Ultrafast Electron and Hole Dynamics in CdSe Quantum Dot Sensitized Solar Cells

A potential candidate for next-generation solar cells is dye-sensitized solar cells (DSSCs). Much attention has been directed toward DSSCs employing nanostructured TiO2 electrodes and organic-ruthenium dye molecules as the light-harvesting media. The high porosity of nanostructured TiO2 film enables a large concentration of the sensitizing dye molecules to be adsorbed. The attached dye molecules absorb light and inject electrons into the TiO2 conduction band upon excitation. The electrons are then collected at a back conducting electrode, generating a photocurrent. DSSCs exhibit high photovoltaic conversion efficiencies of about 11% and good long-term stability. In addition, they are relatively simple to assemble and are low-cost (O’Regan & Grätzel, 1991; Grätzel, 2003; Chiba et al., 2006). However, in order to replace conventional Si-based solar cells in practical applications, further effort is needed to improve the efficiency of DSSCs. A great amount of work has been done on controlling the morphology of the TiO2 electrodes by employing ordered arrays of nanotubes, nanowires, nanorods and inverse opal structures (Adachi et al., 2003; Paulose et al., 2006; Law et al., 2005; Song et al., 2005; Nishimura et al., 2003) in order to improve the electron transport and collection throughout the device. Another important factor in improving the performance of DSSCs is the design of the photosensitizer. The ideal dye photosensitizer for DSSCs should be highly absorbing across the entire solar light spectrum, bind strongly to the TiO2 surface and inject photoexcited electrons into the TiO2 conduction band efficiently. Many different dye compounds have been designed and synthesized to fulfill the above requirements. It is likely that the ideal photosensitizer for DSSCs will only be realized by co-adsorption of a few different dyes, for absorption of visible light, near infrared (NIR) light, and/or infrared (IR) light (Polo et al., 2004; Park et al., 2011). However, attempts to sensitize electrodes with multiple dyes have achieved only limited success to date. Narrow-band-gap semiconductor quantum dots (QDs), such as CdS, CdSe, PbS, and InAs, have also been the subject of considerable interest as promising candidates for replacing the sensitizer dyes in DSSCs (Vogel et al., 1990, 1994; Toyoda et al., 1999, 2003; Peter et al., 2002; Plass et al., 2002; Shen et al., 2004a, 2004b, 2006a, 2006b, 2008a, 2008b, 2010a, 2010b; Yu et al., 2006; Robel et al., 2006; Niitsoo et al., 2006; Diguna, et al., 2007a , 2007b; Kamat, 2008, 2010; Gimenez et. al., 2009; Mora-Sero et al., 2009, 2010). These devices are called QD-sensitized solar cells (QDSCs) (Nozik, 2002, 2008; Kamat, 2008). The use of semiconductor QDs as sensitizers


