Introduction

Surface plasmons are light-driven free electron oscillations at the interface between conductor and insulator with the capacities to break the light diffraction limit, confine light at the deep-subwavelength scale, and enhance near-field light intensity1,2,3. Owing to these fantastic characteristics, surface plasmons play significant roles in optical recording3, integrated logic gates4, ultrasensitive sensing5, angular momentum multiplexing6, optical nonlinearity7, light-field control8, all-optical modulation9, etc. Particularly, plasmonic resonances generated in metallic nanostructures reveal ultrasmall mode volume, exceptional field enhancement, and open configurations, facilitating the reinforcement of light-matter interactions10,11,12. Plasmonic resonances can find crucial applications in the improvement of nonlinear optical conversion10, photodetection11, and light emission12. Recently, plasmonic resonances have been extensively used to achieve the coupling interactions with quantum emitters13,14,15. Transition metal dichalcogenides (TMDs) with structure MX2 (M: Mo or W, X: S or Se) are new-emerging semiconducting nanomaterials (emitters) with fascinating layer-dependent bandgap, high carrier mobility, chemical stability, and flexible integration16,17,18. Owing to the reduced Coulomb screening and 2D confinement, the atomic-layer TMDs present the strong excitonic response (excitons: electron-hole pairs) in the bandgap transition with the high transition dipole moment and large binding energy at room temperature18,19,20. TMD atomic layers offer a promising platform for plasmon-exciton coupling interactions, which enable promising applications in photoluminescence (PL) emission14,21, plasmonic modulation22, interface catalytic reactions23, ultrafast optical nonlinearity24, integrated photonic circuits25, valley polarizations26, etc. Until now, plasmon-exciton coupling behaviors have been extensively demonstrated in the noble metal-based plasmonic systems. Due to the intrinsic drawbacks of noble metals (e.g., poor tunability, incompatibility with standard manufacturing processes, and narrow operating frequency range), optical materials beyond noble metals have drawn special attention for exciting surface plasmons27. Exploring plasmon-exciton coupling in nonmetallic systems will be particularly significant for their practical applications in optoelectronic devices and integration.

Topological insulators (TIs) are a new type of quantum material with a topologically protected conducting surface (or edge) induced by the strong spin-orbit coupling from insulating bulk28. The topological edge state was first reported in the 2D HgTe quantum wells with the generation of the quantum spin Hall effect29. Subsequently, topological surface properties were found in many 3D materials, especially Sb2Te3, Bi2Te3, Bi2Se3, and their compounds30,31,32. Besides excellent electronic characteristics, TIs present particularly promising optical features containing the high refractive index, giant nonlinear optical coefficient, and external tunability33,34,35. Recently, surface plasmons have been unexpectedly observed in the 3D TIs at frequencies ranging from ultraviolet (UV) to terahertz (THz)33,36,37,38. Different from conventional plasmons, TI plasmons possess the advantages of ultrabroad operating frequency range, low loss at optical frequencies, CMOS integration, etc33,37,38. TIs will inevitably promote the development of next-generation photonic and optoelectronic technologies33. Even so, plasmon-exciton coupling behaviors have not been specifically demonstrated in TI systems.

Herein, we experimentally demonstrated the scattering resonance response of surface plasmons in a type of TI metasurfaces and their coupling activities with the excitons in WS2 atomic layers. The TI metasurfaces are composed of disk-shaped nanowells fabricated on the Sb2Te3 single-crystal surface by FIB lithography. The results reveal that the scattering spectrum of plasmonic resonance is dependent on the depth and pitch of TI nanowells. In the TI metasurfaces integrated with atomic-layer WS2, we observed the distinct coupling effect between the TI plasmons and excitons in the WS2 atomic layers. The theoretical analysis illustrates that the plasmon-exciton interactions in the WS2/TI metasurface heterostructures are located in the weak coupling regime with the generation of Fano resonance. The weak plasmon-exciton coupling can give rise to strong PL enhancement of 15 and 25-fold for monolayer and triple-layer WS2 on the TI metasurfaces, respectively. The results will open a new door for plasmon-exciton coupling in nonmetallic systems and applications of TIs in novel optoelectronic devices.

