J . Phys. Chem. 1992, 96, 4558-4563
4558
interactions between the adsorbed thiolate species and is 0.45 ML. Both the maximum coverage and the vibrational spectra indicate that the aromatic ring is tilted substantially away from the surface plane. Acknowledgment. Research was sponsored by the Division of Chemical Sciences, Office of Basic Energy Sciences, US.De-
partment of Energy, under Contract DE-AC05-840R21400 with Martin Marietta Energy Systems, Inc. I thank S.H. Overbury and D. R. Mullins for helpful discussions. Registry NO. PhSH, 108-98-5; Ni, 7440-02-0; C6H6,71-43-2; D2, 7782-39-0; H2, 1333-74-0; C2H5SH,75-08-1; phenyl thiolate conjugate base, 13133-62-5.
Energy Conversion by Photoeiectroiysis of Water: Determination of Efticiency by in Sltu Photocalorimetry Jiirgen K. Dohrmann* and Nils-Soren Schaaf Institut f u r Physikalische und Theoretische Chemie, Freie Uniuersitat Berlin, Takustrasse 3, D- 1000 Berlin 33, Federal Republic of Germany (Received: October 21, 1991; In Final Form: January 4, 1992)
By means of a power balance analysis for a water-splitting photoelectrolysis cell, relations have been derived for determining the energy conversion efficiency, with respect to absorbed monochromatic light, from photocalorimetric in situ measurements at the semiconductor photoanode. It is shown that a quantity readily obtainable from the potential dependence of the heat-monitoring signal represents the upper limit of the energy efficiency attainable with a given photoanode. For a cell with additional power losses associated with overvoltage from hydrogen evolution at the counterelectrode and with ohmic resistance of the electrolyte, the energy efficiency can be determined using the internal photocurrent quantum efficiency obtained by photocalorimetry and the voltage applied to the cell. Results are given for 360-nm photoelectrolysis of water at pH 0.5-13.5 in a cell employing a Ti02 (rutile) thin-film anode and a platinum cathode. Quantum and energy efficiencies are discussed in the light of previously reported findings.
Introduction Since the report by Fujishima and Honda on photoassisted splitting of water in an electrolysis cell utilizing an n-type Ti02 photoanode,' this and similar processes are being considered attractive for converting solar into chemical energy? Unfortunately, a variety of definitions have been used to express the energy conversion efficiency of a hydrogen producing photoelectrolysis ce11,2aJrendering comparison of published data difficult or even impossible. A common feature of the various definitions is that the power of incident light has been taken to represent the radiant energy input. While such expressions are useful for practical purposes, efficiencies defined in terms of monoenergetic radiant power actually absorbed by the semiconductor photoelectrode are of more fundamental significance. It has been shown that in situ calorimetric techniques based on the measurement of temperature change at the illuminated semiconductor electrode can be applied to quantify the energy efficiency of the electrode process with respect to absorbed light quantae6 and to determine the internal quantum efficiency of the p h ~ t o c u r r e n t . All ~ ~ these techniques offer the important advantage of obtaining the desired information without the necessity of measuring the absorbed light energy separately. Similar investigations have been made with solid-state solar cells.8 In previous communications we reported on in situ photocalorimetric studies of various photoanodic oxidation In the present work we analyze the relation between the energy conversion efficiency for photoelectrolysis of water and the heat-monitoring signal measured at the semiconductor photoanode. It is demonstrated that a quantity readily obtainable by photocalorimetry has a particularly simple thermodynamic significance in that it represents the upper limit to the energy conversion efficiency attainable with a given semiconductor photoanode. Results are reported for the 360-nm photoelectrolysis of water as a function of pH in a cell employing a polycrystalline rutile thin-film photoanode and a platinum cathode. The results are To whom correspondence should be addressed.
discussed in the light of efficiency data reported by other workers.
Experimental Section Materials. Ti02 thin-film electrodes were prepared by heating Ti foil (20-mm diameter, 0.08-mm thickness, 99.7%, Johnson Matthey) to red heat in the oxidizing Bunsen flame for 150 s. The oxide film of such electrodes consists of polycrystalline rutile.6 The film thickness was 1.7 pm as estimated from the increase in mass. After the film was abraded on one side for providing electrical contact to a Pt ring, the electrode was used without further pretreatment. The flatband potential and the donor density were 0.3 V vs the hydrogen electrode in the same solution and ca. lOI9~ m - respectively, ~, as obtained from a Mott-Schottky plot of the capacitance. Electrolyte solutions were made up of H2S04, NaOH, NaH2P04/Na2HP04(0.05 M each, pH 6.4), or Na2B407(0.01 M, pH 9) buffers in triply distilled water. The total ionic strength was kept at 1.5 M by adding Na2S04. Analytical grade reagents (1) Fujishima, A.; Honda, K. Narure (London) 1972, 238, 37.
