Letter pubs.acs.org/JPCL
Uncovering Hot Hole Dynamics in CdSe Nanocrystals Cunming Liu,§ Jeffrey J. Peterson,† and Todd D. Krauss*,§,†,# §
Rochester Advanced Materials Program, †Department of Chemistry, and #The Institute of Optics, University of Rochester, Rochester, New York 14627, United States S Supporting Information *
ABSTRACT: Single and multiple exciton relaxation dynamics of CdSe/CdZnS nanocrystal quantum dots (QDs) monitored at the two lowest optical transitions, 1Se−1S3/2 and 1Se−2S3/2, have been examined using ultrafast transient absorption (TA) spectroscopy. For the CdSe/CdZnS QDs studied, the 1Se−1S3/2 and 1Se−2S3/2 transitions are widely separated (∼180 meV) compared to bare CdSe QDs (∼50−100 meV), allowing for clearly distinguishable TA signals attributable to hot hole relaxation. Holes depopulate from the 2S3/2 state with a lifetime of 7 ± 2 ps, which is consistent with the predictions for hole relaxation via a phonon coupling pathway to lower-energy hole states, with possible contributions from hole trapping as well. These results suggest that tuning the surface chemistry of semiconductor QDs is a viable route to measure and possibly control their hot hole relaxation dynamics. SECTION: Physical Processes in Nanomaterials and Nanostructures
D
Here, we report TA data probed at two different transitions 1Se−1S3/2 and 1Se−2S3/2 with various pump intensities using a sample of CdSe/CdZnS QDs in which the manifolds of hole states are widely separated (∼180 meV). The widely spaced hole levels in these QDs have allowed for a successful separation of hole dynamics. Through measurements of the hole relaxation dynamics between the 1Se−1S3/2 and 1Se−2S3/2 transitions, we directly observed hot hole relaxation with a lifetime of 7 ± 2 ps. Further, with significantly intense pump fluences, the relaxation dynamics of biexcitons were also determined (Auger lifetime of 74 ± 2 ps). These results provide new insights into exciton relaxation dynamics of QDs and highlight the importance of QD surface chemistry in controlling hole relaxation dynamics. CdSe/CdZnS QDs in decane were purchased from Invitrogen (CdSe QD 585 ITK organic QDs). Absorption and photoluminescence (PL) spectra of the QDs are shown in Figure 1. The PL peak occurs at 2.12 eV, and the first four transitions in the absorption spectrum occur at 2.18, 2.36, 2.55, and 2.66 eV. A simplified energy level diagram showing the states involved in the four transitions, along with their predicted transition energies, is shown in Figure 1. The predicted transition energies are determined from previously reported experimental determinations of the peak assignments.14 We assign the four peaks as 1Se−1S3/2, 1Se−2S3/2, 1Pe−1P3/2, and 1Se−2S1/2 transitions, respectively. This assignment is confirmed by both low-temperature photoluminescence excitation (PLE) measurements and TA buildup dynamics. We consider
eveloping a detailed understanding of the electron and hole relaxation mechanisms in nanocrystalline materials has been a major focus of semiconductor nanocrystal quantum dot (QD) research during the last 20 years.1−3 Such efforts are not only important for the sake of understanding the fundamental physics of quantum confinement processes but are also critically important for the use of the QDs in technological devices, such as lasers and solar cells.4,5 Specifically, CdSe QDs are the most widely studied material system, and ultrafast transient absorption (TA) spectroscopy has been a useful tool to the develop a detailed picture of their electron and hole relaxation.3,6−8 However, one significant challenge has been extracting the hole dynamics in CdSe QDs from TA signals. In CdSe core QDs, the maximum separation between the two lowest-energy hole states (∼50−100 meV) is comparable to the inhomogeneous line width due to the distribution of QD sizes. Thus, it is often difficult to assign hole contributions to interband TA signals directly.9,10 This problem can be circumvented by performing near-infrared (NIR) intraband TA studies or by employing more complex pump−probe schemes.2,3,10−13 These approaches have found multiple time scale components for hole relaxation, including prominent subpicosecond components and slower 10−30 ps components.2,3,10−13 However, even with these methodologies, it is difficult to know precisely which hole state is probed because the NIR energy (∼600 meV) can match different transitions between multiple hole states.2,10−12 State-specific pump−probe experiments resolve this problem but require probing off resonance of the band edge transition in order to obtain unique transients from which the hole dynamics can be extracted.3,13 © 2014 American Chemical Society