Results and Analysis

Fabrications and measurements of materials and TI metasurfaces

Figure 1a shows the 3D diagram of the TI metasurface integrated with a WS2 atomic layer. Here, the Sb2Te3 TI was employed due to its strong topological property and good capability to support surface plasmons at optical wavelengths30,38. The TI metasurface consists of a disk-shaped nanowell array fabricated on a Sb2Te3 layer. The Sb2Te3 layer was mechanically exfoliated from the single-crystal bulk material and transferred onto a dielectric (Si) substrate. The Sb2Te3 single crystal was grown using the melting and slow-cooling method39. The depth, diameter, and pitch of disk-shaped nanowells are denoted by h, d, and p, respectively. The atomic-layer WS2 was prepared by mechanically exfoliating the WS2 single-crystal bulk material with the scotch tape, and stuck on a polydimethylsiloxane (PDMS) film. The WS2 layer on the PDMS film was transferred onto a TI metasurface with a dry fixed-point transfer method (see Method section). As a typical TMD semiconductor, WS2 exhibits the direct and indirect band gaps for monolayer and multilayers, respectively16. The WS2 layers present the sharp absorption bands and strong excitonic resonances with the generation of intense PL in the visible region13. If there exists a resonant mode in TI metasurface around the WS2 exciton wavelength, the WS2 excitons will couple with the resonant mode, as depicted in Fig. 1b13,14,15. To verify the composition of grown TI, the energy dispersive X-ray (EDX) spectrum of a Sb2Te3 flake was measured using the energy dispersive spectrometer in a transmission electron microscope (TEM), as depicted in Fig. S1. The measured elemental molar ratio of Sb and Te is ~2:3, which is well consistent with the molecular formula of Sb2Te3. For the TEM measurement, the Sb2Te3 thin flakes on a Cu microgrid were prepared with the mechanical exfoliation and chemical etching methods (see Method section). Figure 1c depicts the high-angle annular dark-field (HAADF) image of a thin Sb2Te3 flake and the EDX mapping for Sb and Te spatial distributions. Figure 1d shows the measured selected area electron diffraction (SAED) pattern of the Sb2Te3 flake. The lattice spacing of 110 in the SAED pattern and the spacing of ~0.217 nm in the high-resolution TEM image clarify the high quality and hexagonal structure of the prepared Sb2Te3 single crystal40. The Sb2Te3 TI single crystal is an atomic-layered material that can be easily exfoliated. The layered materials enable the feasible transfer and integration with other substrates or structures. Figure 2a shows the scanning electron microscopy (SEM) image of a Sb2Te3 TI metasurface with disk-shaped nanowells, which was fabricated on a Sb2Te3 layer transferred to a silicon substrate by using FIB lithography (see Method section). An atomic force microscope (AFM) was used to measure the profile of the TI metasurface. As illustrated in Fig. 2b, the measured depth, diameter, and pitch of nanowells are h = 100 nm, d = 300 nm, and p = 700 nm, respectively. Figure 2c depicts the experimentally measured dark-field scattering spectrum of the TI metasurface by employing a confocal Raman spectroscopy system (see Method section). A distinct scattering resonance peak appears at the wavelength of ~587 nm. To verify the experiment result, we numerically simulated the scattering spectrum of the TI metasurface by using a finite-difference time-domain (FDTD) method (see Method section). As depicted in the inset of Fig. 2c, the simulation result is in good agreement with the experimental measurement. Another inset of Fig. 2c shows the distribution of the electric field |E/Ei | 1 nm above one nanowell in the area impinged with the light at the scattering peak wavelength. The electric field is obviously reinforced with the generation of plasmonic resonance in the TI nanowells. The unit size of the TI structure will determine the operating wavelength of surface plasmons. Plasmonic resonances in TI metasurfaces with hundred-nanometer wells work in the visible range, where the TMD (e.g., WS2) exciton wavelengths are located. The existence of plasmonic resonance stems from the negative relative permittivity of Sb2Te3 material in the visible region41. Specifically, the surface state of Sb2Te3 TI presents negative relative permittivities similar to those of metals. At the wavelengths from 300 to 760 nm, the relative permittivity of the bulk state is also negative. Thus, both the surface and bulk states contribute to the formation of TI plasmons at the wavelengths of interest. The topological surface state enables decreasing the optical loss and improving plasmonic field enhancement. The quality factor for surface plasmons of Sb2Te3 TI single crystal is larger than that of gold in the 300–520 nm wavelength range, where surface plasmons of Sb2Te3 TI present relatively low loss. In the TI metasurfaces, the wavelengths of scattering resonance exceed this range. The scattering spectral width of Sb2Te3 TI nanowells becomes slightly broader than that of gold nanowells. The scattering plasmonic resonance of periodic structures relies on the unit size42. We fabricated the TI metasurfaces with different nanowell heights and measured the dark-field scattering spectra. As depicted in Fig. 2d, the scattering resonance peak presents a slight redshift as the nanowell height increases from 100 to 160 nm. Moreover, the scattering resonance is dependent on the pitch for plasmonic arrays43. Figure S2 shows the measured scattering spectrum of a TI metasurface with h = 100 nm, d = 300 nm, and p = 750 nm. The experiment demonstrates that a scattering resonance peak is located at a wavelength of ~623 nm, which agrees well with the simulation result. There may exist a slight fabrication error of 10% for the width of TI nanowells, which will have little effect on the peak position of the scattering resonance.