Introduction
A potential candidate for next-generation solar cells is dye-sensitized solar cells (DSSCs).Much attention has been directed toward DSSCs employing nanostructured TiO 2 electrodes and organic-ruthenium dye molecules as the light-harvesting media.The high porosity of nanostructured TiO 2 film enables a large concentration of the sensitizing dye molecules to be adsorbed.The attached dye molecules absorb light and inject electrons into the TiO 2 conduction band upon excitation.The electrons are then collected at a back conducting electrode, generating a photocurrent.DSSCs exhibit high photovoltaic conversion efficiencies of about 11% and good long-term stability.In addition, they are relatively simple to assemble and are low-cost (O'Regan & Grätzel, 1991;Grätzel, 2003;Chiba et al., 2006).However, in order to replace conventional Si-based solar cells in practical applications, further effort is needed to improve the efficiency of DSSCs.A great amount of work has been done on controlling the morphology of the TiO 2 electrodes by employing ordered arrays of nanotubes, nanowires, nanorods and inverse opal structures (Adachi et al., 2003;Paulose et al., 2006;Law et al., 2005;Song et al., 2005;Nishimura et al., 2003) in order to improve the electron transport and collection throughout the device.Another important factor in improving the performance of DSSCs is the design of the photosensitizer.The ideal dye photosensitizer for DSSCs should be highly absorbing across the entire solar light spectrum, bind strongly to the TiO 2 surface and inject photoexcited electrons into the TiO 2 conduction band efficiently.Many different dye compounds have been designed and synthesized to fulfill the above requirements.It is likely that the ideal photosensitizer for DSSCs will only be realized by co-adsorption of a few different dyes, for absorption of visible light, near infrared (NIR) light, and/or infrared (IR) light (Polo et al., 2004;Park et al., 2011).However, attempts to sensitize electrodes with multiple dyes have achieved only limited success to date.Narrow-band-gap semiconductor quantum dots (QDs), such as CdS, CdSe, PbS, and InAs, have also been the subject of considerable interest as promising candidates for replacing the sensitizer dyes in DSSCs (Vogel et al., 1990(Vogel et al., , 1994;;Toyoda et al., 1999Toyoda et al., , 2003;;Peter et al., 2002;Plass et al., 2002;Shen et al., 2004aShen et al., , 2004bShen et al., , 2006aShen et al., , 2006bShen et al., , 2008aShen et al., , 2008bShen et al., , 2010aShen et al., , 2010b;;Yu et al., 2006;Robel et al., 2006;Niitsoo et al., 2006;Diguna, et al., 2007aDiguna, et al., , 2007b;;Kamat, 2008Kamat, , 2010;;Gimenez et. al., 2009;Mora-Sero et al., 2009, 2010).These devices are called QD-sensitized solar cells (QDSCs) (Nozik, 2002(Nozik, , 2008;;Kamat, 2008).The use of semiconductor QDs as sensitizers has some unique advantages over the use of dye molecules in solar cell applications (Nozik, 2002(Nozik, , 2008)).First, the energy gaps of the QDs can be tuned by controlling their size, and therefore the absorption spectra of the QDs can be tuned to match the spectral distribution of sunlight.Secondly, semiconductor QDs have large extinction coefficients due to the quantum confinement effect.Thirdly, these QDs have large intrinsic dipole moments, which may lead to rapid charge separation.Finally, semiconductor QDs have potential to generate multiple electron-hole pairs with one single photon absorption (Nozik, 2002;Schaller, 2004), which can improve the maximum theoretical thermodynamic efficiency for photovoltaic devices with a single sensitizer up to 44% (Hanna et al., 2006).However, at present, the conversion efficiency of QDSCs is still less than 5% (Mora-Sero´, et al., 2010;Zhang, et al., 2011).So, fundamental studies on the mechanism and preparation of QDSCs are still necessary and very important.In a semiconductor quantum dot-sensitized solar cell (QDSC), as the first step of photosensitization, a photoexcited electron in the QD should rapidly transfer to the conduction band of TiO 2 electrode and a photoexcited hole should transfer to the electrolyte (Scheme 1).Thus charge separation of the photoexcited electrons and holes in the semiconductor QDs and the electron injection process are key factors for the improvement of the photocurrents in the QDSCs.In this sense, the study on the photoexcited carrier dynamics in the QDs is very important for improving the conversion efficiency of the solar cell.To date, the information on the carrier dynamics of semiconductor QDs adsorbed on TiO 2 electrode is limited, although a few studies have been carried out for CdS, CdSe and InP QDs using a transient absorption (TA) technique (Robel et al., 2006(Robel et al., , 2007;;Tvrdy et al., 2011;Blackburn et al., 2003Blackburn et al., , 2005)).Most of them focused on the electron transfer process and the measurements mostly were carried out in either dispersed colloidal systems or dry electrodes.In recent years, the authors' group has been applying an improved transient grating technique (Katayama et al., 2003) to study the photoexcited carrier dynamics of semiconductor nanomaterials, such as TiO 2 nanoparticles with different crystal structures and CdSe QDs absorbed onto TiO 2 and SnO 2 nanostructured electrodes (Shen et al., 2005(Shen et al., , 2006(Shen et al., , 2007(Shen et al., , 2008(Shen et al., , 2010)).The improved TG technique is a simple and highly sensitive time-resolved optical technique and has proved to be powerful for measuring various kinds of dynamics, such as population dynamics and excited carrier diffusion.Comparing to the TA technique, the improved TG technique has a higher sensitivity and measurements under lower light intensity are possible (Katayama et al., 2003;Shen et al., 2007Shen et al., , 2010a)).This fact is very important for studying the carrier dynamics of QDs used in QDSCs under the conditions of lower light intensity similar to sun light illumination.The improved TG technique is also applicable to samples with rough surfaces, like the samples used in this study.This chapter will focus on the ultrafast photoexcited electron and hole dynamics in CdSe QD adsorbed TiO 2 electrodes employed in QDSCs characterized by using the improved TG technique.CdSe QDs were adsorbed on TiO 2 nanostructured electrodes with different adsorption methods.The following issues will be discussed: 1. Pump light intensity dependence of the ultrafast electron and hole dynamics in the CdSe QDs adsorbed onto TiO 2 nanostructured electrodes; 2. Separation of the ultrafast electron and hole dynamics in the CdSe QDs adsorbed onto TiO 2 nanostructured electrodes; 3. Electron injection from CdSe QDs to TiO 2 nanostructured electrode; 4. Changes of carrier dynamics in CdSe QDs adsorbed onto TiO 2 electrodes versus adsorption conditions; 5. Effect of surface modification on the ultrafast carrier dynamics and photovoltaic properties of CdSe QD sensitized TiO 2 electrodes.
Scheme 1. Electron-hole pairs are generated in semiconductor QDs after light absorption.
Then photoexctied electrons in the semiconductor QDs are injected to the conduction band of TiO 2 and/or trapped by surface or interface states.The photoexcited holes are scavenged by reducing species in the electrolyte and/or trapped by surface or interface states.The nanostructured TiO 2 is employed as an electron conductor and electrons transport in TiO 2 to a transparent conductive oxide (TCO) substrate, while the electrolyte is used as a hole transporter and hoels are transported to a counter electrode.

