Letter Cite This: J. Phys. Chem. Lett. 2017, 8, 5680-5686
pubs.acs.org/JPCL
Impact of Element Doping on Photoexcited Electron Dynamics in CdS Nanocrystals Lei Zhang, Qun Zhang,* and Yi Luo* Hefei National Laboratory for Physical Sciences at the Microscale, Department of Chemical Physics, and Synergetic Innovation Center of Quantum Information & Quantum Physics, University of Science and Technology of China, Hefei, Anhui 230026, China S Supporting Information *
ABSTRACT: Element doping plays a key role in achieving desired properties of semiconductor nanocrystals. In the energy-state landscape the doping-induced localized impurity states (LIS) can bring on significant modification of photoelectrochemical effects. It is difficult to retrieve information regarding the doping-induced LIS. Here we report on such information gleaned on a prototypical system of CdS nanocrystals slightly doped with In3+, through joint observations from photoluminescence (PL) and ultrafast transient absorption (TA) spectroscopy. The nonradiative nature of the Indoping induced LIS is revealed by PL. The TA observations, with a set of control experiments, enable us to capture a picture of the photoexcited electron dynamics and unravel the photoexcited electron reservoir (PEER) effect associated with the In-doping induced band gap LIS. This work establishes a fundamental, mechanistic understanding of the significant impact of element doping on the photoexcited electron dynamics in this model system, offering useful inputs for relevant material design and applications.
E
contributed by carbonyl groups on the graphene nanosheet) play a pivotal role in controlling the photoinduced ultrafast charge carrier dynamics therein. Very recently, doping in the topical system of perovskite nanocrystals was also found to modify the band structure and influence the involved charge carrier dynamics.20,21 Despite the existence of scattered reports on doping-related behaviors via spectroscopic investigations,19−25 the current knowledge pertaining to the role of doping-induced LIS, especially the elusive excited-state dynamics aspects, remains rather limited and unclear. Herein, we execute a systematic scrutiny into the role of LIS in a prototypical system of doped nanocrystals, i.e., cadmium sulfide (CdS) slightly doped with indium (In), through joint observations from steady-state absorption and photoluminescence (PL) as well as time-resolved PL and ultrafast transient absorption (TA) spectroscopy. In light of its similar atomic radii to Cd and the fact that In3+ exhibits a best doping effect among other trivalent cations,26 the element In is chosen as the dopant for this study. Our careful analysis of the interrelated spectroscopic data enables fresh insights into the role of Indoping induced LIS in the CdS nanosystem. In addition to the nonradiative nature revealed by the PL measurements, the photoexcited electron reservoir (PEER) effect of the In-doping induced band gap LIS on the ultrafast charge transfer dynamics is unraveled by the TA measurements and further verified by a set of well-designed control experiments.
lement doping holds the key to semiconductor-based technologies.1−3 The impurities (or dopants) intentionally introduced into the host lattice of semiconductor nanocrystals usually bring about new phenomena or effects not found in the bulk.4 Similarly to the bulk situation, the impurities in nanocrystals can significantly modify their electronic, optical, magnetic, and transport properties.5−7 As the promise of nanocrystals for various applications, especially the photoelectrochemical ones, rests on tailoring their properties through doping,8 tremendous efforts have been devoted to optimizing and/or developing controllable, doping-related synthetic methods9 as well as exploring the doping-induced behaviors and functions in semiconductor nanocrystals.10 For instance, in the topical field of solar-to-chemical energy conversion, including applications in solar cells11−13 and photocatalysis,14−17 it has been widely accepted that element doping in semiconductor nanocrystals represents one of the most important strategies to greatly improve the desired photoelectrochemical properties and performances. Nevertheless, one of the challenging tasks in the field lies in how to establish a fundamental, clear understanding of the involved mechanisms, especially the elusive impact of element doping on the photoexcited charge carrier dynamics therein.18 A direct outcome of element doping in semiconductor nanocrystals is the emergence of localized impurity states (LIS). It is highly desirable to glean useful information concerning the role of doping-induced LIS in excited-state dynamics. Recent years have witnessed some progress in this respect by means of ultrafast optical spectroscopy. For instance, in a previous work of our group,19 we revealed that in the as-synthesized graphene oxide (a two-dimensional nanocrystal system of graphene doped with oxygen) the doping-induced LIS (mainly © 2017 American Chemical Society
