Controlling the surface photovoltage on WSe₂ by surface chemical modification

Crystals of transition metal dichalcogenides (TMDCs) are composed of two-dimensional (2D) layers that are bonded with neighbors by the weak van der Waals (vdW) interaction. Such 2D layered materials have been identified as promising components for vdW heterostructure devices.¹ The concept of the vdW heterostructure has been further extended from TMDC layers to metal/TMDC or organic molecules/TMDC layers.² WSe₂, one of the TMDC semiconductors, is known to exhibit a large surface photovoltage (SPV) effect,³ and it has been reported that an even higher SPV is induced by deposition of metals (Rb and In).^4,5 On the other hand, an organic/TMDC heterojunction (C₆₀/MoS₂) has been predicted to show a high quantum yield and a high photovoltaic efficiency for solar cell applications.² Concerning opto-electronic devices, the most fundamental optical response is the generation of photovoltage by the spatial separation of the photo-excited electrons and holes that are created at the surface or the interface region. To make further progress, it is important to perform a detailed study of the surface/interface electronic states and its carrier dynamics.

In this study, time-resolved x-ray photoemission spectroscopy (TRXPS) experiments were performed on WSe₂ to investigate carrier dynamics in relation to the SPV effect. The same measurements were carried out for the surfaces after depositions of potassium (K) atoms or fullerene (C₆₀) molecules. The resulting photoemission band diagrams allow us to draw conclusions about the electronic states and the carrier dynamics at the pristine and modified surfaces, and the derived basic knowledge will be useful for developing TMDC heterojunctions. Figures 1(a) and 1(b) show schematic drawings of the atomic structure. The clean surface of WSe₂ was obtained by cleavage in an ultrahigh vacuum chamber. A metal/WSe₂ surface was prepared by depositing ∼1 ML potassium on the clean surface with a SAES Getters dispenser at room temperature (RT) while an organic molecule/WSe₂ surface was assembled by depositing ∼0.8 ML fullerene on it with a K-cell evaporator at RT. The coverage calibration is explained in the supplementary material.

Photoemission band mapping was performed by angle-resolved photoemission spectroscopy (ARPES) using synchrotron radiation (SR) at beamline 21B of Taiwan Light Source, National Synchrotron Radiation Research Center. The spectra were recorded using a hemispherical analyzer with an energy resolution of 50 meV at photon energies hν = 28 and 42 eV. TRXPS measurements were carried out at BL07LSU at SPring-8 by the pump-probe method.^6,7 The pump light was a laser pulse of hν = 1.55 eV with a pulse duration of 60 fs and a pulse interval of 4.79 μs. The probe light was provided by SR pulses of hν = 253 eV that were generated by F-mode operation of the storage ring with a pulse duration of 50 ps. TRXPS data were obtained using a time-of-flight analyzer. All the photoemission measurements were performed at RT.

Figures 1(d)–1(i) show the changes of core level spectra after surface modifications. After K deposition, both W 4f and Se 3d spectra indicate an energy shift of 0.18 eV toward higher binding energies as shown in Figs. 1(d) and 1(e), respectively. As shown in Fig. 1(f), the appearance of the K 3p peak confirms the presence of the K overlayer. Figures 1(g)–1(i) show the results for C₆₀ deposition. The W 4f and Se 3d core levels show an energy shift of 0.08 eV to the lower binding energies, while the C 1s core-level confirms the presence of C₆₀ adsorbates. Since photoemission spectroscopy probes the near-surface region (hν = 253 eV), the energy shifts of the core levels correspond to variations of the band-bending effect by adsorbates^2,10,11 and indicate the opposite roles of the two overlayers in modifying bulk states near the surface region or the space-charge layer (SCL).

Figures 2(a)–2(f) show ARPES maps of the valence bands of (a) the pristine WSe₂ surface, (b) the K-covered WSe₂ surface (K/WSe₂), and (c) the C₆₀-covered WSe₂ surface (C₆₀/WSe₂). In Fig. 2(a), dispersion curves of the two bulk valence bands, labeled as VB1 and VB2, reach $K¯$ at E − E_F of −1.26 eV and −1.75 eV, respectively. The valence band maximum (VBM) of the clean WSe₂ surface is thus located at E − E_F = −1.26 eV. Since the bulk bandgap of WSe₂ is 1.3 eV,¹⁵ the Fermi level is located very close to the conduction band minimum (CBM), and the surface has an n-type band structure. On the other hand, the VBM within the interior of the sample, VBM_bulk, is evaluated to be E − E_F = –0.22 eV (p-type, see supplementary material for further details) for the current sample; thus, there is a built-in surface band bending of V_s = 1.04 eV in the sample that gives rise to a space-charge layer or a natural p-n junction [see Fig. 1(c)]. With K deposition, VB1 and VB2 shift to higher binding energies by 0.18 eV, as evaluated from the results in Figs. 2(a) and 2(b). This change is consistent with the shift seen in Figs. 1(d) and 1(e). On the other hand, deposition of C₆₀ causes the VBM to move toward a lower binding energy, as seen in Figs. 1(g) and 1(h). The spectral features of VB1 and VB2 appeared faint in Fig. 2(c) due to the C₆₀ overlayer, which strongly attenuates the photoemission signal from the substrate at the photon energy of 28 eV used for the measurement.

