Near-field coupling of interlayer excitons in MoSe₂/WSe₂ heterobilayers to surface plasmon polaritons

2D transition metal dichalcogenides (TMDs) have sparked tremendous interest in the field of emergent electronics for their unique interplay of spin, valley, and degrees of freedom (DOF).^1–3 The vertical van der Waals stacking of homo- or heterolayers add an additional layer DOF and enables the formation of lateral moiré superlattices. The advancement of van der Waals stacking opens a vast space for designing artificial material systems with tailored interactions among all DOFs.^4,5

One of the remarkable aspects of TMDs is the coupling between the valley DOF and the polarization of light field, which allows for the manipulation and control of valley and spin DOFs through light–matter interactions, hence making exciton physics a promising platform to address quantum states in 2D TMDs. However, exciton is spatially circumscribed owing to its low mobility and limited lifetime. This hinders the on-chip manipulation and propagation of the quantum states.

Surface plasmon polaritons (SPPs) are known as electromagnetic surface waves that propagate at the interface between a metal and a dielectric, resulting from the coupling of light with collective electron oscillations.⁶ Its enhanced local fields and large in-plane momentum provide increased light–matter interaction and strong confinement at subwavelength scales.^7–10 As the SPP travels at a fraction of light speed, it is expected that the coherence would persist across a long distance. This may enable the coherent remote coupling of excitons mediated by SPPs.

In this work, we demonstrate the near-field coupling of interlayer excitons in the MoSe₂/WSe₂ heterobilayer to SPPs. We investigate both spin-singlet and spin-triplet interlayer excitons and observe that they both exhibit comparable optical activity and can be efficiently coupled to SPPs. Remarkably, the polarization information of the excitons is successfully transferred to SPPs and propagates over tens of micrometers.

The individual monolayers were exfoliated from bulk single crystals and stacked with a controlled alignment [Fig. 1(b)] to ensure the formation of bright interlayer excitons. The MoSe₂/WSe₂ heterobilayer was encapsulated with a few-layer hexagonal boron nitride (h-BN) single crystals for isolation from absorbates and exciton quenching. The whole heterostructure was transferred onto patterned gold substrates with optical grating trenches as sketched in Fig. 1(a). The 8 × 46 µm² gold strip of 270 nm thickness for SPP coupling was evaporated on the 300 nm SiO₂/Si substrate, and 11 slits (50% duty cycle) were etched at both ends with 95 nm depth, 426 nm width, and 4.5 µm length (sizes measured by atomic force microscope) by the wet-chemical method using a KI solution. The average lateral roughness of about 150 nm at gold trenches was characterized with AFM.

The twisted angle of the heterostructure was characterized with the polarization resolved second harmonic generation (SHG) excited by a 120-femtosecond laser pulse of 800 nm. Figure S1 of the supplementary presents the angle dependence of the SHG intensity from the WSe₂ monolayer and heterobilayer regions, respectively. The angle difference between the monolayer and the stacking region of about 2.8° is extracted by fitting the distorted SHG patterns with the involvement of strain tensor due to the impact of thick gold strip.^11,12 This shows a lattice distortion in the moiré pattern. Together with the decreased intensity in the heterobilayer, the SHG characterization indicates the H-type stacking of our device with a twist angle about 54.4°.

Figure 1(c) shows the absorption and photoluminescence (PL) spectra of the heterostructure at temperature T = 10 K. The interlayer excitons (IX) dominate the PL signal, while the intralayer PL peaks are dramatically quenched at cryogenic temperatures. It implies the efficient interlayer charge transfer of electrons and holes between MoSe₂ and WSe₂ because of the type-II band alignment. The interlayer exciton peaks are absent in the absorption spectrum owing to the weak optical oscillator strengths. In the PL spectrum, the two distinct interlayer exciton peaks at 1.379 and 1.404 eV (linewidth ∼10 meV) with about 25 meV separation are assigned to the spin-singlet and spin-triplet states consisting of the electrons at spin-split conduction bands in MoSe₂ [(as depicted in Fig. 1(e)] and holes at the valence band edge of WSe₂, respectively.^13,14 Spin-triplet excitons are generally electric dipole forbidden in conventional materials. However, in monolayer TMDs, the strong spin–orbit coupling admits a dipole-allowed transition for the spin-triplet excitons with an out-of-plane optical dipole.^15,16 In H-stacked heterostructures, theoretical studies predict that moiré potential modulates the atomic-size potential distribution and further brightens the spin-triplet optical transitions with a nonzero in-plane dipole component.^17,18 Figure 1(c) indicates that the spin-triplet and spin-singlet excitons show a comparable emission strength in the PL spectra even with the regular confocal setup as the spin-triplet occupies the energetically ground state in the H-type configuration.

