Ultrafast Pump–Probe Microscopy on 2D Transition Metal Dichalcogenides

Although microscopic techniques have been used to characterize transition metal dichalcogenides (TMDs), direct observation of charge carrier dynamics distribution in TMDs with diverse shapes remains unexplored. Herein, ultrafast pump–probe microscopy (UPPM) is employed to reveal the carrier dynamics distribution in molybdenum disulfide (MoS2) and tungsten disulfide (WS2) monolayer of four shapes: triangular (t‐MoS2), curved triangular (c‐MoS2), triangular (t‐WS2), and hexagonal (h‐WS2). Monitoring the photon transmission T at 1.55 eV after pumping with a photon energy of 3.1 eV, a negative ΔT/T occurs in t‐MoS2 and c‐MoS2, while a positive ΔT/T is detected in t‐WS2 and h‐WS2 after 3–7 ps time evolution. This distinctive behavior is attributed to deep/shallow defects below the conduction band minimum (CBM) in MoS2 and WS2. Spatial‐independent ΔT/T is observed in t‐MoS2 and t‐WS2, while the ΔT/T in c‐MoS2 has a rapid decay of photoexcited carriers at the vertices and curved edges. Additionally, a threefold symmetry of ΔT/T is revealed in h‐WS2, attributed to the dissimilar occupation of defect states near the h‐WS2 CBM. This work paves the way for examining charge carrier dynamics of various shapes of TMDs and provides a unique microscopic method for studying the charge carrier dynamics in emerging TMDs heterostructures.

Although microscopic techniques have been used to characterize transition metal dichalcogenides (TMDs), direct observation of charge carrier dynamics distribution in TMDs with diverse shapes remains unexplored. Herein, ultrafast pump-probe microscopy (UPPM) is employed to reveal the carrier dynamics distribution in molybdenum disulfide (MoS 2 ) and tungsten disulfide (WS 2 ) monolayer of four shapes: triangular (t-MoS 2 ), curved triangular (c-MoS 2 ), triangular (t-WS 2 ), and hexagonal (h-WS 2 ). Monitoring the photon transmission T at 1.55 eV after pumping with a photon energy of 3.1 eV, a negative ΔT/T occurs in t-MoS 2 and c-MoS 2 , while a positive ΔT/T is detected in t-WS 2 and h-WS 2 after 3-7 ps time evolution. This distinctive behavior is attributed to deep/shallow defects below the conduction band minimum (CBM) in MoS 2 and WS 2 . Spatial-independent ΔT/T is observed in t-MoS 2 and t-WS 2 , while the ΔT/T in c-MoS 2 has a rapid decay of photoexcited carriers at the vertices and curved edges. Additionally, a threefold symmetry of ΔT/T is revealed in h-WS 2 , attributed to the dissimilar occupation of defect states near the h-WS 2 CBM. This work paves the way for examining charge carrier dynamics of various shapes of TMDs and provides a unique microscopic method for studying the charge carrier dynamics in emerging TMDs heterostructures. triangular, tetragonal, and pentagonal shape MoS 2 crystals, due to varying defect distribution in these morphologies. [11] Furthermore, nonuniform alternating domains with bright and dark PL emission apparent in hexagonal tungsten disulfide (h-WS 2 ) are attributed to the tungsten and sulfur vacancies in this morphology. [12][13][14] In catalysis, the shape and edge-structure of WS 2 and MoS 2 is considered a dynamic function for hydrogen evolution reaction (HER) due to the existence of varying metal and chalcogen vacancies in these materials. [11,15] According to the laws of thermodynamics, all crystals grown above 0 K have some level of defects. [16,17] Such defects result in detrimental consequences such as unstable operational circuits, photoluminescence quenching, weak photosensing capabilities, and intense fluctuating of carrier dynamics of 2D materials. [18][19][20][21][22] Although extensive defect studies have been carried out with techniques such as micro-Raman scattering, PL, X-Ray photoelectron spectroscopy, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), and scanning tunneling microscope (STM), [18,[23][24][25][26][27] visualization of the dynamics of charge carriers, which is essential in photonic applications, cannot be realized with these techniques. This has necessitated the employment of a nondestructive and time-efficient technique such as time-resolved optical pump-probe spectroscopy for direct probing and visualization of the charge carrier dynamics in TMDs. Information on the excited-state lifetime is beneficial for practical applications. On one hand, optical switches [28] rely on the short photocarrier lifetimes in materials. On the other hand, light-emitting devices [29] favor the long radiative lifetime of photocarriers for enhancement of luminescence quantum yield. The basic principle of time-resolved pump-probe microscopy is based on triggering the optical response of samples under investigation with a high-intensity pump pulse. The dynamics of the process is investigated by using a weak probe beam delayed with respect to the pump beam. [30,31] Despite of strong Coulomb interactions and strong electron-hole concurrence in TMDs, the dynamics of charge carriers can be visualized to estimate the changes in carrier lifetimes and defect distribution in TMDs. [32] While several ultrafast investigations have been carried out with diverse TMDs such as MoS 2 , [33] WS 2 , [34] MoSe 2 , [35] WSe 2 , [36] and CdSe, [37] little or no work has been done on comparing the charge carrier dynamic distributions in shapedependent MoS 2 and WS 2 samples.
