Prominently enhanced luminescence from a continuous monolayer of transition metal dichalcogenide on all-dielectric metasurfaces

2.1 Structures of TMDC-transferred metasurfaces

Figure 1C shows optical microscopy images of a TMDC-transferred metasurface. The TMDC film was nearly a monolayer that was grown under the condition described in the Methods section, whereas the μm-scale bilayers appear sparsely as triangle shapes. The black scale bar indicates 25 μm. The Inset shows a magnified view with a scale bar of 5 μm, which includes a triangle bilayer at the top left. A further magnified view is shown in Figure 1D, which was taken using a scanning-electron-microscopy (SEM) instrument (SU8230, Hitachi High-Tech, Tokyo, Japan); the scale bar indicates 500 nm. In the top-view SEM image, a hole in the TMDC layer indicated by a triangle is shown, which is in contrast to the transferred TMDC layer that covers approximately the entire top of Si-nanopellet array. Although the Si nanopellets were designed to be circular columns, they were approximated as regular octagons in the nanofabrication and had slight deviation from the perfect circles.

2.2 Optical resonances of the metasurface

To understand electromagnetic (EM) resonant modes of the all-dielectric metasurface itself, the numerically calculated reflectance spectrum is shown in Figure 2A. The metasurface was set in accordance with the actual sample (Figure 1D), such that the periodicity of array was 400 nm and the diameter of Si nanocolumn in the xy plane was 228 nm. Inset shows the unitcell at an xy section including the Si nanocolumn (purple). The optical resonances of the metasurface appear in the wavelength range below 900 nm. High reflectance peaks and deep dips exhibit prominent optical resonances. The four resonances relevant to this study are indicated by triangles B–E at the wavelengths of 592.5, 608.8, 646.9, and 718.1 nm, respectively, which are located near the PL wavelengths of the TMDC monolayer in this study.

Figure 2:

Optical resonances of the all-dielectric metasurface. (A) Computed reflectance spectrum of the metasurface consisting of a 400-nm-periodicity square array of Si nanopellets (or low-height nanocolumn) with a diameter of 228 nm. Triangles indicate resonant wavelengths B–E, corresponding to dip and peaks of the reflectance spectrum. Inset illustrates the unitcell in the xy plane on a z section; dashed line indicates xz section through the center of the Si nanopellet. (B)–(D) Resonant electric- and magnetic-field intensity distributions, |E|² and |H|², at the wavelengths indicated by the triangles in (A). The xz sections correspond to the dashed line in the inset of (A). The xy section is set at half the height of the Si nanopellet. Each intensity is represented by setting the incident field intensity to 1.0. See the Supplementary Material (Figure S1) for the |E|² and |H|² distributions at the condition of triangle E in (A).

The resonant EM-field distributions are shown in Figure 2B–D corresponding to the wavelengths B–D in Figure 2A, respectively. The incident EM fields were set to 1.0. In Figure 2B–D, the interfaces of the Si nanopellets and SiO₂ layer are shown with white lines in the xz section, and the boundary of the Si nanopellets is shown with white lines and circles; the xz sections are set at the center of the Si nanopellets (dashed line in the inset of Figure 2A), and the xy sections are set at the half height of the Si nanopellets.

At the reflectance dip B, the magnetic fields are strongly localized in the Si nanopellet, forming a waveguide mode associated with a 92.1-fold enhancement in the field at the maximum. The electric field shows the greatest enhancement at the sidewall of the Si nanopellet, exceeding 60-fold in comparison to the incident electric field.

The magnetic-field distributions at the reflectance peak C are different from those at the dip B. Multipoles are observed in the Si nanopellet, exhibiting a higher-order magnetic resonance associated with more than a 120-fold enhancement in the field compared to that of the incident field. The electric fields are predominantly distributed at the outermost surface of the Si nanopellet; the intensity exhibits a 127-fold enhancement for that of the incident electric field. Thus, the enhanced electric fields available at the top of the Si nanopellet differ from those on the resonance in Figure 2B.

Although the reflectance peaks D and E are separated by a deep dip at 682.5 nm, their EM-field distributions are similar to each other. Figure 2D shows a strongly localized magnetic mode inside the Si nanopellets. More than 100-fold enhancements in the maximum intensities are observed for both electric and magnetic components, as compared to the incident field intensity. The enhanced electric fields are primarily distributed on the sidewalls of the Si nanopellets as shown in Figure 2D. Figure S1 in the Supplementary Material provides the field distributions at the reflectance peak E.

