Pho tocurrent Spectroscopy of Anodic Oxide Films on Titanium BY JEROME F. MCALEER AND LAURENCE M. PETER Department of Chemistry, The University, Southampton SO9 5NH Received 2nd May, 1980 Photocurrent spectroscopy has been used to examine the early stages of anodic film growth on titanium. Analysis of the photocurrent conversion efficiency as a function of wavelength for films formed at different potentials has shown that the film is essentially pure TiO, above ca. 1.5 V us. SCE (at pH 0). Photocurrent spectroscopy has also been used to follow changes in the structure and thickness of the anodic film during breakdown at higher voltages. The recent upsurge of interest in the photoelectrochemistry of semiconductors owes its origin to the present search for alternative forms of energy conversion, and our understanding of the properties of the semiconductor-electrolyte interface is now advancing rapidly as the potential usefulness of a variety of semiconducting materials is investigated.Although the great majority of recent experimental work has been oriented towards solar energy conversion, at least as a distant goal, it is clear that the concepts and methodology of photoelectrochemistry are more widely applicable. Photocurrent spectroscopy is a powerful in situ method which can be used to study the solid-state and dielectric properties of anodic films on metal~.l-~ These anodic films are important; they can play a role in corrosion and catalysis for instance, and if they are semiconductors they can be characterised by photoelectrochemical methods.In a detailed investigation of the formation of anodic films on titani~rn,~~’ we have combined photocurrent spectroscopy with more conventional electrochemical and optical techniques in an attempt to characterise the passivation process and the properties of the protective oxide layers. The results of these measurements have led to a clearer understanding of the solid-state properties of the anodic films formed on titanium under different electrochemical conditions, and at the same time have helped to clarify the mechanisms of film breakd~wn.~.’ It is appropriate to consider only the photoelectrochemical measurements here; further details of other electro- chemical and optical experiments are presented el~ewhere.~.~ EXPERIMENTAL Titanium disc electrodes, 0.5 cm2 in geometric area, were machined from commercial purity metal (IMI 130) provided by IMI Titanium.The discs were embedded into “ Kel-F ” holders with epoxy resin and mounted on a Teflon rod with an O-ring seal. The electrodes were ground flat and then polished on successively finer grades of diamond. A fresh surface was generated for each experiment by polishing the electrode on 1 pm diamond, followed by 0.05 pm alumina. For the measurements at potentials above 6 V us. SCE, the electrodes were rotated at a speed sufficient to prevent bubble formation during oxygen evolution. Measure-68 PHOTOCURRENT SPECTROSCOPY OF ANODIC FILMS ments were carried out in 1 mol dmV3 H2S04 and 1 mol dm-3 H3P04 solutions prepared from AnalaR grade chemicals and triply distilled water.Photocurrents were excited by chopped monochromatic light and detected by a lock-in amplifier. Absolute conversion efficiencies (electrons per incident photon) were measured using the photocurrent response recorded directly at a chopping rate t l Hz. The power of the incident radiation, which was measured with a calibrated RCA 935 photodiode, was in the range 2-50 pW in the region 300-400 nm. The electrochemical instrumentation was con- ventional, except that a 100 V/1 A potentiostat was modified as a programmable voltage source for measurements in the range 6-60 V. RESULTS GROWTH OF THIN OXIDE FILMS It is convenient to define "thin" oxide films in the present context as those formed on titanium at potentials < 6 V us. SCE (at pH 0).These films are < 30 nm thick, and they behave more simply than films formed at higher potentials. Titanium can be classified as a valve metal6 and in many ways it resembles in its behaviour metals like aluminium and tantalum. Although oxide films on titanium are generally less stable than their counterparts on the other valve metals, there is no doubt that the growth of insulating oxide takes place by the field-assisted migration of ions or vacancies. Titanium differs from other valve metals in the way in which the oxide film is apparently able to recrystallise at low over voltage^,^ and the breakdown of the film is important in applications to corrosion protection and electrocatalysis. The current efficiency for film formation can be less than unity, and this may go some way to explain the wide range of the characteristic growth parameters which have been reported in the l i t e r a t ~ r e .