Interface Engineering Modulated Valley Polarization in MoS₂/hBN Heterostructure

Abstract

Layered transition metal dichalcogenides (TMDs) provide a favorable research platform for the advancement of spintronics and valleytronics because of their unique spin-valley coupling effect, which is attributed to the absence of inversion symmetry coupled with the presence of time-reversal symmetry. To maneuver the valley pseudospin efficiently is of great importance for the fabrication of conceptual devices in microelectronics. Here, we propose a straightforward way to modulate valley pseudospin with interface engineering. An underlying negative correlation between the quantum yield of photoluminescence and the degree of valley polarization was discovered. Enhanced luminous intensities were observed in the MoS₂/hBN heterostructure but with a low value of valley polarization, which was in stark contrast to those observed in the MoS₂/SiO₂ heterostructure. Based on the steady-state and time-resolved optical measurements, we reveal the correlation between exciton lifetime, luminous efficiency, and valley polarization. Our results emphasize the significance of interface engineering for tailoring valley pseudospin in two-dimensional systems and probably advance the progression of the conceptual devices based on TMDs in spintronics and valleytronics.

1. Introduction

Two-dimensional (2D) polarized materials, including ferromagnets [1], ferroelectrics [2], and ferrovalley materials [3,4], demonstrate peculiar behaviors at the quantum realm. Valley pseudospin, which represents the energy band extremes in momentum space, normally exists in periodic solid materials [5,6]. The addressability of valley pseudospin enables the utilization of the momentum states of carriers as a brand-new paradigm in data coding and information handling. Using the research strategies of spintronics [7,8] for reference, a similar concept, valleytronics, arose naturally and vigorously [9]. The novel scientific connotation associated with the manipulation of the valley degree of freedom may result in a transformative impact. Referring to theoretical predictions about the intrinsic properties closely related to the valley pseudospin, rapid experimental advances [10,11,12,13,14] have been performed to observe and manipulate the valley polarization in a way similar to real spin.

Recently, the successful isolation and further experimental characterizations of 2D materials, including but not limited to graphene, hexagonal boron nitride (hBN), and TMDs, enriched our cognition of valley physics [15]. The spatial symmetry breaking along with the time-reversal symmetry enable two sets of the individually addressable valleys, K and K’ points in the first Brillouin zone, for TMDs [16]. Especially when the layered TMDs, such as MoS₂, WS₂, and MoSe₂, are mechanically exfoliated from bulk-phase crystals and thinned down to monolayers, a marvelous transition in their electronic structure occurs, viz the evolution from an indirect bandgap to a direct one. The direct bandgaps of monolayer TMDs normally lie in the near-infrared and visible spectral ranges of approximately 1~2 eV, suitable for the investigation of valley pseudospin-related optoelectronic applications. Van der Waals (vdW) heterostructures composed of 2D materials offer a fascinating research platform for tailoring artificial composite constructions with unique properties, novel phenomena [17,18], and widespread potential applications [19,20,21]. The concept of engineering material properties via fabricating mixed-dimensional vdW heterostructures can also be employed to manipulate valley polarization with great convenience and low cost for spintronics and valleytronics [22,23].

Here, we performed circularly polarization-dependent photoluminescence (PL) measurements upon MoS₂ monolayers transferred atop hBN nanoflakes and SiO₂/Si wafers at room temperature, prepared with the chemical vapor deposition (CVD) method. Substantial variations were observed in these two vertically stacked systems, MoS₂/hBN and MoS₂/SiO₂, especially in their luminous performance. An underlying opposite relationship between the intensity of luminescence and the degree of valley polarization was observed. Specifically, a stronger exciton luminescence was observed in the MoS₂/hBN heterostructure but with a lower degree of valley-polarized emission, while a weaker luminous performance and a higher degree of valley polarization were obtained in the MoS₂/SiO₂ heterostructure. We infer that there exists a connection to be examined between the luminous efficiency and the valley polarization. According to the time-resolved circularly polarization-dependent PL measurements, a lower degree of valley polarization is observed in samples that exhibit longer exciton lifetimes. Our work reveals an inverse relationship between luminescence efficiency and the valley polarization value, and emphasizes the significance of interface engineering for modulating the intrinsic new degrees of freedom in ultrathin 2D polarized materials. It may potentially deepen our understanding of 2D quantum systems and advance the realization of the emerging practical applications for next-generation information storage and processing.

