Exciton–Plasmon Coupling and Electromagnetically Induced Transparency in Monolayer Semiconductors Hybridized with Ag Nanoparticles

Abstract

Exciton–plasmon coupling in hybrids of a monolayer transition metal dichalcogenide and Ag nanoparticles is investigated in the weak and strong coupling regimes. In the weak coupling regime, both absorption enhancement and the Purcell effect collectively modify the photoluminescence properties of the semiconductor. In the strong coupling regime, electromagnetically induced transparency dips are displayed, evidencing coherent energy exchange between excitons and plasmons. Metallic nanostructures support collective surface oscillations of the electrons gas, known as localized surface plasmons (LSP) or localized surface plasmon resonance (LSPR), driven by incident radiation.1-3 The LSPR frequency (ωLSP) of silver and gold nanoparticles (NPs) can be effectively tuned by size, shape, and surrounding dielectric medium.1-3 At the resonance condition, LSPs efficiently confine incident photon energy into deep-subwavelength volumes and dramatically enhance local electromagnetic (EM) fields.1-3 In turn, the enhanced EM field strongly modifies the light–matter interaction in molecules, quantum confined nanostructures, and other excitonic systems residing within the fringing field of the nanoparticle.1-4 Coupling between excitons and LSPs occurs via a coherent dipole–dipole interaction.1, 2 Depending on the lifetime of the LSPs and the excitons, the interaction can be in the weak or strong coupling regime. Strong de-coherence effects, such as the fast relaxation of LSPs on the femtosecond scale, often limit the interaction to the weak coupling regime.1, 2, 5 In this regime, LSPs modify the absorption cross-section and, correspondingly, excitation rate of the excitonic system through a local field enhancement (γexc ∝|E|2, where γexc is the excitation rate and E is the amplitude of electric field). Similarly, the spontaneous emission rate (γrad) is also enhanced due to an increase in local density of optical states, known as the Purcell effect.1, 2 Since radiative and dissipative loss in metal NPs also introduces the possibility of emission quenching by dark modes6 and lossy surface waves,7 the overall enhancement in quantum yield (η = γrad/(γrad + γnonrad)) is therefore determined by the competition of radiative and nonradiative (γnonrad) rates.1, 2 To achieve a large quantum yield enhancement, local field and radiative rate enhancements need to be optimized while minimizing nonradiative losses.1, 2 In the strong coupling regime, a coherent coupling between LSPs and excitons overwhelms all losses and results in two new mixed states of light and matter separated energetically by a Rabi splitting that exhibits a characteristic anticrossing behavior of the exciton-LSP energy tuning.5, 8, 9 In this regime, a new quasi-particle (plexciton) forms with distinct properties possessed by neither original particle. The behavior of plexcitons needs to be described in a quantum electrodynamics scenario. However, most concepts on plexcitons still remain theoretical predictions and experimental demonstrations have been largely limited to a few solid state material systems, including molecules,8-10 quantum wells,11 and quantum dots.5 An intermediate regime, sometimes called Fano interference in the plasmonics field, is also expected when the interaction between excitons and LSPs is not sufficiently strong.2, 5, 8, 9, 12 Notably, Fano interference manifests as transparency dips in the LSP spectrum that is reminiscent of a Rabi splitting; however, the interference effect tends to rapidly fade with increasing detuning.2, 5 Two-dimensional (2D) semiconductors such as atomically thin films of transition metal dichalcogenides (TMDs) are emerging as a new platform to investigate and exploit exciton–plasmon coupling.13-18 Group 6 TMDs are well-known layered materials with the stoichiometry of MX2, where M is group 6 transition metal (Mo, W) and X is chalcogen (S and Se).19 Single layer MX2 are direct gap semiconductors with bandgaps in the visible to near IR frequency ranges. Excitons in single layer MX2 are strictly oriented in-plane,20 strongly bound,21, 22 and exhibit large oscillator strength.14, 21, 23 The robust mechanical structure of these 2D semiconductors allows ease of integration with photonic crystal cavities,16, 24-26 optical microcavities,27 and plasmonic nanostructures,15-17 offering exciting opportunities