PROTECTION OF PHOTOANODES AGAINST PHOTO-CORROSION BY SURFACE DEPOSITION OF OXIDE FILMS : CRITERIA FOR CHOOSING THE PROTECTIVE COATING

Two strategies to solve the problem of instability of photoanodes against photocorrosion have been explored. The photocorrosion of photoanodes generally occurs when they enter the fabrication of efficient photoelectrochemical cells (i.e. showing high values of the open circuit voltage and photocurrent density).


INTRODUCTION
Photoelectrochemistry, i.e. the conversion of solar energy with a semiconductor-electrolyte junction, has several advantages over photovoltaic conversion: elimination of expensive manufacturing steps like differential doping, grid deposition.simplicity of the junction, which is just a semiconductor immersed in an electrolyte.
the photopotential, and hence the e.m.f, of the cell, can be varied as it depends upon the choice of the redox couple.
Many papers and recent patents have thus shown that photo-electro- chemical cells can have high conversion yields, particularly the cells using n-semiconducting photoanodes with high carrier mobility and narrow forbidden band energy gap (n-GaAs, nSi, n-CdSeo.6Teo.3.)3l.
There is, however, an obstacle to the production of high-yield cells with high open circuit voltage (electromotrive force)" n-type semi- conducting photoanodes (with high carrier mobility and a narrow forbidden band) are generally chemically unstable when they are illuminated in an aqueous medium.
Various strategies to solve this instability problem have been explored.The most attractive is protection of the semiconductor by a thin film, resistant to the electrolyte and transparent to sunlight.
In this prospect Horowitz and Garnier indeed obtained a very stable (-100 h) photoanode in aqueous Fe(CN)64-/3-, by surface coating n-GaAs with a 400 film of conducting poly(3-methylthiophene).Moreover, the conversion yield was as high as 10% 2. Note however, that one does not know if the "efficient" junction occurs at the GaAs/polythiophen interface (the electrolyte would then behave as a "transparent contact") or if the electrolyte plays a determining role in the formation of an efficient Schottky-type junction with the photoanode2.
On the other hand, poor results (noticeable drop in the conversion yield, rapid "lifting" of the protective coating) were obtained with semiconductor oxide protectants (Table I)3 '4'5.This difference could be attributed to an infortunate choice of the oxides, listed in Table I, which do not have the appropriate electronic properties.However that may be, there is to our knowledge no model bring- ing forward the mechanisms of charge transfer that must occur in the protective film in order to get high conversion yield.
Note, first of all, that broadly speaking an oxide can be amorphous or crystallized, stoichiometric or defect bearing, conducting or non conducting.It should thus be possible to satisfy the criteria set out below for easy carrier transport in the protective film.

