Hole Surface Trapping Dynamics Directly Monitored by Electron Spin

Dec 1, 2014 - The spin enhancement transients as a function of the pump-orientation delay reveal a fast and a slow hole trapping process with respecti...
0 downloads 3 Views 960KB Size
Letter pubs.acs.org/JPCL

Hole Surface Trapping Dynamics Directly Monitored by Electron Spin Manipulation in CdS Nanocrystals Xiao Li,† Donghai Feng,*,† Haifang Tong,† Tianqing Jia,† Li Deng,† Zhenrong Sun,† and Zhizhan Xu*,‡ †

State Key Laboratory of Precision Spectroscopy, East China Normal University, Shanghai 200062, China State Key Laboratory of High Field Laser Physics, Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai 201800, China



S Supporting Information *

ABSTRACT: A new detection technique, pump-spin orientation-probe ultrafast spectroscopy, is developed to study the hole trapping dynamics in colloidal CdS nanocrystals. The hole surface trapping process spatially separates the electron− hole pairs excited by the pump pulse, leaves the core negatively charged, and thus enhances the electron spin signal generated by the orientation pulse. The spin enhancement transients as a function of the pump-orientation delay reveal a fast and a slow hole trapping process with respective time constants of sub-10 ps and sub-100 ps, orders of magnitude faster than that of carrier recombination. The power dependence of hole trapping dynamics elucidates the saturation process and relative number of traps, and suggests that there are three subpopulations of nanoparticles related to hole surface trapping, one with the fast trapping pathway only, another with the slow trapping pathway only, and the third with both pathways together. SECTION: Physical Processes in Nanomaterials and Nanostructures

C

measurements in order to neglect multicarrier Auger process, the extraction of exclusive electron or hole trapping process is still difficult. In this Letter, we detect the hole surface trapping dynamics in CdS colloidal NCs by monitoring the spin signals of core electrons. The detection principle is as follows. Surface trapping of the hole leaves the core negatively charged, and results in much stronger electron spin coherence amplitude than the case with both electron and hole being in the NC core states.22 Monitoring the transient signal enhancement of electron spin can directly obtain the hole trapping dynamics, while both radiative and nonradiative electron−hole recombination processes will not contribute to the enhanced electron spin signal. In comparison with the often-used time-resolved PL or absorption measurements, the present technique equivalently detects the population buildup process of hole trapped states rather than the depopulation of core states. Weak excitation conditions are not necessary, thus it is feasible to measure the power dependence of hole trapping dynamics and study the saturation process of trap states with enough excitation power. Measurements were performed using a three-pulse, pumporientation-probe (POP)-based time-resolved Faraday rotation (TRFR) spectroscopic technique. Figure 1a shows the configuration of the three-pulse pump−probe setup. A linearly polarized pump pulse generates electron−hole pairs in the

harge carrier trapping at the surface of semiconductor nanocrystals (NCs) has attracted much attention due to its great influence on the optical and electrical properties of materials.1−5 Transition from bulk to nanoparticles leads to an enhanced dominance of surface atoms, where the electron or hole can be readily trapped at those unsaturated surface bonds. Carrier surface trapping generally results in a decreased device performance, and becomes a limiting factor for practical applications of NCs. Great efforts have been devoted to exploring and understanding the role of carrier surface trapping on various physical processes such as carrier relaxation and recombination dynamics,6−10 photoluminescence (PL) quantum yield,11 PL blinking,12,13 optical gain,14,15 charge transport/ transfer efficiency,16 and carrier multiplication as well.17,18 Yet many fundamental issues related to surface trapping, especially in ultrafast dynamics, still remain poorly understood. Carrier trapping dynamics is usually probed in an indirect way using time-resolved PL and absorption spectroscopies,3,9,19,20 which are based on monitoring the depopulation process of the NC core states. Transition from intrinsic core states to surface trap states will cause a variation of the NC absorption and PL intensities. Nevertheless, in addition to carrier trapping, radiative and Auger nonradiative recombination processes of excitons also contribute to the depopulation dynamics of band-edge excitonic states. Each decay time constant of electrons or holes, observed from time-resolved PL and absorption measurements, is an average lifetime that includes all the radiative and nonradiative contributions.9,21 Although weak excitation conditions can be applied in the © XXXX American Chemical Society

