1. Introduction

Two-dimensional (2D) transition metal dichalcogenides (TMDs) exhibit unusual photoelectric properties that can be exploited for novel optoelectronics, such as ultrasensitive phototransistors [1,2], ultrathin photovoltaics [3,4], and flexible photodetection [5,6]. These 2D TMDs assisted by metamaterials also exhibit giant nonlinear optical response [7,8]. For 2D TMDs materials, their energy band structures possess the layer-dependent properties, because there is a transition from an indirect bandgap in the multilayer to a direct gap in the monolayer [9]. Especially, their band structures can be tunable in a variety of ways, such as element doping, defect control, applied strain and electric field etc. [10,11]. What they have in common is that their surfaces are naturally passivated without any dangling bonds, which make it easily to form van der Waals heterostructures or integrate with photonic structures [12]. The difference is that monolayer TMDs absorbs around 10% of vertically incident light at excitonic resonances, while multilayer TMDs can absorb much more of incident light only achieved in the thickness regime [10,13,14]. Therefore, how to balance the relationship between the layer thickness and photoelectric performance is the key for the practical application of 2D materials-based devices. For example, the multilayer TMDs with spiral or pyramid type exhibit stronger second-harmonic intensity than the monolayer TMDs as a novel type of nonlinear optical material [15–18].

The carrier lifetimes are critical to all the proposed TMDs-based optoelectronic devices. The carrier lifetimes directly determine the performance of devices, such as luminous intensity, luminous quantum efficiency and photoelectric conversion efficiency, etc. [14,19–21]. The excited carriers rapidly decay from the high energy state to the bottom of the conduction band to form excitons, and then decay to the ground state through non-radiative or radiative processes [22–24]. So far, most studies have been concentrated on the carrier lifetimes of pure monolayer TMDs [25–29]. For the monolayer TMDs, the lifetime of non-radiative processes is usually in the order of hundreds of femtoseconds to a few picoseconds, which are mainly determined by the Auger electron scattering and exciton-exciton annihilation [11,30,31]. While the radiative lifetime of exciton takes a longer time in tens to hundreds of picoseconds depending on the defect density [27,32]. For the mechanically exfoliated multilayer TMDs, basic decay processes of exciton are similar to the monolayer TMDs, the different is that the phonon-assisted effect and the influence of layer thickness [33]. However, the competition mechanism of decay channels in exciton dynamics of the multilayer TMDs remain poorly understood. Developing a better understanding of the exciton decay channels in the multilayer TMDs is especially important, because the longer exciton lifetime is more conducive to carrier separation, thus improving photoelectric conversion efficiency in the TMDs-based optoelectronic device. To our best knowledge, the exciton dynamics of the stacked multilayer tungsten sulfide (WS₂) prepared by physical vapor deposition (PVD) has not yet been studied.

In this letter, we reported an experimental observation of exciton dynamics in the stacked multilayer WS₂ using a femtosecond pump–probe spectroscopy. The results indicated that exciton decay processes are divided into fast and slow decay processes. The fast decay process is dominated by the exciton-exciton annihilation (EEA) that the lifetime of EEA is decreased initially, followed by an increase with adding layer thickness. The slow decay process is dominated by the defect-assisted recombination (DAR), which is determined by the defect density especially in a high injected carrier density. These results provide a foundation for understanding exciton dynamics in the stacked multilayer WS₂ material.

