J. Phys. Chem. 1993, 97, 1426-1430
1426
Spectroscopic Determination of Flatband Potentials for Polycrystalline Ti02 Electrodes in Nonaqueous Solvents Caretb Redmond and Donald Fitzmaurice' Department of Chemistry, University College Dublin, Dublin 4 , Ireland Received: August 4, 1992; In Final Form: November 16, 1992
The flatband potential Vfiof a polycrystalline Ti02 electrode has been determined, using a newly developed spectroscopic technique, in a range of nonaqueous solvents (MeCN, THF, DMF, MeOH, and EtOH). It has been found that Vfb is significantly more positive for water and nonaqueous protic solvents (MeOH and EtOH) than for nonaqueous aprotic solvents (MeCN, DMF, and THF) due to establishment of a proton adsorptiondesorption equilibrium. Further, we observe that Vh is independent of the electrolyte used in nonaqueous protic solvents, but in nonaqueous aprotic solvents Vfi may depend on the electrolyte cation. Cations which have been found to be potential determining are Li', Na+, and Mg2+. Results indicating that both cation adsorption and intercalation are important in determining Vfi are presented.
Introduction The use of semiconductor electrodesin electrochemical reactors and in solar energy storage and conversion devices offers potential advantages. However, practical applications are likely to require the use of nonaqueous solvents and polycrystallinewide bandgap semiconductorelectrodes. Nonaqueous solvents are preferred as they offer a wider range of electrochemical stability and greater solubility for organic compounds. Polycrystallinewide bandgap semiconductor electrodes are preferred as they are stable and cost effective and have a high degree of surface roughness likely to enhance the efficiency of any solar energy storage or conversion device into which they are incorporated. A particularly attractive electrode in these respects is the transparent polycrystallineTi02 electrode, easily manufactured from medium purity materials, recently described by Graetzel and co-w0rkers.l To permit full realization of their potential an understanding of the factors affecting the band energetics of these transparent polycrystalline Ti02 electrodes in nonaqueous solvents will be essential. Central to an understanding of the band energetics of a semiconductor electrode is an accurate determination of the flatband potential V,. However, accurate determination of V , for semiconductor electrodes in general, and polycrystalline semiconductor electrodes in particular, has proved problematic.2 A new approach to determining V, for a transparent semiconductor electrodehas been developed based on measuring, by visible spectroscopy,the number of electrons present in the conduction band for a given applied potential. The difference between the applied potential and the Vbconsistent with the spectroscopically determined concentration of conduction band electrons is calculated by a fit to the experimental data. This approach has been demonstrated to offer significant advantages over other methods for determination of V, when applied to polycrystalline semiconductor In an attempt to better understand those processes at the interface between a polycrystallineTi02 electrode and the liquid medium supporting the electrolyte, which determine band energetics, V , has been evaluated for a range of nonaqueous solvents dissolving a range of added electrolytes.