Fig. 1: Diagram of Sb2Te3 TI metasurface with atomic-layer WS2 and characterizations of Sb2Te3 material.
Fig. 1: Diagram of Sb2Te3 TI metasurface with atomic-layer WS2 and characterizations of Sb2Te3 material.
Full size image

a 3D diagram of the Sb2Te3 TI metasurface with disk-shaped nanowells covered by a WS2 atomic layer. The height, diameter, and pitch of nanowells are h, d, and p, respectively. b Prototype level system for plasmon-exciton coupling process in the WS2/TI metasurface structures. c TEM-measured HAADF image of a thin Sb2Te3 flake transferred on a Cu microgrid. The right insets are EDX mapping images of the Sb2Te3 flake, which determine the spatial distributions of Sb and Te elements. d TEM image of the Sb2Te3 flake in the area marked with the red dotted rectangle in (c). The right pictures show the SAED pattern and high-resolution TEM image (inset) of the Sb2Te3 flake.

Fig. 2: Fabrications and measurements of Sb2Te3 TI metasurfaces.
Fig. 2: Fabrications and measurements of Sb2Te3 TI metasurfaces.
Full size image

a SEM image of a representative TI metasurface with disk-shaped nanowells fabricated on a Sb2Te3 flake by using the FIB lithography. The scale bar is 2 μm. b Measured height profile along the line in the AFM image (see the inset). c Measured and simulated scattering spectra of the TI metasurface with h = 100 nm, d = 300 nm, and p = 700 nm. The inset below shows the distribution of electric field |E/Ei | 1 nm above one nanowell at the scattering peak wavelength. d Measured scattering spectra of the TI nanowells with different h when d = 300 nm and p = 700 nm.

