TiO2 Interfaces - The Journal of

Nov 13, 2009 - ... in TiO2 nanoparticles, an ionic transport component, and injection of electrons into TiO2 from states of solitary Pd−P molecules ...
0 downloads 0 Views 746KB Size
21090

J. Phys. Chem. C 2009, 113, 21090–21096

Charge Separation at Pd-Porphyrin/TiO2 Interfaces P. Zabel* and Th. Dittrich Helmholtz-Centre for Materials and Energy, Glienicker Strasse 100, D-14109 Berlin, Germany

M. Funes, E. N. Durantini, and L. Otero Departamento de Quı´mica, UniVersidad Nacional de Rı´o Cuarto, Agencia Postal No. 3, X5804 BYA, Rı´o Cuarto, Argentina ReceiVed: June 15, 2009; ReVised Manuscript ReceiVed: September 29, 2009

Charge separation has been investigated at Pd-porphyrin (Pd-P)/TiO2 interfaces by surface photovoltage spectroscopy in the Kelvin-probe arrangement. Ultrathin nanoporous TiO2 layers were covered with Pd-P molecules from highly diluted organic solution starting from the submonolayer range. Mechanisms of charge separation including charge separation in TiO2 nanoparticles, an ionic transport component, and injection of electrons into TiO2 from states of solitary Pd-P molecules not interacting with each other on the surface and from Pd-P molecules interacting with each other and with TiO2 were identified. A developed model for the simulation of surface photovoltage spectra considered mechanisms of charge separation and respective recombination paths. Characteristic charge separation lengths and recombination coefficients were obtained for the different components of separated charge. I. Introduction Interactions and electronic states at interfaces between organic layers with π-electron systems and metal oxides are of fundamental interest and important for a broad area of applications including organic-inorganic hybrid electronics1,2 and photovoltaics.3-5 Porphyrins and their metal complexes, for example, are widely used for sensing,6-9 as light receptors in energy conversion,10 and in molecular electronics and for spectral sensitization of wide band gap semiconductors.11,12 Electronic states at metal oxide/organic layer interfaces with π-electron systems should strongly influence exciton dissociation, recombination, and charge separation. The distribution and behavior of electronic states at metal oxide or semiconductor/ organic layer interfaces depend on chemical interactions as well as polarization effects. There are numerous unresolved questions concerning the nature of interactions and dominating mechanisms of charge separation at organic/inorganic interfaces. It has been shown that charge separation behavior at the interface can be significantly different from the optical density spectrum13 and that the structure of porphyrin layers is different at the interface.14 It is therefore of interest to investigate monolayers and submonolayers of porphyrins. The change of electronic states and their role in charge separation for a metal oxide/organic layer system is the topic of this work. As a model system Pd-porphyrin [palladium(II) 5-(4-carboxyphenyl)-10,15,20-tris(4-methylphenyl)porphyrin;15 see also the inset of Figure 1a] molecules were deposited from a highly diluted organic solution on ultrathin nanoporous TiO2 (np-TiO2) from local submonolayer coverage to coverage by several monolayers. The advantage of the Pd-porphyrin (Pd-P) molecule used here is that its electronic transitions show well-defined so-called Q- and S-bands with only one significant peak with maxima at 2.38 and 2.98 eV (in THF)15 [2.37 and 2.97 eV (in pyridine)16], first, and that it has carboxyl groups * To whom correspondence should be addressed. E-mail: philipp.zabel@ helmholtz-berlin.de.

Figure 1. (a) Optical density spectra of 10-5 M Pd-centered porphyrin (Pd-P) in DCM solution (open circles) and of the same solution after a sample was soaked with an ultrathin nanoporous TiO2 (np-TiO2) layer for 300 min (open triangles). The inset shows the structure of the palladium 5-(4-carboxyphenyl)-10,15,20-tris(4-methylphenyl)porphyrin molecule and the amount of Pd-P molecules adsorbed on np-TiO2 in terms of the surface concentration when assuming the increase of the internal surface area is 10-fold. (b) Normalized optical density spectra of Pd-P in DCM solution (open circles) and of the differences between the untreated np-TiO2 substrate and np-TiO2 soaked for 300 (filled triangles) and 1 (stars) min. The inset shows the optical densities at 2.9 eV of the np-TiO2 samples as a function of the soaking time.

for binding at TiO2 surfaces,17 second. The application of ultrathin np-TiO2 layers allowed performance of homogeneous deposition at an optical density high enough for simple analysis. Surface photovoltage spectroscopy (SPS) in the Kelvin-probe arrangement18 was applied to the investigation of electronic

10.1021/jp905575j CCC: $40.75  2009 American Chemical Society Published on Web 11/13/2009

Charge Separation at Pd-Porphyrin/TiO2 Interfaces

J. Phys. Chem. C, Vol. 113, No. 50, 2009 21091

states, from which light-induced excitation of charge carriers led directly to their separation. The surface photovoltage (SPV) is similar to the voltage of a charged parallel plate capacitor:19