2.3 Enhanced PL spectrum and image

Figure 3A shows the PL spectra measured under excitation at 532.0 nm. The PL spectra of the atomic layer on and off the metasurface are shown with orange and gray curves, respectively. The PL spectrum off the metasurface is magnified by 10-fold, for clarity. The PL EF spectrum, defined as the ratio of the PL spectra on and off the metasurface, is shown with a black curve plotted on the right axis. The peak value of the PL EF spectrum is as large as 300 at 614.5 nm. The peak of the PL spectrum on the metasurface is located at 619.4 nm whereas that off the metasurface is located at 650 nm. Thus, in addition to the large PL EF, a distinct spectral shape change occurs in the PL enhancement on the metasurface.

Figure 3:

PL measurement on the metasurface (MSF). (A) Enhanced PL spectrum measured on the metasurface (orange). PL spectrum on the flat BOX layer (gray), which is 10-fold magnified for clarity. PL EF is plotted for the right axis (black solid). (B) Normalized reflectance spectrum of the TMDC-layer-transferred metasurface, measured at the normal incidence. (C) Confocal PL image, which is shown in orange-hot presentation. White scale bar indicates 200 μm. (D) Photon-number distribution (orange bars) evaluated from the confocal image in (B) and fitted using Poisson distribution (open black bars), which represents a quantum light effect, coherent single-mode emission [44].

We note that the PL peak of WS₂ monolayer on flat substrates can largely shift in a wavelength range from 620 to 670 nm, due to the substrate, transfer, and/or nm-scale strain [8, 9]. Thus, the PL peak at 650 nm off the metasurface in Figure 3A is a plausible result.

The enhanced PL-EF peak shows a good agreement with a reflectance dip at 616 nm in Figure 3B. The normalized reflectance spectrum was measured at the normal incidence on the TMDC-transferred metasurface, showing reflection of the coupled system of the TMDC monolayer and the metasurface. As a result, the reflectance is reduced below 650 nm, compared to the metasurface reflectance in Figure 2A, because of light absorption by the TMDC monolayer. Further, a small reflectance dip at 605 nm in Figure 3B is considered to originate from a reflectance peak C in the computation (Figure 2A). Thus, the measured peak of PL EF appeared near the resonance C of the metasurface. Note that, although there is another resonance of the reflectance peak D in Figure 2A, it does not contribute to the PL enhancement in 620–650 nm. Generally, prominent PL enhancement selectively appears at particular resonances [33, 40, 41]. The small dip at 616 nm is most likely a coupled resonance of the TMDC exciton with the optical resonance C in the metasurface. In short, the prominent PL enhancement at 614.5 nm takes place with help of a particular EM resonance in the metasurface.

Although many reports have addressed TMDC PL enhancement using artificially designed nanostructures [14–27], this type of resonant effect associated with definite changes of spectral shapes has not been reported, except for a previous report using plasmonic nanocavities [14]. This study moreover presents the largest value of PL EF for the TMDC monolayers, compared to those obtained by the dielectric nanostructured platforms previously reported.

We here refer to the optical properties of the as-grown TMDC atomic layers. The Raman-scattering and PL spectra are shown in the Supplementary Material (Figure S2). The Raman spectrum shows that the TMDC consists primarily of WS₂ with a small quantity of Mo. The PL spectrum has a peak at 625 nm, representing A exciton of WS₂. As we mentioned the previous reports [8, 9], transfer of TMDC film often resulted in PL-peak shift, which was observed in this study as well. As shown in Figure 3A, the TMDC transferred onto the SiO₂ layer has a PL peak at 650 nm.

High-spatial-resolution PL images were acquired using an upright confocal FL microscope (Stellaris 5, Leica Microsystems, Wetzlar, Germany). One of the PL images is shown in Figure 3C, where the pseudo-color representation is orange-hot and the white scale bar indicates 200 μm. Rectangular colored area corresponds to the metasurface. Evidently, the PL becomes very dark outside the metasurface though the TMDC atomic layer covers the outside as well. In the confocal image, the PL intensity outside the metasurface is indistinguishable from the zero level in the measurement. The confocal PL image was acquired in a photon-counting manner; therefore, the PL intensity was quantified by examining the number of emitted photons.

