represents a pseudocapacitance (Chapter 10).
When a pseudocapacitance Cp is involved' (see Chapter 10), the following
equivalent circuit represents its behavior to a first approximation:
J-
Pseudocapacitance
circuit
dl
Rs
L..JIINVIMr-.
P:
l__ ~---]
RF
e,
'Here the symbol Cp is used for the pseudocapacitance for consistency with the impedance
literature (cf. C", in chapter 10).
515
516
Chapter 16
devices or systems, provides information on the dynamics of charge accumulation or charge removal from the interfaces of capacitors. The time dependence
of the voltage decline of a capacitor spontaneously discharging through a load
resistance is a related situation that was treated earlier in this chapter.
Since the two techniques of cyclic voltammetry and ac impedance measurement have dominated recent work on capacitor characterization and evaluation, it will be of interest to make some comparisons of the response functions
of the two methods. Some consideration of this question was first given in a paper by Conway, Kozlowska, and Gileadi.21
Basically, in cyclic voltammetry involving potential modulation at a rate
dVldt, over some adjusted voltage excursion, Ll V, the response function is a potential- and time-dependent current:
i =CCdVldt)
(16.116)
}""=jeoC
(16.117)
=LlVI(dVldt)
(16.118)
(16.119)
517
which is analogous to a modulation frequency, w. Now in the case of an alternating voltage signal of amplitude Yo, Vet) = Vo sin wt, there is a maximum rate
of voltage change twice during a full cycle; thus
dV(t)ldt= wVo cos wt
(16.120)
(16.121)
i.e., when the response current is a maximum and the voltage crosses the
half-amplitude value (see later discussion). Thus, some correlation between
the current-response behavior of a system to cyclic modulation of V by repetitive linear sweeps over a constant voltage range and the current response
in av modulation can be realized, if not exactly in an analytical way, then
conceptually. The situation is illustrated schematically in the working diagram
shown as Fig. 16.15. Simplistically, from Eq. (16.121), the ratio of equivalent
sweep rate, s, to w in av modulation is
slw= Vocos
(16.122)
OJt
v(tl
/\ ----r'
,, ,,
"
vltl
C.V.
l!.v -h~;-:-+--#---\----+-~
\.L
a.V.
(dV/dtl mox
\ (dV/dtl mox
\
FIGURE 16.15. Working diagram illustrating relation between a sinusoidal voltage signal and a
cyclic voltammetry voltage signal in relation to frequency and dV/dt in the halfcycles.
Chapter 16
518
(16.123)
Similarly, s at the top or bottom of a sinusoidal signal is zero, as is obvious, although d 2Vldi2 is finite. In addition, the reduced maximum sweep rate is sma/VO,
which is evidently identical with w.
In the case of av modulation addressed to a capacitance, the maximum current flows when dVldt is a maximum, i.e., at wt = 0 or n, while zero current flows
when V = Vo, at the voltage maximum. This corresponds to the current being out
of phase with the voltage by n/2, the well-known behavior at a capacitance.
Conversely, at a pure resistance, the current and voltage vectors are always in
phase.
For the same charging current to be generated by cyclic voltammetry at a
capacitance as the maximum generated by av modulation at the same capacitance, we would have
CdVI dt =-jVowC
(16.124)
(16.125)
(compare Eq. 16.121), noting that the impedance is Z" = -j1wC or the admittance
y" =-jwC (Eq. 16.117).
16.6.5. Impedance of a Pseudocapacitance
519
age, H in the above reaction, of the electrodeposited species, or that of the concentrations of Ox and Red species in a redox reaction. Cp = ql deldV in the general case where ql is the charge required for formation or desorption of the
species or for the total interconversion of Ox to Red in a redox process. A small
but significant solution resistance, R" is usually combined in series with the
Faradaic impedance involving RF and Cpo The circuit elements are combined as
follows:
eel
(a)
z"
10
(b)
FIGURE 16.16. Complex-plane impedance plots for a Faradaic pseudocapacitance. (a) with no
desorption resistance (UPO case), and (b) with desorption resistance, RF (OPO case).
520
Chapter 16
lected parameter values for the circuit elements involved, taking Cd1 typically as
20 flF cm- 2
The interfacial part, Z, of the overall impedance is given by
(16.126)
(16.127)
and Z is combined in series with Rs. From Eq. (16.127) it can be seen that two
limiting cases arise:
1. at very low frequencies (w
i.e., Cd1 and Cp are effectively added as in a simple parallel arrangement and the
measured C =Cd1 + Cpo
2. at sufficiently high frequencies (jwRFCp 1)
(16.129)
(16.130)
or
(16.131)
(16.133)
z= l/jwCd!
521
(16.134)
i.e., the impedance is determined by Cd! at sufficiently high frequencies, combined with Rs if the latter is significant.
In the case where the potential is beyond the UPD range, another Faradaic
or chemical potential-dependent RF arises in parallel with Cp , corresponding to
the equivalent circuit:
II
11
Rs
RF
"
lie
1>
This situation corresponds to overcharge of the pseudocapacitance where a continuous Faradaic reaction takes place (current leakage across Cp ) in the overpotential region for the process, e.g., the H2 evolution process in the HER, the first
step of which formally gives rise to UPD at potentials lower than the thermodynamic potential for the overpotential deposition reaction.
The Faradaic or chemical resistance, R F, in the above circuit corresponds
to the reciprocal of the potential-dependent rate (Tafel behavior) of the process,
which enables the intermediates that give rise to Cp to become desorbed or combined to form an overall product (e.g., H2 or metal crystallites) of the reaction.
This allows a continuous Faradaic current to pass that is potential dependent according to a Tafel relation.
The circuit shown for desorptive leakage of Cp obviously allows continuous direct current to pass through Rs + RF + RF in series in the limit of w ~ 0
for which Zcd and Zcp are 00. There can then be two semicircles and three intercepts on the Z' axis. Note that the first circuit for pseudocapacitance involving
UPD does not allow continuous passage of direct current because of blocking
by Cd! and Cp as w ~ 0, and there are then only two intercepts and a Z" rising
to 00 as w ~ 0 (Fig. 16.16).
The transition to dc leakage, corresponding to overcharge of Cd! and Cp ,
occurs, of course, when the potential difference across the device exceeds the
thermodynamic potential for the overall (OPD) reaction (or some other alternative pathway) for which the UPD process is formally a constituent step (e.g., as
in the HER) at potentials below the reversible potential for that overall reaction.
In capacitor devices, this situation corresponds to overcharge.
522
Chapter 16
or
(16.138)
At sufficiently high w, Z becomes
Z=--+-jwCz jWCd1
(16.139)
i.e., the Cd1 and C2 combine (reciprocally) as a pure capacitance of Cd1 and C2
in series, as is obvious from the above equivalent circuit when w ~ 00, [Z (Cdl>
C2) R F ]
At sufficiently low w, Rp(jwRFCd1 + 1) becomes RF and Z is then the impedance of RF in series with Cz, namely,
(16.140)
523
together with any solution resistance, if significant. This gives a rising vertical
plot of Z" in the complex-plane with decreasing W having an intercept of RF on
theZ'axis(Fig.16.16).
After rationalizing the denominator, the overall impedance, not including
an R" is given by
(16.141)
(16.142)
from which the real and imaginary components of Z are recognized. The above
relation reduces to the limiting cases for W -4 0 or W -4 00 identified above.
Note that unlike the equivalent circuit
considered earlier, the capacitance C2 introduces blocking ofthe alternating current at low w, so Z -400 as W -4 O. In the circuit shown on p. 521, when W -4
00, there is admittance to Cp through Rs and RF in series. In the circuit involving
C2 , an initial semicircle associated with the frequency response of Cdl and RF
with its maximum at We = lIRF Cdb arises with an intercept at Z' = RF (or RF +
Rs, if Rs "# 0); beyond the intercept at RF, with continuing scan of W to lower values, the Z is determined by lIjwC2 , i.e., the vertical line of increasing Z' appears
with decreasing w, as illustrated in Fig. 16.16.
Note that for the above cases and the ones considered earlier, RF will normally be potential dependent according to a Tafel exponential function in Vor
'7. It could be represented in the equivalent circuit as an exponentially variable
Faradaic resistance or, near the reversible potential, by a linearly variable resistance, which is the basis of treatments of small-amplitude av modulation around
the reversible potential of a Faradaic process (Eqns. 16.5 and 16.15).
Finally, it should be noted that self-consistency of analyses and modeling
of frequency-response of various RC or RCL circuits can be usefully made in
terms of the so-called Kramers-Kronig relations,4,22 referred to earlier.
524
Chapter 16
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Chapter 17
In earlier chapters it has been emphasized that the high-area, porous carbon
structures that provide large specific (per gram) capacitance values (in practice,
5 to 50 F g-l) cannot be represented by a simple capacitance or even by a simple
RC circuit. The diagrams in Fig. 17.1 represent a hierarchy of equivalent circuits, from those for a simple capacitor through those involving simple combinations of a capacitor element with either one or two ohmic or equivalent ohmic
resistors, to more complex equivalent circuits involving distributed capacitance
with ohmic elements in series or parallel coupling with the capacitative elements and, for some conditions, with Faradaic leakage resistances in parallel
with the capacitative elements. The latter may also include contributions from
pseudocapacitances associated with surface redox functionalities (Chapter 9) on
the interfaces or edges of carbon particles or the main pseudocapacitance associated with redox oxide capacitor materials (Chapter 11). In some cases, considered later, the equivalent circuit may also include an inductive element, L.
'This Chapter owes much to the derivations and analyses given by Dr. Miller at the various
Deerfield Beach (Florida) seminars on electrochemical capacitors held in recent years.
525
526
Chapter 17
----1 ~
a) Simple capacitor
b)
~
RF
FIGURE 17.1. Hierarchy of equivalent circuits, from that for a pure capacitor, with no esr, to that
for a porous electrode with distributed C and R components.
527
(17.1)
analogous to Ohm's law, where Vet) = Vo sin wt and /(t) =10 sin (rot + rp), are the
time-variant voltages and currents that are modulated according to the frequency w, as expressed earlier. Z(w) may include ohmic (R), capacitative (C),
and sometimes inductive (L) elements; the latter two are dependent on frequency according to Z(w) = l/jwC and Z(w) = jwL, respectively, as displayed
below.
The elements of equivalent circuits and their respective dependences on
frequency are as follows:
Z(w)
Element
Resistor, R
Capacitor, C
Inductor. L
R (dR/dw=O)
l/jwC=-jwC
jwL
where j, as usual, is ~.
In a series arrangement of n impedances
n
~otal =IZ;
(17.2)
;=1
l/Ztotal = I
liZ;
(17.3)
;=1
so that for this case, the smallest of the Zs principally determines the overall Z.
In electrochemical systems the same R, C, and L elements can be recognized as in hardware circuits, except that C is a double-layer electrostatic capacitance or a pseudocapacitance (Ct/J; see Chapter 10) and L is a
pseudoinductance, usually of connecting wires the effect of which is usually
small except at high frequencies or rapid transients. Another qualitative exception is the existence of a so-called Warburg impedance element (represented by
W) that arises when an electrochemical process is or becomes diffusion controlled. The element W contains two components, a diffusional capacitance and a
diffusional resistance (see Chapter 3), each of which is dependent on the squareroot of frequency. In the material that follows, the behavior of various elementary RC and RCL equivalent circuits will be examined, including the case of
distributed capacitance and resistance elements that represent the behavior of
high-area porous electrodes.
528
Chapter 17
Z(w)
= esr + l/jwC
529
(17.4)
which tends to Z(w) = esr at infinite w. This result usually provides the formal
means of evaluation of esr of a capacitor device. Thus, in complex-plane plots
of the imaginary (Z") vs. the real (Z') components of the impedance vector (Z)
at various frequencies (Chapter 16), the effective esr (including any series electrolytic resistance) is revealed as the intercept on the Z' axis as w ~ 00.
The involvement of an esr in electrochemical capacitors is a factor of major
importance in their electrical behavior, especially their power performance
(Chapter 15). The presence of significant esr restricts the rate at which the capacitance can be charged or discharged upon application of a given voltage difference between the electrodes or the alternating currents that can be generated
at a given moment by applying an alternating voltage at a given frequency. The
esr limits the power at which the device can be operated from a given direct or
alternating voltage, and is thus one of the most important properties of a supercapacitor, in addition to its capacitance density in F g-l or F cm-3 . The impedance of the capacitor device then differs from that of a pure capacitor because a
frequency-dependent phase angle (less than 90) is developed between the alternating voltage and the resulting ac modulation, as indicated earlier on the basis of general principles. Power loss and energy dissipation then ensue.
In the case of electrochemical double-layer capacitors, the esr situation is
usually substantially more complex than that illustrated in the equivalent circuit
here. As explained in Chapter 14, distributed electrolytic, interparticle, and intraparticle resistances arise because of the porous, particulate nature of the electrode matrix. Rather similar behavior can arise with oxide-type redox
pseudocapacitors (e.g., Ru02) having high-area microporous matrices.
In the above case there is no unique esr except the component of resistance
ofthe external solution. The impedance behavior, plotted in the complex plane,
exhibits a 45, almost linear relation at elevated frequencies (transmission-line
effect; Chapter 14), with an intercept on the Z' axis as w ~ 00, corresponding to
the resistance of the solution and any separator external to the outer surface region of the porous electrode matrix.
In state-of-the-art double-layer-type capacitors, effective esr values are in
fractions of milliohms up to a few tens of milliohms. The higher values occur
when electrochemical double-layer capacitors are designed for higher voltage
operations using nonaqueous electrolyte solutions, which have intrinsically
lower specific conductances.
The minimization of esr is obviously desirable in constructing electrochemical capacitors that are intended for operation at elevated power levels (see
Chapter 15).
530
Chapter 17
tcutoff = Resr
xC
(17.5)
The initial voltage and current drawn at the start of discharge and hence
their product, the power, P, depend on the energy, E, stored in the capacitor device so that6
PIE
= 21tcutoff
(17.6)
S-1
1000
100
10
LL
......
Cl)
g
u
'r
0.1
a.
0
0.01
0.001
0.01
10
0.1
100
IZI/Ohms
FIGURE 17.2. Impedance plots for ten commercial capacitors showing cutoff of capacitance
with decreasing esr component. I Soleq 470 F; 2 Soleq 70 F; 3 Maxcap 1 F; 4 NEC 0.47 F; 5 Maxcap
0.47 F; 6 Maxcap 0.22 F; 7 NEC 0.1 F; 8 Maxcap 0.1 F; 9 Maxcap 0.047 F; 10 Sprague 0.022 F.
(From Eisenman. 6)
531
The physical construction and materials of the capacitor determine PIE, while
the actual capacitance and maximum charging voltage do not enter as variables. 6 The current U.S. Department of Energy goals specify a PIE ratio of
0.0278 s and a cut-off time of 72 s, i.e., twice the 36-s timeline of the tests. Then
the power at the time of the short circuit would be (after a revision of the Equation in Ref. 6):
P =
(17.7)
532
Chapter 17
(17.8)
(17.10)
The d.f. = Z'IZ" = wRC and dJ. = 1 when w = lIRC or/ = 1I2nRC (2n/ = w).
where/is the frequency in radians/s.
Similar considerations apply to the parallel RC circuit treated in Chapter
16 in order to demonstrate the origin of semicircularity in the complex-plane
plots. For this case, the impedances of the components add up reciprocally:
(17.11)
Z = _R_(1_---'l:-w...".R_C:'-)
1 + w2R2CZ
533
(17.12)
and
(17.13)
As derived in Chapter 16, the real and imaginary components of Z are
(17.14)
and
(17.15)
while Z' and Z" combine, as shown in Chapter 16, to give the complex-plane
plot
(17.16)
RF (1 - jwRFC)
w2 2CZ
1 + RF
(17.17)
where RF is a Faradaic, parallel leakage resistance that may be potential dependent according to a Tafel exponential factor. In this case, a semicircular plot of
Z" vs. Z' is retained, but it is displaced along the Z' axis with an intercept of
Rs. The equation for the semicircle is then
(Z' - Rs - RFI2)2 + (Z')2 =(RFI2)2
(17.18)
Z =R + --"---s
(1
jwR~C
(17.19)
+ w2R~CZ)
from which Z' and Z" are directly recognized and distinguished. The phase angle
tan ifJ = Z"/Z' is found as
534
Chapter 17
+ w2R~CZ)
tan t/J = Rs + RF/(l + w2R~CZ)
jwR~C/(l
-jwR~C
(17.20)
(17.21)
For this case, note the limiting results that for w --t 0, tan t/J --t 0; and
for w --t 00, tan t/J also --t 0, so there is a maximum in tan t/J, as is commonly
observed. When Rs --t 0, then tan t/J = -jwRFC, which varies from 0 to 00 with
0< w < 00, i.e., there is no maximum with increasing w. This is the situation
in dielectric relaxation plots 7 of the real and imaginary components of the dielectric constant as a function of frequency.
When RF --t 00 (no leakage currents), the circuit becomes that for a series
RC connection. For tan t/J = 00, t/J = n12, the phase-angle for a pure capacitance,
while for tan t/J = 0, t/J = O.
Although most circuits involving electrochemical capacitors do not include any deliberatley added inductances, except perhaps for some resonance
purposes, adventitious inductance can be introduced that is significant, e.g.,
when large current pulses are employed in charge or discharge. Then it is the
wiring of the circuit that can introduce a significant self-inductance.
In some other cases, however, where special types of electrode processes
(e.g., metal oxide film formation and passivation) are involved, an intrinsic
pseudo inductance may be exhibited in the impedance frequency response. This
type of situation can arise when inhibition of an electrode process leads to Tafel
behavior that corresponds to a negative resistance (current decreasing with increasing overvoltage), or when surface coverages by some adsorbed intermediates decrease with increasing overvoltage. In the case of corrosion processes
involving onset of passivation, very complicated Z' vs. ZIt complex-plane plots
appear in two or three quadrants, where reentrant curves arise and ZIt enters the
positive quadrant. This is usually due to a situation of decreasing current with
increasing potential, i.e., a negative resistance arises.
In a series R, C, L circuit
Z = R + l/jwC + jwL
=R + j(wL -
lIwC)
(17.22)
(17.23)
and
(17.24)
with
wL - 11wC
tan'f'=---R
535
(17.25)
Rs
two semicircles can arise in the complex plane plot with a high-frequency intercept on the Z' axis of Rs (Fig. 17.4).
In another case for a series/parallel combination, namely,
the complex-plane plot has the form shown in Fig. 17.5, where a single semicircle becomes transformed to a vertical line in the -Z" direction with decreasing frequency; the intercepts on the Z' axis are Rs + R1, and Rs at high frequency.
536
Chapter 17
ESR
Z
<1
o
W
a.
\ /Z
\1
FREQUENCY
FIGURE 17.3. Frequency dependence of the impedance Z of an RCL series network involving
an equivalent series resistance. Rs. that is finite or zero. (After Miller. 1,2)
The vertical element of the plot corresponds to the impedance (ZII) of a pure capacitor (Cl ) as w tends to zero. When w becomes large enough for Zc, to be
small compared with Rs, then the behavior approximates that of the simpler Rs,
R 1 C2 series/parallel circuit considered earlier. Curves of the type shown in Fig.
17.5 are commonly found in a number of studies of the frequency-response behaviors of electrode systems where some film capacitance becomes coupled
with a Faradaic impedance, R 2C 2 in parallel, as in the series double CR circuit
on p. 535.
Miller 2 has considered the behavior of the general RCL capacitor circuit
for various conditions. For the combination
-Z"
FIGURE 17.4. Complex-plane impedance plot for two parallel RC equivalent circuits in series
with each other. and in series with an esr of Rs.
537
.:(0
-1:'
FIGURE 17.5. Complex-plane impedance plots for a series and parallel equivalent-circuit combination, with transition to purely capacitative behavior (vertical line) at sufficiently low frequencies.
RF Rs for practically developed double-layer capacitors where care in engineering and electrochemical design reduces Rs to a small number of milliohms.
For such conditions,
Z =Rs + j[2nfL - l/(2nfC)]
(17.26)
fr = l/[2nCLC)112]
(17.27)
Then a practically important property emerges: such a device will store electrical energy efficiently only for frequencies less than/r . This emphasizes the importance of minimizing design and circuit features that could introduce
significant series inductance. Sincelr is determined by CLC),-1I2 it obviously follows that the condition I <Ir for optimization of energy storage is more easily
attained when L is small. The four- component circuit above reduces to a simpler series circuit for the conditionf <fr and when RF Rs' the practical desideratum referred to earlier. Large RF , of course, will reflect good charge storage
over time since RF determines the self-discharge kinetics of the parallel RFC
combination, it being generally required that RF be maximized or ideally 00 for
an ideally polarizable double-layer capacitor, C.
Figure 17.6 shows test data from Miller2 for a typical practical electrochemical capacitor based on evaluation of Z' and Z" over a wide frequency
range between 1 mHz and 1 MHz showing a self-resonant frequency of 15 kHz.
538
Chapter 17
lk.----------------------------------. lk
2 ..
2
10
E
.c
0
.......
".. ............ .
- _____________
100m
'- "-
10m
1m
____
100
10
:.1.uu~
.c
.......
"-
100m =N
10m
'-
1m
'-
Ik
lOOk 1M
FIGURE 17.6. Impedance spectra of a typical practical electrochemical capacitor. The real and
imaginary parts of the impedance are shown. The dashed lines show the behavior of an ideal resistor-capacitor series combination. Self-resonance frequency, 15 kHz. Original figure based on ca.
100 points. (From Miller?)
The dashed lines show the behavior of an ideal resistor-capacitor series combination, which clearly illustrates the deviation of the practical capacitor from
such ideal behavior.
17.4. DISCHARGE OF A CAPACITOR WITH ESR INTO A LOAD
RESISTANCE, RL
Rs
---Vo
539
For the matching load situation (Rs =RL), the maximum power is found as
P max
= VO/ 4Rs
(17.29)
(17.30)
i.e., it arises at V = V012 or half the potential for full discharge. Note that when
RL = 0, P = 0, and when RL = 00, PL also is zero, while a maximum in P arises
between these conditions for the load resistance. This situation is similar to that
found in the theoretical analysis for Ragone plots of power density vs. energy
density (Chapter 15).
For a series RC circuit with a resistive load, the power supplied in relation
to the maximum power utilized (P max) is given by
P = [4P max r/(1
(17.31)
where Pmax = V6/4Rs, as shown earlier. The parameter r =RL/Rs is the ratio of
the load resistance to the series or equivalent series resistance. r = 1 is referred
to as the "matched load" condition. The time parameter r = tiRsC, where RsC is
540
Chapter 17
the time constant of the Rs' C combination. r is thus a reduced time scale, as
arises in other related problems (Chapter 16, Section 16.3.2).
For a voltage range of Vo (the initial voltage), down to Val2, analytical results calculated by Miller 12 are shown in Fig. 17.7. It is seen that when r = 1
(matched load), PIPmax (ordinate scale of the plot) becomes unity since (l + r)2
is then 4 and exp[-2r/(l + r)], when r = 1 and t =0, is also unity. The supplied
power, P, is equal to the maximum power when r = 1, the matched-load condition.
The origin ofEq. (17.31) is as follows. Consider a capacitor of capacitance
C having an esr of Rs discharging into RL from an initial voltage, Vi, and a voltage during discharge of Vt across the capacitor. Then, at any time,
(17.32)
where the latter two Vs are the voltages developed across Rs and RL, respectively, when a discharge current I(t) passes.
The current for discharge of the capacitor is:
1/ = Ii exp[-tIRC]
(17.33)
also
(17.34)
and
)(
E
a..
.....
0.1
..J
a..
r =10
FIGURE 17.7. Behavior of a series (esr) RC circuit with power delivered into a resistive load
(RL) for discharge from initial voltage Vo down to VoI2 based on results analytically calculated by
Miller. I2) Parameter r =RL/ Rs and r = 1 is a matched load.
541
(17.35)
For an initial charge, qi' on the capacitor,
(17.36)
At t = 0, Vi - IiRS - IiRL = 0 or IiR = Vi where R is written for Rs + R L. Then integrating Eq. (17.36),
=Vi -
=Vi exp[-tIRC]
(17.37)
(17.38)
(17.39)
= ItVt
(17.40)
(17.41)
vr
Rs+RL
(17.42)
Rs+RL
Vr 14Rs.
4Pmax
(17.43)
4Pmax
(17.44)
P =- - . exp[-2tlC(Rs + R L)]
1+r
= - - . exp[-2r/(1 + r)]
l+r
542
Chapter 17
(17.45)
4rPmax
(1 + r)
2 .
exp[-2r/0 + r)]
(17.46)
4Pmax
(1
+ r)
2 .
exp[-2r/(l + r)]
(17.47)
The power drawn from the capacitance minus that lost in Rs is the power
developed across the load, i.e.,
Peap -
Ps = 4Pmax
O+r)
4Pmax
2 exp[-2r/(1 + r)] 2 exp[-2r/(1 + r)]
(1 + r)
(1 + r)
(17.48)
4Pmax r
(1
+ r)
2 .
exp[-2r/(1
(17.49)
+ r)]
which is the required time dependence of power developed in the load from the
discharge of the capacitor.
The behavior of several RC or RCL systems that may be equivalent to that
of any practical capacitor has been simulated in calculations by Miller. 2 A firstorder model for any capacitor is (cf. p. 537), in the simplest analysis:
c
L
Rs
where Rs and Rp are series and parallel resistive components coupled with C. L
mayor may not be significant. Rp can be a Faradaic (potential-dependent) reaction resistance when C is a double-layer capacitance. Rp is much greater than
Rs for practical capacitors except when Faradaic leakage currents arise, leading
to self-discharge (see Chapter 18). Z is then (cf. Eqn. 17.23)
543
Z = Rs + j[wL - lIwC]
(17.50)
10
II)
E
.c
II)
100m
N
I
10m
: ......
E
.c
......
10m N
1m
1m
100m
10
100
1k
10k
lOOk
Frequency I Hz
FIGURE 17.8. Z" andZ' components ofthe impedance of the RCL equivalent circuit as a function
of frequency over a wide range (log scale) for a constant real component of I ohm. (Resonant frequency,JR = 1I[2n(LC)1I2) is ca. 3 x 102 Hz). (From Miller?)
544
Chapter 17
Zp =
d
~ ~ j) x coth [t -J wC
2nn r KWC !
Kr
(1 + j)]
(17.51)
where n is the number of pores in the electrode, r is the radius of a typical pore,
assumed to be cylindrical, K is the conductivity of the electrolyte, w is the angular frequency as usual, Cd! is the double-layer capacitance per unit area, and
t is the length of a typical cylindrical pore. Details of the treatment of impedance of a pore, based on de Levie's analysis, were given in Chapter 14.