(2) For recent reviews, see: (a) Pleskov, Yu. V. Solar Energy Conuersion:
A Photoelectrochemical Approach; Springer-Verlag: Berlin, Heidelberg, New York, 1990. (b) Getoff, N. Znt. J . Hydrogen Energy 1990.15.407. (c) Hill, R.; Archer, M. D. J . Photochem. Photobiol. 1990, 51, 45. (d) Bard, A. J. Ber. Bunsen-Ges. Phys. Chem. 1988,92, 1187. ( e ) Memming, R. Top. Curr. Chem. 1988. 143, 79. (3) Ang, P. G. P.; Sammells, A. F. J . Electrochem. SOC.1984, 131, 1462. (4) Fujishima, A.; Maeda, Y.; Honda, K.; Brilmyer, G. H.; Bard, A. J. J . Electrochem. SOC.1980, 127, 840. ( 5 ) Maeda, Y.; Fujishima, A.; Honda, K. Bull. Chem. SOC.Jpn. 1982, 55,
3373. (6) Rappich, J.; Dohrmann, J. K. Ber. Bunsen-Ges. Phys. Chem. 1988, 92, 1342. (7) Wagner, R. E.; Mandelis, A. Phys. Rev. B 1988, 38, 9920. (8) Cahen, D.; Buchner, B.; Decker, F.; Wolf, M. ZEEE Trans. Electron Deu. 1990, 37, 498. (9) Rappich, J.; Dohrmann, J. K. J . Phys. Chem. 1989, 93, 5261. (IO) Rappich, J.; Schaaf, N.-S.; Dohrmann, J. K. J . Electroanal. Chem. 1990, 279, 123. (11) Rappich, J.; Dohrmann, J. K. J . Phys. Chem. 1990. 94, 7735.
0022-365419212096-4558$03.00/00 1992 American Chemical Society
The Journal of Physical Chemistry, Vol. 96, No. 11, 1992 4559
Energy Conversion by Photoelectrolysis of Water (Merck p.a.) were used throughout. The pH was measured by means of a glass electrode. Instrumentation and Method. Measurements were made in a three-electrode photoelectrolysis cell equipped with a flat quartz window opposite the TiOz working electrode. The TiOZelectrode was mounted in the Teflon holder of the assembly described previously for pyroelectric detection of the thermal signal.'O The area exposed to the electrolyte was 1 cm2. The potential of the T i 0 2 electrode was controlled with respect to that of an Ag/ AgCl/O.l M KCl reference electrode by means of a home-built potentiostat. The Haber-Luggin capillary of the reference electrode was placed close to the T i 0 2 electrode to provide for negligible uncompensated resistance. The counterelectrode was a Pt foil (3 cm2). The catholyte was separated from the anolyte by a ceramic frit to prevent reduction of photoanodically generated oxygen at the Pt counterelectrode. Prior to and during the measurement, the catholyte was flushed with hydrogen. Electrolyte solutions in the working electrode and counterelectrode compartments were identical. The light source was a 450-W high-pressure xenon lamp (Osram XBO 450 W/1) in a housing with an elliptical reflector (Photochemical Research Associates; London, Ontario, Canada, Model ALH 220, f12.5). The lamp was operated at 300 W by a stabilized dc-power supply (AMKO, Tornesch, FRG, LPS 1000). A 360-nm light was isolated by an interference filter (Schott AL, 46% maximum transmission, 18-nm half-width) from the radiation prefiltered by an aqueous 0.1 M CoSO, solution (10-cm path length), a broad-band UV filter (Schott UG 5 ) , and a neutral-density filter (30%transmission). The light could be chopped a t a 1:l light-to-dark ratio (ORTEC Brookdeal 9479). Quartz lenses and a quartz diffusor plate provided for virtually uniform illumination of the T i 0 2 electrode. The intensity of the 360-nm light striking the electrode was ca. 2 mW.cm-2 as measured by means of a thermopile (Kipp & Zonen, Model CA-l) between the electrode and the fully opened chopper. The intensity corresponded to that in the 315-400-nm range of AM 2 solar radiation.2b Thus, by this optical configuration, which employed several filters, standardized illumination conditions were approximated and pH gradients otherwise generated by too large photocurrents were minimized. Photoelectrochemical and photocalorimetric measurements were made simultaneously under chopped illumination at 5.3 Hz. The photocalorimetric detector detailed previously1° employed a polyvinylidene difluoride foil (DT 1-028 K, Kynar piezo film, Pennwalt, King of Prussia, PA) as a pyroelectric sensor. The absolute values of both the modulated photocurrent and the sensor signal were measured using two two-phase lock-in amplifiers (EG&G PAR 5210) synchronized with the light modulation. For further improvement in the signal-tenoise ratio, band-pass filtering was applied. Measurements were taken as a function of the potential of the T i 0 2 electrode at 0.1-V steps. In addition, the cell voltage supplied by the potentiostat during the light period was recorded as a function of the potential of the Ti02electrode. An A/D converter (Analog Devices RTI-820) and a PC-AT were used for process control and data collection. '
Thermodynamic Considerations The objective of this section is to derive the relation between the energy conversion efficiency of photoelectrolysis of water and the quantities measured by photocalorimetry at the n-type semiconductor photoanode of the following cell M'/SC/02/H20/H2/M
(1)
Metal M' is in contact with the photoanode, SC, at which oxygen is evolved. M is a noble-metal cathode operating as a hydrogen electrode. The aqueous solution is saturated with oxygen and hydrogen under atmospheric pressure in the anode and cathode compartment, respectively. The pH is the same in both compartments. Corrosion and luminescence need not be considered in the present case of a polycrystalline titanium dioxide (rutile) photoanode. Upon illumination of the semiconductor anode with above-bandgap light and application of the voltage U to the cell,
0
T
Metal iM')
n-type Semiconductor
Electrolyte
Meta: (M)
Figure 1. Energy diagram for photoelectrolysis of water using a cell biased by voltage U between metal (M')-supported n-type. semiconductor photoanode and noble-metal cathode (M): E'F,EF,Fermi levels; Ec, EV, EHz H+, EH ojo,, energy levels of the conduction band, valence band, H2/H+- and H20/0,-redox couples, respectively; E,, photon energy; O E , standard emf of H 2 / 0 2 fuel cell.