In this 54.4° twisted MoSe₂/WSe₂ heterobilayer (twist mismatch Δθ ∼ 5°), the lattice mismatch δ of ∼0.1% leads to the hexagonal moiré superlattice with a period up to several tens of nanometers,^17,19 as clearly illustrated in Fig. 1(d). In a H-type moiré unit cell, each stacking configuration transitions through three atomic registries $HhM$⁠, $Hhh$⁠, and $HhX$ with threefold rotational symmetry along the lateral translation.^17,18 As for optical band edge transitions at K and K’ points, spin triplet (singlet) of interlayer excitons centered at these H-type registries presents diverse selection rules, which can, respectively, couple to σ⁻, σ⁺, and z (z, σ⁻, and σ⁺) polarized photons.¹⁸ However, upon the occurrence of local lattice deformation, the ideal moiré pattern will experience a transformation, leading to the appearance of periodically reconstructed patterns. According to previous theoretical and experimental studies, only $Hhh$ stacking is dominant after lattice reconstruction,^18,20–23 which is consistent with our observation that there are only two prominent interlayer exciton peaks in the PL. As presented in Fig. 1(e), the above optical selection rules lead to opposite valley polarization for the spin-singlet and spin-triplet interlayer excitons. Furthermore, the existence of the atomic reconstruction or external strain could modify the quasi zero-dimensional (0D) moiré traps into one-dimensional (1D) strips or 2D domains.^23,24 The distorted SHG patterns imply the strain to the heterobilayer, and it may result in the absence of spectrally narrow moiré peaks in the PL spectra.

Beyond the optical signatures of interlayer excitons, we observed the efficient coupling of these excitons to SPPs via the near-field coupling. Figure 2(a) shows the photoluminescence image of the heterostructure under the marked point-excitation where the laser beam is focused to a micron-size spot. The photoluminescence emerges from the whole 46 µm long device on the gold strip. Figure 2(b) presents the representative spectra collected at the corresponding points, while the excitation is focused on the marked spot. As the exciton’s diffusion length is limited to 1–2 microns at maximum and the emission pattern from the heterobilayer to monolayer is not continuous, it surely eliminates the possibility of exciton diffusion for photoluminescence across the area of tens-microns.²⁵ The photoluminescence out of the excitation spot must result in the remote coupling of the SPPs: the laser excitation pumps intralayer and interlayer excitons across the heterostructure, and these various excitons couple to the gold strip via near-field coupling and form distinct SPPs. The SPPs spread across the gold strip and consequently transfer the energy to the heterobilayer and monolayer WSe₂ outside of excitation area.

The other evidence of the SPP coupling is the emission from the slits at the both ends of the gold strip, around 10–20 microns from the excitation spot. This emission results from the scattering of the SPPs by the gold trenches, which convert SPPs to free space photons. The momentum match is materialized by both the optical gratings and the roughness of the trenches. The PL spectra at the slits carry all the exciton energies, including intralayer excitons from MoSe₂ (X_Mo) and WSe₂ (X_W) as well as interlayer excitons IX. Compared with the PL spectrum at the excitation spot, the weight of the X_Mo peak is suppressed and the peak X_W experiences a redshift of 4–8 meV at the slits. This likely originates from the contrasting coupling strength with SPPs. Owing to the inherent transverse-magnetic (TM) wave of SPPs, the out-of-plane optical dipoles have more efficient coupling to SPPs compared with the in-plane dipoles. The spin-triplet excitons in monolayer TMDs possess out-of-plane transition dipoles as a result of optical selection rules, which display a significantly higher efficiency in coupling with SPPs compared to the in-plane polarized spin-singlet excitons.²⁶ The intralayer excitons at MoSe₂ X_Mo are the band edge excitons with pure in-plane optical dipoles, while the intralayer excitons at WSe₂ X_W consist of two spin-singlet and spin-triplet excitons with in-plane and out-of-plane dipoles, respectively. The spin-singlet exciton is dominant in the PL spectrum at normal collecting angles, while the spin-triplet exciton is the band-edge exciton and radiates along the plane. The orientation of the optical dipoles determines the coupling efficiency with SPPs: X_Mo and the spin-singlet X_W excitons couple with SPPs in a much weaker strength than the spin-triplet X_W exciton. This is well consistent with the emission at the slits.

