|
11. |
On the kinetics of semiconductor-electrode stabilization |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 153-164
Felix Cardon,
Preview
|
PDF (681KB)
|
|
摘要:
On the Kinetics of Semiconductor-electrode Stabilization BY FELIX CARDON Laboratorium voor Kristallografie en Studie van de Vaste Stof, Rijksuniversiteit Gent, Krigjslaan 271, B-9000 Gent, Belgium AND WALTER P. GOMES, FERNAND VANDEN KERCHOVE, DANIEL VANMAEKELBERGH AND FRANK VAN OVERMEIRE Laboratorium voor Fysische Scheikunde, Rij ksuniversiteit Gent, Krijgslaan 271, B-9000 Gent, Belgium Received 17th April, 1980 Experimental results are presented, showing that the stabilization of illuminated n-type 111-v semiconductor electrodes by reducing agents is light-intensity dependent. A kinetic analysis of the stabilization process is made. The experimental data are discussed in the framework of this analysis. In recent years, there has been considerable interest in the problem of the stabiliza- tion of illuminated n-type semiconductor electrodes by a competing hole reaction with dissolved reducing agents, in connection with the possible use of such systems in photoelectrochemical solar cells.In many cases, the stabilization ratio, s, de- fined as the fraction of the photocurrent associated with the anodic oxidation of the reducing agent, was found experimentally to be significantly less than unity, and most of the research was hitherto directed, for a given electrode material, towards the de- pendence of s upon the nature and concentration of the reactants. In this paper, experimental research on n-type 111-v semiconductor electrodes will be presented, demonstrating that the stabilization ratio with a given reducing agent not only depends on the concentration, but also on the total photocurrent and hence on the light intensity.Starting from this fact, a kinetic analysis of the stabilization process will be made, the results of which will be compared with the experimental data, the objective being to gain insight into the mechanisms of photoelectrochemical reactions on semiconductor electrodes. RESULTS The systems studied were n-GaP/Fe(CN)%- (pH = 9.2), n-GaP/Fe"-EDTA and n-InP/Fe"-EDTA (pH = 3.8) in an aqueous medium. In all cases, the (TIT) face was exposed to the electrolyte. The semiconductor samples were mounted as the disc in a rotating ring-disc set-up, the ring being Au or amalgamated Au. Before each experiment, the electrode was chemically etched. The stabilization ratio, s, was determined from the measurement of the reduction current on the ring of the oxidizing agent Z formed by photoelectrochemical oxidation of the dissolved reducing agent Y at given total photocurrent, taking into account the collection efficiency of the r.r.d.e.(rotating ring-disc electrode) set-up. The collec- tion efficiencies were calculated from the characteristic radii. In the case of Gap, the calculated values were in agreement with those obtained from reduction currents on the ring under circumstances where, from earlier work,lS2 s was known to be very close to unity. The experiments were carried out at voltages at which the photocurrent was nearly voltage-independent. Only steady-state values of the photocurrent as well as of the154 SEMICONDUCTOR-ELECTRODE STABILIZATION stabilization ratio were considered.Both the photocurrent and s were also checked to be independent of the rotation speed of the r.r.d.e., indicating that effects due to diffusion-limitation can be excluded. At given concentration of reducing agent, the photocurrent was found to be proportional to the light intensity from a mercury lamp. A typical result is shown in fig. 1, where the stabilization ratio is plotted against the 0 50 100 150 200 FIG. 1.-Stabilization ratio, s, as a function of total photocurrent density, jph, at different concen- trations, y, of solved reducing agent Fe'I-EDTA. n-Gap, (TIT) face surface area 19.6 mm2. Disc potential 1.0 V us. SSE (saturated sulphate electrode), ring potential - 1.4 V us. SSE, rotation speed lo00 r.p.m.Fe" EDTA concentration: A, 1.0; X, 2.0; @, 5.0; m, 7.6 mmol Further electrolyte composition: 0.25 mol dm-3 K2S04 + 0.1 mol dm-3 KH-phthalate, pH 21 3.8. total photocurrent density at the GaP electrode for different concentrations of Fe"- EDTA. It can be seen that, at given concentration, s decreases with increasing photocurrent density, whereas at given photocurrent density, s increases with increas- ing concentration. Similar results have been obtained for the systems n-Gap/- Fe(CN);- [ref. (3)] and n-InP/Fe"-EDTA, showing that the observed behaviour is not restricted to a particular semiconductor nor to a particular reducing agent. KINETIC CONSIDERATIONS We will now investigate different kinetic models which may lead to light-intensity- dependent stabilization.The discussion will be restricted to those simple cases, frequently observed experimentally, in which neither the photocurrent density j,, nor the stabilization ratio s is diffusion-controlled and in which j,, is proportional to the light intensity L. It will be assumed that the proportionality factor between j,, and L is unaffected by the addition of the stabilizing reactant Y, hence excluding inter- ference with possible surface recombination as well as current-doubling effects. We will also limit ourselves to cases where Y does not react with the semiconductor in darkness nor adsorbs at its surface (the latter point may for example be checked experimentally by capacitance measurements). Furthermore it will be assumed that the anodic oxidation of Y is one-equivalent, that it occurs through one specific charge-transfer step and that the reverse of this step can be neglected. Since the evidence for light-intensity dependent stabilization has hitherto been restricted toF.CARDON et al. 155 n-type semiconductors, only these materials will be considered. It will be assumed that the overall electrochemical oxidation of the n-type semiconductor AB involves n valence-band holes (in the case of GaP e.g., n II 6).,v4 Evidently, the decomposition mechanism must be complex and has to involve at least n elementary electrochemical steps and a certain number of reaction intermediates. In what follows, we will denote by X1 the one-equivalent oxidation product formed in the first electrochemical step of the reaction between a hole h+ and the semiconductor surface (1) ki k-1 (AB)s + h+ Xi.In part 1 of our discussion, we will assume that no specific sites are involved in the formation of X1. 1. MECHANISMS NOT INVOLVING SPECIFIC SITES In this case, X, represents an electron missing from a surface bond. This defect may be either mobile within the surface, in a way comparable to the free holes in the bulk, or fixed at a given bond. We will subsequently investigate the kinetics in both cases. 1.1. Xi IS MOBILE It will be assumed in this case that all decomposition steps beyond X, are with the Hence, omitting possible non-electrochemical mobile species X, and not with h+. steps, the following reaction sequence can be considered : x2 + x1* x3} X n - , + X, % decomposition products. (4) We denote by j = jph/e the number of holes consumed in electrochemical reactions per unit of surface area and per second.The fractions of j used in the stabilization and in the decomposition are then given by s j and (1 - s)j, respectively. Under steady-state conditions, the net formation rates of X2 and of all higher oxidation products are equal and given by (1 - s)j/n = k2x; - k12~2 (5) i (1 - s)j/n = k3x2x1 - kL3x3 ----- (8) k’, = k-t + k”ly and in which xi represents the surface concentration of X i and y the volume concen- tration of the stabilizing agent Y . The rate constant k f f , is only different from zero if Y is oxidized by Xi through: (9) Y + xi -3- z + xi-,.156 SEMICONDUCTOR-ELECTRODE STABILIZATION In order to solve the problem, one more equation is generally needed.In many cases, the following one, defining j under the given circumstances, is appropriate: (10) The rate constant ko only differs from zero if Y is oxidized by holes directly through j = klp - k 1 x l + koyp. Y + h+k"-Z. (1 1) The value of x2 can be obtained by solving the linear set of equations given by eqn (6) and (7), leading to : + ..*). (12) X 2 = W -+--+-2L 1 1 kL3 1 k' kL n (k3x1 k3X1 k4X1 k3X1 k4X1 k5xl From eqn (5) and (12), one obtains an equation for x l : We will now investigate different cases, depending upon the values of certain of the rate constants involved. 1.1.1. Direct reaction between Y and h+ : ko # 0. It follows that: sj = ~ O Y P 1.1.1.1. The formation of X1 is irreversible: k - N 0. From eqn (lo), in this case From eqn (14) and (15), it follows j = (k1 + ~ O Y ~ P .koY kl + koY. S = (14) Hence, this mechanism leads to light-intensity independent stabilization. 1.1.1.2. These conditions amount to quasi-equilibrium in reaction (l), so that From eqn (14) and (17), one obtains The formation of X1 is reversible: k - l > 0. klp = k-lxl. (17) (1 8) . kok-1 SJ = - Y X 1 . Elimination of x1 from eqn (13) and (18) results in a relationship between s, j and y . Consider the simple case that kW2 N 0, which is probable from the chemical point of view. Since we have assumed that Y reacts through one reaction only and since in the given case, ko # 0, hence kL2 = 0 and, according to eqn (8), kL2 = k - , = 0. From eqn (13) and (18), it then follows that kl For k F 2 > 0, more complicated light-intensity-dependent expressions for s are ob- tained which cannot be written as a function of (y2/') only.F.CARDON et at. 157 1.1.2. will subsequently discuss the possibilities that Xi = X1, X2 or X3. No direct reaction between Y and h+ : ko = 0. The electrochemical oxidation of Y then involves a reaction of the type (9). We 1.1.2.1. Reaction between Y and XI : Y + X1 2 Z + (AB),. In that case, sj = klllyxl Assuming again that k 1 2 = k 21 0, it follows from eqn (13) and (21) that 1 - s n k , s2 - --- 1 kll: 6) The similarity between eqn (19) and (22) results from that between eqn (18) and (21), combined with the fact that in both cases the assumption k - 2 N 0 has been made. In an analogous way, for k-2 > 0, expressions are obtained of the same type as those mentioned at the end of section 1.1.1.2.1.1.2.2. Reaction between Y and X2: kIl2 # 0. This case might not be very realistic from the chemical point of view, since it implies, assuming the bonds to be covalent, that Y would donate an electron to a completely broken bond. It might be interesting, however, briefly to take it into consideration, since under special conditions it appears to lead to relations between s and (y2/j) differing from the type given by eqn (19) and (22). For example for k - 2 = 0 and k-3 = 0, eqn (13) yields, considering eqn (8): Furthermore, From eqn (S), (8) and (24): s j = k1,,yx2. From eqn (23) and (25): Finally, from eqn (25) and (26): 1.1.2.3. Reaction between Y and X3: kL3 # 0. Assuming k-2 = k-3 = k--4 = 0, it follows that s2 1 k2k'L23 y2 -=--- (3).1 - s n k$ Other cases lead to more complicated behaviour.158 SEMICONDUCTOR-ELECTRODE STABILIZATION 1.2. XI IS IMMOBILE The mobile species is now supposed to be h+, so that eqn (2)-(4) have to be re- placed by eqn (29)-(31) in order to describe the anodic decomposition: Xi + h+ T- X2 (29) x2 + he ----- "y X,-, + h+ + decomposition products. (3 1) For simplicity, the same symbols will be used for the rate constants of formation and back-reaction of the intermediates as in section 1.1. Under steady-state condi- tions, the net formation rates of XI and of all higher oxidation products corresponding to the reactions (I), (29), (30) and (31) are given by: (1 - s>j/n = klp - kL,xl (1 - s)j/n = k2xlp - kL2x2 (1 - s)j/n = k,x,,-lp.I ----- From eqn (33) and (34), one calculates xl as a function of p : Elimination of x , between eqn (32) and (35) yields 1.2.1. Hence, ko corresponding to eqn (1 1) is different from zero, and s is given by eqn (14). The relation between s, j and y is generally found by eliminating p from eqn (14) and (36). Direct reaction between Y and h+. Some typical cases will be discussed. 1.2.1.1. From eqn (14) and (36): The formation of X , is irreversible: k - , = kL, 21 0. i.e., s is independent of the light intensity. 1.2.1.2. The formation of X, is reversible: k - , > 0. k - k2P Assuming that in eqn (36), -' 1 and that k - , r kL2 = 0, it follows from eqn (14) and (36) that For k- > 0, more complicated light-intensity-dependent expressions of the type mentioned at the end of section 1.1.1.2.are found.F. CARDON et al. 159 1.2.2. No direct reaction between Y and h+ : ko = 0. 1.2.2.1. from eqn (21), (33) and (36): Reaction between Y and XI. Assuming that k - l = 0 so that kL, = k”,y and that k - 2 = kL2 N 0, it follows (39) From the mechanistic point of view, this is one of the simplest cases leading to More complicated expressions are obtained when reverse steps are taken into light-intensity dependence which can be envisaged. consideration. 1.2.2.2. one obtains from eqn (33), (36) and the general expression for sj, that is to say Reaction between Y and Xi (i > 1). Considering the simple case where k - l kLl = 0 and k-ci+ll f kLLi+ll = 0, sj = k!. tyxi that Obviously, such an expression can only hold within a limited range of (y/j). 2.MECHANISMS INVOLVING SPECIFIC SITES From literature data on structure-sensitive stabilization, it follows that structural defects such as kink sites may participate in photoelectrochemical reactions. Restrict- ing the discussion to binary compounds AB, two types of kink sites have to be taken into consideration, associated with the components A and B, respectively. The non- oxidized forms of these kink-site species will be denoted by A. and Bo, respectively. Electrochemical oxidation occurs through reactions with a species which will be denoted by X, and which can either be h+ or the mobile surface species X1 discussed above. The electrochemical decomposition reactions then are : k-%A Ap-l + X L”,A_ product *I + ----- =A2} J ----- kv,B Bv.-l + X + product (43) (44) (45) in which160 SEMICONDUCTOR-ELECTRODE STABILIZATION For simplicity, it has been assumed that both sets of reactions [(42)-(44) and (45)- Possible reactions with the stabilizing agent Y are : (47)] are mutually independent.Y + X L Z Ai + Y '2 Ai-l + Z ( i > 0) B, + Y "lilB B,-l + Z ( j > 0). Analogously to eqn (8), we define: Under steady-state conditions, the net formation rates of A,, B1 and of the higher oxidation products are given by (1 - S)j/n = k1,AaoX - k'-,,Aa, (54) (1 - s)j/n = kv,Bbv-lx (59) in which a, and bj are the surface concentrations of Ai and B,, respectively, and x the concentration of X. In what follows, it will be assumed that, throughout the decomposition process, the total number of kink sites per unit of surface area remains constant, which may for example be the case when the kink sites are associated with screw dislocations.Mathematically, this amounts to : One calculates from eqn (54)-(56) that (a0 + a1 + + a p - 1 ) + (bo + bl + - + b y - 1 ) = K. (60) k i C i + l l , A k', +- k t + LAX ki + 2,AX ki +3,AX An analogous expression for bj is obtained from eqn (57)-(59). From eqn (60), (61) and the analogous expression for bj, one obtains: } = K. k L C 1 + l I . B I C ) - U + 2 I , B + . . . kj+l,B kj+2rB k j + 3 , B 0F . CARDON et al. 161 (Note that kL,,* and k',,,, are equal to zero and cause the above series to be truncated.) In what follows, we will assume that, in case Y reacts with a mobile species X (either h+ or Xl), the same mobile species also enters into the reactions (42)-(47). 2.1 Y REACTS WITH x Hence sj = kyx and, from eqn (52) and (53), respectively: kL i,A EE k- i,A kLj,B E5 k-j,B 2.1.1.All decomposition steps are irreversible. It then follows from eqn (62) and (63) that that is to say s is independent o f j . 2.1.2. Limiting the discussion to cases where the quadratic term in l/x in eqn (62) is dominant (e.g., this would be the case if there is only one important reverse step, or else when non-consecutive reverse steps are involved), one obtains : The decomposition process involves reversible steps. 2.2. Y REACTS WITH ONE OF THE INTERMEDIATES A,, A2.. . or B1, B 2 . . . For example Y + At + Z + At-1. Hence, sj = ktt,,dai. 2.2.1. All decomposition steps are irreversible. Eqn (62) reduces to the terms in 1/x and to a term in 1/x2 proportional to k',i,A = k",AY.2.2.1.1. The term in llx is dominant. In that case, from eqn (59, (62) and (67), 0 0162 SEMICONDUCTOR-ELECTRODE STABILIZATION 2.2.1.2. The term in 1/x2 is dominant. It follows from eqn (59, (62) and (67) that Note that in all cases under section 2.2.1 ., s = f(u/'). 2.2.2. We again limit the discussion to cases where, in eqn (62), the quadratic term in l/x is dominant. This term, however, now contains a term corresponding to the re- action with Y as well as terms corresponding to the reversible decomposition steps. The decomposition involves reversible steps. 2.2.2.1. The term corresponding to the reaction with Y is dominant. This case is equivalent to case 2.2.12, provided that k-ci+l],A = 0.2.2.2.2. The term corresponding to the reaction with Y is negligible. Again assuming that k-ci+ll,A = 0, it follows from eqn (55), (62) and (67) that 0 0 DISCUSSION OF EXPERIMENTAL RESULTS The cases considered in the above theoretical development lead to expressions for the stabilization ratio which are either independent of the total photocurrent density and hence of the light intensity, or which can be written as a function of the variable (y/jph) or of the variable (f/jph) only (it is recalled here that j p h = je). Since ex- perimentally, s is undoubtedly light-intensity dependent, the first group of mechanisms, i.e., those mentioned under sections 1.1.1.1., 1.2.1 .l. and 2.1.1, can be excluded. In other words, the hitherto generally accepted mechanism, in which the decomposition involves irreversible steps only whereas the reducing agent is irreversibly oxidized by free holes, is apparently not valid in the cases considered.In fig. 2, the results of fig. 1 are replotted as log s against log (Yljph), demonstrating that s cannot be considered as a function of the variable (Y/jph) only. On the other hand, fig. 3 demonstrates that all experimental data are grouped reasonably well around a single curve when plotting log s against log (y2/jph). Analogous results were obtained with the systems n-GaP/Fe(CN)g- and n-lnP/Fe"-EDTA. As all mechanisms leading to such a relationship except for one rather improbable one [case (1.1.2.2.)] are of the type s2/(1 - s) - (y2/jph), we have replotted the same data according to this relationship on a double-logarithmic scale in fig.4. Considering that the experimental errors become relatively important for s either close to zero or to one when considering the variable s2/(1 - s), the results indicate that the light- intensity dependence may indeed be of the type just mentioned. From the section dealing with kinetic considerations, it appears that reversibility in at least one of the decomposition steps is a necessary condition for this type of relationship, except for the mechanisms mentioned under sections 1.1.2.1. and 1.1.2.3. Hence, if for some reason reversible decomposition steps can be excluded, one is forced to accept theF . CARDON et al. 163 1 1 I I -5.5 - 5.0 - 4.5 -4.0 - 3.5 -3.0 log10 (YbPh) FIG. 2.-Plot of loglo s against log,, (Yljph); same data as in fig.1 ; y/jph in arbitrary units. mechanism in which the decomposition involves a bimolecular step between two mobile surface holes X, and in which the reducing agent reacts by donating an electron to a partially broken surface bond, thus restoring the bond. If on the other hand inde- pendent experimental evidence, such as for example, the influence of surface pre- treatment upon the stabilization, suggests that specific sites are involved, the present kinetic study implies at least one reversible step in the decomposition reaction (sec- tions 2.1.2. or 2.2.2.2.). It should be remarked here that only electrochemical steps have been considered in the decomposition mechanism. Obviously, the mechanism may also involve purely chemical steps (e.g., reaction with OH- ions) as well as physical steps (e.g., 0 -0.5 b 2 2 -1.0 - -1.5 I I I I -8.0 - 7.5 -7.0 -6.5 - 6.0 - 5.5 log10 ( y b p h ) FIG. 3.-Plot of log,, s against loglo (Y/jph); same data as in fig.1 ; y2/jph in arbitrary units.164 SEMICONDUCTOR-ELECTRODE STABILIZATION surface restructuring). However, it can be shown that the appropriate set of equa- tions can be transformed to purely " electrochemical " ones such as those discussed above by elimination of intermediate products from the rate equations and by adapta- tion of rate constants. In particular, in this way, an irreversible chemical or physical step results in an irreversible " electrochemical " step. In all light-intensity-dependent cases emerging from our discussion, the stabilization ratio was found to decrease with increasing light intensity, a fact which is basically associated with the multi-equivalence of the anodic decomposition reaction. This 1 1 T 1 I I -31 A - 8.0 -7.5 -7.0 -6.5 -6.0 - 5 . 5 lO€hO(Y2 / j P d FIG. 4-Plot of loglo [s2/(1 - s)] against log,, (y2/jph); same data as in fig. 1 ; y2/jph in arbitrary units. light-intensity dependence may also constitute a serious problem when semiconductor/ electrolyte systems are used as solar cells, in particular when, for economic reasons, the cells are operated under concentrated sunlight. It results in a more rapid deterioration of the semiconductor for the same solar-energy input, provided at higher light in- tensity. This problem may even be present in cases where s is very close to unity so that the electrode is considered as being " stable ". Fundamental studies such as the present one may help to better understand the mechanism of competing photo- electrochemical reactions and hence to solve the problem of stabilization of photo- electrochemical solar cells. M. J. Madou, F. Cardon and W. P. Gomes, Ber. Bunsenges. phys. Chem., 1977,81, 1186. R. Memming, J. Electrochem. SOC., 1978, 125, 117. F. Van Overmeire, F. Vanden Kerchove, W. P. Gomes and F. Cardon, Bull. SOC. chim. belg., 1980,89, 181. K. Kohayakawa, A. Fujishima and K. Honda, Nippon Kagaku Kaishi, 1977,6,780.
ISSN:0301-7249
DOI:10.1039/DC9807000153
出版商:RSC
年代:1980
数据来源: RSC
|
12. |
Luminescent tellurium-doped cadmium sulphide electrodes as probes of semiconductor excited-state deactivation processes in photoelectrochemical cells |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 165-175
Arthur B. Ellis,
Preview
|
PDF (888KB)
|
|
摘要:
Luminescent Tellurium-doped Cadmium Sulphrde Electrodes as Probes of Semiconductor Excited-state Deactivation Processes in Photoelectrochemical Cells BY ARTHUR B. ELLIS, BRADLEY R. KARAS AND HOLGER H. STRECKERT Chemistry Department, University of Wisconsin-Madison, Madison, Wisconsin 53706, U.S.A. Received 29th April, 1980 Correlations among quantum efficiencies for photocurrent (rpx), emission (rp,), and non-radiative recombination (rp,,) are discussed with reference to data from single-crystal, n-type, 100 and 1000 p.p.m. CdS : Te-based photoelectrochemical cells (PECs) employing aqueous sulphide electrolyte. These materials emit while they serve as PEC electrodes. The assumption that the proportionality of rpr to rpnr is unaffected by potential, leads to a simple expression relating rp, to rp, for monochromatic excitation.Calculated and observed emission data are in reasonable agreement ; sources of devia- tion are discussed. Polychromatic excitation is shown to yield photocurrents and emission intensities which are approximately a weighted average of the values obtained with theconstituent monochromatic frequencies. Practical implications of the qx correlation with rp, are described, as are related results from other PECs. Luminescence, traditionally used to characterize excited-state properties, has found only limited use thus far in photoelectrochemical cells (PECs). Early studies involving emissive semiconductor photoelectrodes have employed n- and p-type GaP,’p3 n-type Zn0,293 n-type CdS,3 and n- and p-type G ~ A S . ” ~ We have focused our attention recently on n-type, tellurium-doped CdS (CdS : Te).5-7 Like the photoanode in PECs employing aqueous polychalcogenide electrolytes, single-crystal, 100 and 1000 p.p.m.CdS : Te emit while effecting the oxidation of polychalcogenide species. Emission from CdS:Te is believed to involve intraband-gap states which are introduced by the lattice substitution of Te for S; the band gap of CdS:Te is esti- mated to be about equal to that of undoped CdS (ca. 2.4 eV; ca. 515 nm).*-I2 Holes trapped at Te sites can coulombically bind an electron in or near the conduction band; the radiative collapse of this exciton leads to emission. The emissive spectral distribution (uncorrected A,,, !z 600 and 650 nm for 10 and 1000 p.p.m. CdS:Te, res- pectively) was found to be unperturbed when the material was used as a photoelec- trode in a PEC.5-7 Importantly, the doped electrodes mimic the electro-optical pro- perties of undoped CdS-based PECS.~?’~ Photocurrent and emission represent competing excited-state deactivation pro- cesses.The semiconductor excited state consists of a photogenerated conduction- band electron and valence-band hole (e’-h+ pair). This e’-h+ pair can either separate to yield photocurrent or recombine in a radiative or non-radiative fashion.14 Emission thus serves as a probe of the recombination process. In this paper we examine relationships between photocurrent and emissive quantum efficiencies. We demonstrate that the relationships which we find to exist for the CdS:Te-based PEC are consistently interpreted in terms of band bending, optical penetration depth and the competitive nature of the excited-state decay processes.166 LUMINESCENT CdS:Te ELECTRODES EXPERIMENTAL Plates of vapour-grown, single-crystal, 100 and 1000 p.p.m.CdS:Te, ca. 5 x 5 x 1 mm and oriented with the 5 x 5 faces perpendicular to the c-axis, were purchased from Cleveland Crystals, Cleveland, Ohio. Resistivities were ca. 2 C2 cm (four-point probe method). Con- centrations of Te are estimates based on starting quantities. Samples were etched with 1 : 10 (v/v) Br2 + MeOH before use. Electrode and electrolyte preparation as well as electro- chemical and optical instrumentation have been described previou~ly.~ Front-surface emissive properties were characterized by placing the PEC in the sample compartment of an emission spectrometer, as sketched in fig.1. The Coherent Radiation FIG. 1 .-Top view of the experimental arrangement used for observing front-surface emission from a CdS:Te photoelectrode: A, laser; B, monochromator; C, beam expander followed by slit; D, emissive CdS:Te photoelectrode, the site of S2- oxidation (not pictured are a Pt counter-electrode at which Hz evolution occurs, a SCE, and a potentiostat/programmer to which all three electrodes are connected); E, aqueous, N,-purged, sulphide (1 mol dm-3 OH- + 1 mol dm-3 S2-) electrolyte; F, emission detection optics. CR-12 Ar ion laser excitation source was used with multiline mirrors. This provided various combinations of excitation wavelengths in a single ca. 3 mm diameter beam.Mono- chromatic excitation was achieved by passing the multiline beam through an Oriel model 7240 monochromator whose grating was blazed at 500 nm. Operation of the monochroma- tor in zero order allowed for polychromatic excitation; the composition of the polychromatic beam was determined with the aid of a Bausch and Lomb 33-86-07 monochromator. Light intensities were adjusted with laser power and neutral density filters. In all experiments the laser beam was 10 x expanded and masked to fill the electrode surface. Incident intensity measurements required removal of the PEC from the spectrometer sample compartment ; after reconstruction of the PEC, the light intensity was reduced to give matching photocurrent and subsequently measured with a Tektronix 516 radiometer (56502 probe head; flat re- sponse 450-950 nm).RESULTS The experimental arrangement pictured in fig. 1 can be used to obtain current- luminescence-voltage (iL V ) curves as a function of excitation wavelength. Since the emissive spectral distribution is unaffected by either electrode potential or laser excitation ~avelength,~ emission intensity is readily monitored at a single wavelength, generally the emission band maximum. In fig. 2-4 and 5-7 we present iLV curves for 100 and 1000 p.p.m. CdS:Te-based PECs, respectively, employing transparent, aqueous, sulphide electrolyte. Although the data shown were obtained point-by-point with equilibration periods of ca. 20 s, similar curves were also observed with sweep rates up to 100 mV s-l. All data were obtained with ca.0.2- 1 .O mW cm-’ monochromatic or polychromatic laser excitation. With regard to monochromatic excitation, there is a strong distinction between 514.5 nm and the shorter laser wavelengths. For comparable intensities 514.5 nm light (fig. 2 and 5) generally yields substantially less photocurrent and greater emis- sion intensity than is found with il < 500 nm of which 488.0 nm excitation is illustra-A . B . ELLIS, B . R . KARAS A N D H . H . STRECKERT 167 h - -3 l a ) v x *z 2 16 +d f.ll -9 0 12- Il$ I I ._ 4 8 - 8 2 4 - f2 0 - c (d - I I 1 1 1 1 1 1 1 1 1 c - 28 - 2 4 - 20 - 4 I -16 2 I I - I I * I -12 4 4 3 5 I - a 5 / b l - I - 0 I: I I 1 I I I I I I I I I I I 400 W I 0 0 (d -3 0 - - -1.5 -1.3 -1.1 -0.9 -0.7 -0.5 -0.3 -1.5 -1.3 -1.1 -0.9 -0.7 -0.5 -0.3 !i potential/V us.SCE FIG. 2.-Emission intensity monitored at 600 nm (a) and photocurrent (6) us. potential for a 100 p.p.m. CdS : Te single-crystal electrode in sulphide electrolyte. The ca. 0.18 cmz exposed electrode area was excited with 0.1 1 mW at 514.5 nm. The error bars for each measurement are plotted and are primarily due to laser intensity fluctuations. Open circles in (a) represent the calculated emission intensity (see text); the open-circuit calculated value has been arbitrarily set at the midpoint of the observed emission intensity error bar. Midpoints of the photocurrent error bars have been used for the emission intensity calculation at the other potentials shown. Although not shown, little change in emission intensity is observed even at potentials several hundred mV cathodic of open circuit. potential/V us.SCE FIG. 3.-Emission intensity monitored at 600 nm (a) and photocurrent (b) us. potential for 488.0 nm excitation (0.14 mW) of the PEC of fig. 2. Because the geometry is unchanged from fig. 2, the emission intensity and photocurrent from the two excitation wavelengths may be directly compared when corrected to matching incident intensities (ein s-'). The error bars and open circles ( a ) have the same meaning as in fig. 2.168 LUMINESCENT CdS:Te ELECTRODES -1.5 -1.3 -1.1 -0.9 -0.7 -0.5 -0.3 -1.5 -1.3 -1.1 -0.9 -0.7 -0.5 -0.3 potential/V vs. SCE FIG. 4.-Emission intensity monitored at 600 nm (a) and photocurrent (6) us. potential for poly- chromatic excitation (0.20 mW) of the PEC of fig.2 and 3. The polychromatic light consists princi- pally of 514.5 nm (44 % by power), 496.5 nm (16%), 488.0 nm (32%) and 476.5 nm (8%). The PEC geometry is unchanged from fig. 2 and 3. Error bars represent the range of measured values. Open circles in (a) stand for emission intensities calculated by treating the exciting light as mono- chromatic; triangles are values calculated from a weighted average of the emission intensities obtained for the constituent monochromatic wavelengths (see text). The weighted average at open circuit, when corrected for the increased incident intensity used in this experiment, is ca. 40% above the midpoint of the observed emission intensity and has been arbitrarily scaled down to match this value; the emission intensities calculated in this manner at the other potentials were then also scaled down by this same factor.I 2 .4 -1.5 -1.3 -1.1 -0.9 -0.7 -0.5 -0.3 -1.5 -1.3 -1.1 -0.9 -0.7 -0.5 -0.3 potential/V us. SCE FIG. 5.-Emission intensity monitored at 670 nm (a) and photocurrent (b) us. potential for a 1000 p.p.m. CdS:Te single-crystal electrode in sulphide electrolyte. The ca. 0.18 cm2 exposed electrode area was excited with 0.040 mW at 5 14.5 nm. Error bars for each measurement are shown. Open circles are calculated emission intensities as described in the text and in fig. 2. Although not shown, little change in emission intensity is observed even at potentials several hundred mV cathodic of open circuit. I I I lbl -1.6 - 5 -1.2 g a -0.8 ,O - 0.4 -0 4-0 0 - c a II I I I I I I l l l l l l l l l l l lA .B . ELLIS, B . R . KARAS AND H . H . STRECKERT 169 h U .C( 5 3 0 0 - 3 250- -e 3 200- f 100 x Y .3 U .C( v1 v1 .* 3 150- a .3 - 2 5 0 - 2 x Y .M 3 100 c) i ~ l ~ l ' l ' l ' l ' l i l l J l ~ l ~ l ~ ~ l l 1 - 1 0 l a I 1 1 I I - 8 I - 6 - 4 - - 2 I - d - I c) - 0 I L L I I - O A T O 0 0 O O 8 4 I I l b l - . d a O 0 o 0 I - 0 1 1 1 I 1 1 1 1 1 1 1 1 1 I l l l l l l l l l l l l , potential/V us. SCE FIG. 6.-Emission intensity monitored at 670 nm (a) and photocurrent (6) us. potential for 488.0 nm excitation (0.038 mW) of the PEC of fig. 5 . The geometry is unchanged from fig. 5 so that emis- sion intensity and photocurrent from the two excitation wavelengths may be directly compared when corrected to matching incident intensities (ein s-'). Error bars and open circles have the same meaning as in fig.2. tive (fig. 3 and 6). A second recurrent feature is that, despite its relatively large intensity, emission from 5 14.5 nm excitation is scarcely potential-dependent (fig. 2 and 5). Contrast this (fig. 3 and 6) With the weaker emission intensity from 488.0 nm excitation which shows a marked potential dependence; in passing from -0.3 V vs. SCE to open circuit, the emission intensity increases by factors of almost 3 in fig. 3 and 6. FIG. 7.-Emission intensity monitored at 670 nm (a) and photocurrent (b) us. potential for poly- chromatic excitation (0.047 mW) of the PEC of fig. 5 and 6. Composition of the polychromatic light is principally 514.5 nm(29 "/o by power), 496.5 nm (19%), 488.0 nm (37 %) and 476.5 nm (14%).The PEC geometry is invariant with respect to fig. 5 and 6. Error bars, open circles and triangles have the same significance as in fig. 4. The actual weighted average at open circuit, when corrected for incident intensity variations in fig. 5-7, is ca. 20% below the midpoint of the observed emission intensity and has been scaled up to match this value; the emission intensities calculated in this manner at the other potentials were then also scaled up by this same factor.170 LUMINESCENT CdS:Te ELECTRODES Although only 488.0 nm iLV curves are pictured, our general observation has been that all Ar ion laser wavelengths from 457.9-501.7 nm give similar iLV proper tie^.^-^ We will subsequently refer to these as ultraband-gap wavelengths, since their energies exceed the nominal band-gap energy; absorptivities for these wavelengths are ca. so that designation of this line as band-gap edge is more appropriate.Polychromatic laser excitation was used for fig. 4 and 7. Its power composition was roughly 35% 514.5 nm and 65% ultraband-gap wavelengths, principally 488.0, 496.5 and 476.5 nm. As might be expected, the photocurrent and emission intensity from polychromatic excitation are bracketed by the values observed for pure 5 14.5 nm and ultraband-gap excitation. In both the monochromatic and polychromatic excitation experiments, the emission intensity (open circuit) and photocurrent (-0.3 V us. SCE) varied linearly with light intensity over the relevant intensity regime. 105 cm-l . 8-12 Absorptivities for 5 14.5 nm excitation are in the 1 03- 1 O4 cm- ' range,l2 DISCUSSION MODEL We will consider three routes for excited-state deactivation : separation of e--h+ pairs to yield photocurrent, and recombination of e--h+ pairs either radiatively or non-radiatively. The quantum efficiencies for these processes can be symbolized by p?, Pr, and Pnr, respectively.Treating these as the only decay routes yields the relationship ~x + Pr + Pnr = 1- We have sought to determine how PEC experimental parameters influence the relative and absolute magnitudes of these quantum efficiencies. While absolute values for qX can be measured, we have had to content ourselves with relative p, measurements, owing to the experimental difficulties inherent in accounting for the spatial and spectral distribution of emitted light.7*'s Measurements of qnr can be obtained with the technique of photothermal spectroscopy (p.t.s.),16-20 but we presently lack this information for the systems under scrutiny.Simplification of the excited-state PEC description occurs at open circuit. In this case no e--h+ pair separation can occur, making px = 0. The ratio of pr to qn, is easily shown to be dependent on excitation wavelength by comparing the emission intensities of fig. 2 with fig. 3 and fig. 5 with fig. 6 at open circuit. We have also previously shown this ratio to be temperature dependent at open circuit : q, decreases with increasing temperature.21 The photoelectrode may thus be considered as having a characteristic open-circuit ratio of radiative to non-radiative recombination (pro/ pnr0), eqn (2), where k depends at least on (1) Pnro = kPr0 (2) the excitation wavelength and temperature employed.The effect of now bringing the PEC into circuit is to permit some fraction of e---h+ pairs to separate rather than recombine. Values of p, and q,,, can be potential de- pendent in the sense that different fractions of e--h+ pairs are diverted from recombin- ation, depending on the magnitude of the photocurrent, px, Correlating px with p, and pnr also involves knowing the relative extent to which radiatively and non- radiatively recombining e--h + pairs are prevented from recombining, i.e., their rela- tive contributions to photocurrent. We will consider three simple schemes : photo- current, qx, interconverts (1) exclusively with pnr; (2) exclusively with q,; (3) with both qr and ynr such that qnr = kp, for any value of qx.A .B . ELLIS, B . R . KARAS AND H . H . STRECKERT 171 Scheme (1) predicts that emission intensity will be independent of potential. Although fig. 2 and 5 appear to be in accord with this prediction, it is contradicted for ultraband-gap excitation, as typified by fig. 3 and 6. Scheme (2) is untenable because of the relative magnitudes of vr and qx. We have measured q, and 9, to be ca. and l O - l , respectively, for 514.5 nm e~citation.~ In pasing from -0.3 V us. SCE to open circuit in fig. 2 and 5, scheme (2) would predict a ca. 100-fold increase in emission intensity in contrast to the invariance observed. A similar argument can be made for ultraband-gap excitation where qx is even larger.Scheme (3), although not perfect, is most compatible with our data. The ratio of emission intensi- ties at any two potentials, 1 and 2, can be computed by combining eqn (1) and (2) to give eqn (3). For the case where the first potential is open circuit, 1 - 9 x 1 - Prl 1 - 9x2 P 2 -- q r o - 1 . eqn (4) obtains q r 1 - ~x Table 1 consists of a compilation of qr,/qr ratios as a function of qx. TABLE 1 .-RELATIONSHIP BETWEEN px AND pro/pra (3) (4) 0.001 0.01 0.05 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1 .oo 1 .oo 1.01 1.05 1.11 1.25 1.43 1.67 2 .oo 2.50 3.33 5 .OO 10.00 co a Calculated from eqn (4) where px is the photocurrent quantum yield, and pro/pr is the ratio of emission quantum yields between open circuit and the potential where px is measured, We should point out that our treatment of pr and qx is not without precedent.Both Gap- and ZnO-based PECs have been examined in this regard.'.' The effects of optical penetration depth as well as carrier lifetime and diffusion length were con- sidered in these studies which will be discussed below. More recently p.t.s. has been utilized to establish relationships between qnr and qX.l6*l8 In a very broad sense the competition among qr, qnr and qx can be likened to a Stern-Volmer analysis22 wherein photocurrent fills the role of the quencher in homogeneous molecular systems. EXPERIMENTAL TESTS OF THE MODEL The predictive value of eqn (4) for monochromatic excitation is analysed with the aid of table 2 which compares direct measurements of qx at -0.3 V us.SCE with172 L u MI N E s c E N T CdS : Te EL E c TROD E s indirect determinations from tpr,/tpr for each of the experiments in fig. 2-7. Agreement is seen to be reasonably good. A more detailed comparison can be made by using each point of the iLV curve: After arbitrarily matching the calculated and observed open-circuit emission intensities, the measured qX at each potential was used to calcu- late [eqn (4)] the corresponding emission intensity. Calculated intensities are indicated by the unfilled circles in fig. 2, 3, 5 and 6. Looking first at overall changes, band-gap edge 514.5 nm excitation (fig. 2 and 5, table 2) gave a maximum p, of ca. 0.1. From table 1 we expect and observe little potential dependence of emission intensity. Low values of 9, are consistent with the comparatively large penetration depth of this wavelength ; a substantial fraction of the incident light will be absorbed beyond the depletion region whose width is typically TABLE 2.