Ceramic electrodes for photoelectrolytic decomposition of water

CERAMICS

INTERNATIONAL,

Vol.

55

7. n. 2. 1981

Ceramic Electrodes for Photoelectrolytic Decomposition of Water J.M. KOWALSKI Department

of Materials

Science

and Engineering

and H.L. TULLER

Massachusetts

In the last decade a number of semiconducting oxides have been identified which serve as essential components in the photoassisted electrolytic decomposition of water to produce hydrogen fuel. To date, this attractive process still suffers from relatively low solar energy conversion efficiencies. In this paper we outline the operating principles characterizing photoelectrolytic cells, and discuss those materials properties which appear most relevant to their efficient operation. Recent data obtained for BaTiO, and other photoelectrodes are presented which illustrate the importance of materials preparation.

Institute

- INTRODUCTION

The fuel shortage of recent years has stimulated the search for alternate energy sources and means for energy storage. One of the more attractive schemes now being investigated is the photoelectrolytic decomposition of water. In this process, one of the metal electrodes in a standard electrolysis cell is replaced with a semiconductor electrode (most often an oxide). Upon illumination of the semiconductor with the appropriate light source, oxygen gas may be generated at the photo-anode (n-type semiconductor) and hydrogen gas at the cathode without need for an external voltage source. Advantages associated with the use of hydrogen as a fuel and as a raw material for producing chemicals are well known. The source of hydrogen in the form of water is nearly limitless and readily accessible. HI fuel may be be readily stored and shipped, it may be burned to produce heat or power vehicles and perhaps as importantly can be used in conjunction with fuel cells to generate electricity efficiently and without pollution. Generation of H> with conventional electrolysis [requiring applied potentials of > 1.23 volts) however, is prohiitive given the present high cost of electricity. If sunlight can be efficiently harnessed to drive the water splitting reaction hti >

Hi0

l/2

Ol+HI

photoelectrode the combined advantages of the hydrogen economy ‘, solar energy conversion, and normally high conversion efficiencies of electrochemical devices could be utilized. The feasibility of photo-assisted electrolysis of water most directly depends on identifying a semiconducting material which exhibits the necessary physical and chemical characteristics and can also be readily fabricated. We begin our discussion by outlining the operating principles characterizing such a cell and then focus on those materials properties which appear to be most

??

A

tional

major part of this work was presented at the Fourth InternaSt. Vincent, Italy, May 1979. Meeting on Modern Ceramics.

Cambridge,

Ma 02139, USA

relevant for its efficient operation. Only oxide semiconductors are considered here; they constitute the only group of photoelectrodes to maintain their chemical stability upon illumination within an electrolytic bath Finally, we conclude by presenting data recently obtained in our laboratory for BaTiOX photoelectrodes which illustrate the importance of materials preparation on final cell efficiency. 2 - MODEL

1

of Technology,

FOR

CELL

OPERATION

Whenever two substances are brought into intimate contact a redistribution of charge occurs so as to establish a new condition of equilibrium. In particular, if an n-type semiconductor is placed into a liquid electrolyte with Fermi energy lower than that of the semiconductor, electrons will flow into the electrolyte until the two Fermi energies become degenerate. This charge redistribution results in a depletion of majority carriers within the semiconductor in the vicinity of the semiconductor-electrolyte interface. As in common p-n junctions this depletion of charge results in the bending of energy bands so as to block further diffusion of majority carriers towards the interface. Analysis of the protoresponse of semiconductorelectrolyte junctions has shown them to be analagous to simple Schottky barriers I. Consequently, relations derived describing the I-V and C-V characteristics of Schottky barriers can be used to characterize semiconductor-electrolyte itnerfaces and are so used later. Figure 1 illustrates the energy states of the semiconducting photoanode, liquid electrolyte, and metal counter electrode within a typical photo-electrolytic cell in the dark and upon illumination. In operation photons with above-band-gap energies generate an excess of electron-hole pairs in the vicinity of the depletion region by photoexcitation. Because of the built in potential within the depletion region the photoexcited electrons are swept down the potential barrier into the semiconductor bulk and towards the metal catode while holes are swept up the potential barrier towards the semiconductor interface. Without this built-in mechanism for charge carrier separation excess electrons and holes would recombine almost instantly returning the photo energy in the form of thermal and/or radiative energy without contributing to electrolysis. The degree of band bending characterized by the energy difference between the conduction band edge at the surface, EL, and the conduction band edge in the bulk, Erb defined by Ebb= E,, -

Etb

is therefore a measure of how efficient ration in the depletion zone is likely to Under illumination electron and hole are shifted relative to those under thermal (in the dark). The new distributions may

charge sepabe. distributions equilibrium be characte-

J.M.