The observed p-type band bending of the pristine WSe₂ surface counters our expectation that a pure crystal of WSe₂ should exhibit flat bands due to the absence of intrinsic surface states on the prefect vdW crystal surface.^11,12 However, previous ARPES studies have also revealed similar surface band bending effects. It has been argued that the actual surfaces might have various types of defects, such as atomic steps or dislocations.³ Buck et al.⁴ reported that these factors induce the large band-bending effect and the SCL width of WSe₂ extends to ∼120 nm. The WSe₂ SCL has now attracted much interest as a carrier reservoir for photovoltaics.

Focusing on the photoemission map of K/WSe₂ near the Fermi level [Fig. 2(e)], an apparent spot at k_ǁ = 0.65 Å⁻¹ and a broad dispersive feature are observed in the spectra. The former can be assigned to the CBM located at $T¯$ [Fig. 2(d)] of WSe₂, while the latter likely arises from a K-derived metallic band. The K-derived electronic state can be regarded as a quantum-well state (QWS)¹³ in the overlayer and gives rise to a parabolic dispersion near the Brillouin zone center [Fig. 2(e)]. This assignment is consistent with previous studies¹³ and the fact that no K intercalation appeared between the WSe₂ layers⁸ [see Fig. 1(f)]. The K overlayer behaves as an electron donor to the WSe₂ layer.

For C₆₀/WSe₂, the highest occupied molecular orbital (HOMO) and the next highest occupied molecular orbital (NHOMO) of the fullerene are observed at E − E_F = −2.62 and −4.06 eV, respectively, in the ARPES spectrum [Fig. 2(f)]. The LUMO position of the C₆₀ molecule is estimated to be around −1.02 eV.¹⁴ The spectral feature of LUMO can be specified from the gentle slope, as shown in the inset of Fig. 2(f). Comparing the clean surface data shown in Fig. 2(a) with the C₆₀ deposited surface data shown in Fig. 2(c), the LUMO level is located within the bandgap of WSe₂ and the energy position of HOMO is lower than the VBM of WSe₂. It is likely that the LUMO orbital is responsible for the acceptor-type⁹ character of the C₆₀ overlayer on WSe₂. The LUMO position is above the VBM but below E_F, and this is supported by a previous model suggested by a photoluminescence study.⁹ 

To probe the dynamical behavior of the WSe₂, K/WSe₂, and C₆₀/WSe₂ surfaces, we carried out TRXPS measurements of the W 4f core level by the pump-probe method. In the experiment, electron carriers are photo-excited by the ultrafast laser pulses of hν = 1.55 eV that is larger than the indirect bandgap (1.3 eV).^10,15 Figure 3(a) shows the energy shifts of the core-level peaks by the SPV effect with respect to the power density of the pump pulses. At a pump power density of 1 mJ/cm²/pulse, the measured photovoltage is about 0.15 V for both K/WSe₂ and WSe₂ and about 0.05 V for C₆₀/WSe₂. Figure 3(b) shows the W4f_7/2 XPS spectra of K/WSe₂ at different delay times. The time evolution of the SPV is obtained from the energy difference between the peak position as a function of the delay time and the peak position measured at equilibrium (without pump laser). Figure 3(c) displays SPV as a function of delay times for WSe₂, K/WSe₂, and C₆₀/WSe₂. Evidently, the relaxation time for the SPV is different for the three cases, suggesting differences in the dynamical behavior.^16,17

Conventionally, SPV relaxation can be described by the thermionic emission model^18–20 if thermionic emission dominates the carrier recombination process. The illumination power dependence and the temporal relaxation derived from the rate equation of hole (electron) charge density and the relation between SPV and hole (electron) density are as follows:

where k_B, T, and V₀ are the Boltzmann constant, the sample temperature, and the SPV shift at t = 0. I and t are the fluence of the pump laser and the delay time between pump and probe pulses. Parameters η, γ, and τ_∞, are the ideality factor, the efficiency of the optical excitation, and the relaxation time of SPV. The data points in Figs. 3(a) and 3(c) are well fitted by Eqs. (1) and (2). Table I shows a summary of the fitting parameters for WSe₂, K/WSe₂, and C₆₀/WSe₂. The pre-factor of η was around 2 for the three surfaces. The quantity η is related to a ratio of thermal cross-section for electrons and holes, and it typically ranges between 0.5 and 2, indicating appropriateness of the model.¹⁹ The relaxation time, τ_∞, of WSe₂ was longer than those of K/WSe₂ and C₆₀/WSe₂, suggesting a faster electron-hole pair recombination rate caused by the overlayers.