SPV )

d Q εrelε0

(1)

where d is the distance between the centers of negative and positive charges separated in space, ε0 is the dielectric constant of a vacuum, εrel is the relative dielectric constant, and Q is the amount of charge separated in space (C/cm2). The dielectric constant depends also on the mechanism of charge separation since different materials may be involved. From the experimental point of view effects introduced by changes in d or εrel can be hardly distinguished. Therefore, the ratio d/εrel was considered as an effective charge separation length. A model was developed in which SPV spectra were simulated by taking into account various mechanisms of charge separation and respective recombination processes. Charge separation was considered to be significantly faster than recombination, and transport equations were not taken into account in the model due to a limited set of measurable parameters. From this point of view transport-related properties appeared indirectly in parameters describing recombination rates and effective charge separation lengths. SPV spectra were calculated from the optical density spectra by adjusting the effective charge separation lengths and parameters of recombination coefficients of different processes. II. Experimental Section Ultrathin np-TiO2 layers were deposited by spin coating from a suspension (the average diameter of TiO2 nanoparticles was 20 nm, Solaronix) on glass coated with conductive SnO2:F (TEC15). The thickness and the roughness factor of the npTiO2 layers were about 150 and 10-15 nm, respectively. The area of the samples was 1 in. by 1/2 in. The substrates were fired in air at 450 °C for 30 min before soaking in 10-5 M Pd-P in dichloromethane (DCM; Sigma-Aldrich) solution. Pd-P was obtained as previously described.15 The soaking time (tsoak) varied between 8 s and 300 min. Each sample was soaked in a separate solution with a volume of 5 mL each. The optical transmission spectra of the 10-5 M Pd-P in DCM solution and of the solutions after soaking for different times were measured. For this purpose the solutions were filled into a quartz cuvette. A quartz prism monochromator (SPM2, Carl Zeiss Jena) with a Si photodiode was used for the transmission measurements. Figure 1a shows the optical density (OD) spectra of the solution before and after soaking for 300 min. For the given soaking time, the OD decreased by about 50% due to adsorption of Pd-P molecules on the substrate. The amount of Pd-P molecules adsorbed on np-TiO2 was calculated from the decrease of the Pd-P concentration in the solution. The surface concentration of Pd-P molecules on npTiO2 was calculated from the amount missing from the Pd-P solution after soaking and by taking into account the sample geometries and increased surface area by a factor of 10. The inset of Figure 1a shows the tsoak dependence of the Pd-P surface concentration on np-TiO2. The Pd-P surface concentration increased from about 1013 cm-2 for tsoak ) 1 min to about 3 × 1014 cm-2 for tsoak ) 300 min. The Pd-P surface concentration depends on tsoak by a power law with a power coefficient of 0.5. This indicates that the deposition of Pd-P molecules on np-TiO2 is limited by molecular diffusion in the

solution. The specific area of a Pd-P molecule is about 1 nm2.12 Therefore, the (local) coverage of TiO2 nanoparticles varied from the submonolayer range to the order of several monolayers. The optical transmission was measured on the np-TiO2 substrates before and after soaking in Pd-P solution. The OD was taken as the difference in optical transmission between the soaked and unsoaked np-TiO2 substrates. Figure 1b compares normalized OD spectra of the 10-5 M Pd-P solution with npTiO2 after soaking for 1 and 300 min. The OD spectra are broadened for the np-TiO2 samples soaked in Pd-P solution, and the broadening increases with increasing tsoak. The OD of soaked np-TiO2 increased with tsoak by a power law with a power coefficient of 0.5, giving additional evidence for the wellcontrolled increase of coverage with Pd-P molecules. Surface photovoltage spectra were measured with a Kelvin probe (Besocke Delta Phi). A light-induced change of the contact potential difference (∆CPD) between the sample surface and the vibrating gold mesh was obtained. Since there is no photochemical activity, ∆CPD is equal to SPV.18 The diameter of the vibrating gold mesh was 3 mm. Illumination was performed under conditions similar to those used for optical transmission measurements. The spectrum of the excitation light was measured with a pyroelectric detector (EMM). The height of the exit slit of the monochromator was limited to 1.5 mm. The excitation light was focused with a quartz lens on the vibrating gold mesh of the Kelvin probe placed in a vacuum chamber. The intensity of the excitation light was calibrated with a Si photodiode (Hamamatsu) placed behind the vibrating gold mesh instead of a sample. Before the SPV measurement was started, the vacuum chamber was evacuated to 2 × 10-3 mbar and then flooded with Ar to 600-700 mbar. The samples were kept in an Ar atmosphere for about 15 min before the measurement was started. The measurements started at the lowest photon energy in the dark with a closed exit slit of the monochromator. The exit slit was opened at a photon energy of 0.8 eV. The spectra were measured between 0.5 and 3.3 eV with a step of 20 meV. The total measurement time was 1200 s for the spectral range between 0.9 and 3.3 eV for all samples. III. Surface Photovoltage Measurements A. Passivated np-TiO2. Figure 2a shows the ∆CPD spectrum of np-TiO2 soaked in pure DCM for 10 min and measured, as compared with the np-TiO2 soaked in Pd-P solution, between 0.8 and 4 eV with a step of 50 meV. The value of ∆CPD started to decrease at a photon energy of 3.2 eV due to separation of photogenerated charge carriers in np-TiO2. The photon energy of 3.2 eV corresponds well to the onset of optical absorption in np-TiO2.20 The decrease of ∆CPD is caused by a separation of photogenerated electrons toward the SnO2:F back contact. It has to be pointed out that there was no evidence for charge separation caused by transitions due to defect states in the forbidden band gap of TiO2. This seems surprising in comparison with previous SPV measurements on np-TiO2 samples21 which were not soaked in DCM and shows that TiO2 surfaces can be passivated by DCM, avoiding the formation of surface states in oxygen and/or water-deficient ambience. B. np-TiO2 after Soaking for 8 s. Figure 2b shows the ∆CPD spectrum of np-TiO2 soaked in Pd-P solution for 8 s. At photon energies below 2 eV this sample showed an upward drift of the ∆CPD, indicating that the sample surface had not yet stabilized. At 2 eV, ∆CPD increased more sharply than the drift. Therefore, 2 eV corresponds to the onset of an optical transition leading to separation of photogenerated positive charges toward the SnO2:F back contact. At about 2.25 eV the