Setting an analyzing region on the metasurface (Figure S3), the distribution of photon numbers was obtained, as shown in Figure 3D. The vertical axis represents number of photon-emission events. The photon numbers (orange bars) are small and primarily limited to five or fewer, including 14.4 % zero-photon and 29.7 % one-photon events. The distribution was fitted using the Poisson distribution and reproduced fairly well, being found to have mean photon number of 1.47. Small deviation is probably due to structural imperfection of the TMDC monolayer: the triangular bilayers and the broken holes as seen in Figure 1C and D, respectively. The fitted Poisson distribution is plotted in Figure 3D using open black bars. The photon distribution obeying Poisson distribution represents coherent single-mode photon emission [44], which is direct evidence for a quantum light emitter. It thus turns out that the confocal PL image (Figure 3C) directly visualizes a TMDC quantum light emitter. As a corollary of the coherent single-mode emission, the degree of second-order coherence g ⁽²⁾ is derived to be unity, irrespective of the mean photon numbers [44]. Note that the confocal PL imaging is more informative than the conventional g ⁽²⁾ measurement, which is reported frequently as experimental evidence of single-photon emission, and completes in much shorter time of tens of seconds than the time-consuming g ⁽²⁾ measurement.

2.4 Decay-time analysis of enhanced PL

Figure 4A shows the PL decay profile in a ps range. The PL was excited using laser pulses with 2-ps duration and 540-nm wavelength and was detected around the PL-peak wavelengths. The instrumental response function (IRF) of the laser pulses in the setup used is shown with a green curve, and it has a full width at the half maximum of 10 ps. The initial response of the PL decay curve follows the IRF. The ultrafast response suggests that photocarriers are immediately consumed via a nonradiative process. In the time range later than 15 ps from the peak of the laser pulse, the measured data (orange closed circles) are fitted using an exponential curve (black solid curve) such that

where C ₀ is a proportional constant, x ₀ is time offset, and τ _s is decay time in the ps range. The fitting results as shown in Figure 4A revealed the value of τ _s = 22.3 ± 0.3 ps. The PL lifetime under the excitation condition was approximately 10 ps in a previous time-resolved study [45]. It was also reported that a WS₂ monolayer grown on SiO₂/Si had the decay time of 11.4 ps and a partially oxidized WS₂ monolayer had that of 31.5 ps  [11]. Thus, the measured τ _s lies in a plausible range.

Figure 4:

Temporal profiles of measured PL. (A) And (B) fast and slow components in ps and ns ranges, respectively. Measured data are shown with orange closed circles and fitted curves with solid curves in both (A) and (B). The fitting exponential functions, Equations (1) and (2) for (A) and (B), respectively, are described in the text. In (A), the PL intensity is shown on the logarithmic scale, and the temporal profile of ps excitation laser pulse is drawn with a green dotted curve for reference. (C) PL profile measured off the metasurface (orange crosses). IRF to the ps laser pulse (or the measured temporal profile) is also shown with green curves in (A) and (C).

To analyze the temporal profile of PL in a ns range, in Figure 4B, the time-integrated PL intensity over [0, T] (orange dot) is plotted for the gate time T of a photon-counting detector. The growing temporal profile I(T) was defined, using an integrated double-exponential function, such that

and fitted the measured data (solid black curve) where a relation τ _L ≫ T was assumed to derive Equation (2). C, D, and τ _s are the fitting parameters and the fitted curve is shown with a black curve, which well reproduces the PL-growing profile. The short lifetime τ _s was evaluated to be 0.41 ± 0.16 ns; consequently, for gate time longer than 5 ns, the profile I(T) dominantly depends on T and shows an approximately linear increase in Figure 4B. This result indicates that the τ _L is sufficiently longer than the maximum gate time of 11 ns. In Figure 4B, the total PL intensity coming from the long lifetime component, Dτ _L , is larger than that from the short component, Cτ _s , whereas the ratio of populations in the fast and slow components, i.e., C : D is determined to be 11.8 : 1 using by Equation (2). These results imply that there are two possible excitonic transitions in the TMDC monolayer. A theoretical study on the exciton bands of monolayer MoS₂ showed that the A excitons have K–K direct and K–Γ intervalley transitions at approximately the same energy [46], where K and Γ denote the high-symmetry points in the Brillouin zone. Because WS₂ is considered to have similar exciton bands to those of MoS₂, the two excitonic transitions are attributed to the two observed PL components.