Received: November 4, 2014 Accepted: December 1, 2014

4310

dx.doi.org/10.1021/jz502340w | J. Phys. Chem. Lett. 2014, 5, 4310−4316

The Journal of Physical Chemistry Letters

Letter

σ− orientation pulses, respectively (see the Supporting Information for more details related to POP measurements). We first discuss the case of 471 nm pump for simplicity. Upon the pump pulse excitation, band-edge electrons and holes are initially generated in the NC core. Afterward, little by little, they can be directly annihilated via radiative or nonradiative recombination, or transfer to the NC surface via electron trapping, hole trapping, or both. Following the pump pulse, a circularly polarized orientation pulse is introduced and generates the electron spin signal (hole spin dephases very fast, and is beyond detectable24). Figure 1b shows TRFR signals for different pump-orientation pulse delays Δt. When the orientation and pump pulses coincide in time, the spin signal is small, being equal to that without the pump being on. Increasing the delay time, the spin coherence amplitudes increase, while the oscillation phase and frequency keep the same. The enhanced spin amplitude with the delay is directly related to the temporal evolution of the pump-excited electrons and/or holes. Recombination induced electron−hole annihilation can be excluded for the explanation, as the initial point (Δt = 0) and final point (Δt = ∞, equivalent to the case without the pump on) of this process cannot have an equal spin signal. Directly generating the spin polarization of surface-trapped electrons by orientation pulses is impossible due to negligible transition dipole moments. During the whole time scale of pump-orientation and orientation-probe delays shown in Figure 1b, no change of the electron g factor (manifested by an invariant oscillation frequency peak; see Fourier transform spectra of the Faraday rotation signals in the Supporting Information, Figure S6) further indicates that electron surface trapping is not responsible for the spin enhancement with pump-orientation delay times. The only explanation is hole surface trapping. Surface trapping of the hole leaves the core negatively charged.17 Following the hole surface trapping, more and more net electrons are resident in the NC core. Electron spin polarization is then generated by the orientation pulse via polarization-selective electron-to-trion excitation. The trion singlet ground state consists of two electrons with opposite spin orientations and a single hole with a spin of ±3/2. According to optical selection rules, spin-up (spin-down) resident electrons will be excited to spin-up (spin-down) trion states where the total spin is defined by the hole spin, leaving a net part of spindown (spin-up) polarized resident electrons. The spinpolarized electrons precess in an external transverse magnetic field, inducing periodical oscillations of Faraday rotation signals. The small electron spin signal without the pump on is due to scarce intrinsic electrons resident in the NC. At zero pumporientation delay, the pump-induced electrons are bound with holes. In this case, electron−hole exchange interaction induces fast spin flip at room temperature on a time scale as little as a few hundred femtoseconds.25,26 The fact that only the spin signal from resident electrons but not excitons has been detected22−24 when without the pre-excitation of the pump pulse confirms a very short exciton spin relaxation time, which is not resolvable in the present experimental time window. Figure 2 shows the spin coherence amplitude as a function of the pump-orientation pulse delay. There are three stages shown in the curve, which correspond different time sequences between the pump, orientation, and probe pulses. In stage 1, the pump pulse is introduced after both the orientation and probe pulses, thus the pump has no effect on the probed signal, which is the same as the case without the pump on. In stage 2, the pump pulse is after the orientation but before the probe

Figure 1. (a) Scheme of the three-pulse time-resolved Faraday rotation (TRFR) measurements in CdS nanocrystals with the pump/ orientation/probe configuration. (b) TRFR signals for different pumporientation pulse delays. Plots are offset for clarity. The pump, orientation, and probe laser wavelengths are set at 471 nm, Ppump = 1.0 W/cm2 and Porientation = 1.5 W/cm2.

nanoparticle, and a circularly polarized orientation pulse is used to polarize the spin of pump-excited electrons. The subsequent electron spin evolution is accessed by monitoring the rotation angle of the polarization plane of a linearly polarized probe pulse. TRFR signals were recorded using an optical polarization bridge combined with lock-in detection.23,24 All measurements were performed at room temperature and in a transverse external magnetic field B of 0.5 T (Voigt geometry, where B is perpendicular to the spin polarization direction). Commercially available colloidal CdS NCs in a toluene solution (Lumidot CdS 480, Sigma-Aldrich Corporation) were used for the measurements without further sample processing. The sample was surface stabilized with oleic acid coating, with a particle diameter size ∼5.6 nm. The orientation/probe light wavelengths were fixed at 471 nm for all measurements, which were located near the absorption edge of the sample (see the Supporting Information, Figure S3). And the pump light wavelength was either 471 or 400 nm, which generates cold or hot excitons, respectively. Femtosecond visible laser pulses were generated by parametric nonlinear processes and frequency mixing from a Ti:sapphire regenerative amplifier (800 nm wavelength, 50 fs pulse duration, and 1 kHz repetition rate). Additionally, 400 nm laser pulses were generated through frequency doubling of the amplifier output. The pump, orientation and probe laser beams were focused to the same spot on the sample, and had a decreased focus diameter of ∼700, 350, and 200 μm, respectively, for the sake of better detection homogeneity. Only the orientation beam was chopped for lock-in detection by a mechanical optical chopper. In order to get rid of any nonprecessing background of spin coherence signals, we performed the measurements twice by changing the orientation pulse with light polarization of σ+ and σ− in sequence.22 The final Faraday rotation data were obtained by making the difference of two signals created by the σ+ and 4311

