Pulsed Laser-induced Photoelectrochemistry at Polycrystalline and Single-crystal Semiconductor Electrodes BY SUZANNE B. DEUTSCHER AND JEFFERY H. RICHARDSON Chemistry Department, Lawrence Livermore Laboratory, Box 808 L-325, Livermore, California 94550, U.S.A. AND SAM P . PERONE, JEFF ROSENTHAL AND JAMES ZIEMER Department of Chemistry, Purdue University, W. Lafayette, Indiana 47907, U.S.A. Receiced 6th May, 1980 Our studies have focused on the production and detection of transient phenomena related to u.v.-visible photoelectrolysis at various types of semiconductor electrodes. For this work we have utilized xenon flash-lamp and pulsed dye-laser sources. We are reporting on two significant aspects of this work : a theoretical/experimental description of the time dependence of pulse photocurrents and experimental studies directed at the detection of transient intermediates in photoelectrolysis pro- cesses.Two experimental measurement approaches have been investigated. One of these involves controlled-potential chronoamperometry, utilizing potential steps synchronized with pulsed irradia- tion for detection of transient photoproducts. The other approach involves photocoulostatic poten- tiometry, where open-circuit photovoltage excursions are studied with sub-microsecond time resolu- tion. The former approach allows selective monitoring of various potential regions for qualitative identification of transient photoproducts, whereas the latter approach allows much faster time resolu- tion of photo-induced events. These studies were conducted with single-crystal n-TiO,, p-Gap, p-GaAs and polycrystalline n-Ti0,.Because of the much lower ohmic resistance, the polycrystalline electrodes allowed the best time resolution for controlled-potential experiments. A discussion of the limitations of time-resolved measurements, as well as the nature and significance of transient photo- induced processes is presented. Most photoelectrochemical studies of semiconductor electrode processes have employed continuous or chopped light sources resulting in relatively long time experi- ment~.’-’~ As a prerequisite to a complete understanding of the processes taking place, studies of the transient effects should be conducted. It is the purpose here to demonstrate that flash irradiation can provide a new perspective on semiconductor photoeffects which complements observations made with continuous sources. We have reported previously on the limited time-resolution of potentiostatic photocurrents measured with pulsed At best one can achieve resolu- tion of ca. 500 ns, with ca.50 ps more typical if correction for induced charging cur- rents is made.15 The results reported here for potentiostatic measurements fall in this range. Barker et a1.16 proposed coulostatic-flash measurements to obtain micro- second time-resolution in photoemission studies with mercury electrodes. We recently achieved 200 ns time resolution in coulostatic-flash mercury photoemission studies.” The present work demonstrates that we can observe photopotential tran- sients at semiconductor electrodes with at least 15 ns time resolution.We have analysed the electrical and electronic characteristics of photoelectrochemical cell and measurements circuitry to ensure the validity of short-time measurements; and we have suggested possible interpretations for the observed behaviour.34 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY Various types of electrodes were studied, single-crystal, polycrystalline, n-type and p-type. One objective was t o develop an understanding of the response of an elec- trode/solution interface to potential step and potentiostatic-flash irradiation perturba- tions. Another objective was to examine the transient sub-microsecond photopoten- tial excursions obtained with coulostatic-flash experiments. EXPERIMENTAL POTENTIOSTATIC-FLASH INSTRUMENTATION The excitation source used was a S321 xenon flash lamp manufactured by Xenon Corp.The total energy per flash was 200 J, with a pulse width of 10 ps. Transient behaviour was monitored with a potentiostat constructed from selected components [see ref. (1 8)J; unity- gain bandwidth was ca. 5 MHz. Current measurement was made with a Burr-Brown 36225 differential-input instrumentation amplifier, for which the 3 db roll-off point is 2 MHz at a gain of 100. Current-time curves were obtained from oscilloscope traces or from a mini- computer data acquisition system, for which the maximum sampling rate for a 10-bit word was 167 kHz. Coulomb measurements were obtained by numerical integration of the data acquired by the minicomputer system. Currents which decayed too rapidly to be monitored accurately by the computer were recorded by oscilloscope, and the areas under the traces were measured by planimetry.RC bridge experiments and cyclic voltammetry were con- ducted using a PAR 173 potentiostat. COULOSTATIC-FLASH INSTRUMENTATION The pulsed laser source was a Molectron model UV 1000 nitrogen laser (337 nm) and Molectron model DL 200 dye laser with 10 ns pulse width. The coulostatic-flash instrumen- tation has been described," but some modifications to the measurement approach have been made for this study. They are as follows: ( a ) for photopotential transients in a time domain from ns to 10 p s , the working electrode and Pt quasi-reference electrode were directly con- nected (and a.c.-coupled) to a Tektronix type 7A13 (105 MHz bandwidth) differential ampli- fier plug-in which was mounted in a Tektronix model 7623A oscilloscope.All control/ monitoring electronics were located inside the Faraday cage enclosing the cell. The y-axis output from the scope back panel was connected externally to a type 7A19 (225 MHz band- width) plug-in amplifier (50sZ input, a.c.