-cOMPARISON BETWEEN MEASURED AND CALCULATED VALUES OF vx electrode (fig.)" excitation I/nmb vX, meas.c px, calc.d 100 p.p.m.CdS:Te (2) (3) (4) (4) lo00 p.p.m. CdS:Te (5) (6) (7) (7) 514.5 488.0 P P 514.5 488.0 P P 0.07 0.49 0.28 0.28 0.12 0.74 0.60 0.60 0.00 0.64 0.04 0.31' 0.04 0.66 0.3 1 0.56" a The indicated electrode was used in the PEC shown in fig. 1 . Table entries are based on experi- mental results in the indicated figures. Excitation wavelength. An entry of P denotes poly- chromatic laser excitation with compositions given in the fig.4 and 7 captions. Photocurrent quantum efficiency measured at -0.3 V DS. SCE by the procedure described in the Experimental section. Entries are estimated to be accurate to f5 %, uncorrected for reflective losses and electro- lyte absorption. Photocurrent quantum efficiency calculated at -0.3 V DS. SCE with eqn (4) from the observed ratio of open-circuit emission intensity to the emission intensity at -0.3 V us. SCE. The midpoints of the emission intensity error bars were used for these calculations. We estimate that values are accurate to f 1 5 p/, based on the full error bars. Photocurrent quantum efficiency calculated at -0.3 V us. SCE using a weighted average of the measured px values for the individual monochromatic laser excitation lines which make up the polychromatic excitation.For fig. 4 the composition is roughly 44 % 514.5 nm and 56% ultraband-gap excitation; for fig. 7,29 % 514.5 nm and 71 % ultraband-gap excitation. The 488.0 nm qx value was used in each case as representative of the ultraband-gap wavelengths. 10-4-10-5 cm.I4 The lack of band bending beyond the depletion region favours e--h+ pair recombination. For the 488.0 nm excitation experiments maximum p, values are 0.49 and 0.74 for fig. 3 and 6, respectively. The large p, values observed at these potentials and wavelengths are expected, since almost all of the light for ultraband-gap wavelengths is absorbed within the depletion region. Between -0.3 V us. SCE and open circuit, corresponding two- and four-fold increases in emission intensity are anticipated (table 1). There is some discrepancy between the calculated and observed emission data, although they generally differ by < 20%.The polychromatic excitation experiments of fig. 4 and 7 were analysed by two methods. Open circles in these figures correspond to calculations based on the measured value of q, and use of eqn (4). In essence this amounts to treating the exciting beam as monochromatic light. The inappropriateness of this treatment is apparent from the marked deviation of the calculated curves from those observed.A . B . ELLIS, B . R . KARAS A N D H . H . STRECKERT 173 We find, however, that the measured photocurrent and emission intensities from polychromatic excitation are in good agreement with a weighted average of the photocurrents and emission intensities of the constituent monochromatic lines.This can be seen by comparing fig. 4 with fig. 2 and 3, and fig. 7 with fig. 5 and 6; the 488.0 nm data are treated as representative of the ca. 56 and 71% ultraband-gap composition of the polychromatic light in fig. 4 and 7, respectively. Open triangles in fig. 4 and 7 reveal that emission intensity calculated by this method is in much better agreement with the observed data, Implicit in the weighted-average procedure are the assumptions that the excitation wavelengths act independently of one another and that px and qr are independent of intensity. The latter assumption was verified (vide supra). Although fig. 2-7 demonstrate reasonable agreement between calculated and obsetcved emission intensity, we have seen iL V curves which cannot be accommodated by the simple scheme used thus far.For example, some iLV curves display “ humps ” wherein the luminescence-potential curve is characterized by maximum intensity at a potential other than open circuit. In this case there is a potential regime where both yX and VI, are declining. We have also seen iLV curves where the photocurrent is constant and saturated over a potential range, but the emission intensity continues to decline in the direction of more positive potential. Here qx is constant but qr and qnr are changing with respect to one another, i.e., their proportionality constant is no longer independent of potential. One possible explanation for these failures of the model is electroabsorption or potential-dependent abs~rptivity.~~ If electroabsorption were operative, it would mix the effects of optical penetration depth and band bending.Absorption of light in different regions of the electrode can give rise to different combinations of q,, qnr, and qx not only because of the differences in band bending, but also because the local environment in which e--h+ pairs are formed may vary because of lattice de- fects, impurities, etc. Although we have not observed any direct evidence for electro- absorption effects, we have not been able fully to discount the possibility either.’ Properties of the semiconductor surface represent a second possible source of deviations from model predictions. Although luminescence from CdS : Te electrodes is primarily a volume effect, there is a contribution from holes trapped at Te sites near the surface.The magnitude of this contribution may be reflected in the decline in emission intensity with decreasing optical penetration depth ; sites for non-radiative recombination are likely to be most prevalent near the surface. The filling and emptying of surface states with changes in potential could also influence 03, when surface contributions to emission are not negligible. Another possible surface effect arises in connection with our iLV curves which exhibit constant qx with declining qr values at positive potentials. This observation is reminiscent of data obtained in the ZnO- based PEC study where lower values of q, than expected were observed at positive potentials.2 This was thought to be due to a deficiency in the electron concentration needed for recombination near the surface; in turn, this was ascribed to the consider- able band bending present.In general we would predict that surface effects would be most significant with less penetrating ultraband-gap wavelengths. APPLICATIONS Examination of relationships among qx, qnr, and qr has both theoretical and prac- tical implications. In a theoretical sense we have tried to determine how the PEC experimental variables of excitation wavelength and potential influence the manner in174 LUMINESCENT CdS:Te ELECTRODES which the semiconductor excited state partitions input optical energy. The assumption that the proportionality of qr to qnr is unaffected by potential leads to a simple expres- sion [eqn (4)] which fits much of our data for CdS:Te-based PECs.As described above, however, there are iLYcurves which clearly violate the assumptions of the model. Whether or not eqn (4) turns out to be a good prognosticator of relationships between qr and q,, it serves to illustrate the potential utility of such a correlation. For the systems at hand which appear to obey eqn (4), 9, can be determined for monochromatic excitation simply by the use of a linearly-responsive light detector. That is, if the ratio between open-circuit and in-circuit emission intensities is known, then so too, is yx at the in-circuit potential. The additional knowledge of the photo- current permits calculation of the absorbed light intensity without having to correct measured light intensities for reflective losses, electrolyte absorption, etc.One difficulty with the technique is that, as table 1 indicates, it really only becomes sensitive when qx exceeds ca. 0.10. The insensitivity of qr to potential with 514.5 nm excitation is a case in point (fig. 2 and 5). We circumvented this problem in one in- stance by exploiting the negative temperature dependence of the CdS: Te band-gap. Sufficiently large values of 9, were obtained with 514.5 nm excitation at elevated temperatures to yield qro/qr values well in excess of unity.2' A method not requiring a change in PEC parameters is differential luminescence which was used to detect quenching by charge transfer in p-GaP when qx was only ca. 0.01.' A strategy employing p.t.s. to correlate temperature changes with qx has recently been described.16*'* This technique permits the simultaneous determination of qx and energy efficiency without a calibrated light source.It also provides a method for independently monitoring qnr and should be useful for examining our assump- tions regarding the potential dependence of qr and ynr. A key question related to our studies of excited-state decay routes in CdS:Te- based PECs is their applicability to other systems. It is gratifying to see that similar relationships between qx and qr are obtained for both 100 and 1000 p.p.m. CdS:Te in aqueous sulphide electrolyte. The excited-state manifolds of these species differ considerably: Lightly doped samples such as 100 p.p.m. CdS:Te are believed to have states ca. 0.2 eV above the valence-band edge; in addition to these states, more heavily doped samples such as 1000 p.p.m.CdS:Te have states ca. 0.4-0.6 eV above the valence-band Interestingly, both samples also exhibit an emission band at ca. 510 nm, near the band-gap edge; this transition is observable in electrolumines- cence and photoluminescence experiments under certain conditions.24 It is quenched by photocurrent roughly in parallel with the lower-energy emission band.25 Although systems with multiple excited states can potentially complicate the model, their study should particularly enhance our understanding of energy migration through the excited-state manifold. Besides CdS:Te, n-type ZnO, Zn0:Cu and p-type GaP have been examined with respect to relationships between photocurrent and emission intensity.Whereas CdS : Te electrochemistry consisted of oxidation of an electrolyte reductant, ZnO underwent photoanodic decomposition (to Zn2+ and 0,) and p-GaP gave reduction of water in the electrolytes employed.'*2 A rigorous treatment of photocurrent and emission intensity in terms of absorptivity, depletion region width, surface properties, carrier density, lifetime and diffusion length was presented in these studies. At least in the limiting cases of low and high q,, the potential dependence of pr was consistent with our results: besides the small dependence described above for p-Gap, no potential dependence of 9, was observed in the ZnO system with band-gap-edge excitation; on the other hand, complete extinction of emission could be obtained with ultraband-gap excitation at positive potentials, consistent with a qx value near unity (table 1).AA . B . ELLIS, B . R . KARAS AND H . H. STRECKERT 175 mirror-symmetry in the iLV curve, predicted from both eqn (4) and a derivation pre- sented in the ZnO study, was approximated by some of the ZnO experimental data. The foregoing observations indicate that there may, indeed, be general correla- tions among qx, qr, and qnr. Incorporation of other data (absorptivity, carrier proper- ties, etc.), examination of other systems, and the use of an independent probe for qnr such as p.t.s., should enable us to construct a refined model for excited-state decay processes of PECs. We are grateful to the Office of Naval Research for support of this work. We also thank Prof.Allen J. Bard for a preprint of ref. (18) and stimulating discussions. K. H. Beckmann and R. Memming, J. Electrochem. SOC., 1969, 116, 368. G. Petermann, H. Tributsch and R. Bogomolni, J . Chem. Phys., 1972, 57, 1026. B. Pettinger, H.-R. Schoppel and H. Gerischer, Ber. Blmsenges. phys. Chem., 1976,80,849. D. J. Benard and P. Handler, Surface Sci., 1973, 40, 141. A. B. Ellis and B. R. Karas, J. Amer. Chem. SOC., 1979, 101, 236. A. B. Ellis and B. R. Karas, Ado. Chem. Ser., 1980, 184, 185. B. R. Karas and A. B. Ellis, J. Amer. Chem. SOC., 1980, 102, 968. D. Dutton, Phys. Rev., 1958, 112, 785. A. C. Aten, J. H. Haanstra and H. devries, Philips Res. Rep., 1965, 20, 395. D. M. Roessler, J . Appl. Phjs., 1970, 41, 4589. lo J. D. Cuthbert and D. G. Thomas, J. Appl. Phys., 1968, 39, 1573. l2 P. F. Moulton, Ph.D. Dissertation (Massachusetts Institute of Technology, 1975). l 3 A. B. Ellis, S. W. Kaiser and M. S. Wrighton, J. Amer. Chem. SOL., 1976, 98, 6855. l4 H. Gerischer, J . Electroanalyt. Chern., 1975, 58, 263. l5 L. R. Faulkner and A. J. Bard, Electroanalytical Chemistry, ed. A. J. Bard (Marcel Dekker, Inc., New York, 1977), vol. 10, chap. 1, p. 1. l6 A. Fujishima, G. H. Brilmyer and A. J. Bard, Semiconductor Liquid-Junction Solar Cells, ed. A. Heller (The Electrochemical Society Softbound Proceedings Series, Princeton, N. J., 1977), p. 172. l 7 G. H. Brilmyer, A. Fujishima, K. S. V. Santhanam and A. J. Bard, Analyt. Chem., 1977, 49, 2057. A. Fujishima, Y. Maeda, K. Honda, G. H. Brilmyer and A. J. Bard, J. Electrochem. Soc., 1980,127, 840. l 9 A. Fujishima, H. Masuda, K. Honda and A. J. Bard, Analyt. Chem., 1980,52,682. 2o G. H. Brilmyer and A. J. Bard, Analyt. Chem., 1980, 52, 685, 21 B. R. Karas, D. J. Morano, D. K. Bilich and A. B. Ellis, J. Electrochem. SOC., 1980,127,1144. 22 N. J. Turro, Modern Molecular Photochemistry (Benjamin/Cummings. Menlo Park, Calif., 23 D. F. Blossey and P. Handler, Semiconductors and Semiinetals, ed. R. K. Willardson and A. C. 24 H. H. Streckert, B. R. Karas, D. J. Morano and A. B. Ellis, J. Phys. Chem., in press. 25 B. R. Karas, H. H. Streckert, R. Schreiner and A. B. Ellis, J. Amer. Chem. Soc., submitted. 1978), chap. 8, p. 232 and references therein. Beer (Academic Press, New York, 1972), vol. 9, chap. 3, p. 257,
ISSN:0301-7249
DOI:10.1039/DC9807000165
出版商:RSC
年代:1980
数据来源: RSC
|
13. |
Photoelectrochemistry of mercury (II) sulphide |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 177-187
R. Stephen Davidson,
Preview
|
PDF (811KB)
|
|
摘要:
Photoelectrochemistry of Mercury(I1) Sulphide BY R. STEPHEN DAVIDSON AND CHARLES J. WILLSHER Department of Chemistry, The City University, Northampton Square, London EClV OHB Received 16th May, 1980 Some photoelectrochemical reactions of red mercury(I1) sulphide powder coated on a platinum mesh are presented. The sulphide is an n-type semiconductor stable to photocorrosion in suitable electrolytes, but in aqueous halides can undergo darkening. In certain electrolytes (halides, thio- cynate) light-induced decomposition also occurs. The blackened sulphide, when used in an electrolyte that does not induce solubilisation, is more sensitive than the red form, in that larger photocurrents can be generated. The role of the electrolyte is emphasised, with respect to potential-determining ions, those which influence the semiconductor space-charge layer, and those which " block " surface states used in electron transfer.Photo- acoustic spectra show metacinnabar to be present, and the red-black transformation is treated in terms of rehybridisation of mercury under the influence of illumination and strongly interacting solution species. Band edges for red mercury(rr) sulphide are presented, and the potential of the material to harness solar energy evaluated. The blackening is discussed. Semiconductor photoelectrochemistry is an area of research currently receiving much attention because of the potential of utilising semiconductor-electrolyte systems to convert and store solar energy. A good deal of work has been done on the funda- mental principles; energy conversion with stable wide band gap (ca.3 eV) materials, stabilisation of narrower gap materials (1.5-2.5 eV) against photocorrosion, the exten- sion of wavelength response of wide gap compounds, and exploratory research to characterise new semiconductors-1 We have shown that red mercury(I1) sulphide is one of the few n-type semiconductors that responds to wavelengths corresponding to peak solar irradiance, and remains stable under specified conditions.' In this Dis- cussion, we present further results, with particular reference to the role of the elec- trolyte. Hitherto, the part played by the electrolyte has been discussed in terms of semiconductor stabilisation with redox c o ~ p l e s , ~ oxidation of specific materials at illuminated n-type semiconductors,4 and competitive oxidation of ions at photo- generated positive holes.s In other words, the electrolyte is treated as a source of, and sink for, electrons. It is clear that a particular electrolyte-semiconductor contact can act uniquely; the sulphate anion, for example, plays a part in titanium dioxide photocorrosion.6 Sulphate and hydrogen sulphate, often deemed " stable ", are capable of oxidation to peroxydisulphate at photogenerated valence band holes in titanium d i ~ x i d e , ~ and the sulphate radical anion has been pin-pointed as the corrosive species.' Tt is common to attribute the effects of ion adsorption on semiconductors to " changing the Helmholtz double-layer potential " and " shifting the surface band edges ".We consider that ions in the electrolyte have functions equally important to determining potentials and acting as electron banks.178 PHOTOELECTROCHEMISTRY OF MERCURY(II) SULPHIDE EXPERIMENTAL Electrodes are fabricated from commercial, as-received, mercury(I1) sulphide powder.A platinum mesh is coated with sulphide from an aqueous suspension. Details of this, and other experimental methods, have been previously described? Photoacoustic spectra were obtained with an OAS 400 spectrometer (EDT Research, London). All salts are of AnalaR or equivalent grade, and aqueous electrolytes made up with deionised water. RESULTS Previous experimental results2 are briefly reviewed. In pH 7 aqueous 0.1 mol dmV3 sodium nitrate, a negative relative photovoltage (- 150-- 300 mV) is obtained, indicating the sulphide is n-type.Anodic relative photocurrents obtain for biasing potentials positive of -0.25 V (us. a saturated calomel electrode SCE): this onset potential varies with pH by ca. -0.06 V per pH unit and is taken as the flat-band potential. W cm-2, a relative photocur- rent of ca. 3 pA is produced (effective area of illumination of electrode is ca. 2.25 cm2). The photocurrent is directly proportional to light intensity, and the photovoltage varies with In (intensity). The dependence of photoeffects on wavelength indicates the funda- mental band gap of ca. 2 eV controls the photoresponse; light of wavelength >600 nm is ineffective. No drop in photocurrent with time or visible deterioration of the sulphide occurs, except infrequent " blackening " of the irradiated surface.Blackening can be readily induced in aqueous electrolytes containing certain reducing agents, for example iodide, bromide, chloride and thiocyanate. Decomposi- tion of the sulphide usually accompanies the colour change, although a blackened electrode, when washed, dried and used in a nitrate electrolyte, generates photo- currents at least five times larger than those produced by the red electrode under identical experimental conditions. If mercury(I1) sulphide decomposes according to the equation HgS -+ Hg2+ + S + 2e, the passage of current should lead to solubilisa- tion. After passage of charge which should have resulted in detectable mercury solubilisation, no mercury could be detected by atomic absorption spectroscopy. Also no significant weight loss of HgS was observed. We have seen sustained gas (hydrogen) evolution at the platinum counter-electrode, but the photocurrent often drops with time.As little gas (oxygen) was evolved at the HgS electrode, this drop in photocur- rent is attributed to adsorbed oxygen. The properties of blackened red HgS corres- pond to those of the red form, except for an extended wavelength response to 700 nm. The performance of these electrodes depends upon blackening conditions ; optimum parameters for potassium iodide as blackening agent are iodide concentration = 0.1 mol dm-3, irradiation time = 60 min, and the red sulphide electrode held at 0.0 V us. SCE in a closed circuit, and illumination from a 1.6 kW xenon lamp filtered by passing through 11 cm of aqueous CuCl, (having 90% transmission at 525 nm in a 1 cm cell) and focused as prescribed.2u Blackening gives a more sensitive electrode, so it is unwise to conclude that in- creases in photocurrent with an oxidisable electrolyte are due to oxidation of such electrolyte species. However, we infer that this oxidation does take place, from the following observations. Irradiation of a suspension of red mercury(r1) sulphide in an oxidisable electrolyte (e.g., bromide, iodide) results in mercury being solubilised (de- tected by atomic absorption analysis of the electrolyte).Stirring the sulphide in the same electrolyte in the absence of light causes no mercury solubilisation. It appears necessary for irradiation to directly cause solubilisation, or generate a species that With an input light power of ca.2 xR. S. DAVIDSON AND C. J . WILLSHER 179 effects decomposition. When the electrolyte is potassium chloride, of molarity < 0.1 mol dm-3, blackening, but no sulphide decomposition, occurs on illumination. Halogens are known to attack red mercury(I1) ~ulphide,~ but the lack of mercury solubilisation in chloride means no chlorine is generated on irradiation. We conclude that, in general, the halide ion causes darkening, and the halogen produced from photoinduced oxidation of the halide causes solubilisation. The increased photo- current in these electrolytes is due to oxidation of the halide anion. We now present further experimental results to show how the electrolyte influences the photoelectro- chemistry of red mercury@) sulphides.Table 1 gives results of red mercury(I1) sulphide electrodes in pyridine containing TABLE 1 .-PHOTOEFFECTS PRODUCED BY RED HgS IN PYRIDINE CONTAINING LITHIUM NITRATE AND HALIDES electro- dark potential relative relative photocurrent in pA at com- lyte" potential on illu- photo- given bias/V us. SCE ments (0.1 mol /mV us. mination potential dm-3in SCE /mVvs. /mV SCE -0.2 0.0 +0.2 +0.4 + 0.6 ~ LiNOSb +220 4-75 -145 -5.0 0.0 +4.0 $8.0 +12.0 no colour change LiCl' +30 -125 -155 -0.8 +0.4 +1.7 +2.7 +6.2 no colour change LiBrd +60 -105 -165 -0.2 +1.3 +1.9 +4.0 no colour change LiIe -280 -420 -140 +1.6 +3.0 current in dark no = +260 ,uA @ colour $0.1 v change a Pyridine is doubly-distilled from KOH pellets and distilled a third time immediately prior to use.The electrolyte is purged with N2 passed through silica gel. * LiN03 is " ultrapure " and dried under vacuum before use. LiCl is dried under vacuum before use. d p e Used as received (fresh samples). lithium salts. All the electrodes gave the same relative photovoltage in pyridine, and with added lithium halides, photocurrents were largest for the iodide and smallest for the chloride. Mercury solubilisation occurred for lithium iodide and bromide, imply- ing iodine and bromine are produced. No solubilisation took place in lithium chloride or nitrate. The most striking result is that no blackening is seen in the halide-pyridine electrolyte. All possible precautions were taken to ensure the absence of water, al- though it is possible that trace amounts remain and contribute to photocurrents, but there is insufficient to induce the darkening that occurs in halide-water electrolytes.Fig. 1-4 show the variation in red mercury(I1) sulphide electrode potentials (us. SCE) in aqueous electrolytes with different molarities of potassium iodide, bromide, chloride and thiocyanate. These plots indicate that halide ions and thiocyanate can be potential-determining, in that HgS potentials us. SCE are controlled by salt con-180 PHOTOELECTROCHEMISTRY OF MERCURY(II) SULPHIDE 1U4 10-2 10 O [KBr]/mol dm-3 FIG. 1 .-Variation in initial dark potential ( x), final dark potential (O), and potential on illumination (+) of mercury@) sulphide electrodes in electrolytes of various potassium bromide concentrations. 200 0 Fil 52 2 4 B. -200 > ." 4- 5 c1 a -4 00 ( \ X X X + 104 loo [KI]/mol dm-3 FIG.2.-Variation in initial dark potential ( x), final dark potential to), and potential on illumination (+) of mercury(I1) sulphide electrodes in electrolytes of various potassium iodide concentrations.R . S . DAVIDSON AND C . J . WILLSHER 181 -2001 , , , 10" 10-2 100 [KClJ/mol dm-3 FIG. 3.-Variation in initial dark potential ( x ), final dark potential (0), and potential on illumination (+) of mercury(r1) sulphide electrodes in electrolytes of various potassium chloride concentrations. centration with a linear variation of ca. -80 mV per log,, [anion]. This is close to potentials predicted by the Nernst equation for a one-equivalent reducing agent :- RT [ oxidised form] nF reduced form * E = E* + -In E is the observed potential, Ee is the potential when species in the logarithmic function are in a defined standard state.R, T, n and P have their usual meanings. [oxidised form] is taken to be one, since no oxidised material is initially present, and RT/nF = -0.059 V when n = 1. E = E* - 0.059 log,, [halide, or SCN-1. Thus 400 pJ 200 # > E ._ 0 3 0 8 --- - Y 4- a -200 x x X X X \ O\, X loo [KSCN]/mol dm-3 FIG. 4.-Variation in initial dark potential ( x ),final dark potential (0), and potential on illumination (+ ) of mercury(I1) sulphide electrodes in electrolytes of various potassium thiocyanate concentrations.182 P HO TOE L E c TR o c HE MI s T R Y o F M ER CUR Y ( I I ) s u LP H ID E -80.- -60 w t; > E -40- $ . 8 +-' - 0 -20 TABLE 2.-vARIATION IN POTENTIALS US.SCE AND RELATIVE PHOTOPOTENTIAL FOR A RED MERCURY(II) SULPHIDE ELECTRODE IN ELECTROLYTES OF VARIOUS SODIUM NITRATE CONCENTRATIONS - - [NaN03] dark potential" potential on relative /mol dm-3 /mV us. SCE illumination photopotential /mV us. SCE /mV 10' + 30 - 100 -130 1 oo +155 - 5 -160 lo-' + 25 - 120 - 145 + 15 - 130 - 145 10-3 0 -185 - 185 10-4 - 5 - 135 - 130 10-5 + 140 + 20 - 120 a Initial and final dark potentials are the same. Table 2 shows potential in sodium nitrate electrolyte; no trend in HgS potential is seen with various nitrate concentrations, and the relative photovoltage is independent of [NaNO,]. The principal conclusion to be drawn is that each ion can act quite uniquely in its adsorption to HgS, and in some cases (e.g. SCN-) can control the band bending and hence the sign and magnitude and photovoltage.When oxidation of the reducing agent occurs, the oxidised half of the redox couple may also be potential- influencing, and will modify E according to the Nernst equation. When taking the photopotential, it has been noted that for platinum electrodes coated with powdered titanium dioxide, the voltage decay on ending illumination is 'X \x \ 0 ' I I I 0 1 2 3 4 In (timels) FIG. 5.-Voltage decay on ending illumination at open circuit for blackened red mercury(I1) sulphide in pH 7 0.1 mol dm-3 sodium nitrate: x , unstirred electrolyte; 0, electrolyte rapidly stirred by a magnetic Teflon-coated stirrer bead. The unstirred electrolyte plot has slopes of 15 mV/ln time and 7.5 mV/ln time, and slopes for the stirred electrolyte are 19 mV/ln time and 13 mV/ln time.R.S . DAVIDSON AND C . J . WILLSHER 183 influenced by the electrolyte.1° Rate constants (second-order) were evaluated in a number of electrolytes, but for mercury(I1) sulphide electrodes, the decay of photo- voltage obeyed neither first- nor second-order kinetics. Rather, the only near-linear decay plot was noted for voltage against the logarithm of time (see fig. 5 ) . Although there is a break in the linear decay, the fact that straight lines are obtained implies ion adsorption controls photovoltage decay. Note that stirring the electrolyte speeds up the decay. For a decay being controlled by chemisorption, the following relationship holds lo AN = A In (t/to + 1) where AN is the change in concentration of chemisorbed species, t is the time, and 1 and to are constants.The light-induced darkening of red HgS has been observed in vermilion pigments," cinnabar mineral l2 and various explanations as to its nature offered. Colloidal mer- cury,13 mercury-deficient HgS l4 and metacinnabar l5 have been postulated as causing the black colour. ESCA analysis of HgS blackened by irradiation in aqueous potas- sium iodide shows only Hg" and S" to be present, with no signals corresponding to zero-valent mercury or sulphur, iodide/iodine, or oxygen. Thus the superficial black colour is considered to contain black mercury(r1) sulphide, metacinnabar. Electrodes bearing entirely metacinnabar are hardly photoresponsiveZc and the authentic, black sulphide has been found to undergo solubilisation on irradiation in sodium nitrate.2c Thus the metacinnabar produced from illuminating red HgS must exist in such a form This is dealt with in greater detail in the Discussion section.X \ 1 \x 0 2 .o 2.2 2.4 2.6 \ X - x-x- x I I I 500 550 600 650 wavelengthlnm wavelength/pm FIG. 6.-Photoacoustic spectra in (a) visible and (b) infrared regions of red mercury(r1) sulphide (- x -1, black mercury(I1) sulphide (- - - -), and blackened red mercury(r1) sulphide (---I.184 PHOTOELECTROCHEMISTRY OF MERCURY(I1) SULPHIDE that its unfavourable properties are not realised. A solid solution of red and black sulphide is postulated. X-ray powder photographs of blackened HgS show patterns characteristic of cinnabar, and evidence for the presence of metacinnabar, in some form, is inferred from photoacoustic spectra.Fig. 6(a) indicates spectra for visible wavelengths. The absorption edge of cinnabar is clearly defined, but absorption is unchanged from 500 to 650 nm for blackened HgS and metacinnabar. Fig, 6(b) shows detail in the i.r. region. Between 2000 and 2600 nm the cinnabar spectrum is featureless, but metacinnabar has characteristic signals. These signals are contained in the spectrum of blackened cinnabar. DISCUSSION Results of pho toelectrochemical experiments with red mercury(1r) sulphide indicate this material possesses interesting properties, which have not been noted in any other semiconductor. These properties are now discussed in terms of influence of the electrolyte, blackening, solubilisation and potential usage as a harnesser of optical energy. INFLUENCE OF ELECTROLYTE Dissolved salts that form a conductive solution can not only be oxidised at light- produced positive holes in mercury(1r) sulphide, but can also influence the Helmholtz double-layer potential (fig.1-4). These phenomena must come about by a strong adsorptive interaction between specific ions and the sulphide surface. This inter- action is reflected in the way the potential of the sulphide electrode against the SCE varies with the salt concentration. For potassium iodide and bromide, potentials become more negative at higher concentrations, although in chloride this is not so straightforward, and for thiocyanate, only the final dark potential follows this trend. It is clear from table 2 that nitrate ions are not significantly adsorbed so as to become potential-determining.Indeed, the relative photopotential is invariant with nitrate concentration, confirming no influence to band-bending. However, the relative photovoltage for thiocyanate solution varies by 64 mV per log,, [SCN] (fig. 4). This implies that band bending is controlled by the thiocyanate anion, and can be altered sufficiently to generate positive photo-induced voltages at concentrations > 0.1 mol dm-3. For chloride, bromide and iodide electrolytes, the salt concentration does affect photo-induced voltages, but not in such a straightforward manner. It is accepted that OH- and H+ are adsorbed on to semiconductor surfaces and shift valence and conduction bands, plus Fermi level, by -59 mV/pH, and that Eredox for aqueous electrolytes shifts negatively equally, so no change in band-bending occurs with electrolyte pH variations. From the results here with halides and thio- cyanate, it appears that potential-determining ions must be divided into (at least) two categories : those which alter only the Helmholtz double-layer potential and those which alter both the double-layer potential and the potential drop in the semiconductor space-charge layer.with respect to EFermi will alter the band- bending, but when ions are very strongly interacting with a semiconductor, is it correct to assume an interaction between the bulk electolyte and the semiconductor? Rather, one should postulate a localised (E*redox) immediately at the surface that equilibrates with the semiconductor Fermi level to give the solid-liquid junction. &dox, as ob- tained from tables, should not be considered in such a junction, but E*redox used in- stead, based on very high local ion concentrations. The presence of adsorbed ions on the electrode surface will influence the move- Now a change ofR.S. DAVIDSON AND C. J . WILLSHER 185 ment of electrons across the solid-liquid interface, particularly if surface states mediate such electron transfer, and solution species are strongly adsorbed to " block " surface states. When the level photo- voltage is obtained, the semiconductor Fermi level is raised close to the flat-band po- tential and equilibrium exists between electron-hole pair generation from irradiation, and recombination due to the small space-charge layer.On ceasing illumination, all positive holes will be rapidly filled, but an excess of electrons remains in the conduc- tion band. Electrons must depart from this band, but since the valence band is now filled, they have to be transferred to the electrolyte. This is relatively simple, provided little surface barrier exists (this will be true immediately after ending illumination), and an acceptor level lies near or just below the surface conduction band edge. If no such level is present, then the electrons have to tunnel through the surface barrier (which increases as more electrons leave the conduction band and the semiconductor Fermi level falls), and may be transferred to the electrolyte via surface states. If these surface states are " blocked " by adsorbed ions, then the photovoltage decline will be impeded. Fig.5 indicates a " sorption '' process controls voltage drop, and, most significantly, a faster decline is noted in a stirred electrolyte. This suggests that agi- tation aids desorption and allows better electron transfer from the semiconductor to the electrolyte. Since NaNO, is not potential-determining, and therfore Na+ and NO3- are not significantly adsorbed, it must be OH-/H+ adsorption that blocks sur- face states. These are termed " weakly adsorbed ", since stirring (partially) removes them; but photovoltage decay in thiocyanate is unaffected by stirring; it is concluded that SCN- is '' strongly adsorbed." Blackening of red mercury(I1) sulphide occurs in some electrolytes, particularly halides.However, the results obtained from lithium salts in pyridine show that the blackening process is solvent-influenced. Solubilisation of the sulphide in pyridine, like in aqueous solutions, is dependent on the type of anion. Optoacoustic spectra imply that the dark material is metacinnabar, and ESCA re- sults indicate no species other than Hg" and S"- are present, at least on the surface. Thus, in certain aqueous electrolytes, a light-induced change in structure can be brought about. The principal change that occurs is in the co-ordination of mercury. In cinnabar, near-linear 2-co-ordination is found, with infinite -S-Hg-S-chains present.16 Metacinnabar has both mercury and sulphur in tetrahedral co-ordination. (Although the co-ordination of sulphur increases from two to four in going from cinnabar to metacinnabar, the bond angle at sulphur hardly changes.) It is suggested that in the presence of light and strongly interacting ions, a change in hybridisation occurs at mercury, from sp to sp3, and new bond formation to give metacinnabar then takes place.Now, as has been observed, both solute and solvent influence influence this darken- ing. Aqueous potassium chloride, bromide and iodide solutions effect blackening, but pyridine solutions of the corresponding lithium salts cause no colour change. The nature of solvation of active ions and/or the sulphide surface must be just as crucial as the presence of the ions themselves. It is often found that ions which induce blackening also cause solubilisation, and these ions are easily oxidisable.In some cases, no solubilisation is observed without illumination, and here it is concluded that the oxidised form of the redox couple causes decomposition. This is certainly true for halides; in particular, iodine in iodide solu- tion is known to decompose mercury(r1) sulphide according to : Fig. 5, showing photovoltage decay, exemplifies this. HgS + 2KI + I2 --f K2Hg14 + S.186 PHOTOELECTROCHEMISTRY OF MERCURY(II) SULPHIDE But results in pyridine electrolyte indicate that darkening and solubilisation are not necessarily concurrent ; this illustrates the importance of the electrolyte in determining photochemical reactions. POTENTIAL USAGE OF RED MERCURY(II) SULPHIDE TO HARNESS OPTICAL ENERGY A material that is efficient in harnessing solar energy should have stability, or be stable under known conditions, have a band gap that allows usage of optimal solar wavelengths, and be available in a suitable form for electrode preparation.Red mer- cury(r1) sulphide fulfils these conditions ; it has probably not attracted much research hitherto due to the difficulty in synthesising single crystals of high purity. We have illustrated it to be possible to obtain a photosensitive electrode with commercial ver- milion powder. Thus must represent the material at its worst, since the resistivity is undoubtedly high, doping is uncontrolled, and the platinum-coated electrodes, in their present form, are short-circuited to a large extent. The key to mercury(i1) sulphide acting as an efficient photoanode lies in the band positions relative to solution redox couples, and in the flat-band potential We estimate2 this potential to be -0.25 V us.SCE at pH 7, varying by -0.059 V per pH unit. This value falls within the limits set by Bockris and Uosaki17 for a useful n-type semiconductor. We have calculated the energies of the conduction and valence bands2c according to Butler and Ginley's method, from atomic electronegativities.18 However, the valence band edge has also been obtained from photo-emission ~tudies.'~ The band edges obtained by these two means vary by 1 eV (see fig. 7). In (a) the conduction band (E,) lies 4 1 E&/H20 electrolyte Ec ---- FIG. 7.-Postulated band edges for red mercury(1r) sulphide (a) calculated from the method in ref.(18), and (6) obtained from ref. (19). at 5.10 eV us. vacuum and the valence band (E,) at 7.10 eV. Water decomposition is a likely reaction of economic significance, and the relevant solution redox levels are indicated. Note that the level for H+ reduction lies above the conduction band, meaning a non-photon energy input is necessary for hydrogen evolution. In (b) E, lies at 3.9 eV and E, at 5.9 eV. Here, the water decomposition redox levels lie entirely within the band gap, which is favourable for photoelectrolysis unassisted by any other energy input. We have been unable to fix the band edges by chemical means (e.g., obtaining E, by noting oxidation of species with Ee negative of the band edge in the dark, etc.) since (i) bare platinum at the sulphide electrode renders such experiments meaningless, and (ii) the blackening undoubtedly results in a change in band edges.R.S . DAVIDSON AND C. J . WILLSHER 1 a7 Butler and Ginley’s method is a theoretical one, whereas the band edges shown in fig. 7(b) are derived from experimental results. The valence band is taken as the photoemission cut-off at low energie~.’~ This, more correctly, corresponds to the work function, but the authors of ref. (19) equate work function with valence band top. This is contrary to normal practice, where the work function and Fermi levels are equated. When this is done, E, and E, approximate to those calculated by Butler and Ginley’s procedure, and the band edges of fig. 7(a) hold true. We have observed that it is usually necessary to apply a bias of > +0.2 V to achieve hydrogen evolution, although with some electrodes, gas production occurs at zero bias.Thus we conclude that the band edges of mercury(1r) sulphide are unfavourably placed for hydrogen evolution. However, for other useful energy-harnessing reactions, mercury(r1) sul- phide may prove a useful photoanode, especially as its blackening can improve the photosensitivity. We thank the N.R.D.C. for a Fellowship to C. J. W., and Prof. J. Wightman of the Chemistry Department, Virginia Polytechnic Institute and State University, Blacks- burg, Virginia, U.S.A. for performing the ESCA analyses. (a) A. J. Nozik, J. Cryst. Growth, 1977, 39, 200; (6) H. P. Maruska and A. K. Ghosh, Solar Energy, 1978,20,443 ; (c) A. J. Nozik, Ann. Rev. Phys. Chem., 1978,29,189; ( d ) K.Rajeshwar, P. Singh and J. Dubow, Electrochim. Acta, 1978, 23, 1117; (e) M. Tomkiewicz and H. Fay, Appl. Phys., 1979, 18, 1; (f) A. Heller and B. Miller, Electrochim. Acta, 1980, 25, 29; ( g ) R. Memming, Electrochim. Acta, 1980, 25, 77. (a) R. S. Davidson and C. J. Willsher, Patent 7913420 (18 Apr. 1979); (b) R. S. Davidson and C. J. Willsher, Nature, 179,278,238; (c) R. S. Davidson and C. J. Willsher, J.C.S. Faraday I, 1980,76,2587. M. S . Wrighton, J. M. Bolts, A. B. Bocarsly, M. C. Palazzotto and E. G. Walton, J. Vac. Sci. Technol., 1978, 15, 1429. F. Vanden Kerchove, J. Vandermolen and W. P. Gomes, Ber. Bunsenges. phys. Chem., 1979, 83, 230. A. Fujishima, T. Inoue and K. Honda, J . Amer. Chem. Soc., 1979,101, 5582. L. A. Harris and R. H. Wilson, J. Electrochem. SOC., 1977, 124, 839. K. Kohayakawa, T. Yamabe, A. Fujishima and K. Honda, Nippon Kaguku Kaishi, 1978, 10, 1351 (See Chem. Abs., 89,223083t). J. W. Mellor : A Comprehensive Treatise on Inorganic and Theoretical Chemistry (Longmans, Green and Co., London, 1946), vol. IV, p. 942 et seq. H. H. Chambers, R. S. Davidson, R. R. Meek and R. M. SIater, J.C.S. Faraday I, 1979, 75, 2517. R. L. Feller, Nat. Gall. Art, Report and Studies in Hist. Art (Washington D.C., 1967), p. 99. a M. Gerstner, J. Electrochem. Soc., 1979, 126,944. l2 W. H. Cropp, Proc. Austral. Inst. Mining Met., 1923, 52, 259. l3 R. M. Dreyer, Amer. Mineral, 1939, 24, 457. l4 R. W. Potter and H. L. Barnes, Geol. SOC. Amer. Abs. Programs, 1974, 3, 674. C. Brosset, Naturwiss., 1936, 24, 813. l6 K. L. Aurivillius, Acta. Chem. Scand., 1950, 4, 1413. J. O’M. Bockris and K. Uosaki, Conf: Proc. 1st World Hydrogen Energy Conf., 1976,2, 5B-1. l8 M. A. Butler and D. S. Ginley, J . Electrochem. Soc., 1978, 125, 228. l9 N. J. Shevchik, J. Tejeda, D. W. Linger and M. Cardona, Phys. Status Solidi B, 1973,60,345.