2. Results and discussion

2.1 Basic characterization of the stacked WS₂ layers

The basic properties of the stacked WS₂ layers on a Si/SiO₂ substrate prepared by PVD method were first characterized. Figure 1(a) is the optical microscope images of stacked WS₂ layers. The samples with triangular shape are typical properties of 2D WS₂ layers [26,34,35]. As shown in Supplement 1 Fig. S1, the constructed monolayer WS₂ has a side length of approximately 41 µm. Such large-size samples offer a powerful condition for accurate studies of layer-related optics and exciton dynamics features. The corresponding number of layers and measurement positions are marked with symbols (1, 2, 3, 4 and M) in Fig. 1(a). During the subsequent pump power dependent and probe wavelength dependent exciton dynamics examinations, the measurement positions were kept unchanged to minimize the measurement errors caused by position changes. The thickness of the WS₂ layers is characterized by atomic force microscopy (AFM) as shown in Fig. 1(b). The height profile of ∼ 7 Å indicates that a monolayer was stacked on the top [23,35,36]. Layer numbers and thickness are indicated in the appropriate positions. The measured microscopic reflectance and photoluminescence (PL) spectra are presented in Fig. 1(c) and (d). The reflectance spectra of stacked WS₂ layers are shown in Fig. 1(c). The absorption peaks of excitons appear as valleys in the reflectance spectrum. Due to the limits of the spectral apparatus, the overall linear baseline of the reflection spectrum is altered, but the location of the reflection valley of the exciton is unaffected. All layers display two main valleys of X_A and X_B excitons. The X_A and X_B excitons originates from the top splitting of the valence band maximum (VBM) caused by strong spin-orbit coupling and interlayer coupling (only suitable for few-layer WS₂) [10,33]. The reflection valley of the X_A exciton is approximately 638 nm (∼1.94 eV) in the monolayer and grows with the number of layers which reaches about 648 nm (∼1.91 eV) in the 4-layer WS₂. The reflection valley of the X_A exciton is, conversely, blue-shifted to 629 nm (∼1.97 eV) in the multilayer WS₂. The X_B exciton is located at ∼525 nm (∼2.36 eV) for all layers. The PL intensity following excitation at 3.1 eV was normalized around the emission peak of monolayer. The emission peaks of X_A and X_B excitons were clearly presented in the PL spectra around 642 nm (∼1.93 eV) and 534 nm (∼2.32 eV) which can be observed in all different WS₂ layers. The PL intensities of X_A and X_B excitons decrease as the number of layers increases while the splitting energy (∼400 meV) is almost independent of layer numbers [37]. It needs to be noticed that the PL intensity here is relatively lower compared to the PL from exfoliated WS₂ monolayers which indicated the PVD growth stacked WS₂ layers have lower quantum efficiency and more defects [10,30]. The PL of bilayer WS₂ prepared by the classical method is 2 orders of magnitude less than that of the monolayer [10]. Whereas the PL intensity of bilayer or even multilayer WS₂ prepared by PVD method here is of the same order of magnitude as the monolayer due to the 0° stacking [36,38]. Referring to the results of the literature [36] and combining the experimental data of the optical image and PL spectra in this work, we have reason to infer that the stacked multilayer WS₂ is a 3R-like phase stacking structure. As the layer number increase, the PL intensity of the 3R-like phase sample weaken more slowly, which mainly attributed to different interlayer coupling strengths [36,39,40]. The interlayer electronic coupling leads to the transition from direct to indirect band gap, the lower the indirect band gap, the stronger the coupling strength [41]. From Fig. 1(d), it is reasonable to infer that the PL intensity of the stacked multilayer WS₂ exhibit slower attenuation, which mainly originated from the weak interlayer coupling strengths. The relationship of them and the relevant topics need further study.

 

Fig. 1. Characterization of the stacked WS₂ layers on a Si/SiO₂ substrate. (a) Optical microscope images of stacked WS₂ layers with scale bar of 10 µm. (b) AFM image of the investigated WS₂ layers within the white dotted box in (a). The inset depicts a monolayer AFM picture matching to the black dotted frame in (a). (c) Measured microscopic reflectance spectra of stacked WS₂ layers shown in (a). The exciton peaks X_A and X_B are indicated. (d) Measured microscopic PL spectra of stacked WS₂ layers shown in (a). The PL peaks of X_A and X_B excitons are indicated.

Download Full Size | PDF

2.2 X_A exciton dynamics versus layer number

The strategy that is followed to investigate the X_A exciton dynamics in stacked WS₂ layers consists of exciting the samples with photon energies above their band gaps, then probing the decay dynamics of the X_A exciton. The experimental configuration of dichromatic coaxial microscopic pump-probe measurement shown in Fig. S2 is as follows; the excitation photon energy is selected to be 3.1 eV (400 nm) while the probing photon energy is ∼2 eV (∼620 nm). This excitation photon energy can pump electrons with energy levels lower than VBM to energy levels higher than conduction band (CB) bottom level. Noticed that the electron-hole pairs generated by the injection of a 400 nm pump pulse into WS₂ layers will rapidly generate excitons due to the large exciton binding energy, which is included in the rising edge of the differential reflection signal [42,43]. Therefore, the photocarrier systems caused by the pump pulse is mainly composed of excitons. From the pump fluence, we can estimate the injected exciton density by considering the absorption on the order of 10% per layer in the visible region of the optical spectrum and assume that every pump photon absorbed excites one exciton [28,44]. The probe energy (∼2 eV) is tuned to around the X_A exciton resonance of WS₂ layers, so that the dynamic process of X_A exciton can revealed. All results reported here were obtained at room temperature ambient conditions. No samples were damaged during the entire study.

