Semiconductor electrodes. 13. Characterization and behavior of n-type zinc oxide, cadmium sulfide, and gallium phosphide electrodes in acetonitrile solutions

The photoelectrochemical behavior of n-ZnO, n-CdS, and n-GaP single crystal semiconductor electrodes was investigated in acetonitrile which contained various electroactive compounds whose standard redox potential varied by over 3.2 V. The cyclic voltammograms of the n-type semiconductors in the dark and illuminated were compared to the Nernstian behavior at a Pt disk electrode. The photodissolution of the semiconductor electrodes did not occur until the electrode potential was well positive of the flat band potential. An underpotential was developed for the photooxidation of solution species at each semiconductor. A model for electron transfer at the semiconductor-solution interface is proposed which incorporates a surface state or intermediate level which can mediate electron transfer and surface recombination. The model is used to discuss previous work with semiconductors, electron transfer, and stabilization of the semiconductor surface. Key factors in the utilization of semiconductor electrodes in electrochemical cells and devices, including those for solar energy conversion, are knowledge of the relative locations of the energy levels in the semiconductor and solution and an understanding of the role that surface states or intermediate levels in the band gap region may play in the charge transfer processes. This information has been obtained by studying the electrochemical properties of the semiconductor electrode in solutions containing different redox couples and assuming that the model of Gerischer' can be used to rationalize the observed current-potential (i-V) behavior. This model basically requires that charge transfer occur isoenergetically from the conduction or valence bands of the semiconductor to solution species. A number of such studies of different semiconductors in aqueous solutions2-I0 have shown that while this model is useful in locating the band positions, it is necessary to invoke intermediate levels or surface states within the band gap region to explain the i -V and spectroscopic behavior. W e have previously' I s i 2 discussed the advantages of using a nonaqueous solvent, such as acetonitrile (ACN), in studying the behavior of the semiconductors Ti02 and Si. These include a much wider potential range (Le., effective solvent band gap) for investigating the semiconductor behavior and the availability of a large number of well-characterized couples which undergo reversible one-electron transfer reactions at potentials throughout this range. Thus the studies on the n-Ti02/ACN interface demonstrated the occurrence of reductions and oxidations of species whose standard potentials were located above the flat-band potential (V,) via the conduction band, and reduction of species, via an intermediate level or surface states, a t potentials positive of Vfi. We should also mention the work of Landsberg et al.13 which was primarily concerned with reductions a t n-GaP in A C N a t potentials negative of Vfi. The work reported here deals with a study of the behavior of n-type ZnO, CdS, and G a P with A C N solutions. The location of the bands, the determination of V,, and the mapping of the band gap region is described and the behavior of these materials (which photodecompose in aqueous solutions) under irradiation is discussed. Finally a model which relates the interfacial charge transfer rate to competitive rates involving the conduction and valence bands and surface states is proposed. Experimental Section Low-resistivity single crystal semiconductors about 1 m m thick and polished with 0.5 1 alumina were used. The n-GaP and n-ZnO were obtained from Atomergic Chemicals (Long Island, N.Y.) and the n-CdS from National Lead. The (OOl), (OOl), and (1 11) faces were used for the n-ZnO, n-CdS, and n-Gap, respectively. Two ohmic contacts were made on each by electrodepositing indium from a 0.1 M InC13 solution on the back of the crystal. The n-GaP was heated in a hydrogen atmosphere for about 1 h following the In deposition at 400 "C to make the contact ohmic. In each case the current measured between the two contacts was directly proportional to the applied voltage and independent of polarity. A copper wire was connected to the ohmic contacts with conducting silver epoxy cement (Allied Products Corp., New Haven, Conn.). The back and sides of the crystal were then covered with insulating 5-min epoxy (Devcon Corp., Danvers, Mass.) and mounted on a small glass disk connected to a glass tube with silicone adhesive (Dow Corning, Midland, Mich.). The glass tube provided an insulated internal path for the wire from the electrode. There was no apparent chemical attack on the silicone adhesive from either the A C N or the etching solutions. The n-ZnO was first etched in H3P04 and then in concentrated HC1, each for 15 s.14 The n-CdS and n-GaP were etched in 11 M HCl for 1 min. Most of the compounds used in this study were obtained from commercial sources. The Ru(TPTZ)*(C104)3 (TPTZ = 2,4,6-tripyridyl-s-triazene) and Ru(bpy)3(C104)2 were prepared as described previ0us1y.l~ Most of the compounds were recrystallized several times from an appropriate solvent and all have been previously characterized in this laboratory. Controlled potential coulometry was used with some compounds to transform them to the desired oxidation state. Polarographic grade tetra-n-butylammonium perchlorate (TBAP), which was dried for 3 days under vacuum, was used as the supporting electrolyte. The ACN was dried and purified as previously described.I6 A cyclic voltammogram was obtained a t a platinum disk electrode a t the beginning of each experiment to ensure the purity of the solutions and to locate the standard potential with respect to the reference electrode. A three-compartment cell, similar to the previous design, I was used for the electrochemical measurements. The reference electrode was either a silver pseudoreference electrode, consisting of a silver wire in the supporting electrolyte solution and separated from the test solution by a medium porosity fritted-glass disk or a silver wire in a 0.1 M A g N 0 3 solution separated from the main compartment by a porous Vycor glass junction. All values a re reported vs. the aqueous SCE. The counter electrode for voltammetric measurements was a coiled platinum wire separated from the main compartment by a Kohl, Bard / n-Type ZnO, CdS, and GaP Electrodes