Ultrafast Direct and Indirect Electron-Injection Processes in a


Ultrafast Direct and Indirect Electron-Injection Processes in a...

4 downloads 431 Views 156KB Size

J. Phys. Chem. B 2004, 108, 12583-12592

12583

Ultrafast Direct and Indirect Electron-Injection Processes in a Photoexcited Dye-Sensitized Nanocrystalline Zinc Oxide Film: The Importance of Exciplex Intermediates at the Surface Akihiro Furube,*,† Ryuzi Katoh,† Toshitada Yoshihara,† Kohjiro Hara,† Shigeo Murata,† Hironori Arakawa,† and M. Tachiya‡ Photoreaction Control Research Center (PCRC), National Institute of AdVanced Industrial Science and Technology (AIST), Tsukuba Central 5, 1-1-1 Higashi, Tsukuba, Ibaraki 305-8565, Japan ReceiVed: March 19, 2004; In Final Form: June 16, 2004

The processes involved in ultrafast electron injection from a photoexcited novel coumarin dye (NKX-2311), an efficient photosensitizer for TiO2-based dye-sensitized solar cells, into the conduction band of a nanocrystalline ZnO film were investigated by observing the femtosecond transient absorption in the visibleto-infrared range (600-5000 nm). After photoexcitation of adsorbed NKX-2311 dye, the stimulated emission and absorption of the singlet excited dye decayed with a 500-fs time constant. These were followed by rises in absorptions of the oxidized dye and conduction-band electrons, indicating a direct electron-injection process. In addition, indirect electron-injection processes involving intermediates were identified. The intermediates showed stimulated emission at longer wavelengths than that of the excited dye; very broad absorptions in the near-IR region (900-1300 nm) were observed immediately after excitation, and they decayed with 1- and 10-ps time constants, leading to further rises in the absorption of electrons in the conduction band. On the basis of their dynamics and spectral features, the species decaying with 1- and 10-ps time constants were assigned to neutral and ionic exciplexes, respectively. The multiplicity of electron-injection processes is due to the presence of various adsorption sites for the dye on the ZnO surface.

1. Introduction The mechanism of interfacial charge injection from molecularly localized electronic levels into delocalized continuum levels in solids is an important problem in fundamental science, as well as in terms of realizing a high performance in various applications, such as solar energy conversion (based on dyesensitized semiconductors1-3 and organic heterojunctions), silver halide photography,4 photocatalysis, molecular electronics, and sensing.5,6 Among these applications, dye sensitization is a useful technique for utilizing visible sunlight to generate charge carriers in wide band gap semiconductors. For example, in the case of silver halide photography, J-aggregates of cyanine dyes work as sensitizers to initiate reduction of silver ions in silver halide crystals by photoinduced electron injection from the excited dye to the conduction band of the silver halide crystals.4 In dyesensitized solar cells (DSSCs), Ru-complex dyes and many types of organic dyes are known to work as efficient sensitizers when they are adsorbed on a semiconductor surface such as TiO2 or ZnO.1,2 Electrons injected into the conduction band are used to generate electrical power. In these reactions, the electronic coupling between the excited dyes and the continuous levels in the conduction bands is important for efficient charge separation to occur. In addition, similar or stronger coupling can be expected between the excited dyes and the localized states that exist on the surface. This situation is similar to the well-known state of donor-acceptor molecules in solution, because the surface state has a localized electronic state. Electron transfer between localized electronic * Corresponding author. E-mail: [email protected]. † Photoreaction Control Research Center (PCRC). ‡ National Institute of Advanced Industrial Science and Technology (AIST).

