Ultrafast charge separation dynamics in opaque, operational dye-sensitized solar cells revealed by femtosecond diffuse reflectance spectroscopy

Efficient dye-sensitized solar cells are based on highly diffusive mesoscopic layers that render these devices opaque and unsuitable for ultrafast transient absorption spectroscopy measurements in transmission mode. We developed a novel sub-200 femtosecond time-resolved diffuse reflectance spectroscopy scheme combined with potentiostatic control to study various solar cells in fully operational condition. We studied performance optimized devices based on liquid redox electrolytes and opaque TiO2 films, as well as other morphologies, such as TiO2 fibers and nanotubes. Charge injection from the Z907 dye in all TiO2 morphologies was observed to take place in the sub-200 fs time scale. The kinetics of electron-hole back recombination has features in the picosecond to nanosecond time scale. This observation is significantly different from what was reported in the literature where the electron-hole back recombination for transparent films of small particles is generally accepted to occur on a longer time scale of microseconds. The kinetics of the ultrafast electron injection remained unchanged for voltages between +500 mV and –690 mV, where the injection yield eventually drops steeply. The primary charge separation in Y123 organic dye based devices was clearly slower occurring in two picoseconds and no kinetic component on the shorter femtosecond time scale was recorded.


Time-resolved diffuse reflectance spectroscopy
State of the art pump-probe diffuse reflectance spectrometer Figure S1 depicts the schematic of developed femtosecond time-resolved diffuse reflectance spectrometer. The new design of light collection configuration based on coupled off-axis parabolic mirrors allowed us to achieve a time resolution of sub-200 fs and collection of light with a big solid angle. The technique enables, the study of opaque solid films as well as highly absorbing systems. Quantitative data analysis and optical modeling Figure S2 shows the pump-probe diffuse reflectance signal at a delay time of 50 ps as a function of excitation intensity. The amplitude of the signal shows a good linear response to the excitation intensity tuned over a wide range. Figure S2. Linearity test. Transient diffuse reflectance measurements over a broad excitation pulse energy range are shown. The sample is a dye-sensitized double layer TiO 2 film. The pump beam energy is changed from 0.047 µJ to 0.95 µJ. The y-axis is the amplitude of the pump-probe signal at 50 ps after pulse excitation.
The amplitude of the signal versus the excitation intensity shows a linear behavior.
The Kubelka-Munk formalism is applied to the time-resolved diffuse reflectance measurements according to equation (3) of the main text. The Kubelka-Munk function, which is representative of the concentration of absorbing species in the film, shows a perfect linear response over the excitation energy of pump beam.

Influence of excitation intensity on the kinetics in Z907 sensitized complete photoanode
Intensity dependence of diffuse reflectance measurements on Z907 sensitized complete DSC photoanode (double layer) in the presence of MPN solvent is presented in Figure S4. The signals can be fitted with single exponential function. The observed kinetics at 840 nm is assigned to the early back recombination of photoinjected electrons with oxidized dye molecules. It should be noted that in kinetics studies all the measurements are performed at very low excitation intensities (below 300 nJ/ pulse). Excitation pulse energy is changed by one order of magnitude from 300 nJ to 1000 nJ. Consequently, the amplitude of the signal is raised, and the decay kinetics is accelerated. Rate constants of single exponential decay function fit to traces measured at 400 nJ, 780 nJ, and 1000 nJ are respectively, 0.0014 × 10 12 s -1 , 0.0023 × 10 12 s -1 and 0.0028 × 10 12 s -1 . This depleting kinetics represents an early back reaction of electrons with oxidized molecules.

Effect of probe wavelength
The evolution of oxidized dye molecule is also probed in the visible wavelength region at 670 nm and is compared with the measurements at 840 nm. The early back recombination kinetics is again observed when the diffuse reflectance measurements kinetics is recorded at 670 nm. The signal is fitted with single exponential function. The first order rate constant for measurement at 670 nm is 0.0008

Effect of TiO 2 film morphology and presence of electrolyte
To understand the effect of morphology on the kinetics of charge separation process, we also investigated the electron injection profiles in DSC devices based on different opaque nanostructured TiO 2 films like TiO 2 nanofibers, and TiO 2 nanotubes prepared by anodization of Ti foil. Similar measurement on Z907 dye-sensitized anodized TiO 2 nanotubes film in the presence of MPN solvent and redox electrolyte is also presented. It is interesting to observe that the electron injection is still in the ultrafast regime, and the recombination features are still present. It should be noted that the picosecond recombination feature is occurring in nanotubes with different relative amplitude compared to that of scattering particles.
Transient absorption studies on the band gap excitation of bare TiO 2 nanocrystalline films is performed by Furube and co-workers 1 . They have observed that after ultrafast formation of electron hole in TiO 2 particles, the surface trap electrons and surface trap holes forms in 200 fs (in the limit of time resolution) and relax to deep bulk traps in 500 ps. However, our measurements situation is slightly different with those studies in the sense that our samples are dye-sensitized TiO 2 films. In our system, the electrons are in the TiO 2 particles and holes are in the oxidized dye molecules.
Although the photoinjected electrons can trap, in the same way, as what was observed for bare TiO 2 particles by groups of Colombo and Furube 1,2 .
In anodized TiO 2 nanotube film, the trap states are energetically deeper than in the TiO 2 nanoparticle film as it was measured previously by modulated voltage-current techniques and terahertz spectroscopy 3,4 . So, the trap state distribution difference could play a vital role in decreasing the observed decreased relative amplitude of the picosecond recombination feature (i.e. the loss of electrons due to back recombination) in the study with nanotubes compared to nanoparticle film (black and green traces in 4a and 4b). Therefore, in nanotubes films, electrons that are trapped in the films are less free to undergo the early picosecond recombination in comparison to what is seen in particles.