Obviously this model is seriously oversimplified since, e.g., there is a distribution of pore lengths and diameters and there may be fractal structures that
represent the pore geometries better than a simple cylindrical one. However,
only with such a simplified model can an analytical result for Zp be derived and
that, it is seen, is already quite complex. More complex geometries were treated
by Keiser, Beccu, and Gutjahr4 (Chapter 14).
Nevertheless, two limiting cases are informative and useful:
1. the high-frequency limit
Zp -
(l-j)
2nn ...J r3KWCd!
(17.52)
and
2. the low-frequency limit
(17.53)
=-j/wC
(17.54)
since C is the overall double-layer capacitance for a total surface area of S =2nr
t nand C =SCdl .
The complex impedance of an electrode composed of uniform pores (as
an approximation) is as shown in Fig. 17.9 in which Q is the ionic solution
resistance within the porous structure (typically tens of milliohms or less in
well-engineered capacitors) and Rs is the (external) equivalent series resistance. It is seen that at low enough frequencies the impedance behavior becomes that of a pure capacitor, with Z" increasing as l/wC in a vertical line
having an extrapolated intercept on the Z' axis for infinite w of Rs + Q, while
the actual experimental high-frequency intercept is Rs at the high-frequency
limit of the 45 line, corresponding to transmission-line behavior below the
knee of the curve. The distance between the two intercepts in Fig. 17.9 gives an
estimate of the internal resistance Q, which can then be optimized to maximize
good power output performance.
545
w
J
:
N
FIGURE 17.9. Complex impedance behavior (approximation) of an electrode composed of un iform pores. Rs is the esr and Q the electrolyte resistance within the pores. (From Miller.' Reproduced by permission of The Electrochemical Society, Inc.)
0.1 uH
1.1
0.2
0.44
174
1.5
r-~+'~~--~--'-~~.-~~-r--~--'-~N--'
I~
________- L_ _ _ _L -_ _- L_ _ _ _
_ _~
FIGURE 17.10. Five time-constant equivalent circuit for a porous electrode situation with indicated selected values of R and Cd', and with L =0.1 JlH. Various other circuit models are possible.
(From Miller. ',2 Reproduced by permission of The Electrochemical Society, Inc.)
546
Chapter 17
~---------------------------------------, 100
E
o
III
E
.s:::.
o
.......
N
.s:::.
10
100m
.....
III
100m
10 m
10m
100m
10
100
1k
10k
lOOk
...CIJ
N
IN
Frequency / Hz
FIGURE 17.11. Frequency-response behavior of the five-time-constant network of Fig. 17.10.
Original figure based on ca. 100 points. (From Miller. 1.2 Reproduced by permission of The Electrochemical Society, Inc.)
I k ----------------------------------------~
lk
100
100
10
10
.c.
,~
100 m
100m
10m
10m
_......
N
1m
10m 100m
10
100
Ik
10k
lOOk
1M
FREQUENCY I Hz
1. 1.-7
.... '
.~
-~
'N\
..,.
. 69
,.
'"oM.
2 .2
4 '3
-- .'.'~ .,
FIGURE 17.12. Frequency-response of a memory backup electrochemical capacitor and the corresponding five-element equivalent circuit which gives a good fit to the frequency dependence of
the impedance components 2' and 2". (From Miller. 1,2 Reproduced by permission of The Electrochemical Society, Inc.)
547
element, C (Z" = lIjwC), while the ascending part is due to the inductive element, L, for which Z" =jwL. The minimum (at the self-resonant frequency) corresponds to a phase angle of zero where the opposing contributions of C and L
to Z" cancel. Then the impedance is just that due to the ohmic component, R.
Even with L = 0.1 ,uH, because C is very large, the self-resonant frequency is
relatively low, in the range 1 to 10 kHz.
(17.55)
where Vet) is the voltage after time t and Vi is the initial voltage before discharging starts. Equation (17.55) is a general one expressed in terms of a reduced time
scale, tIRC, as explained earlier.
Note that the time constant for each capacitative element is calculated by
multiplying the capacitance of that element by the series sum of all the resistive
elements from that RC element to ground, i.e., CR = C x ER (to ground). The
RC elements furthest from the load (Fig. 17.12) have the longest time constant
and hence the slowest frequency response. The fastest response is given by the
RC element directly coupled with the load resistance.
This circuit has a fairly wide range in its frequency response-on the order
of 0.5 kHz to ca. 0.05 Hz. Since all the C elements are accessed through a series
548
Chapter 17
457
% TOTAL
CAPACITANCE
0.0426
Load
(At)
RCllME
CONSTANT ($)
2.9
6.3
13.7
2.4
23.2
0.40
37.8
0.19
22.4
0.070
FIGURE 17.13. Cumulative capacitance of the five-element equivalent circuit required to simulate the wide-band frequency response of an Evans Capattery device. (From Miller?)
549
mtJ
(17.56)
where
(17.57)
Here a' = aAtS where A = cross-section of pore in cm2 and S = surface area per
unit length of pore. At sufficiently high frequencies, w ~ 00 or Imtl ~ 00; the
coth term ~ 1 so Eq. (17.57) becomes reduced to the Warburg form (i.e., analogous to the frequency dependence of the Warburg diffusional impedance):
Z( w ~ 00)
1- j
=---.:--1
(2SCd1w)
12
(17.58)
550
Chapter 17
I
I
I
I
I
I
I
I
I
I
FlfliIe
Porous
Eleclrode
::
N
I
00-+0
IPlanar I
z
FIGURE 17.14. Comparative impedance plots for a redox pseudocapacitance in the complex
plane for a planar electrode (ideal case), and finite porous and semi-infinite porous electrodes.
(From Tilak and Chen. 5)
giving rise to a phase angle of 45. However, this result arises because of the distributed resistance of the equivalent circuit representing the porous electrode
(transmission-line model) rather than any involvement of diffusion.
A wide-ranging series of forms of the complex-plane plots can arise for a
simple redox reaction, depending on pore length, conductivity in pores, magnitude of the Faradaic charge transfer resistance, and the associated pseudocapacitance. The resulting behaviors represent several features of the frequency
response of RuOz pseudocapacitor electrodes.
In Eqs. (17.56) and (17 .57), p is the Laplace transform variable; S is the
surface area of a pore of length per unit length; (J' is the specific conductance
of the solution (ohm-1cm- 1); A is the cross-sectional area; Rt is the Faradaic
charge transfer resistance of the electrode process, related reciprocally to the exchange current density, io; and RI is (io nF)/(RT) (l + q*/nf) where q* is determined by the component variations of the charge density on the metal with the
surface excesses of the redox species and R in the n-electron redox process,
i.e., q* is related to the electrosorption valencies of the and R species.
t,
551
4r-------r-~------------------~
I
1
1
1
1
1
11.0
N
1
CT'
=- 100
1
1
Z'
FIGURE 17.15. Typical plots of impedance for imaginary vs. real components for a porous electrode based on Eqs. (17.56) and (17.57), used in the calculations of Tilak and Chen5 for three values
of conductance, (1', of the solution and two values of the pore length, l
These calculations gave a useful comparative idea of the impedance behavior of electrodes of various forms (see Ref. 4) in relation to the experimental frequency response of Ru02 pseudocapacitor-type devices. As may be intuitively
expected, a porous redox pseudocapacitor behaves similarly to a distributed
double-layer capacitance and connected resistance networks as is demonstrated
by Tilak and Chen. 5
In relation to the results of Tilak and Chen,5 discussed above, it is useful
to demonstrate the progression of frequency-response behavior for an RFC
pseudocapacitance couple with increasing frequency from 1 to 105 Hz. This is
shown in Fig. 17.16 for the complex-plane plots and in Fig. 17.17 for the corresponding phase-angle plots. With decreasing frequency, the impedance behavior approaches that of a pure capacitance, as for a porous electrode.
A general summary of frequency-response behaviors for equivalent circuits of increasing complexity is presented diagrammatically in the series 1 to
7 in Fig. 17.18.
10
. i.
"
15
20
R-5
-2
-22
-42
-62
o
10
20
R-10
-30
-50
-70
-90
-110
-130
20
40
R=30
-120
-170
-220
-270
-320
-370
50
100
R-loo
FIGURE 17.16. Series of complex-plane plots for a Faradaic pseudocapacitance with various values of the charge transfer
resistance, RF in ohms, and the pseudocapacitance, C.p, illustrating transition from an RC frequency response to a pure C response with decreasing frequency. (Calculations by Qian and Conway, unpublished.) (RF shown as "R").
-2
-12
-22
-32
-42
-52
-62
~...
....-..I
o
::rIII
(J1
10
100
100
1000
1000
100000
10000
100000
R=100
10000
R=30
FIGURE 17.17. Series of plots of phase angle vs. log frequency in hertz corresponding to the four series of
complex-plane plots in Fig_ 17.16 for the same four RF values. (Calculations of Qian and Conway, unpublished.)
10
C_2000~F
0
100000
0
10000
-10
-10
1000
-20
-20
100
-30
-30
10
-50
-40
-60
-60
-40
-70
-50
-80
-70
-10
-10
-90
-20
-20
-80
-30
-30
-90
-40
100000
-50
-40
10000
-60
-60
-50
1000
-70
-70
100
80
80
10
90
90
-I
C11
C11
Co)
c:
a"
1/1
c:
-...
o
...
<
AI
::J"
(1)
(1)
(')
::J
AI
3"
9-
It
::J
(1)
554
Chapter 17
EQUIVALENT CIRCUIT
-z" jro
/
ro -co
o
2. Interfacial capacitance with
ohmic series resistance. Rs.
e.g. solution or circuit resistance.
-z
.
o
z'
-z"
ro-co
Rs
ro=o
'--!!_ _ _ _
R --..40~/
Rs
z'
-z
~~---+~~~~~
ro=o
ro __ co
FIGURE 17.18. Hierarchy of equivalent circuits of increasing complexity and their corresponding frequency-response behaviors plotted in the complex plane. (RF and R~ in circuits 4 and 7, respectively, have potential (V)-dependent values). (Diagrams are schematic).
555
:z:'
o
6.
-Z'
As
11
R, L __ R,\jl
i
7.
As
556
Chapter 17
REFERENCES
1. J. Miller and B. E. Conway, notes for an Electrochemical Society short course on
electrochemical capacitors at the Electrochemical Society meeting, Chicago, II. October 1995,
Electrochemical Society, Pennington, N.J. (1995).
2. J. Miller, in Proc. Second IntI. Symposium on Electrochemical Capacitors and Similar Energy
Storage Devices, S. P. Wolsky and N. Marincic, eds., Horida Educational Seminars, Boca
Raton, Ha. (1992).
3. R. de Levie, Electrochim. Acta, 8, 751 (1963).
4. H. Keiser, K. D. Beccu, and M. A. Gutjahr, Electrochim. Acta, 21, 539 (1976).
5. B. V. Tilak and Y. Chen, in Proc. Second Inti. Seminar on Double-Layer Capacitors and
Similar Energy Storage Devices, S. P. Wolsky and N. Marincic, eds., Horida Educational
Seminars, Boca Raton, Ha. (1992).
6. E. T. Eisenmann, in Proc. Fifth Inti. Seminar on Electrochemical Capacitors and Similar
Energy Storage Devices, S. P. Wolsky and N. Marincic, eds., Horida Educational Seminars,
Boca Raton, Ha. (1995).
7. K. S. Cole and R. H. Cole, 1. Chem. Phys., 9, 341 (1941).
Chapter 18
Self-Discharge of Electrochemical
Capacitors in Relation to that at Batteries
18.1. INTRODUCTION
Charged capacitors, like charged batteries, are in a state of high free energy
relative to that of the discharged state, so there is a pseudo-driving force for their
self-discharge. The driving force is thermodynamically the same as that which
provides current and power when the capacitance is discharged through a load
resistance, but then the resulting process is one of controlled discharge, which
is different from that for spontaneous self-discharge on open circuit. The thermodynamic situation of instability (high free energy) of the charged state is illustrated in Fig. 18.1. The same kind of considerations apply to self-discharge
of battery systems, where a similar phenomenon is well known.
The phenomenon of self-discharge of an electrochemical capacitor diminishes its performance characteristics (e.g., power and energy density) because
there is a decline in voltage on open circuit after charging. This introduces an
unreliability factor in energy storage by a capacitor, depending on the type of
use to which it is put. The self-discharge characteristics of a capacitor device are
important in evaluating its performance and commercial specifications. Knowledge of the mechanisms of self-discharge and their control is therefore important for evaluation and improvement of such devices.
When capacitors or batteries are charged and then left on open circuit for
some time, a certain degree of self-discharge can set in, depending on the chem557
558
Chapter 18
>a:
w
z
CJ
: Driving "force"
: for self-discharge
istry and electrochemistry of the system, the purity of reagents and electrolyte,
and, significantly, the temperature. The rate of self-discharge, which usually diminishes with time, determines what is called the "shelf life" of batteries, which
is especially important in the case of primary cells, which are sold in the charged
condition and then stored as inventory in retail or wholesale outlets to the battery market. Some battery systems have excellent shelflife (e.g., Li-SOCI 2 primaries); others, such as NiOOH-Cd or NiOOH-Zn (which are usually
employed as secondary rechargeable systems) have a substantially poorer lifetime in the charged state because of the self-discharge processes that occur, especially at the nickel oxide cathode or zinc anode.
Similar self-discharge can arise in the case of electrochemical capacitors
in the charged state, but for somewhat different reasons, in detail, and with different mechanisms from those for the case of charged batteries. However, the
practical result is the same: after the capacitor stands for some period in the
charged condition, the charge may become lost and the interplate voltage diminished, causing the capacitor to approach a disfunctional condition, (e.g., where
"standby" voltage and power are required), necessitating use of a large capacitance for these purposes. Thus, the self-discharge behavior of electrochemical
capacitors is a matter of major practical significance in their operation and the
types of function they may be required to fulfill. However, to date, very little
work has been done to elucidate the mechanisms of self-discharge of capacitor
devices, although self-discharge rates are often measured and specified.
Charged electrochemical capacitors usually exhibit greater rates of selfdischarge than do most battery electrode or battery cell systems at a given temperature. This difference of stability of voltage is attributable to the fact that the
potentials of battery electrodes (in the charged state) are usually thermodynami-
559
560
Chapter 18
1. The first is if the capacitor (or battery cell) has been overcharged beyond the respective (plus or minus potential) decomposition limit of the electrolyte. Then the self-discharge is really the spontaneous decline of the
overcharge overpotential, which will not decay further than the respective reversible potential for the H2 or the O 2 evolution reactions in water if the latter
is the solvent. This spontaneous decline of the overvoltage, 11, which is due to
the overcharging current, occurs on account of continuing discharge across the
double layer until 11 ~ O. The mathematical representation of this situation and
its consequences is treated later in this chapter.
The leakage process corresponds to a Faradaic, potential-dependent charge
transfer reaction. Its rate therefore tends to progressively decrease as the selfdischarge continues-exponentially with the declining potential according to a
Tafel relation, as indicated below. In terms of equivalent circuit representation
of the behavior of the electrode interphase of the double layer, the charge leakage process corresponds to a potential-dependent Faradaic resistance, RF , operating in parallel across the double-layer capacitance, its value increasing with
declining potential (Chapter 3).
2. If the capacitor material, and/or its electrolyte contains impurities that
are oxidizable or reducible over the potential range corresponding to the potential difference attained at the capacitor on charge, then the capacitor becomes to
some extent nonpolarizable, again with Faradaic charge leakage currents that
are potential dependent. If these impurities are present in only small concentrations, then the self-discharge redox process(es) is (are) diffusion controlled. For
some impurities, their effect can be like that of a shuttle, with "Red" and "Ox"
species interchanging and diffusing in both directions between the positively
and negatively charged matrices of the electrochemical capacitor. Trace quantities of transition metal ions (e.g., Fe3+, Fe 2+) can commonly occur as impurities
in some carbons as well as in redox-active organics. Also, oxygen, which is usually adsorbed on carbon materials, can be reduced to H 20 2 or HOi, which can
then be either reduced or reoxidized, providing a shuttle.
3. In the case of double-layer capacitors, self-discharge can also occur by
internal redox reactions involving some of the functional groups, e.g., those of
quinonoid kinds, commonly residing on the edges of graphitic particles. The
types of groups present have been identified (as reported by Kinoshita2) and are
listed in Chapter 9. However, since such reactive groupings are sparsely distributed on a molar or per gram basis of the active carbon material, and some are
not redox active, their reaction on self-discharge could hardly consume more
than about 2 to 5% of full charge at, say, a 1.3-V potential difference between
the plates in an aqueous electrolyte.
4. In case (l), if H2 and O 2 are generated as a result of adventitious overcharge at the negatively and positively polarized half-cell electrodes, then situation (2) can be set up, with cross diffusion of H2 (to the positive electrode) and
561
O 2 (to the negative one). Depending on the electrocatalytic properties of the capacitor electrode material, H2 or O2 mayor may not be easily oxidized or reduced, respectively. At carbon, any evolved O 2 can be reduced to H 20 or
HOl: ion (or H 202> depending on pH), which can then behave either as an oxidizing or reducing agent in subsequent shuttle depolarization reactions, leading
to self-discharge, as noted in (2). This mechanism is probably an important one
since O2 from the air is commonly adsorbed on high-area carbons and thus introduced into the capacitor device unless the material is vacuum degassed when
the capacitor is being constructed.
5. If there are adventitious ohmic leakage pathways between the two electrodes of a charged capacitor cell (e.g., through fibrils passing through separator
films, or on account of shunt leaks around the seals of bipolar electrode plates
in capacitors based on bielectrode configurations) then self-discharge will take
place as though the capacitor were discharged slowly through a load resistor.
562
Chapter 18
recording instrument in such cases will distort the desired measurement. This is
a simple but important point.
Another simple and practically convenient procedure, as noted earlier, is
to apply a float current after charging that will maintain the full-charge potential
difference across the capacitor cell electrodes. This float current then just
matches the self-discharge rate at that potential. Similar measurements can be
made at lower potentials down the discharge curve. However, this procedure
gives no information about what is going on at one electrode or another. Table
18.1 contains a summary of the preferred methods for measuring self-discharge.
(18.1)
where i('l) is the density of current in A cm- 2 or A m-2 at an overcharge overvoltage of '1; io is the exchange current density (i.e., the current passing reversibly in both anodic and cathodic directions) at equal values per square
centimeter or square meter at the reversible potential [" =0] for which i('l) =0;
and a is the so-called "transfer coefficient" related to the effect of electrode potential on the activation process in the electrode reaction and its involvement in
the mechanism of the self-discharge process (see Chapter 3). a determines the
TABLE 18.1. Preferred Methods for Self-discharge Measurements on
Electrochemical Capacitors
1.
2.
3.
4.
5.
6.
563
slope of the electrode polarization (Tafel) relation between YJ and log i, and also
the slope of the self-discharge relation (see later discussion).
In the simplest case, the self-discharge process is a continuation of the
overcharge reactions (e.g., generation ofH 2 and O2 from the solution above its
decomposition voltage), which cause a discharge of the double-layer capacitance, Cd}, by the continuing passage of charge across the double layer until the
driving force (the overvoltage, YJ) declines to zero. This occurs as a result of the
self-discharge of the polarization across the double layer that is associated with
the collapse of the charge separation in this layer. If the rate of decline of YJ with
time is -dYJldt, it then follows that 1,3-5
(18.2)
-C dVldt = io exp(aVFIRT)
(18.3)
obtains where the right-hand term is the Tafel function (Eq. 18.1) for the potential dependence of the current passing. For integration, the above equation is rearranged to
io
C
(18.4)
-exp(-aVFIRndV = - . dt
cr]
RT
io
io [ t+-.
-exp(-aVFIRT)=-t+r=aF
C
C
~
(18.5)
564
Chapter 18
-aVFIRT=ln [ aFiO]
RTC +In (
crl
t+To
(18.6)
Hence,
(18.7)
Thus the potential V declines with log t (Fig. 18.2) plus a term involving the integration constant, r. It should be noted that the slope of this relation is the negative of the Tafel slope, RT/aF, of the initial polarizing process. The result in Eq.
(18.7) applies only if C is independent of V or 11 over the potential range of the
observed self-discharge.
If the Faradaic depolarization current is a parasitic impurity current rather
than an overcharge current, the same analysis still applies as long as the kinetics
are activation controlled in that process (i.e., a Tafel relation obtains), but are
different from those for the Faradaic charging current leading to overcharge
(cases 1 and 2).
The integration constant r can be evaluated from the initial condition at t
= 0, where V = Vi and i = ii, the initial polarization conditions. Then
r
RT
=-aF
exp(-aV;lRT)
(18.8)
(18.9)
o
FIGURE 18.2. Time dependence of potential on open-circuit self-discharge (Faradaic process involved). Vj =initial potential at t =0 (log t =-00). Linear and log plots in time, Eq. (18.7).
r=(RT)~
aF
565
(18.10)
Ij
(low fJ)
(18.11)
or
c =K exp(-VFIRT)
(high fJ)
(18.12)
c = K exp(VFIRT)
(18.13)
It should be noted here that the potential dependence of coverage, fJ, by an intermediate that gives rise to this potential dependence of the pseudocapacitance,
C (Chapter 10), is exactly the same effect that leads to a change of Tafel slope
for a desorption-controlled reaction step as increases from a small value to
near unity with increasing overpotential.
Introducing this V-dependent C in Eq. (18.3) instead of the constant C
gives
(18.14)
566
Chapter 18
where a may differ from the a of Case 1, depending on the mechanism of the
Faradaic process (Chapter 3). When the latter involves appreciable V-dependent
coverage by the intermediate, a = 1 + a for low 8 and a =a for 8 ~ 1 (see
Chapter 3).
Rearranging Eq. (18.14) to combine the exponential terms gives
io
-[exp(VF-aVF)IRT] dV= Kdt
(18.15)
io
-[exp(1 - a) VF IRT] . dV = K . dt
(18.16)
or
RT
io
io [ t+-.KT)
exp(I-a)VFIRT=Kt+T=K
(1- a)F
10
(18.17)
(a+ 1) VFIRT=-ln [
(a + 1)
RTK
(18.18)
so that
dVId In t = RTI( 1 - a) F
(18.19)
dVId In t as -2RTIF
(= - 11812.3 mV at T= 298 K)
and for large evalues (~ 1), -ve sign in Eq. (18.19), and a = 112, Eq. (18.19)
gives
dVld In t as -2 RTI3F
(= - 40/2.3 mV at T= 298 K)
567
In the general case, the capacitance, C, is usually some function of electrode potential. In the case of Hg, it is a quite complex variation, depending on
the type of electrolyte and solvent (Chapter 6), as shown by Grahame and others. Also, for various carbon materials, as illustrated in Chapter 9, the capacitances are quite dependent on potential. For such materials therefore the
capacitance has to be written as a 3-, or 4-term power series of electrode potential [e.g., C(V) = Co + aV + bV 2 + cV 3 + ... =fey)] for which the self-discharge
relation can be written formally in the usual way (cf. Eq. 18.3), with the above
C(V) function substituted in that equation, which can be integrated on the basis
of some empirically evaluated coefficients a, b, c, inj(V).
In the case of activation-controlled self-discharge, we have seen that Vet)
declines logarithmically with time, with a slope determined by the negative of
the Tafel slope factor RT/aF or the combined factor RT/( 1 - a)F. In the diffusion-controlled case (for a planar electrode; see Section 18.8), the decline of
V is linear in tll2, with a slope determined by the parameters of the diffusion
process, 2zFADll2n"2cofC (see Eq. 18.38), i.e., it is quite different from the result for the Tafel-determined self-discharge case. Note that the slope in this case
is proportional to the initial concentration, Co, of the depolarizing substance in
solution.
It is seen that for the case when the capacitance increases with declining
electrode potential, the rate of self-discharge is smaller with the log of time
(relative to the constant C case), while for the case where the capacitance declines with declining potential, the rate of self-discharge [in terms of the Vet) vs.
log time behavior] coincidentally works out to be the same as for the constant
C (Eq. 18.7) if a = a. In both cases, the decline of potential is logarithmic in
time (Fig. 18.2), originating from the exponential factors in potential.
The importance of the self-discharge arising from overcharge-type processes (referred to as side reactions in Ref. 6) in double-layer capacitors was recognized in a recent paper by Pillay and Newman.6 The reactions involved are
electrolytic evolution of H2 and 02, but with the possibility that such processes
take place below the normal decomposition limits of aqueous solution, i.e.,
within the otherwise expected ideal polarizability range (namely, double-layer
charging or discharging). This can occur within porous electrodes as a result of
the Hz or O2 partial pressures being below the standard (unit fugacities) values.
However, other impurity redox reactions, including shuttle processes involving
H 20 2 derived from adsorbed 02' should be recognized. These processes can go
on within the stability range of water. Related calculations have been provided
by Ksenzhek and Stender7 using the single-pore model of de Levie.
568
Chapter 18
4H+ + 4e + Oz
it is seen from the Nernst relations that when PH 2 and Po 2 are below their standard
values, the H2 potential will move to more positive values and the Oz potential
to less positive values, diminishing the difference in potential required for water
decomposition.
The buildup of Hz and O2 can then lead to self-discharge by the Faradaic
diffusion-controlled mechanism discussed earlier in this chapter, and the
achievable energy density is thereby diminished. A Ragone-type plot for a symmetric capacitor having 50-,um thick electrodes was constructed. 6
The extents of buildup of Hz and Oz were calculated for a duty period of
1000 cycles over a voltage range of 1 V, and the cumulative amounts were 0.218
and 0.008 mol m-z, respectively. These quantities translate into amounts that exceed the solubility of the two gases in the electrolyte in a 125-,um-thick capacitor device, so appreciable gas partial pressures can be generated.
18.7. COMPARISON WITH A REGULAR CAPACITOR DISCHARGING
THROUGH AN OHMIC LEAKAGE RESISTANCE
The equivalent circuit for a leaky regular capacitor is (there may also be an
equivalent series resistance, esr, Rs):
C
R
involving an ohmic equivalent load resistance, R; its self-discharge characteristics will be determined then only by the CR of the above combination. The
"self'-discharge behavior is then given by:
-C dV = i through R = ~
dt
R
(18.20)
569
dV dt
--v=RC
(18.22)
(18.23)
(18.24)
and then
In(V;lv)
=tlRC
(1S.25)
or
VIVi =exp(-tIRC)
(1S.26)
570
Chapter 18
Load current
diminishing
logV/V\
-ve
FIGURE 18.3. Time dependence of potential on discharge of a capacitance into a load resistance.
Vi =initial potential at t =O. Log [potential) is linear in t, Eq. (18.25).
that for solvent stability. This situation corresponds to cases (2) and (4) identified earlier.
