Discrete two-terminal single nanocluster quantum

optoelectronic logic operations at room temperature

Tae-Hee Lee and Robert M. Dickson*
School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta, GA 30332-0400

olecularly engineered electronics hold great promise for

producing functional nanoscale devices (13), yet difficulties in wiring three independent electrodes in nanoscale dimensions continue to limit advances and even preclude measuring
polarity in such devices (47). As the active elements shrink
toward the molecular scale (810), standard transistor-based
circuit size remains limited by electrode fabrication and addressability. Even with current nanodevices (1115) roomtemperature operation using discrete energy levels has not been
demonstrated. Such room-temperature nanoscale quantum electronic devices would enable more complex calculations with
vastly simpler circuits than possible with standard transistors
(1618). Here we report two-terminal optoelectronic logic operations using quantized single Ag nanocluster (19) energy levels
at room temperature. Achieved through pulsed operation, the
first pulse gates electron injection into discrete molecular
nanocluster energy levels by the second pulse. This pulsed,
energetically discrete, two-terminal operation greatly simplifies
nanoscale electronics device fabrication, yielding room temperature AND, OR, XOR, and NOT logic operations and even a full
adder from individual parallelized single Agn (n 28 atoms)
nanoclusters (19) with inherent electroluminescent readout (20).
The difficulties associated with electrically isolating similar or
different single molecules are circumvented through optically
reading the unique electrical response of each molecule.
Readily formed in both chemically prepared and radio
frequency-sputtered thin Ag films, 1D arrays of nanoscale break
junctions (21) are fabricated in situ by computer-controlled
passage of voltage-limited current through the films for 30 s.
This gentle electromigration process (21, 22) reproducibly forms
arrays of spatially isolated electroluminescent Agn nanoclusters
in the resistive break junction region. Formed Agn nanoclusters
are likely to be closer to the anode because of the momentum
transfer from the electron wind during electromigration (22).
Voltages applied to the anode with the cathode as ground
produce strong electroluminescence (EL) from current flow
through multiple electrically written individual Ag nanoclusters
spanning insulating silver oxide layers on silver electrodes (Fig.
1) (20). Knowledge of this device structure enables construction
of a detailed energy-level diagram that governs the EL process
(Fig. 2). Experimental (2325) and theoretical energy levels for
the electrode (Ag) (23), barrier oxide (Ag2O) (24), and active
species (Ag2Ag8) (2527) dictate the necessary voltages to
www.pnas.orgcgidoi10.1073pnas.0635474100

inject electrons into Agn excited states through the Ag2O conduction band. Because Agn nanocluster ground states are within
the Ag2O bandgap, injected holes are trapped by the oxide layer
(Fig. 2). Independent of electrode separation, the entire applied
voltage is dropped at the emissive junction. The narrow recombination zone leads to the high-frequency enhancement attainable in these discrete single-molecule elements (20).
To probe the time scales relevant to the EL process and
perform logic operations, two consecutive high repetition rate
narrow pulses (25 MHz) were used to initially inject holes then
reinject electrons. EL intensities were recorded as functions of
pulse width, polarity, amplitude, and interpulse delay. Contrary
to dc studies, (20) pulsed excitation introduces polarity; EL is
observed only for positive- followed by negative-pulse combinations. Emission is most enhanced when narrow positive and
negative pulses (3.5 ns, full width half maximum) are sequentially injected within 4 ns of each other. Although some
variability in pulse widths arises in different devices (2.83.5
ns), interpulse delay is always limited by the two pulse widths and
corresponds to the ac enhancement frequency. The first pulse
establishes a field across the junction to inject holes into the
nanoclusters by tunneling through the oxide layer, essentially
concentrating charges in the nanocluster region for a short time.
Because of the junction capacitance and the higher energy
nanocluster excited states relative to the Ag electrode Fermi
level, only when a negative second pulse follows a positive first
pulse can the second pulse reach the nanocluster junction and
directly reinject electrons into the EL-producing Agn nanocluster-excited electronic energy levels. Because the oxide barrier at
the junction prevents direct electron reinjection into the nanocluster ground states, quenching of EL is avoided (Fig. 2).
Obviating the need for a third electrode, the first short pulse acts
as a gate by concentrating holes in and around the active
molecules for a short time. Establishing a field to extract
electrons, the first-pulse amplitude does not interact with discrete Ag nanocluster energy levels, but relates to a probability of
hole injection in the active medium. Acting as a gate in this
two-terminal device, increasing first-pulse amplitude increases
the probability of hole injection in each individual molecule,
thereby increasing total EL (Fig. 3).
When applied within 4 ns of the positive first pulse (gate),
the negative second pulse reinjects electrons into discrete Agn
nanocluster-excited electronic energy levels. When properly
applied, these pulse combinations enable current to flow with EL
resulting from electronhole recombination within each nanocluster. Intrinsically molecular species, different nanoclusters
produce strong EL only when the second-pulse amplitude precisely corresponds to excited-state electronic energy levels.
Because the first pulse is for hole injection, only the second pulse
directly injects electrons into the discrete energy levels of the
single Agn nanocluster junctions composing the recombination
zone. Measurements of EL intensity (IEL) from individual
This paper was submitted directly (Track II) to the PNAS office.
Abbreviation: EL, electroluminescence.
*To whom correspondence should be addressed. E-mail: dickson@chemistry.gatech.edu.

