Photoelectrochemistry: Inorganic Photochemistry Mark S. Wrighton Massachusetts Institute of Technology, Cambridge, M A 02139 Electron-transfer reactions can he induced by the photoexcitation of inorganic substances. Such reactions are presently of great technological importance in photography. Research during the past decade has established that inorganic photoredox processes may be important in the generation of chemical fuels using sunlight as the driving force to convert low-eneigy starting materials, such as Hz0, to highenergy product mixtures, such as Hz 11202 from HzO, eqn. (1) (1).
+
runlisht
Hz0
H2 + 1/20:,
(1)
maximum power point, somewhere between open-circuit (i = 0; E v = Ev(oc)) and short-circuit (i = is,; E v = 0). The maximum value of 7,vmax, occurs a t the maximum power point. Photogalvanic cells for the conversion of light to electricity less than 1%for the conversion of have small values of sunlight. There is also little prospect for improving such devices so that their efficiencies are sufficiently high to make them practical. The basis for the low efficiency associated with photogalvanic cells can be appreciated from consideration of a specific example. The sequence represented by eqns. (3) and (4)
While solar enerrv is not the only application of inorganic photoredox proc&ses, the potential f& long-term impact is so great that an understanding of the state-of-the-science is important. At the present time, the demonstrated efficiency for the direct conversion of solar to stored chemical energy exceeds 10% using devices called photoelectrochemical cells. This efficiency is sufficiently high that amajor energy technology based on such devices cannot he ruled out. However, now the devices are far too expensive to he useful in practical chemical energy generation from sunlight. A consideration of the principles associated with photoelectrochemistry will reveal where major progress has been made and where the major difficulties lie. Photogalvanic Cells In one type of photoelectrochemical cell, species dissolved in an electrolvte solution are the lieht absorbers. Such devices are called phbtogaluanic cells (2).% a photogalvanic cell the absorotion of light bv some s ~ e c i ein s the solution can chance the distrihution'of rcdox-artiie marerids in thr vicinity of uk. of tht. elrrtrodw. That is, n change in 111erariool'redux-ac~i\r material; attay irtm the thenno&mami~.equilibrium position is oosslbk because the light i~hsorbedcreates rxrited species that an.generally more potent oxidants and more potent rrductant* than are the ground d a t e s~xwcs.The l~ghr-mdurcd change in the ratio ofredox materLals in the vicinity of one electrode compared to the other electrode gives rise to a potential difference between the two electrodes a t open-circuit, Ev(oc), assuming that the electrodes equilibrate with the redox materials in solution. The value of Ev(oc) depends on the extent to which the ratio of redox materials can he changed by the light. At open-circuit there can he no electrical power from the device, since no current, i, passes at open-circuit and electrical power from the device is given by output photovoltage times the photocurrent, Ev X i. The objective is to obtain an efficiency, 7,as high as possible where 7 in per cent is given by eqn. (2). (Ev X i) X 100% (2) ?= Input Optical Power At short-circuit the value of i is finite, is,, hut the power from the device is again zero since this situation corresponds to E v = 0 because two electrodes shorted together are a t the same potential. Thus, for some electrical load in series with the two electrodes there is a maximum value of E v X i. The point on a plot of i versus E v where E v X i is maximum is called the
is a situation where the [Ru(bpy)32+]*excited species reduces the Fe(OHz)fi3+and becomes oxidized to form ground state Ru(hpy)?++. The R ~ ( h p y ) , ~ + / F e ( O H ~photoredox )6~+ products represent high-energy products that could be useful in a photogalvanic cell. However, the back reaction, eqn. (5), occurs very rapidly in homogeneous solution. When such hack reaction occurs, both i and E v are small leading to low power output from the cell. In homogeneous solution it is difficult to inhibit the back reaction of the energetic redox products formed photochemically. Unless this problem can be overcome, photogalvanic cells will not become efficient. Semiconductor-Based Photoelectrochemical Cells The second type of photoelectrochemical devices are those where one, or both, of the electrodes absorbs the input optical power, Figure 1( I ) . Since a solid material is the light absorber, such photoelectrochemical cells are a kind of photovoltaic device. Indeed, the first photovoltaic device was a photoelectrochemical cell and was discovered over a century ago (3). The most efficient photoelectrochemical