ISSN:0301-7249
DOI:10.1039/DC9807000177
出版商:RSC
年代:1980
数据来源: RSC
|
14. |
Photoelectrochemical behaviour of layer-type transition metal dichalcogenides |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 189-205
Helmut Tributsch,
Preview
|
PDF (1270KB)
|
|
摘要:
Photoelectrocheniical Behaviour of Layer-type Transition Metal Dichalcogenides BY HELMUT TRIBUTSCH Laboratoire d’Electrochimie Interfaciale du C.N.R.S., 1, Place Aristide Briand, 92190 Meudon, France Received 19th May, 1980 The photoelectrochemical properties of semiconducting layer-type disulphides and diselenides of transition metals belonging to Groups IV (zirconium, hafnium), VI (molybdenum, tungsten) and VIII (platinum), which have energy gaps between 1 and 2 eV, have been studied systematically. Characteristic reaction mechanisms and differences between compounds were found which can be interpreted in terms of the contribution of d-orbitals to the valence band in which photoreactive holes are generated. Evidence is provided that the behaviour of d-band semiconductors is critically influenced by a specific light-induced interaction of the interface with the redox system.This inter- action leads to surface states which are apparently charged to an electric potential which is propor- tional or equivalent to the difference between the valence band edge and the redox potential of the electron donor (AE). It is the source of a photoelectrochemical reaction enthalpy which is useful for energy conversion. d-Band semiconductors thus have behaviour which is characteristically differ- ent from that of classical semiconducting compounds where hole formation is equivalent to the destruction of existing chemical bonds and the energy difference AE is dissipated as heat. The parti- cular photoelectrochemical properties of transition metal dichalcogenides enable a larger variety of new applications to be explored.While the compounds of Mo and W can be used as reasonably stable electrodes for regenerative and fuel producing solar cells, those of Zr and Hf could be developed for the combined conversion and storage of solar energy by means of light-induced intercalation reac- tions. PtS2, finally, could be used as a sensitive and highly selective photoelectroanalytical probe for electron-exchanging ions and molecules. Various theoretical aspects ranging from photoelectrode stability to water decomposition with visible light are considered in this context. It is well known from biochemical research that all essential electron-transfer reactions in living systems which play a role in energy conversion are mediated by enzymes with transition metal complexes as their catalytic reaction centres. This is true for nitrogen fixation, hydrogen evolution and oxygen reduction, which are dark reactions, as well as for light-induced oxygen evolution from water in photosynthesis, which is known to be accomplished by a protein containing manganese.Little is established about the molecular and electronic nature of these electron-transfer reactions but it is clear that they involve transition metal d-orbitals and a simultaneous reorganization of chemical bonding which is controlled by the oxidation state of the transition metal centre. It is this interaction between electron transfer and chemical bonding which makes electrochemical mechanisms based on d-states of transition metals especially interesting for research.Since control of catalytical properties is essential for the use of the energetic advantage of electrochemical mechanisms, it might ultimately also have been the reason for the selection of transition metal centres during evolution. Apparently the first attempt to identify semiconductors which allow photoreac- tions to occur via energy bands derived from d-orbitals of transition metals was made in 1977.l This directed attention to approximately 40 lamellar compounds in which190 TRANSITION METAL DICHALCOGENIDES monolayers of transition metals (M) are sandwiched between monolayers of sulphur, selenium or tellurium, these layers (SMS) being arranged into stacks which are held together by van der Waals forces. Essentially alternating with the Group of the Periodic Table they form either semiconductors (Groups IV, VI and VIII) or metals (Groups V and VII), the electrochemical or photoelectrochemical properties of which are characterized by the existence of energy bands which are derived from transition metal d-orbitals.Depending on the degree of filling of these energy bands and on differences in the symmetry of crystalline organization (octahedral or trigonal pris- matic packing), the photogeneration of holes and electrons occurs at electronic states which are typically different for each of the groups (fig. 1). Electrons are excited from octahedron ( t 1 - 4-dk s-p 2g I::::.:.:.:.:.:.:. . . . . . . _ . . hv.-.-. . . . . . . . . . . . _ . _ . :::::::: . . . . . . _ . . _ .. . ~ . . . . . _ _ . _ Group: I V compound ZrS2,HfS2 tr igonal octahedron prismatic 2 V I v I11 M o S ~ , W S ~ PtS2 ZrSeZ. HfSe2 MoSe2,WSe2 Pt Se2 FIG. 1 .-Schematic structure, crystal symmetry and energetic distribution of electronic d states of the transition metal (M-d) in semiconducting dichalcogenides of group IV, VI and VIII. an energy band derived from sulphur orbitals into a d-energy band (t2g) in zirconium and hafnium dichalcogenides, between energy bands derived mainly from d-orbitals (d,z -+ dxy, dx2 - ,z) in molybdenum and tungsten dichalcogenides and from a d-band (t2g) to higher non-bonding mixed states in platinum dichalcogenides. The very first photoelectrochemical s t ~ d i e s ~ * ~ were made with natural crystallineH. TRIBUTSCH 191 samples of MoS2 (molybdenite).By exhibiting good photoeffects and energy conver- sion efficiencies up to ca. 1 % (in the regenerative mode) they served to pinpoint some of the remarkable properties of these materials. These results with mineral samples suggest, moreover, that photoelectrochemical energy conversion with molybdenite and Fe2+/3+ as a redox couple also occurs as a natural geophysical phenomenon. Further studies have been continued with better defined synthetic material from vari- ous laboratories. They concentrated mainly on MoSe;-' and WSe?*'O or included the corresponding ~ulphides.l'-'~ Recently MoSe," and WSe,'69'7 have been studied as electrode materials for electrochemical solar cells by additional research groups. So far, photoelectrochemical studies of semiconducting layer type transition metal compounds of Group IV (ZrS,, ZrSe2)fsv19 and Group VIII (PtS2),0 appear to have been made only by the present author.It is the aim of this work to provide a comparative study of the photoelectro- chemical reaction behaviour of semiconducting layer-type transition metal dichalcoge- nides and to elaborate the more important conclusions for research and application. Scientific investigations on this large group of compounds is still hampered by the lack of information on solid-state properties although many details have been published and quite extensively reviewed.21-24 EXPERIMENTAL The layer-type crystals used for this study have been synthesized in different laboratories : Prof. R. Nitsche, Freiburg (ZrS,, ZrSe, HfS,, HfSe2 and TiS,), Prof.Bucher, Constanz (MoS,, MoSe,, MoTez and WSe2), Dr. Levy, Lausanne (MoSe, and WSe,) and Dr. 0. Goro- chov, Paris-Meudon (PtS2). They were electrically contacted by means of silver or platinum paste and mounted on a cylindrical Teflon electrode which exposed the van der Waals surface ( l c ) to the electrolyte, The area of the electrode surface ranged from 1 to 10 mm2. It was regularly renewed by attaching and removing an adhesive tape which served to peel off a thin outer layer. The experiments were performed in a standard electrochemical cell using poten- tiostatic and linear sweep conditions. The light source which was usually employed was a 150 W xenon lamp. Spectral dependences were studied with a tungsten lamp using a Jobin Yvon HRS 2 monochromator, lock-in measurements were made with a PAR 121 lock-in amplifier and a PAR 125 A chopper.RESULTS MOLYBDENUM AND TUNGSTEN DICHALCOGENIDES PHOTOELECTROCHEMICAL PROPERTIES When the van der Waals surface (kc) of an electrode of a disulphide, diselenide or ditelluride of molybdenum or tungsten in contact with an aqueous electrolyte is illu- minated, photocurrents can be drawn at a quantum efficiency between 0.6 and 0.8. Most of the remaining light intensity is lost because of reflection from the metallic bright electrode surface. As with other semiconductors, photocurrents are either anodic or cathodic, depending on whether the electrode material is n- or p-type con- ducting, which can be controlled by properly selecting the conditions for crystal growth or doping with either rhenium or niobium.As seen from table 1, which com- pares the electronic and photoelectrochemical properties of different layer compounds, the principal reaction product of the anodic photoreaction is sulphate or selenate, res- pectively. This reaction behaviour is in marked contrast to semiconductors such as CdS or CdSe, which provide holes in valence bands derived from chalcogenide orbitals,192 TRANSITION METAL DICHALCOGENIDES and photocorrode to molecular sulphur and elemental selenium, respectively. The anodic photoreaction of molybdenum and tungsten dichalcogenides thus involves water and is suppressed when the aqueous electrolyte is replaced by an organic solvent which resists oxidation (e.g., acetonitrile). In this case only the addition of water or a reducing agent, hydroquinone for example, can restore the phot~current.~ The spectral sensitivity of layer-type molybdenum and tungsten compounds covers the entire visible and part of the near infrared spectral range.Fig. 2 and 3 show the FIG. 1 2 3 photon energy/eV -Anodic photocurrents (full line, electrode potential: 0.8 V; electrolyte: O.va mc dm-3 HzS04 + 0.5 mol dm-3 K,S04) of MoS2 and MoSe, and corresponding absorption coefficients a [logarithmic plot, dotted lines, adapted from ref. (3 l)] at ambient temperature. photocurrent spectra, plotted as a function of the photon energy in eV for MoS,, MoSe,, WSe, and MoTe,. Superimposed on the pictures are corresponding absorp- tion constants of the same materials at ambient temperature, which clearly reveal light absorption by excitonic transitions (the exciton binding energies of the A structure are estimated to be 50,67 and 50 meV in MoS,, MoSe, and WSe,, respectively).24 The most interesting observation which can be made is that the excitons, certainly exciton B, do not contribute to the generation of the photocurrent.Another is that the photo- current density is not proportional to u (the absorption constant), as one would expect if the diffusion length L, for holes L, were small compared with l/u. It could, in fact, be estimatedI4 that the diffusion length for holes perpendicular to the layers of theH . TRIBUTSCH 193 crystals would have to exceed cm in order to account for the characteristic be- haviour of the photocurrent as seen in its voltage and spectral dependence.This is a surprisingly large value if one considers a hopping mechanism across the van der Waals gaps of the layer-type material. The theoretical analysis of charge-carrier transport in these compounds is not only complicated by a strong anisotropy of mobi- lity which can amount to up to three orders of magnitude, but also by a lack of infor- mation on physical constants and detailed band structures, so that more research will be needed to clarify details. This is equally true for the value of the indirect band gap, which, as seen in table 1, has still not been thoroughly investigated, with values scat- I ........ ..... .... o.o&l I .3 O , O 2 I - 0 . 0 o L i! $ 0.5 - c a * 0 . 2 5 - ........... ........ W MoTe 0.5 1 2 3 FIG.3.-Anodic photocurrents and corresponding absorption coefficients for WSe2 and MoTe2 (conditions as in fig. 2). tering considerably depending on the experimental method used for its determinati~n.~~ It seems to be clear that the effective photoelectrochemical energy gap of these com- pounds is smaller than originally a ~ s u m e d . ~ . ~ Nevertheless, it is high enough for efficient solar energy conversion, as can be seen from the photopotentiaIs which, in the presence of I-/Iy, can reach values as high as 0.57 V with MoSe, and 0.7 V with WSe,. The photocurrent-voltage characteristics of layer-type molybdenum and tungsten dichalcogenides are very similar and are qualitatively explained in fig. 4 for n- and p- type WSe,. In the absence of a reducing agent in the electrolyte the anodic photo- current of n-type samples starts only at an elevated electrode potential (near 0.5 V us.SCE). In the presence of a redox system like [-/I<, however, there is a considerable shift of photocurrents towards negative electrode potentials. A less pronounced nega-194 TRANSITION METAL DICHALCOGENIDES tive displacement of the photocurrents is observed in the presence of Fe2+I3+ or Br-/Br2. This photocurrent shift, induced by the presence of redox systems, is paralleled by a shift of the flat-band potential of the layer-type electrode, which is thus dependent on the redox potential and the composition of the electrolyte. In contact with an aqueous electrolyte containing only an inert salt, n-type molybdenum com- pounds assume a flat-band potential near or slightly more positive than the hydrogen potential (-0.24 vs.SCE) whereas tungsten compounds assume flat-band potentials which are a few hundred mV more negative. The strong cathodic photocurrents TABLE 1 .-COMPARISON OF PHYSICAL PROPERTIES AND PHOTOELECTROCHEMICAL BEHAVIOUR OF SEMICONDUCTING TRANSITION METAL DICHALCOGENIDES anodic cathodic energy (photo) (photo) reaction photocurrent possible gap electron conduction reaction reaction of redox efficiencies applica- compounds lev transition type products products systems reached tion H& Ha ZrSa' 1.68 (ind.) p --f d p-type sc (?) So, ZrIV intercal. products b n-type SC 0.1-0.2 e n-type SC Has. H2 HfSep 1.96 (ind.) p + d p-type sc (?) So, Hf I V intercal. products t0.05 e N H2Se, Ha HfSp 1.13 (ind.) p --+ d n-type p-type sc sc (?) Se, HflIv intercal.products 50.05 e < 10-6 degen. SC so, TiIv intercal. products TiS2" 0.83 (ind.) p -+d FS lopl Sop, MoVI 0.6-0.8 f lower oxida- 0.6-0.8 f states of M o MoS2 1.13-1.6 (?) d + d n-type p-type SC Oa (small Has amounts) (ind.) MoSe, 1.0g-1.35 d + d :?$ SC SO:-, MoVI H2Se ind.) MoTe, FS 1.02 (ind.) d + d SC Tea:- MOVI HJe 0,; com- platinum (small amounts) PtSa 1.03 (ind.) d+d n-type SC plexes of Ha d 0.6-0.8 h ' Measurements with crystals of moderate photoelectrical quality; non-specific reaction with holes; specific photoreaction (e.g. with I-); * photocurrents shift characteristically with redox potential; photointercalation : solar-energy conversion and storage; regenerative electrochemical solar cells; photoelectroanalytical probe. photodecomposition of HI into +Hz + +Iz; observed with p-WSe, (fig.4) are due to hydrogen evolution. Accurate data on flat- band positions of layer-type electrodes are difficult to obtain because of deviations from the Mott-Schottky re1ati0n.l~ The comparison of electrode materials of very different surface quality has re- vealed the extraordinary importance of surface steps exposing surface areas to the electrolyte for the characteristic shape of the current-voltage curve as well as the mag- nitude of dark currents.''-13 While electrode surfaces exhibiting a high concentra-H . TRIBUTSCH 195 tion of steps are characterized by high dark currents and shallow slopes of photocur- rent curves, negligible dark currents and steep slopes of photocurrent dependences are observed with nearly perfect van der Waals surfaces.The influence of step sites (Ilc) on the electrode characteristic has been discussed in terms of different band struc- tures, surface states, a narrowing of the space-charge layer, which decreases charge separation, and the strongly asymmetric mobility, which favours recombination pro- cesses at such steps.12*16 Ip-HI 1 -1.0 - 0.5 0 0.5 1 .O electrode potential/V us. SCE FIG. 4.-Photocurrent-voltage behaviour of n- and p-type WSez as an example for the properties of Group VI materials. The shading indicates how the characteristic curves may vary between samples of different preparation and quality. The inserted pictures show how energy can be converted in regenerative and fuel producing solar cells, respectively. SOLAR- ENER GY CONVERSION Fig.4, which describes the (photocurrent, voltage) dependence of n- and p-type WSe,, can also serve to explain the possibilities which exist for solar-energy conversion. When a metal counter-electrode is placed in contact with an I-/Iy solution in a solar cell together with a n-type photoelectrode (top left scheme in fig. 4) it will assume an open-circuit potential equivalent to the redox potential of the electrolyte [Eo(1-/I3)]. The open-circuit potential of the illuminated n-WSe, electrode will be approximately 0.7 V more negative where the photocurrent becomes zero. Depending on the cha- racteristic shape of the photocurrent curve and the external and internal resistances of the cell the photocurrent generated by the solar cell under working conditions (IT-~/IJ will correspond to an actual electrode voltage somewhere between these potentials.Solar cells of layer-type n-conducting materials utilizing the I-/I: redox system in the196 TRANSITION METAL DICHALCOGENIDES electrolyte have demonstrated a good performance. A MoSe, (I-/I:) solar cell has been operated at a current density of ca. 10 mA cm-, for 10 month^.^ Even consider- ing the high number of 14 holes for the oxidation of MoSe, to MoV1 and 2 Se0;- this would correspond to 5000 times the charge required to dissolve the whole electrode. Nevertheless at the end of the experiment the material was still present, although the surface appeared to be changed. The efficiency (uncorrected for reflection and elec- trolyte absorption) of the MoSe, solar cell was determined to be 8% for near-infrared light.7 For the WSe, solar cell, which provides a higher voltage output,'O Bard and collaborators obtained an efficiency of 14% for light (A > 550 nm) from a xenon lamp.16 For the same system solar energy efficiencies between 3.7 and 5.2% have been reported by Heller and co-worker~.'~ However, the open-circuit photovoltage of their crystals ((0.6 V) remained clearly below the observed maximum of 0.7 V7*'0 and apparently did not have the optimal doping.The dependence of photopotentials on the redox potential of the electrolyte, which has been measured systematically7* lo shows that there is no simple linear dependence, and that iodine produces the highest effect, clearly exceeding that of other couples with comparable redox systems.We therefore do not exclude the possibility that other electron donors could be found which are even more favourable. A very attractive route of research would also seem to be the utilization of organic electrolytes in solar-energy devices, which thus eliminates the possible side reaction with water leading to sulphate or selenate. Regenerative solar cells can also be operated with p-conducting layer-type semi- conductor~.~ In this case the redox couple is not very useful, but Fe2+I3+ can shift the cathodic photocurrents considerably into positive direction (up to 0.7 V in the case of high quality p-WSe,), thus enabling energy conversion at reasonably high cell photovoltages.The high stability of layer-type semiconductors in the presence of iodine containing electrolytes together with the large photocurrent densities which can be obtained (65 mA cm-, at 150 mW ern',) l6 suggest another interesting possibility for energy con- version (bottom right scheme in fig. 4). In the case of WSe,, in which the conduction band edge is at a more negative potential than the hydrogen redox couple, p- and n- type material could favourably be combined to convert -HI into -$ H, and 3 I2 for energy storage and subsequent combustion in fuel cells, (current in fig. 4). A solar collector consisting of many microscopic patches of p- and n-type WSe, embedded into a conducting matrix could, when operated with a fuel cell, thus serve as a system for the combined generation of heat, fuel and electricity.The technological prob- lems which would have to be surmounted to reach such an aim are however quite sub- stantial. No cheap method of producing photoelectrically efficient layer-type ma- terial has yet been developed. ZIRCONIUM AND HAFNIUM DICHALCOGENIDES PHOTOELECTROCHEMICAL PROPERTIES As can be seen in fig. 1, transition metal dichalcogenides of Group IV permit only light-induced hole generation on a valence band derived from sulphur p-orbitals. Photocurrents could actually be found with sulphides and selenides of zirconium and hafnium. Most work was done with zirconium compounds which yielded photo- currents ca. 1 mA cm-2 (fig. 5).18919 ZrS, was found to be only n-conducting, which is in agreement with crystal-chemical studies of non-intentionally doped material ; ZrSe, was also obtained as a compensated or predominantly p-type semiconductor.H.TRIBUTSCH 197 Photocurrents were even obtained for TiS, which is a degenerate n-type material with an electron concentration approaching lo2’ ~ m - ~ . As to be expected, the quantum efficiencies were three orders of magnitude smaller (fig. 5). - 0.5 0 0.5 1 electrode potential’v us. SCE FIG. 5.-Photocurrent-~oltage behaviour of n-type ZrSz, n and p- (compensated) ZrSea and TiSz (note much smaller quantum efficiency; lock-in technique, 150 Hz, 10 mVs-I). Note that addition of iodide does not shift the photocurrent. (a) n-ZrSez (0.5 mol dm-3 KCl), (b) n-ZrSez (0.1 mol dm-3 KI), (c) n-ZrS2 (1 mol dm-3 HCl), (d) p(n)-ZrSez (1 mol dmd3 HCI), (e) TiSz(n-deg.) (0.1 mol dm-3 KOH).The most striking behaviour of these materials as compared to Group VI com- pounds was the generation of molecular sulphur and elemental selenium, respectively, as anodic reaction products (table 1). A second important observation was that redox systems like I-/I, did not change the photocurrent-voltage characteristic. It is interesting to note that the photo-threshold for ZrS, is at a higher energy than that of MoS, (6 eV as compared to 5 eV) 24 which indicates that hole reactions involving water should be energetically even more favourable with that material. The anodic photo- electrochemical behaviour of Group IV semiconductors thus resembles more closely that of CdS or CdSe (which also provides holes on a valence band derived from chal- cogen orbitals) than that of layer type molybdenum and tungsten compounds.SOLAR-ENERGY CONVERSION Since zirconium is the 1 1 th most abundant element on earth, being more abundant than most technically useful metals (Cu, Zn, Pb) and since it is non-toxic, its compounds might become interesting for solar-energy devices. In contrast to solid-state solar cells their utilization as electrode material in conventional regenerative electrochemical solar cells appears to be unattractive because of their anodic instability. However, as we proposed recently,ls an entirely different reaction mechanism could be utilized for a combined conversion and storage of solar energy: Layer-type Group IV transition metal compounds can easily be intercalated by alkali and alkaline-earth metals, transi-198 TRANSITION METAL DICHALCOGENIDES tion metals, organometallic complexes and Lewis-base molecules [for a review see ref.(26)]. Such an intercalation can conveniently be performed by cathodic polariza- tion27*28 and is the basis of the high-energy Li-TiS, battery How a solar powered intercalation battery could work is demonstrated in fig. 6. Its upper part shows an energy scheme of a p-conducting layer-type zirconium or haf- nium sulphide or selenide electrode in an electrochemical cell arrangement. Below, M' p ZrSz,ZrSeZ electrolyte M 0 S, Se; 0 Zr ; @ intercalating metal FIG. 6.-Above: energy scheme representing light-induced cathodic intercalation of a p-type Group IV transition metal dichalcogenide (e.g., p-ZrS2, ZrSe2) in a photointercalation solar cell. Below: a structural representation of the same mechanism.the organisation of atoms in the layer structure is given to visualize the intercalation reaction. Illumination of the p-type material produces a cathodic photocurrent (shown in fig. 5 for p-ZrSe,). In the presence of ions (M+) which can be reduced to form inter- calation compounds the intercalation reaction proceeds and changes the free energy of the electrode. It is significant that the electrode (ZrM,S,) remains semiconducting when the intercalated species are for example Cu or Fe with 0 < y < 0.22, but the energy gap decreases, depending on y.30 An important difference from a conventional intercalation battery is also the nature of the counter-electrode which should not have a negative potential like lithium but a positive one like copper.A thermodynamic analysis of energy conversion and storage by photointercalation indicates a good chance of getting practically interesting solar-energy devices provided predictable material problems can be successfully handled. It is important to note that a rever- sible mechanism of intercalation and de-intercalation (as in the case of the Li-TiS,H . TRlBUTSCH 199 system) should not involve corrosive electrode processes. Photointercalation bat- teries utilizing organic electrolytes are not the only possibility. The capacity for solar-energy storage in the form of photogenerated ion gradients could also be ex- ploited utilising solid electrolytes or simply p-n junctions of intercalated substances.This novel mechanism of solar-energy conversion and storage which is based on the combined photoelectronic and ionic properties of layer-type semiconducting material is to some degree similar to that applied in biological membranes, where (photo- induced) electron transfer generates energy-storing ionic gradients. PLATINUM DISULPHIDE PHOTOELECTROCHEMICAL PROPERTIES PtS, is a layer compound, the physical properties of which have been little investi- gated.32 Photoelectrochemically it is not only interesting because of its d-band char- acter (fig. 1) but also because of the particular catalytic properties of Pt and because this noble metal, in contrast to W and Mo, cannot easily leave the electrode as a positive ion, thus decreasing the possibility of anodic corrosion.Pt could, however, form a large variety of complexes with suitable ligands if they are present in the electrolyte or produced at the electrode. The photoelectrochemistry of PtS, (Eg = 1.03 eV) has been studied in greater detail2' but only the more striking properties can be sketched in the context of this paper. The general current-voltage characteristic of PtS, is very similar to that of Pt, with hydrogen evolution in the negative, and oxygen evolu- tion in the positive potential region. Upon illumination anodic photocurrents lead- ing to the evolution of oxygen are observed which can exceed 100 mA ern-,. Such large photocurrents are, however, obtained only near the potential region where oxygen evolution is also possible in the dark.Towards lower potentials photocur- rents are still observed, but they are much smaller. No sulphur is liberated as anodic reaction product. Even after passing large anodic currents for days the surface re- mains metallic bright. Equally, photocurrents do not decay but stay reasonably con- stant, which excludes the formation of a platinum film on the surface (parallel to a formation of soluble SO4,-). Like Pt electrodes, which pass through different oxida- tion stages when positively polarized, PtS2 equally exhibits different oxidation states which influence its electrochemical and photoelectrochemical properties. It is not yet clear how steps in the surface affect the observed behaviour of the material, but it is certain that there should be a pronounced difference in the reactivity of both surfaces (Itc and Lc).The photoelectrochemical behaviour of PtS, is also very interesting in the presence of redox systems in the electrolyte. As can be seen in fig. 7, both the onset potential and the characteristic dynamic shape of the photocurrent curves vary extraordinarily from system to system. There is clearly a systematic correlation between the onset of photocurrents and the equilibrium redox potential of the redox couples involved. POSSIBLE ANALYTICAL APPLICATION Since light can be modulated and photocurrents measured with high accuracy by phase-sensitive techniques it is obvious that the PtS, electrode could be developed as a new probe for electron exchanging substances. As we have observed that organic substances also (e.g., ascorbic acid) generate photocurrents we feel that the method might be more generally applicable.We do not exclude the possibility that p-type200 TRANSITION METAL DICHALCOGENIDES O n C Y .C1 Y 5 t: 20 CI a 0 0 = electrode potential/V us. SCE FIG. 7.-Photocurrents (relative units, magnitudes not comparable) generated by different redox systems at a PtS, electrode; current densities between 10 and 150pA cm-’, potential sweep: lOmV s-l. Lock-in technique 150 Hz. I-(C = 5 x 1O-j rnol dmP3, 2C = lo-’ mol dm-3; electrolyte: 0.05 rnol dm-3 H2S04 + 0.5 mol dm-3 K2SO4). Br-(5 x mol dm-3 electrolyte: 0.05 rnol dm-3 HzS04 + 0.5 mol dm-3 &Sod); Fe(CN)64-/3-(10-2 mol dmP3; electrolyte: 0.5 rnol dm-3 K2S04) Fe2+(J0-l rnol drn-j; electrolyte: 1 mol dm-3 H2SO4).(A) 31- +- 13- + 2e-, (B) I- + HzO + HI0 + H+, + 2e-, (C) Fez- + Fe3- + e-, (D) Br- + +Brz + e-, (E) Fe(CN6)4- + Fe(CN6)3- + e - , (F)Cl- +-)Clz + e-. PtS,, provided it can be synthesized, could show the similar photocurrent behaviour in presence of oxidizing substances. Since perfectly clean PtSz electrodes can easily be produced by detaching very thin layers of material from the surface by means of adhesive tape the system could also fulfill strict criteria for purity. DISCUSSION Although it is probable that the energy scheme (fig. 1) describing the contribution of d-orbitals to the valence and conduction bands of semiconducting layer-type tran- sition metal dichalcogenides is oversimplified and needs further refinement, manyH . TRIBUTSCH 20 1 t h e ~ r e t i c a l ~ ~ - ~ ~ and also experimental results3’ fit into such a picture so that it can be considered a reasonable basis for photoelectrochemical models.Our experimental results concerning the electrochemical and photoelectrochemical reaction behaviour of these materials are in agreement with such an electronic scheme since they permit a clear distinction between a hole reaction over a valence band derived mainly from d- orbitals and one over a valence band derived from sulphur or seleniump-orbitals. AS compared in table 1, the photoinduced reaction with water leading to sulphate and selenate is limited to semiconductors which provide holes in energy bands derived from d-orbitals. Even more striking is the pronounced shift of photocurrents which redox systems in the electrolyte induce at the electrodes.It is not possible to explain such a considerable variability of the double layer and the flat-band potential in terms of a simple adsorption model. Electrodes with identical crystal structures and identi- cal surface atoms (Ic) such as PtS, and ZrS, cannot be expected to produce entirely different surface states or defects. They do, however, provide holes in different energy bands. The striking similarity between the photoelectrochemical behaviour of PtS, on one hand and WS, and MoS, on the other hand (which are all d-band photoelec- trodes anodically reacting with water and showing characteristic shifts of the photo- current curves in presence of redox systems) supports our conclusion. The most important question which remains to be answered is that of the molecular background of an interaction between a redox electrolyte and a layer-type photoelectrode which is apparently systematic.There is interestingly no evidence that steps in the surface, which expose crystal surfaces, are responsible for this phenomenon, since photoeffects improve with the crystalline quality of the van der Waals type electrode surface and do not lose their characteristic properties.12 We have therefore to conclude that the photoreaction on the van der Waals surface is causing changes which are dependent on the redox potential of the electrolyte. In order to comprehend the phenomenon in a qualitative way we can proceed as follows: It is clear that a systematic displace- ment of the photocurrents with the redox potential of the electrolyte is not to be expected simply by taking into account conventional electron-transfer theory for semi- conductor electrolyte junctions, since the relative positions of energy levels in the electrode and the electrolyte are neither dependent on the electrode potential nor on the redox potential of the electrolyte. In this respect a PtS, photoelectrode behaves more like a metal electrode in the dark.According to L e v i ~ h ~ ~ the essential difference between the electron transfer at metal electrodes and semiconductor electrodes is the absence of the overvoltage q to the second power in the exponential in the rate con- stant for electron transfer to the semiconductor. It is therefore sufficient to consider these terms when searching for the significance of the photocurrent shifts.The rate constant for the anodic electron transfer to a semiconductor is: (Ev + eoU,,o + k z exp - ( 4AkT where Ev + eoU,,, is the energy difference between the valence band edge and the redox energy of the electron donor and 3, the reorganization energy (fig. 8). The corresponding rate constant for the electron transfer to a metal where rp is the potential in the outer Helmholtz layer measured against the bulk of the electrolyte and q the overvoltage. Since d-band photoanodes behave qualitatively202 TRANSITION METAL DICHALCOGENIDES useful energy (, i. Ev+ Q U R,* : -energy -: (I lost specific surface interaction FIG. 8.-Energy scheme depicting a semiconductor in the dark alternatively in contact with two dif- ferent redox systems (left-hand side) as well as after illumination for the case of a classical semicon- ductor (p-band) (upper picture).What is important is the different significance of the energy differ- ence Ev + eURlo for the photoreaction. like metal electrodes, exhibiting an electrochemical reaction enthalpy (AH) which is dependent on the overpotential, we can formally write eoy 3- eoq = AH + TAS = Ev 3- eoUR,, (3) (4) and AG = -e,E,, = - (Ev + eoUR,,) where AG is the free-energy change and EMF the electromotive force). This means one would be able to explain the behaviour of these layer-type photoelectrodes by assuming that during the interaction with a redox system free energy (AG) equiva- lent to an electromotive force (EMF) is liberated.The energy difference between the edge of the valence band and the redox energy, which is dissipated as heat in classical semiconductors, is thus accessible as a source of energy. It should, however, be em- phasized that relations (1) and (2) have been derived for electron transfer in the absence of a specific interaction with the electrode surface, thus making our conclusions quali- tative. Since it is difficult to imagine that electron donors of very different chemical character (e.g., I - , Fez+) can produce systematic molecular changes in the electrodeH . TRIBUTSCH 203 double layer, it is more probable that they charge surface states according to the existing redox potential. As dangling bonds should be scarce on van der Waals surfaces it is probable that surface states are formed during the photoreaction.When charged to a degree which depends on the redox potential of the electrolyte they may produce a potential jump which controls the flat-band potential of the electrode, as shown in It is reasonable to assume that the specific interaction between electron donor and electrode surface which is responsible for the remarkable effect of redox systems on the current-voltage behaviour of d-band semiconductors also facilitates the photoreaction with water. Since the electrochemical energy needed to oxidize water or OH- ions to radicals is not available at the electrode surface, it has to be assumed that a chemical- bond formation facilitating electron transfer from a surface bound intermediate is taking place. Apparently, holes on d-states (apart from being less accessible for cor- rosion mechanisms) also provide kinetically more favourable conditions for the dis- charge of OH- ions.Since such a difference in reactivity between holes on d- and p-states cannot be explained in terms of thermodynamic considerations on photoelec- trode stability as discussed by G e r i ~ c h e r ~ ~ and Bard and Wright~n,~' we have recently attempted another theoretical approach starting from a kinetic equation for oxidation reactions derived for metal electrodes and modifying it for illuminated semiconductor- electrolyte Junctions.19 The relatively complicated result can be simplified to the fol- lowing condition, which would have to be fulfilled for an anodic photoreaction (photo- corrosion, oxidation of water or of a reducing agent) to occur.fig. 8. The first term in this relation is the electric energy available in the interface (qSE is the potential drop in the double layer of the semiconductor-electrolyte interface). The second term is the activation energy needed for the oxidation reaction. The third term is the energy gap and the fourth the difference between the Fermi energy of the electrode and the energy corresponding to the reversible potential of the proceeding electrochemical reaction [redox reaction ( E r e d o x = - F u r e d o x ) or decomposition reac- tion (&dox = -FU,,,,,,)]. The constant on the right-hand side is dependent on the reaction entropy, the work function of the electrode and the light intensity. It would thus only be a constant for electrodes with a similar free energy of electrons involved in the same type of electrochemical reactions.Relation (5) permits the following qualitative conclusions : (1). When MoS, and ZrS, are compared, all quantities apart from the first two can be considered to be comparable. The different reaction behaviour is thus basically of kinetic origin: ZrS, has a low activation energy AH# for photo-oxidation to mole- cular sulphur and MoS, a low activation energy for the photoreaction with water. (2). Surface areas [I c which are exposed at crystal steps produce surface states with donor character which can accumulate excessive positive charge which will affect the potential distribution in the same way as thin metal deposits on semiconductors.12 The potential drop qSE at such interfaces will be bigger than that at van der Waals surfaces and since it enters the first term in relation (5) this will favour anodic reactions. Since activation energies could equally be lower at such surfaces, they should be the principal sites of corrosive anodic attack, which is experimentally confirmed.Rela- tion (5) also permits some more general conclusions which could be helpful during the search for photoelectrodes which are either stable or oxidizing water: since the energy gap A& enters (which may vary between 1 and 4 eV) it is evident that kinetic factors as expressed in the activation energy (the magnitude of which will rarely exceed 1.2204 TRANSITION METAL DICHALCOGENIDES eV) will lose their importance for large gap semiconductors such as Ti02 or SrTi03.If their decomposition potential is sufficiently positive, oxygen evolution from water will be more favourable than photocorrosion even with a high activation barrier for the first and a low one for the latter. If however, visible light of lower energy is to be used for photoelectrolysis or photoassisted electrolysis kinetic factors will become predominant. In this case it will be necessary to search for electrodes which have a high activation energy for corrosive decomposition and a low one for the photo-oxida- tion of water. PtS2, with an energy gap of ca. 1 eV, apparently has a high activation energy for decomposition. It can therefore photoreact with water and evolve oxygen. However, because of the charging of the interface, which we discussed before, appa- rently stemming from a relatively high overpotential for oxygen evolution (which is also typical for Pt electrodes), reasonably high photocurrent efficiencies for O2 evolu- tion are reached only at higher electrode potentials.20 Semiconductors with com- parable electronic structures but originating from metals which are especially active catalysts for the oxidation of water like Ru, 0 s or Ir should be more favourable in this respect.We have shown in this report that, because of their peculiar electronic structures, layer-type transition metal dichalcogenides constitute an interesting testing ground for photoelectrochemical mechanisms and concepts. The promising possibilities con- cerning practical application for solar energy conversion and analytical methods which become evident already at this very early and rudimentary stage of photoelectrochemi- cal knowledge deserve further detailed studies.The author is supported by a fellowship of the “ Heisenberg-Program ” of the He thanks Dr. R. Parsons for stimu- Deutsche Forschungsgemeinschaft, G.F.R. lating discussions. H. Tributsch, Z . Naturforsch., 1977,32A, 972. H. Tributsch and J. C. Bennett, J . Electroanalyt. Chem., 1977, 81, 97. H. Tributsch, Ber. Bunsenges. phys. Chem., 1977, 81, 361. H. Tributsch, J . Electrochem. SOC., 1978, 125, 1086. H. Tributsch, Ber. Bunsenges. phys. Chem., 1978, 82, 169. J. Gobrecht, H. Gerischer and H. Tributsch, J . Electrochem. SOC., 1978, 125, 2085. H. Tributsch, T. Sakata and T.Kawai, Electrochim. Acta, in press. J. Gobrecht, H. Gerischer and H. Tributsch, Ber. Bunsenges. phys. Chem., 1979, 82, 1331. 655. ’ H. Tributsch, Solar Energy Mater., 1979, 1, 257. lo H. Tributsch, H. Gerischer, C. Clemen and E. Bucher, Ber. Bunsenges. phys. Chem., 1979, 83, l1 J. Gobrecht, Thesis (Technical University, W. Berlin, 1979). l2 W. Kautek, H. Gerischer and H. Tributsch, Ber. Bunsenges. phys. Chem., 1979, 83, 1000. l3 W. Kautek, Thesis, (Technical University, W. Berlin, 1980). l4 W. Kautek, H. Gerischer and H. Tributsch, J. Electrochem. SOC., 1980, 127, 2473. l5 B. Scrosati, University of Rome, personal communication of results, 1980. l6 F. Fan Fu-Ren, H. S. White, B. Wheeler and A. J. Bard, J . Electrochem. SOC., 1980, 127, 519. l7 H. J. Lewerenz, A. Heller and F. J. Di Salvo, J. Amer. Chem. SOC., 1980, 102, 1877. l9 H. Tributsch, J . Electrochem. Soc., submitted. 2o H. Tributsch and V. Gorochov, Electrochim. Acta, submitted. 21 J. A. Wilson and A. D. Yoffe, Adv. Phys., 1969, 18, 193. 22 A. D. Yoffe, Festkorperprobleme, 1973, XIII, 1. 23 B. G. Silbernagel, Mater. Sci. Eng., 1977, 31, 281. 24 Physics and Chemistry of Materials with Layered Structures, ed. E. Mooser (D. Reidel, Dor- 25 C. R. Whitehouse and A. A. Balchin, Phys. Status Solidi (a), 1978, 47, K 173. 26 F. R. Gamble and T. H. Geballe, Treatise on Solid State Chemistry (Plenum Press, New York, H. Tributsch, Appl. Phys., in press. drecht, 1979), 5 volumes. 1970, vol. 3, p. 89.H . TRIBUTSCH 205 27 G . V. Subbor Rao and J. C. Tsang, Mater. Res. Bull., 1974, 9, 921. 2 9 M. S . Whittingham, Science, 1976, 192, 1126. 30 B. G. Yacobi, F. W. Boswell and J. M. Corbett, J . Phys. C, 1979,12, 21 89. 31 J. A. Wilson and A. D. Yoffe, Adv. Phys. 1969, 18, 193. 32 C. Mankai, G . Martinez and 0. Gorochov, Phys. Rev. B, 1977, 16,4666. 33 L. F. Mattheiss, Phys. Rev. 1973, B8, 3719; and Phys. Rev. Letters, 1973, 30, 748. 34 R. V. Kasowski, Phys. Rev. Letters, 1973, 30, 1175. 35 K. Wood, and J. B. Pendry, Phys. Rev. Letters, 1973,31, 1400. 36 C. Y. Fong, and M. Schliiter, Electronic Structures of Some Layer Compounds in Physics and Chemistry of Materials with Layered Structures, ed. E. Mooser (D. Reidel, Dordrecht, 1979), vol. 3. V. G. Levich, in Advances in Electrochemistry and Electrochemical Engineering, ed. P. Delahay (Academic Press, New York, 1966), vol. 4. M. S . Whittingham, U.S. Patent 4 040 917, 1977. 37 R. S. Title and M. W. Schafer, Phys. Rev. Letters, 1972, 28, 808. 39 H. Gerischer, J . Vac. Sci. Technol, 1978, 4, 15. 40 A. J. Bard, and M. S . Wrighton, Proc. Electrochem. Soc., 1977, 77, 3.
