Low in-plane symmetry two-dimensional (2D) materials have attracted great interests in recent years due to their unique optoelectronic properties, such as valley-selective optical coupling and strong nonlinear optical responses [1–3]. This class of material has recently been verified to exhibit anisotropic electronic transport properties and remarkable light–matter interactions, making them a promising candidate for a wide range of applications, such as 2D field-effect transistors (2D-FETs) and photodetectors [2,4–6]. As a member of low-symmetry semimetallic 2D materials, Ta2NiSe5 has a direct band structure with a narrow band gap of 0.36 eV, thereby facilitating the transport of electrons within the band structure [7–10]. Besides, Ta2NiSe5 has a monoclinic structure stacked by weak van der Waals interactions, where the underlying one-dimensional [TaSe6]2 chains exhibit a significant in-plane anisotropy ratio [11,12]. These characteristics show Ta2NiSe5 can enhance the performance of different nano-devices owing to their better carrier transport capability and photoelectric conversion efficiency. However, some important optoelectronic properties, such as anisotropic photo-excited carrier behaviors, of Ta2NiSe5 have yet to be totally investigated, which is critical for developing new electronic devices with higher efficiency and improved performance.

Previously, Ta2NiSe5-based FETs have been reported to show a field-effective mobility of 161cm2V−1s−1 along the armchair direction. A polarization-dependent Raman spectroscopy study showed the anisotropic optical responses [13]. However, these methods can only study the steady state optoelectronic properties rather than revealing the evolution of carrier transportation with high spatial resolution (<1μm) and temporal resolution (<1ns). Besides, macroscopical sample defects and impurities may result in a significant discrepancy between the device mobility and the true value of the material [14–16]. Pump-probe microscopy is a powerful tool to monitor ultrafast processes with temporal resolution down to tens of femtoseconds. It also has a diffraction-limited spatial resolution of several hundred nanometers, which makes it possible to characterize the transient carrier behaviors of Ta2NiSe5.

Regarding carrier dynamics, Zhao et al. conducted studies using pump-probe microscopy and confirmed the carrier lifetime of black phosphorus is ∼100ps after 20 ps, while it can reach ∼180ps for MoS2 [17,18]. Yue et al. experimentally studied the semiconducting cubic boron arsenide and determined an ambipolar diffusion coefficient of 39cm2s−1 [19]. These studies show that pump-probe microscopy is an effective tool for monitoring carrier diffusion processes for 2D materials and semiconductors with a fixed scanning trajectory. However, it is still challenging to acquire in situ anisotropic carrier diffusion coefficients via a conventional pump-probe microscope that uses a scanning mirror to steer the laser, since the probe beam needs to precisely follow designed scanning paths (e.g., armchair and zigzag directions). Although rotating the specimen to specific angles or inducing a pair of scanning mirrors may partially address the issue, the method is time-consuming, and motion errors may be introduced during repeated operations. To address this issue, we developed a spatially resolved pump-probe microscope based on a random-access digital micromirror device (DMD) scanner, where the 3D position of the laser focus can be changed by directly switching holograms on the DMD chip, and the trajectories of the probe beam can be arbitrarily programmed based on binary holography, making it possible to perform in situ anisotropic carrier diffusion characterization more flexibly [20].

In this work, we study the dynamics of photocarriers in multi-layer Ta2NiSe5 using a DMD-based pump-probe microscope. To track the anisotropic photocarrier diffusion process, we designed two sets of holograms to command the DMD for scanning the specimen along the armchair and zigzag directions, respectively. By analyzing the evolution of photocarriers, we experimentally measured the transport performance and in-plane anisotropy of Ta2NiSe5, which showed rapid carrier diffusion behaviors in the first 5 ps and an abrupt decline after that. These results confirm the superior capability of DMD-based pump-probe microscopy to investigate ultrafast dynamic phenomena of optoelectronic materials and devices.