states has been studied extensively in photochemistry. In solution, donor and acceptor molecules sometimes form a complex (exciplex) as a result of charge-transfer (CT) interaction.7,8 Excited molecules on a surface can interact in a similar way with localized states on the surface. The excited molecules can therefore interact with many electron-acceptor states, including continuous, delocalized states in the conduction band and discrete, localized states at the surface. Complicated reactions, including CT complex formation, can occur during charge separation at the surface. Because of this complexity, the mechanism of interfacial electron transfer is not well understood. Ultrafast spectroscopy is a powerful experimental technique for studying interfacial electron transfer, and many recently published reports discuss its mechanism, particularly for the case of dye-sensitized semiconductors.9-28 The electron-injection times from excited molecules to conduction bands are of the order of femtoseconds to nanoseconds. Attempts have been made to understand the electron-transfer mechanism by controlling the physical parameters involved, such as the coupling strengths between donors and acceptors, the energy levels of donors and acceptors, and their environments. For instance, the donor-acceptor distance can be controlled by introducing a molecular spacer between the chromophore and the anchoring group chemically connected to the semiconductor surface.16,29 This spacer can control the electronic coupling strength of the electron-transfer reaction. Electrochemical methods can be used to determine and control the conduction-band level relative to the LUMO level of the dye molecules when nanocrystalline semiconductor films are prepared on electrodes. The environments of such films are easily varied from vacuum to various gases or liquids.

10.1021/jp0487713 CCC: $27.50 © 2004 American Chemical Society Published on Web 07/24/2004

12584 J. Phys. Chem. B, Vol. 108, No. 33, 2004

Furube et al.

SCHEME 1: Schematic Showing Energy Levels in Exciplexes

Nevertheless, the mechanism is still not well understood, mainly because of the difficulty in obtaining spectral information on electrons transferred to acceptor states in the semiconductor. Lian and co-workers have successfully applied a femtosecond IR probe technique to observe electrons injected into the conduction band, which give a strong IR absorption band due to intraband transitions within the conduction band.16,17,27,30 Turner and co-workers used ultrafast terahertz spectroscopy to measure the conductivity of the injected electrons directly.26 These studies have revealed the processes for the generation of “conductive” electrons in the conduction band; however, electron-acceptor states are believed to consist of not only the continuum levels in the conduction band but also surface states.12,17 When a localized electron donor in the excited state (D*) is coupled with an acceptor through a charge-transfer (CT) interaction, a complex called an exciplex can form. In the exciplex, the electronic states of a donor-acceptor pair (D*A), with a wave function ΦN, and an ion pair (D+A-), with a wave function ΦI, mix together through CT interaction.31-33 As presented in Scheme 1, when D*A is energetically more stable than D+A- by E0, the exciplex formed has a neutral character (presented as [D*A]ex with an approximate wave function ΦN + δΦI), and the exciplex is energetically stabilized relative to the D*A state. At the same time, D+A- changes into a less stable state, [D+A-]ex (δΦN + ΦI). The optical transition from [D*A]ex to [D+A-]ex with an energy difference of ECT corresponds to a CT transition that transfers an electron from D* to A. Therefore, [D*A]ex shows absorption bands similar to those of D* [locally excited (LE) state] and the CT band in transient absorption spectroscopy.31,34-36 On the other hand, when D+Ais energetically more stable than D*A, an exciplex composed of an ion pair, [D+A-]ex (δΦN + ΦI), is formed and shows absorption bands similar to those of D+ and A-, together with the CT band where an optical transition from [D+A-]ex to [D*A]ex occurs.31,34,35,37 In this paper, we call the exciplex having a neutral character “neutral exciplex” and that composed of an ion pair “ionic exciplex”. A dye-sensitized semiconductor system that has been widely investigated by ultrafast spectroscopy is the efficient dyesensitization system composed of N3 dye [cis-bis(4,4′-dicarboxy-2,2′-bipyridine)dithiocyanato ruthenium(II), [Ru(dcbpy)2(NCS)2]] anchored to a TiO2 nanocrystalline film. The injection in this system was found to be as fast as tens of femtoseconds for the major path, although it was accompanied by slower reaction paths operating on the tens-of-picoseconds scale.13 The contribution of this reaction was proved by two independent