ISSN:0301-7249
DOI:10.1039/DC9807000189
出版商:RSC
年代:1980
数据来源: RSC
|
15. |
Photoelectrochemical systems with energy storage |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 207-222
Peter G. P. Ang,
Preview
|
PDF (1158KB)
|
|
摘要:
Photoelectrochemical Systems with Energy Storage BY PETER G. P. ANG AND ANTHONY F. SAMMELLS Institute of Gas Technology, 3424 South State Street, IIT Center, Chicago, Illinois 60616, U.S.A. Received 14th May, 1980 Photoelectrochemical cells having the capability of energy storage are covered. Photoanodes used for these cells were selected from MoSe,, GaAs and CdSe. Redox electrolyte species present at the semiconductor-electrolyte interface included Br-/Brz (MoSez), S’-/Sf- (CdSe) and Se2-/Se$- (GaAs). Electroactive materials at the counter-electrode of the storage systems were 1-12, Se2-/Se$- and Cd. The performance of these photoelectrochemical storage cells for charge upon photoanode illumi- nation and their subsequent electrochemical discharge is discussed. Photoelectrochemical cells possess several potentially significant advantages over their solid-state photovoltaic counterparts which may give us the opportunity for more convenient storage of the incident photon energy as chemical species.Such chemical species could then be electrochemically discharged for the production of electricity. The formation of the Schottky-like barrier at a semiconductor-liquid junction in these cells is produced simply by immersion of a semiconductor electrode into a redox electrolyte. This is in comparison with a sequence of highly-controlled steps necessary for the formation of, for example, a p-n junction in a solid-state cell. To minimize electron-electron hole recombination reactions in solid-state devices, the use of high-purity materials is required to reduce the number of trap sites.For silicon-based solid-state cells, solar conversion efficiencies have been increased by the use of narrower junction depths together with lower surface doping concentrations and anti-reflection treatments. Power conversion efficiencies achievable with photo- electrochemical cells are beginning to approach those values realized for solid-state photovoltaic devices. Efficiencies for polycrystalline and single-crystal gallium arsenide liquid-junction devices have been reported, respectively, at 7.8 and 19%.2 Regenerative photoelectrochemical cells, where no overall change in the redox electrolyte composition occurs, can be used for the direct generation of electricity. Cells of this type usually consist of an illuminated n-type photoanode and a counter- electrode.Photopotentials and corrosion stabilities for p-type photocathodes have, in general, been lower than for n-type materials. For cells directly producing electricity, as in the case of solid-state photovoltaic devices, energy storage would probably involve external storage batteries. For conventional battery storage only lead-acid and nickel-iron would appear technically feasible, at the present time. Many advanced batteries are presently under development for utility load levelling storage, which, if successful, would be available for the direct storage of photovoltaic energy. Much work remains to be performed, however, before advanced batteries would be made available for photovoltaic system integration. Redox batteries may be attractive for storage of photovoltaic energy since they are capable of incrementally varying their storage capacity.Liquid-junction devices can be used not only for the direct generation of electrical208 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE energy, but also for the production of either oxidized or reduced chemical species. As a consequence, the photogeneration of such redox species gives us the opportunity of designing and operating photoelectrochemical cells with a built-in storage capacity. Redox battery technology is compatible with the photoelectrochemical storage systems discussed here. The origin of photopotentials (and corresponding photocurrents) upon illumina- tion of a photoelectrochemical cell has been adequately documented by other^,^,^ and requires no detailed explanation here.Empirically the process involves a bending of the semiconductor valence and conduction bands at the interface, which occurs due to an equilibration of Fermi level differences between the semiconductor and the redox electrolyte, when the semiconductor is immersed in the electrolyte. This forms a space-charge layer near the surface of the semiconductor which, because of its in- herent electric field, can effect a separation of photoproduced electrons and electron holes. The solar-conversion efficiency is partly dependent upon the useful light absorbed within the space-charge The width of the space-charge region, which is influenced by doping concentration and the voltage drop in the space-charge region, should ideally be equal to, or larger than, the mean reciprocal absorption coefficient.Illumination of photoanodes creates a negative photovoltage by the creation of a net negative charge on the semiconductor. Hence, redox species present at the interface can be oxidized at potentials negative of their equilibrium value. The photopotential, which is the difference between the potential of the illuminated photoanode and the redox equilibrium potential of the counter-electrode, allows electrical power to be extracted from the device upon application of an external load. To a first approximation, the optimum cell potential is determined by the difference between the flat-band potential and the redox potential of the electrolyte. Since the flat-band potential is the most cathodic value the illuminated photoanode can attain, this parameter should ideally be as negative as possible for optimization of the cell voltage.The flat-band potential can be quantitatively related to the electron affinity of the semiconductor,8 which can be calculated from the atomic electronega- tivities of the constituent atoms. In practice, because of losses caused by thermal and entropy effects within the space-charge layer upon photon absorption, the initial available energy to drive the photoelectrochemical oxidation or reduction reactions is dictated by the difference between the electron and electron-hole quasi-Fermi l e v e l ~ . ~ Several excellent reviews have recently appeared covering the field of photo- electrochemi~try.~~~~-~~ Upon substitution of a single electrolyte photoelectrochemical cell for one con- taining two redox or electroactive materials which are separated by an ion selective membrane, the energy of the incident photon at the photoelectrode can be stored as chemical energy.Design of such cells requires both a stable low resistant separator which minimizes direct chemical reaction of the electroactive species, and the selected redox couples must be able to stabilize the semiconductor. As is well known, photoelectrode stability is dependent upon the relative energy-level positions between electrons, electron holes, and the decomposition potential of the photoelectrode itself. For an illuminated n-type semiconductor, photoanodic corrosion can occur if the quasi-Fermi level for holes is positive of the semiconductor decomposition p ~ t e n t i a l .~ ? ' ~ For a two-electrolyte photoelectrochemical cell to achieve optimum stability, the relative energy levels of the donor redox species (Eredox 1) at the photoanode and that of the acceptor redox species or electroactive material at the counter-electrode (Eredox 2), will be required to lay, respectively, negative of the semiconductor decomposition potential and below the lower edge of the conduction band, as shown in fig. 1. If surface states14*15 for the capture of electron holes are present within the semi-P . G . P . ANG AND A. F . SAMMELLS 209 conductor band-gap, reaction of the holes with a redox species may become pinned at a potential more negative than expected, and therefore protect the semiconductor from photoanodic corrosion.The stability of TiO, as a photoanode for oxygen evolution, has been attributed to this effect. Increased stability for narrow band-gap photoanodes has been demonstrated also by derivatization of their surface. Most work has been with n-silicon, with surface attached ferrocene derivatives.16 The I I I EBG metal n - type semiconductor i I rc B I I E r e d o x l I I I [decornp I I I fvs I I electro- I electro- lytet I 1yte2 I I metal \separator FIG. 1 .-Energy-level diagram of an illuminated n-type semiconductor1 redox I lredox 2 Icounter- electrode system. (A diode in the circuit prevents self-discharge through the photoanode in the dark). objective here is to achieve rapid kinetics for oxidation of the surface derivatized group, compared with those for semiconductor oxidation. Illumination of such photoanodes results in the transient oxidation of the surface attached group, followed by oxidation of the redox species of interest. For many semiconductors, their vulnerability to photoanodic corrosion effects can, in part, be attributed to the photoexcitation of electrons from their valence to con- duction band, involving bonding orbitals.This can lead to transient bond rupture. Recently,17 a new class of semiconductors has been suggested for photoelectrochemical cells based upon the transition-rnetal dichalcogenide layer compounds. The promise of increased stability to photoanodic corrosion by these materials, over other narrow band-gap candidates comes from the orbitals involved in the electron excitation pro- cess.Here, photon-induced electron transitions from valence to conduction bands occur predominantly via non-bonding d-d orbital~,l~-*~ and consequently photo- transitions do not involve bond rupture. n-Type MoSe, appears as an attractive material for the efficient photoelectrochemical oxidation of halide redox couples.210 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE Having a band-gap of 1.4 eV, it has a good match to the solar spectrum. By intro- duction of reversible redox couples, cathodic of the oxygen potential, long lifetimes have been demonstrated.'* Several photoelectrochemical cells which might facilitate storage have been sug- gested. Photoelectrochemical cells using third storage electrodes have been pro- In the first, a storage electrode (Ag/Ag2S) separated from the photoanode and counter-electrode by a cation specific separator was charged by an illuminated polycrystalline CdSe photoanode, increasing the cathodic activity of the metal sulphide for later electro- chemical discharge via a counter-electrode.In a second cell configuration, reported by the same workers, two illuminated photoelectrochemical cells were connected electrically in series to achieve sufficient potential for the cathodic deposition of either tin or zinc onto a third electrode. This approach to energy storage has problems inherent in many electrochemical devices, where chemical phase changes occur upon cycling, resulting in electrode shape change and the inception of dendritic growth, possibly leading to cell shorting.For the systems covered in this work, the oxidized species produced at the photo- anode interface and the corresponding reduced species at the counter-electrode are Two general cell configurations have been reported. 5 FIG. 2.-Schematic diagram of the photoelectrochemical storage cell system. 1 , two cuvettes clamped together; 2, separator; 3, n-type photoanode; 4, counter-electrode; 5 , discharge electrodes ; 6, sealed caps. stored for later electrochemical discharge. The electroactive material at the counter- electrode can be either a soluble redox species or an insoluble active material. Of the insoluble materials we have evaluated Cd since it has demonstrated good electro- chemical cycle performance in alkaline cells. Selection of redox couples for photoelectrochemical cells in this work was based partly upon their previous demonstration of fast electrode kinetics.The redox species also should have high solubilities and ionic conductivities in the electrolyte. Photon absorption in the wavelength region of interest by the electrolyte should be low, Work reported here is on photoelectrochemical cells with storage capability for their photoproduced electroactive species. Cells based upon the semiconductor materials MoSe,, GaAs and CdSe are discussed, together with their photocharge and electrochemical discharge characteristics.P. G . P . ANG AND A . F . SAMMELLS 21 1 EXPERIMENTAL Single-crystal n-type MoSe, was obtained from Prof. Dr. Rudolf Nitsche, Universitat Freiburg, West Germany.The crystals were grown by the halogen transport method. The large faces are [OOl], formed by hexagonal al and a, axes. Contact was made by electro- plating indium. A platinum wire was then attached to this side using a conducting silver epoxy (Cerac, Inc.). The silver epoxy was also allowed to contact the edges of the front sur- face to reduce series resistance. An insulating layer of paraffin-based wax was later applied to cover the silver epoxy and the platinum wire so that only the illuminated front surface of the crystal was exposed to the electrolyte. Single-crystal n-type GaAs (Si-doped) was obtained from Crystal Specialties, Inc. The surface was oriented in the [l 111 direction and had an unoptimized dopant concentration of 1.3 x lo'* ~ m - ~ . Ohmic contact was made by electroplating indium and gold and annealing it at 400 "C under hydrogen for 2 h.The crystal was attached to a platinum or a nickel wire as described above. The photoanode was etched for 10-15 s with a 1:l mixture of 30% Hz02 + H2S04 to matte black and dipped in 0.01 mol dm-3 RuCl, + 0.1 mol dm-3 HNO, solution for 10 s to improve its performance.28 n-type polycrystalline CdSe was made by thermal vacuum evaporation of CdSe powder (99.999%, Cerac, Inc,) on a porous titanium substrate (Gould P/M Ti 6525). This was accomplished by using an Edwards 306 thin-film evaporator at a pressure of Torr. The electrode was annealed in air at 400 "C for 30 min and cooled down slowly. The electrode was then contacted to a nickel wire with silver epoxy and the assembly was coated with epoxy for insulation.The storage cell consisted of two cuvette cells as shown in fig. 2 . A hole of ca. 0.8-cm diameter was drilled in one face of each cuvette and a thin layer of paraffin wax was applied around the hole. The two cuvettes with the separator/membrane in between were clamped together using Plexiglass frames. The outside of the cuvette near the region where the separator was being held was sealed with paraffin wax. Nafion, the cell separator materials, were obtained from Du Pont de Nemours. Nafion consists of a hydrophylic perfluorosulphonic acid. Negatively charged sulphonic groups are attached to the fluorocarbon polymer chains within the Nafion. The membrane is permeable to cations and relatively impermeable to anions.A platinum elec- trode (1 cm2 Pt foil attached to Pt wire) was used in each compartment of the redox cell. Typically, ca. 3 cm3 of redox electrolyte was placed into each cuvette. The cuvettes were closed with caps and sealed with paraffin wax. The membrane was equilibrated with the electrolyte solutions for about one day before measurements were performed since wetting of the membrane reduced cell resistance. Another set of platinum electrodes (1 cm2), one in each compartment, was present, either for monitoring the cell voltage or for cell discharge. The photoelectrochemical properties of the semiconductors were investigated using a Wenking ST-72 potentiostat and an Oriel 150 W xenon light source. The electrode potential was measured using a Wenking PPT-70 electrometer.A Tacussel-type GSTP2B pulse-sweep generator was used to control the potentiostat. The current-voltage relationships of the electrodes were recorded on a Hewlett-Packard 7046 X- Y recorder. The long-term MoSe, experiments were performed using an H3 quartz-halogen lamp (12 V, 55 W). The light was heat-filtered and focused onto the crystal. A small piece of separator was mounted over the hole in one cuvette. RESULTS AND DISCUSSION Photoelectrochemical cells with storage will be discussed using single-crystal MoSe,, GaAs and polycrystalline CdSe as photoanodes. The photoresponse of the photoanodes and their charge-discharge characteristics in a storage-cell system will be described individually.212 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE n-MoSe, SINGLE- CRY ST AL PHOTO ANODES Molybdenum diselenide has previously been shown to be a stable photoanode in iodide/iodine redox electrolyte, where photopotentials of ca.-500 mV were demon- strated. Here, solar-energy conversion efficiencies of 8 % in monochromatic light, with lifetimes over 1 year at current densities of 10 mA cm-, were achieved.18*21 This work was performed with single-crystal material. For layer compounds with imperfect surface morphologies there is some evidence that as a consequence of the large anisotropy in the conductivity, a large fraction of electron holes might become deflected to edge recombination sites29 rather than participating in oxidation of a donor species at the interface. Little success has been achieved to date in fabricating polycrystalline layer compounds having a good photoresponse, quite possibly result- ing from this recombination effect.To identify a photoelectrochemical cell capable of being charged, using n-MoSe, in 1-/1, redox electrolyte, it is necessary to identify acceptor redox couples having standard potentials within 500 mV negative of 1-/12 that also possess fast kinetics for electron exchange. It would also be desirable that the acceptor have similar charge polarity to that of the donor so that effective separation across the cell might be achieved. Up till now no suitable acceptor redox couple meeting the above criteria has been identified. As a consequence, the Br-/Br2-I-/12 redox storage system was considered in a cell. Here, the photoanode was in contact with the bromine redox electrolyte, while the iodine redox electrolyte at the counter-electrode was at ca.500 mV negative of the bromine potential. The actual cell voltage will, of course, be dictated by the relative activities of the redox species present. We have investigated the photoelectrochemical performance of n-MoSe, in bromine both in alkaline and in acid electrolytes. For the alkaline solution 6 mol dm-3 KOH + 1 mol dm-3 KBr + 1 mol dm-3 Br, photopotentials were significantly lower (ca. -200 mV) than those found for the acid redox electrolyte. This may be explained by the presence of Br0,- species in alkaline electrolyte (redox potential of the electrolyte was +415 mV us. SCE) which probably changed the photo-oxidation reaction mechanism at the photoanode relative to that in acid electrolyte.It is well known that conduction parallel to the surface of the van der Waals22 plane is higher than conduction from layer to layer in the crystal. It was anticipated that partial front surface contact would reduce the series resistance of the electrode assembly. Therefore, the silver epoxy was also allowed to contact the edges of the front surface with appropriate insulation. This photoanode was then evaluated in the redox electrolyte 1 mol dm-3 HBr + 1 mol dm-3 Br, inside a small glass cuvette. Initial open-circuit potential us. a platinum reference electrode in the same electrolyte was -130 mV in the dark and -480 mV under 200 mW cmw2 illumination. The crystal was then subjected to an anodic polarization treatment at ca.1 V whereby a current of ca. 5 mA cm-2 flowed in the dark. After anodizing for ca. 1 h, the open- circuit potential in the dark became -30 mV us. platinum, i.e., approaching more ideal behaviour. This anodization technique may have passivated the steps and other imperfections of the crystal surface, perhaps by the formation of a passivating oxide layer. The current-potential characteristics of this crystal are shown in fig. 3. Here the open-circuit potential under 200 mW cm-, illumination was -490 mV us. the platinum electrode. Short-circuit current densities of ca. 50 mA cm-, were achieved. The power conversion efficiency was 6.2% and the fill-factor 0.55. The long-term performance of n-MoSe, single-crystal in the acid electrolyte (1 mol dme3 HBr + 1 mol dm-3 Br2) under an applied constant anodic current density Work reported here is on single-crystal MoSe,.P .G . P . ANG A N D A . F . SAMMELLS 213 of 5 mA Using 100 mW cm-, illumination (quartz-halogen lamp with a Schott KG-2 filter) the potential of the photoanode vs. the platinum electrode in this regenerative system stayed close to -450 mV for over 70 days (fig. 4), demonstrating good photoanode stability in this acid electrolyte. The open- circuit photopotential and current-voltage characteristics remained essentially con- stant during this period. The performance of n-MoSe, was also evaluated in 6 mol dm-3 HCl with various was evaluated. - 500 -2 50 E us. Pt/mV 0 FIG. 3.-Current-potential and power-potential characteristics of single-crystal MoSe, in 1 rnol dmd3 HBr 4- 1 mol dm-3 Br,.200 mW cm-’ xenon light illumination. (E dark = -30 mV; E Iisht = -490 mV DS. Pt; Eet = 810 mV us. SCE; q = 6.2%; FF = 0.55; is, = 49 mA cm-’). metal redox couples, including Fe2+/Fe3+, Cu+/Cu2+, Sb3+/Sb5+ and Sn2+/Sn4+. Here photopotentials on the order of 100-200 mV were observed. In particular, anodic photocurrents were found to be smaller for the Fe2+/Fe3+ and Sb3+/Sb5+ redox couples than with Cu+/Cu2+ and Sn2+/Sn4+. In general, however, photo- potentials and short-circuit current densities obtained with these couples were inferior to those achieved with Br-/Br, acid electrolyte. Single-crystal n-MoSe, was used to charge a bromine-iodine photoelectro- chemical redox storage system. A cell of configuration MoSe211 mol dm-3 HBr + 0.01 mol dm-3 Br,lNafion 31511 mol dm-3 HI + 0.18 mol dm-3 1,IPt was assembled214 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE in a two-cuvette cell (fig.2) as described in the experimental section. The illuminated photoanode surface area was 0.2 cm2 and that of the platinum foil electrode in the iodine compartment was 2 cm2. Another set of platinum electrodes (2 cm2), in each compartment, was used for monitoring the cell voltage and for cell discharge. Upon shorting the MoSezphotoanode to itscounter-electrode in the other compartment, a cell -500 -I > E 8 . + 3 0 -300 I 1 r 1 I 0 20 40 60 80 time/day FIG. 4.-Voltage of single-crystal MoSe, us. Pt in 1 mol dm-3 HBr + 1 mol dm-3 Br, during a long- term experiment under constant-current load of 5 mA cm-, [quartz-halogen illumination, heat- filtered (100 mW cm-').Electrolyte was not stirred. Temperature 28 "C]. discharge current of ca. 0.2 mA passed. However, upon illumination with approxi- mately 200 mW cm-2 xenon light, the current direction reversed and a photocurrent of ca. 0.8 mA flowed, resulting in charge of the bromine-iodine cell. Thus, the system could be charged photoelectrochemically and discharged electrochemically upon demand. A sequence of charge-discharge cycles is shown in fig. 5. Here, the redox cell was initially charged photoelectrochemically for ca. 6 h. During this period the charging photocurrent dropped from 0.8 to 0.5 mA, while the cell voltage, as measured 8 4 3 - 0 + 4 < 4 8 12 0.6 3 0.4 8 0.2 > 0.0 G I I I I I ! I I Ii li 2 4 6 0 0.5 1 0 4 8 12 0 1 time/h A 2 FIG.5.-Current and voltage of the redox storage system MoSe,ll mol d ~ n - ~ HBr + 0.01 mol dm-3 Br,INafion 31511 mol dm-3 HJ + 0.18 mol dmP3 I,IPt during photoelectrochemical charge upon illumination (200 mW cm-, xenon light) of the MoSe, (0.2 cm2) and during discharge through a pair of platinum (2 cm2 each) electrodes with 10 R load. Cell voltage was measured between the platinum electrodes. Electrolyte (2 cm3 in each compartment) was unstirred, temperature 27 "C.P . G . P . ANG AND A . F . SAMMELLS 215 by the auxiliary platinum electrodes, rose from 0.45 to ca. 0.49 V. Discharge of the cell using a 10 0 load across the auxiliary platinum electrodes resulted in an initial current of 15 mA. After ca. 1 h this current dropped to 1 mA.The cell voltage between the auxiliary platinum electrodes also decreased from 0.49 to ca. 0.03 V. Upon returning to open-circuit the cell voltage was recorded as 0.45 V. The system was charged again by illuminating the photoanode for ca. 16 h. Subsequent to this extended photocharge, the initial cell current and voltage during electrochemical discharge were improved, as illustrated on the right-hand side of fig. 5. This was a consequence of the higher bromine and iodide activities present after electrochemical charge. The state of charge of the Br-/Br, electrolyte could also be seen visually. During charge the initially yellow redox electrolyte became darker in appearance. By using a 22.5 15 7.5 N I ; o 3 - 7.5 -15 1 I 0 250 500 750 1000 -22.5 E/mV FIG.6.-Potentiodynamic (50 mV s-') current-voltage characteristics of the PtlBr-IBrz - I-II21Pt redox cell using 0.5 cmz Nafion 315 separator. The electrolytes were 1 mol dm-3 HBr + 0.01 mol dm-3 Br, and 1 mol dm-3 HI + 0.18 mol dm-3 12, respectively in each cuvette. The solutions were unstirred ; temperature 27°C. smaller load for discharge, e.g., 200 s1, cell currents of ca. 1 mA at 0.3 V could be drawn for many hours from the system. The current-voltage characteristics of a Br- I Br2-1- [I, redox storage cell upon electrochemical charge and discharge using smooth platinum electrode is shown in fig. 6. Nafion 3 15 was found quite effective as a cell separator for preventing direct chemical reaction between the redox species. From the slope we calculated a polariza- tion resistance of 20 SZ cm-2.Upon performing over 90 continuous charge-discharge cycles (48 min duration per half-cycle) at 5 mA cm-2, electrochemical energy effi- ciencies of 75% were achieved.216 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE n-GaAs s I N G L E - c R Y ST A L PHOTO A NODE The photoelectrochemical performance of a gallium arsenide photoanode was evaluated in the redox electrolyte 1 mol dm-3 Na2Se + 0.1 mol dm-3 Se + 1.0 mol dm-3 NaOH. The current-voltage characteristics are shown in fig. 7. The open- circuit potentials in the dark and under 100 mW cm-2 xenon illumination were, respectively, -190 and -500 mV. Short-circuit current of 16 mA cm-2 and power conversion efficiency (uncorrected for redox electrolyte absorption) of ca. 4% and fill factor of 0.53 was measured.The doping and quality of this photoanode was not I light dark -500 -250 E us. Pt/mV 0 FIG. 7.-Current-potential characteristics of single-crystal n-GaAS in 1 mol dm-3 Na2Se + 0.1 mol dm-3 NaOH under 100 mW cmd2 xenon illumination (Edark = - 190 mV; Ellght = - 500 mV us. Pt ; Ept T- - 690 mV us. SCE; q = 476; FF = 0.53; is, = 16 mA cm2 uncorrected for light absorption loss in the electrolyte). optimum. Large efficiencies have been previously reported with single-crystal GaAs treated with ruthenium ion chemisorption where solar conversion efficiencies of 12% have been observed.28 It appears, however, that the greatest enhancements in efficiencies after ruthenium treatment have been observed with polycrystalline materials. Introduction of ruthenium ions onto the surface of a chemically vapour- deposited polycrystalline GaAs has resulted in solar-energy conversion efficiencies up to 7.8X.l Increased fill factors after such treatment3* has been shown to result from reduction of the number of carrier trap sites, by the use of charge-collection scanning electron microscopy t e c h n i q ~ e s .~ ~ With photopotentials of around -500 mV observed at this photoanode, redox couples which are within - 500 mV of the selenidelpolyselenide equilibrium redox potential could be used in the counter-electrode compartment. This restricts the number of feasible counter-electrode species available for charge by this photoanode. The redox potential of cadmium in 2 mol dmW3 NaOH is ca. -1050 mV us. SCE or ca. 300 mV negative of the selenide/polyselenide couple in alkaline solution.Cad- mium electrodes in alkaline electrolytes have demonstrated excellent cycling and life-P. G . P. ANG AND A . F. SAMMELLS 217 time characteristics in nickel-cadmium batteries. The cadmium electrode used in these studies was selected from this type of battery. A storage cell of configuration GaAsll rnol dm-3 Na,Se + 0.1 mol dmW3 Se + 1 mol dm-3 NaOH{Nafionl2 mol dm-3 NaOHlCd was assembled in a two-cvvette cell. Initial open-circuit potential of the n-GaAs electrode us. Cd was 350 mV in the dark and upon illumination became -100 mV. The initial cell voltage is, of course, dependent upon the charge state of the electroactive materials. At 0 mV us. Cd counter-electrode (ie., short-circuit) a cathodic current density of 0.3 mA cm-2 passed in the dark, as shown in fig. 8.Upon 0 200 time/s 4 00 FIG. 8.-Current of n-GaAs in the system 1 mol dm-3 Na,Se + 0.1 mol dm-3 Se + 1 mol dm-3 NaOHINafionl2 mol dm-3 NaOHlCd at 0 mV between GaAs and Cd. The current reversed its sign upon illumination, charging the system. illumination of the photoanode (0.1 cm2) an anodic charging current of ca. 8 mA cm-2 passed initially. This demonstrates the feasibility of illuminated n-GaAs for charging the Se2-/Seg--Cd system. After ca. 1 min, the current density dropped to half this value due to mass-transfer limitation. 15 I I I 1 1 I t N I : a 2 i l l . I I I . EIV FIG. 9.-Potentiodynamic (40 mV s-') current-voltage characteristics of the system Ptl 1 rnol dm-3 Na2Se + 0.1 mol dm-3 Se + 1 rnol dmb3 NaOHINafion 31512 mol dm-j NaOHICd at 27°C.Un- 0 0.1 0.2 03 0.4 0.5 0.6 0.7 stirred electrolyte. Electrodes were platinized platinum (1 cmZ) and porous cadmium (1 cm*).218 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE The current-voltage characteristics of the redox cell platinized Pt[ 1 mol dm-3 Na,Se + 0.1 mol dm-3 Se + 1 rnol dm-3 NaOH[Nafion[2 mol dm-3 NaOHlCd are shown in fig. 9. As can be seen from the polarization curve, this redox couple shows good electrochemical reversibility, although with the unstirred electrolyte used significant polarization losses were observed. POLYCRYSTALLINE n-CdSe PHOTOANODE Power conversion efficiencies for polycrystalline CdSe in the presence of chalco- genide/polychalcogenide redox have been reported by others as 5%32 and 6%.33 Although encouraging lifetimes have been demonstrated, some deterioration in photo- anode performance has been attributed to chalcogenide exchange 34 reactions between the photoanode and the electrolyte.Such exchange reactions have been reduced by the introduction of selenium to the sulphide/polysulphide solution which can pre- cipitate any photoproduced cadmium ions back to lattice sites in the cadmium sele- nide.l0 For single-crystal CdSe, efficiencies of 8% have been reported.35 30 1 5 - N I 3 4 E 2 0 - 1 I I I -450 -350 -250 -150 -SO +50 E us. Pt/mV FIG. 10.-Current-potential characteristics of CdSe, thermal vacuum-evaporated on porous Ti electrode (electrolyte 1 mol dm-3 Na,S + 1 rnol dm-3 S + 1 rnol dm-3 NaOH + 0.1 mol dm-3 Na,Se + 0.01 rnol dm-3 Se, unstirred, 100- mW cm-2 xenon illumination, efficiency = 4%, fill factor = 0.45, annealed in air at 400 "C for 30 min.slow cooling, etched in 6 mol dm-3 HC1 for 15 s). In the work reported here performance and stability of thermally evaporated n-CdSe was found to be dependent on the nature and morphology of the substrate used. The photo- electrochemical performance of n-CdSe is shown in fig. 10, where photopotentials of -400 mV were achieved under 100 mW cm-2 xenon light illumination. A short- circuit current density of ca. 20 mA cm-2 and power conversion efficiency of 4% were calculated, The use of porous substrates Best results were obtained using porous titanium as a substrate. The fill factor, however, was ca. 0.45.P .G . P . ANG AND A. F . SAMMELLS 219 had the advantage of greater CdSe adhesion, and more effective photoanode area available for illumination. Here the redox potential of the alkaline sulphide/poly- sulphide was ca. -700 mV us. SCE. Thus, with a photopotential of -400 mV, electroactive species with redox potentials between -700 and -1100 mV us. SCE might be suitable in the counter-electrode compartment of a storage cell. The two electroactive materials for the counter-electrode evaluated in this work have been Se2-/Sei- (-800 mV us. SCE) and Cd (- 1050 mV us. SCE). The two-compartment storage cell n-CdSell mol dm-3 Na,S + 1 rnol dm-3 S + 1 rnol dm-3 NaOH INafionll rnol dm-3 Na,Se + 0.1 rnol dm-3 Se + 1 rnol dm-3 NaOHIplatinized Pt was assembled. Surface areas of the photoanode and platinum electrodes were, respectively, 0.25 and 1 cm2.The initial open-circuit voltage of this cell in the dark was ca. 100 mV. Upon illumination (100 mW cm-, xenon) of the n-CdSe, a current of 2.7 mA passed initially between the photoanode and the platinum counter-electrode in the second compartment. Concurrently, the cell voltage measured via the two auxiliary platinum electrodes, rose as shown in fig. 11. Here the cell was charged 1 1 I 1 I I I I I G. $ zooi 100 0 0 1 2 3 0 1 2 3 4 5 time/h FIG. 11.-Current and voltage of the system 1 rnol dm-3 Na2S + 1 rnol dm-j S + 1 mol dm-3 NaOHJNafion 31 5 11 rnol dm-3 Na2Se + 0.1 rnol dm-3 Se + 1 mol dm-3 NaOH during photocharge through 0.25 cm2 CdSe/l cm2 Pt electrodes and discharge through 1 cm2 platinized platinum electrodes with 100 i2 load.photoelectrochemically for ca. 3 h. Upon discharge with a 100 SZ load between the two auxiliary platinum electrodes a current of ca. 0.5 mA at 60 mV was obtained for ca. 5 h. The current-voltage characteristics of a sulphur/selenium redox cell are illustrated in fig. 12. Although good reversibility in manifest, this system possesses an inherently low voltage, which would be of little interest for practical devices. To increase the cell voltage of n-CdSe, sulphide/polysulphide based storage cells, Cd was evaluated as the electroactive species at the counter-electrode. The cell n-CdSell mol dm-3 NaOH + 1 mol dm-3 Na2S + 1 rnol dmW3 S + 0.1 rnol dm-3 Na,Se + 0.01 rnol dm-3 Se(Nafionl2 rnol dm-3 NaOHICd have an initial open-circuit potential of + 150 mV (CdSe electrode us.Cd) in the dark. Upon illumi- nation (100 mW cm-2 xenon light) the voltage between CdSe and Cd became -200 mV. By shorting the illuminated cell a charging photocurrent of 2.1 mA220 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE EIV FIG. 12.-Potentiodynamic (40 mV s-l) current-voltage characteristics of the system Ptll rnol dm-3 Na2S + 1 mol dm-3 S + 1 mol dm-3 NaOHINafion 31511 mol dm-3 Na2Se + 0.1 mol dm-3 Se + 1 mol dm-3 NaOHlPt at 27 "C. Unstirred electrolyte, platinized platinum electrodes; area of separator = 0.5 cm2. 0 0 % % ; ; ; ; ; 6 ( 1 time/h FIG. 13.-Current and voltage of the system 1 mol dmP3 NaOH + 1 rnol dm-3 NazS + 1 rnol dm-3 S + 0.1 mol dm-3 Na,Se + 0.01 rnol dm-3 SelNafion 31512 rnol dm-3 NaOHICd during photocharge through 0.25 cm2 CdSe[ 1 cm2 Cd electrodes and discharge through 1 cm2 platinized platinum elec- trodesll cm2 Cd with 100 Cl load.P .G . P . ANG AND A . F . SAMMELLS 22 1 I I I 1 I was recorded as shown in fig. 13. The cell was photocharged for 5 h and then left at open-circuit for 20 min. Afterwards the cell was discharged using a 100 R load between the platinum and cadmium electrodes. A current of ca. 2 mA at a cell voltage close to 175 mV flowed for 5.5 h until the capacity of the sulphide/polysulphide system was exhausted. In the dark a short-circuit discharge current of 0.15 mA was recorded between the CdSe and Cd. A polarization curve for a S2-[Si-ICd redox i I I I cell is shown in fig. 14. In- creasing the activity of the polysulphide species can be expected to result in larger cell voltages.Photoelectrochemical storage cells discussed here and their charge-discharge performance are summarized in table 1. The photopotential observed with n-MoSe, was found to be just sufficient to effect charge of the bromine-iodine redox cell, where the equilibrium cell voltage was between 0.4 and 0.5 V. However, from a kinetic Here an open-circuit voltage of 310 mV was observed. TABLE 1 .-PHOTOELECTROCHEMICAL CELLS WITH STORAGE iphot ochatgel idischargel cell system EImV" Eph/mVb mA cm-2 mA cm-, ~~ MoSe, IBr - IBr, 11 - 11, IPt 400-500 490 4 at 200 mW cm-2 20.6 GaAs [Se2- ISeZ- ICd 250-350 500 8 at 100 mW cm-, 14.6 CdSelS2- ISn'- ISe2- ISeZ- IPt 50-200 400 8 at 100 mW cm-2 0.75 CdSelS2- IS,'- ICd 300-400 400 10 at 100 mW cm-2 8.25 Voltage range of storage cell.Typical photopotential of the photoanode. Typical charge short-circuit photocurrent when the photoanode was illuminated at the listed light intensity. Short- circuit discharge current of the storage cell in the dark using platinized platinum electrodes (1 cmz) in unstirred electrolytes at 27 "C. standpoint this redox system is attractive since large short-circuit discharge current densities were obtained. The other storage systems have in general lower cell volt- ages ; however, polycrystalline GaAs and CdSe photoanodes can be readily fabricated. For n-GaAs the selenium-cadmium redox species only produced cell voltages between222 PHOTOELECTROCHEMICAL SYSTEMS WITH ENERGY STORAGE 250 and 350 mV, which is below the charging capacity achievable from the present state-of-the-art photoanodes.2p28 If one uses zinc instead of Cd (Zn redox potential at - 1500 mV us.SCE) a higher cell voltage could be realized. Photocharge of such a cell, however, would not be feasible using n-GaAs as the photoanode. Cells using polycrystalline CdSe do reasonably well for charging the sulphur-cadmium system where charging currents of 10 mA cm-2 could be achieved. The charge-discharge characteristics of our photoelectrochemical storage cells were obtained under un- stirred electrolyte condition. We expect that in a stirred or a flowing electrolyte system the current densities that can be achieved should be higher. In addition, higher voltage photoelectrochemical storage cells will require the introduction of suitable photocathode materials into the counter-electrode.This will increase the range of potential electroactive species that might be conceived for such devices. A. Heller and B. Miller, Abstract no. 81, ACS Division of Colloid and Surface Chemistry, 179th ACS National Meeting, Houston, Texas, March 1980. R. Noufi and D. Tench, J. Electrochem. SOC., 1980, 127, 188. H. Gerischer, J. Electroanalyt. Chem., 1977, 82, 133. A. J. Bard, J. Photochem., 1979, 10, 59. W. Schottky, 2. Phys., 1942, 18, 539. E. Spenke, Elektronische Halbleiter (Springer-Verlag, Berlin, Heidelberg, 2nd edn, 1965), p. 1 1 5. J. F. Dewald, Bell Systems Tech. J, 1960, 39, 615. M. A. Butler and D. S. Ginley, J. Electrochem. SOC., 1978, 125, 228.Illumination, ed. A. Heller (Electrochemical SOC., Princeton, N.J., 1977), p. 1. ' H. Gerischer, Proc. Con$ Electrochemistry and Physics of Semiconductor Liquid Interfaces under lo A. Heller and B. Miller, Electrochim. Acta, 1980, 25, 29. l1 R. Memming, Electrochim. Acta, 1980, 25, 77. l2 A. J. Nozik, Ann. Rev. Phys. Chem., 1978, 29, 189. l3 A. J. Bard and M. S. Wrighton, J. Electrochem. SOC., 1977, 124, 1706. l4 D. Laser and A. J. Bard, J. Electrochem. SOC., 1976, 123, 1828. l5 S. M. Frank and A. J. Bard, J. Amer. Chem. SOC., 1975, 97, 7427. l6 J. M. Bolts, A. B. Bocarsly, M. C. Palazzotto, E. G. Walton, N. S. Lewis and M. S. Wrighton, l7 H. Tributsch, 2, Naturforsch., 1977, 32A, 972. l8 H. Tributsch, Solar Energy Muter., 1979, 1, 257. l9 H. Tributsch and J. C. Bennet, J. Electroanalyt. Chew., 1977, 81, 97. 2o H. Tributsch, Ber. Bunsenges. phys. Chem., 1977, 81, 361. 21 J. Gobrecht, H. Tributsch and H. Gerischer, J. Electrochem. SOC., 1978, 125, 2085. 22 H. Tributsch, J. Electrochem. SOC., 1978, 125, 1086. 23 W. Kautek, H. Gerischer and H. Tributsch, Ber. Bunsenges. phys. Chem., 1979, 83, 1000. 24 H. Tributsch, H. Gerischer, C. Clemen and E. Bucher, Ber. Bunsenges. phys. Chem., 1979,83, 655. 25 J. Manassen, G. Hodes and D. Cahen. Semiconductor Liquid-Junction Solar Cells, ed. A. Heller (Electrochem. SOC., Princeton, N.J., 1977), vol. 77-3, p. 34. z6 J. Manassen, G. Hodes and D. Cahen, J. Electrochem. SOC., 1977, 124, 532. 27 G. Hodes, J. Manassen and D. Cahen, Nature, 1976, 261, 403. 29 H. J. Lewerenz, A. Heller and F. J. DiSalvo, Abstract no. 103, ACS Division of Colloid and 30 W. D. Johnston, H. J. Leamy, B. A. Parkinson, A. Heller and B. Miller, J. Electrochem. Soc., 31 H. J, Leamy, L. C. Kimerling and S. D. Ferris, Scanning Electron Microscopy, ed. 0. Johari 32 M. A. Russak, J. Reichman, H. Witzke, S. Deb and S. N. Chen, J. Electrochem. SOC., 1980, 33 R. Noufi, P. Kohl and A. J. Bard, J. Electrochem. SOC., 1978, 125, 375. 34 A. Heller, J. P. Schwartz, R. G. Vadinsky, S. Menezes and B. Miller, J. Electrochem. Sac., 1978, 35 A. Heller, K. C. Chang and B. Miller, J. Electrochem. SOC., 1977, 124, 697. J. Amer. Chem. Soc., 1979, 101, 1378. B. A. Parkinson, A. Heller and B. Miller, J. Electrochem. SOC., 1979, 126, 954. Surface Chemistry, 179th ACS National Meeting, Houston, Texas, March, 1980. 1980, 127, 90. (LIT Research Institute, Chicago, 1976), part IV, vol. 1 . 127, 725. 125, 1156.