Received: September 14, 2017 Accepted: November 6, 2017 Published: November 7, 2017 5680
DOI: 10.1021/acs.jpclett.7b02449 J. Phys. Chem. Lett. 2017, 8, 5680−5686
Letter
The Journal of Physical Chemistry Letters The undoped and In-doped CdS samples are facilely synthesized via a modified solvothermal method.27 Both samples exhibit a nanorod (NR) morphology with a diameter of ∼50 nm and a length of about a few hundred nanometers (Figure S1 in the Supporting Information). With the aid of inductively coupled plasma mass spectrometry (ICP-MS) analysis, the molar composition of the cations (Cd2+:In3+) in doped CdS NRs is determined to be ∼0.98:0.02 (sample denoted Cd0.98In0.02S NRs hereafter). This molar ratio is also confirmed by X-ray photoelectron spectroscopy (XPS) (Figure S2) and energy-dispersive X-ray spectroscopy (EDS) (Figure S3), demonstrating the uniform distribution of the In3+ dopants in the lattice of CdS. Notably, since nearly no difference is discerned for the powder XRD patterns recorded on CdS NRs and Cd0.98In0.02S NRs (Figure S4), such a slight doping turns out not to affect the crystalline phase of CdS (indexed to a wurtzite structure; refer to JCPDS No. 41-1049). It is wellknown that the valence-electron XPS spectral analysis can help determine the band structure of semiconductor materials.28,29 From such an analysis it is found that the two samples possess a nearly identical band structure (Figure S5). Further, the conduction-band minimum (CBM) of the samples is estimated to be about −3.4 eV (vs vacuum) with the aid of Mott− Schottky measurement (Figure S6). The above morphology and structure characterizations clearly demonstrate the successful doping of a very small amount of In3+ into the CdS lattice. Certainly, the obtained real-space information is insufficient for directly correlating the In-doping effect with the resulting photoelectrochemical properties of CdS. It is desirable to obtain energy-landscape information (both steady-state and dynamical) associated with the In-doping induced LIS for this purpose. Therefore, we systematically interrogate the doped Cd0.98In0.02S NRs (in reference to the undoped CdS NRs) by steady-state absorption and PL as well as time-resolved PL and ultrafast TA spectroscopy. Figure 1a shows the steady-state absorption spectra of CdS NRs and Cd0.98In0.02S NRs recorded in the visible region (450− 770 nm). As for CdS NRs, the two peaks observed at ∼500 nm (2.48 eV) and ∼480 nm (2.58 eV) can be ascribed to two quasi-discrete excitonic states (ES), Σ and Π bands, respectively, similarly to the conventional notation used for describing smaller quantum-dot nanosystems.30−33 As is seen, such quasi-discrete ES features are smeared out in the spectrum of Cd0.98In0.02S NRs. This is due to the fact that the In-doping gives rise to new states in the vicinity of the ES, and the resulting broadened absorption deteriorates, to a certain extent, the spectral resolution as the merging of the doping-induced defect states with the ES would make the ES-related absorption features less distinct.21 Figure 1b shows the steady-state PL emission spectra (450−770 nm) of CdS NRs and Cd0.98In0.02S NRs, excited at 400 nm. As for CdS NRs, the two spectral profiles peaking at ∼510 nm (2.43 eV) and ∼675 nm (1.84 eV) can be assigned to the CdS band-edge and defect-state emissions, respectively.34−36 As for Cd0.98In0.02S NRs, there emerges a red-shift (∼25 nm, or ∼113 meV) for the CdS bandedge emissions, suggesting that the In-doping also leads to the formation of new states in the vicinity of the CBM and that a portion of these new states should be located below the CBM. Additionally, nearly no spectral shift is observed for the CdS defect-state emissions. From the CBM position determined above [−3.4 eV (vs vacuum)] together with the PL observations, the energy positions for the valence-band
Figure 1. (a) Steady-state absorption, (b) steady-state PL emission, and (c) time-resolved PL spectra of CdS NRs and Cd0.98In0.02S NRs. The excitation and emission wavelengths are annotated in the plots. IRF, instrument response function.
