Haifeng Xu*
School of Mechanical and Electronic Engineering, Suzhou University, Suzhou 234000, PR China. E-mail: xuhaifeng@ahsztc.edu.cn
First published on 26th April 2017
MoS2 monolayer is a member of transition metal dichalcogenides (TMDs), which has shown great potential for applications in light-emitting devices due to a direct band gap in its electronic structure. However, a region of atmomic thickness remains a challenge due to its weak light-matter interaction. Most approaches to improve the light–matter interaction of MoS2 have been devoted to plasmonic nanostructures and photonic crystal. Herein, we demonstrated a method based on the insertion of a gold mirror layer between the dielectric layer and the substrate to enhance the optical absorption and emission of the MoS2 monolayer. The hybrid nanostructure comprised a 35 nm SiO2 dielectric spacer and 40 nm golden film. The overall photoluminescence intensity was amplified nearly 4 times, resulting from enhanced optical absorption and Fabry–Perot cavity reflection. We further demonstrated that the improvement in the PL intensity can also be modulated by selecting a spacer material and changing the layer thickness. This study is broadly applicable to enhance the light–matter interaction of TMDs for applications in potential devices such as light-emitting devices, lasers, and photodetectors.
In this study, we demonstrated that a mirror layer in a substrate played a significant role in the optical gain of the MoS2 monolayer. MoS2 monolayers were successfully transferred onto SiO2/Si and SiO2/Au/Si substrates via wetting transfer technology. The spectral change due to different substrates was systematically investigated to elaborate the mechanism of the mirror layer-enhanced light–matter interaction. The absorption spectra were dramatically improved in the case of a hybrid substrate consisting of 35 nm SiO2 and a 40 nm gold mirror layer, which was due to Fabry–Perot interference-induced multi-reflection light absorption. Moreover, the exciton recombination rate was dramatically increased with the field enhancement, resulting in a 4-fold PL enhancement of MoS2 on the hybrid substrate. We proposed that by performing electromagnetic simulations, light–matter interaction of MoS2 monolayer can be engineered by selecting the spacer material and spacer thickness for broadband system. This study provides an opportunity to optimize the light–matter interaction of 2D materials and further improve the performance of 2D optoelectronic devices.
Fig. 1c and d show the optical images of the MoS2 monolayer transferred on different substrates. CVD-grown MoS2 on the 35 nm SiO2/Si substrate can be clearly observed as the blue triangular region, and some dark blue dots were the two-layered MoS2 domains.20 The color of the MoS2 monolayer was dark red with poor optical contrast, which was related to the Fabry–Perot interference. The side length of the triangular MoS2 monolayer was about 100 μm, and some wrinkles and defects were caused during the transfer process. The properties of the substrates, as shown in Fig. 1e and d, were studied at the tilted view of 45°, where a cross section of the substrate was cut using focused ion beam (FIB) with a Ga source. The hybrid layer of 30 nm SiO2 and 40 nm Au was clearly observed as the dotted region shown in Fig. 1e at low resolution, and the scale bar is 300 nm. The dotted region was amplified using a high magnification objective lens, as shown in Fig. 1f, with the scale bar of 30 nm. The Au layer is a conductive film and appears as a bright layer in the SEM image. Although the surface was slightly rough, the average thicknesses of SiO2 and Au were about 35 nm and 40 nm, respectively.
The MoS2 monolayer samples were grown via a chemical vapor deposition (CVD) method.21 The sulfur (S) and molybdenum oxide (MoO3) powder were prepared as the S and Mo source, respectively. The S powder was located at the upstream region of the furnace, and the MoO3 powder in a grown boat was placed in a fused quartz tube at the center of the CVD furnace. The samples were first grown on a 300 nm SiO2/Si substrate and were transferred onto the prepared substrates via a wetting transfer process. The SiO2 film and Au film were both deposited using an e-beam deposition system. A home-built spectrometer was used for obtaining Raman and PL spectra using an iHR550 Raman spectrometer (Horiba) and a CCD imaging system. The objective lens had 50× magnification, and the excitation laser wavelength was 488 nm. The working gratings were 600 g mm−1 for PL detection and 2400 g mm−1 for high resolution Raman acquisition. Fig. 2a shows the Raman spectra of the MoS2 monolayer on the SiO2/Au/Si substrate. The E12g peak was obtained at 389.0 cm−1, representing in-plane vibrations, and the A1g mode peak was acquired at 410.3 cm−1, representing the out-of-plane mode. The difference in the characteristic frequency between E12g and A1g phonon modes of the MoS2 structure was determined to be 20.7 cm−1.22 From both the optical image contrast and Raman frequency difference, the samples were verified as monolayers. A Raman peak map of the MoS2 monolayer integrated at the A1g mode is shown in Fig. 2b, confirming the homogeneous properties of the CVD-grown samples.
The spectral properties of the MoS2 monolayer on different substrates were further investigated, and the results are shown in Fig. 3. Fig. 3a shows the optical absorption of the MoS2 monolayer on different substrate, and the thickness of the MoS2 monolayer was about 0.7 nm. The overall spectral intensity (red line) significantly enhanced on the SiO2/Au substrate with the obvious exciton A and B peaks near 660 nm and 615 nm, respectively. When the MoS2 monolayer was placed on the substrate without a gold mirror layer, no clear spectral peak was observed due to poor absorption. The maximum enhancement of optical absorption was nearly 6-fold, and this contrast could be attributed to the Fabry–Perot cavity. Fig. 3b shows the difference in the absorption cross-section simulated by (Finite-Difference Time-Domain) FDTD solutions. The dielectric constant of the MoS2 monolayer was utilized by parameterizing the experimental data into a band exciton transition (BET) model, as reported in the previous work.23 The absorption cross section of the MoS2 monolayer was enhanced more than two times, which was simulated by FDTD solutions and corresponded to the experimental phenomenon. Because of the absorbance enhancement, light–matter interaction was increased. This generated more exciton in the band gap. Therefore, PL intensity dramatically amplified under the 488 nm excitation laser.