21092

J. Phys. Chem. C, Vol. 113, No. 50, 2009

Figure 2. Contact potential difference spectra of np-TiO2 soaked in pure DCM (a) and of np-TiO2 soaked in 10-5 M Pd-P in DCM solution for 8 s (b), for 10 min (c), and for 100 min (d).

∆CPD started to decrease, which is the signature of a transition causing charge separation of photogenerated electrons toward the SnO2:F back contact. At about 2.37 eV, the value of ∆CPD started to increase again; i.e., a second optical transition causing a preferential separation of photogenerated positive charges toward the SnO2:F back contact started. Above 2.5 eV, the ∆CPD signal tended to saturate and started to decrease at 2.7 eV, marking the onset of a second optical transition causing charge separation of photogenerated electrons toward the SnO2:F back contact. A tendency to saturation was observed for photon energies larger than 3.06 eV, and at 3.19 eV the value of ∆CPD started to decrease sharply due to the fundamental absorption in TiO2 nanoparticles and separation of photogenerated electrons toward the SnO2:F back contact. Optical absorption of the Q- and S-bands of Pd-P molecules in solution starts at about 2.3 and 2.8 eV. The charge separation onsets observed at 2.0 and 2.37 eV and at 2.25 and 2.7 eV belong to transitions from states related to shifted energy positions of the Q- and S-bands in the solution. Therefore, the observed transitions of np-TiO2 soaked for 8 s are related to at least three mechanisms of charge separation, including charge separation inside TiO2 nanoparticles. C. np-TiO2 after Soaking for 10 min. Figure 2c shows the ∆CPD spectrum of np-TiO2 soaked in Pd-P solution for 10 min. At photon energies below 2.1 eV, this sample showed a similar drift of ∆CPD to higher values. At 2.1 eV, ∆CPD started to decrease strongly and tended to saturate at photon energies above 2.41 eV. This region can be assigned to transitions caused by absorption in the Q-band of Pd-P molecules. At 2.7 eV, ∆CPD started to decrease again and tended to saturate at photon energies above 3.05 eV. This is the spectral range for transitions in the S-band of Pd-P. The transitions from states related to the Q- and S-bands led to preferential charge separation of electrons toward the SnO2:F back contact. Charge separation related to absorption in TiO2 nanoparticles started at 3.2 eV. As a remark, the behavior of np-TiO2 soaked for 1 min in Pd-P solution was very similar to that of the sample shown here.

