Plasmonically enhanced photoluminescence of monolayer MoS 2 via nanosphere lithography-templated gold metasurfaces

: We demonstrate a simple, cost-effective method to enhance the photoluminescence intensity of monolayer MoS 2 . A hexagonal symmetric Au metasurface, made by polystyrene nanosphere lithography and metal coating, is developed to enhance the photoluminescence intensity of monolayer MoS 2 . By using nanospheres of different sizes, the localized surface plasmon resonances of the Au metasurfaces can be effectively tuned. By transferring monolayer MoS 2 onto the Au metasurface, the photoluminescence signal of the monolayer MoS 2 can be signi ﬁ cantly enhanced up to 12-fold over a square-centimeter area. The simple, large-area, cost-effective fabrication technique could pave a new way for plasmon-enhanced light-mater interactions of atomically thin two-dimensional materials.


Introduction
Atomically thin two-dimensional transition metal dichalcogenides (2D TMDCs) have been intensively investigated in recent years due to their unique physical and chemical properties [1][2][3][4].With rapid development of nanofabrication and nanosynthesis techniques, their optical, electrical, electrochemical, and structural properties have been constantly improved [5][6][7][8][9][10]. Being a direct band gap semiconductor, monolayer (1-L) MoS 2 promises great potential for a wide range of applications such as field effect transistors [11], photodetectors [12,13], and flexible electronics [14,15].However, due to its low dimension, the light-matter interaction for MoS 2 is extremely weak; in addition, the defects are inevitable during the process of fabrication.As a result, the quantum efficiency inside the MoS 2 is usually less than 1% [16,17] and the photoluminescence (PL) of MoS 2 is extremely weak as well, limiting its potential applications to a great extent.
Various strategies have been put forward to enhance the PL emission of 2D TMDCs, which can be mainly divided into two categories: chemical treatment and physical assistance.Chemical treatment for the 1-L MoS 2 improves the PL intensity by fixing the atom vacancies and modifying the electron density [18][19][20].By chemical treatment on 1-L MoS 2 , the quantum efficiency can be improved to near unity from 0.6% and the PL intensity can experience an over 150 times enhancement.However, the chemically treated samples require extremely dry air environment, which is not favorable for the subsequent device fabrication and processing.Physical assistance mainly depends on the localized electromagnetic effects that can be created by either plasmonic [21][22][23][24][25][26][27][28][29], dielectric [30][31][32][33][34][35] or metaldielectric hybrid [36] nanoparticles/nanostructures.Comparatively, plasmonic nanomaterials provide large PL enhancement factor and wide tunability since the plasmonic resonances are highly dependent on the size, the shape and the spatial arrangement of nanoparticles/ nanostructures.In particular, plasmonic metasurfaces (2D subwavelength metallic nanostructures) are highly advantageous due to their large-scattering cross section and local electromagnetic field enhancement near the surface of the metallic nanostructure.As a result, they can greatly enhance the light-matter interactions between the plasmonic metasurface and 2D TMDCs, leading to the enhanced PL emission [37].As an ultrathin active layer, 2D monolayer TMDC becomes optically thick but physically thin to preserve the monolayer properties, thereby potentially leading to many optoelectronic applications with ultrathin, compact and integratable features.
However, due to the subwavelength feature of plasmonic metasurfaces, it still remains a grand challenge to achieve large-area, cost-effective fabrication strategy.Most of the state-of-the-art plasmonic metasurfaces have been made through the sophisticated nanofabrication techniques, such as electron-beam lithography (EBL), focusedion beam (FIB) milling, self-assembly, etc.For examples, Butun et al. prepared Ag nanodisk arrays using EBL to enhance the PL intensity of 1-L MoS 2 [22].By adjusting the size of Ag nanodisks (106-227 nm), the PL intensity was increased by 12 times.They proved that the PL enhancement came from the enhancement of the excitation field at the wavelength of pump laser and the effective scattering at the wavelength of exciton emission.Lee et al. proposed the MoS 2 -Ag "bow-tie" resonant array system [23].By adjusting the spacing of Ag nanoparticles in the silverbowtie nanoantenna array, a strong coupling between MoS 2 excitons and the surface plasmons on the grid surface was achieved.Through localized surface plasmon resonance (LSPR) effect and Fano resonance, the PL intensity of 1-L MoS 2 had been increased by 40 times.More economical methods to enhance PL intensity were demonstrated by randomly attaching Au nanoparticles to TMDCs [24,25].Upon excitation, the created LSPRs from the Au nanoparticles will modify the local density of states of the TMDCs, hence causing the enhanced PL intensity.Although the PL of TMDCs can be largely enhanced via the random assembly of plasmonic nanoparticles, there is still room for improvement in terms of large-area uniformity and enhancement factor.
The aforementioned EBL and FIB techniques require complicated, costly processes and intensive, skillful labor, which are nonpreferred for large-area fabrication.
Comparatively, self-assembly is a simple, cost-effective technique but lacks the accurate control of patterning, especially for plasmonic nanoparticles.Large-area self-assembly of polystyrene (PS) nanosphere can be achieved with the nanosphere size of 200 nm [38].However, additional complicated surface modifications for colloidal nanosphere are required.In our previous report [39], we have developed an improved self-assembly method to obtain a large-area, high-quality monolayer PS nanospheres with simple additional steps compared to the conventional self-assembly method.By exploiting the monolayer PS nanospheres as a templated mask, we can fabricate a variety of periodic three-dimensional hierarchically ordered metastructures using cyclic inductively coupled plasma etching technique.In this work, hexagonal Au metasurfaces are fabricated using the nanosphere lithography technique, which is based on the templated masks via the selfassembled polystyrene nanospheres.By transferring the monolayer MoS 2 onto these Au metasurfaces, the PL signal of the monolayer MoS 2 can be significantly enhanced.Upon excitation by a 405 nm laser beam, the plasmonically enhanced monolayer MoS 2 exhibits a 12-fold increase in PL intensity over a large area.Experimental results show that our proposed approach is a simple, large-area, and costeffective fabrication technique, which could pave a new way for plasmon-enhanced light-mater interactions such as photoexcitation and nonlinear optics of 2D TMDCs.