The dynamics of X_A excitons under stacked WS₂ layers are monitored by measuring differential reflection of the time-delayed 620 nm probe as shown in Fig. 2(a). The pump fluences E_s here is 0.62 µJ/cm² which can be derived by the equation ${E_s} = {P_{ex}}/({f \times S} )$, where P_ex is the average power of the incident light, which is directly measured by the power meter, S is the area of the pump spot and f is the pulse frequency of the laser. The number of photons n in a single pump pulse can be calculated from equation ${E_s} = nhv$, where h is the Planck’s constant and v is the photon frequency [45]. By considering the absorption on the order of 10% per layer and assume that every pump photon absorbed excites one exciton [28], the corresponding injected carrier density is about 1.24 × 10¹¹cm⁻².The differential reflection is defined as $\Delta R/{R_0}\; = \; ({R\; - \; {R_0}} )/{R_0}$, where R and ${R_0}$ are the reflection of probe signal with and without the presence of the pump injection, respectively. Since differential reflection signal is related to the exciton density as discussed before, its decay reflects exciton dynamics. As shown in Fig. 2(a), the differential reflection signal decays quickly in the first few ps shown in the transparent light blue areas and then slowly over several hundred ps in all layers obviously. Figure 2(b) is the signals extracted from (a) near zero-time delays with 100 fs interval steps. To focus on the rising edge of the signals, with more stacking layers, X_A excitons form more slowly. This result means that it takes excited electrons more time in the CB to undergo intraband relaxations before reaching the lowest allowed energy level in the CB to form X_A excitons when the layer number increases. The exciton formation is illustrated in Fig. 4(a). Across all layers, the formation time is within 1 ps, and in monolayer, the formation time is around 0.5 ps which is in good agreement with previous literature [29,42,43,46].

 

Fig. 2. (a) Differential reflection signals of stacked WS₂ layers measured with pump fluences of 0.62 µJ/cm² at the center probe spot with 500 fs interval steps. (b) Differential reflection signals extracted from (a) near zero-time delays with 100 fs interval steps. (c) Peak differential reflection signal as a function of WS₂ layer number. (d) Dependence of τ_EEA and τ_DAR of the dynamics of X_A excitons on the WS₂ layer number. Error bars are standard deviations of the fitted results. The dashed lines are guides for the eye.

Download Full Size | PDF

The peak $\Delta R/{R_0}$ data versus layer number are extracted from Fig. 2(a) shown in Fig. 2(c). As the pump pulse only changes the probe signal less than 0.7% in all layers, the saturation effect of the pump absorption can be ignored [28,47]. According to the optical configuration, the intensity of the peak $\Delta R/{R_0}$ signal reflects the X_A excitons population in the surface atomic layer of the WS₂ sample after pump pulse injection [28]. As layers are stacked, the X_A excitons population increases; however, with multiple layers, it is attenuated. The reason behind this is due to the linear absorption differences of the layers shown in Fig. 1(c). Since the differential reflection is related to the exciton density, its decay reflects the X_A excitonic dynamics. Here, a biexponential decay function was used to fit the decay signal of the X_A excitons in Fig. 2(a). The specific fitting process and attention details are shown in Supplement 1 S3. In 2D materials, exciton-exciton annihilation (EEA) can be triggered when the initial injected exciton density exceeds 10¹⁰cm⁻² [24,31]. The injected exciton density here reaches up to 10¹¹cm⁻² which can easily trigger the EEA process. Hence, the fast decay lifetime (τ_EEA) here is considered mainly to be the EEA process as shown in Fig. 4(a). Noted in few and multilayers, which are indirect semiconductors, momentum conservation requires assistance from phonons in the EEA process [33]. As for the slower decay channel, defect-assisted recombination (DAR) is generally considered to be the dominant process since the stacked WS₂ layers prepared by PVD method contain many defects which can easily capture excitons into defect state as shown in Fig. 4(b). In Fig. 2(d), the fit values of the fast and slow decay times (τ_EEA and τ_DAR) are shown in black and red points, respectively. The dashed lines are guides for the eye. The τ_EEA decreased from ∼0.7 ps initially, followed by an increase up to ∼1 ps with adding layer thickness. The decrease τ_EEA can be attributed to the phonon-assistance in few layers which increasing the EEA probability. Whereas, with the further increase of the layer number, the defect density increased which widen the distance between excitons and weaken the exciton diffusion inducing the increase τ_EEA. Hence, the variation of τ_EEA can be attributed to the competition between phonon-assisted effect and defect effect. In all layers, the τ_DAR is on the order of hundred picoseconds, showing a long-lived time for X_A excitons under defect-assisted recombination. We also tested the three others stacked WS₂ layers to maintain data accuracy and rigorousness as shown in Supplement 1 S4. The results of more systematic exciton lifetime tests and the underlying physical mechanism will be discussed in the following part.

