Anda di halaman 1dari 19

Electrc&imica

Acta.

1968.

Vol.

13, pp.

1005 to 1023.

Pergamon

Press.

Printed

in Northern

Ireland

HOMOGENEOUS AND HETEROGENEOUS OPTICAL AND THERMAL ELECTRON TRANSFER*


N. S. HUSH Department of Inorganic Chemistry, The University, Bristol 8, England Abstract-A unified theory of optical and thermal outer-sphere electron-transfer processes is outlined, in which the equations are obtained as special cases of general expressions for radiative and radiationless transition probabilities. For a particular donor-acceptor ion system, this leads to correlations between the rate of homogeneous thermal electron exchange and the frequency and bandwidth of the corresponding optical intervalence transfer absorption. These in turn can be correlated with thermal and optical transfer at a metal/solution interface. The relationship between the transmission coefficient in adiabatic thermal exchange and the transition probability in the corresponding optical transfer is discussed, and a number of predictions are made for the optical transfer of electrons between a molecule or ion and an electrode, which has so far not been observed. R&m&-Esquisse dune thBorie des processus de transfert des Hectrons des sphkres extemes optique et thermique: les Equations sont obtenues comme cas spkiaux dexpressions g&&ales propres aux Pour un syst&ne particulier donneurprobabilitBs de transitions rayonnantes et non rayonnantes. accepteur dion, ceci conduit d des corr&lations entre la vitesse dkhange thermique homo&ne de 18lectron et la frequence ainsi que la largeur de la bande dabsorption propre %lintervalence optique correspondante. Ceci & son tour peut ttre coordonne avec le transfert thermique et optique ZL une interface m&al/solution. Discussion du rapport existant entre le coefficient de transmission dun tchange thermique adiabatique et la probabiIit6 de transition du transfert optique correspondant. Des prbdictions sont faites pour le transfert optique des electrons entre une mol&ule ou un ion et une electrode, non observe jusquici.
Zusammenfassung-Man entwickelt eine Theorie des optischen und thermischen Elektronentibergangs in Aussenschalen. Die erhaltenen Gleichungen stellen Spezialf;ille von allgemeinen Gesetzeny d& Wahrscheinlichkeitsfunktionen fiir Uberglnge mit und ohne Strahlung dar. Fiir ein spezielles Donor-/Akzeptorionensvstem ergeben sich Beziehungen zwischen der Geschwindigkeit des homogenen therm&hen Elel&ronena%tausches und der Frequenz und Bandbreite der-entsprechenden optischen Absorption fiir den Intervalenziibergang. Diese ihrerseits kiinnen mit den thermischen und optischen Uberggngen an einer MetalI/LGsungsgrenzfl&che in Verbindung gebracht werden. Die Beziehung zwischen dem Transmissionskoeffizienden des adiabatischen thermischen Austausches und der tfbergangswahrscheinlichkeit im entsprechenden optischen Vorgang wird diskutiert. Man macht eine Reihe von Vorhersagen fiir den optischen Elektroneniibergang zwischen einem Ion oder Molekiil und einer Elektrode, welcher bis heute noch nicht festgestellt werden konnte. 1. INTRODUCTION

ELECTRON transitions may in general proceed either by radiative or radiationless mechanisms. A familiar example of a combination of these processes is found in molecular phosphorescence. A molecule is photoexcited to the lowest excited singlet state l rl * and subsequently undergoes a radiationless conversion to the lowest excited triplet level 3r1*, after which the system returns to the ground state with emission of a quantum of energy E(31l*) - E(ll?,). In this example the excited electron is essentially localized, ie the transition dipole moment is intramolecular. In the solid state, it is also possible not only to excite an impurity hole or electron, such as an electron trapped at an F-centre, but also to ionize it to a free state in the conduction band. This can be accomplished either optically or thermally. The * Presented at the 18th meeting of CITCE, Elmau, April 1967; manuscript received 23 August,
1967. 1005

1006

N. S.

HUSH

thermal excitation or ionization processes can be regarded as many-phonon jumps, and for many years the mechanism of these radiationless processes was obscure. There was no doubt that (together with the reverse dissipation of light energy on de-excitation) they resulted from coupling of electronic and vibrational motion, but the detailed reasons for this were unknown. According to an early theory of Frenkel, slight differences in the lattice vibrations in electronically excited states from those in the ground state are the cause of the many-phonon emission when the electron returns to the ground state. Subsequent work made it clear that differences of lattice equilibrium positions in the two states were often more important. There is an extensive literature on this subject, both for processes in insulators on semiconductors and in ionic crystal~.~ Chemical or electrochemical electron-exchange processes can also be considered from this point of view. In a thermally-activated electron transfer between two ions in a polar medium at the high-temperature limit, an electron can be considered to hop from one nearly localized site to an adjacent one (or to the conduction band of a metal) by an adiabatic mechanism. There is usually a considerable dispersion of equilibriumconfigurations and to a lesser extent of vibration frequencies accompanying such a process. These can be interpreted, in the high-temperature limit, in terms of absolute reaction-rate theory. Little attention has yet been paid to the corresponding optical transfer processes *. Here we shall firstly briefly outline the theory of the connexion between homogeneous optical and thermal electron transfers. A number of aspects of these will be discussed, and finally the types of optical transfer which can be predicted to occur between an interfacial ion or molecule and a metal will be outlined. It is possible to make reasonably definite predictions about the frequencies and characteristics of the absorption envelope of a variety of heterogeneous transitions of this kind, none of which have yet been observed. We shall assume localized (or near-localized) electronic functions in the subsequent discussion. For adiabatic processes, there are many advantages in the use of delocalized (M.O.) functions (c$ ref. 14). However, this would complicate the development of a theory not restricted to the high-temperature limit.
2. ENERGY-DISTRIBUTION FUNCTIONS

The essential features of the radiative and non-radiative transition processes will now be briefly outlined. The discussion is not restricted to electron-transfer processes; the solutions of the equations for these involve only a particular choice of parameter values. In recent years, the properties of electrons in polar media have been discussed mainly in terms of polaron theory,2 which originated with the work of Landaus and Fr61ich.6 This provides a general theoretical framework applicable to weak, intermediate and strong coupling of the motion of the electron with that of the lattice. The situation in which we are interested is that in which electrons are essentially localized in the fields of individual ions, so that the system electron + accompanying lattice vibration constitutes a small polaron. For our purposes here, the theoretical approach of Kubo et a17*8 (cf also Huang and Rhyse) is particularly appropriate, even though it is less general in scope. In particular, the use of co-ordinate and
l The existing literature on the colours and absorption spectra of mixed-valence compounds and solutions has been reviewed..