experiments in which ultrashort visible and IR laser probes were used to observe the generation of oxidized dye13,21 and injected electrons,16,30 respectively. Both these products showed an ultrafast rise within 50 fs as a major component and an additional picosecond rise. These multiple paths imply a diversity of interactions between the excited dye and electronacceptor states. Recently, we pointed out the importance of localized surface states in the electron-injection process from photoexcited N3 dye into the conduction band of a ZnO nanocrystalline film.38 We identified the presence of indirect electron injection, that is, electron injection via an intermediate state. The intermediate showed an immediate absorption rise in the near-IR region after photoexcitation of N3 dye, and the rise was followed by a slow multiexponential decay on a 1-100-ps time scale. The decay was then followed by a rise in the absorption of the conductionband electrons at 2 µm. We assigned the intermediate to an exciplex formed at the ZnO surface and having a CT band in the near-IR region. The generation of the conduction-band electrons is considered to be retarded by the formation of the intermediate through CT interaction between the excited N3 dye and the surface state. The stepwise reaction process indicates that the excited N3 dye interacts strongly with the surface states rather than with the conduction band when the dye is adsorbed on ZnO. The rapid (∼50 fs) generation of electrons in the conduction band for N3 dye on TiO2 is mainly due to the higher density of states in the conduction band of TiO2 compared with ZnO.16 Similar experimental results showing slow and multiexponential generation of conduction-band electrons have been reported for the same system and for similar systems.39,40 The existence of the intermediate could be revealed by near-IR spectroscopy, which is a suitable technique for obtaining information on intermolecular interactions. In this paper, we report clearer evidence of stepwise processes for electron injection into ZnO via exciplex intermediates at the surface, and we discuss the electronic structure of the intermediates. In the previous report for N3 dye on ZnO, we could not determine whether the observed exciplex was neutral or ionic. The dye used in this study is a newly synthesized coumarin derivative (NKX-2311) that was recently reported to be a highly efficient photosensitizer for a dye-sensitized nanocrystalline TiO2 solar cell showing a 6% solar energy-toelectricity conversion efficiency.41,42 The molecular structure is shown in Chart 1. A methine unit (-CHdCH-) connecting the cyano (-CN) and carboxyl (-COOH) groups is introduced into the coumarin framework. Photoexcitation at around 500 nm is considered to cause an intramolecular CT transition that

Electron Injection in Dye-Sensitized ZnO Film

J. Phys. Chem. B, Vol. 108, No. 33, 2004 12585

CHART 1: Molecular Structure of the Coumarin Derivative NKX-2311

transfers the electron from the coumarin framework to the methine unit.41,42 Because the dye is closely anchored to the ZnO surface through the carboxyl group, a strong interaction between the excited dye and ZnO is expected. The LUMO of NKX-2311 is known to be located about 0.2 eV above the bottom of the conduction band of ZnO.43 We have investigated the electron-injection dynamics from the photoexcited NKX-2311 dye to the conduction band of a ZnO nanocrystalline film by measuring the transient absorption in the spectral range from 600 to 5000 nm to observe groundstate absorption bleach, stimulated emission, absorption of the excited singlet state of the dye, that of the oxidized dye, CT absorption band appearing in the near-IR region, and absorption of conduction-band electrons in the IR region. We found multiple reaction paths, including direct and indirect injection processes. No intermediate was observed in the direct process, whereas exciplex intermediates were clearly observed in the indirect processes. We consider that photoexcited dye can interact both with localized states at the surface and with continuum levels in the conduction band. The multiplicity of electron-injection paths and times is discussed in terms of the diversity of the adsorption sites and the change in the density of states (DOS) for the ZnO conduction band with the energy levels. 2. Experimental Section 2.1. Samples. The samples measured are 2-cyano-5-(1,1,6,6tetramethyl-10-oxo-2,3,5,6-tetrahydro-1H,4H,10H-11-oxa-3aazabenzo[de]-anthracen-9-yl)penta-2,4-dienoic acid (NKX-2311, prepared by Hayashibara Biochemical Laboratories) adsorbed on ZnO nanocrystalline films. The detailed procedure for the synthesis of the novel coumarin dye is shown in the Supporting Information for Hara et al.42 A HOMO-LUMO calculation indicated that the electron moves from the coumarin framework to the -CHdCH- unit under visible-light excitation of the dye.42 The ZnO film was prepared on a glass substrate by a similar procedure reported elsewhere,43 and its thickness was about 5 µm. The dye was loaded by immersing the bare ZnO film into the dye solution for several hours. After drying the film, steady-state and time-resolved experiments were performed. This sample is designated NKX-2311/ZnO. The LUMO of NKX-2311 is known to be located about 0.2 eV above the bottom of the conduction band of ZnO.43 The monochromatic incident photon-to-current conversion efficiency from 420 to 600 nm for a DSSC based on NKX-2311 and ZnO was several tens of percentage points. 2.2. Fluorescence Lifetime Measurements. A streak camera (Hamamatsu, Streak Scope C4334) was used to measure photoluminescence lifetimes of the dye in solution. The excitation light source was the second-harmonic light (400 nm) of an amplified femtosecond mode-locked Ti:sapphire laser (Spectra Physics, Super Spitfire). The sample solution was put into a quartz cell (1 cm × 1 cm), and the fluorescence light was introduced into the streak camera through a monochromator at a 90° angle with respect to the excitation beam. The time resolution of the measurements was 30-50 ps. 2.3. Nanosecond Transient Absorption Measurements. Nanosecond transient absorption measurements were made to

