Characterization of the Semiconductor Electrolyte Interface of p-GaAs

curves and the Mott-Schottky plots obtained in the light indicates a negative potential shift in the position of the band edges upon illumination. The...
2 downloads 0 Views 656KB Size
J. Phys. Chem. 1996, 100, 5509-5515

5509

Characterization of the Semiconductor Electrolyte Interface of p-GaAs/GaInP2 Electrodes in Acetonitrile Solutions David K. Watts and Carl A. Koval* Department of Chemistry and Biochemistry, UniVersity of Colorado, Boulder, Colorado 80309-0215 ReceiVed: NoVember 17, 1995; In Final Form: January 22, 1996X

The interfacial properties at a rotated p-GaAs/GaInP2 electrode were investigated in acetonitrile solutions of metallocene redox couples. Current-voltage and interfacial capacitance measurements display minimal band edge movement with changes in solution potential, indicating a well-passivated surface with few surface states within 1 eV above the valence band edge. However, analysis of the shape of the current-voltage curves and the Mott-Schottky plots obtained in the light indicates a negative potential shift in the position of the band edges upon illumination. The extent of band edge movement depends on the concentration of the oxidized form of the redox couple in solution. Greater band edge movement upon illumination was observed at low concentrations, while less movement was observed at higher concentrations, indicating the accumulation of photogenerated electrons. Possibilities for the existence of the accumulated charge in states just below the conduction band are discussed.

Introduction GaAs potentially offers several advantages over Si in solar conversion devices, such as a higher theoretical conversion efficiency, enhanced radiation resistance, and a higher temperature of operation.1 However, the efficient performance of GaAs conversion devices requires some sort of surface treatment to (a) minimize surface recombination, (b) enhance minority carrier lifetimes, (c) increase the rate of electron transfer, and/or (d) minimize photocorrosion at the surface. Modification of the GaAs surface in photoelectrochemical cells (PECs) has involved metal ion complexation techniques and the use of high concentrations of chalcogenide redox couples.2,3 One of the best understood and most efficient PECs is the n-GaAs/KOHSe2--Se22- system.4 In solid-state conversion devices, AlGaAs has been shown to provide nearly perfectly passivated GaAs surfaces;5,6 however, this surface is highly prone to oxidation which adversely affects the long-term performance of such devices. GaAs/GaInP interfaces can potentially provide more stable surfaces than GaAs/AlGaAs junctions and have been shown to exhibit even lower recombination velocities than comparable GaAs/AlGaAs interfaces.7,8 High-efficiency (14%) GaAs solar cells utilizing a thin GaInP passivating layer have been observed.9 More recently, the first photoelectrochemical cells utilizing p-type GaAs electrodes passivated by a thin (3050 Å) GaInP layer were reported to exhibit near-perfect passivation of surface energy levels (surface states) with low surface recombination velocities, leading to nearly ideal kinetic behavior.10 However, there is very little reported in the literature on the photoelectrochemical behavior of GaInP-capped GaAs electrodes as a probe of the nature of the semiconductor/ electrolyte interface (SEI). An understanding of the interfacial energetics at SEIs is crucial to the interpretation of results from photoelectrochemical studies of electron transfer kinetics. In this study, we investigate the photoelectrochemical behavior of an 80 Å Ga0.52In0.48Pcapped p-GaAs, rotating disk electrode in acetonitrile solutions containing metallocene redox couples with reduction potentials ranging between -0.3 and -1.2 V versus a Ag/0.1 M AgNO3 reference electrode. Metallocenes are known to be nonadsorbing X

Abstract published in AdVance ACS Abstracts, March 1, 1996.