Homogeneousand heterogeneousoptical and thermal electron transfer

1007

momentum variables rather than creation and annihilation operators for the polarization field results in an intuitively simpler representation of the physical problem. We consider the Hamiltonian H of a system comprising an electron localized initially in a discrete bound state at a site a in a medium L. The medium may be polar or non-polar. In general this can be written as H=H,+He+Hr where HL and He are the sum of kinetic and potential energy operators for L and for the electron, respectively, and HI represents the interaction energy of the electron with L. The analytical form of these operators naturally depends critically on the nature of L and in particular on whether L is an insulating or a conducting medium. Writing nuclear and electronic co-ordinates as Q and r respectively, the energy El(Q) for the lfh electronic state with fixed Q is defined by [He + Hr(r,

Q)ld,(r, Q>= E,(QM,(r, Q). Q) = +,(r, QKv(Qh

(2)
(3)

The wavefunction of the system L + electron can then be written as


yro(r, where I and u are electronic and vibrational quantum numbers. This expresses the Born-Oppenheimer condition, which is usually a good approximation for stationary states. In adiabatic theories of thermal electron-transfer reactions it is assumed that departures from electronic/vibrational separability are so small that they will make little difference to the activation free energy, but are large enough to permit the transmission coefficient to be high. According to (2) and (3), the nuclear vibrational energy E,, in the electronic state 1 is given by [HL + E~(QL(Q) = J%,,(Q). (4) However, it is easily shown that HYJr, interaction potential HI is non-vanishing. mlc(r, Q) is not exactly equal to E,,Y,, In fact, if the

Q) = GY,, + [HL~,,+,- d&,L,l = E,Y,, + HY,,.

(5)

With time-dependent perturbation theory, the last term in (5) leads to a finite value for the transition probability Wlr,.. for the thermal excitation 1+ I. Summed over ail vibrational levels, this is

with E + AE > AE > E. The Boltzmann factor of the initial state (i) is M-~. For the corresponding optical transition I -+l the optical absorption constant k(v) for frequency v is k(v) = (8&Vi/3nc) eEz 2 ~~l(l~(MI1u)j2. v*v* (7)

In (7), Ni is the concentration of electrons per unit volume, c and n are the velocity
of light in vacua and the refraction index of L. M is the electric dipole moment operator eR which may be a function of Q. This is assumed to dominate the optical probability.

1008

N. S. HUSH

In the high-temperature limit, the general features of the distribution functions are as follows.,s For an arbitrary adiabatic potential U(Q), such that the kinetic energy can be expressed in unit quadratic form but not otherwise specified, the thermal distribution function for M #f(Q) is

:=

IWWv)12

Jexp - {BU(x)}G{(E - U(x)) + U(x)} dx J exp - {@U(x)) dx

(84

where x = *(Q + a> and p = (kT)-l. This shows that the most probable optical transition is a vertical transition obeying the Franck-Condon principle. On the other hand, the corresponding distribution function for the thermal process is proportional to W, where
w, = .f

exp- @u(x)}~{~(x) - u(x)>dx. Sexp- @Wx)} dx

@b)

From this it is clear that the radiationless path (at high temperature) proceeds through the point of intersection of the two adiabatic surfaces. This agrees with the expression obtained by Marcus lo for the thermal electron-transfer case. It is also to be noted that there is a close connexion between the distribution functions for the thermal transfer and for the optical transfer in the zero-frequency limit (E = 0). This in fact forms the basis of the detailed interconnexion of the optical and thermal excitations. The connexion between homogeneous and heterogeneous thermal electrontransfer reactions and other types of radiationless processes was first pointed out by Levich and co-workers.lr Their results will be discussed after the general approach is outlined.
3. THERMAL ELECTRON TRANSFER

The considerations outlined above apply generally to electron-excitation processes. We now obtain more detailed expressions, and consider the special case of electron transfer between ions in an ionizing medium or between an ion and an electrode and discuss the thermally-activated process in the high-temperature limit. The adiabatic theories of MarcusI and Hush, although different in approach, are in agreement* (apart from points of detail) for outer-sphere transfers, which are the only types to be discussed here. The non-equilibrium statistical mechanics of these systems has been explored in considerable detail by Marcus. We can reproduce the main features of these results by considering the general equations for the probability of a radiationless transition. In this way it is possible to see clearly the connexion * This is true for homogeneous reactions. For heterogeneous reactions, the role of the image A usual experimental condition is force is not entirely clear in solutions of finite concentration. that in which the electrolyte concentration is sufficiently high for the diffuse double layer potential to be very nearly suppressed. If the reactant ion is located at the outer Helmholtz plane in such a solution, the inner potential at the reactant plane is the same as that in the interior of the solution, and hence the ion is not subject to an image force as it would be in an infinitely dilute solution. For this reason, the image contribution is neglected in calculating an activation free energy by the authors method, whereas Marcus includes it. Marcus predicted activation energies are therefore lower. It is probable that these are limiting approximations, and a detailed analysis of the effect of a discrete double layer is needed.