dx.doi.org/10.1021/jz502340w | J. Phys. Chem. Lett. 2014, 5, 4310−4316

The Journal of Physical Chemistry Letters

Letter

Figure 3. Faraday rotation signals as a function of the orientationprobe delay time (the pump-orientation pulse delay is fixed to 10 ps). Red curve: exponentially decaying cosinusoidal fit. The laser parameters are the same as for Figure 1.

Figure 2. Electron spin signal as a function of pump-orientation pulse delays. The orientation-probe delay is fixed at the time point corresponding to the first oscillation peak in Figure 1. The experimental laser parameters are the same as for Figure 1. Black and red curves are experimental data and biexponential fit, respectively. Inset: The time sequence for the pump, orientation, and probe pulses in stage 1, 2, and 3 as denoted for experimental curves.

the orientation-probe delay time with Δt = 10 ps. The timedependent oscillation amplitude θF(t) can be described as ⎛ t ⎞ θF(t ) = θF(0) cos(ωLt ) exp⎜ − ⎟ ⎝ T2* ⎠

pulse. In this case, the pump pulse reduces the orientationinduced spin signal, as it re-excites the spin-polarized electron to trion states.22 In stage 3, the pump goes before the orientation and probe pulses. As discussed above, this case detects the hole surface trapping process. The induced enhancement transients of spin amplitude S(t) can be well fitted by a biexponential function, S(t ) = S(∞) − A1 exp( −t /τ1) − A 2 exp( −t /τ2)

(2)

where θF(0) is the initial Faraday rotation amplitude following the excitation of the orientation pulse, ωL corresponds to the oscillation frequency (i.e., Larmor precession frequency of electron spin along a transverse magnetic field), and T2* is the spin dephasing time of an ensemble system. The electron spin coherence dynamics in Figure 3 can be well fitted by a single exponential decay function with a well-defined oscillation frequency, yielding T2* ∼ 1.1 ns. The decay rate will be a summation due to all the dephasing mechanisms, including electron−hole recombination, electron−nuclear hyperfine interation,23,27,28 and inhomogeneous dephasing.29,30 As one of the dephasing mechanisms, the recombination time of core electron−hole pairs has to be at least above 1.1 ns, which is much longer than both of the hole surface trapping times τ1 and τ2. In order to show the difference between the technique based on POP configurations and conventional time-resolved absorption (TRA) pump−probe spectroscopy, we performed POP and TRA measurements in parallel to investigate the hole trapping process for both initially cold (471 nm pump) and hot hole states (400 nm pump). Figure 4a shows the hole trapping dynamics monitored by the POP technique. Figure 4b shows TRA measurement results, with the same pump/probe laser conditions as for Figure 4a (the orientation pulse is blocked off in TRA measurements). Table 1 shows time constants and relative amplitudes extracted from the measurements shown in Figure 4. For 400 nm pump light excitation in the POP measurement, two time constants of 2.5 and 46.4 ps are obtained, showing slightly faster trapping rate compared with cold holes.31 In comparison, TRA measurements only show 76.8 and 499 ps decay components in the time window of 1.8 ns. The 76.8 ps component is attributed to Auger recombination process, as verified by pump-power dependent measurements shown in Figure 4c. When reducing the pump power to 6 mW/cm2, no clear decay can be detected in the time window of 300 ps in Figure 4c. The contribution of the Auger process increases with increasing the pump power. The

