Conduction Band Mediated Electron Transfer Across Nanocrystalline

Apr 20, 2007 - ... carrier” solar cells that could exceed the Shockley−Queisser limit.11 ... by the Kohlrausch−Williams−Watts (KWW) model, whi...
0 downloads 0 Views 196KB Size
6822

J. Phys. Chem. B 2007, 111, 6822-6828

Conduction Band Mediated Electron Transfer Across Nanocrystalline TiO2 Surfaces† Aaron Staniszewski, Amanda J. Morris, Tamae Ito, and Gerald J. Meyer* Departments of Chemistry and Materials Science and Engineering, Johns Hopkins UniVersity, 3400 North Charles Street, Baltimore, Maryland 21218 ReceiVed: January 17, 2007; In Final Form: March 2, 2007

Mesoporous thin films comprised of interconnected nanocrystalline (anatase, 20 nm) TiO2 particles were functionalized with [Ru(bpy)2(deebq)](PF6)2, where bpy is 2,2′-bipyridine and deebq is 4,4′-diethylester-2,2′biquinoline, or iron(III) protoporphyrin IX chloride (hemin). These compounds bind to TiO2 with saturation surface coverages of 8 (( 2) × 10-8 mol/cm2. Electrochemical measurements show that the compounds first reduction occurs prior to or commensurate with the reduction of the TiO2 electrode. Apparent diffusion constants, Dapp, abstracted from chronoabsorption data measured in acetonitrile were found to be dependent on the applied potential and the electrolyte used. The Dapp values for reduction of Ru(dcbq)(bpy)2/TiO2, where dcbq is 4,4′-(COO-)2-2,2′-biquinoline, increased with decreasing surface coverage. At near saturation surface coverage, the apparent diffusion constant was 9.0 × 10-12 m2/s after a potential step from -0.61 to -1.31 vs Fc+/0. The Dapp varied by over a factor of six with applied potential for the oxidation of [Ru(dcbq-)(bpy)2]-/TiO2 to Ru(dcbq)(bpy)2/TiO2. Complete reduction of hemin/TiO2 to heme/TiO2 was observed under conditions where the heme surface coverage was about 1/100 of that expected for monolayer surface coverage. The hemin reduction rates were strongly dependent on the final applied potential. The rates for heme to hemin oxidation were less than or equal to the hemin to heme rates in the presence and absence of pyridine. This behavior was opposite to that observed with Ru(dcbq)(bpy)2/TiO2 where reduction was slower than oxidation. A Gerischer-type model was proposed to rationalize the rectifying properties of the interface.

Introduction The redox and excited state behavior of molecules anchored to TiO2 nanocrystallites (anatase) interconnected in mesoporous thin films are of considerable interest with possible applications in solar conversion and photo- and electro-chromic materials.1 Examples of lateral intermolecular ‘hole’ transfer across the anatase surface are now well documented and include ruthenium and osmium polypyridyl coordination compounds as well as triaryl amines.2-6 Voltammetric and spectroelectrochemical measurements have provided keen insights into the mechanism of hole transfer. The redox chemistry is initiated by oxidation of molecules anchored to the conductive substrate that supports the nanocrystalline thin film. This is usually a transparent conductive oxide (TCO) like fluorine-doped tin oxide. Oxidation of the molecules anchored to the TiO2 nanoparticles then occurs by lateral intermolecular hole transfer. A percolation threshold of about 50% saturation surface coverage is required for complete oxidation of all the molecules in the thin film.4 Therefore, hole transfer does not involve the TiO2 valence band. Indeed, similar redox behavior has been observed on insulating ZrO2 and Al2O3 that likely have very different valence band positions. Much less is known about lateral ‘electron’ transfer across anatase surfaces. However, the rapid reduction of surface anchored viologen molecules suggested a conduction band mediated process.7 The significant color changes that can be achieved on millisecond time scales has attracted attention for ultrafast display applications.8 Lateral ‘hole’ and ‘electron’ transfer processes at these interfaces are also relevant to solar energy conversion.9,10 Visible light excitation of heteroleptic Ru(II) compounds such as Ru†

Part of the special issue “Norman Sutin Festschrift”.

(dcbq)(bpy)2, where dcbq is 4,4′-(COO-)2-2,2′-biquinoline and bpy is 2,2′-bipyridine, anchored to TiO2, produced a remarkably long-lived charge separated state, eq 1.9