Homogeneousand heterogeneousoptical and thermal electron transfer

1009

with the corresponding optical process. To simplify the algebra at this point, we assume throughout this section that all vibration frequencies are the same in initial and final states of the system. The frequency tensor associated with the vibrations (assumed to be harmonic) of all atoms of the medium L is designed Q. The displacement vector B = Qof Qt is the vector difference of equilibrium positions of atoms in the final and initial states of the system.? Then, in the high-temperature limit, the activation energy E* of the thermal transition (between adjacent sites for electron transfer) with over-all energy E,, is (with mass normalized to unity) E* = (UQA + Eo)~
2AR2A

For a solution electron-transfer process, the total activation energy is E* plus the energy APE of bringing the ions together in solution from the initial state (i) to form the close-contact precursor state (P).~ On the continuum approximation, in which the medium L is regarded as a continuous dielectric with static and optical dielectric constants K and ~~ the general expression for E* is
E* =

(%+

EoY

4x where8

(10)
In (lo), E,. and EIn are the electric fields due to the charge distributions & and 4&s respectively. An improvement to this approximation can be made by regarding the medium as continuous in an outer volume V and as consisting of a discrete set of oscillators in the complementary inner volume V, with no correlation between the nuclear motions in the two sub-volumes. Writing
Xinner = +A&PA,

and
Xo,,ter =; (; - ~)~& - &12dv,

we have E* = [x inner + Xouter + This can also be written


E* = I:[x + E,,],

Eo12

(12)

4[Xinner + Xouterl

where 1: = *[x +
and x is now Xouter+ xtnner.

Eollx

(13)

We are considering here a process which in the outer-sphere electron-transfer case corresponds to transfer between ions which have been brought together to the t An early and illuminating discussion of the orientation of normal coordinates in an electronic transition was given by Duschinskyls. I am indebted to Professor M. Kasha for providing a
translation of this paper.

1010

N. S. HUSH

close-contact internuclear separation. The over-all activation energy, starting with ions at infinite separation, is that given by (13) plus the work of bringing the ions together, and E,, is the over-all energy change at the close-contact separation. As we are assuming here that the frequency tensor Q and IR are identical, one can write the total free energy of activation for the outer-sphere electron transfer as
A@+ =

~APG

(x +4yG)2

UW

where i, p and q refer to the initial, precursor and successor states. The formal similarity of this equation to previous expressions for adiabatic electron transfer is apparent ; in particular, it is formally identical with the expression of Marcus:i2 in Marcus terminology, APG = w, x = 3, and PAqG = AF,). The expression of Levich and co-worker@ is based on the continuum approximation (neglecting xi,& and is formally similar to (13a) with the appropriate approximation of (10) for xouter. These authors also obtain a low-temperature limiting equation which is essentially identical with that of Kubo and Toyozawa* for the general low-temperature limit for a radiationless transition. The main point we wish to stress here is that all equations up to (13) are general equations for radiationless transitions of the type specified earlier in this section and contain no special features arising from the fact that they are applicable to the electron-transfer process. These special features come in through the particular approximations subsequently introduced in interpreting the x terms. We have discussed the leading features of the high-temperature approximation, in which the radiationless transition is regarded as proceeding at the intersection point? of two adiabatic surfaces. If the temperature is lowered, or the vibration frequencies become very high, the activation energy will decrease until it is finally equal to the over-all energy E,,. At this point, the reaction proceeds entirely by vibrational tunnelling to the ground state of the final state system (or successor state in the case of electron transfer) (cfFig. 1). This is closely connected with the narrowing of the absorption band width of the corresponding optical process when the temperature is lowered ($4). For a thermoneutral process (E, = 0) we can regard E* as arising from the dispersion of equilibrium distances if, as here, the difference of frequency tensors is neglected. This would formally be regarded as a contribution to the free energy of activation (PA%) in analysis of a rate constant. This will vary considerably with temperature at lower temperatures, in spite of the assumption of equality of the frequency tensors. Considering the maximum value of E in the Laplace transform of the energy dispersion exp (-_BE) due to d*, one obtains for the above conditions
E =

kTAQ
2ti

tanh

and

(14a)

7 Strictly speaking, at the lowest intersection point. There may be more than one such point if the frequency tensors are different in initial and final states, in which case the equation for B, (13), becomes a quadratic (cf (146)).

Homogeneous and heterogeneous optical and thermal electron transfer

1011

Both are identical (setting E = E*) with (9) for E, = 0 at high temperature, but not at lower temperatures. This points to the complexity of the temperature-dependence of electron-transfer rates if the conditions are such that either the temperature is low or some vibration frequencies are very high. This may be called the intermediate region, in contrast to the high and very low temperature regions, Investigation of the intermediate region would be fruitful for the theory of both homogeneous and heterogeneous reactions. It is worth noting that Huang and Rhyse have obtained a general expression for the rate of a radiationless transition that covers the complete temperature range, for the special case of a single frequency, as is assumed, for example, in the pure continuum approximation. This gives a useful guide to the behaviour of the rate constant in the intermediate region, for the electron-transfer case. It has so far been assumed in this section that R = U. If this restriction is relaxed, it is found that the free energy of activation for the radiationless transition is given by the maximum value of the expression @(l - 2)AE2{(1 - @X2 + izQ2}-Q2A +@Ztr In {P2((1 - A)S22 + KI2)) + 1E,,. (14b)

This is obtained from the Laplace transform of the transition probability,8 with 1 as a dimensionless variable. With a number of simplifications, whose importance cannot be discussed here, this can be brought to the approximate form of (13a) with the term APG added for the electron-transfer case. The term xinner is now given bY
Xinner =
A$Z(Q2 + c-J2)-R2~, W)

This last expression is identical with that of Marcus l2 for the electron-transfer case. it is assumed for further simplicity In actual calculations of inner-shell terms ,12*14 that the non-diagonal frequency elements are zero. For metal/solution electron exchange, a similar approach yields (xinner) electrode = +(xinner) homogeneous, (xouter) electrode = &out,& homogeneous (image potential included, with ion-metal distance equal to half ion-ion distance in homogeneous case), (xouter) electrode = (~~~3 homogeneous (image potential neglected). If the image potential term is included, we have simply
Xelectrode =+ Xhomogeneous. If

the image potential is neglected, we obtain


%electrode = khomogeneous + 3(XouterJ homogeneous,

giving rise to a larger activation energy.