2. Materials and Methods

Large-area, continuous MoS₂ monolayers with a lateral size of several millimeters were routinely prepared with a standardized CVD process in a commercial dual-temperature-zone furnace [24,25] using high-purity S and MoO₃ powders as the solid-state powder sources. Mechanically exfoliated hBN flakes were transferred atop SiO₂/Si wafers with a 285 nm oxidation layer before CVD growth. Highly purified argon was employed as the transport gas. The growth temperature and the required time parameters we used for the CVD process are shown in the inset of Figure 1a. The programming temperatures for the two respective zones, in which the S and MoO₃ powders were placed, were elevated from room temperature to 160 and 650 °C in 40 min and remained unchanged for 5 min. Once the reaction and deposition processes finished, it was cooled naturally. Generally speaking, the typical growth of MoS₂ crystals atop SiO₂/Si normally results in equilateral triangle shapes under thermodynamically stable conditions. As shown in the optical microscopy image, the MoS₂ samples were successfully deposited onto the 285 nm SiO₂/Si wafer with some mechanically exfoliated hBN flakes randomly distributed atop its surface (Figure 1a). It can be observed that numerous discrete MoS₂ monolayers connected to each other and merged into a continuous single-layer film. The continuous single-layer film typically exhibits a lateral dimension up to several hundreds of micrometers, beneficial for probing many discrete locations within a sample via an optical means. Yu et al. previously reported that, unlike the growth of MoS₂ on SiO₂ substrates, the growth of MoS₂ on hBN normally follows a lattice alignment epitaxial growth mode [26], where two structurally equivalent 0° and 60° stacked MoS₂/hBN are presented.

Figure 1b presents the Raman scattering spectra acquired from six distinct locations, corresponding to the MoS₂ samples prepared atop the hBN and SiO₂ substrates, respectively. The characteristic feature located at approximately 520.7 cm⁻¹ is attributed to the underneath Si substrate used for wavenumber calibration. Due to the reliability and repeatability of our synthetic strategy, we can observe two characteristic peaks of MoS₂, E¹_2g and A_1g, as well as the signal from hBN (≈1366.4 cm⁻¹), as shown in the Raman spectra labeled as H1–H3, confirming that MoS₂ is successfully grown on the hBN flake. The peak interval between the two characteristic features of MoS₂, E¹_2g and A_1g, can help us determine the number of layers of the MoS₂ samples quantitatively and rapidly (Appendix A). One accessible mode for hBN is the in-plane E_2g mode located at approximately 1366.4 cm⁻¹ (Appendix B).

The vdW heterostructures comprised of 2D materials serve as an inspiring platform for tailoring physicochemical properties, exploring novel quantum effects, and ultimately fabricating conceptually new devices [17,18]. A classical paradigm is the demonstration of the fascinating Hofstadter butterfly observed in artificial vertically stacked moiré superlattices comprised of graphene and hBN [27,28]. The strategy to tailor material properties with interface engineering is also available for the vertically stacked structure comprised of TMDs and hBN, which enables the exploration of valley polarization. The circular polarization-sensitive PL measurement was performed in a home-made micro-zone setup as displayed in Figure 1c. The linearly polarized excitation light is converted to the circularly polarized one by a broadband quarter-wave plate. A 50× objective (N.A. 0.55) is used to collect the emission signals from the MoS₂ monolayers. The left-handed and right-handed circularly polarized light signals (σ+ and σ−) are converted to horizontal and vertical linearly polarized light signals (I_σ+→σ+ and I_σ+→σ−), respectively, via a broadband quarter-wave (1/4λ) plate. The two linearly polarized light beams can be separated in real space with a Wollaston prism and then focused to two spots positioned at the entrance slit of the spectrometer equipped with the charged coupled device (CCD) cooled at −75 °C. The intensities of these two orthogonally linearly polarized beams, corresponding to the σ+ and σ− components of the PL signal, are recorded simultaneously with the CCD. By doing so, we greatly reduce the error caused by laser power fluctuation. The steady-state circularly polarized PL spectra were captured under resonant excitation with a continuous-wave 633 nm laser. For the time-resolved PL measurements, a pulsed linearly polarized Ti:sapphire laser with the wavelength of 405 nm was employed.

3. Results and Discussion

3.1. Crystal Structure of MoS₂ and the Coupled Valley-Spin Excitonic Transition Rules

Transmission electron microscopy (TEM) provides a powerful tool for examining the morphology and lattice structure of low-dimensional materials. As shown in Figure 2a, a MoS₂ monolayer triangle was transferred atop a carbon-film-coated TEM copper microgrid in the presence of a relatively complete geometric morphology. The recorded selected area electron diffraction (SAED) spots (inset, Figure 2a) exhibit a typical hexagonal pattern, consistent with that of MoS₂. The lattice spacings of approximately 0.27 nm and 0.16 nm are clearly visible along the {100} and {110} planes of MoS₂, respectively (Figure 2b). The atomic-level resolution TEM image along with the corresponding SAED patterns demonstrate that the CVD-synthesized MoS₂ monolayer possesses excellent crystal quality with a hexagonal lattice structure.