In this case the leakage current is not io exp [a VFIRT], but a diffusion-limited current
(18.27)
where CR is the concentration of the diffusible redox species, R, and dCRldx is
the gradient of concentration of R along the direction x in the diffusion boundary
layer near the electrode surface and normal to it; D is the diffusion coefficient
of R in solution. Then
dV
-Cdt
=zFD (dCRldx)
(18.28)
The solution of this type of equation is more difficult than for Case I (Eq.
18.3) since CR and dCRldx are time dependent so that
dCR) . dt
-G- dV=zFD [d;
(18.29)
CR and dcRldt can be obtained by solution of Fick's equations, say, for semi-infinite diffusion to a plane capacitor electrode. Solutions for porous bodies or
within a cylindrical pore are more complicated, but the diffusion path length
will generally be small, on the order of microns or less in such electrodes.
The time dependence of CR and dCRldx at a particular distance x from a
plane electrode is generally an inverse square-root in t, so that integration ofEq.
(18.29) would lead to a self-discharge relation for the t dependence (decline) of
571
(18.30)
Most problems involve solutions to Fick's second law for particular boundary
conditions. Fick's second law, as a partial differential equation, is
(18.31)
i.e., the time dependence of concentration at a particular laminar at x, normal to
the concentration gradient, is proportional to the first derivative of the concentration gradient. Solutions can be obtained by numerical methods and are available in the literature, e.g., in the monographs by CrankS or Carslaw and Jaege~
on heat transfer and diffusion.
At a smooth planar electrode, the concentration gradient can be recorded
directly by means of the Schlieren optical method as used in electrophoresis and
ultracentrifugation instruments, or by means of interferometry. Unfortunately,
in porous electrodes of practical significance in capacitor technology, only indirect electrochemical or analytical procedures are available.
For the case of so-called semi-infinite diffusion, the concentration, c, of a
reacting diffusing substance at the electrode surface is given as a function of distance, x, and time, t, by
c(X,t)
(18.32)
where Co = c(x,t) = 0 and erf represents the error function integral (also arising
in statistics) defined by
A
erflAI
=_2_
f exp(-Z2) . dz
;rr112
(18.33)
572
Chapter 18
1= iA
(18.34)
(18.35)
dt
or
-C dV =zFA DII2 Co
]tll2 t -112 .
dt
(18.36)
(contrast Eq. 18.29) where Co is now the initial concentration in mol cm-3 of the
depolarizing impurity.
Integration of the above equation leads to
(18.37)
when t
is the potential decay relation for self-discharge under linear (semi-infinite) diffusion-controlled conditions. Thus the potential decline Vi - V (at time t) increases with t l12 (Fig. 18.4).
The mathematics for the reaction at a cylindrical electrode have also been
worked out and involve Bessel functions. This case is relevant to a porous electrode made, e.g., of carbon fibers. The results are
(18.39)
for small values of t/J (see later discussion), or
1=
[1
zFADco
Y
r
In(4t/J) - 2y - [In(4t/J) - 2y]2 ...
(18.40)
for large t/J values, where t/J is the dimensionless factor Dt!? and y = 0.57722; r
is the radius of the cylinder. The time variable for this case is in the parameter
t/J.
573
'-r-
Self-discharge
current
diminishing
~
o
FIGURE 18.4. Decline of potential in time, as (112, for a diffusion-controlled self-discharge process (semi-infinite diffusion model; Eq. 18.38).
Again, the current I can be combined with -CdVldt and a self-discharge relation for V(t) can be derived by integration. This is easy for the case of small
values of (short times), but is more complex for large s since t is then an implicit function in the inverse log term. Numerical evaluations are then easier.
The above functions are dealt with in Delahay's well-known monograph.lO
In the case of impurity-controlled self-discharge, it should be noted that the
reversible potentials for the respective impurity redox processes will be well
displaced from their standard (EO) values, probably up to about 360 mV (e.g., 6
magnitudes multiplied by 2.3 RTIF for a Ie redox reaction), owing to the dilution of the species. For a reduction reaction (e.g., from to Fe3+ to Fe2+), this displacement would be in the positive direction in the initial virtual absence of
Fe 2+. However, the redox potential will actually depend on the ratio of activities
of the oxidized and reduced forms of the redox couple species.
The component iF increases as the electrode potential increases, usually according to a Tafel relation (Chapter 3), and at sufficiently high overpotentials, where
decomposition of the solution takes place, iF ie and i ~ iF' This situation is
574
Chapter 18
the opposite of self-discharge, while for the ideally polarizable electrode or capacitor, iF = 0 and i == i e
For optimum behavior, iF should be a minimum over the whole range of
potentials that correspond to the charging process over the operating potential
range of the capacitor: practically, this is about 1.3 V for aqueous electrolytes
and 3.0-3.5 V for nonaqueous, aprotic solvent solutions. If iF is appreciable toward the end of the charging half-cycle of a capacitor, there will be a corresponding appreciable self-discharge leakage current (at least initially) following
termination of charging when the overall i is cut to zero.
18.10. SELF-DISCHARGE OF DOUBLE-LAYER-TYPE
SUPERCAPACITOR DEVICES
Self-discharge of a double-Iayer-type electrochemical capacitor device differs in a significant way from that which may occur in a leaky regular capacitor
or at a single electrode. The essential difference is that a double-Iayer-type capacitor consists of two interfacial double-layer capacitances separated by an
electrolytic resistance, R e, plus (usually) a resistance, (Rs' in series) due to the
separator. The equivalent circuit is then
Self-discharge processes can go on at either or both of the double-layer interfaces, depending on overcharge conditions or the presence of redox impurities. However, it is evidently possible, in principle, for asymmetric
self-discharge to occur (e.g., across one double layer or the other), depending
on the type and redox potential of the diffusible impurity.
Self-discharge would then be represented by the equivalent circuit:
Cd! (1)
-r1~
Cd! (2)
RF(1)
h-
RF(2)
where RF.1 may not be the same as RF.2, not to mention that one side of the capacitor will be a positively charged and polarized electrode while the other will
be complementarily negati ve, so that different types of Faradaic leakage current
575
processes would tend to be operative-one cathodic, the other anodic-probably having different kinetic or diffusional characteristics.
This is an interesting but little understood situation since to date very little
work, except that currently under way in the author's laboratory, has been done
on self-discharge of half-cell electrochemical capacitor electrodes using a separate counter electrode and a reference electrode against which to scale the declining potential of a single electrode. Table 18.2 contains a summary of
self-discharge effects originating from various mechanisms that provide diagnostic criteria for identifying these effects.
576
Chapter 18
interfacial regions in the absence of current flow. This will also be accompanied
by diffusive ionic transport processes as the distributed interfacial charge densities relax to a common value within the matrix. The principle involved in this
situation can be represented in an elementary way by a simple equivalent circuit
with three or four capacitative and resistive elements, as shown below.
+
WE
Cl
CE,(a)RE
I
I
C2
{3
I
C4
I
I
577
11
+
10
~ RI
.!!!
"0
>
I.
~~):
~ H.u
6
5
.. ..
('J,:.::::>..
R3: 99k
ClC2 o C3-C.135
/(d)
IlA
charge curren! 50
Al -! Oak
R2 a 99k
20
40
60
eo
100
120
140
160
IJF
180
200
lime I seconds
FIGURE 18.5. Time variation of potentials on open circuit across four capacitors in a transmission-line type of network following charging at 50 /lA. Experimental responses on a hardware R-C
network. (From Pell and Conway, in press, 1. Power Sources, 1999).
20
-<
:1.
.:;
!:!
'u
i:
10
'"0
0.
c:
0
:r~.,..:
'10
:;
u
20
0.
30
2
-u
<II
<II
40
charge curranlSO
t-t.
t---5:~
10
20
30
40
50
60
IlA
RI'OOk
R2-S9k
R3-99k
CI "C2-C3C.~1351JF
70
80
90
100
time I seconds
FIGURE 18.6. Time variation of charging currents on open circuit after charging at 50 /lA for
70 s; calculated from C. dVldt from the responses in Fig. 18.5, of the capacitors at (a), (b), (c), and
(d). (From Pell and Conway, in press, 1. Power Sources, 1999).
578
Chapter 18
1(T) = A exp[-WIRT]
(18.42)
(18.43)
so that log leT) is a linear relation in liT, the slope of which gives -MtlR.
For Faradaically controlled self-discharge processes, M+ is on the order
of 40 to 80 kJ mol-I, while if they are diffusion controlled, Mt is around 16 to
20 kJ mol-I, corresponding to the activated motion of the depolarizing impurity
or redox shuttle species in the bulk of the solution toward the electrode surface.
Covalent Associates, Inc., for example, report an E+ for self-discharge of 28 kJ
mol- 1 for one of their capacitors. This is one of the few data available. The general and important practical consequence of the temperature effect is that selfdischarge of charged capacitors or batteries is a much more serious problem at
60-70C than it is at 0-25C.
With an activation energy of, e.g., 40 kJ mol- 1 (a typical value), the ratio
of self-discharge currents at 70C compared with OC is easily calculated from
Eq. (18.42), i.e.,
i(70C) i(343 K)
i(OC) == i(273 K)
[-40,000)( 1
1)
=exp 8.31 343 - 273
(18.44)
= 37 or 101.56,
(18.45)
579
(18.46)
(18.47)
with
2MH~2M+Hz
(18.48)
and do not need any electron transfer to or from the external circuit, as is required for a self-discharge process. Simultaneously, the charge on the double
layer becomes reduced in relation to the declining potential of the electrode on
open circuit. The same applies to all battery-type self-discharge processes; they
must be able to be written equivalently as chemical reactions not involving any
net flow of electrons. At a double-layer capacitor electrode, however, it is sufficient that a direct single electrode process occur leading to self-discharge, e.g.,
S[z-] + Fe3+ ~ S[(z-)-l] + Fez+ where S is, say, a negatively charged surface
580
Chapter 18
ofthe electrode (its double layer) having its charge density, ze, diminished here
by Ie owing to reaction with a redox species such as an Fe3+ ion or more commonly 02 and H 20 2.
This situation was analyzed in detail in papers by Tilak and Conway4 and
by Harrington and Conway. 11 The relation between the decline of potential with
time and that of coverage by the electro active species is illustrated in Fig. 18.7.
First the potential declines with discharge of the double-layer capacitance until
a potential is reached where the buffering effect of the mixed process causes an
arrest of potential for a period during which the two coupled processes operate
to discharge the pseudocapacitance.
In the case of pseudocapacitance associated with the potential dependence
of the oxidation states of an oxide 1 such as NiOOHJNi0 2 or PtOlPt0 2, a similar
mixed potential situation can arise, leading to 02 evolution on open circuit, if
the oxidation state upon completion of the charge corresponds to attainment of
an anode potential above the reversible potential for 02 evolution. Conway and
Bourgault! analyzed this situation in terms of coupled electron transfer between
Ni4+ and Ni 3+ states in the surface region of thin oxide film on Ni, and electron
transfers in the steps of electrochemical 02 evolution. The reaction scheme also
requires coupled proton transfer in the hydrous oxide lattice between OH- and
02- ions. It is illustrated in Fig. 18.8.
0.5
0.4
1.0
0.8
0.6
0.3
>
CD
<1l
0'
0
0.4
0.2
...
Q)
>
0
0.1
"'lp-----~
I
I
0.2
0.0
-8
log (Tp)
-2
2.0
FIGURE IS.7. Relation between decline of potential (curve BCD) and change of coverage (curve
A) by an electroactive species, giving rise to a pseudocapacitance, Cp, when the latter is self-discharged. 11 p is the plateau potential associated with discharge of Cpo B is the Cdl discharge region.
Ip is the time to onset of Cp discharge. Reprinted from D. A. Harrington and B. E. Conway, 1. Electroanal. Chern., 221, 1 (1987), with permission from Elsevier Science.
581
Hp .. Ni 3+(surface )
OH '(surface )
H+
OR
Ni 4+(surjace )
H+
OH- .. Ni 3+(surjace )
OH '(surface )
OH-(lattice surjace)
0 l-(bulk)
!F 0 2-(lattice
surface)
OH-(buik)
COUPLED WITH:
Ni 3+(Jattice)
Ni 4+(lattice surface)
Ni 4+(/attice)
Ni 3+(lattice surface)
OVERALL:
Ni(IIl)'Ni(IV)(0)3'OH
Hp .. 2Ni(1ll)0OH
40H' .. 2H10
OH'
01
FIGURE 18.8. Schematic diagram of the processes of coupled electron and proton transfer in the
self-discharge of an Ni(III) or Ni(IV) oxide electrode. (Based on Conway and Bourgault. I)
When reducing impurities are present (e.g., molecular H2 or low oxidationstate transition metal ions such as Fe2+), oxide films in high oxidation states can
become reduced through a self-discharge reduction reaction. An example of this
was studied by Shibata 12,13 for the case of electrocatalytic reduction of Pt oxide
films by H 2 Such processes have the form
(18.49)
(18.50)
the potential being determined by the oxidation and reduction potential profile
as a function of reduction charge, e.g., as known from cyclic voltammetry experiments. 14
Processes such as the above two can also be represented as mixed, coupled
electrochemical reduction and oxidation reactions; thus:
(18.51)
(18.52)
or correspondingly
582
Chapter 18
PtO + 2H+ + 2e
-t
Pt + H 20
(18.53)
(18.54)
each of which pair adds up to the chemical steps shown in Eqs. (18.49) and
(18.50). Note again that no transfer of electrons need take place to or from the
external circuit.
In the case of RuOz films, and probably other transition metal oxides, it is
observed that some self-discharge also takes place. In the case of overcharge to
above the O 2 evolution potential, the self-discharge arises for well-understood
reasons, like that with nickel oxide. 1However, below the O2 reversible potential
and in the absence of molecular H2 (cf. Shibata13), self-discharge still seems to
arise, a phenomenon that is not so easy to explain except in terms of reactions
with some other reducing impurity reagents, adventitiously present, or inhomogeneity in the initial state of oxidation of RU02 after charging; see later discussion.
With oxide films that are generated over several oxidation states, as with
nickel oxide (Niz+, Ni 3+, and probably Ni4+), a gradient of oxidation state of the
oxide film can be developed from the outer surface into the bulk. On termination
of charging, some relaxation of this gradient by a diffusion-controlled mechanism can occur, leading to an apparent decline in the oxide film potential (at the
outside of the film) with time on open circuit, following termination of charging. Such a mechanism could also occur with Ru02 where Ru z+, Ru 3+, and Ru 4+
(or even Ru 6+) states are involved in the anodic development of RuOz films (see
Section 18.15).
583
parative experiments were also made on a packaged "Gold Cap" Panasonic unit
of 3.5 F. The results are outlined and discussed in the following paragraphs; in
particular, the diagnostic tests derived earlier (Table 18.2) were applied to the
results.
The carbon electrodes studied were prepared from high-area activated
carbon fiber fabric (Spectracorp Spectracarbon 2225, surface 2500 m2/g). The
apparent surface area of the electrode used in this work was 0.15 cm2. The electrochemical behavior of the films was characterized by evaluation of the anodic
and cathodic reduction charge in voltammograms or by dc galvanostatic charge,
discharge, and recharge cycles at given currents.
Self-discharge and potential recovery measurements were conducted in a
three-compartment electrochemical cell containing 0.5 M H2S04 prepared from
high-purity Millipore water. The solutions were bubbled with Orfree Nz
throughout the experiments, some of which were continued for a number of
days. Galvanostatic charging of carbon electrodes was conducted up to various
potentials in the range 0.5 to 1.25 V and then the open-circuit decline of voltage
was recorded through a high input impedance electronic instrumentation system.
Cathodic discharging was also conducted at controlled currents with potential being taken down to limits of 0.30 or -0.10 (RHE). After completion of
such discharge curves, the electrodes were held on open circuit and recovery of
potential was observed.
Self-discharge data were recorded over three decades of time (in seconds;
Fig. 18.9) from several initial polarizing (charging) potentials, Vi. In this case
the VI data plotted out as excellent linear relations to log(t + r) (Fig. 18.9) (cf.
Eq. 18.7), but from various VI values separate, nonconvergent, log time relations
were observed having slopes dependent on Vi, unlike the results at nickel oxide. l
This was probably due to the potential dependence of the double-layer capacitance of this electrode and is currently the subject of further study. The direct
relationship of VI to the log of time clearly indicates a Faradaic process as the
mechanism of self-discharge.
After a cathodic current is imposed, the potential recovers, also logarithmically in time (Fig. 18.9, scale on rhs). Similar to self-discharge, from various
Vi values, separate, nonconvergent, log-time relationships were observed. The
slopes of these relationships increased with increasing cathodic Vi' for reasons
not yet understood.
At the same carbon fiber electrode, charged negatively, and with the solution saturated with 02, the self-discharge was in t 1l2 , clearly indicating diffusion
584
Chapter 18
Vln" 1.24 V
1.2
~
(5
0.7
1 .1
0.6
>
1.06 V
1.0
0.9
~
O.B
0.5
0.4
0.90 V
-33 mV
0.3
43 mV
0.7
'0
>
......... -~~.;-
0.1 0 _________
_ ...... -...
0.2
.. --" ............
0.6
10 2
0.1
4 5
10 3
4 5
10 4
4 5
time I seconds
FIGURE 18.9. Potential decay and recovery behavior at a charged and discharged carbon felt single electrode plotted logarithmically in time. (From Pell, Liu, and Conway. 16)
585
(3
>
1.2
(3
4.5
1.0 .::
4.0
0.8
CI)
3.5
Q)
(/)
3.0
2.5
2.0
0.6
0.4
0.2
OJ
cU
(3
>
>..
iii
>
o
u
L......J.._~-'---~--'-~--'-~--L-~-'500.0
10
20
30
40
time / hours
.-..
CI)
C)
0.75
0.70
(3
0.65
...
0.60
>
CI)
C)
cU
.r:
0.55
:0
0.50
(/)
r-r~-.------r~--,--.......~--,--.,.........,.~-,-.,....,--r"1
(b 1.2
Q)
..
0.0
...
o
>
>..
CI)
>
o
u
CI)
-0.6 ..=.
OJ
.2
0.45
0.40
0.35
OJ
0.6
L.....LO~--'-~J........--'-~--'-~...........--'-~-'-~L...........J...-'
1 0 15
20
25
30
35
40
-1.2
45
time / hours
FIGURE 18.10. Diagnostic test plots for evaluation of mechanism of self-discharge observed at
a Panasonic 3.5-F Gold Cap capacitor. Potential recovery relationships after forced discharge are
also shown. (From Pel!. Liu, and Conway. 16)
586
Chapter 18
en
5.5
>
5.0
.~
Q)
CD
!!
4.5
4.0
CD
3.5
Q)
.~
--
3.0
'C
(j)
en
1.2
'0
1.0 >
-Q)
'0
>
~
.r:.
u
(c) 1.4
2.5
2.0
0.8
0.6
0.4
0.2
100
200
300
400
Cl
!!
'0
>
>Q;
>
0
U
0.0
'0
>
Q)
Cl
!!
'0
5.5
(d) 1.4
5.0
1.2
4.5
1.0
O.B
Cl
3.5
0.6
--.
3.0
0.4
2.5
0.2
Q)
~
.r:.
u
.~
'C
(j)
en
en
>
Q)
4.0
>
...
'0
2.~ 00
101
102
103
104
105
Cl
,!g
'0
>
>Q;
>
0
0.0
time / seconds
FIGURE 18.10. (continued)
18.15.1. Background
587
Chapter 18
588
1.5
(a)
:c
a:
'"
>
>
1.2
0.9
c:
---
Q)
'0
c=>
~
- -
0.6
.~
c=
Q)
c-
0.3
0.0
6
8
Time I hours
1.5
'"
>
>
14
1.2
0.9
0.6
"E
Q)
12
(b)
:c
a:
10
--
-- -
(5
c-
'5
~
'u
C:
(l)
c-
0.3
0.0
8 10 12 14 16 18 20 22 24
Time f hours
FIGURE IS.11. Open-circuit potential decay following anodic charging to 1.40 V (RHE) and
open-circuit potential recovery following cathodic discharge to 0.1 V (RHE) for (a) thermally prepared RU02 and (b) electrochemically prepared RU02 electrodes in 0.5 mol dm- 3 H2S04 at 298 K.
589
as that attained in potential decay from the charged state. Similar recovery behavior was observed at the electrolytically formed films [Fig. 18.11(b)].
Plots of decaying or recovering potential (even after inclusion of the integration constant time r; cf. Ref. 1 and Eq. IS.6) vs. log [time] are not linear, indicating that a Faradaic process of self-discharge, of the kind treated in Refs. 1
and 4 and examined earlier here, is not applicable.
3. The most remarkable observation is that following the recovery of the
potential to ca. 0.6 V, a further cathodic discharge down to 0.10 V, involving
passage of an appreciable further cathodic charge, can be conducted; then, following that further discharge, the potential again recovers to near the 0.6 V
value [Fig. IS.12(a)]. This discharge-potential-recovery cycle can be repeated
many times and the successive recovery potentials remain almost the same. A
"perpetual" source of charge is, of course, impossible, so presumably there is a
sufficient bulk reservoir of oxidized Ru to maintain a surface region, which is
evidently potential determining and sufficiently oxidized that (through diffu-
0.8
(a)
0.6
w
:::c
a:
en
>
>
Cii
0.4
'E
Q)
'0
c...
0.2
4
Time I hours
FIGURE 18.12. Multiple potential recovery transients at RU02 in aq. H2S04 at 298 K following
successive galvanostatic discharges, without interim recharging: examples of successive recoveries
of potential at (a) a thick, thermally formed RU02 film;
590
Chapter 18
+e
Ru IIl
+e
Ru ll
0.8
(b)
0.6
w
:x:
II:
en
>
>
cti
0.4
~
Q)
"0
a..
0.2
100
200
300
Time I hours
FIGURE 18.12.
400
591
at the electrode surface on discharge, coupled with a redistribution of the oxidation state of Ru in the oxide by electron hopping following such a discharge,
according, e.g., to:
Ru III bulk + RuII surface ~ Ru II bulk + RuIII surface
probably involving electron and proton exchange (cf. Ref. 1); and also
RuIV bulk + RuIII surface ~ RuIII bulk + RuIV surface
Coupled with such redox processes must be adjustment of the local distribution
of 0 2- and OH- ions in the lattice by proton transfer, namely
OH- (surface) + 02-(bulk)
Proton transfer equilibria with H30+ and H 20 in the surface region eventually will also be involved
0 2- (surface) + H30+ (solution) ~ OH- (surface) + H 20 (solution)
Evidently, either from self-discharge or potential recovery, the surface region
eventually provides a fairly stable thermodynamic potential that is dependent on
the previous cycle history of the oxide film. Similar time-dependent adjustments of potential took place at electrolytically formed Ru02 films, as was
shown in Fig. 18.11 (b), but over rather longer time scales; the recovery phenomenon was also similar to that observed at the thermally formed oxide film
[Fig. 18.11(a)].
18.15.3. Model for Potential Recovery
592
Chapter 18
593
established. It is about 0.7 V, or 0.9 V, RHE, as seen in Fig. 18.11, differing between the thermally and electrolytically formed RU02 films.
In order to elucidate what kinds of equilibrium redox processes might be
involved in establishing these potentials, the pH dependence of the steady potentials was determinedY The results are shown in Fig. 18.13 from which
dE/dpH = -30 m V was reproducibly found for the thermally formed film compared with -60 m V for the electrochemically formed one. These data provide a
pH/potential kind of Pourbaix relation for the potential-determining processes
at the two oxide materials following self-discharge and potential recovery. Evidently the oxide in its supposedly RU02 state at 1.4 V (or at 1.3, 1.2, 1.1 V) is
not stable in aqueous solution at any pH between 0 and ca. 10 (Fig. 18.13), and
prefers to establish a substantially less positive steady potential. While a -30m V dependence on pH has not previously been reported, Siviglia, Daghetti, and
Trasatti 23 found that the open-circuit potential of a thermal RU02 electrode exhibited a -60-mV dependence on pH.
In Fig. 18.13 the potentials asf(pH) have been corrected for the 60-mV pH
dependence of the hydrogen reference electrode used, so the pH dependence of
the potentials plotted can be interpreted in terms of half-cell processes involving
two oxidation states of Ru in the potential-determining process, together with a
stoichiometric number of involved protons and a corresponding number of electrons.
0.9
w
::r:
:z
en
>
0.8
>
0.7
:;:;
c
'"
0.6
"0
0.
0.5
Q)
::>
.~
0.4
'c..=
0.3
Q)
0.2
10
12
pH
FIGURE 18.13. pH dependence of open-circuit decay and recovery potentials after approach to
steady values for _ thermally and. electrochemically prepared thick film of RuOz on electrodes.
Reprinted from Pell et aI., Electrochim. Acta, 42,3541 (1997), with permission from Elsevier Science.
594
Chapter 18
595
The unambiguous difference between the Nernstian slope for the thermally
prepared compared with the electrolytically generated Ru02 film is of considerable interest in relation to the chemical constitution of the surfaces (or bulk as
well, in the case of the electrolytic film) of the two materials despite the qualitative similarity between their respective potential decay and recovery behavior
(Figs. 18.11 and 18.12).
The above thermodynamic and potential decay information indicates that
the self-discharge process does not originate (as it does at Ni0OH-Ni02 electrode surfaces I) from continuing O 2 evolution since (1) the potential decay process continues down to ca. 0.75 V rather than 1.23 V (the OrH20 reversible
potential) and (2) it is also observed from several initial potentials below that
reversible potential. Hence the potential decay process must involve readjustment of Ru redox couples in the surface region of the oxide. The same argument
also applies to the process involved in potential recovery following cathodic
discharge down to 0.1 V, or other potentials less than 0.75 V.
It is well known that in a series stack an unmatched faulty component capacitor may undergo overcharge or overdischarge if it is adventitiously of lower
C than that of the other capacitor units with which it is unmatched. Related problems can arise if one or more components in a stack suffers self-discharge at
rates greater than others, so uniformity of self-discharge behavior is desirable,
if not essential. High-level manufacturing control conditions are therefore desirable.
REFERENCES
1. B. E. Conway and P. L. Bourgault, Can. J. Chem., 37, 292 (1959); Trans. Faraday Soc., 58,
593 (1962).
2. K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, Wiley, New York
(1988).
3. 1. A. V. Butler and M. Armstrong, Trans. Faraday Soc., 29,1261 (1933).
4. B. V. Tilak and B. E. Conway, Electrochim. Acta, 21, 745 (1976); 22,1167 (1977).
5. B. E. Conway, in Power Sources, vol. 16 (Brighton Power Sources Conference Proceedings)
(1997), Inti. Power Sources Symposium Ltd.