4. HOMOGENEOUS OPTICAL (INTERVALENCE ELECTRON TRANSFER) TRANSFER

The corresponding optical electron-transfer process is symbolized by the vertical transition in Fig. 1. For the special case of zero frequency difference between initial

1012

N. S. HUSH

and final states, it is evident that the maximum absorption occurs at a frequency kn,X = E,, + &AQ2A (15)

There is thus a simple relationship between the optical frequency maximum and the activation energy E* of the thermal process, viz,

React icn co-ordinate

FIG. 1. Schematic representation of intersection of adiabatic surfaces for electronic states I and I, with adiabatic potentials U and U and over-all energy difference Eo. Vertical transition occurs with most probable frequence hvmax. Thermal activation energy in high-temperature limit is E*. At low temperature, the probability of nuclear tunnelling from point A increases for the thermal transfer. Electron transfer between pairs of adjacent ions in which the electron localized on one ion or the other in initial and final states (intervalence exhibits some unusual features. Let us denote by p an electron-phonon constant, defined as p = ~AC12A/Eo. is nearly transfer) coupling (17)

With this parameter, the frequency maximum is &n,, = E,(l +


P>. (18)

For many transitions, p is small; this is particularly so in ionization of trapped electrons to the conduction band of a semiconductor, as the nearly free electron will have little effect on the lattice equilibrium configurations. It is larger in F-centre excitations, where it may be of the order of 0.5. In intervalence transfer absorption, p is usually very large. For the special case of thermoneutral transfers in which E,, (arising from splitting of the levels of adjacent ions) will normally be a very small fraction of an eV, p is nearly infinite. Intervalence transfer is thus a many-phonon jump with an unusually large electron-phonon coupling constant. If it were not for

Homogeneousand heterogeneousoptical and thermal electron transfer this effect, an optical transition such as Fez+ + Few -+ Fe3+ + Fez+

1013

would occur at almost zero frequency. Experimentally, it is found that intervalence transfer absorption usually occurs in the visible region of the spectrum, although examples of ir and uv bands are also known. In Table 1, calculated frequencies are listed for outer-sphere transfer between pairs of adjacent aquo transition ions at the close-contact separation. These are obtained from the calculated values of E*.14 In
TABLE 1. CALCULATEDAC~VATIONENERGIES(E*) FORTHERMALOUTER-SPHERRELECXXONTRANSFER PEAK FRBQUENCIES (h,,,) AND BAND BETWEEN IONS IN AQUEOUS SOLUTION" AND CALCULATED HALF-WIDTHS (Al,*, 300K) FOR THE CORRESPONDING HOMOGENEOUS OPTICAL INTBRVALENCE TRANSFER.

E*

Ion system
Tja+_Tiz+

Kcal/mol 9.5 9.0


19.5 13.4 12.3 11.5 1@7 17.6*

hl,* cm-l

A1/r cm-l 5500 5300


so00 6600 6300 6100 5900

va+-ve+
Cr*+Xr+ Mns+-Mnz+ Fes+-Fee+ Co*+-Co~+ Pu4+-Pu*+ TI*+-TI+

13,300 12,300
27,300 18,800 17,200 16,100 15,ooo

29,400

7500

* This is the experimental value; the value of E. for the process Tl*+ + Tl+ -+ 2T18+ has previously been calculated~ from this to 13.7 Kcal. This value of E,, is used in calculating hv,,,.

practice it is difficult to obtain sufficiently high concentrations of adjacent outersphere ion pairs of these elements for these predictions to be directly verified. However, outer-sphere transitions for a number of non-transitional ions are known,3*4 such as Sb(III)C1,3--Sb(V)Cl,and Pb(IV)Cl,4--Pb(II)Cl,2in the solid state. Most intervalence transfer bands for transition elements have been observed in inner-sphere environments.3*4 However, the frequencies are not far from those calculated for the simpler case; Fe(III)-Fe(H) transfers, for example, usually give rise to absorption in the region 16-18 Kk in a number of dissimilar environments. 3*4 Further details of the transfer can be obtained from the shape of the absorption envelope. The band envelope will usually have a somewhat more complicated shape than is suggested by the simple expression of (15a). This is because of differences in frequency tensors and the neglect of higher moments than the second of the energy distribution function. If F(E) is the distribution function, the moments are defined by

pFCn

s
--m

m E,F,(E) J-,"

dE

(19)

F,(E) dE

1014

N. S. HUSH

The first three moments for the optical distribution function are* ,ul = aAlR2A + t tr(CF
pa =

coth (@XY/2)r) + E,,,

;b*ap

coth (j?tiCI/2))a2A+ T tr ({W coth (~KY/2)I}2) , coth (/%Q/2) IX-Y-r coth (/3KY/2)) SY2A + $ tr (U-l coth (/3AQ/2)I2) + p tr ({a-l coth (@KY/2)r}3). (20)