0022-3654/96/20100-5509$12.00/0

and to exhibit outer-sphere electron transfer at metal and semiconductor electrodes.11 By varying the solution reduction potential and the concentration of reduced and oxidized redox species and studying the effect on the band edge positions in the dark and under illumination, insight can be gained into the energetics of this SEI. Experimental Section All photoelectrochemistry was done in a Vacuum Atmospheres helium-filled drybox. UV-grade acetonitrile (BurdickJackson) and supporting electrolyte tetrabutylammonium fluoroborate, TBAF (Southwestern Analytical), were purified as described previously.12 All solutions used were 0.1 M in TBAF. Decamethylyferrocene, DFER0 (Strem), was further purified by sublimation. Cobaltocenium hexafluorophosphate, COB+ (Strem), was recrystallized from ethanol and acetone. Dicarbomethoxycobaltocenium hexafluorophosphate, DCMC+, was synthesized as described by Rausch and Sheats13 and then further purified by recrystallization from methanol and acetone. The reduced form of the two cobaltocenes, DCMC0 and COB0, and the oxidized form of decamethylferrocene, DFER+, were produced by bulk electrolysis at a platinum mesh electrode. The novel semiconductor material used in this study was generously provided by the research group headed by Dr. A. J. Nozik at the National Renewable Energy Laboratory (NREL). The wafers consisted of a 0.5 µm layer of p-GaAs (Zn doped, 4 × 1017 cm-3) grown by MOCVD onto a heavily doped p-GaAs substrate, followed by a thin 80 Å Ga0.52In0.48P epilayer. This material will henceforth be referred to as p-GaAs/GaInP2. Two wafers of ∼6 mm × ∼6 mm were provided and used to make two rotating disk electrodes. The two semiconductor rotating disk electrodes were made by machining each semiconductor wafer to fit the disk assembly of a rotating Pt ring-removable disk electrode (Pine Instrument Co., Model AFDTI-36PTPTT) as described previously.12 The surface of each semiconductor electrode was protected with Apiezon W wax during the machining process. This was removed by dissolution in dichloromethane prior to placing the electrodes in the drybox. Both electrodes were made with a geometric area of 0.17 cm2. These two electrodes will be referred to as electrode 1 and electrode 2. Controlled rotation © 1996 American Chemical Society

5510 J. Phys. Chem., Vol. 100, No. 13, 1996

Watts and Koval The photocurrent and dark current were measured as a function of potential. The equation for the diffusion limited current at a rotated disk electrode is17

ID,1,c ) 0.620nFπr12D2/3ν-1/6ω1/2Cox

(2)

where r1 is the radius of the disk. By relating eqs 1 and 2

ID,1,c ) r12/(r33 - r23)2/3IR,1,c

Figure 1. Rotated Pt ring limiting current versus concentration of cobaltocenium hexafluorophosphate. The slope was used to determine all reported solution concentrations from the ring limiting currents.

of the electrodes was accomplished using a Pine Instrument Co. rotator (ASR2) with speed control. Illumination of the disk electrode was accomplished with a Melles-Griot, 5 mW, 632.8 nm, He-Ne laser. The GaInP2 layer was optically transparent at this wavelength of light, so light absorption predominantly occurred in the 0.5 µm p-GaAs layer. Controlled chopping of the laser light was achieved with a shutter and driver (Vincent Associates, Uniblitz, Model T132). The laser was defocused to an area larger than the disk electrode. Light intensity was adjusted with two Oriel neutral-density filters (nominal density 1.0 and 0.6) to ensure a photon-limited current. The light intensity was measured using a power meter (Newport, 1815C) and detector (818-SL). The detector head was equipped with a mask, in the center of which was a hole the same size as the electrode. Using an absorption coefficient for GaAs of 3 µm-1 14 and a refractive index of 3.65,15 the light intensity was then corrected for a 21.5% loss due to reflection at the semiconductor/electrolyte interface.16 The calculated internal quantum yield at the maximum photocurrents in all solutions was 100 ( 10%. The minimum quantum yield with no correction due to reflection was 75 ( 10%. It must be noted that while these electrodes provide significantly greater stability over comparable AlGaAs-capped GaAs electrodes, the surface was still subject to oxidation at anodic potentials, changing the current-voltage behavior and possibly the nature of the surface. Prior to inserting the disk electrode, the platinum ring electrode was polished on a Buehler Ltd. polisher sequentially with 5-, 1-, 0.3-, and 0.05-µm alumina particles. The Pt electrode was used to determine the solution potentials and concentrations. The solution potential, Vredox, was taken as the potential of zero current. All potentials are reported versus a Ag/0.01 M AgNO3, 0.1 M TBAF reference electrode (-0.10 V vs ferrocene). In a rotated ring voltammogram, the limiting cathodic current, IR,l,c, is expected to be proportional to the concentration of the oxidized form of a redox couple in solution, Cox, by the equation17

IR,1,c ) 0.620nFπ(r33 - r23)2/3D2/3ν-1/6ω1/2Cox

(1)

where r2 and r3 are the inner and outer radii of the Pt ring, respectively, D is the diffusion coefficient of the oxidized species, ν is the kinematic viscosity, and ω is the rotation frequency. Four solutions of known COB+ concentration were made, and the limiting currents at a rotation rate of 1000 rpm were plotted versus concentration (Figure 1). The slope of 289 µA/mM was used to determine all reported solution concentrations from the ring-limiting current obtained at 1000 rpm.