Templated masks preparation
The PS nanosphere latex solutions (2.5 wt% dispersion in water) with different nanosphere sizes in diameter (350, 400, 500, 600, and 750 nm) were purchased from Alpha Aesar Chemical Co., LTD (Shanghai, China) and Shanghai Aladdin Bio-Chem Technology Co., LTD (Shanghai, China).Prior to making templated masks by nanosphere self-assembly, the glass substrates were subsequently cleaned with acetone, ethanol, isopropanol and deionized (DI) water in an ultrasonic cleaner (KQ-250DE, Kun Shan Ultrasonic Instruments Co., Ltd), and then followed by 15 min treatment in a UV-O 3 cleaner (BZS250GF-TC, HWOTECH).The templated masks for nanosphere lithography consisted of a monolayer of PS nanospheres, which were formed on the glass substrate via an improved self-assembly method at the air-liquid interface [33].

Gold metasurface fabrication
Once the monolayer-PS-nanosphere templated masks were ready, a 2 nm thick Cr layer and a 50-nm-thick gold film were sequentially deposited on the templated masks by an electron-beam evaporator (TF500, British HHV).The ultrathin Cr film served as an adhesion layer in our experiments.To reduce the surface roughness and strengthen the adhesion of the Au film simultaneously, the evaporation rates in the deposition process were controlled to be ∼0.2Å/s for the Cr film and ∼0.5 Å/s for the Au film, respectively.During the deposition process, the vacuum level inside the evaporator chamber was controlled at 5 × 10 −6 Torr.After deposition, the Au-coated templated masks were soaked into the DI water to remove the PS nanospheres with the help of ultrasonic cleaner.The removing process was carried out at the power of 90 W and lasted for 1 min.Finally, we achieved the gold metasurface on the glass substrates.

Transfer of 1-L MoS 2
The CVD-grown 1-L MoS 2 samples on the Si substrates (SIX-Carbon Inc., China) were used in our experiments.The MoS 2 was a continuous film with a size of 1 × 1 cm 2 , which can be easily stripped off from the Si substrate without any additional treatment.The Si substrate with the 1-L MoS 2 was slowly immersed into the DI water with a tilt angle of ∼45°.Due to the surface tension of water, the 1-L MoS 2 atomic film separated from the Si substrate and then floated on the water surface.Meanwhile, the Si substrate sunk into the water bottom.After that, the hydrophilically treated Au metasurface was used to pick up the floating MoS 2 film.We immersed the target substrate into the DI water, kept an angle of ∼45°with respect to the water surface, and pick up the MoS 2 film gently.Please also refer to the supplementary videos of the transfer process in the supporting information.