Experimental Section Measurement of Flatband Potentials. The electrode used in all experiments consisted of a 4 pm thick layer of fused Ti02 particles, having an average diameter of 15 nm,supported by a 0.5 pm thick layer of fluorine doped Sn02 on glass. Preparation of these Ti02 electrodes has previously been described in detail
by Graetzel and co-workers.' Briefly, however, Ti02 was prepared by hydrolysis of titanium isopropoxide. The resulting dispersion wasconcentrated toabout 160 g/L, andcarbowax (40%wt equiv of TiO2) was added yielding a white viscous liquid used to form a 4 pm thick layer on conducting glass. This layer, allowed to dry in air for 1 h, was fired at 500 OC for 12 h, and the electrode was transferred immediatelyto a vacuum desiccator where it was stored until required. The Ti02 electrode formed the working electrode (2.0 cm2surface area) of a closed three electrode signal compartment cell, the counter electrode being platinum and the reference electrode a saturated calomel electrode(SCE) connected via a salt bridge. Experiments performed using Ag as a pseudoreferenceelectrodewere found to be in excellent agreement. We therefore limit our report to those results obtained using the saturated calomel electrode, and all potentialsare reportedversus SCE. Potential controlwas provided by a Thompson Electrochem Ministat precision potentiostat in association with a Hewlett Packard 3310A function generator. The above cell was incorporated into the samplecompartmentof a Hewlett Packard 8452A diode array spectrometer. Absorbance was measured at 780 and 356 nm, the applied potential being scanned at 0.005 V s a - ] unless otherwise stated. For each determination of V, a new working electrode and freshly prepared electrolyte solution were used. Preparation of Solvents. All aqueous solutions were prepared using distilled deionized water. Acetonitrile (MeCN) supplied by Rathburn was refluxed over PzOs for 3 h and recovered by distillation, and the middle fraction was retained for immediate use. Tetrahydrofuran (THF) supplied by Merck was refluxed over CaH2 for 3 h, refluxed over sodium benzophenone ketyl for 2 h, and refluxed over PzOs for 3 h. The solvent was recovered by distillation, and the middle fraction was retained for immediate use. Dimethylformamide(DMF) supplied by Merck was placed on an A1203 column, allowed to stand over Cas04 for 24 h, decanted off, and distilled under vacuum, the middle fraction being retained for immediate use. Methanol (MeOH) supplied by BDH was refluxed over CaH2 for 3 h and recovered by distillation, and the middle fraction was retained for immediate use. Absolute ethanol (EtOH) supplied by Merck was dried over CaS04 for 24 h and recovered by distillation, and the middle fraction was retained for immediate use. Preparation of Solutes. Tetrabutylammonium perchlorate (TBAP), tetrapropylammonium perchlorate (TPAP), tetraethylammonium perchlorate (TEAP), and tetramethylammonium perchlorate (TMAP), supplied by Fluka, were used following drying at 150 O C under vacuum for 48 h. Lithium perchlorate,
0022-365419312097-1426$04.00/0 0 1993 American Chemical Society
Spectroscopic Determination of Flatband Potentials
I
The Journal of Physical Chemistry, Vol. 97,No. 7,1993 1427
0.25
E
E 0
co
b Q
8E Q
e
E 0
co
0.15
CL
b
-2
pH 2.0(aq) pH 11.0 (aq)
0.15
U
Q
al 0
E
0.05
w
-
0.25
e P -1
0
0.05
-3
Potential (V, SCE)
-2
-1
0
Potential (V, SCE) 0.25
E E
0.25
3 b
0.15
c
Q
8E
0.15
0
e 8
0.05
P
a -0.05 -2
0.05
-0.05 2
-1
0
Potential (V, SCE) -1
0
Potential (V, SCE) Figure 1. (a) Absorbance measured at 780 nm as a function of applied potential for a polycrystalline Ti02 electrode at pH 2.0 and pH 1 1.O in aqueous 0.1 M TMAP solution. (b) Same as a with addition of 0.1M LiC104.
sodium perchlorate, potassium perchlorate, and magnesium perchlorate, supplied by Aldrich, were used following drying at 150 OC under vacuum for 48 h. Tetrabutylammonium bromide (TBABr) and tetrabutylammonium iodide (TBAI), supplied by Aldrich, were used following drying at 80 OC under vacuum for 48 h, while tetrabutylammoniumchloride (TBACI), also supplied by Aldrich, was used following drying at 30 OC under vacuum for 1 week. Lithium bromide, lithiumchloride,andlithium iodide, supplied by Merck, were used following drying at 150 OC under vacuum for 48 h.