(3)

the diffusion-limited cathodic current at the rotated semiconductor disk electrode of known radius can be calculated for all solutions from the diffusion-limited current at the Pt ring. In all cases, the photocurrent observed at the semiconductor disk was less than the calculated diffusion limited current by more than a factor of 2. The impedance of the semiconductor/solution interface was measured as a function of potential in various redox solutions. Measurements were carried out at a frequency of 5000 Hz using a PAR 179 potentiostat, a PAR 5204 lock-in analyzer, and a User Friendly 286 computer with a Data Translations 2801-A analogue and digital board. The data acquisition and reduction software was provided by John Turner at NREL. The impedance measurements were analyzed assuming the Helmholtz double-layer capacitance was much larger than the space charge capacitance, and the Mott-Schottky equation was applicable:18

1/Csc2 ) (2/0qNA)(V - Vfb - kT/q)

(4)

where  is the relative dielectric constant, 0 is the permittivity of free space, q is the electronic charge, NA is the doping density, V is the applied potential, and Vfb is the flatband potential. From a plot of 1/Csc2 versus V, the doping density can be calculated using a dielectric constant of 13.1.14 Vfb can be obtained from the intercept of the potential axis. The energies of the valence band, Ev, and conduction band, Ec, can then be assigned using the equations

Ev ) EF - kT ln(Nv/NA)

(5)

Ec ) Ev + Eg

(6)

where EF is the Fermi level energy given by qVfb, Nv is the density of states in the valence band, 7.0 × 1018 cm-3,19 and Eg is the bandgap energy of GaAs, 1.42 eV. Results and Discussion Dark Studies. Impedance measurements on electrode 1 were obtained in the dark in a blank acetonitrile, electrolyte solution. The Mott-Schottky plot gave a flatband potential of -0.18 V. This flatband potential was consistent with those reported previously for p-type GaAs in nonaqueous solutions.19,20 In solutions containing three poised redox couples, COB+/0, DCMC+/0 and DFER+/0, with potentials of -1.11, -0.60, and -0.35 V, respectively, impedance measurements were conducted on electrode 1 both in the dark and light. Typical MottSchottky plots are shown in Figure 2. Plots obtained in the dark were linear and showed no hysteresis over a range of more than 0.5 V. Average flatband potentials in the dark for each solution were -0.30, -0.20, and -0.18 ( 0.02 V, respectively. These flatband potentials were used to determine the conduction and valence band edges of the semiconductor in the dark. The band edge positions in acetonitrile solutions containing electrolyte only, DFER+/0, and DCMC+/0 were virtually identical within experimental error, indicating no Fermi level pinning within this range of solution redox potentials.

Semiconductor Electrolyte Interface of p-GaAs/GaInP2

J. Phys. Chem., Vol. 100, No. 13, 1996 5511

Figure 3. Depiction of the semiconductor electrolyte interface in the dark and under illumination. The 80 Å GaInP2 epilayer has been omitted for clarity. Energies based on solution potential of the three metallocens are shown as Eredox. The most probable energies for the oxidized state of each metallocene are also shown.

Figure 2. Mott-Schottky plots of p-GaAs/GaInP2 in solutions of (a) decamethylferrocene+/0, (b) dicarbomethoxycobaltocene+/0, and (c) cobaltocene+/0.