56 rized

by their

quasi-fermi

energies,

E+,, defined

by

nn* E,, = Ef + kT In

1

+

-

n,

An* El,, = E, -

kT In

1 +

PO

1

where An* and up* are the excess electron and hole concentrations generated during excitation while n, and p” are the equilibrium concentrations. The shift of these quasi-Fermi energies with illumination has two important consequences. First, the upward shift of Et,, under illumination will result in transfer of electrons from the cathode to the HI/H20 level if E,, is sufficiently above this level as indicated in Figure 1. Injection of electrons into the H,/H?O level of the electrolyte results in the reducion of water molecules and the generation of hydrogen gas by the reaction 2H10 + 2e- ---+

Hz? + 20H-

KOWALSKI

and

H.L.

TULLER

dicting the relative stability of various compounds under illumination yet remains to be established. Since solar energy is the major source of energy in the photoelectrolytic decomposition of water, the semiconducting electrode should be capable of absorbing as large a fraction of the incident radiation as possible consistent with the water splitting requirements. As observed earlier, for photoelectrolytic decomposition the Hz/Hz0 and OH-/O2 levels, which lie 1.23 eV apart must be straddled by the semiconductor conduction and valence band edges respectively (see Figure il. Allowing for cell overvoltages and ohmic losses one comes up in this way with a minimum of - 1.5 eV. Two factors combine to determine the maximum band gap which can be tolerated. First, since only 1.23 eV can be stored in the form of hydrogen, the fractional component of incident energy used effectively in water splitting decreases as the photon energy increases. This combined with the fact that the fraction of solar photons with energy > 2.5 eV falls off rapidly places an upper limit on the band gap of Z 2.5 eV for reasonable conversion efficiencies ‘.

It may be noted that the equilibrium Fermi energy in the dark normally lies below the HI/HI0 level and so does not contribute to hydrogen gas evolution. Secondly, minority hole carrier concentrations are greatly enhanced in the depletion layer upon illumination. This results in the sharp drop in Et, illustrated in Figure 1. This now allows electron transfer from the OH-/O, level to the semiconductor valence band resulting in the evolution of oxygen gas by reaction 20H-+2p.

--_)

l/2

Olf+

HzO.

As illustrated in Figure 1, this reaction is assumed to proceed via surface states with energy E, lying in the vicinity of the OH-/Hz level. The importance of surface states for this reaction will be considered in greater detail at a later stage of this paper. Combining the above relations wih the photoexcitation reaction 2hu -+

cH-+2p-e~+02t

+li+

Electrolyte

hktal counter electrode 2H+ t 2e’+

l-lzt

FIGURE 1 . Energy level diagram for a photoelectrochemical cell in the dark and under illumination. Reactions describing evolution of hydrogen and oxygen are indicated at the appropriate electrode.

2p‘ + 2e-

one obtains

for the overall

HI0 +

l/2

reaction

O,? + H$

which describes the decomposition drogen and oxygen gases.

3 - SEMICONDUCTING MENTS

n-type semcaductore!ectr0&

of water

PHOTOELECTRODE

into

hy-

REQUIRE-

Early studies with photoelectrolytic cells attempted to demonstrate that the concept of photo-assisted electrolysis of water was in fact feasible”. Now that a numer of semiconducting oxides have been tested and shown to be capable of dissociating water attention is being focused on the semiconductor properties necessary for efficient solar to chemical conversion. To maintain cell operation, whatever the overall efficiency, the semiconductor must remain stable and inert in the often corrosive electrolytes used in these studies. Although many semiconductors do not measurably dissolve in solution under quiescent conditions, they corrode rapidly upon illumination. This happens to almost all of the III-V semiconductors tested’ and thus makes them unsuitable for this application. On the other hand, most oxides studied to date remain stable even upon illumination and are therefore presently the materials of choice. A general theory capable of pre-