Modeling of the laser induced photovoltage in the solar cell and dark current correction
In this model, a chemical capacitance is assigned to the DSC. The capacitance of the cell is measured by impedance spectroscopy, at each bias voltage. The chemical capacitance in the TiO 2 film increases (as the applied forward bias is increased). The amount of photoinjected charge into this capacitor is estimated based on quantification of the transient diffuse reflectance spectrum.
Kubelka-Munk formalism is integrated on the diffuse reflectance spectrum, as it is proportional to the concentration of absorbing species. Having: The photovoltage induced by each laser pulse is estimated to be about 10 µV at 700 mV bias voltage and 38 µV when the cell is biased at 520 mV. However for laser spectroscopy measurements on the cell at short circuit condition or small bias voltage, the amount of laser induce shift in the quasi-Fermi level position is in the order of some mV.
A good approximation of dark current correction is to consider the cell series resistance and according to Ohm law draw the voltage drop in the cell due to dark current.

Comparison of the photovoltaic and optical response of two different DSC devices made based on transparent layer, and scattering layer film of same thickness
Two types of DSC devices based on two different TiO 2 films with same thickness sensitized with Z907 dye were made. The first device is based on a 5 µm-thick TiO 2 film made of scattering particles and the second device is made of a 5 µm-thick transparent TiO 2 layer. All preparation and test situation were identical for both devices. The morphological parameters, optical properties and photovoltaic performances of the devices are depicted in Table S1.
As depicted in Table S1, in the scattering TiO 2 layer, the BET surface area is 27 m 2 /gr and the roughness factor of the film is 36.4 /cm 2 µm. These values are much smaller than those in transparent layer film being 85 m 2 /gr and 98 /cm 2 µm respectively. Therefore a smaller amount of dye is adsorbed on the scattering layer film. Therefore, the value of the maximum light absorptance ( ) in the scattering layer is 0.66 and is smaller than that in the transparent film being 0.88. The photocurrent generated in a DSC device made of transparent layer is 11.4 mA/cm 2 while this value for scattering TiO 2 layer based device is only 5.9 mA/cm 2 . The difference in the photocurrent is due to the amount of adsorbed dye molecule in the two films. For a fair comparison, the photocurrent of devices is normalized to the absorptance of the film. The obtained value for the transparent layer is 12.95 while for scattering layer is only 8.93. Therefore, the amount of J/ (current normalized to the absorptance) for the scattering film is 30% less than that of the transparent layer. This observation indicates that in the scattering layer based device about 30% of the photo-generated electrons are lost. This can be in agreement with our laser spectroscopy results that for big particles we witnessed loss of electrons due to prompt back recombination.   Figure S8 illustrates the steady-state absorbance spectrum of Y123 dye measured in solution. The first absorption peak of the dye is at 530 spectral regions. Figure S8. Optical analysis. Steady-state absorbance spectrum of the Y123 dye in solution. Figure S9 depicts the measurements on Y123 dye-sensitized different TiO 2 films in the presence of only MPN solvent probed at 840 nm. Comparing with the measurements in the presence of the electrolyte, which is shown in figure 6 of the main text, no obvious difference is observed, and the kinetics are independent of the environment.   Figure S10 shows the measurements on double layer film immersed for two days in acetonitrile, which is the same sample for which the kinetics is shown in the red trace in figure 6 of the main text.

Effect of liquid environment
We observe that after immersing in acetonitrile, the slow rise kinetics is removed, and a flat kinetics is replaced. This is due to dissolving of the extra dye molecules attached to the surface of the film in the form of dye aggregates. nm can be a tail of the same absorption feature, which is partially, covered by laser pulse excitation at 600 nm. The positive peak can be assigned to the excited state absorption of the Y123 dye. This deactivation of the Y123 dye excited state can be fitted to a single exponential function with almost the same lifetime of about 50 ps and, therefore, has a mirror-like kinetics to ground state bleaching.
A small negative peak at the NIR region about 750 nm is assigned to the emission of the dye.  sensitized TiO 2 transparent film, in NIR wavelength region. The excitation wavelength is at wavelength 600 nm. Excitation intensities are 300 nJ/ pulse for films and 1000 nJ/ pulse for measurement in solution. Figure S11c and S11d depict the NIR transient absorbance spectrum of the Y123 dye in solution and anchored on TiO 2 film. The excitation wavelength at these measurements is 600 nm. In the transient spectrum of the dye in solution, the negative peak at NIR region up to 950 nm can be due to the emission of the dye. The spectrum of the Y123 sensitized TiO 2 film shows a contribution of both absorption of oxidized dye molecules and dye emission.