Ru(bpy)2(dcbq)**/TiO2 + RuII(bpy)2(dcbq)/TiO2 f [RuIII(bpy)2(dcbq)]+/TiO2 + [RuII(bpy)2(dcbq-)]-/TiO2 (1) The quantum yields for formation of this state were shown to be strongly excitation wavelength dependent. The yields more than doubled when blue light was substituted for green.10 A mechanism was proposed wherein ultrafast electron injection into TiO2 was followed by trapping of the “hot” electron by another Ru(II) compound. Interestingly, the charge separated state stored more free energy than the emissive excited state, Ru(dcbq)(bpy)22+*. This occurs because ultrafast injection is from a vibrationally hot excited state, abbreviated Ru(bpy)2(dcbq)**/TiO2, and the charge separated state stored some energy that would otherwise have been lost as heat through nonradiative relaxation to the emissive excited state. These photoinduced redox reactions are thus of relevance to “hot carrier” solar cells that could exceed the Shockley-Queisser limit.11 Isoenergetic intermolecular charge transfer provides a mechanism by which charge can ‘hop’ across the semiconductor surface, Scheme 1. Lateral “hole” MIII/II and/or “electron” M(dcbq0/-) hopping can lead to encounters of the reducing and oxidizing equivalents and hence a pathway for charge recombination to yield ground state products. Intermolecular electron hopping between ligands of coordination compounds anchored to anatase nanoparticles has not been previously studied. This

10.1021/jp070413n CCC: $37.00 © 2007 American Chemical Society Published on Web 04/20/2007

Conduction Band Mediated Electron Transfer

J. Phys. Chem. B, Vol. 111, No. 24, 2007 6823

SCHEME 1

is due to the fact that large Faradaic currents from TiO2 obscure reduction of the coordinated 4,4′-(COO-)2-2,2′-bipyridine (dcb) ligands that are most commonly used.1,2 To our knowledge, there is only one report of electron transfer to coordinated dcb ligands, and this was measured on insulating colloidal Al2O3 thin films.6 The dcbq ligand, on the other hand, is about 400 meV more easily reduced than dcb and, under appropriate conditions, occurs prior to reduction of the TiO2. It is these interfacial energetics that enabled the novel charge separation process to occur.10 The time-resolved absorption changes that correspond to charge recombination, eq 2, were well described by the Kohlrausch-Williams-Watts (KWW) model, which is a paradigm for charge transport in disordered media ∆A ) R exp -(kt)β.12,13

[RuIII(bpy)2(dcbq)]+/TiO2 + [RuII(bpy)2(dcbq-)]-/TiO2 f 2 RuII(bpy)2(dcbq)/TiO2 (2) An advantage of this model is that the normalized data could be quantified with only two variables: β, which is inversely related to the width of an underlying Levy distribution of rate constants, and k which is the rate constant at the maximum amplitude of the distribution. The average rate constant for a large number of samples were found to be 1.0 ( 0.8 × 105 s-1. The β values varied between 0.2 and 0.3, which corresponded to a highly skewed distribution of rates with significant amplitude over 6 orders of magnitude. A broad distribution would be expected if the donors and acceptors were located at variable distances from each other. The long ∼millisecond lifetimes observed could correspond to redox equivalents generated on different TiO2 nanocrystals for example. A disadvantage of the KWW model is that it does not provide molecular insights into the recombination mechanism(s). An issue that naturally arises concerns which charges move faster, the ‘electrons’ or the ‘holes’. Hole transfer involves the metal t2g orbitals in the formal oxidation states of RuIII (t2g5) and RuII (t2g6). Electron transfer involves the π* orbitals of the diimine ligand. For Ru(bpy)32+ in fluid acetonitrile solution, the RuIII/II(bpy)33+/2+ self-exchange rate constants are 2 × 109 M-1 s-1 and the Ru(bpy0/-)(bpy)22+/+ are 8 × 108 M-1 s-1.14,15 Sutin has previously analyzed the large RuIII/II self-exchange rate constants and the role of π backbonding in these compounds.15 The solution self-exchange rate constants suggest that ‘hole’ and ‘electron’ transfer rates across nanoparticle surfaces should be very similar. However, the situation may be very different at semiconductor interfaces where the reorganization terms are

expected to differ, and the electronic properties of the semiconductor may play an important role. Herein the first spectroelectrochemical and chronoabsorption measurements of electron transfer in Ru(II) diimine compounds anchored to TiO2 are reported. To test generality, the reduction of heme catalysts was also quantified. The intense absorption of hemes enabled electron-transfer studies at surface coverages that were about ∼1/100 of that expected for a monolayer. The mechanisms for electron transfer were found to be significantly different from that previously reported for hole transfer.3-7 Compelling evidence for conduction band participation was observed. In fact, it appears that all the spectroelectrochemical data reported herein can be adequately explained by a conduction band mediated processes with no evidence for lateral electron transfer. We note that hemes anchored to highly reduced TiO2 nanoparticles are known to drive multi-electron-transfer reduction of organohalides16,17 and may also be useful for proton and CO2 reduction.18,19 Experimental Section Materials. Methanol (Aldrich, >99.9%; or Fisher, ACS), pyridine (Fisher, ACS), hemin (Fluka, 98.0%), tetrabutylammonium perchlorate or TBAP (Fluka, 99.0%), tetrabutylammonium hexafluorophosphate or TBAH (Fluka, 99.0%), DMSO (Fisher, ACS), CH3CN (Burdick & Jackson, spectroscopic grade), and LiClO4 (Fluka, 99.0%) were used as received. Ru(deebq)(bpy)2(PF6)2 was available from previous studies.10 TiO2 Thin Films. Mesoporous thin films (10 µm) of ∼20 nm anatase TiO2 nanocrystallites were prepared by a sol-gel technique that has been previously published.3 Freshly prepared films were soaked in a concentrated DMSO solution of hemin for ∼5 min until the desired surface coverage was reached. These films were then washed with methanol and utilized. For pyridine studies, the hemin/TiO2 materials were placed into a cuvette of neat pyridine, purged with N2 gas for at least 30 min, and photolyzed with UV light to yield heme(py)2/TiO2. The thin films were then removed from the pyridine solution, rinsed with methanol, and used. For experiments with Ru(deebq)(bpy)2(PF6)2, the TiO2 thin films were first pretreated with an aqueous pH 11 (NaOH) solution, rinsed in acetonitrile, and then placed in millimolar acetonitrile solutions of Ru(deebq)(bpy)2(PF6)2 overnight. Electrochemistry. A BAS model CV-50W potentiostat was used in a standard three-electrode arrangement with a glassy carbon working electrode, a Pt gauze counter electrode, and Ag/AgCl as the reference electrode. Cyclic voltammetry of the compounds anchored to TiO2 was performed with surface