~ Our own measurement^^*^ have shown that za, the product of ion-charge and half-jump distance, is 0.9 nm for films grown in the range 0-2 V us. SCE in 1 mol dmW3 H2S04. The exchange current, i,,, for the injection of ions at the metal-oxide interface was found from galvanostatic and potentiodynamic measurements in the same potential range to be 6 x lo-" A cm-2, in reasonable agreement with the results of Ahrens et aL8 The interpretation of electrochemical measurements of oxide film growth is not easy. Calculation of the film thickness requires knowledge of the film stoichiometry, current efficiency, the density of the deposit and surface roughness.Similarly, evaluation of capacitance data is only possible if both the relative permittivity and surface roughness are known. For this reason, direct optical methods for the deter- mination of films thickness are valuable. Interferometry has been used for thick oxide films on t i t a n i ~ m , ~ * ~ whereas ellipsometry1°-12 can be used even for very thin films. Careful analysis of photocurrent data provides an alternative way in which to determine the thickness of the oxide layer, while at the same time providing informa- tion about solid-state properties. Ti02 is an insulating or semiconducting material, which occurs widely ip nature and can crystallise in three non-cubic forms, anatase, brookite and rutile. The photoelectrochemistry of reduced specimens of the crystal form rutile has been widely studied.The material is an n-type semiconductor with a band-gap of 3.06 eV, and it shows a characteristic anodic photocurrent at potentials more positive than Efb, the flat-band potential. Anodic films on titanium are also p h o t o ~ e n s i t i v e , ~ ~ ~ ' ~ ~ ~ although no detailed quantitative study of their photoelectrochemistry appears to have been made. Fig. 1 shows a cyclic voltammogram and the corresponding photocurrent measured during the growth of the oxide film on titanium at 100 mV s-l in H,P04. The photocurrent rises linearly with potential beyond 0.5 V vs.J . F . MCALEER A N D L . M. PETER 69 SCE, and on the reverse sweep the photocurrent falls again almost linearly with potential, although a tail can now be seen which extends as far as -0.3 V us.SCE. The observation that the photocurrent increased linearly with potential immediately suggested that the film thickens at a uniform rate, as would be expected for growth by the high field migration of ions or vacancies. On the other hand, the almost linear decrease of photocurrent with potential on the reverse scan is less readily understood, since an approximately square-root dependence of photocurrent on potential might be expected for a highly doped semiconductor. We have established that the properties of the anodic oxide film on titanium depend greatly on the rate at which the film is grown. The dielectric behaviour and donor distribution of films grown at sweep I -i -0 i 2 3 i 5 E/Vvs.SCE FIG. 1 .-Cyclic voltammogram (left-hand scale) and corresponding photocurrent iph for a titanium electrode in 1 mol dm-3 H3P04. Excitation wavelength 335 nm. Sweep rate 100 mV s-I. rates above 5 mV s-l are anornalo~s,~*~ and for the present we shall therefore focus attention on the properties of films grown at lower sweep rates (1 mV s-I). The photocurrent response observed during the slow potentiodynamic formation of the anodic film is shown in fig. 2. Two points are immediately striking. First, the increase of photocurrent with potential on the forward scan is divided into two linear regions, with a change of slope at ca. 1.5 V us. SCE. Secondly, the photocurrent on the reverse potential scan has the more familiar shape observed with single crystal semiconductor electrodes.We consider first the photocurrent during growth. At the high fields which are needed to overcome the potential energy barriers for ion migration (>lo6 V cm-I), the efficiency of charge-carrier separation must be essen- tially unity since the characteristic thickness of the space-charge region, the Schottky length, L,, is greater than the film thickness itself. Even if space-charge is present, the field at the metal-oxide contact must still be large enough to move ions across the70 PHOTOCURRENT SPECTROSCOPY OF ANODIC FILMS interface. Under these conditions, the whole of the anodic film is active in the photo- electrochemical conversion process, and if reflection at the metal-oxide contact is taken into account, it follows that the conversion efficiency, @, is given simply by @A = 1 - (exp - 2ccALf), (1) where t c ~ is the absorption coefficient of the oxide at the wavelength 1 and Lf is the film thickness.For sufficiently thin films, the exponential term may be linearised, and a linear dependence of @ on Lf is then predicted. In order to use the observed variation of @ with potential during growth of the film to derive the thickness potential relationship, the appropriate value of CCA is needed. -1 0 1 2 3 4 5 E/V us. SCE FIG. 2.