Ta2NiSe5 has a monoclinic crystal structure stacked by van der Waals interaction along the b axis [Fig. 1(a)], where the Ta double chain and Ni single chain are parallelly aligned with the a axis (armchair direction) and are alternately arrayed along the c axis (zigzag direction). The in-plane anisotropic structures are formed via the periodic connection of TaSe6 octahedral chains and NiSe4 chains along the zigzag direction [13]. High-resolution transmission electron microscopy (TEM) and selected area electron diffraction pattern measurements showed that the prepared specimen has well-defined monoclinic crystal structures, confirming the specimen is suitable for the planned ultrafast experiments [Fig. 1(d)].

In the transient absorption measurements, the multi-layer Ta2NiSe5 specimen was attached to a quartz substrate via Scotch tape. A 400-nm (3.10 eV) pump laser with a pulse duration of 200 fs was focused on the specimen, and the photon energy is high enough to generate free carriers [21,22]. The probe beam, which was controlled by the DMD, was an 800-nm (1.55 eV) laser with a pulse duration of 200 fs, and the polarization is determined by the rotation angle of the half-wave plate, as shown in Figs. 1(b) and 1(c). Both the pump and probe beams were directed to the specimen, where the probe beam (∼1.6μm) was set to be slightly larger than the pump beam (∼0.8μm). Detailed DMD-based pump-probe microscopy optical configuration is presented in Appendix A.

To study the dynamics of the induced photocarriers at the picosecond time scale, we measured the differential reflection signals by probing the sample with a designed time delay, Δt, after the pump beam excited the sample [23]. The differential reflection signals, i.e., ΔR/R, are defined as the relative intensity changes between the initial and reflected probe beams. The results are presented in Fig. 2(a), which show that the differential reflection signals peak at around 0 ps and then decay in two distinctive ways, reflecting the different carrier diffusion behaviors. First, from 0 to 5 ps, the signals exponentially decay with a time constant of ∼1ps; after that the signals enter a slow decaying stage with a time constant of ∼150ps. (See Appendix C for more details.) To validate that the measured time constants are intrinsic to the material, the experiments were repeated six times using different pump laser powers (50–300 μW), and the probe power was 100 μW (pulse fluence, 4μJcm−2). The results are plotted in Fig. 2(b), where the red dots, which represent the peaks of |ΔR/R|, are proportional to the pump power, and the cyan dots, which represent the time constant, are independent of the pump power at ∼1ps in the first 5 ps. This result provides an experimental basis for the anisotropic carrier dynamics studies in the following sections as it indicates that the dynamic processes of carriers are intrinsic physical properties of Ta2NiSe5 and will not be influenced by the densities of the induced electrons or carriers. To ensure a high signal-to-noise ratio and avoid sample damage, the power of the pump beam was set at 200 μW in the following experiments (pulse fluence, ∼80μJcm−2).

Conventionally, the anisotropic optical response of Ta2NiSe5 has been studied by polarization-dependent Raman spectroscopy [13]. Here, we further applied our DMD-based pump-probe microscope to monitor ΔR/R variations using the probe beam of different polarizations. Figure 3(a) presents the contour color map of the polarization-resolved (θ=0°−180°) differential reflection as a function of the probe beam time delay. We selected four polarization angles, i.e., 0°, 30°, 60°, and 90°, from Fig. 3(a) and found that the curves of different polarizations share the same decaying trend [Fig. 3(b)]. The 0° polarization corresponds to the armchair direction of Ta2NiSe5, which shows the capability of this method to identify the crystalline direction of the flake [13,24]. Figure 3(c) shows that for the same time delay (i.e., 0 or 10 ps), |ΔR/R| intensities exhibit periodic variations with a period of 180° and display a symmetrical trend along the zigzag (90°, 270°) direction. The measured angle-dependent data are in good agreement with previous studies, which are fitted by the red solid and dashed lines [13,25,26]. Considering the polarization-resolved carrier lifetime [Fig. 3(d)], we can confirm that the carrier lifetime is ∼1ps and is independent of the probe beam’s polarization. This can be attributed to the fact that the polarization of the probe beam mainly affects the probing efficiency, rather than the evolution of carrier dynamics [17,24].