ISSN:0301-7249
DOI:10.1039/DC9807000207
出版商:RSC
年代:1980
数据来源: RSC
|
16. |
Surface modification in semiconductor liquid-junction cells |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 223-232
Barry Miller,
Preview
|
PDF (686KB)
|
|
摘要:
Surface Modification in Semiconductor Liquid-junction Cells BY BARRY MILLER, ADAM HELLER, SHALINI MENEZES AND HANS J. LEWERENZ Bell Laboratories, Murray Hill, New Jersey 07974, U.S.A. Receiued 30th April, 1980 Modification of the photoactive junction in semiconductor based photoelectrochemical cells to improve or stabilize the current-voltage characteristics under illumination has been attempted in many ways in the brief history of such energy converting devices. We discuss two such efforts with n-type gallium arsenide substrates. In one, polycrystalline films incorporating submonolayers of metal ions in the surface region show significant enhancement of power output in selenide-polysele- nide cells. In the second, single crystals electroplated with noble metals and contacted by a redox solution to the resulting metal-semiconductor Schottky junction are examined for solar conversion prospects.From the recent extensive interest in semiconductor-electrode-solution interfaces as sites for light-modulated chemical reactions for energy conversion or electrosynthe- tic ends, it has become clear that the intricacies of this junction are critical to deter- mining both the products and energy efficiency of this process. In practice the diverse possibilities of surface modification have been, on the one hand, opportunities to direct the system efficiently into desired paths and, on the other, natural impediments to the initial output and/or long term stability of the active interface. We consider here two special cases representing apparent extremes : first, sub- monolayer surface composition change with chemisorbed metal ions on polycrystal- line substrates and second, the interposition of an electroplated metallic phase between single-crystal semiconductor and solution. In the first we seek to alter the kinetics of charge transfer under illumination for a semiconductor-redox couple pair to pro- mote the energy efficiency of the regenerative reaction sequence which also competes against photocorrosion of the substrate.In the second we effectively create a new photoactive junction away from chemical attack, using the redox system and counter- electrode as a contact. We recount here some recent work in our laboratories on these aspects of the broad subject of surface modification, especially with respect to solar energy conversion schemes.Earlier related reports have been g i ~ e n . ~ - ~ EXPERIMENTAL Two forms of n-GaAs substrates were employed, polycrystalline chemically vapor depo- sited (c.v.d.) specimens of n-GaAs/n+-GaAs/W/graphite s t r u ~ t u r e , ~ kindly supplied to us by Dr. S. S. Chu, and single crystals of 2.4 x 1015 C M - ~ carrier concentration with (111) or (100) faces exposed, The polycrystalline specimens have grain-size distribution of 1-20 pm and an average linear dimension of 9 pm. These c.v.d. film electrodes were mounted in epoxy and run in the n-GaAsl0.8 mol dm-3 K2Se + 0.1 mol dmP3 K2Se2 + 1 mol dm-3 KOHlC cell previously de~cribed.~.’ Metal ions Surface etching of c.v.d. films was carried out with 1% bromine in methanol.224 SEMICONDUCTOR LIQUID- JUNCTION CELLS were chemisorbed from solutions of 0.01 mol dm-3 RuC13 in 0.2 mol dm-3 HC1 and of 0.01 mol dm-3 PbO in 0.1 mol dm-3 KOH, in both cases following exposure of the etched substrate to the selenide-polyselenide electrolyte. All steps were separated by deionized water rinsing.The single crystals were given ohmic back contacts and then mounted as rotating disc electrodes.8 They were etched in 1 : 1 HzS04 + 30% Hz02 prior to electroplating of Au, Rh and Pt from standard baths.9 A front contact to metal-plated electrodes was made with an indium ring in the non- electrochemical experiments. Electrolytic solutions were prepared with reagent chemicals and triply distilled water. Light sources, solar-efficiency measurements and cell methodology are as earlier de~cribed.~*'~ RESULTS POLY CRY STALLI NE GaAs P HOTOA NOD ES Earlier reported results with substrates of n-GaAs prepared by metal-organic" or Ga-AsH,-HCl' c.v.d.showed solar conversion efficiencies of 4.8% l2 and 7.3%13 in the selenide-polyselenide cell when ruthenium was incorporated in their surfaces. The effect of ruthenium on both these types of electrodes with average crystallite sizes (10 pm is much more dramatic than with single amounting to a few hundred per cent improvement over prior etching alone with samples of poorer non-modified activity.I2 Enhancements of current paralleling the power output im- provements seen in photoelectrochemical cells were found after evaporating gold Schottky barriers on Ru modified substrates by charge collection scanning electron microscopy.12 With single metal ions adsorbed from 0.1 mol dm-3 acid solutions on single-crystal n-GaAs electrode^,^^ ruthenium had the most pronounced and permanent positive effect, but others such as Pb" temporarily increased efficiency in the selenide cell.We now find that when Pb" is adsorbed as plumbite ion from alkaline solution, it is both more effective and lasting in its n-GaAs electrode modification properties than in the acid adsorption case. The initial etching step that is applied to the c.v.d. electrode surface before the metal ion adsorption is also important, both as to its effect on final activity and com- TABLE 1 .-GaAs MODIFICATION AND NORMALIZED CELL POWER OUTPUT surface treatment u r n ~In)Il*rm. methanol + bromine etch for 30 s Pb" from alkaline solution; 45 s, 50 "C Ru"' from acidic solution; 45 s, 50 "C 0.55 0.89 0.93 1 .o Ru"' (acidic) then Pb" (alkaline); 45 s each, 50 "C patibility in etch rate with the use of these ca.25 pm layers. Bromine + methanol is more effective with the polycrystalline electrodes both in final efficiency achieved and avoiding pin-hole generation than the 1 : 1 H2S04 + 30% H202 employed on single crystal^.'^ Etch-solution variations are treated in a more complete experimental description; l5 we describe here the metal ion influence after the bromine + methanol step which allows the highest overall result. Voltammetric data under illumination, shown in fig. 1, and maximum power pro-B. MILLER, A. HELLER, S. MENEZES AND H. J .LEWERENZ 225 duct data calculated from such curves and collected in table 1, summarize results with separate ruthenium and plumbite, and multiple ion adsorption (Ru --+ Pb, ruthenium followed by lead) for otherwise comparable photochemical cell runs. The differences in efficiency, given normalized, are consistently reproduced. They indicate a rank order with respect to cell conversion efficiency of Ru -+ Pb :* Pb -+ Ru Ru > Pb > etching only. Pb --f Ru sequence data are not shown but are found to be essentially indistinguishable from those of the Ru step alone. The contrast to the transient effect of acid adsorbed Pb" on the ce11I4 emphasizes 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 cell voltage FIG. 1 .-Output characteristics under tungsten-halogen illumination of polycrystalline n-GaAslSe-* + Se2-3 + OH- IC cell after MeOH + Br, etching of photoelectrode, and then after ruthenium(rrr) and ruthenium(rrr) followed by lead(rr) treatment (Ru -+ Pb).that the chemical environment alters strongly the net effectiveness of individual metal ion treatment. The plumbite solution step produces a stable enhancement for at least several days. The earlier relative assessment of the effect on cell activity of 20 metal ions in common oxidation states adsorbed from 0.1 mol dm-3 acid l4 is valid only with respect to the particular solutions employed. Varying the conditions may enlarge the list of ions whose surface incorporation leads to a decrease in the inter- facial recombination of electrons and holes presumably responsible for the lower con- version efficiency of the " unmodified " GaAs surface in the selenide cell.Apparent from the ranking is the synergistic improvement obtained with multiple adsorption of metal ions. While it is clear that the effect of Ru is primary, the further improvement after plumbite treatment indicates that additional sites on the GaAs surface, for which the Ru species has low affinity and on which Pb(0H)j is adsorbed, can contribute to a further reduction of the recombination rate in the selenide cell. That the order of chemisorption with multiple ions is significant may have many226 S EM1 CON D UC TOR LIQUID- J UNCTION CELLS sources, including chemical and morphological surface heterogeneity, surface-charge distribution, ionic charge and equilibria, and adsorption kinetics.The retention of cell activity after the combined treatment has not been subjected to the testing reported for Ru a10ne.l~ In both these cases specifics of cell design and possible electrolyte doping with the metals could be significant. Similar trends amounting to ca. 5% increase in relative efficiency have been mea- sured for (100) faces of single-crystal n-GaAs with the Ru + Pb treatment over Ru alone, following the matte black surface conversion with sulphuric acid + hydrogen peroxide etchant that produced a 12% solar-energy conversion efficiency.' Ruther- ford back-scattering measurements have proven the presence of both Ru and Pb in the surface region after the multiple adsorption sequence with these crystals. The solar-energy conversion efficiency reached with the Ru --+ Pb sequence on methanol + bromine etched polycrystalline n-GaAs is shown in the cell curve of fig.0 0.1 0.2 0.3 0-4 0.5 0.6 0.7 cell voltage FIG. 2.-Response of fig. 1 cell with Ru+ Pb treatment under 53.7 mW cme2 solar input (Pi). Maximum output (Po) indicated. Power conversion efficiency = 100 Po/P, = 7.8%. 2, where a 7.8% value is attained. Comparison with the previously reported value of 7.3 % on similar substrates with the ruthenium alone is consistent with the relative laboratory data of fig. 1 and table 1. M ETA L- F I L M ED G a A s P HOTOANOD ES A limiting case for protection of the semiconductor from chemical or photoanodic reaction is the removal of the active junction from electrolyte contact. If a semi- transparent and unreactive metal film is interposed between the two p h a s e ~ , ~ ~ l ~ - ~ ~ the resulting metal-semiconductor junction must form a Schottky barrier as the new photoactive junction and the electrolytic redox couple with counter-electrode becomes the front ~ 0 n t a c t .l ~ Such cells are suitable models for light-flux and mass-transfer interactions that occur in all types of electrolyte junction system~.~ Here we move from an example of this approach to examine the properties of electrodeposited metal films on n-GaAsB . MILLER, A . HELLER, S . MENEZES A N D H . J . LEWERENZ 227 with electrolytic contacts as energy converters. In the latter we compare some of the cell junction characteristics with those with solid ohmic contacts to the Schottky bar- rier metal.By electrodeposition with available plating baths which, to varying degrees, are reactive to GaAs, low porosity films were difficult to achieve at the level of average thickness 200 8, semitransparent layers. However, the films obtained were sufficiently stable in photocell electrolytes not corrosive to the semiconductor substrate that elec- trodeposition was applied exclusively in these experiments. Plating procedures would be very suitable in a practical scheme to cover large substrate areas. Current-voltage behaviour of an illuminated gold plated n-GaAs and a pure gold disc electrode under mass-transfer limitation for Fe(CN);4 oxidation in both cases are shown in fig. 3. Conventional logarithmic analysis indicating 1 -electron slopes PI I -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 EDlV US.SCE FIG. 3.--ControlIed potential scans (ED US. SCE) of Au plated n-GaAs (laser illuminated) and pure Au (dark) disc electrodes in 2 mmol dm-3 Fe(CN)6, + pH 5 acetate buffer. Log analysis plots are superimposed, with IL = convective diffusion limited current at the 1600 r.p.m. rotation speed. for both plots is superimposed. The waves are essentially distinguishable at this light level only by the abscissa separation of 0.6 V, attributable to the photovoltage deve- loped at the Au-n-GaAs junction. A more complex arrangement of metal-plated semiconductor disc electrode and solution combining a mass-transfer limiting step followed by a light-limited irrever- sible solution oxidation process has the iD against ED behaviour under illumination shown in the inset to fig.4. The behaviour of the Pt-n-GaAs junction and 1 mmol dm-3 I - + 100 mmol dm-3 NO, solution simulates processes (solvent oxidation, photocorrosion) competitive with a desired one (I- oxidation) in a regenerative cell. The superposition of sinusoidal hydrodynamic modulation 2o on the rotation speed in this example gives the current component Aia responsive to the mass-transfer limited process which disappears as light-flux dominance (plateau above 0.4 V in iD trace) converts the cell to essentially galvanostatic control. In the region prior to the Aim peak, the I- oxidation is selective, as is proved by ring-disc electrode measurement~.~ In the main part of fig. 4 are plotted relative currents (iD at 0.4 V and the peak value of Aiw) as a function of normalized input power, PIPmax.The light-flux limited cur- rent, iD at 0.4 V, is linear with P/P max whereas peak Aim is constant above the light flux equivalent to the mass-transfer limited current density, and drops off sharply below it. The significance of these experiments, when applied to directly interacting semicon- ductor-electrolyte junctions, is that the presence of a region of Aiw response under228 SEMICONDUCTOR LIQUID-JUNCTION CELLS appropriate selection of input power, redox concentration and rotation speed4*9*21 is clear and necessary evidence of the selective reaction and thus successful competition of the desired redox system for light-generated minority carriers against photocorro- sion or solvent oxidation or reduction. The absence of Aim response means that the 0 f D ( V Y S SCE 10” 1 PIPmnx FIG.4.-Inset: Disc current, iD, and modulated disc current component, Aim, against scanned poten- tial, ED, at Pt plated n-GaAs disc in 1 mmol dm-3 NaI + 0.1 mol dm-3 NaNOz + pH 5 acetate buffer. Rotation speed = 1600 r.p.m. modulation amplitude 2 (r.p.m.)+ at frequency of 2 Hz. Main plot: Aim (peak) and iD at 0.4 V as a function of varying intensity, P, of laser beam. All values normalized to the response at unattenuated beam power, Pm,,. photoelectrode reaction is not capable of sustaining successful regenerative cell opera- tion. The comparative behaviour of metallically and electrolytically contacted metal filmed GaAs electrodes is seen in the normalized short-circuit current spectra in fig.5 of Au and Pt on n-GaAs. The GaAs band edge at ca. 870 nm for all electrodes, in- creased transmission for Au at ca. 500 nm, and the region of reduced response fromB . MILLER, A . HELLER, s. MENEZES AND H . J . LEWERENZ 229 h/nm FIG. 5-Au 1 and Pt 1 : Short-circuit photocurrent spectra of plated n-GaAs electrodes with indium contacts. Au 2 and Pt 2: same with 5 mmol dm-3 Fe(CN);4 solution. Relative scale arbitrarily normalized to compare curve shapes. (-) Au 1, (- - .) Pt 1 ; (- - -) Au 2 (a a ) Pt 2. solution absorbance below 600 nm of coloured Fe(CN)z4 electrolyte contacts may be noted. The unilluminated forward (or cathodic) current-voltage behaviour of both types of Au plated diode structures is basically similar, with slopes of 65-67 mV per decade current over several orders of magnitude.Voltammetric response under 632.8 nm He-Ne laser and solar illuminations is shown for Pt and Rh plated n-GaAs in fig. 6. The external energy conversion efficiency is shown on each trace. The 2.2- 5.3% solar values are typical of these specimens and depend at this point mostly on the quality and thickness of the electroplating. v c e I 1 FIG. 6.-Outputs of cells with photoanodes of n-GaAs plated with indicated metal with either solar irradiance (Pt, 93 and Rh, 90 mW cm-') or 632.8 nm laser source. Carbon counter-electrode and 0.25 mol dm-3 Fe(CN); + 0.01 mol dm-3 Fe(CN)Z in pH 5 acetate buffer. Maximum power conversion efficiency of each cell indicated on the curves.230 SEMI C 0 N D U CTOR LIQUID- J UNCTION CELLS Parallel experiments with gold-plated electrodes are shown in fig.7, along with a log plot of short-circuit current, I,,, against gold thickness, obtained under fixed laser illumination. Extrapolating the linear part of the log I,, against thickness plot to zero thickness gives an intercept corresponding to 60 10% quantum efficiency for the incident laser beam. The value at the thickness used in the gold-plated solar curve is ca. 32%. In the pH 5 acetate buffer the Au plated n-GaAs electrode is itself stable to pro- longed photo-oxidation of Fe(CN), '. However, the solution is rapidly degraded by thickness of Au plate/A 1000 6.3 Ofo He-Ne I I -4 " O 0 o 2ol - 50 - 2 0 7 -10 4 & - 5 z I3 --. * 4.1% \ I SUN I I I 0 0.2 0 .I4 0.6 I Vcell FIG. 7.-Solid lines identical to those of fig.6 but with Au plated electrode and solar irradiance of 69 mW cm-*. Open circles are short-circuit current, plotted logarithmically, as a function of Au plate thickness with laser illumination, same solution conditions. light and electrolyte replacement is required to restore the original current level at short-circuit to a carbon cathode. Other solutions such as the colourless Fe" + FeIII + pH 2 phosphate medium (which would essentially eliminate the short-wave- length deficit in the spectra of fig. 5 ) undermine the metal films by attack of the under- lying GaAs within pores upon illumination, and an output decline with time is ob- served. Thus activity is not restorable without etching and replating. Neutral solutions are required to make the GaAs surface sufficiently passive to prevent under- cutting and the subsequent loss of adherence and activity.SUMMARY AND CONCLUSIONS As is well illustrated here with the n-GaAsISe-2 + SeF2 + OH-IC cell, sub- monolayer metal modification of semiconductor-electrolyte interfaces can produce large variations of effective electrochemical reaction rate under photogenerated charge-transfer conditions. The influence of grain boundaries is of greater import- ance with semiconductors, where such factors as interaction of crystallite size with, bulk carrier diffusion length and light absorption depth, as well as enhanced inter- facial recombination rates at boundaries, will affect photocurrent yield. The competitive possibilities for photoelectrochemical cells over solid-state devices depend in part on better utilization, in principle, of polycrystalline substrates which retain the properties of spontaneous junction formation with the liquid and extraction of charge directly at the illuminated surface. The passivation of grain boundariesB .MILLER, A . HELLER, S . MENEZES AND H . J . LEWERENZ 23 1 meaning here reducing hole-electron recombination in these regions, by chemisorptive alteration of surface-state distribution, is achieved to an important degree in these n-GaAs films by a simple chemical process. The extension to multiple-ion adsorption, as in the Ru -+ Pb case, is evidence of chemical options deserving further exploration, particularly for III-V compounds. Many schemes involving selective removal and exchange of constituents can be visual- ized in these compounds to achieve modified surface-state characteristics.Adsorption of selenide species before metal treatment increases the rate of reaching stable enhance- ment on return of the specimen to the cell environment. Systematic prediction of use- ful single- or multiple-adsorption candidates and chemical schemes awaits further understanding. The use of semiconductors directly exposed to barrier-determining redox electro- lytes is dependent on overcoming kinetically, as above, the inherent interface stability p r o b l e m ~ . ~ ~ ' ~ ~ Hybrid cells of the metal-filmed semiconductor type represent a com- promise where the surface stability of a noble metal and independence of the redox potential of the solution are gained at the expense of light loss in the metal layer and additional complexity of fabrication. Compared with that in the solid-state version, the metal layer may, in principle, be much thinner since lateral conductivity to a col- lector is not required, the solution-counter-electrode representing a parallel path.Integrity of such thin layers and prevention of access to the semiconductor through pores constitute the preparative problem. The present work has extended earlier treatments of the hybrid J ~ n c t i o n ~ * ' ~ - ' ~ to fabrication of cells and measurements in sunlight with efficiencies of at least 5% achieved. Combination of these results with semiconductor thin-film preparation methods would be a possible sequential route to economic substrates of moderate efficiency.Plating of polycrystalline substrates at semi-transparent metal thicknesses is even more difficult for uniform coverage. Preliminary experiments using c.v.d. films show that open-circuit potentials, comparable with those observed on the single crystals, are attained, but degradation under load is indeed more rapid, as would be expected from additional porosity. Although we have described the two cases treated in this paper as extremes of sur- face modification, less-than-monolayer coverage of adsorbed ions to phase junction formation, there are actually situations of practical importance in photoelectrochemi- cal energy conversion which operationally cover this gamut. For example, n-CdSe electrodes in sulphide solution may undergo light-induced exchange reactions * which can have drastic effects on efficiency.The sequence of possibilities covers the whole range from partial substitution to eventual surface replacement and formation of a semiconductor-semiconductor junction, paralleling the submonolayer to buried- junction range considered here in discrete examples. Each stage in the CdSe case has a different overall effect on stability and efficiency, and the system can be modified, as by selenium addition to the to keep the surface in a more desirable state. From the perspective of light-modulated chemical conversion at electrodes, the experiments touched upon here sample the wide oppor- tunities in surface manipulation which remain to be taken. H. Gerischer, J. Electronalyt. Chem., 1975, 58, 236. A.Heller and B. Miller, in Interfacial Photoprocesses: Energy Conversion and Synthesis, ed. M. S. Wrighton, Ado. Chem. Series, no. 184 (Amer. Chem. SOC., 1980), p. 215-231. A. Heller and B. Miller, Electrochim. Acta, 1980, 25, 29. B. Miller and A. Heller, in Proc. Symp. Electrode Materials and Processes for Energy Conversion and Storage, ed. J. D. E. McIntyre, S. Srinivasan, and F. G . Will, PV 77-6 (The Electrochemical SOC., 1977), p. 91-104.232 S EM1 C 0 N D U C T 0 R LIQUID- J U N C T I ON CELLS S. S. Chu, T. L. Chu, and M. S. Lan, J. Appl. Phys., 1979, 50, 5805. K. C. Chang, A. Heller, B. Schwartz, S. Menezes and B. Miller, Science, 1977, 196, 1097. B. A. Parkinson, A. Heller and B. Miller, Appl. Phys. Letters, 1978, 33, 521. B. Miller, J . Electrochem. Soc., 1969, 116, 1677. B. Miller, S. Menezes and A. Heller, J. Electrochem. Soc., 1979, 126, 1483. lo A. Heller, K. C. Chang and B. Miller, J. Electrochem. Soc., 1977, 124, 697. H. M. Manasevit, J. Electrochem. Soc., 1971, 118, 647. l2 W. D. Johnston, Jr, H. J. Leamy, B. A. Parkinson, A. Heller and B. Miller, J. Electrochem. SOC., 1980, 127, 90. l3 A. Heller, B. Miller, S. S. Chu and Y. T. Lee, J. Anzer. Chem. Soc., 1979,101,7633. l4 B. A. Parkinson, A. Heller and B. Miller, J. Electrochem. SOC., 1979, 126, 954. A. Heller, H. J. Lewerenz and B. Miller, Ber. Bunsenges. phys. Chem., 1980, 84, 592. l6 Y. Nakoto, T. Ohnishi and H. Tsubomura, Chem. Letters, 1975, 883. I7 Y. Nakoto, K. Abe and H. Tsubomura, Ber. Bunsenges. phys. Chem., 1976, 80, 1002. I8 R. H. Wilson, L. A. Harris and M. E. Gerstner, J. Electrochem. SOC., 1977, 124, 1233. l9 L. A. Harris, M. E. Gerstner and R. H. Wilson, J. Electrochem. SOC., 1977, 124, 1511. 'O S. Bruckenstein and B. Miller, Accounts Chem. Res., 1977, 10, 54. B. Miller in Physiochemical Hydrodynamics, ed. D. B. Spalding (Advance Publications, London, 1977), p. 973-987. '* H. Gerischer, J. Vac. Sci. Technol., 1978, 15, 1422. 23 A. J. Bard and M. S. Wrighton, J. Electrochem. SOC., 1977, 124, 1706. 24 A. Heller, G. P. Schwartz, R. G. Vadimsky, S. Menezes and B. Miller, J. Electrochem. SOC., 1978,125, 1156.
ISSN:0301-7249
DOI:10.1039/DC9807000223
出版商:RSC
年代:1980
数据来源: RSC
|
17. |
Evaluation and reduction of efficiency losses at tungsten diselenide photoanodes |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 233-245
Bruce A. Parkinson,
Preview
|
PDF (1055KB)
|
|
摘要:
Evaluation and Reduction of Efficiency Losses at Tungsten Diselenide Photoanodes B Y BRUCE A. PARKINSON, THOMAS E. FURTAK, DUANE CANFIELD, KAM-KEUNG KAM AND GERALD KLINE Ames Laboratory, U.S. Department of Energy,t Iowa State University, Ames, Iowa 5001 1, U.S.A. Received 20th May, 1980 Single crystals of the layered semiconductor tungsten diselenide have been grown by chemical vapour transport and evaluated as the photoanode in a n-WSe21 1 mol dm-3 KI + 0.1 mol dm-3 IzlPt photoelectrochemical cell. In situ topographic photogenerated carrier collection analysis with a scan- ning laser spot technique has verified that crystal edges exposed to the electrolyte are a major source of efficiency losses in the light to electrical energy conversion. Surface states due to edge sites have been detected with sub-bandgap photocurrent spectroscopy and are located 0.2 eV below the conduction band edge.Several approaches have been demonstrated for the specific chemical treatment of the edge sites to reduce the effect of the surface states and increase the energy conversion efficiency. Iodine reduction (dark current) tlia these surface states is another factor in limiting cell efficiency. The original work by Tributsch and coworkersl-ll demonstrating that certain properties of the semiconducting transition metal dichalcogenides make these materials attractive for use in photoelectrochemical energy conversion schemes has also stimu- lated several other recent investigation^.^^-^^ The beneficial properties of these materials include the ability to produce both n and p conductivity types, bandgaps in the region of optimal solar conversion efficiency and stability against photocorro- sion reactions.Photoexcitation in these materials results in very little photo- corrosion due to the intrametallic d-d nature of the transitions which have only a small effect on the bonding in the semiconductor lattice and as a result of the multitude of electrochemical and chemical steps in the photocorrosion reactions in aqueous solu- tions. The two-dimensional layered structure of these compounds, which are composed of metal atoms sandwiched by chalcogenide layers, results in considerable anisotropy in the chemical and physical properties of the single crystals. For example, the electronic conduction is much greater along the layers than perpendicular to the layers.16 This fact has been useful in the interpretation of the different electro- chemical l7 and photoelectrochemical reactivity of edges as compared with the van der Waals surfaces in cells constructed from single crystal transition metal dichalco- genides.Photogenerated holes accumulate at crystal edges which are exposed to the electrolyte of a photoelectrochemical cell. The surface-state-rich edge supplies states for the recombination of carriers with a resultant loss in the efficiency of solar con- Surface states on a semiconductor photoelectrode operated near its maximum power output will also contribute to dark currents which oppose the photogenerated current and result in less power output.18 t Operated for the U.S.Department of Energy by Iowa State University under contract no. W-7405-Eng-82.234 EFFICIENCY LOSSES AT WSe, PHOTOANODES The layered compounds provide a very useful model for the chemical nature of semiconductor surface states.” The van der Waals surface, composed of close- packed selenium atoms, has no “ dangling bonds ” as do the surfaces of materials with three-dimensional order and thus could be expected to be rather free of surface states within the bandgap. A very high energy conversion efficiency has been re- ported for a tungsten diselenide photoelectrochemical cell made from a sufficiently small area of a crystal to ensure a defect-free surface.I4 The edges of the crystal would be expected to be rich in surface states, however, because of the termination of the lattice which results in coordinatively unsaturated transition metal sites.We report the preparation, characterization and behaviour as photoanodes of single crystals of n-type tungsten diselenide. We have verified that photogenerated carriers recombine at edge sites with the use of an in situ scanning laser spot analysis of the topographical carrier collection efficiency on a crystal surface. We have detected and established the position of the surface states responsible for the re- combination with the use of sub-bandgap photocurrent spectroscopy and established their importance in dark-current processes. We have also attempted with some suc- cess the minimization of losses due to the surface states by the use of several edge- specific chemical treatments of tungsten diselenide crystals.EXPERIMENTAL MATERIALS Tungsten powder (99.98%), selenium shot (99.999%) and tellurium tetrachloride were obtained from Alfa Research Chemicals. Natural molybdenite crystals from a South Korean mine were generously donated by Climax Molybdenum Company. The organic chemicals used for edge treatment studies were obtained from Aldrich and used as received. All other chemicals were reagent grade and used without further purification. Water was triply distilled with the second distillation from alkaline permanganate. APPARATUS The light source used for most experiments was a 10 mW polarized helium-neon laser from Hughes Laser Products. The laser power was measured with a Metrologic radiometer. Current-voltage curves were obtained with a PAR model 173 potentiostat with an external triangle wave generator.The crystals were illuminated through a Pyrex optical flat which was also the cell bottom. The scanning laser spot apparatus is described in detail in a recent publication.20 High resolution sub-bandgap photocurrent spectroscopy was performed on a high-intensity system based on a 1000 W tungsten-halogen lamp coupled to anfl3.6 monochromator. The grating dispersion provided a resolution of 6.6 nm mm-’ at the exit slit which was set at 1.5 mm. The monochromatic beam was mechanically chopped at 23 Hz to establish the photo- current dark-current subtraction. The light was focused on the sample in a two-electrode sample chamber with platinum as the inert electrode. A potentiostat was used to control the cell volatage and to convert the cell current to a voltage signal.The signal was fed to a pre- amplifier and then to the input of a lock-in amplifier which was referenced to the mechanical chopper. A microprocessor-controlled interface enabled the signal to be averaged for 3 s per data point while at the same time controlling the wavelength drive of the monochromator. This system allows us to record reliable, high-resolution spectra from 700 to 1500 nm. In this region, an Oriel filter G772-7000 was used to eliminate contributions from second- order radiation. The incident photon flux accounting for window loss and transmission through the solution was measured with a photoacoustic cell fitted with a lampblack target. The photoacoustic signal was fed into the same data handling electronics system. The resultPARKINSON, FURTAK, CANFIELD, KAM, KLINE 235 was a spectral representation of the incident power which was converted to a curve propor- tional to the spectral flux.The spectral flux could be stored in the memory unit of the micro- processor controller and could be recalled for normalization of the recorded photocurrent. CRYSTAL PREPARATION Single crystals of tungsten diselenide were prepared by chemical vapour transport 21 in sealed quartz ampoules with diameters ranging from 20 to 44 mm. The transport was carried out for ca. two weeks in a multizone furnace with a 50 "C temperature gradient and the hot zone at 950-1000" C. The transporting agent was chlorine gas at 0.5 atm and was introduced to the ampoule either directly or by the addition of TeC14 which decomposed to produce C1, at the transport temperatures.The starting material (WSe2) was prepared from the elements by reaction of stoichiometric amounts in an evacuated quartz ampoule at 950 "C for ca. one week which yielded a glittery micro-crystalline product. ELECTRODE PREPARATION The tungsten diselenide crystals were carefully removed from the transport tubes and placed on a glass plate and a mechanically sound back contact was prepared by suspending a copper wire just above the crystal and conductive silver paint was allowed to flow down the wire and complete the contact. The crystal was then mounted at the end of 6 mm glass tub- ing and all surfaces except the face were insulated with Torr Seal epoxy resin which was then cured at 80 "C for 24 h.Crystal surfaces were washed with methanol and triply distilled water before use in an electrochemical cell. For edge-treatment studies crystal faces with many step sites were deliberately selected to enable us to ascertain the effects of chemically treating the edges. RESULTS AND DISCUSSION CRYSTAL CHARACTERIZATION Tungsten diselenide samples obtained from synthetic ampoules and transported crystals were analysed for trace impurities by spark-source mass spectrometry. Table 1 shows a representative sample of the impurities in the WSe, and in a natural crystal of molybdenite (MoS,). Although the natural crystal had a well defined layered struc- ture and a shiny metallic lustre there were quite high levels of impurities present.The TABLE 1 .-MASS SPECTROSCOPIC ANALYSIS OF SAMPLES natural MoS2 WSe2 (untransported) WSe2(TeCI,) WSe2(C12) (all numbers in p.p.m.) C Si Te Ni Cr Fe A1 Ca Na c1 S Pb 0 2000 4000 <3 20 0.8 400 1800 1200 1000 0.8 160 large - 30 <5 <0.2 <2 3 1 <5 3 <0.3 20 <6 10 - 30 50 500 40 20 9 <1 t 5 t 2 3 <4 t l 10 100 10 4 30 20 30 <0.1 9 10 50 <3 30 -236 EFFICIENCY LOSSES AT WSe, PHOTOANODES fact that appreciable photoeffects have been observed on these22 and other natural molybdenite crystals 1*2*13 suggests that the impurities are segregated into islands. Uniform doping at such levels (> 10 2o ~ m - ~ ) would result in a very narrow space- charge layer and subsequent inefficiency of photogenerated charge-carrier separation. The large levels of impurities in the natural crystals discourage their use for definitive investigations of the intrinsic chemical and physical properties of the layered materials.Some interesting observations can be made by comparing the impurity levels in the tungsten diselenide samples listed in table 1. The untransported material has in general the lowest level of impurities while the transported crystals contain more car- bon, silicon and transition metals. A revealing entry in table 1 indicates that 500 parts per million of tellurium has been incorporated into the crystal lattice when tellurium tetrachloride was used as a source of transport gas. Because the effect of tellurium doping at these levels on the intrinsic properties of tungsten diselenide is unknown the direct addition of chlorine gas to the quartz ampoule was used in sub- sequent crystal growing attempts.Table 2 shows behaviour of a randomly selected sample of tungsten diselenide TABLE 2.-BEHAVIOUR OF TUNGSTEN DISELENIDE CRYSTALS AS PHOTOANODE. SEE TEXT FOR CELL CONDITIONS. sample no. o.c.v./V" s.s.c./pAb f.f." 4v/mWd Vl%e 35 33 34 26 20 38 36 22 24 37 0.36 0.155 0.005 0.1 3 0.23 0.24 0.17 0.36 0.038 0.16 17 16 1 18 56 27 25 88 9 18 0.55 0.45 0.27 0.33 0.37 0.34 0.31 0.39 0.24 0.36 0.13 0.12 0.13 0.10 0.51 0.12 0.13 0.44 0.041 0.16 2 1 1 1 2 1 1 0.2 0.6 -a a Open-circuit voltage; short-circuit current; fill factor; light intensity (632.8 mm); conversion efficiency. crystals as the photoanode in a photoelectrochemical cell with a solution of 1 mol dm-3 KI + 0.1 mol dm-3 I2 as the electrolyte and a platinum counter-electrode.There is considerable variation in the solar cell parameters of the cells made from these crystals, all of which were grown with C12 as the transporting agent. The faces of the crystals were all smooth and shiny to the eye; however, examination with an optical or electron microscope reveals many steps and dislocations on the surface. Crystal edges exposed to the electrolyte have been implicated as a major source of efficiency losses by Lewerenz et aZ.I2 and workers at the Fritz-Haber Institute.l* For the pur- poses of this study no attempt was made to cleave the crystals to obtain edge-site-free surfaces because we will attempt the chemical modification of these sites later in this study.PHOTOGENERATED CARRIER COLLECTION TOPOGRAPHY To verify that surface structure is a factor controlling the carrier collection on tungsten diselenide crystals, a device was constructed which scans a 25 pm laser spot across the crystal in situ.18 The photoresponse from the 632.8 nm laser light isB C FIG. 1 .-Scanning laser spot map of the carrier collection efficiency of a n-WSe, photoanode taken in situ (left) and photomicrograph of the same region of the crystal (right). The area labelled A is a relatively defect-free region of the crystal while area B is rich in exposed edges. C indicates the epoxy resin in which the crystal was mounted and provides a zero collection reference. A defect to the left of region A is also clearly mapped. The laser power was 9.6 pW of 632.8 nm light.PARKINSON, FURTAK, CANFIELD, KAM, KLINE 237 superimposed on a voltage proportional to the spot position on the sample to produce a " map " of the carrier collection efficiency like that shown in fig.1. The photo- response map can be compared with the photomicrograph and it is obvious that the carrier collection on the smooth surface is higher and more uniform than on the portion of the crystal with many exposed edges. A large defect in the crystal is also shown quite clearly in the scanning laser spot map. No photoresponse is obtained when the spot is off the crystal and this provides a reference point from which to compare the relative response of the different crystal areas. The interpretation of the scanning laser spot maps is not always as straightforward as is shown in fig.1. Some crystals which have areas apparently free of surface defects when observed with optical microscopy (40 x ) can still show considerable variation in carrier collection as the spot is scanned across the surface. These results could arise from non-linear optical effects from the high light intensities obtained (up to 500 W cm-2) when the laser beam is sharply focused. Subsurface defects which act as recombination centres or an inhomogeneous impurity distribu- tion which would alter the depth of the space-charge layer across the surface could also be responsible for non-uniform carrier collection on apparently defect-free sur- faces. Recent results of Gerischer 23 with the layered semiconductor gallium selenide have shown a large effect of the angle of incidence and polarization of the incident light on the photocurrent spectra of this material.Similar effects would be expected on WSe, if the crystal surface is not always perfectly normal to the polarized laser beam. SUB-BANDGAP SURFACE STATE SPECTROSCOPY High resolution sub-bandgap photocurrent spectroscopy was employed in an effort to observe the direct optical excitation of transitions involving the surface states within the bandgap.lg We expect that the surface states associated with the edge sites will exist close to the conduction band due to the 5 dxy and 5 d,t-,,r nature of both the unsaturated transition metal bonds and the semiconductor conduction band. At energies below the onset for interband transitions, photoexcitation creates a quasi-bound electron-hole pair which is immediately available at the surface, pro- vided surface states are present.Charge transfer to the redox couple in solution, which is in competition with the recombination process, can produce a current which can be identified with the direct population of the surface state. In fig. 2 the results of such an experiment are shown. Curve (b) is the photo- current spectrum near the short-circuit potential of a sample which was nearly free of surface steps and defects while curve (a) is the spectrum of a crystal which was mounted such that only the edge of the crystal is exposed to the electrolyte. Both curves have been corrected for lamp spectra and then normalized to the photogenerated carrier collection at a point above the onset of the direct interband transition to eliminate the effect of recombination which is much higher on the edge-mounted sample.The edge-exposed sample has a pronounced photocurrent from sub-bandgap radiation which is quite clearly shown by the difference spectrum [curve (c)]. The difference spectrum shows that unoccupied surface states are present in a broad band at the edge of the conduction band and extend ca. 0.2 eV into the gap. The lack of hysteresis in the current-voltage curves indicates that the surface state is in rapid communication with the conduction band. Fig. 3 shows band diagrams at short-circuit condition and at a potential near238 EFFICIENCY LOSSES AT W Se2 PHOTOANODES where a solar cell would have its maximum output and our interpretation of the surface-state position. The processes in which the surface state participates are a function of the band bending in the semiconductor.In part A of the figure there is considerable band bending at the semiconductor-electrolyte interface which allows tunnelling from the unfilled surface state to the conduction band, a process which will lower the voltage output of the cell. The tunnelling process also allows electrons to communicate with the surface and provides a pathway for recombination. Also 1.0 0.9 0.8 0.7 a^ 0.6 24 2 0.5 3 0.4 0.3 0.2 0.1 .- c1 .,a 1.20 1.10 I .oo 0.90 0.80 0.70 JlW FIG. 2.-Photocurrent spectra of n-WSe2 single crystals in a 1 mol dm-3 KL + 0.1 mol dm-3 Iz solu- tion mounted with the edge exposed (a) and van der Waals surface exposed (b).The spectra were normalized to eliminate the recombination losses on the edge-mounted sample and the difference spec- trum is (c). The arrow indicates the position of the direct band edge for WSel (1.34 eV). shown is the direct photoexcitation to this state as was discussed above. If the poten- tial of the semiconductor is close to the flatband situation (fig. 3B), as when the cell is operated for maximum solar to electrical energy conversion, the Fermi level is high enough to populate the surface state with electrons. These electrons are sufficiently reducing to react with tri-iodide ion which is likely to be adsorbed on the crystal edge site^.^^^^^ The dark reduction of iodine (tri-iodide) on crystal edges and simultaneous photo-oxidation of iodide on the crystal face is essentially a chemical route for re- combination.SELECTIVE SURFACE CHEMISTRY It is apparent from the scanning laser spot measurements and the sub-bandgap photocurrent spectroscopy that surface states attributable to edge sites are responsible for the reduction of energy conversion efficiency in layered semiconductors. The production of a practical solar converter from these materials would then require the production of large areas of defect-free surfaces. The high reflectivity of such aPARKINSON, FURTAK, CANFIELD, KAM, KLINE 239 perfect surface (up to 40%)24 would then present the further problem of finding an effective antireflection coating which does not degrade the performance of the cell. An alternative strategy would be to reduce the losses associated with the edge sites by a selective chemical treatment of the atoms responsible for the surface state.This strategy has been successful in reducing surface recombination 1 9 v 2 5 and grain boundary A B FIG. 3.-Band diagrams illustrating the position of the surface state (&) relative to the conduction band (&), valence band (Ev) and fermi level (EF) of the semiconductor. Diagram A shows the situa- tion near the short-circuit potential of the cell where photogenerated holes can recombine cia the sur- face state in competition with the redox couple oxidation process. The wavy line illustrates the direct photoexcitation of the surface state. A is the tunnelling distance through which the surface state is in communication with the conduction band.Diagram B is at a potential close to the maximum power point of the cell. In this case the Fermi level (EF) is high enough to allow electrons to populate the surface state and then react via redox couple reduction, a process which reverses the photo-oxidation by holes. recombination 26 in gallium arsenide photoelectrochemical cells. Strong interaction of ruthenium atoms with a gallium arsenide surface state resulted in the splitting of the surface-state energy and removing the state from the region of the conduction band edge.18 If the co-ordinately unsaturated tungsten atom at an edge site were reacted with a strong electron donor, the surface-state energy could be shifted out of the semicon- ductor bandgap. Several donating ligands which might be expected to be selectively reactive towards the edge site with various means of co-ordination are shown schemati- cally in fig.4. The current-voltage behaviour of photoelectrochemical cells and solid-state240 EFFICIENCY LOSSES AT WSe, PHOTOANODES photovoltaic devices in the solar cell region (from short circuit to open circuit) is very sensitive to the position and density of surface states located within the band- gap.18p27*28 Therefore the solar cell response provides us with a sensitive handle on the reactivity of various reagents towards the metal atoms which terminate the semi- conductor lattice. Fig. 5 shows the effect on the solar cell response and dark current of successive 6 A B C O = S e .=c Q = N @ =s O=H o= w FIG.4.-Schematic illustration of the direct binding of ligands to the crystal edge sites in WSe2: A, half of a diphos molecule bonded; B, bonding of phenyl isocyanide; C, chelating of dimethyl dithiocarbamate. treatments of a tungsten diselenide crystal with many exposed edges in a saturated solution of bis( 1,2-diphenylphospino)ethane (diphos) dissolved in alcohol at 60 "C. On each successive treatment the dark current is reduced and a corresponding increase in the open-circuit voltage and power output of the cell is observed. This result is consistent with the co-ordination of the diphos to the tungsten atom in a manner similar to that depicted in part A of fig. 4. Blocking this site would eliminate the inner phere pathway for iodine (tri-iodide) reduction which is responsible for the dark current which limits the efficiency of this cell.Unfortunately further treatment of this crystal with diphos does not further reduce the dark current. When the crystalPARKINSON, FURTAK, CANFIELD, KAM, KLINE 24 1 is illuminated over an extended period (12 h), the response reverts to a condition where the dark current and corresponding cell response is worse than the initial response of an untreated crystal and does not improve with further treatments with diphos. This behaviour can be attributed to an irreversible oxidation of the phosphine ligand by the highly oxidizing photogenerated holes. The oxidation products remain at the edge sites and create a higher density of surface states within the bandgap than were cell voltage/V FIG.5.-Effect on the solar cell response (above) and dark current (below) of an n-WSe, photoanode to successive one hour treatments with a saturated alcohol solution of diphos at 60 "C. The solid line is the initial response, the long dashes is the first treatment, short dashes the second and dots indi- cate after the third treatment. 12 h operation near the maximum power point (ca. 0.2 V) results in a response with higher dark current and irreversibly degraded photoresponse. present initially. This behaviour is much like that of some adsorbed metal ions on gallium arsenide surfaces which degraded the response of that Similar results were obtained when crystals were treated with isocyanide l i g a n d ~ , ~ ~ the bonding of which is schematically illustrated in fig.4B. The treatment with dithio- carbamates (fig. 4C) has little or no effect on the solar cell response of most crystals. Another approach to the introduction of donors specifically near edge sites is to intercalate a molecule which will not penetrate to the bulk. Tungsten diselenide is known not to form bulk intercalation compounds because the dzz orbitals on the tung- sten atom are filled. Partially filled dz2 orbitals in prolific intercalaters such as tanta- lum diswlphide are implicated as a d-band acceptor in a charge-transfer complex with a z-donor intercalate. The van der Waals surfaces near the edge of a tungsten diselenide crystal have acceptor character due to the unsaturated metal d orbitals at the edge and242 EFFICIENCY LOSSES AT WSe, PHOTOANODES 0 = w = c 0 = N 0 = H FIG.6.-Schematic illustration of the interaction of t-butyl pyridine with the edges of a WSez crystal. --2- ------- -----_ cell voltage/V FIG. 7.--Tlluminated (above) and dark (below) current-voltage curves in the solar cell region of a WSez photoanode in 1 mol dm-3 KI + 0.1 mol dm-3 I2 solution before treatment in neat t-butyl pyridine (solid lines), after a 1 h treatment at 80 "C (short dashes) and after an additional 1 h treatment (long dashes).PARKINSONy FURTAK, CANFIELD, KAM, KLINE 243 so one may be able to selectively intercalate only near edge sites. Fig. 6 shows a highly schematic interpretation of '' semi-intercalation " of tungsten diselenide by the molecule 4-t-butyl pyridine (TBP). The tertiary butyl group is an additional steric anchor to assure the molecule is held at the edge sites and will also help to make the edge sites hydrophobic and insulate them from the solution to reduce dark currents due to reduction of solution species. If a tungsten diselenide crystal with a stepped surface is treated in neat TBP at 80 "C increases in both the short-circuit current and open-circuit voltage are obtained (fig.7), resulting in a >35% increase in the power output in this case. In some cases the efficiency of very poor electrodes (< 1 %) can be improved to over 2%. Like the treatment with the phosphine ligands, further treatments result in further improvements and running the cell at the maximum power point results in a gradual loss of the improvement over many hours. Unlike the phosphine treatment the decay in the improvement brought about by TBP treatment is to the original electrode treatment running t i m e l h 1 2 1 2 3 4 I I 1 I I I 1 1 I I I i 30' treatment time/ h 1 2 1 2 3 4 5 6 I I I I l l I 1 FIG.8.-Time dependence of the power output (above) and dark current measured at 0.175 (below) of a n-WSe21mol dm-3 KI + 0.1 mol dme3 I,]Pt solar cell monitored during treatment (0, m) and while running (A, A ) with 0.24 mW of 632.8 nm laser light. Two different 2 h treatments and the subsequent decay are shown. The first point (left) is the initial response of the cell before treatments. response and retreatment will restore the improvement (fig. 8). The magnitude of the dark currents upon TBP treatment is very crystal dependent and may increase, decrease as in fig.8 or remain nearly the same as in fig. 7. We attribute the differences in the response of the tungsten diselenide photoanodes to treatments with strong ligands (phosphines, isocyanides) and TBP to a different mode of binding in the two cases: the ligands bind as shown in fig. 4 and the TBP semi-intercalates as depicted in fig. 6. The introduction of donor states by semi- intercalation is much like n-n+ doping in a solid-state solar cell where increases in the244 EFFICIENCY LOSSES AT WSe, PHOTOANODES short-circuit current and open-circuit voltage can be obtained.,* The utility of this doping scheme has also been demonstrated recently in photoelectrochemical cells.3o The introduction of donors at edge sites would help to counteract the field which drives the holes to the recombination prone edges.12 The n+ doping would also raise the Fermi level near the edges which would tend to increase the dark current; however, this may be counteracted by the bulkiness of the hydrocarbon group which may restrict the access of iodine to the edge sites. Treatment of a cleaved crystal with TBP which already has a reasonably good photo- conversion efficiency, such as the one shown in fig.9, results in virtually no improve- 1.2r cell voltage/V FIG. 9.-Current-voltage curve of a n-WSe21mol dm-3 KI + 0.1 mol dme3 IzlPt solar cell which shows a 5% monochromatic conversion efficiency. Virtually the same response is observed after treatment with t-butyl pyridine. The fill factor is ca. 0.4. 632.8 nm He-Ne laser; input power 5 mW.ment in the illuminated or dark current-voltage characteristics. This fact is con- sistent with the presumptions that good conversion efficiency results from defect-free surfaces and that the edge treatment chemistry is highly selective. The lack of im- provement also suggests that parameters other than recombination at edges may also play a role in limiting the response of photoelectrochemical cells made from layered compounds. It must be emphasized again that there is considerable variation of the response to the chemical edge treatments and to semi-intercalation from crystal to crystal and that the results presented here are illustrative of the general trends we have observed when a variety of reagents were used to treat many different crystal samples. We are continuing our efforts to discover a reagent or treatment which will result in morePARKINSON, FURTAK, CANFIELD, KAM, KLINE 245 dramatic and prolonged efficiency improvements and thus open the way for poly- crystalline devices made from the layered semiconductors.CONCLUSIONS The large anisotropy of the layered material has allowed us to identify the chemical species identified with a surface state at the semiconductor-electrolyte interface. By topographic analysis of the carrier collection we have verified that recombination occurs via the surface states associated with crystal edges exposed to the electrolyte. The energy levels of the surface state were found to extend ca. 0.2 eV below the con- duction band edge. We have succeeded in doing reversible and irreversible chemistry on the chemical site responsible for the surface state and have demonstrated that temporary improvements in the light to electrical conversion efficiency can be achieved by some chemical treatments.We propose that such a well defined surface state provides a useful model for surface states which exist at the semiconductor-electrolyte interface for most crystallographic orientations of semiconducting materials with three-dimensional order. Helpful discussions with H. J. Lewerenz, Prof. A. J. Bard and H. S. White are gratefully acknowledged. The spark-source mass spectroscopy and microscopy services of the Ames Laboratory were very helpful. This research was supported by the Solar Energy Research Institute under contract no.XP-9-8 198-1. H. Tributsch and J. C. Bennett, J. Electroanalyt. Chem., 1977, 81, 97. H. Tributsch, 2. Naturforsch, 1977, 32A, 972. H. Tributsch, Ber. Bunsenges. Phys. Chem., 1977, 81, 361. * H. Tributsch, Ber. Bunsenges. Phys. Chem., 1978, 82, 169. ’ H. Tributsch, J. Electrochem. SOC., 1978, 125, 1086. J. Gobrecht, H. Gerischer and H. Tributsch, Ber. Bunsengesphys. Chem., 1978, 82, 1331. J. Gobrecht, H. Gerischer and H. Tributsch, J. Electrochem. SOC., 1978, 125, 2085. H. Tributsch, Solar Energy Materials, 1979, 1, 705. H. Tributsch, H. Gerischer, C. Clemen, E. Bucher, Ber. Bunsengesphys. Chem., 1979, 83,655. lo W. Kautek, H. Gerischer and H. Tributsch, Ber. Bunsengesphys. Chem., 1979,83, 1OOO. T. Kawai, H. Tributsch and T. Sakata, Chem. Phys. Letters, 1980, 69, 336. l2 H. J. Lewerenz, A. Heller and F. J. DiSalvo, J. Amer. Chem. SOC., 1980, 102, 1877. l3 L. F. Schneemeyer and M. S . Wrighton, J. Amer. Chem. SOC., 1979,101,6496. l4 F. R. F. Fan, H. S. White, B. Wheeler and A. J. Bard, J. Electrochem. SOC., 1980, 127, 518. l6 R. Fivaz and P. E. Schmid, Physics and Chemistry of Materials with Layered Structures, ed. l8 B. A. Parkinson, A. Heller and B. Miller, J. Electrochem. SOC., 1979, 126, 955. l9 F. Garcia-Moliner and F. Flores, J. Phys. Chem., 1976, 9, 1609. 2o T. E. Furtak, D. Canfield and B. Parkinson, Appl. Phys. Letters, submitted. 21 H. Schafer, Chemical Transport Reactions (Academic Press, N.Y., 1964). 22 B. Parkinson, unpublished results. 23 H. Gerischer, paper presented at the Houston meeting of the American Chemical Society, March 24 P. A. Lee (ed.), Physics and Chemistry ofMaterials with Layered Structures, Vol. 4, Electrical 25 R. J. Nelson, J. S. Williams, H. J. Leamy, B. Miller, H. C. Casey, Jr., B. A. Parkinson and 26 W. D. Johnston, Jr., H. J. Leamy, B. A. Parkinson, A. Heller and B. Miller, J. Electrochem. SOC., S. Menezes, F. J. DiSalvo and B. Miller, J. Electrochem. SOC., in press. P. A. Lee (D. Reidel, Dordrecht, Holland, 1976, vol. 4. S. M. Ahmed and H. Gerischer, Electrochim. Acta, 1979, 24, 705. 1980. and Optical Properties (D. Reidel, Dordrecht, Holland, 1976). A. Heller, Appl. Phys. Letters, 1980, 36, 76. 1980, 127, 90. P. J. Boddy, J. Electroanalyt. Chem., 1965, 10, 199. H. J. Hovel, Semiconductors and Semimetals, vol. 11, Solar Cells (Academic Press, N.Y. 1973). 29 B. Parkinson and D. Canfield, in preparation. 30 R. No& and D. Tench, J. Electrochem. SOC., 1980, 127, 188.