6824 J. Phys. Chem. B, Vol. 111, No. 24, 2007

Figure 1. Visible absorption spectrum of Ru(dcbq)(bpy)2/TiO2 immersed in a 100 mM TBAP-acetonitrile solution (black line). Shown as a dashed red line is the absorption spectrum observed after stepping the potential to -1.31 V vs Fc+/0. This spectrum is assigned to the reduced form of the compound, [Ru(dcbq-)(bpy)2]-/TiO2.

functionalized TiO2 films deposited on FTO glass as the working electrodes submerged in 0.1 M TBAP-acetonitrile. The CV experiments were carried out at room temperature under argon. Spectroelectrochemistry of TiO2 thin films was performed in a three-electrode custom designed long cuvette with a sensitizer/ TiO2/FTO film as the working electrode, a Pt gauze as the counter electrode, and a Ag/AgCl reference electrode, all submerged in 0.1 M TBAP-CH3CN under a nitrogen atmosphere. In some experiments methanol was used instead of CH3CN and Li+ in place of TBA+. All potentials reported are versus ferrocenium/ferrocene present as an internal standard. Absorbance Measurements. UV-vis absorbance measurements were made on a Hewlett-Packard 8453 diode array or a Carey 14 spectrophotometer. A 385 nm long pass filter was placed between the lamp and the sample for most diode array measurements.

Staniszewski et al.

Figure 2. Absorption changes for Ru(dcbq)(bpy)2/TiO2 electrodes measured at 470 nm after stepping the potential from -0.61 V to -1.31 V. The electrodes had different concentrations of the ruthenium compound: near saturation surface coverage (upside down triangles), 2 /3 of saturation coverage (triangles), 1/3 saturation coverage (circles), and no sensitizer (squares) in 0.1 M TBAP-CH3CN. The inset shows the raw experimental data, and the larger plot is this same data normalized to a unit absorbance change. The absorption change corresponds to the reduction of Ru(dcbq)(bpy)2/TiO2 to [Ru(dcbq-)(bpy)2]-/TiO2.

Results Shown in Figure 1 is the absorption spectrum of Ru(deebq)(bpy)2(PF6)2 anchored to TiO2, abbreviated Ru(dcbq)(bpy)2/ TiO2, immersed in an acetonitrile electrolyte. Two metal-to-ligand charge transfer (MLCT) absorption bands are present in the visible region. Resonance Raman spectroscopy has shown that the higher energy 444 nm band is Ru f bpy in nature while the 539 nm band is Ru f dcbq. Previous IR studies have shown that the ethyl ester groups of the deebq ligand are hydrolyzed to carboxylates upon reactions with basic TiO2.10 The reduction potential for Ru(dcbq)(bpy)2/ TiO2 was -1.01 V in 0.1 M TBAH acetonitrile electrolyte.10 Also shown in Figure 1 is the absorption spectrum measured after stepping the potential to -1.31 V vs Fc+/0. Comparative studies with closely related compounds in fluid solution demonstrate that the spectral changes are due to reduction of the coordinated dcbq ligand to yield [Ru(dcbq-)(bpy)2]-/TiO2. The positive absorption bands at 400 and 1100 nm are characteristic of the reduced dcbq ligand.10 Figure 2 shows the absorption change monitored at 470 nm as a function of time after a potential step from -0.61 V to -1.31 V. Experimental data shown in the inset is for an unsensitized TiO2 thin film and sensitized films with nearly unit, 2/ , and 1/ saturation surface coverage. The normalized data 3 3 reveals that complete reduction of the films required about 20 s and became faster as the surface coverage was lowered. For high surface coverages, the time-scale for the absorption change was found to be independent of the monitoring wavelength. A Ru(dcbq)(bpy)2/TiO2 electrode could be switched between the