One of the major problems limiting initiation of practical photoelectrolysis is associated with the fact that only semiconducting electrodes wih wide band gaps of 3eV or greater maintain chemical stability during illumination. Efforts are therefore now being directed towards finding corrosion resistant oxides with larger smaller band gap’ and/or means of sensitizing band gap oxides’ so as to enable them to utilize a larger fraction of the solar spectrum. The requirement that the band gap of a semiconductor be greater than 1.5 eV is not in itself sufficient to ensure photoelectrolysis. As indicated earlier the relative alignment of semiconductor and electrolyte levels must also be such so as to insure transfer of carriers between appropriate states without need for externally applied potentials. The energy level diagram given in Figure 1 is a reasonably good representation of the position of levels in an oxide such as SrTiO,. One observes that the OH-/O, level is substantially above the valence band edge and so transfer of electrons from the OH-/O: level to the valence band should at least from an energetic standpoint cause little difficulty. Mavroides. et al.’ have recently demonstrated that charge transfer across the interface from the OH-/O: level to the valence band is assisted by the existence of surface states in SrTiO: -2.2eV above the valence band. These states (E. in Fi-

CERAMIC

ELECTRODES

FOR

PHOTOELECTROLYTIC

photoelectrochemical

Material

properties

sand gap (ev)

Fe203 cu20* cue WO3 ZnO* SnO2 v205 PbO BipOg YFe03* C=203 coo Cd0 bO2 sro CdFep04 PbFel2019 PbqTijWOl3 HwTa207 HwNb207 ZrO2 ?? Tap05 * m205 CaTiO3

* ??*

K-%77mQ23O3** PbTh.

of

5Wr~ 506.5 ??

t

??

selected

Flat band wtential JS. Hz/H20

3.0 3.2 3.3 3.5 2.2 2.1 2.2 2.8 2.2 2.2 1.7 2.7 3.2 3.5 2.75 2.8 2.8 2.6 1.4 0.5 2.3 0.3 5.7 2.3 2.3 2.4 1.8 1.8 5.0 4.0 3.4 3.4 3.2 2.4

TiOp SrTiO3 BaTi KTaO3 FeTiO3* FepTiOq Fe2Ti05 BaTil_xFex03_xF:

DECOMPOSITION

Referenced Referenced

0.05 - .2 .l - .2 1.1 1.2 1.5 0.7 1.0 0.5 0 0.5 1.2 0.45 0.7 0.5 0.8

0.8 1.0 0.4 1.2 1.1 -1.0 -0.4 0 -0.5 -0.4 0.4 in paper in paper

OF WATER

semiconductor

57 oxides.

Remarks stable stable II 1,

Reference 3

in base

leaching of Fe leaching of Fe stable photo response stable at pH > corrodes 11

atoms atoms

11,27 12,13 14 15 15 15 16 17,18 19 17 13,18 20 21 17 17 17 22 17 17 17 17 17 13 13 13 13 13 23 23 23 24

noted noted

quenced 6

stable at low pH corrodes stable corrodes 1, II stable corrodes, poor corrodes, poor no response 0 II

response response

poor response but stable stable at high pH stable at high pH stable II I, I, II II II

I

TABLE I - Photoelectrochemical properties of selected semiconductor oxides.

II

by Maruska and Ghosh by Harris and Wilson

gure 1) being nearly degenerate with the OH-/O1 level result in the high tunneling probability necessary for efficient charge transfer across the interface The situation is not nearly as clear at the upper energy states, given the near coincidence of the conduction band edge and the HI/H20 level. For efficient evolution of hydrogen the degree of band bending must be sufficiently large to insure continued electron-hole separation even under illumination. Taken to its extreme, a flat-band condition, characterized by the elimination of space charge at the interface, results under intense illumination. The higher the flat band energy, Efb, (i.e.. the less negative it is relative to vacuum), the greater is the resultant band bending, Ebb. This insures more efficient electron-hole separation as well as electron transfer from semiconductor to Hz/Hz0 level. When such requirements are not met, for example, as in TiOl, photocurrents do not flow spontaneously upon illumination. Upon application of a small anodic bias, however, sufficient band bending may occur driving the reaction forward. To achieve optimal band bending without bias the difference in work functions between semiconductor and electrolyte should be maximized. For an n-type semiconductor this requires that the work function or its near equivalent in heavily doped materials, the electron affinity, X. be small. Since electron affinities cannot be calculated from first principles one usually compares measured flat band potentials, Vrb, to determine which semiconductors are likely to exhibit improved properties. One may obtain values for both, Vfb, to determine which semiconductors are likely to exhibit improved properties. One may obtain values for both Vfb and the dopant