We mention that the τ _s is approximately one-order longer than that found in the ps range. This is because the temporal resolution in Figure 4B is limited by the minimum gate time of 0.5 ns. Thus, the evaluated τ _s in the ns range suggests that τ _s is shorter than 0.4 ns and does not imply inconsistency with the analysis in the ps range (Figure 4A). The precise evaluation of τ _s is conducted in the ps range.

The PL decay was also measured off the metasurface and then, the TMDC layer was on placed the SiO₂ layer. These data are shown with orange crosses in Figure 4C. The ps-laser-pulse IRF is shown together with a green curve, being in agreement with the PL profile. The PL decay was also fitted by an exponential curve (not shown here) and the decay time was 4.7 ± 0.1 ps. These results indicates that the PL decays together with the incident laser pulse and strongly suggests the PL predominantly decays within 5 ps via ultrafast nonradiative process(es) because the PL intensity is heavily reduced in comparison with that of the TMDC monolayer on the metasurface. In other words, the photocarriers in the TMDC film off the metasurface are immediately consumed through the nonradiative path(s).

2.5 Resonantly enhanced PL processes

Figure 5A and B shows schematics of the PL dynamics on and off the metasurface, respectively. Blue lines denote the TMDC monolayers. Schematics of the structures and energy diagrams are drawn on the left and right sides, respectively. On the metasurface (Figure 5A), photoexcited carriers are considered to have two major relaxation paths based on the analysis of PL decay time: one is the nonradiative process(es) and the other is the excitonic transitions. Green arrows indicate excitation using the laser light. Excited-state transfer is represented with solid black arrows. Radiative and nonradiative transitions are shown with red and dashed black arrows, respectively. The thickness of the arrows represents the probability of each process; the thicker arrows denote higher probabilities. As shown in the temporal profiles (Figure 4), the excitonic transitions have fast and slow components. However, these components are indistinguishable in the PL spectrum, because the transitions occur at approximately the same energy [46]. In Figure 5A, the vertical red arrow denotes the direct excitonic transition and the oblique red arrow does the indirect (or intervalley) transition.

Figure 5:

Schematic of the PL dynamics: (A) and (B) TMDC monolayers (blue) on and off the metasurface, respectively. Structural schematics are drawn on the left side; excitons are represented with orange rounds. Energy diagrams are depicted on the right side. Green arrows represent photoexcitation. Solid red and black arrows denote radiative and excitation transfer processes, respectively.Dashed black arrows indicate nonradiative decay. See more details in the text.

In contrast, off the metasurface (Figure 5B), the PL dynamics are attributable to dominant nonradiative process(es), indicating that nearly all the photocarriers are consumed without emitting PL. A damping oscillator model [47] clarified that the excitation transfer on the top of dielectric becomes drastically fast, when the distance between the excited state and the surface of dielectric is close within a few nm [40]; the transfer rate exceeds 10⁵ at the distance of 0.1 nm, is more than 10³ at 0.5 nm, and is in the one-digit order at 5 nm and more. Although the excited states in the TMDC are not simple isolated oscillators, it is probable that the excitation transfer from the TMDC layer to the SiO₂ layer takes place in a similar manner, as indicated by the black arrows in Figure 5B.

2.6 PL EF

The PL EF in the experimental configuration is expressed as [33, 40, 48]

where the subscripts m and 0 denote on and off metasurface, respectively. N is the photoexcited populations, η is the quantum yield, and γ(k) is the radiative decay rate dependent on outgoing wavevector k. In particular, η = γ/(γ + γ _NR) where γ _NR means nonradiative decay rate. The ratio γ _m /γ ₀ is often referred to as the Purcell factor [49].

The electric-field intensity at the excitation wavelength (Figure S4) shows that the TMDC monolayer was excited efficiently by a maximum of 5.3-fold, which gave the ratio N _m /N ₀. Other terms in Equation (3) were based on the PL processes and Purcell effect.