cells are those that employ semiconductors as the light absorbing electrode maexterials. The key reason for the ability to achieve an,,,q ceeding 10% for conversion of solar power is that there is a mechanism for inhibiting the back reaction of energetic redox products at a semiconductorfliquid electrolyte interface. Simply immersing a semiconductor electrode into an electrolvte solution can result in a strong electric field across the semknduru,r result~ngin a region,ear the surface exposed to the liauid, where ~hotoexcited&ctrons, r .and holes, ht. are literally driven in opposite directions, ~ i ~ u2.r For e n-type semiconductors the majority charge carriers are electrons, and holes are not available in the dark, whereas for p-type semiconductors holes are the majority charge carrier and electrons are not available in the dark. In both cases, the desired situation occurs when the electrolyte solution is such that the majority charge carrier is driven away from the solution. The minority carrier of the semiconductor, holes for n-type and electrons for p-type, can be created hy absorption of photons Volume 60 Number 10 October 1983
877
Figure 2. Energy diagrams for n-type (top) and p-type (bottom) semiconductors in contact with an liquid electrolyte having an electrochemical potential. Ere= The energetics are shown at different electrode potentials, E,, and illumination conditions. Excitation of ec-h+ pairs in the band bending region gives charge separation in the sense indicated resulting in the minority carrier being driven to the interface. At high light intensityand open-circuit, the bands are flattened and the maximum Ev(oc) is obtained as shown at the right.
Figure 1. Semiconductor-based cells for the conversion of light to electricity (a) (from reference (7))or to chemical energy in the form of H2and O2(b) (from reference (9)).
sufficient in energy to raise an electron from the valence band to the conduction band of the semiconductor. This energy is the optical band gap, E,, of the semiconductor. As illustrated in Figure 2, the photogenerated minority carrier is driven to the semiconductorAiquid electrolyte interface where the minoritv carrier is available to effect some redox process. The electrons have the reducing power associated withthe position of the bottom of the conduction band,Em, and holes have the oxidizing power associated with the top of the valence band, E ~ BWith . reference to Figure 2, the noteworthy finding is that whenever there is greater than -0.3 V of band bending, the minoritv carrier redox reaction can compete almost completely successfully with all processes that result in net recombination of the electron-hole pairs created by light. Recombination is the analogue of back reaction in homogeneous solution that limits the efficiency of photogalvanic cells. The point is that, whenever the field across the semiconductors is sufficiently strong, there can be high quantum efficiency for electron flow in the external circuit to give a good photocurrent. i . Quantum efficiency here refers to the number of electronspassing in the external circuit per photon absorbed. For mod semiconductor photoelectrochemical cells the quantum efficiency is approximately one. As noted above, for photogalvanic cells a good value of i does not insure a good val"e of 1),since the power from a device is given by i X Ev. For the semiconductor-based cells the maximum value of Ev(oc) is given by the so-called barrier height, EB. The value of EB is the difference between Ed,,, the electrochemical potential of the solution, and ECBor EVBfor n- or p-type semiconductors, respectively. An objective is to make E R as close as possible to Egin magnitude. In principle, this can be accomplished by using very negative (reducing) redox couples in contact with p-type semiconductors or by using very positive (oxidizing) redox couples in contact with n-type semiconductors (4). The variation of Eredo. for a given semiconductor should yield a variation in the maximum value of Ev(oc). When a cell is actually being used to convert optical power, the electrode potential of the semiconductor, Ef, is not a t the value associated with maximum Ev(oc). When maxi878
Journal of Chemical Education
Figure 3. Current-voltage curves for a reversible electrode in the dark compared to illuminated n- and p-type semiconductor electrodes in an electrolyte solution containina redox active A and At. The extent to which the onset of anodic current
The photoelectrodes give a plateau in me current because the cunent becomes limited by the rate of exciting electrons (light intensity)once the band bending exceeds approximately 0.3V. Thus, the photocurrent is independentof potential by approximately 0 3 V beyond the onset of photoanodic or photocamodic current for n- or p- type semiconductors. respectively.