* Referenced in paper by Maruska and Ghosh (26). ** Referenced in paper by Harris and Wilson (4).

25 13

.(26)

(

4)

density, No, from measurement capacitance, C,,, as a function This relation is given by’ 1 -= C’,‘

,.:

N

of the space charge of applied potential, V.

(I’-V,b-kT/qj

0 D

in which E is the relative dielectric constant. When flat band potentials are reported measured relative to the Hz/H20 level, then negative values imply photoelectrolysis may proceed spontaneously without bias. In Table I we present flatband voltages as well as band gap energies obtained for a number of oxides tested as photoanodes. Few exhibit negative flatband values which would make them capable of photoelectrolysis without external bias. A number of other materials parameters important in determining the overall efficiency of such cells are best examined by consideration of the photocurrent in a Schottky barrier-like interface. Butler and Ginley lo show that the photocurrent density, J, can be related to the depletion layer width, WO, the applied and flat band potentials, V and Vfb, the minority carrier diffusion length, L,, the optical absorption coefficient, a, and the photon flux, %. by the following expression exp I--aW,

I J = q@aJ,-1

I

1 + EL,

I,2 1

The depletion layer across the junction, 2E wo=

__

[

(V -

@Jo

V,,)“‘l

1 1

width, W,. defined given by

for

one

volt

J.M.

58 is seen to be inversely proportional to the donor density, No. To obtain optimum charge separation it is desirable that electron-hole pairs be generated within the depletion layer. It thus follows that the absorption length I/a should be 5 to the depletion layer thickness (W = W0 (V-VV,~)“). For wavelengths for which a is small (near the band edge), larger values of W are desirable. This may be accomplished by doping less heavily with smaller No values. On the other hand, ohmic losses within the semiconductor electrode must be kept at a minimum. Since the electrical conductivity is proportional to ND, this requires that ND be maintained at least at some minimum level. A compromise choice for No must therefore be found to accommodate both requirements. Few studies to date have systematically studied the effects of variations in materials parameters such as crystal purity, composition, choice of donor and microstructure, each of which is likely to effect some of the variables considered above. In the following section we review the results of studies which demonstrate the possible effects such variables may have on photoelectrolytic efficiency. 3.1 - Microstructure One of the major requirements for the establishment of a viable solar conversion device is the ability to produce the photosensitive element in large surface areas cheaply. Even if single crystalline materials with high efficiencies could be identified it is unlikely that could be produced in large scale economically. Perhaps the most attractive features of the photoelectrochemical approach is (a) that the junction need not be fabricted but is produced simply by insertion of the electrodes into the electrolyte and (b) that polycrystalline and thin film electrodes appear to approach the operating efficiencies .of single crystalline materials. For example, comparisons between FelOl photoanodes produced by 1) flame oxidized iron sheet, 2) sintered powders, ” 3) sintered powders doped with TiOl” and 4) single crystalline Fe201 ” show collection efficiencies of < lo%, < lo%, - lo%, and - 18% respectively with 400 nm light and applied bias of + 0.5 V. Similar results have been obtained for TiO, and BaTiO,. Maximum quantum efficiencies of -60% have been reported for single crystalline TiOl”“, as well as for polycrystalline rutile ” and polycrystalline anatase TiOl ‘I. Studies of single ” and polycrystalline ” BaTiOl show the polycrystalline material to have even a higher quantum efficiency than that of the single crystalline material. This surprising result very likely is an artifact of the experiment. Neverheless, the main feature remains, i.e., polycrystalline materials exhibit conversion efficiencies approaching that of single crystals. The similarity between the photoelectrochemical behavior of single and polycrystalline materials suggests grain boundaries do not have a substantial deleterious effect on the minority carriers in these systems. This may be a result of small diffusion lengths of minority carriers (ICL reported for holes in single crystalline TiO, ‘). No careful studies of grain boundary effects have yet been performed which attempt to explain the manner in which the microstructure affects efficiency. 3.2 - Surface