From the top view of the crystal structure in MoS₂, a hexagonal honeycomb lattice structure can be observed that generates two sets of degenerate-but-not-equivalent valleys, K and K’, at the edges of the first Brillouin zone (left panel, Figure 3a). These valleys that are degenerate in energy exhibit a huge splitting in the valence band (Δ_v = ~148 meV) induced by spin-orbit coupling and a much smaller one in the conduction band (Δ_c = ~3 meV), which is also depicted for completeness (right panel, Figure 3a), for the MoS₂ monolayer [29,30,31,32]. The K and K’ valleys are differentiated by the opposite spin orientations corresponding to the valence band maximum (VBM) and conduction band minimum (CBM). As schematically displayed in the right panel of Figure 3a, combined with the time reversal symmetry, the spin orientations of the K and K’ valleys are anti-symmetric, enabling a locking of the spin and the valley degree of freedom. The remarkable difference guarantees the valley-dependent optical selection rules that the direct excitonic transitions should obey, including A and B, in the MoS₂ monolayer. To be specific, the circularly polarized lights with left-handed helicity (σ+) excite the excitonic transitions in the K valleys exclusively, whereas those lights with right-handed helicity (σ−) only couple to the K’ valleys. In our previous report [24], we reported that the A excitonic emission from the MoS₂ monolayer on the hBN flake exhibited a ubiquitous enhancement compared with that on SiO₂/Si. As plotted in Figure 3b,c, the intensities of the PL signals from MoS₂ on hBN are much stronger than those obtained from MoS₂ on SiO₂, both under 633 nm and 488 nm excitation.

3.2. Steady-State Circularly Polarized PL Spectra

The valley polarization-resolved luminous property appears a typical representative among the abundant unique physical properties for the MoS₂ monolayer. The vertical stacking of MoS₂ and hBN enables us to explore the valley polarization utilizing our home-made circular polarization PL measurement system. We further performed the PL measurements upon the MoS₂ monolayer on hBN and SiO₂ under resonant excitation with a 633 nm laser, widely employed in valley pseudospin-related luminous property investigation. The value for the degree of valley polarization obtained from the MoS₂ monolayers with different substrates underneath was calculated based on the measured polarization-resolved PL spectra involving σ− and σ+ components. As shown in Figure 4a and b, we obtained the PL spectra by using the 1.96 eV (633 nm) laser radiation as excitation, that is with the left-handed helicity (σ+) on resonance with the A excitonic transition. We determined the degree of valley polarization quantitatively [10,11] as

$$P=\frac{I_{\sigma +\to \sigma +}-I_{\sigma +\to \sigma -}}{I_{\sigma +\to \sigma +}+I_{\sigma +\to \sigma -}}$$

(1)

where I_σ+→σ+ and I_σ+→σ− denote the intensities of the left-handed and right-handed circularly polarized PL signals, respectively, which are excited with a left-handed circular excitation laser. As plotted in Figure 4c, the average valley polarization values, ranging from 656 nm to 676 nm, were calculated to be 0.207 for MoS₂ on SiO₂ and 0.080 for MoS₂ on hBN. It can be observed, among the wavelength range of the direct A excitonic emission for the MoS₂ monolayer, the valley polarization for MoS₂ on hBN is always lower than that observed from the MoS₂ monolayers on SiO₂.

Based on the numerical analyses of 25 MoS₂ monolayer samples on hBN, the statistical average for the degree of valley polarization is P = 0.130 ± 0.046 (Figure 4d), which demonstrates a relatively low degree of valley polarization in the MoS₂/hBN heterostructure at room temperature (300 K). In stark contrast to that, the degree of valley polarization for the MoS₂ monolayer deposited atop the SiO₂ wafer under the identical measurement conditions is relatively high. The statistical average value equals approximately 0.198 ± 0.020 from which an important message can be delivered that the interaction widely existent in 2D materials/supporting substrates may play a critical role in the modulation of valley pseudospin in 2D polarized materials. Combined with the measurement results mentioned above, an enhanced PL intensity and a reduced valley polarization are observed in the MoS₂/hBN heterostructure. Why does there exist a noticeably opposite relationship between luminous intensity and valley polarization? To answer this question and further clarify the underlying mechanism, we further performed the time-resolved circularly polarization-dependent PL measurements at room temperature.