ResulQ Determinition of V, for a Polycrystalline Ti02 Electrode in AqueousElectrolyteSolutions. Absorbance measured at 780 nm as a function of applied potential for an aqueous solution of 0.1 M TMAP at pH 2.0 (added HC1O4)and pH 11.O (added KOH) isshown in Figure la. The separationof potentialscorresponding to 0.1 AU is 0.54 V, equivalent to a shift of 0.06V per pH unit. This is in excellent agreement with our previously published data measured in a similar f a ~ h i o n . ~ -Repetition ~ of the above measurement with an electrolyte solutioncontaining0.1 M added LiC104resulted in the traces shown in Figure 1b. As before, the potential shift is 0.06V per pH unit, but we note also that each trace is within 0.01 V of the correspondingtrace shown in Figure la. This observation accords with the general assertion that Vp, for Ti02 in aqueous solution is independent of the electrolyte used.* On this basis we can assumethat our previously determined relationship between pH and Vp, for a polycrystalline TiOz electrode in aqueous electrolyte solution is applicable.5
Vp,= -0.40 - ( 0 . 0 6 ~ p H )(V, SCE) (1) Using eq 1 we calculate Vp,for polycrystalline Ti02 electrodes
Figure 2. (a) Absorbance measured at 780 nm as a function of applied potential for a polycrystalline Ti02 electrode at pH 2.0 and pH 11.0 in aqueous 0.1 M TMAP solution and in MeCN (0.2 M TBAP). (b) Absorbance measured at 780 nm as a function of applied potential for a polycrystalline Ti02 electrode at pH 2.0 and pH 11.0 in aqueous 0.1 M TMAP solution and in MeOH (0.2 M TBAP).
in aqueous electrolyte solutions at pH 2.0 and pH 11.O to be -0.52 and -1 -06V, respectively. Determination of Vn for a Polycrystalline Ti01 Electrode in
Nonaqueous ElectrolyteSolutions. Vp,for a polycrystallineTi02 electrode in dry MeCN (0.2 M TBAP) was determined by measuring absorbanceat 780 nm as a functionof appliedpotential and plotting this data on the same scale as the data in Figure la, see Figure 2. Assuming that the extinction of the free conduction band and trapped electrons is the same for an electrode in water and MeCN we calculate V b to be -2.04 V. Experiments were also performed to determine Vp,for polycrystalline Ti02 electrodes in DMF (0.2 M TBAP) and THF (0.2 M TBAP) yielding values of -2.04 and -2.34 V, respectively. The trace recorded for a similar experiment in MeOH (0.2 M TBAP) is shown in Figure 2b yielding a value for Vp,of -1.12 V. The value determined for EtOH under similar conditions was -1.39 V. The effect on Vp,of addition of LiC104to electrolyte solutions was examined for nonaqueous solvents. As seen in Figure 1, addition of LiC104to an aqueous solution of TMAP has no effect on Vp,. The results presented in Figure 3a, however, indicate that addition of LiClO, has a pronounced effect on the Vp,of a polycrystalline Ti02 electrodein MeCN (0.2 M TBAP). Addition M LiC104 produced a reproducible positive of as little as shift AVp, of 0.03 V, with larger additions producing correspondingly larger AVp,. The activity of LiC104 in MeCN was calculated for each concentration and plotted against Vb*,see Figure 3b. The appearance of a step in the potential scan at higher added LiC104 concentrations is discussed below. Repetition of similar experiments for THF and DMF yielded qualitatively similar behavior. The results of these experiments are collected in Table I. The same experiment when repeated for MeOH produced no significant change in Vb,while the behavior observed for EtOH was intermediate between that observed for MeOH and MeCN, see Table I.
Redmond and Fitzmaurice
1428 The Journal of Physical Chemistry, Vol. 97,No. 7, 1993
TABLE W. Flatbmd Potential of PolycryswUaeTi02 Electrode in Acetonitrile Containing 0.2 M Tetnbutylammonium Salt and Added Li Salt
E E
0
R
Vb
[LiX] = 0.0 M -2.04 -1.98 -2.03 -2.01
C
Q
TBAP (X = Clod) TBACl (X = CI) TBABr (X = Br) TBAI (X = I)
;
2
-0.05
-3
-2
0
-1
Potential (V, SCE)
[LiX] = 1 x 10-3 M -1.97 -1.97 -1.98 -1.97
[UX] = 1 X 10-I M -0.90 U
-0.85 -0.81
a Not possible to fully dissolve 0.1 M LiCl in MeCN containing 0.2 M TBACI.