treatment

Often results reported by various investigators show substantial discrepancies concerning collection efficiencies. A recent study by Wilson, et al. ‘A suggests that surface treatment plays an important role in quantum efficiencies. They showed that polishing of the anode surface leads to a decrease in photocurrent. It is suggested that polishing creates a disturbed layer

KOWALBKI

and H.L. T’JLLER

with an increased density of recombination centers. Subsequent annealing of the polished surface results in photocurrents roughly five times larger. Similar results have been reported by Ghosh and Maruska ‘.

3.3 - Dopant

effects

Some groups have attempted to improve photocatalytic behavior of the anode surface by additions of various dopants. Ghosh and Maruska ’ demonstrated that small additions of Al lead to an increase in the hole diffusion length. Reasons for this increase are unclear. Improved efficiencies have also been reported for Be doped TiOl ” but these are suspect since levels of Be added were well beyond the solubility limit of Be in Ti02 “. Furthermore, the suggestion of these authors that Be, should stabilize V; is inconsistent with the donor nature of Be;. In the work of Schleich et al. I6 single crystalline BaTiOJ was simultaneously doped with Fe in an attempt to decrease the bandgap. Although absorption began at lower frequencies, the photoresponse was quenched. Reasons for this behavior are unclear. Other studies” have investigated the affects of various additives but the materials were insufficiently well characterized to be able to make any reasonable conclusions. A related question concerns the nature of the donor states necessary for inducing highly n-type semiconducting properties. Nearly all n-type semiconducting oxides studied as photoelectrodes in photoelectrolytic cells have been renderd semiconducting by reduction in reducing atmospheres (often HZ )at elevated temperatures. The donor states thus generated are usually accepted as being oxygen vacancies in compounds such as Ti02, SrTiOl, BaTiO,, etc. although titanium interstitials are at times also reported in the case of TiOl.‘” Doping with pentavalent impurities (e.g. Ta, Nb), which substitute for Ti) also result in n-type semiconductors but without the accompanying lattice defects (i.e., oxygen vacancies) otained upon reduction. Early photoemission studies of absorbed water on species TiO, surfaces ” suggest that the chemisorbed interact with the surface of the semiconductor via the oxygen atom. It is therefore, reasonable to hypothesize that reduced and doped surfaces may offer different types of sites for chemisorption. We have carried out preliminary measurements on pure-reduced and Nb doped BaTiO, single crystals to test this hypothesis. Nb-doped and undoped BaTiO, single crystal specimens grown by a top seeded solution technique 3( were prepared with resistivites of 50 ohm-cm and 10 ohm-cm, respectively. The undoped BaTiOl electrode was rendered n-type by reduction in H2 at 1100°C for several hours. Measurements of photocurrent versus incident wavelength for both reduced and Nb doped BaTiO, were made in 5M NaOH by conventional techniques’. Both BaTiOl electrodes exhibited appreciable photocurrents with no applied voltage as do SrTiO, electrodes. This may be contrasted with TiOl electrodes which require a bias to operate. Photocurrents were first observed at for each of the specimens wavelengths of -370nm indicating identical bandgaps of -3.3eV for the two specimens. In Figure 2 we plot photocurrent versus applied voltage (corrected to SCE) for both BaTiOJ electrodes. This data was taken using the full output of a 200 watt Hg arc lamp with IR filter in NaOH electrolyte with pH = 14. For the case of Nb doped BaTiO,, positive photocurrents begin at - 1.25 volts while for reduced BaTiOl they begin at - 1.24 volts (vs. SCE). Both values are more negative than previously reported values of - 0.9 volts ” and - 0.95 volts ” (vs. SCE) respectively. The outstanding feature of these curves is the large difference in photoresponse exhibited by the

