14514
J. Phys. Chem. B 2010, 114, 14514–14520
Charge Carrier Resolved Relaxation of the First Excitonic State in CdSe Quantum Dots Probed with Near-Infrared Transient Absorption Spectroscopy† Eric A. McArthur, Adam J. Morris-Cohen, Kathryn E. Knowles, and Emily A. Weiss* Department of Chemistry, Northwestern UniVersity, 2145 Sheridan Rd., EVanston, Illinois 60208-3113 ReceiVed: March 8, 2010; ReVised Manuscript ReceiVed: May 5, 2010
This manuscript describes a global regression analysis of near-infrared (NIR, 900-1300 nm) transient absorptions (TA) of colloidal CdSe quantum dots (QDs) photoexcited to their first (1Se1S3/2) excitonic state. Near-IR TA spectroscopy facilitates charge carrier-resolved analysis of excitonic decay of QDs because signals in the NIR are due exclusively to absorptions of photoexcited electrons and holes, as probe energies in this region are not high enough to induce absorptions across the optical bandgap that crowd the visible TA spectra. The response of each observed component of the excitonic decay to the presence of a hole-trapping ligand (1-octanethiol) and an electron-accepting ligand (1,4-benzoquinone), and comparison of time constants to those for recovery of the ground state bleaching feature in the visible TA spectrum, allow for the assignment of the components to (i) a 1.6 ps hole trapping process, (ii) 19 ps and 274 ps surface-mediated electron trapping processes, and (iii) a ∼5 ns recombination of untrapped electrons. Introduction This paper describes a global regression analysis of nearinfrared (NIR, 900-1300 nm) transient absorptions (TA) of the first excitonic state (the 1Se1S3/2 transition1,2) of 4.1 nm diameter colloidal CdSe quantum dots (QDs). We compared the timeresolved NIR spectra of CdSe QDs with their native ligands, which include long-chain alkylphosphonates3 and alkylamines,4-7 to the spectra of QDs exposed to various concentrations of 1,4benzoquinone (an electron-accepting ligand) and 1-octanethiol (a hole-trapping ligand), and subsequently fit the time dependence of the spectra with a global regression algorithm. Unlike time-resolved photoluminescence (PL) measurements and TA measurements in the visible region (400-750 nm), the observed signal in the NIR TA experiment originates directly from absorption of probe photons by the individual charge carriers of the exciton. The absence of ground state and biexcitonic features in this region makes it amenable to global regression analysis; therefore, even though the NIR contains heavily overlapping signals from hole and electron transitions, we could obtain wavelength-resolved analysis of dynamics as a function of the surface chemistry of the QDs, and thereby propose mechanisms for each component of the decay of the first excitonic state. We use global analysis of NIR TA spectra to determine that the band-edge excitonic state decays via a fast hole trapping process (1.6 ps), two surface-mediated electron trapping processes (19 and 274 ps), and a recombination of untrapped electrons (∼5 ns); the dynamics of the recovery of the ground state bleach (in the visible TA spectrum) support these assignments. The combination of NIR TA and global regression is a valuable tool for determining the roles of surface ligands in nonradiative relaxation processes. This information is critical for solar cell, photocatalytic, and biosensor applications, which rely directly on either extraction of both charge carriers from the QDs, or high-yielding PL from the QDs.8 †
Part of the “Michael R. Wasielewski Festschrift”. * Corresponding author. E-mail:
[email protected].