TABLE 111: Flrtbrnd Pot" of POI c s t d l h Ti02 Electrodes in Acetonitrile Containing 0.g Tetraalkylammonium Perchlorate and Added LiClOl VI%
8
~
c
c
= 0
P
G
,
-'"i -
i
TBAP" TPAP TEAP
i
-5
-2.04 -2.09 -2.14
-1.97 -2.07 -2.13
-0.90 -0.91 -0.91
TABLE I V Flatbad Potential of Pol crystalline Ti02 Electrodes in Acetonitrile Containing 0.g M Tetrrbutylammonium Perchlorate and Added Metal Perchlorates
. -4
-3
2
1
-2
-1
Activity)
TABLE I: Flatband Potential of Polycqstalliw Ti02 Electrodes in Nonaqueow Solvents Contaming 0.2 M Tetrabuhrlammonium Perchlorate and Added LiClOd Vb
[LiClOa] = 0.0 M
[LiC104] = 1 x 10-3 M
[LiC104] = 1 x 10-I M
-2.04
-1.97 -2.02 -2.44 -1.37 -1.11
-0.90 -1.09 -1.20 -1.10 -1.10
-2.04 -2.34 -1.39 -1.12
[LiCIO4] = 1 X 10-I M
a TMAP was not sufficiently soluble in MeCN at room temperature for comparable experiments to be performed.
Figure 3. (a) Absorbance measured at 780 nm as a function of applied potential for a polycrystalline Ti02 electrode in MeCN (0.2 M TBAP) and with addition of 0.001 and 0.1 M LiC104. (b) Vb against log of activity of LiC104 added to MeCN (0.2 M TBAP).
MeCN DMF THF EtOH MeOH
[LiC104] = 1 x 10-3 M
Vb
2
log (LiC104
solvent
[LiCIO4] = 0.0 M
The role of the electrolyte in determining V b for a polycrystalline Ti02 electrode in nonaqueous solvents was examined. A first series of experimentsconcerned the dependenceof V b on the electrolyte anion. V b was determined for a range of tetrabutylammonium salts (perchlorate, bromide, chloride, and iodide) prior to and following addition of le3and 10-1 M amounts of the corresponding Li salt. The results of these experiments are found in Table 11. A second series of experiments examined the dependenceof V b on the electrolyte cation. V b was determined for a range of perchlorate salts (TBAP, TPAP, and TEAP) prior toand following addition of lO-3and lo-' M LiC104. The results of these experiments are found in Table 111. Finally, V, was determined for TBAP electrolyte prior to and following addition of lO-'and lo-' M LiClOs, NaC104,and Mg(C104)~.The results of these experiments are found in Table IV. The appearance of a step in the potential scan at added metal perchlorate concentrations of greater than about 10-3 M in nonaqueous aprotic solvents, see Figure 3a, led us to examine the corresponding behavior at 355 nm. At this wavelength the measured absorbance is expected to decrease as the applied potential is scanned to more negative values due to the Burstein
LiClO4 NaC104 Mg(CQ)z 0
[X(Cl04)] = 0.0 M
[X(Cl04)1 = 1x 1 0 - 3 ~
[X(ClO,)] = 1 x 10-I M
-2.04 -2.04 -2.04
-1.97 (0.0009). -1.98 (O.OOO9) -2.02
-0.90 (0.0253) -1.65 (0.0283) -0.70
Activities of perchlorates in MeCN, calculated as indicated in ref 8.
shift that results from occupation of the available conduction band states under accumulation conditions.'" As can be seen from Figure 4a a step in the potential scan is also observed at 355 nm. The amplitude of this step is dependent on the rate at which the applied potential is scanned to more negative values, see Figure 4b.