Zabel et al. The differences of the ∆CPD values at 2.06 and 2.45 eV as well as at 2.66 and 3.1 eV amounted to 0.06 and 0.035 V, respectively, and these values correspond to the surface photovoltage under the given illumination conditions. The photon fluxes at 2.2 and 2.8 eV were 1.6 × 1014 and 3 × 1013 cm-2 s-1, respectively. The ratio between the ODs of the Q- and S-bands is about 0.08. Considering the photon fluxes and an identical mechanism of charge separation for the transitions related to the Q- and S-bands, one would expect an SPV ratio of 0.5 between the values at 2.2 and 2.8 eV. However, the measured ratio was about 4 times larger. Therefore, an additional mechanism of charge separation that brings out an asymmetry is hidden in the ∆CPD spectra. D. np-TiO2 after Soaking for 100 min. Figure 2d shows the ∆CPD spectrum of np-TiO2 soaked in Pd-P solution for 100 min. This sample did not show a drift of ∆CPD at lower photon energies. The ∆CPD signal started to decrease already at about 1.6 eV. A shoulder appeared in the ∆CPD spectrum at 2.17 eV. At photon energies larger than 2.34 eV the ∆CPD signal started to increase and tended to saturate at about 2.5 eV. The ∆CPD signal decreased up to a photon energy of 2.85 eV and increased again at photon energies higher than 2.85 eV. No significant onset of the ∆CPD related to absorption in TiO2 nanoparticles was observed, probably due to broadening of the spectrum and/or superposition of different mechanisms of charge separation. Remarkably, the np-TiO2 soaked for 300 min in Pd-P solution showed a quite similar behavior. The OD of np-TiO2 soaked for 100 min in Pd-P solution had a minimum in the spectral range between 2.34 and 2.5 eV, where the ∆CPD signal increased. The appearance of an additional charge separation mechanism causing a blue shift of ∆CPD with an opposite sign in the range with very low OD seems unlikely. For one given transition, a ∆CPD spectrum can be understood in first approximation as a superposition of OD spectra and partially integrated OD spectra due to different behaviors of recombination in relation to the spectrum of the photon flux. Therefore, for np-TiO2 with tsoak ) 100 min a much higher recombination rate than for np-TiO2 with tsoak ) 10 min is assumed in the following considerations. IV. Mechanisms and Model of Charge Separation A. Mechanisms of Charge Separation. Figure 3a shows a schematic of the layer illustrating the asymmetry, which results in a measurable surface photovoltage. Contrary to the effective bulk region, in the external surface region the coverage of npTiO2 with Pd-P is asymmetric. This causes a preferential movement of the negative center of charge toward the back contact when electrons are injected from the molecules into the TiO2. The observed light-induced changes of the contact potential can be well described assuming four mechanisms of charge separation (see Figure 3). The first charge separation mechanism is related to the photogeneration of electrons and holes in TiO2 nanoparticles and the preferential separation of electrons toward the SnO2:F back contact (Figure 3b). The remaining three mechanisms may explain the dominating effects related to the Q- and S-bands. The initial process of charge transfer is caused by electron injection from Pd-P molecules into TiO2 nanoparticles, leading to a positive charge on a Pd-P molecule and a negative charge inside the TiO2 nanoparticle. Preferential separation of photogenerated positive charge toward the SnO2:F back contact appeared exclusively for npTiO2 soaked in Pd-P solution for 8 s. After annealing in air, the TiO2 nanoparticles are covered by hydroxide and phys-

Charge Separation at Pd-Porphyrin/TiO2 Interfaces

J. Phys. Chem. C, Vol. 113, No. 50, 2009 21093 the TiO2 nanoparticles have one common center of charge. Each component of positive charge has its own center of charge, and the SPV signal can be expressed as the following superposition:

SPV )

Figure 3. Schematic of the idealized layer structure (a) and of the mechanisms of charge separation assumed in the model (b-e). The asymmetry between external and internal surfaces of the ultrathin npTiO2 (L about 100 nm) layer causes differences in charge separation that result in a measurable surface photovoltage. The charge separation lengths dTiO2 (b), dH2O (c), dPd (d), and dPd′ (e) are perpendicular to the surface of the vibrating electrode, and they depict charge separation inside TiO2 nanoparticles (b), charge separation by ionic transport due to adsorbed water molecules (c), and charge separation by injection from states in the Pd-P molecules (d) and from states arising due to interaction between clustered Pd-P molecules and between Pd-P molecules and TiO2 (e).

isorbed water molecules. A very short dipping of np-TiO2 in the Pd-P solution was not sufficient to replace the majority of physisorbed water molecules. Therefore, ionic transport in a surface layer may play an important role. Water molecules are donorlike; i.e., they can form hydronium ions by giving an electron, e.g., to a positively charged Pd-P molecule. We assume that the dominating transport of protons on a homogeneous wet surface layer of interconnected nanoparticles caused the increase of ∆CPD for np-TiO2 soaked for 8 s (Figure 3c). Injection of photoexcited electrons from Pd-P molecules into TiO2 nanoparticles can appear from noninteracting (Q- and S-bands) and interacting Pd-P molecules (Q′- and S′-bands). This consideration opens the opportunity to introduce an asymmetry between light-induced changes of ∆CPD for the two bands and shifted onsets of photogeneration. Isolated Pd-P molecules are noninteracting (Figure 3d), and they dominate charge separation in the case of submonolayer coverage of TiO2 nanoparticles by Pd-P molecules. Interacting Pd-P molecules become important when clustering of molecules starts (Figure 3e). In such a case, the effective charge separation length may change significantly. Interaction between Pd-P molecules and TiO2 could not be distinguished from the intermolecular (Pd-P)-(Pd-P) interaction in our experiments. B. Basic Equations. The np-TiO2 layer is considered as an effective medium with a thickness L. The volume densities of the positive charges are related to the density in TiO2 nanoparticles (pTiO2), in a layer of adsorbed water molecules at TiO2 nanoparticles (pH2O), in states of noninteracting Pd-P molecules (pPd), and in states of interacting Pd-P molecules (pPd′). All negative charge carriers appear in the TiO2 nanoparticles so that there is a common density of electrons (nTiO2). The electrons in

[( )

L d ε0 εrel

TiO2

pTiO2 +

( ) d εrel

H2O

( ) ( ) ]

pH2O +

d εrel

d εrel

pPd +

Pd

pPd′

(2)