Figure 1. UV-vis absorption (thin line) and fluorescence (thick line) spectra of NKX-2311 in ethanol and tert-butyl alcohol/acetonitrile (1: 1), respectively. The inset shows the decay curve of the emission, indicating a lifetime of 1.9 ns.

obtain basic spectral data for the NKX-2311 dye in solution, the bare ZnO film, and the dye adsorbed on the ZnO film. A second- or third-harmonic pulse (532 or 355 nm, duration 8 ns) from a Nd3+:YAG laser (Continuum, Surelite II) was used as the pumping light. A halogen lamp (50 W) was used as the probe light source. The probe light transmitted through the sample specimen was detected with a Si, InGaAs, or mercury cadmium telluride (MCT) photodetector after being dispersed with a monochromator (Acton Research, SpectraPro-150). The signal at a particular wavelength was recorded with a digital oscilloscope (Tektronix, TDS680C) and analyzed by a computer to obtain the change in absorbance with wavelength. The details of this system have already been reported elsewhere.43 2.4. Femtosecond Transient Absorption Spectroscopy. The light source for the femtosecond pump-probe transient absorption measurements was a regenerative/multipath double-stage amplifier system of a Ti:sapphire laser (800-nm wavelength, 50-fs fwhm pulse width, 1.4 mJ/pulse intensity, 1-kHz repetition; Spectra Physics, Super Spitfire) combined with two optical parametric amplifiers (OPAs; Spectra Physics, OPA-800). To probe the IR wavelength at 5 µm, another regenerative amplifier system of a Ti:sapphire laser (800-nm wavelength, 160-fs fwhm pulse width, 1.0 mJ/pulse intensity, 1-kHz repetition; Spectra Physics, Hurricane) combined with two OPAs (Quantronix, Topas) was used. For a pump pulse, the output of the OPA at a wavelength of 540 nm with an intensity of several µJ per pulse at a 500-Hz modulation frequency was used; and for a probe pulse, the output of the other OPA or the white-light continuum generated by focusing the fundamental beam (800 nm) onto a sapphire plate was used. The probe beam was focused at the center of the pump beam on the sample and then detected by a Si, InGaAs, or MCT photodetector after passing through a monochromator (Acton Research, SpectraPro-150). The polarization of the pump and probe beams was set to the magic angle condition. A part of the probe beam was split before the sample and detected by another set comprising a monochromator and a photodetector. The two output signals of the photodetectors were analyzed with three boxcar integrators and an analogue processor (Stanford Research) to obtain the absorbance change, typically 10-3 OD. All measurements were performed at 295 K. 3. Results 3.1. Spectroscopic Characterization of NKX-2311 Dye in Solution. Figure 1 shows the UV-vis absorption and fluorescence spectra of NKX-2311 in ethanol and tert-butyl alcohol/ acetonitrile (1:1) mixed solvent, respectively. The maxima of absorption and fluorescence are located at 490 and 555 nm,

12586 J. Phys. Chem. B, Vol. 108, No. 33, 2004

Furube et al.

Figure 3. Steady-state absorption spectra of NKX-2311 adsorbed on a ZnO nanocrystalline film and in ethanol.