Characterization and measurements
Surface morphologies of the Au metasurfaces were investigated using the field-emission scanning electron microscopy (FESEM, Merlin, Zeiss) and atomic force microscope (AFM, Ahpha300, WITec).The optical spectra of the Au metasurfaces were collected by a UV-Vis-NIR microspectrophotometer (CRAIC technologies Inc.).We carried out the Raman spectra measurement and Raman mapping to confirm the transferred 1-L MoS 2 on the gold metasurface using a confocal Raman system (Alpha300, WITec).We used a 100× objective lens to focus the excitation laser light (working wavelength: 532 nm) onto the samples and collect the Raman signal to the spectrometer.For PL test, a laser with the working wavelength of 405 nm was used for the MoS 2 excitation, and a 100× objective with the numerical aperture of 0.8 was used to focus the laser beam onto the 1-L MoS 2 sample and collect the emitted PL signals simultaneously.A 450 nm long-pass filter was put in the PL collection light path to block the reflected laser light.The PL spectra were measured under a microscopy-combined spectrometer system (MA 01720, Princeton Instruments).To ensure the repeatability of the PL spectra and exclude the subtle differences between different areas on each sample, the PL signal was taken and averaged at randomly selected 10 positions within an area of 1 × 1 cm 2 on every single sample under the same experimental conditions.

Simulation
The optical fields and transmission spectra were calculated using the finite-difference time-domain (FDTD) method.The dispersion of gold is based on Johnson and Christy model.The simulation area was set to allow only one period of the gold metasurface in the x-y plane and enough space for the light source and power monitors.Auto-uniform meshing with the finest meshes (2 nm minimum mesh step) was used to achieve the accurate results.The light source was selected with the wavelengths ranging from 400 to 1300 nm.Three power monitors were separately located 5 nm above the Au metasurface, 1 μm above the light source and 3 μm below the gold metasurface to collect electric field intensity, reflected and transmitted optical signals.