-Photocurrent measured for a titanium electrode in 1 mol dmW3 H3PO4 during a cyclic sweep at 1 mV s-l. Excitation wavelength 335 nm. The reliability of this approach to the determination of film thickness was tested in the following way. First of all, a series of photocurrent excitation spectra were measured under the " steady-state " conditions obtained by holding the electrode at a given potential for the 10 min.Fig. 3 shows a set of photocurrent spectra converted to give the wavelength dependence of the conversion efficiency. These spectra all show a smooth increase of @ with photon energy, and the absence of any reduction of efficiency at short wavelengths suggests that surface recombination is negligible under these conditions. These conversion efficiencies were then used to construct the plots of --In (1 - 0) against electrode potential shown in fig. 4. Provided that is independent of film thickness, these plots should provide a direct measure of Lf as a function of potential. The absorption spectrum of the anodic oxide film is unknown, but Mollers et aZ.13 have reported the absorption spectrum of CVD films of TiOz, and we have used their values of aA to convert the data in fig.4 into a plot of Lf against electrode potential. The success of the method is evident in fig. 5 , which shows that the data obtained at different wavelengths fall onto a common straight-lineJ . F. MCALEER AND L. M. PETER 71 hlnm FIG. 3.-Spectral dependence of the photocurrent conversion efficiency, Q,, at a titanium electrode in 1 mol dmW3 H2SOo. The electrode potentials were (vs. SCE): (1) 1.0, (2) 1.5, (3) 2, (4) 3, ( 5 ) 6 V. Slit resolution 9 nm. plot which exhibits the same change of slope seen in the linear sweep measurement of photocurrent (fig. 2). The relationship between film thickness and voltage is often referred to as the " anodising ratio ".The concept arises from the inverse logarithmic law for high- field oxide growth,6 since it can be shown that the rate of thickening at constant poten- tial becomes negligible in a fairly short time. Reported anodising ratios for titanium lie in the range 1-5 nm V-l; the rather large scatter probably reflects the different techniques and assumptions used. The analysis of the photocurrent data in fig. 5 O.1 0 2 4 6 FIG. E/V us. SCE 4.-Treatment of the photocurrent spectra in fig. 3 according to eqn( 1). The wavelengths chosen for analysis are (a) 320, (b) 330, (c) 340 and (d) 350 nm.72 PHOTOCURRENT SPECTROSCOPY OF ANODIC FILMS gives an anodising ratio of 6 nm V’I up to 2 V and 3.6 nm V-l above 2 V. These results may be contrasted with the values calculated from fig.2, viz. 6.4 nm V-I below 1.5 V and 3.3 nm V-I from 1.5 to 5.4 V. We have also measured “ steady-state ” anodising ratios by ellipsometry using McCrackin’s program.14 For 10 V films, an anodising ratio of 3 nm V-l was found. Dynamic anodising ratios under linear scan conditions were also calculated from the za and io values obtained for the growth of the oxide in the potential region from 0 to 2 V us. SCE. At 1 mV s-’, the calculated a E/V us. SCE FIG. 5.-Apparent film thickness as a function of electrode potential. The plot was constructed from the data shown in fig. 4, using the absorption coefficients for TiOz given in ref. (13). Eqn (1) shows that the axis parameter is equivalent to the film thickness, Lr.Wavelengths as as follows: 0, 330; +, 340; A, 350 nm. anodising ratio was 3.5 nm V-l, very close to the value obtained from the photo- current measurements at potentials above ca. 2 V us. SCE. Rather lower values around 2.5 nm V-’ have been measured by reflectance and ellip~ometry,’-~~ but this is not surprising since the data were obtained at higher sweep rates than those used here. Since we have established 4*5 that the electron-donor distribution is non- uniform, it is not easy to relate measurements made at different sweep rates. A change in the properties of the film was also seen when the geometric capacity of the film was measured during growth at 1 mV s-l, although in this case the change in shape occurred at a lower potential (fig. 6). A problem arises immediately when fig.5 and 6 are compared. Whereas fig. 5 suggests that the film thickness is