(1)

where A1,2 and τ1,2 are the corresponding increase amplitudes and time constants, respectively. The biexponential fit gives τ1 = 7.0 ps and τ2 = 92.3 ps, related to two hole trapping processes. Hole trapping with two different time constants has also been found in CdSe NCs.4,9 Knowles et al.9 applied a combination technique of transient absorption and time-resolved PL spectroscopies, and found two hole trapping processes with respective time constants of a few and tens of picoseconds, which were assigned to different subpopulations of CdSe NCs. Also, in a recent calculation using the atomistic semiempirical pseudopotential approach, Gómez-Campos and Califano4 identified two kinds of hole trap distributions in core-only CdSe NCs, which is associated with single or double Se dangling bonds. Due to different trap efficiencies, their calculated trapping time could range from one picosecond to nanoseconds or more. With a given orientation excitation power, the increase of electron spin amplitude is proportional to the net negative charge populations in the NC core, and thus to the surface trapped hole populations. Therefore, the two amplitudes A1 and A2 extracted from the biexponential fit in Figure 2 reflect the hole populations trapped in two different surface traps. When Δt = 10 ps, the increase of the electron spin amplitude is around half of the total amplitude (A1 + A2), thus there are around 50% holes in the core and 50% holes trapped at the surface. Due to the fact that the enhanced spin signal is from the photoexcited core electrons, the recombination of core electrons with both core and surface holes will contribute to the decay of electron spin after the orientation pulse. Figure 3 shows the time-resolved Faraday rotation signal as a function of 4312

dx.doi.org/10.1021/jz502340w | J. Phys. Chem. Lett. 2014, 5, 4310−4316

The Journal of Physical Chemistry Letters

Letter

that TRA with pump−probe wavelengths degenerate at 471 nm can now detect hole trapping contribution. The reason is as follows: 471 nm is subresonant to the band edge exciton (which corresponds to ∼458 nm with a full width at halfmaximum (fwhm) bandwidth of ∼20 nm) and the probe light has a ∼12 nm fwhm bandwidth, thus the detected signal comes from the sum of 471 nm pump/B1 probe and 471 nm pump/ A1 probe, where B1 and A1 denote the first bleaching and absorptive peaks in transient absorption spectra, respectively.33 471 nm pump/B1 probe is only sensitive to electron signals, and 471 nm pump/A1 probe sensitive to exciton charge distribution which resolves both electron and hole trapping dynamics.34 Note that the Auger recombination process is detected only in TRA but not in POP measurements, thus resulting in the data difference of time constants and relative amplitudes for sub-10 ps and sub-100 ps components between TRA and POP measurements under 471 nm pump, e.g., slightly larger time constants and increased ratio of fast to slow components in the POP measurement. As an analogy, the transients for 400 nm pump also contain A1 probe signals. However, it fails to detect hole trapping dynamics. The origin is that 400 nm pump induces strong biexciton interactions, as investigated by Kambhampati and coworkers using state selective pump/probe measurements.34,35 For the 400 nm pump/A1 probe, the signal in the initial point of the decay dynamics is absorptive due to biexciton interactions, and the evolving distribution of hot excitonic states partially cancels the hole-trapping induced absorption evolution, thus making the fast component in the dynamics level off.35 In comparison, the 471 nm pump only excites cold excitons, and the initial decay point is of an attenuated bleach character due to a small biexciton binding energy, thus no influence to the hole trapping dynamics. The pump power dependences of the hole surface trapping dynamics for 471 nm pump in the POP measurements are shown in Figure 5a−c. Generally, the hole trapping process with different pump power shows a similar kinetic behavior, where all measured dynamic curves could be well fitted by biexponential functions.36 The change of excitation power gives no evident influence on both of the hole trapping time constants τ1 and τ2, as shown in Figure 5b. The increase amplitudes of the electron spin due to two different hole trapping processes, i.e., A1 and A2 which are extracted from biexponential fits, are shown in Figure 5c as a function of the pump power. On increasing the pump power, there are more photoexcited electron−hole pairs. After the hole trapping, more net negative charges leave in the NC core, yielding stronger electron spin coherence under the orientation pulse excitation. Therefore, both A1 and A2 will increase with the pump power, until the hole trapping sites are fully filled or the absorption of the pump light is saturated. The fast amplitude A1 levels off above the pump power of ∼1.0 W/cm2, significantly less than the value of ∼4.0 W/cm2 for the slow amplitude A2. It is clearly that the leveling-off of the fast amplitude A1 results from the full-filling of hole trapping sites. The leveling-off of the slow amplitude A2 is also mainly due to the fact that the hole trapping sites for the slow process are fully filled, as can be evinced by the fact that the absorption of the pump light is still unsaturated as shown in Figure 5d. The two hole trapping processes shown in Figure 5a−c may exist in the same nanoparticle together or separately in different subpopulations of the NC sample.37 If all NCs have the fast trapping pathway, the trapping transients will only be

Figure 4. (a) Hole trapping dynamics revealed by pump-orientationprobe measurements. The pump light wavelength is either 471 or 400 nm, and all the orientation and probe light wavelengths are fixed at 471 nm. Pump power for 400 and 471 nm excitation, P400 = 0.2 W/ cm2 and P471 = 0.4 W/cm2, respectively. (b) Time resolved absorption (TRA) measurements, with the essentially identical experimental conditions as for panel a. (c) Normalized TRA signal under 400 nm pump for different pump power.