Discussion
Vfiof a Polycrystalline Ti02 Electrode in Nonaqueous Electrolyte Solutions. The principal difference, with respect to VR,, between a polycrystallineTi02 electrodeimmersed in an aqueous and a nonaqueous electrolyte solution is the potential absence of a proton adsorption-desorption equilibrium. In nonaqueous aprotic solvents the absence of such an equilibrium will result in large uncertainties in the measured values of Vh,due principally to the fact that trace amounts of protic impurities, for example water, will lead to significant positive shifts in VR,.Therefore, when considering V b in nonaqueous solvents, we would expect significantly different behavior for nonaqueous protic and aprotic solvents. We note that our value of V b for MeCN (-2.04 V) is in good agreement with that obtained by Schumacher and co-workers (-2.2 V) from Mott-Schottky plots for single crystal rutile with TBAP as the supporting electr~lyte.~JJ Both values are significantly more negative than those previously reported for MeCN and tabulated by Finklea.2 However, Schumacher has observed that additionoftraceamountsof water lead toa significant pasitive shift in VR,.~ This shift was attributed to specific adsorption of H+ions at the surface of the metal oxide electrode. Therefore, the close agreement of our value for VR,with that of Schumacher is likely due to the considerable care taken in drying the solvents and electrolytes used. The values of V b determined for DMF
The Journal of Physical Chemistry, Vol. 97, No. 7, 1993 1429
Spectroscopic Determination of Flatband Potentials
l -
!
-0.5 4 -1.75
I -1.25
-0.75
-0.25
Potential (V, SCE)
-a-
.-cnc
Abs.at356nm
C
3
-0.54 -1.75
.
-1.25
-0.75
-0.25
1
Potential (V, SCE) Figure 4. Absorbance measured at 780 and 356 nm as a function of applied potential for a polycrystalline Ti02 electrode in MeCN (0.2 M TBAP) containing0.1 M added LiC104. The potential scan rates were (a) 0.001 V s-I and (b) 0.005 V s-l.
(-2.04 V) andTHF (-2.34 V) are alsosignificantlymore negative than those previously determined by Fox and co-workers using both Mott-Schottky and photocurrent onset potential methods.” The conclusion of Fox and co-workers that V b for nonaqueous aprotic solvents shows no apparent systematic variation is consistent with the data contained in Table I. That V b in an aprotic solvent is constant for a given electrolyte solutionsuggests that Vp, is determined, in the absence of a proton adsorptiondesorption equilibrium, largely by the physical and chemical properties of the semiconductor electrode. The value of V b determined for MeOH (-1.12 V) is 0.92 V more positive than that for MeCN (-2.04) but significantlymore negative than the values previously reported for this solvent by Tamura and co-workers.12-14 This would suggest specific adsorption of H+ ions at the electrode surface. The autoprotolysis constant for MeOH is 2 X 1&I7, as opposed to a value of 3 X for MeCN, on which basis we would predict a value for V b of about -0.90 V for MeOH assuming a proton adsorptiondesorption eq~ilibrium.’~Failure to observe any shift in V b following addition of LiC104to MeOH (0.2 M TBAP), the same behavior observed upon addition of LiC104 to aqueous 0.2 M TMAP, further supports the view that in nonaqueous protic solvents preferential adsorption of H+ determines Vb. The autoprotolysis constant for EtOH is 8 X and we would expect a V b more negative than that for MeOH and for V b to show some dependence on the concentration of added LiClO4. That is observed to be the case. Thus it appears to be more useful to classify behavior at the electrode-clectrolyte solution interface according to whether the supporting solvent is protic or aprotic rather than aqueous or nonaqueous. In protic solvents the existence of a proton
adsorption4esorption equilibrium determines Vb. As a consequence Vta is independent of the nature of the added electrolyte. This is the case provided that the autoprotolysis constant of the proticsolvent isgreater than about 10-17,for example, for MeOH. For solvents such as EtOH, with an autoprotolysis constant of 8 X 1C2O,an intermediatebehavior is observed. V,I,is significantly more positive for EtOH than for MeCN but more negative than for MeOH, however, addition of LiCI04 shifts V b to somewhat more positive potentials. Finally, for aprotic solvents such as MeCN, with an autoprotolysis constant of 3 X V b is very negative due to the absence of a proton adsorption-desorption equilbriumand consequently shifted to significantlymore positive potentials following addition of LiC104. Dependence of V, for a