water is split into hydrogen and oxygen with unity photocurrent yield
hu,U
H20(1)
H2W + 1/202(g)
(2)
The power balance for the photoelectrolysis cell is given by
9 a + IphU =
AoGf,HtOnH20 +
g
(3)
where 9, is the power of monochromatic light absorbed by the semiconductor electrode, I p h is the photocurrent, U is the voltage applied to the cell, A o G f , H 2 0 is the molar free energy of formation of H20(1), tiHz0 denotes the rate of photoelectrolysis (moles of H 2 0consumed per unit time), and g represents all the losses, per unit time, in the process of converting input energy into chemical free energy stored in the photoelectrolysis products, e.g., heat from radiationless electron-hole recombination in the photoanode. The energy conversion efficiency, qG,a,for photoelectrolysis of water in the cell described above will be defined as (4) denotes the standard emf of the reversible Hz/OZfuel cell ( O E = 1.23 V). By use of Faraday's law, tiHto= -Iph/2F, and of the relation AoGf,HzO = - 2 P E , eqs 3 and 4 can be combined to give O E
=
- k/9a
(5) As shown below, the term g/& contains the information obtainable by photocalorimetry. It should be noted that, for practical reasons, photoelectrolysis efficiencies given in the literature usually refer to the power of incident light,28*3J2 while the efficiency, eq 4, has been defined with respect to the power of absorbed light, in keeping with the power balance of the cell and the physical principle of ph~tocalorimetry.~-~ The power conversion loss, g, taken as a positive quantity in eq 3 can be evaluated by inspecting the energy diagram of the photoelectrolysis cellI3 biased by the external voltage U (Figure 1). After generation of an electron-hole pair by absorption of a photon having an energy EA(process a in Figure l), losses arise from radiationless electron relaxation (b) and electron-hole recombination (d) in the semiconductor electrode. The contribution of these photophysical processes to the power conversion loss is g S C , P = f i b ( E h - Eg) + fidEg (6) Nb and N d are the number of electrons passing, per unit time, channels b and d, respectively. E is the bandgap energy. Further, there are losses originating from t i e photoelectrochemical processes of charge separation in the space-charge layer (c), of electron transfer from the conduction-band level, Ec, to the Fermi level, %,a
(12) Ghosh,A. K.; Maruska, H. P. J. Electrochem.Soc. 1977,124, 1516. (13) Tomkiewicz,M.; Fay, H. Appl. Phys. 1979, 18, 1.
4560 The Journal of Physical Chemistry, Vol. 96, No. 11, 199’2 E’F, of metal M’ (e), and of electron transfer from the energy level, EH20,02,of the H 2 0 / 0 2couple to the energy level, of the hole at the semiconductor surface (f). These processes give rise to &,E
= Ne(Eg + E H 2 0 / 0 2 -
(7)
Ne is the number of photoelectrons flowing, per unit time, through the external circuit. Equation 7 is obtained by summing the contributions to gSC,E of processes c, e?and f and considering the continuity of photoelectron flux, Ne = N, = Np In addition, kinetic hinderance of hydrogen production (g) by electron transfer from the Fermi level, EF, of the metal cathode M to the energy level, EH2/H+,of the H2/H+couple contributes to the conversion loss as gM
= I(T,(EF - EH2/Hf)
(8)
The total power conversion loss from eqs 6-8 is g = &,,(EA- EJ + &,Eg+ Ne(Eg + EH2O/O2 - EH2/H+ + EF- E’F) (9) The power, d,, of monochromatic light absorbed by the semiconductor electrode is expressed by
da = Nph,aEA
Since in the loss processes b-f no work other than pu-work is done, the heat evolved can be expressed by Aq = Ag TAs where Ag and As are the changes in free energy and entropy, respectively. In keeping with thermodynamic sign convention, Aq and Ag are considered negative quantities. The sensor signal, P, generated during the illumination period, Af, is therefore proportional to At(d(-Aq)/dt). At nonzero photocurrent, the sum of the power conversion losses, gsc,p(eq 6) and gsc,E(eq 7), represent the rate of change in -Ag. Hence the sensor signal, P(esc), a t electrode potential esc can be expressed by
+
K is an instrumental parameter. Ne is the rate of photoelectron flux in the external circuit. QpE = -TAS denotes the Peltier heat per electron transferred in the photoelectrochemical process. In the present case, AS comprises the changes in entropy associated with photoanodic oxidation of water and transport of charge carriers in the electrode and the ele~trolyte.~ Using eqs 6 and 7, the potential-dependent internal quantum efficiency, qa(esc), of the photocurrent, eq 1 1 , and electrode potentials instead of energies on the Fermi scale, eq 14 can be written as