In this article, we employ time-resolved pump-probe microscopy to reveal the shape-dependent carrier dynamics in MoS 2 and WS 2 morphologies. In these experiments, transient transmittance changes ðΔT=TÞ are used to observe the ultrafast dynamics in four morphologies of CVD grown TMDs comprising triangular MoS 2 (t-MoS 2 ), curved triangular MoS 2 (c-MoS 2 ), triangular WS 2 (t-WS 2 ), and hexagonal WS 2 (h-WS 2 ). Ultrafast probing of the carrier dynamics in t-MoS 2 and t-WS 2 revealed position-independent ΔT=T. However, the ΔT=T in c-MoS 2 is sensitive to position, where a quick decay of photoexcited carriers occurs at the curved edge and vertices. In addition, positiondependent ΔT=T and threefold symmetry is observed in h-WS 2 , with defect states as the proposed mechanism. Additionally, with the same pump-probe transmission, a negative transient transmittance is observed in both t-MoS 2 and c-MoS 2 while a positive transient transmittance occurs in both t-WS 2 and h-WS 2 samples after 3-7 ps (defect-dependent), which we attribute to the different defect states in the two materials. The results of carrier dynamic distribution in this work will provide knowledge on specific areas of optoelectronic device fabrication when using MoS 2 or WS 2 of desired shapes. Figure 1a shows the schematic depiction of the optical ultrafast pump-probe microscopy (UPPM) setup used for the ultrafast dynamic measurements of MoS 2 and WS 2 samples (See experimental section for details). For the spatial-resolved ultrafast carrier dynamics, a scanning microscope combined with the pump-probe system is adopted to map the MoS 2 or WS 2 . By controlling the time delay between pump and probe pulses, the position-dependent transient transmittance changes (ΔT/T ) can be obtained. The measured ΔT/T can be expressed as [38] ΔT

Results and Discussion
where σ r ðσ i Þ is the real (imaginary) part of the sample optical conductivity, n s % 1.45 is the refractive index of the substrate, and η 0 is the free space impedance. When η 0 σ r and η 0 σ i are ( 1 at the probe energy, [32] the second term on the right-hand side in Equation (1) becomes much smaller than the first term. Therefore, the differential transmission is predominantly affected by changes in the real part of the optical conductivity as expressed in the first term on the right-hand side, as a result of intraconduction band and intravalence band. [32] To check the sample quality, standard micro-Raman mapping measurements on an as-grown t-MoS 2 were carried out. The data shown in Figure 1b reveals a uniform spatial map of the in-plane E 1 2g phonon mode. In addition, the Raman spectrum of E 1 2g and A 1g modes are extracted from the vertex, edge, and center of t-MoS 2 , which confirms the uniformity of the t-MoS 2 sample (see Figure S1a,b, Supporting Information). All samples are characterized in the same fashion prior to pump-probe measurements. Three individual ΔT/T measurements are then performed at the vertex, edge, and center of t-MoS 2 (inset of Figure 1c). As plotted in Figure 1c, all three single-point ΔT=T measurements exhibited near-identical absorption strength and decay behavior, indicating a homogeneous carrier dynamic distribution. Transient transmittance microscopy of the same sample further revealed a uniform transmission behavior distribution from 0 to 40 ps (Figure 1d), confirming a similar transmittance relaxation time over the entire area of t-MoS 2 flake. As a result, the uniform photoexcited carrier dynamic distribution of t-MoS 2 is directly observed with our UPPM.