Plasmon-exciton coupling in WS2/TI metasurface heterostructures

Subsequently, we integrated a WS2 atomic layer with a Sb2Te3 TI metasurface (see Fig. S3) to study the plasmon-exciton coupling interactions. Figure 3a depicts the optical microscope image of the fabricated TI metasurface with the transferred WS2 atomic layer. The AFM image in Fig. 3b reveals that the WS2 layer mostly covers the TI metasurface. Figure 3c shows the AFM-measured height profiles along the lines in Fig. 3b. The depth, diameter, and pitch of the TI metasurface are h = 135 nm, d = 300 nm, and p = 700 nm, respectively. To verify WS2 layer number, we measured the Raman shift spectrum of WS2 on the TI metasurface by impinging a 532 nm CW laser, and fitted the spectrum with a multi-Lorentzian model, as can be seen in Fig. 3(d). For WS2, the Raman modes of A1g(Γ) and E2g1(Γ) correspond to an out-of-plane atomic vibrational mode and an in-plane vibrational mode, respectively. In Fig. 3d, the A1g(Γ) and E2g1(Γ) modes are located at 416.7 and 354.7 cm−1, whose difference is 62 cm−1. This frequency difference is consistent with the reported value of monolayer WS244. Meanwhile, we measured the scattering spectrum of the WS2/TI metasurface heterostructure, as depicted in Fig. 3e. A distinctly asymmetric spectral profile was observed around the scattering resonance peak, which resembles a Fano resonance response in optical systems45,46,47. Different from traditional Fano resonances, this Fano-like resonance in the WS2/TI metasurface system is attributed to the coupling between the TI plasmons and WS2 excitons48. The spectrally broad plasmonic resonance of TI metasurface offers a quasi-continuum state, and the spectrally narrow WS2 exciton resonance works as a discrete state. As depicted in Fig. 1b, the coupling interference between the quasi-continuum and discrete states gives rise to the appearance of Fano resonance lineshape49,50. The coupling effects are common phenomena, which generally exist in atomic-layered materials/structures51,52. To clarify the mechanism, the coupling interaction between the TI plasmons and WS2 excitons can also be described using the coupled oscillators model53. The TI plasmons and WS2 excitons can be regarded as the two oscillators coupled with each other. In accordance with the description in Fig. 1b, ωex and ωre are assumed as the resonance frequencies of WS2 excitons and TI plasmons, respectively. γex and γre represent the linewidths (damping rates) of excitonic and plasmonic resonances, respectively. g is the coupling strength between the two oscillators. When the harmonic wave E(ω)=E0e-iωt is incident onto the WS2/TI metasurface heterostructure, the amplitudes of plasmons and excitons in the coupled system can be governed by53

$$\left\{\begin{array}{l}{\ddot{x}}_{re}+{\gamma }_{re}{\dot{x}}_{re}+{{\omega }_{re}}^{2}{x}_{re}={E}_{0}{e}^{-i\omega t}-2g{\dot{x}}_{ex}\\ {\ddot{x}}_{ex}+{\gamma }_{ex}{\dot{x}}_{ex}+{{\omega }_{ex}}^{2}{x}_{ex}=2g{\dot{x}}_{re}\end{array}\right.$$
(1)
Fig. 3: Fabricated WS2/TI metasurface heterostructure, scattering spectra, and PL emission.
Fig. 3: Fabricated WS2/TI metasurface heterostructure, scattering spectra, and PL emission.
Full size image

a Optical microscope image of the TI metasurface with a transferred WS2 atomic layer. The light-blue dashed curves mark the area of WS2. The scale bar is 2 μm. b AFM image of the WS2/TI metasurface heterostructure. c AFM-measured height profiles of the metasurface along the lines in (b). d Raman shift spectrum of the WS2 atomic layer and fitted spectra with a multi-Lorentzian model. e Experimentally measured and theoretically fitted scattering spectra of the WS2/TI metasurface heterostructure. f PL emission spectra of the Sb2Te3 TI, WS2/TI, and WS2/TI metasurface structures. The inset shows the PL intensity mapping spectrally integrated from 615 to 635 nm in the marked rectangular area of (a). The scale bar is 300 nm.