Figure 3. Absorption changes measured at 470 nm for a [Ru(dcbq-)(bpy)2]-/TiO2 electrode at near saturation surface coverage after stepping the potential from -1.31 V to the indicated potentials in 0.1 M TBAPCH3CN. The absorption change corresponds to the oxidation of [Ru(dcbq-)(bpy)2]-/TiO2 to Ru(dcbq)(bpy)2/TiO2.

ground and reduced state over 10 times with no measurable decomposition provided that the electrode was not maintained in the reduced state for more than a minute. An interesting absorption signature was observed between 500 and 550 nm for films that had low surface coverages: Upon stepping the potential to -1.31 V, the absorption first increased slightly and then decreased to a steady state value as the film was reduced. The initial increase was also observed in the absence of the RuII compound and was due to reduced TiO2, TiO2(e-), which has a weak absorption in this region.20 Significant desorption and some decomposition was observed when a [Ru(dcbq-)(bpy)2]-/TiO2 electrode was held at -1.31 V for extended periods of time. The desorption was accelerated at more negative applied biases. Therefore it was difficult to determine if the reduction rates were dependent on the final applied potential. The bias dependence of the reverse process, [Ru(dcbq-)(bpy)2]- + TiO2 f, was quantified. Figure 3 shows the absorption changes measured at 470 nm after stepping the potential from -1.31 V to potentials between -0.21 and -0.81 V. The oxidation rate showed a strong dependence on the final

Conduction Band Mediated Electron Transfer

J. Phys. Chem. B, Vol. 111, No. 24, 2007 6825

TABLE 1: Apparent Diffusion Constants for Reduction of TiO2 and Ru(dcbq)(bpy)2/TiO2a Γ (× 10-8 mol/cm2)

Dapp (× 10-11 m2/s)

0.0 3.3 6.0 9.1

2.8 2.1 1.2 0.90

a All measurements performed at room temperature in 0.1 M TBAPCH3CN after a potential step from -0.61 V to -1.31 V vs Fc+/0.

TABLE 2: Apparent Diffusion Constants for the Oxidation of [Ru(dcbq-)(bpy)2]-/TiO2 as a Function of the Final Potential Vfinal

Dapp (× 10-10 m2/s)

-0.21 -0.31 -0.41 -0.51 -0.61 -0.71 -0.81

3.9 3.2 2.5 1.8 1.3 0.90 0.61

Figure 4. Absorption changes measured at 470 nm for a Ru(dcbq)(bpy)2/TiO2 electrode with saturation surface coverage after stepping the potential from -0.61 V to -1.31 V and then back to -0.61 V in 0.1 M TBAP (black squares line) and 0.1 M LiClO4 (blue circles line) in CH3CN. Superimposed on the data are best fits to eq 3. The absorption change corresponds to the reduction of Ru(dcbq)(bpy)2/TiO2 to [Ru(dcbq-)(bpy)2]-/TiO2.

a All measurements were performed at room temperature in 0.1 M TBAP-CH3CN after a potential step from -1.31 V to the indicated potential vs Fc+/0.

potential. The absorption spectra observed 10 s after the potential step corresponds to the ground state compound. The oxidation of [Ru(dcbq-)(bpy)2]-/TiO2 was thus faster than the initial reduction. The data shown in Figures 2 and 3 were fit to the following model from which the apparent diffusion constants, Dapp, were abstracted, eq 3.4,5

∆A )

2AmaxDapp1/2t1/2 dπ1/2

(3)

Here Amax is the maximum absorption change, t is the time in seconds, and d is the film thickness which was taken to be 10 µm. As has been previously reported for hole transfer, about 60% of the absorption change could be fit to this model.4-6 The Dapp values abstracted reflect what is seen in the raw experimental data; they increased slightly with surface coverage and strongly with applied positive potential, Tables 1 and 2. The apparent diffusion constants measured in acetonitrile electrolyte were found to be sensitive to the nature of the electrolyte. Shown in Figure 4 are absorption changes measured after stepping the potential from -0.61 V to -1.31 V and then back to -0.61 V in 0.1 M Li+ and TBA+. Analysis of this data with eq 3 showed that Dapp was 3.7 × 10-12 m2/s in the Li+ electrolyte and increased to 8.6 × 10-12 m2/s when TBA+ replaced Li+. Figure 5 shows absorption spectra of a hemin/TiO2 thin film before and after band gap excitation in methanol. The product of the photolysis has been assigned to heme/TiO2, where the iron center was photoreduced from a formal oxidation state of III (hemin) to II (heme).21 The inset of Figure 5 shows absorption spectra recorded over the first 20 s after the applied potential was stepped from -0.61 to -1.21 V in 0.1 M TBAHmethanol. Four sharp isosbestic points were observed, and the final spectra recorded indicated quantitative formation of the reduced compound, heme/TiO2. The coordination of pyridine to heme/TiO2 induces characteristic changes in the visible absorption spectrum of the porphyrin.22 Hemin/TiO2 in neat pyridine or concentrated pyridine methanol solutions was found to autoreduce to yield heme(py)2/TiO2 after several hours. The FeIII/II reduction