Figure 2. (a) S-S absorption spectra of NKX-2311 in an acidic CD3OD solution at the delay times of 1.0 and 25 ps and the time profile at 1300 nm (inset). The excitation wavelength was 540 nm. (b) The absorption spectrum of the oxidized form of NKX-2311 (dye+) measured by adding an electron acceptor, 1,4-benzoquinone, to a CD3OD solution of NKX-2311 (closed circles) using a nanosecond laser system.

respectively. The lifetime of the fluorescence was determined to be 1.9 ns from the decay curve obtained with the streak camera, as shown in the inset to Figure 1. We measured the S-S absorption spectra of NKX-2311 after the femtosecond laser excitation to be able to discuss the ultrafast excited-state dynamics of NKX-2311 on ZnO through a comparison with the condition in which electron injection does not occur. Transient absorption of the dye in acidic CD3OD solution was measured in the case where the pump wavelength was 540 nm and its intensity was about 5 µJ/pulse. The observed transient absorption spectra at 1.0 and 25 ps and the time profile at 1300 nm are shown in Figure 2a. The spectral peak is at 1260 nm at 1.0 ps, and spectral blue-shift and sharpening with a time constant of 8 ps are observed. These dynamics are considered to be a consequence of vibrational cooling and solvation processes. The contribution of the latter process is considered to be large, because the photoexcitation induces an intramolecular CT state. Because these dynamics and the spectral peak position were slightly dependent on whether the solution was acidic or basic, it is considered that the proton in the carboxyl group affects the electronic properties of the coumarin chromophore to some extent. Actually, in basic CD3OD, the spectral peak was located at a 100-nm longer wavelength than in acidic solution, and the spectral evolution was analyzed with a 7-ps time constant. In this case the dye exists as an anion in the solution; therefore the spectra measured in the acidic condition, where the dye is neutral, will be more suitable as reference data in the experiment of the dye-sensitized film. We have examined the electron distribution of the LUMO and HOMO of the dye on a semiconductor surface based on a DFT

calculation and FTIR measurements in our previous study,42 and in a series of calculations we confirmed that the electron distribution of the LUMO and HOMO for an adsorbed dye is almost similar to that for a neutral dye. To obtain the absorption spectrum of the oxidized form of NKX-2311 (dye+), we measured the transient absorption of a CD3OD solution containing NKX-2311 and an electron acceptor, 1,4-benzoquinone. The concentrations of benzoquinone and NKX-2311 were 0.1 and 1 × 10-5 M, respectively. The absorption spectrum in the near-IR range was obtained (Figure 2b) at 100 ns after 532-nm light excitation. There are two characteristic peaks at 875 and 1010 nm. The radical anion of benzoquinone does not have absorption in this spectral range,44 and the lifetime of the excited singlet state of NKX-2311 is 1.9 ns, as mentioned above; therefore, the observed spectrum can be ascribed to the absorption of oxidized NKX-2311. The absorption of oxidized NKX-2311 was also measured by using another electron acceptor, methyl viologen, whose molar extinction coefficient in the reduced form is known. By comparing the absorptions of the transient species, the molar extinction coefficient of NKX-2311 in the oxidized form was estimated to be 6000 mol-1‚dm3‚cm-1 at 1010 nm.45 3.2. Spectroscopic Characterization of NKX-2311/ZnO Film. Figure 3 shows the steady-state UV-vis absorption spectrum of NKX-2311/ZnO. The spectrum of NKX-2311 in ethanol solution is also shown for comparison. Adsorption of the dye onto ZnO leads to spectral blue-shift and slight broadening. These changes in the spectrum may be due to electronic interaction between the dye and the ZnO surface and, probably, to inhomogeneity of the interaction strength. The strong absorption below 400 nm of NKX-2311/ZnO is a result of the band gap transition of ZnO. Our group has already reported the nanosecond transient absorption spectra of electrons in the conduction band (e-CB) of ZnO under band gap excitation (355 nm) of a bare ZnO film and under visible-light excitation (532 nm) of NKX-2311/ZnO.36 Both experiments gave the same absorption band of e-CB. The band intensity increases with wavelength, and the absorption was assigned to the transition of electrons from free or shallowly trapped states near the conduction-band bottom to the higher states in the same band. Recently, we have estimated the efficiency of electron injection of NKX-2311/ZnO as 0.8 ( 0.1.38 Figure 4 shows the transient absorption spectrum in the visible and near-IR ranges of NKX-2311/ZnO excited at 532 nm with the nanosecond laser. The spectrum consists of absorption spectra of dye+ and e-CB. The absorption maxima of dye+ for NKX-2311/ZnO are red-shifted by 10 nm compared with those of dye+ in solution. The absorbance is converted to the molar absorption coefficient by assuming that the value for dye+ in NKX-2311/ZnO is the same as that in solution at the peak wavelengths.