These are measures of the peak frequency, the band width and the skewness of the absorption envelope, respectively. Higher moments cannot be ignored in an exact treatment-for example, the fourth moment measures the kurtosis (flattening of the peak and increase of intensity in the wings relative to a pure Gaussian distribution), which is clearly visible on a number of experimental bands. However, in first approximation it is adequate to consider only the second moment. In this approximation, with neglect of variation of frequency tensors (ie, I = 0) the band width can be calculated from (20) if the frequency maximum is known. In the high-temperature limit, this is proportional to T-j2. Estimates of the band half-widths obtained in this way for typical transfers are also listed in Table 1. These agree qualitatively with the observed fact that homogeneous intervalence transfer bands are characteristically broad owing to the large dispersion of equilibrium distances. It is useful to know, for the purpose of practical band analysis, the conditions under which it is permissible to make the approximation coth x M x-l in the moment expressions. In Table 2, values of the quantity gr = (Z)coth(&) for a single frequency u) at 293K are listed. Also, the ratio of square of band halfwidths at 293 and 90, {A1,2(293)/A1,2(90)}2, f or a transition involving only a single frequency arelisted. (This conforms to the continuum model with no localized modes).
TABLE 2. RATIOOF HALF-BANDWJDTHSAT 293K AND 90K~s FUNCXIONOF~U$%T~(T~ AND g-FACTOR AT 293K, ON SINGLE-FREQUENCYMODEL(SEETEXT)
= 293K)

hw/2kT,
1

coth (hw/2kT,) 1.3130 1.3961 l-5059 1.6546 1.8620 2.1640 2.6319 3.4327 5.0665 10.0333 1.313 1.387 1.490 1.620 1.795 2.005 2.260 2.575 2.830 3.140* 1.313 1.257 I.205 1.158 1.117 1.082 1.053 1.030 1.015 1@03t

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2


0.1 * limiting value 3.256 t limiting value 1

Homogeneous

and heterogeneous

optical and thermal electron transfer

1015

A rough idea of the average frequency involved can be obtained from experimental measurements of the half-widths at these two temperatures. Then, if x is calculated from the band width using the high-temperature expression for the moment pz, a first-order corrected value is XCOlT(T) = X(T)/&P This expression is of use in the absence of detailed information about the frequency tensors. Useful correlations can be made between the activation energy of the thermal process and the frequency maximum and band-width of a corresponding optical process. For identical conditions, these correlations are at present limited to solidstate transfers, where the concentration of pair sites is high. In this case the activation energy of the thermal exchange is in principle obtainable from the temperature coefficient of semiconduction, assuming this to proceed by a phonon-assisted hopping mechanism. Direct evidence for the operation of this mechanism in a particular case is really necessary before definite conclusions can be drawn, however. It should be noted that recent work on the Hall effecP and Seebeck coefficient9 of alkali-metal-doped NiO have cast some doubt on the interpretation of the semiconduction process as NP+Niw thermal hopping, although the situation is not completely resolved. A further point of interest here is that for an isolated two-site system, thermal hopping will be expected to lead to a Debye loss, which could be obtained by measurement of the complex dielectric constant. There is some evidence for such an effect in a-Fe,O, at low temperatures from the work of Volger.lt
5. THERMAL TRANSMISSION ABSORPTION COEFFICIENT PROBABILITY AND OPTICAL

We shall also expect to find a correlation between the transmission coefficient of a thermal process and the dipole strength of the corresponding optical transition. This may be useful in estimating transmission coefficients, as the optical intensities are in principle easily measurable. The transitions we discuss here are those in which the degeneracy of the two states at the intersection point of the adiabatic surfaces is removed by resonance interaction, so that the separation of upper and lower surfaces at this point is of the form &I = 2((cl H lb) - (a I b) @l-H la>> 1-(alb)2
P-9

where u, b represent orbitals centered on adjacent sites and H is an appropriate one-electron perturbation. In the high-temperature limit, the transmission coefficient of the thermal transition is 4rr2&2 I K = 1 - exp (21) hv Igrad (U - V),lj where v is the average thermal velocity and the quantity grad (U - U), is measured at the intersection point. In order to estimate this, the reaction co-ordinate must t See note added in proof p. 1023.

1016

N. S. Husk

be known. This can either be (i) a displacement in the dielectric, or (ii) an innershell normal co-ordinate. Estimates of both can be obtained. On the continuum model, with mass normalixed to unity, we have
-

lgrad (u - u,I = ~42~~~

= 2/2x.

(22)

If an inner shell mode is taken as the transfer co-ordinate, the frequency and the gradient of (U - U), can be calculated* from the vibrational potential functions.lan14 The problem common to either choice is calculation of the energy splitting 2s. It is useful to have the magnitude of this for typical cases. Consider an outer-sphere transfer between two six-co-ordinated aquo-ions, with a transitional metal ion at the centre of each octahedron. It is assumed that the splitting arises through overlap of the oxygen 2p, orbitals of neighbouring ions in close contact. Two cases can be distinguished. If the transferring orbital is of type dy*(ie, eB*) in Oh symmetry (eg C!?+-Cr2+), the most favourable configuration for overlap is apical contact, Fig. 2. On the other hand, for d&* (ie t,,*) transfer (eg Fe3+-Fes+), a skew configuration is more favourable.
Y

1_

(a)

(b)

FIG. 2. u-overlap of octahedrally co-ordinated transition text for orbital notation).

ions in apical contact

(see

This arises because the wave-functions of the transferring electrons are of the form Iy* > = a d,,_,, - &(l - ~G)l~(c~4 o, + 0, + u4), [.s*> = /3 d,,_,, - 6(1 - B2Y2(r1 + 2 +
773 + n44),

(23)

where a(1 - a2>, a(1 - /3) measure the delocalization of d-electron density on to each of the appropriate oxygen (Tor r oxygen orbitals, respectively. Estimates of these amplitudes are available from electron-resonance experiments. In each case, the central problem is calculation of the interaction energy a of two oxygen p-functions * If we consider a volume integrated coefficient, ie K = J K(r)4r(r)L dr, we obtain, on expanding the exponential in (22),
K =$q~e+ This is the expression invalid, as an average expanded exponential radiationless transition ;;j$: _ E/)2 du)lllR1dr

given originally by Levich and Dogonadzel However, this procedure is integral total rate should be considered. An expression similar to (21), with and a slightly different numerical factor has been derived8 for the general with resonance interaction of the degenerate states.