Polycryswllne Ti02 Electrode in Nonaqwcnrs Aprotic Solvents 011 Added Electrolyte Schumacher’s observationthat V b for a polycrystallineTi02 electrodein MeCN is dependent on the concentration of added H2O and consequent proton adsorption9suggestedthat the value of V b in a nonaqueous aprotic solvent may depend on the nature of the added electrolyte assuming the absence of any other protic solute. That V b is dependent on the nature of the electrolyte present in dry MeCN is evident from the shift of V, to more positive potentialsfollowing addition of LiC104 to MeCN (0.2 M TBAP), Figure 3a. As the concentrationof added LiC104is increased the AVb also increases, Figure 3b. Similar behavior is observed for DMF and THF, see Table I. That V b is shifted to more positive values following addition of either H20 or LiC104 suggests that Li+ is the potential determining ion in the latter case.l+l* Support for this comes from the data presented in Tables I1 and 111. In Table I1 AVb observed upon addition of and 10-I M LiC104, LiCI, LiBr, and LiI are similar, indicating that V b is not dependent on the nature of the anion present. Further evidence indicating that V b is not dependent on the nature of the anion present is agreement of the values of V, determined for TBAP, TBACI, TBABr, and TBAI. The data presented in Table I11 suggests that while large cations are potential determining only to a very limited extent, if at all, they are equally so for a polycrystalline Ti02 electrode in a nonaqueous aprotic solvent with Vrt,for TBAP, TPAP, and TEAP being in close agreement. Furthermore, AVb following addition of 10-3and 10-1M LiC104 to TBAP, TPAP, and TEAP are similar. Finally, AV, observed following addition of lW3 and 10-I M LiC104, NaC104, and Mg(Cl04)~are compared in Table IV. For 10-3 M added perchlorate the shifts in V b are similar. However, for 10-1 M added perchlorate the observed shift in V b for Na+ is significantly smaller than that observed for Li+,while the observed shift in V, for Mg2+is significantlylarger than that observed for Li+. It is noted that these differences cannot be accounted for by differences in the activities these electrolytes.* Li+ as a Potential Determining Ion in Nonaqueous Aprotic Solvents. Insight into the mechanism by which Li+ ions (also Na+ and Mg2+) shift V b to more positive values is gained by examining a plot of V b against the log of LiC104 activity, see Figure 3b. This plot indicates that a concentrations of added LiC104 up to 10-3 M there is a gradual 40 mV per decade shift, suggestingspecific adsorption of Li+ ions at the electrodesurface. Thereforeat low concentrationsit appears that surface adsorption is the mechanism by which the electrodbelectrolyte interfacial potential is lowered. This is supported by the fact that AV, observed following addition of less than M LiClO4, NaClO4, and Mg(C104)2 are similar. The rapid increase in V b observed at higher added LiC104 concentrations is attributed to intercalation of Li+ close to the Ti02 crystal lattice surface. The data presented in Table I1 indicating that AV, is independent of the counteranion of Li+ is consistent with intercalation of Li+ close to the electrodesurface. If the intercalatingcation had a tendency
1430 The Journal of Physical Chemistry, Vol. 97, No. 7 , 1993
to penetrate the crystal structure to a significant extent it would be expected that electrostatic attraction between the counteranion and the intercalating cation would lead to a smaller shift for LiCl than for LiC104. This was not observed to be the case. Li+ intercalation is a well-established phenomenon and would be expected to lead to a positive shift in the potential of the conduction andvalence band edges a t theelectrodeelectrolyte interface.1’21 AVb attains a constant value a t Li+ Concentrations greater than about 10-1 M, equivalent to the doping level in Ti02 a t room temperature, about lO-I9 This might be expected as it is these free carriers which facilitate maintenance of electrical neutrality at the electrode-electrolyte solution interface. Such an interpretation is also consistent with the AV, data found in Table IV. AVb observed following addition of 10-1M of these perchlorates is larger for Li+ than Na+ which accords with the fact that they have ionicdiameters of 1.18 and 2.04 A, respectively. The shift observed for Mg2+ is larger than that observed for Li+, presumably because while Mg2+is a dication it has a diameter similar to that of Li+. That adsorption should dominate a t lower concentrations and intercalation at higher concentrations has previously