6. B. Pillay and 1. Newman, J. Electrochem. Soc., 143, 1806 (1996).
596
Chapter 18
7. O. S. Ksenzhekand V. V. Stender, Dokl. Akad. Nauk. SSSR, 106, 487 (1956); see also Russian
1. Phys. Chem., 27, 1089 (1963).
8. 1. Crank, The Mathematics of Diffusion, The Clarendon Press, Oxford (1956).
9. H. S. Carlslaw and J. C. Jaeger, Conduction of Heat in Solids, Oxford Univ. Press, Oxford,
(1947).
10. P. Delahay, New Instrumental Methods in Electrochemistry, Chapter 3, Interscience, New
York (1954).
11. D. A. Harrington and B. E. Conway, 1. Electroanal. Chem., 221, I (1987).
12. S. Shibata and M. P. Sumino, Electrochim. Acta, 20, 739 (1975).
13. S. Shibata, 1. Electroanal. Chem., 89, 37 (1978).
14. F. G. Will and C. A. Knorr, Zeit. f Elektrochem., 64, 258 (1960).
15. W. G. Pell, T.-C. Liu, and B. E. Conway, Electrochim. Acta, 42, 3541 (1997).
16. W. G. Pell, T.-C. Liu, and B. E. Conway, in Proc. Sixth IntI. Symposium on Electrochemical
Capacitors and Similar Energy Storage Devices, S. Wolsky and N. Marincic, eds., Florida
Educational Seminars, Boca Raton, Fla. (1996).
17. S. Trasatti and G. Buzzanca, 1. Electroanal. Chem., 29, App. 1 (1971).
18. D. Galizzioli, F. Tantardini, and S. Trasatti, 1. Appl. Electrochem., 4, 57 (1974).
19. S. Hadzi-Jordanov, H. A. Kozlowska, M. Vukovic, and B. E. Conway, 1. Electrochem. Soc.,
125, 1471 (1978).
20. S. Hadzi-Jordanov, H. A. Kozlowska, and B. E. Conway, 1. Electroanal. Chem., 60, 359
(1975).
21. J. P. Zheng and T. R. Jow, 1. Electrochem. Soc., 142, L6 (1995); 1. P. Zheng, P. J. Cygan, and
T. R. Jow, 1. Electrochem. Soc., 142,2699 (1995).
22. T. Arikado, C. Iwakura, and H. Tamura, Electrochim. Acta, 22, 513 (1977).
23. P. Siviglia, A. Daghetti, and S. Trasatti, Coli. Surfaces, 7, 15 (1983); see also 1. Electroanal.
Chem., 122,395 (1981).
24. M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, Nat!. Assoc.
Corrosion Engineers, Houston, Tex. (1974 reprint).
25. L. D. Burke and 1. F. Healy, 1. Electroanal. Chem., 124, 327 (1981).
26. L. D. Burke, M. E. Lyons, E. J. M. O'Sullivan, and D. P. Whelan, 1. Electroanal. Chem., 122,
403 (1981).
Chapter 19
19.1. INTRODUCTION'
The key components in an electrochemical capacitor are the electrode, electrolyte, and the separator materials. Most electrochemical capacitors, except the
hybrid type, l are made with identical electrodes, i.e., preparation of only one
type of electrode is necessary, unlike in batteries. However, as the specific capacitance (e.g., of a carbon electrode) when positively polarized is usually different from that of the same material when it is negatively polarized, some
allowance for this can be made by using a different quantity (weight or volume)
of the active material in the two electrodes. Then the resulting device should be
polarity labelled.
The packaging for capacitors depends on their type, size, and ultimate use.
Coin-type cans and electrolytic capacitor-type housings are frequently used for
PC-board applications. Larger cells are of prismatic construction, with heavywalled containers and robust terminal connections. The end plates must provide
good electrical contact in order to minimize esr, and some compression may be
advantageous in construction, but not at the expense of squeezing out significant
quantities of electrolyte. The focus of this chapter is on laboratory-scale construction and testing of capacitor cells. An example follows.
*This section is based on material contributed by Dr. B. V. Tilak (Occidental Chemical Corp.),
Grand Island, N. Y. and Dr. S. Sarangapani, ICET Inc., Norwood, Mass., and edited for this volume
by the author.
597
598
Chapter 19
599
placed over a Toray H060 (Toray Marketing, New York) carbon paper (or the
Spectracorp equivalent) and gently rolled with a rolling pin between several layers of filter paper to remove excess water. The electrode is then sandwiched between several layers of dry filter paper and pressed in a hydraulic press at
100-400 psi for 1-2 min. At this stage, the microporous filter membrane can
be removed by gently peeling off the layer from the electrode. The electrode is
then placed between PEP Teflon films and pressed at 100C at 100-400 psi for
1-2 min. The electrode is then dried at 100C for 1 h, baked at 270C, and sintered at 340C for 20 min.
The hydrophobic nature of such electrodes prevents the full realization of
the maximum available capacitance during the initial testing period. However,
these electrodes undergo a "break-in" phenomenon, during which period the
electrolyte seeps into the electrode and wets the internal finer pores. Vacuum
filling of the electrolyte is desirable to speed up the break-in process, and/or a
trace of surfactant can be added to the electrolyte.
A titanium foil is immersed in hot oxalic acid for 2-3 min; it is then ultrasonically cleaned in distilled water. A solution of the following composition is
prepared (South African Patent, 680,834):
n-butyl alcohol:
concentrated aqueous HCI:
RuClf
6.2 cc
0.4 cc
1g
This solution is either sprayed or brushed onto the clean titanium foil. The
foil is air dried and the process of brushing and drying is repeated until a sufficient loading of the ruthenium is obtained, typically 12 times. Once a sufficient
thickness of the coating is built up, the foil is heat treated in air at a temperature
of 350-500C for about 5 min. Care should be exercised to not build up too
thick a layer, since thick layers tend to spall on heat treatment. Other transition
metal oxide electrodes (e.g., Ir02, C030 4 ) can also be prepared by this procedure
that is based on the recipe for preparation of DSA RU02-Ti anodes for the
chloralkali industry (Beer patents).
For comparative examination of electrochemically formed RU02 films, Ru
metal (bulk or electrodeposited on C or Au) is subjected to repetitive anodic and
cathodic cycling under cyclic voltarnmetry conditions over a potential range of 0.05
to 1.40 V (RHE) in aqueous H2S04 for a number of hours. This leads to the formation of quite thick films of hydrous ruthenium oxide (up to microns), which are easily visible under an SEM or optical microscope (see Chapter 11). The hydrous
oxide exhibits appreciable redox pseudocapacitance, while the thermally
formed DSA type has a larger component of double-layer capacitance.
600
Chapter 19
A paint of ruthenium oxide or ruthenium on a carbon catalyst powder (Aesar) in a solution of 1100 EW Nafion ionomer (Solution Technology, Mendenhall, Pennsylvania or Aldrich) is prepared by suspending and dispersing the
catalyst particles in the Nafion solution. An amount of 25-30 wt. % N afion (dry
weight basis) may be adequate to prepare a satisfactory paint. An oxidized niobium foil is coated with this paint and dried in an oven to remove the alcohol
solvents. A Nation 117 membrane, in the acid form and well equilibrated with
water, is placed in between two such coated niobium foils and hot pressed at 110C
for 1-2 min. The membrane electrode assembly is then removed from the press,
soaked in distilled water, and subjected to charge and discharge cycling.
. - - - - - - - Electrode 1
. - - - - - Separator
Electrode 2
Heater
Pad
__ .1 ~
~ '/4.1r-~1/4'
1/2"
1/2'
FIGURE 19.1. Sectional diagram of laboratory-scale electrochemical cell for testing capacitor
electrodes at various temperatures.
601
glass fiber paper (Whatman) soaked with 6 M sulfuric acid electrolyte. Suitable
end plates made of resin-filled graphite (Carbone of America, JP1345 or equivalent) can be used as current collectors. A simple framework that consists of two
stainless-steel end plates, with pedestals machined into them, serves as the cell
housing, as shown in Fig. 19.2. This arrangement is useful for laboratory screening of potential capacitor materials in successive series of tests. The cell assembly is placed between the two end plates, which are clamped together with bolts
and wingnuts. Care should be taken to apply insulating sleeves on the bolts together with nylon washers to prevent shorting of the two electrodes. Sufficient
pressure can be applied to the cell by varying the thickness of the gaskets in the
cell andlor tightening the nuts. This design also permits cells with nonaqueous
electrolyte (assembled inside a glove box) to be tested under ambient conditions. A spring-loaded cell construction suitable for the present application has
been described by Kronenberg, Stein, and Codd. 2 For the assembly and testing
of solid polymer electrolyte capacitors, the cell assembly shown in Fig. 19.2 can
be used (Sarangapani et a1. 3.4).
Gold Plated
Tab
Cell Assembly
Teflon Gasket
Insulated
Bolt
FIGURE 19.2. Cell assembly based on arrangement of electrodes illustrated in Fig. 19.1, but for
a solid polymer electrode.
602
Chapter 19
Cyclic voltammetry (Chapters 3 and 10) is useful as a quick screening procedure to identify potential capacitor materials. The experimental procedure involves potential cycling between two voltage values preselected for a given
electrolyte. The experimental details of conducting a cyclic voltammetric experiment can be found in textbooks, e.g., those by Gileadi 7 or Conway.s A simple electrochemical cell and electrode holder are shown in Figs. 19.3 and 19.4,
respectively. Note the inclusion of a reference electrode (Fig. 19.3).
19.6.2. Impedance Measurements
The frequency-response characteristics of a capacitor and its equivalent series resistance are important in an evaluation of a capacitor and are dependent
upon: (1) the intrinsic nature of the electrode material; (2) the pore-size distribution of the high-area material used in the fabrication of the electrodes; and (3)
the engineering parameters used in the preparation of the electrodes, e.g., the
thickness of the active material of the electrode and the nature of the particleparticle contact, as well as the macropore distribution as determined by the applied pressure. Impedance spectroscopy provides a convenient way to assess the
frequency -response characteristics of a capacitor material, especially the powerlimiting internal resistance.
Impedance measurements as a function of frequency can now be readily
made, even down to low frequencies of 0.01 Hz, e.g., by means of a Solartron
Frequency Response Analyzer or an EG and G/PAR lock-in amplifier. The
analyses of the resulting data are usually not simple and have been treated in
Chapters 16 and 17. However, various software packages are available for processing the results and performing equivalent-circuit simulations of recorded impedance spectra.
Impedance measurements are more useful for evaluating capacitor behavior than that of batteries because the capacitance component of the impedance
is the main quantity of interest and is directly determinable as l/jwC (Chapters
16 and 17). In addition, the combination of C with any series resistance (external
and esr) can be easily evaluated and the role of porous-electrode behavior
(Chapter 14) readily perceived from the frequency-response behavior of the im-
603
Electrode Holder
1 1 - - - Counter
Electrode
DHE or HgS04
Reference
Electrode
Bridge
FIGURE 19.3. Electrochemical three-electrode cell (thermostated) for screening the behavior of
capacitor electrodes by means of cyclic voitammetry. DHE signifies dynamic hydrogen (reference)
electrode.
604
Chapter 19
Noryl Base
Viton Pad
Gold Foil
Test E.lectrode
~ Viton Washer
~
~
/ ~;w
I
~
Noryl
FIGURE 19.4. Simple test-electrode holder for mounting in the electrochemical cell shown in
Fig. 19.1.
605
606
Chapter 19
REFERENCES
1. T. C. Murphy, G. H. Cole, and P. B. Davis, in Proc. Fifth IntI. Seminar on Electrochemical
Capacitors and Similar Energy Storage Devices, S. P. Wolsky and N. Marincic, eds., Florida
Educational Seminars, Boca Raton, Fla. (1995).
2. M. L. Kronenberg, B. J. Stein, and B. P. Codd, 1. Electrochem. Soc., 141, 2587 (1994).
607
Chapter 20
Technology Development
20.1. INTRODUCTION
In presenting this chapter on technology development, it must be stated
that attempts to make an up-to-date technology survey are restricted, unfortunately, by the proprietary nature of recent advances, details of which are often
unavailable in the public domain, especially information on the preparation of
electrode materials and electrode fabrication. Consequently this chapter is based
substantially on material garnered from conferences and seminars, and conference proceedings (see Appendix, Section 20.18) on the technological aspects of
the developing field of electrochemical capacitor energy storage. Especially
valuable are the conference proceedings volumes of the electrochemical capacitor seminars organized by Florida Educational Seminars, Inc., which are referenced in the text that follows by the abbreviation FES and the date.
Some technological information has, however, also been obtained directly
from companies currently active in the field, and a summary review of a selection of patents is given in Chapter 21. Nevertheless, limitations on the availability of production and process details, owing to intellectual property and
proprietary information restrictions, have naturally led to a less than comprehensive account of technology development in this chapter. However, the main
and general lines of development over the past 5 to 7 years are available and are
reviewed here. The sources of information from various conferences and their
proceedings volumes are listed in the appendix.
Around 1990-1991, developers of electrochemical capacitors became
aware of a new market for relatively large units required for power trains of
electric vehicles based on hybrid battery-capacitor power sources. This devel609
Chapter 20
610
In very recent years, the following possibilities for new and expanded utilization of such devices, including substantially scaled-up systems, both in actual
size and energy storage capacity, as well as in operating voltage, have been recognized:
Hybrid power systems, with batteries, for electric vehicles;
Cold, or heavy-load (diesel locomotive) starting assist;
Utility load leveling (large units);
Power-supply low-frequency smoothing;
Large capacitance ballast;
Automotive subsystems such as catalytic converter preheating, and possibly power-steering assist;
Military and medical (defibrillator) applications;
Actuators and motor drives;
Possible applications in robotics, probably in hybrid systems;
Technology Development
611
612
Chapter 20
1957
1969
1975
University
of Ottawa
Group
(funded by
Continental
Group, Inc.)
Mixed
oxides
(RuOr
Ta20s)
Construction Tar-lump
blacksulfuric
acid
electrolyte
Carbon
pastesulfuric
acid
electrolyte
Market
No market
for
low-voltage
capacitors
No market Military
for
low-voltage
capacitors
Principal
energy
storage
Double
layer
Double
layer
1978
1984
NEC/Matsus Pinnacle
hita
Research
Activated
carbon
fiber-cloth
organic
electrolyte
Memory
backup
Double
Redox
pseudocapaci layer
tance
1990
Matsushita!
Isuzu
Mixed
Activated
carbonoxides
(based on
aluminum
UOI
foil organic
Continental electrolyte
project)
Military,
Ultra-high
high rate
current
power
capacitor
charge and
dic,(;harge
Pseudocapac Double
itance plus layer
double layer
613
Technology Development
TABLE 20.2.
Developer
Electrode-electrolyte
Material
Panasonic
Pinnacle Research
Institute
Maxwell Laboratories
Maxwell Laboratories
Livermore National Lab
Carbon-organic
Mixed metal
oxides-aqueous
Carbon-organic
Carbon-organic
Aerogel
carbon-Aqueous
Sandia National Lab
Synthetic carbonaqueous
Los Alamos National Lab Conducting
polymer-aqueous
Wh/kg a
Whlliter
2.2
0.8
2.9
3
6
7
2
<2
<2
9
9
1.5
1.7
Wlkg- 1
Voltage
400
500
3V
28V
2500
2000
2000
24V
3V
IV
1000
IV
>500
0.75 V
Source: From Burke (FES 1995). Reprinted with the permission of S. P. Wolsky and N. Marincic, Florida
Educational Seminars.
a Weight based on cell, including current collector, active material, separator, and electrolyte.
electrostatic capacitors where the series stacking requirement involved in highvoltage electrochemical capacitor devices would be minimized. These systems
are described in a later section of this chapter.
As in most electrochemical technology development, the materials requirements are usually of paramount importance. Here, for capacitor development, they can be specific to the end use of the device, but some general
requirements applicable to most systems can be formulated as indicated in the
following sections.
20.4.1. Electrodes
614
Chapter 20
Optimized pore-size distribution for maximum specific area but minimized internal electrolyte resistance (minimum esr);
Good wettability, hence favorable electrode/solution interface contact angle (depends on pore structure);
Minimized ohmic resistance of the actual electrode material and contacts;
Capability of material being formed into electrode configurations having
mechanical integrity (e.g., compressed powders with binder, fiber or
woven fiber matrices, aerogel materials, glassy-carbon structures) and
minimum self-discharge on open circuit.
In relation to the above ideal requirements, the active area, or active weight
or volume, is a common concept involved in descriptions or specifications of
battery electrodes. It refers to the active mass of electrode materials that can undergo the required electrochemical redox reactions of the battery. A similar term
can be applied to electrochemical capacitors; in the case of double-layer-type
capacitors, it refers to the real area of electrode matrices that can undergo electrostatic charge acceptance or charge depletion. In the case of redox-pseudocapacitance devices, the terms have the same significance as for battery systems,
but the real accessible electrode area is also usually important because this determines power performance.
The electrochemically significant active area in an electrochemical capacitor device is usually significantly less than areas of carbon materials determined
by BET measurements in the dry state. As referred to earlier, this is due to problems of wetting and invasion by the electrolyte into the finer pores of the pore
distribution. A useful summary of the characteristics of various electrode materials is shown in Table 20.3.
TABLE
Electrode material
Carbon-metal composites
Aerogel carbon
Cellulose-deriv. foamed
carbon
Doped polymer
(polyaniline)
Anhydrous ruthenium
oxide
Hydrous ruthenium oxide
Specific Cap.
(F g-l)
Surface area
(m2 g-l)
Densi~
(g.cm- )
Resistivity
(ohm-em)
100 to 200
120 to 160
70 to ISO
100 to 1500
500 to SOO
500 to SOO
0.5 to 0.7
0.4 to O.S
O.S to 1.0
<0.01
<0.01
0.Dl5
400 to 500
200 to 400
0.5 to 0.9
0.01
100 to 150
100 to 150
2.5 to 3.0
<0.001
600 to 750
SOO to 100
2.3 to 2.5
0.002 to
0.005
1.4 to 2.0
12 to 17
Aqueous electrolytes
Organic electrolytes
Source: From Burke (PES 1994, 1995).
US to 1.24
1.1 to 1.2
Technology Development
615
The general behavior and properties of carbon preparations for electrochemical capacitors of the double-layer type were reviewed in Chapter 9. The
main practical requirements are high specific areas, optimized porosities and
pore-size distributions, and good electrochemical stability on cycling. Generally, preparations having very high specific areas tend to be the least stable because they have the greatest surface-to-volume ratios and often the largest
presence of oxygen-type surface functionalities (Chapter 9).
The required porosities may be obtained by phase separation,3-9 gelation4-6.lO
emulsification,3, 11, 12 aerogel-xerogel formation,3,13,14 replication, 15 and by general activation procedures. Another original procedure for preparation of activated C plus carbon composite electrodes having controllable sizes and volume
fractions of open pores involves adding raw polymer particles [e.g.,
poly(methyl methacrylate) spheres, 20 f1m in diameter], that are subsequently
burned out. The double-layer capacitances achieved can be up to 234 10 F g-I,
which are rather favorably high values.
Work by Delnick, Ingersoll, and Firish (FES 1993) concentrated on an impedance evaluation of one of these specialized C materials, namely, microcellular carbon foam derived from activated cellulose carbon, used as a
double-layer capacitor electrode. The impedance analysis followed the de
Levie l6 treatment of porous electrodes (see Chapters 14 and 17).
Photomicrographs of the cellulosic porous electrode material are shown in
Fig. 20.1. This material gave a complex-plane impedance plot typical for a porous electrode according to the transmission line model (a 45 plot of -Z" vs. Z'
at high frequencies with transition to a simple capacitance behavior corresponding to -2" increasing with diminishing frequency with almost constant Z').
The material studied l7 . 18 had a volume-specific capacitance of 58 to 150 F
cm-3, depending on the surface modification achieved, pore structure, and poresize distribution. For 0.025-cm-thick electrodes, the effective series resistance
in 1.0 M aqueous sulfuric acid was 0.24 ohm/cm2, a relatively high value.
20.4.3. Activation Procedures for Carbon Particles and Fibers
Generally, "raw" carbon powder or fiber materials are not satisfactory for
double-layer capacitor fabrication without some activation procedure conducted at an elevated temperature in a controlled gas atmosphere (Chapter 9).
Pretreatment procedures for activation of carbon materials (powders, polymers,
fibers) are usually necessary in order to achieve the required large specific areas
(l00-2000 m 2 g-I) or to modify the morphologies of fiber materials (e.g., see
Lipka, FES 1993-1995), as well as to modify the distribution of surface functionalities. Such techniques are often proprietary so that details of procedures
are not usually available in the public domain. However, generally, pro-
Chapter 20
616
CLIIFCMM-040
dens
area
880mglcm3
830 m 2/g
Technology Development
617
TABLE 20.4.
Expectations
Maximum excess charge density corresponds to perhaps
0.15-0.20 electrons per surface atom (see Chapter 2)
For 2000 m2/g and 25 /1F/cm 2, maximum (theoretical)
capacitance is 500 FIg
500 FIg
166 FIg
618
Chapter 20
The most investigated oxide electrochemical capacitor has been the RuOr
aqueous H 2S04 system, treated in detail in Chapter 11. Ru02 thermally formed
on Ti or on C felt, and electrochemically formed on bulk or electroplated Ru
metal by anodization, has been extensively studied at the University of Ottawa
Laboratory and developed by Continental Group, initially at Hooker Research
Center and later by Pinnacle Research and the U.S. Army group at Fort Monmouth. The RU02 electrodes thermally formed by the DSA procedure have
properties significantly different from those formed electrochemically; the latter
are hydrous oxides and exhibit a larger ratio of oxide redox pseudocapacitance
to double-layer capacitance than that for the DSA-type Ru02.
A useful improved technology is the direct binding of RU02 onto an
ionomer separator, developed by Sarangapani at Giner, Inc. and described in
Section 20.5.2. The procedure is reminiscent of ionomer fuel cell electrode design and maximizes energy and power density, with minimized esr. Various
other oxide and oxide-mimicking (e.g., Mo nitrides) redox pseudocapacitance
materials that are much cheaper than RU02 are known [e.g., oxide films on Mo,
W, and Co (C0 30 4)], but they have too small an operating charge and discharge
range-only about 0.7 V.
20.4.5. Conducting-Polymer Electrodes
Technology Development
619
620
Chapter 20
which Ru02 was bonded to an ionomer electrolyte/separator to make a solid redox pseudocapacitance device (see description in Section 20.5.2).
(20.2)
This is n times the energy stored in each unit of capacitance Cj , each charged to
a potential difference Vi' Thus, although n capacitors combined in series have
together lin of the individual unit capacitance value, the overall stored energy
is n times that for each unit, Eq. 20.2. Reflection will show that this result is a
necessity arising from the law of conservation of energy since the simple series
Technology Development
621
1
1
- = - + - + - + .. etc.=I, lie;
en e, ez e3
(20.3)
However, the potential differences across each unit are different but add up to
the total, V. Thus the potential differences are
n
(20.4)
(20.5)
and the charge held by the assembly at overall voltage V is accommodated
among the capacitor units. Depending on the respective unit capacitance values,
e b ez, e3, etc., qb qz ... are the respective distributed charges. It is therefore
seen that for a given q, the potential difference across each capacitor element is
inversely related to its respective capacitance. In an array of capacitors in series,
it is this individual cell potential difference that should not exceed the decomposition potential of the electrolyte in that given cell element.
The principal practical difficulty in engineering a multiunit series capacitor
of a double-layer or redox type is to match n capacitors so they each have the
same capacitance values and, if possible, other similar performance parameters
such as esr. If one or more capacitors in a bank in the series differ in rating, there
is a danger, on charge or discharge, of overcharging or overdischarging, with the
result that electrolyte decomposition, with the possibility of gassing, could occur, usually with undesirable consequences. This occurs because one or more
capacitor cells could suffer a polarity voltage difference greater than the decomposition voltage of the electrolyte. Analogous problems arise, of course, in a
regular battery containing one or more unmatched cells.
One useful procedure in charging a bank of capacitors in series is to actually charge them in parallel, so that each is individually charged. Then, by a simple switching arrangement, they can be discharged in series, giving a voltage
multiplication n times the individual cell voltage. This device, although convenient and well known for static dc voltage multiplication, would present electrical
622
Chapter 20
Seals
--"1HHHr--~~
//\
III f
Case
Case
FIGURE 20.2. Schematic diagrams of bipolar electrode arrangements for an electrochemical capacitor showing a combination of capacitor units equivalent to a bipolar stack.
___ Ieakage
I II I
pathway
FIGURE 20.3. As in Fig. 20.2 but illustrating current leakage or shunt effect at a leaky seal in a
bipolar electrode array.
623
Technology Development
discharge. The effect is the same as that of putting the electrically joined cathode
and anode of two neighboring cells of a multicell battery into a common volume
of electrolyte; then short-circuit discharge between two such electrodes of
course rapidly occurs. The situation is closely analogous to the behavior of a
galvanic couple, e.g., Cu and Zn metals in contact, immersed in an electrolyte.
It must be understood that there is a double layer on each side of each bipolar electrode, but with a polarity opposite to that of the solution (Fig. 20.4).
Down the array of cells, from EI to E 2, there is a progressive series of potential
drops (Eq. 20.4) at each electrode interface along the cell stack. The bipolar geometry can be imagined to arise from bringing plates A and B in Fig. 20.4a directly into back-to-front contact corresponding to the bipolar plates AB in Fig.
20.4b. Electrolyte resides in the space between the end plate EI and the first
plate A and then between B and the next plate A, etc., to the end plate, E2 The
configuration b requires the ends of each bipolar plate to be sealed to an outer
confining case so that each region is electrolytically separated from its neighbor.
Depending on the mode of construction of the bipolar assembly and the rigidity
of the plates, the bipolar configuration mayor may not require conducting separators between E, A, B, and E2 , and between the repeated series of electrode
pairs, AB, AB, etc.
One of the practical developers of carbon-based bipolar electrochemical
capacitors in a major but exploratory way is Maxwell Laboratories (San Diego,
California). They were funded in 1992 under the DOE Ultracapacitor Development Program, which aimed at units capable of providing at least 5 Wh kg -I and
power capability in the range of 500-1000 W kg-I. Another bipolar, carbonbased capacitor, employing a nonaqueous electrolyte, and rated at 70 V, was
demonstrated by Redox, Inc. (Marincic, FES 1995).
20.4.10. Current Distribution in Capacitor Devices
Current distribution problems arise in two ways in an electrochemical capacitor in discharge or recharge. The first is in a microscopic way, within pores
(a)
Unipolar stack
(b)
Bipolar electrodes
(seals to case not shown)
FIGURE 20.4. Change from a multi-unipolar stack to an equivalent bipolar electrode stack configuration (schematic).
624
Chapter 20
Technology Development
625
some gas flow and water management arrangements can assist with the management of heat produced from irreversibility of the cell reactions and PR heating
during power production (Section 20.8).