Results and discussion
Figure 1a is a schematic illustration of the 1-L MoS 2 -covered Au metasurface.The surface morphologies of all the metasurfaces fabricated from the templated masks using the PS nanospheres with different sizes (350, 400, 500, and 600 nm in diameter) have been characterized under the SEM, as shown in Figure S1.As mentioned above, we used continuous monolayer MoS 2 rather than single crystalline one during the transfer process.Such a large-scale, CVD-grown MoS 2 monolayer covers the Au metasurface evenly.In addition, there was no requirement of a support layer during the wet transfer, hence avoiding pollution effectively [40].The purchased 1-L MoS 2 samples and the fabricated metasurfaces were characterized using AFM, as shown in Figure S2.The measured thicknesses of 1-L MoS 2 and metasurfaces are ∼1 and ∼50 nm, respectively.Figure 1b shows the optical image of 1-L MoS 2 -covered Au metasurface, where the metasurface was fabricated from the templated mask using the PS nanospheres with the diameter of 750 nm.A clear boundary of the MoS 2 is marked by the black dashed line.
During the transfer process, the 1-L MoS 2 could be folded, stacked or damaged due to improper handling.To ensure that the transferred MoS 2 was still a monolayer, we carried out Raman spectra measurement on the transferred samples.As known, for the 1-L MoS 2 , the E 1 2g (in plane) and A 1g (out of plane) are the main vibration modes that locate at ∼384 and ∼403 cm −1 , respectively [41,42].Depending on the experimental environment, the specific positions of these two characteristic Raman peaks of 1-L MoS 2 might be slightly different, however, their spacing is usually fixed.Figure 1c shows the Raman spectra of 1-L MoS 2 transferred onto the Au metasurface.We can see that after transfer, the wavenumber spacing between the A 1g and E 1 2g modes still remains 18 cm −1 , confirming that the transferred MoS 2 onto the Au metasurface is still a monolayer.In order to verify that the 1-L MoS 2 transferred onto Au metasurface still remains good quality, a 10 × 10 μm 2 Raman mapping covering the boundary was carried out, as shown in Figure 1d.The continuity and integrity of transferred 1-L MoS 2 can be directly resolved by the distribution and variation of the A 1g signal intensity.In the mapping results, the distribution of A 1 peak and E 1 2g peak are almost the same.
The fabricated Au metasurfaces consist of hexagonally arranged triangle prisms.Each triangle prism is expected to produce strong LSPRs upon excitation, which will then enhance the absorption and emission efficiency of the 1-L MoS 2 .In our study, we fabricated the Au metasurfaces with different periods and unit sizes, which were highly dependent on the templated masks by the self-assembled nanospheres of different sizes.It is straightforward that the LSPRs of Au metasurfaces are highly dependent on structural parameters.Figure 2 shows the measured and simulated spectra of the Au metasurfaces that were fabricated from the templated masks with different nanosphere sizes (350, 400, 500, 600, and 750 nm in diameter).From the simulated results in Figure 2d, for the Au metasurface made by nanospheres with the size of 350 nm, the absorptance peak locates ∼670 nm.With the nanosphere diameter changing from 350 to 750 nm, the period of the Au metasurface increases, and the individual Au unit structure becomes large as well (see Figure S1).As a result, the absorptance peaks of the Au metasurfaces demonstrate a clear red shift from ∼670 to ∼1200 nm, showing a broad absorption tuning range.Meanwhile, the transmittance and the reflectance spectra exhibit the similar trend as the absorptance ones.The experimental results are also in reasonable agreement with  the simulation and show a similar trend as the simulated results.It is worth mentioning that there are also discrepancies between experimental and simulated results, which could be mainly attributed to the fabrication errors and the defects in the fabricated samples.
We subsequently investigated the effect of Au metasurfaces on the PL emission using a semiconductor laser (405 nm) as the pumping source.We pumped the 1-L MoS 2covered Au metasurfaces at various pumping intensities.As a control experiment, we also studied the PL of the nontransferred 1-L MoS 2 as a function of excitation laser intensity, as shown in Figure 3a.It can be clearly seen that as the laser intensity increases, the PL intensity becomes strong.The primary PL peak locates at 672 nm.At a higher excitation intensity, we also notice an emergence of a secondary PL peak, locating at ∼620 nm.It is well-known that these two main peaks locating at 672 and 623 nm correspond to the A1 and B1 band transitions of 1-L MoS 2 [43,44].Compared to B1 band transition related to E 1 2g (in plane) mode, A1 band related to A 1g (out of plane) transition has higher PL intensity and more sensitive to the Au metasurface [18].Figure 3b shows measured PL intensity of the 1-L MoS 2 on the Au metasurfaces that were fabricated from selfassembled nanospheres with different sizes at a fixed pumping intensity.At the fixed pumping intensity of 7.8 mW/cm 2 , compared Figure 3b to Figure 3a, there is significant enhancement of A-exciton emission due to the addition of the Au metasurface compared to the nontransferred sample.It can be also clearly seen from Figure 3b that as the size of the PS nanospheres decreases, PL enhancement becomes significantly prominent.In addition, the secondary PL peak at ∼620 nm (corresponding to B-exciton emission) has been greatly suppressed, which might be attributed to the absorption of Au metasurfaces.
The PL spectra of the 1-L MoS 2 on different Au metasurfaces under various pumping intensities have been also investigated and the results are summarized in Figure S3.From Figure S3 and Figure 3b, we can see that there is different enhancement of PL signals for each Au metasurface.It is obvious that among the Au metasurfaces in our experiments, the one made by nanospheres with the size of 350 nm in diameter provides the largest enhancement.To directly illustrate the enhancement effect of Au metasurface, we have plotted the PL peak intensity as a function of the excitation laser intensity, as shown in Figure 3c.It is obvious that for all cases, the PL peak intensity consistently changes linearly with the excitation laser intensity.Therefore, with the PL intensity of the nontransferred case as a baseline, the slope of each fitting line indicates the enhancement factor provided by the Au metasurfaces.
The enhanced PL from the 1-L MoS 2 could be attributed to the enhanced absorption or emission or both.Upon the laser excitation, the Au metasurfaces could be resonantly excited.As a result, strong LSPRs are induced around each unit, producing a large |E| 2 near the 1-L MoS 2 and subsequently enhancing the absorption efficiency of the 1-L MoS 2 .
On the other hand, the strong LSPRs could also modify the local density of states in the 1-L MoS 2 due to the well-known Purcell effect [45][46][47], resulting in an additional emission pathway for the MoS 2 and further enhancing PL.We also derived enhancement factor of each Au metasurface from the slopes of each fitting line, as shown in Figure 4a.Among the investigated Au metasurfaces in our experiments, we can see from Figure 4a that as the nanosphere diameter increases, the enhancement factor gradually decreases.In our experiments, the Au metasurfaces made by the nanospheres with the size of 350 and 750 nm provide the maximum and minimum enhancement factors (i.e., 12× and 2×), respectively.We have also numerically calculated the local field enhancement of the Au metasurfaces under the excitation and emission wavelengths (405 and 677 nm, see Figure 4b and c), respectively.The calculation details can be found in the supporting information.We have also numerically calculated the near-field electric field distributions of the Au metasurfaces at the excitation and emission wavelengths, as shown in Figure S4 in the supporting information.From Figure S4, we can confirm that the PL enhancement mainly arises from single Au triangle plasmonic modes rather than the coupling effect between two triangle prisms because the triangles are spaced too far to form strong couplings.It is worth noting that the gap between the adjacent triangles can be further tuned via etching nanosphere first before the gold film deposition [48].By precisely controlling the gap size, strong plasmonic couplings can be achieved to further enhance the PL signal.It is also worth mentioning that we only numerically calculated local field enhancement around the Au triangles in the Au metasurfaces at the excitation and emission wavelengths in Figure 4b and c, meaning that we neglected the other areas where there are no Au triangles in the Au metasurfaces.
From the SEM images of Au metasurfaces in Figure S1, the Au triangles only account for a small fraction of the whole metasurface area.As a result, the achieved enhancement factor becomes smaller by taking into account of PL contribution from the whole area of Au metasurface.As shown in Figure 4b and c, |E avg | 2 decreases as the nanosphere diameter increases for both excitation and emission wavelengths.Interestingly, |E avg | 2 for both cases experiences the coincidently same trend as the measured enhancement factor, indicating that the strong LSPRs generated from the Au metasurfaces indeed enhance both absorption and emission of the 1-L MoS 2 .From Figure 4, we could foresee that as the nanosphere size further decreases, the enhancement factor will become larger.For example, the numerical results show that the enhancement factor will reach 23 upon using the nanospheres with the diameter of 200 nm.However, in our experiments, the fabrication of the large-area Au metasurfaces becomes challenging when the diameter of the nanospheres is less than 300 nm.The further improvement of the self-assembly process and precise gap tuning with small nanospheres is still undergoing.