Figure 5. Absorption spectra of a hemin/TiO2 electrode in methanol before (black solid line) and after (red dashed line) band gap excitation in methanol. The absorption spectra after band gap excitation was assigned to heme/TiO2. The inset shows absorption spectra recorded after the potential of a hemin/TiO2 electrode was raised from -0.61 V to -1.21 V in 0.1 M TBAH-methanol. The absorption changes in the inset correspond to the reduction of hemin/TiO2 to heme/TiO2.

potential for heme(py)2 was measured to be -0.47 V in methanol electrolyte. The heme(py)2/TiO2 did not desorb at negative applied potentials with only minor decomposition that could be avoided by limiting the time the negative potential was applied. These properties enabled us to look at the bias dependence of hemin/TiO2 and hemin(py)2/TiO2 reduction. Shown in Figure 6 are time-resolved absorption spectra recorded after stepping the potential applied to a hemin(py)2/ TiO2 electrode from -1.61 to -0.46 V. The initial spectrum measured is that expected for heme(py)2/TiO2 as indicated by comparative studies in fluid solution. The time required to reduce the iron center was found to be strongly dependent on the final applied potential. The inset of Figure 6 shows that the reduction was complete within 5 s when the applied potential was -1.61 V. Similar behavior was observed for the reduction of hemin/TiO2. The time required to oxidize heme/TiO2 (or heme(py)2/TiO2) to hemin/TiO2 was found to be greater than or equal to the time required for the initial reduction, hemin/TiO2 + e- f heme/ TiO2. This was opposite to the behavior observed for Ru(dcbq)(bpy)2/TiO2, where reduction was always much slower than the reoxidation. Figure 7 shows the time required for reduction and

6826 J. Phys. Chem. B, Vol. 111, No. 24, 2007

Figure 6. Absorption spectra of a heme(py)2/TiO2 electrode recorded every 60 s after stepping the potential from -1.61 V to -0.46 V. The arrows indicate the direction of the absorption change that corresponds to the oxidation of heme(py)2/TiO2 to hemin(py)2/TiO2. The inset shows the absorption change at 417 nm of a hemin/TiO2 electrode after the applied potential was stepped from -0.46 V to -1.61 V (black squares), -1.41 V (red circles), -1.31 V (green triangles), and -1.21 V (blue upside down triangles). The absorption changes in the inset correspond to the reduction of hemin(py)2/TiO2 to heme(py)2/TiO2.

Figure 7. Absorption changes measured for three consecutive -0.61 to -1.31 to -0.61 V potential steps of (a) TiO2 (800 nm) in 0.1 M TBAH-MeOH, (b) Ru(dcbq)(bpy)2/TiO2 (470 nm) in 0.1 M TBAPCH3CN, and (c) hemin/TiO2 (420 nm) in 0.1 M TBAH-MeOH. The spectral changes were assigned to reduction of (a) TiO2 to TiO2(e-), (b) Ru(dcbq)(bpy)2/TiO2 to [Ru(dcbq-)(bpy)2]-/TiO2, and (c) hemin/ TiO2 to heme/TiO2.

reoxidation of (a) TiO2, (b) Ru(dcbq)(bpy)2/TiO2, and (c) hemin/ TiO2. The TiO2 data was measured in methanol electrolyte, however similar behavior (i.e., reduction faster than oxidation) was also observed in acetonitrile electrolytes. For hemin/TiO2 and TiO2 alone, the magnitude of the absorption change decreased with increasing cycling. Interestingly, for both heme/ TiO2 and heme(py)2/TiO2 the apparent diffusion constants were a function of the number of times the electrodes were cycled between the reduced and oxidized forms. The Dapp values for reduction of hemin/TiO2 to heme/TiO2 were found to decrease with cycling while the Dapp values for the reverse process increased. Discussion The results obtained in this study provide compelling evidence for a conduction band mediated electron-transfer process. Unlike hole transfer reactions,3-6 no percolation threshold was observed and the rates actually increased slightly with lower surface