For large capacitor units for which there is no self-ventilation, heat management requires special design engineering and control. Continuing work in
this area is required for hybrid electric vehicle motive-power systems, especially those for high-power applications. Further details are discussed in Section
20.7.
The transition from laboratory-scale modules to larger, practical commercial systems, especially in the case of electrochemical capacitors for battery load
leveling in electric vehicles, involves a number of well-known electrochemical
problems, as follows:
1. change of current distribution, related to change of geometry and length
of ohmic pathways;
2. hence, usually, increase of esr;
3. possibility of localized overcharge;
4. when the scaleup involves multiple bipolar electrode configurations,
the avoidance of shunt leakage pathways, and
5. emergent problems of heat management, especially with stacked, multicell bipolar designs.
Several of these problems also arise in well-known ways in larger scale battery
designs and in the technology of electroplating.
Scale-up factors for use of organic electrolyte electrochemical capacitors
in hybrid configurations with lead acid and zinc-bromine batteries were considered by Marincic (PES 1995) in some tests and modeling. A large bipolar plated
carbon-felt capacitor, rated at 70 V, was tested. Good plastic sealing of the edges
of the electrodes in the multiplate assembly was achieved, leading to virtually
zero shunt effects.
The energy density and the power density were shown plotted in a useful
comparative way against a common axis of capacitor weight fraction in the two
battery hybrid systems (Figs. 20.5 and 20.6). Capacitor power density increases,
as expected, relative to energy density of the hybrid combination because a
greater weig'lt fraction of the capacitor component is present. Note that the
crossover point arises at a higher capacitor fraction in the case of the Zn-Br2 than
for the lead acid hybrid system. These working diagrams provide a useful basis
for the design and optimization of performance of hybrid capacitor-battery
power systems for electric vehicles where scaleup becomes an important con-
626
Chapter 20
80 r-----------------------------------------------~600
LEAD-ACID HYBRID
101
x.
.s::
400 "-
"-
>-
>t:
f-
U)
300 ~
W
0
>-
200
C)
a::
w
z
w
a::
w
~
100
__
_ _ _ _L -_ _
0.2
_ _ _ _L __ _
_ _ _ _L __ _
0.4
_ _ _ _L __ _
0.6
__
c..
~O
0.8
sideration for the high voltage and high energy, high power levels of operation
required for such applications.
With carbon-felt electrodes oflarge size in monopolar configurations that
do not need to be supported on a separate current collector matrix, scaleup can
lead to increasing voltage losses across the felt plate to the edge current collector
connections relative to those in a smaller scale electrode of the same, say rectangular, geometry. These are the same factors that also apply in scaleup of the
size of battery electrodes. This ohmic scale-up factor depends on the specific re-
80
r-------------------------------------, 500
Zn/Br 2 HYBRID
'",""
.c:. 50
;:
"-
t:
rJ)
Z
40
W
Cl
>\!l
a:
w
Z
w
200
5
'!:
20
100 Cl.
L-__L -_ _L-~__~~~--~~-L--~~~--~O
00
0.2
0.4
0.6
0.8
FIGURE 20.6.
As in Fig. 20.5 but for hybrid with a Zn-Br2 battery. (From Marincic, FES 1995.)
Technology Development
627
628
Chapter 20
only a part of the stored energy is available (see Chapter 17); e.g., in discharging
from 400 to 200 V (or in general from Vinitial to (1I2)VinitiaI>, 75% of the stored
energy becomes available since stored energy is (1I2)CV 2 at any voltage, V. The
Maxwell sintered fused composite electrode structures are capable of meeting
these specifications with the stated energy-density requirement and are now figuring prominently in electric vehicle power-train technology (see Section
20.13).
Deposition of other materials, such as the electroactive polymers, onto
carbon-fiber support material has also been employed, and was later investigated in the author's own laboratory27 for the case of Ru02 in order to provide
more extended films.
Nishino and Yoshida (PES 1991) described the direct use of activated carbon fibers as the electrode material for a double-layer capacitor. The technology
involved plasma spraying aluminum directly onto the surface of a compacted fiber matrix to provide good external circuit contract.
It has been emphasized in several chapters of this volume that for successful design of double-layer capacitors, high-area carbon materials with accessible fine pore structures are required. Delnick, Ingersoll, and Firish (PES 1993)
have evaluated a variety of microcellular carbon foams (Fig. 20.1) derived by
controlled pyrolysis and carbonization of polymers such as polyacrylonitrile
(PAN), polymethacrylonitrile (PMAN), resorcinol-formaldehyde (RR),
divinylbenzene cross-linked methacrylonitrile (DVBPMAN), phenolics (furfurylaicohol), and cellulose polymers such as rayon.
Further results of tests by Andrieu et al. (FES 1993, 1994) (Aicatel
Alsthom Recherche) on liquid organic electrolyte electrochemical capacitors of
D-size having a capacitance of 175 F (75 g) and operating voltage of 3.0-3.1 V
(internal resistance 25 mohms), and manifesting an energy density of 2.9 Wh
kg- 1 and power density of 1.20 kW kg- 1 are reported in PES (1994). The separators used were 50-,um microporous polyolefin films, and the active surface
was ca. 600 cm2. The electrodes consisted of a thick coating of activated carbon
black on a perforated aluminum foil with a PTPE binder.
The materials have been examined in some detail by Lipka (PES 1993);
they possess a high degree of crystalline anisotropy. Depending on processing
and important pretreatment conditions, a variety of specific properties, useful
for electrochemical capacitor fabrication, can be generated. In the work described by Lipka (PES 1993), pitch-based carbon was used in as-received and
activated preparations, as well as some materials containing PAN-based carbon
fibers and epoxy matrices. The nominal fiber diameters were 10 ,urn. Capacitance behavior was characterized by ac impedance and cyclic voltammetry.
An important relatively high-area carbon material, an alternative to higharea carbon powders, is carbon fiber or felts of carbon fibers that give specific
629
Technology Development
Voltage bias
(V)
Total cell
capacitance at
I mHz(F)
Specific
capacitancea
(F/g C)
0.01
0.99
0.01
3
6.6
4.0
3.9
1.6
241
147
286
116
Activation
treatment
esr(mohm)
270
427
capacitance values as listed in Table 20.5. These have the advantage of easier
handling and mechanical management.
The (proprietary) pretreatment conditions can give rise to specific differences in the morphology of the resulting modified carbon materials so that, to
some extent, tailormade structures of controllable morphology can be prepared
(Lipka, FES 1993). This offers important opportunities for advancing the technology in this field, leading to desirable materials properties of the basic carbon
structures required for double-layer-type, electrochemical capacitors.
Figures 20.7 and 20.8 show SEM micrographs of two carbon fiber materials activated in two proprietary ways, "A" and "B." It can be seen that major
changes in the accessible inner area of the carbon fibers can be obtained, depending on the mode of pretreatment.
Lipka and Zawacki (FES 1995) continued their earlier work (FES 1994)
by examining the cycling and other aspects (Tables 20.5 and 20.6) of the performance of carbon fiber materials for electrochemical capacitors. Activated
carbon fibers of 10 11m diameter prepared from pitch were used in chopped
forms having a range of aspect ratios (length to diameter ratios). The activation
procedure (proprietary) leads to multiple splitting (see SEM photos, Figs. 20.7
and 20.8) of the fibers, producing an enhanced area. The initial fiber material
(so-called Type 1, short fiber), as received, had an area of 0.35 m2 g-1 while after
activation the specific area was increased to 0.73 m 2 g-I.
TABLE 20.6. Differential Capacitance for Carbon Fiber Epoxy Composites
Fiber orientation
Parallel
Perpendicular
Random
Source: From Lipka (FES 1993).
aCapacitance per area of carbon fiber measured at 0.05 V vs. SCE.
Capacitance/area (I1F/cm2)a
167
380
371
630
Chapter 20
FIGURE 20.7. Scanning electron micrographs of mesophase pitch carbon fibers and PAN fibers before and after activation treatments: (a) Mesophase pitch
carbon fiber A, as received (fiberlength). (b) Mesophase pitch carbon fiber A, as received (fiber end). (c) PAN carbon fiber A, as received (fiber end). (d) Mesophase pitch carbon fiber A, after activation process A (fiber end). PAN = polyacrylonitrile polymer precursor; Proprietary activation processes A and B are distinctively different, but proprietary, activation treatments. (SEM photos kindly provided by Professor S. Lipka of Aorida Atlantic University, FES 1993.)
Ol
W
.....
:s
'g
0"
3
III
'<
cc
o
0"
:s
fiber A, after activation process C (fiber length/end). (From Lipka, FES 1993.)
FIGURE 20.8. As in Fig. 20.7 but for two activated carbon fibers. (a) Mesophase pitch carbon fiber B, after activation process B (fiber length). (b) PAN carbon
en
..
DI
oJ
Co)
I\)
Technology Development
633
Io.1mA
FIGURE 20.9. Comparison of cyclic voltammograms for activated and unactivated carbon fiber
of Type I in 30 wt.% aqueous H2S04 taken ata sweep rate 0[200 mV s-l. (From Lipka, FES 1994.)
634
Chapter 20
Technology Development
635
A useful and original technology for development of a solid-state electrochemical capacitor of a different type was reported by Bruder and Owens (PES
1993) and Oxley (FES 1991, 1993). Early devices of this kind, based on the Ag+conducting solid electrolyte RbAg41, were described some 25 years ago. Newer
technology, based on the solid polymer electrolyte, Nafion, has led to favorable
performance except with respect to energy and power densities, the latter due to
the limited specific conductance of polymeric ion film electrode separators.
However, some backup, low-power, low-voltage applications can be realized.
One aspect of the use of RbAg41 is that decomposition of the electrolyte to 12 at
the anode occurs between 0.5 and 0.66 V (vs. Ag-Ag+), depending on conditions, but the adsorbed 1/1- couple leads to a pseudocapacitance. 28
The solid-state device is trade marked as Hypercap and has a different type
of configuration from that of either a regular two-pole carbon capacitor or an anode-cathode battery. The system (Fig. 20.10) utilizes the solid electrolyte
RbAg415 in which the high solid-state conductivity of this salt was designed into
the performance of a multistack, high-voltage energy storage system.
The device is neither a pure capacitor nor a battery, but rather a hybrid consisting of a silver anode separated from an activated-carbon cathode (double-
CATHODE
PLATING SILVER
ON
SILVER
..
ELECTROLYTE
ELECTROLYTE
DEPLETION
ANODE
DOUBLE LAYER
FORMATION
e-
636
Chapter 20
637
Technology Development
......
>- 200.- .. .............
!::
(/)
zw
o
100
a::
w
~
a.
( 0)
.... ......
... ".,
,
800
lOt
~
600
>-
I-
in 400
Z
.,
r
'
..... ,
...
'.
... ...
....
.,
.... , ,
200
3:
a.
0
(b)
..........
a::
-.
O~------~--------~------~~"'e--
..........
,,
..... ,
'.,
'.
...
....
...
2
3
ENERGY DENSITY I Wh kg- 1
1
FIGURE 20.11. Modeling of Ragone plots for a double-layer capacitor having a thin-film configuration and electrolyte conductivity of 5 x 10-5 S cm -1, which is typical of most solid polymer
electrolytes. Capacitance value: 40 F g -I. (a) 1OO-,um electrodes, 50-,um electrolyte. (b) 50-,um electrodes, 30-,um electrolyte. (From 8audry et aI., FES 1994.)
made from tantalum bearing a Ta20S barrier oxide film) coupled with a lowvoltage, high energy-density electrochemical redox pseudocapacitance electrode (formed e.g., from RU02) serving as the cathode of the electrode pair.
Operating voltages of around 150-230 can be achieved with capacitances up to
100 mF. Capacitance values remain constant within about 2% over a voltage
range of 50 to 230 V. Complex-plane impedance plots are typical of a transition
638
Chapter 20
DIELECTRIC
N
T
I
A
D
E
L
+
-------
MMO
\\
I
E
L
E
C
T
R
L
Y
T
E
C
A
T
H
D
E
FIGURE 20.12. Voltage profile normal to the electrodes in an Evans Hybrid-type capacitor using
a mixed metal oxide cathode material (schematic).
Technology Development
639
where R; is the sum of the internal electrolyte and separator resistances. Normally such a capacitor is operated with recognition of its polarity, the high voltage being established over the thin oxide film of Al 30 3 or Ta205 at the anode
exhibiting capacitance Ca <Co see above). The oxide film is the dielectric in
Ca and the pseudocapacitance of RU02 gives rise to Ce . At each oxide/electrolyte interface, double-layer charge separation also occurs, but the overall capacitance for such electrolytic capacitors does not arise primarily from the
capacitance of those double layers as it does in an electrochemical double-layer
capacitor.
In order to provide some examples, a selection of performance data for 50-,
28-, and 32-V units of the Evans "Hybrid" capacitor having capacitances of 18,
18.1, and 8 mF are listed in Table 20.7. Each unit has an attractive high operating voltage capability, but their capacitances are on the order of 1000 times
smaller per gram than those of currently developed regular carbon-based double-layer or RU02 redox capacitors. However, the energy storage capability
must be considered: at 50 V, a 20-mF capacitor can ideally store (50)2 x (20/2)
X 10-3 joules, i.e., 25 J. An RU02 pseudocapacitor of, say, 10 F g-1 operating,
say, at 0.5 V (half of a I-V span between two similar electrodes) would store
only 1.25 J per gram. Thus the Evans high-voltage device would have better energy storage capability than a small (1 g) RU02 electrochemical redox pseudocapacitor. It also would have the advantage of substantially lower esr and
consequently higher pulse power capability, with a low RC time constant.
Weight, volume, and energy density of a hybrid Evans capacitor were
compared with these quantities for a corresponding aluminum electrolytic ca-
TABLE 20.7.
Item
2
3
Capacitance
No. of (120
Hz) anodes
(mF)
18
18.1
8
Voltage
(volts)
50
28
32
Dimensions
diameter X
Anode
height
pellet
(mm)
weight (g)
Weight (g)
56
30
20
36 x 12
36x 8
36x7
12
6
12
No. of
anodes
3
3
640
Chapter 20
FIGURE 20.13. Photo of an Evans Capattery capacitor capable of operation at selected voltages
of 15, 30, 60, or 120 V. (Photo provided by the Evans Co.)
pacitor, both having specifications of 18 F and 50 V operating voltage. Clear advantages of the hybrid device were demonstrated by Evans (PES 1995); the hybrid has 2.5 times higher gravimetric energy density and 11 times higher
volumetric energy density than a corresponding aluminum electrolytic.
Note that the high operating voltage of the Evans "Hybrid" capacitor is
achieved by most of this voltage falling across the insulating oxide barrier-layer
film, with only about 1 V remaining across the electrolyteIRu02 interface (Fig.
Technology Development
641
5
4
>
.....
.J
IZ
W
<t
b
a..
Log [tim.e/h]
FIGURE 20.14. Self-discharge curves (log plots in time, cf. Chapter 18) for a commercial Panasonic double-layer capacitor from four states of charge., 48 h; 0, 3 h; x,O.5 h; and D., 0.05 h. Voltage, 5-V, dc.
642
Chapter 20
An example of self-discharge of a commercially available packaged capacitor, manufactured by Panasonic, is shown in Fig. 20.14 in which the declining potential is plotted empirically as a function of log [time]. This latter type
of plot is, however, fundamentally based (see Chapter 18) and originates when
a continuing charge transfer process, following the Tafel polarization relation,
discharges the single electrode interface.
In the evaluation test experiments by Lipka (FES 1993) on his carbon fiber
capacitor electrodes, described in Section 20.5.1, measurements were also made
of their self-discharge behavior. Substantial discharge (-90%) from charging to
3 V for 2 h took place after 100,000 cycles over 4 days on open circuit. The decline of voltage with time approximated a square-root relation in time, indicating (see Chapter 18) the role of diffusion-controlled depolarization by redox
impurities.
As emphasized in Chapter 18, self-discharge is a general and fundamental
phenomenon, observed with both batteries and supercapacitors. It is particularly
significant with the latter devices, where a 50% loss in charge can take place in
a few days or weeks, accelerated, as may be expected, at elevated temperatures.
Attention has been given to empirical evaluation of the self-discharge of
packaged electrodes containing one or more pairs of electrodes, e.g., in tests by
Evans (on their capacitor devices), Giner, and especially by Burke and Murphy
in the INEEL testing program (see FES 1991-1993). In a recent paper by Conway with Pell,27 a diagnostic examination of self-discharge mechanisms was
formulated quantitatively, with analytical tests being performed on individual
charged carbon and RU02 electrodes in a three-electrode measuring cell configuration (so that the potential decay at the initially charged electrode could be
followed digitally) against the potential of a reference electrode in the same solutions over a period of several days. This is the preferred procedure for fundamental study of the self-discharge phenomenon, i.e., without the complication
of what may be occurring at a second, oppositely polarized, conjugate electrode.
Sarangapani et a1. 29 used a different procedure for characterizing the selfdischarge behavior of their solid polymer electrolyte, RuOrbased capacitor device. The energy efficiency, in terms of the ratio of the energy stored on charge
to that recovered on discharge after various durations of self-discharge (0 to 302
s), was evaluated. This ratio declined from 72 to 54% over the above period of
time. The effect of temperature on the energy efficiency over a range of -4 to
+46C, was also determined. The energy efficiency increased from 29 to 66%
over this range, in this case for direct charge and discharge, i.e., without time
being allowed for significant self-discharge to take place. These results indicate
relatively high rates of self-discharge of RU02 in the polymer electrolyte device.
More work on this topic is needed since the self-discharge phenomenon restricts
practical applications of electrochemical capacitors for certain purposes and impairs the quality of commercial specifications of capacitor devices.
Technology Development
643
All electrochemical energy storage devices experience changes of temperature upon charging and discharging. In the case of batteries and fuel cells,
the thermal effects are of two kinds: (1) from [2Rjoule heating and (2) from the
heat change generated from the Faradaic cell reaction. Ideally, for this latter
case, all of the Gibbs energy is available as electrical energy, tlG = -zFE, and
this is the sum of the enthalpy and - T times the entropy, tlS, of the cell reaction:
M-I - TtlS = -zFE. However, this relation is true only if the cell reactions take
place reversibly, i.e., at quite low current densities. At the practical discharge
rates required for high power battery or fuel cell operation, the Faradaic reactions do not proceed reversibly and appreciable overvoltages arise, e.g., 0.30.4 V in an H2-0 2 fuel cell. Under such conditions, heat is generated that is
equivalent to Ttimes the entropy generation associated with the irreversible behavior of the processes.
In an electrochemical capacitor on discharge and recharge, factor (1) is
again important and [2R joule heating occurs from the passage of current
through the capacitor cell involving mainly the equivalent series resistance of
the device plus that of any interconnects or current collectors. In an electrochemical double-layer-type capacitor, at least of an ideal type, no Faradaic processes take place on charge or discharge below the potential difference for
solution decomposition; that is, heat effects of type (2) would occur only on
overcharge or overdischarge at appreciable rates. However, in charging a double-layer capacitor, the double layer usually suffers a negative entropy change,
-tlSch , and this is associated with a Gibbs energy change of -T(-tlSch ). Normally there is no enthalpy change, Mi, except for some practical cases. The energy is of opposite sign upon discharge, of course, but the joule heating effect
is the same in both directions for a given discharge and recharge current.
In a high-voltage, multicell, stacked capacitor unit, probably configured ,
with bipolar plate electrodes, the heat dissipation problem will be worse than for
single-cell units of the same individual electrode areas and geometry since it is
the surface-to-volume ratio that determines the rate of heat loss to surroundings
or to a ventilation system.
In use, the heat management problems with a capacitor are much like those
for a battery and depend on the power level of operation. In practice, optimum
heat loss can be achieved by maximizing surface-to-volume ratio, but most applications (including the large-scale one in which a capacitor stack is hybridized
with a large rechargeable battery for electric vehicle power trains) require minimum volume, i.e., a geometrical arrangement for maximizing overall volumetric power density. These two requirements are obviously in opposition and
compromises must be sought in the form of forced ventilation or use of a cooling
fluid and heat exchanger, as with an internal combustion engine.
644
Chapter 20
-23C
23C
20
>
"0
...c 16
Q)
...0
a..
MAXWELL 24V
12
0
20
40
60
80
Time / h
FIGURE 20.15. INEEL self-discharge test at three temperatures on Maxwell 24-V unpackaged
capacitor. (From Murphy, Cole, and Davis, FES 1995.)
645
Technology Development
10.--------.--------.--------.--------r-------,
Q)
Cl
.s 7
o
>
300 A
4 L _______-L______
~L_
______
15
10
Discharge time (5)
______
~~
20
______
25
FIGURE 20.16. Effect of temperature on discharge curves for a Panasonic 3-V, 1500-F double.
layer capacitor at 100 and 300 A discharge rates. (From INEEL test, Murphy, Cole, and Davis, FES
1995.)
646
Chapter 20
120r---------------------------------------------------~
M 50C - __ _
C 50C
---
"0
Q)
.~
Qj
---- ---
100
----------------
~
c
g
u
0.
80
'15
Q)
60
Q)
~
40~----------~~----------~~--------~~~--------~
0.00
0.25
0.50
0.75
1.00
Technology Development
647
The second property offundamental significance is the actual specific double-layer capacitance (F cm-2). Its temperature dependence is determined by
how (1) the effective dielectric constant of the solvent in the double layer depends on temperature (it will generally decrease with increasing temperature,
especially for associated, dipolar solvents); and (2) the effective thickness, d, of
the double layer depends on temperature. Like condensed-state substances, the
interphasial fluid of the double layer tends to expand with a rise in temperature,
so the effective average thickness of the compact part of the double layer increases and its specific capacitance decreases. The change is about 10-15%
over 60C at Hg electrodes, so this is not a major effect having serious practical
consequences. In a more detailed fundamental way (Chapters 6 and 7), the dependence of double-layer capacitance on temperature is substantially more
complicated because factors other than the thickness of the compact region are
also involved.
In the case of capacitors depending in part on Faradaic charge and discharge of a redox pseudocapacitance, temperature effects occur on account of
the finite activation energy associated with the kinetics of the charge transfer
process (e.g., e and H+ hopping and transfer) in the usual way for electrochemical reactions.
In the case of self-discharge, the temperature dependence arises from the
significant heat of activation, tM/+, associated with the electrochemical kinetics
of such processes; tM/+ can have values between about 20 and 60 kJ mor l ,
which will lead to substantial differences in self-discharge rates between, say,
-30C (low) to +60C (usually quite high rates). Since this effect is an intrinsic
aspect of chemical and electrochemical rate processes (Chapter 3), there is little
that can be done to diminish such temperature effects except to lower the actual
rates by reducing the concentrations of species (usually redox impurities or 02)
that give rise to self-discharge, commonly by shuttle redox reactions (Chapter
18).
Some evaluations of the effects of temperature on capacitor performance
have been made by Murphy, Cole, and Davis (FES 1995) in INEEL testing procedures. Some results of their tests at Idaho National Engineering and Environmental Laboratory follow. An example of the temperature dependence of
self-discharge of a Maxwell 24-V (unpackaged) capacitor over the range -23C
to +60C is shown in Fig. 20.15 and that for a Panasonic capacitor in Fig. 20.14.
It is can be seen that there is a major effect corresponding to a tM/+ of about 40
kJ mol-I. The results of tests on a commercially available Panasonic 3-V, 1500F capacitor discharged at 100 and 300 A between -20 and +65C are shown in
Fig. 20.16 (Murphy, Cole, and Davis, FES 1995). The capacitances achieved
(reciprocal slopes of the lines) remain almost independent of temperature, but
the voltages diminish with lowering of temperature, which is due mainly to increased esr because of a conductance decrease.
648
Chapter 20
649
Technology Development
1.0 r---"~I.---------------t
~3Vt1500F
0.8 r-
W
"W
~/3Vt500F
.~
0.6 f-
0.4
.,
.'----L.---L--'----'---J1L----L.-....I1---'----L-----I
500
1000
FIGURE 20.18. Constant power discharge characteristics of 3-V, lS00-F and 3-V, SOO-F Panasonic capacitors shown as relative energy stored /100 vs. power density (INEEL tests, from Burke,
FES 1993).
Constant power charging or discharging is a third mode that has been employed, e.g., Fig. 20.18. Generally, for this procedure, the current must be decreased as the voltage increases so that I x V remains constant. This provides
better homeostasis of the conditions inside the packaged device.
20.S.3. Rate Effects on Charge or Discharge
In the case of double-layer capacitors, limitation of charging or discharging rates occurs mainly on account of the distributed resistance in the pore structures, as discussed in Chapter 14, and the associated range ofRC time constants,
giving rise to a power spectrum. With electrochemical capacitors based on
pseudocapacitance, dispersion of capacitative response at high rates of charge
or discharge occurs, however, for electrode-kinetic reasons, as discussed in
Chapter 10, which are based on the polarization principles treated in Chapter 3.
Electrolyte conductance is, of course, also a parallel factor.
Generally, rate effects of the above kinds cannot be eliminated because they
occur for unavoidable fundamental reasons. However, they can be effectively minimized by careful choice of design, materials, and operating conditions.
650
Chapter 20
subject to environmental and safety regulations extant in various countries, especially in the United States. However, in actual use, capacitors are less likely
to be hazardous than some batteries, especially regarding fire and explosion incidents that can occur, e.g., with various types of Li batteries or fuel cells.
Dangers can arise, however, as with batteries, if adventitious overcharge
or overdischarge takes place, resulting in gassing, especially in the case of aqueous solution devices. In the case of the nonaqueous solution, double-layer types
of electrochemical capacitors, overcharge can result in the production of noxious gases from the decomposition of, e.g., propylene carbonate and/or tetraalkylammonium salts. The actual local environmental safety hazard will depend
on whether such gases are vented in significant volume. In the aqueous case,
where either strong KOH or H 2S04 solutions are used as electrolytes, the usual
burn hazards associated with such solutions are recognized if casings become
opened by buildup of excessive gas pressure, or an explosion hazard arises from
generation of stoichiometric 2H2 + O2 gas mixtures.
From an environmental safety point of view, the problems are probably
more significant in the workplace, where handling of large quantities of very
fine carbon particles or fibers could lead to lung irritation or even, under some
conditions, to dust explosions. Also, good ventilation systems must be provided
in dry rooms to minimize significant dispersion of carbon particles or to avoid
buildup of noxious solvent vapors. While propylene carbonate has a low vapor
pressure at ordinary temperatures, some added cos 01 vents (e.g., dimethoxyethane) are more volatile and also noxious; they are also more of a fire
hazard.
