Monitoring Morphological Changes in 2D Monolayer Semiconductors Using Atom-Thick Plasmonic Nanocavities

Nanometer-sized gaps between plasmonically coupled adjacent metal nanoparticles enclose extremely localized optical fields, which are strongly enhanced. This enables the dynamic investigation of nanoscopic amounts of material in the gap using optical interrogation. Here we use impinging light to directly tune the optical resonances inside the plasmonic nanocavity formed between single gold nanoparticles and a gold surface, filled with only yoctograms of semiconductor. The gold faces are separated by either monolayers of molybdenum disulfide (MoS2) or two-unit-cell thick cadmium selenide (CdSe) nanoplatelets. This extreme confinement produces modes with 100-fold compressed wavelength, which are exquisitely sensitive to morphology. Infrared scattering spectroscopy reveals how such nanoparticle-on-mirror modes directly trace atomic-scale changes in real time. Instabilities observed in the facets are crucial for applications such as heat-assisted magnetic recording that demand long-lifetime nanoscale plasmonic structures, but the spectral sensitivity also allows directly tracking photochemical reactions in these 2-dimensional solids.


A) AFM of CdSe and MoS 2 films on Au planar surfaces
In the samples prepared on Au, the MoS 2 monolayer crystals cover ~50% of the gold surface. Similarly, the CdSe platelets form a monolayer with a surface coverage of ~60% ( Fig. S1).

B) SERS of MoS 2 and CdSe in the NPoM geometry
In order to ensure that the nanoparticles are placed directly on top of the semiconductor layers, we compare the Raman scattering of the CdSe and MoS 2 spacer without nanoparticles and in the NPoM geometry. We employ excitation wavelengths of 532nm for the CdSe and 633nm for the MoS 2 and 1mW optical power. Both wavelengths lead to significant plasmonic activity in the plasmonic MIM-waveguide nanocavity and enable excitation near the bandgap of the respective material, leading to resonant Raman effects. The strongly enhanced SERS observed clearly demonstrates that the AuNPs are on top of the semiconductor layers and that they do not introduce significant strain or doping. Both spacer materials exhibit clearly enhanced Raman scattering when placed into the NPoM geometry compared to spectra taken on a flat Au surface covered by the semiconductor layers (Fig. S2). In the case of the MoS 2 spacer, the Raman scattering amplitude is amplified by a factor of ~10. We note that this overall enhancement actually stems from only the tiny field-enhanced volume in the gap whose lateral diameter is given by = √ (R nanosphere radius, d gap distance). 1 Taking into account that the collection area diameter is ~1μm, we estimate a SERS-enhancement in the junction of >10 5 .
In the case of the CdSe spacer, the observed enhancement is approximately a factor of 10 lower due to the less intense plasmon field at 532nm. While absence of NPs only allows observation of the (LO) 1 phonon near 200cm -1 , the nanogap environment additionally reveals the first overtone 2(LO) 1 near 400cm -1 .
The photoluminescence of MoS 2 is virtually unaffected by UV laser irradiation (Fig. S2c). In order to fully confirm the plasmonic nature of the dipolar infrared mode, we investigated the polarisation of its far-field radiation (Fig. S3a). A λ/2-plate was placed in the scattered beam while the incident supercontinuum light was unpolarised. The intensity of the scattered light shows a clear dependence on the analyser position as expected. The spectral position of the dipolar mode is highly sensitive to the initial facet size of the NP but in all NPoMs is located between 1100nm and 1200nm with good reproducibility (Fig. S3b).

D) DF scattering of NPoM without semiconductor spacer
Omitting the semiconductor spacer leads to conducting contact of the AuNP with the surface. A charge oscillation across nanoparticle and image charge is present, leading to only a single coupled mode (as seen in the typical spectra of Fig. S4 observed on many NPoMs).
As the charge separation between the AuNP and the image is suppressed, no mode splitting can be observed in contrast to the semiconductor-spaced NPoM.

E) Photoluminescence decay and phonon shift
We investigate the kinetics of the photoluminescence (PL) decay and the phonon shifts on CdSe platelets in the NPoM geometry using a Renishaw inVia Raman microscope with multiple excitation sources.
Using a 488nm excitation laser, we record the photoluminescence amplitude over time.
Although the PL spectral peak does not shift in energy, it decays following a double exponential function and always stabilizes at 15%-30% of the initial amplitude, depending on the irradiation power ( Fig. S5a,b). Since power levels even higher than those applied in the DF-irradiation experiments do not lead to a complete PL drop, we exclude the possibility of complete CdSe ablation. The optical phonons in the CdSe platelets are further investigated using 532nm excitation.
The Raman scattering peak intensity of the CdSe (LO) 1 phonon decays along with the photoluminescence, but also a significant fraction remains visible even after long irradiation times. We note a spectral shift from 201cm -1 to 208cm -1 after irradiating with the blue laser for 30 seconds (Fig. S4c). Such phonon shifts are known in core-shell type quantum dots 2 and are generally attributed to increased strain in the crystal as a result of a lattice mismatch between core and shell.