Experimental Advantages of Transient Absorption in the NIR. Measuring the NIR TA spectra of CdSe QDs offers four major advantages over time-resolved PL and visible TA spectroscopy for determining excitonic dynamics: (i) Transient absorptions in the NIR are due exclusively to absorptions of the excited state of the QD because the energies of the probe beam in the NIR are not large enough to induce ground state absorptions across the bandgap. In contrast, excited state signals in the visible TA spectrum are obscured by strong signals from ground state absorption of the probe; consequently, signals in the visible region primarily reflect ground state recovery rather than excited state decay.2,9 (ii) The ability of NIR TA to specifically probe the excited state, from which each charge carrier can be excited independently, allows separation of the excitonic decay pathways due to transitions of the hole from those due to transitions of the electron. In contrast, in the visible region, the probe signal results from the creation of a new exciton via excitation across the bandgap, and therefore is not specific to either carrier. Time-resolved PL spectroscopy also cannot resolve transitions of individual charge carriers since signal acquisition necessarily requires recombination of both charge carriers of the exciton to produce a photon. The time constant for decay of the PL signal is given by a product of exponential decay functions for all contributing decay processes, rather than a sum as in the TA experiment.10 The dynamics obtained from time-resolved PL therefore primarily reflect the dynamics of the charge carrier with the fastest nonradiative decay, while the NIR TA experiment yields time constants for the decay of individual charge carriers. (iii) The excited stateonly character of the NIR TA spectrum makes it ideal for application of global regression analysis of decay dynamics. In the visible region, excited state absorptions (Stark-shifted due to an electric field produced by the pump-induced exciton11) overlap with the ground state bleaches, so extracting quantitative information by regression analysis, global or otherwise, is difficult. (iv) Scattered pump light that can interfere with NIR TA signals can be filtered out with a long-wave-pass filter, whereas, for QDs excited to the first excitonic state in the visible TA experiment, the pump is often at the same wavelength as
10.1021/jp102101f 2010 American Chemical Society Published on Web 05/27/2010
Analysis of NIR TA of Photoexcited Colloidal CdSe QDs the probe and cannot conveniently be filtered out. Similarly, the small Stokes shifts of QDs (a few hundred wavenumbers) require that PL signals of QDs excited to the first excitonic state be monitored at wavelengths within ∼10 nm of the pump wavelength, which is less than the typical bandwidth of a modern tunable ultrafast laser. For both visible TA and PL, then, scattered pump light is difficult to minimize. The main disadvantage of NIR TA is that the signal strength is 1 order of magnitude weaker in the NIR than in the visible region of the TA spectrum. We were, however, able to acquire data in this region with a signal-to-noise ratio of g10, which is adequate for the application of a global regression analysis. Experimental and Data Analysis Strategy. Our NIR TA experiment differs from that of previous authors,10,12-16 who scan or use a single probe wavelength to measure TA in the NIR, because we use a continuum of wavelengths (from 900 to 1300 nm) as a probe, and linear array detection that allows us to collect TA signals at all of these wavelengths simultaneously. Our ability to acquire contiguous data over a 400 nm range during a single collection period warrants the application of a global regression analysis. This analysis involves the mathematical convolution of an instrument response function with a material response functionshere a Gaussian and sum of exponentials, respectivelysas a least-squares approximation that simultaneously describes all the overlapping decay processes contributing to the TA signal. To fit this compound function to the data, we use a global regression algorithm that requires the exponential time constants in the sum to be shared among all time-resolved curves in the data set. The output of the global analysis then reflects how the relative amplitudes of these shared exponential decay functions vary across the spectral region and results in the decomposition of the spectral amplitude into its constituent components. In this study, we are interested only in the decay processes of a single exciton from the first excitonic state to the ground state. Our experimental strategy therefore includes (i) excitation of all samples to the first excitonic state (1Se1S3/2)sthat is, from the HOMO of the QD to the LUMO of the QDsso that hot carrier relaxation from higher energy excited states does not contribute to the observed relaxation processes; (ii) limitation of the pump fluence (pulse energy per area) to produce an expected 10% excited state population, 〈N〉 ) 0.1, which is well below the reported threshold for multiple carrier generation.10,11,17 This limit ensures that there is a