(10)
where Nph,a is the rate of photon absorption. By making use of the internal quantum efficiency, ‘la, of the photocurrent defined as ‘la = Ne/Nph,a
Dohrmann and Schaaf
(11)
and of the quantum efficiencies of the photophysical processes b, fi+,/&, h,a = 1, and d, N,j/Nph,a= 1 - ‘la,it follows from eqs 9 and 10 that ‘la
g/da = 1 + -(EH20/02 - EH2/H++ EF- E’F) (12) EA The internal quantum efficiency, ‘la,of the photocurrent is a function of the semiconductor electrode potential and can be determined by photocal~rimetry.~~ As seen from Figure 1, EH@/02 - EH2/H+= -leoJoEand EF - E $ = leolUwhere leal is the absolute value of the elementary charge. The standard emf, ‘E, corresponds to the difference of the equilibrium potentials of the H 2 0 / 0 2and H2/H+redox electrodes, t 0 , H 2o2~- EO,H~/H+, in the aqueous electrolyte at the same pH value. +he bias voltage can be expressed by U = tSC- t M + AtR, where tsc and t M are the potentials of the semiconductor anode and the metal cathode, respectively, measured versus an arbitrary reference electrode, and A e R is the voltage drop across the ohmic resistance of the electrolyte solution. Hence, from eqs 12 and 5 one obtains
At zero photocurrent, i.e., la= 0 at the illuminated electrode, only photophysical loss processes contribute to the sensor signal. This condition is met by measuring the signal a t the flatband potential, efb, of the semiconductor electrode under otherwise identical conditiom6 Defining the potential-dependent relative change, L(tsc), of the sensor signal, P(esc), with respect to the value, P(tb), a t the flatband potential and using eq 15, one obtains
By use of eq 176 or of similar relations derived for analyzing potential-dependent temperature change^,^-'^^ internal quantum efficiencies of the photocurrent for various semiconductor-electrolyte systems4-” and Peltier heats6Se1’ have been determined, including those for photoanodic oxidation of water as a function of pHe9 Equation 17 can now be combined with eq 13 to establish the relation between the quantities obtained by photocalorimetry at the semiconductor electrode and the energy conversion efficiency of the photoelectrolysis cell. By rearrangement of eq 17 LG(%c)E -L(%c)
‘la(%C)QPE/EA=
(13)
It is seen that the energy conversion efficiency, ‘lG,a, of the photoelectrolysis cell is made up of three terms, the first of which is governed by properties of the semiconductor anode, Le., the internal quantum efficiency and the undervoltage, t 0 , ~ 4 -~ CSC, / ~ 2 with respect to the reversible H,O/O, electrode while the remaining terms account for losses by hydrogen overvoltage, t M H+ < 0, and the ohmic drop, AQ > 0. d e first term of eq 13 can be determined by in situ photocalorimetry. In this technique the heat generated at the illuminated semiconductor electrode is monitored by a detector sensing K.4-8J0 changes in temperature of the order of some Thermistors in solution near the mirage-effect deflection of a laser beam directed parallel to the electrode ~ u r f a c e , ~ and a microphone6V8or a pyroelectric filmlo coupled to the metal support (M’ in Figure 1) of a thermally transparent semiconductor electrode assembly have been used for this end. The following analysis is an extension of that given previously.6 It applies to detection of the thermal signal transmitted through the metalsupported semiconductor electrode.
and comparison with eq 13, the energy conversion efficiency, ‘lG,a, can be expressed as
Evidently, the quantity &(ew) corresponds to the energy efficiency of photoelectrolysis of water in a cell employing a thermodynamically reversible hydrogen counterelectrode (EM = ~ o , H ~ / H + ) and a zero-resistance electrolyte solution (AcR = 0). &(esc) thus represents the upper limit to the energy efficiency attainable with a given semiconductor photoanode in a real cell. Determination of &(eSC) by in situ photocalorimetry is therefore considered a useful technique for testing the efficiency of photoelectrodes irrespective of the properties of the counterelectrode. As with other photothermal technique^,^,^ there is the additional advantage of obtaining & without the necessity of measuring the light intensity. In our previous work, the quantity -L(esc), eq 17, was taken as a measure for the energy efficiency of the single photoanodic r e a c t i ~ n . ~ , ~The - ’ ’ present analysis reveals that LG(tSC),eq 18,
The Journal of Physical Chemistry, Vol. 96, No. 11, 1992 4561
Energy Conversion by Photoelectrolysis of Water 0 2,
0.2,
,4
UIV 3
0
10.8
.