Preparation methods of CdSe QD adsorbed TiO 2 nanostructured electrodes
The method for preparing the TiO 2 electrodes has been reported in a previous paper (Shen et al., 2003).A TiO 2 paste was prepared by mixing 15 nm TiO 2 nanocrystalline particles (Super Titania, Showa Denko; anatase structure) and polyethylene glycol (PEG) (molecular weight (MW): 500,000) in pure water.The resultant paste was then deposited onto transparent conducting substrates (F-doped SnO 2 (FTO), sheet resistance = 10 Ω/sq).The TiO 2 films were then sintered in air at 450 ºC for 30 min to obtain good necking.The highly porous nanostructure of the TiO 2 films (the pore sizes are of the order of a few tens of nanometers) was confirmed through scanning electron microscopy (SEM) images.CdSe QDs can be adsorbed onto the TiO 2 nanostructured electrodes by using the following methods: 1. Chemical bath deposition (CBD) method (Hodes et al., 1994;Shen et al., 2008) Firstly, for the Se source, an 80 mM sodium selenosulphate (Na 2 SeSO 3 ) solution was prepared by dissolving elemental Se powder in a 200 mM Na 2 SO 3 solution.Secondly, 80 mM CdSO 4 and 120 mM of the trisodium salt of nitrilotriacetic acid (N(CH 2 COONa) 3 ) were mixed with the 80 mM Na 2 SeSO 3 solution in a volume ratio of 1:1:1.The TiO 2 films were placed in a glass container filled with the final solution at 10 ºC in the dark for various times to promote CdSe adsorption.2. Successive ionic layer adsorption-reaction (SILAR) method (Guijarro et al., 2010a) In situ growth of CdSe QDs using the SILAR method was carried out by successive immersion of TiO 2 electrodes in ionic precursor solutions of cadmium and selenium.
Aqueous solutions were employed in all cases.A 0.5 M (CH 3 COO) 2 Cd (98.0%,Sigma-Aldrich) solution was used as the cadmium source, while a sodium selenosulfate (Na 2 SeSO 3 ) solution was used as the selenium precursor.The sodium selenosulfate solution was prepared by heating under reflux for 1h a mixture of 1.6 g of Se powder, 40 mL of 1 M Na 2 SO 3 (98.0%,Alfa Aesar) and 10 mL of 1 M NaOH.The resulting solution was filtered and mixed with 40 mL of 1M CH 3 COONa (99.0+%,Fluka) solution, and finally was stored in the dark.The pH of the sodium selenosulfate solution was optimized to improve the QD deposition rate by using 0.25 M H 2 SO 4 and/or 0.1 M NaOH stock solutions.3. Direct adsorption (DA) of previously synthesized QDs (Guijarro et al., 2010b) Colloidal dispersions of CdSe QDs capped with trioctylphosphine (TOP) were prepared by a solvothermal route which permits size control.DA of CdSe QDs was achieved by immersion of TiO 2 electrodes in a CH 2 Cl 2 (99.6%,Sigma Aldrich) CdSe QD dispersion, using soaking times ranging from 1 h to 1 week.4. Linker assisted adsorption (LA) of previously synthesized QDs (Guijarro et al., 2010b) LA was performed employing p-mercaptobenzoic acid (MBA; 90%, Aldrich), cysteine (97%, Aldrich), and mercaptopropionic acid (MPA; 99%, Aldrich) as molecular wires.First, the linker was anchored to the TiO 2 surface by immersion in saturated toluene solutions of cysteine (5 mM) or MBA (10 mM) for 24 h.Secondly, these electrodes were washed with pure toluene for ½ h to remove the excess of the linker.Finally, the modified electrodes were transferred to a toluene CdSe QD dispersion for 3 days, to ensure QD saturation.The procedure for modification of TiO 2 with MPA has previously been reported (Guijarro et al., 2009).After the CdSe QD adsorption, the samples were coated with ZnS by alternately dipping them three times in 0.1 M Zn(CH 2 COO) and 0.1 M Na 2 S aqueous solutions for 1 minute for each dip (Yang et al., 2003;Diguna, et al., 2007a;Shen et al., 2008).

An improved transient grating (TG) technique
The transient grating (TG) method is a well-established laser spectroscopic technique of four wave mixing (Eichler et al., 1986;Harata et al., 1999).In the TG method, two time-coincident short laser pulses (pump beams) with the same wavelength and intensity intersect with an angle in a sample to generate an optical interference pattern at the intersection.Interaction between the light field and the material results in a spatially periodic modulation of the complex refractive index, which works like a transient diffraction grating for a third laser pulse (probe beam) incident to the photoexcited region.Then, by measuring the time dependence of the diffraction light of the probe beam, dynamics of the transient grating produced in the sample can be monitored.The TG technique is a powerful tool for detecting population dynamics (Rajesh et al., 2002), thermal diffusion (Glorieux et al., 2002), diffusion of photoexcited species (Terazima et al., 2000), energy transfer from photoexcited species to liquids (Miyata et al, 2002), structural or orientational relaxation (Glorieux et al., 2002), the sound velocity of liquids (Ohmor et al., 2001), and so on.Although this technique provides valuable information, it presents some technical difficulties for general researchers (Harata et al., 1999).First, the three beams must overlap on a small spot, typically within a spot diameter of less than 100 μm on a sample, and each beam must be temporally controlled.This is very difficult for pulsed laser beams, especially for those with a pulse width of ~100 femtoseconds.Secondly, since the diffraction of the probe beam, namely the signal, is quite weak, it is difficult to find the diffraction beam during the measurements.Thirdly, for a solid sample, the surface must be optically smooth.It is almost impossible to measure a sample with a rough surface by using the conventional TG technique.
In 2003, Katayama and co-workers (Katayama et al., 2003;Yamaguchi et al., 2003) proposed an improved TG technique (it was also called a lens-free heterodyne TG (LF-HD-TG) or a near field heterodyne TG (NF-HD-TG) technique in some papers), which overcomes the difficulties that exist in the conventional TG technique.The improved TG technique features (1) simple and compact optical equipment and easy optical alignment and (2) high stability of phase due to the short optical path length of the probe and reference beams.This method is thought to be versatile with applicability to many kinds of sample states, namely opaque solids, scattering solids with rough surfaces, transparent solids, and liquids, because it is applicable to transmission and reflection-type measurements.The principle of the improved TG technique has been explained in detail in the previous papers (Katayama et al., 2003;Yamaguchi et al., 2003) and is only described briefly here.Unlike the conventional TG technique, only one pump beam and one probe beam without focusing are needed in the improved TG technique.The pump beam is incident on the transmission grating.Then, the spatial intensity profile of the pump beam is known to have an interference pattern in the vicinity of the other side of the transmission grating, and the interference pattern has a grating spacing that is similar to that of the transmission grating.When a sample is brought near the transmission-grating surface, it can be excited by the optical interference pattern.The refractive index of the sample changes according to the intensity profile of the pump light and the induced refractive index profile functions as a different type of transiently generated grating.When the probe beam is incident in a manner similar to that of the pump beam, it is diffracted both by the transmission grating (called a reference light) and the transiently generated grating (called a signal light).In principle, the two diffractions progress along the same direction; therefore, these two diffractions interfere, which is detected by a detector positioned at a visible diffraction spot of the reference beam.In the improved TG technique used for studying the ultrafast carrier dynaimcs of semiconductor QDs, the laser source was a titanium/sapphire laser (CPA-2010, Clark-MXR Inc.) with a wavelength of 775 nm, a repetition rate of 1 kHz, and a pulse width of 150 fs.The light was separated into two parts.One of them was used as a probe pulse.The other was used to pump an optical parametric amplifier (OPA) (a TOAPS from Quantronix) to generate light pulses with a wavelength tunable from 290 nm to 3 µm used as pump light in the TG measurement.The probe pulse wavelength was 775 nm.