Conclusion
In summary, we have demonstrated a simple, cost-effective approach to enhance the PL signal of the monolayer MoS 2 .Hexagonal Au metasurfaces were fabricated from the templated masks that were formed by self-assembled polystyrene nanospheres.By transferring the monolayer MoS 2 onto these Au metasurfaces, the PL signal of the monolayer MoS 2 can be significantly enhanced.Under a 405 nm laser beam excitation, the LSPR-enhanced monolayer MoS 2 exhibited a 12-fold increase in PL over a large area.We expect our proposed approach could pave the way for many potential applications based on plasmonenhanced 2D TMDCs.

Figure 1 :
Figure 1: (a) Schematic of a 1-L MoS 2 -covered Au metasurface on a glass substrate, and the crystal structure of 1-L MoS 2 is shown in the inset.(b) Optical image of 1-L MoS 2covered Au metasurface.A clear boundary of the MoS 2 is marked by the black dashed line.The arrow points out the crack on MoS 2 .Inset shows a typical SEM image of the Au metasurface fabricated from the templated mask using the PS nanospheres with the diameter of 750 nm.(c) Raman spectra of MoS 2 transferred onto metasurface.With the peaks fitting, the Raman shift between A 1g mode and E 1 2g remains 18 cm −1 .(d) Raman mapping for boundary area with the collected signal of A 1g peak.

Figure 3 :
Figure 3: The measured PL signals of (a) the nontransferred 1-L MoS2 as a function of the pumping laser intensity (0.8-7.8 mW/ cm 2 ), and (b) the transferred 1-L MoS2 on different Au metasurfaces made by the nanospheres with different sizes in diameter (350-750 nm) under the fixed pumping laser intensity of 7.8 mW/cm 2 .(c) The scatter plot for the PL intensity of A-exciton in 1-L MoS2, including samples nontransferred and samples transferred onto various metasurface.The straight line in the figure is the result of linearly fitting for the spots in the scatter plot.The color of the line corresponds to the color of the fitted spots.

Figure 4 :
Figure 4: (a) Enhancement factor for each structure to enhance PL intensity of a 1-L MoS 2 , and the average E-field intensity of each structure.The numerically calculated local field enhancement of the Au metasurfaces per unit area under the excitation (b) and emission (c) wavelengths (405 and 677 nm), respectively.