been suggested by Murray and co-workers when examining adsorption of group I and I1cations a t Mn02surfaces.20 We tentatively suggest that observation of a stepin the potential scan at Li+ concentrations greater than M is due to competitive filling of acceptor states resulting from Li+ intercalation. Failure to observe this step at lower concentrations would be due to filling of these states by free conduction band electrons present at room temperature; this process would be expected to be incomplete a t Li+ concentrations greater than 10-3 M for a free electron concentration of electrons cm-3. The fact that at slower scan rates this step is not observed is consistent with our suggested explanation. At these lower scan rates the Li+ acceptor states are filled completely a t potentials very close to the flatband potential and no step is observed. Comparisonwith OtherMethodsfor Determinationof V,. The two most widely used methods for determining Vbare preparation of a MottSchottky plot and determination of the photocurrent onset potential. The principal advantage of a MottSchottky plot is that Vbcan bedetermined in the absenceof direct bandgap irradiation. The principal disadvantages are frequency dispersion and sensitivity to surface defects. For polycrystalline electrodes thesedisadvantages are particularly pronounced, and consequently MottSchottky plots have proved of limited use in characterizing polycrystalline Ti02electrodes.22The other widely used method for determination of Vbis measurement of the photocurrent onset potential. This method has as its principal advantage that measurements are easily made. Its principal disadvantage is a large uncertaintyin the value determined for the flatband potential due to difficulties in deciding at what potential the photocurrent is perceptible over the dark current. By comparison the method used here is simple, does not require direct bandgap excitation, can be applied to polycrystalline materials which have a high concentration of defects and to electrode surfaces which have been treated by adsorption of sensitizers or other molecules provided they do not overlap completely absorption by the conduction band electrons. Its principal disadvantages are that a semiconductor electrode of good optical quality must be prepared, the extinction coefficient of the trapped and free electrons must be known, and the doping level must be determined. Finally, experiments have been performed by Kavan and coworkers using transparent Ti02 electrodes prepared by aerosol pyrolysis on conducting glass. These electrodes are sufficiently defect free that V b has been determined using both the MottSchottky method and the spectroscopic method described herein; the observed agreement is very satisfactory.23
Redmond and Fitzmaurice
Conclusions The flatband potential V , of a polycrystalline Ti02 electrode has been determined, using a newly developed spectroscopic technique, in a range of nonaqueous solvents (MeCN, THF, DMF, MeOH, and EtOH). It has been found that for water and nonaqueous protic solvents (MeOH and EtOH) Vbis significantly more positive than for nonaqueous aprotic solvents (MeCN, DMF, and THF) due to establishment of a proton adsorption-desorption equilibrium. Further, while it is found that in nonaqueous protic solvents Vb is independent of the electrolyte used, in nonaqueous aprotic solvents Vbmay depend on the electrolyte cation. Cations which have been found to be potential determining are Li+, Na+, and Mg2+. It appears that a t low cation concentration adsorption is responsible for the observed positive shift in V b while a t higher cation concentrations intercalation close to the electrode surface is important. This suggests that when considering the role of the electrolyte in determining the band energetics of a semiconductor electrode it is more useful to distinguish between protic and aprotic solvents than between aqueous and nonaqueous solvents. Our current work is directed toward quantifying to what extent the observed shift in Vh is due to adsorption and to what extent it is due to intercalation. More detailed information concerning the mechanism of intercalation is also sought. Acknowledgment. This paper is dedicated to Dr. Hector Rubalcava. The authors thank Professor M. Graetzel, Dr. L. Kavan, and Dr. E. Waghorne for their useful comments in the course of the preparation of this manuscript. We also thank Hewlett Packard (Ireland) and Eolas (The Irish Board for Science and Technology) for their support. References and Notes ( 1 ) ORegan, B.; Graetzel, M. Nature 1991,353, 737.