The above analyses present initial results of the photocarrier dynamics with fixed pump and probe beams. In this section, we present the characterization of the anisotropic carrier coefficient of Ta2NiSe5. In the spatiotemporal dynamic measurements, the 400-nm pump beam was first focused on the specimen to generate photocarriers with high density. After that, the photocarriers rapidly diffused within the flake. The diffusion process was monitored by reflectivity variations over time across the sample via the probe beam. The fast sample scanning was realized by the DMD-scanner. By uploading designed holograms, the DMD can control the probe beam to scan over the pump beam along the armchair and zigzag directions, respectively, as shown in Fig. 1(e) [27–29]. Figure 4(a) presents the measured spatiotemporal differential reflectivity along the armchair direction, and the probe beam polarization was θ=0°. By selecting four different time delays (i.e., 0, 1, 10, and 20 ps), we can obtain |ΔR/R| as a function of the probe position [Fig. 4(b)], where the solid lines show curves fitted by a Gaussian function (R2∼0.98).

A prerequisite for determining carrier diffusion coefficients is to confirm that the obtained reflectivity profiles broadening at different time delays are indeed caused by carrier diffusion, rather than other factors (e.g., photocarrier recombination). Figure 4(c) presents the differential reflectivity at four selected probe positions (i.e., 0, 2, 3, and 5 μm) as a function of the probe beam time delay. From the results, one may observe that all dynamics of photocarriers at different distances from the pump beam center decayed exponentially with a time constant of ∼1ps regardless of the detection position. This demonstrates that the broadening of the reflectivity profiles is solely attributed to carrier diffusion instead of photocarrier recombination, in which case the time constant will vary with the detection position [17]. Figure 4(d) presents the FWHM2 as a function of the time delay, where the FWHM2 values are obtained from fitting the data in Fig. 4(a) with a Gaussian function. From the results, one may observe that the time delay curves increased bi-linearly, where the transition occurred at approximately 5 ps. Specifically, the FWHM2 value increases rapidly in the first 5 ps, followed by a slower increase.

Based on the FWHM2 values in Fig. 4(d), we next calculate the carrier diffusion coefficient via the carrier diffusion model, expressed in Eqs. (1) and (2) [30]: FWHM2(t)=FWHM2(t0)+16ln(2)D(t−t0),(1)dFWHM2(t)dt=16ln(2)D,(2)where D is the carrier diffusion coefficient. From the red line in Fig. 4(d), we can extract two slopes of 0.55μm2ps−1 and 0.06μm2ps−1, which correspond to the fast and slow diffusion coefficients of 500±30cm2s−1 (Δt=0−5ps) and 56±8cm2s−1 (Δt>5ps) calculated by Eq. (2), respectively. Similarly, from the carrier diffusion measurement along the zigzag direction [blue lines in Fig. 4(d)], we can calculate the fast and slow diffusion coefficients as 100±10cm2s−1 (Δt=0−5ps) and 35±5cm2s−1 (Δt>5ps), respectively. It is worth noting that during the first 5 ps period, the carrier diffusion coefficient in the armchair direction is about five times higher than that in the zigzag direction, which confirms the strong anisotropic carrier mobility characteristic of Ta2NiSe5. The corresponding photocarrier diffusion map along the zigzag direction is presented in Appendix E.