Pyrolytic procedures for formation of carbon fibers from polyacrylonitrile,
rayon, phenolics, or cellulose can lead to formation of noxious gases. PAN is a
common precursor in the preparation of activated carbon fibers, and the formation process leads to evolution of gases, including H 20, NH3 , COb and Nb together with CO, HCN, and nitriles; the latter three are, of course, an
occupational health hazard for operatives in plants or laboratories where this
material is prepared. The carbon fibers are themselves irritants and pose hazards
from inhalation of their dust. Related chemical hazards can also arise in the
preparation of carbon foams or aerogels from precursors such as formaldehyde
and resorcinol.
Probably the most hazardous situation, especially with the nonaqueous solution type of electrochemical capacitor when used for hybrid vehicle load leveling, would arise in vehicle crashes or sudden shocks. Then shorting of
high-voltage systems will occur with overheating and fire hazard, together with
the probable generation of noxious gases and vapors.
Some shock tests on packaged electrochemical capacitors have been conducted at Sandia National Laboratories. The effects of vibration, elevated temperature (-30 to +75C), and high shock (-7000 g) on capacitor behavior have
Technology Development
651
been examined in testing programs, e.g., by Clark at Sandia (PES 1994). The
Pinnacle Research and the Evans types of capacitor devices were the subjects
ofthese tests, which were conducted (to some extent for military reliability purposes) on systems used for computer memory backup.
The disposal of small, carbon-type capacitors probably involves fewer environmental pollution problems than those arising with domestic alkaline or Leclanche MnOrZn primary batteries, although some electrochemical capacitor
devices may contain Ni metal in the form of powder, filaments, or X-Met as current collector. Such sources of Ni are now regarded as potentially carcinogenic,
though only after multiple exposures.
Finally, potential hazards arising in the transportation of appreciable quantities of packaged electrochemical capacitors have to be recognized. Deliveries
of such materials are thus subject to various transportation regulations in some
states or across borders. However, in the discharged condition, capacitors are
much less hazardous in quantity than primary batteries or charged secondary
ones.
The safety, environmental, and health issues involved in the fabrication
and use of electrochemical capacitors, especially for advanced electric vehicle
applications, have recently been reviewed in a useful report from the National
Renewable Energy Laboratory (NREL) (Golden, Colorado) for the U.S. Department of Energy.
652
Chapter 20
R-N!0\-R1
3
R2
A1ky1-imidazolium ion
Technology Development
653
functional groups. The tubular structures are remarkably uniform and have appreciable lengths. The morphologies of the carbon nanotubes are such that electrodes can be fabricated having properties that are almost impossible to obtain
with other carbon materials. Capacitors constructed with these materials gave
excellent frequency response and a power density of 8 kW kg-I. Such capacitors
are capable of providing high pulse power at large repetition rates. The impedance behavior was that of a porous electrode performing according to the classical transmission-line model. Remarkably, virtually all the available surface
area was accessible for charge acceptance. This is probably because the intertube pore structure (visible under the TEM) is relatively open and quite different
than that of other porous carbon materials exhibiting a distribution of macro,
meso, and micro 1 nm) pores. This material can now be prepared in commercially significant quantities (Hyperion Catalysis International Inc., Cambridge,
Massachusetts).
Molybdenum nitride has recently been proposed as an inorganic, high surface-area material for use in electrochemical capacitors as an alternative to
Ru02 or Ir0 2. It obviously has the advantage of being relatively cheap, and can
be easily prepared by chemical vapor deposition (CVD) by reaction between
MoCl 5 and NH3 between 500 and 800C, or by reactive conversion from Mo0 3
by a spray pyrolysis procedure. A procedure for its preparation has been given
by Jaggers (Ph.D. thesis, University of California, Berkeley, 1988; UMI order
No. 8916713).
Rather simple capacitor modules were fabricated by Finello (PES 1994)
for use with either a nitromethane plus KI electrolyte or an aqueous 32 wt. % sulfuric acid solution. Relative to a comparable RuOz electrode, about 40 to 50%
of the capacity of the RuOz-based capacitor was realized. Few further details
have been published, but in the work by Jaggers it was observed that the cyclic
voltammogram obtained with molybdenum nitride was not reversible (i.e., it
had a nonmirror-image form for directions of charging and discharging) in the
way Ru02 is (see Fig. 11.2 in Chapter 11). It is therefore not, in that stage of
preparation, able to substitute Ru02 or conducting polymers (Chapter 12) as a
material approaching behavior of an ideal capacitor.
It has been suggested in a conference discussion (FES 1994) that the active
material of molybdenum nitride is actually Mo0 3 arising from hydrolysis in the
acid medium. For example, Conway31 found that an Mo0 3 film (Fig. 20.11 in
Ref. 31 and Fig. 11.12 in Chapter 11) gave good mirror-image cyclic voltammograms (see Chapter 10) over ca. an 0.8-V range, such behavior being the requirement for satisfactory capacitative behavior. Later work (1998) by Conway
in collaboration with Roberson (in press) showed that the CVD films of Mo 2N
and MoN on Ti could give reversible responses in cyclic voltammetry (Fig.
20.19), with almost constant capacitance as a function of potential, but only over
654
Chapter 20
0.2
N
E
0
<E
---
0.1
>.
VI
c:
0.0
Q)
.....c:
....
....
:::J
Q)
-0.1
()
-0.2
0.0
0.1
0.2
0.3
0.4
0.5
Potential I V
FIGURE 20.19. Cyclic voltammogram for CVD preparation at 4()()OC of M02N showing behavior like that of Ru02 but over a shorter operating voltage range. [From Liu, Conway, and Roberson,
J. Electrochem Soc., 145, 1882 (1998)].
ca. 0.7 V, beyond which film decomposition took place, with some of the film
spalling off.
During 1996, work at Pinnacle Research Institute (PES 1996) was extended to inorganic conducting solids, which are analogous to Ru02 but substantially cheaper. Good capacitative performance has been demonstrated, for
example, with nitrides ofMo (M02N, MoxN), Ti (TiN and TiON in KOH), and
Fe (Fe4N). These species provide materials much cheaper (5.8 to $1.8 kg-I) than
RuOrTa205 systems ($1000 kg-I!). y-M0 2N has, however, an operating voltage range of only 0.8 V, but Fe4N in aqueous KOH can reach 1.3 V with a capacitance density of about 32 F g-I. Other advances in this field reported by
Oesten (FES 1996) have involved investigations of stoichiometry changes in
perovskites, for example, by using some ruthenates that have good electronic
conductivity (e.g., SrRu03) and exhibit capacitative behavior like that of RU02'
depending on doping, e.g., with La3+.
Russian developments in high power electrochemical capacitor technology have been slow to appear in the Western literature, but an account of some
recent developments at Econd-Sojus was given at a conference by Ivanov, Poliashov, and Radinov (PES 1993). The application is mainly to starter-assist for
Technology Development
655
656
Chapter 20
TABLE 20.8.
LONG TERM
250
600
150
(200 desired)
400
135
300
80
(" 100 desired)
200
Life (years)
10
600
1000
20%
20%
<$150
<$100
Operating environment
-30 to 65C
-40 to 85C
Recharge time
<6 hours
3 to 6 hours
Continuous discharge
in I hour (no failure)
75% (ofrated
energy capacity)
Long term
Efficiency
C/3 discharge
6 hr charge
75%
80%
Self-discharge
<15% in 48 hrs
Maintenance
no maintenance
no maintenance
Thermal loss
(for high temperature batteries)
service by qualified
personnel only
3.2 W/k Whr
15% of capacity
48 hour period
tolerant
service by qualified
personnel only
3.2 W/kWhr
15% of capacity
48 hour period
tolerant
minimized by on board
controls
minimized by on board
controls
Abuse resistance
Specified by contractor:
packaging constraints
environmental impact
safety
recyclability
reliability
overcharge/over-discharge tolerance
657
Technology Development
Broad application
Memory backup
Electric vehicle load leveling
Space
Military
Automotive subsystems
Industrial
Source: From Barnett and Wolsky (FES 1994).
has also provided a survey of the 1995 performance status of various developers' products. Some of his data are collected in Table 20.2. A brief but useful
survey of the historical aspects of capacitor development and perceived applications has been prepared by Barnett and Wolsky (FES 1994).
Table 20.10 contains a list of current uses and explorations of capacitor applications. Table 20.4 listed the energy storage expectations from carbon double-layer capacitor devices. Table 20.11 categorizes capacitor types according
to their provision of general power requirements related to typical discharge
times and corresponding types of application. Table 20.12 summarizes qualitatively the performance requirements for several applications.
A capacitor-battery (sealed lead acid, 15 Ah) test configuration for simulation of a hybrid power source for an electric vehicle was reported by Burke
(PES 1993) together with a FUDS test regime. The latter also enabled the current splits between the battery and the capacitor at 79 W kg- 1 peak power to be
evaluated.
Another potential use in future years for electrochemical capacitors having
large energy (charge) storage capacity is in the storage of electrical energy generated from solar photo voltaic systems. Each unit of a photovoltaic device is essentially a low-voltage system, which would therefore be compatible with
Capacitor category
Milliseconds or less
Pulse power
Bridge power
<5 s
Load leveling
Standby power
<2 min
Hours at low rate
Typical application
SDI
Automotive subsystems, smart
weapons
Automotive subsystems
Memory protection
658
Chapter 20
Performance requirements
Low power, low voltage
High power, high voltage
High power, high voltage, reliability
Mediumlhigh power, high voltage, reliability
Mediumlhigh power, high voltage
Mediumlhigh power, medium voltage
aqueous electrochemical capacitor units, which are themselves low-voltage devices. Also, banks of low-voltage capacitor units could be charged photovoltaically in parallel and be discharged in series (voltage multiplication principle)
often providing higher voltage availability.
As with other uses of capacitors, minimization of self-discharge rates
would be desirable (see Chapter 18), but in such solar photovoltaic applications,
overnight storage of electric energy, in the absence of sunlight, would be the
principal requirement. Under such conditions, most electrochemical capacitors
would not suffer serious self-discharge and, in fact, optimization for minimization of short-term self-discharge would be possible.
However, this application is a projection some years into the future when
solar photovoltaic energy collection becomes practically, economically, and environmentally more developed, although one substantial system has already
been demonstrated by Arco in the United States, but not yet in any hybrid configuration with capacitors being employed to store the acquired energy or
charge.
Other recent projections for electrochemical capacitors envisage them being employed as relatively large-scale units for backup in power failures at electric utility installations, and as hybrid systems for gasoline-powered vehicles to
diminish emissions during otherwise high power drains. A combination with
advanced fuel cells in a load-leveling function, as for battery-powered vehicles,
is also being investigated. In the back-up function, it is known that about 90%
of utility power failures last for less than 3 s, so although high energy, voltage,
and power are required in such applications, it is only for very short periods of
time.
Although commercial development of electrochemical capacitor technology has been rather slow, there are a number of products and activities that
should be recognized.
659
Technology Development
500, 1.8
60
175 to 350
<100
<45
>5, > 18
> II, >40
20 to 50
60 to 120
Continuous
>100,000
<1000
90%
2.3
660
Chapter 20
Voltage (V)
Current (A)
0-3
3
3-0
0
0-3
3
3-0
0
0-3
0
3-0
0
25
0
25
0
50
0
50
0
50
0
100
0
60
20
60
20
30
20
30
20
30
20
15
20
Technology Development
661
662
Chapter 20
ducted with discharge currents as high as 400 A at power levels of23 kW. Selfdischarge power was below 1 W maximum. The power performance tests were
conducted by measurements using three types of loads: resistance, current sink,
and power sink. Figure 20.13 shows a photograph of this capacitor. The 13-inch
diameter end plates are separated by 6.5 inches. Electrically isolated tension
rods are connected to intermediate cell current collectors to allow operation at
tappings of 15, 30, 60, or 120 V. By this means, the Evans prototype EC can be
configured for 15-,30-,60-, or 120-V operation using external bus connections.
This original design of capacitor has attracted much attention, especially for
high-power applications. See photo in Fig. 20.13.
A five-element (cf. Miller's calculations cited in Chapter 17) equivalent
circuit of the transmission-line type was examined in simulation of the Evans
capacitor's impedance behavior in a series-connected, two-unit 1.5-F cell stack.
Other possible commercial sources of capacitors are Alcatel Alsthom
(SAFf) in France (1995) and later developments from NEC, who are producing
an aqueous sulfuric acid carbon double-layer capacitor in a series-stacked, 18cell unit that can provide an energy density of 1.3 Wh kg-I and a usable power
density of 500 W kg-I. Asahai Glass Co. in Japan is another supplier of lowpower capacitors that is moving to produce higher power and larger units of
3300 and 4300 F employing an organic electrolyte with carbon electrodes.
10,000
...
r"'"
lat
~
PULSE
_Flywheel
I I d .
HEV PRIMARY
Slpoar
ea aCI
~
',000
C
I{pulse
power)
USABC
41
Motor/controller-1993
/--'0
....
ICE /tron;l..-.......
-Flywheel
41
Neor-term .
.......Bipolar lead ocid
~
ultrocopacltor
(primary storage)
Il.
L-I
goal
/
/
at
Monoblock nickel-cadmium-x
0
"
..1
100,
'00
10
Log [Energy density IWh kg-1]
500
FIGURE 20.20. Energy vs. power requirements for various energy storage technologies (1993).
(From Burke, FES 1993.)
Technology Development
663
These have an energy density on the order of 6-8 Wh kg- I and a power density
of 500 W kg-I. Figure 20.20 shows energy vs. power requirements for various
energy storage techniques as of 1993.
20.13. CAPACITOR-BATTERY HYBRID APPLICATION FOR ELECTRIC
VEHICLE DRIVE SYSTEMS
Capacilance (F)
5
100
500
1500
Voltage (V)
Energy (kJ)
35
10
7
4
1
1.5
8.3
664
Chapter 20
TABLE 20.16.
>3 kWlkg
<20s
>100,000 cycles
>95%
For the normal vehicle, the power and energy requirements are, of course,
usually only for SLI purposes. For the electric vehicle, competitive cost is ca.
$10 (U.S.) per kilowatt-hour for future commercial viability.
Double-layer capacitor design targets for electrically heated exhaust catalysts (EHC), as well as projected electrically powered steering (EPS) and electric vehicle power trains, were evaluated as shown in the table.
The interface electronics systems required for energy-efficient hybridizing
between electrochemical capacitors and batteries form a very important component of the overall power train for electric vehicles. Developments have been reviewed by Dorcker at INEL (cf. Burke, FES 1994), but any detailed description
or discussion of this aspect of power-train technology is outside the scope of this
monograph. It is essentially hardware electrical engineering.
A useful comparison of energy vs. power requirements (Ragone-type diagram, Chapter 15) for various energy storage systems was also given by Burke
(FES 1993) (Fig. 20.17) and a record of interface electronic development
(Solectria Corp., Wilmington, Massachusetts) was published by Burke, together
with a schematic diagram of an electronics circuit for an electric vehicle drive
control. A single point is given for the near-term (1993) electrochemical capacitor goal and this is presumably for the fully charged condition. However, as dis-
Interface
Electronics
75kW
Technology Development
665
50
(Wh)cap
Battery
Power (kW) 25
500
o
25
Capacitor
Power (kW) 0
Capac.
SOC
1.0
0.5
-.~
.-.....-...
./'_.....,..,...J......~"",.....:.,.-~I--~..
FIGURE 20.22. Effect of introduction of a load leveling electrochemical capacitor (500 Wh) in
hybrid combination with a battery for the motive power of an electric vehicle as determined in a
Federal Urban Driving Cycle. (From Burke, FES 1993, 1994.)
666
Chapter 20
at that date were tabulated by Farahmandi (FES 1993). More demanding specifications now exist.
Technology Development
667
held (e.g., the FES series) over the past 8 to 10 years or so. More attention, however, is desirable (Anderson, FES 1995) on such aspects as mass-production
techniques and "real-world" applications.
Market shares for various types of capacitors as perceived in 1996 (Zogbi,
FES 1996), are estimated to be as follows: ceramic $4.2 billion; AI(oxide electrolytic) $3.0 billion; dc film $1.5 billion; ac film $.0.9 billion and electrochemical $0.08 billion (anticipated to increase to $0.225 billion by the year 2000). The
latter remains a small market fraction, but for most envisaged applications, electrochemical capacitors do not suffer competition from the types listed here since
they have specialized applications.
20.14.2. Market Status and Future Opportunities
As has been stated earlier, research and development in the field of electrochemical capacitors has until recently (ca. 1994) been limited largely to optimization of fabrication and development of testing procedures for such
devices, with the exception of commercial production by some Japanese companies. As of 1992, the global market was estimated at 130 million units with a
market value of between 50 and $80 million.
Sales in the United States in 1992 were on the order of3 million units, representing about 4% of world (mainly in Japan) production, but a later estimate
for 1993 quotes a 40-50% increase, with projections for substantial increases
to 1996 and onward. It is expected that this monograph will increase the awareness of the opportunities for using high-rated electrochemical capacitors among
power-source and design engineers, as well as those directly in the profession
of electrochemical engineering.
Section 20.2.1 (see also Section 20.11) listed currently developing areas of
application of supercapacitors. Major market increases are being projected, with
the possibility of an overall market volume of about $1.35 billion by the year
2000. However, this projection is predicated on the assumption of continued socioeconomic support for substantial expansion of the use of electric vehicles, especially in large conurbations.
668
Chapter 20
soker Associates) in 1994 under the title "The International Technology and
Market Study of Electrochemical Capacitors" (see Ref. 34). This is a very thorough assembly, not only of information on the patent history of the development
of technology in this field, but also of specifications, performance data, and applications of Japanese patents, which are not easy to trace otherwise. They are
listed in this publication together with the North American and European literature. Market study information is also included in various forms. Based on the
patent overview,34 a figure of 78.8% has been estimated for the extent to which
Japanese patents dominated the technology development in 1994, compared
with only ca. 8.6% in the United States and less than 1% in Canada.
The early patents of Becker (1957) and of Sohio, noted in the text elsewhere, should be specially recognized here. There are also the European and Canadian patent claims of D. R. Craig 38 for the first Ru02 pseudocapacitor
technology developed in collaboration with the present author, beginning in
1975 at the University of Ottawa under contract with Continental Group, Inc.
Technology Development
669
Um
(MV'm- l )
Ebdc
Eb ac (rms)
(MV m- I)
(J cm- 3)
II
2.2
3.5
3.6
2.9
770(83)
730(33)
594(65)
470(13)
720(19)
220(83)
493(26)
289(20)
300(4)
400(20)
28.9
5.2
5.5
5.2
6.7
Source: T. R. Jow and P. J. Cygan. Conference Record. IEEE IntI. Symp. on Electrical Insulation. pp. 181-187.
Institute of Electrical and Electronics Engineers. New York (1992).
Note: PVDF = polyvinylidene fluoride, PP = polypropylene. PET = polyethylene terephthalate. PI-polyimide,
PC = polycarbonate. Values in parentheses are standard deviations.
TABLE 20.19.
Characteristics
TypeKM
TypeLM
Technology
Delivered energy (kJ)
Energy density (JIg)
PP/ME a
PVDFIMEa
J/cm 3
Energy efficiency (%)
Voltage range (kV dc)
50
0.77
1.0
95
4-44
125
1.5
3.0
65
4-44
Source: From Jow (FES 1992). based on technical data from Aerovox Inc . New Bedford. Massachusetts.
aME-metallized electrodes; PP and PVDF dielectrics; see footnote to Table 20.18.
670
Chapter 20
devices overcome one of the problems of stacking units of low voltage electrochemical capacitors to achieve high operating voltages, e.g., for electric vehicle
power systems. Dielectric constants of various polymer film formulations were
investigated and values from 3 to 16 were realized. Good performance was retained over a temperature range of -50 to +125C, with a capacitance change
of about 25%.
These somewhat new capacitor variants serve to bridge the gap between
conventional capacitors, including electrolytics, and the developing family of
electrochemical capacitors. The main and advantageous difference is the high
operating voltages that are offered by this new family of electrostatic thin-film
capacitors. Large-scale, high-power applications are envisaged using materials
of excellent structural strength. The possibility of operating voltages being as
high as 500 V offers a good energy storage opportunity on account of the quadratic dependence, E = (l/2)CV2, of energy density on voltage across the capacitor plates, a situation that is achievable with electrochemical capacitors only by
multiple unit series stacking.
Maxwell's technologies have extended to developments in high-voltage,
energy storage and pulsed discharge electrostatic capacitors and have been reviewed by Ennis, Haskell, and Sevigny.36 These high energy-density pulse capacitors are contributing to the introduction of new military, industrial and
commercial systems of original kinds. The advances are based largely on new
dielectric materials and dielectric systems, but not of the electrochemical capacitor kind.
A paper by Laghari and Sarjeant37 discussed the technology for pulsed
power and energy storage by regular dielectric capacitors rather than those of
the double-layer type, so the material covered in that paper is mainly outside the
scope of this volume. However, some ceramic materials (e.g., barium titanate)
having dielectric constants between 500 and 6000, with quoted breakdown voltages of some 50 V per mil, are of interest for the development of solid-state devices. For example, for a dielectric constant of 1000 and a thickness of 10 mil,
the achieved energy storage density at 500 V would be about 1.2 x 10-3 J cm-2
or 1.2 x 10- 1 J cm-3, but this is a rather small value compared with the doublelayer capacitor energy storage densities that have been achieved, but at much
lower voltages.
Technology Development
671
years with applications of small devices for computer memory backup and lowfrequency filtering being recognized, while substantially bigger devices have
been developed for possible larger-scale uses in the future for capacitor-battery
hybrid power systems for electric vehicle drive trains and for low-temperature
power assists for starting diesel engines. Another projected larger-scale application is for back-up power in the electrical grid system when short-period
power outages occur.
Substantial research and development of technology has been achieved by
companies such as Panasonic-Mitsushita, Maxwell, Evans, Pinnacle Research,
and others referred to earlier in this chapter. Notably, important advances have
been made in the fabrication and pretreatment of high specific-area carbon materials for double-layer capacitors, especially in preparation of carbon fibers,
carbon felts, and carbon aerogels, as well as some glassy carbon materials.
In addition to this work and that involving Ru02' important advances of a
different kind have been made using conducting polymers such as polyaniline,
polypyrrole, and especially poly thiophene derivatives as electrochemical capacitor materials, sometimes supported on high-area carbon fiber substrates, as
in the work and publications by the Los Alamos group (Gottesfeld et al.) and that
at the University of Texas at Austin (Bard et al.). Conducting oxide and nitride materials are also under examination as less expensive alternatives to Ru02.
From a fundamental point of view, as discussed in several earlier chapters,
there are opposing effects that arise between the requirements of high energy
density (scaling with high specific-surface area for double-layer capacitors) and
high power density, necessitating low internal resistance (esr). This is a central
problem with the use of high specific-area porous electrodes as discussed in
Chapters 14 and 15. It has led to much interesting work on characterization of
the impedance and pulse-response behavior of such systems and to new levels
in understanding of the relation of pore-size distribution to real specific-area
values, their connection to electrolyte accessibility (wetting) of pores, and the
specific capacity values obtained.
The remaining problems in the technology of electrochemical capacitors
are connected to the engineering of high-rate production lines for fabrication of
very well-matched units for stacking in multicell, high-voltage devices; the production of reliably sealed, low esr, bipolar, stacked devices; and the broader
general problem of the appreciable self-discharge rates that most electrochemical capacitors suffer, especially at elevated temperatures.
The principal sources of technological information on electrochemical capacitor development have appeared mostly in symposium proceedings volumes
672
Chapter 20
over the past 7 years or so, apart from a multitude of patent claims, some of
which are summarized in the next chapter. Other information has been provided
to the author in private communications.
In the foregoing pages, material has been selectively abstracted from papers presented at seminar conferences. Mainly, topics of technological interest
in the developing field are referred to, especially those organized by Florida
Educational Seminars, Inc. and designated as PES in the text. The author wishes
to thank Drs. S. P. Wolsky and N. Marincic for permission to use information
from the seminar preprint books and reproduce various diagrams. These seminars were held in Florida in the years 1991-1996 and have continued.
In this chapter, these symposia have been referred to for brevity by PES 1,
2,3,4,5,6, and 7, respectively, which took place annually (and continue) in the
period 1991-1997. For these publications, available from Florida Educational
Seminars, Inc., 2300 Glades Road, Suite 307E, Boca Raton, Florida 33431,
there are no volume or page numbers; hence only the author attributions and the
above designations are given in the references.
Papers on electrochemical capacitors have also been presented at conferences held by the Institute of Electrical and Electronics Engineers (IEEE) and
at the bi-annual Brighton (UK) Power Sources Conferences, published in proceedings volumes of those meetings (now in J. Power Sources).
One of the first IEEE power-sources conferences at which a section was
specifically devoted to electrochemical capacitors was the Thirty-fourth International Power Sources Symposium held in June 1990, and sponsored by the Industry Applications Society of the IEEE (proceedings available as catalog No.
90 CH 2863-9) and by the U.S. Army Laboratory Command. Seven papers were
presented on pulse power batteries and electrochemical capacitors; three were
devoted to capacitor electrochemistry. The present author opened the session
with a general comparative examination of batteries and capacitors in a paper
on the transition from supercapacitor to battery behavior in electrochemical energy storage. The subject matter of this contribution has been dealt with in
Chapters 2 and 10 in this volume and in Ref. 39. At this conference, Burke delivered one of the earliest papers specifically addressed to the application of socalled ultracapacitors to electric vehicle propulsion systems.
Besides the Florida Educational Seminars proceedings, other principal
symposium proceedings that contain information on electrochemical capacitors, are as follows:
Proc. 34th IntI. Power Sources Symposium, IEEE, catalog No. 90-CH2863-9.
Proc. Electrochemical Society Symposium on Electrochemical Capacitors, vol. 95-29, F. Delnick and M. Tomkiewicz, eds., The Electrochemical Society, Pennington, N.J. (1995).
Technology Development
673
REFERENCES
1. T. C. Murphy and P. B. Davis, in Proc. Electrochemical Society Symposium on
Electrochemical Capacitors, no. 95-29, p. 269, The Electrochemical Society, Pennington, N.J.
(1996); see also T. C. Murphy, O. H. Cole, and P. B. Davis, FES (1995).
2. S. Sarangapani, B. V. Tilak, and C. P. Chen, in Electrochemical Society Proceedings on
Electrochemical Capacitors, F. Delnick and M. Tomkiewicz, eds., vol. 95-29, The
Electrochemical Society, Pennington, N.J. (1995).
3. 1. D. LeMay, R. W. Hopper, L. W. Hrubesh, and R. W. Pekala, Materials Research Society
Bull., December, 19,43 (1990).
4. 1. H. Aubert and A. P. Sylwester, Chemtech, 21, 234-290 (1990).