Fermi level pinning refers to the movement of the semiconductor band edges with changing solution potential caused by surface states acting as sources of sinks for storing charge.3,21 This causes the electric field within the space charge region at equilibrium to become independent of solution potential. In photoelectrochemical cells, Fermi level pinning causes the opencircuit photovoltage, Voc, to be independent of solution potential. Complete Fermi level pinning may be observed with less than 1% of a monolayer of surface states, 1013 cm-2.3 Fermi level pinning at p-GaAs electrodes has been used to explain the saturation of the onset photovoltages in THF solutions of various redox potentials.19 It has also been used to justify the observation of photoeffects in solutions containing redox couples with potentials lying above the conduction band edge, for experiments in which the band edge position was determined by impedance measurements in acetonitrile solutions containing electrolyte only.20 Fermi level pinning has also been observed by Cachet and co-workers for n-GaAs in various redox couples.22 The authors demonstrated minimal band edge movement for couples located close to the valence band and deduced that there were no surface states within 0.3 eV of the valence band. However, for more negative redox couples, shifts in flatband potential

comparable to the corresponding shift in the solution potential were observed, indicating Fermi level pinning. The virtually unchanged band edge positions in solutions of electrolyte only, DFER+/0, and DCMC+/0 indicate a surface free of surface states within 0.5 eV of the valence band. In the more negative COB+/0 couple, only a 0.1 V further shift in flatband potential was observed for a corresponding 0.51 V change in solution potential. This indicates that there is partial Fermi level pinning with very few surface states within 1 eV of the valence band. Thus the GaInP2 layer does indeed appear to provide significant passivation of the GaAs surface. A depiction of the semiconductor solution interface, based on the model proposed by Gerischer,23 is shown in Figure 3 (the GaInP2 cap has been left out for clarity). The reduction potentials of the three redox solutions are shown. The DFER+/0 solution is poised at a potential significantly closer to the valence band than DCMC+/0 and COB+/0. This allows for the injection of holes into the valence band from DFER+ states in solution as depicted in Figure 3. As a result, significantly more cathodic dark current is expected in solutions of DFER+/0 at potentials negative of the solution redox potential than would be expected for DCMC+/0 and COB+/0 solutions. This is borne out by the current-voltage curves obtained in the three solutions mentioned above (Figure 4). Illumination. The photocurrent-voltage curves shown in Figure 4 are also consistent with minimal Fermi level pinning when the solution potential is changed. It was difficult to determine the open-circuit photovoltage, Voc, at the low light intensities used here, and the nonideal shape of the photocurrent-voltage curves (discussed below) made it difficult to estimate Voc using the correction to ∆E1/2 based on the theory of Santangelo et al.24 However, we can estimate the onset

5512 J. Phys. Chem., Vol. 100, No. 13, 1996

Watts and Koval

Figure 5. Photovoltage versus solution redox potential. A slope ) 1 (dashed line) is indicative of no Fermi level pinning, and a slope ) 0 would be observed for complete Fermi level pinning.

Figure 4. Photocurrent and dark current at p-GaAs/GaInP2 versus applied potential in (a) decamethylferrocene+/0, (b) dicarbomethoxycobaltocene+/0, and (c) cobaltocene+/0.

photovoltage, Vph, from the current-voltage curves by

Vph ) Von - Vredox

(7)

where Von is the voltage at which photocurrent is first observed. This qualitative method has been used before.19 Figure 5 shows Vph as a function of Vredox. A line of slope ) 1 (shown for comparison) would be expected if no Fermi level pinning (no band edge movement) occurs, and a slope ) 0 is expected for complete Fermi level pinning. In this case, a slope of 0.9 is observed for the change from DFER to DCMC, and 0.7 from DCMC to COB. These results are consistent with the extent of band edge movement observed from the impedance measurement results reported above. Illumination of the electrode provided some interesting observations in the impedance measurements. The MottSchottky plots in the light gave two distinct slopes (Figure 2). The capacitance at the most positive potential is similar in both the dark and the light. However, as the potential is swept negative, the illuminated curves display smaller slopes than are found in the dark. Eventually, at still more negative potentials, the plots in the light parallel the curves obtained in the dark. This behavior is consistent with a negative shift in flatband potential upon illumination. The negative shift only occurs at potentials where significant photocurrent is observed. Hence