From the experiments in Figs. 2(a) and 3(b), two distinctively different dynamic behaviors of the photocarriers have been observed; i.e., when Δt<5ps, the measured |ΔR/R| decayed rapidly, and when Δt>5ps, the measured |ΔR/R| decayed much slower. Similarly, in the transition stage at ∼5ps, an abrupt slope decrease of FWHM2(t) in Fig. 4(d) was observed. These behaviors are attributed to the interplay between the hot electron relaxation and phonon scattering effects. In other words, the high-energy carriers initially achieved a much higher transportation speed upon the laser excitation, gradually replaced by the vibration of phonons due to the dissipation of energy [31]. (See Appendix C, Fig. 8, for frequency analyses of in situ transient absorption measurements that confirm the presence of the phonon scattering effect [32].) Experiments and simulations also show that the nickel films in the 0–10 ps have similar transient dynamics compared to Ta2NiSe5 [33]. Based on these results, we can deduce that the metal atoms in Ta2NiSe5 play a crucial role in the rapid relaxation of hot carriers, which results in a small transition time. The alternating crystal structure among Ta, Ni, and Se atoms contributes to the pronounced anisotropy, as shown in the transient polarization-resolved experiments and carrier diffusion measurements along the armchair and zigzag directions.

From the inset in Fig. 4(d), one may observe that the hot electron diffusion coefficient is independent of the pump power (as it increases from 100 to 300 μW). This is because the initial coefficient is mainly influenced by the ratio of the electron thermal conductivity (ke) and heat capacity (Ce) instead of the pump fluence or carrier density [19,21,34]. As ke of Ta2NiSe5 was reported to have values of 20.5Wm−1K−1 and 4.5Wm−1K−1 along the armchair and zigzag directions, respectively, and Ce remains constant [35], the diffusion coefficient ratio between the armchair and zigzag directions is calculated to be ∼4.6, which agrees with our experimental result (i.e., diffusion ratio=5 in 0–5 ps).

Lastly, we studied the 2D photocarrier diffusion evolution. The results are presented in Fig. 5. In the experiments, the DMD commanded the probe beam to perform random-access scanning (to avoid damage on the sample) over an area of 16μm×16μm with a step size of 400 nm, while the pump beam remained stationary, and Figs. 5(a)–5(c) plot the measured |ΔR/R| as a function of the probe time delay (i.e., 0, 5, and 10 ps). (Note that the intensity gradient is plotted logarithmically for better displaying the low-intensity components in the image.) The ovality (O) of the intensity profiles in Figs. 5(a)–5(c) was calculated based on the gray dashed lines plotted by connecting the data points with an intensity value of ∼1.8×10−5. From the results, one may observe that for a given polarization (i.e., 0°), the peaked intensity decreases rapidly from 0 ps to 5 ps and decreases slowly from 5 ps to 10 ps. This is consistent with the results presented in Fig. 2(a). Note that the ovality in Figs. 5(a)–5(c) increases from 0.6% to 4%, where the major axis of the ellipse is aligned with the armchair direction. This is because the excited photocarriers travel faster in the armchair direction than in the zigzag direction, which well shows the in-plane anisotropic characteristics of Ta2NiSe5. Figure 5(d) presents a Gaussian function-fitted profile of the intensity data in Fig. 5(a) (R2∼0.98).

In this work, we investigated the ultrafast dynamics of photo-excited carriers in a multi-layer Ta2NiSe5 specimen via a DMD-based pump-probe microscope. Specifically, the evolution of the carrier diffusion was studied, which shows two distinctive dynamic processes, i.e., a rapid diffusion process dominated by the hot electrons relaxation and a slow diffusion process dominated by the phonon scattering. The polarization-dependent transient absorption measurements are consistent with the trends of polarized Raman spectra. The associated carrier diffusion coefficients along the armchair and zigzag crystalline directions were measured to be ∼500cm2s−1 and ∼100cm2s−1 in the first 5 ps, respectively. The obtained diffusion coefficient variations demonstrate semimetallic properties of Ta2NiSe5, which supports the development of Ta2NiSe5-based optoelectrical devices. Compared with the conventional Raman spectroscopy and FET-based techniques, the DMD-based system offers a sub-picosecond resolution for studying different ultrafast dynamics. Additionally, the DMD platform can be readily applied to study other ultrafast phenomena in the micro- and nano-scale optoelectronic devices, e.g., photonic chips and solar cells, enhancing the fundamental understanding of different 2D materials, perovskites, nanocrystals, and so on.