maximum of one photoinduced exciton produced per QD; (iii) for each set of samples with added ligand, measurement of a “reference” sample of QDs with their native ligands, handled and purified in an identical way to the samples with added ligand in order to isolate the influence of the added ligand on the decay dynamics. Experimental Methods Synthesis, Purification, and Ground-State Characterization of CdSe QDs. We prepared CdSe QDs with diameter d ) 4.1 nm, using a procedure developed by Qu et al. with minor modifications.7 We injected 1 mL of 1 M trioctylphosphine selenide (TOPSe) precursor, prepared and stored in a drybox under nitrogen atmosphere, into a mixture of cadmium sterate (CdSt2) (MP Biomedicals, 118 mg), trioctylphosphine oxide (TOPO) (90% Sigma Aldrich, 1.94 g), and hexadecylamine (HDA) (90% Sigma Aldrich, 1.94 g) at 320 °C. The CdSe QDs were then grown at 290 °C, after which the reaction mixture was removed from heat and quickly cooled by the addition of 10 mL of chloroform (ACS grade, VWR). After the reaction mixture sat unstirred for two hours, addition of methanol to the
J. Phys. Chem. B, Vol. 114, No. 45, 2010 14515 solution (2:1 methanol/chloroform by volume) caused the QDs to precipitate from solution, and centrifugation at 3500 rpm for five minutes produced a red pellet of QDs at the bottom of the tube. The cloudy liquid containing excess ligands was decanted and discarded. Redispersion of the red pellet in 4 mL of hexane and centrifugation produced a white pellet of excess ligand from the reaction mixture and a red supernate containing the QDs. We decanted the red supernate and treated the solution with 6 mL of methanol, which precipitated the QDs. Centrifugation, decantation and redispersion in 3 mL of chloroform produced an optically clear dispersion of QDs. Finally, we precipitated the QDs with 6 mL of methanol, isolated a red pellet through centrifugation and decantation of the supernate, and redispersed the pellet of QDs in ∼10 mL of CCl4 (g99.5% Sigma Aldrich). The Supporting Information contains the ground state absorption and PL spectra of the QDs with their native ligands in CCl4. The first absorption peak has a maximum at 550 ( 1 nm, where the error originates from batch-to-batch variations. According to a calibration curve we constructed from TEM images of QDs synthesized in our lab,18 this bandgap corresponds to CdSe QDs with diameters of 4.1 nm with a dispersion of 10-15% in diameter. The preparation described above produced QDs with cadmium enriched surfaces (Cd/Se ) 3.25 ( 0.07 by ICP-AES) consistent with our previous study.5,18 The QDs have a PL emission peak maximum at 559 nm with a fwhm of 26 nm. We observe no deep trap emission. Ligand Exchange. We purchased 1-octanethiol (OT), 1,4benzoquinone (BQ) and CCl4 from Aldrich and used them asreceived. We diluted the purified QDs in CCl4 to a concentration of ∼5 µM, as measured from the ground state absorption spectrum,19 and added 1 mL of this solution to each of twelve 25 mL scintillation vials. We then added 10 µL of solutions of OT or 100 µL of solutions of BQ of appropriate concentrations in CCl4 to each of the vials so the final mixtures had a total of 0, 1, 10, 100, 1000 or (70,000 for OT; 2500 for BQ) ligands per QD. We stirred the solutions in the dark for 24 h to allow the ligands and QDs to equilibrate. The UV/vis absorption spectra of the samples before and after exchange were identical except for the sample with 70000 OT:QD, for which the first absorption peak shifted to lower energy by 12 meV.20,21 Transient Absorption Measurements. We split the 2.5 mJ output of a commercial amplified Ti-sapphire laser (Solstice, 1 kHz, 100 fs, Spectra Physics), and guided 95% to an optical parametric amplifier (TOPAS-C, Light Conversion) used to produce the pump wavelength for sample excitation, and 5% to a commercial TA spectrometer (Helios, Ultrafast Systems) for use as the probe. Within the spectrometer, a single filament broadband continuum of wavelengths was generated in a 1.2 cm thick sapphire plate and then passed through a long-wavepass filter to isolate NIR wavelengths above 850 nm. The pump and probe were recombined at the sample, in a 2 mm quartz cuvette. The pump spot size was expanded to greater than twice the size of the probe spot to compensate for any imperfections in translation stage alignment. All samples were prepared with an absorbance of 0.1 ( 0.02 au at the maximum of the first absorption feature, corresponding to the first excitonic state resonant with the pump light, to ensure that both pump and probe light interact with the CdSe QDs uniformly across the sample thickness. The pump light was depolarized to prevent intentional photoselection so that measurements reflect only population dynamics. Incident pump fluence was adjusted to produce an expected excited state population of 10%. The solution was stirred with a small magnetic stir bar assembly to minimize local heating. The transmitted probe signal was passed
14516
J. Phys. Chem. B, Vol. 114, No. 45, 2010