Next, c-MoS 2 is examined in the same fashion. As shown in the micro-Raman mapping image of the in-plane E 1 2g mode of c-MoS 2 , the intensity of the E 1 2g is low at the vertices and at the edges compared to the center where a fairly uniform intensity is obtained (Figure 2a). Representative Raman spectrum extracted from spots A-C reveal varying Raman intensities at the vertices, edge, and center (see Figure S2, Supporting Information), which could be related to defect-induced strain. It should be noted that, though defect-induced strain causes changes in the vibrational modes of MoS 2 flakes, no apparent shift in the vibrational modes could be observed in our Raman measurements. This could be due to insufficient strain from low defect concentrations that could not be detected in these Raman measurements. [12,13]   This Raman response of defect-induced strain suggests that photoexcited carriers can be trapped with these additional defect states. Figure 2b depicts three individual ΔT/T transients measured at the vertex, edge, and center of c-MoS 2 (see inset of Figure 2b). The relaxation time of photoexcited carriers at the vertex and the edge (location A and B) are shorter than the relaxation time in the center (location C), indicating a non-uniform distribution of localized trapping states in c-MoS 2 . As a result, a nonuniform negative ΔT/T mapping by UPPM is obtained for c-MoS 2 flake and shown in Figure 2c. It is apparent from the transient transmittance images that the relaxation times of photoexcited carriers are shorter at the vertices and the edges than in the center of c-MoS 2 , visualized by the decreasing density of red points with increasing time. UPPM hence confirms this inhomogeneous carrier dynamic distribution in c-MoS 2 . To provide a much clearer view of the difference in carrier dynamics distribution in t-MoS 2 and c-MoS 2 , the ΔT/T spectra at each pixel within t-MoS 2 (Figure 1d) and c-MoS 2 (Figure 2c) are fitted by a double-exponential decay to obtain the spatial distribution of relaxation time (τ 1 and τ 2 ) (See Figure S3, Supporting Information). It is apparent that both t-MoS 2 and c-MoS 2 show uniform distribution in shorter relaxation time (τ 1 ) of several picoseconds. However, for a longer relation time (τ 2 ) of several tens of picoseconds, the decay at the center (red-dashed circle in Figure S3d, Supporting Infromation) of c-MoS 2 is significantly longer than those at the vertices (pink-dashed circle in Figure S3d, Supporting Information). This nonuniform distribution is clearly absent in t-MoS 2 as shown in Figure S3b, Supporting Information.
Furthermore, the dynamics of charge carriers in WS 2 morphologies are examined. Similar to t-and c-MoS 2 samples, Figure 3a reveals a uniform micro-Raman mapping of the inplane E 1 2g mode of t-WS 2 flake. Raman spectra with the same intensity and peak position were also observed in both the E 1 2g and A 1g modes (see Figure S4, Supporting Information). Time-resolved ΔT/T was then applied to three locations of the t-WS 2 flake, and the measured data are depicted in Figure 3b. Before 4 ps delay time, ΔT/T exhibits a similar behavior compared to the data shown in Figure 1c from t-MoS 2 , a quick relaxation of negative ΔT/T at all locations. Interestingly, after 4 ps, a positive ΔT/T persists for relaxation time longer than 40 ps.
Contrary to the negative ΔT/T relaxation distribution observed in both t-MoS 2 and c-MoS 2 as expected from Equation (1), UPPM reveals a positive ΔT/T distribution in t-WS 2 after 3 ps with a slow relaxation as shown in Figure 3c. We attribute the sign change of ΔT/T t-WS 2 to the partially filled shallow defect states near conduction band minimum (CBM). A more detailed mechanism will be elaborated on later.