maximum (VBM) and band gap defect states can be estimated to be about −5.8 and −4.0 eV (vs vacuum), respectively. These new states that arise as a result of the In-doping are illustrated in Figure 2 (see the colored shadow regions: gray for
Figure 2. State energetics (vs vacuum) involved in the investigated system of Cd0.98In0.02S NRs, where VBM, CBM, ES, and LIS stand for valence-band maximum, conduction-band minimum, excitonic states, and localized impurity states, respectively. Also shown are the mechanisms related to electron dynamics in the absence (black arrows) and presence (purple and yellow arrows) of electron-accepting molecules, i.e., BQ and MV2+, whose LUMO energies are −3.85 and −4.16 eV, just higher and lower than the band gap LIS position (−4 eV), respectively. 5681
DOI: 10.1021/acs.jpclett.7b02449 J. Phys. Chem. Lett. 2017, 8, 5680−5686
Letter
The Journal of Physical Chemistry Letters the ES; brown for the states at around the CBM and within the band gap). Most of the state energetics are scaled to the lefthand energy bar (from −3 down to −6 eV), except for the ES levels that are schematically depicted for clarity as they are located only ∼0.05−0.15 eV higher than the CBM (the corresponding ES levels near the VBM are not shown for clarity). Considering that the In-doping turns out not to give rise to new PL emissions and that no spectral shift is manifested in the CdS defect-state emissions, we can safely reach a conclusion that, energetically, the In-doping-induced LIS are embedded in the CdS band gap defect states, as labeled “LIS” as a whole in Figure 2 for simplicity. Moreover, it is seen from Figure 1b that the In-doping leads to a significant PL quenching for both the band-edge and defect-state emissions, by roughly 12- and 2-fold intensity decrease, respectively. This observation suggests that the new near-CBM states and In-doping induced band gap LIS are of a nonradiative nature. Figure 1c compares the time-resolved PL spectra for CdS NRs and Cd0.98In0.02S NRs, excited at 400 nm and monitored at 510 nm (band-edge emissions). The average PL lifetimes obtained through a multiexponential fitting (Table S1) are ∼52 and ∼9 ns for CdS NRs and Cd0.98In0.02S NRs, respectively. Apparently, the In-doping induces a roughly 6-fold decrease in PL lifetime, which is commensurate with the PL quenching effect observed in the steady-state measurements. Note that the PL lifetimes of the 675 nm emissions responsible for the band gap defect states (or LIS) are not given here, since such lifetimes are found to exceed the measurement limit (∼1 μs) of our PL spectrometer. The above PL results reveal the nonradiative nature of the In-doping induced LIS. To further explore how such states influence the photoexcited charge carrier dynamics of CdS, we carry out ultrafast TA spectroscopy characterizations37−42 on doped Cd0.98In0.02S NRs in reference to undoped CdS NRs. In our TA measurements, a femtosecond time-resolved pump− probe configuration is adopted (details in Supporting Information). Figure 3a,b shows the TA spectra (450−750 nm) taken at several probe delays under 400 nm excitation for CdS NRs and Cd0.98In0.02S NRs, respectively, both of which feature a probe bleaching of dominantly the ES and near-CBM bands (manifested as negative TA signals). From the comparison of the two TA spectra, one can see that the Indoping leads to nearly no changes in terms of the dip position (∼500 nm) and the temporal evolution of spectral profile. Nevertheless, there occurs a pronounced broadening; for instance, in terms of the TA profiles taken at 2 ps the fullwidth half-maximum (fwhm) is ∼30 and ∼41 nm for CdS NRs and Cd0.98In0.02S NRs, respectively. Again, such a spectral broadening verifies that the In-doping brings about new states in the vicinity of the ES and CBM, echoing well to the above discussions. Figure 3c presents the TA kinetic traces for both CdS NRs and Cd0.98In0.02S NRs, recorded at 500 nm. The analysis of their early time kinetics (Figure S7) reveals that for both samples the exciton generation (due to the relaxation of the 400 nm photoexcited hot electrons within the CB) takes a characteristic time constant of ∼0.3 ps, a process that corresponds to the initial build-up process43,44 (not expanded in Figure 3c) marked with a gray, downward arrow in Figure 2. The subsequent relaxation processes, marked with three black, downward arrows in Figure 2, can be described by three time constants: τ1 = 100 ± 1 ps (78%), τ2 = 685 ± 132 ps (8%), and τ3 = 3.8 ± 0.4 μs (14%) for CdS NRs; τ1 = 42 ± 1 ps (53%), τ2 = 234 ±
Figure 3. TA spectra of (a) CdS NRs and (b) Cd0.98In0.02S NRs, taken at several probe delays (pump at 400 nm). (c) TA kinetic traces for both CdS NRs and Cd0.98In0.02S NRs, recorded at 500 nm. The signal amplitude of the former is normalized to that of the latter. The inset of panel c shows the zoom-in of the time window 0.3−50 μs. The TA signals (i.e., the absorbance changes, or ΔAbs.) are given in the unit of mOD (OD, optical density).