Staniszewski et al. coverages. With heme surface coverages as low as 1/100 of that expected for a full monolayer, complete reduction of all the surface bound hemes was observed. The apparent diffusion constants were consistently 3-10 times larger than those previously reported for hole transfer in related TiO2 materials3-6 and for electron transport on insulating Al2O3.6 The apparent diffusion constants showed a strong bias dependence that is not easily rationalized without invoking conduction band states. Finally, the characteristic absorption of TiO2 electrons was observed under conditions where rapid electron transfer occurred.20 In addition, the reduction rates could be tuned by electrolyte modification. Taken together, these observations are most consistent with a conduction band mediated reduction mechanism. Interestingly, the rates for reduction and reoxidation were found to be different and dependent on the redox active molecule, behavior that has not been previously reported for hole transport. This data and the measured reduction potentials are given in Table 3. Below we discuss these findings in the context of previous literature results. The anaerobic reduction of TiO2 by chemical, electrochemical, and photochemical methods yields a bluish-black colored material.24 The spectroscopic and redox properties of reduced anatase TiO2 have been attributed to conduction band electrons20 and TiIII states.25 We refer to the reduced nanocrystalline thin film materials studied here simply as TiO2(e-). Rothenberger and co-workers have proposed an accumulation layer model and have utilized it to estimate anatase flat band potentials in organic electrolytes.20 A key assumption in this model is that the TiO2 band edges remain fixed as the Fermi energy of the electrode is raised. In the electrochemical reduction of anatase thin films, the onset potential was dependent upon the TiO2 electrolyte and solvent. Flat band potentials have been estimated to be -1.2 V in methanol and -2.04 V vs SCE in acetonitrile with tetra-alkyl ammonium salts.26,27 Assuming that the TiO2 is intrinsically doped, the conduction band edge, Ecb, would be a few hundred millivolts more negative. Interestingly, a strong cation dependence to the flat band potentials was noted in acetonitrile but not in methanol. With Li+ cations in acetonitrile, the flat band potential was -1.0 V vs SCE. It remains unknown how the binding of dye molecules or other compounds influences the energetic position of the conduction band. One might anticipate that the reactions with carboxylic acid groups, such as those present in hemin, would shift the conduction band edge positive on an electrochemical scale.2 In methanol, it is clear that the conduction band edge is more negative (closer to the vacuum level) than the FeIII/II reduction potentials of the surface bound hemes. For both hemin and hemin(py)2 anchored to TiO2, band gap excitation results in the reduction of the iron center to the +2 formal oxidation state. Methanol serves as a sacrificial donor to consume photogenerated valence band holes. With prolonged light excitation, the characteristic absorption spectrum of TiO2(e-) appears but only after all the iron centers were reduced.17 This is clear evidence that the reaction TiO2(e-) + hemin f TiO2 + heme is thermodynamically favored. The energetic position of Ecb relative to the reduction potential of Ru(dcbq)(bpy)2 was found to be ionic strength dependent in acetonitrile. Visible light excitation of Ru(dcbq)(bpy)2 in the presence of sacrificial donors, such as triethylamine, lead to the appearance of the reduced compound in acetonitrile, [Ru(dcbq-)(bpy)2]-/TiO2.10 Likewise pulsed light excitation of Ru(dcbq)(bpy)2/TiO2 in neat acetonitrile transiently generated this same reduced compound.9,10 The electrochemical reduction in acetonitrile electrolytes showed the intense absorption of

Conduction Band Mediated Electron Transfer

J. Phys. Chem. B, Vol. 111, No. 24, 2007 6827

TABLE 3: Summary of Reduction Potentials and Apparent Diffusion Constants Dapp (m2/s)a

compound

E1/2 (V vs Fc+/0)

electrolyte

C+e fC

Ru(dcbq)(bpy)2/TiO2 hemin/TiO2 hemin(py)2/TiO2

-1.01b -0.68c -0.47c

0.1 M TBAP-CH3CN 0.1 M TBAH-MeOH 0.1 M TBAH-MeOH

(8 ( 2) × 10-12 (5 ( 2) × 10-12 (7 ( 2) × 10-12

-

-

C- f C + e (7 ( 2) × 10-11 (2 ( 1) × 10-13 (2 ( 1) × 10-13

a The apparent diffusion constants calculated with eq 3 in the indicated electrolytes. The errors represent standard deviations from at least three separate measurements. The potentials were stepped from -0.61 to -1.31 and back for Ru(dcbq)(bpy)2/TiO2 and hemin/TiO2. For hemin(py)2/TiO2 the final potential was +0.54 V vs Fc+/0. b The reduction potential for a Ru(dcbq)(bpy)2/TiO2 electrode as described in ref 10. c Half-wave potentials measured in fluid solution with the indicated electrolyte.