Electron Injection in Dye-Sensitized ZnO Film

Figure 4. Transient absorption spectrum of an NKX-2311/ZnO film, measured by the nanosecond transient absorption system under 532nm excitation.

Figure 5. Top: Transient absorption spectra of NKX-2311/ZnO between 600 and 1350 nm at delay times of 2 (thick solid line), 10 (thick dotted line), and 100 ps (thin solid line) after 540 nm excitation by the femtosecond pulse with the intensity of about 10 µJ. The spectrum at 50 ns, consisting of dye+ and e-CB absorptions, is also shown (thin dotted line) following normalization to the peak intensity of the spectrum at 100 ps. The dot-dashed line indicates an estimated CT band shape of an ionic exciplex. Bottom: Reference spectra shown for comparison; S-S absorbance (at 1.0 ps) in solution (thin solid lines), oxidized dye (dye+) absorbance in solution (thick solid line), dye fluorescence of a solution sample (dashed line), and steady-state absorption of NKX-2311/ZnO (dotted line). The last two spectra are shown in inverted form.

3.3. Femtosecond Transient Absorption of NKX-2311/ZnO Film. Figure 5 shows the femtosecond time-resolved transient absorption spectra of NKX-2311/ZnO in the spectral range between 600 and 1350 nm at delay times of 2, 10, and 100 ps after excitation at 540 nm with an intensity of about 10 µJ and a focus diameter of about 0.5 mm. The spectrum at a delay time longer than 50 ns is also presented following normalization to the peak intensity of the spectrum at 100 ps. For comparison, spectra shown already, that is, the S-S absorption of dye* in solution (at the 1.0-ps delay time), dye+ absorption in solution, fluorescence of the dye in solution, and steady-state absorption of NKX-2311/ZnO, are presented in the lower panel in the figure. The last two spectra are plotted upside down with respect to the vertical axis because in transient absorption measurements, stimulated emission and absorption bleaching appear on the negative side. In the femtosecond measurements in the visible region, negative signals are observed that are due to transient bleaching of the ground-state absorption (up to ∼630 nm) and to stimulated emission on its longer wavelength side (up to ∼780 nm). The stimulated emission seems to extend toward the longer wavelength side compared with the steady-state emission of the

J. Phys. Chem. B, Vol. 108, No. 33, 2004 12587 dye in solution. At 100 ps, the transient bleaching still remains while the emission has already disappeared. In the near-IR region, two peaks at 900 and 1050 nm are clearly observed at 2 and 100 ps. The spectral shape is identical with that of dye+ in CD3OD solution except for a slight redshift by about 40 nm (0.04 eV). Actually, the spectrum slightly shifts with time toward shorter wavelengths, and at 100 ps, it is almost the same as that at 50 ns. We will discuss this spectral shift later. From the appearance of the dye+ absorption band, it is obvious that the dye is oxidized immediately after the photoexcitation and that dye+ persists over nanoseconds. The transient absorption spectrum of dye+ at 2 ps seems broader than that at 100 ps. This is because a broad absorption band, shown by a dot-dashed line, overlaps with the dye+ absorption at early times. The analysis for obtaining the dot-dashed line spectrum is described latter. It is also worth noting that the intensity of the dye+ band itself is quite similar at 2 and 100 ps, judging from the height of the spectral maximum measured from the minimum between the two peaks. The intensity of the broad band, on the other hand, decreases with time. The time profiles for the transient absorption at some specific wavelengths were measured to examine the electron-injection dynamics from dye* to the conduction band of ZnO. The wavelengths 600, 700, 1050, 1150, 1300, and 2630 nm were chosen to permit selective observation of the stimulated emission (and the bleaching of ground-state absorption), the same emission on the longer wavelength side of the fluorescence peak in solution, dye+, the broad absorption band, the excited singlet state of the dye, and e-CB, respectively. Correspondences between probe wavelengths and observable transient species are summarized in Table 1. The excitation intensity was almost the same for all these measurements. Transient absorption at around 5000 nm was also probed by the second laser system (Hurricane with TOPAS) to observe e-CB. These results are shown in Figure 6 by thin solid lines for the time region from -3.0 to 30 ps. Although all the observed time profiles exhibit multiexponential behavior, they can be analyzed by using three representative time constants of 500 fs, 1.0 ps, and 10 ps together with a time-independent component. The fitting function was