-coupled), mounted in a Tektronix model 7844 oscilloscope. (b) For photopotential transients in the time domain from 1 ps to seconds, the Pt quasi-reference electrode was capacitively coupled (0.47 pF) to the saturated calomel reference electrode (SCE), the SCE and working electrode potentials were input to a wide- band pass (5 MHz) high-input-impedance (10" n) differential amplifier (4.6 x), with d.c. offset bias so that the d.c.level of transients could be conveniently measured. Its output was in turn connected to a Tektronix type 7A16-A amplifier plug-in, with 1 Mi2 input impedance, and 225 MHz bandwidth. Traces were recorded photographically. The philosophy behind the above measurement approaches is documented The important point is that the Pt quasi-reference electrode reflects accurately any transient changes in the cell potential in a time domain of <10 ps. Capacitive coupling of the two reference electrodes allows measurements of photopotential transients referenced t o the SCE for times > I ps. However, when only the magnitude of photopotential transients was required the quasi- reference electrode could be monitored directly, without coupling to the SCE. CELL AND ELECTRODES The photoelectrochemical cell design is basically as described previously.20 Four different single crystal n-Ti02 electrodes were used for coulostatic-flash studies, representing a rangeDEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 35 TABLE 1 .--SUMMARY OF ELECTRODE CHARACTERISTICS (n-TiOz) ~~ electrode A E,,(MS)IV us.SCE - 0.4 ET,(C.W./U.V./d.C.)/v US. SCE -0.8 ETH(pulse iJ/V us. SCE -0.55 ETH(C.f*)/V ?IS* SCE -1.3 ETH(c.w./u.v./a.c.)/v US. SCE - 1.1 N,( MS)/cm - C,(E = + 1 .O V)/pF R,(E = +1.0 vyn 1.6 x 10" 0.14 3 30 area/cm2 0.1 1 B -0.75 -1.1 -1.3 3.0 x lo2' 2.2 29 0.14 C -0.53 -0.88 -0.80 - 0.50 -0.90 1.4 x 10'' 1.8 38 0.20 D -0.80 -0.93 -0.70 -0.50 -1.1 1.5 x lozo 1.8 N 86 0.21 ~~ All data obtained with 1.0 mol dm-3 KN03 electrolyte. EFB = flat-band potential; ETH = thres- hold potential; ND = donor density; MS = Mott-Schottky plot; c.w./u.v.refers to photocurrent against E experiments with C.W. U.V. laser source; a.c. refers to lock-in detection; d.c. refers to direct current measurement; c.f. refers to coulostatic-flash experiments; R,, C, refer to impedance bridge measurements of cell resistance and capacitance at 1 kHz. TABLE 2.-cHARACTERISTICS OF SEMICONDUCTOR ELECTRODES USED IN POTENTIOSTATIC- FLASH STUDIES material used dopant and type dopant le~el/cm-~ orientation flat-band potential (at pH 7)/V us. SCE mean capacitancea / p F cm-2 contact stable potential range /V us. SCE manufacturer preparation pretreat men t GaAs (111) 0.20 2 silver epoxy +0.2, - 1 .o General Diode 1 4 Ti02 Ti(n) <OOl) -0.74 4 x 1019 5 indium f1.0,-0.5 NL Industries 2 5 pol y-Ti02 Ti(n) 6 x 10l6 -0.50 - 0.01 silver epoxy + 1 .O, -0.8 Alfa Products 3 none pol y-Fe,O Feb) 1 x lo2' - 0.40 20 silver epoxy + 1 .o, -0.1 Alfa Products 3 none GaAs was used as received from General Diode, already doped.A pure crystal of Ti02 was doped in a reducing atmosphere(H,) for 30 min at 650 "C. The oxides were formed on a piece of metal foil (either Ti or Fe) by heating in a dry oxygen atmosphere to 1000 "C and allowed to anneal in the furnace as it cooled to room temperature. GaAs was etched in H,SO, (95%)/H202 (30%)/- H 2 0 (3: 1 : 1) for 90 s" before a set of runs. Ti02 crystals were rinsed in methanol before use. (1 Mean capacitance is defined as the average capacitance of the electrode over the potential range studied.of resistivities and donor densities. characteristics are summarized in table 1 . tiostatic-flash studies are described in table 2. These were designated as electrodes A-D, and their The semiconductor electrodes used in the poten- PHOTOELECTROCHEMICAL PROCEDURES The coulostatic-flash experimental procedure has been described in detail e1~ewhere.l~ It involves first adjusting the cell voltage to the desired initial potential, E,, for the semi- conductor working electrode using the polarograph. A solid state electronic switch which opens the cell circuit is triggered at 5 s intervals. After ca. 30 ps time delay the pulsed laser is triggered and the open-circuit potential difference between the working and reference36 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY electrodes is monitored as described above. After a 2 s interval the cell circuit is closed again and the intial potential reimposed by the polarograph.Pulse photocurrent measurements were made as described previously.20 RESULTS AND DISCUSSION POTENTIOSTATIC-FLASH EXPERIMENTS Since potential steps can be used to identify intermediates and products of photo- redox reactions,21 an understanding of the response of a semiconductor/electrolyte interface to a potential step must be formulated. In doing this, first consider the equivalent circuit for a semiconductor-electrolyte interface given in fig 1 . 2 2 3 2 3 The Rsc FIG. 1 .-Analogue representation of an electrochemical cell. Rsol : represents the cell impedance Cdf the double-layer capacitance, C,, the space-charge capacitance, R,, and C,, the surface-state impedance and capacitance, respectively, R,, and R, the impedance of the space-charge region and semiconductor bulk, respectively.Rr represents the faradaic impedance. The total current passed through the cell is denoted by isol which is equal to the sum of the currents passed through the space- charge region (isc) and surface states (i)SS. impedance to ion transport through the cell is represented by the double-layer capacitance (or Helmholtz capacitance), c d , ; the capacitance of the space-charge region of the semiconductor electrode, Csc ; the impedance to charge transport through the space-charge region and the bulk of the semiconductor electrode are represented by R,, and R,, respectively. The product R,,C,, is the charging time-constant of sur- face states at the interface.The faradaic impedance is represented by Rf, which is set to infinity for the potential step theory developed here. When a potential step is applied across this equivalent circuit the changing potential across the space-charge region will cause the space charge capacitance to change in a manner expressed by the Mott-Schottky relationship lo, 11*24 where E is the permittivity of free space, E , is the dielectric constant of the semicon- ductor, e is the charge of an electron, N is the number of donors or acceptors in the semiconductor, VSc is the potential drop across the space-charge region and 9 = kT/e. To determine the importance of this effect some simplifications in the equivalent circuit of fig.1 will help. First, assume that the effects of surface states can be neg- lected, eliminating R,, and C,, from the circuit. This leaves the space-charge capaci- tance and double-layer capacitance in series so that the total capacitance, CT = c , , c d l / ( c , c + c d l ) . In general we can impose the constraint that C,, <C,,, leaving the capacitance of the semiconductor-electrolyte interface to be representedDEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 37 by only Csc. Finally, the impedances of the space-charge region, the bulk of the semiconductor and of the solution can be combined into one net impedance. The simplified circuit is shown in fig. 2 and consists of only two elements, the interfacial R FIG.2.