Pd′

The effective charge separation lengths may depend on time and on the distributions of carrier concentrations in general. However, in such a case it will be practically impossible to develop a suitable model of charge separation. Effective charge separation lengths are dominated by the given mechanism of charge separation. Therefore, we assume that the dependence on the distribution of carrier concentrations is a second-order effect. In the following, the effective charge separation lengths are assumed to be independent of time and distributions of carrier concentrations. The values of positive and negative charges depend on the generation and recombination rates. The positive charge in the layer of adsorbed water molecules at TiO2 nanoparticles can be considered as independent with a certain lifetime. The positive charges pTiO2, pPd, and pPd′ as well as the negative charge nTiO2 depend on each other. The generation rates are GTiO2, GPd, and GPd′ for absorption in TiO2 nanoparticles, states of noninteracting Pd-P molecules, and states of interacting Pd-P molecules, respectively. The recombination coefficients BTiO2, BPd, and BPd′ describe recombination between nTiO2 and pTiO2, nTiO2 and pPd, and nTiO2 and pPd′, respectively. Transport processes, such as electron diffusion across a TiO2 nanoparticle or hopping of holes across clusters of Pd-P molecules, are not considered explicitly. They are involved indirectly in the constant values of the effective charge separation lengths. The continuity equations can be written as

dpTiO2

) GTiO2 - BTiO2nTiO2pTiO2 dt dpPd ) GPd - BPdnTiO2pPd dt dpPd′ ) GPd′ - BPd′nTiO2pPd′ dt dnTiO2 ) GTiO2 - BTiO2nTiO2pTiO2 + GPd - BPdnTiO2pPd + dt GPd′ - BPd′nTiO2pPd′ (3)

The generation rates are given by

GTiO2 ) Φ(ODTiO2) GPd ) Φ(ODPd) GPd′ ) Φ(ODPd′) GH2O ) GPd′ ) Φ(ODPd′)

(4)

where Φ is the measured spectrum of the photon flux and

ODexptl ) ODPd + ODPd′

(5)

21094

J. Phys. Chem. C, Vol. 113, No. 50, 2009

Zabel et al.

Figure 4. Optical density spectra of the transitions of noninteracting (Q, S) and interacting (Q′, S′) Pd-P molecules for np-TiO2 soaked for 8 s, 10 min, and 100 min (a-c, respectively) in 10-5 M Pd-P in DCM solution. Optical densities for noninteracting and interacting molecules add up to the measured optical densities. The shapes of the individual optical density spectra followed from fitting of simulated SPV spectra.

ODexptl is the difference between the measured OD of soaked np-TiO2 and the measured OD of unsoaked np-TiO2. For transitions related to the Q- and S-bands for tsoak ) 8 s and for the transition related to the Q-band for tsoak ) 1 min the sensitivity of the measured OD spectra was insufficient so that the shapes of the OD of Pd-P in solution were taken. The superposition given in eq 5 provided a relatively large degree of freedom for fitting with Gaussians (usually three Gaussians were used for fitting of a transition related to the Q- or S-bands). For the given model, absorption below the band gap of TiO2 is important. Typical values of the absorption coefficient of nanocrystalline TiO222 were taken into account. ODTiO2 is described by an exponential function:

(

ODTiO2 ) 0.875 exp

pω - 3.4 Et

)

(6)

corresponding to an optical layer thickness of 100 nm. The energy of the exponential tail is denoted by Et with a typical value of 0.08 eV. One has to remark that eq 6 is rather simplified and may not exactly express the behavior of ODTiO2. The relations in eq 3 were integrated over time with a MATLAB/Simulink program, using an energy ramp corresponding to a measurement speed of 20 meV/10 s as an input to the photon flux and optical densities. The SPV signal was calculated from the resulting carrier concentrations according to eq 2. The carrier concentrations never exceeded 1017 cm-3. V. Charge Separation in the Pd-P/np-TiO2 System Parts a-c of Figure 4 show the optical density spectra used as input to calculate simulated ∆CPD spectra which fit to the measured ∆CPD spectra of np-TiO2 soaked in Pd-P solution for 8 s, 10 min, and 100 min, shown in parts b-d of Figure 2, respectively. The optical density spectra were selected as superpositions of Gaussians in such a way that the dominating