5. C. L. Renschler and A. P. Sylwester, Materials Sci. Forum, 52-53, 301 (1989).
6. H. Hatori, Y. Yamada, and M. Shiraishi, Carbon, 30, 303 (1991).
7. C. Arnold, 1. H. Aubert, R. L. Clough, P. B. Rand, and A. P. Sylwester, U.S. Patent 4,832,881
(May 23, 1989).
8. R. L. Clough and A. P. Sylwester, U.S. Patent 4,832,870 (May 23, 1989).
9. R. R. Lagasse, P. K. Leslie, K. R. Thompson, and R. 1. Weagley, Advances in Preparing and
Characterizing Low-Density PAM-Carbon Microcellular Foam, SAND 93-0172, Sandia
National Laboratories, Albuquerque, NM (1993).
10. R. R. Lagasse, in Extended Abstracts, 21st Bienniel Con! on Carbon, vol. 158, American
Chemical Society, Washington, D.C. (1993).
11. D. Barby and Z. Haq, European Patent No. 0-060- \38 (Sept. 3, 1982).
12. C. 1. C. Edwards, D. A. Hitchen, and M. Sharples, U.S. Patent 4,775,655 (Oct. 4, 1988).
13. R. W. Pekala, U.S. Patents 4,873,218 (Oct 10, 1989) and 4,997,804 (March 5, 1991).
14. R. W. Pekala, C. T. Alviso, and 1. D. leMay, 1. Noncryst. Solids, 125,67 (1990).
15. R. W. Pekala and R. W. Hopper, 1. Mater. Sci., 22,1840 (1989).
16. R. de Levie in Advances in Electrochemistry and Electrochemical Engineering, P. Delahay,
ed., vol. 6, Interscience Publishers, New York (1967).
17. F. M. Delnick, C. D. leager, and S. Levy, Chem. Eng. Commun., 35, 231 (1985).
18. F. M. Delnick, in Proc. Symp. on the Science of Advanced Batteries, No. 8-9, D. Scherson,
ed., Case Western Univ. for The Electrochemical Society, Pennington, N.l.(l995).
19. N. Nawa, T. Nogami, and H. Mikawa, 1. Electrochem. Soc., 113 (6), 1457 (1984).
20. D. Kohler,1. Zabasajja, and B. Tartarchuk, 1. Electrochem. Soc., 137 (I), 136 (1990).
21. D. Kohler, 1. Zabasajja, F. Rose, and B. Tartarchuk, 1. Electrochem. Soc., 137 (6), 1750
(1990).
22. S. T. Mayer, 1. Kaschmitter and R. W. Pekala, Abstract No. 500, in Extended Abstracts of the
186th Electrochemical Society Meeting, vol. 94-2, The Electrochemical Society, Pennington,
N.J. (1994).
674
Chapter 20
23.
24.
25.
26.
27.
28.
T. P. Hoar and J. N. Agar, Disc. Faraday Soc., 1, 291 and 297, (1947).
J. Newman, Electrochemical Systems, Prentice-Hall, Engelwood Cliffs, N.J. (1991).
S. Sekido, Y. Yoshina, T. Muranaka, and M. Mori, Denki Kagaku, 48, 40 (1980).
N. Marincic, private communication (1996).
W. Pell, T. C. Liu, and B. E. Conway, Electrochim. Acta, 42, 3541 (1997).
1. E. Oxley, in Extended Abstracts, J4Ist Electrochemical Society Meeting, Houston, Texas
(1972).
S. Sarangapani, P. Lessner, J. Forchione, A. Griffith, and A .. B. LaConti, in Proc. 25th
Intersociety Energy Conversion Conference, vo!. 3, p. 137; see also S. Sarangapani, 1. A.
Kosek, and A. B. La Conti, in Handbook of Solid-State Batteries and Capacitors, M. Z. A.
Munshi, ed., p. 601, World Scientific Pub!. Co., Singapore (1995).
1. Robinson and R. A. Osteryoung, J. Am. Chem. Soc., 101, 323 (1979); see also J.
Electrochem. Soc., 130, 1968 (1983).
B. E. Conway, in Proc. of the Symposium on Electrochemical Capacitors, F. M. Delnick and
M. Tomkiewicz, eds., vo!. 95-29, p.15, The Electrochem. Society, Pennington, N.J. (1995).
R. W. Pekala, in Ultrastructure Processing of Advanced Materials, D. R. Uhlmann and D. R.
Ulrich, eds., pp. 711-717, Wiley, New York (1992).
1. R. Miller, in Proc. 11 th IntI. Seminar on Primary and Secondary Battery Technologies and
Applications, S. P. Wolsky and N. Marincic, eds.; Horida Educational Seminars, Boca Raton,
Ha. (1994).
S. P. Wolsky and R. S. Wissoker, The International Technology and Market Study of
Electrochemical Capacitors, Horida Educational Seminars, Inc., Boca Raton, Ha. (1994).
T. R. Jow, P. J. Cygan, J. Carter, S. Guilman, and L. Krzastek, in Extended Abstracts, I83rd
Electrochemical Society Meeting, The Electrochemical Society, Pennington, N.J. (1993).
1. B. Ennis, D. K. Haskell, and 1. A. Sevigny, in Proc. 34th Inti. Power Sources Symposium,
p. 392, C. Berger, ed., Inst. Electrical and Electronic Engineers, Piscataway, N.J. (1990).
J. R. Laghati and W. J. Sarjeant, in Proc. 34th Inti. Power Sources Symposium, p. 380, C.
Berger, ed., Inst. Electrical and Electronic Engineer, Piscataway, N.J. (1990).
D. R. Craig, Canadian patent 1,196,683, (1985) and Electrochemical Energy Storage Devices,
European Patent Application, 82,\09,061,0 submitted September 30, 1982, pub!, no.
0,078,404, May 11th (1983).
B. E. Conway, in Proc. 34th IntI. Power Sources Symposium, p. 319, C. Berger, ed., Industry
Applications Soc. ofinst. of Electrical and Electronics Engineers, Piscataway, N.J. (1990).
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
Chapter 21
Patent Survey
5,442,197 (8/15/95)
675
676
Chapter 21
5,430,606 (7/4/95)
3.
5,429,893 (7/4/95)
4.
5,420,168 (5/30/95)
5,260,855 (1119/93)
5.
5,392,191 (2121195)
6.
5,388,025 (217/95)
7.
5,383,089 (1117195)
(continued)
6n
Patent Survey
5,369,546 (11129/94)
9.
5,351,164 (9/27/94)
10.
5,319,518 (617194)
11.
12.
5,245,512 (9/14/93)
5,212,622 (5/18/93)
13.
5,172,307 (12115192)
14.
5,168,433 (1211192)
15.
5,161,094 (1113/92)
5,136,478 (8/4/92)
5,047,899 (9110/91)
16.
5,428,501 (6/27/95)
5,151,848 (9129/92)
17.
5,150,283 (9122192)
NH1
18.
5,144,537 (9/1/92)
19.
5,136,474 (8/4/92)
678
Chapter 21
TABLE 21.1. Continued
5,136,473 (8/4/92)
21.
5,136,472 (8/4/92)
5,115,378 (5119/92)
22.
5,079,674 (1/7/92)
23.
4,896,249 (1123/90)
4,876,631 (1124/89)
24.
4,862,328 (8129/89)
25.
4,766,522 (8123/88)
26.
4,737,889 (4112188)
4,597,028 (6124/86)
4,562,571 (12131185)
27.
4,730,239 (3/8/88)
28.
4,633,372 (12130/86)
(12116186)
29.
4,609,972 (912186)
Patent Survey
679
4,605,989 (8/12/86)
31.
4,438,481 (3/20/84)
32.
4,414,607 (11/8/83)
33.
4,337,501 (6/29/82)
34.
4,327,400 (4/27/82)
35.
2,800,616 (7/23/57)
5,429,893 (7/4/95)
2.
5,426,561 (6/20/95)
3.
5,418,682 (5/23/95)
Electrolyte
Propylene carbonate + dimethoxyethanol with LiAsF6 ,
(CH 3)4 NCI0 4, LiCI0 4 (see also Table 21.1).
Capacitor with electrode layers formed of a mixture of
carbon and an electrolyte binder in nonaqueous
solvent-salt system where the solvent is taken from a
group consisting of propylene carbonate-diethyl
carbonate, 1,3-dioxolane, and tetrahydrofuran; and the
salt consisting of tetrabutyl ammonium tetraphenyl
borate, tetraethylphosphonium tetrafluoroborate and
tetraethylammonium tetrafluoborate. The electrode
layers were formed on metallized polymer film, the
metal forming the current collector. Energy densities of
>10 kWhlkg and power densities of>1 kWlkg at a
volumetric energy densities of> 15 WhIliter were
reported.
Solvents: acetonitrile, succinonitrile, glutaronitrile,
propylene carbonate, ethylene carbonate. Salts: alkali
metal-tetraalkylammonium cations with CF3 SO;,
N(CF3S0 z 2, BF:! , PF6' , AsF6' and CIO:!. Accessible
potential range of -4 V achieved with energy densities of
10 WhIliter.
(continued)
680
Chapter 21
TABLE 21.2.
5,303,118 (4112194)
5.
5,121,301 (6/9/92)
6.
5,038,249 (8/6/91)
7.
4,748,542 (5/31/88)
8.
4,725,927 (2116/88)
9.
4,725,926 (2116/88)
10.
4,713,731 (12115/87)
11.
4,622,611 (11111186)
12.
4,363,079 (1217/82)
13.
3,652,902 (3128172)
Continued
Electrolyte
Patent Survey
681
TABLE 21.3. Authors, Date, and Assignee for Patents in Tables 21.1 and 21.2
Authors
Assignee
ELECTRODE MATERIALS
5,442,197 (8/15/95)
5,430,606 (7/4/95)
5,429,893 (7/4/95)
5,420,168 (5/30/95)
S. T. Macer, J. L. Kaschmitter,
and R. W. Pekala
5,260,855 (1119/93)
1. L. Kaschmitter, S. T. Macer,
and R. W. Pekala
G. Thomas, S. Prasad, and J. E.
Nerz
1. L. Davis, 1. K. Arledge, T. 1.
Swirbel, and M. Williams
M. Williams and 1. L. Davis
T. Saito, Y. Shimizu, and 1.
Tabuchi
I. I. Grigortchak, K. D. Tovstjuk,
Z. D. Kovalyuk, I. D. Kozmik,
and B. P. Bahmatyuk
R. A. Blood
R. S. Rounds
5,392,191 (2121195)
5,388,025 (2127/95)
5,383,089 (1117/95)
5,369,546 (11129/94)
5,351,164 (9/27/94)
5,319,518 (6/7/94)
5,245,512 (9/14/93)
5,212,622 (5/18/93)
5,172,307 (12115/92)
5,168,433 (1211192)
5,161,094 (11/3/92)
5,136,478 (8/4/92)
5,047,899 (9/10/91)
5,428,501 (6/27/95)
5,151,848 (9/29/92)
5,150,283 (9/22192)
5,144,537 (9/1/92)
5,136,474 (8/4/92)
5,136,473 (8/4/92)
5,136,472 (8/4/92)
D. R. MacFarlane, 1. R. Tetaz,
and M. Forsyth
1. Tabuchi, T. Saito, A. Chi, and
Y. Shimizu
H. Mukouyama, M. Mmaesaka,
K. Watanabe, and M. End
J. F. Bruder and L. N. Swink
(continued)
682
Chapter 21
TABLE 21.3.
U.S. Patent No. (Date)
5,115,378 (5/19/92)
5,079,674 (1/7/92)
Continued
Authors
Y. Tsuchiya, K. Kurabayashi, S.
Kito, and Y. Niida
F. P. Malaspina
4,542,444 (9117/92)
A. Shirata, K. Kurabayashi, Y.
Tsuchiya, and Y. Niida
K. Kurabayashi, Y. Tsuchiya, A.
Yoshida, and H. Koizurni
R. C. Boland
3,648,126 (317m)
3,634,736 (1111172)
3,536,093 (10127170)
4,896,249 (1123/90)
4,876,631 (1124/89)
4,862,328 (8/29/89)
Ibid.
T. Morimoto, Y. Sanada, S.
Oshahi, and Y. Kimura
1. McHardy, F. A. Ludwig, L. R.
Higley, A. Kindler, and C. W.
Townsend
A. Nishimo, I. Tamahashi, and A.
Yoshida
A. Yoshida, A. Nishino, I.
Tanahashi, and Y. Takeuchi
P. Deman and J. Potage
J. C. Currie, L. F. DiFranco, and
P. D. Bennett
J. L. Calahan, E. A. Cuellar, M. 1.
Desmond, and J. C. Currie
E. A. Cuellar and M. J. Desmond
M. Edeling and U. Gebhardt
5,077,634 (12/31191)
5,072,337 (12/10/91)
4,766,522 (8/23/88)
4,737,889 (4112/88)
4,597,028 (6/24/86)
4,562,571 (12/31185)
4,730,239 (3/8/88)
4,633,372 (12/30/86)
4,630,176 (12/16/86)
4,609,972 (8/2/86)
4,605,989 (8/12/86)
4,438,481 (3/20/84)
4,327,400 (4/27/82)
2,800,616 (7/23/57)
4,414,607 (1118/83)
4,337,501 (6/29/82)
Assignee
Ibid.
Motorola, Inc., Schaumburg, Ill.,
U.S.
Isuzu Motors, Ltd., Tokyo, Japan
Ibid.
Standard Oil Company,
Cleveland, Ohio, U.S.
Ibid.
Ibid.
Ibid.
Murata Manufacturing Co., Ltd.,
Japan
Ibid.
Asahi Glass Co., Ltd., and Elna
Company Ltd., Fujisawa, Japan
Hughes Aircraft Co., Los
Angeles, Calif., U.S.
Matsushita Electric Industrial
Co., Ltd., Kadama, Japan
Ibid.
Thomson CSF, Paris, France
Stemcor Corp., Cleveland, Ohio,
U.S.
Standard Oil Company,
Cleveland, Ohio, U.S.
Ibid.
Siemens Aktiengesellschaft,
Munich, Germany
Standard Oil Company,
Cleveland, Ohio, U.S.
United Chemi-Con, Inc.,
Rosemont, Ill., U.S.
Matsushita Electric Industrial
Co., Ltd., Kadama, Japan
Siemens AktiengerseIIschaft,
Berlin and Munich, Germany
Matsushita Electric Industrial
Co., Ltd., Kadama, Japan
Genera! Electric Co. Scenectady,
N.Y., U.S.
ELECTROLYTES
5,429,893 (7/4/95)
G. Thomas
5,426,561 (6/20/95)
(contmued)
683
Patent Survey
Authors
4,713,731 (12115/87)
1. S. Lundsguard
T. Morimoto, K. Hiratsuka, Y.
Sanada, and H. Aruga
T. Morimoto, K. Hiratsuka, Y.
Sanada, and H. Ariga
D. L. Boos and A. V. Marse
4,622,611 (11111/86)
4,363,079 (1217/82)
3,652,902 (3128n2)
4,725,926 (2116/88)
Assignee
Rockwell International Corp.,
Seal Beach, Calif., U.S.
NEC Corp., Tokyo, Japan
Isuzu Motors, Ltd., Tokyo, Japan
Colgate-Palmolive Co.,
Piscataway, N.J., U.S.
Mead Corp., Dayton, Ohio, U.S.
Asahi Glass Company and Elna
Co., Ltd., Fujisawa, Japan
Asahi Glass Co., and Elna Co.,
Ltd., Fujisawa, Japan
Standard Oil Co., Cleveland,
Ohio, U.S.
Ibid.
Matsushita Electric Industrial
Co., Ltd., Kadama, Japan
IBM Corp., Armonk, N.Y., U.S.
Index
p,43
Backward and forward currents, 44
Barrier symmetry factor, 43
Batteries
charge for, 445
comparison with
capacitors, 11,29, 30, 31
supercapacitors, II
types of, 15
Battery requirements for vehicles, 471
Bipolar electrode
systems, 622
leakage at, 622
Blocking capacitance, 522
Bridges, A.C., 502
Bnjmsted relation. and charge transfer. 43
Brush electrode
model, 392
transmission line behavior, 392
impedance of. 392
Butler Volmer equation, 45
linearization, 45
near equilibrium, 45
diagrammatic, 47
C-rate
calculations of, 446
definitions of, 443, 444
in capacitor discharge, 444
Capacitance
behavior
of conducting polymers, 299
at Mercury, principal types of, 148
blocking, 522
charging curves, 54
components, reciprocal relation for, 130
685
686
Capacitance (cont.)
of chlorinated pyrolytic graphite, 20 I
and coverage, relations to electrode potential,
illustrated,228
dependence on electrolyte concentration,
470
diagrams, for double-layer capacitance, 132
and dielectric constant, 88
diffuse layer, 112
of diffuse-layer, calculated, 129
dispersion of, in pores, 387
of double-layers, 12, 105
hump, 147
hump at Mercury, 145
ideally polarizable, 121
measurement techniques, 502
overall, theoretical, 131
relation to sweep-rate, 25
units of, 88
Capacitative behavior,
instrumental requirements, 59
response functions. 53
Capacitative properties. of ruthenium oxide films,
265
Capacitor
advantages for EV drive lines, 664
applications, exploratory. 657
vs battery
charging curves, 20
energy storage, 19
and battery charging, illustrated, 23
battery hybrid for electric vehicles, 663
characteristics. of Evans types, 639
constant power discharge of, 459
discharge. mto load resistance, 538
electrode interphases, electrostatics of, 67
electrodes
of ruthenium oxide, A.C. and D.C. responses,
285,286
for small scale testing, 598
of molybdenum nitride, 653
energy storage, 3
energy stored in, 3, 83
entropy in, 463
"Hybrid," 466
interfaces, double-layer at, 105
load leveling in drive cycle, 665
model circuit of, 455
performance,
temperature dependence of, 644
at various temperatures, illustrated, 644,
645
power density rating of, 436
Ragone plots for, 429
Ragone plots for, calculated 439
requirements, for electric vehicles, 472
stacking, 620
stacks, 465
Index
Capacitor (cont.)
systems
current distribution in, 623
thermal management of, 643
technology, state-of-the-art, 627
temperature dependence, factors in, 645
test cell, plane form, 61
test modes for, 456
Capacitor-battery hybrid systems, 448
Capacitor-battery hybridization, calculations of,
451,452
Capacitors
carbon fiber technology for, 629
characterization test, exemplified, 660
classes of, 610
commercial development and testing, 658
impedance plots for, 530
comparison with batteries, 11, 29, 30, 31
developing uses of, 610
development of technology of, 609, 610
electrochemical, defined, 7
electrode materials for, 614
historical aspects, 3
illustrated, 4, 5, 640
impedance behavior of, 479
matenals for, 17
materials requirements for, 613
non-ideally polarizable, 28
nonaqueous electrode type, 169
based on porous electrodes, 376
power
and energy-density relationships, 433
requirements and applications, 657
practical design aspects of, 620
pre-treatment of carbons for, 203
preparation, practical aspects of. 620
properties of compared, 29
based on pseudocapacitance, 221
of ruthenium oxide, formation of, 270
ruthenium oxide materials for, 634
selected performance requirements, 658
self-discharge of, 557
mechanisms at, 559
sol id state type, 635
types of, IS
usage bases of, 655
using ruthenium oxide, 259
Capacity, of a capacitor, 74
Capattery, Evans, illustrated, 640
Carbon
for capacitors, 6
chemistry, historical aspects of, 183
double-layer at, 183
general reading references, 220
intercalation
behavior, tabulated, 218
effects at, 217
monograph on, by Kinoshita, 186, 219
Index
Carbon (cont.)
oxidation of, 196
porous, lOS
double-layer at, 105
surface functionalities, 183
surface-area modification of, by heat treatment,
204
Carbon fiber
electrodes, activated, cyclic voltammograms for,
633
technology, for capacitors. 629
Carbon fibers
activated treatment of, SEM photos. 630, 631,
632
heat and gas-phase treatment of. 207
SEM photos. 630
XPS spectra of. 192
Carbon foam. for capacitors. SEM photos. 616
Carbon matenal requirements. for electrochemical
capacitance. 208
Carbon materials
activation procedures for. 184.615
for electrochemical capacitors. 185. 206. 208
pretreatment of. by heat and chemical methods,
203
semiconducting properties. 195
surface functionalities. 186
illustrated, 189
surface properties. 186
typical values for double-layer capacitance of.
206
voltammograms for. 191
Carbon surfaces
ESR of, 209
interaction with oxygen, 212
work function and surface potentials of. 213
Carbonaceous materials, modifications of, 203
Carbons
ESR spectra of, 192
heats of immersion of and pH. 216
Cell, all glass for electrode kinetic experiments,
60
Cell design, 17
Chapman theory, of double-layer. III
Characteristic frequency, 494
and time constant, 496
Charge
discharge, rate effects on, 649
self energy of, 76
Charge capacity values, for batteries, 445
Charge density
in double-layer, 114
numerical, liS
Charge-storage, densities of, 18
Charge-transfer. partial, 14
Charges in three dimensions, 75
Charging and discharging of ruthenium oxide.
mechanism of, 276
687
Charging curves
for batteries and capacitors, 19
and capacitance, 54
comparison with cyclic voltammetry, 22
illustrated,21
for Li batteries, 21
Charging equations, 379
Charging
of capacitor, Gibbs, energy, 84
entropy of, 85
modes, constant current, constant potential. 648
of non-ideally polarizable electrode. 573
rates, distribution of, in pores, 388
of RC networks, 379
Chemisorption
aspects, of anion adsorption, 135
of water, and orientation of. 149
Chemistry of conducting polymers. 304
Circle in Nyquist plot. origin of. 493
Circuit
for experimental measurements. 62
model for frequency responses, of five-element
networks. 546
Circuits. equivalent, 17
Cluster models. for water adsorption, 143
Cole and Cole. dielectrics classic paper. 491
Commercial capacitor, self-discharge behavior of.
584
Complex equivalent circuit, impedance plot for. 536
Complex molecules, dipole moments of~ 82
Complex plane plot. 484
admittance, 495
for Faradaic pseudocapacitance, 552
impedance, 484. 491
involving diffusion, 508
origin of semi-circular form, 491
for pseudocapacitance, with leakage current, 519
for redox pseudocapacitance. 550, 551
top point on, 494
Complex quantities, mathematics of, by Gauss,
491
Components of equivalent circuits, 18,489, 525
Computer interfacing. 62
Concentration polarization
onset of. 441
power limitation by, 440
Condensed phase, dielectric polarization in, 98
Condensers, illustrated, 4,5
Conditions for underpotential deposition, 225
Conductance, summary of determining factors,
347
Conducting polymer
capacitors, favorable costs of, 300
chains, ion association with, 303
charging mechanisms. related to semi-conductor
doping. 300
electrodes, 618
film propagation models. 329
Index
688
Conducting polymer (cont.)
films
conductivities of, 317
electronic spectra of, 317
ellipsometric studies of growth, 327, 330
grown on fibrous carbon paper, 323
Hall effect at, 325
ionization, contrasted with polyelectrolytes, 301
processes, schematic diagrams of, 308, 309,
311,312
Conducting polymers, 299
capacitance behavior of, 299
conductivities of,
cyclic voltammograms of, 314
historical aspects of,
ionization at, 303
Lewis acid, base oxidation reduction processes,
300
mechanisms of oxidation and reduction, 300
pseudocapacitance, 312
redox processes of, 304
Conductivities
of quaternary and lithium salts, 373
tabulated, of various ion-transport systems, 349
Conductivity, of electrolytes, 335
Constant potential charge or discharge, 605
Constant power
charge or discharge, 605
discharge
for capacitor, 459
example plotted, 649
test, 457, 458
Correlation factor in dielectrics, 100
Coulomb's Law, 68
formula, 71
Coverage dependence, of adsorption
pseudocapacitance, 226, 227
Coverage potential relations, 50
Craig, project on capacitors, 6
Critical frequency of RC circuit, 381
Crystal ionic radii, 113
Current and rate, 41
Current distribution
in capacitor devices, 623
in fuel-cell electrode, 398
in pores, 396
Current, voltage phase relations, 487, 488
Current near equilibrium, problematic, 47
Current-response transient, 57
Cyclic voltammetry, 12
comparison with charging curves, 22
diagrammatic, 58
method,513
relation to alternating voltage measurements, 515
response current behavior, 513
at ruthenium oxide, compared with lead chloride, 289
testing of capacitors, 602
Cyclic voltammogram
for molybdenum nitride, 654
of poly thiophene derivative, 318
Cyclic voltammograms
for activated carbon fiber electrodes, 633
for carbon materials, 191
for conducting polymers, 314, 315
de Levie, 392
model,382
Debyc length, 52
width of diffuse layer, 404
Debye-Langevin, function for average moment,
99
Decomposition voltage
limits, for various solutions, 352
windows, for electrolytes, 350
Densities, of charge storage, 18
Design aspects, of capacitors, practical, 620
Design, of capacitor cells, 17
Dielectric
correlation factor in, 100
of double-layer, solvent, 136
in a tield, 77
polarization in, 91
Dielectric behavior, in double-layer, 102
Dielectric constant, 68
of associated solvents, 96
and capacitance, 88
details of, 70
in double-layer, 102
interphase, 138
and ion pairing, 344
and polarization, 97
and relative pennittivity, 89
and stored energy, 90
tabulated, 346
Dielectric polarization
in condensed phase, 98
diagrammatic, 93
of interacting dipoles, 100
of molecules, 81
of non interacting orientable dipoles, 99
theories of, 87
Dielectric properties, of molecules, 81
Dielectric saturation, 103
in double-layer, 140
Macdonald theory of, 140
Dielectric theory, Kirkwood, 100
Dielectrics
in capacitors, 87
electrostatic theory of, 92
Diffuse double-layer
overlap of, 163
treatment of, 127
Diffuse layer
capacitance of, calculated, 129
half-width of, 404
Index
Diffuse layer (COlli.)
potential. 52
profile across. 160
theoretical treatment oC 127
Diffusion control
in impedance measurement. 506
in kinetics. 48
onset of. diagrammatic. 49
in self~discharge. 56'!