16 ps (24%), and τ3 = 3.7 ± 0.4 μs (23%) for Cd0.98In0.02S NRs. On one hand, the statistical weights for the τ2 and τ3 components are increased by a factor of 3 and 1.6 for Cd0.98In0.02S NRs as compared to CdS NRs (i.e., 24% vs 8% for τ2 and 23% vs 14% for τ3), arising as a result of the In-dopinginduced, near-CBM states and band gap LIS. On the other hand, the In-doping induces a nearly 3-fold acceleration for the former two decays (i.e., 42 ps vs 100 ps for τ1 and 234 ps vs 685 ps for τ2), while the final charge recombination process seems unaltered (i.e., ∼ 4 μs). This observation indicates that both the coupling between the ES and the near-CBM states (characterized by τ1) and the coupling between the near-CBM states and the band gap LIS (characterized by τ2) get enhanced, justifying again the In-doping induced increase of density of states therein.21,44 These doping-induced new states open up new channels for the cascading electron transfer (ET) from the upper states involved. Following a similar treatment documented in the literature,45−47 the efficiency of such new channels can be estimated by the relevant ET rate, kET, that equals [1/τi(Cd0.98In0.02S) − 1/τi(CdS)] (i = 1, 2). In this particular case, the kET values are 1.4 × 1010 and 2.8 × 109 s−1 for the cascading τ1 and τ2 processes, respectively. The opening of such additional, efficient ET channels can help deplete effectively the photoexcited electrons at the ES and near-CBM states, thereby suppressing the competitive, nonradiative Auger recombination in CdS NRs.48−50 5682
DOI: 10.1021/acs.jpclett.7b02449 J. Phys. Chem. Lett. 2017, 8, 5680−5686
Letter
The Journal of Physical Chemistry Letters Markedly, an interesting phenomenon is observed in the relaxation kinetics. As shown in the inset of Figure 3c (a zoomin of the time window 0.3−50 μs), for Cd0.98In0.02S NRs starting from ∼3 μs, there emerge pronounced, positive TA signals whose amplitude reaches a maximum at ∼0.25 mOD, which is, however, absent for CdS NRs. Generally, two possible origins are responsible for the appearance of such long-time, positive TA signals: either due to photoinduced absorption by photoexcited charge carriers trapped in defect states or due to nonlinear interactions such as Stark-induced effect and exciton−exciton annihilation.43,51−55 Considering that the latter usually becomes significant only under a high-fluence pumping condition, we can safely rule out the nonlinear effects in this case since the pump fluence used in our TA measurements is relatively low (typically 3 μs) relaxation kinetics, a natural outcome of photoinduced absorption by the retained electrons therein. To examine the validity of the above hypothesis (i.e., the proposed PEER effect), we resort to a set of control experiments on the In-doped samples (i.e., Cd0.98In0.02S NRs), in which small amounts of two well-known electronacceptor reagents, i.e., benzoquinone (BQ) and methyl viologen (MV), are separately used as additives in the sample solutions. Note here that the concentrations of BQ and MV are ∼2 and 1.5 mg/mL, roughly 7- and 2-fold the concentration of CdS sample, respectively. In similar works using BQ or MV as electron acceptor, their concentrations were usually much higher than what we used here; for instance, ∼ 100-fold the concentration of semiconductor sample.43 Once added, the BQ and MV molecules will be adsorbed on the surface of Cd0.98In0.02S NRs (note that MV will be in the form of MV 2+).56−58 The lowest unoccupied molecular orbital (LUMO) energies of BQ and MV2+ are approximately −3.85 and −4.16 eV (vs vacuum), respectively,59,60 which are just higher and lower than that of the band gap LIS [−4 eV (vs vacuum)], respectively, as illustrated in Figure 3. It is worth mentioning here that the highest occupied molecular orbital (HOMO) energies of both BQ and MV2+ are rather low,61,62 and hence the hole transfer processes can be ignored in the mechanistic picture. Upon photoactivation, the photoexcited electrons will be rapidly transferred to the electron acceptors BQ and MV2+, and, as a result, the charge-separated states (usually denoted Cd0.98In0.02S+−BQ− and Cd0.98In0.02S+−MV+•) are formed. Such a unique energy relationship [i.e., E(BQ) > E (band gap LIS) > E(MV2+)] allows us to evaluate the PEER effect associated with the In-doping induced band gap LIS through monitoring the variation of the involved electron dynamics in the presence of charge-separated states with higher and lower energies relative to the In-doping induced band gap LIS. Figure 4a,b shows the TA kinetic traces (recorded at 500 nm) for Cd0.98In0.02S NRs in the presence of BQ and MV2+, respectively. The relevant TA spectra are presented in Figure S8. It is important to point out here that all of the TA signals are acquired under exactly the same conditions except for the
Figure 4. TA kinetic traces (400 nm pump and 500 nm probe) for (a) Cd0.98In0.02S−BQ and (b) Cd0.98In0.02S−MV2+, in both of which the trace for Cd0.98In0.02S NRs (taken from Figure 3c) is also plotted for comparison. The insets show the zoom-in of the time window 0.3−50 μs. The TA signals (i.e., the absorbance changes, or ΔAbs.) are given in units of mOD (OD, optical density).