PNAS March 18, 2003 vol. 100 no. 6 30433046

PHYSICS

Readily formed at nanoscale break junctions, arrays of individual

spatially isolated, strongly electroluminescent Ag2Ag8 nanoclusters perform complex logic operations within individual twoterminal nanoscale optoelectronics devices. Simultaneous electrical excitation of discrete room-temperature nanocluster energy
levels directly yields AND, OR, NOT, XOR, and even full addition
logic operations with either individual nanoclusters or nanocluster
pairs as the active medium between only two electrodes. Imaged
in parallel, noncontact electroluminescent readout obviates the
need for electrically isolating individual features. This gated,
pulsed, two-terminal device operation will likely drive future
nano and molecular electronics advances without complicated
nanofabrication.

CHEMISTRY

Edited by Allen J. Bard, University of Texas, Austin, TX, and approved December 12, 2002 (received for review September 9, 2002)

Fig. 1. Microscopic images of an Ag nanocluster junction. (A) Scanning electron micrograph (Hitachi S-3500H, 15-keV electron energy) of a typical silver
nanocluster junction. (B) Optical image (0.2-s charge-coupled device exposure collected with a 1.4 numerical aperture, 100 oil-immersion objective) from the
region bounded by the white box in A. White circles represent the EL areas corresponding to those further magnified in the scanning electron micrograph images
(C and D). Formed in situ by dc excitation (20), ac-excited emission [163-MHz sine wave with 5-Vpp (2.5 V) amplitude] arises from species too small to be observed
with scanning electron micrography (5 nm).(E) Degree of oxidation with distance from the junction as determined with energy dispersive spectroscopy
(ThermoNoran, Middleton, WI, attached to Hitachi S-3500H) from 3-m diameter regions. The poor energy dispersive spectroscopy spatial resolution was
necessary to obtain sufficient signals, but it still clearly indicates a silver-rich junction surrounded by silver oxide between the pure silver electrodes. Numbers
on the curve (AgO ratio) are the inverse of the y axis (OAg ratio) to more clearly indicate approximate AgO atomic ratios in different regions.

nanoclusters while adjusting the second-pulse amplitude from 0

to 3.15 V are thus expected to yield several peaks in the IEL vs.
voltage curves. However, because the number and position of
peaks strongly depend on the electronic energy levels of individual nanoclusters, integration of the whole EL signal from all
of the different nanoclusters in one device (100) yields strong,
first-pulse gated EL with two-terminal operation, but without
observation of discrete levels as a function of second-pulse
amplitude. Even with 30 molecules, the discrete nature of
single molecule energy levels is blurred, yielding a nearly monotonic increase in EL with increasingly negative pulses after the
positive gate pulse (Fig. 3B). Only the EL signals from individual
nanoclusters or from groups of those with similar properties
exhibit discrete nanocluster energy levels (Fig. 3C). EL from
each individual nanocluster and that summed from similar

Fig. 2. Composite energy diagram governing silver nanocluster EL. The

photoelectric work function of Ag(111) (4.74 eV) (23), x-ray and inverse
photoelectron spectroscopy of the Ag2O band structure (24), gas-phase ion3044 www.pnas.orgcgidoi10.1073pnas.0635474100

ization potential (5.66 eV) (25), and calculated electronic energy levels (26, 27)
of Ag3 all are plotted on the same energy scale to illustrate EL from silver
trimer. Other small nanoclusters also have both visible emission and molecular
energy levels favorable for EL in this device geometry (25). (A) This diagram
shows that holes can be trapped in silver nanoclusters because the hole (h)
is located within the oxide band gap. After the hole is formed by field
extraction, election reinjection causes either nonradiative thermal decay (B) or
excited-state injection with subsequent emission (C) (20). Optimal hole formation and electron reinjection times are found to range between 2.8 and 3.5
ns from two-pulse excitation experiments.