zero at 0.5 V vs. SCE, it is clear from fig. 6 that the inverse capacitance relationship is obeyed down to at least 0 V; clear evidence that the film is already growing at this potential. It is difficult to establish exactly where film growth commences, since a natural oxide film is already present at the beginning of the electrochemical experiment. We can conclude from the comparison of fig. 5 and 6 that the first few monolayers of oxide on titanium do not give rise to a photocurrent. There are two possible reasons for this. First, if the oxide is very thin, photo-excited carriers will be quenched by electronJ . F . MCALEER AND L . M. PETER 73 exchange with the metal.This mechanism is probably restricted to the first one or two monolayers of oxide. Secondly, the film formed at low overpotentials may not be TiO, at all. X.P.S. and Auger of the anodic film on titanium have sug- gested that the oxygen/titanium ratio in films formed below 2 V is only 1.6-1.7. The oxide film formed in air appears to be close in stoichiometry to TiO, whereas at low potentials Ti203 appears to be formed.” The photocurrent results suggest that a photo-inactive oxide grows on the titanium electrode at potentials less than 0.5 V us. SCE. This oxide appears to behave as a normal dielectric, since the plots of inverse capacity are linear. The unusually high value of the anodising ratio obtained from the analysis of the photocurrent below 2 V can be explained if it is assumed that growth is taking place at the surface between the lower oxide and the Ti02 film -1 0 1 2 3 4 5 E/V us.SCE FIG. 6.--Inverse capacitance of a titanium electrode measured during growth of the oxide film at 1 mV s-’ in 1 mol dm-j H3P04. Frequency 110 Hz. as well as at the TiO,/electrolyte interface. The TiO, film could then grow at the expense of the underlying photoinactive oxide. Alternatively, TiO, may be nucleated as islands in the lower oxide phase.” Whatever the mechanism below 1.5 V, it seems reasonable to conclude that the film is essentially TiO, beyond 2 V, and that further growth is restricted to the oxide/solution interface, ions being transported across the film by the electrical field. Armstrong and Quinn lS have observed an interesting peak at 1.7 V in the linear sweep voltammogram obtained with 20 nm evaporated Ti films in 1 mol dm-3 HC104. This peak, which is also associated with an abrupt change in surface conductance, may be related to a restructuring of the film which occurs when it is entirely converted to TiO, from the lower oxides.It is striking that this peak occurs close to the potential at which the photocurrent-voltage relationship changes its slope (fig. 2). The capacitance data were also used to estimate the dative permittivity of the oxide at potentials above 2 V us. SCE. The surface roughness of the electrodes, although remarkably reproducible, was unknown, so that an arbitrary surface rough- ness factor ,%’ = 1.5 was chosen. The relative permittivity, Ef, of films grown at 1 mV s-’ was then calculated to be 60. The cf values found for films grown at higher sweep rates was reproducibly lower, and changes in film structure are reflected in an increase of geometric capacitance of films with The changes are slow; films grown at 1 V s-’ and then left overnight at the formation potential were indis- tinguishable from films grown at 1 mV s-’.74 PHOTOCURRENT SPECTROSCOPY OF ANODIC FILMS PHOTOCURRENT-VOLTAGE CHARACTERISTICS OF THIN FILMS The photocurrent recorded during the reverse scan in fig.2 represents the current against voltage relationship for an oxide film of fixed thickness. In the case of the slowly grown film, an analysis of the curve was made taking reflection at the metal- oxide interface into account.In the absence of surface recombination, the photo- current conversion efficiency is given by l6*I7 exp ( - 2 4 ) 1 + 2aL, ' @ 2 ! 1 - where L,, the Schottky length, is related to the potential drop, Ay, in the semiconductor by and Lp is the diffusion length for holes. A plot of -ln(l - @) against A@ should therefore be a straight line with slope proportional to Ng3 and intercept -ln(l + 2aLp). The donor density can also be obtained from the well-known Mott-Schottky relationship and fig. 7 is the corresponding experimental plot for a film grown at 1 mV s-'. The linear relationship between C-2 and E observed experimentally indicates that the *1 6 N I L44 1 -1 0 1 2 3 4 5 EIV us. SCE FIG. 7.