ISSN:0301-7249
DOI:10.1039/DC9807000233
出版商:RSC
年代:1980
数据来源: RSC
|
18. |
Analysis of the mass and charge transfer limits of copper phthalocyanine photoelectrochemistry |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 247-254
William M. Ayers,
Preview
|
PDF (388KB)
|
|
摘要:
Analysis of the Mass and Charge Transfer Limits of Copper Phthalocyanine Photoelectrochemistry BY WILLIAM M. AYERS University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A. Department of Chemical and Biochemical Engineering, Received 19th May, 1980 Copper phthalocyanine, sublimed on a platinum rotating-disc electrode, is examined as a semi- conducting photoelectrode. The carrier photogeneration and mass-transfer limits to the photo- current, flat-band potential and cell power are determined in Fe(CN6) with KCl supporting electrolyte. The open circuit photovoltage dependence on hydrogen ion and disolved oxygen concentration are also examined. The catalytic and photoconductive properties of the metal phthalocyanines have attracted the attention of investigators for many ~ e a r s .l - ~ Most investigations of the photoconductive and photovoltaic properties of these materials have been carried out in gaseous environments or in vacuum. More recently, electrochemical studies with copper phthalocyanine (CuPc) sublimed on gold,4 platinum’ and SnOZ6 supports have been reported. Meshitsuka and Tamaru’ also examined the action spectrum and light intensity dependence of the photoassisted oxidation of oxalate. Although these authors considered the sublimed CuPc films to be semiconductor electrodes, the properties associated with semiconductor electrochemistry (flat-band potential, interfacial energy levels and current-voltage rectification) have not yet been reported. Thus in these investigations the interfacial capacitance as a function of bias potential, dark and photocurrent-voltage behaviour, as well as the carrier photo- generation and mass-transfer limits to the photocurrent are determined. In addition, the dependence of the open-circuit photovoltage on solution pH and dissolved oxygen content are examined.Copper phthalocyanine sublimes in the or-crystal form.’ A band model for con- ductivity is supported by the observations that (1) the conductivity increases with temperature,s (2) the activation energy for extrinsic conduction is dependent on the oxygen content of the crystal,* (3) the Hall-effect voltage changes sign near the temperature at which intrinsic conduction predominatess and (4) the optical absorp- tion and photoconductive action spectra show close corresponden~e.~J~ It has been suggested that the large carrier mobilities of CuPc ( p - = 88 cm2 V-l s-l, p+ = 131 cm2 V-’ s-l at 300 K) are due to the copper atoms acting as a point of strong inter- action between nitrogen electrons of neighbouring molecules.s These mobilities are approximately a thousand times greater than those for metal-free phthalocyanine.EXPERIMENTAL The copper phthalocyanine was purchased from Eastman Organic Chemicals. Two sublimations in a vertical tube furnace brought the absorption coefficients up to those of published values.11 The final CuPc crystals were needle-shaped and ca. 0.1 x 0.1 x 0.1248 COPPER PHTHALOCYANINE PHOTOELECTROCHEMISTRY mm3. The purified CuPc was sublimed onto the platinum disc of the rotating-disc electrode in a Kinney evaporator at 1.9 x Nm-2.Approximately 25 min at the sublimation temperature (ca. 820 K) were required to produce a suitable coating on the electrode. The thickness of the CuPc coating was ca. lo5 A as measured by the optical absorption of a CuPc deposited on a glass slide in the same manner. The Plexiglass electrochemical cell and associated equipment for measuring the CuPc photoelectrode current, voltage and interfacial capacitance are shown in fig. 1. The rotating- calomel ,/ reference electrode redox electrode l a m p r o t a t i n g disc surface w i t h CuPc f i l m FIG. 1 .-Schematic diagram of electrochemical cell with rotating-disc electrode. The 0.46 cm2 disc surface coated with copper phthalocyanine is shown in the insert. disc electrode surface with CuPc coated platinum disc is shown in the insert.The electrode (Pine Instrument Co., DT6) was attached to a constant-speed motor and the rotation speed checked with a frequency counter. A Sargent-Welsh polarograph provided the bias potential for the disc circuit. A con- stant bias or a bias sweep of approximately 2 mV s-l was sent through the circuit with the polarograph. The potential of the CuPc relative to the saturated calomel reference electrode was measured with a Corning 110 digital pH meter in the millivolt mode. The current through the circuit was measured with another Corning 110 meter R input) detecting the voltage drop across a lo4 R resistor in series with the cell. The counter-electrode was a platinum wire coil. The interfacial electrode capacitance was measured with a General Radio 1608A bridge operating at 250, 500 or 1000 Hz.The amplitude of the bridge signal in the cell circuit was 100 mV. The intensity, 76 mW cm-2, uncorrected for solution absorption, was measured with a Spex 1448 power meter. The illumination intensity was varied with Oriel neutral density filters. The dissolved oxygen content of the electrolyte solution was measured with a Y.S.I. oxygen probe. The oxygen content was varied by bubbling nitrogen through the solution. Illumination of the CuPc was with a tungsten microscope lamp. RESULTS AND DISCUSSION The effects of hydrogen ion and dissolved oxygen concentration on the open-circuit photovoltage are shown in fig. 2 and 3. The photovoltage decreases approximatelyW. M. AYERS 300 L4 4 > E 5 3 8 0 - x .0 M 4 > 0 c a U 249 I 1 - 3.9 4.0 1.1 4.2 4.3 ril ' 400 2 > E 2 300 9 . b) U c1 + -2 200 , I I , : I 1 I l 1 2 3 4 5 6 7 8 9 PH FIG. 2.-Open-circuit photovoltage dependence on pH. The hydrogen ion content of the 0.1 mol dm-j KCI was adjusted with dilute HCl. 4 - 3 !.k N 3 2 . n W \ 0% 0 1 h I 1 1 I 1 100 200 300 400 500 600 700 800 potential/mV us. SCE 0, 500; A, lo00 Hz. FIG. 4.-Mott-Schottky plot of CuPc electrode in 0.1 mol dm-j KCl. Bridge frequency: 0, 250;250 COPPER PHTHALOCYANINE PHOTOELECTROCHEMISTRY 100 200 300 400 500 600 700 800 900 potential/mV us. SCE FIG. 5.-Mott-Schottky plot of CuPc electrode in 0.1 mol dm-3 KCI + 0.001 mol dm-3 Fe(CNs)-3. Bridge frequency: 0,250; 0, 500 Hz. by 16 mV per pH unit. This is about one quarter of the 60 mV Nernstian change observed with materials (such as T i 0 3 that react with hydroxyl ions.12 The open- circuit photovoltage increases by ca.0.35 mV for each percent decrease in dissolved oxygen content. The dissolved oxygen was removed from the solution at a rate of ca. 2 % min-l. Thus, the change in photovoltage is most likely caused by a decrease in oxygen adsorbed on the CuPc surface. The interfacial capacitance of the CuPc electrode as a function of the electrode x I -3.0 -5.0 - -5.02Fe(CN), - 3/-4 -5.54 FIG. 6.-Energy-level diagram of CuPc electrode in 0.1 mol dm-3 KCI + 0.001 mol dm-3 Fe(CNs).W. M. AYERS 25 1 -4 -5 -6- potential under bias is shown in fig. 4 and 5. The data are plotted for the Mott- Schottky calculation of the flat-band potential.The flat-band potentials in either the 0.1 mol dm-3 KCl or KCI plus 0.001 mol dm-3 Fe(CN)6-3 are ca. 0.80 V us. the calo- mel reference electrode. The slope of the interfacial capacitance plots suggests that the CuPc is a p-type material. An energy-level diagram based on the flat-band potential and the properties of CuPc in uacuo, is shown in fig. 6. The bottom of the conduction band (electron affi- nity, x ) is taken as -3.0 eV.13914 The top of the valence band is at -5.0 eV and the oxygen acceptor level at -4.6 eV. Both of these values are based on bandgap mea- surements of Heilmeier and Harrison.’ Also shown in fig. 6 is the equivalent energy level for the Fe(CN)6’3’-4 redox electrolyte in KCl. The dark and photocurrent-voltage behaviours of the CuPc electrode in the KCl - 0 A - 0 plus Fe(CN6) electrolyte are shown in fig.7. There 0 A is little difference between the potential/mV us. SCE FIG. 7.-Current-voltage behaviour in 0.1 mol dm-3 KCI + 0.001 mol dm-3Fe(CN6): 0, dark cur- rent; A, photocurrent. Electrode rotation rate, 100 r.p.m. Electrode surface area, 0.46 cmz. The equilibrium redox potential at a platinum electrode is E‘. dark and photocurrents. This photoenhancement of the reduction rate would be expected from a p-type depletion region at potentials less than the flat-band potential. The rectification, however, indicates that the cathodic current is more bias-dependent than the anodic current, as would be expected from an n-type material with redox energy levels overlapping the conduction band. The substantial involvement of both electrons and holes in the dark conductivity of CuPc and the span of Fe(CN6) energy levels across the small bandgap of CuPc makes interpretation of the current-voltage behaviour in terms of n-or p-type behaviour difficult.The photocurrent and cell voltage under several external resistance loads are The greatest difference occurs at cathodic potentials.252 COPPER PHTHALOCYANINE PHOTOELECTROCHEMISTRY \ o\o 0 \ 1 10 20 30 40 cell photovoltagelmV bias, 0 r.p.m. FIG. 8.-Photoelectrochemical cell power in 0.1 mol dm-3 KC1 + 0.001 mol dm-3 Fe(CNb). No shown in fig. 8. The maximum cell power of 1.5 x W cm2 occurs near 25 mV. The open circuit photovoltage is approximately five times that observed with unpuri- fied CuPc in vacuo.lo The possible limiting conditions for the photocurrent are the rate of arrival of the redox electrolyte at the electrode surface and the rate of photogeneration of excess carriers within the CuPc.For an interfacial reaction rate that is first order in redox concentration, the mass-transfer-limited current is proportional to the square root of the angular velocity of the rotating-disc e1ectr0de.l~ A plot of the photocurrent against the square root of the angular velocity at several electrode bias potentials is shown in fig. 9. Under these conditions, redox mass-transfer limits the reaction rate only at potentials >0.3 V us. the calomel electrode, or currents >ca. 6 pA cmV2. 0 0 0-0- 0- I I 1 2 3 4 5 6 7 8 (angular velocity)+/rad+ s+ G. 9.-Photocurrent dependence on electrode angular velocity in 0.1 mol dm-3 KCl + 0.001 mol dm-3 Fe(CN6).Electrode bias us. SCE: 0, 204; 0,287; A, 358 mV.W. M. AYERS 253 3 8 CI 0 “ a 6 4 2 10 20 30 40 50 60 70 80 90 100 relative light intensity PA) FIG. 10.-Photocurrent dependence on white light intensity (0) and square root of light intensity(C1) in 0.1 mol dm-3 KCI + 0.001 mol dm-3 Fe(CNa). No bias, 0 r.p.m. The photocurrent dependence on white light intensity is shown in fig. 10. The linear dependence of the photocurrent on the square root of the light intensity is usually interpreted as evidence of carrier depletion by a second-order process. That is, if the steady-state continuity balance for photogenerated holes (or electrons) is approximated as 0 = a I,, e-ar - k [&)I2 (1) then the surface hole concentrations, p(O), and photocurrent, iph, will be proportional to the square root of the intensity, lo.Here a is the absorption coefficient and k is a constant for the depletion process. Since eqn (1) assumes that diffusional and migrational flux of carriers within the CuPc is negligible, and the boundary concentration is constant, the apparent applica- bility of eqn (2) to fig. 10 suggests that the photocurrent of the unbiased CuPc electrode is limited by the internal carrier generation and depletion processes. The assistance of A. R. McGhie and A. Varma in the purification and characteriza- tion of the copper phthalocyanine and R. M. Hochstrasser and S. E. Harrison for discussions on photoprocesses in molecular crystals is gratefully acknowledged. F. Gutmann and L. Lyons, Organic Semiconductors (John Wiley, New York, 1967). H. Meier, Organic Semiconductors (Verlag Chemie, Weinheim, 1974). M. Calvin, Accounts Chem. Res., 1978, 11, 369. H. Tachikawa and L. R. Faulkner, J . Amer. Chem. Soc., 1978, 100,4379. S. Meshitsuka and K. Tamaru, J.C.S. Faraday 11, 1977, 73, 236. V. R. Shepard and N. R. Armstrong, J . Phys. Chem., 1979,83, 1268.254 COPPER PHTHALOCYANINE PHOTOELECTROCHEMISTRY ’ S. E. Harrison and K. H. Ludewig, J . Chem. Phys., 1966,45,343. * G. H. Heilmeier and S. E. Harrison, Phys. Rev., 1963, 132, 2010. lo S. C. Dahlberg and M. E. Musser, J. Chem. Phys., 1979, 70, 5021. l1 G. E. Ficken and R. P. Linstead, J. Chem. SOC., 1952,4846. l2 M. A. Butler and D. S. Ginley, J. Electrochem. SOC., 1978, 125, 228. l3 M. Pope, J. Chem. Phys., 1962,36,2810. l4 W. Pong and J. A. Smith, J. Appl. Phys., 1973,44, 174. l5 J. S. Newman, ElectrochemicaZ Systems (Prentice-Hall, Englewood Cliffs, N. J., 1973). S. E. Harrison, J . Chem. Phys., 1969, 50,4739.
ISSN:0301-7249
DOI:10.1039/DC9807000247
出版商:RSC
年代:1980
数据来源: RSC
|
19. |
General discussion |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 255-283
H. Gerischer,
Preview
|
PDF (2437KB)
|
|
摘要:
GENERAL DISCUSSION Prof. H. Gerischer (Berlin) said: I would like to explain three aspects of the sub- mitted paper in somewhat more detail. First, the thermodynamic stability criteria recalled in section I. 4 can be modified by including single steps in the sequence of the decomposition reaction in this picture. This has recently been described at another occasion (IUPAC Conference on Photochemistry at Seefeld, Austria, July 14-1 8, 1980, in course of publication within the conference volume). One can attribute an equilibrium Fermi level to all reaction steps in which electrons or holes are involved as reactants. Since the decom- position potential represents an average over all single-reaction steps, there will be some reactions having a larger, and others having a smaller value of the equilibrium Fermi levels.The most important step for the decomposition process will be that one having the most energetic position for the electron or hole. This step will mainly control the rate of decomposition. A number of different situations can result which are summarized in fig. 1 . This figure contains the decomposition Fermi levels for the overall reaction leading to Two such steps are shown in fig. 1 of my paper. E S + e S+h E S+e S + h 1’- FIG. 1 .-Energy diagram for anodic and cathodic decomposition reactions h+ e- s __f I -+ PI -f Pz; s __+ 1 ’ 3 P; -f Pi: (a) thermodynamically stable, (6) thermodynamically unstable, kinetically stable, (c) totally unstable. the final product P and the Fermi level for the step having the highest free energy. To shorten the argument, the equivalent situation for the cathodic and the anodic process is shown in the same part of the picture.From this model one can dis- tinguish between three different situations instead of two. Real thermodynamic stability is shown in part (a). Here the position of the Fermi level for the overall re- actions and consequently also for the most difficult steps are located within the res- pective energy bands. Part (b) shows a situation which one can call kinetic stability because the Fermi level of the most difficult reaction step is located within the respec- tive energy band. Part (c) shows a situation where all reaction steps have Fermi levels within the band gap, and the semiconductor is therefore totally unstable. This more detailed treatment of the decomposition process also modifies the dis- cussion of the stabilization by competing redox reactions.The competition really occurs between the redox system in solutions and the slowest electrochemical step in the decomposition reaction. Fig. 2 outlines the different possibilities which may be256 fE GENERAL DISCUSSION t’ S + h “i FIG. 2.-Stabilisation of semiconductors by competing redox reactions: (1’) S + e + Ox’JS + Red’(fast) (3’) I + Ox’ -+ S + Red’(fast) left-hand side: right-hand side (1) S 3- h + R e d 3 S + Ox (fast) (3) I + Red + S + Ox (fast) (2) S + h + I (slow) (2’) S + e 3 I’ (slow) (4) I + h 3 P (slow) (4’) I + e 3 P (slow) anodic cat hod ic found in real systems. This concept demonstrates that even in a thermodynamically very unfavourable situation a fast electron-transfer reaction with components of the solution has a chance to reduce the decomposition rate to a very low level if the de- composition process contains one step which needs a much higher free energy of the electrons or af the holes than the thermodynamic average indicates for the net reac- tion.Secondly, in the kinetics of the anodic decomposition reaction as given in eqn (19)-(23) of my paper, the potential drop in the Helmholtz double layer is one impor- tant magnitude for the reaction rate. If AyH varies, the position of Ev varies too and the decomposition reaction rate is strongly affected. The same is true for the cathodic decomposition reaction, the equations of which are not given here explicitly.It is shown in section 11. 4, as in the papers of Prof. Bard and others, that in the presence of surface states the potential drop in the Helmholtz double layer can be changed con- siderably if the system is illuminated. This can have serious consequences for the decomposition reactions. The stability criteria of eqn (1 8) can in combination with eqn (14) and (15) be expressed as critical concentrations of electrons or holes which have to be surpassed in order to get a driving force for decomposition. This means for the cathodic process that decomposition is possible if and for the anodic process that decomposition is possible if ( -pEdecomp- “) kT ps* > Nv x exp In these equations the energy positions of the conduction band or the valence band appear in the exponential terms.Any change in the position of the band edges at the interface therefore drastically alters the critical concentrations of the electronic car- riers at the surface. The conclusions for the effect on stability are listed in table 1. Usually, the shift of the band-edge energy positions at the interface between a semi- conductor and an electrolyte caused by illumination will go into an unfavourable direction. There might, however, in some cases be a chance to get an improvement by this effect.GENERAL DISCUSSION 257 TABLE CO CONCLUSIONS FOR THE EFFECT ON STABILITY OF MOVING THE BAND EDGES shift of band edges anodic decornp. cathodic decornp. upwards downwards retarded enhanced enhanced retarded (Communicated): The use of the Fermi level in redox electrolytes has been criticized by Prof. Bockris in his rCsum6.First of all, the Fermi level is the equivalent of the electrochemical potential of electrons. There can be no doubt that the use of this term is justified in this thermodynamic sense. The objection of Prof. Bockris obvi- ously was based on the assumption that Fermi statistics could not be applied to such systems. It was, however, shown 20 years ago that the distribution of electrons over the electronic states of a redox system must necessarily follow Fermi statistics if the electronic energy levels of the occupied (reduced) and unoccupied (oxidized) species are considered as one ensemble. Any distribution function of the energy levels for the reduced and oxidized species in a redox system must give a combination which in the time average results in a Fermi distribution for the electrons.It can easily be shown that the use of Gaussian functions for the energy distribution of the indi- vidual redox energy levels fulfils this condition. The Fermi-level description for the equilibrium situation in redox electrolytes is therefore fully justified. 2. phys. Chem. (Frankfurt), 1960,26, 223. Dr. A. J. Nozik (Colorado) said: I would like to ask Prof. Gerischer if he would consider the question of how to describe the existence of quasi-Fermi levels in the electrolyte. In his paper, the quasi-Fermi levels are shown in the semiconductor, but they stop at the interface. It seems that if one chooses to use quasi-Fermi levels in the solid, then they must extend continuously into the liquid and merge to a common Fermi level at the metal electrode, as well as in the semiconductor bulk.Prof. H Gerischer (Berlin) said: The concept of quasi-Fermi levels is also applic- able in redox electrolytes, if two redox systems are present therein which are not in equilibrium. They correspond for instance to the two different redox potentials which one has in photogalvanic cells under illumination. The Fermi levels of electrolytes having no redox system added correspond to the redox potentials of the solvent itself. For example, water has two Fermi levels, one for its oxidation and another one for its reduction reaction. Prof. H. 0. Finklea (Virginia) said : I would like to raise a point concerning Prof. Gerischer’s method of determining thermodynamic decomposition levels for semi- conductors in electrolytes. For semiconductors with homogeneous composition from the bulk to the surface, such as the metal oxides, the method is certainly valid, but for non-metal oxides, such as GaP or GaAs, there exists a surface layer of oxide.If the decomposition reactions occur via the surface oxide layer, then perhaps a more valid method would be to use chemical equations involving oxides rather than the bulk semiconductor composition for determining the thermodynamic decomposition poten- tials. Would Prof. Gerischer please comment on this point ? Prof. H. Gerischer (Berlin) said: The formation of an oxide or any other compound on the surface of a semiconductor can be considered as a decomposition process,258 GENERAL DISCUSSION where the product is insoluble in the electrolyte. If such a layer is formed in connec- tion with an oxidation or a reduction reaction in which holes or electrons are taking part, there will exist a critical decomposition Fermi level for these processes.The kinetics of growth in thickness of such layers can be controlled by transport processes, by chemical reactions or also by electrochemical steps provided the layer has some electronic conduction. Only in the latter case can the concept of Fermi levels or quasi-Fermi levels be of some use in describing the reactions in the layer or at the contact between the surface layer and the electrolyte. If redox reactions occur at the surface of such a layer, there will be some relation between the Fermi level at the interface to the electrolyte and at the contact to the sub- strate.However, if we have to deal only with very thin layers or even monolayers, I would suppose that the Fermi level at the contact to the semiconducting substrate will more or less also control the Fermi level at the surface of such a layer, for instance via tunnelling pro- cesses. The decomposition rate, however, might be mainly controlled by the dissolu- tion rate of such a surface layer. In detail the situation will usually be extremely complicated. Prof. W. J. Albery (London) said: I would like to suggest that it may be possible for a significant potential difference to develop across the Helmholtz layer of a decom- posing semiconductor because of kinetic considerations. For instance, consider a decomposition reaction MX -t h+ --+ M2+ + 3 Xz- where the rate-limiting step for the dissolution is the oxidation of X by the holes.The removal of X will lead to an accumulation of positive charge on the excess M in the surface of the lattice. The potential would then be redistributed under kinetic control until the flux of M (increased by being driven across the Helmholtz layer) was equal to the flux of X which would be decreased because h+ would be more repelled from the surface. The generation of this field in a steady state to balance the fluxes has some similarities with the generation of liquid-junction potentials. Prof. H. Gerischer (Berlin) said : I agree that the potential distribution between the Helmholtz double layer and the space-charge layer should vary, if an electrochemical decomposition reaction is going on.The chemical composition of the surface will be changed to some extent in the steady state of decomposition, until the species which transfer the electric charge into the electrolyte can pass the Helmholtz double layer at the same rate as they are generated on the surface by holes or electrons. This effect can be considered as a generation of surface states in connection with a decomposition reaction. However, > 10l2 charged states cm-2 will be needed to obtain a considerable change in the charge distribution under potentiostatic condi- tions. To confirm this effect experimentally will be difficult because one has to dis- tinguish between changes in the charge distribution in the space-charge layer due to an accumulation in surface charge or to a deviation from the equilibrium distribution in the volume.Careful capacity measurements including the frequency dispersion might offer a chance to find such differences. Dr. J S. Curran (Lyon) said: I would like to comment on Prof. Gerischer’s in- teresting discussion of the consequences of the accumulation of charge at the semi- conductor-electrolyte interface under illumination. The change in the potential drop across the Helmholtz layer thus induced will, in the presence of a redox system, normally accelerate both the rate of photodecomposition and redox stabilizationGENERAL DISCUSSION 259 reactions. The overall consequences for the stability of the electrode then depend on the relative sizes of the electrochemical transfer coefficients (cc) for each reaction.Should it prove possible, as Prof. Gerischer suggests, to find a system in which the added surface charge is of opposite sign to the minority carriers, the stabilizing effect of the added charge may even in the worst case, where sc (redox) = cc (corrosion), be zero. However, one might expect cc (redox) < cc (corrosion) since the redox reaction is generally very downhill, in which case the system should indeed show enhanced stabi- lity under illumination. Prof. H. Gerischer (Berlin) said: The consequence outlined in this comment is cor- rect that an accumulation of surface charge with the same sign as the minority carriers will accelerate the rate of photodecomposition as well as the rate of redox reactions and that this will be unfavourable for the decomposition reaction.This is parti- cularly true if the semiconductor is stable in equilibrium with the redox reaction and the decomposition Fermi level can only be exceeded at some overvoltage for the redox reaction. However, this depends not only on the ratio between the charge-transfer coefficients cc but also on the difference between the redox Fermi level and the decom- position Fermi level. In the hypothetical case where the accumulation of surface charge with a sign opposite to the minority carriers causes a shift of the band edges relative to the elec- trolyte one could take advantage of this effect for a stabilisation of the electrode against decomposition. In this case, one could use a redox system with a redox Fermi level in the stable region further apart from the decomposition Fermi level.The consequence would be a smaller band-bending in the dark and, without a shift of the band edges, a smaller photovoltage. However, the shift of the band edges would in this case allow the steady-state photovoltage to increase without a corresponding shift of the minority-carrier Fermi level and could prevent the latter from passing the decomposition Fermi level. Dr. J. S. Curran (Lyon) said: I would like to describe some recent results which, albeit for a different semiconductor, tend to support Prof. Gomes’ finding that for n-type 111-V photoelectrodes the stabilization by a reducing agent in solution does not take place by a direct reaction with a hole.We have found that n-cadmium telluride in contact with organic electrolytes can be stabilized to some extent against photocorrosion by the reducing agent ferrocene. More to the point, the solvent effect on this stabilization is very considerable: propylene methyl solvent pyridine methanol acetonitrile carbonate nitrate i (redox) i (corrosion) (1.0 < 1 .o (1.0 103 104 Using Prof. Gomes’ notation, for the kinetics of stabilization by the direct mechan- ism: Y + h + - + Z , i (redox)/i (corrosion) = koy/k,. The free-energy change for the redox step, AGO, in eV is given by: AGO = (EyO - V,, - EJ 2: -1.0 eV260 GENERAL DISCUSSION where E is the standard redox potential for ferrocene, vfb is the flat-band potential of n-Cd Te, and Eg is the band gap. The effect of a change of solvent (A to B) on the rate constant can be written in terms of a linear free-energy relationship AAGo # = a.AAGo In ( k t / k t ) = a. AAGo/RT whereAAGo = (AG! - AGt) = EOy*B - EOy*A + V k - V$ SAG$ and a. = ___ SAGo According to the Marcus theory a. is equal to the electrochemical transfer coeffi- cient for the same redox reaction. An analogous expression for kl can be written: In (klA/klB) = alAAG1/RT. For the redox reaction, to a first approximation AAGZ, = 0, since (a) AEO, = 0 in the particular case of ferrocene is an assumption widely used in estimating free ener- gies of transfer, (b) AVfb 21 0 since only the small solvent dipole orientation contribu- tion may change, (c) a. 21 0 according to the Marcus theory since AGO is large and nega- tive.Thus the entire solvent effect must be accounted for by changes in kl. Free energies of transfer of metal ions between the solvents concerned have been measured by others, permitting an approximate evaluation of AAGl. For Cd2+ for example AG (acetonitrile -+ methyl nitrate) is approximately 0.5 eV. However, the kl step presumably creates a surface species XI with a single positive charge. Free energies of transfer for univalent metal ions are typically smaller (ca. 0.2 eV) and fur- thermore the intermediate X1 is not fully solvated. AAG1 = AGO (Xl, A -+ B) < 0.2 eV Hence klA/klB < 50. Hence we can write: and a1 < 0.5 for a downhill reaction. The solvent effects listed above, after correction for concentration (y) of ferrocene, are significantly larger than this analysis predicts.If on the other hand the redox re- agent reacts with a partially solvated intermediate Xi there will also be a solvent effect on ko since now: y + x, 2+ xi4 + z AGO = Ei - Go (Xi) + Go (xl-1) AAGo 21 -AGO (Xi) + AGO (Xl-1). Hence 0 > AAGo > -0.2 eV and 0 < a. < 0.5 since the redox reaction is less downhill: 1 > k f / k t > 1/50. Thus the large solvent effects can be explained since there is a simultaneous in- The situation is straightforward if ki + is a " unimolecular " step : crease of ko ( x 50) and decrease of ki +1 ( x 50). x, kr+lkXi+l i (redox)/i (corrosion) = ko Y/k, and solvent effects up to 2.5 x lo3 become feasible. example if ki+l was the rate constant for an activated desorption step. This would be the case for However, ifGENERAL DISCUSSION 26 1 the k , + step involves a hole, expressions like eqn (41) given by Prof.Gomes and his collaborators will arise. Hence : kl ko i (redox)/i (corrosion) 21 - and at constant light flux solvent effects can be as large as those listed above if the sol- vent effect on k i , l is more important than on k l . It would be impossible to distin- guish between the latter two cases without additional evidence; hence all that can be readily concluded is that the direct mechanism is improbable. Prof. W. P. Gomes (Gent) said: As stated in the paper, our kinetic results do not allow us to distinguish between two types of mechanism for the stabilization process, i.e., the type containing a reversible decomposition step and the type Y + Xi, respec- tively.Therefore, Dr. Curran’s comment is particularly interesting, not only because from independent reasoning (concerning the magnitude of solvent effects on stabiliza- tion), the same general conclusion is reached ( i x . , that the competition does not occur through the mechanism involving irreversible decomposition steps only and in which the reducing agent reacts by Y + h+), but also because it tends to support the parti- cular mechanism Y + Xi. It would be interesting to study the kinetics, and more specifically the light-intensity dependence, in the systems discussed by Dr. Curran. Prof. J. O’M. Bockris (Texas) said: The analysis of Prof. Gomes and his fellow authors has been carried out by neglecting chemical surface reactions. This is hardly satisfactory, because in semiconductor decomposition there may be reactions which are analogous to those met during the dissohtion of metals.Here, surface diffusion reactions, which are not dependent upon potential in any direct way, are sometimes rate-determining at potentials near to the reversible region. Furthermore, there may be surface reactions or chemical radical reactions which play an important part. Without taking these into account the analyses may be incomplete. Of course, intermediate chemical reactions would make no difference to the final equations, were they not to be rate-controlling. But there is evidence that chemical surface reactions do sometimes control the rate of metal dissolution, and it would be unexpected if nothing similar happened in semiconductors.l J.O’M. Bockris and G. Razumney, Electrucrystallizatin (Plenum Press, New York, 1967); J. O’M. Bockris and A. Despic, Encyclopaedia Brittanica, 1979, 6 , 644. Prof. W. P. Gomes (Cent) said: The possible role of chemical steps was mentioned at the end of the discussion section of our paper. We will illustrate our statement by an example. Consider case 1.2, and let us suppose that the capture of one hole by XI does not lead to X2 immediately, but to an intermediate Q which would then form X2 by a purely chemical step. Hence, eqn (29) would have to be replaced by XI + h+ - kQ Q (294 k-Q kc (29b) QT X2- Under steady-state conditions, the net rates of production of Q and X2 are equal, and given by eqn (33a) and (33b), respectively:262 GENERAL DISCUSSION in which q represents the surface concentration of Q, and where, for simplicity, it has been assumed that neither Q nor X2 reacts with Y .By eliminating q from eqn (33a) and (33b), the following expression is obtained: One can see that eqn (33c) is of the same form as the original eqn (33), which illustrates our statement that the incorporation of chemical steps into the reaction mechanism merely leads to an adaptation of rate constants. It can further be seen that an irrever- sible chemical step formally results in an irreversible electrochemical step, since putting k - , = 0 results in the cancellation of the second term in the right-hand side of eqn (33c). This example shows that, although such chemical steps do not affect the general derivations made in this paper, they may indeed, as Prof.Bockris points out, affect the rates, since the " chemical " rate constants are incorporated in the overall " electrochemical " ones. We have deliberately refrained from constructing detailed reaction mechanisms comprising chemical steps however, considering that a kinetic study such as the one under consideration does not allow one to determine the exact chemical nature of the reaction intermediates. Additional information on this point might be obtained from experiments in which, e.g., the influence of surface pretreat- ment (etching, cathodic current) on the stabilization kinetics would be studied. The subject is presently under investigation. Dr. D. E. Williams (Harwell) said: I should like to bring out the point that the chemistry of surface states is a matter which should be given some consideration.Al- though Prof. Gerischer has implied here that electron transfer between surface states and redox couples of appropriate energy in solution is a rapid process, there is a view which considers that this may in fact be more of a chemical reaction-a slow, activated process perhaps involving some reorganisation of the surface state itself. This view is embodied in some of the reaction schemes given by Prof. Gomes, where some of the intermediates, X, might be identified with surface states. In discussions of the anodic evolution of oxygen, for example on n-TiO,' or on oxidised metals,2 an intermediate which can be considered as a form of surface oxygen vacancy has been postulated, and the subsequent reaction of this intermediate with a water molecule is considered to be the rate-determining step.Other studies of oxygen evolution3 have made it clear that this same intermediate is involved in the anodic dissolution reaction which occurs in parallel with oxygen evolution. It is tempting to consider that some of the surface states postulated from photoelectrochemical studies of oxide semiconductors might be surface oxygen vacancies of some sort, also. Superstructures observed on the surfaces of compound semiconductors give rise to surface valence and conduction bands4- surface states again; it is quite conceivable that interaction with solvent or solution components will change these superstructures and hence the surface states in a way not readily predictable.It is also clear, I think, that the chemistry associated with such surface states is not necessarily readily predicted from consideration of the electronic structure of the bulk of the material, and furthermore that it is in fact the chemistry of the surface states which is actually ob~erved.~ Another question is at what stage does one make a distinction between a surface state and a new surface phase? One very definite point is that it is not sufficient, as might have been done in the past, simply to draw in an energy-band diagram a line in the band gap at the semiconductor surface, call it a surface state and then proceed to attempt to ignore it. In my view, the chem- istry of surface states is a central problem.GENERAL DISCUSSION 263 P.J. Boddy, J. Electrochem. Soc., 1968, 115, 199. A. Damjanovic and B. Jovanovic, J. Electrochem. Soc., 1976, 123, 374. D. E. Williams, unpublished work. P. Mark in Electrode Processes in Solid State Ionics, ed. M. Kleitz and J. Duprey (Reidel, Dordrecht, 1976, p. 19. J. Horkans and M. W. Shafer, J. Electrochem. Soc., 1976, 124, 1202. Prof. W. P. Gomes (Gent) said : I agree with Dr. Williams that surface states might well be associated with some of the intermediates involved in our reaction schemes. We have made this suggestion in a preliminary report on this work [ref. (3) of our paper]. Prof. W. J. Albery (London) said: Prof. Ellis has shown that he can correlate his photoemission data with the results from the photocurrents. Has he also tried to explain, quantitatively if possible, the shapes of the photocurrent-voltage curves in fig.2-7 of his paper? Prof. A. B. Ellis (Wisconsin) replied: We have not analysed the shapes of our (i, Y ) curves (fig. 2-7) in any detail. We can say that the figures are fairly representative of a large number of curves obtained for both nominally undoped CdS as well as for CdS : Te. Problems we perceive in making a quantitative analysis of our data include characterizing the electrode surface and establishing the role of recombination. For example, we are presently unable to assign with confidence the potential at the semi- conductor surface; Schottky-Mott plots indicate a strong interaction between chal- cogenide electrolyte species and CdS, suggesting that surface states play a significant role in this system.The low fill factors of our (i, Y ) curves provide evidence that re- combination processes are important. Besides e--h+ pair recombination in the semi- conductor (for which we use luminescence as a probe), we can also have short-circuit- ing at the electrode surface: photo-oxidized products can be reduced by surface-state or conduction-band electrons. We feel that once the aforementioned phenomena are more clearly understood we will be in a much better position to quantitatively treat and interpret the shapes of our (i, V ) curves. Dr. J. H. Richardson (California) said : We have completed some preliminary luminescence studies of tellurium-doped CdS electrodes in alkaline sulphide elec- trolyte. These results are complementary to the reported steady-state luminescence measurements of Prof.Ellis and co-workers. We only examined the long-wavelength (A 3 590 nm) luminescence resulting from band-gap or supra-band-gap irradiation. Generally the observed decay is not a single exponential; the nominal lifetime varies from <lo to 3250 ns as a function of laser power, wavelength and electrode potential. Fig. 3 illustrates the increase in relative importance of the longer-lived luminescence component at negative potentials. These data were taken with supra-band-gap irradiation and small laser-peak powers (476 nm and 15 W, respectively). At wavelengths closely corresponding to the band gap, the luminescence lifetime greatly increases (ca. 250 ns) regardless of electrode potential; to a great extent this is undoubtedly due to bulk excitation, At high laser-peak powers (kW) the luminescence lifetime greatly decreases regardless of elec- trode potential; to a great extent this is undoubtedly due to the expected increase in recombination rate because of the laser-induced high concentration of electrons and holes.These time-resolved luminescence measurements are consistent with and comple- mentary to our time-resolved laser-induced coulostatic measurements made at CdS264 GENERAL DISCUSSION electrodes. First, a slow cou- lostatic response (ca. 200 ns) with band-gap irradiation was interpreted as bulk excita- tion and slow electron-hole separation ; this result and interpretation is consistent with the long luminescence lifetime observed with band-gap excitation. Secondly, at high laser-peak powers the coulostatic signal can be driven past the flat-band potential Two examples are given to illustrate this agreement. 6 0 5 0 4 0 30 2oE 10 20 15 10 -1.5 -1.0 -0.5 0 0.5 1.0 FIG.3.-Potential dependence of (a) luminescence intensity measured at 0 and 20 ns delay relative to the maximum intensity and (b) pulsed photocurrent : A,, = 476 nm, laser-peak power ca. 15 W, scan rate = 100 mV s - l . in <lo ns, but quickly returns to the flat-band potential in <50 ns. It can be con- cluded, because of the similarity of the two rates at the same laser intensity (i.e., luminescence decay and potential decay), that this return to the flat-band potential is consistent with a mechanism based largely on electron-hole recombination.Prof. A. B. Ellis (Wisconsin) asked: Could Dr. Willsher provide a more detailed description of the HgS used in these experiments? I would be particularly interested in knowing the chemical nature and estimated concentration of impurities in the as- received HgS. Additionally, how reproducible are the photoelectrochemical proper- ties of the samples employed? Dr. C. J. Willsher (London) said: One of the main objectives of our work has been to see if commercially available samples of mercury(r1) sulphide could be used in photoelectrochemical cells. By means of the blackening procedure we have been able to use materials from a wide variety of sources and to produce electrodes having similar photoelectrochemical properties. We have not attempted to obtain ultra- pure mercury(rr) sulphide and study the properties of this material.Dr. M. D. Archer (Cambridge) said: The quasi-Nernstian response of the dark potentials with respect to bromide concentration in fig. 1 of Prof. Davidson's paper imply that the Br,/Br' couple is reasonably reversible at HgS. Surely, however, it is not reasonable to put oxidised form =[I] in a halogen-free medium. I would like to suggest that the dark potentials observed are mixed potentials rather than equili- brium potentials. Presumably the substantial dark current reported in Prof. Davidson's table 1 for 0.1 mol dm-3 LiI at HgS is due to iodide oxidation.GENERAL DISCUSSION 265 Prof. A. J. Bard (Texas) said: I draw attention to Dr. Archer’s comment. It is not clear that the Nernst equation is the most appropriate one to describe the mercuric sulphide electrode in the rather non-poised solutions that are being employed.The activity of oxidized form in these solutions is very low and perhaps the system is better described by ideally polarized electrode behaviour. Dr. C. J. Willsher (London) said: We agree that the Nernst equation may not be the best way of describing the systems involving halides and thiocyanate ions. How- ever, potential variation with electrolyte pH has been found to fit the Nernst equation (for red HgS), and it is on the basis of this observation that the equation is employed here. The oxidised form is given the value unity to give the Nernst equation a meaningful arithmetic form. The observed dark potentials reflect the halide ion, not halogen, since no chemical reaction is assumed to occur before illumination.It is correct to assume that the large dark current noted in table 1 is due to iodide oxidation. Prof. J. O’M. Bockris (Texas) said: In the praiseworthy attempt of Davidson and Willsher to consider solution effects on photoelectrochemical kinetics, they have overlooked the probable connection between specific adsorption and surface states. For example, they have assumed on the third page of their paper that the flat-band potential can be identified with the onset potential. In the presence of sufficient sur- face states (likely as a result of changes of solution composition in certain directions) this may not be an admissible assumption. Dr. C. J. Willsher (London) said: We agree that identification of the flat-band po- tential with onset potential is not entirely admissible, since surface states, the presence of which we have invoked, will render these two potentials unequal.Perhaps a better statement would be “ the flat-band potential lies in the region of the photocurrent onset potential, with surface states and specific adsorption accounting for any difference between the potentials ”. Dr. L. Peter (Southampton) said: Dr. Isabel da Silva Pereira and I have studied the photoelectrochemical properties of mercuric sulphide layers grown electrochemically on a mercury anode in sulphide solutions.’ The results are interesting since they appear to throw some light on the solid-state properties of the black metacinnabarite phase of HgS.Fig. 4 illustrates the rather complex electrochemistry of mercury in a buffered sulphide solution, details of which have been discussed elsewhere.* The important point to note here is that interrupted illumination of the anodic HgS film gives rise to an anodic photocurrent; in other words the deposit is an n-type semi- conductor. The red cinnabar form of HgS has a band gap of ca. 2 eV, whereas the metacinnabarite modification is reported to be a ~emi-metal.~ Our results suggest that this is not so. Fig. 5 demonstrates that the photoresponse of the black HgS film grown on mercury extends out to 1300 nm, and the inset shows how we were able to estimate the band gap to be 0.92 eV (the plot corresponds to the data treatment appropriate for an indirect-band-gap material.4 We conclude that the black form of HgS is an n-type semiconductor with an indirect band gap of 0.92 eV.We have also considered the photostability of HgS. The appropriate decomposi- tion reaction in the pH range 2-14 is HgS + 20H- + 2h+ --+ HgO + H20 + S.266 GENERAL DISCUSSION FIG. 4.