(2) Finklea, H. Semiconductor Electrodes; Elsevier: New York, 1988. (3) O’Regan, 8.;Graetzel, M.; Fitzmaurice, D. Chem. Phys. Lett. 1991, 183, 89. (4) ORegan, B.; Graetzel, M.; Fitzmaurice, D. J. Phys. Chem. 1991,95, 10525. ( 5 ) Rothenberger, G.;Graetzel, M.; Fitzmaurice, D. J. Phys. Chem. 1992,96,5983. (6) Kavan, L.; Stoto, T.; Graetzel, M.; Fitzmaurice, D.; Shklover, V. J. Phys. Chem., submitted for publication. (7) Graetzel, M.; Fitzmaurice, D. J . Phys. Chem., manuscript in preparation. ( 8 ) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions, 2nd ed., Butterworths: London, 1959. (9) Heinzel, A. B.; Teschner, D. M.; Schumacher, R. Ber. Bunsen-Ges. Phvs. Chem. 1981.85. 1 1 17. 1IO) Schumacher, R.;Teschner, D.; Heinzel, A. B. Ber. Bunsen-Ges. Phys. Chem. 1982,86,1153. ( 1 1 ) Kabir-ud-Din.; Owen, R. C.; Fox, M. A. J . Phys. Chem. 1981,85, 1679. ( I 2 ) Yoneyama, H.; Toyoguchi, Y .;Tamura, H . J . Phys. Chem. 1972,76, 3460. (13) Miyake, M.; Yoneyama, H.; Tamura, H. Electrochim. Acta 1976, 21, 1065. (14) Miyaka, M.; Yoneyama, H.; Tamura, H. Electrochim. Acta 1977, 22, 3 19. (15) Coetzee, J. F.; Padmanabhan, G.R. J. Phys. Chem. 1962,66,1708. (16) Green, M. J . Chem. Phys. 1959,31, 200. (17) Bard, A. J.; Bocarsly, A. B.; Fan, F. F.; Walton, E. G.;Wrighton, M. S. J . Am. Chem. SOC.1980. 102, 3671. (18) Bard, A. J.; Fan, F. F.; Gioda, A. S.; Nagasubramanin, G.; White, G. S. Faraday Discuss. Chem. Soc. 1980,70, 19. (19) Tennakone, K.; Wickramanyake, S.W.; Samarasekara, P.; Fernando, C. A. Phys. Status Solidi 1987,104, K57. (20) Murray, D. J.; Healy, T. W.; Fuerstenau, D. W. Advances in
Chemistry; Gould, R. F., Ed.; American Chemical Society: Washington, DC. 1968, Vol. 97, Chapter 7. (21) Ooi, K.; Miyai, Y.; Sakakihara, J. Lmgmuir 1991, 7, 1167. ( 2 2 ) Unpublished results of experiments performed while one of us (D.F) was a visiting scientist in Professor M. Graetzel’s laboratory at EPFL, Switzerland. (23) Kavan, L.; Graetzel, M. Electrochim. Acta, submitted for publication.