-Simplified analogue equivalent to an electrochemical cell with a semiconductor working electrode. R represents the combined cell and semiconductor impedance and C the interfacial capacitance. capacitance (C) and the net solution-semiconductor impedance (R). Since faradaic current is not considered in understanding the potential step response, Rf is allowed to go to infinity and is not included in fig. 2. The interfacial capacitance is shown as a variable capacitor since it is voltage dependent. For a potential step applied to the simplified circuit of fig. 2 the following boundary values are defined, t < to; V = V,, i = 0 and V,, = V, Vf - vo t = to; V = Vf, i = i,,, = ~ R (2) and Vsc = Vo (3) Vf - R t > to, v = vf, i = -‘Sc Substitution of the relationship and eqn (1) into yields _- - dt. -~ di 2 idvf-q- Ri Integration of eqn (8) results in 4 < I = - - R (‘f - 9) m z where dvo - q l - d V f - y‘vo - q F d G t = exp( - t/RC,) and (4)38 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY A plot of In i against t is curved, with an intial slope of l/RCo and a final slope of l/RCf.But, for small potential steps or steps through a potential region for which the space-charge capacitance changes very little, a In i against t plot is nearly linear. With this in mind we can now proceed to the more complicated and accurate equivalent circuit and study its response to a potential step. If we assume that Csc changes very little for the potential step to be applied to the cell, the following set of equations re- sult from the application of Kirchhoff's laws to the circuit in fig.1, assuming Rf is infinite. 1 1 %c - = 0 c s c j s s R s s + 4ss - - css All resistances and capacitances retain the definitions given previously. The ion flux, represented as current, is denoted by isol ; current through the space-charge capacitance, isc; current through the surface states, iss. The magnitude of the poten- tial step is Vf - Vo, the difference of the final and initial potentials. The charge on each of the capacitors in the circuit is denoted by q with the appropriate suffix. [Recall that i = dq/dt, so that differentiation of eqn (12) and (13) will replace the q terms with i. J This system of equations is solvable by differentiation and separation into a system of linear o p e r a t o r ~ .~ ~ The result is as follows, where Rr2 + l / C d l Rr2 + 1/c , 6 2 = - r l = - +(a - d m , r2 = - + (a + da2 - 4b), R = Rsol + R,, + R, and 1 1 c=-+-. C d , c s c The important observation here is that the total cell current decays as a function of the sum of two exponential functions. Furthermore, if C,, and Cs, < c d , and Rss 9 R (i.e., the surface states possess a long time constant and the double-layer capacitance is large compared with the surface-state and space-charge capacitances,DEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 39 which are not unreasonable hold : the following two simplifications (16) (17) 1 - = (R + Rss)Css and r1 1 -= RC,,. r2 Although it is difficult to test eqn (15) rigorously, experimental observations do demonstrate two different regions of exponential decay.An example of this beha- viour is shown in fig. 3. Table 3 lists the initial and final slopes (l/rl and 1/r2) from the semilog plots of fig. 3, where several different potential steps were applied to a - 6 -7 7 -8 E n 3 -9 U C - -10 -11 I 1 I I I I 1 0 2 4 6 a 10 12 14 tlms FIG. 3.-Semilog plot of the current response to a potential step for GaAs. Potential steps are from -1,000 to -0.800 V (a), -0.600 V (b), -0.400 V (c), -0.200 V ( d ) and 0.000 V (e). UU-U wwrmng cicurvuc. fiiong wirn rnese values are iisrea rne KL proaucts or me cell (an RC value is the product of the cell impedance and capacitance as determined by the bridge technique). Notice that the initial and final decay constants do not Y - ____ - - tions which lead to eqn (16) and (17) is not maintained.Nevertheless, the type of behaviour described by eqn (1 5) is at least qualitatively followed. Moreover, there is TABLE 3.-cOMPARISON OF THE INITIAL AND FINAL DECAY CONSTANTS FROM EQN (15) TO THE RC OF THE CELL DETERMINED BY THE RC BRIDGE METHOD AT VARIOUS POTENTIALS. THE WORKING ELECTRODE IS GaAs IN A pH 7 BUFFER SOLUTION. 1 .ooo 0.800 0.50 12 27 1 .ooo 0.600 1.60 14 29 1 .ooo 0.400 2.30 17 30 1 .ooo 0.200 3.60 24 33 1 .ooo 0.000 3.80 29 3540 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY reasonably close agreement between l/r, and the RC value obtained from the bridge technique. Bear in mind, however, that the above derivation and discussion provides us with no information about the energetic location of surface states at the interface (relative to the valence and conduction bands); nor is there any information available to evaluate the density of surface states or relative numbers of surface states with different time constants.We can only demonstrate the existence of surface states which are slow in comparison with the space-charge time constant (which is deter- mined by RC,,). Also Slow surface states are evidenced by the large values of l/rl. K a .,E . -- 1 0 1 0 U 0 10 20 30 40 50 FIG. 4.