Lastly, Figure S5c, Supporting Information, shows a uniform micro-Raman mapped image of the in-plane E 1 2g mode of the h-WS 2 sample. Raman spectra of the in-plane E 1 2g and out-of-plane A 1g mode analyzed at different locations exhibit the same Raman peak positions and intensities as shown in Figure S5b, Supporting Information. The discrepancy in the threefold symmetry for h-WS 2 in our micro-Raman mapping and previous reports [12][13][14] could result from the low defect concentrations, leading to insufficient strain and therefore, hardly affects the Raman spectra at different domains, as revealed in new and aging WS 2 flakes. [39][40][41][42] As shown in Figure 4a, UPPM measurements reveal a threefold symmetry of negative (red) ΔT/T and zero (empty) ΔT/T in h-WS 2 before 5 ps, which correlates to the threefold symmetry visible in the optical image of the measured h-WS 2 flake (inset of Figure 4b).
We assigned the red regions as α domains and empty regions as β domains. [12][13][14] These domains were further revealed in ADF-STEM as shown in Figure 4c-e. Figure 4c displays the Our results are consistent with the previous h-WS 2 study, [14] which also uncovered the defect concentration/sizes with alternating α and β domains by STEM measurements. Upon relaxation after 7 ps, positive ΔT/T slowly emerges in α domains, similar to the time evolution in Figure 3c. It is worth noting that before 5 ps, the ΔT/T signal in β domains is too low, exhibiting an empty region in UPPM. Meanwhile, the noises of ΔT/T in β domains continuously emerge as the blue points in UPPM and require further examination to exclude these contributions. Therefore, three locations in the alternating bright (α) and dark (β) domains of the apparent three-fold symmetry were examined and ΔT/T plotted in Figure 4f. As shown in the inset of Figure 4b, ΔT/T at β domain shows no decrease or increase (location C) through pump-probe process. In α domains, in contrast, position-independent ΔT/T relaxation behavior was revealed (location A and B), similar to the ΔT/T results in Figure 3b, showing negative ΔT/T relaxation before 7 ps and a persistent positive ΔT/T over 40 ps. These results support our direct observation of the carrier dynamic distribution in alternating domains of h-WS 2 by UPPM.
In this section, we present a model to interpret the negative and positive transmittance changes in Figure 5. For all samples, the differential transmittance signal would be zero when only probe laser pulse with a photon energy of 1.55 eV is used (below the bandgap of materials), showing a constant transmittance depicted as t < 0 in Figure 5a,b. When the probe laser pulse is applied after a pump laser pulse with a photon energy of 3.1 eV, electrons in the valence band (VB) are excited to the CB, and the transmission of the material changes temporarily, as shown in t ¼ 0 in Figure 5a,b. The resulting change in transmission depends on the relaxation behavior of these photoexcited carriers.
In Figure 5a, for t-MoS 2 , the pump laser pulse excites electrons from the VB to the middle of the conduction band, filling the energy states near the CBM with relaxation time τ 0 . At the same time, electrons from the CBM relax to the deep defect states with relaxation time τ 1 or directly to the VB via phonon assist (or to deeper defect states) with long relaxation time τ 2 , during 0 < t < 2 ps. Because τ 0 is shorter than τ 1 , more excited electrons relax back to the CBM before they have a chance to relax into defect states. The accumulated electrons at CBM contribute to negative ΔT/T even after t > 5 ps. Based on this model, more defects exist at the edge and the vertex for c-MoS 2 , creating additional relaxation channels for electrons at CBM. As a result, τ 1,edge/vertex should be shorter than τ 1,center (still longer than τ 0 ), consistent with the slower ΔT/T recovery at the center (location C) observed in Figure 2b. Consequently, for both t-MoS 2 and c-MoS 2 , only negative transmittance recovery to the original transmittance is observed by UPPM shown in red to empty points with time evolution in Figure 2c.