1 73
:op, I ,““ ‘
,
flame-oxidized pH 1 8 360 nm
0
05
10 E,,/
15
20
0.1. -02
0
0
0
I 25
V vs HESS
Figure 2. Photocurrent density, iph,photocalorimetric signal, P,and cell voltage, U,as a function of the potential, csc, of the n-Ti02 photoanode (rutile thin-film) for 360-nm photoelectrolysis of water at pH 1.8 (1 mW.cm-2 average incident light power during chopped illumination at
5.3 Hz). provides a thermodynamically well-defined expression for the energy efficiency of photoassisted electrolysis. Other determinations of the energy efficiency by photothermal methods4v5apparently did not take the power conversion loss by Joule heat evolved in the semiconductor electrode (process c in Figure 1) into account. This would give rise to an unrealistically large value of the energy efficiency, as discussed in the following section.
Results and Discussion The modulated photocurrent measured upon intermittent illumination of the TiO, thin-film electrode at pH l .8 (I mW.cm-2 average incident power at 360 nm) is shown in Figure 2 as a function of potential relative to that of the hydrogen electrode in the same solution (HESS). The photocurrent under continuous 2 mW-cm-2 illumination was twice as large. The dark current was negligibly small. The photocurrent onset potential is in good agreement with the measured flatband potential of 0.3 V (HESS), thus affording a maximum undervoltage of -0.9 V with respect to the potential of the reversible H20/02-electrode at 1.23 V (HESS). Flatband potentials reported for various TiOz electrodes fall between +0.35 and -0.3 V (HESS).14 The external quantum efficiency of the photocurrent, referring to incident photons, is 0.47at photocurrent saturation. Plateau values of the quantum efficiency for single-crystal TiO2 electrodes can vary between 0.1 and 0.815at and below 360 nm. Obviously, singlecrystal material is not necessarily superior to the polycrystalline T i 0 2 film. Also shown in Figure 2 are the signal, P(csc), of the heatmonitoring sensor (see eq 15) and the cell voltage, U,as a function of the potential of the TiOz photoanode. At zero photocurrent, P is proportional to the light energy absorbed by the TiOa electrode. Upon photocurrent onset, P diminishes since part of the light energy is converted into chemical energy by photoanodic water oxidation (process f i n Figure 1). As seen from eq 15, the minimum arises from the contrary effects to P(csc) of increasing photocurrent quantum efficiency and decreasing undervoltage for oxygen evolution. The minimum occurs at a potential cathodic with respect to that of the oxygen electrode, t 0 , ~ ~ 0=/ 01.23 ~ V (HESS), where the first term of eq 13 is positive as desired for energy conversion. The further increase of P(esc) in the plateau region of the photocurrent is effected by Joule heat from charge separation in the photoanode (process c in Figure 1). With the present setup a relative change in P of ca. 1% can be resolved. The photon energy absorbed during the illumination period of 0.2 s is 0.3 mJ considering the reflection loss of ca. 0.2 (see below). Hence, the sensitivity of the photocalorimetric measurement is ca. 3 pJ under the present conditions. Since a hydrogen counterelectrode was used in the electrolyte under study, the cell voltage, U,measured at zero photocurrent is identical to the potential eSc (vs HESS) of the TiO, electrode (14) Finklea, H.0. In Semiconductor Electrodes; Finklea, H. O., Ed.; Studies in Physical and Theoretical Chemistry, Vol. 55; Elsevier: Amsterdam,
1988. (15) Frank,
5
A
S.N.; Bard, A. J. J. Am. Chem. SOC.1977, 99,4667.
-0.11
0
v 0.5
\
1.0 E,
15 2.0 / V vs. HESS
I
I 25
Figure 3. Relative change, L (eq 16), of the photocalorimetric signal, , energy efficiency, & internal photocurrent quantum efficiency, T ~and (eq 18), for photoelectrolysis of water as a function of the potential of the n-Ti02photoanode (360 nm, pH 1.8). Broken curve: & for qn - csc step function (see text).