According to the coupled oscillator equations, the scattering spectrum of the coupled system could be described as53,54

$${\sigma }_{\mathrm{scat}}(\omega )=F{\omega }^{4}{\left|\frac{{\omega }^{2}-{{\omega }_{ex}}^{2}+i\omega {\gamma }_{ex}}{({\omega }^{2}-{{\omega }_{re}}^{2}+i\omega {\gamma }_{re})({\omega }^{2}-{{\omega }_{ex}}^{2}+i\omega {\gamma }_{ex})-4{\omega }^{2}{g}^{2}}\right|}^{2},$$
(2)

where ω is the angular frequency of incident photons, F is the scattering amplitude. By using the above equation, we can theoretically fit the measured scattering spectrum of the WS2/TI metasurface heterostructure. As depicted in Fig. 3e, the fitted curve is in good agreement with the experimental measurement. According to the fitting results, the value of g can approach ~27.9 meV, and γre is equal to 404 meV. For monolayer WS2, γex is derived as ~43.8 meV, which agrees well with the reported value55. The large plasmonic linewidth may stem from the intrinsic absorption (loss) of Sb2Te3 bulk state and limited light confinement of TI nanowells at the wavelengths around 600 nm. The coupling strength g is much less than the threshold strength gt = (γre-γex)/4 = 90.5 meV. According to the coupling criterion (g<gt), the plasmon-exciton coupling is located in the weak-driving regime with the generation of Fano resonance. Due to the Purcell effect, the weak coupling will contribute to the PL emission enhancement of 2D semiconductors54. Figure 3f shows the PL emission spectra of Sb2Te3 TI, WS2/TI, and WS2/TI metasurface structures, which were measured with the confocal Raman spectroscopy with a 532 nm excitation laser. It demonstrates that the PL emission peak around the WS2 A exciton wavelength is attributed to the WS2 atomic layer, not the Sb2Te3 TI. The inset of Fig. 3f depicts the PL intensity mapping spectrally integrated from 615 to 635 nm in the marked rectangular area of Fig. 3a. Obviously, the PL intensity is stronger from the monolayer WS2 on the TI metasurface. It shows in Fig. 3f that the PL intensity can be enhanced by a factor of 15 compared with the PL of monolayer WS2 on the TI. Moreover, the PL peak presents a redshift of 9 nm for monolayer WS2 on the TI metasurface. To clarify it, we employed a multi-Lorentzian model to fit the PL spectra of WS2/TI metasurface and WS2/TI structures56. The portion of A- trions in WS2 can distinctly increase when the laser is impinged onto the WS2/TI metasurface heterostructure, as can be seen in Fig. S4. The wavelength of A- trions is larger than that of A excitons, thus can give rise to the redshift of the PL peak. This may result from the electron doping (injection) of WS2 induced by TI plasmons56. Generally, the structural strain of the TMD layer will also result in the redshift of PL peak with the generation of trions57. There may exist a strain for monolayer WS2 integrated with the TI mesurface, which could give rise to the PL redshift. The PL peak presents a slight redshift with a narrower spectral width as the temperatue rises. The laser heating may exert little influence on the redshift and broadening of PL spectrum.