CRITERIA FOR CHOOSING PROTECTIVE OXIDE FILMS FOR THE PHOTOANODE
Either insulating or conducting oxides can be envisaged, provided their properties differ from those of the "low-performance" materials listed in Table I.
Protection by an non-conducting oxide: EIS structure (Electrolyte-Insulator-Semiconductor).We first note that the non-conducting oxides listed in Table I have large band gap: Eg(TiO2) 3.0eV Eg(AI203) 4.0eV Eg(Nb2Os)-- 2.9eV (6) (7)   Their valence bands are, therefore, below that of the photo-anode they protect, which have small optical gaps (Eg < 2eV).Figure 1 is an energy diagram of the photoanode-film-electrolyte junction.It clearly shows that the tunnel effect is the only way for a photo- produced hole to cross the protective film and oxidize reduced species in the electrolyte.
If the electrolyte may reasonably be thought to behave as a metal (relative to the non-conducting oxide and to the photoanode), it is then clear that useful EIS structures must have an insulating coating of less than 30A: we indeed recently showed that in metal-insulator- semiconductor (MIS) structures, the insulator thickness for tunneling- charge-transport to occur should not exceed 20/7.Photo-generated holes cannot tunnel through the oxide film if its thickness, e, exceeds 10 .
This was not the case for the heterojunctions listed in Table I, since the films were over 40 thick3 '4'5.Hence it was very difficult for photo-holes to tunnel through them (Figure 1).Confinement of the photo-hole at the photoanodefilm interface is then unavoidable.This either produces a high recombination rate of the photo-carriers, and hence a drop in the performances, or initiates corrosion of the photo- anode leading to lifting of the film3'4'5.
EIS structures should thus have oxide layers of less than 20.This is technologically difficult to perform.Thicker oxide films would be acceptable, however, if carrier confinement could be prevented by making the photo-holes migrate in the valence band of the oxide (Figure 2).It implies that the top of the valence band of the oxide should fall either above or at the same level as the top of the valence band of the photoanode (Figure 2).This seems difficult to fulfill in practice, because we know of no transparent oxide in which the top of the valence band (of anionic character, as it occurs for the oxides listed in Table I) falls high enough.
G. CAMPET et al. redox electrolyte FIGURE 2 An "ideal" EIS solar cell.Photo-generated holes cross the insulator via the valence band.
Note, however, that charge transport through a protective non- conducting film could occur in a valence band, provided we use not the intrinsic anionic valence band of the oxide, but cationic, extrinsic or intrinsic, valence band or valence levels (Figure 3).This opens the field of investigation, since a wide range of oxides could satisfy this condition.
Table II gives some examples of significant oxides which should efficiently protect n-GaAs photoanodes from photocorrosion: the cationic valence levels quoted in Table II are indeed close to the top of the valence band of GaAs (Ev 5.4eV8,9-14).
The Ce3+:4f energy states should indeed be located slightly above the valence band edge of GaAs (Table II) and thereby allow the transfer of the photogenerated holes through the film (Figure 3).
For comparison, the n-GaAs/SrTiO3 heterojunctions sharing undoped SrTiO3 as a coating, were considered.
An r.f.sputtering deposition technique was used to deposit the Sro.gsNao.oCeo.o1TiO3and SrTiO3 films, on the (100) face of the n- GaAs substrate, as it generally gives rise to highly compact films adhering tightly to the substrate.
To obtain reproducible conditions the following process was used (for both undoped and Ce-doped samples): Details concerning the films analysis and texture, the conductivity and photoconductivity measurements will be reported separately.
The non-conducting (o(25oc) 10-1f2-cm-l) deposits were par- tially crystallized but had the expected stoechiometry.Finally, the photoc0nductivity spectra of the films gave evidence of subbandgap CeB+:4f energy levels at about 1.6eV below the conduction band7.
Figure 4 shows the photocurrent as a function of the potential difference across the cells which were used as generators.The significant current observed with the n-GaAs/Sr0.98Nao.olCe0.0TiO3heterojunction confirms the validity of the model.It is probable that greater photocurrents could be obtained by increasing the amount of Ce3+.practically no response.This could be expected since SrTiO3 is not here doped by Ce3/, and it is virtually impossible for photoholes to tunnel through the film (film thickness > 20A (Figure 1)).
The disappointing results obtained by previous authors with the heterojunctions n-GaAs/n-SnO2, n-Si/n-SnO2 and n-CdSe/n-SnO2 (Table I) may be interpreted on the basis provided by Figure 5, i.e., the thickness (W) of the space-charge region could be done eli- minated by protecting the n-GaAs with a degenerate oxide coat having a consequently metallic conductivity.
ITO (90% In203 + 10% SnO2) could be a suitable choice since it is the "perfect" degenerate semiconductor, combining both transpar- ency and metallic conductivity.Unfortunately, we have shown recently that, quite unexpectedly, the metallic conductivity of ITO does not prevent the formation of a space charge region at the ITO/ h Ph ot;oanocle r-SnCI 2 elcgr'olyt; film FIGURE 5 n-GaAs/n-SnO2/electrolyte junction.electrolyte interface7.This space-charge region, which is over 30A thick even for heavily n-type doped films (102-1021 cm-3), is obviously incompatible with efficient carrier tunneling across the ITO/electrolyte interface7.In fact, such a space-charge probably occurs for most of transparent degenerate semiconductors (n+In203, n+T1203... ).
It thus appears that.photoanodes cannot be "efficiently" pro- tected, neither by an ITO coating nor by most of others transparent coatings having metallic conductivity.
Despite these comments, successful protection of photoanodes by a conducting oxide is not improbable.In fact, photogenerated holes at the semiconductor surface should, in principle, cross the protective film to react with the electrolyte, provided that the carrier transport occurs close to the Fermi level, EF, of the film via: --either a partially filled band, or a sufficiently high density of localized states, within which Ev is located.
The oxide should also have a low electron affinity so that an electric field gradient (AE) could be set up, to activate the transport of holes towards the electrolyte (Figure 6).
Oxides such as those below could be suitable: where M 3+ Cr 3+ for example.
The electronic affinity of these titanate oxides which are likely to have the illmenite structure, defined by the energy of the bottom of the Ti:3d(tzg) conduction band, should indeed be small 15.Further- more, the Fermi level should be trapped in the Mn or Co band, as it is partially filled.It follows that efficient photo-hole transport in the film should use this band according to the process shown in Figure 6.
Note that a deposit of n-SrTiO3 could also be suitable because: when it is prepared under suitable conditions (a 40% H2 + 60% Ar mixture can be used as a sputtering atmosphere; the other deposi- tion parameters are identical to those mentioned above for undoped and Ce-doped SrTiO3), the Fermi level can be trapped within the forbidden band by Ti'3d(tzg) energy states s.
We have thus demonstrated the second strategy of protection by depositing n-SrTiO3 on n-GaAs.As expected, we obtained promising results with the following cell: n-GaAs/n-SrTiO3/0.1MIOf/0.1MI-/0.1MKOH/Pthybrid electrode Electrolyte Cathode structure Figure 7 indeed shows conversion yields as high as 18% for an il- lumination of 5 mW/cm2.

FIGURE 6
FIGURE6 Charge transport mechanism in the conducting protective film.