(broken line in Figure 2 ) . At finite photocurrent the hydrogen overvoltage and the voltage drop in the electrolyte contribute to the cell voltage. This contribution is primarily due to the voltage drop across the ceramic frit separating the anolyte from the catholyte ( 2 kQ in the electrolyte solution at pH 1.8), while the hydrogen overvoltage is less than 50 mV. Both these voltage drops diminish the energy conversion efficiency, eq 13, for the photoelectrolysis cell as a whole. The relative change, L(csc), of the sensor signal with respect to the value at the flatband potential (eq 16) and the internal quantum efficiency of the photocurrent, qa(csc), as determined from L(esc) by means of eq 17 are shown in Figure 3 for pH 1.8. The internal quantum efficiency approaches a plateau value, qa,s = 0.62. Considering the external quantum efficiency, qs = 0.47, the reflection loss, R = 1 - q/qa, is ca. 20% at the Ti02 electrode. Values of qa,s obtained previously by applying photoacoustic methods to photoanodic water oxidation at similarly prepared Ti02 thin-film electrodes are 0.47 (360 nm, pH 2)9 and 0.8 (325 nm, unbuffered neutral electrolyte).16 By photothermal spectroscopy, qaawas found to vary between 0.1 and 0.7 for a singlecrystal Ti02 electrode (340 nm, 1 M H2S04) depending on the degree of red~ction.~ The model of GBrtner” leads to the following relation between the quantum efficiency and the diffusion length, Lp, of minority carriers q8 = 1 - exp(-aW)/(l
+ aLP)
where CY is the optical absorption coefficient of the semiconductor and W is the depletion layer width given by
Nd and c are the donor density and the dielectric constant of the semiconductor, respectively, and eSC - cb is the band bending. Equation 20 has previously been applied for estimating Lp in various TiO, photoanodes.I2J8 In the present case, qa,s = 0.62 at a band bending of 2.2 V which corresponds to a depletion layer width of 50 nm for Nd = lo2’m-3. Taking CY = 6 X lo4cm-I for 360-nm light,I9 the hole diffusion length from eq 20 is 0.14pm. This value is 1 order of magnitude less than that reported for a single-crystal T i 0 2 electrode,12however 10-fold larger than the diffusion length estimated for a Ti02 thin film prepared by heating Ti in air at 600 OC.lEb Oxidation of Ti in the flame apparently produces T i 0 2 crystallites of much better quality for photoelectrochemical processes. In fact, the T i 0 2 electrode studied here meets the condition for efficient electron-hole separation,18a1/CY (16) Yoshihara, S.;Aruchamy, A.; Fujishima, A. Bull. Chem. SOC.Jpn. 1988,61, 1017. (17) Gartner, W. W. Phys. Rev. 1959, 116, 84. (18) (a) Salvador, P. Solar Energy Mater. 1982,6,241. (b) Ahlgren, W. L. J . Electrochem. SOC.1981, 128, 2133. (19) Eagles, D. M., Jr. Phys. Chem. Solids 1964, 25, 1243.
4562 The Journal of Physical Chemistry, Vol. 96, No. 11, 199'2
-< W + Lp,quite well. It is further noted that the optical penetration depth, l/a,of 0.17 gm is smaller than the thickness of the TiO, film as required for photocalorimetric measurements. The Peltier heat, QPE= -0.23 eV, derived from L(csc) at pH 1.8 is in good agreement with the value reported previously for a similar TiOz p h ~ t o a n o d e .From ~ L(tsc), qa(tsc), and QPEthe function LG(tsc)shown in Figure 3 was calculated by means of eq 18. As detailed in the previous section, & represents the upper limit of the energy conversion efficiency for photoelectrolysis of water attainable with the photoanode under study. As Seen from eq 18, the efficiency is controlled by both the potential-dependent internal quantum efficiency, qa(tsc), and the potential difference, t0,H20/02- csc, Hence there is a maximum in the conversion efficiency which, in the present case, corresponds to L G , , = 0.056 at tsc 0.76 V (HESS). At ESC t0,H20p2,& is zero. At this potential the energy conversion gain by photoassisted oxidation of water and concomitant reversible hydrogen production at the counterelectrode is compensated by the loss from conversion into heat of electrical energy supplied to the photoanode for transport of charge carriers. It is further seen that operation in the potential region of photocurrent saturation does not improve the energy efficiency, although the quantum efficiency is being maximized. In fact, & becomes negative at tSC> t0,H20/02. This means that the thermodynamic advantage of light-assisted production of oxygen from water at a potential cathodic with respect to ~ O , H ~ / O , is not exploited. However, for increased energy efficiency a more cathodic photocurrent onset potential and a steeper rise of the current-potential curve would be desirable. In their photothermal investigation of semiconductor photoelectrochemical cells, Fujishima et a1.4 defined the single electrode, monochromatic energy efficiency as =
EA- QSC/Nph.a
(22) Ex where EA is the photon energy, Qsc is the heat via recombination and other radiationless processes, and Nph,a is the number of photons absorbed (see eq 13 of ref 4). In a later analysisSit was shown that qe can be written as (23) q e = tla.s(/Eh - ERD/EA qe
Dohrmann and Schaaf
i 02
\, ,
06
C4
0.8
1C
,
,
,\1
12
14
E,,N vs HESS
Figure 4. Energy conversion efficiencies & (eq IS), vog (eq 4), and ~ 0 (eq 24) for photoelectrolysis of water as a function of the potential of the
,
~
n-TiO, photoanode.