Received: July 23, 2014 Accepted: August 18, 2014 Published: August 18, 2014 3032
dx.doi.org/10.1021/jz5015554 | J. Phys. Chem. Lett. 2014, 5, 3032−3036
The Journal of Physical Chemistry Letters
Letter
coating and long-chain alkyl ligands, although the exact identities of these components are undisclosed.15 The increased separation between the first two peaks in the absorption spectrum is consistent with expectations for quasi-type-II core− shell QDs, such as CdSe/CdS QDs,16 in which the hole wave function is strongly confined to the core, thereby broadening the separation between hole states, while the electron wave function penetrates deeply into the shell.17 Furthermore, literature reports have shown that the peak separation is dependent on both the QD’s surface capping ligands and its crystal structure. 18−21 Although a variety of potential contributions exist, the salient point is that we have clearly identified the second peak as the 1Se−2S3/2 transition. To illuminate the hot hole dynamics and exciton relaxation in this sample, the 1Se−1S3/2 and 1Se−2S3/2 transitions were probed following 400 nm pump excitation at varying pump fluences corresponding to average exciton populations (⟨N0⟩) of 0.30−2.80. The TA signals probed at the 1Se−1S3/2 transition as a function of pump fluence are shown in Figure 2a. The TA signals at low pump fluence are dominated by the single-exciton dynamics. As the pump fluence increases, multiple excitons are created in the QDs, and a fast, biexciton Auger component in the TA signals concomitantly appears.2,22 Because the 1S state is two-fold degenerate,2,23 we use a simple kinetic scheme that considers an initial population of single
Figure 1. Absorption and PL spectra of the CdSe/CdZnS QDs. Red bars are the allowed optical transitions assigned according to ref 14. (Inset) Energy level diagram illustrating the relevant energy of electron/hole states and allowed optical transitions (diagram not drawn to scale).
two probable assignments of the peaks in the absorption spectrum, one in which the second peak is assigned as the 1Se− 2S3/2 transition and one in which it is assigned as the 1Pe−1P3/2 transition. Low-temperature (∼80 K) PLE spectroscopy narrows the absorption peak line widths by reducing both thermal broadening and inhomogeneous broadening from the distribution in QD sizes, clearly resolving multiple peaks in the spectrum and allowing them to be easily identified (see Supporting Information (SI) Figure S1a). The observed PLE transition energies are in excellent agreement with the assignment of the second peak as the 1Se−2S3/2 transition and are inconsistent with its assignment as the 1Pe−1P3/2 transition (see SI Figure S1b). In the case of TA buildup dynamics, one expects that buildup dynamics should be comparable for the 1Se−1S3/2 and 1Se−2S3/2 transitions because they share a common electron state and that both should be slower than that for the 1Pe−1P3/2 transition.2,10 This behavior is precisely observed when the second absorption peak is assigned as the 1Se−2S3/2 transition and the third peak as the 1Pe−1P3/2 transition (see SI Figure S2). These independent observations are self-consistent and provide full confidence in the assignment of the second peak in the absorption spectrum as the 1Se−2S3/2 transition. The ∼180 meV separation between the first two peaks in the absorption spectrum is notable. Because the electron state involved in the 1Se−1S3/2 and 1Se−2S3/2 transitions is the same, the energy separation between the first two absorption peaks is indicative of the separation between 1S3/2 and 2S3/2 hole states. In previous studies using CdSe QDs, this peak separation was ∼50−100 meV, and thus, the 1Se−2S3/2 transition overlaps with the band edge transition 1Se−1S3/2 (see SI Figure S3).2,9 The wide separation between hole states in the current sample enables new access to the hole dynamics via conventional visible TA spectroscopy. Several factors likely contribute to the unusually large 1S3/2− 2S3/2 hole energy level spacing in these QDs. The commercially available Invitrogen CdSe QDs are known to consist of a core− shell structure, likely CdSe/CdZnS, surrounded by a polymer