59

CERAMIC ELECTRODES FOR PHOTOELECTROLYTIC DECOMPOSITION OF WATER

Nbdqcd

BaTQ

FIGURE 2 - Measured photocurrent vs. applied voltage for reduced and Nb doped BaTiO, electrodes in NaOH electrolyte, pH = 14.

two electrodes. The photoresponse of the doped BaTi electrode is roughly an order magnitude greater than that of the reduced electrode over almost the entire range of anodic bias. The large discrepancy between the doped and reduced BaTiO, photoresponse suggests that donor type, i.e., oxygen vacancy versus impurity dopant can play an important role in determining the overall efficiency of a photoelectrochemical cell. Similar results have been reported for Fe-doped TiOl”. Changes in surface states or in minority diffusion lengths may perhaps be responsible for the improved behavior but more detailed studies are necessary to establish reasons for the observed discrepancy.

REFERENCES 1. Proc. of Hydrogen Economy Miami Energy Conference, Miami Beach, Fla. (1974). ed. T. Nehat Veziroglu. N.Y.: Plenum Press. 2. H. GERISCHER. Electroanal. Chem. and Interfac. Electrochem. 58 [1975] 263. 3. A. FUJISHIMA and K. HONDA, Nature 238 (1972) 37. 3-a. MS. WRIGHTON. D.S. GINLEY, P.T. WOLCZANSKI. A.B. ELLIS, D.L. MORSE and A. LINZ, Proc. Nat. Acad. Sci. 72 (19751 1518. 4. L.A. HARRIS and R.H. WILSON, Ann. Rev. Mater. Sci. 8 (19781 99. 5. see reference 4. 6. P.A. KOHL, S.N. FRANK, A.J. BARD, J. Electrochem. Sot. 124 (1977) 225. A.K. Ghosh and H.P. Maruska, Proc. Int. Symp. Solar Energy. ed. J.B. Berkowitz. I.A. Lesk, p. 92-100. Princeton, N.J.: The Electrochem. Sot. 7. A.K. GHOSH and H.P. MARUSKA. J. Electrochem. Sot. 124 [I9771 8.

9. 10.

11.

12. 13. 14. 15. 16. 17. 18.

1516.

J.G. MAVROIDES. V.E. HENRICH. H.J.ZEIGER. G. DRESSELHAUS, J.A. KAFALAS and D.F. KOLESAR. Proc. Symp. Electrode Mat. 8 Processes for Energy Conversion and Storage, ed. J.D.E. MC Intyre. S. Srrmvasan. F.G. Will, p. 45-53. Philadelphia, PA. The Electrochem. Sot. V.A. MYAMLIN and Y.V. PLESKOV. 1967. Electrochemistry of Semiconductors. New York: Plenum, p. 58. M.A. BUTLER and D.S. GI’NLEY. Proc. Symp. Electrode Mat. 8 Processes for Energy Conversion and Storage, ed. J.D.E. MC Intyre. S. Srinivasan. F.G. Will. p. 54-65. Philadelphia, PA: The Electrochem. Sot. M.S. WRIGHTON. A.B. ELLIS, P.T. WOLCZANSKI, D.L. MORSE, H.B. ABRAHAMSON and D.S. GINLEY. J. Am. Chem. Sot. 98 11976) 2774. R.D. NASBY and R.K. QUINN, Mat. Res. Bull. 11 (1976) 985. H.H. KUNG. H.S. JARRETT. A.W. SLEIGHT and A. FERRETTI, J. Appl. Physics 48 (19771 2463. A.B. ELLIS, S.W. KAISER and M.S. WRIGHTON, J. Phys. Chem. 80 (1976) 1325. D.S. GINLEY and M.A. BUTLER, J. Appl. Phys. 48 (19771 2019. D.M. SCHLEICH, C. DERRINGTON, W. GODEK, D. WEISBERG and A. WOLD. Mat. Res. Bull. 12 (19771 321. K.L. HARDEE and A.J. BARD, J. Electrochem, Sot. 124 (1977) 215. R.D. OUINN. R.D NASBY and R.J. BAUGHMAN. Mat. Res Bull 11 Cl9761 1011