through an 850 nm long-wave-pass filter and collected into an optical fiber. The output of the fiber was dispersed onto an array detector. The output differential absorption spectrum (∆A) was obtained through active background subtraction of the ground state spectrum by chopping the pump at 500 Hz. The instrument response, given by the cross-correlation of the pump and probe, was measured by inducing the optical Kerr effect in a quartz microscope coverslip at the position of the sample, and it was found that signal rises are instrument limited in all cases. Each data set was mathematically corrected for the temporal dispersion of the probe continuum by using the peak maximum of the pump-probe cross correlation data set to determine the relative time offset of each probe wavelength. The Supporting Information contains details of the spot size measurements, excitation calculation, and a continuum spectrum. Separate TA measurements of QD reference samples (no added ligand) from different batches are similar in shape, and produce statistically similar time constants from regression analysis. For each sample, averaged individual dynamic traces (25 scans per data set) overlap in signal strength and shape; their similarity indicates that the laser pump light did not degrade the sample during the experiment. Identical ground state UV/ vis absorption spectra of the samples acquired before and after TA measurements further support this conclusion. To avoid crowding the NIR region of the TA spectrum with absorptions due to vibrational overtones of the solvent, we measure the spectra of QD samples dispersed in CCl4, which has no absorptions in this region. Experimentally we find that it is possible to measure NIR TA dynamics in other solvents, such as chloroform and hexanes, with the only readily apparent difference of added noise to the TA signal in the regions where solvent overtone absorptions are most dense and partially extinguish probe light. The signals from vibrational overtones of the ligands bound to the QD do not complicate the spectrum because these transitions have absorption cross sections on the order of 10-21 cm2, whereas the cross sections of excited state transitions of the QDs, akin to free carrier and solvated electron cross sections, are on the order of 10-17 cm2.16,22,23 Global Regression Analysis. We applied a global regression analysis to determine the temporal evolution of the TA spectra in the range of 900-1300 nm using a commercial data analysis software package (Origin 8, Origin Lab Corporation). The signal-to-noise ratio was too small, 1300 nm. Probe light in this region does not possess enough energy to promote electrons across the optical band gap, so these positive features originate from transitions of either the electron or the hole of the exciton to a more energetic state (energy level diagrams 3 and 4 in Figure 1). The dynamics of these signals are specific to the individual charge carriers of the excited state. From previous TA studies in the NIR, it is known that the feature that peaks at g1300 nm originates from intraband hole transitions,10,13-15 but the 926 nm feature has not been reported previously. We present evidence in the following sections that suggests it is due to intraband electron transitions. Description of the Dynamics of Excitonic Charge Carriers in the NIR Region. Relaxation of the first excitonic state inWolWes four decay pathways. Inspection of the decay curves from the TA spectra of all reference samples (QDs passivated by their native ligands) at wavelengths from 900 nm to 1300 nm shows that the absorptions in this region have at least three visibly distinct exponential decay components. Residual plots corresponding to fits of the data using a sum of three exponentials (convoluted with an instrument response function), however, exhibited unacceptable curvature. Addition of a fourth exponential to the sum resulted in both convergence and a uniform scattering of the residuals across a zero line (see Supporting Information). The global fitting yields relative
Analysis of NIR TA of Photoexcited Colloidal CdSe QDs
J. Phys. Chem. B, Vol. 114, No. 45, 2010 14517
Figure 2. Decomposition of the spectrum of the reference QD sample showing the individual contributions of each exponential component to the spectral envelope shown in the inset to Figure 1. The Supporting Information contains NIR spectra of the QDs at various probe delay times and fits to these spectra.
amplitudes and time constants for each exponential in the sum. Multiplication of the relative exponential amplitudes by the full TA spectral signal envelope at the completion of the rise decomposes the spectrum into its individual components (Figure 2): component C1 with time constant τ1 ) 1.6 ( 0.6 ps (red), which is primarily associated with the feature peaked at ∼1300 nm, and components C2 with τ2 ) 19.4 ( 2.3 ps (green), C3 with τ3 ) 274 ( 9 ps (blue), and C4 with τ4 ) 5.4 ( 1.1 ns (black), which is not completely resolvable within our experimental delay window of 3 ns. The error bars on the time constants are standard deviations of measurements made on four separate samples from two different batches of QDs. This error originates from sample-to-sample variation: for all four time constants and amplitudes, the statistical error of the global fit is