Power dependence of carrier dynamics in CdSe QDs adsorbed onto TiO 2 electrodes
As described in depth in previous papers (Katayama et al., 2003;Shen et al., 2005Shen et al., , 2007)), the TG signal is directly proportional to the change in the refractive index occurring in the sample (n(t)) upon photoexcitation.In the timescale of these experiments less than 1 ns, assuming Drude's model, the refractive index change, i.e., the TG signal intensity for CdSe QD adsorbed samples, will be a linear function of the concentration of photogenerated carriers (electrons and holes in CdSe QDs, i.e., N e,CdSe and N h,CdSe , and injected electrons in TiO 2 , i. e., N e , TiO2 ) , as follows (Guijarro et al., 2010a(Guijarro et al., , 2020b)): (1) where  p is the radial probe frequency, e is the elementary charge,  0 is free-space permittivity, and n 0,CdSe (2.7) and n 0,TiO2 (2.5) are the refractive indices of CdSe and TiO 2 , respectively.The important feature of the TG signal is that both the photoexcited electron and hole carrier densities contribute to the signal.In principle, the exact contribution on Δn(t) by each carrier depends inversely on its carrier effective mass.According to the Drude theory (Kashiski et al., 1989;Kim et al., 2009), it can be considered that only free photoexcited electrons and holes are responsible for the population grating signals.For CdSe, the effective masses of electrons and holes are 0.13m 0 and 0.44m 0 (m 0 is the electron rest mass), respectively (Bawendi et al., 1989), so both the photoexcited electron and hole carrier densities in the CdSe QDs contribute to the signal.It is known that the effective mass of electrons for TiO 2 is about 30 m 0 , which is about two orders larger than that for CdSe.Therefore, the TG signal due to the injected electrons in TiO 2 (no holes injected into TiO 2 ) can be ignored (Shen et al., 2005(Shen et al., , 2006a(Shen et al., , 2007(Shen et al., , 2008a(Shen et al., , 2010a)).For a semiconductor material, usually, there are three kinds of photoexcited carrier relaxation dynamics.The first one is a one-body recombination, which is trapping and/or transfer of photoexcited electrons and holes.In this case, the lifetimes of the photoexcited electrons and holes are independent of the pump light intensity.The second one is a twobody recombination, which is a radiative recombination via electron and hole pairs.The third one is a three-body recombination, which is an Auger recombination via two electrons and one hole, or via one electron and two holes.In the latter two cases, the lifetimes of photoexcited electrons and holes are dependent on the pump light intensity.In order to determine what kinds of photoexcited carrier dynamics are reflected in the TG kinetics, we first confirmed many-body recombination processes such as the Auger recombination process existed or not under our experimental conditions.For this purpose, we measured the dependence of the TG kinetics on the pump intensity (from 2.5 to 16.5 μJ/pulse) for CdSe QD adosrobed TiO 2 electrodes (Figure 1) (Shen et al., 2010a).As shown in Fig. 1, we found that the dependence of the maximum signal intensity on the pump intensity was linear, and that the waveforms of the responses overlapped each other very well when they were normalized at the peak intensity.These results mean that the decay processes measured in the TG kinetics were independent of the pump intensity under our experimental conditions, and many-body recombination processes could be neglected.Therefore, it is reasonable to assume that the decay processes of photoexcited electrons and holes in the CdSe QDs are due to one-body recombination processes such as trapping and transfer under our experimental conditions.