The Purcell effect was often discussed with PL enhancement of the TMDC atomic layers [20, 27]. In a simple scheme describing two-level systems in optical resonators, the spontaneous emission is expedited through the resonant electric fields. We numerically explored the Purcell effect in this configuration using the all-dielectric metasurfaces; the details are noted in the Section 4.6. In case of z polarization, the total Purcell factor was as large as 180 at the maximum, whereas, for the in-plane (i.e., x or y) polarization, it was approximately 2 in the range of 600–650 nm (Figure S5). The total Purcell factor contains both radiative and nonradiative effects. The A exciton in the TMDC is limited to the in-plane polarization [50]. Further, a radiative effect is observed in the PL measurement. Overall, the factor γ _m /γ ₀ is at most 2 in Equation (3) and the other contribution comes from the ratio η _m /η ₀, which is estimated to be 30 or less. As illustrated in Figure 5, the change in exciton relaxation dynamics is a major factor yielding the large PL EF (Figure 3A).

4.1 TMDC growth

Atomic-layer WS₂ films were grown on c-plane sapphire substrates via cold-wall chemical vapor deposition at 800 °C and 50 Torr for 60 min. H₂S and WOCl₄ were used as sulfur and tungsten precursors, respectively. The flow rate of H₂S was 1 sccm. The feed rate of N₂ carrier gas through the WOCl₄ canister was 300 sccm, and the temperature and pressure of the canister were maintained at 45 °C and 760 Torr, respectively. The total N₂ flow rate in the reactor was 2500 sccm. Additionally, 0.1 sccm O₂ was injected during the deposition of WS₂. Dragontrail glass serving as a catalyst reservoir was placed upstream of the sapphire substrate during growth, providing Na and K catalysts and promoting the lateral growth of WS₂ with enlarged grain size.

Due to the previous usage of the chamber for the growth of MoS₂ monolayers, a trace amount of Mo was found in the Raman spectrum of the as-grown film (Figure S2); however, because the ratio of Mo is estimated to be small (10 % or less) [13] and plays a minor role in the optical properties, we simply refer to the TMDC as WS₂.

4.2 Transfer process onto metasurfaces

The as-grown monolayer WS₂ films were transferred from the sapphire onto the metasurface. The procedure is schematically illustrated in Figure S6. Before the delamination, the WS₂/sapphire sample was coated with polymethyl methacrylate (PMMA) to support and protect the ultrathin WS₂. Subsequently, the PMMA/WS₂ stack was peeled off from the sapphire surface by allowing deionized water to penetrate the interface between the sapphire and WS₂. Finally, the sample of PMMA/WS₂ on the above-mentioned metasurface was immersed in acetone at 50 °C for 60 min to remove the PMMA and then cleaned in isopropanol, followed by drying with a nitrogen blow.

After the transfer, rapid temperature annealing was performed, which started at room temperature; the temperature was raised to 300 °C and kept for 10 min under N₂ gas flow at 120 sccm; subsequently, the sample was cooled down to room temperature under N₂ gas flow. This annealing process improved the PL intensity a few times at the maximum.

4.3 Nanofabrication of metasurfaces

Metasurfaces were fabricated based on the Si-on-insulator wafers; the top Si layer was 200 nm thick, the middle SiO₂ layer was 375 nm, and the base Si wafer was 675 μm. The top Si layer was crystalline with the (100) surface and had a p-type dopant. The nanofabrication procedures using electron-beam lithography and selective dry etching of the Si layer have been described elsewhere in detail [34, 42].

4.4 Optical measurement

The PL of the TMDC on the metasurfaces was measured at room temperature using a micro-PL setup for the spectrum acquisition and a confocal microscope for the spectrally resolved PL images. In the micro-PL setup, single-mode continuous-wave laser light of emission wavelength 532.0 nm (Action532Q-0050, AOTK, China) was focused on the TMDC monolayer on the metasurface using a 50× objective lens with a numerical aperture (NA) 0.55 (M Plan Apo, Mitsutoyo, Kawasaki, Japan) and a working distance of 13.0 mm. The PL from the TMDC monolayer was collected using the objective lens. The laser power was set to 0.5 mW or less at the entrance of the objective lens. The collected PL was passed through a monochromator (ISO-160, Teledyne Princeton Instruments, Trenton, NJ, USA), and the PL spectra were acquired using a cooling CCD camera (ProEM1024HS, Teledyne Princeton Instruments). The laser power was set to 0.5 mW or less at the entrance of the objective lens. When comparing the PL intensities, we kept the laser power to be a constant value.