3.4 - Surface

modeling

Surface states at semiconducting TiOj/electrolyte Interfaces are believed to play an important role in charge transfer and thereby the efficiency of photoelectrochemical processes at such interfaces. ” In order to test this hypothesis, theoretical calculations were recently performed by the authors using the SCF-X&W method to determine the position and character of surface states at various characteristic interfaces. At the TiO>/water interface, antibonding surface states were found which when occupied would explain the experimentally observed dissociation of water into hydroxyl groups at n-type semiconducting Ti02 surfaces”O. Similarly, antibonding surface states were found at the TiO>/OH- interface which when occupied would tend to destabilize the OH bond. A likely mechanism for the dissociation of water and decomposition of certain photoanodes in photoelectrochemical cells based on the above results was presented. Further such calculations may suggest the role that dopants and degree of reduction may have on the character of such surface states.

ACKNOWLEDGEMENT Support by the Office of Naval Research (Contract # N00014-78-CO3661 is gratefully acknowledged. Single crystalline samples supplied by Dr. A. Linz and assistance by A. Bocarsly in performing cell measurements are greatly appreciated. Helpful discussion with K. H. Johnson, D. Epstein, A. Linz, and M. Wrighton are acknowledged.

19. R. WILLIAMS, J. Vat. Sci. Technol. 13 (1976) 12. J. Electrochem. Sot. 113 (1966) 1174. 20. H. GERISCHER. WRIGHTON. D.L. MORSE, A B. ELLIS, D.S GINLEY and 21. MS. H.B. ABRAHAMSON, J. Am. Chem. Sot. 98 (1976) 44. BUTLER, D.S. GINLEY and M EIBSCHUTZ. J. Appl Phys 22. M.A. 48 (1977) 3070. J.R. MARTIN, R. OLIER and C. VALLOUY. C.R. 23. P. CLECHET. Acad. Sci. Paris 282C (1976) 887. 24 D.I. TCHERNEV. 1976. Int. Conf Photochem. Conv. Storage Solar Energy, Abstr. C3. London, Ontario. and M.S. WRIGHTON. J. Phys. Chem 80 (19761 25. J M. BOLTS 2641. 26. H.P. MARUSKA and A.K. GHOSH, Solar Energy 20 (1976) 443. 27. J.G. MAVROIDES, J.A. KAFALAS and D F. KOLESAR. Appl. Phys. Lett. 28 cl9761 241. 28. J.S. CURRAN and W. GISSLER. J Electrochem. Sot. 126 (19791 56. 29. K.G. MC CREGOR, M. CALVIN and J.W. OTVOS. J. Appl. Phys. 50 (1979) 369. 30. T. OHNISHI. Y. NAKATO and H. TSUBOMURA, Ber Bunsenges Phys. Chem. 79 (1975) 523. 31 C STALDER and J AUGUSTYNSKI, J Electrochem Sot 126 (1979) 2007. 32. J.H. KENNEDY and K W FRESE Jr, .I Electrochem Snc. 123 (1976) 1683. 33. R.H. WILSON, L.A. HARRIS and ME GERSTNER, ; Eleitrti them. Sot. 126 (1979) 844. 34. E.M. LEVIN. CR. ROBBINS and H.F. MC MURDIE, Phase Diagrams for Ceramists. Am. Ger. Sot.. Inc., 1964, p, 101. 35. J.F. HOULIHAN. D.B. ARMITAGE, T. HOOVER, D. BONAOUIST. D.P. MADACSI and L.N. MULAY, Mat. Res. Bull 13 (19781 1205. 36. J.R. AKSE and H.B. WHITEHURST. J. Phys. Chem. Solids 39 (19781 457. 37. W.J. LO, Y.W. CHUNG and G.A. SOMORJAI, Surf, Sci. 71 (1978) 199. 38. V. BELRUSS. J. KALNAJS and A. LINZ. Mat, Res. Bull. 6 (1971) 899. 39 S.N. SUBBARAO, Y.H. YUN, R. KERSHAW, K. DWIGHT and A. WOLD. Mat Res. Bull. 13 (1978) 1461. 40. J.M. KOWALSKI. K.H. JOHNSON and H.L TULLER. J. ElectroChem Sot. 127 (1980) 1969.

Recerved

May

3,

1979,

revised

text

received

August

29,

1960