[Ru(dcbq-)(bpy)2]- which obscured the much weaker TiO2(e-) absorbance in the visible and near IR regions (400-1100 nm). However, visible photolysis in the presence of 0.1 M LiClO4 and the sacrificial donor resulted in only the buildup of TiO2(e-) with no evidence for the reduced ruthenium compound. These results strongly suggest that in neat acetonitrile Ecb is more negative on an electrochemical scale than the first reduction of Ru(dcbq)(bpy)2, Eo ) -1.01 V.10 However, at high ionic strength, like the conditions used for reductive electrochemistry measurements, both TiO2(e-)s and the reduced compounds were present. Based on literature Ecb values, one might anticipate that conduction band mediated reduction of Ru(dcbq)(bpy)2/TiO2 would require application of very negative potentials to access the conduction band states (∼ -2.2 V and -1.0 V vs SCE in 0.1 M TBAP and LiClO4 in acetonitrile, respectively). However, this was not observed. In fact, electron transfer occurred at much more positive potentials. Furthermore, electron transfer was faster in the presence of TBA+ then Li+ with surface coverages well below the percolation threshold, Figure 3. Such data cannot be explained by literature values of the energetic position of Ecb on unfunctionalized TiO2. Instead there appears to be a sufficient density of TiO2(e-) present at -1.00 V vs SCE to reduce Ru(dcbq)(bpy)2 in both Li+ and TBA+ containing acetonitrile electrolytes. These more easily reduced TiO2 states could have been introduced when the sensitizers were anchored to the nanoparticle surface. In any case, the faster rates measured in the presence of TBA+ compared to Li+ may reflect a more favorable outer-sphere reorganization energy for electron transfer. Lithium cations are strong Lewis acids that are known to adsorb and intercalate anatase TiO2.28,29 The bulky TBA+ cation, on the other hand, has not been shown to adsorb on TiO2 and is too large to intercalate into the anatase lattice. We note that intercalation of cations in materials such as WO3 has been attributed to sluggish redox chemistry and slow time constants for color changes in electrochromic windows.30 The different rates for reduction and reoxidation of the surface bound compounds can be attributed to involvement of the TiO2 conduction band. The electrochemical reduction of unfunctionalized TiO2 was found to be slower than the reoxidation in both methanol and acetonitrile electrolytes. Similarly with Ru(dcbq)(bpy)2/TiO2, the apparent diffusion constants for reduction were an order of magnitude slower than the corresponding oxidation. This data is consistent with TiO2(e-) and TiO2 mediation of these processes respectively. However, with hemes, the situation was reversed: the apparent diffusion constants for hemin + TiO2(e-) f were always greater than or equal to those measured for heme + TiO2 f. A possible explanation for this was that different axial ligands were present resulting in both high- and low-spin iron centers. In methanol, the iron is likely to be high spin for both the FeII and FeIII formal oxidation states. Coordination of pyridine to heme yields six-coordinate lowspin FeII compounds22,23 that show the same electrochemical behavior. Therefore, while changes in coordination number and

SCHEME 2

spin state may accompany the FeIII/II redox chemistry, these changes alone are not sufficient to explain the sluggish heme oxidations rates. A Gerischer-type model consistent with the experimental data is shown in Scheme 2.31 Interfacial electron transfer is isoenergetic and will not occur unless there is significant overlap of the conduction band with the energy levels of the reduced or oxidized compounds. It is likely a sharp conduction band edge does not exist for these nanocrystalline thin films and the scheme is written to reflect this.1,2 For Ru(dcbq)(bpy)2/TiO2, overlap exists for both the reduced (W(Ru(dcbq-)L2)-) and the oxidized (W(Ru(dcbq)L2)) levels such that the conduction band can mediate both reduction and oxidation, Scheme 2a. Reduction of Ru(dcbq)(bpy)2/TiO2 to [Ru(dcbq-)(bpy)2]-/TiO2 is dependent on the concentration of TiO2(e-) and is hence expected to be potential dependent. The oxidation of [Ru(dcbq-)(bpy)2]-/ TiO2 to Ru(dcbq)(bpy)2/TiO2 occurs more rapidly due to the high density of unfilled conduction band states. The formal reduction potentials of hemin/TiO2 and hemin(py)2/TiO2 (Eo(FeIII/II)) are more positive relative to the conduction band, Scheme 2b. There is still sufficient overlap for FeIII to FeII reduction; however, oxidation is unfavored. In fact, if a welldefined conduction band edge existed for these materials, it might have been possible to trap all the hemes in the FeII formal oxidation state, even when the Fermi energy of the TiO2 was very positive. The sluggish heme oxidation observed presumably occurs through a relatively low density of TiO2 surface states that have appropriate energetics. For molecules like Ru(dcbq)(bpy)2, whose reduction potential lie just above Ecb in acetonitrile electrolytes, TiO2(e-) and TiO2 can mediate both the reduction and oxidation reactions, respectively. Conclusions Clear evidence for conduction band mediated electron transfer was obtained from a spectroelectrochemical analysis of the reduction of hemin and Ru(deebq)(bpy)2(PF6)2 anchored to nanocrystalline TiO2 thin films. Complete reduction was observed under conditions where the surface coverage was about 1/100 of that expected for a monolayer. There was no evidence for a percolation threshold like that previously reported for hole transfer.3-6 The apparent diffusion constants were 3-10 times larger than hole transfer on these same materials. The Ru2+/+ self-exchange that have been measured in fluid solution are therefore of no obvious relevance to these TiO2 interfaces.14,15