Absorbance(t) ) A0 + A1 exp(-t/0.5) + A2 exp(-t/1.0) + A3 exp(-t/10) (1) where t was the delay time in picoseconds and A0-A3 were the fitting parameters for the amplitudes of the constant, 0.5-, 1.0-, and 10-ps kinetics. Their amplitudes for each time profile are summarized in Table 1 and shown in Figure 7. The fitting curves drawn by thick solid lines in the figure agree well with the experimental data. At 600 nm, a rapid decay of the stimulated emission with a major decay time of 500 fs was observed. This corresponds to ultrafast quenching of dye*. At this wavelength, we also expect to observe the bleaching of the ground-state absorption just after the excitation. Indeed, the bleaching was recognized clearly at times later than 20 ps. In addition, a slow (10-ps) signal change was seen as a minor component. This signal change may be due to an overlapping positive transient absorption band that decays with a 10-ps lifetime. At 700 nm, stimulated emission on the longer wavelength side was observed. It initially showed a decay slower than that at 600 nm. The major decay time was 1.0 ps. The time profile also contained a 10-ps component and finally approached zero. At 1050 nm, an instantaneous rise within 200 fs was followed by a slow rise with time constants of 0.5 and 1.0 ps, and then

12588 J. Phys. Chem. B, Vol. 108, No. 33, 2004

Furube et al.

TABLE 1: Correspondences of Probe Wavelengths to Observable Transient Species for the Transient Absorption Measurement of NKX-2311/ZnO (SE ) Stimulated Emission)a λprobe

600 nm

700 nm

1050 nm

1150 nm

1300 nm

2630 nm

5000 nm

species

SE, bleaching

SE on the red side

dye+

broad abs (dCT bond)

S-S abs

e-CB

e-CB

-8.48 -19.5 -9.21 7.8

-0.173 -1.31 -14.1 -8.01

11.5 -7.26 -1.96 7.05

1.39 3.98 6.26 6.18

1.08 11.3 5.56 4.18

12 -2.94 -2.69 -4.23

9.63 -4.41 -8.42 -8.36

const 500 fs 1.0 ps 10 ps a

Amplitudes correspond to the best fit to the transient absorption profiles of NKX-2311/ZnO at each wavelength from 600 to 5000 nm. The fitting curve is a multiexponential function with fixed time constants of 500 fs, 1.0 ps, and 10 ps, together with a constant component. The amplitudes obtained are listed in 10-3 OD, with those for the 2630-nm data being multiplied by 3.

At 2630 nm, the absorption band of e-CB in ZnO showed a gradual rise up to 30 ps. It is clear that the generation process of e-CB takes place on all the time scales: 500 fs, 1.0 ps, and 10 ps. This indicates the existence of multiple paths for the electron-injection reaction. The absorbance measured was constant from 30 to 100 ps. The time profile at 5000 nm appeared to be almost the same as that at 2630 nm, except at early times, where instantaneous negative signals were observed. Detailed discussions about the negative signals are given later. 4. Discussion

Figure 6. Transient absorption time profiles of NKX-2311/ZnO at probe wavelengths of 600, 700, 1050, 1150, 1300, 2630, and 5000 nm after 540-nm excitation by a femtosecond pulse with an intensity of about 10 µJ. Thick solid lines indicate the result of a fitting procedure (see text for details).