Electronic states at surfaces play various roles in controlling the performance of photovoltaic materials. One is to change the degree of band bending that is directly linked to the spatial separation of the photo-excited carriers or generation of photovoltage. The other is to regulate the recombination process of the photo-excited electron-hole pairs. A heterojunction of WSe₂ with the donor-type (K) overlayer enhances the surface potential such that SPV also increases relative to that of the pristine WSe₂ surface at lower pumping laser power density. This may be attributed to the fact that the space charge layer near the surface now includes a region with electron occupation in the conduction band. The electrostatic potential becomes constant within this region, and there is no drift force within to drive the photo-exited carriers. Concerning the relaxation time, τ_∞ was suppressed after K deposition. When the WSe₂ surface makes contact with the acceptor-type (C₆₀) overlayer, the surface potential and the SPV are both reduced, as shown in Figs. 1(g), 1(h), and 3. Furthermore, the relaxation time becomes short, compared to the pristine WSe₂ surface.

In order to describe the different relaxation times by the surface modification, we recall the surface recombination model that approximates the SPV relaxation time as^19,20 $τ∞=τs∼WS$⁠, where τ_s, W, and S are the surface recombination time, width, and surface recombination velocity, respectively. The SCL widths for our WSe₂, K/WSe₂, and C₆₀/WSe₂ are evaluated to be about 120 nm from the value of the band bending. Thus, variations in W are not the key factor of the significant changes in relaxation time. On the other hand, the surface recombination velocity can vary with the trapped states at a surface having the 2D density. The surface recombination velocity can be written as^21,22 $S=Ntvthσp e−eVs/ηkBT,$ where N_t, v_th, σ_p, and V_s are the trapped carrier density near surfaces, the thermal carrier velocity, the hole carrier capture cross-section, and the surface potential, respectively. The parameters N_t, σ_p, and V_s can vary with surface modifications, but v_th should remain as a constant. In this study, V_s changes within ±20% in three of the surfaces (see Table I); thus, the key factors are the density of trapped states and the hole carrier capture cross-section. It is inferred that the QWS and the occupied bulk state near CBM at the K/WSe₂ surface, as observed in Fig. 2(e), may become recombination centers, leading to an increase in N_t or σ_p. In the K/WSe₂ case, one may conclude that the increase in charges (states) at a surface enhances the band-bending effect but it may also enhance the electron-hole pair recombination rate. Concerning the case of C₆₀/WSe₂, the system also has shorter relaxation time than that of the pristine WSe₂ surface. It can be explained by the decrease in V_s [Fig. 1(g)] and the increase in N_t or σ_p [Fig. 2(f)]. The LUMO states located within the bulk bandgap of WSe₂ likely dominate the recombination of the photo-excited electron-hole pairs.

Photoemission band mapping and time-resolved photoelectron spectroscopy were systematically performed on surfaces of a p-type WSe₂ crystal and the modified surfaces with two types of adsorbates, K and C₆₀. For the K/WSe₂ surface, the potassium layer acts as a donor and it enhances the band-bending effect, while the fullerene layer at C₆₀/WSe₂ behaves as an acceptor and it reduces the effect. Transient surface photovoltage effects were observed for the three surfaces. The larger SPV is observed on K/WSe₂ due to the larger initial band bending. The shorter SPV relaxation time after the formation of the overlayers indicates that adsorbate-induced electronic states likely become recombination centers of photo-excited carriers. This system offers a good model to illustrate a trade-off between the magnitude of the SPV and the SPV relaxation rate and the multi-parameter dependencies of SPV generation and relaxation. These observations should provide critical insights into designing TMDC-based optoelectronic devices in the future.

See supplementary material for the coverage estimation of C₆₀ and K at the WSe₂ surface and the band bending profile and the space charge layer of p-type WSe₂.

This work was supported by the Grant-in-Aid for Scientific Research (KAKENHI) (Grant Nos. 16H03867 and 16H06027) and the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Division of Materials Science and Engineering (Grant No. DE-FG02-07ER46383 for T.-C.C.). The experiment was carried out as a joint research in the Synchrotron Radiation Research Organization and the Institute for Solid State Physics, the University of Tokyo (Proposal Nos. 2013B7454, 2014B7480, 2016A7503, and 2016B7523). T.-C.C. received a visiting professor appointment at the University of Tokyo during the experiment.