In contrast, for t-WS 2 during 0 < t < 3-7 ps (defectdependent), photoexcited electrons relax to CBM with a relaxation time τ 0 while the electrons at CBM relax to the shallow defect states with shorter relaxation time τ 1 (τ 1 < τ 0 ) shown in www.advancedsciencenews.com www.adpr-journal.com Figure 5b. For the case of h-WS 2 , the τ 0 is 1.5 % 2.3 ps, which is much longer than 0.4 ps in MoS 2 . Due to this slow intra-band relaxation of the photoexcited electrons in CB and fast relaxation of the electrons from the CBM into shallow defect states, the unbalance (output rate > input rate) and missing CBM electrons contribute to a positive ΔT/T observed after 3-7 ps (defectdependent) in Figure 3b. It is important to note that because of more defects at the edges of t-WS 2 , CBM electrons at these locations relax to defects earlier than CBM electrons at the center, shown in red to blue points with time evolution by UPPM in Figure 3c. To provide a much clearer view of lifetimes between t-MoS 2 and t-WS 2 samples, the lifetime fitting and fitting parameters for all samples are provided (See Figure S6, Supporting Information). Applying the same absorption mechanism from t-WS 2 for h-WS 2, we describe the threefold symmetry revealed from UPPM as a result of pumping efficiency difference. In α (bright) domains, photo-excited electrons relax to shallow defects or the VB similarly as the photoexcited electrons relax in t-WS 2 as shown in Figure 5b; whereas in β (dark) domains, the pumping efficiency is too low because of the low absorption of visible light range in WS 2 with W-vacancy [14] and the limited amount of photo-excited electrons result in non-observable transmittance change as indicated in Figure 5c. Thus, we have developed a useful optical tool UPPM, and corresponding transient absorption mechanism to study and visualize carrier dynamic distribution for 2D TMDs morphology. For the future, carrier dynamic distribution of various shapes, grain boundaries, or lateral heterostructure studies on TMDs can be of great interest by applying UPPM.

Conclusion
In this work, we investigated the shape-dependent ultrafast dynamics of tailored CVD MoS 2 and WS 2 morphologies and showed that carriers in t-MoS 2 and t-WS 2 flakes behave homogenously in these two samples. A position-dependent relaxation behavior of photoexcited carriers is revealed in c-MoS 2 and h-WS 2, which we attribute to varying defect states in these morphologies. Additionally, with the same transmission probe photon energy, a negative/positive transmission change observed in MoS 2 /WS 2 sample is discussed with the existence of shallow defect states below CBM of WS 2. The faster relaxation of CBM electrons to shallow defect states than the relaxation of photoexcited electrons to CBM leads to the positive transmission change in WS 2 samples. Furthermore, a threefold symmetry of photoexcited carriers is observed in h-WS 2 flake, which we attributed to S-/W-vacancy defect states near the CBM, resulting in dissimilar pumping efficiency in the alternating domains. Our results provide a paradigm shift in device fabrication, such that, with the revealed carrier behaviors in MoS 2 and WS 2 samples, one can precisely choose TMDs of desired shapes and carrier dynamics for suitable optoelectronics.

Experimental Section
Controlled Growth of Triangular and Curved Triangular MoS 2 Flakes: Triangular and curved triangular MoS 2 monolayers were synthesized via chemical vapor deposition (CVD) under atmospheric conditions, using a two-zone furnace fitted with a quartz tube of 7 cm in diameter. Prior to growth, a pre-annealed double-side polished 10 Â 10 mm 2 sapphire was cleaned and used as the growth substrate. 0.002 M NaMoO 4 .2H 2 O as Mo source was coated on the pre-cleaned sapphire. The substrate was then placed at the center of the quartz tube and sufficient sulfur powder was placed upstream, 34 cm from the growth substrate. The quartz tube was then evacuated to 10 À6 torr and subsequently purged with N 2 gas until an atmospheric condition is attained. Next, the oven was heated up to 850 C in 10 min, and maintained at this temperature for 15 min. For the growth of MoS 2 flakes, Ar with a purity of 99.99% was used as the www.advancedsciencenews.com www.adpr-journal.com carrier gas with 60 standard cubic centimeters per minute (sccm) under an inert atmosphere. To obtain triangular MoS 