Photoinduced Interfacial Electron Transfer (Heterogeneous Electron Transfer)

Since their discovery in early 1990s, dye sensitized solar cells (DSSCs) have been extensively investigated because of their potential in wide commercial applications for conversion of sunlight to electricity. The key interaction of light and matter in DSSCs is light induced interfacial charge transfer from transition metal complex dye sensitizers to semiconductor nanoparticles.101 108 Extensive studies of energetics, kinetics and structural dynamics of the inter­facial charge transfer processes have generated a wealth of information,101 108 but structural evolution of the adsorbed dye sensitizer and the nanocrystal surface associated with the electron density shift during and after the interfacial charge injection remains vague and is actively pursued by several groups with theoretical modeling.109 111 The example below demonstrates the current syn­chrotron XTA capabilities in solving transient electronic and geometric structures of a ruthenium complex dye sensitizer undergoing interfacial pho­toinduced charge separation, mimicking the electron injection process in DSSCs. This is a heterogeneous system where the hybrid material composed of RuN3 and TiO2 nanoparticles is in a suspension that may be difficult to study by OTA because the decorated particles may form larger aggregates causing visible light scattering. XTA, on the other hand, is not sensitive to visible light transparency of the sample and hence is an ideal tool for studying such a system.

In a DSSC, Run(dcbpy)2(NCS)2 [or RuN3 or N3, dcbpy = bis(4,4′-dicar – boxy-2,2′-bipyridine] was attached to the surface of TiO2 nanocrystals in a film as a working electrode. The metal-to-ligand-charge-transfer (MLCT) state of RuN3 injects one electron into the conduction band of TiO2, leading to an interfacial charge separated state, (TiO2)n /RuN3+ (Figure 12.8). The electron

injection dynamics of this process has been characterized by a dual-exponential function with time constants of <100 fs and 1 ps to 100 ps depending on the sample environment.104,112,113 Because the current XTA time resolution at a synchrotron source is about 80 ps, only the RuN3 transient structure of the (TiO2)n /RuN3+ state after photoinduced interfacial electron injection can be captured and the structure of the RuN3 MLCT state itself could only be resolved with fs X-ray pulses from X-ray free electron source, such as the Linear Coherent Light Source (LCLS).

Figure 12.9 displays the XANES spectra as well as difference XANES spectra at the Ru K-edge without and with the laser pump pulse. At the 50 ps delay time, the excited RuN3 molecules have already completed electron in­jection into the nanoparticles and are mostly in the oxidized form, RuinN3+, according to OTA studies. The edge energy of the laser-on XANES spectrum is slightly shifted to a higher energy as expected for RuIIIN3+ after the electron injection to the TiO2 lattice.

The data analysis was complicated by simultaneously having two types of ligand (NCS and dcbpy) and two different states (the ground and charge separate states) in RuN3 at this delay time. Because the core-hole lifetime decreases as the atomic number of the element increases,114 the bandwidth of a second and third row transition metal is broader than those of the first row transition metals. Consequently, the transition edges of high Z element appear to rise more slowly with an increase in the incident X-ray photon energy, which smears characteristic transition features. By taking the difference spectra between the laser pump without and with pulses, the changes caused by the generation of the transient species become more distinct, as seen in Figure 12.9. Therefore, we chose to use the multidimensional interpolation approach (MIA)42 implemented in the FitIt code43 and full multiple scattering (FMS) calculations of XAS using FEFF8.2115 This method focuses on the difference spectrum which can be extracted precisely when the laser-on and laser-off spectra are collected strictly under the same detection and X-ray beam conditions.

The data fitting includes the following steps: (1) calculating the ground state spectrum and optimizing non-structural parameters (e. g. transition edge pos­ition) if it is necessary; (2) calculating a series of transient state spectra by varying different structural parameters and constructing the interpolation polynomial; and (3) minimizing the discrepancy between the two sets of dif­ference spectra, one corresponding to the experimental difference spectrum and the other corresponding to the theoretical difference spectrum in order to re­trieve the best fitting parameters. The best fit can be found using Equation (12.4):

(12.4)

where mape^on and mlaSSp~off are experimental absorption coefficients measured with and without the photoexcitation, respectively; msheor and mg‘leor are the theoretically calculated absorption coefficients for the transient and ground states, respectively, f is the fraction of transient species (i. e., the charge separ­ated state, (TiO2)n /RuN3+) and E1 and E2 define the energy range of spectral comparison.