-Current response of the cell to a potential step (ca. 100 mV) for each type of semiconductor electrode, (a) poly-TiO,, (6) GaAs, (c) sc-Ti0, and (d) Fe203. All solutions are in pH 7 buffer. The potential of each electrode is + 1 .O, - 1 .O, + 1 .O and + 1 .O V vs. SCE, respectively. note that no information regarding Cdl is obtained from the above analysis of current- time behaviour.A graphic summary of surface contributions to the charging current profiles for each type of electrode studied is presented in fig. 4. Notice that for the polycrystal- line TiO, electrode the charging time is very short. This suggests that either the cell and electrode resistance (R) is very small and/or that the surface states are very fast (i.e., R&, is small). Semilog plots of current against time do not show two distinct slopes. At the other extreme, GaAs, single-crystal TiO, and polycrystalline Fe,O, demonstrate a considerable contribution from slow surface states (relative to the space- charge decay constant). Semilog plots for these electrodes show two distinct decay constants (see fig.3). The dependence of photocurrent on potential for flash irradiation is fundamentallyDEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 41 the same as that for continuous irradiation, as shown in fig. 5. Procedurally, the process of flash irradiation can be broken down as follows. Before illumination a potential is applied to the cell so that a certain amount of band-bending is present in the space-charge region of the semiconductor. This causes a majority carrier defi- ciency at the surface relative to the bulk. At the onset of irradiation electron-hole pairs are formed and, since a potential gradient exists across the space-charge region, the newly formed electrons and holes migrate in opposite d i r e ~ t i o n s ~ ~ s ~ ~ until a steady- state concentration gradient is formed.This happens very quickly compared with the duration of the flash and ion migration in the solution.28 In an n-type semicon- ductor the electrons promoted to the conduction band are received in the bulk of the //O GaAs / 0.5 # iJmA cm-2 FIG. 5.-Photocurrent-voltage profile for each type of electrode taken at the current maximum after the flash. A11 solutions are in pH 7 buffer. The potential of each electrode is the same as that in fig. 4. semiconductor, while the holes in the valence band migrate to the surface. If the energetics are correct water will be oxidized to give oxygen. The electrons of a p-type semiconductor, however, are accumulated at the surface in the conduction band and may reduce the solution, while the holes in the valence band migrate into the bulk.As a result of this process two sources of perturbation on the original potential field exist. First, the steady-state redistribution of electrons and holes in the space-charge region creates a new potential distribution which has already been referred to as a p h o t o ~ o l t a g e . ~ ~ * ~ ~ Secondly, transport of charge across the double layer in the form of faradaic current discharges the double-layer capacitance, changing the potential drop across it. After termination of the irradiation the remaining electron-hole pairs recombine on a time-scale fast relative to the flash d ~ r a t i o n . ~ ~ ~ ~ ~ Hence the electrode production of electron-hole pairs ceases and no further chemical reaction occurs at the interface.All that remains at non-equilibrium is the potential drop across the double-layer and space-charge capacitances, which will be recharged at a rate go- verned by eqn (1 5). Quantitative evidence for this is lacking for the semiconductor- electrolyte interface. Notice, however, that the same trends are observed for the response of the electrode to a potential step or flash irradiation. An electrode which demonstrates a short decay-constant for a potential step will re-establish equilibrium very rapidly after a flash, and an electrode with a long potential step decay-constant will require a relatively long time to establish equilibrium after a flash (cf: fig. 4 and 6).42 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY One set of experiments involved flash irradiation with a delayed potential step.For these experiments an electrode is held at a potential favourable for a photo- current to flow upon flash irradiation. After the flash is over (25 p s delay), a potential step of a predefined magnitude is applied to the cell so that at the new potential any products or intermediates of the photoreaction might be detected electrochemically. The timing of the experiment for an n-type electrode is clarified in fig. 7. At time to I 0 1 0 0 50 100 tips FIG. 6.-Typical photocurrent response for poly-Ti02 (b), GaAs (c), Sc-Ti02 ( d ) and Fe203 (e) com- pared to the Xe flash lamp profile (a). The photocurrent responses are from the output of the current measurement amp of the potentiostat. The flash lamp profile was measured with a Motorola MRDSOO photodiode (rise-time = 1 ns).All solutions are pH 7 buffer solution, and the electrode potentials are the same as in fig. 4. the flash is initiated with a positive potential applied, and the current response of the cell is monitored. At time t l , after a delay of 7, a negative potential step is applied sufficient to reduce photoproducts formed by the flash. If z is large enough, the cur- rent flow through the cell induced by the delayed potential step will be identical to the current induced without the preceding flash. That is, only step-charging current is observed. As z decreases, however, fewer of the products produced as a result of the flash will have had a chance to diffuse away from the electrode, so that there will be a farradaic contribution to the step-induced current.The temporal behaviour of theDEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 43 faradaic current can be obtained by subtraction of the response due to the potential step alone from the response observed with the flash and step. In other words, the faradaic and charging-current contributions are additive [this is not precisely true at times which are small compared with the RC of the cell; see ref. (21) and (34) for more on this subject}. The faradaic contribution should vary inversely as the square FIG. 7.-Timing diagram for a flash-potential step experiment. The initiation of the flash is at to. After a delay time (z) the potential step is applied at tl. root of z, by analogy to heat-transfer as the electrochemically active product diffuses away from the electrode.Another consideration coupled to a decrease in z is the amount of current still flowing as a result of the initial flash. When z is so small that a potential step is ini- tiated while a significant amount of current is still present, as a result of the flash, the potential drop across the semiconductor/electrolyte interface will be unknown. 