-Cyclic voltammogram at 2 mV s-' of mercury in 0.1 mol dm-3 NazS + 1 mol dm-3NaHCOJ under periodic illumination with the white light (100 W tungsten lamp). We calculate a value of pEdecomp at pH 7 of +0.501 V us. NHE and at the pH of our system this becomes +0.442 V us. NHE. Fig. 6 shows the energy-level diagram which we have derived for HgS in sulphide solutions from experimental data, and contrasts it with equivalent diagrams for CdSS and Bi2S3.4 It is clear that HgS should be ther- modynamically unstable with respect to anodic photodecomposition, and it shares this property with the other sulphides. However, an effective stabilisation of HgS in sulphide solutions is guaranteed because the solubility product of HgS is consider- ably lower than that of HgO. In the absence of sulphide, however, photodecomposi- i 1 i O.'O 1 I A/nm FIG. 5.-Wavelength dependence of the photocurrent conversion efficiency (B) of a mercuric sulphide film grown at +0.9 V us.Hg/HgS. The inset shows how the indirect band gap of the material was determined.GENERAL DISCUSSION 267 tion should give rise to HgO in the pH range 2-14, and the solubility of this product may be sufficiently low to preclude its detection in solution. M. I. da Silva Pereira and L.M. Peter, to be published. ' M. I. da Silva Pereira. Ph.D. Thesis (Southampton, 1979). G. G. Roberts and R. Zallen, J. Phys. Chem., 1971, 4, 1890. L. M. Peter, J. Electroanalyt. Chem., 1979, 98, 49. L. M. Peter, Electrochim. Acta,, 1978, 23, 1073. 1 [black HgS) p E d e c I i -VB p E d ec LVB Cd S [red HgS) Bi2S3 FIG. 6.-Band energies on the hydrogen scale determined for films of HgS, CdS and Biz$ grown on the parent metal and measured in 0.1 mol dm-3 NazS (buffered to pH 9 in the case of HgS and CdS). The figure also shows the calculated anodic (,Ed,,) and cathodic (nEdec) decomposition potentials (see text). Dr. C. J. Willsher (London) said: The work of Drs. Peter and da Silva Pereira is most interesting, especially with respect to the formation of a mercury sulphide which responds out at 1300 nm.We too have observed extension of response beyond 650 nm. Fig. 5 of these remarks does, however, indicate maximum photocurrent con- version efficiency at ca. 480 nm, which implies red HgS is also present. Thus, a mix- ture of red and black sulphides exists, and since the starting materials are mercury metal and the sulphide anion, it is likely that many non-stoichiometric Hg-S systems exist-the authors note that mercury in a buffered sulphide solution exhibits complex electrochemistry. They, quite correctly, conclude a black form of HgS is present, an indirect-band-gap, n-type material (Eg = 0.92 eV). Our blackening was effected by an entirely different means, and we conclude the black material contains metacinnabar, the semi-metal.Regarding the photostability of HgS, we have no evidence for HgO formation, as suggested in the comments. Electrodes bearing both red and yellow HgO (authentic) have been investigated photoelectrochemically and results indicate HgO is very un- stable and exhibits p-type behaviour. No p-type behaviour has been noted for red or blackened HgS in aqueous electrolytes containing non-oxidisable salts. Thus, HgO formation in the absence of sulphide would be detected by a marked change in photoelectrochemical properties. We have failed to detect any oxygen in blackened HgS by ESCA.2 It was found that 0.1 mol dm-3 Na,S was a most unsuitable electrolyte to investigate HgS, since the material dissolves fairly readily. We do not exclude the existence of other forms of black HgS.We have never observed this.268 GENERAL DISCUSSION We have determined the relative positions of band edges and decomposition levels for red HgS,lv3 based on HgS + 2h+ -+ Hg2+ + S HgS + 2e --f Hg + S2- and fig. 7 illustrates band edges and decomposition levels at pH 9, which should be compared to fig. 6 of Dr. Peter’s comments. The band edges are quite similar in both figures, as is the relative position of the decomposition levels to the bands. and C. J. Willsher, Ph.D. Thesis (University of Leicester, 1980). R. S. Davidson and C. J. Willsher, J. C. S. Dalton, 1980, 833. R. S. Davidson and C. J. Willsher, J. C. S. Furuduy I, 1980,76! w 0-- 51 2 s + I - - > 2587. FIG. 7.-Band edges for red HgS in a pH 9 electrolyte (aqueous sodium nitrate) and decomposition levels based on : HgS + 2h+ --+ Hg2+ + S (anodic) HgS + 2e --+ Hg + S2- (cathodic).and Dr. D. S . Ginley (New Mexico) said: I am afraid I have to be a bit defensive. It is gratifying that our electronegativity model is being applied to so many systems, but some of the discrepancies observed may not be due to faults in the model. One of the fundamental assumptions of the model is that the flat-band predicted by the electronegativity calculations is that at the point of zero zeta potential (PZZP), i.e. the point where the adsorbed cationic and anionic charges are equal and the net surface charge is zero. There can be a dipole contribution as well across the Helmholtz layer from water or other neutral molecules but this is normally quite small.Unless the measured V,, for an electrode is corrected to the PZZP for the appropriate potential determining ions, we would find agreement between theory and experiment fortuitous indeed. Prof. R. S. Davidson (London) said: We are not attributing any discrepancies in flat-band potentials to the model of Butler and Ginley. We realise that the PZZP is not accounted for when applying the model to obtain the conduction band energy of mercury(1r) sulphide, but employed it to provide a guide to the energies of the band edges.GENERAL DISCUSSION 269 Prof. J. O’M. Bockris and Dr. V. Guruswamy (Texas) said : We want to state our appreciation of Dr. Tributsch’s publication of the d/d transition as an important mode of absorption in semiconductors used in photoelectrochemistry.The contribution has given rise to a broadening of perception of what compounds may be available for us to deal with. We, however, want to report that Mr. Lee Handley at the Flinders University of South Australia made the same suggestion early in 1977. His thoughts (the rational seeking of d-d transitions in transition metals within perovskite lattices) gave rise to the development by us of the “ lanthanum chromite ” electrode (electrodes formed by mixing lanthanum oxide and chromium oxide on the surface of titanium and heating for ca. 24 h in argon). This type of preparation has given rise to valuable electrodes which give more energy conversion and better practical quantum efficiency than Ti02. There is evidence in some of them that lanthanum chromate is active, and X.P.S.analysis carried out by Dr. Vanecia Young in the Department of Chemistry at Texas A & M University has shown that the surface of these electrodes is a mixture of lan- thanum chromate and titanium chromite. Dr. H. Tributsch (Meudon) said: I am very pleased to see oxide materials with d-d transitions emerging. They should have clear advantages due to their potential resistance against oxidation which would be needed for light induced oxygen evolution. Before concentrating on sulphides and selenides I have myself tried to identify possible oxide materials. I failed since I was unable to get sufficient information on detailed electronic structures. Prof. H. Gerischer (Berlin) said: Semiconductors of this class with layer structure are very resistant against corrosion, if the surface consists of perfect van der Waals planes.This stability against photocorrosion is attributed in the paper to the fact that d-d transitions control the light absorption in the energy range above the band gap. These transitions have indeed little effect upon the bonding between the metal atoms and the chalcogen atoms. However, this does not explain the stability against photocorrosion. It is obvious from the structure of these materials that the wave- functions of the metal atoms are considerably shielded against interaction with reac- tants from the solution. This does not prevent electron transfer, since very little elec- tronic interaction (overlapping of the wavefunction) is needed to get a high rate of electron transfer.If, however, new chemical bonds have to be formed in order to compensate for the loss of a strong interaction with other atoms or molecules, much more overlapping of the wavefunctions between the reactants is necessary. This is prevented in these materials for the metal atoms by the presence of the filled electronic states of the chalcogen atoms and therefore no attack should occur on the perfect van der Waals planes. The situation is completely different at edges and probably also at dislocations on the surface, since there the electronic d-orbitals of the metal atoms can largely overlap with donor orbitals of reactants in solution. Oxidative attack can very easily occur at such surface sites, as has been shown in a number of experiments by several groups.The unfortunate situation for semiconductors with layer structure is that poly- crystalline materials are therefore very susceptible to corrosion. Only if measures can be found to stabilize these active sites on the surface of such materials will there be some hope that this type of semiconductors can be used in practical devices for the conversion of solar energy.270 GENERAL DISCUSSION One brief remark may be made on fig. 1 of Dr. Tributsch's paper. The picture of the electronic constitution of these materials given there is only a very crude approxi- mation. In reality there is always some mixing of the d-states of metal atoms with the p-states of the chalcogens. This also explains that these materials can be photo- oxidized in presence of water with the formation of sulphites and sulphates or selenites and selenates as final products if neither elemental sulphur nor selenium is obtained.The formation of S-0 or Se-0 bonds caused by the accumulation of holes at the surface indicates that the hole on the surface has partial p-character which has been confirmed by the analysis of photoelectron spectra obtained from these materials. Dr. H. Tributsch (Meudon) said: My arguments against the concept proposed by Prof. Gerischer are the following: Since the chemical products of photocorrosion are sulphates and selenates, respectively, it is the photoelectrochemical reaction with water which deserves our special attention. Bond formation is certainly more favourable at step sites since they expose dangling orbitals. However, while anodic dark currents (in the presence of water) disappear with negligible step-site concentra- tions photocurrents reach their highest values and most negative onset potentials [ref.(12) of my paper]. Since electron transfer from OH- ions or water needs, for energetical reasons, chemically bonded reaction intermediates or products we have therefore to assume that the illuminated van der Waals interface is chemically reactive and not only an acceptor for electrons. My opinion is that it is not the existence of a van der Waals surface which reduces the susceptibility of layer materials to photocorro- sion in regenerative solar cells but the fact that reducing agents with fast electron- transfer kinetics like I- are generating photocurrents at more negative electrode poten- tials (cf.fig. 4 of my paper). When operated in a regenerative mode the solar cell therefore draws most of the current from the iodide oxidation and little or negligible current from the photoreaction with water, which would lead to photocorrosion. Photocorrosion depends on the concentration of step sites mostly in an indirect way through the (current, voltage) curves which depend on the concentration of step sites. The working potential of the solar cell should also have a crucial influence on the rate of photocorrosion. The MoSe, crystal which we used for a long-term experiment (10 months) on a solar cell [ref. (7) of my paper] had a relatively high concentration of steps and a quite low fill factor. We do not think that anyone has yet shown that a polycrystalline layer-type material is very susceptible to corrosion.The main prob- lem is that it shows considerable recombination losses. It was only intended as a comparative picture to indicate the position of the energy gap in different transition-metal compounds. There should be some mixing of d-states with p-states of chalcogens. However, the experimental observations that only negligible amounts of molecular sulphur or elemental selenium are anodically liberated, that photocur- rents disappear in contact with acetonitrile [ref. (13) of my paper], the observation by L. F. Schneemeyer and M. S. Wrighton' that even natural crystals of MoS, (which should have many defects) can sustain the photo-oxidation of CI- in organic media without corrosion, make it difficult for me to believe that the contribution of p-states to the photoreactions is kinetically important.More research on this subject is needed. I agree that fig. 1 of my paper shows a scheme which is too crude. J. Amer. Chem. SOC., 1979, 101, 6496. Dr. H. S . Jarrett (Delaware) said: I should like to call attention to another class These are of compounds possessing d-states similar to the metal dichalcogenides.GENERAL DISCUSSION 27 1 Ti 02 I distorted perovskites (RE)Rh03, where (RE) is an element in the first rare-earth series.l The 4dt2 sub-band of Rh+3 is filled with six d-electrons, and it appears to lie at the top of the valence band. The bottom of the conduction band is probably formed by the empty 4de levels of the Rh+3. LuRh03, which has a band gap of 2.2 eV, is the most efficient of the series, because the 4f shell is filled, and no partially filled states that could promote electron-hole recombination are present in the energy gap.Hole mobility in the valence band is found to be ca. 2 cm2 V-l s - ~ . ~ This high mobility is believed to be a result of the larger spatial extent and hence greater orbital overlap of second-row transition elements. The oxide, as synthesized, is p-type but may be doped n-type by addition of thorium during synthesis2 The open-circuit voltage of p-LuRhO, under illumination is ca. -0.4 V vs. SCE at pH 12.6. The current characteristic of p-LuRhO, compared with n-TiO, is shown in fig. 8. There is sufficient photopotential developed so that a cell comprising -- 2 w 2 i l l , [ I I / 1, FIG.&-Current characteristics of p-LuRh03 and n-Ti02 (0.1 mol dm-3 NaOH). n-Ti02 as the anode and p-LuRhO, as the cathode will decompose water when both electrodes are illuminated with light of band-gap energy or greater. The short-circuit current i, and cell potential are also shown. A small amount of power can also be delivered to a load resistance while maintaining photoelectrolysis.2 U.S. Patent no. 4, 144, 147. ’ H. S. Jarrett, A. W. Sleight, H. H. Kung and J. L. Gillson, J. Appl. Phys., 1980,51, 3916. Mr. P. R. Trevellick (Oxford) said: I would like to question the economic feasibi- lity of the solar batteries introduced in this paper. As an example, I will take the case of a small device capable of supplying 500 W for a maximum period of 8 h between charging cycles. This can be envisaged as the smallest possible capacity that would be sufficient to supply a single household with all of their electricity requirements.Straightforward arithmetic shows that 500 W for 8 h corresponds to 1.44 x lo7 J.272 GENERAL DISCUSSION The best storage cells in this paper generate a potential of 0.5 V, and so the charge storage capacity necessary at this working potential is 2.88 x lo7 C. Division by the Faraday constant reveals that 300 moles of redox couple are consumed during charge or discharge cycles. In fact, any practical device will require a quantity of redox couple in excess of this figure. This is because near-exhaustion of either oxidised or reduced species would cause a violent swing of redox potential, and the cell would not function efficiently.It would seem that the considerable quantity of electrolyte required will make solar batteries prohibitively expensive with all but the cheapest of redox couples. Cer- tainly, 300 moles of iodide is an unrealistic requirement for a cell providing only 4 kW h of storage capacity. Prof. A. J. Bard (Texas) said: With regard to the previous question, solar storage systems do appear to be economically reasonable. For example, a system under intensive investigation at this time is the Texas Instruments system based on produc- tion of hydrogen and bromine at silicon-cell panels. The stored reactants are later recombined in a fuel cell to produce electricity. An analysis by the Texas Instruments group suggests that this approach is practically and economically viable.Let me also mention some of our recent work on redox flow cells and solar storage systems. A major requirement in such systems is the discovery of electrolytes which are highly soluble (> 1 mol dm-3), show rapid reactions at the electrodes, are highly stable and have standard potentials in the required range for the semiconductor or counter-electrode. Moreover, in practical systems, these electrolyte solutions must be safe, non-corrosive and inexpensive. Ideally, the solutions in both half-cells would also be compatible so that intermixing through the membranes would not be particularly deleterious. Dr. K. S. V. Santhanam and Y. W. Chen at our laboratories have been investigating iron (+3/+2) couples.By appropriate choice of ligands, one can shift the potential over a range of ca. 1.3 V, depending upon whether the iron@) or iron(Ir1) form is more highly stabilized. Many of these couples are highly soluble, although colour of the complex species appears to be a problem. With regard to intermixing of the solutions, I would like to ask the authors if they have had any problems with transport of either sulphide or cadmium ions through the membrane thus causing precipitation of cadmium sulphide during extensive use. Prof. H. Gerischer (Berlin) said: I appreciate that such a storage system has been investigated by Drs. Ang and Sammells. These systems have the advantage that redox couples having low overvoltages can be used. The energy losses at the semiconductor electrodes can therefore be kept much smaller than in the devices for water photo- electrolysis.A problem is certainly how to reach a high enough storage capacity. One needs electrolytes with a high concentration of the redox system but still remaining transparent enough for the incident light. The position of the membrane imposes an important problem for the device. In an earlier paper1 I proposed a possible arrangement where the membrane is located normal to the incident light. A somewhat more refined construction of this type is shown in fig. 9. In order to get a high enough storage capacity, it will certainly be necessary to store the electrolyte outside the solar cell and to use another device providing the power output for the consumer. A difficult problem seems to be to find suitable redox systems with negative redox potentials This makes the Cd2+/Cd reaction impractical.GENERAL DISCUSSION 273 which do not decompose water.may still be questionable, it appears to me very important to explore them. However, although the economics of such systems H. Gerischer, Solar Power and Fuels, ed. J. R. Bolton (Academic Press, New York, 1977), p. 77. insul f ra ,sting me - elec I t r i c h rn&al\serniconductor contact electrode hlectrode line FIG. 9.-Solar battery with two separated xedox electrolytes A and B. Semiconductor-metal diodes act as voltage generators. Dr. A. F. Sammells (Illinois) said: The use of the Cd2+/Cd couple as an acceptor species was chosen because it was one of the few candidates having an equilibrium potential compatible for photocharge by the CdSe/sulphide, polysulphide or GaAs/ selenide, polyselenide photoanode, donor redox species.I agree, however, that for a practical system the redox reactants and reaction products should be soluble so that they are compatible for their convenient removal from the photoelectrochemical cell for storage. The Texas Instruments system mentioned by Prof. Bard is certainly very interesting and is, of course, at a more advanced stage of development than our systems. How- ever, the cost for hydrogen storage and the long-term stability of practically sized fuel- cell electrodes in this application are probably still unanswered questions. The use of porous flow-through electrodes for the electrochemical discharge of photoelectro- chemically produced redox species may have inherent advantages both in lifetime and in overall cost.The solubility of cadmium complexes can be expected to be fairly low in the elec- trolyte used, but certainly there is the possibility of cadmium sulphide formation either by direct mixing of donor and acceptor species or by microscopic precipitation within a given separator. With separators used in this work, however, we have seen no evi- dence of this being a problem, although cells have only been operated for up to l month. The present emphasis of our work is on soluble redox couples. For longer cell tests, such a problem may become manifest.274 GENERAL DISCUSSION Prof. J. O’M. Bockris and Dr. V. Guruswamy (Texas) said: In view of the title of the paper by Drs. Ang and Sammels, we would like to report a 6% conversion of water to hydrogen at oxygen using “ lanthanum chromite ” with platinum.An out- side potential source was used to change the energy level within the semiconductor, but the 6% is calculated after allowing for the use of energy from the battery. Cor- respondingly, we would like to report unpublished measurements which show that lanthanum rhodate and gallium arsenide give rise to chlorine from sodium chloride, and that we have synthesized formic acid and oxalic acid from CO, in DMF to which a little water has been added. It seems likely that an analogy can be made between the action of enzymes and the potentialities offered by a series of micro-photoelectrochemical couples (the analogues of semiconductor chips) of varying cell potential, placed in contact with complex chemical systems, and insulated.Such contrived photoelectrochemical multi- synthesis seems an avenue for recycling of, e.g., rubbish and sewage, as the cost of other forms of energy rises. Prof. L. R. Faulkner (Illinois) said: The thicknesses of the gold overlayers used in Are any data available to indicate the minimum thick- Miller’s work are substantial. ness of gold required to manifest the Schottky junction in the electrochemistry? Dr. B. Miller (New Jersey) said: The gold layers were made sufficiently thick to reduce the effect of porosity and the consequent degradation by attack of the exposed semiconductor and undercutting of the metal. The required thickness for the barrier formation is less than even a few monolayers and, if either the film were perfectly con- tinous or the semiconductor adequately passive in the electrolyte, such thin layers should function well.Dr. L. M. Peter (Southampton) said: Could Dr. Miller give some idea of the valence state and coverage of the ruthenium adsorbed at the surface of n-GaAs? Presum- ably it is present in covalent rather than ionic form if the surface coverage is appreci- able. If so, I wonder whether it is fair to treat such a surface layer as a “ surface state ”; it seems more reasonable to assume that some kind of ruthenium doped selenide/oxide layer is present, and it would be interesting to know something more about its properties and how they affect recombination processes. Dr. B. Miller (New Jersey) replied: The ruthenium surface coverage is about 1/3 of a monolayerl though the oxidation state of the species has not been identified.Rutherford backscattering and Auger data indicate that the Ru is non-substitutionally adsorbed at the n-GaAs surface rather than being dissolved in an oxide or selenide phase. Auger results2 show that, as oxide grows onto Ru treated n-GaAs, the Ru stays at the GaAs-oxide interface. In our interpretation the chemisorbed ruthenium stabilizes the freshly exposed surface, following stripping of most possible surface films and also metallic As3 by basic selenide-polyselenide. As a result, donor surface states like those present on the untreated GaAs surface are removed. These latter states are likely loci for recombination.Recent studies by Kar et aL4 establish the removal of donor states. R. J. Nelson, J. S. Williams, H. J. Leamy, B. Miller, H. C. Casey Jr, B. A. Parkinson and A. Heller, Appl. Phys. Letters, 1980, 36, 76. J. E. Rowe, personal communication. A. Heller, in Photoeffects at Semiconductor-Electrolyte Interfaces, Adv. Chem. Ser., ed. A. Nozik, in press.GENERAL DISCUSSION 275 S. Kar, S. Jain, A. Heller, S. Ashok and S. J. Fonash, Abstracts IEEE International Electron Device Meeting, Washington, D.C., Dec. 8-10, 1980. Dr. M. D. Archer (Cambridge) asked: Is it possible to detect adsorbed Ru"' or Pb" on the GaAs surface by cyclic voltammetry, or to decide by any other means whether valence changes could be a significant feature of the surface incorporation of these ions? Dr.B. Miller (New Jersey) said : The presence of a submonolayer amount of these metal ions on the semiconductor surface could not readily be detected electrochemi- cally in either the 0.01 mol dm-3 solutions frGm which they are adsorbed or in the rela- tively concentrated selenide-polyselenide cell electrolyte. Attempts to do these in another electrolyte after the adsorption step and surface rinsing have not been made. Surface methods such as ESCA, which are sensitive to oxidation state, have been difficult to apply because of mutual interferences of the elements involved and the question of the mechanism of incorporation remains open. Mr. M. P. Dare-Edwards (Oxford) said: I would like to report on some work on chemical modification of an electrode surface, similar to that of the Bell group but in this case on p-type GaP electrodes.' Fig.lO(a) illustrates the basic problem : despite potential/V us. SSE FIG. 10.-Cyclic voltammagrams of p-GaP (TIT) in 0.5 mol dm-3 H S 0 4 illuminated with white light from a high-pressure xenon arc lamp (a) - clean, as etched surface, (b) - - - following 20 s dip in 0.01 mol dm-3 RuCI3 in 0.1 mol dm-3 HN03. Ufb := 0.65 V. gallium phosphide possessing a flat-band potential of +0.65 V vs. SSE (SSE = standard silver electrode = -0.040 V us. SCE), the photocurrent due to hydrogen evolution does not rise significantly until some 0.6 V away from the flat-band, i.e., the fill factor of any photoelectrolysis cell employing GaP for hydrogen evolution will be exceedingly low. Results taken at varying light intensities show that in this potential region near flat-band, absolute photocurrent efficiency decreases as light intensity increases. This is in contrast to the expected behaviour from systems possessing a fixed concentration of inherent surface states.We have produced a kinetic model invoking photogener-276 GENERAL DISCUSSION ated intermediates in the hydrogen evolution reaction which act as parasitic recom- bination centres (i.e., photogenerated surface states) : electronlhole pair generation hv 3j e z + P& in space-charge layer surface faradaic process redox potential = E ~ ( ~ ) evolution of hydrogen redox potential = E ~ ( ~ ) ect + Hi&q/ads k,_ Hids e ~ < + + H&q/ads 5 H2 Hids + p:b k,\ Ha+q,ads { parasitic surface recombination migration of hydrogen atoms into Hids A H k k semiconductor surface However, a method has been discovered that reduces the effect of the parastic recombination reaction (k4) above.Using a chemical pretreatment similar to the Bell group (0.01 mol dm-3 RuCl,, in 0.1 mol dm-3 HN03 or 0.1 mol dm-3 HCl), the electrode performance is dramatically improved in the potential range near flat-band [see fig. lo@)]. Note that there is no significant change in the saturation efficiency of the electrode. Action spectra demonstrate that there is no significant change in spectral response of the electrode save in the sub-band-gap region where photocurrents are enhanced by a factor of two. This difference is however insufficient to account for the observed change in fill factor.Detailed impedance analysis of the clean and " ruthenated " electrodes do how- ever show considerable differences. Results were fitted by least squares to a five- component circuit that allows for additional contributions to interfacial capacitance from surface states : Rss css csc 1 1 where Rb = solution + bulk semiconductor resistance, R, = space-charge + fara- daic resistance, C,, = space-charge capacitance, C,, = surface-state capacitance and R,,C,, = time constant for surface states. Using results obtained from such impedance analyses, plots of Csc-2 against vol- tage produce straight-line graphs as expected for Mott-Schottky behaviour. Identical intercepts (+0.675 V us. SSE) are obtained for clean and " ruthenated " surfaces and so no simple change in flat-band potential can explain the observed fill-factor behaviour.With a clean surface, there is a small peak in surface-state capacitance (C,,) at +0.35 V us. SSE [fig. 1 l(a)]. This agrees well with the supposition that the loss of efficiency near flat-band is due to surface states lying up to 0.6 eV above the valence band edge. We have also further evidence from a.c. cyclic voltammetry that illumination will cause an increase in the contribution to interfacial capacitance from photogenerated surfaceGENERAL DISCUSSION 277 states. On " ruthenation ", the surface state capacitance near flat-band increases by nearly an order of magnitude [fig. ll(b)]. This appears to be in opposition to the expected correlation of increased efficiency of hydrogen evolution near flat-band being related to a decrease in surface-state density.In consequence of this increased concentration of surface states following the ruthenium treatment, we can only explain the increased efficiency of H2 evolution by a corresponding decrease in recombination rate (k, J. ) and not by a catalysis of the H2 potential/V us. SSE FIG. 11.-Variation of CSb with potential for the p-GaP (100) surface; (a) 0, clean surface; (b) 0, surface after ruthenium treatment. evolution reaction (k3 t ). We have some evidence to suggest that this decrease in recombination rate may be due to a positive shift in the band edges when the electrode is held at potentials between 0 and +0.6 V us. SSE. Although crude ruthenium(1r) produces the greatest magnitude of improvement in response, many other metal ions also have some lesser effects (see table 2 below).Note the poor effectiveness of Pt dips supports the claim that pure catalysis of hydro- gen evolution cannot provide a satisfactory explanation of the efficiency enhance- ments: As to the nature of the ruthenium species responsible for the above behaviour, some experiments were performed using " pure " ruthenium(rr1) compounds, notably K2RuC15*H20. Results from predips of this material dissolved in HCl showed much smaller improvements in response than crude RuCl,. However, when dissolved in HN03, in which the orange complex decomposes and darkens to produce various nitrosyl and nitro derivatives, the improvement in response showed a direct correla- tion with the amount of decomposition.Indeed the crude RuCl,, as purchased, is well-known to contain quite high concentrations of nitrosyl complexes. It would therefore seem likely that the active ingredient in these electrode modifications is a nitro or nitrosyl derivative of ruthenium which may complex directly with phosphorous or adsorbed hydroxyl groups at the semiconductor surface. Although this system is in many ways different from the n-GaAs/selenide cells, I would be interested in any comments that Dr. Miller has on these observations. In278 GENERAL DISCUSSION particular, it is interesting to note the significant contrast between the model put for- ward by the Bell group to explain their metal-ion-induced efficiency enhancements with the experimentally supported model proposed in this work on p-type 111-V ma- terials.Might the experimental techniques used here be applied also to the n-GaAs system to provide additional insight into the detailed behaviour of that electrode/elec- trolyte interface? M. P. Dare-Edwards, A. Hamnett and J. B. Goodenough, J. Electruanalyt. Chern., in press. TABLE 2.-EFFECTS OF VARIOUS METAL IONS ON HI-V SEMICONDUCTORS FOR Hz GENERATION electrode positive effect null effect negative effect All solutions are 0.01 mol dmW3 in metal ion dissolved in 0.1 mol dm-3 HCl or 0.1 mol dm-j HN03. " Ru"' " corresponds to solutions made up from crude RuC13-xHz0. Dr. B. Miller (New Jersey) responded: We see no reason for analogy between results on n- and p-type photoelectrodes of different 111-V materials or for photo- electrodes in electrical power and H2/0, generating cells.Dissimilar surface chemis- tries and environments are involved and the kinetic issues for regenerative and hydro- gen-producing cells are quite different. We have previously noted that, while R u + ~ chemisorption significantly modifies surface combination velocities, carrier lifetimes and luminescence intensities at n-GaAs air interfaces, no measureable effects were observed with p-GaAs. Yet spectroscopic ellipsometry' shows the surface chemistry of both n-and p-type GaAs in air, following each step of our treatment procedures, to be the same. The location with respect to the valence band and conduction band edges and type of surface states (acceptor or donor) is critical to the consequences for recombination and pinning.For example, donor states appear to be removed from the proximity of the conduction band after the ruthenium-n-GaAs treatment and accep- tor states added near the valence band. This would reduce recombination at photo- anodes but might well worsen the problem at photocathodes. Metal ions also introduce significant catalytic effects in the systems reported by Dare-Edwards and Hamnett, but these niay have little to do with the subject of our own paper. D. E. Aspnes, personal communication. Prof. A. J. Bard (Texas) said: I am pleased to see that Dr. Parkinson is reporting crystals of WSe, which show an efficiency of 10.2% in sunlight. We, previously, have reported' an efficiency of 14% under red light illumination (A < 590 nm) for n-WSe, in an iodide/iodine medium.A requirement for such high efficiencies on these elec-GENERAL DISCUSSION 279 trodes, as Dr. Parkinson also demonstrates, is the absence of edges on the electrode surface. A useful method of characterizing such edge-free electrodes is by capacitance plots. A high-efficiency electrode will show linear Mott-Schottky plots which are independent of frequency over the range 500 Hz to 10 kHz as shown by the results of Dr. Fu-Ren Fan from our laboratory (fig. 12 and 13). ' A. J. Bard et al., J. Electrochem. SOC., 1980, 127, 518; J. Amer. Chem. SOC., 1980, 102, 5142. 1 .o 0.6 0.2 -0.2 - 0.6 potential/V us. SCE FIG. 12-Mott-Schottky plot for n-WSez electrode in aqueous 0.5 mol dm-3 Na2S04 (pH 4.5) at the following frequencies: 0,200 Hz; 0 , 5 0 0 Hz; 0 , 2 kHz (similar behaviour is found for frequencies up to 10 kHz).Dr. B. A. Parkinson (Iowa) said: I would like to make some comments about the use and measurement of efficiencies of photoelectrochemical devices. Because &- ciency numbers published in various reviews and papers are quoted by scientists who are not in this field and often even by the lay press I believe some standardization should be attempted. Photoelectrochemical devices, although not at the point of immediate ' O k 8 .o potential/V us. SCE FIG. 13.-Mott-Schottky plot for p-WSe2 electrodes in aqueous 0.5 mol dm-3 NaZSO, (pH 4.5) at the following frequencies A, 1 kHz; 0 , 2 kHz; 0, 5 kHz; 0, 10 kHz.280 GENERAL DISCUSSION application, are also competing with several other emerging technologies and it would be to our advantage if only meaningful efficiencies were quoted and compared.My personal feeling is that real solar efficiencies should be the primary number to measure and quote for several reasons. First it is the ultimate source for which the device should be designed. Secondly, although there are some changes from locale to locale in both spectral distribution and intensity, it is a light source which is avail- able to everyone including our third-world colleagues who may not have the resources to obtain standard sources etc. but are able to obtain a total radiant flux measuring device at a much lower cost. And finally it offers a great excuse to spend a sunny afternoon outside in the sun rather than in the lab! A secondary measurement would be a monochromatic measurement.The power and wavelength of a monochomatic source is easily duplicated from laboratory to laboratory and would provide an easy assessment of a material or systems perform- ance in your hands relative to those of other workers. Prof. H. 0. Finklea (Virginia) said: What evidence does Dr. Parkinson have indi- cating that the difference photocurrent spectrum shown in fig. 2 of this paper is due to a surface electronic transition as opposed to a bulk electronic transition? Do the ligand or the “ semi-intercalation ” treatments affect the photocurrent spectrum at the same wavelengths ? Prof. H. Gerischer (Berlin) said: I wonder about the very large size of the photo- currents attributed in fig. 2 of the paper to transitions from the valence band to surface states.Usually, the matrix elements for such transitions are small and the quantum yield should therefore be very low. Has it been checked by studying the potential dependence whether this is really a surface state and not a transition in the bulk? Dr. B. A. Parkinson (Iowa) said: Prof. Gerischer and Dr. Finklea bring up a very important point and I have to say that we are in the process of doing measurements on some well-formed WSe, crystal faces, other than van der Waals faces, on crystals that we have recently grown. The measurement of the photocurrent spectroscopy on these faces with p and s polarized light and the effect of surface treatments on the spectra will help to answer these questions more definitely than the spectra on these rough as-grown edges employed in this study.Another possibility for explaining the larger sub-band-gap response on an edge- mounted crystal would invoke the indirect transition in this material which many workers have measured to be at ca. 1.1 eV. The conduction anisotropy of the material might allow for the more efficient collection of carriers formed by the deeper light penetration associated with an indirect transition at wavelengths > 930 nm. We hope to distinguish these interpretations with further measurements. Dr. L. M. Peter (Southampton) said : Since we have heard quite a lot about “ organic semiconductors”, I would like briefly to examine the usefulness of this concept. Non-polar organic compounds are generally characterised by sufficiently weak inter- molecular forces that the optical absorption properties of the solid are fairly similar to those of the molecule in solution.The optical excitation processes are described in terms of Frenkel excitons, and interband optical transitions do not play a significant role in charge-carrier generation. I would like to take the criteria of semiconductivity listed by Prof. Ayer, and ask if they are sufficient to distinguish between a true semi- conductor (with band-widths >kT) and an insulator with a high density of traps. Criterion (1) is that the conductivity should increase with temperature, also appliesGENERAL DISCUSSION 28 1 to trapped carriers, since their thermalisation will control the conductivity. Criterion (2) will also be fulfilled if oxygen acts as an electron trap.Criterion (4) is direct evid- ence against the use of the band model in the description of optical transitions in orga- nic solids-the absorption spectrum corresponds in fine structure to a molecular excita- tion broadened and shifted by intermolecular interaction. In other words, the optical excitation gives rise in the first instance to an exciton, not to a charge-carrier pair. These mobile excitons (singlet, triplet or charge-transfer states) can diffuse through the insuIator and interact either in the bulk or at the surface. If the excitons interact with trapped carriers, a photocurrent known as a photoenhanced current is produced. If the exciton interacts at the surface of the material, on the other hand, a photoinjec- tion current results.It is therefore essential to distinguish between bulk and surface photoeffects, and this can only be done if the properties of the contacts to the solid are well-understood. In order to observe true photoinjection currents it is important that charge carriers should not be injected in the dark, i.e. blockingcontactsareneeded. Exciton-trapped carrier interactions have been studied quantitatively, for example in the case of tetracene,’.’ and a clear distinction between bulk and surface exciton pro- cesses can be made. These processes have not been adequately characterised in the case of the phthalocyanines, and I would therefore suggest that application of the excited-state concepts of the band model should be limited to covalent semiconductors. W.Arden, M. Kotani and L. M. Peter, Chem. Phys. Letters, 1976, 40, 32. W. Arden, M. Kotani and L. M. Peter, Phys. Stat. Sulidi, 1976, 75, 621. Dr. W. M. Ayers (New Jersey) said: Copper phthalocyanine (CuPc) is fundament- ally different from the weakly intereacting molecular crystals that you mention such as tetracene. The C U + ~ in the solid, is hexacoordinated to four nitrogens of the parent Pc molecule and two nitrogens of nearest neighbours above and below the parent molecule. Details of this intermolecular coordination and its role in CuPc semiconductivity can be found in Harrison and Assours’ e.s.r. study of the material. It is necessary to separate the question of applicability of band theory from the question of mechanism of free-carrier generation. Excitons (> 1 eV binding energy), and the formation of free carriers from excitons, occur in many indirect-gap materials, e.g., GaP,2*3 to which band theory can certainly be applied.I agree that the mechanism of free-carrier formation in phthalocyanines and the role of contacts needs further study. However, the purpose of the observations in the introduction is to support application of band theory to CuPc. Further support for this viewpoint can be found in ref. (4)-(6). S. E. Harrison and J. M. Assour, J. Chem. Phys., 1963, 40, 365. J. I. Pankove, Optical Processes in Semiconductors (Dover Publications, New York, 1975). R. A. Smith, Seniiconductors (Cambridge University Press, Cambridge, 1978). G. H. Heilmeier and G. Warfield, J. Appl. Phys., 1963, 34, 2278. H. Hochst, A.Goldrnan, S. Hufner and H. Malttr, Phys. Stat. Solidi b, 1976, 76, 559. ‘ R. 0. Loutfy and Y. C. Cheng, J. Chem. Phys., 1980,73,2902. Prof. 2-H. Liu (Peking) (communicated): In addition to Dr. Ayers’ result, I give some experimental details of our own work concerning porphyrin electrode photo- voltaic cells. (1) The solution side of the electrode often provides significant phenomena in the course of light conversion to electricity. For instance, in our semi-transparent meso- tetraphenylporphine (TPP) electrode photovoltaic cell where the electrode was pre- pared by subliming TPP onto Nesa glass, it was found that by adding some surfac- tant to the solution of the redox electrolytes Fe3+/Fe2+, the short-circuit photocurrent282 GENERAL DISCUSSION Isc is greatly enhanced, as shown in table 3, although the surfactant itself does not take part in the redox reaction.The effect of the adsorption of the surfactant is a shift of the charge from the diffuse layer to the Stern layer, the SOS being adsorbed in what might still be called the inner Helmholtz layer.2 Hence there is a cooperation action of SOS- and Fe3+ ions to facilitate the reduction of Fe3+ ions to Fe2+ ions which then migrate to the opposite electrode. (2) The coating of organic dyes onto Nesa glass forming a semi-transparent elec- trodel provides a means of conquering the drawback of the large electrical resistance of the dye layer, which limits the thickness that can be a p ~ l i e d , ~ because we can con- struct a multilayer photovoltaic cell with a stack of thin-layer semitransparent elec- trodes. TABLE 3.-EFFECT OF SODIUM OCTYL SULPHATE (sos) at 25 "c ~~ ~ ~~ ~ [SOS]/ mol dm-3 0 1 0 - 7 1 0 - 5 1 4 Isc/pA cm-2 18 25 26 31 19 For instance, we have constructed such a cell by using seven parallel TPP-coated electrodes in a stack and obtained >2 V open-circuit photovoltage while connecting the terminals in series. A 3-electrode assembly, shown in plate 1, delivered nearly 1.4 V. The interesting point is that this assembly seems to be simulating somewhat the utilization of sunlight by green leaves in which piles of semi-transparent thylakoids constitute a grana in the chloroplast, and TPP is similar to chlorophyll in the main skeleton structure. As to the main differences, we may mention that TPP is much more light-fast than are chlorophylls and the layer of TPP may be coated much more thickly than the natural chlorophyll monolayer, thus facilitating light conversion in our photovoltaic cell. In our TPP sublimed layer the microstructure contains threadlike assemblies of TPP molecules having a diameter (3) The sublimed dye coating is not always uniform. TABLE 4.-EFFECT OF THE SUBSTRATE AT 25 "c substrate [SOS]/mol dm-3 V,,,/mV I S h A ~~~ 0 366 81 1 0 - 3 420 118 0 348 121 430 215 SnO ITO" IT0 is a mixture of SnOz and In2O3. of a few hundred A as shown in plate 2. Only instantaneous weak electron diffraction patterns could be seen by using transmission electron microscopic technique, this is due to the destructive effect of the intense electron beam used. The assemblies show that there is a strong coherent interaction between TPP molecules. The microstruc- ture of the TPP layer affects the photoresponse of the photoelectrode appreciably in a complex manner and this will discussed later. The main role of the Sn02 substrate is the separation of electric charge produced by the light-excitation of the dye mole- cules, which prevents the electron from travelling to the connecting wire via the SnO,PLATE 1 .-A three semi-transparent electrode assembly. PLATE 2.-Microstructure of TPP thin layer revealed by electron microscopy (100 000 magnification). [To face page 282GENERAL DISCUSSION 283 layer due to its semiconductor characteristics. In fact, it delivers electrons to com- bine with TPP+, a product of the excited TPP molecule. The effect of the nature of the substrate is shown in table 4. (4) The recombination character of the dye layer. In the case of a photoconductor the relationship between the short-circuit photocurrent I,, and the intensity of light F often fits the equation I,, = Fy where yis a characteristic of the system having a value between 0.5 and 1 in many systems, Along the same line, for our TPP photoelectrode we obtain y = 0.6 which is the slope of the straight line shown in fig. 14. This value differs slightly from that 1 .oo 1.50 log F 2.00 FIG. 14.-Photocurrent dependence on light intensity. obtained by Dr. Ayers, whose value according to the above equation will also be 0.5. Our value lies nearer to 0.5 than to 1, and therefore, according to the simplified light- excited kinetics of the photoelectrode, the recombination of charge carriers in the TPP thin layer plays an important role in the light-conversion process. Z. H. Liu, 2. C. Bi, D. S. Cheng, Y. N. Zhu, Y. S. Li and H. Ti Tien, Kexue Tongbao, 1979,24, 1027. R. H. Ottewill and M. C. Rastogi, Trans. Furaday SOC., 1960, 56,886,880. H. Gerischer, in Solar Power and Fuels (Academic Press, London, 1977), p. 103.