for engineering light–matter interaction and realizing novel photodetectors,28 photovoltaics,13, 29 light-emitting diodes,30 and lasers.24, 25 Several recent studies15-18 showed that hybridizing monolayer MoS2 with metal nanostructures results in various degrees of exciton–plasmon coupling and correspondingly enhanced light absorption and emission. While clear evidence of plasmon enhancement has been demonstrated, the material is largely limited to MoS2 and the extent of exciton–plasmon coupling remains elusive. In this communication, we report on plasmon-induced modification of light–matter interaction in 2D MX2 (MoS2, WS2, MoSe2, and WSe2) layers hybridized with Ag NPs. The hybrid structures were prepared by thermal evaporation of a metal thin film on top of monolayer MX2 and subsequent annealing. The size and morphology of the metal NPs were tuned through the thickness of the metal film, annealing temperature, and the surface condition of MX2 sheets. We show that photoluminescence (PL) intensity enhancement is achieved via both optical field enhancement and spontaneous emission rate enhancement. Comparative studies of different MX2 materials reveal that the emission quantum yield of the bare material determines the competition between enhancement and quenching effects. It was found that the quantum yield enhancement in originally bright materials such as WS2 requires a dielectric spacer layer to minimize quenching effects. Furthermore, we report room temperature observation of a distinct electromagnetically induced transparency dips that appear in the scattering spectra of a MoSe2-NP system, which indicates a coherent dipole–dipole interaction between exciton and LSP in the strong coupling regime. Strong exciton–plasmon coupling even at room temperature highlights the potential of this unique hybrid system for electro-optical modulation devices. Monolayer MX2 sheets were obtained by mechanical exfoliation of bulk single crystals onto quartz and SiO2/Si substrates. Bulk crystals (natural) of MoS2 were obtained from SPI Supplies, while WS2, MoSe2, and WSe2 crystals were grown by chemical vapor transport.31, 32 Figure 1a shows the optical and fluorescence images of a monolayer MoS2 sample on a SiO2 (90 nm)/Si substrate. All exfoliated samples were annealed in vacuum (10−5 mbar) at 150–200 °C to remove surface contaminations. Subsequently, a thin film of Ag (6 nm thick) was deposited on the samples under ultrahigh vacuum (≈8 × 10−7 mbar) by thermal evaporation followed by mild annealing at 150 °C for 40 min under nitrogen atmosphere. Upon annealing, the silver adlayer agglomerated into disk-shaped nanoparticles on monolayer MX2 surface (inset of Figure 1b and Figure S1 of the Supporting Information), presumably due to surface diffusion, similar to previously reported observations.33, 34 As shown in Figure 1a,b, the optical appearance of MoS2 changed dramatically after formation of Ag NPs. The PL intensity measured under identical conditions was found to be significantly stronger for MoS2 with Ag NPs as clearly seen in the fluorescence microscope images shown in Figure 1c,d. This enhancement is attributed to plasmon-induced enhancement of incident light absorption in addition to emission rate enhancement as will be discussed below. In order to eliminate contributions from interference due to the SiO2/Si substrate, we focus the following discussions on the results from samples prepared on a transparent quartz substrate. Figure 2a,b shows optical images for a monolayer MoS2 sample on quartz without and with Ag NPs, respectively. Disk-shaped Ag NPs of various sizes ranging from 20 to 120 nm with thickness of ≈10 nm were found to be uniformly distributed on MoS2 surface as shown in Figure 2c. (Note that the morphology of the particles on MoS2 was similar on both SiO2/Si and quartz substrates.) The particles were highly packed with interparticle spacing of several tens of nanometers or less. Electric field distributions for different optical excitation energies were simulated using finite-difference time-domain (FDTD) simulations. The structure of experimentally obtained nanodisks (Figure 2c) was used as input parameters for the simulations. Figure 2d shows the simulated 2D mapping of EM fields. Strong enhancement in EM field up to 50 times was found to occur in “hotspots” that are located at the narrow gaps of particles where strong dipole–dipole interactions between adjacent LSPs are expected.1-3 The average enhancement