Table 1. Comparison of Time Constants and Relative Amplitudes between the POP and TRA Measurements Shown in Figure 4 feature POP TRA POP TRA

(400 (400 (471 (471

nm) nm) nm) nm)

τ1 (ps)

τ2 (ps)

2.5

46.4 76.8 82.2 64.2

6.1 4.0

τ3 (ps)

A1

A2

0.36

0.64 0.58 0.37 0.36

499 679

0.63 0.30

A3 0.42 0.34

time scale of ∼100 ps for the Auger process agrees well with recent measurements performed by Kobayashi et al. using transient absorption spectroscopy.32 From the comparison between POP and TRA measurements, one can conclude that TRA measurements under 400 nm pump and 471 nm probe cannot resolve hole trapping processes. This is reasonable, as the state-filling-induced bleaching of interband optical transitions is dominated by the electron population.10 On the other hand, two time constants of 6.1 and 82.2 ps are obtained for 471 nm pump light excitation using the POP technique, while three time constants of 4.0, 64.2, and 679 ps are resolved in TRA measurements. The interesting point is 4313

dx.doi.org/10.1021/jz502340w | J. Phys. Chem. Lett. 2014, 5, 4310−4316

The Journal of Physical Chemistry Letters

Letter

Figure 5. Pump power dependence of hole surface trapping dynamics under 471 nm degenerate pump-orientation-probe configurations for (a) hole trapping transients, (b) hole trapping time constants, and (c) hole trapping induced electron spin enhancement amplitude. I0 = 1.04 W/cm2. In panels a−c, the power density of the orientation pulses is fixed at 1.5 W/cm2. (d) The pump light absorption intensity (proportional to the bleaching amplitude) as a function of the pump light power, which was measured by normal two-pulse pump−probe spectroscopy (with blocking off the orientation light).

1.5 or P2/(P1 + P2 + P1,2) ≈ 0.4. The excitation power for the full-filling point is ∼1.0 W/cm2 and ∼4.0 W/cm2 for the fast and slow trapping processes, respectively. It corresponds to an absorption ratio of 1:3 as shown in Figure 5d. The ratio of N2 and N1 should then be equal to a value between 2 and 3, taking account of the fact that the subpopulation of NCs with τ1 and τ2 processes together needs a larger saturation power than the one with only a τ2 process. For a qualitative analysis, we take N2/N1 ≈ 2.5.38 Note that, multitraps in the fast or slow trapping process may be not necessary in the same energy level. It includes all the trap levels involved together in the respective saturation process of the fast or slow trapping. For example, especially in the slow trapping process, an initial surfacetrapped hole could quickly leak to another adjacent deeper state. Nevertheless, the saturation means all the involved trapping sites are fully filled, which have to be counted together as the trap numbers. After arriving at the saturation points, the ratio of the final amplitude A1 and A2 (∼3.5 as can be seen from Figure 5c) will be dependent on both of the NC particle’s and trap numbers, thus there is (N2P2 + N2P1,2)/(N1P1 + N1P1,2) ≈ 3.5. From above, the relation of P1: P2: P1,2 ≈ 1.6:4:4.4 can be obtained. Therefore, there are three subpopulations of nanoparticles related to hole surface trapping in the present sample, two of which have respective trapping time of τ1 and τ2, while the third has both trapping processes together. In summary, we applied a novel pump-orientation-probe ultrafast spectroscopy to elucidate the hole trapping dynamics in colloidal CdS NCs. Under the framework of the three-pulse technique, the pump pulse excites electron−hole pairs in the

