The Effect of Preparation Conditions on Raman and Photoluminescence of Monolayer WS2

We report on preparation dependent properties observed in monolayer WS2 samples synthesized via chemical vapor deposition (CVD) on a variety of common substrates (Si/SiO2, sapphire, fused silica) as well as samples that were transferred from the growth substrate onto a new substrate. The as-grown CVD materials (as-WS2) exhibit distinctly different optical properties than transferred WS2 (x-WS2). In the case of CVD growth on Si/SiO2, following transfer to fresh Si/SiO2 there is a ~50 meV shift of the ground state exciton to higher emission energy in both photoluminescence emission and optical reflection. This shift is indicative of a reduction in tensile strain by ~0.25%. Additionally, the excitonic state in x-WS2 is easily modulated between neutral and charged exciton by exposure to moderate laser power, while such optical control is absent in as-WS2 for all growth substrates investigated. Finally, we observe dramatically different laser power-dependent behavior for as-grown and transferred WS2. These results demonstrate a strong sensitivity to sample preparation that is important for both a fundamental understanding of these novel materials as well as reliable reproduction of device properties.

The photoluminescence was investigated for laser excitation in addition to 532nm. The PL intensity map acquired using laser excitation (λ exc ) of 488 nm exhibits clear intensity variations (Fig. S1), with lowest intensity extending from center outward to the three corners. This pattern is analogous to that obtained using 532 nm excitation (presented in the main text Fig. 4(j)) and shows the intensity variations are independent of laser excitation wavelength.   , no paQern is present in peak posi7on of E 1 2g or A 1g . As evident by op7cal images taken (e) before scanning and (f) a5er scanning, the sample becomes slightly out of focus during the long map. This is likely the cause of the slight intensity reduc7on present in the boQom third of Fig. S2  In conjunction with the photoluminescence characterization, Raman maps were acquired for λ exc = 488nm. In contrast to the clear spatial variations observed in PL intensity (for both 488nm and 532nm excitation), the dominant in-plane and out-of-plane Raman peaks display no discernible pattern ( Fig. S2 (a,b)). We do observe a slight decrease in overall intensity near the bottom third of the Raman maps, most likely caused by modifications to zposition of the sample. The optical images acquired before (Fig. S2(d)) and after (Fig. S2(e)) performing the Raman map show the sample has drifted away from the focal point. The observed decrease is consistent with a gradual defocussing during the course of the scan, as maps proceed from top left to bottom right. E 1 2g and A 1g peak positions ( Fig. S2 (c,d)) are steady across the sample. The uniformity observed in Raman peak positions and intensities suggest structural defects (as opposed to local variations in strain or electronic doping) are the source of the observed variations in PL.
Care is taken to ensure all acquisition conditions are below the damage threshold, particularly for the power-dependent investigations (presented in Fig. 5 and Fig. 6 of the main text), as high-power laser exposure is capable of damaging monolayer TMDs. Spectra in the main text are presented as the laser power is swept from low (6nW) to high (140µW). After which, the power is returned to 6nW and a final spectrum is acquired. A direct comparison of spectra obtained at low power and high power are presented in Fig. S3. Additionally, spectra  Figure S3: PL spectra acquired at low and high laser power. Spectra are normalized to the X 0 intensity. (a) As-grown WS 2 exhibits only minor differences between 6nW and 140µW excitaDon power. A small red-shiE (~4meV) and increased FWHM is observed at higher power, most likely from sample heaDng. Both (b) PMMA and (c) PDMS transferred WS 2 are highly sensiDve to laser power. Emission from the neutral exciton, X 0 , dominates at low power, but transiDons to T dominated emission with increasing laser power.
Asgrown PMMA PDMS a"er before a"er before a"er before Figure S4: Comparison of PL spectra before and a>er power sweep. Photoluminescence is measured using 6nW laser excita8on before and a"er exposure to 140µW laser. Nearly iden8cal spectral shape and emission energy are obtained for (a) as-grown, (b) PMMA, and (c) PDMS samples, indica8ng the WS 2 samples are not damaged by laser powers u8lized in this work.
obtained for 6nW excitation are presented before and after the power sweep is completed (Fig. S4) and exhibit nearly identical spectral shape and emission position, indicating the samples are unchanged by the 140µW laser exposure. PL and Raman spectra are measured at the same location for as-grown WS 2 on fused silica, Si/SiO 2 , and c-sapphire substrates and presented in Fig. S5. As discussed in the main text, PL emission energy is sensitive to the strain in the as-grown WS 2 , with increased strain resulting in a decreased band-gap and a red-shift in PL. The distinctly different emission energies indicate the largest amount of strain is present for WS 2 grown on fused silica, whereas the smallest strain is present in c-sapphire (Fig. S5 (a)). The variation in Raman E 1 2g peak further supports the connection between strain and growth substrate, as the position of E 1 2g is known to red-shift with increasing strain (Fig S5 (b)).
In addition to the AFM image acquired on as-grown WS 2 (Fig. 1b of the main text), AFM data is obtained on PDMS and PMMA transferred WS 2 . Figure S6 presents the AFM acquired from three representative PDMS x-WS 2 samples. In all three images, small particles (white spots in Fig. S6 a-c) are present on both the WS 2 and SiO 2 substrate. Such particles are not present in as-grown samples, and are most likely residues from the PC stamp and/or processing chemicals. Imperfections such as small tears (Fig. S6a) and microscopic wrinkles (Fig. S6c) are observed in some regions. The WS 2 step height for each sample is measured along the black dashed line and displayed in the inset. All three samples exhibit a step height of ~1nm, which is slightly larger than the 0.8nm measured for as-grown WS 2 . The AFM acquired from several PMMA x-WS 2 samples (Fig S7 a-c) show features that are qualitatively similar to those of PDMS x-WS 2 . Again, surface particles and imperfections are evident. Line cuts along the dashed line are displayed in the inset for each sample. Of note is the relatively large step height for PMMA x-WS 2 , with measured values ranging from 1.7nm to 1.9nm for monolayer WS 2 . Several factors may contribute to the step height value, and include effects such as increased sample-substrate distance, water layers trapped between the monolayer sample and SiO 2 substrate, 1   AFM for these types of samples, the modification of the Raman and PL spectra compared to the as-grown samples is the same as for the PDMS x-WS 2 . Therefore, our conclusions remain unchanged, regardless of the source of the extra step height.
To assess the crystalline quality, monolayer WS 2 is imaged using high-resolution transmission electron microscopy. While the sample is predominantly monolayer, the HAADF image of a terraced region is purposefully displayed (Fig. S8 a,b) and establishes a clear intensity contrast between monolayer and multilayer WS 2 . Images acquired from a single layer region exhibit a uniform, defect-free, single-crystalline hexagonal atomic structure (Fig. S8c). The measured intensity depends on the atomic number (Z) of the imaged atom as Z 1.64 . 3 Therefore the bright spots correspond to tungsten atoms (Z=74) with darker contrast indicating the position of sulfur atoms (Z=16).
The chemical composition of as-grown and transferred WS 2 is analyzed using X-ray photoelectron spectroscopy. We investigate two different as-WS 2 , one PDMS x-WS 2 , and two separate PMMA x-WS 2 samples. All samples exhibit the same tungsten and sulfur core levels, demonstrating the chemical composition is the same for as-grown and transferred WS 2 .