Separation of the ultrafast electron and hole dynamics in the CdSe QDs adsorbed onto TiO 2 nanostructured electrodes
Figure 2 shows a typical kinetic trace of the TG signal of CdSe QDs adsorbed onto a TiO 2 nanostructured film (prepared by CBD method for 24 h adsorption) measured in air.The vertical axis was plotted on a logarithmic scale.Three decay processes (indicated as A, B, and C in Fig. 2) can be clearly observed.We found that the TG kinetics shown in Fig. 2 could be fitted very well with a double exponential decay plus an offset, as shown in eq. ( 2): 12 // 120 tt y Ae Ae y where A 1 , A 2 and y 0 are constants, and τ 1 and τ 2 are the time constants of the two decay processes (A and B in Fig. 2).Here, the constant term y 0 corresponds to the slowest decay process (C in Fig. 2), in which the decay time (in the order of ns) is much larger than the time scale of 100 ps measured in this study.The time constants of the fast (τ 1 ) and slow (τ 2 ) decay processes of photoexcited carriers in air are 6.3 ps and 82 ps, respectively (Table 1).As mentioned above, τ 1 and τ 2 are independent of the pump intensity under our experimental conditions, so the three decay processes are mostly due to one-body recombination such as carrier trapping and carrier transfer.The vertical axis is plotted on a logarithmic scale.Three decay processes A, B and C can be clearly observed (Shen et al., 2010a).
In order to separate the photoexcited electron dynamics and hole dynamics that make up the TG kinetics, the TG kinetics of the same sample was measured both in air and in a Na 2 S aqueous solution (hole acceptor) (1 M) (Shen et al., 2010a).As shown in Fig. 3, a large difference can be clearly observed between the TG responses measured in air and in the Na 2 S solution.By normalizing the two TG responses at the signal intensity of 90 ps, we found that they overlapped with each other very well for time periods of longer than 15 ps, but the fast decay process apparently disappeared when the time period was less than 10 ps in the TG kinetics measured in the Na 2 S solution (hole acceptor).This great difference can be explained as follows.In air, both hole and electron dynamics in the CdSe QDs could be measured in the TG kinetics.In the Na 2 S solution, however, photoexcited holes in the CdSe QDs will transfer quickly to the electrolyte and only electron dynamics should be measured in the TG kinetics.Therefore, the "apparent disappearance" of the fast decay process in the Na 2 S solution implies that the hole transfer to S 2-ions, which are supposed to be strongly adsorbed onto the CdSe QD surface, can be too fast in these circumstances as indicated by Hodes (Hodes, 2008) and therefore could not be observed under the temporal resolution (about 300 fs) of our TG technique.This observation is particularly important, because the result directly demonstrated that the transfer of holes to sulfur hole acceptors that are strongly adsorbed on the QD surface could approach a few hundreds of fs.An earlier study on the dynamics of photogenerated electron-hole pair separation in surface-space-charge fields at GaAs(100) crystal/oxide interfaces using a reflective electro-optic sampling method Fig. 3. TG responses of the CdSe QDs adsorbed onto nanostructured TiO 2 films measured in air (-) and in Na 2 S solution (□) as well as their "difference response" (○) (Shen et al., 2010a).
showed that the hole carrier transit time was faster than 500 fs (Min et al., 1990).We believe an ultrafast hole transfer time from the QDs to hole acceptors that are strongly adsorbed on the QD surface is a more feasible and reasonable explanation, since photoexcited holes can more easily reach the surface of QDs with diameters of a few nm.The TG response measured in the Na 2 S solution, which is considered to only relate to electron dynamics as mentioned above, can be fitted well with eq. ( 2).As shown in Table 1, besides the slower decay process, a faster decay process with a decay time of 9 ps was also detected in the TG response measured in the Na 2 S solution.Such a faster decay process with a characteristic time of a few picoseconds in the TG response measured in the Na 2 S solution was considered to correspond to electron transfer from the QDs in direct contact with the TiO 2 (first layer of deposited QDs) (Guijarro et al., 2010a(Guijarro et al., , 2010b)).It is worth noting that the relative intensity A 1 (0.07) measured in the Na 2 S solution is much smaller compared to the A 1 (0.39) measured in air and it could be ignored here.The slower relaxation process in the TG response was not influenced by the presence of the Na 2 S solution, as shown in Fig. 3.The decay time τ 2 (85 ps) and the relative intensity A 2 (0.31) for the slower decay process in the TG response measured in the Na 2 S solution are almost the same as those measured in air (Table 1).The slower electron relaxation mostly corresponds to electron transfer from the CdSe QDs to TiO 2 and trapping at the QD surface states, in which the decay time depends to a great extent on the size of the QDs and the adsorption method that is used (Guijarro et al., 2010a(Guijarro et al., , 2010b;;Shen et al., 2006Shen et al., , 2007;;Diguna et al., 2007b).The slowest decay process (with a time scale of ns) may mostly result from the non-radiative recombination of photoexcited electrons with defects that exist at the CdSe QD surfaces and at the CdSe/CdSe interfaces.
The difference between the two TG responses measured in air and in Na 2 S solution (normalized for the longer time), which was termed the "difference response", is believed to correspond to the photoexcited hole dynamics in the CdSe QDs measured in air.As shown in Fig. 3, the difference response decays very fast and disappears around 10 ps and can be fitted very well with a one-exponential decay function with a decay time of 5 ps (Table 1).Thus, we did well in separating the dynamics of photoexcited electrons and holes in the CdSe QDs and found that the hole dynamics were much faster than those of electrons.Some papers have also reported that the hole relaxation time is much faster than the electron relaxation time in CdS and CdSe QDs (Underwood et al., 2001;Braun et al., 2002).In air, the fast hole decay process with a time scale of about 5 ps can be considered as the trapping of holes by the CdSe QD surface states.This result is in good agreement with the experimental results obtained by a femtosecond fluorescence "up-conversion" technique (Underwood et al., 2001).Thus, by comparing the TG responses measured in air and in a Na 2 S solution (hole acceptor), we succeeded in separating the dynamic characteristics of photoexcited electrons and holes in the CdSe QDs.We found that charge separation in the CdSe QDs occurred over a very fast time scale from a few hundreds of fs in the Na 2 S solution via hole transfer to S 2- ions to a few ps in air via hole trapping.films measured in air and in Na 2 S solution (hole acceptor) as well as their "difference response" as shown in Fig. 3 with a double exponential decay equation (eq.( 2)).τ 1 and τ 2 are time constants; A 1 , A 2 and y 0 are constants (Shen et al., 2010a).