Lee and Dickson

nanoclusters definitively show that EL continuously increases

with first-pulse amplitude (gain), but exhibits clearly resolved
peaks with second-pulse amplitude (Fig. 3C). Thus, with direct
EL readout from selected individual nanoclusters, this device
can be used as a time- and voltage-gated, two-terminal quantized
single molecular switch when operated with properly designed
gate and input signal pulses into a single electrode. Because only
two electrodes are necessary, even the simple analog gating with
the first pulse makes these elements useful as single molecule
devices without complicated nanofabrication of source, drain,
and gate electrodes. The varied but discrete second-pulse voltage
dependence of EL from different nanoclusters can be further
used with parallel EL readout to perform significantly more
complex operations within a single device than possible with
standard transistor-based electronics.
This two-terminal, parallel, single-molecule excitation with extremely sensitive EL readout (Fig. 3A) results from electron injection into discrete energy levels and enables room-temperature
circuit construction with much greater simplicity than possible with
standard electronic components. Using the inherently parallelized
collection of quantized EL features and the simple gated emission
Lee and Dickson

CHEMISTRY
PHYSICS

Fig. 3. Quantum optoelectronic characteristics of typical Ag nanocluster

junctions. (A) Typical Agn EL region with two consecutive pulse excitation
(25-MHz repetition rate) exhibiting blinking (28 30) and dipole emission
patterns (19, 20, 31) characteristic of individual molecules. The EL shows a
strong pulse polarity dependence and is detected only with a positive followed by a negative pulse with proper widths (2 4 ns), amplitudes, and
interpulse delay (4 ns). Other two-pulse polarity combinations yield no
detectable emission. The EL intensity also depends strongly on the amplitudes
of the positive first and negative second pulses. (B) Integrated EL from 30
molecules exhibits nearly monotonically increasing emission with both firstand second-pulse amplitudes. Positive pulses 4 ns ahead of the negative
pulses, each with 2.6-ns pulse widths and 1.0-ns transition times, were used.
Onoff ratios of 15 were obtained with 2.1-V signal pulses (the second pulse)
after either 1.59-V (off) or 3.15-V (on) gate pulses. (Inset) The lowest voltage
case, showing the threshold behavior, is expanded. (C) Integrated emission
from five similar molecules exhibits three clear peaks along the second-pulse
amplitude at 1.6, 2.1, and 2.85 V. EL increases monotonically with
first-pulse (gate) amplitude, showing no discrete features except as a function
of second-pulse amplitude. (Inset) Expansion of the lowest first-pulse amplitude trace to show the high contrast obtained and therefore the discrete
nature of the individual molecularnanocluster energy levels.

Fig. 4. Quantum optoelectronic logic operations with coupled individual silver

nanoclusters at room temperature. (A) Using 2.6-ns pulses at 25-MHz repetition
rates, a 2.2-V first pulse was always added before the second composite pulse,
which itself was constructed from three separate input pulses (P1, P2, P3), each
with either 0.0- or 1.05-V amplitude. Voltages described below refer to the
composite second-pulse amplitudes. (B) Nanocluster S (sum) is turned on (an
output of 1) at 1.05 and 3.15 V and turned off (an output of 0) at 2.1 V.
Nanocluster C (carry) is turned on at 2.1 and 3.15 V and turned off at 1.05 V.
When only two input pulses with 0 V (an input of 0) or 1.05 V (an input of 1)
amplitude are used, S and C act as XOR and AND gates, respectively to constitute
a half adder. For example, if only P2 and P3 are used as inputs, C [(P2 OR P3) AND
NOT(P2 AND P3)] and S (P2 AND P3). The onoff ratios for XOR and AND
molecules are 6.0 and 12.2, respectively. The same molecules can also be fed with
three inputs to work as a full adder. Input pulses and results for arithmetic
addition of the corresponding three bits are shown (A). (C) The discrete ELproducing energy levels of these two nanoclusters enable operation as a full
adder. Nanoclusters S and C act as the output nodes in the logic diagram,
incorporating nine basic binary logic gates typically requiring at least 25 standard
field effect transistors. In the full adder implementation demonstrated in A,
nanocluster S [(P2 XOR P3) XOR P1] and nanocluster C [P1 AND (P2 XOR P3)]
OR (P2 AND P3), in which XOR is defined as: a XOR b [(a OR b) AND NOT
(a AND b)].