-
\G 0
a,
6 01
3
2
4
I
8 1 0 1 2 1 4
6 PH
Figure 5. Dependence on pH of maximum energy efficiencies obtained for 360-nm photoelectrolysis of water employing a n-Ti02 thin-film photoanode and a Pt cathode: &, efficiency, from photocalorimetry, for zero internal resistance and reversible cathode reaction, eqs 18 and 19; t ~ ~ efficiency , ~ , including loss due to electrolyte (frit) resistance and cathodic overvoltage, eqs 4 and 19; qG,O,efficiency from I,, - U char~ to the acteristic and power of incident light, eq 24. (& and V G , refer power of absorbed light.) The bars indicate the typical experimental error.
where Eh is the energy level corresponding to the flatband potential is zero. At this potential the additional losses (distance AB in and ER is that of the redox system. Apparently, the energy loss Figure 4) are given by the second term of eq 19 for qa (0.87 V) associated with Joule heat, qa(tsc)leol(esc - cw), evolved in the = 0.44 (see Figure 3). Solving this term for (Q,H?/H+ - eM) + A ~ R semiconductor upon flow of photocurrent (process c in Figure l), gives 0.4 V corresponding to distance BC in Figure 4. has not been taken into account. On the basis of eq 22, an energy For further comparison, the energy efficiency qG,O,defined by efficiency of 0.37 has been reported for 340-nm photoanodic oxidation of water in 1 M H2S04at a single-crystal TiOz electrode? which is much larger than the efficiency &,,, of ca. 0.05 determined in the present work. The discrepancy seems to derive mainly from the neglect of the energy conversion loss by Joule is shown in Figure 4. 4o is the power of the incident light, and heat in eqs 22 and 23, which is not justified in view of the energy the remaining quantities have the same meaning as in eq 4. The balance analysis given here. In terms of eq 18, qc is identical to efficiency, eq 24, is widely used in photoele~trochemistry.~~~~~~~ the maximum value of LG for a photocurrent-potential characqG,owas measured potentiostatically upon continuous illumination. teristic defined by the step function qa(tSC< em) = 0,va(tsc L The values of qG,oare smaller than those of qG,aas expected from era) = For such an idealized characteristic, L G = the reflection loss of ca. 20% (see above). qa.sleOl(eO,H,O 0 2 fSC)/EA and &,m corresponds to ~ a . ~ ~ e O l ( ~ O , H ~ 0 / 0 ~ The various energy efficiency data were determined in the pH - tfi)/EA. dhis is illustrated by the dashed curve in Figure 3. In range 0.5-13.5. The maximum values are shown in Figure 5 as the limiting case of no band bending, tSC= th, no Joule heat is a function of pH. &,,, rises slightly with pH and reaches the generated in the semiconductor. On replacing IEm- ERI in eq largest value near p H 12. A similar tendency was observed fdr - th) it is seen that qe = &,,, for the idealized 23 by le01(t0,H20/o, the pH dependence of the internal quantum efficiency at constant qa - tSCcharacteristic defined above. band bending of below 1 V. It is known that oxidation of hyThe energy conversion efficiency, qG,a,defined for the phodroxide is favored kinetically and thermodynamically over that toelectrolysis cell by eq 4,is shown in Figure 4 as a function of reflects of water at TiO, electrodes.143 The energy efficiency km the potential of the TiOZanode. qG,a was determined from the this effect. &,,,represents the maximum efficiency attainable cell voltage U (Figure 2) and the internal quantum efficiency with the present TiO, thin-film electrode for energy conversion (Figure 3) making use of the relation I h/4a= qaleol/EA.Comby 360-nm photoelectrolysis of water. However, this efficiency parison of qG,awith LG demonstrates tke additional conversion can only be approached in conjunction with a thermodynamically losses associated with hydrogen overvoltage and ohmic resistance (see eq 19). No attempt has been made to minimize these losses (20) (a) Desplat, J.-L.J. Appl. Phys. 1976, 47, 5102. (b) Salvador, P. J . by platinizing the counterelectrode and using a low-resistance Elecrrochem. Soc. 1981, 128, 1895. (c) Tributsch, H. J . Electrochem. Soc. diaphragm. The maximum value of qG,ais 0.028. At the cell 1981. 128, 1261. (d) Norton, A. P.; Bernasek, S. L.; Bocarsly, A. B. J . Phys. voltage of 1.23 V, corresponding to tSC= 0.87 V (HESS), ) I ~ , ~ Chem. 1988, 92, 6009.
J. Phys. Chem. 1992,96, 4563-4567 reversible hydrogen electrode in a cell having zero ohmic resistance. With the cell employed, at least 50% of the maximum possible efficiency was exploited as seen from the ratio of ) I ~to, ~ L.G.The gap between these two efficiencies is smaller in strongly acid and alkaline solution where hydrogen overvoltage and ohmic resistance are smallest. In a comparison of the maximum values of & and qG,+ it should be recalled that in the whole pH range the anode potential corresponding to the maximum of qGa was more cathodic than that pertaining to the maximum of LG (Figure 4). This implies that at the optimum energy conversion point of the cell the internal quantum efficiency was smaller than under the condition of determining the maximum of LG. Nozik2' expressed the energy efficiency for photoelectrolysis of water by a relation which can be written as
AoHfF@is the molar heat of combustion of H2and the remaining quantities are as defined above. Values of qHp obtained by 20-mW broad-band (300-400-nm) photolysis employing a single-crystal TiOz anode and a platinized Pt counterelectrode were ca. ~
~
~~
(21) Nozik, A. J. Nature (London) 1975, 257, 383.