at potentials positive of the photocurrent onset (see Figure 4), the capacitance in the dark and light are similar. However, at more negative potentials, when the electric field in the space charge region is great enough to allow for a significant flow of electrons to the surface, there can be a significant shift in the position of the band edges of the semiconductor. This shift is most pronounced in the COB+/0 solution and least evident in the DFER+/0. The band edge movement upon illumination is likely a result of the accumulation of photogenerated negative charge at the interface. As discussed above, this trapped charge must be located in states with energies greater than 1 eV above the valence band or in the conduction band. The solutions each have a similar concentration of oxidized states in solution (∼1 mM). However, assuming a reorganization energy for these metallocenes of about 1 eV,25-27 the Gaussian distribution of the oxidized states18,28 can be used to determine the most probable electronic energy level for each species (shown as EDFER+, EDCMC+, and ECOB+ in Figure 3). EDFER+ is located just below the conduction band edge, i.e., nearest to the likely energies of the photogenerated minority carriers. This gives the DFER+ species a higher “effective concentration” and allows carriers to react at a more rapid rate than they are generated. Consequently, the band edge movement is minimal in the DFER+/0 and greatest in the COB+/0 for similar concentrations of oxidized species. The effect of concentration on the band edge movement can be studied by varying the concentration of DFER+, as shown in Figure 6. At a relatively low concentration of the oxidized form, it is observed upon chopping the light source that there is an increasing cathodic dark current induced by illumination (Figure 6a). As mentioned above, the DFER+/0 redox potential is close to the valence band of the semiconductor, allowing for hole injection from vacant DFER+ states in solution into the valence band. Hole injection would be expected to increase as the band edges shift negative under illumination, making the valence band more accessible to the vacant DFER+ states in solution (Figure 3). Addition of the reduced form (DFER0) does not significantly affect this photoinduced dark current (Figure 6b), although there is an interesting “crossover” in the photocurrent on the return scan. This can be attributed to the p-type semiconductor going into accumulation at a more negative potential due to the photoinduced band edge movement. The factor of 10 higher concentration in DFER0 allows for a significantly greater anodic current in Figure 6b than in Figure

Semiconductor Electrolyte Interface of p-GaAs/GaInP2

J. Phys. Chem., Vol. 100, No. 13, 1996 5513

Figure 6. Dark current, photocurrent, and chopped photocurrent versus applied potential at p-GaAs/GaInP2 in decamethylferrocene+/0 with three different ratios of Ox/Red.

6a. At more positive potentials, the photocurrent decreases, and the band edges shift positive again, so the current returns to its original position. Addition of the oxidized form of the couple, DFER+, removes the photocurrent crossover, and the photoinduced dark current is also decreased. Upon chopping of the light, the current begins to return to the baseline dark current as the concentration of DFER+ approaches 1 mM (Figure 6c). This indicates that as the solution concentration of DFER+ is increased, carriers begin to react at a rate which is competitive with the rate at which they are generated. It should be noted that at low DFER+ concentration, when the light is turned off, the photoinduced dark current decays slowly (a few seconds) back to the dark current baseline. This photoinduced dark current is not observed in the DCMC+/0 or COB+/0 couples because these couples have potentials significantly negative of the valence band, and thus a small shift in the band edge position does not allow for significantly greater hole injection into the valence band. Instead, the negative shift in the band edge positions causes the electric field of the space charge region for these couples to become smaller. Consequently there is less charge separation of electron hole pairs, so the photocurrent decreases and hysteresis appears as the band edges shift negative (Figure 4b,c). Using electrode 2, the effect of the COB+ concentration on the shape of the photocurrent-voltage curve was studied. Increas-

Figure 7. Photocurrent and dark current versus applied potential at p-GaAs/GaInP2 in four concentrations of cobaltocenium hexafluorophosphate.

ing the concentration caused a decrease in photocurrent hysteresis, as shown in Figure 7. We can consider the possibility that mass transport effects account for the shapes of the photocurrent-voltage curves by examining the theoretical steady-state voltammetry at an ideal

5514 J. Phys. Chem., Vol. 100, No. 13, 1996

Watts and Koval

Figure 8. Mott-Schottky plots of p-GaAs/GaInP2 in the dark and under illumination immersed in four solutions with different concentrations of cobaltocenium hexafluorophosphate.