of multiple spatially isolated individual molecules, a variety of logic

functions can be performed. Although examples of logic operations
based on the simple gating behavior with the first pulse are not
PNAS March 18, 2003 vol. 100 no. 6 3045

explicitly shown here, the fundamental data are given in Fig. 3. The
high nanocluster emission onoff ratio is a strong function of
first-pulse (gate) and second-pulse amplitudes and polarity, turning
on only with specific second-pulse voltages. Provided that the
positive first pulse for field-dependent electron extraction (hole
injection) is always added at the prescribed time before the second
pulse, the nonmonotonic IEL-V characteristics enable combination
of several input signal pulses as a composite second pulse to yield
different logic gates. In this manner, AND, OR, NOT, and XOR
gates, as well as complicated combinations of these logic gates,
are readily produced from different molecules within individual
two terminal devices (Fig. 4). Although the output is light and
therefore difficult to propagate to subsequent devices, these optoelectronic devices are an important advance as they use discrete
room-temperature energy levels to form true single molecule logic
devices.
All emissive nanoclusters in all such devices exhibit similar
discrete injection behavior and can perform different logic
operations. Differently designed input pulses can even yield
different operations from the same molecule by being on or off
resonance with EL-producing excited nanocluster energy levels.
For example, one individual EL nanocluster (Fig. 4C, molecule
S) was gated with a 2.2-V first pulse followed by different
composite negative second pulses. Because this particular molecule produces EL when excited with a second pulse of 1.05 V,
but not at 2.1 V, combinations of two signal pulses each either
0.0 or 1.05 V (i.e., 0 or 1, respectively) produce four possible
composite second pulses with amplitudes of 0 V (one combination), 1.05 V (two combinations), or 2.1 V (one combination). Thus, only a 0 combined with a 1 in either channel yields
EL. This nanocluster, therefore acts as an XOR gate with two
0- or 1.05-V input signal pulses overlapped to make one
composite second pulse (Fig. 4, molecule S). Contrary to serial
concatenation of multiple logic gates (typically using eight
standard three-terminal field-effect transistors) to construct one
XOR gate, the single XOR nanocluster between only two
terminals successfully performs the parallel XOR operation.
This same molecule also operates as a NOT gate with more
complex pulse sequences. By repetitively applying a composite
gate pulse consisting of a 2.2-V first pulse 4 ns ahead of a
trailing 1.05-V pulse, single-molecule EL is continuously ob-

served. Combination of this gate pulse with an additional

1.05-V signal pulse overlapping that incorporated with the
composite gate pulse yields no EL. In the same manner, AND
and OR gates can also be implemented with different nanoclusters and different gate and input signal pulse amplitudes corresponding to the proper energy levels for each individual molecule logic gate. More strikingly, by injecting electrons into
excited states with energies more negative than (i.e., higher
energy than) 2.1 V in the molecules shown in Fig. 4 so that
three or more input pulses can be fed into the two-terminal
integrated device, more complicated logic operations can also be
performed. The discrete but coupled responses of two different
Agn nanoclusters enable their EL to perform full addition
operations. Using standard circuitry, a full adder would consist
of nine basic binary logic gates (requiring at least 25 standard
three-terminal field-effect transistors; ref. 18), but is successfully
implemented with two single nanoclusters in a single twoterminal integrated device (Fig. 4A).
Whereas different nanoclusters have different energy levels, the
IEL vs. V2nd pulse for each can be characterized and subsequently
used to readily yield logic gates with more refined pulse input
requirements. Thus, although formed in situ in the nanoscale break
junction, this is a general method of using nanomaterials to form
two-terminal quantum electronic logic gates with noncontact EL
readout. Such pulsed operation with properly designed electrode
systems greatly simplifies the construction of nanoscale electronics
components. Electron injection into discrete electronic energy
levels offered by these robust Ag nanomaterials at room temperature opens the way to vastly simplified circuit design with naturally
parallel optoelectronic information processing. Because the active
regions of most single-molecule and monolayer devices are extremely narrow, two-terminal device operation will be extremely
important to future advances in nanoelectronics without complicated nanofabrication. The simplification of logic gates afforded
through using the discrete room-temperature molecular energy
levels of properly designed single moleculemonolayer devices is
likely to fuel novel nanoelectronics advances.