characterized by a fast process for a weak excitation power, which is below the trap full-filling point. Therefore, the distinct existence of the slow process at weak excitations means that there must be one subpopulation of NCs in which only a slow trapping pathway (no fast one) exists. The observed transients in Figure 5a at weak excitations are assigned to the superposition of dynamics from different subpopulations of NCs, some with time constant τ1 and some with τ2 due to the heterogeneity of NC surface states. Nevertheless, at weak excitations, we cannot exclude the possibility of that part of the particles have both fast and slow processes together, as only one observable process with a time constant of τ1τ2/ (τ1 + τ2) ≈ τ1 is detected in this case. One needs to compare both low and high excitation conditions and analyze the power dependence in order to confirm its existence, owing to the fact that increasing the excitation power above the saturation point of the fast process can lead to the emergence of the slow trapping. Now supposing there exist three subpopulations of NCs, the first one with time constant τ1 and particle number P1, the second one with time constant τ2 and particle number P2, and the third one with two time constants τ1 and τ2 together in one particle for which the particle number is P1,2 (see the Supporting Information for the illustration of the decay pathways of band-edge hole states in CdS NCs, Figure S7). The trap number per particle in the fast and slow processes is assumed as N1 and N2, respectively. Taking account of the fact that the hole recombination time is much longer than both τ1 and τ2, and the ratio between the fast and slow amplitudes at a weak pump level A1: A2 is around 1.5, there is (P1 + P1,2)/P2 ≈ 4314

dx.doi.org/10.1021/jz502340w | J. Phys. Chem. Lett. 2014, 5, 4310−4316

The Journal of Physical Chemistry Letters

Letter

(7) Logunov, S.; Green, T.; Marguet, S.; El-Sayed, M. A. Interfacial Carriers Dynamics of CdS Nanoparticles. J. Phys. Chem. A 1998, 102, 5652−5658. (8) Knowles, K. E.; Frederick, M. T.; Tice, D. B.; Morris-Cohen, A. J.; Weiss, E. A. Colloidal Quantum Dots: Think Outside the (Particlein-a-)Box. J. Phys. Chem. Lett. 2012, 3, 18−26. (9) Knowles, K. E.; McArthur, E. A.; Weiss, E. A. A Multi-Timescale Map of Radiative and Nonradiative Decay Pathways for Excitons in CdSe Quantum Dots. ACS Nano 2011, 5, 2026−2035. (10) Klimov, V. I.; Schwarz, Ch. J.; McBranch, D. W. Ultrafast Dynamics of Inter- and Intraband Transitions in Semiconductor Nanocrystals: Implications for Quantum-Dot Lasers. Phys. Rev. B 1999, 60, R2177−R2180. (11) Norberg, N. S.; Gamelin, D. R. Influence of Surface Modification on the Luminescence of Colloidal ZnO Nanocrystals. J. Phys. Chem. B 2005, 109, 20810−20816. (12) Galland, C.; Ghosh, Y.; Steinbrü c k, A.; Sykora, M.; Hollingsworth, J. A.; Klimov, V. I.; Htoon, H. Two Types of Luminescence Blinking Revealed by Spectroelectrochemistry of Single Quantum Dots. Nature 2011, 479, 203−208. (13) Cordones, A. A.; Bixby, T. J.; Leone, S. R. Direct Measurement of Off-State Trapping Rate Fluctuations in Single Quantum Dot Fluorescence. Nano Lett. 2011, 11, 3366−3369. (14) Klimov, V. I.; Mikhailovsky, A. A.; Xu, S.; Malko, A.; Hollingsworth, J. A.; Leatherdale, C. A.; Eisler, H.-J.; Bawendi, M. G. Optical Gain and Stimulated Emission in Nanocrystal Quantum Dots. Science 2000, 290, 314−317. (15) Cooney, R. R.; Sewall, S. L.; Sagar, D. M.; Kambhampati, P. State-Resolved Manipulations of Optical Gain in Semiconductor Quantum Dots: Size Universality, Gain Tailoring, and Surface Effects. J. Chem. Phys. 2009, 131, 164706. (16) Bakulin, A. A.; Neutzner, S.; Bakker, H. J.; Ottaviani, L.; Barakel, D.; Chen, Z. Charge Trapping Dynamics in PbS Colloidal Quantum Dot Photovoltaic Devices. ACS Nano 2013, 7, 8771−8779. (17) Califano, M. Photoinduced Surface Trapping and the Observed Carrier Multiplication Yields in Static CdSe Nanocrystal Samples. ACS Nano 2011, 5, 3614−3621. (18) Tyagi, P.; Kambhampati, P. False Multiple Exciton Recombination and Multiple Exciton Generation Signals in Semiconductor Quantum Dots Arise from Surface Charge Trapping. J. Chem. Phys. 2011, 134, 094706. (19) Jones, M.; Lo, S. S.; Scholes, G. D. Signatures of Exciton Dynamics and Carrier Trapping in the Time-Resolved Photoluminescence of Colloidal CdSe Nanocrystals. J. Phys. Chem. C 2009, 113, 18632−18642. (20) McArthur, E. A.; Morris-Cohen, A. J.; Knowles, K. E.; Weiss, E. A. Charge Carrier Resolved Relaxation of the First Excitonic State in CdSe Quantum Dots Probed with Near-Infrared Transient Absorption Spectroscopy. J. Phys. Chem. B 2010, 114, 14514−14520. (21) Wheeler, D. A.; Zhang, J. Z. Exciton Dynamics in Semiconductor Nanocrystals. Adv. Mater. 2013, 25, 2878−2896. (22) Feng, D. H.; Shan, L. F.; Jia, T. Q.; Pan, X. Q.; Tong, H. F.; Deng, L.; Sun, Z. R.; Xu, Z. Z. Optical Manipulation of Electron Spin Coherence in Colloidal CdS Quantum Dots. Appl. Phys. Lett. 2013, 102, 062408. (23) Feng, D. H.; Li, X.; Jia, T. Q.; Pan, X. Q.; Sun, Z. R.; Xu, Z. Z. Long-Lived, Room-Temperature Electron Spin Coherence in Colloidal CdS Quantum Dots. Appl. Phys. Lett. 2012, 100, 122406. (24) Tong, H. F.; Feng, D. H.; Li, X.; Deng, L.; Leng, Y. X.; Jia, T. Q.; Sun, Z. R. Room-Temperature Electron Spin Generation by Femtosecond Laser Pulses in Colloidal CdS Quantum Dots. Materials 2013, 6, 4523−4531. (25) Huxter, V. M.; Kovalevskij, V.; Scholes, G. D. Dynamics within the Exciton Fine Structure of Colloidal CdSe Quantum Dots. J. Phys. Chem. B 2005, 109, 20060−20063. (26) Scholes, G. D.; Kim, J.; Wong, C. Y. Exciton Spin Relaxation in Quantum Dots Measured Using Ultrafast Transient Polarization Grating Spectroscopy. Phys. Rev. B 2006, 73, 195325.