Subsequently, we fabricated another TI nanowell metasurface on a Sb2Te3 layer, as depicted in Fig. S5a. The AFM measurement denotes that the height, diameter, and pitch of TI nanowells are h = 200 nm, d = 300 nm, and p = 700 nm, respectively (see Figure S5b). We transferred a WS2 multilayer on the TI metasurface to demonstrate the plasmon-exciton coupling, as shown in Fig. 4a. The result in Fig. 4b illustrates that the Raman spectrum of the transferred WS2 multilayer possesses the obvious peaks at 354.4 and 418.9 cm−1 for E2g1(Γ) and A1g(Γ) modes, respectively. The frequency difference between the two Raman modes is 64.5 cm−1, which agrees well with the value of triple-layer (3L) WS258. As depicted in Fig. S5c, the PL emission peaks of WS2 on the TI layer are located at the wavelengths of ~624 and 772 nm, which are attributed to the direct excitonic transition and indirect band gap emission of 3L WS2, respectively59. We experimentally measured the scattering spectrum of WS2/TI metasurface heterostructure. As depicted in Fig. 4c, the scattering spectrum also presents an asymmetric resonance line shape. A distinct dip appears at the wavelength of WS2 A excitonic transition in the scattering spectrum of plasmonic resonance. With the coupled oscillators model, we theoretically fitted the measured scattering spectrum by employing Eq. (2). As shown in Fig. 4c, the theoretical result agrees well with the experimental measurement. The fitting results illustrate that g is ~48.9 meV, close to \(\sqrt{3}\)(the square root of WS2 layer number) times as much as the coupling strength between TI plasmons and excitons in monolayer WS259. The damping rate of TI plasmons γre is ~383.1 meV. The damping rate of A excitons γex is ~75.8 meV. The coupling strength g is less than the threshold value (gt = 76.8 meV), thus the coupling effect is still located in the weak-driving regime. The asymmetric lineshape is attributed to the coupling interference between the TI plasmons and WS2 excitons. Figure 4d depicts the PL emission spectra of 3L WS2 on the TI layer and metasurface under the excitation of 532 nm laser. The PL intensities can be strongly enhanced for the 3L WS2 on the TI metasurface compared to the 3L WS2 on the TI layer. The PL intensity can be enhanced by ~25-fold. The PL emission peak around A excitons exhibits a redshift of ~12 nm for 3L WS2 on the TI metasurface. This redshift of the PL peak for the 3L WS2 is larger than that of monolayer WS2 on the TI metasurface. It may stem from the stronger plasmon-induced electron doping (injection) or the strain effect of 3L WS2 on the TI metasurface. They will give rise to the higher proportion of A- trions in the 3L WS2 (see Figure S5d), which results in the larger redshift of PL peak wavelength around A exciton. In this work, the strong plasmon-exciton coupling was not achieved due to the relatively large mode volume (or low quality factor) of TI plasmons in Sb2Te3 nanowells. The coupling strength could be further increased by optimizing the TI structures for smaller mode volume or by adding the WS2 layer number for more excitons.

Fig. 4: Fabricated TI metasurface with a WS2 multilayer, scattering spectra, and PL emission.
Fig. 4: Fabricated TI metasurface with a WS2 multilayer, scattering spectra, and PL emission.
Full size image

a AFM image of the Sb2Te3 TI metasurface with a WS2 multilayer. The scale bar is 1 μm. b Raman shift spectrum of the WS2 multilayer and fitted spectra with the multi-Lorentzian model. c Experimentally measured and theoretically fitted scattering spectra of WS2-integrated TI metasurface with h = 200 nm, d = 300 nm, and p = 700 nm. d PL emission spectra of the WS2 multilayer on the TI layer and metasurface structure.

Discussion

Surface plasmons in metallic systems with strong field enhancement and open configurations offer a promising technology for light-matter interactions, especially their coupling interactions with quantum emitters. The atomically thin TMD semiconductors with excellent electronic, chemical, and optical characteristics play a crucial role in exploring plasmon-exciton coupling activities and optoelectronic applications. Plasmon-exciton coupling attracts broad attention in metallic systems with the realization of various optical functionalities. Exploring plasmon-exciton coupling interactions in nonmetallic systems is essential to enrich their applications in optoelectronic integrated devices. As new-emerging materials, TIs have presented fantastic electronic and photonic features, especially the dynamic tunability, integration, and capacity to excite surface plasmons in an ultrabroad range from UV to THz. Herein, we have demonstrated the plasmon-exciton coupling interactions in the TI system integrated with WS2 atomic layers. By measuring the scattering spectrum, plasmonic resonance has been observed in the TI metasurfaces, which consist of disk-shaped nanowells fabricated on the Sb2Te3 layer by using the FIB lithography. The experiment results demonstrate that the scattering resonance wavelength presents a redshift with increasing the nanowell depth and pitch. Moreover, the WS2 atomic layers have been innovatively integrated with the Sb2Te3 TI metasurfaces. The coupling behaviors between the TI plasmons and WS2 excitons have been studied in the WS2/TI metasurface heterostructures. Based on the coupled oscillators model, we have theoretically analyzed the coupling behaviors. It is found that the plasmon-exciton coupling is located in the weak-driving regime with a coupling strength of ~27.9 meV for monolayer WS2. The plasmon-exciton coupling-induced Fano resonance gives rise to the strong PL enhancement of 15-fold in monolayer WS2. The coupling strength can approach ~48.9 meV for 3L WS2 on the TI metasurface with a PL enhancement of 25-fold. Our work will pave a new avenue for plasmon-exciton coupling interactions and applications of TIs in compact optoelectronic devices.