Q and S spectra added with the small Q′ and S′ spectra form the measured optical density spectra. In the simulation, the Q and S peaks were related to charge separation by injection of electrons from states in noninteracting Pd-P molecules into the TiO2. The Q peak maxima were located at 2.3 eV (8 s) and 2.35 eV (10 and 100 min). The S peaks had their maximum at 3.02 eV for all samples. A. Submonolayer Deposition of Pd-P on np-TiO2 with Residual Water. The Q′ (maximum at 2.15 eV) and S′ (at 2.44 eV) peaks in Figure 4a were related to charge transfer due to an interaction with physisorbed water molecules. This led to an increase of ∆CPD at 2.0 and 2.37 eV (Figure 2b). Exponential relaxation with a time constant of about 200 s was taken into account. The drift in the beginning of the ∆CPD spectrum is simulated as a linear function of time. B. Submonolayer Deposition of Pd-P on np-TiO2. In Figure 4b, the Q′ and S′ transitions that correspond to interacting states were red-shifted to 2.3 and 2.83 eV. Similar red shifts have been reported to be caused by formation of aggregates.23 The effective charge separation length for transitions related to interacting states (Q′ and S′ peaks) was about 1 order of magnitude larger than the effective charge separation length for transitions related to noninteracting states. The optical density of interacting states was larger for transitions related to the Q′band than for transitions related to the S′-band. The recombination coefficient of the interacting states BPd′ was about one-third of the recombination coefficient of the noninteracting states BPd, which was similar to the recombination inside the TiO2 (BTiO2 ≈ 10-21 cm3/s). BPd is the highest recombination coefficient; overall all coefficients are very similar. C. Clustering of Pd-P on np-TiO2. In Figure 4c, the Q′ peak was red-shifted to 2.10 eV, which was much stronger than for np-TiO2 soaked for 10 min. The S′ peak of np-TiO2 soaked for 100 min does not consist anymore of one peak with a welldefined position but of at least three peaks with positions at 2.52, 2.77, and 3.2 eV. The existence of the peak at 3.2 eV suggests a modification of the surface states at TiO2 nanoparticles close to the band gap due to interconnected Pd-P agglomerates. The Q′ and S′ peaks in the optical density spectrum are responsible for the largest changes in the ∆CPD spectrum. The optical density of the S′ peak is larger than that of the Q′ peak. The effective charge separation lengths of the noninteracting states (d/εrel)Pd and inside the TiO2 (d/εrel)TiO2 are of the same order of magnitude. The effective separation length of the interacting states (d/εrel)Pd′ is about an order of magnitude larger than both. The recombination coefficients of BPd and BPd′ are practically the same, whereas BTiO2 is about 1 order of magnitude larger than BPd and BPd′. D. Effective Charge Separation Lengths. Figure 5a compares the values of the effective charge separation lengths for the four separation mechanisms as a function of tsoak, i.e., local coverage of TiO2 nanoparticles with Pd-P molecules. The value of (d/εrel)H2O corresponding to charge separation via ionic transport was quite high compared to those of the other mechanisms, especially taking into account the high dielectric constant of water (εrelH2O ≈ 7824,25). The largest effective charge separation length was obtained for injection from interacting states in np-TiO2 soaked for 1 min [(d/εrel)Pd′ ) 7 × 10-8 cm)]. The effective charge separation length for the interacting states (d/εrel)Pd′ was about 1 order of magnitude larger than the effective separation length for the noninteracting states (d/εrel)Pd. The latter was about 1 order of magnitude larger than the effective separation length in TiO2 (d/εrel)TiO2 for the samples which are soaked for 8 s, 1 min, and 10 min. The estimated charge

Charge Separation at Pd-Porphyrin/TiO2 Interfaces

Figure 5. Dependence of the quotient between the charge separation length and the dielectric constant (a) and the recombination constant (b) on the soaking time of np-TiO2 in 10-5 M Pd-P in DCM solution.

separation lengths for the noninteracting states dPd and interacting states dPd’ are less than 1 and 10 nm, even though only the first ca. 1 nm of porphyrin layers on TiO2 contributes to charge separation.26 We point out that these are not local charge separation lengths but the difference between the centers of the distributions of positive and negative charges. A local charge separation length within one nanoparticle is significantly larger. Further, the resulting effective charge separation lengths are reciprocal to the estimated value of the effective layer thickness, which does not necessarily have to be similar to the layer thickness. While this limits the significance of the absolute values to the order of magnitude, the validity of the relative changes between samples remains unaffected. The lowest effective charge separation lengths were found for charge separation in TiO2 nanoparticles in np-TiO2 soaked for 8 s, 1 min, and 10 min [(d/εrel)TiO2 between 4 × 10-10 and 6 × 10-10 cm, resulting in d ≈ 0.1 nm for εrel ≈ 3327]. The value of (d/εrel)TiO2 increased to a range between 1.3 × 10-9 and 3 × 10-9 cm for np-TiO2 soaked for 50, 100, and 300 min. The local coverage with one monolayer of Pd-P molecules at TiO2 nanoparticles was reached between the soaking times of 10 and 50 min. Interacting states started to dominate the ∆CPD spectra and changes in electronic states related to TiO2 appeared after the longer soaking times. In conclusion, coverage of TiO2 nanoparticles with more than one monolayer of Pd-P molecules caused a stronger charge separation of electrons and holes toward the center and the surface of TiO2 nanoparticles, respectively. The effective charge separation length (d/εrel)Pd reached a minimum for np-TiO2 soaked for 50 min, (d/εrel)Pd ) 1.5 × 10-9 cm, which was even less than (d/εrel)TiO2 for the given soaking time. Therefore, the generation of additional defect states in TiO2 seems to decrease the effective charge separation length for injected electrons. However, (d/εrel)Pd increased again