factor of EM fields over the simulated area was less than 10 since a large fraction of the surface is effectively inactive. In another set of samples, we modified the sample preparation method to obtain Ag NPs with significantly smaller sizes. Specifically, we washed the surface of exfoliated MX2 sheets with acetone and isopropanol prior to metal deposition. The presence of trace amount of solvent residues significantly affected the surface migration of Ag, resulting in formation of semispherical particles with an average size of around 34 nm. The smaller particle size is most likely due to surface impurities acting as nucleation centers and pinning sites.33 The morphology of the particles did not change with annealing (Figure 2g and Figure S1 of the Supporting Information). We refer to samples with the large (57 nm average size) nanodisks and small (34 nm average size) semispheres as LNP (large NP) and SNP (small NP), respectively, in the following text. FDTD simulation of Ag SNP shows interparticle hotspots with strong EM field enhancement but with a lower overall enhancement similar to the case of Ag LNP. Experimental extinction spectra of monolayer MoS2 with and without Ag NPs are shown in Figure 2i,j along with the FDTD-simulated spectra of LNP and SNP. The extinction spectra of the hybrid system consist of features that can be attributed to MoS2 excitons and LSP of the NPs. A broad peak centered at ≈700 nm observed for MoS2-Ag LNP hybrid and a broad background peak at ≈500 nm for MoS2-Ag SNP hybrid can be attributed to LSPR of Ag NPs based on the agreement with our FDTD simulation results. The broad LSP resonance spanning the visible to near IR spectrum can be explained by the large distribution of particle sizes, suggesting that both excitation optical field and spontaneous emission rate enhancement can be responsible for the observed PL enhancement. According to Fermi's golden rule, the radiative rate is proportional to the local photonic density of states, that is, γrad ∝ ρ(ω).1, 2 Therefore, the spontaneous emission enhancement should follow approximately the extinction spectra of LSPs.35, 36 In Figure 3a, PL spectra of bare MoS2 and MoS2-LNP hybrid are shown along with their intensity ratio (gray plot). The emission spectrum is dramatically enhanced at the low energy side (above 670 nm) for the hybrid system, revealing a wavelength-dependent enhancement spectrum, which agrees well with the coupled system's extinction spectrum. In contrast, the emission spectrum profile was not significantly modified for MoS2-SNP hybrid system. Notable changes were observed only below 650 nm (see Figure S6 of the Supporting Information). In this system, the local field enhancement is mainly responsible for PL enhancement due to the small overlap between LSPR and exciton emission peaks. Time-resolved PL is an effective approach for quantitatively studying the spontaneous emission rate enhancement due to Ag NPs.2, 35 However, the short lifetime of MX2 excitons (<100 ps) owing to the dominance of fast nonradiative relaxation37 as well as partial quenching by the metal NPs presents significant challenges in the measurement and interpretation of data. In order to eliminate the effects of EM field enhancement, we studied the emission enhancement in the excitation saturation regime. As shown in Figure 3b, the PL intensities of both the bare and hybrid samples exhibit a sublinear increase as a function of laser power at a 633 nm excitation. While the full saturation regime was not explored in this experiment to avoid sample damage, the experimental data were instead fitted with the empirical equation for saturation: , where IPL is the PL intensity, P is the excitation intensity, and a and b are constants. By extrapolation, the PL EF in the saturation regime is determined to be about 6.8, which is effectively the enhancement of quantum yield. The related radiative rate enhancement, i.e., Purcell factor for the radiative component, is also estimated to be around this value as will be discussed below. In order to investigate the EM field enhancement, we studied the excitation wavelength dependence of PL EF for monolayer MoS2 with Ag LNP and SNP as shown in Figure 3c,d, respectively. The EM field enhancement factors were extracted from Raman scattering enhancement, which scales as where β is a constant,1, 2, 38 and plotted along with the PL EF. The obtained EM field enhancement factors are found to be relatively small (<5) and only