1. Aviram, A. & Ratner, M. A. (1974) Chem. Phys. Lett. 29, 277283.

2. Alivisatos, A. P., Barbara, P. F., Castleman, A. W., Chang, J., Dixon, D. A.,
Klein, M. L., McLendon, G. L., Miller, J. S., Ratner, M. A., Rossky, P. J., et al.
(1998) Adv. Mater. 10, 12971336.
3. Kagan, C. R., Mitzi, D. B. & Dimitrakopoulos, C. D. (1999) Science 286,
945947.
4. Reed, M. A., Zhou, C., Muller, C. J., Brugin, T. P. & Tour, J. M. (1997) Science
278, 252254.
5. Klein, D. L., McEuen, P. L., Katari, J. E. B., Roth, R. & Alivisatos, A. P. (1996)
Appl. Phys. Lett. 68, 25742576.
6. Chen, J., Wang, W., Reed, M. A., Rawlett, A. M., Price, D. W. & Tour, J. M.
(2000) Appl. Phys. Lett. 77, 12241226.
7. Collier, C. P., Wong, E. W., Belohradsky, M., Raymo, F. M., Stoddart, J. F.,
Kuekes, P. J., Williams, R. S. & Heath, J. R. (1999) Science 285, 391394.
8. Schon, J. H. & Bao, Z. (2002) Appl. Phys. Lett. 80, 847849.
9. Liang, W., Shores, M. P., Bockrath, M., Long, J. R. & Park, H. (2002) Nature
417, 725729.
10. Park, J., Pasupathy, A. N., Goldsmith, J. I., Chang, C., Yaish, Y., Petta, J. R.,
Rinkoski, M., Sethna, J. P., Abrun
a, H. D., McEuen, P. L. & Ralph, D. C.
(2002) Nature 417, 722725.
11. Huang, Y., Duan, X. F., Cui, Y., Lauhon, L. J., Kim, K. H. & Lieber, C. M.
(2001) Science 294, 13131317.
12. Huang, Y., Duan, X. F., Wei, Q. Q. & Lieber, C. M. (2001) Science 291, 630633.
13. Duan, X. F., Huang, Y., Cui, Y., Wang, J. F. & Lieber, C. M. (2001) Nature 409,
6669.
14. Rochefort, A., Ventra, M. D. & Avouris, P. (2001) Appl. Phys. Lett. 78, 25212523.
15. Bachtold, A., Hadley, P., Nakanishi, T. & Dekker, C. (2001) Science 294,
13171320.

16. Esaki, L. (1958) Phys. Rev. 109, 603604.

17. Brown, E. R., So
derstro
m, J. R., Parker, C. D., Mahoney, L. J., Molvar, K. M.
& McGill, T. C. (1991) Appl. Phys. Lett. 58, 22912293.
18. Huber, J. L., Chen, J., McCormack, J. A., Zhou, C. & Reed, M. A. (1997) IEEE
Trans. Electron Dev. 44, 21492153.
19. Peyser, L. A., Vinson, A. E., Bartko, A. P. & Dickson, R. M. (2001) Science 291,
103106.
20. Lee, T.-H., Gonzalez, J. I. & Dickson, R. M. (2002) Proc. Natl. Acad. Sci. USA
99, 1027210275.
21. Park, H., Lim, A. K. L., Alivisatos, A. P., Park, J. & McEuen, P. L. (1999) Appl.
Phys. Lett. 75, 301303.
22. Schimschak, M. & Krug, J. (1998) Phys. Rev. Lett. 80, 16741677.
23. Kittel, C. (1996) Introduction to Solid-State Physics (Wiley, New York).
24. Tjeng, L. H., Meinders, M. B. J., Elp, J. V., Ghijsen, J., Sawatzky, G. A. &
Johnson, R. L. (1990) Phys. Rev. B 41, 31903198.
25. Alameddin, G., Hunter, J., Cameron, D. & Kappes, M. M. (1992) Chem. Phys.
Lett. 192, 122128.
26. Bonacic-Koutecky, V., Pittner, J., Boiron, M. & Fantucci, P. (1999) J. Chem.
Phys. 110, 38763886.
27. Bonacic-Koutecky, V., Veyret, V. & Mitric, R. (2001) J. Chem. Phys. 115,
1045010460.
28. Dickson, R. M., Cubitt, A. B., Tsien, R. Y. & Moerner, W. E. (1997) Nature
388, 355358.
29. Lu, H. P. & Xie, X. S. (1997) Nature 385, 143146.
30. Nirmal, M., Dabbousi, B. O., Bawendi, M. G., Macklin, J. J., Trautman, J. K.,
Harris, T. D. & Brus, L. E. (1996) Nature 383, 802804.
31. Bartko, A. P. & Dickson, R. M. (1999) J. Phys. Chem. B 103, 1123711241.

3046 www.pnas.orgcgidoi10.1073pnas.0635474100

We greatly benefited from stimulating discussions with J. Janata. We

gratefully acknowledge funding from the Dreyfus and Sloan Foundations
and the National Science Foundation Career Program (CHE-9984507).

Lee and Dickson