J. Phys. Chem. C, Vol. 113, No. 50, 2009 21095 for the longest soaking time, probably due to additional charge transfer across Pd-P molecule clusters. The behavior of (d/ εrel)Pd′ was very similar to that of (d/εrel)Pd, pointing to the similar nature of carrier injection. E. Recombination Coefficients. Figure 5b shows the recombination coefficients BTiO2, BPd, and BPd′ corresponding to the three recombination mechanisms (eq 3) as a function of the soaking time in Pd-P solution. The recombination coefficient BTiO2 was 2 × 10-21, 1.5 × -21 10 , and 1 × 10-21 cm3/s for soaking times of 8 s, 1 min, and 10 min, respectively. The value of BTiO2 increased by 2 orders of magnitude to 2 × 10-19 and 4 × 10-19 cm3/s for soaking times of 50 and 100 min, respectively. However, BTiO2 decreased to 4 × 10-20 cm3/s for the longest soaking time of 300 min. At soaking times below 50 min the recombination coefficient BPd was about 3 times larger than BTiO2. The value of BPd increased strongly by a factor of 13 between the soaking times of 10 and 50 min; i.e., the increase was less by more than 1 order of magnitude in comparison to BTiO2. BPd decreased to 2 × 10-20 cm3/s after a soaking time of 100 min and increased to 3 × 10-19 cm3/s after a soaking time of 300 min. This qualitative behavior of BPd is just the opposite of the behavior of BTiO2. The values of BPd′ behaved quite similar to those of BPd for soaking times below 300 min. The value of BPd′ did not increase after 300 min. The values of BTiO2 are very low in comparison with conventional semiconductors. The reason for the very low values of BTiO2 is local charge separation and trapping. It is known from transient absorption measurements that trapped electrons or holes create deep electronic states in TiO2.28 The recombination probability of deeply trapped electrons and holes decreases exponentially with increasing distance between the trapped electron and hole. After annealing in air, the surfaces of TiO2 nanoparticles contain a large amount of -OH surface bonds acting as electron traps. More and more -OH surface bonds are transformed into binding groups of -O-(Pd-P) with increasing coverage of the TiO2 surface with Pd-P molecules. With the ongoing removal of -OH surface bonds, (d/εrel)TiO2 increased by a factor of 5 and BTiO2 increased by about 400fold. This was the most striking change in all experiments. For the longest soaking time the values of (d/εrel)TiO2 and BTiO2 decreased by a factor of 2 and 20, respectively. This fact seems to be caused by a new process appearing due to increasing charge transfer across interconnected clusters of Pd-P molecules. For charge separation from noninteracting and interacting Pd-P molecules the effective charge separation lengths did not increase between the soaking times of 10 and 50 min in contrast to the recombination coefficients. This is not surprising since charge separation by injection of electrons from surface molecules into TiO2 nanoparticles, unlike recombination, is practically unaffected by trapping. The values of BTiO2 and (d/εrel)TiO2 correlate quite well with each other. This is demonstrated in Figure 6. The correlation corresponded to a power law with a power coefficient close to 3. A three-dimensional model including local and effective charge separation as well as distance-dependent recombination will be needed for the interpretation of the power coefficient. As a remark, the values of BPd and (d/εrel)Pd and of BPd′ and (d/εrel)Pd′ did not show any comparable correlation. VI. Conclusions Photoexcited charge carriers can be separated in space in the Pd-P/np-TiO2 system. By simulating SPV spectra and fitting them to measured data, distributions of states from which

21096

J. Phys. Chem. C, Vol. 113, No. 50, 2009

Zabel et al. Pd-P molecules with holes remaining on Pd-P clusters, would increase the number of free parameters in the model. Additional experimental methods will be necessary to distinguish processes in a more detailed recombination model. Acknowledgment. We are grateful to Y. Shapira for critical remarks. The DAAD (Grant D/06/33824) and CONICET (L.O. and E.N.D.) are acknowledged for financial support. M.F. thanks ANPCYT for a research fellowship. References and Notes

Figure 6. Correlation between the recombination constants and the effective charge separation lengths for recombination and charge separation in TiO2.

separation takes place were obtained, as were effective charge separation lengths and recombination coefficients for involved processes. Surface conditioning of TiO2 nanoparticles is very important for the dynamics of charge separation and recombination. In our simple model, photoexcited electrons on Pd-P molecules could be separated only by injection into TiO2 nanoparticles and consistent parameters were obtained. Changes in the recombination coefficients and effective charge separation lengths could be explained by the role of adsorbed water molecules for motion of positive ionic charge, of surface -OH bonds as electron traps for limiting the motion of electrons in TiO2, and of clustering of Pd-P molecules for positive charge transfer. Electronic states, from which electrons photoexcited on Pd-P molecules could be injected into TiO2 nanoparticles, depended on specific interactions. The optical density was split into contributions from noninteracting and interacting molecules without specification of interaction details. The effective charge separation length was larger by 1 order of magnitude for excitation from interacting states in comparison to excitation from noninteracting states. This is not surprising assuming interacting states on clusters of molecules (decreased εrel in comparison to that of TiO2) and noninteracting states on isolated molecules. At the same time, this would mean that clusters of Pd-P molecules appeared even on np-TiO2 soaked for 1 and 10 min. Despite the low concentration, it is not unlikely that the dissolved Pd-P molecules are in aggregate form to some extent. It is also possible that molecular clusters are forming during adsorption on TiO2. The large increase of the recombination coefficients between soaking times of 10 and 50 min followed directly from the change of ∆CPD to more positive values at photon energies larger than 2.2-2.3 eV and larger than 2.7-2.8 eV for soaking times of 50, 100 and 300 min. Here, the disappearance of -OH surface bonds as electron traps was proposed as the reason for the strong increase of B. The implementation of additional processes, such as recombination of electrons photoexcited on