For the steady-state conditions excited with a continuous wave (CW) laser, the degree of valley polarization, P, can be determined with

$$P=\frac{P_0}{1+\frac{2{\tau }_e}{{\tau }_v}}$$

(2)

under a rate model, where P₀ denotes the initial polarization, and τ_e and τ_v represent the exciton and valley relaxation times, respectively. The derivative process is provided in Appendix C. The value of P increases with either an increase in valley lifetime τ_v or a decrease in exciton lifetime τ_e, as clearly observed from Equation (2). Previous reports did not identify the influence of substrates on the valley relaxation time τ_v for MoS₂, among which the measured values are close numerically and even the supporting substrates are different. All the as-synthesized MoS₂ monolayers shown in Figure 4 were excited with exactly the same excitation wavelength, pumping power, and exposure time, and exposed to nearly the identical environment. Thus, it is assumed that the valley relaxation time τ_v and the initial polarization P₀ are the same. It can be inferred that the degree of valley polarization P will decrease if the exciton relaxation time increases.

3.3. Time-Resolved Circularly Polarized PL Spectra

According to the theory in semiconductor physics, the exciton relaxation time τ_e is closely related to both radiative and non-radiative recombination times through the following relational expression:

$$\frac{1}{{\tau }_e}=\frac{1}{{\tau }_r}+\frac{1}{{\tau }_{nr}}$$

(3)

The time parameters, τ_r and τ_nr, represent the radiative and non-radiative lifetimes, respectively. For transition metal dichalcogenides, the non-radiative lifetimes can be orders of magnitude shorter than the radiative recombination times. In other words, the magnitude relationship τ_nr << τ_r exists in MoS₂ monolayers. Hence, there exists an approximation relation τ_e ≈ τ_nr. For a certain material, the radiative recombination time τ_r can be regarded as a constant. Since the luminous intensity is in direct proportion to quantum yield (QY),

$$QY=\frac{1}{1+\frac{{\tau }_r}{{\tau }_{nr}}}\approx \frac{1}{1+\frac{{\tau }_r}{{\tau }_e}}$$

(4)

If the radiative recombination time remains unchanged, a longer exciton relaxation time will lead to a higher QY and, thus, an enhanced PL intensity.

We further performed time-resolved photoluminescence (TRPL) measurements to examine the exciton dynamics and capture the corresponding fluorescence lifetime with which our hypothesis mentioned above may be verified. The measured TRPL spectra are plotted in Figure 5. Two systems, including the MoS₂ monolayers deposited atop hBN and SiO₂, were measured, which behave remarkably different from each other in the intensity of luminescence and the degree of valley polarization. A pulsed, linearly polarized Ti:sapphire laser with the wavelength of 405 nm was employed to pump the direct excitons into the two degenerate-but-not-equivalent valleys, K and K’, in the MoS₂ monolayer simultaneously via an optical means. The pulsed optical pumps were realized with an optical parametric amplification. The subsequent luminescent signals created from the K and K’ valleys could then be collected and directed to a time-resolved CCD detector. The dynamic experimental results were quite similar to those observed in the steady-state PL measurement results.

As mentioned above, the MoS₂ monolayers on hBN and SiO₂ display noticeably different PL intensity as well as distinguishing valley polarization behaviors. The time-resolved emission spectra measured from these samples, as presented here, also exhibit significantly different exciton attenuation kinetics. MoS₂ on hBN possesses a longer exciton lifetime than MoS₂ on SiO₂. Compared with SiO₂, hBN is chemically inert and has no defect states or hanging bonds on its surface. This may result in less disorder to the MoS₂ monolayer [33], which is mainly stemmed from extrinsic effects, for example, defects and trap states. The difference in the exciton relaxation time explains the enhanced PL intensity and lower valley polarization of MoS₂ on hBN.

4. Conclusions

In conclusion, we successfully synthesized a MoS₂ monolayer on hBN nanoflakes via the CVD method and performed room-temperature circularly polarized PL measurements upon the MoS₂ samples with different substrates underneath. We observed that there exists an apparent inverse correlation between the intensity of PL signals and the value of valley polarization, which originates from the variations in exciton relaxation time. Compared to MoS₂ monolayers atop SiO₂, the MoS₂ monolayers atop hBN exhibit a relatively low degree of valley polarization. An enhanced PL intensity and a longer exciton lifetime were also experimentally consolidated in the MoS₂/hBN system. Our discovery suggests a pathway to tailor carrier dynamics via crystal modification, which can be realized by introducing an appropriate amount of defects or non-radiative recombination sites. By doing so, the room-temperature valley polarization can be effectively modulated, and the coupling between degenerate valleys can be attenuated, which provides new strategies for the realization of state-of-the-art devices in spintronics and valleytronics.