TABLE 1: Data for θ and |EP3/4 - EP1/4| at Various Concentrations of COB+ [COB+] (mM)

θ

|EP3/4 - EP1/4|a (mV)

0.6 1.7 2.9 5.9

4 10 18 34

240 (60) 146 (58) 126 (57) 118 (57)

a Values in parentheses are those expected based on theory by Santangelo et al.24

rotating semiconductor disk as reported by Santangelo et al.24 The shape of simulated current-voltage curves were shown to depend on the parameter θ:

θ ) Ipm/(IL + I0)

(8)

where Ipm is the diffusion-limited current at a Pt disk electrode of comparable area to the semiconductor, IL is the photocurrent, and I0 is the reverse saturation current at the semiconductor disk electrode. Table 1 shows the θ values corresponding to our experimental conditions. For a solution containing 1 mM of oxidant at 1000 rpm, Santangelo et al. calculate the parameter, |EP3/4 - EP1/4|, to approach that of a metal electrode, 56 mV, at these large θ values. However, the values of |EP3/4 - EP1/4| obtained here on the forward scan are significantly larger (Table 1). At 0.6 mM of COB+, |EP3/4 - EP1/4| is 240 mV, a factor of 4 greater than the theory predicts. As the concentration of the oxidized form of the couple is increased, |EP3/4 - EP1/4| decreases to 118 mV at 5.9 mM, still a little over a factor of 2 greater than the theoretical value. These large values of |EP3/4 - EP1/4| and the shapes of the photocurrent-voltage curves are

not consistent with theoretical predictions at the acceptor concentrations and light intensity used, so mass-transport effects cannot account for the anomalous behavior in the currentvoltage curves. Concurrent impedance measurements show that band edge movement explains the photocurrent-voltage results. Electrode 2 had a more negative flatband potential than electrode 1; nevertheless, as expected, the negative shift in the flatband potential upon illumination becomes less with increasing concentration of the oxidized form of the redox couple. Figure 8 shows that the change in flatband potential, ∆Vfb, is 150 mV at a concentration of COB+ (Cox) of 0.6 mM (Figure 8a), whereas at Cox ) 5.9 mM, ∆Vfb is only 70 mV (Figure 8d). Again, at the higher concentrations of acceptor states in solution the band edge shift becomes less. Conclusions The studies on the variation of solution potential within the GaAs bandgap indicate that the GaInP2 surface layer does appear to minimize surface states in the middle and lower regions of the semiconductor bandgap. However, band edge movement under illumination indicates trapping of photogenerated electrons. Figure 9 depicts two possibilities for conduction band electron traps. There may exist high-energy surface states, predominantly within 0.4-0.5 eV of the conduction band, that are filled by photogenerated minority carriers (electrons) at a rate jt and empty at a rate, jt′, dependent on the concentration of acceptor species in solutions (Figure 9a). Using time-resolved photoluminescence (TRPL), Nozik and co-workers have shown that the GaInP2-passivated surfaces exhibit very low surface

Semiconductor Electrolyte Interface of p-GaAs/GaInP2

J. Phys. Chem., Vol. 100, No. 13, 1996 5515 was funded by the Department of Energy (BES) Solar Photochemistry Program, DE-FG03-94ER14434. References and Notes

Figure 9. Possible mechanisms for electron transfer and surface charging leading to band edge movement at the p-GaAs/GaInP2/ electrolyte interface.

recombination velocities (SRV < 200 cm/s).10 These results indicate that the GaAs/GaInP2 junction is almost completely free of defect sites that may contribute to surface recombination. Another possibility is the existence of an intermediate layer of GaInAsP at the GaAs/GaInP2 interface (Figure 9b). Evidence for such an intermediate layer has been seen before.29 GaInAsP has a smaller bandgap than GaInP2, so photogenerated electrons may be trapped in this layer since the GaInP2 may provide a significant barrier to electron transfer. These results have important implications about the use of these electrodes in studying the kinetics of electron transfer to solution. At an ideal semiconductor electrolyte interface, electron transfer occurs at the conduction band edge with a rate jet. However, the presence of interfacial trap sites such as surface states allows for trap-site-mediated electron transfer (jt and jt′), and their role must be taken into account in any kinetic study.30 For example, previous studies in this laboratory on majority carrier electron transfer at n-WSe2 have shown that even a low density of surface states, 1012 cm-2, can have a dominating contribution to the mechanism of electron transfer to acceptor states in solution.31 Acknowledgment. We would like to thank Dr. Art Nozik, Brad Thacker, and the rest of the Nozik group for providing the semiconductor material and for many helpful discussions. We also thank Dr. John Turner for writing the software for impedance measurement acquisition and analysis. This work