As shown in Fig. 3c, total PL intensity increased nearly 4-fold with the laser power of 1.6 mW. The main PL peak is composed of primary exciton A0, trion A− (two electrons bounding a hole), and valence band splitting-induced B exciton. The peak energy of the exciton A0, exciton A−, and exciton B was fitted by Lorenz peaks near 1.83 eV (678 nm), 1.80 eV (689 nm), and 1.96 eV (633 nm), respectively.24 Moreover, the binding energy of the exciton A0 is ∼40 meV, which agrees with the value reported earlier. In addition, the spectral shape of the MoS2 monolayer was further altered with the gold mirror layer, which was due to the selective amplification of the primary exciton A0. Moreover, the intensity of exciton recombination was obviously attenuated, as shown in the fitting curves of Fig. 3c. Fig. 3d shows the electric field enhancement at the surface of the substrate, where the exciton was treated as the dipole radiation source. The electric field was increased nearly three times, as observed from FDTD simulations.
In a dipole–dielectric–metal structure model, the dipole emission is directly associated with the electric-field enhancement, which is known as the Purcell effect. The Purcell enhancement factor can be defined as FP ≡ Γ1/Γ0,25 where Γ1 and Γ0 are the radiative decay rates on SiO2/Si and SiO2/Au/Si substrates, respectively. Basically, Γ is directly proportional to the square of electric (E) and electric dipole (d). Fig. 3e and f show the ultrafast pump–probe signals of the MoS2 monolayer on different substrates. In the ultrafast pump–probe measurement, wavelengths of pump and probe were 400 and 675 nm, respectively. A 100 fs laser pulse with a pump fluence of 13.2 μJ cm−2 excited the electron–hole pairs in the MoS2 monolayer. The decay signals could be fitted by decay functions, as expressed by y = y0 + Ae(−t/τ), wherein the emission rate can be estimated from the experimental results. The decay rates of MoS2 on different substrates were obtained as 25.6 ps (Γ1) and 11.1 ps (Γ0). As the emission rate Γ1 was larger than Γ0, the spontaneous emission of MoS2 monolayer was enhanced and restricted the formation of trion and exciton B due to limited band-filling effect.18,26 In MoS2 semiconductor, hot carriers excited by laser illumination were thermalized to the lowest exciton energy level and recombined via radiative or nonradiative decay. Under intense light, when the optical carrier generation rate was much lower than the recombination rate, thermalized carriers accumulated from the band edge and quasi-Fermi level shifted deeper into the conduction and valence bands in a steady state. Although multi-reflection strengthens the light–matter interaction of the MoS2 monolayer, providing more excitons, there were still not enough excitons to fill the exciton states due to field enhanced emission rate. Moreover, MoS2 primary exciton emission was selectively amplified, which resulted in the overall PL enhancement.
To systematically investigate the enhancement of the light–matter interaction of the MoS2 monolayer via Fabry–Perot interference, simulations of electric field and optical absorption depending on mirror materials and dielectric layer thickness were conducted at normal light incidence. In the present study, Fabry–Perot interference originated from multiple interfaces among air/SiO2 and SiO2/Au contact surface, indicating destructive and constructive interference in the SiO2 layer. In Fig. 4a, Si substrate, metal film (40 nm), and SiO2 layer (35 nm) were stacked in adjacent regime from bottom to top along vertical direction. The MoS2 monolayer was located at the surface of the SiO2 layer. The calculated map of the electric field intensity (|E|2) is displayed in the x–z plane with the normal incident light wavelength of 680 nm. The resulting |E|2 profiles in air were found oscillatory along the z direction with a period of 340 nm. The refractive index of air, SiO2, Au at 680 nm was considered as 1, 1.5, and 0.2, respectively. The dependence on the metal materials was further investigated among Al, Au, Ag, and Cu. The layer thickness was set constant, and the calculated electric field intensities (|E|2) on the surface of the SiO2 layer are shown in Fig. 3b. The intensity increased in the order of Ag, Cu, Au, and Al because Ag exhibited ultra-low dissipation and strong reflectance in the visible range. A proper mirror material in the proposed substrate can be optimized to achieve strong light–matter interaction of the MoS2 monolayer.
In the classical Fabry–Perot cavity, constructive interference occurred at the condition 2nd cosθ = mλ, where n = 1.5 is the refractive index of SiO2 between two mirror surfaces, d = 35 nm is the thickness of the SiO2 layer, θ is the angle of the incident light measured from the vertical line against the reflecting surface, m is the order of interference, and λ = 680 nm is the wavelength of the radiation in vacuum. The separation difference between two consecutive constructive interferences (m = 1) was estimated to be 226 nm (Δd ≈ λ/3). However, the interference condition was different in the current case because there was only one metallic reflecting layer at one side. The simulations of MoS2 absorption at the surface can be performed to optimize the results. Fig. 4c shows the map of MoS2 absorption intensity distribution from 540 to 700 nm, which was plotted with different thickness of the SiO2 layer ranging from 30 to 310 nm. All the substrates with 30–40 nm and 240–310 nm SiO2 film showed dramatic enhancement of both absorption peaks. There appears a period of absorption maximum, which is consistent with the changing order of the diffraction of the Fabry–Perot interference. These results can be applied for improving the light–matter interaction of 2D materials by designing a hybrid substrate.
This journal is © The Royal Society of Chemistry 2017 |