Methods

Growth and fabrication of Sb2Te3 TI

The Sb2Te3 TI material was grown using the melting and slow-cooling method39. The Sb and Te powders were mixed with an atom ratio of 2:3 in a quartz tube and heated up to 900 °C until they completely melted. Subsequently, the temperature of the melted powders dropped to 650 °C and slowly declined to 550 °C. Then, the mixed material was naturally cooled to room temperature. Thereby, the Sb2Te3 material can be prepared with the growth orientation along the [001] direction. The Sb2Te3 flakes were mechanically exfoliated from the grown Sb2Te3 bulk material by repeatedly tearing with scotch tape. The TI metasurfaces with disk-shaped nanowells were fabricated on the Sb2Te3 layer by employing the FIB lithography (FEI Helios G4 CX) with 24 pA current and 30 kV voltage. The WS2 atomic layers were mechanically exfoliated from the WS2 single crystal (SixCarbon Technology) with the scotch tape. The WS2 layers were transferred onto the TI metasurface through the dry fixed-point transfer processing in a home-made optical microscope and micromanipulation system.

Material and structure characterizations

For preparing the TEM sample, the exfoliated Sb2Te3 TI flakes were transferred onto a SiO2/Si substrate, and then covered by a Cu microgrid. The SiO2/Si substrate underneath the Cu microgrid can be etched by a KOH solution droplet with ~2 mol/L concentration. After the SiO2 layer was etched, the Sb2Te3 flakes can be stuck on the Cu microgrid. The EDX spectrum, HAADF image, high-resolution TEM image, and SAED of a selected Sb2Te3 flake were measured using the TEM instrument (FEI Talos F200X) with 200 kV voltage. The SEM images of the fabricated TI metasurface were captured employing the FIB-integrated SEM with 21 pA current and 5 kV voltage. The surface morphologies of TI metasurfaces were measured by the AFM equipment (Bruker Dimension FastScan). The dark-field scattering spectra of TI metasurfaces were measured using the confocal Raman spectroscopy (WITec Alpha 300R). The Raman shift and PL emission spectra were tested employing the confocal Raman spectroscopy with a 532 nm CW laser.

Calculation methods

The dark-field scattering spectra and field distributions of TI metasurfaces can be numerically simulated by the full-vector 3D FDTD method60,61. The perfectly matched layer (PML) absorbing boundary condition was set at each side of the computational space. The dark-field light source was built with two Gaussian light beams with the spot sizes of 2300 and 1900 nm as well as the different NA (e.g. NA = 0.5 and 0.495) for outer and inner lens. The phase difference between the two beams was set as π to construct an annular light source. The space area in the x-y plane was set to be larger than the Gaussian beams (e.g. 2500 nm × 2500 nm). A finite-size power monitor was properly placed above the source to detect the light backscattered from the structures, which could simulate the scattering light collected through the objective lens60. In the simulations, the relative permittivity of Sb2Te3 TI was set as the experimental data41. The surface and bulk states of Sb2Te3 TI can be simultaneously considered with a layer-on-bulk model35,41.