Figure 2. (a) TA signals of the 1Se−1S3/2 transition at various pump fluences. The red lines are the fits with eq 1. (b) The relative weights A1 and A2 corresponding to the long component τ1 and the short component τ2, respectively, as a function of pump fluence. Uncertainties in the weights are less than 1% and are masked by the symbol size. The red lines are theoretical predictions calculated using eqs 10 and 11 in the SI. 3033
dx.doi.org/10.1021/jz5015554 | J. Phys. Chem. Lett. 2014, 5, 3032−3036
The Journal of Physical Chemistry Letters
Letter
exctions and biexcitons, as well as the decay of biexcitons to single excitons, to describe the 1Se−1S3/2 TA data. Rate equations and further details of the model are provided in the SI. The 1Se−1S3/2 TA data are fit to the equation ⎛ τ ⎞ ΔT (t )1S = ⎜2BX 0 − BX 0 1 ⎟e−t / τ2 T0 τ1 − τ2 ⎠ ⎝ ⎛ τ ⎞ + ⎜X 0 + BX 0 1 ⎟e−t / τ1 τ1 − τ2 ⎠ ⎝
(1)
where X0 and BX0 are the initial populations of single excitons and biexcitons, respectively, and τ1and τ2 are the lifetimes of single and biexciton species. The data are globally fit, with τ1 and τ2 shared and leading coefficients X0 and BX0 floated among all transients. The TA signals in Figure 2a are described very well by eq 1. A long component τ1 equal to 0.9 ± 0.1 ns is found, consistent with nonradiative relaxation of electron to surface traps.2,23 The short component τ2 is equal to 74 ± 2 ps. A large range (20−100 ps) of Auger lifetimes has been reported for CdSe QDs with similar absorption peak energies, and the current value is consistent with these measurements.22−25 The relative weights of these two components as a function of pump fluence are well described by the expected Poisson populations of single excitons and biexcitons (Figure 2b), further supporting the conclusion that the fast component is responsible for biexciton annihilation. Details of the data fitting to Poisson populations are described in the SI. It is notable that the theoretical fits in Figure 2b have no adjustable parameters and still provide good fits to the experimental data. TA signals probed at the 1Se−2S3/2 transition are shown in Figure 3a. In previous studies, the 1Se−1S3/2 and 1Se−2S3/2 TA dynamics were indistinguishable because the 1Se−1S3/2 and 1Se−2S3/2 transitions share an electron state (1Se) and the hole states are closely spaced.9,10 However, here a new fast component in the first ∼15 ps is observed that is clearly not present in the 1Se−1S3/2 transients. The fast component is apparent in the lowest fluence data (Figure 3b), where any multiexciton effects are negligible. In order to characterize the prominent fast component in the 1Se−2S3/2 data (Figure 3a), the transient signals are fit to the following equation ΔT ΔT (t )2S = Xh e−t / τh + C (t )1S T0 T0
Figure 3. (a) TA transients of the 1Se−2S3/2 transition at various pump fluences. The red lines are the fits using eq 2. (b) Comparison of TA signals from the 1Se−2S3/2 and 1Se−1S3/2 transitions under ⟨N0⟩ = 0.30, after being normalized at long decay times. The red lines are the fits using eq 2. (Inset) The hot hole relaxation extracted between the 1Se−2S3/2 and 1Se−1S3/2 transitions (the red line is a single-exponential fit). (c) The relative weights of the prefactors in the exciton relaxation dynamics of the 1Se−2S3/2 transition as a function of pump fluence. A1 and A2 correspond to the long component τ1 and short component τ2 for electron relaxation of the 1Se−1S3/2 transition, respectively. Ah corresponds to the hole component τh of the 1Se− 2S3/2 transition. Uncertainties in the weights are less than 1% and are masked by the symbol size. The red lines are fits by eqs 13−15 in the SI.
(2)
where Xh and C are constants related to the population of holes and the ratio of oscillator strengths between 1Se−2S3/2 and 1Se−1S3/2 transitions, respectively, and τh is the lifetime for the hole relaxation component. τh is fit globally, and the lifetimes of the Auger recombination and electron trapping components, as well as the relative ratio of their amplitudes, are fixed to the values determined from the 1Se−1S3/2 TA analysis. Using this methodology, we find that the fast component for the hole relaxation τh is equal to 7 ± 2 ps. Alternative fitting procedures (e.g., triexponential fitting with floated amplitudes or subtractive approaches) produce slightly greater hole relaxation lifetimes, in the range of 10−15 ps. The corresponding hole cooling rates are on the order of 0.01−0.02 eV/ps, depending on whether or not the relaxation occurs to the 1P3/2 (ΔE = 130 meV) or 1S3/2 (ΔE = 180 meV) hole state and on the exact magnitude of the hole relaxation lifetime. The hole relaxation lifetime that we have determined is somewhat different than that measured in previous studies.