6828 J. Phys. Chem. B, Vol. 111, No. 24, 2007 In fact, there was no evidence for intermolecular electron transfer across the nanoparticle surfaces, and all the data can be adequately explained with TiO2(e-) as the reducing agent. The results suggest that current rectification will occur for molecules with formal reduction potentials at the conduction band edge: fast conduction band mediated reduction and slow oxidation that is presumably mediated by surface states. We emphasize that rectification of this type is not a special property of nanocrystalline TiO2 but has been observed at polycrystalline and single-crystal TiO2 electrodes.32 The results suggest that the conduction band may indeed mediate lateral electron transfer for charge separated states generated with visible light, Scheme 1.9,10 The color changes that can be rapidly induced with an applied potential may have applications as displays and as photoand electrochromic windows.8 In addition, the high concentration of TiO2(e-) may be exploited for multi-electron-transfer photoand/or electro-catalysis.17 Acknowledgment. This research was supported by grants from the Division of Chemical Sciences, Office of Basic Energy Sciences, Office of Energy Research, U.S. Department of Energy (DE-FG02-96ER14662). References and Notes (1) Gra¨tzel, M. Nature 2001, 414, 338. (2) Meyer, G. J. Inorg. Chem. 2005, 44, 6852. (3) Heimer, T. A.; D’Arcangelis, S. T.; Farzad, F.; Stipkala, J. M.; Meyer, G. J. Inorg. Chem. 1996, 35, 5319. (4) Bonhote, P.; Gogniat, E.; Tingry, S.; Barbe, C.; Vlachopoulos, N.; Lenzmann, F.; Comte, P.; Gra¨tzel, M. J. Phys. Chem. B 1998, 102, 1498. (5) Trammel, S. A.; Meyer, T. J. J. Phys. Chem. B 1999, 103, 104. (6) Wang, Q.; Zakeeruddin, S. M.; Nazeeruddin, M. K.; HumphryBaker, R.; Gra¨tzel, M. J. Am. Chem. Soc. 2006, 128, 4446. (7) Cummins, D.; Boschloo, G.; Ryan, M.; Corr, D.; Rao, S. N.; Fitzmaurice, D. J. Phys. Chem. B. 2000, 104, 11449.

Staniszewski et al. (8) Gra¨tzel, M. Nature 2001, 409, 575. (9) Hoertz, P. G.; Thompson, D. W.; Friedman, L. A.; Meyer, G. J. J. Am. Chem. Soc. 2002, 124, 9690. (10) Hoertz, P. G.; Staniszewski, A.; Marton, A;, Higgins, G. T.; Incarvito, C. D.; Rheingold, A. L.; Meyer, G. J. J. Am. Chem. Soc. 2006, 128, 8234. (11) Shockley, W.; Queisser, H. J. J. Appl. Phys. 1961, 32, 510. (12) Kohlrausch, R. Leibig’s Ann. Chem. 1847, 5, 430. (13) Williams, G.; Watts, D. C. Trans. Faraday Soc. 1970, 66, 80. (14) Chou, M.; Creutz, C.; Sutin, N. J. Am. Chem. Soc. 1977, 99, 5615. (15) Sutin, N.; Creutz, C. AdV. Chem. Ser. 1978, 168, 1. (16) Stromberg, J. R.; Wnuk, J.; Pinlac, R. A. F.; Meyer, G. J. Nano Lett. 2006, 6, 1284. (17) Obare, S. O.; Ito, T.; Meyer, G. J. J. Am. Chem. Soc. 2006, 128, 712. (18) Bhugun, I.; Lexa, D.; Saveant, J. J. Am. Chem. Soc. 1996, 118, 1769. (19) Grass, V.; Lexa, D.; Saveant, J. J. Am. Chem. Soc. 1997, 119, 7526. (20) Rothenberger, G.; Fitzmaurice, D.; Gra¨tzel, M. J. Phys. Chem. 1992, 96, 5983. (21) Obare, S. O.; Ito, T.; Meyer, G. J. EnViron. Sci. Technol. 2005, 39, 6266. (22) Kadish, K. M.; Tabard, A.; Lee, W.; Liu, Y. H.; Ratti, C.; Guilard, R. Inorg. Chem. 1991, 30, 1542. (23) Brault, D.; Rougee, M. Biochemistry 1974, 13, 4591. (24) Semiconductor Electrodes; Finklea, H. O., Ed.; Studies in Physical and Theoretical Chemistry 55; Elsevier: New York, 1988. (25) Cao, F.; Oskam, G.; Searson, P. C.; Stipkala, J.; Farzhad, F.; Heimer, T. A.; Meyer, G. J. J. Phys. Chem. 1995, 99, 11974. (26) Redmond, G.; Fitzmaurice, D. J. Phys. Chem. 1993, 97, 1426. (27) Enright, B.; Fitzmaurice, D. J. Phys. Chem. 1994, 98, 6195. (28) Lunnell, S.; Stashans, A.; Ojamae, L.; Lindstrom, H.; Hagfeldt, A. J. Am. Chem. Soc. 1997, 119, 7374. (29) Kavan, L.; Kratochilova, K.; Gra¨tzel, M. J. Electranal. Chem. 1995, 394, 93. (30) Large-Area Chromogenics: Materials and DeVices for Transmittance Control; Lampert, C., Cranquist, C., Eds.; SPIE Optical Engineering Press: Washington, DC, 1990. (31) Gerischer, H. Pure Appl. Chem. 1980, 52, 2649. (32) Frank, S.; Bard, A. J. Am. Chem. Soc. 1975, 97, 7427.