Enhanced and directional light emission from two-dimensional excitons using Mie voids

Abstract

Controlling light emission at the nanoscale has important applications in solid-state lighting, displays, and quantum light sources. Achieving this control requires both enhanced local electromagnetic fields to boost emission intensity and engineered radiation patterns to direct photons efficiently. Mie voids, consisting of an air cavity surrounded by a high-index semiconductor, are particularly suited for this purpose because they expose their strongest fields in an accessible region for nearby emitters while supporting resonances that shape directional emission through interference. Here, we demonstrate an all-van der Waals nanophotonic platform that couples excitons in atomically thin WS$_2$ to Mie void resonators formed in WSe$_2$. Guided by electromagnetic simulations, we identify void geometries that maximize photoluminescence through synergistic enhancement of excitation and emission processes. We also develop a two-step fabrication strategy that enables independent control of void diameter and depth, providing a route to systematically tune the optical response. Experimentally, we observe up to a 600-fold increase in photoluminescence intensity from monolayer WS$_2$ placed on individual voids compared to on an unstructured WSe$_2$, along with pronounced out-of-plane beaming of light that yields a forward-to-off-axis enhancement of 2.6 dB. Our results establish Mie voids in van der Waals semiconductors as a new platform for controlling light-matter interactions and realizing compact, directional, and efficient nanoscale light sources.

Figures (5)

• Figure 1: Controlling light emission using void resonators.a, Schematic of the device displaying an hBN encapsulated monolayer of WS$_2$ (zoomed) on top of a void resonator constructed in WSe$_2$. b, Field localization within the air region enabled by the void resonance. c, Emitter-void interaction that results in enhanced and out-of-plane directional emission.
• Figure 2: Simulation of enhanced and directional light emission.a, Simulated electric field distribution $E_x$ in the $yz$-plane for a void with diameter $d=540$ nm and depth $h=160$ nm. The excitation is a normally-incident plane wave polarized along $x$ with a wavelength of $\lambda_\mathrm{exc}=532$ nm. The black border separates the air and WSe$_2$ domains. The white dashed line marks the top surface of the void. The black dot denotes the center of the void opening, where the electric field intensity is evaluated to obtain the excitation efficiency. b, Simulated electric field distribution $E_x$ from an $x$-oriented dipole (yellow dot) emitting at a wavelength of $\lambda_\mathrm{emi}=610$ nm and placed at the center of the void opening. The void dimensions are $d=500$ nm and $h=200$ nm. c, Photoluminescence enhancement as a function of void diameter $d$ and depth $h$ calculated using Eq. (\ref{['eq:PLenh']}). d–e, Excitation and emission efficiencies as a function of void dimensions, respectively. The black lines depict the depth vs. diameter relation given by Eq. (\ref{['eq:constinter']}). f, Simulated Poynting vector amplitude of dipole-on-void system normalized by dipole on flat WSe$_2$. The void dimensions are the same as in b.
• Figure 3: Solving fabrication barrier.a, Schematic of conventional lithographic fabrication of Mie voids in a WSe$_2$ flake (pink) with a thickness exceeding the targeted void depth. The etched void diameter can exceed the nominal diameter $d_\mathrm{nom}$ defined by the resist mask opening (blue) due to lateral etching. b, Atomic-force microscopy measurements of voids fabricated using identical etch times show that larger mask openings $d_\mathrm{nom}$ produces voids with both larger diameters $d$ and larger depths $h$. The surface is at $h=0$ nm. c, Measured void depth versus diameter for the conventional process, demonstrating a near-linear dependence on the void dimensions due to the mask-opening-dependent etch rate. d,e, Mie voids fabricated using a two-step process to independently control the diameter and depth. A WSe$_2$ flake with a thickness matching the target depth ($\sim200$ nm) is lithographically patterned (d). The patterned flake is subsequently transferred onto an optically-thick WSe$_2$ that serves as a substrate to provide the bottom interface of the Mie void (e). f, Measured void depth versus diameter for the two-step fabrication process. The void diameter is controlled by the mask opening, while the depth remains constant at $h=212$ nm.
• Figure 4: Void-enabled photoluminescence enhancement.a, Bright-field micrograph of the WS$_2$ monolayer on voids with different diameters. The red border indicates the monolayer. b, Corresponding optical image of PL emission showing much brighter emission from the voids relative to flat WSe$_2$. The scale bars for both the images are 10 µ m. c, Spectrally resolved PL signal with and without the void resonator. d, Experimental PL enhancement of monolayer-on-void structure as a function of the position from the void center. Top-right inset shows two-dimensional PL mapping of the same structure (scale bar: 500 nm). Top-left inset shows experimental peak PL enhancement of voids with varying diameters. e, Simulated PL enhancement as a function of the dipole position. The inset shows the normalized electric field profile at the void opening at an excitation wavelength of 532 nm. The black line indicates the axis along which the PL enhancement is calculated.
• Figure 5: Void-enabled directional emission.a, Optical image of a WS$_2$ monolayer on a void array. Scale bar: 5 µ m. b, PL map of the same region, showing brighter emission from monolayer-on-void regions. Scale bar: 5 µ m. c, Maximum PL count of individual voids covered by monolayer WS$_2$. The black dashed line indicates the sample mean ($\mu$) of the maximum PL intensity, and the shaded band shows the standard deviation ($\pm \sigma$) around the mean. d, Back-focal-plane image of a monolayer-on-array configuration. e, Back-focal-plane image of a monolayer-on-flat WSe$_2$ configuration. f, Azimuthally-averaged 1D angular PL plots (normalized) calculated from the back-focal-plane images in d,e.