mum Ev(oc) is realized (high light intensity, open-circuit), the bands of the semiconductor are flattened and Ef is a t the socalled flat-band potential, EFB. There is no driving force separating electron-hole pairs when Er = EFBand electronhole pairs recombine giving no photocurrent. This results in the degradation of optical energy to heat. However, as the resistance between the photoelectrode and counterelectrode is lowered from the infinite resistance associated with opencircuit, band bending can occur and the quantum efficiency can approach unity when the band bending is -0.3 V. Whenever there is photooxidation current a t a more negative electrode potential than Eredo. a t an n-type semiconductor or photoreduction current a t a more positive electrode potential than Eredoxa t a p-type semiconductor, the value of 7 is positive. The extent to which E i is more negative or positive is the output photovoltage, Ey. Figure 3 summathan ETedo. rizes the situation for photoanodes (n-type semiconductors) and photocathodes (p-type semiconductors) compared to a reversible (metallic) electrode for a reversible redox couple A+/A.
e
i
rEv1 YVVV
ELECTRICAL LOAD
Figure 4. Circuit and energy diagram tor a semiconductor-basedphatoeiectrochemical call for the conversion of light toelectricityas in the cell sketched in Figure la.
-
Focusina on the ~hotoanode,the kev is that a light-induced oxidation process, A A+, can occur when Eris such that A+ A should, in fact, occur in the dark. That is, the electrons ofthe srmiconductr~rhave tht*thernicdynamic wducingpower hdes needed tu redurr .I- to .4. Hwvwer, the ~hoto~enc~rated have the potency to oxidize A A+. he hole(minority carrier) process can he successful because the electrons are inhibited from reacting with the A+ created by the holes. The inhibition comes about because the electrons are driven away from the surface. The electrons would have to go over a harrier or tunnel through it in order to reduce A+ to A. The barrier is too large when the band bending is 20.3 V. Recall that the hand bending results as a consequence of immersing the semiconductor in an appropriate solution. There is no analogous effect associated with molecules in solutions. I t is imto note that the hand bendine tends to inhibit the nortant prompt recombination of electron-hole pairs created by ahsorption of a photon. However, this is not the crucial aspect of the semiconductor band bending, since photoexcited molecules can be completely efficient with respect t o the quenching of the excited state to produce energetic products as in eqn. (4) (5).The point is that once the energetic products are made they are not inhibited from hack reacting as a t the surface of the semiconductor: the excited electron of the semir~mductorisclriven away from Ar, lnhlbiting hack reaction. T h r ~hoturacltedelectron result. In current flow, and reductionbf A+ can occur a t the counterelectrode (see Fig.
-
Electmltye
n-~fliO~+~queour
-
.
4).
In terms of sultlr energy conversion, semicondurtor-based ohotut+?ctrochtmical cells should he 3l1le to eiw :I solar eifi&ency of -30% when one of the electrodes is aphotoelectrode (6). The recent literature reports efficiencies in the vicinity of 15% for the solar conversion of light to electricity or to chemical fuel in the form of isolated redox products (7.8). In calculating the efficiency for a solar conve;sion, eqn. (2) can be used where Ev is the output photovoltage relative to the Errdoxfor the half-cell reaction a t the photoelectrode. In fuel-forming reactions the half-cell occurring at one electrode is not necessarily the reverse of that occurring at the other electrode. For example, in splitting water according to eqn. (1) 0 2 is evolved a t the anode and H2 a t the cathode. In a photoelectrochemical cell for electricity generation, the half-cell reaction at one electrode is the reverse of what occurs a t the other as illustrated in Figures 1and 4. The high efficiencies obtained establish that semiconductor-hased photoelectrochemical cells represent the best chemical-based systems for the direct conversion of light to electricity or chemical fuel. However, there is still considerable room for improvement and costs must be lowered if practical energy conversion systems are to he realized. A consideration of some of the problems associated with semiconductor photoelectrodes will now he given using specific examples.
3.
P t Cathode
Wavelength Response and Stability of Photoelectrodes In 1976, the photoassisted oxidation and reduction of Hz0 according to i n . ( 1 ) wab reported using an ,I-type Srl'iOJ photoanude-hssed cell. Figure 5 (91 interestia~ly,the (lrvice ctn~ldbe used 1 8 , yenemtv H: and 0. without :tnv e n e r p input other than thv light to i.xc.tte the vlectrons in the SrTiO ,. 'l'he wnvdeneths of'lieht that will be e f k t i v e a r t those that will excite el&ons f r k the valence hand to the conduction hand. That is, only photons of energy 2E, = 3.2 eV will he effective for SrTiOs. Unfortunately, the photons of 23.2 eV in energy corresnond to ultraviolet lieht.. a repion .. of the solar smctrum that comprises