Homogeneous

and heterogeneous

optical and thermal electron transfer

1017

I.

-4-

-5

2 d mr %

ii

FIG. 3. Separation of energy surfaces (2~) as a function of distance d,, of adjacent oxygen atoms for a-overlap of octahedrally co-ordinated transition ions with a* = @7. This value is appropriate to the Cr*+-CP+ couple. The pointp indicates the close-contact separation, at which the distance d m is equal to the diameter of a water molecule.

at a distance

A,, overlapping

as in Fig. 2. The energy E is then (dy * transfer). E = &(I - c?)& E = $(l - @2)& (d&* transfer). (24)

In a one-electron approach, E is given by eN= Z,,,((u jr,-l lb) - (~~b)(alr,-~~a))(1 - (~(b)~)-~, where Z,,, is the effective nuclear charge of oxygen. Taking Z,,, = 4.55 and using Slater functions, E is easily computed as a function of the distance d,. In Fig. 3, a plot of log (2.5) (ie the energy separation of the surfaces at the intersection point) is shown for the particular case of a2 = 0.7, which is that appropriate to Cr(I-I,0),3f. The close contact O-O separation can be taken to be twice the radius of a water molecule, ie 2.8 A. At this separation 2~ is approximately O-046 eV. This is in fact of the order of magnitude that has been implicitly assumed in previous work. The energy separation varies very sharply with distance, and is less dependent on the ionic nature of the bond (as measured by (1 - a2) or (1 - p>l). This connects with the oscillator strength of the corresponding optical process. The first-order electronic wave-functions for the states &, and & are?

t The interaction energy E is assumed to be identical in precursor (P., and transition states in this estimate.

1018

N . S . HUSH

Taking hvm~x = (U ~ ' - U')p, and neglecting the overlap integral between ff~. and 4~z., the transition moment of the optical intervalence transfer is thus

where R is the interionic separation and g is the number of equivalent transferring electrons per centre [g---- 1 for both FeS+-Fe~+ and CrS+-Cr~]. The oscillator strengthfis f= 4.32 10-9J r dv (27)

= 1.085 10-Sh~,m~xIMI~/eL

where h*'ma x is expressed in wave-numbers and ]Ml/e is in Angstroms3 For a band of Gaussian shape, the excitation coefficient of the peak maximum is f 109 emax -- 4.6Al/2 , (28)

where emax is the molar decadic extinction coefficient, with the band half-width At/2 in wavenumbers. Thus, for the Cra+-Cr a+, aquo-ion optical transfer, using the theoretical values in Table 1, we calculate em~x = 11. The experimental values so far obtained for the oscillator strengths of intervalence transfer bands suggest that the extent of electron delocalization in the ground state is usually quite small, and comparable with that calculated above for the Cr(H~O)es+Cr(H20)6 ~" close-contact pair. For example, the f-value for the TP+(C1-)e-Ti~-(CI-)x intervalence transfer (20"41 KK) in concentrated HC1 is estimated as 0.0026, corresponding to a value of about 0.11 h for the dipole strength [Ml/e./.is This implies that { e ' / ( U ~ -- U')} 2, (25), is 0"5 10-8 for this chlorine-bridged inner sphere TiS+-Ti ~complex, ie that the electron is 99.95 ~o localized on Ti ~- orbitals and 0.05 ~o delocalized on to Ti a+ in the ground state. A similar calculation from the intensity of the Fe(III)t2gaeg~(6S) + Fe(II)t2g6(xS)-,.
Fe(II)tzo4e~2(~D) + Fe(IlI)tzg6(2I)

transition in Prussian Blue (14.1 kk) shows that the Fe(II)t~g electrons are 99.03 per cent associated with the donor Fe(II) ion in the crystal. 4a9 In this case, each donor ion is surrounded by six equidistant acceptors, resulting in a relatively large total delocalization. Also the factor g is 6 for the low-spin Fe(II)t2g 6 configuration, so that the final intensity of the band (summed over all exciton states) is comparatively high. In the theoretical estimate of optical transfer oscillator strength we have not considered the possible Q-dependence of the transition moment. This will be important for transfer to some excited states, in which the static moment vanishes by symmetry. Also, other sources of intensity (eg intensity borrowing from neighbouring charge-transfer states) may be present in particular systems. However, it is clear that for systems in which it is permissible to assume that the above model is adequate, a close relationship exists between the transmission coefficient of the radiationless transfer and the oscillator strength of the corresponding optical transfer.

Homogeneous and heterogeneous optical and thermal electron transfer 6. OPTICAL ELECTRON TRANSFER BETWEEN AN INTERFACIAL ION OR MOLECULE AND AN ELECTRODE

1019

These considerations are easily extended to the case of optical transfer between a species present in the electrode interface and an electrode. The type of experiment envisaged here is one in which a transparent electrode (eg a stannic oxide Glm) is used, preferably mounted on the face of a prism in a multiple-path attenuated total reflexion cell. To begin with, we shall assume that the centre of the ion is located at the outer Helmholtz plane, so that there is no specific adsorption and that the ionic strength of the solution is sufficiently high for the diffuse doublelayer potential to be suppressed. It is also assumed that the difference r is zero. Under these conditions it is easily deduced that optical transfer at zero overpotential (denoted by the zero right-hand superscript, below) will occur at a maximum frequency hv:,, homogeneous, (29) where E* is the activation energy PAE) for the symmetrical homogeneous thermal process, & + a=1 -+ al-1 + &. This is predicted to be independent of the nature of the metal. The approximate equality sign in (29) indicates the uncertainty about the magnitude of the electrical image term, a complication which will also arise in a slightly different form (as the electrode dielectric constant is no longer infinite) for semiconductor electrodes. However, this uncertainty will usually be about 0.1 eV, or less than 1 KK. It is therefore assumed here that the equality sign holds. Calculated values of hv;,, and band-widths for ion metal transitions for the aquo ions discussed earlier are shown in Table 3. Again, calculated values of the quantity x(=4E*) have been used. For typical transition-ion transfers, the optical bands at zero overpotential will lie in the ir, often at very low frequencies. (It would be necessary to use D,O as solvent for many of these systems). The bands will be narrower than those observed in homogeneous transfer in the ratio Al,z(homogeneous)/A1,z(heterogeneous) = 2/2. Perhaps the most interesting feature of these predicted transitions is that the peak frequencies will vary linearly with overpotential, q, according to
hv;,,
TABLE 3. CALCULATZD ION sr. ELECTRODE

