Large-scale quantum-emitter arrays in atomically thin semiconductors

Quantum light emitters have been observed in atomically thin layers of transition metal dichalcogenides. However, they are found at random locations within the host material and usually in low densities, hindering experiments aiming to investigate this new class of emitters. Here, we create deterministic arrays of hundreds of quantum emitters in tungsten diselenide and tungsten disulphide monolayers, emitting across a range of wavelengths in the visible spectrum (610–680 nm and 740–820 nm), with a greater spectral stability than their randomly occurring counterparts. This is achieved by depositing monolayers onto silica substrates nanopatterned with arrays of 150-nm-diameter pillars ranging from 60 to 190 nm in height. The nanopillars create localized deformations in the material resulting in the quantum confinement of excitons. Our method may enable the placement of emitters in photonic structures such as optical waveguides in a scalable way, where precise and accurate positioning is paramount.

Height profiles across the lines over nanopillars 1 (pink), 2 (blue) and 3 (green). The full-width half-maximum measured for the nanopillars with no flake on top (1 and 2) are ~250 nm, while that of site 3 is ~500 nm, larger by as much as a factor ×2 due to the tenting of the flake over the nanopillar. We correlate the dark field microscopy (DFM) images with AFM scans. As mentioned in the main text, non-pierced pillars appear as brighter spots in DFM due to a larger scattering area, compared to the dimmer pierced sites. We carried out measurements on a total of over 80 nanopillar sites for the different nanopillar heights, observing no clear dependence of emission wavelength on nanopillar height, except a trend towards a narrower distribution of emission wavelength with increasing height. c, Fine structure splitting values of the 1L-WSe2 QEs, which lie within the range 200-700 µeV. We measured one data point for the 60 nm nanopillars due to the QEs having linewidths in the range of 1 meV, generally greater than the fine structure splitting.

Supplementary Note 1. Raman and PL material characterisation
Room temperature Raman and PL measurements are performed as discussed in the main text.
Supplementary Figures 1a,b plot, respectively, the Raman and PL spectra of 1L-WS2, as preliminarily identified by optical contrast, after transfer on the nanopillars. The Raman peaks at ~358 and ~419 cm -1 correspond to the E' and A'1 modes, respectively 1 . The separation between the two peaks is thickness-dependent 2 , and increases with increasing number of layers 2 . Our value of ~61 cm -1 is consistent with one-layer 2 . To further confirm this, we analyze its PL spectrum (Supplementary Figure 1b). A single peak at ~615 nm, corresponds to the neutral unbound exciton at the direct optical transition, a signature of 1L-WS2 3 . We label this exciton X 0 , following the notation used for TMDs by Ref. S4. We note that in literature one can find an alternative notation (e.g. in Ref. S3), where the letter A is used to distinguish it from a higher energy direct optical transition at ~520 nm (called "B") due to the spin-split valence band top. Supplementary Figures 1c,d (red lines) plot the Raman spectrum of 1L-WSe2, as initially identified by optical contrast, after transfer on the nanopillars. For comparison we also measure in Supplementary Figures 1c,d (blue lines) the spectrum of a 2L-WSe2 flake on Si+285nm SiO2, as identified by optical contrast. Supplementary Figure 1c indicates that in the low frequency Raman region two additional peaks appear at ~17 and ~26 cm -1 in 2L-WSe2.
The first peak, called C, is a shear mode caused by the relative motion of the layers, while the second peak is due to layer breathing modes 5,6 and can only appear in multi-layers. In Supplementary Figure 1d, red line, the peak at ~251 cm -1 , with full-width at half maximum (FWHM) ~2 cm -1 , is assigned to the convoluted A'1+E' modes 1,2 , degenerate in 1L-WSe2 1,2 , while the peak at ~262 cm -1 belongs to the 2LA(M) mode. Due to the A'1 and E' degeneracy, we do not use the separation between peak positions as fingerprint of the number of layers. In shows two components, at ~760 nm (orange line) and ~800 nm (purple line). The first corresponds to the direct optical transition, A 3 , of 2L-WSe2 , while the second is due to its indirect optical transition, I 3 .

Supplementary Note 2. QE creation in 1L-WSe2 using nanodiamonds
We deposit nanodiamonds, milled from bulk HPHT diamond (NaBond), of average diameter ~100 nm onto SiO2/Si substrates. We do this via a standard drop-casting technique, whereby we suspend the nanodiamonds in ethanol and deposit a drop onto the substrate using a pipette.
The drop is left on the substrate for 1 minute and then washed with de-ionised water, leaving behind only those nanodiamonds stuck to the surface of the substrate. We then place 1L-WSe2 flakes on them using the same viscoelastic technique as reported in the main text. These create similar protrusions or deformations in the flake as the nanopillars, but of varying sizes owing to the size and shape dispersion of the nanodiamonds. Supplementary Figure 8a is an AFM scan of a 1L-WSe2 flake on nanodiamonds. We take a height profile (shown in Supplementary Supplementary Figure 8c shows an integrated PL raster scan of the same sample taken at 10 K.
The flake is highlighted by the white lines. There is an increase in PL intensity at the nanodiamonds site, similar to the effect seen with the nanopillars. Supplementary Figure 8d shows a spectrum taken at this location, at 10 K and under 532 nm laser excitation, showing a sub-nm peak. About 20 such nanodiamonds-induced QEs were measured.