2.3 X_A exciton dynamics versus pump fluences

The differential reflection was measured on all layers with different pump fluences to establish a precise correlation between the differential reflection and the exciton density. Figure 3(a) and (b) are the 2D plots of differential reflection signals versus pump fluence for monolayer and multilayer WS₂, respectively. The $\Delta R/{R_0}$ scales are normalized from 0 to 1% for both plots. A comparison of Fig. 3(a) and (b) indicates that the τ_DAR of multilayer WS₂ is much longer than that of monolayer. The differential reflected signals under different pump fluences for the remaining samples are shown in Supplement 1 S5. In Fig. 3(c), the peak $\Delta R/{R_0}$ data points under different injected carrier densities are extracted from different layers which are distinguished by different colors. The peak $\Delta R/{R_0}\; $ signals show similar trend with the change of injected carrier densities in all layers. Based on a saturable absorption model, the relationship can be accurately described [28].

By assuming that every pump photon absorbed excites one exciton, the N and N_s here can be regarded as the X_A exciton density and the saturation exciton density, respectively. As shown in Fig. 2(c), the solid lines represent the fitted curves of the saturable absorption model where the X_A exciton density saturates more rapidly on multilayer WS₂ compared to others. The fitted results N_s in each layer are represented in Fig. 2(d). As layers are stacked, the saturation X_A exciton density decreases nonlinearly, from nearly (1.46 ± 0.09) × 10¹²cm^-2 in monolayer down to (0.23 ± 0.02) × 10¹²cm^-2 in multilayer, which is consistent with previous literature [48]. The average X_A exciton distance is converted from about 8 nm in monolayer increased to 26 nm in multilayer. Considering that the saturation carrier density of monolayer MX₂ is in the order of 1 × 10¹³cm^-2, the results is reasonably consistent with these calculations [28,48]. The impact of surface defect density increase in the uppermost layer is the primary factor causing the saturation exciton density to drop as the number of layers increases [30]. Due to the PVD method, the uppermost atomic layer defect density would grow as layers were stacked, reducing the effective regions where excitons might develop.

 

Fig. 3. 2D plot of differential reflection signals versus pump fluence for monolayer (a) and multilayer (b) WS₂. The $\Delta R/{R_0}$ scales are normalized from 0 to 1% for both diagrams. (c) An overview of the peak $\Delta R/{R_0}$ signals as a function of the injected carrier density in all layers. The solid lines represent the fitting curves. (d) Variation of saturation carrier concentration with the number of layers. Error bars are standard deviations of the fitted results.

Download Full Size | PDF

The decay curves of differential reflection signals under varied injected carrier densities were collected and fitted with biexponential functions using the 2D diagrams of WS₂ layers in Fig. 3 and Supplement 1 S4. EEA and DAR processes illustrated in Fig. 4(a), respectively, are dominant in the fitted lifetimes of fast and slow decay processes, as was previously noted. The individual τ_EEA and τ_DAR as a function of the injected carrier density are represented in Fig. 4(b) and (d). As observed in Fig. 4(b), τ_EEA generally keeps at 0.6 ps in the case of few layers stacking with just 1 to 4 layers, and is therefore basically independent of the injected carrier density. For the multilayer WS₂, τ_EEA declines nonlinearly from 1.6 ps to around 0.6 ps as the injected carrier density rises, which finally is equivalent to 1-4 layers. In multilayer WS₂, the additional defect denistiy widen the space between excitons and slow down exciton diffusion [30]. At low carrier injection densities, these effects lead to a lower probability of exciton-exciton annihilation and are attenuated at high injection carrier densities. Figure 4(d) shows the variation of τ_DAR with the injected carrier density under all layers. Compared with monolayer, the superposition of layers leads to higher defect density and more relaxation of indirect excitons are caused by defect-assisted recombination, thus prolonging the lifetime of τ_DAR [24,32,46]. As the injected carrier density increases for WS₂ with few layers, τ_DAR drops from 300–600 ps to below 200 ps. This is due to the fact that more non-radiative recombination pathways get involved in exciton recombination as the density of injected carriers climbs [24,33]. However, the τ_DAR of the multilayer exhibits a phenomenon of nonlinear increasing from ∼200 ps to ∼ 800 ps. On the multilayer, the probe spot covers several stages. The actual result is the superposition of the exciton lifetime at the edges of different layers in the multilayer. More defects in the multilayer and the isolation of exciton diffusion between layers can be the root cause of the rising long τ_DAR [32]. As a result, the rate of exciton recombination is falling, and the lifetime of excitons is further increasing. The entire process can be very complex and competitive.