0 ‘flash ‘f ‘0 FIG. 8.-Relationship of charge passed during a step, flash and Aash-step. The total charge passed through the cell by the potentiostat as a result of a flash at a potential Vo is (Ql + Qz); for a step from Yo to V,, -Qz; for a flash at V, followed by a step to Vr, Ql. Consequently the time dependence of the charging current is unknown and a point by point subtraction, retaining temporal integrity, is not valid.By reference to fig. 8 it can be shown, however, that the faradaic and charging current contributions are separable by subtraction of the total charge passed. In fig. 8 the total interfacial capacitance is plotted as a function of the potential drop across it. The area under the curve between two potentials is the number of coulombs accumulated on the capacitor when the potential is changed from one to the other. As the capacitance44 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY against voltage curve is plotted in fig. 8 for a potential step from V, to V,, the amount of capacitative charge passed will be, qs = - Q2. The amount of charge passed as the result of a flash can be represented by a potential step from Yo to Vflash (assuming an instantaneous flash generating an initial photovoltage of Yflash - V,) where the total number of coulombs passed will be, q F = Ql + Q2.(Note, however, that the mea- sured current is opposite in sign to qs, because it results from the potentiostat attempt- ing to change the potential from Yflash to Yo after the flash.) For a flash coupled with a step the final potential will be Vf and the charge passed, qFS = el, if no faradaic processes are initiated as a result of the potential step. Thus, subtraction of the charge passed during a step alone from the charge passed during a flash with delayed step should be equal to the charge passed for a flash alone. This can be summarized by defining qNET such that qNET = qFS - q S - qF.If no faradaic process is initiated by the delayed potential step, qNET = 0. However, if the products of the flash are reducible or oxidizable at the new potential, qNET # 0. Due to the assumption of an instantaneous flash however, eqn (18) is only precisely true for electrode/electrolyte systems with a long time-constant. In other words, the longer it takes a system to respond to a flash perturbation the more instantaneous the flash will appear. Fig. 9 summarizes the dependence of qNET on the potential stepped to (V,) for all four electrodes. The assumptions which lead to the calculation of qNET via eqn (18) are strictly followed for only the slower responding electrode, Fe,O, (see fig. 6). How- ever, the trends observed for iron oxide and the other electrodes are the same.Notice (18) FIG. 9.-Plot of the qNET against potential for each type of electrode, sc-Ti02 (a), Fe203 (b), poly- Ti02 (c) and GaAs (d). The residual q is obtained by subtracting qs from qFS and then subtracting the amount of charge passed for a flash at the initial potential from the result. that qNET is non-zero at all potentials, except the initial potential, for which the poten- tial step is zero. To test the presumption that the non-zero qNET is the result of faradaic current a set of experiments were conducted with a constant potential step, but variable delay time, z. By doing this, the initial difference between the current due to a potential step and a potential step applied after a flash could be measured over a wide range of delay times.Polycrystalline TiOz was chosen for this study since it responds most quickly to a flash, leaving little charging current to interfere with theDEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 45 delayed potential step (see fig. 6). The result follows typical Cottrell behaviour (fig. 10). That is, the net current detected by a potential step after the flash decays as l / d z . This result is consistent with the detection of a stable photoproduct (probably oxygen) which is diffusing away from the electrode. The lifetime of this photo- product is > 15 ms. Because there are no sharp steps in the plots of qNET against potential (fig. 9), as in a polarogram, it appears that significant back reaction of product occurs at the elec- trode, even at potentials favourable for photo-breakdown of the solvent.(It should 600 i I 1 0 50 100 z-*/s-f FIG. 10.-Cottrell plot for poly-Ti02. The potential of the electrode before and during the flash is + 1 .000 V us. SCE. The potential stepped to is -0.400 V us. SCE. The current (i) is measured at the initiation of the step. be pointed out that the fact that qNET goes to zero at the initial potential is inherent in the design of the experiment and does not imply that the product is not being con- sumed by the electrode.) The evidence here suggests that at an n-type semiconductor electrode some of the photo-oxidation product is reduced. The driving force for the backreaction of the photoproduct at the electrode can be broken down into two contributions. First the charge transfer which takes place during irradiation drives the potential across the space-charge region toward flat-band.30*31 This results in a cathodic shift for an n-type electrode which is more favourable for reduction.Likewise for a p-type electrode, the shift in potential across the space-charge region is anodic. For the electrodes used, the amount of charge passed as a result of a flash experiment is about the same as a potential step of 400 to 800 mV. Secondly, the use of surface states for electron transfer between the product and the electrode is suggested here since suitable overlap between the product and the conduction band is unlikely. When the valence-band energy and the energy of the donor orbital of OH-46 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY are in close proximity the energy level of the oxidized form of OH- (0,) will be too high to interact favourably with the valence band.29p32 For example, a cyclic voltam- mogram of oxygen saturated electrolyte with a polycrystalline TiO, electrode shows no distinct O2 wave, only a gentle rise that begins at ca.0.3 V vs. SCE (also true for other n-type electrodes used here). It has already been reported in the literature that sur- face states can furnish a significant amount of ~ u r r e n t . ~ ~ * ' ~ * * ~ Also, it has been re- ported that etching can reduce surface-state contributions 23 and that etching increases the efficiency of the semiconductor electrode's ability to convert radiant to electrical en erg^.