M 2E* M $(hv,,J

= hv;,,

+ FT.
(300K)
FOR

(30)

FREQUENCY MA~~AANDBANDI-MLF-WIDTHS OPTICALELE~TRONTRANSFERATZEROOVERPOTENTIAL

Jdmxt
Ion system*
Ti*+-Tj*+

cm-l 6,600 6,100 13,600 9,400 8,600 8,m 7,500 14,700

A 11s cm-l
-

va+-vs+ Crsf-Cra+ Mn*+-Mnx+ Feat-Fe*+ co*+-Coe+ Pu+-pus+ TP+-TI+ * Aquo-ions in aqueous solution.

3,900 3,800 5,700 4,700 4,500 4,300 4,200 5,300

t Source of theoretical values for x as in Table 1.

1020

N. S. HUSH

This will also aid in the identification of the band, and it will also make it possible to shift it away from absorption regions of the individual ions. Thus, an Fe(H,0)69+Fe(H,0)2+ band would be expected at 8.6 KK at zero overpotential and to shift to 16.67 KK at a positive overpotential of 1 V. On the other hand, the band half-width should be invariant with potential. This behaviour is illustrated schematically in Fig. 4 for Few-Few and Tlw-Tl+ optical transfers at the interface with an aqueous solution.

$v):

+I

Fe 2t- Fe 3c

AA

r)(v):

-I

ti

Tl+-T13+

KK FIG. 4. Predicted frequencies and approximate band envelopes for Fes+-Fe*+ and TP+-T1+ optical transfer at a metal-aqueous solution interface as a function of overpotential. The relative intensities are arbitrary.

The band intensities (and hence the extinction coefficients) are expected to be sensitive to changes of overpotential. The relative intensity 8, of a metal-to-ion transfer at overpotential 11to the total intensity at 7 = 0 is approximately

eq = (1 + exp (-F#T))-lT vq (M,12


Vmax ( lwll = 1

(1 + exp (-F~/RT))-~$+.

nlax
In this expression, the first factor expresses the dependence of the acceptor-ion concentration on overpotential. If x is of the order of 2 eV, 8, is 413 at 0.5 V positive overpotential and 0 at O-5 V negative overpotential. For the reverse ionto-metal transfer, the same relationships hold with the sign of r reversed. For the ions discussed, the absolute intensities will be very low. With a value of 100 as an estimate of a typical molar extinction coefficient, by analogy with the

Homogeneousand heterogeneousoptical and thermal electron transfer

1021

homogeneous case and with a probable path length of ca 5 A, the optical density for a 1 M ion solution is only 5 x IO- 6. There is thus a problem analogous to that encountered in the study of homogeneous transfers. It is clear that in order to obtain bands of moderate intensity, closer interaction with the metal is desirable. This may be effected in a number of ways, such as (i) the use of large organic ligands which have extensively delocalized n-systems (eg dipyridyl, o-phenanthroline) (ii) the use of bridging ligands such as N3-, or others in which there is a possibility of forming a bond to the metal surface through specific adsorption. In organic systems, it is often likely that specific adsorption will occur, which will greatly enhance the intensity. The dependence on overpotential will now be more complex, however, owing to the finite potential gradient in the inner Helmholtz layer. At finite overpotential, the forward and reverse transfers az+ + e(M) -+ a(-l)+ az-l)f -+ aa+ + e(M) will occur with different energy maxima. In the high-temperature limit, the difference is given by /W;,,(I) - hv;,,(2) = 2Fy. (31) (11;

Thus, at small overpotentials it may be possible to distinguish two peaks, for metal-to-ion and ion-to-metal transfer, respectively. As the concentration ratio [az+]/[a(z-l)+] is given by exp (-Fq/lU), the intensity of (ii) will rapidly diminish at positive overpotentials while that of (i) will rapidly diminish at negative overpotentials. Thus, at positive overpotentials, the transfer will be almost entirely (i), ie transfer to the ion, while at negative overpotentials the reverse ion-to-metal transfer will predominate. The actual intensities can be calculated by an extension of the method outlined above, taking into account the relative concentrations of oxidized and reduced ions as a function of overpotential. The experimental observation of optical transfer at electrodes would be of great interest, as we would then have systems in which the optical and thermal rates could simultaneously be studied with continuous variation of the over-all energy difference. In addition, photocatalysis of the thermal electron-transfer step, analogous to that observed in some homogeneous electron-transfer systems, could also be examined. 7. CONCLUSION In outlining a unified theoretical treatment for optical and thermal transfer processes, it has been necessary to restrict the treatment to the general features of the phenomena. There are many detailed aspects which have not been discussedfor example, the operation of spin conservation rules. However, the essential features of the correlations seem clear. Further experimental work is necessary before it will be profitable to pursue them in greater detail. REFERENCES
Phys. Rev. 37, 17, 1276 (1931). 2. Pohrons and Excitons, ed. C. G. Kuper and G. D. Wbitfield. Oliver and Boyd, Edinburgh and London (1963).
1. J. FRENKEL,