Electron injection from CdSe QDs to TiO 2 nanostructured electrode
To investigate the electron transfer rate from CdSe QDs to TiO 2 electrodes, Shen and coworkers measured the TG kinetics of CdSe QDs adsorbed on both nanostructured TiO 2 electrodes and glass substrates under the same deposition conditions (Shen et al., 2008).
Figure 4 shows the TG kintics (pump beam wavelength: 388 nm) of CdSe QDs adsorbed on a TiO 2 nanostructured electrode and on a glass substrate (CBD method).The average diameter of the CdSe QDs was about 5.5 nm for both kinds of substrates.The TG kinetics were fitted to a double exponential decay function (Eq.( 3)) using a least-squares fitting method, convoluting with a 1ps Gaussian representing the laser pulse.where A 1 and A 2 are constants, and τ 1 and τ 2 are the time constants of the two decay processes, respectively.The fast decay time constants τ 1 were obtained to be 2.3 ps for both the two kinds of substrates.The slow decay time constants τ 2 were obtained to be approximately 140 ps and 570 ps for the CdSe QDs adsorbed on the TiO 2 electrode and on the glass substrate, respectively.It was found that τ 1 was the same for both kinds of substrates.However, the value of τ 2 for the CdSe QDs adsorbed on the nanostructured TiO 2 electrode was much smaller than that on the glass substrate.Under the experimental conditions (pump intensity < 2 μJ/pulse in this case), the waveforms of the responses overlapped each other very well when they were normalized at the peak intensity, as mentioned above.Therefore, it is reasonable to assume that the two decay processes in the CdSe QDs adsorbed on these two kinds of electrodes were due to one-body recombination, such as trapping, decay into intrinsic states and/or a transfer process.Because the fast decay time constant is independent of the substrates and it is known that hole trapping time is much faster than that of electron relaxation (Underwood et al., 2001;Braun et al., 2002), we believe that the fast decay process with a time scale of 2-3 ps mainly corresponded to a decrease in the photoexcited hole carrier density due to trapping at the CdSe QD surface/interface states or relaxation into the intrinsic QD states, as mentioned earlier.
Fig. 4. TG kinetics of the CdSe QDs (average diameter: 5.5 nm) adsorbed on a nanostructured TiO 2 electrode and on a glass substrate under same adsorption conditions (Shen et al., 2008).
This result is in good agreement with the experimental results obtained with a femtosecond fluorescence up-conversion technique and transient absorption measurements (Underwood et al., 2001).The slow decay process was considered to reflect the photoexcited electron relaxation processes.For the glass substrate, no electron transfer from the QDs to the glass substrate could occur.Thus, the slow decay process only corresponded to the relaxation of photoexcited electrons in the CdSe QDs, which was mostly due to trapping at the CdSe QD surfaces.For the nanostructured TiO 2 substrate, the CdSe QDs were adsorbed onto the surfaces of the TiO 2 nanoparticles and electron transfer from the CdSe QDs to the TiO 2 nanoparticles could occur, which had been confirmed by the photocurrent measurements (Shen et al., 2004a(Shen et al., , 2004b)).As mentioned above, the TG signal due to the change in the refractive index () nt  of the TiO 2 nanoparticles resulted from the injected free electrons from the CdSe QDs would be very small and can be ignored.Therefore, the slow decay process of the CdSe QDs adsorbed on the nanostructrued TiO 2 film mostly reflected the decrease of the photoexcited electron densities in the CdSe QDs due to both electron trapping at the CdSe QD surfaces/interfaces and electron transfer from the CdSe QDs to the TiO 2 nanoparticles.This is the reason why the τ 2 obtained for the CdSe QDs adsorbed on the nanostructured TiO 2 electrode was much smaller than that obtained for the CdSe QDs adsorbed on the glass substrate.For the CdSe QDs on the nanostructured TiO 2 electrode, the decay rate k 2 (k 2 = 1/τ 2 ) can be expressed as where k r is the intrinsic decay rate (mostly trapping) in CdSe QDs and k et is the electrontransfer rate from CdSe QDs to TiO 2 .The intrinsic decay time of electrons in the CdSe QDs adsorbed on the nanostructured TiO 2 electrode was assumed to be the same as that in the CdSe QDs adsorbed on the glass substrate.In this way, we can estimate the electron transfer rate from the CdSe QDs into the nanostructured TiO 2 electrode to be approximately 5.6 x 10 9 s -1 using the values of τ 2 obtained from the CdSe QDs adsorbed on these two kinds of substrates.This result is very close to the electron transfer rates from CdSe QDs to TiO 2 substrates measured recently using a femtosecond transient absorption (TA) technique by Kamat and co-workers (Robel et al., 2007).