2 , the growth substrate was placed face-up in a narrow crucible placed inside the 7 cm diameter quartz tube, to serve as a narrow reaction chamber. Curved triangular MoS 2 was grown by placing the coated sapphire substrate face-up on the back of a quartz boat in the 7 cm diameter quartz tube to serve as a wide reaction chamber. Controlled Growth of Triangular and Hexagonal WS 2 Flakes: Triangular and hexagonal WS 2 monolayers were synthesized via CVD under atmospheric condition, using a two-zone furnace fitted with a quartz tube of 5 cm in diameter. Prior to growth, a pre-annealed double-side polished 10 Â 10 mm sapphire was cleaned and used as the growth substrate. 0.008 M NaWO 4 .2H 2 O solution was mixed with 60% w/v solution of iodixanol in water (OptiPrep, density gradient medium (Sigma-Aldrich, D1556)) to serve as the W source and coated on the pre-cleaned sapphire. The coated substrate was then placed at the center of the quartz tube and sufficient sulfur powder was placed upstream, 16 cm from the growth substrate. The quartz tube was then evacuated to 10 À6 torr and subsequently purged with ultra-pure Argon gas till atmospheric pressure is attained. Next, the oven was heated up to 830 C in 10 min, and maintained at this temperature for 15 min. For the growth of WS 2 flakes, Ar with a purity of 99.99% was used as the carrier gas with 60 sccm under an inert atmosphere. To obtain triangular WS 2 , the growth substrate was placed face-up in a narrow crucible placed inside the 5 cm diameter quartz tube, to serve as a narrow reaction chamber. Hexagonal WS 2 was grown by placing the coated sapphire substrate face-up on the back of a quartz boat in the 5 cm diameter quartz tube to serve as a wide reaction chamber.
Micro-Raman Scattering: Raman scattering was performed with a 442 nm Ar-Ion laser focused on the sample using a 100x long-range objective, resulting in a laser spot size of 1 μm 2 at the sample surface. The Raman spectra were obtained with a nitrogen-cooled charge-coupled device (CCD) camera and the spectrum resolution is 0.4 cm À1 .
Optical UPPM: The light source is a Ti:sapphire laser (Femtosource scientific XL300, Femtolaser) with a central wavelength of 800 nm, a repetition rate of 5.2 MHz, and a pulse width of 100 fs. For the pump pulses, a β-BaB 2 O 4 nonlinear crystal is used to convert the wavelength from 800 nm to 400 nm after an acousto-optic modulator, which modulates the pump beam for the reference frequency of a lock-in amplifier. The polarization of the pump and probe pulses were set to be perpendicular to eliminate the coherent spike around the zero-time delay. [43] Finally, the pump and probe pulses are combined by a dichroic beam splitter and collimated to a microscope system. For the spatial-resolved ultrafast carrier dynamics, a scanning microscope combined with the pump-probe system is adopted to map the MoS 2 or WS 2 . Both pump and probe pulses are focused on the sample by an objective after a galvo mirror, which serves the function of spatial mapping. The spatial resolution is about 1.5 μm, and the typical pump and probe fluences amount to 266 and 165 μJ cm À2 , respectively. By controlling the time delay between pump and probe pulses, the positiondependent transient transmittance changes (ΔT/T ) can be obtained.
ADF-STEM Sample Preparation and Characterization: The WS 2 was grown by CVD on a sapphire substrate. The substrate was spin-coated by PMMA with 3000 rpm for 1 minute. The edges of the substrate were scratched using a razor blade to expose the sapphire surface, allowing faster etching. Then, the substrate was floated up onto sodium hydroxide solution with the volume ratio 1:10, NaOH and H 2 O respectively. The sapphire substrate was immersed into the solution after being etched off by the solution and the PMMA film with WS 2 flakes stayed onto the solution. The film was transferred to de-ionized (DI) water 3-4 times to clean the residual solution, followed by being fished out by a TEM grid and air-dried. The TEM grid was annealed in a CVD tube furnace under a mixture of 90% N 2 % and 10% H 2 at 200 sccm to remove the PMMA. The temperature was increased up to 40 C with the rate of 1 C min À1 , held at 40 C for 2 h, and cooled down to room temperature. All ADF-STEM images were taken in a Thermo Fisher Titan Themis operated at 300 kV.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.