1022

N. S. Husrr

3. G. C. ALLEN and N. S. HUSH, in Progress in Inorganic Chemistry, Vol. 8, p. 357, ed. F. A. Cotton. Interscience, New York (1967). 4. N. S. HUSH, in Progress in Inorganic Chemistry, Vol. 8, p. 391, ed. F. A. Cotton. Interscience; New York (1967). 5. L. D. LANDAU, Phys. Z. Sowjetunion 3,644 (1933). 6. H. FR~LICH, Proc. roy. Sot. A160,230 (1937). 7. R. Kuno, Phys. Rev. 86,929 (1952). 8. FL Kuno and Y. TOYOZAWA, Prog. theor. Phys. 13,160 (1955). 9. K. HUANG and A. Rws, Proc. Roy. Sot. A204,406 (1950). 10. R. A. MARCUS,J. them. Phys. 24,966 (1956). 11. V. G. LEVICHand R. R. DOGONADZE, Dok. Akad. Nauk SSSR 124, 123 (1959); Proc. Acad. Sci. USSR Phys. Chem Sect. 124,9 (1959). 12. R. A. Mucus, Ann. Reo. Phys. Chem. 15,155 (1964). 13. F. DUSCHINSKY,Acta Physiochim. USSR 7,551 (1937). 14. N. S. HUSH, Trans. Faraday Sot. 57,557 (1961). 15. V. P. ZHUZE and A. I. SHELYKH,Sou. Phys. Solid State 5, 1278 (1963). 16. A. J. BOShiAN and C. Cmmcmxm, Phys. Reu. 144,763 (1966). 17. G. VOLGER, Disc. Faraday Sot. 23,63 (1957). 18. C. K. JORGENSEN, Acta Chem. Scand. 9,405 (1955). 19. M. ROBM, Znorg. Chem. 1,337 (1962). DISCUSSION Z. R. Grabowski.-You speak of an optical transition of an electron initially localized on one atom to another orbital localized on the second atom. In the case of an intervalence transfer between like atoms, cannot the upper state correspond to a delocalized electron, common to both atoms? N. S. Hush.-The term intervalence transfer is used for transitions in which the electron is very nearly localized on the donor ion in the ground state. Thus, the electronic wave-function of the ground state can be written as IO> = $d + ~da+a, where +d and $B are the donor and acceptor wave-functions, for the excited state is correspondingly and
Ada

is small.

The wave-function

Provided that the orbital overlap is small (as will always be the case in intervalence transfer) we can set &a = Aad. This shows that the excited state contains only a small admixture of the donor function. The amplitude I2 is often of the order of 0.0005 for typical intervalence transfers, which usually have comparatively small oscillator strengths per electron. These expressions are indeed quite good approximations in the high-temperature limit. R. A. Marcus.-The comparison of the energies involved in thermal and light-induced electron transfers and made by Dr. Hush is an interesting one. By examining the polar solvent contribution to these energies, one can show that the simple relation is obtained between the two, with or without dielectric continum theory, if a dielectric unsaturation approximation is used and specitic interactions are neglected. From (68) and (14) with appropriate substitutions, one finds

(1)
humax = A+ A, (2) where* A = AFO + wp - wr, vmBX is the frequency for maximum absorption and the other symbols have usual meaning.* When vibrational motions of the reactants are included and treated as harmonic the same results obtain, but now rZis lo + i(r, detined elsewhere, instead of lo.

* If an electronically excited state of the product is found in (1) or (2) the AFO in that equation is the standard free energy of formation of that state from the reactants in the prevailing medium and at the prevailing temperature. 1. R. A. MARCUS, J. them. Phys. 43,679 (1965). 2. R. A. MARCUS, J. them. Phys. 43, 1261 (1965). 3. R. A. MARCUS, Electrochim. Acta 13, 995 (1968). 4. R. A. MARCUS, Ann. Rev. phys. Chem. 15, 155 (1965).

Homogeneous

and heterogeneous

optical and thermal electron transfer

1023

Note added in proof: (30 April 1968) There is now definite evidence for electron-transfer Debye loss in a number of inorganic solids. S. Van Houton and A. J. Bosman (InformaI Proc. Buhl Int. Conf. Materials, Pittsburgh 1963, p. 123. ed. E. R. SCHATZ. Gordon and Breach, New York, 1964) have studied Li-doped NiO and have observed dielectric loss attributed to the thermoneutral exchange Ni*+ + Ni+ -+ Nis+ + Nix+ between ions adjacent to the substitutional Li+ ion. The similar transfer

co*+ + co*+ + con+ + co*+ has been studied in alkali-metal doped COO. [BOWAN and C. CRE~ECXXEUR, J. Phys. Chem. Soli& 29,109 (1968)). In the authors laboratory [N. S. HUSH and R. C. &APPELL, to be published], similar results have been obtained for Li-doped NiO; in addition, Debye losses in Nb-doped TiO, and in non-stoichiometric ferric ferricyanide have been observed, and are attributed respectively to the exchanges Tia+ -C Ti4+ + Ti4+ + Tiaf Fe8+ + Fe*+ -+ FeB+ + FeS+ An interesting feature of this effect is that the method is the analogue for homogeneous systems of the determination of heterogeneous electron transfer rates by measurement of faradaic admittance. Electron-transfer Debye loss will be less easily observed in stoichiometric compounds. Here the transfer is not thermoneutral (EO # 0). The dielectric loss E at the Debye peak will be smaller than that for a corresponding thermoneutral transfer by the factor sech* (&/XT). This is about l/100 for EO = 6kT, so that high sensitivity will be generally required even for systems in which E,, is comparatively small. The possibility of resonance absorption at low temperatures as a result of transition to a nonadiabatic vibrational tunnelling mechanism [cf. Fig. 11 is also being explored.

Anda mungkin juga menyukai