Changes of carrier dynamics in CdSe QDs adsorbed onto TiO 2 nanostructured electrodes versus adsorption conditions
The photoexcited carrier dynamics of CdSe QDs, including the electron injection rates to the TiO 2 electrodes, depends greatly on the size of the CdSe QDs, the adsorption method (mode of attachment), the adsorption conditons such as adsorption time, and the properties of the TiO 2 electrodes such as crystal structure.In the following, two examples will be introduced.
The first example is the carrier dynamics of CdSe QDs adsorbed onto TiO 2 electrodes with SILAR method (Guijarro et al., 2010a).Figure 5 shows the normalized TG kinetics of CdSe adsorbed TiO 2 electrodes with different SILAR cycles and an example of the fitting result of the TG kinetics with eq. ( 3) to determine the parameters A 1 , A 2 , τ 1 and τ 2 shown in Fig. 5.
Figure 6 shows the dependence of τ 1 and τ 2 on the number of SILAR cycles.As shown in Fig. 6, τ 1 increases from 2.8 to 6.3 ps and τ 2 increases from 83 to 320 ps upon increasing the number of SILAR cycles from 2 to 9-15.With the SILAR cycle increasing, both the average QD size and the number of QDs increase.Therefore, for a low number of cycles, all the CdSe QDs would be in direct contact with the TiO 2 electrode, favouring electron injection.3)) (b) (Guijarro et al., 2010a).
However, for a large number of deposition cycles, a large fraction of CdSe QDs would not be in direct contact with the TiO 2 electrode, resulting in an increase in electron injection time.As discussed in depth by Guijarro and co-workers (Guijarro et al., 2010a), in this case, the fast decay component was attributed to direct electron transfer from the first layer of deposited CdSe QDs besides the hole trapping, and the slow one to electron injection into the TiO 2 from CdSe QDs in the outer layers (and trapping at surface and interfacial QD/QD states).For the fast decay component, the increase in QD average sizes resulted from the increase of SILAR cycles likely plays a key role in determining the values of τ 1 .The results were in qualitative agreement with those reported by Kamat and co-workers (Robel et al., 2007) showing that electron injection into TiO 2 from CdSe QDs becomes faster as the QD size gets smaller.It was considered to be due to an increase in the driving force as the CdSe conduction band shifts toward higher energies.A plot of τ 1 versus the number of SILAR cycle numbers (Fig. 6(a)) shows approximately a linear dependence.For the slow decay component, the electron relaxation time τ 2 also increases upon increasing the number of SILAR cycles (Fig. 6(b)), which can be explained as follows.The increase in the number of SILAR cycles results in not only an average increase of the CdSe quantum dot size but also an increase in the average distance of the CdSe QDs to the oxide particle.Correspondingly, the concentration of QDs not in direct contact with the TiO 2 would also grow.This leads to an increase in the average time needed for transferring the photoexcited electrons from the QDs to the oxide.Therefore, the values of A 1 and A 2 mainly reflected the relative contribution of the fast electron transfer (from the first QD layer in direct contact with the TiO 2 ) and the slow one (from the QDs in the outer layers), respectively.A 1 decreases from Fig. 6.Dependence of τ 1 (A) and τ 2 (B) on the number of SILAR cycles.Dashed lines correspond to linear fittings including all data, whereas the solid line shows a linear fitting forced to pass through the origin without considering the point corresponding to 15 SILAR cycles (Guijarro et al., 2010a).
The second example is the carrier dynamics of CdSe QDs adsorbed onto TiO 2 electrodes with linker-assisted (LA) method (Guijarro et al., 2010b).Figure 7 shows the TG kinetics for TiO 2 nanostrucuted electrodes modified by linker-adsorbed (using cysteine, pmercaptobenzoic acid (MBA) and mercaptopropionic acid (MPA)) CdSe QDs (average size: 3.5 nm), together with the TG kinetics of the CdSe QD solution used for the adsorption.The TG kinetics of the LA adsorption samples were not sensitive to adsorption time.It indicates that the specific interaction between the thiol group (-SH) of the three kinds of linkers and the QD leads to homogeneous adsorption, in which all the QDs are adsorbed directly on the TiO 2 surface with no aggregation.As shown in Fig. 7, strong dependence of the decay of the TG kinetics on the linker nature was observed.This result suggests that, not only the length, but also other factors such as the dipole moment, the redox properties or the electronic structure of the linkers may play a role in the carrier dynamics.The TG kinetics of the LA adosprtion samples were fitted very well to eq. ( 3),

B.
 and electron transfer rate constants were calculated according to eq. ( 4) by assuming that the intrinsic decay rate constant in adsorbed TiO 2 electrodes is the same as that in the CdSe colloidal dispersion (Table 2).The values of the electron transfer rate constant from CdSe to TiO 2 via MPA are in good agreement with the results obtained by Kamat's group (Robel et al., 2007), ranging from 1.0 x 10 9 to 2 x 10 11 s -1 .With respect to the effect of QD size, it was found that the smaller the QD, the faster the electron injection.Very interestingly, a direct correalation between the ultrafast carrier dynamics and the incident photon current conversion efficiency (IPCE) values measured in the absence of electron acceptors in solution for CdSe QDSCs was found.(Guijarro et al., 2010b).Fig. 7. TG kinetics for CdSe-sensitized TiO 2 electrodes using cysteine, MPA and MBA as linkers.
All the curves are normalized to the maximum value.The TG kinetics corresponding to the CdSe solution used for LA adsorption is also given for comparison (Guijarro et al., 2010b).