addition of a small amount of BQ or MV molecules. Similarly to the case in the absence of electron-acceptor reagents (also plotted in Figure 4), the relaxation processes can also be described by three time constants (fitting results listed in Table 1). Table 1. Relaxation Time Constants for the Three Investigated Systemsa system Cd0.98In0.02S Cd0.98In0.02S−BQ Cd0.98In0.02S−MV2+
τ1/ps
τ2/ps
42 ± 1 (53%) 234 ± 16 (24%) 23 ± 1 (50%) 136 ± 6 (30%) 25 ± 1 (49%) 202 ± 7 (29%)
τ3/μs 3.7 ± 0.4 (23%) 3.0 ± 0.5 (20%) 8.8 ± 1.4 (22%)
The corresponding statistical weights from the triexponential fitting are given in the parentheses. a
As for the Cd0.98In0.02S−BQ system (Figure 4a), the addition of BQ results in a nearly 2-fold acceleration for the cascading τ1 and τ2 processes (i.e., 23 ps vs 42 ps for τ1 and 136 ps vs 234 ps for τ2). Similarly, such an acceleration can be ascribed to the opening of new ET channels by BQ (highlighted by the upper two purple arrows in Figure 2, left panel), and the corresponding kET values are 1.9 × 1010 and 3.1 × 109 s−1 for the τ1 and τ2 processes, respectively. Inevitably, a portion of the electrons received by BQ will be transferred to the band gap LIS due to the higher energy of the former than the latter (refer 5683
DOI: 10.1021/acs.jpclett.7b02449 J. Phys. Chem. Lett. 2017, 8, 5680−5686
Letter
The Journal of Physical Chemistry Letters to the shortest, purple arrow in Figure 2, left panel). Such injected electrons into the band gap LIS are expected to levitate the photoinduced absorption. As expected, starting from ∼1 μs there emerge more pronounced, positive TA signals (maximum ∼0.38 mOD) for Cd0.98In0.02S−BQ than for Cd0.98In0.02S (maximum ∼0.25 mOD), as compared in the inset of Figure 4a. As for the Cd0.98In0.02S−MV2+ system (Figure 4b), it is not surprising to see that the presence of MV2+ also leads to a nearly 2-fold acceleration for the τ1 process (i.e., 25 ps vs 42 ps). However, introducing MV2+ turns out not to significantly affect the τ2 process (i.e., 202 ps vs 234 ps), which implies that the τ2 process is dominated by the coupling between the nearCBM states and the band gap LIS. Given its lower energy than the band gap LIS, MV2+ tends to receive electrons from the latter, giving rise to variation of the observed long-time, positive TA signals associated with the In-doping-induced band gap LIS. Indeed, such positive TA signals disappear utterly, as clearly seen in the inset of Figure 4b. In addition, the charge recombination process (characterized by τ3) is found to be decelerated by more than 2-fold (i.e., 8.8 μs vs 3.7 μs). This is most likely caused by the effective coupling between the band gap LIS and the low-lying LUMO of MV2+ (refer to the shortest, yellow arrow in Figure 2, right panel), which slowly depletes the electrons accumulated in the band gap LIS. It is worth mentioning here that the addition of an excess amount of MV2+ usually produces positive TA signals at ∼600 nm due to photoinduced absorption by MV2+ itself.56,57 In the current study, MV2+ is added in a very small amount (1.5 mg/mL) and hence the TA signals from MV2+ itself is found to be negligible under the relatively low pump fluence (typically