Normal reflectance in Figure 3B was measured using the setup for the micro-PL measurement above. A 10× objective lens of NA 0.28 (M Plan Apo, Mitsutoyo) was used. Light source was a halogen lamp and two apertures were inserted at the entrance of the objective lens and an imaging lens (MT-L, Mitsutoyo). The apertures were set to be narrow and the incident angle of incoming beam was limited to be less than 0.3°, ensuring the normal incidence practically. Reference signals were measured using a calibrated Al mirror. Thus, we quantitatively evaluated reflectance of the TMDC–metasurface coupled system.

For the confocal PL image shown in Figure 3C, the excitation wavelength was set to 514.0 nm and a 10× objective lens of NA 0.32 (HC PL Fluotar, Leica Microsystems) was used. The lateral resolution was 819.2 nm. The PL growth in the ns range as shown in Figure 4B was measured by varying the gate width of the photon-counting detector in the confocal microscope.

PL decay-time measurements were conducted using a 76-MHz-repetition pulsed laser (MIRA-OPO-X ps, Coherent, Santa Clara, CA, USA) with an emission wavelength of 540 nm and a pulse width of 2 ps. The pulsed laser light was focused using a 50× objective lens with NA 0.8 and a working distance of 1.0 mm (MPFLN50X, Olympus, Tokyo, Japan). The laser power injected in the objective lens was 20 and 100 μW for the TMDC monolayer on and off the metasurface, respectively. The PL was collected using the same objective lens. The detection wavelengths were set to 615–620 and 640–650 nm on and off the metasurface, respectively, in accordance with the PL peaks (Figure 3A). The PL in the ps range was acquired using a synchronously scanning streak camera (C5680, Hamamatsu Photonics, Hamamatsu, Japan), attached to a monochromator (250is, Chromex, Albuquerque, NM, USA). The IRF with a full width at half maximum of 10 ps (Figure 4A and C) was determined by measuring the time traces of the excitation pulses.

4.5 Computation of reflectance spectra and EM-field distributions

The reflectance spectrum in Figure 2 and the emittance spectrum in Figure 3 were computed based on the rigorous coupled-wave analysis (RCWA) method [51], which was combined with a scattering-matrix algorithm [52] for numerical stability. The RCWA code was implemented on a supercomputer. The EM-field distributions in Figure 2 were output using the RCWA method as well. A material parameter, that is, the permittivity of crystalline Si constituting the metasurface were taken from literature [53]. The permittivities of air and SiO₂ in the wavelength range of interest were set to their representative values of 1.00054 and 2.1316, respectively.

4.6 Numerical evaluation of purcell factor

To evaluate the Purcell factors in Equation (3), the excitons in the WS₂ layer were approximated as two-level energy quantum systems. Their Purcell factors were calculated using the macroscopic quantum electrodynamic theory. The excitation created by an electric dipole source is connected in quantum mechanical terms with the relaxation of the exciton. In the real terms, the relaxation of the exciton was obtained by solving the Maxwell equations for a point dipole excitation in a complicated environment, which includes the EM response of the involved materials. A commercial finite-difference time-domain (FDTD) software (Ansys, Canonsburg, PA, USA) was used to solve the Maxwell equations. The optical responses of the different materials were determined from their experimentally measured dielectric permittivities [53].

Physically, the Purcell factor, Γ(r, ω), gave the enhancement or inhibition of the relaxation rate of the exciton when the WS₂ monolayer was placed in the nanostructured environment, compared to the reference relaxation value of the WS₂ monolayer. A reference value was used for the case where the WS₂ layer was placed in a homogeneous media. In the reference media, the WS₂ excitons followed an exponential relaxation, exp(−t/τ _ref), from the excited to the ground state, where τ _ref is the reference lifetime. In the presence of the nanostructured environment, the relaxation was accelerated by the Purcell factor value to exp(−Γt/τ _ref). The Purcell factor is given by Γ = τ _ref/τ _NS, where τ _NS is the lifetime of the exciton in the nanostructured environment. The factor Γ can also be expressed as Γ = γ _m /γ ₀, which is a factor on the right-hand side of Equation (3).