By using MIA, XTA spectra were constructed from different structural parameters for RuN3 dye, which shows a high sensitivity in fitting over structural variations. A key in searching for the best structural parameters is to match the shape of the spectrum in the XANES region that account for the multiple scattering effect in the molecule. Extracting a correct fraction of the transient structure created by the pump laser pulse is often a key to determining

correct structural parameters of the transient state. In this particular case, iterative fits with varying fraction and structural parameter were used. The results show that the conversion from the ground to the charge separate states causes different responses in the Ru-N bonds with different ligands, resulting in the average Ru-N(NCS) bond length being shortened by ~0.06 A, from 2.05 A to 1.99 A, although the average Ru-N (dcbpy) bond length shows almost no changes. Other combinations of structural variations were used in analyses, but the results were significantly worse in the difference spectra, as shown in Figure 12.9(C).

These results can be rationalized by a recent time-dependent density functional theory (TDDFT) calculation on a RuN3-TiO2(lattice) system which suggested two steps in photoinduced electron injection (Figure 12.8(B)), (1) shifting electron density from the highest occupied molecular orbital (HOMO), assigned as Ru t2g character with a sizable contribution from the NCS ligand orbitals, to the dcbpy ligands and (2) another electron density shift from dcbpy bridging ligands to the TiO2 lattice.109,110 Consequently, a net electron density loss at the Ru-NCS moiety and a net electron density gain at the surface of TiO2 lattice take place, while the dcbpy ligands anchored on the TiO2 act as an electron density relay with a minimal net change in the electron density. The example demonstrates the XTA capabilities for studying inter­facial charge and energy transfer which are key processes in many DSSCs, hybrid solar conversion systems and catalytical systems. We learned from the results that anchoring the dye molecule in the charge injection direction aligned with the electron density gradient orientation was important to the electron injection efficiency and rate. Because high efficiency DSSCs require both fast electron injection and minimal germination charge recombination, it is not entirely clear whether little or no structural reorganization could accelerate the charge injection because certain structural reorganization could also prevent the germination recombination of the charges and enhance the device efficiency.

This work has extended the XTA studies from those of homogeneous systems to heterogeneous interfacial charge separation systems, which is a very active research area relevant to future developments of photovoltaic devices and heterogeneous catalysis. In order to study these surface specific interfacial light and matter interactions, solid film samples need to be studied by XTA, raising the challenge of minimizing the radiation damage of the stationary sample and circulating the film samples. Grazing angle incident XAS conducted in XTA will be pursued by additional low temperature studies.

On the scientific front of the DSSC research, alternative dyes with cheaper first row transition metals and other organic molecules have been investi – gated.116 122 One type of alternative are copper(I) diimine complexes that have been extensively studied for their interesting photochemistry and structural dynamics.123 130 These complexes have amazingly similar UV/vis absorption spectra to those of Ru(II)-trisbipyridyl complex and its derivatives and also undergo the MLCT transition where Cu(I) diimine becomes a Cu(II) diimine complex with one of the ligands becoming an anion. One such example is [Cu(I)(dmp)2]1 + hv — [Cu(II)(dmp)(dmp) ]1 (dmp = 2,9-dimethyl-1,10-phenan – throline) which shifts the electron density from the Cu(I) (3d10) center to the ligands, resulting in a transient Cu(II) (3d9) center.72,131 148 What distinguishes Cu(I) diimine complexes from the Ru(II) complexes is their susceptibility to Jahn-Teller distortion as a result of the MLCT transition, which transforms the pseudo-tetrahedral coordination geometry of the Cu(I) center in the ground state to a flattened coordination geometry in the MLCT excited state. Meanwhile, the formation of an “exciplex” between the Cu(II) center in [Cu(II)(dmp)(dmp) ]1 and a solvent molecule lowers the energy of the MLCT state and shortens its lifetime from hundreds of ns to <1 ns (Figure 12.10).

In order to prolong the MLCT state lifetime and preserve the initial energy of the MLCT state, we have started structural dynamics studies on a series of Cu(I) diimine complexes with variable structural hindrances for the Jahn-Teller distortion and solvent accessibility.149 One important part of this study is the direct structural determination of the MLCT state by XTA to identify the oxidation state of the copper center as well as its coordination geometry. Our studies found that the intersystem crossing (ISC) time constant in these com­plexes is strongly coordination symmetry dependent. The ISC time constant is <1ps if the MLCT state has two ligand planes orthogonal to each other, but is >10 ps if the relative orientation of the two ligand planes is far from orth­ogonal.148,150 In the context of DSSC-related issues, the study intends to de­termine which structure, the one with orthogonal ligand planes or the one with flattened tetrahedral coordination geometry, favors electron injection into the semiconductor nanoparticle electrode and if we can use structural constraints to prevent Jahn-Teller distortion or solvent accessibility to the excited state and hence to modulate the excited state properties. Recently, we have found that one of the water soluble Cu(I) diimine complexes with sulfonic groups151 performs photoinduced electron injection into TiO2 nanoparticles. The struc­tural dynamics of similar complexes hybrid with TiO2 nanoparticles are under investigation from energetic, dynamic and structural control aspects.