 

Fig. 4. (a) Schematic of X_A exciton dynamics in defective WS₂ layers. (b) and (d) are the τ_EEA and τ_DAR as a function of the injected carrier density in all layers, respectively. The biexponential fitted values with standard deviations are shown as solid color points with error bars. (c) The lifetime weights of τ_EEA and τ_DAR, marked as A1 and A2, as a function of the injected carrier density in each layer independently. The solid lines serve as eye-guides while the intersection positions are marked with green points.

Download Full Size | PDF

A1 and A2 in Fig. 4(c) are the amplitudes of the biexponential components which represents the lifetime weight of EEA and DAR process, respectively. The variations of Ai with increasing injected carrier density can reveal information on the predominant decay channel for excitons because the amplitudes Ai are closely correlated with the excited population [32]. The lifetime weight A1 of the EEA process for few layers decreases as the injected carrier density increases, whereas A2 has the opposite tendency. This competition phenomenon has been reported in the literature before [32], the trend can be explained by as follows. When in low exciton density, the exciton decay channel is dominated by the phonon-assisted EEA process for the defects can be nearly negligible. As the density of excitons rises, there are many defect states available inducing the DAR pathway is preferred while the phonon-assisted effect is attenuated leading the weight of EEA process falls. For few layers, the intersection position of A1 and A2 advances with the increase of the layer number, that is, the increase of defect density further increases the weight of DAR, lending further credence to the aforementioned explanation [32]. For multilayer, the weights of A1 and A2 remain basically unchanged with the injected carrier density for the extremely higher defect density and lower saturation exciton density.

2.4 X_A exciton dynamics versus probe wavelength

In addition to studying the variation of X_A exciton dynamics with layer number and pump fluence, this work continues to further explore the X_A exciton dynamics as a function of probe wavelength. Here, a low carrier injection density of 0.62 µJ/cm² with injected photoenergy of 3.1 eV at all probe wavelengths were maintained. The probe wavelengths, which covered absorption peaks of the X_A exciton in each layer, varied from 610 nm (2.03 eV) to 650 nm (1.91 eV). From (a) and (b) in Fig. 5, it is clear that the change in peak $\Delta R/{R_0}$ intensity with probe wavelength corresponds to the reflectance spectrum. Please see Supplement 1 S6 for the 2D diagrams of all layers. This indicates that pump injected excitons lessen the strength of the exciton transition in all stacked layers at the responding absorption resonance peaks [28,32,46]. The individual τ_EEA and τ_DAR as a function of the probe wavelength are shown in Fig. 5(c) and (d). Figure 5(c) shows that in the case of few layers stacking with just 1 to 4 layers in the overall probe photoenergy, the whole τ_EEA decreased initially, followed by an increase with adding layer thickness. The reason has been revealed in the former part. The overall lifetime of the DAR process in Fig. 5(d), which is measured to be hundreds of ps and has minor dependence on the probe wavelength, does not appear to be regularly distributed. This result shows that the EEA and DAR processes of stacked WS₂ layers prepared by PVD method can be obviously detected when the detecting energy near the X_A exciton energy band. The lifetime of two processes is consequently unchanged with probe wavelength at low carrier injection densities, implying that minor variations in the observation energy levels have no obvious impact on the characterization of these processes. Additionally, the detection spot is restricted to the matching layer surface and the moving region is very limited, the diffusion related to the exciton has not been examined in this work.