Figure 7. Bar graph of the absorbance amplitudes listed in Table 1.

a slight decay with a 10-ps time constant. The absorption due to the long-lived oxidized dye remained. When the broad absorption band was selectively monitored at 1150 nm, no detectable rise component other than the instantaneous rise was found. The observed rise at 1050 nm with the 0.5- and 1-ps time constants is therefore considered to be due to the generation of dye+. At 1300 nm, a rapid decay of the S-S absorption with a 500-fs major decay time was observed. This decay time is the same as that of the stimulated emission at 600 nm. There was a small contribution of the broad absorption band, which was clearly seen at delay times longer than 2 ps.

4.1. Assignment of the Transient Absorption Bands and Reaction Paths. We first discuss the generation process of e-CB observed in the IR region, because in the visible and near-IR regions many transient species overlap each other. Absorption signals of e-CB were observed both at 2630 nm and at around 5000 nm. Their picosecond behaviors at the two wavelengths are almost the same, although bleaching of a vibrational band, probably arising from CdO or CtN stretching, seems to overlap at around 5000 nm and makes the signal negative just after excitation. This result indicates that the picosecond dynamics (after 1 ps) observed at these two wavelengths are simply due to the generation of e-CB, and any relaxation processes, such as cooling and trapping of electrons in the conduction band, which cause temporal spectral change, are negligible. The time profile at 1050 nm (the peak of the dye+ band) differs from that of e-CB at 2630 nm. At 1050 nm, no slow rise with a 10-ps time constant was seen, whereas faster-rising components within the time resolution and with time constants of 500 fs and 1.0 ps were observed. In addition, a decay behavior with a 10-ps time constant was observed at this wavelength. We therefore concluded that the generation of e-CB lasts longer than that of dye+. The different time profiles at 1050 and 2630 nm clearly show that the electron injection in NKX-2311/ZnO is not a simple charge separation process, suggesting that intermediate states are involved, as in our previous study of N3/ZnO. We previously reported that electron injection occurred stepwise through an intermediate state in N3/ZnO. The intermediate, rising within 100 fs and decaying on the 1-100-ps time scale, was assigned to an exciplex, since it showed a CT absorption band in the near-IR range. The transient absorption spectrum at 2 ps in Figure 5 shows a broad absorption band overlapping with the absorption band of dye+. The amplitude of dye+ absorption was constant between 2 and 100 ps, whereas the broad absorption band decreased with time. We therefore conclude that the 10-ps decay at 1050 nm results only from a decrease in the broad absorption band. Taking into consideration that the e-CB signal in the IR region showed a 10-ps rise, we suggest that the electron ejected from the excited dye initially forms an intermediate complex with dye+, which dissociates and releases the electron into the conduction band.

Electron Injection in Dye-Sensitized ZnO Film From these experimental results and the following consideration, the intermediate that decays on the 10-ps time scale is ascribed to an ionic exciplex. The absorption spectrum of the ionic exciplex generally shows absorption bands of ions (cation and anion) as well as a CT band. It would be expected that the absorption band of dye+ forming an ionic exciplex would be practically indistinguishable from that of free dye+. Therefore, the amplitude of the dye+ band should be constant during the dissociation of the ionic exciplex into the free dye+ and e-CB. In this period, the CT band should decrease and the e-CB band should increase. Such behaviors were indeed observed in the present study, so we tentatively assigned the broad absorption band to the CT band. An anion band, that is, an absorption band of an electron trapped at a surface state, may also absorb in the near-IR spectral range. The time profiles of the CT band and the trapped electron would be the same so that it would be difficult to separate them; however, we think that the absorption coefficient of the trapped electron is relatively small, because the nanosecond transient absorption experiment of band gap excitation of the ZnO film did not show any absorption peak around this wavelength region. Although electrons trapped at the surface are expected to exist in the bare film, only a weak absorption tail of the conductive electrons was observed between 1000 and 1500 nm. Thus, we conclude that most of the broad absorption band in the near-IR region for NKX-2311/ZnO consists of a CT band. (The absorption coefficient of the CT band is discussed later.) In addition, careful inspection of the transient absorption spectra showed that there was a slight temporal shift of the dye+ band from 2 to 100 ps, as already mentioned. This fact also supports the interpretation involving an ionic exciplex, because an electron close to the dye+ in the ionic exciplex should affect the electronic structure of dye+. The reaction scheme for 10-ps injection is written as