weakly dependent on excitation energy for both LNP and SNP hybrids in agreement with simulation results (see the Supporting Information section for simulation results, Figure S7). This was consistent with the simulated enhancement factors of the metal NP substrates. PL EF was consistently greater than EM field EF at all excitations for both hybrid systems. The observation is consistent with the contribution from quantum yield modification discussed above. We note that deposition of Ag NPs on monolayer WS2, which is significantly more luminescent than monolayer MoS2 in the bare form,32 consistently led to a reduction in the PL intensity as shown in Figure 4a. Minor enhancement (≈2) was achieved only by using a dielectric spacer layer such as hexagonal boron nitride as shown in the insert of Figure 4b. This is attributed to the quenching component of the Purcell effect.1, 2 We explain below that the striking contrast between the two materials originates from differences in their intrinsic quantum yield. Equation 4 implies that materials with originally low quantum yield are only weakly affected by quenching effects of the metal. Indeed, in the limit of , the quantum yield enhancement is linearly proportional to the radiative Purcell factor for the spontaneous emission only. In the other limit of quantum yield approaching unity, quenching is unavoidable as can be seen from Equation 3. Thus, plasmon enhancement of moderately luminescent materials requires careful tuning of the ratio between Frad and Fnonrad by controlling the spacing between the emitter and the metal.1, 2 In Figure 4a, the relative quantum yields for monolayer MoS2 and WS2 with and without Ag NPs are compared. Monolayer WS2, which is 100 times more luminescent than monolayer MoS2, is significantly more prone to quenching when placed in contact with metal NPs. In contrast, monolayer MoSe2 exhibits sharp exciton features (see Figure 5 and Figure S13 of the Supporting Information) and optical gap frequency in resonance with the LSP of LNP, which is ideal for strong coupling. We found that the MoSe2-LNP hybrid exhibits a large (∼25%) transparency window in the vicinity of the A exciton peak as shown in Figure 5 WSe2-LNP hybrid showed similar features as shown in Figure S14, Supporting Information. The extinction features with a large transparency dip are similar to those previously observed in molecules hybridized with plasmonic nanostructures, where the features were attributed to Rabi splitting and emergence of plexcitons.5, 9, 10 Strictly speaking, such hybridization modes (plexciton) appear when κ > ΓLSP,Γexciton is satisfied, in the strong coupling regime.1, 5, 10 While we do not observe a clear evidence of mode splitting, the clear transparency dips in the extinction spectrum reveal effective plexcitonic coupling in this system. In summary, we showed that exciton–plasmon coupling can be tuned in monolayer MX2-Ag NP hybrid systems. Significant PL enhancement was achieved in monolayer MoS2 by tuning LSPR to excitation and exciton energy. The low quantum yield of monolayer MoS2 allows enhancement effects to overcome quenching even when the metal particles are in direct contact with the emitter. The PL enhancement in monolayer WS2, which tends to exhibit high PL quantum yield, was only achieved when using a dielectric spacer layer to minimize quenching by the metal. Furthermore, exciton-induced transparency dips were observed indicating the coherent dipole–dipole coupling between excitons of monolayer MX2 and LSP of Ag NPs. Our results highlight that integration of 2D semiconductors with plasmonic nanostructures is a viable approach to engineering light–matter interaction in these materials and to enable their implementations to photonics. Tunable coupling between excitons and LSP from weak to strong coupling regimes is of fundamental interest and opens up prospects for novel technological applications. G.E. acknowledges Singapore National Research Foundation for funding the research under NRF Research Fellowship (NRF-NRFF2011-02) and medium-sized center program. G.E. and K.O. acknowledge support from JSPS under its JSPS-NUS joint research project scheme (Award No. R-143-000-611-133). K.O. acknowledges support from JSPS KAKENHI (26289109). K.M. was supported by a Grant-in-Aid for Scientific Research from the Japan Society for Promotion of Science (Grants Nos. 26107522, 15K13500, 25400324, 25246010). As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.