-Mott-Schottky plot for the oxide film grown at 1 mV s-I to 5.4 V us. SCE in 1 mol dm-3 H3P04 (see fig.6). The slope corresponds to a donor density of ca. 1 x lozo ~ r n - ~ (for Ef = 60, 92 = 1.5). distribution of electron donor states is spatially homogeneous. The intercept does not give the flat-band potential directly; the potential drop in the Helmholtz layer is not negligible since the field strength at the surface is appreciable at such a high donor density (lo2* ~ m - ~ ) . Correction for the potential drop in the Helmholtz layer can beJ . F. MCALEER A N D L . M. PETER 75 made if the capacity of the electrical double-layer is known,l* but in the present case the flat-band potential in 1 mol dm-3 H3P04 was estimated from the photocurrent onset to be -0.3 V us. SCE. The double-layer capacity was then calculated from the Mott-Schottky plot to be ca.40 pF cm-2. The Fig. 8 illustrates the treatment of the photocurrent according to eqn (2). 0.2 n 8 I I H W - 0.1. 0 / / /’ 0 ./. / / / I ’ 1 I I 1 2 3 Ap+/V+ FIG. 8.-Treatment of the photocurrent-voltage curve for the oxide film grown at 1 mV s-’ in 1 mol dm-3 H3P04 [see eqn (2)]. (-.--.--.) Theoretical line for L, = 0. (- - - -) Theoretical line for L, = 20 nm. experimental data correspond to an L, value of ca. 2 nm, although this estimate may not be reliable since surface recombination causes the current to decrease as the elec- trode potential is lowered. Similar low values of L, have been obtained in this way by Kennedy and Fresel’ for cc-Fe203, but values for thermally grown TiOa films are generally higher.” An interesting feature of these results is that they show that the Schottky length coincides with the film thickness at the formation voltage.This is clear in fig. 8, where the theoretical lines were drawn for the experimentally determined values of E and Nd. This result suggests that the film grows by high-field migration until the field at the metal-oxide interface is reduced to zero by the space charge in the film, i.e., until L, = Lf. The growth limit of anodic films at constant potential is therefore determined by the donor density, and a true steady-state thickness can be defined. The importance of surface recombination in the determination of 4~ is clear from fig. 8. The photocurrent begins to fall rapidly when the electrode potential is re- duced below 1 v. We therefore examined the photocurrent response to check that a true steady-state photocurrent was measured by the lock-in amplifier.The photo- current transients in fig. 9 show that at potentials above ca. 0.5 V, the photocurrent response is sufficiently ideal to justify quantitative evaluation. At lower potentials,76 PHOTOCURRENT SPECTROSCOPY OF ANODIC FILMS the photocurrents show some overshoot and decay, probably associated with trap- ping and recombination. Interestingly, very similar photocurrent transients were obtained with the oxide film grown at higher sweep rates, although in this case the photocurrent against voltage relationship indicates a very high density of trapping and recombination centres. Evidently the characteristic time constants for these bulk processes is very much shorter than the time-scale of the experiment (seconds).The very low value of L, (two to three orders of magnitude lower than for single crystal rutile) 2o shows that the minority-carrier lifetime in slowly grown anodic films 1 v 0.75 V 0.5 V 0.5 V 0.3 V 0.2 v 0.1 v H 1 s FIG. 9.-Photocurrent response to chopped illumination for an oxide film grown in 1 mol dm-3 H3P04 at 1 mV s-I to 5.4 V us. SCE. Excitation wavelength 335 nm. is small. It is usual to assume that the time taken for a minority carrier to cross the space charge region is very much smaller than q,, the hole lifetime. While this as- sumption is still justified in the case of the slowly grown films, it is no longer valid for films grown at sweep rates above 10 mV s-l. The effective mobility of charge car- riers appears to be reduced sufficiently that only a certain fraction of carriers generated inside the space-charge layer escape recombination or trapping.The almost linear dependence of the photocurrent on potential suggests that carrier transport in the space-charge region is dominated by field emission from traps.4 A quantitative treatment of the problem of photocurrents in amorphous anodic films is not available at present, although Williams and Wright,21 for instance, have shown that pulsed photopotential measurements may offer a useful experimental approach. THICK OXIDE FILMS ON TITANIUM The breakdown of oxide films on valve metals generally takes place at rather high voltages, and it appears to be accompanied by recrystallisation.Titanium behavesJ . F. MCALEER AND L . M. PETER 77 differently; oxygen evolution is already appreciable at potentials around 2 V us. SCE (at pH 0), and changes in film structure