F) Photochemistry
The photochemistry of CdSe is well known in a variety of configurations, and we summarise how this relates to the current observations. We suggest a photoinduced transformation of the CdSe surface monolayer according to the scheme 3 CdSe + 2h +  Cd 2+ + Se.
As a result of the light-induced separation of electron-hole pairs at the interfaces, the CdSe is converted into atomic Se and Cd 2+ ions. The Cd ions further react with environmental water to Cd(OH) 2 and thence further to CdO. Evidence for this is the observed phonon shift described in section (E). Further we note the occurrence of a CdO Raman peak near 270cm -1 .
Once the top layer of the platelet has reacted, it acts as a protective layer preventing further oxidation.  This yields an effective refractive index which is very sensitive to the gap size, and the dielectric filling (Fig. S8). for such narrow gaps, and hence this is a reasonable assumption. By using this boundary condition on a circular rim, we find the mode quantisation as given in the main text. We solve simply using Equation (S1) to extract eff and use this to find the precise wavelengths. This allows the ratio of the tuning rates of the different lines to be extracted, so ̇3 /5 = 5 5 / 3 3 ∼1.4, which matches the experimental observations. We can also use this formula to estimate the number of atoms arriving at the edge of the facet, = / 2 , giving the formula for ̇ in the main text.

H) Effect of electron microscopy on the optical NPoM scattering
The optical scattering of NPoMs is investigated before and after imaging in a scanning electron microscope (SEM) (Fig. S9). Figure S9 | a, Scattering spectra of AuNP before (black) and after (purple) SEM examination. b, Control measurement: scattering spectra of a NP on the same sample that has not been irradiated with the electron beam before (black) and after (purple) the sample has been in the SEM.
The electron beam exposure completely changes the optical scattering spectrum (Fig. S9a) while scattering from a particle on the same sample that has not been directly irradiated with the electron beam remains unchanged (Fig. S9b). As always seen in electron microscopy, carbon contamination is deposited in the scanning area of the beam 6,7 , affecting the scattering spectrum but also the Au adatom mobility. Interrogation with an electron beam hence introduces irreversible changes into the system which prevents direct structural detection of such laser-induced morphological changes. 8

I) Thermal activation of plasmon tuning
As well as using optical irradiation, we find thermally-activated changes in the NPoM scattering when heating the samples and simultaneously recording the scattering spectra (Fig.   S10). In these experiments the AuNP and Au surface are instead spaced apart by a thin insulating ligand layer. Heating of the substrates is performed on a heating stage mounted on the microscope in 10°C steps with a rate of 10°C/ min and after each heating step the temperature is kept constant for 10min. Once heated above 60°C, redshifts of the coupled plasmonic mode are observed similar as discussed in the main paper, while the -mode is little affected. We conclude that thermal energy also leads to increased atomic mobility in the AuNPs and an increasing facet size. Heating the sample above 100°C leads to sudden blueshifts of the modes from the IR as the AuNP and surface fuse together forming charge transfer plasmons ( , ′ ) ( Fig.   S10b) with increasing conductivity. [9][10][11][12] The ′ mode shifts rapidly from the IR towards the blue at ~6 nm / s and merges into the -mode. A slower shift is observed for the mode which reaches an equilibrium state at a wavelength of ~750nm. This proves that Au atom migration can occur, leading in this case to conductive contact with the substrate, which is not possible with the impermeable inorganic semiconductor platelets.

J) Cross-sectional NPoM imaging with DF-STEM
To investigate the contact area of the facetted AuNP with the surface underneath, we use dark-field scanning transmission electron microscopy (DF-STEM) for cross-sectional imaging of the NPoM construct which is sectioned by focussed ion beam milling at the position of a single nanoparticle. The sample fabrication process is illustrated in Fig. S11. The slice is mounted on a TEM grid by evaporating a thin Pt layer.
First, a NPoM is embedded in a platinum (Pt) matrix which is evaporated onto the sample surface. Focussed ion beam milling (FIB) is used to isolate a cross-sectional slice from the sample, which is then mounted on a TEM grid with a nanomanipulator via local deposition of a Pt layer at the touching points. The slice is further thinned to a thickness of ~100nm using FIB and finally imaged via DF-STEM in a Hitachi S-5500 at 30kV acceleration voltage. 13 While the facet and the formation of the MIM-cavity with the surface is clearly seen (Fig. 1 in the manuscript), the atom migration in the gap cannot be resolved. The imaging technique is highly invasive due to the deposition of the Pt matrix as well as carbon contamination, preventing the same nanoparticle being imaged before and after optical irradiation.