NC core. The surface trapping of the hole leaves the core negatively charged, resulting in an enhancement of the electron spin coherence amplitude generated by the orientation pulse. The enhancement transient of electron spin signals thus reflects the hole trapping dynamics. The measurements reveal a fast and a slow hole trapping process with time constants of sub-10 and sub-100 ps, respectively. Both the trapping rates are orders of magnitude faster than that of carrier recombination. Increasing the excitation power to some level, the trapping sites become fully filled. The power dependence reveals three subpopulations of nanoparticles that differ in hole trapping pathways: the first one characterized by the fast process, the second one characterized by the slow process, and the third one with both processes coexisting in one particle. The trap number per particle for the slow trapping process has been shown to be ∼2−3 times more than that for the fast trapping process.



ASSOCIATED CONTENT

S Supporting Information *

Absorption and PL spectra of the sample, more details related to pump-orientation-probe measurements, Fourier transform spectra of the Faraday rotation signals, and illustration of the decay pathways of band-edge hole states are provided. This material is available free of charge via the Internet at http:// pubs.acs.org.



AUTHOR INFORMATION

Corresponding Authors

*E-mail (D.F.): [email protected]. *E-mail (Z.X.): [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was partially supported by National Key Project for Basic Research of China (Grant No. 2010CB923203), National Natural Science Foundation of China (Grant Nos. 10904038, 11374099 and 11274116), and the Shanghai Municipal Science and Technology Commission (Grant 11JC1403500).



REFERENCES

(1) Cordones, A. A.; Leone, S. R. Mechanisms for Charge Trapping in Single Semiconductor Nanocrystals Probed by Fluorescence Blinking. Chem. Soc. Rev. 2013, 42, 3209−3221. (2) Mooney, J.; Krause, M. M.; Saari, J. I.; Kambhampati, P. A Microscopic Picture of Surface Charge Trapping in Semiconductor Nanocrystals. J. Chem. Phys. 2013, 138, 204705. (3) Jones, M.; Lo, S. S.; Scholes, G. D. Quantitative Modeling of the Role of Surface Traps in CdSe/CdS/ZnS Nanocrystal Photoluminescence Decay Dynamics. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 3011−3016. (4) Gómez-Campos, F. M.; Califano, M. Hole Surface Trapping in CdSe Nanocrystals: Dynamics, Rate Fluctuations, and Implications for Blinking. Nano Lett. 2012, 12, 4508−4517. (5) Kern, S. J.; Sahu, K.; Berg, M. A. Heterogeneity of the ElectronTrapping Kinetics in CdSe Nanoparticles. Nano Lett. 2011, 11, 3493− 3498. (6) Garrett, M. D.; Dukes, A. D., III; McBride, J. R.; Smith, N. J.; Pennycook, S. J.; Rosenthal, S. J. Band Edge Recombination in CdSe, CdS and CdSxSe1−x Alloy Nanocrystals Observed by Ultrafast Fluorescence Upconversion: The Effect of Surface Trap States. J. Phys. Chem. C 2008, 112, 12736−12746. 4315