4563
0.03-0.04, independent of pHa2' The efficiencies qH,oand qG,o, eq 24, are connected by qH,O
= %,o -
Iph 0
TAoSf,H20
2F
(26)
where AoSf+H20 is the standard entropy change for formation of H20(1),-163 J.mo1-'SK-'. Taking vG,o= 0.02 at pH 1.8 from the present study, qH,ois 0.03, close to Nozik's values.
Conclusion In situ photocalorimetry employing pyroelectric monitoring of heat evolved at a semiconductor photoelectrode is a technique useful for studying the energy efficiency of photoelectrolysis with respect to absorbed radiant power. As with other photothermal methods, no additional measurement of the light intensity is required. Both the maximum possible energy efficiency attainable with the photoelectrode in conjunction with a thermodynamically reversible counterelectrode and the efficiency for a cell with additional losses from overvoltage at the counterelectrode and from electrolyte resistance can be determined. Acknowledgment. Support of our work by Deutsche Forschungsgemeinschaft and, in part, Fonds der Chemischen Industrie is gratefully acknowledged.
Infrared Spectroscopy as a Probe of the Adsorption and Electrooxidation of a Cyanide Monolayer at Platinum under Aqueous Electrochemical Conditions Vicki Berger Paulissen and Carol Korzeniewski* Department of Chemistry, The University of Michigan, Ann Arbor, Michigan 481 09 (Received: November 5, 1991; In Final Form: January 28, 1992)
Adsorbed cyanide was observed on platinum electrodes using infrared spectroscopy. In situ spectral analyses were performed in a cyanide-free electrolyte solution following cyanide monolayer preparation according to an electrochemical procedure developed for earlier LEED and AES studies of cyanide adsorption on platinum electrodes. This approach minimized spectral interferences associated with reactive cyanide species in bulk solution and simplified the assignment of bands associated with surface species. Two forms of the adsorbed cyanide surface layer were identified. With the electrolyte solution at pH 12.5, a potential-dependent band appears at about 2058 cm-I and is assigned to the vibrations of a surface layer of CN-. With the electrolyte solution at pH 8, a potential-dependent band appears at about 2147 cm-I and is assigned to a surface cyanide layer in which CNH is the predominate species.
Introduction In recent years, considerable progress has been made in understanding molecular adsorption at the electrode/solution interface.' Electrochemical and spectroscopic methods have provided valuable insights into the effect of solvent, electrolyte, and electrode potential on adsorbate structure and bonding. Infrared spectroscopy is recognized as a valuable tool for examining these interfacial interactions, as recent in situ studies of molecular adsorption and electrooxidation on electrodes of well-defined surface structure will confirma2 The present work employs infrared spectroscopy to examine the adsorption of cyanide on platinum electrodes. The majority of vibrational spectroscopic studies concerning cyanide electrosorption have employed silver and gold electrodes. Early studies utilized surface-enhanced Raman spectroscopy (SERS)? and later work employed infrared ~pectroscopy.~These vibrational studies assign a potential-dependent band centered at about 2100 cm-'to the *N stretching mode of cyanide adsorbed Linearly through the carbon atom and report vibrational band shifts
* To whom correspondence should be addressed.
with applied potential in the range 30-35 cm-'/V. Additional bands are observed in the infrared spectra and have been assigned to solution-free cyanide and metal-cyanide complexes formed by dissolution of the electrode surface upon oxidation.s Fewer vibrational studies have been reported for cyanide adsorption on platinum electrodes. It appears that spectral features of the adsorbed ion are obscured by products of the cyanide oxidation reaction,6 and as a result, vibrational assignments of (1) (a) Hubbard, A. T. Chem. Reu. 1988, 88, 633. (b) Hubbard, A. T. Acc. Chem. Res. 1980, 13, 177. (2) Chang, S.;Weaver, M. J. J. Phys. Chem. 1991, 95, 5391. (3) (a) Kunimatsu, K.; Seki, H.; Golden, W. G.; Gordon 11, J. G.; Philpott, M. R. SurJ. Sci. 1985, 158, 596. (b) Kunimatsu, K.; Seki, H.; Golden, W. G.; Gordon 11, J. G.; Philpott, M. R. Lungmuir 1988, 4, 337. (c) Gao, P.; Weaver, M. J. J. Phys. Chem. 1989, 93, 6205. (4) (a) Kunimatsu, K.;Seki, H.; Golden, W. G. Chem. Phys. Len. 1984, 108, 195. (b) Corrigan, D. S.; Gao, P.; h u n g , L. H.; Weaver, M. J. Lungmuir 1986, 2, 744. (c) Seki, H. Electrochemical Surface Science; Soriaga, M . P., Ed.; American Chemical Society: Washington, DC, 1988; Vol. 378, Chapter 22. ( 5 ) Jones, L. H.; Penneman, R. A. J. Chem. Phys. 1954, 22, 965. (6) Hinman, A. S.; Kydd, R. A.; Cooney, R. P. J. Chem. SOL.,Faraday Trans I 1986, 82, 3525.
0022-365419212096-4563%03.00/0 0 1992 American Chemical Society