Previous studies have measured multiple time scale components for hole relaxation, including prominent subpicosecond components and slower 10−30 ps long components.2,10−13,26 The corresponding hole cooling rates, 0.20−1.0 eV/ps, are generally reported to be much faster than what we have observed in these QDs.2,13,26 We note that the probe delay time 3034
dx.doi.org/10.1021/jz5015554 | J. Phys. Chem. Lett. 2014, 5, 3032−3036
The Journal of Physical Chemistry Letters
Letter
state-specific population filling pathways would further complicate the analysis.9 State-specific pump−probe experiments would provide further insight regarding these mechanisms.6−9 The relative weights of three components in the 1Se−2S3/2 transients as a function of pump fluence are well described by the same Poisson model used to describe the 1Se−1S3/2 TA data. The model (described in the SI) assumes an initial hole population in the 2S3/2 state arising from a Poisson population of photoexcited holes and includes only one adjustable parameter C′ due to the different oscillator strengths between 1Se−2S3/2 and 1Se−1S3/2 transitions. Because Auger recombination occurs on a much slower time scale (∼74 ps), it is not expected that the fast hot hole relaxation (∼7 ps) would be influenced by multiexciton effects under a relatively high pump fluence. Consequently, the relative weight of the hot hole relaxation component from the 2S state is nearly constant (∼26%) under different pump fluences, agreeing well with studies of hole relaxation measured via the intraband TA in the NIR.11 In summary, we have examined the exciton relaxation dynamics in CdSe/CdZnS core−shell QDs with the visible TA spectroscopy. The 1S3/2 and 2S3/2 hole states are widely separated in these QDs, enabling the direct observation of hot hole relaxation dynamics. We found that the 2S3/2 state depopulates with a lifetime of 7 ± 2 ps, which is consistent with a relaxation pathway involving both intraband relaxation via multiphonon coupling and hole trapping to intrinsic defect states. This work introduces CdSe QDs with widely separated hole states as a useful system to study relaxation dynamics. Such control could be advantageous when optimizing QDs for use in optoelectronic technologies that benefit from longerlived charge carriers (e.g., photovoltaics). Future work employing state-specific pump−probe schemes, as well as investigating the QD size dependence of the observed TA signals, would be useful to help further elucidate details of the relaxation channels of these systems.
window used here is ∼10× longer than those typically used, which may contribute to these differences.10,11 There are three main pathways by which a hot hole may relax have been proposed, ligand-induced nonadiabatic transitions,3,13 hole−longitudinal optical (LO) phonon coupling,27−29 and relaxation into intrinsic trapping states.10,30 The total hole relaxation rate is governed by the sum of each of these rates, and we consider their possible contribution in the current QDs below. Ligand-induced nonadiabatic transitions are important in systems with strong overlap between the hole and surface ligand wave functions.13 They enable hole relaxation on 100 fs time scales, corresponding to hole cooling rates on the order of 0.30 eV/ps.13 Because the current commercial CdSe QDs contain a thick semiconductor shell layer of CdZnS that spatially separates the hole and surface ligands, we expect that such ligand-induced interactions are negligible. In the phonon coupling relaxation pathway, a hot hole can relax via emitting multiple phonons by the Fröhlich interaction.27−29 Rates (1/τr) of this process are calculated using a golden rule expression, which include factors such as the exciton−phonon coupling strength and energy state separation and are summed over multitude phonon modes and energy states.31 The expressions are numerically evaluated and are found to reach a maximum value (when the phonon energy matches the energy separation) and undergo exponential-like decay as the energy separation increases. The lifetime of hot hole relaxation, therefore, is commonly estimated as τr = ω−1 LO exp(ΔE/kBT), where τr is the hot carrier cooling lifetime via the multiple LO phonon relaxation, ωLO is the frequency of the LO phonon (∼210 cm−1 in CdSe),4,13 ΔE is the gap between the hole states, kB is the Boltzmann constant, and T is the temperature. The predicted lifetime of hole relaxation from the 2S3/2 hole state via phonon coupling is 25−180 ps, depending on whether the relaxation occurs to the 1P3/2 hole state or directly to the 1S3/2 hole state. The corresponding hole cooling rates are on the order of 0.01 eV/ps or less. Finally, no general theoretical framework exists to predict hole relaxation rates to intrinsic trapping states; however, these processes are empirically observed to occur on picosecond time scales.2,10 Considering these relaxation mechanisms, the ∼7 ps lifetime for the 2S3/2 hole state is consistent with a pathway dominated by phonon-coupled relaxation of photoexcited holes to the band edge, with possible contributions from hole trapping as well. Although the observed 7 ps lifetime is generally smaller than that predicted by the phonon coupling model (i.e., a 7 ps relaxation time corresponds to a 1S3/2−2S3/2 hole energy spacing of ∼100 meV), a number of effects could further reduce the predicted lifetime. Strong Coulomb interaction of electron−hole pairs, crystal field effects, and shape anisotropy have been reported to split the hole states by an additional 5− 30 meV.2,30 Such effects are sufficiently large to reduce the predicted hole relaxation lifetime to the experimentally observed value. We further note that there is no observed ∼7 ps rise in the 1Se−1S3/2 TA signal, as would be nominally expected if this decay time corresponds to the dominant relaxation channel for hot holes to the 1S3/2 hole state. This absence may suggest a larger contribution from hole relaxation to intrinsic trapping states compared to band edge relaxation. However, the absence of a ∼7 ps rise at the 1Se−1S3/2 transition may also indicate that the simple model for hole electronic states is inadequate. For example, fast hole trapping (