dx.doi.org/10.1021/jz502340w | J. Phys. Chem. Lett. 2014, 5, 4310−4316

The Journal of Physical Chemistry Letters

Letter

(27) Syperek, M.; Yakovlev, D. R.; Yugova, I. A.; Misiewicz, J.; Sedova, I. V.; Sorokin, S. V.; Toropov, A. A.; Ivanov, S. V.; Bayer, M. Long-Lived Electron Spin Coherence in CdSe/Zn(S,Se) SelfAssembled Quantum Dots. Phys. Rev. B 2011, 84, 085304. (28) Liu, W. K.; Whitaker, K. M.; Smith, A. L.; Kittilstved, K. R.; Robinson, B. H.; Gamelin, D. R. Room-Temperature Electron Spin Dynamics in Free-Standing ZnO Quantum Dots. Phys. Rev. Lett. 2007, 98, 186804. (29) Gupta, J. A.; Awschalom, D. D.; Peng, X.; Alivisatos, A. P. Spin Coherence in Semiconductor Quantum Dots. Phys. Rev. B 1999, 59, R10421−R10424. (30) Stern, N. P.; Poggio, M.; Bartl, M. H.; Hu, E. L.; Stucky, G. D.; Awschalom, D. D. Spin Dynamics in Electrochemically Charged CdSe Quantum Dots. Phys. Rev. B 2005, 72, 161303(R). (31) Sewall, S. L.; Cooney, R. R.; Anderson, K. E. H.; Dias, E. A.; Sagar, D. M.; Kambhampati, P. State-Resolved Studies of Biexcitons and Surface Trapping Dynamics in Semiconductor Quantum Dots. J. Chem. Phys. 2008, 129, 084701. (32) Kobayashi, Y.; Nishimura, T.; Yamaguchi, H.; Tamai, N. Effect of Surface Defects on Auger Recombination in Colloidal CdS Quantum Dots. J. Phys. Chem. Lett. 2011, 2, 1051−1055. (33) Klimov, V. I. Optical Nonlinearities and Ultrafast Carrier Dynamics in Semiconductor Nanocrystals. J. Phys. Chem. B 2000, 104, 6112−6123. (34) Kambhampati, P. Hot Exciton Relaxation Dynamics in Semiconductor Quantum Dots: Radiationless Transitions on the Nanoscale. J. Phys. Chem. C 2011, 115, 22089−22109. (35) Sewall, S. L.; Cooney, R. R.; Anderson, K. E. H.; Dias, E. A.; Kambhampati, P. State-to-State Exciton Dynamics in Semiconductor Quantum Dots. Phys. Rev. B 2006, 74, 235328. (36) There may be some heat accumulation at higher excitation power, but its quantity and influence should be small in our system. As verified, for example, after the measurement with Ppump = 5.2 W/cm2, remeasuring the experiment with Ppump = 1.0 W/cm2 can reproduce the corresponding experimental results. (37) Klimov, V. I.; McBranch, D. W.; Leatherdale, C. A.; Bawendi, M. G. Electron and Hole Relaxation Pathways in Semiconductor Quantum Dots. Phys. Rev. B 1999, 60, 13740−13749. (38) N2/N1 = 2 is for the case of P2 = 0, which is impossible as P2/(P1 + P2 + P1,2) ≈ 0.4 shown in the text. N2/N1 = 3 for the case of P1,2 = 0 is also impossible, since N2/N1 = 3 will lead to a relation of P1:P2:P1,2 ≈ 3:4:3 using the calculation method described in the text. N2/N1 ≈ 2.5 is an appropriate value for the case P2 ≅ P1,2, which is in agreement with the final calculated relation of P1:P2:P1,2 ≈ 1.6:4:4.4.

4316

dx.doi.org/10.1021/jz502340w | J. Phys. Chem. Lett. 2014, 5, 4310−4316