^ Thus, from the observations in the literature and this study it appears reasonable that backreaction uia surface states of the product of the photo-induced decomposition of water can be significant.COULOSTATIC-FLASH STUDIES In contrast to pulse photocurrent measurements, coulostatic (open-circuit) photo- potential (E,) measurements made with a pulsed laser source are not limited in response time by the cell time c o n ~ t a n t . ' ~ * ~ ~ The response is limited by the RC time-constant defined by the cell impedance (R = R, + R,,) and the combined value of the inter- electrode stray capacitance, the input capacitance of the measurement electronics, and cable capacitance. This combined value, C,, is typically less than 100 pF. Thus, for electrodes A to D (table l), respectively, RC, <4 <3 <4 and <5 ns, where R is the value measured at high frequency with a vector impedance meter in 1.0 mol dm'3 KN03.These response time constants are comparable to or less than the pulse- width of the laser source (10 ns), thus allowing time-resolved studies in a time domain of 210 ns as shown below. Fig. 11 and 12 show typical photopotential (E,) behaviour for coulostatic flash experiments in 1.0 mol dm-3 KN03, using the nitrogen laser (337 nm). Irradia- tion with the nitrogen pumped dye laser (520 nm) resulted in photopotential excur- sions at least four orders of magnitude lower (using comparable laser intensities). Several features are worth noting. First of all, it is clear that the rise-time (tg5%) is ca. 12 ns, which suggests that it is limited by the laser pulse-width and/or the bandwidth of the measurement electronics.The potential changes sharply in a more negative direction, but then decays back towards the initial potential, Ei. When Ei is very positive of EFB, the final photopotential, (Ep)F, is reached after ca. 1-20 ms for the electrodes studied here. When Ei is near EFB, not only is there a much smaller initial excursion of the photopotential, but (Ep)F + Ei ; also, (E,JF is reached in < ca. 1 ms. The net magnitude of photopotential excursion achieved after the decay, (AEp)F, is taken to be indicative of the net quantity of charge transferred, or " photocharge " (Q,). Qp can be calculated from the relationship, Qp = C,(AE,)F where C, is the combined capacitance of C,,, C,, and Cdl. An indication of the efficiency of the photoelectrolysis process can be obtained from the ratio Q,/A, where A is the electrode area.The data in table 4 clearly show that electrode B is the most efficient and elec- trode A is the least efficient. By comparison with table 1, the trend in efficiency cor- relates well with the donor density. (However, because the laser pulse intensity varied &25% due to optical alignment, the efficiency data can only be interpreted semi-quantitatively .) The times, tF, required for the photopotential to decay to (AEJF were only esti- mated from visual inspection of photographic traces, but do indicate a qualitative in- verse correlation between tF and the donor density. Finally, the quantum efficiency Table 4 summarizes the time-dependent behaviour of four electrodes.DEUTSCHER, RICHARDSONy PERONE, ROSENTHAL AND ZIEMER 47 P EP a P J P _i 1 1 1 1 1 1 - / a / l 5 mV I time time L _I time FIG.11 .-Coulostatic-flash photopotential transients at positive potential. Electrode C; E, = + 1 .O V us. SCE; 1.0 mol dm-3 KNOB; light intensity cu. 10 k W cm-z.48 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY time 1 time time FIG. 12.-Coulostatic-flash photopotential transients near EFB. Electrode C; Ei = -0.5 V us. SCE; 1.0 mol dm-3 KNOJ; light intensity EZ 10 k W cm-2.DEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 49 (Q,/total photons incident) can be estimated to be ca. This value is rather low, but certainly includes the uncertainties in measuring the total charge transferred and the number of photons absorbed in the depletion layer during each 10 ns laser pulse.Several different factors may be involved. The first factor is that the initial photopotential is caused by electron- hole separation, with the subsequent decay caused by electron-hole recombination. The rates of such processes in the bulk for highly doped semiconductor materials are typically found to be in the sub-microsecond time domain.36 However, longer The nature of the Ep decay is certainly of interest. TABLE 4.-sUMMARY OF TIME-DEPENDENT BEHAVIOUR OF SEVERAL n-Ti02 ELECTRODES WITH PULSED-LASER COULOSTATIC-FLASH EXPERIMENTS IN 1 .O mol dm- KNO, (A) For E1 = +1.5 V us. SCE A B C D z"/ns 12 10 15 12 tF/ms N 20 - 2.2 - 1 - 1 (C\E,)F (AEp)ma x (bQp/A)/pC cm-f 0.024 0.56 0.10 0.12 0.15 0.76 0.59 0.41 CS/@ 0.12 1.96 1.56 1.46 (B) For Ei near EFB for each electrode A B C D z"/ns 12 10 12 12 t F b -lo00 -50 -175 -150 0.10 0.18 0.13 0.20 (AEPlF (AEp)max ('Qe,/A)/pC 0.017 0.067 0.044 0.032 EilV US.SCE -0.7 -0.6 -0.5 -0.8 cs IPF 0.53 3.5 3.1 3.4 "z = tgs% = rise time of photopotential in ns, based on a smoothed fit to the rising portion of the = (AE,)FC,; A = working electrode area in cm'. (The C, value used was taken from curve. MS plots or was taken as the maximum measured value from the MS plot when Ei < EFB.) lifetimes might occur in the space-charge region at a semiconductor-electrolyte inter- face (band bending at the interface facilitates the separation of electrons and holes). The second factor is oxidation by holes migrating to the surface (of either the sol- vent or species at the electrode surface).Note that this process would tend to increase AEp with time. An opposite effect would be the dark reduction of photoproducts; i.e., the photopotential decay might represent the dark back reaction of photo-oxidation products produced by the flash. This would be consistent with the observations in the potentiostatic-flash studies that such back reactions may occur, probably via sur- face states, even when the electrode potential is quite positive of EFB. The rate (or efficiency) of the back reaction appears to increase as the potential is made less posi- tive, and this is consistent with the observation here that the decay time constant decreases as Ei approaches E F B (table 4). The fourth possible explanation is that the very high rate of charge injection50 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY effected by the intense 10 ns laser pulse causes a transient non-equilibrium expansion of the space charge region in the semiconductor, followed by a relaxation back to its equilibrium dimensions, with associated overshoot and decay of the photopotential.