ISSN:0301-7249
DOI:10.1039/DC9807000255
出版商:RSC
年代:1980
数据来源: RSC
|
20. |
Sensitisation of semiconducting electrodes with ruthenium-based dyes |
|
Faraday Discussions of the Chemical Society,
Volume 70,
Issue 1,
1980,
Page 285-298
Martin P. Dare-Edwards,
Preview
|
PDF (1031KB)
|
|
摘要:
Sensitisation of Semiconducting Electrodes with Ruthenium-based Dyes B Y MARTIN P. DARE-EDWARDS, JOHN B. GOODENOUGH, * ANDREW HAMNETT, KENNETH R. SEDDON AND RAYMOND D. WRIGHT Inorganic Chemistry Laboratory, South Parks Road, Oxford OX1 3QR Received 8th May, 1980 Practical photoelectrolysis of water by sunlight requires the development of suitable semiconductor electrodes. One approach is the photosensitisation of wide-bandgap oxide semiconductors by the chemical attachment of a suitable dye to the surface. We report on the photosensitisation of n-TiOz, n-SrTi03 and n-Sn02 with chemically attached Ru(bipy)2(bpca) in which the 2,2’-bipyridine-4,4’ dicarboxylic acid (bpca) is chemically attached to the semiconductor surface by two ester linkages. Kinetic analysis reveals a surprisingly low quantum efficiency (qe x 0.25%) for electron injection from the photoexcited dye into the semiconductor and a low rate constant for reoxidation of the dye.A suggestion is made how the dye molecule might be designed to improve the situation radically. Some deterioration in performance was observed after illumination of the sensitised electrode for many hours. Large-scale, non-agricultural utilization of solar energy in Northern Europe prob- ably awaits an economical method of converting sunlight into chemical energy, chemi- cal energy being the most versatile form of stored energy. Hydrogen is a clean fuel as well as an important chemical feedstock; it is now ob- tained largely from fossil fuels. Two options are available for the conversion of sun- light to hydrogen: (a) use of photovoltaic cells to power the electrolysis of water and (b) direct photoelectrolysis of water.This paper is concerned with the evaluation of one technique for realizing the latter process. In a photoelectrolysis cell, the photovoltaic component is incorporated into one or both electrodes of an electrolysis cell. Photoactive electrodes are semiconductors ; electron-hole pairs photogenerated in the semiconductor near the electrode/electrolyte interface are separated by a depletion-layer field in the semiconductor as illustrated in fig. 1. The Fermi energies in the two electrodes may be shifted with respect to each other by an applied voltage, V. If only the oxygen electrode is photoactive, fig. l(a), the Fermi energy of the cell lies at or above the H+/H2 level, E,,(H+/H,), of the aqueous electrolyte during hydro- gen evolution.To obtain a depletion layer of sufficient width in the semiconducting oxygen electrode, the semiconductor must not be too heavily doped n-type; the Fermi energy of the disconnected electrode is adjusted to lie 0.3-0.4 eV below the conduction- band edge E,. Therefore, a depletion layer of sufficient field strength (aE,/ax, where x is the distance from the surface) for efficient electron-hole separation requires a conduction-band edge at the surface, E,”, at least 0.7 eV above E,(H+/H,). The energy difference Ez - E,(H+/H,) is insensitive to pH because E,” depends upon the net sur- face charge, and hence on the surface protonation of the semiconductor. Finally, chemical stability for an oxygen electrode implies use of an oxide semiconductor In operation a Y = 0 is desired.286 SENSITISATION OF SEMICONDUCTORS ( 0 ) t ANODE ’ ELECTROLYTE ’ CATHODE FIG.1 .-Energy diagram for two photoelectrolysis cells: (a) a semiconductor photoanode and metallic cathode, (6) two semiconductor electrodes. containing immobile cations. However, the 02- 2p6 bonding states of an oxide repre- sent too stable a valence band (VB). The VB edge at the surface, E,”, lies sufficiently far below the energy EO(O2/H20) to ensure that photogenerated holes do not build up in bonding states at the surface; but the energy Eo(02/H20) - E,” represents a loss to be minimised. With only an active oxygen electrode, fig. l(a), the constraints on E; make the energy gap Eg = (E, - Ev) > 3.0 eV if the VB is 0’- 2p6; this is too large an energy for the conversion of most of the sunlight.(The solar spectrum peaks at ca. 2.5 eV.) A recognized strategy for photosensitising such a large-bandgap semi- conductor to visible light is the use of a suitable dye attached to the surface of the semiconductor. It is this strategy that is investigated in this paper. Fig. 2 illustrates the energy-level scheme for an oxygen electrode photosensitised by a dye molecule D having a ground-state energy ED, a ground-state energy ED+ for the oxidized state D+, and excited-state energies Eg’ and Ez corresponding to the unrelaxed and relaxed excited states D*’ and D*. With this configuration, sub- bandgap radiation of energy Av = (ED’ - ED) .: E, excites the dye, and separation of photogenerated electron-hole pairs requires injection of the excited electron to the semiconductor conduction band (CB) and capture by D+ of an electron from the electrolyte to evolve 0, and return the dye to D.The energy losses in this scheme are Ez‘ - E0(H+/H2) and Eo(Oz/H20) - ED. The energy difference EE - Eo(H+/H,) > 0.7 eV required for an adequate depletion-layer field is a built-in loss. In the more likely event that excited electrons are injected into the CB from D* rather than D*’, the energy difference A* = (ED’ - ED) becomes an additional loss that must be minimised. Since electron capture by D+ occurs with the dye at the energy ED+, the energy loss A = (E; - ED) must also be minimised. Unless we can find (A + A*) < [E,(O,/H,O) - E,” - E ~ ] , where co is the energy loss for oxygen evolution at an oxygenDARE-EDWARDS, GOODENOUGH, HAMNETT, SEDDON, WRIGHT 287 ELE CT RO LY T E FIG.2 .-Ener gy-level scheme for d ye-semi t isa tion. electrode, use of a dye becomes a failed strategy. In addition, separation of the photogenerated electron-hole pair must compete against luminescent decay of the dye back to its ground state as well as the recapture of a CB electron (probably via a sur- face state). In order to test the feasibility of this sckLeme, we make the optimistic assumption that multiple reflections at a rough surface will permit capture of most of the light by a monolayer of dye having a large extinction coefficient. With this assumption, the critical parameters to be measured are (a) the extinction coefficient aA of the chemically attached dye D, (b) the quantum efficiency re of electron injection into the semicon- ductor per photon absorbed by the dye, (c) the sensitivity of qe to the relative positions of Eg', E& and E& (d) the rate of water oxidation by electrolyte-electron capture by D +, (e) the lifetime z of the dye in its excited state before luminescence occurs, cf) the rate of CB-electron recapture by D+, ( g ) the changes in the action spectrum of the dye due to chemical attachment, (h) the long-term stability of the dye, and (i) the extent to which all these parameters can be changed relative to one another by exchange of the dye ligands and/or the semiconductor.This paper presents a partial evaluation of a single-dye system.EXPERIMENTAL Ti02 and SrTiOJ were obtained as single-crystal slices cut to expose the (001) and (100) faces, respectively. They were reduced in flowing hydrogen at 600 and 1100 "C, respectively, for 2 h, polished with 6 pm diamond paste and etched as reported previously.' The carrier densities were 1018-1019 ~ m - ~ . Films of Sn02 were prepared as described by Gleria and Memming ; they included 3 % antimony. Solutions were 0.05 mol dm-j H2S04 of AnalaR or AristaR quality made up with de- ionised, doubly distilled water. Ru(bipy),(bpca) was prepared and its chemical attachment effected in two ways as reported by Anderson et aL3 The electrode assembly and electrochemical equipment have been described el~ewhere.~ THE DYE The oxidised complex D+ = [Ru(bipy)$ +, where bipy represents 2,2'-bipyridine, is known to oxidise water; and the excited state D* of the reduced complex D = [R~(bipy)~]~+ is known to be a triplet state with a long lifetime z.In solution, the absorption spectrum of D peaks at a wavelength 1 = 450 nm, corresponding to a photon energy kv = (E%' - ED) x 2.7 eV.5 The most extensively studied oxygen electrodes for photoelectrolysis are the oxide semi-288 SENSITISATION OF SEMICONDUCTORS conductors n-TiO, (Eg = 3.0 eV), n-SrTi03 (Eg = 3.2 eV), and n-SnO, (Fg = 3.5 eV); these have an E,“ - (H+/H2) of 0.14, 0.4 and -0.1 eV, respectively. Most investigations have, like this one, been confined to Schottky barrier electrodes of the type illustrated in fig. 1 and 2. Mott barrier electrodes are also possible; in these, the bulk semiconductor may be heavily doped because the depletion layer is replaced by an intrinsic region across which the voltage drops.A recent investigation6 of n-TiO, having a Mott barrier indicated that the density of surface states in the energy gap is sharply reduced at an intrinsic layer. Although detri- mental to electrode performance in the absence of a dye, such a feature should, on the con- trary, be advantageous to a chemically modified electrode utilising an attached dye. In this study, we used n-TiO,, n-SrTi03 and n-SnO, electrodes having a Schottky barrier at the electrode/electrolytic interface. Clark and Sutin’ were the first to investigate the use of D = [R~(bipy)~]~+ for the photo- sensitisation of n-Ti02. They used [Ru(bipy),]C12 in acidic aqueous solution.Photosen- sitisation by a D in solution requires that the lifetime T of the excited state D* be sufficiently long that the majority of excited complexes are within a diffusion length (DT)* of the electrode surface, D being the chemical diffusion coefficient. Nevertheless, they reported an anodic sensitisation (nett photocurrent corresponding to electron transfer from electrolyte to elec- trode via D*) to chopped light. However, Hamnett et af.4 demonstrated this to be a purely transient phenomenon at potentials near flatband (Vfb), the applied potential where E, =Ez throughout the electrode. On making the potential Vmore cathodic, a sizeable dark current sets in as Vapproaches Vfb, and photoexcitation of surface-state electrons to the CB modu- lates this dark current so as to produce a nett cathodic d.c.photocurrent. Only at potentials (V - Vfb) > 0.5 V could any anodic d.c. sensitisation be seen. Nevertheless, Clark and Sutin’s experiment was of great importance as it demonstrated that in acidic solution D = [R~(bipy)~]~+ has E,*’ > ED” > EE(Ti02), A* z 0.6 eV, and an EZ sufficiently below Eo (02/H20) to oxidise water. The requirement (LIZ)* may be relaxed to k,t > 1, where k, is the rate constant for injection of an electron from D* to the CB, if D is either adsorbed or chemically attached to the electrode surface. The simplest experiment is to adsorb a water-insoluble dye on the surface. Significant photosensitisation currents have been observed for phthalocyanine dyes on n-Ti02* and for [R~(bipy)~(AQ)Ru(bipy)~]~ + where AQ is 1,5-dihydroxy-anthroquinone on glassy ~ a r b o n .~ Memming and Schroppel lo found that monolayers of several surfactant analogues of [R~(bipy)~]~+ deposited on n-Sn02 electrodes gave, at A = 488 nm, quantum efficiencies qq = 15% at pH = 9. Their Ru complexes, if oxidised by flash photolysis, were reduced by water nearly six orders of magnitude faster than the [R~(bipy)~]~+ complexes pro- duced chemically by Creutz and Sutin.” This surprising result remains unexplained. In this paper, we report on the photosensitisation of n-TiO,, n-SrTi03 and n-SnO, by a chemically attached derivative of [Ru(bi~y)~]~ + . The electrode surface was derivatised with [R~(bipy)~(bpca)](bpcaH, = 2,2’-bipyridine-4,4’-dicarboxylic acid) by two different chemical routes, both involving reaction with electrode-surface sites that react like hydroxyl groups : the surface was either reacted with [R~(bipy)~(bpca)] in the presence of dicyclohexylcarbodi- imide in a one-step process or with 2,2’-bipyridine-4,4‘-di(carbonylchloride), followed by cis-[Ru(bipy),C12], in a two-step process; Ti Ti I I 0 =ODARE-EDWARDS, GOODENOUGH, HAMNETT, SEDDON, WRIGHT 289 These chemically attached dyes gave long-term photosensitisation indicative of somewhat more rapid oxidation kinetics than the low value (k E s-’) found by Creutz and Sutin for chemically oxidised dyes in solution.” Chemical attachment substantially alters the energy levels of the dye as well as those of the semiconductor surface states.Therefore, the action spectra obtained for chemically attached molecules may not correspond so closely to the absorption spectra of the isolated dyes in solution as do those of adsorbed dyes. In fact, significant differences were observed in the action spectra of dyes attached to Sn02 as compared with TiOz and SrTi03.3 These differences are explored here in more detail. For a single-crystal surface with a monolayer coverage of a dye having an extinction coefficient ct x lof7 cm2 mo1-’, the maximum photocurrent generated by an incident-light intensity of 10- * ein cm-2 s-I is ca. 1 p A cm-2. In fact, we observed much smaller photocurrents because of an unexpectedly poor quantum yield. Finally, the photocurrents measured in this work are small. RESULTS AND DISCUSSION Regardless of the dye used, the following sets of reactions must be considered in 1.Absorption of light by the dye followed by non-radiative decay to the long-lived any analysis of dye sensitisation. state D*, 2. Decay of D* to D by luminescence or other parasitic processes, D*% D 3. Injection of an electron from D* into the CB ke D* --+ D+ $- eGB 4. Oxidation of a species R in solution by D+, D+ + ~ k o ’ D + R+ 5. Recapture of an electron from the CB by D+, kr D + + e,;--+ D. (3) (4) (5) The combined reaction (1) can be used provided k,, is much greater than any of the other rate constants; it is then characterised by the extinction coefficient CQ. The assumption of a large k,, can be tested by measuring the change in electron-injection quantum efficiency qe for Es’ > E$ > E,” as compared with ED*’ > E,” > EZ.These reactions have been investigated for Ru(bipy),(bpca) attached as previously described3 to single-crystal n-TiO, and n-SrTiO, as well as to transparent films of polycrystalline n-Sn02. The two-step dye attachment gave essentially the same photochemical results as the single-step preparation. However, if the two-step pro- cess was interrupted after only the bpca ligand had been attached to the surface, no photosensitisation was observed. Direct evidence for ruthenium attachment to the surface in the photosensitised samples has been provided by a-back-scattering spectro- metry, which suggests a coverage close to 100% of a monolayer.12 Other techniques attempted failed to provide conclusive evidence for Ru attachment : surface contami- nation vitiated X.P.S.data, and the anticipated [Ru(bipy)J3+/*+ couple could not290 SENSITISATION OF SEMICONDUCTORS be detected on n-TiO, by a.c. cyclic voltammetry, although it was found on analo- gously treated n-SnO, using this technique.', Reflectance spectra have not yet been obtained for the single crystals, but chemically modified surfaces of finely powdered hydrous TiO, gave spectra markedly different from a simple mixture of dye and oxide.14 The experiments performed included: 1. Measurement of the action spectra for inci- dent light in the sub-bandgap region to characterise changes in the dye characteristics due to chemical bonding to the surface. 2. A study of photocurrent transients to determine the various rate constants appearing in reactions (1)-(5) in both the presence and absence of dye supersensitisers.3. Comparison of the quantum efficiencies determined for different semiconductors and pH to obtain some indication of how the energy levels of a chemically attached dye move with pH changes relative to 15," and E0(H+/H2) and of the reliability of some assumptions made in the analysis. 1 . ACTION SPECTRA The action spectra for [Ru(bipy),(bpca)] chemically attached to the three electrodes are shown in fig. 3. Unlike the corresponding spectra obtained by solution sensitisa- tion7 or by monolayer adsorption using the Langmuir-Blodgett surfactant technique," chemical attachment has caused a distinct broadening of the spectral features. In- deed, very little structure is evident from the chemical modification of n-Ti02 in the absence of a supersensitiser.If 0.5 mmol dm-, 1,4-dihydroxybenzene is added as a supersensitiser to the electrolyte to increase ko of reaction (4), the net anodic photo- current is enhanced and begins to show some structure [fig. 3(a)]. Somewhat en- hanced structure is evident for n-SrTiO,, and relatively sharp peaks can be seen in the action spectrum associated with n-Sn0,. The observations can be understood qualitatively with the aid of fig. 4. In the case of Ti02 and SrTiO,, the CB is a n* band derived from Ti 3dorbitals of f2* paren- tage; these overlap strongly with the TC and n* orbitals of the complex system. The 5s CB of SnO,, on the other hand, is orthogonal to the TC and n* systems of the dye, which minimizes perturbation of the D and D*' states by dye attachment.The exist- ence of a well-defined structure in the sub-bandgap action spectrum of modified n-Sn02 is clear evidence that electrons are not photoexcited from D directly into the conduc- tion band, which is consistent with a large k,, in reaction (1). The situation is not so straightforward for n-TiO, and n-SrTiO, where chemical interactions at the surface create semiconductor/dye n* states having a surface density of states that overlaps the CB edge E,". Two types of sub-bandgap optical transitions to these states are possible : from the VB at the surface or from the dye ground state D. The first is expected to produce a long tail on the long-wavelength side of the absorption edge; the second may give rise to a broad peak centred approximately at the wavelength of the absorption peak or, for stronger surface interactions, may show no structure at all.For n-TiO,, with a bandgap of 3.0 eV sensitised with a dye whose absorption peak lies at 2.7 eV, the action spectrum without supersensitiser shows little structure, as can be seen in fig. 3(a). The dye-excitation contribution may, however, be considerably en- hanced by adding 1,4-dihydroxybenzene, and in this case a broad shoulder becomes evi- dent at about 490 nm. For n-SrTiO,, with a slightly larger bandgap (3.2 eV), the supersensitised action spectrum [fig. 3(b)] now shows the shoulder resolved into a very broad peak, again centred at ca. 490 nm. This evidence of structure, however broad- ened, is consistent with the assumption of a large k,, in reaction (1).The action spectrum observed will be the superposition of these two effects.DARE-EDWARDS, GOODENOUGH, HAMNETT, SEDDON, WRIGHT 291 3 .O 0 0 OOOO oooo 0 0 OOO 0 0 OOOnn 2 . 0 k * I 6 0 .. -. -. O'?OO L20 LLO L60 L80 500 520 540 560 580 600 wavelength/nm FIG. 3.-Action spectra for dye sensitised electrodes: (a) Ru(bipy),(bpca) attached to Ti02 (@) with- out 1,Cdihydroxy benzene in solution, (0) with 0.5 mmol dm-3 1,4-dihydroxybenzene in solution; (b) Ru(bipy),(bpca) attached to SrTi03 with 0.5 mmol dm-3 1,4-dihydroxybenzene in solution; (c) Ru(bipy),(bpca) attached to Sn02 without 1,4-dihydroxybenezene in solution. All spectra were taken in 0.05 mol dm-3 H2S04 and the electrodes were held at + 1.0 V (SSE).FIG. 4.-Nature of the bonding of Ru(bipy),(bpca) to Ti02 and Sn02.292 SENSITISATION OF SEMICONDUCTORS To test whether these effects were dye-specific, similar experiments were carried out on n-TiO, with chemically attached Rhodamine-B. The action spectrum was quite featureless, showing no trace of the solution absorption peak, whereas attach- ment to SnO, has been reported5 to give an action spectrum showing the essential absorption profile of Rhodamine-B in solution. The clear distinction between a 5s and a n* conduction band is again evident, though the larger bandgap of SnO, also separates sub-bandgap structure due to the dye from band-edge tailing. 2. TRANSIENTS The photocurrent-voltage plot for sub-bandgap radiation (450 < A/ nm < 670) on n-Ti0, without dye attachment exhibits a large anodic peak 200 mV wide centred near Vfb.It is entirely transient and appears to be associated with the excitation of hydro- gen atoms at and just inside the surface.I6 The chemically modified electrode exhibits, in addition to this peak, an anodic a.c. photocurrent extending to the highest anodic potentials measured. At + 1.0 V us. SSE (the standard silver electrode whose potential is -0.040 V us. SCE and to which all potentials are referred in this paper) the photo- current consists essentially of only this second component. Therefore, we chose to study the contribution due to dye attachment by monitoring the photocurrent at +1.0 V (SSE) after cycling to -0.2 V. Cycling to -0.2 V floods the surface with electrons, thereby ensuring that all the dye molecules are in the reduced D state before the light is switched on.Fig. 5 shows the photocurrent response to sub- bandgap I I . . . , I , I , 0 40 80 320 0 40 80 120 time/s FIG. 5.-Transient effects observed for Ru(bipy),(bpca) attached to TiOz : (a) potential held at + 1.0 V (SSE); (6) potential held at +O.O V (SSE). All results were taken in 0.05 mol dm-3 H2S04 without 1,4-dihydroxybenzene in solution. Illumination from a 150 W xenon lamp with high and low pass filters at 450 and 670 nm, respectively. radiation (450 < A/ nm < 670) of single-crystal n-TiO, having D = [Ru(bipy),(bpca) chemically attached to an (001) surface with no supersensitiser in the 0.05 mol dm-3 H,SO, electrolyte and (a) V = + 1.0 V (SSE), (b) Y = $0.0 V (SSE).Both curves exhibit an important transient component and a residual d.c. component. To avoid drift problems, we used light chopped at 80 Hz to measure the magnitude of the photocurrent iph as a function of time t . Fig. 6 displays the data of fig. 5(a) asDARE-EDWARDS, GOODENOUGH, HAMNETT, SEDDON, WRIGHT 293 log[ip, - i;] against t , where i& is the limiting value of iph after a long irradiation time. The curve fits a single exponential decay for t > 40 s. Analysis of the data proved straightforward if the following assumptions are made for t > 40s: 1. The surface charge has reached a steady-state equilibrium, and any increase in the fraction 6 ; of dye molecules in the D+ state is compensated by an equivalent loss of surface protons.In addition, any energy transfer from D* to excite a surface-state I 2’ I I I I 1 0 20 40 60 80 100 120 140 time/s FIG. &-Logarithmic plot of the transient part of the observed photocurrent against time for Ru(bipy), (bpca) attached to Ti02. The electrode was held at +1.0 V in 0.05 mol dm-3 H,SO, without I ,Cdihydroxybenzene in solution. Illumination as for fig. 5. electron to the CB (Forster quenching) is followed by electron tunnelling from D to the surface-state hole created, the net process being equivalent to a contribution to reaction (3). It then follows from reactions (1)-(5) that where 6, and 6; are the fractions of dye molecules in states D and D*, ka zz 2.303~lI/N (8) contains the decadic absorption coefficient ul/cm2 mol-I for light of wavelength J.incident with intensity Ilphoton cm-2 s-’, N is Avogadro’s number, and n, z nbexp- (-pJv,l) is the effective surface-electron density for a CB-electron density nb and a normalised potential u, = (V - Vfb)e,/kT modulated by a constant p. This implies that the excited-state fraction 0; is much smaller than either of the two ground-state fractions &, 6&, so 2. In eqn ( 6 ) and (7) k, and k, % ka, k,, krns. e, + o,+ z 1. (9) This will not apply if the light intensity becomes very large.294 SENSITISATION OF SEMICONDUCTORS 3. If 8: < OD and 02, it is reasonable to suppose that This is, in effect, the steady-state appro~imation,'~ and we have, from relation (6) 0: x kaqeeD/ke (1 1) re kel(kd + ke). (12) krn,8D+ z 0. (13) where we define the quantum efficiency for electron injection from D* to the CB to be 4.The applied potential Y is sufficiently positive at + I .O V (SSE) that With these assumptions, it follows from eqn (7), (10) and (12) that aO',/at x kaqe(l - 8,') - k,O,' iph(t > 40 S ) w eoNTkaqe(l - 0;) % eol\r,(kaqe/ki)[ko + kaqeexp(-kit)l. (16) (1 7) (1 8) ki kave + k, = 0.018 S-'. (19) This result is consistent with fig. 6, giving ln[iph(t > 40 s) - iih] z const - kit iih x eoNdkg velki)ko and from the slope of fig. 6 we obtain For an I x 1017 photon cm-2 s-' and an aA x 1.5 x lo7 cm2 mol-', we obtain a value of ka x 6 s-'. If a 1,4-dihydroxybenzene (Q) supersensitiser is added to the electrolyte, the tran- sient effects are found essentially to disappear. Since the supersensitiser should in- crease the rate of reaction (4) without strongly influencing kaqe, we conclude that ko' $ (ko + kaqe), where k: refers to k, in the presence of the supersensitiser.From eqn (18) and from the measured photocurrents in the range 450 < ;I/ nm < 670 with and without supersensitiser, we obtain the ratio iih(Q)/ipOh X (kaqe + ko)/ko 7 (20) (21) which, with eqn (18), gives k, w 2.6 x S-' and kaqe s 0.015 s-'. Since ka z 6 S-', it follows that ?,Ie W 0.0025 and kd S 400ke. From eqn (18) we can also define a quantum efficiency for conversion of absorbed light into oxygen as qs = i i /ii,(max) x vek,/(kaqe + k,) x 4 x (23)DARE-EDWARDS, GOODENOUGH, HAMNETT, SEDDON, WRIGHT 295 where igh(max) = eoNTka is the maximum possible photocurrent corresponding to re = 1 and k, > ka.In the presence of the supersensitiser, eqn (23) reduces to qq N" re and provides an independent estimate of the dye coverage NT E igh/eokare % 6 x 1013 cm-2, which is close to a monolayer in agreement with the cc-back-scattering experiment. Unless k, --f ka, qq will remain low even for a large qe, and an estimated k0/ka of 4 x Drastic increases in both k, and qe appear to be required. The value of the rate constant k, for the oxidation of water calculated above is 2.6 x s-' at pH 1. This is to be compared with a value of 1.41 x s-' (pH 3 and 4.8) measured for the rate of reduction of chemically produced [Ru(bipy),13+ by water'' and a value of 9 x lo2 s-l (pH 1) measured for the rate of reduction of photo- chemically produced [Ru(bipy),13+ .18 Our own rate constant clearly lies much closer to that of Creutz and Sutin'l even allowing for the possible change in k, with pH. It can be seen from the above analysis that a rate constant of 9 x lo2 s-l would eliminate any transient effects.If we make the applied potential less anodic, our assumption (4) no longer holds, and the net photocurrent is the algebraic sum of two terms: is discouragingly small. iph = e,NTkeBg - eoNTkrnsBD+. (24) ki -+ ki kaqe + k, + krns (25) iph eoNT(&aqe/ki)[ko + (kaqe + krns)exp(-git)]- (26) Under steady-state conditions, we find that in eqn (14) and, from eqn (10) and (1 2) Comparison with eqn (16) shows that both iih and the time for iph to decay to l/e its initial value are reduced by the factor ki/ki. This effect can be seen by comparing fig.5(a) and (b). Moreover, when the light is switched off, the photocurrent becomes which predicts the fast-decaying negative transient also evident in fig. 5(b). These results demonstrate the need for an adequate depletion-layer electric field and hence an I?: - E,(H+/H2) > 0.7 eV. 3. OTHER SEMICONDUCTORS We have seen that the transient component of the photocurrent iph(t > 40 s) at V = +1.0 V (SSE) for Ru(bipy),(bpca) attached to a (001) face of single-crystal n-TiO,, fig. 5(a), can be eliminated by the addition of a 1,4-dihydroxybenzene super- sensitiser to the solution. We have also identified this transient with the establishment of a steady-state 0; value under the steady-state approximation (%);/at) z 0. Its elimination is associated with the much larger value of k, in the presence of the super- sensitiser.An alternate method of increasing k, is to raise the Eo(02/H20) level rela- tive to the energy ED+. Whereas the band edges E: and Ez of the semiconductor remain fixed relative to E0(H+/H2) and Eo(02/H20) with changing pH, due to adjustments in surface protonation, the energy levels of a chemically attached dye need not do so. The energy levels of the dye in solution show a less marked change with pH;19s20 to what extent those of a chemically attached dye will change depends on the local electrical field and hence on the dye's location relative to the Helmholtz layer at the solid/liquid interface. Since most of the local field at a semiconductor/liquid interface is across the depletion layer, we may anticipate that the energy E0(02/H20) - E+,296 SENSITISATION OF SEMICONDUCTORS increases with pH.Experiments in alkaline solution have shown that k, becomes large enough to eliminate transients, but the quantum efficiency ve is reduced even further. Reduction in re could be anticipated from the work of Clark and Sutin,’ which located E& just above E,” for n-TiO, in acid solution, but just below E: in alka- line solution. Since k, varies sensitively with the density of CB states at the energy E:, a change to alkaline solution reduces re even if the shift in E$ relative to E,” is not so great for a chemically attached dye as for a dye in solution. These results support our assumption (1) that k,, is large since Eg’ is sufficiently far above E,” that the efficiency of injection from this level would not be changed significantly. Use of another semiconductor allows examination of the change in E,” com- pared with E: by some other method than pH change.It also permits examination of changes in qe with a change from a n* to an s-like CB. Chemical attachment of Ru(bipy),(bpca) gives rise to a sub-bandgap photocurrent density about half as large as that observed for n-Ti02. This difference is almost certainly due to a decreased v,, since all other factors (except NT) should be unchanged. It is consistent with the fact that E: lies about 0.25 eV higher in SrTi03 than in TiO,. Support for this view comes from the observation by Clark and Sutin that solution-sensitisation of TiO, by [Ru(bipy)J2+ was reduced by the same factor if E,, was raised 0.2 eV by altering the pH.One difference between SrTiO,, and Ti0, is the retention in SrTiOJ of a small initial transient in the presence of a 1,4-dihydroxybenzene supersensitiser. The origin of this feature is not known. The action spectra of n-SnO, sensitised by either Rhodamine-B or Ru(bipy),(bpca) show sub-bandgap peaks closely resembling those of the isolated dyes, a reflection of decreased bonding of the molecular system to the semiconductor (see above). How- ever, the unmodified film of SnO, made by the spray technique always showed, unlike the single-crystal studies, substantial sub-bandgap photocurrents, 21 and the action spectra of the chemically modified films are a superposition of these and the dye- injection currents.By using filters to select light of wavelengths lying between 450 and 670 nm, most of the photocurrent observed for a film modified with Ru(bipy),- (bpca) may be ascribed to the dye. Interestingly, addition of a supersensitiser caused a much smaller effect in the case of modified SnO, than TiOz, a factor of only 2 being observed in the increase in photocurrent as compared to 7 for TiO,. A kinetic analy- sis, similar to that carried out for TiO,, gave k, “N 0.1 s-l and re z 0.02. This efficiency is much larger than that for TiOz, and it is difficult to understand the increase on the basis of the relative positioning of the conduction-band edge in SnO, unless, as already indicated by the action spectra of fig. 3, chemical attachment creates semi- conductor/dye n* states having a surface density of states that overlaps the CB edge E,, in TiO,, but not in SnO,.Electrons trapped in these surface states at energies below E,” would have a high probability of recombining with the D+ hole to recreate D. Preliminary absorption spectra of free and chemically attached dyes indicate that the 7c* orbitals of the (bpca) ligand lie at a lower energy than those of the (bipy) ligands,14 so the low value of re cannot be attributed to a preference of the D* state to have the excited electron in the solution-oriented (bipy) rather than the solid- attached (bpca) n* orbitals. The unacceptably large value of kd would therefore appear to reflect electron trapping associated with the method of chemical attachment of the dye.Direct attachment of the ruthenium to the oxide surface layer should eliminate this problem. If this proves to be the case, we should then be able to deter- mine the extent of any concentration quenching due to interactions between dye mole- cules of the chemisorbed layer. The behaviour of n-SrTi03 is closely analogous to that of n-TiO,.DARE-EDWARDS, GOODENOUGH, HAMNETT, SEDDON, WRIGHT 297 Finally, the very long-term stability of chemically attached Ru(bipy) (bpca) was investigated with high-intensity illumination applied over a period of many hours, A slow deterioration in performance was observed; it was quite irreversible, and the calculated quantum efficiency for this decay is ca. The mechanism is being actively investigated and may be related to that discussed by Abrufia et a1.22 CONCLUSIONS Kinetic analysis of the d.c.photocurrents obtained from n-type semiconducting oxides modified by the chemical attachment of [Ru(bipy),]'+-based dyes to the surface gives information on the critical rate constants and quantum efficiencies operative when these systems are used as oxygen electrodes in a photoelectrolysis cell. Comparison of our results with those of Memming and Schroppello for adsorbed dyes on n-SnO, suggests that the parameters qe and k, can be significantly improved by a different choice of ligands for the chemically attached dyes. Recent experiments indicate that the spectral properties of dyes of the form Ru(Ll),L2, where L1 and L2 are different derivatives of bipyridine, are approximately a weighted superposition of the spectra of R U ( L ~ ) ~ and Ru(L,),. It may therefore prove possible to modify L2, the ligand by which the dye is attached to the electrode, to increase qe and simultaneously to adjust L, to increase k,.On the other hand, even with ED* close to E," and ED too close to E,(O,/H,O) for a large ko, the absorption peak for the dye occurs at hv = E$' - ED z 2.7 eV, which is unacceptably large. Clearly a significant reduction in the energy losses E3' - E* and ED - ED is also required. Drastic improvement of qe may permit injection of electrons into the solid from D*', a condition that would eliminate the energy loss This ED*'- E$. Finally, the stability of the chemically attached layer needs to be improved. should prove possible once the reason for the deterioration is known. We are indebted to Dr. H. J. Scheele of IBM, Zurich, for the gift of single-crystal M.P. D.E. and R. D. W. thank the S.R.C. for student- samples of TiOz and SrTiO,. ships and R. D. W. and K. R. s. the C.E.G.B. for financial support. M. P. Dare-Edwards and A. Hamnett, J . Electroanalyt. Chem., 1979, 105, 283. M. Gleria and R. Memming, Z . Phys. Chenz. (N.F.), 1975, 98, 303. S. Anderson, E. C. Constable, M. P. Dare-Edwards, J. B. Goodenough, A. Hamnett, K. R. Seddon and R. D. Wright, Nature, 1979,280, 571. A. Hamnett, M. P. Dare-Edwards, R. D. Wright, K R. Seddon and J . B. Goodenough, J . PhJv. Chem., 1979, 83, 3280. F. Felix, J. Ferguson, H. U. Gudel and A. Ludi, C'hem. Phys. Letters, 1979, 62, 153. H. Morisaki, I. Baba and K. Yazana, Phys. Rev., 1980, 21B, 837. ' W. D. K. Clark and N. Sutin, J . Anier. Chenz. Soc., 1977, 99, 4676. F.-R. F. Fan and A. J. Bard, J. Amer. Chem. SOC., 1979, 101, 6139. K. Pool and R. P. Buck, J . Electroanalyt. Chem., 1979, 95, 241. l o R. Memming and F. Schroppel, Chem. Phys. Letters, 1979, 62, 207. c'. Creutz and N. Sutin, Proc. Nat. .4cad. Sci., 1975, 72, 2858. l2 C. G. Wilson, J. Day and K . R. Seddon, unpublished research. l 3 P. E. Pickup, H. A. 0. Hill and K. R. Seddon, unpublished observations. l4 J. Day and K. R. Seddon, unpublished observations. l 5 T. Osa and M. Fujihira, Nature, 1976, 264, 349. I 6 L. A. Harris, M. E. Gerstner and R. H. Wilson, J . Electrocherti. Soc., 1979, 126, 844, 850. P. W. Atkins, Physical Chwzistry (Oxford University Press, Oxford, 1 978). R. Memming, F. Schroppel and U. Bringman, J. Electroarzalyt. Chrrn., 1979, 100, 307.298 SENSITISATION OF SEMICONDUCTORS l9 G. Clantelli and F. Pantani, Ricerca sci., 1968, 38, 706. 2o F. P. Dwyer, J . Proc. Roy. SOC. (N.S. W.), 1949, 83, 134. 21 H. Kim and H. A. Laitinen, J. Electrochem. SOC., 1975, 122, 53. 22 H. D. Abrufia, T. J. Meyer and R. W. Murray, Inorg. Chem., 1979, 18, 3233.
ISSN:0301-7249
DOI:10.1039/DC9807000285
出版商:RSC
年代:1980
数据来源: RSC
|
|