Diffusion efTect
onset oC 441
on power limitation. 440
in redox pseudocapacitance. 250
Diffusion impedance. Warburg. 385
Diffusion layer
diagrammatic. 49
illustrated,442
Nernst.507
Diffusion
and complex plane plot, 508
cylindrical, 389
cylindrical vs. plane, 389
phase-angle relation, 388
at plane electrode, 389
in Randles circuit, 508
Dipole
arrays. of water in double-layer. 150
in field, 95
Dipole moment
average, 99
of complex molecules, 82
induced. 95
Dipole orientation
in double-layer, Bockris, Devanathan and
Mtiller treatment of, 141
molecular treatment of. 99, 141
Dissipated energy
vs. delivered energy, 455
and power, 457
Dissociated ion. mobility of. 343
Distribution
of oxidation states of ruthenium oxide films, 279
of pore sizes. 407
DOE
689
Double-layer (COlli.)
capacitance of (cant. )
for edge oriented graphite. 202
jellium model, 157
at mercury, 132
metal electron contribution, 156
and pseudocapacitance, distinction between,
225
capacitor
self-discharge of, 574
Sohio, 185
capacitors
energy storage expectations. 617
in nonaqueous media. 170
capacity in various solvents, compared, 178,
179
Chapman theory of, II I
charge density, 114
components of potential across, 130
development at conducting polymer chain, 303
dielectric behavior in, 102
diffuse component of, 127
dipole orientation in, 141
theory of, 141
effects.
historical, 52
in kinetics, 51
in electrochemical capacitors, 106
electron
density profile in, 157
field in, 119
equivalent circuit of, 123
frequency response, 525
general reading references, 168
Gouy
model,109
theory, III
Grahame review on, 168
Gurney cosphere overlap, 173
Helmholtz model, 109
hydrogen bonded lattice models, 148
ion charge density in, I 16
modeling
calculations, recent. 152
at electrodes, 125
models of structures ot: 108
in nonaqueous electrolytes. 169
orientation of water In, chemisorption. 149
polarizatIOn of solvent in, 81
in porous electrodes, 161. 402
Stern model, 109
structure of. 105
theoretical treatment oC 125
treatments of, hierarchy of, 126
types of solvent in, 137
in various nonaqueous solutions, compared, 175
water adsorption, cluster models, 143
water polarization in, 139
690
Double-layers
in electrochemical capacitor, 107
and electrostatic treatments, 67
overlap. in porous electrodes, 403
Double-pulse procedure, Gerischer, 56
Driving cycles, 417
Electric
field, in double-layer I 19
polarization
diagrammatic, 78
molecular, 78
references on, 86
tension, 77
equation for, 77
vehicle drive line, schematic, 664
vehicles, capacitor battery hybrid power systems, 663
Electrical
energy
energy density for, 417
power density for, 417
storage, modes of, II, 12, 15, 16
response functions, 53
units, 68
Electro-osmotic
effects in porous electrode charging
in pores, 414
Electrocatalysis and work function, 39
Electrochemical
Bnjmstcd relation, 43
capacitance, based on pseudocapacitance, 221
capacitor
goals (1995) by DOE, 659
potential profile across. 107
double-layer of, 161
practical Impedance spectra of, 538
capacitors
carbon materials for, I R5
carbon materials requirements for. 208
detined, 7
development history of, 612
double-layer in, 106
electrolytes for, 351
evaluation of, 597
impedance behavior of, 479
market share, 666
power limitation at, 382, 435
status of(1995), 613
measurements, for two-electrode devices, 63
rate equation, 43
Electrochemical Society Symposium, as information source, 672
Electrochemically oxidized carbon, XPS spectra
of, 192
Electrode
arrangements, bipolar, 622
interface, model of, 119
Index
Electrode (cont.)
kinetic experiments, cell for, 60
kinetics
and double-layer, 51
general references, 64
materials
characteristics of, for capacitors, 614
patent survey. 675
potential
and activation energy, 42
and energy factors, 37
processes
energetics of, 34
kinetics ot~ 33
under diffusion control, 48
reference, 59
Electrodes
bipolar, 622
conducting polymer type, 618
porous, 376
de Levie's treatment, 382
Electrolyte
conductivity and available area in
supercapacitor electrodes, 360
low temperature molten, 342
conductivities of, 342
solid polymer type, 636
solutions
decomposition voltage windows, 350
factors determining properties of, 354
general reading references, 375
selected graphically represented data, 367372
tabulations of data, 366
systems in technology development, 618
Electrolytes
conductance and ion pairing, 337, 33R
conductivity of, 335
for electrochemical capacitors, 351
Ion
paIring in. 335
solvation tactor in capacitor charging
molten, 350
nonaqueous, 347
for pseudocapacitance
solvation of, 335
special new anions, 341
weak and strong. 340
Electrolytic capacitors, energy-density of, 464
Electron
densities, per atom, 18
density
emergent at surface, I 18
lellium model, 117
profile, at electrode surface, 157
historical,3
name, 2
spin resonance at carbon surfaces, 209
Index
Electron (COllI.)
transfer
and energy barrier. 43
energy conditions for. 38
processes following. 50
Electronic
instruments. 61
polarization. 78
transfer. 35
work function. 37
Electrons. involved in matter. 34
Electrosorption
and pseudocapacitance. 224
valence. 14. 253
Electrostatic
capacitors. high voltage. 668
field. 71
force. 71
potential. 71
theory. of dielectrics, 92
Electrostatics
of electrode interphases. 67
elements of, 67
Ellipsometry, applied to polyaniline films, 324,
330
Energetics of electrode processes, 34
Energy
barrier diagram for charge transfer, 43
of a charge
conditions, for electron transfer. 38
densities for batteries and capacitors, 19
density, 417
of electrolytic capacitors, 464
optimization of, 448
factors, and electrode potential, 37
storage
per atom, 19
by capacitors, II. 16. 83
by electrostatic capacitors, 668
expectations for double-layer capacitors
(1994),617
by flywheel systems, 474
technologies compared, 662
vs. power requirements, for various technologies,662
Energy-power relations in hybrid systems, 626
Entropies of hydration of ions, tabulated. 364
Entropy
of charged capacitor, 463
of charging, 85
Equations
for kinetics, 4 I
for porous electrode, 383
Equilibrium. currents near, 45, 46
Equivalent circuit
components, 18
of double-layer, 123
elements, 481
691
Equivalent circuit (cont.)
hardware model of, 576
with inductance, 542
for pore, illustrated, 390
of a pore containing electrolyte, 381
for pseudocapacitance, 514
Equivalent circuits, 17
and frequency responses, 554
hierarchy of, 526
selected, impedance of, 532
two stage, 535
Equivalent series resistance, ESR in impedance,
528
in impedance spectroscopy, 528
and impedance plot for capacitor, 545
power limitation by, 382, 426, 435
values and Ragone plots, 471
ESR and porous electrode impedance, 545
ESR
basis of technique, 210
of free radicals at carbon surfaces, 209
spectra of carbons, 192
EV power systems. DOE options for, 665
Evans
Capattery, illustrated, 640
"Hybrid," 466
type capacitors, 466
Exchange current and heat of adsorption, 40
Exchange current-density
and work function, 39
diagrammatic, 47
Exchange currents
graphical representation, 46
volcano relation, 40
Experimental measuring circuit, 62
Faradaic
impedance, 483
processes, 13, 14
and self-discharge, 562
pseudocapacitance, complex plane plot for, 552
reaction resistance, 46
resistance, 483
determination ot: 46
In impedance, 46
Faraday
Lecture. 2
scientific biography, 64
Fermi level. 35
Fick's equations, for diffusion, 571, 573
Field
dipole in, 95
high, dielectric saturation in, 103
influence, on a dipole, 79
due to surface of charges, 74
Fields, due to charge distribution, 72
Florida Educational Seminars, as information
sources, 672
692
Flywheel systems
energy storage by. 474
Ragone diagrams for, 475
Forward and backward currents. 44
Fractal
growth. of conducting polymers.
surfaces
illustrated. 405
in porous electrodes. 404
Frequency
dependence
of equivalent-circuit elements. 527
of RCL network. 536
response
analyzers. digital. 505
of double-layer capacitance. 525
equations, for pseudocapacitance, 520
of RC networks. 379
simulated by five-element circuit, 548
selt~resonance, 543
spectrum. of phase-angle values, 393
FUDS
cycle, 417
and energy delivery, 473
and EV performance, 473
Fuel-cell electrode
current distribution in, 398
overpotential for, 431
Fullerenes, 184
Future market status, 667
Galvani,2
Gauss's relation, 74
Gerischer procedure, in double-layer charging, 56
Gouy theory of double-layer. I I I
Grahame results, for double-layer capacitance, 132
Grahame. seminal review on double-layer, 124
Graphical representation. of exchange current. 46
Graphite
crystals. properties of, 212
edge and basal plane properties. 199
mtercalation compounds, 193
structure of. 188
Gurney cosphere overlap at electrode. 173
Half-cycle charge and discharge of polymer electrode, schematic, 319, 320, 321
Health and safety hazards, 649
Heat
of adsorption. and exchange currents, 40
treatments, of carbon. 207
Heats of immersion of charcoals. and oxygen content, 214
Helmholtz layer
capacitance of, 112
capacitance and ion adsorption, 133
Parsons model, 146
results of Parsons' treatment, 147
Index
Helmholtz layer (cont.)
for various conditions, 145
at various metals, 153
Heyrovsky, Nobel prize. 513
Hierarchy of equivalent circuits, 526
High area carbons, pseudocapacitance at. 255
Historical
development, of electrochemical capacitor, 612
overview, I
references, 9
Hybrid
battery-capacitor systems, energy delivery, 473,
474
systems
of battery capacitor modules, 448
energy-power relations in, 626
type capacitor, Evans, illustrated, 638
"Hybrid" capacitors. impedance plot for. 467
Hydration energies, of ions tabulated, 364
Hydrogen
bond, lattice lattice models of double-layer. 148
bonding to metal, 41
evolution reaction. impedance of, 510
Ideally polarizable electrode, 121
Imidazole type salts. use of, 652
Imidazolium ion structure, 652
Imide ions, 341
Impedance
behavior
of accessible pores. 41 I, 412
of pseudocapacitance, 549
of various circuits, 525
of brush electrode, 392
complex plane plots, 484
components, relations to frequency compared. 489
diagram
complex plane, for porous electrode. 381
for porous electrode, 402
of hydrogen evolution reaction, 510
Faradaic. 483
kinetic interpretation of, 506
measurement
Lissajous figures
for testmg of capacitors. 602
of selected cirCUIts. 532
plot for "Hybrid" capacitor. 467
plots, for commercial capacitors. 530
of a pseudocapaCltance. 518
spectra, of practical electrochelTIlcal capacitor.
538
spectroscopy
and ESR, 528
references for. 524
by Solartron. 482
Induced dipole moment, 95
Induction. of polarization. 94
IN EEL. testing procedures by. 647
Index
Information sources, 671
[nstrumental requirements, for capacitor characterization,59
Instruments, electronic, 61
[ntercalation
behavior at carbons, tabulated, 2 [8
at carbon, 217
compounds, of graphite, 193
electrodes, 25
Interionic distance, in double-layer, 116
Interphasial, related to interfacial, 108
Ion
adsorption
and compact layer capacitance, 133
in double-layer, Stem treatment of, 133
charge density, in double-[ayer, 116
pairing
and dielectric constant, 344
in electrolytes, 335
models for, 337
and solvent donicity, 344
potential and field at, 72
Ionic radii, I 13
Ions, electrostatic treatments, 67
Iridium oxide, voltammograms for, 263
Irreversibility, of pseudo capacitance, onset of 245
Isotherm
equation for pseudocapacitance, 225
for Lithium sorption, 27
Jellium model
charge distribution, 15R, 159
of double-layer, 117
Kinetic
analYSIS method
equatIOns for, 509
for impedance, 509
interpretation, of impedance behavIOr, 506
parameters, for power source evaluation, 43 I
theory, of pseudocapacitance, 236
Kinetics
of electrode processes, 33, 41
and work function, 37
Kinoshita, monograph on carbon, 186,219
Kirkwood, dielectric theory, 100
Lead acid hybrid battery system, 626
Lead chloride, voltammograms compared with ruthenium oxide, 289
Leakage
current
and non-ideal polarizability, 55
in pseudocapacitance, 5 19
and self-discharge, 605
currents, diagrammatic, 55
e[ectrical, at bipolar electrodes, 622
Leiden jar: see Leyden jar, I
693
Leyden jar, I
illustrated,4
invention of, I
Lines of force, 73
Faraday's concept, 73
Lissajous figures, 503
illustrated, 504
Lithium
electrodes, intercalation, 25
intercalation, illustrated, 26
pseudocapacitance, 25
Load [eve ling in driving cycle by capacitor, 665
Load resistance, capacitor discharge into, 53R
Lock-in amplifiers, 504
Market
aspects, for capacitor sales, 666
share, for electrochemical capacitors, 666
status, in future, 557
Matched load situation, 538, 539
Materials
for capacitors, [7
recent advances in, 651
requirements, for capacitor development, 613
Measurement techniques for capacitance, 502
Mechanisms
of oxidation, reduction of conducting polymers, 300
of self-discharge of capacitors, 559
Mercury, double-layer capacitance at, 132
Metal electron contribution, to double-layer capacitance, 156
Methide ions, 341
Methods for self-discharge studies, tabulated, 562
Microscopic polarization, 46
Miller, contributions to impedance spectroscopy, 525
Mirror image, voltammograms, 12, 22, 262
of ruthenium oxide, 262
Mobility of dissociated ions, 343
Model
of double-layer, general, 114
of electrode Interphase, I 19
Models
of double-layers, diagrammatic, 109
Modern developments, 627
Molten electrolytes, 350
Molybdenum nitnde
behavior Similarity to ruthenium oxide, 654
cyclic voltammogram ot~ 654
for capacitor electrodes, 653
Moments
resolution of, 82
vectorial addition of, 81
Non-Faradaic processes, II, 13, 14
Non-ideal polarization, and leakage currents, 55
Non-ideally polarizable electrode, charging of 573
Nonaqueous electrolytes, 347
and double-layer, 169
694
Nonaqueous
media, double-layer in, 170
solutions, outlook for use of, 180
solvents
acceptor numbers for, 359
classification of, 174
donor numbers for, 340
Nyquist plots, 484
Ohmic potential, in porous electrode, 398
Orientable dipoles, dielectric polarization of, 99
Orientation polarization, 96
Ottawa group, capacitor project, 6
OverpotentiaL 44
current-density relations, for porous electrode,
399
current relation for fuel-cell electrode, 431
distribution, for fuel-cell electrode, 398
Oxidation
of carbon, 196
states in ruthenium and iridium oxide electrodes,277
Oxide films,
conductive, general reading references, 297, 298
various, exhibiting redox capacitance, 286
Oxygen contents, and heats of immersion of
charcoals, 214
Oxygen, interaction with carbon surfaces, 212
Parsons' treatlnent
of volcano relation, 40
of double-layer, water clusters in, 143
Particle size effects in capacitance, 469
Patent literature, for technology, 667
Patent survey
authors dates and assignees for patents, 681
electrolytes, 679
Pauling ionic radii, I 13
Penetration depth
de Levie calculation, 387
of signal in pore, 387
of signal in pore, tabulated, 395
Permanent dipole, in a field, 79
Permittivity
relative, 89
of space, 69
value of, 70
pH
and heats of immersion, 216
and work function differences of carbons, 215
Phase-angle
plots for Faradaic pseudocapacitance, 553
in porous electrode, frequency spectrum 393
related to diffusion, 388
Phase relations, between current and Voltage, 487
Phase-sensitive detection, 504
Poisson's equation, 75, 76
Polarity induction, model of, 94
Index
Polarization
cooperative, 96
dielectric, illustrated, 93
in dielectrics, 91
and dielectric constant, 97
electronic of atoms and molecules, 78
and field, 97
orientation, 96
of power source, 430
of solvent, in double-layer, 81
relations for power source, 432
Polarography, A.C, 508
Polyaniline film growth, EQCM results, 331
Polyaniline films, ellipsometry applied to, 324,
235,330
Polyelectrolyte ionization, 302
Polymers, conducting, 299
Pore
accessibility, impedance behavior, 411
current distribution in, 396
element of equivalent circuit, 390
equivalent circuit of, element, 383
model with equivalent circuit, 348
penetration depth into, 395
of signal, 387
of a porous electrode, electrically represented, 381
size, 407
sizes, classitlcation of, 408
structure, complex in porous electrode, 399
surface areas, for carbon fibers, 410
illustrated, 409
Pores, distribution of charging rates in, 388
Porous electrode, 545
behavior, simulation ot: 547
complex plane impedance diagram, 402
complex plane plot for, 545
complex structure in, illustrated, 400
cumulative pore area, 165
current distribution in, 396, 397
double-layer in, 402
double-layer overlap in, 403
equations for, 383
and ESR, 545
fractal surfaces of, 404
impedance, and ESR, 545
overpotential-current density relations, 399
pore width related to pore area, 165
response function in time, 386
Porous electrodes
of carbon, 376
complex structure in, 399
de Levie model, 382
de Levie's treatment, 382
electrochemical behavior at, 376
electrochemistry at, 555
high area, 376
internal electrolyte conductivity, 360
real areas of, 410
695
Index
Porous electrodes (coni.)
redistribution of charge in, 575
roughness factors of, 414
Porous platinum, impedance of, 413
Potential
and field
at an ion, 72
related, 72
decay, relation to potential recovery, 583
decline in self-discharge, slopes of, 567
in ditTusion layer, 52
profile, across diffuse layer. 160
recovery
model for ruthenium oxide film, distribution
of oxidation states, 592
muillple stages oC at ruthenium oxide, 590
in relation to potential recovery, 583
step polarizallon, 57
Potential-sweep response, diagrammatic, 58
Potentials, due to charge distribution, 72
Power
from capacitors, limitation by ESR, 382, 435
delivery
efficiency, 454
maximum, 452
-density
and C-rate, 443
application aspects of, 468
energy density, equations for, 426
energy density, treatment of, 423, 425
optimization of: 448
related to energy density, 427
time relations, 437
energy density, for capacitors, 423, 425, 433
energy relations for battery and capacitor, compared, 438
limitation
and concentration polarization, 440
and diffusion effect, 440
losses, ohmic and kinetic, 433
production vs, current, polarization relations for,
432
source
polarization in, 430
polarization relations, reduced variables, 432
Precursors and resulting types of carbon, 205
Pretreatments of carbon materials, for capacitors,
203
Pretreatments, chemical, of carbon materials and
capacitance, 208
Principles, of electrostatics, 68
Properties of capacitors, compared, 29
Propylene carbonate salt solutions, 373
Proton transfer within ruthenium oxide films, 281
Pseudocapacitance, 221
of adsorbed H, 224
application to supercapacitors, 253
behavior of pyrite, 234
Pseudocapacitance (coni.)
characteristic parameters for, 229
of conducting polymers, 312
correlation of types, 27
defined,7
and double-layer capacitance, distinction between, 255
electrolytes for, 335
equations, 223, 226
equivalent circuit, 514
formal origin of. 222
frequency response equations, 250
of H adatoms at single-crystal surfaces, 232
at high area carbons, 255
impedance of. 518
for intercalation process, 252
for ion adsorption process, 254
for lithium intercalation, 252
in lithium intercalation, 26,27
intercalation type, 248
IOvolv1I1g chemisorption of anions, 253
kinetic theory of. 236
kinetic theory of, relation to cyclic voltammetry,
238
redox equation for, 249
redox, relation to decomposition voltage ofwater. 247
of redox system, relation to diffusion effect, 250
redox at other oxide films, 286
redox, of transition metal oxide films, 293
redox type, 248
relations to potential, examples, 230
reversibility and irreversibility of, 241
self-discharge of. 579
self-discharge, theoretical calculations of, 580,
581
significant ranges, 246
of sorption into a fullerene, 252
theoretical treatments of, 224
voltammetry curves, characteristics, of, 239
Pseudocapacitor
systems, oxide redox, 618
stored energy in, 222
Pyrite
cyclic voltammograms of, 234
pseudocapacitance behavior of, 234
Quadrupole, in a field, 80
Quartz crystal microbalance
for polyaniline film growth, 331
work on ruthenium oxide, 284
Quaternary salts, conductivities of, 373
Radiationless transition, 38
Ragone
diagram
for capacitor hybridization, 449
for flywheel systems, 475
Index
696
Ragone (cont. )
plots
and SR values, 471
for batteries and capacitors compared, 450
for capacitor, calculated, 439
for capacitors, 418
for solid-polymer capacitor electrodes, 637
illustrated, 419, 420
shape of, 428
Randles circuit, 46, 508
Rate-controlling step, after electron transfer, 50
Rate effects on charge or discharge, 649
Rationalized units system, 69
RC
equivalent circuits, for simulation of porous
electrode behavior, 547
networks, charging and frequency response, 379
and response times, 498
time constant, 379
and characteristic frequency, 494
in complex plane plot, 494
distribution in pore, 388
and equations, 379, 380
formal significance, 50 I
significance of, 497
RCL
circuit, 542
impedance diagram for, 543
network, impedance behavior, 536
Reaction resistance, Faradaic, 46
Real areas
and capacitance. 410
of porous electrodes, 4 I 0
Recyclability. of capacitors. 12
Redistribution
of charge in porous electrodes, 575
of potentials and CUITents In porous electrodes, 577
Redox
Born-Haber cycle, 36
processes, 35
pseudocapacitance
complex plane plot, 550
impedance behavior, 549
pseudocapacitors, ruthenium oxide for, 634
Reduction potentials, relative to water and
acetonitrile, 355
Reference electrodes, 59
Relative permittivity, 89
and dielectric constant, 89
Resistance, Faradaic, 483
Response
current, in cyclic voltammetry, 514
function, for porous electrode, 386
time scales and RC, 498
Reversibility, and irreversibility in
voltammograms, 24
Root-mean-square currents, 498
Rotating vector diagram, for alternating Voltage, 486
Roughness factors, of porous electrodes, 414
at batteries. 557
of capacitors. practical aspects, Ml
of carbon capacitance. measurements on, 582
of a commercial capacitor, 584
under diffusion control, 5Il9
of double-layer capacitor, 574
driving force for. 558
of electrochemical capacitors, 557
by Faradaic processes, 562
leakage through ohmic resistance, 568
log plot in time, 564
measurements, methods for, 561
mechanisms, 559
plot
for diffusion control, 573
for kinetic control, 572
for ohmic leakage, 570
practical phenomenology, 557
697
Index
Self-discharge (cont.)
of a pseudocapacitance, 579
of pseudocapacitance, mechanisms of, 579
and recovery
at commercial capacitor, 585
at ruthenium oxide, 586
slope parameters for, 567
in a stack, 595
surface coverage effects in, 565, 566
temperature effects on, 578
theory of, 563
Self energy, derivation of, 77
Selt~resonance frequency, 543
SEM photos
of carbon tlbers, 630
of carbon foam material, 616
of ruthenium oxide films, 274. 275
Semi-circular form of complex plane plots, 491
Semiconducting properties, of some carbon materials, 195
Semiconductor, space-charge in, 110
Series relation, of capacitance components, 112
Single-crystal surfaces
platinum, voltammetry of. 232
voltammetry of metal deposition, 233
Sluyters' cirClllt. 492
classic paper by. 491
Sohio
brochure, 185
capacitor development by, 185
Solartron
Impedance publication, 482
system. block diagram for.
Sol id metals
Helmholtz layer capacity at, 153
solvent adsorption capacitance of~ 151
Solid-polymer
electrodes. Ragone plots for, 637
electrolyte technology, 636
Solid state, capacitor systems, 635
Solvation
energies, in various solvents, 356
energy, Born calculation, 363
of electrolytes, 335
Solvent
adsorption at solid metals, 151
aprotic in double-layer, 137
classes for capacitors, 349
as dielectric in interphase, 136
effects on reduction potentials, 355
names, abbreviations of, 373
nonaqueous in double-layer, 137
nonaqueous, for higher voltage capacitors,
137
orientation
3-dimensional representation ol~ 154, 155
distribution, 154, 155
in double-layer. 114
at ion and electrode, compared, 175
Solvents
high dielectric constant, 100
nonaqueous classification of, 174
various
potential scales. 355
solvation energies, 356
Space-charge
capacitance of basal plane graphite, illustrated,
200
distribution, 110
Stability ranges for various solutions, 352
Stack of capacitors, self-discharge in, 595
Stacking, of capacitors, 454, 620
Stern theory of double-layer, I 12
Stoney, electron name, 2
Stored energy
per atom, 19
and dielectric constant, 90
in double-layer and pseudocapacitor, 222
Strongly interacting dipoles, dielectric polarization
ol~ 100
Suburban driving, power, energy requirements,
418
Supercapacitor
Ibattery transition, 27
charging and cation, anion separation, 361
defined, 7
Supercapacitors, comparison with batteries. II
Surface
analysis of. ruthenium r.ranilllll oxide Illms,
290
area distribution, In porous electrodes, 408
of charges. field due to. 74
electron density, 118
functionalities
on carbon, 189
at high area carbons, 186
produced by oxidation of carbon, 197
functionality of carbon materials, illustrated,
189
potentials of carbon surfaces, 213
Tafel
relation, 44
relations, for various steps, 51
slope and coverage relations, 50
slopes, 45
Technology development, 609
comparative summary, 670
general reading references, 674
information sources on, 671
references, 673
USABC criteria for, 656
Technology summary, patent literature, 667
Temperature dependence, of capacitor performance, 644
Temperature effect
on capacitance, 462, 644
on self-discharge, 578
698
Test
cell
assembly, for laboratory tests, 60 I
for laboratory, 600
with reference electrode, 603
-electrode holder for electrochemical cell. 604
modes for system performance. 456
procedures. compared, 606
Theoretical treatments, of pseudocapacitance,
224
Theories of dielectric polarization. 87
Thermal management. of capacitor systems, 643
Thermochemical preparation of ruthenium oxide,
265
Thermodynamics of batteries and capacitors, 30
Thiophene and thiophene derivative conducting
polymers. 319
Thomas-Fermi
screening distance, 109
diagrammatic, I 10
Time constant
and characteristic frequency, 496
RC,3RI
Index
Vectorial addition of moments, 8 I
Vehicles
battery requirements for, 471
capacitor requirements for, 472
electric, hybrid power systems for 663
FUDS test for, 473
Volcano relation
for exchange currents, 40
and metal-H bond strength, 41
Parsons' treatment. 40
Volta, I
Voltaic electricity. 2
Voltammetry
and alternating voltage compared,S I 7
behavior, comparative, of various oxide films.
diagrammatic, 294, 295
cyclic, 12
of pseudocapacitance, relation to D.C. charge
and discharge, 243
Voltammogram
for lead/lead chloride, 24
for ruthenium oxide, 24
Voltammograms
for carbon materials, 19 I
of ruthenium oxide, after film thickening, 262
ofruthenium oxide, initial stage, 261
mirror image type, 262
Warburg diffusion impedance, 385
Water
dipole arrays, model, 150
orientation, ice like, at electrodes, 153
oriented in double-layers, 150
polarization of, in double-layer, 139
radial distribution functions for, 172
structure of, 172
Work function
of carbon surfaces, 213
electronic, 34
Work functions
of carbons, as function of pH, 215
and exchange current density, 39
X-ray plots for thermally formed ruthenium oxide,
274
XPS spectra, of carbon fibers, 192
Zinc--bromine hybrid system, 626