(This decay may or may not involve hole-electron pair recombination, depending on the survival of excess surface holes.) Such a relaxation effect has been observed with coulostatic electrical charge injection at a mercuryldilute-electrolyte i n t e r f a ~ e ~ ~ n ~ ~ and has been described in terms of a transient over-expansion of the Helmholtz layer (interfacial double layer) followed by a relaxation back to an equili- brium thickness, with corresponding changes in the measured double-layer voltage analogous in nature and time dependence to those observed here.The nature of steady-state photopotentials generated under conditions where charge transfer to solution is blocked has been described previously; 42-44 and experi- mentally observed instantaneous photovoltaic response has been d e ~ c r i b e d . ~ ~ ’ ~ ~ G e r i ~ c h e r ~ ~ suggested that short light pulses should be used for these measurements to obtain a better characterization of the space-charge region free from charge-transfer and surface effects. However, no reliable measurements have been reported on a time-scale comparable with our studies. The possibility that a transient photopoten- tial overshoot and decay might be observed has been suggested p r e v i o ~ s l y .~ ~ ~ ~ ~ Laser and Bard28 also pointed out that the response of the semiconductor space-charge layer to coulostatic electrical charge injection is analogous to that predicted by Feldberg41 for diffuse double-layer relaxation, except that it should occur in a much shorter time-domain. Unfortunately, the pulsed laser-induced photopotentiometry experi- ments performed here do not correspond exactly to the electrical coulostatic charge injection arrangement described by Laser and Bard.28 Nor do the conditions agree with those used to simulate steady-state photo potential^^^ or to simulate the relaxation of photogenerated free carriers.47 However, ref. (47) does demonstrate that distinct transient minima and maxima in the concentration-distance profiles of minority and majority charge carriers can be generated, perhaps leading to the kind of relaxation processes observed here.The time-scale of the transient behaviour predicted 47 is several orders of magnitude shorter than observed here. However, the conditions chosen for the simulations correspond to light fluxes and donor densities several orders of magnitude lower than those used here. To investigate further the merits of the fourth explanation for the photopotential decay, the time dependence was compared with that predicted theoretically for double- layer r e l a ~ a t i o n . ~ ~ In that case, assuming equal mobilities for positive and negative charge carriers, theory predicts that the observed double-layer potential should follow a t - 2 / 3 dependence during the latter part of the decay when the potential is near EpZC (potential of zero charge).Essentially the same theoretical concepts can be imposed on the response of the semiconductor space-charge region to an instantaneous light- induced charge injection. Holes and electrons are considered to replace cations and anions, with their associated mobilities and concentrations; the thickness of the deple- tion layer replaces the Helmholtz-layer thickness, and EFB replaces Epzc. The suggestion that Helmholtz-layer relaxation could also contribute to the time dependence observed here is of course appropriate. However, the effects have not been observed at mercury electrodes with concentrated electrolyte( > -0.05 mol d r n 7 and relatively small charge injection.37 Thus, we should not expect to see Helmholtz- layer relaxation effects here, with 1.0 mol dm-3 KN03 electrolyte and Q,/A < 1 pC cm-2.The value of Ei for the experimental data in fig. 13 was near EFB. The data plotted are for only the latter part of the decay, where a linear dependence on t‘2/3 was observed. This A typical plot of E, against t -2/3 is shown in fig. 13 for electrode D.DEUTSCHER, RICHARDSON, PERONE, ROSENTHAL AND ZIEMER 51 50 40 30 20 10 I I I I I 1 0 0.1 0.2 0.3 0.4 0.5 0.6 t-213pS-Z13 FIG. 13.-Transient photopotential us. t-213 for El near E F B . 1.0 mol dm-3 KN03, electrode D, -0.5 V us. SCE, gain = 4.6. behaviour is analogous to double-layer relaxation observation~,~~ where linear depen- dence of cell potential on t -2/3 is predicted after coulostatic charge injection when Ei is near Epzc.41 The behaviour shown in fig.13 was typical of all the electrodes used in these studies. Non-linear dependence on t - 2 / 3 was observed at short times, as TABLE 5.