(1) Robertson, N.; McGowan, C. A. Chem. Soc. ReV. 2003, 32, 96– 103. (2) Tsuda, A.; Osuka, A. Science 2001, 293, 79–82. (3) Kay, A.; Gra¨tzel, M. J. Phys. Chem. 1993, 97, 6272–6277. (4) Wang, Q.; Campbell, W. M.; Bonfantani, E. E.; Jolley, K. W.; Officer, D. L.; Walsh, P. J.; Gordon, K.; Humphry-Baker, R.; Nazeeruddin, M. K.; Gra¨tzel, M. J. Phys. Chem. B 2005, 109, 15397–15409. (5) Imahori, H.; Kimura, M.; Hosomizu, K.; Fukuzumi, S. J. Photochem. Photobiol., A 2004, 166, 57–62. (6) Lee, S.-K.; Okura, I. Anal. Chim. Acta 1997, 342, 181–188. (7) Andersson, M.; Holmberg, M.; Lundstro¨m, I.; Lloyd-Spetz, A.; Ma˚rtensson, P.; Paolesse, R.; Falconi, C.; Proietti, E.; Di Natale, C.; D’Amico, A. Sens. Actuators, B 2001, 77, 567–571. (8) Pedrosa, J. M.; Dooling, C. M.; Richardson, T. H.; Hyde, R. K.; Hunter, C. A.; Martı´n, M. T.; Camacho, L. J. Mater. Chem. 2002, 12, 2659– 2664. (9) Suslick, K. S.; Rakow, N. A.; Kosal, M. E.; Chou, J.-H. J. Porphyrins Phthalocyanines 2000, 4, 407–413. (10) Kobuke, Y. Eur. J. Inorg. Chem. 2006, 2333–2351. (11) Fungo, F.; Otero, L.; Sereno, L.; Silber, J. J.; Durantini, E. N. J. Mater. Chem. 2000, 10, 645–650. (12) Gervaldo, M.; Fungo, F.; Durantini, E. N.; Silber, J. J.; Sereno, L.; Otero, L. J. Phys. Chem. B 2005, 109, 20953–20962. (13) Zidon, Y.; Shapira, Y.; Dittrich, Th.; Otero, L. Phys. ReV. B 2007, 75, 195327. (14) Zidon, Y.; Shapira, Y.; Shaim, H.; Dittrich, Th. Appl. Surf. Sci. 2008, 254, 3255–3261. (15) Scalise, I.; Durantini, E. N. J. Photochem. Photobiol., A 2004, 162, 105–113. (16) Rogers, J. E.; Nguyen, K. A.; Hufnagle, D. C.; McLean, D. G.; Su, W.; Gossett, K. M.; Burke, A. R.; Vinogradov, S. A.; Pachter, R.; Fleitz, P. A. J. Phys. Chem. A 2003, 107, 11331–11339. (17) Ma, T.; Inoue, K.; Yao, K.; Noma, H.; Shuji, T.; Abe, E.; Yu, J.; Wang, X.; Zhang, B. J. Electroanal. Chem. 2002, 537, 31–38. (18) Kronik, L.; Shapira, Y. Surf. Sci. Rep. 1999, 37, 1–206. (19) Mora-Sero´, I.; Dittrich, Th.; Garcia-Belmonte, G.; Bisquert, J. J. Appl. Phys. 2006, 100, 103705. (20) Wang, Z.; Helmersson, U.; Ka¨ll, P.-O. Thin Solid Films 2002, 405, 50–54. (21) Dittrich, Th.; Neumann, B.; Tributsch, H. J. Phys. Chem. C 2007, 111, 2265–2269. (22) Kumar, P. M.; Badrinarayanan, S.; Sastry, M. Thin Solid Films 2000, 358, 122–130. (23) Horiuchi, T.; Miura, H.; Sumioka, K.; Uchida, S. J. Am. Chem. Soc. 2004, 126, 12218–12219. (24) Uematsu, M.; Frank, E. U. J. Phys. Chem. Ref. Data 1980, 9 (4), 1291–1306. (25) Ferna´ndez, D. P.; Mulev, Y.; Goodwin, A. R. H.; Levelt Sengers, J. M. H. J. Phys. Chem. Ref. Data 1995, 24 (1), 33–69. (26) Kroeze, J. E.; Savenije, T. J.; Warman, J. M. J. Photochem. Photobiol., A 2002, 148, 49–55. (27) Busani, T.; Devine, R. A. B. Semicond. Sci. Technol. 2005, 20, 870–875. (28) Bahnemann, D.; Henglein, A.; Lilie, J.; Spanhel, L. J. Phys. Chem. 1984, 88, 709–711.

JP905575J