(1) Hovel, H. J. Solar Cells; Academic Press: New York, 1975. (2) Ryba, G. N.; Kenyon, C. N.; Lewis, N. S. J. Phys. Chem. 1993, 97, 13814. (3) Tan, M. X.; Laibinis, P. E.; Nguyen, S. T.; Kesselman, J. M.; Stanton, C. E.; Lewis, N. S. Prog. Inorg. Chem. 1994, 41, 21. (4) Tufts, B. J.; Abrahams, I. L.; Casagrande, L. G.; Lewis, N. S. J. Phys. Chem. 1989, 93, 3260. (5) Melonkamp, L. W.; van’t Blik, H. F. J. J. Appl. Phys. 1988, 64, 4253. (6) Yablonovitch, E.; Bhat, R.; Harbison, J. P.; Logan, R. A. Appl. Phys. Lett. 1987, 50, 1197. (7) Olson, J. M.; Ahrenkiel, R. K.; Dunlavy, D. J.; Keyes, B.; Kibbler, A. E. Appl. Phys. Lett. 1989, 55, 1208. (8) Harris, J. H.; Sugai, S.; Nurmikko, A. V. Appl. Phys. Lett. 1982, 40, 885. (9) Olsen, G. H.; Ettenberg, M.; D’Aiello, R. V. Appl. Phys. Lett. 1978, 33, 606. (10) Nozik, A. J.; Thacker, B. R.; Bertness, K.; Rosenwaks, Y. J. Phys. Chem. 1995, 99, 7871. (11) Strelets, V. V. Coord. Chem. ReV. 1992, 114, 1. (12) Torres, R.; Koval, C. A. J. Am. Chem. Soc. 1993, 115, 8368. (13) Rausch, M. D.; Sheats, J. E. J. Org. Chem. 1970, 35, 3245. (14) Sze, S. M. The Physics of Semiconductor DeVices; John Wiley and Sons: New York, 1981. (15) Swaminathan, V.; Macrander, A. T. Material Aspects of GaAs and InP Based Structures; Prentice Hall: Englewood Cliffs, NJ, 1991. (16) Koval, C. A.; Segar, P. R. J. Am. Chem. Soc. 1989, 111, 2004. (17) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and Applications; John Wiley & Sons: New York, 1980. (18) Morrison, S. R. Electrochemistry at Semiconductor and Oxidized Metal Electrodes; Plenum: New York, 1980. (19) DiQuarto, F.; Bard, A. J. J. Electroanal. Chem. 1981, 127, 43. (20) Kohl, P. A.; Bard, A. J. J. Electrochem. Soc. 1979, 126, 59. (21) Bard, A. J.; Bocarsly, A. B.; Fan, F. F.; Walton, E. G.; Wrighton, M. S. J. Am. Chem. Soc. 1980, 102, 3671. (22) Gabouze, N.; F., B.; Gorochov, O.; Cachet, H.; Yao, N. A. J. Electroanal. Chem. 1987, 237, 289. (23) Gerischer, H. AdV. Electrochem. Electrochem. Engr. 1961, 1, 139. (24) Santangelo, P. G.; Miskelly, G. M.; Lewis, N. S. J. Phys. Chem. 1989, 93, 6128. (25) Marcus, R. A. J. Phys. Chem. 1990, 94, 1050. (26) Smith, B. B.; Koval, C. A. J. Electroanal. Chem. 1990, 277, 43. (27) Yang, S. H.; Chan, M. S.; Wahl, A. C. J. Phys. Chem. 1980, 84, 3094. (28) Koval, C. A.; Howard, J. N. Chem. ReV. 1992, 92, 411. (29) Nittono, T.; Sugitani, S.; Hyuga, F. J. Appl. Phys. 1995, 78, 5387. (30) Gerischer, H. J. Phys. Chem. 1991, 95, 1356. (31) Howard, J. N.; Koval, C. A. Anal. Chem. 1994, 66, 4525.

JP953390Y