-sUMMARY OF t-213 ANALYSIS OF PHOTOPOTENTIAL DECAY FOR & NEAR EFB A electrode B C D EJV US. SCE 0.7 -0.6 - 0.7 -0.5 - 0.8 "S/V ps-2I3 0.089 0.012 0.0025 0.027 0.01 3 "SIPS N 30 -4 -4 N 15 N 15 dqp/pC cm-2 0.01 7 0.067 0.021 0.044 0.032 "CZ/lO4 ~ r n - ~ 0.027 5.0 5.0 2.33 2.5 '( AEpMmV 3.6 2.7 0.78 2.8 2.0 *105y 0.22 1.21 1.74 1.75 1.59 S is the slope of a AEp against t - 2 / 3 plot. begins. ' density = C, (AEp)F/A (see tables 1 and 2 for C,, A). ( =6 x ts is the time at which linear dependence on t - 2 / 3 is the intercept of the AE, against t - ' I 3 plot as t + 03.'q, is the photocharge Cz = ND/L, where L is Avogadro's number 7 = SCz3/2/qp5'3. indicated in table 5. Non-linear dependence on t - 2 / 3 was observed at all times when Ei was very positive of EFB (e.g., at + 1.0 V us. SCE), for all electrodes; the slope of a linear least-squares fit on a log-log plot of Ep us. t becomes increasingly more positive as Ei becomes more positive of EFB. CONCLUSIONS These studies demonstrate for the first time that photoinduced electrode processes can be observed reliably in the nanosecond to millisecond time domains. The ob- served limitations on time resolution are consistent with those predicted. The time dependence of transient potentiostatic photocurrents is clearly related to the cell time52 PULSED LASER-INDUCED PHOTOELECTROCHEMISTRY constant (microseconds) ; whereas the time dependence of transient coulostatic photo- potentials can be studies with ca.10 ns time resolution. The fact that the rise-time is ca. 12 ns in 1.0 mol dm-3 electrolyte may indicate that photoinduced transfer of charge to the electrolyte solution is essentially instantaneous with a 10 ns laser pulse. The observed subsequent photopotential decay may be related to a phenomenon of space-charge relaxation which exhibits a decay time varying from ca. 1 to ca. 20 ms as the potential becomes more positive of EFB. A significantly shorter and less dramatic decay is observed with increased doping density. Alternative explanations suggested here include photoproduct back reaction or hole-electron recombination. Because these other processes are known to occur to some extent, their possible con- tributions to the initial photopotential transient behaviour should be investigated fur the r .These studies have provided a foundation on which to base further investigations of transient behaviour with other semiconductor photoelectrodes and other photo- electrolysis processes. It should be possible to study the dynamics of redox stabilized photoelectrodes as well as dye-sensitized photoelectrolysis on a previously inacces- sible time-scale. Modification of the local environment (e.g., non-aqueous solvents) may be necessary to fully exploit this technique for studying chemical processes as opposed to processes inherent to the semiconductor electrode.The authors thank Jackson Harrar and Lloyd Steinmetz for their contributions to this study. This work was supported by the U.S.D.O.E. contracts no. W-7405- ENG48(LLL) and DE-AC02-77ER04263. A002 (Purdue). L. Hollan, J. C. Tranchart and R. Memming, J. Electrochem. SOC., 1979, 126, 855. J. O’M. Bockris and K. Uosaki, J. Electrochem. Soc., 1977, 124, 1348. K. Nakatani and H. Tsusomura, Bull. Chem. SOC. Japan, 1977,50,783. A. B. Bocarsly, J. M. Bolts, P. G. Cumins and M. S. Wrighton, Appl. Phys. Letters, 1977,31, 568. R. H. Wilson, L. A. Harris and M. E. Gerstner, J. Electrochem. Soc., 1979, 126, 844. L-S. R. Yeh and N. Hackerman, J. Electrochem. Soc., 1977, 124, 833. K. L. Hardee and A. J. Bard, J . Electrochem.SOC., 1976, 123, 1024. J. H. Kennedy and K. W. Frese Jr, J. Electrochem. SOC., 1978, 125, 709. 419. a J. S. Curran and W. Gissler, J. Electrochem. SOC., 1979, 126, 56. lo S. M. Wilhelm, K. S. Yun, L. W. Ballenger and N. Hackerman, J. Electrochem. Soc., 1979,126, l1 L. A. Harris and R. H. Wilson, Ann. Rev. Muter. Sci., 1978, 8, 99. l2 R. L. VanMeirhaeghe and F. Cardon, Ber. Bunsenges. phys. Chem., 1979, 83,236. l3 D. Laser and S. Gottesfeld, J. Electrochem. SOC., 1979, 126, 475. l4 S. S. Fratoni and S. P. Perone, Analyt. Chem., 1976, 48,287. Is K. F. Dahnke, S. S. Fratoni and S. P. Perone, Analyt. Chem., 1976, 48,296. l6 G. C. Barker, A. W. Gardner and G. Bottura, J. Electroanalyt. Chem., 1973, 45, 21. 17J. H. Richardson, S. B. Deutscher, A. S. Maddux, J. E. Harrar, D.C. Johnson, W. L. Schmel- zinger and S. P. Perone, J. Electroanalyt. Chem., 1980, 109, 95. S. S. Fratoni, Ph.D. Thesis (Purdue University, W. Lafayette, IN, 1976). l9 C. C. Herrmann, G. G. Perrault and A. A. Pilla, Analyt. Chem., 1968, 40, 1173. 2o S. P. Perone, J. H. Richardson, B. S. Shepard, J. Rosenthal, J. E. Harrar and S. M. George, in New Applications of Lasers to Chemistry (A.C.S. Symp. Ser., Washington, D. C., 1978), vol. 85, pp. 126-170. 21 B. S. Hall, K. F. Dahnke, S. S. Fratoni Jr, and S. P. Perone, J. Phys. Chem., 1977, 81, 866. 22 S. Toshima, Progress in Surface Membrane Science, ed. J. F. Daniel], M. D. Rosenberg and z3 M. Tomkiewicz, Semiconductor Liquid-Junction Solar Cells, ed. A. Heller (The Electrochemical 24 F. M. Delnick and N. Hackerman, J. Electrochem. SOC., 1979, 126, 732. 25 Any number of good books on differential equations will demonstrate the technique used to D. A. Cadenhead (Academic Press, New York, 1974), vol. 4, pp. 231-297. SOC., Princeton, N. J., 1977), proceedings vol. 77-3.DEUTSCHER, RICHARDSON, PERONE, ROSENTHAL A N D ZIEMER 53 derive eqn (15). An example is Introduction to Diflerential Equations by Shepley L. Ross (Xerox, Lexington, MA, 1974). 26 M. Nishida, Nature, 1979, 277, 202. 27 W. Schmickler, Ber. Bunsenges. phys. Chem., 1978, 82, 477. 28 D. Laser and A. J. Bard, J. Electrochem. Soc., 1976, 123, 1828. 29 H. Gerischer, Physical Chemistry: An Advanced Treatise, ed. H. Eyring (Academic Press, New 30 H, Reiss, J. Electrochem. Soc., 1978, 125, 937. 31 H. Gerischer, J. Electrochem. SOC., 1966, 113, 1174. 32 F. Williams and A. J. Nozik, Nature, 1978, 271, 137. 33 R. J. Nelson and R. G. Sobers, J. Appl. Phys., 1978, 49, 6103. 34 S. S. Fratoni, Jr., and S. P. Perone, J. Electrochem. SOC., 1976, 123, 1672. 35 H. S. Carslaw and J. C. Jaeger Conduction of Heat in Solids (Oxford University Press, 2nd edn, 36 S. M. Sze, Physics of Semiconductor Devices (Wiley-Interscience, New York, 1969). 37 F. C. Anson, R. F. Martin and C. Yarnitzky, J. Phys. Chem., 1969, 73, 1835. 38 C. Yarnitzky and F. C. Anson, J. Phys. Chem., 1970, 74, 3123. 39 J. Newman, J. Phys. Chem., 1969, 73, 1849. 40 R. P. Buck, J. Electroanalyt. Chem., 1969, 23, 219. 41 S. W. Feldberg, J. Phys. Chem., 1970, 74, 87. 42 C. G. B. Garrett and W. H. Brattain, Phys. Rev., 1955, 99, 376. 43 E. 0. Johnson, Phys. Reu., 1958, 111, 153. 44 D. Laser and A. J. Bard, J. Electrochem. Soc., 1976, 123, 1833. 45 P. J. Boddy and W. H. Brattain, J. Electrochem. SOC., 1963, 110, 570. 46 P. J. Boddy and W. H. Brattain, Ann. N. Y. Acad. Sci., 1963, 101, 683. 47 D. Laser and A. J. Bard, J. Electrochem. Soc., 1976, 123, 1837. 48 R. K. Rhodes and R. P. Buck, J. Electroanalyt. Chem., 1979, 103, 19. York, 1970), vol. 9A. 1959).