Effect of surface modification on the ultrafast carrier dynamics and photovoltaic properties of CdSe QD sensitized TiO 2 solar cells
As mentioned above, the energy conversion efficiency of QDSCs is still less than 5% at present (Mora-Sero', et al., 2010;Zhang et al., 2011) and more studies are necessary for improving the photovoltaic properties of QDSCs.The poor photovoltaic performance of QDSCs usually results from three kinds of recombination.The first one is the recombination of the injected electrons in TiO 2 electrodes with electrolyte redox species.The second one is the recombination of photoexcited electrons in the QDs with electrolyte redox species.The third one is the recombination of photoexcited electrons and holes through surface and/or interface defects.To improve the photovoltaic performance, surface passivation should be carried out.In the previous work (Shen et al., 2008;Diguna et al., 2007a), it was found that ZnS surface coating on the CdSe QDs could imporve the photovoltaic properties of CdSe QDSCs significantly.In addition to ZnS, the authors also tried surface mosifications with Zn 2+ and S 2-surface adsorption.The effects of the three kinds of surface modification (ZnS, Zn 2+ , and S 2-) on the photovoltaic performances and ultrafast carrier dynamics of CdSe QDSCs have been investigated.The three kinds of surface modifications for CdSe QD adsorbed TiO 2 nanostructured electrodes (24 h adsorption at 10 ℃ using CBD method) were carried out as follows: (1) the samples were coated with ZnS by alternately dipping them two times in 0.1 M Zn(CH 2 COO) and 0.1 M Na 2 S aqueous solutions for 1 minute for each dip; (2) the samples were adsorbed with Zn 2+ by dipping them in 0.1 M Zn(CH 2 COO) for 6 min; (3) the samples were adsorbed with S 2-by dipping them in 0.1 M Na 2 S for 6 min.For characterization of the photovoltaic performances, solar cells were assembled using a CdSe QD sensitized TiO 2 electrode as the working electrode and a Cu 2 S film as the counter electrode (Hodes et al., 1980).Polysulfide electrolyte (1 M S and 1 M Na 2 S solution) was used as the regenerative redox couple (Shen et al., 2008b).The active area of the cells was 0.25 cm 2 .The photovoltaic characteristics of the solar cells were measured using a solar simulator (Peccell Technologies, Inc.) at one sun (AM1.5, 100 mW/cm 2 ).For the TG measurements, the pump and the probe wavelengths are 530 nm and 775 nm, respectively.Figure 8 shows the photocurrent density-photovoltage curves of CdSe QDSCs, in which the QD surfaces were not modified and modified with Zn 2+ , S 2-, and ZnS, respectively.The photovoltaic properties of short circuit current density, open circuit voltage and fill factor (FF) were improved after modifying the surface with ZnS or Zn 2+ , especially the short circuit current density was improved significantly (Table 3).As a result, the energy conversion efficiencies of the samples with ZnS or Zn 2+ surface modification increased as many as 3 times compared to the sample without surface modification.The improvement of the photovoltaic performances by means of the ZnS or Zn 2+ surface modification may be due to (1) the decrease of surface states of TiO 2 and CdSe QDs and ( 2) the formation of a potential barrier at the CdSe QD/electrolyte and TiO 2 /electrolyte interfaces.However, there is little effect of the modification with S 2-on the photovoltaic properties.So, maybe Se 2-is rich on the CdSe QD surfaces.
Figure 9 shows the TG kinetics of the CdSe QD adsorbed TiO 2 electrodes before and after Zn 2+ and ZnS surface modification.From the TG kinetics, two decay processes were observed.The fast and slow decay processes were mostly attributed to photoexcited hole trapping and electron transfer dynamics, as discussed earlier.For the sample before and after Zn 2+ surface modification, the fast decay times were 4.5 and 2 ps, and the slow decay times were 53 ps and 25 ps, respectively.It indicates that the hole and electron relaxation became faster after the surface modification.Similar results were also obtained after the ZnS and S 2-surface modifications as shown in Fig. 10.The faster electron transfer due to the surface modification may result from the decrease of the surface trapping states.Further studies on the mechanism of the effects of the surface modifications on the photovoltaic performances and the carrier dynamics are in progress now.3. Photovoltaic properties of CdSe QDSCs before and after surface modifications with Zn 2+ , S 2-, and ZnS, respectively.J sc is the short circuit current density, V oc is the open circuit voltage, FF is the fill factor, and η is the energy conversion efficiency.

Conclusion
In summary, the photoexcited carrier dynamics in CdSe QDs adsorbed onto TiO 2 nanostructured electrodes have been characterized by using the improved TG technique.It has been demonstrated that both photoexcited electron and hole dynamics can be detected by using this improved TG technique.By comparing the TG responses measured in air and in a Na 2 S solution (hole acceptor), the dynamics of photoexcited electrons and holes in the CdSe QDs has been sucssesfully separated from each other.It was found that charge separation in the CdSe QDs occurred over a very fast time scale from a few hundreds of fs in the Na 2 S solution via hole transfer to S 2-ions to a few ps in air via hole trapping.On the other hand, the electron dynamics in the CdSe QDs, including trapping and injection to the metal oxide electrodes, depends greatly on the QD size, and adsorption methods (such as CBD, SILAR, DA and LA adsorption methods) and conditons (such as adsorption time and SIALR cycle number).In addition, surface modifications such as ZnS coating and adsorption with Zn 2+ have been demonstrated to improve the photovoltaic properties and have a great influence on the ultrafast carrier dynamics of CdSe QDSCs.Detailed studies on the correlations between the carrier dynamics and the photovoltaic properties in QDSCs are in progress now.

Fig. 1 .
Fig. 1.Dependence of the TG kinetics of CdSe QDs adsorbed onto nanostructured TiO 2 films (CBD method and the CdSe adsorption time was 24 h) on the pump light intensity (a); Dependence of the TG peak intensity on the pump light intensity (b); Normalized TG kinetics measured with different pump light intensities (c)(Shen et al., 2010a).

Fig. 2 .
Fig. 2. TG kinetics of CdSe QDs adsorbed onto a nanostructured TiO 2 film measured in air.The vertical axis is plotted on a logarithmic scale.Three decay processes A, B and C can be clearly observed(Shen et al., 2010a).

Fig
Fig. 8. Photocurrent density-photovoltage characteristics of CdSe QD-sensitized solar cells without and with surface modifications.

Table 1 .
Fitting results of the TG responses of CdSe QDs adsorbed onto nanostructured TiO 2 Ultrafast Electron and Hole Dynamics in CdSe Quantum Dot Sensitized Solar Cells 299 0.84 to 0.41 while A 2 increases from 0.22 to 0.54 as the SILAR cycle increases from 2 to 9-15.For a low number of SILAR cycles, A 1 is large, indicating the predominance of electron injection from QDs in direct contact with the TiO 2 .As the SILAR cycle increases, A 2 increases because the contribution of the slower electron transfer through CdSe/CdSe interfaces becomes more and more important.
. 8. Photocurrent density-photovoltage characteristics of CdSe QD-sensitized solar cells without and with surface modifications.