 

Fig. 5. 2D plots of differential reflection signals versus probe wavelength for monolayer (a) and multilayer (b) WS₂. The $\Delta R/{R_0}$ scales are normalized from 0 to 1.43% for both diagrams. (c) and (d) are the τ_EEA and τ_DAR as a function of the probe wavelength in all layers, individually. The biexponential fitted values with standard deviations are shown as solid color points with error bars. The dashed lines serve as eye-guides.

Download Full Size | PDF

3. Conclusion

In summary, a thorough investigation of the X_A exciton dynamics in stacked WS₂ layers grown by PVD method was studied. Firstly, as the layer stacks, the defect density increases, the saturation X_A exciton density drops nonlinearly, and the exciton formation time rises. All WS₂ layers present the fast and slow decay processes, which are dominated by the EEA and DAR processes, respectively. Secondly, the lifetime of EEA decreased from ∼0.7 ps initially, followed by an increase up to ∼1 ps with adding layer thickness, which can be attributed to the competition between phonon-assisted effect and defect effect. Thirdly, the lifetime of DAR is on the long-lived timescale of hundreds of picoseconds (200∼800 ps), which is mainly determined by the defect density, extremely in a high injected carrier density. Lastly, the layer number and injected carrier density affect the weights of the EEA and DAR processes. This is primarily because the competition of decay channel between phonon-assisted effect and defect density. The study of the competition mechanism of exciton decay process for the stacked WS₂ materials, which is useful for its optical characteristics and 2D TMDs-based optoelectrical devices. The choice of relaxation channels and lifetime of exciton directly affect its luminescence characteristics, including the PL, second or third harmonic emission etc. Moreover, the photoelectric conversion efficiency is a key indicator for the applications of 2D TMDs-based optoelectrical devices. Extending the carrier lifetimes is more beneficial to the collection of charge. This work revealed that the carrier lifetimes can be efficiently modulated by controlling the exciton decay channels, which lays foundation for the analysis of the carrier relaxation process at the heterojunction interface, and is also helpful to guide the design of 2D TMDs-based optoelectronic devices.

4. Experimental section

Preparation of the stacked 2D-WS₂ layers: An alumina boat loaded with WS₂ powder was first placed at the heating center of a 1-in. quartz tube, and a piece of Si wafer (with 300 nm of SiO₂) was placed 13 cm downstream of the center of the furnace. Argon gas flow was introduced into the system at a flow rate of 60−100 sccm for 60 min to exhaust the oxygen inside the tube before heating. The furnace was then rapidly heated to 1150 °C and maintained at this temperature for about 30 min, and WS₂ nanosheets with stacked multilayer structures were deposited on the SiO₂ surfaces.

Microscopic Reflectance and PL spectroscopy measurements: A customized microscopic optical spectroscopy system was used to measure the reflectance and PL spectra which were recorded by the spectrometer (Princeton Instruments SP2500) with a × 100 objective lens (Olympus MPlanFL, NA = 0.9). The samples were located on an automated XYZ stage with 0.5 µm step size with commercial Kohler lighting system. The reflectance spectra were represented by the relative reflectivity, which was calculated by dividing the reflected intensity of the sample to that of the substrate. The same procedure was performed for the PL spectra with excited photoenergy of 3.1 eV. Suitable filters were applied on the optical path to remove unnecessary signals.

Dichromatic coaxial microscopic pump-probe spectroscopy: The entire pump-probe measuring apparatus was constructed by ourselves, and Fig. S2 shows the schematic diagram that goes with it. The 800-nm fundamental frequency light generated in Chameleon Ti:Sapphire Laser with a repetition rate of 80 MHz and a pulse duration of 100 fs was passed into the Chameleon Compact OPO-vis device. High quality 400 nm pump light and tunable probe light were emitted from the two ports of the instrument, individually. The pump light chopped at 1000 Hz went through a linear polarizer with the polarization direction along the x axis then passed through a delay line controlled by a stepper linear stage (Newport Motion Controller ESP301). The two laser beams were collinear through a dichroic mirror and normally incident on the sample at room temperature by a × 40 objective lens (0.65 NA). The spot diameter of laser on the sample surface is about 3 µm. The sample can be precisely positioned via a three-axis high-precision translation stage with a self-made Kohler illumination device. Reflected probe light from the sample was detected by a high-sensitivity photomultiplier (Thorlabs PMM02), while reflected pump light was filtered by a long pass filter of 450 nm. The SNR was further improved through the chopper with lock-in amplifier SR830.The acquisition of transient differential reflection signals were controlled by Matlab program.