are evident at 10 V. Leach and Sidgewick7 have reported recently that the oxide film formed on titanium at low current densities recrystallises, and they have identified crystallites of brookite embedded in an amor- phous matrix. On the other hand, di Quorto et aZ.22 have shown that a minimum field strength is required to stabilise thick oxide films on titanium; a reduction of the current density below a critical value was found to result in a catastrophic breakdown of the film. Under these circumstances it is not possible to distinguish between re- crystallisation as cause or effect of the breakdown.The clearest evidence of film breakdown can be seen in Iinear sweep experiments, and fig. 11 shows that breakdown phenomena are evident on the forward and reverse 0.8 0.6 QP 0.4 0.2 0 280 300 3 20 340 360 380 h/nm FIG. 10.-Time dependence of the spectral response of a titanium electrode held at 30 V in 1 mol dm-3 H2S04; (1) immediately after film formation, (2) after 10 min and (3) after 15 min. sweeps, We have characterised both types of breakdown by a number of tech- n i q u e ~ , ~ . ~ including interferometry, impedance analysis and photocurrent spectro- scopy; a complete discussion is outside the scope of this paper which considers only the relevant results obtained by photocurrent spectroscopy. The " forward breakdown ", which occurs as the potential is increased in the linear sweep, appears to be a process of continued microfracture and repair of the film, whereas the " reverse breakdown ", which takes place when the potential is relaxed, appears to be a catastrophic process which is initiated by the nucleation and propagation of fractures in the film; the whole film appears to " explode ", and a burst of oxygen bubbles can be seen to form on the electrode.We first became aware that the oxide film on titanium was unstable during studies of the excitation spectra for films formed at constant potential. Above ca. 10 V, the magnitude of the conversion efficiency and its spectral distribution were found to be time dependent. The effect became more obvious as the formation voltage was increased, and fig.10 shows as an example how the spectral response of a 30 V film changed when the electrode was held at the formation voltage for 15 min. Clearly, the film has more than doubled in thickness in this time, and also the response at short wavelengths has changed considerably, so that the conversion efficiency now passes78 PHOTOCURRENT SPECTROSCOPY OF ANODIC FILMS 0 10 20 30 40 50 cell voltagelv FIG. 1 1 .-Cyclic voltammogram for titanium in 1 mol dm -3 H2S04, showing the " forward break- down " (1) and " reverse breakdown " (2). Sweep rate 1 V s-l. through a maximum. This behaviour is characteristic of surface recombination ; photons of higher energy are absorbed close to the surface where the effects of surface recombination have a predominant influence on the photocurrent.Surface recom- bination centres can be generated, for example, by the mechanical abrasion of single- crystal surfaces,23 and in the present case it seems reasonable to conclude that film breakdown gives rise to areas which are mechanically damaged. We have shown above (see, for example, fig. 2) that the photocurrent increases smoothly during film growth as more light is absorbed. If the oxide films were stable, the conversion efficiency should approach unity when virtually all incident light is absorbed in the film. In fact, this behaviour was not observed experimentally. At potentials above ca. 7 V, the photocurrent began to decrease and then levelled 0.2, a) 0.1 0 0 10 20 30 40 50 cell voltage/V FIG. 12.-Photocurrent conversion efficiency during a cyclic voltage scan (see fig.11). The electrode was rotated at 20 Hz during the experiment. Sweep rate 1 V s-l.J . F . MCALEER AND L. M. PETER 79 out at a constant but lower value. This effect was not simply due to the scattering of the incident light during oxygen evolution since the electrode was rotated in order to prevent bubble formation. Fig. 12 shows how the photocurrent changed during a linear sweep experiment. On the forward sweep, the photocurrent decreased to a roughly constant value, and on the reverse sweep it increased abruptly. A comparison of fig. 11 and 12 makes it clear that the " forward breakdown " is responsible for the initial reduction of the photocurrent, and we attribute this to the increased rate of recombination and trapping in the stressed and damaged oxide.The reverse scan is particularly interesting. The photocurrent increases suddenly when " reverse break- down " takes place, and impedance measurements show that the capacity of the film also increases abruptly at the same point. Apparently the oxide films breaks down completely at this point, and a new barrier film of about half the original thickness is formed underneath the damaged original layer. This type of film breakdown is particularly easy to miss in an experimental study. Ex situ studies of the oxide film necessarily involve the inadvertent formation of a film which has suffered the reverse FIG. 13.