We further explore the near-field coupling of the interlayer excitons of the MoSe₂/WSe₂ heterobilayer to SPPs under a cryogenic condition (10 K). Figure 3 demonstrates the comparison of the PL spectra collected directly from the focal spot (canonical far-field detection) and from the scattering at slits. The emission from both spin-triplet neutral excitons (X_D) and trions (X_DT) in WSe₂ becomes prominent at the slit accompanied with the suppressed spin-singlet exciton (X_A and X_AT), resulting from the contrasting coupling with SPPs as discussed above. Unlike in the TMD monolayer, the spin-triplet interlayer exciton shows the same order of magnitude transition dipole with the spin-singlet state.¹⁸  Figure 3(b) shows that both spin-singlet and spin-triplet interlayer excitons exhibit comparable optical activity and successful coupling to SPPs with different strengths. We distinguish the intensity distribution from the well-fitted Lorentz peaks and find that the spin-singlet exciton becomes dim in the slit areas. Figure 3(c) depicts the ratio of the intensity at the slit to that at the excitation spot, where the average ratios of triplet and singlet states are ∼0.255 and 0.185, respectively. It reflects their efficiencies of the SPP coupling, showing the contrasting out-of-plane transition dipole strengths of the exciton. In addition, the PL peak shape exhibits a slight blueshift, accompanied by an increase in intensity and broadening on the high-energy side from IX_S [Fig. 3(d)].

Meanwhile, we also examine the valley polarized information of interlayer exciton from both the excitation spot and the slits at 10 K. Figure 4(a) presents the polarization resolved PL spectra of the interlayer exciton under circularly polarized excitation at the excitation spot. The spin-triplet exciton exhibits the same polarization as the pump laser, while the singlet state displays an opposite helicity, which agrees with the optical selection rules in the $Hhh$ stacking.^17,18 The degree of circular polarization (P_C) is defined as (I_σ+ − I_σ−)/(I_σ+ + I_σ−), where I_σ+ (I_σ−) represents the intensity of the co-polarized (cross-polarized) PL component with respect to the excitation. As in Fig. 4(d), the triplet exciton peak shows P_C ∼ 14.7%, while singlet exhibits P_C ∼ −16.5%.

Remarkably, we observed that the valley polarization of the interlayer excitons is successfully coupled to SPPs without significant attenuation P_C in the up-slit after 20 microns propagation, as shown in Figs. 4(b) and 4(e). The down-slit exhibits a noticeable decrease in polarization, yet still maintains a P_C ∼ 8% (−6%) for triplet (singlet) states [Figs. 4(c) and 4(f)]. The depolarization is likely attributed to the contamination scattering on the SPPs’ propagation path as seen in the optical image. The lateral roughness of the slit edge provides the possibility for circularly polarized emissions.^27–30 The valley polarization of the interlayer exciton is impressively preserved, even tens of micrometers away from the focal region.

In summary, we demonstrate the near-field coupling of monolayer TMD and TMD heterostructures to surface plasmon polaritons. Both excitons with in-plane and out-of-plane optical dipoles could couple with SPPs, though with different coupling efficiencies. Remarkably, the polarization of the excitons is successfully coupled to SPPs and propagates over tens of micrometers. This work provides an approach to achieve remote delivery of valley polarization.

The data and analysis of the polarization resolved second harmonic generation (SHG) experiments are presented.

The work was supported by the National Key R&D Program of China (Grant No. 2020YFA0309600), Guangdong-Hong Kong Joint Laboratory of Quantum Matter, and the University Grants Committees/Research Grants Council of Hong Kong SAR (Grant Nos. AoE/P-701/20, 17300520, and 17301223). K. W. and T. T. acknowledge the support from the Elemental Strategy Initiative conducted by the MEXT, Japan (Grant No. JPMXP0112101001), and JSPS KAKENHI (Grant Nos. 19H05790, 20H00354, and 21H05233).

The authors have no conflicts to disclose.

Xiong Wang: Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Validation (lead); Visualization (lead); Writing – original draft (lead). Zemeng Lin: Investigation (equal); Methodology (equal). Kenji Watanabe: Resources (equal). Takashi Taniguchi: Resources (equal). Wang Yao: Funding acquisition (equal); Investigation (equal); Validation (equal). Shuang Zhang: Investigation (equal); Project administration (equal); Validation (equal); Writing – review & editing (equal). Xiaodong Cui: Conceptualization (lead); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (lead); Supervision (lead); Writing – review & editing (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.