-Photocurrent conversion efficiency for a titanium electrode held at 50 V in 1 mol dmW3 HzS04; (1) before the reverse breakdown and (2) after the reverse breakdown.breakdown. The film appears to be stabilised in situ by the electrical field until a certain critical minimum value is reached when the film collapses. The sudden in- crease in photocurrent at this point probably shows that the total film thickness increases as a fresh barrier film grows under the fractured original. The effect of the reverse breakdown is also evident on subsequent potential cycles. It quite clearly results in the formation of a thicker film with a high density of recom- bination centres, and this is best seen in the corresponding changes in the photocurrent conversion efficiency and its dependence on wavelength. Fig. 13 illustrates the effect80 PHOTOCURRENT SPECTROSCOPY OF ANODIC FILMS of reforming a film which has undergone reverse breakdown.When the electrode is brought back to the original formation potential, the photocurrent spectrum changes considerably. First, it is clear from the response at longer wavelengths that the film thickness has increased by at least a factor of two. Secondly the fall in response at short wavelengths shows that breakdown and reformation gives rise to a high density of surface states, and electron microscopy has confirmed that blistering and cracking of the film do occur. These results show that photocurrent spectroscopy is a convenient in situ technique which is sensitive to changes in the thickness and structure of the oxide film on titanium. The need for such a technique is pressing when ex situ methods are excluded by the irreversible changes which can occur when the control of electrode potential is interrupted. Sensibly combined with other optical and electrochemical techniques, photocurrent spectroscopy has shown great potential as a method for the study of the growth and properties of semiconducting anodic films on metals.This study has also shown why anodised titanium films are unsuitable for photoelectrolysis ; thin films, although perfect in other respects, absorb too little light, whereas thicker films are mechanically damaged and are characterised by a high density of trapping and recombination centres. We thank IMI Titanium for advice and the supply of the titanium electrodes. J. M. also thanks the S.R.C. for financial support. L. M. Peter, Electrochim. Acta, 1978, 23, 1073. L. M. Peter, J. Electroanalyt. Chem., 1979, 98, 49. L. M. Peter, Surface Sci., in press. J. F. McAleer, Ph.D. Thesis (University of Southampton, 1980). J. F. McAleer and L. M. Peter, to be published. L. Young, Anodic Oxide Films (Academic Press, London and New York, 1961). ' J. S. L. Leach and D. Sidgewick, 5e Journe'es d'Etude du Titane et de Ses Alliages, Nantes (1978). M. Ahrens, K. D. Allard and K. E. Heusler, Werkt. und Korros., 1975, 26, 694. L. Arsov, M. Froehlicher, M. Froment and A. Hugot-Le-Goff, Compt. rend. C, 1974, 279, 485. lo C. K. Dyer and J. S. L. Leach, J. Electrochem. SOC., 1978,125, 1032. l1 D. Laser, M. Yaniv and S. Gottesfeld, J. Electrochem. SOC., 1978, 125, 358. l2 S. Gottesfeld, S. Srinivasan, M. Yaniv and D. Laser, J. Physique, 1977, C5, 145. l3 F. Mollers, H. J. Tolle and R. Memming, J. Electrochem. SOC., 1974, 121, 1160. l4 F. L. McCrackin, Nut. Bur. Stand. Tech. Note, 479, 1969. l6 V. A. Myamlin and Y. V. Pleskov, Electrochemistry of Semiconductors (transl.) (Plenum, New l7 Z. A. Rotenberg, T. V. Dzhavrishvili, Yu. V. Pleskov and A. L. Asatiani, Sou. Electrochem., la B. Pettinger, H. R. Schoppel,! T. Yokoyama and H. Gerischer, Ber. Bunsenges. phys. Chem., l9 J. H. Kennedy and K. W. Frese, J. Electrochem. SOC., 1978, 125, 709. 2o R. H. Wilson, Semiconductor Liquid-Junction Solar Cells, in Proc. Airlie Con$, 1977 (Electro- 21 D. K. Williams and G. A. Wright, Electrochim. Acta, 1979, 24, 1179. 22 F. di Quarto, H. Gerischer and K. Doblhofer, Electrochim. Acta, 1978, 23, 195. 23 H. Gerischer, F. Hein, M. Lubke, E. Meyer, B. Pettinger and H-R. Schoppel, Ber. Bunsenges. N. R. Armstrong and R. K. Quinn, Surface Sci., 1971, 67,451. York, 1967). 1977. 13, 1803. 1974,78,1024. chemical SOC., Princeton, N.J.). phys. Chem., 1973,77,284.