Effects of High-k Dielectric Encapsulation and Carrier Density on Raman Scattering in Synthetic Monolayer WS₂

Two-dimensional (2D) semiconductors have gained significant interest due to their atomically thin structure, theoretically pristine van der Waals interfaces, good carrier mobility, and potential utility for future opto-electronics. Among 2D semiconductors, tungsten disulfide, WS₂ is particularly interesting, as it can exhibit electronic ambipolarity at monolayer and bilayer thicknesses [1].

Raman spectroscopy is a fast, non-destructive characterization technique that enables rapid, large-area analysis of 2D semiconductors. Previous work has utilized Raman spectroscopy to quantify the strain and carrier density in monolayer MoS₂ [2], as well as evaluate the damage induced in MoS₂ from metal deposition [3]. However, these characterization techniques have not been experimentally examined in WS₂, which features different Raman signatures from MoS₂.

Here, we investigate the effects of dielectric environment and carrier density on the Raman spectrum of WS₂. We start with three samples of monolayer WS₂ grown by chemical vapor deposition, one on SiO₂/Si and the others on sapphire. We then transfer the two sapphire samples onto 125-µm thick polyethylene naphthalate (PEN) containing a patterned high-k/metal-gate stack as described in [4]. We encapsulate the as-grown SiO₂/Si sample and one PEN sample with a 1.5 nm Al seed layer + 10 nm Al₂O₃ [4], then the final PEN sample with 1 nm Si seed layer + 10 nm Al₂O₃ [5].

We find that, of the primary peaks in the WS₂ Raman spectrum, the 2LA(M) peak at ~350 cm^-1 is the most sensitive to encapsulation. The as-grown, unencapsulated WS₂ on SiO₂ sample exhibited a 2LA(M) peak position ~3.2 ± 0.3 cm^-1 higher than the transferred, unencapsulated WS₂ on Al₂O₃ back-gate sample. After encapsulation with the Al+Al₂O₃ layer, the 2LA(M) peak in the SiO₂ sample redshifts by an additional ~1.0 ± 0.4 cm^-1 while that of the Al₂O₃ sample redshifts by ~4.5 ± 0.5 cm^-1. We also find a 2× stronger red-shift for the 2LA(M) peak in the Si seed sample (~8.1 ± 0.8 cm^-1) than in the Al seed encapsulation sample. This indicates that the 2LA(M) peak may be useful as an optical marker for interfacial quality. In comparison, the first-order E’ and A’ peaks shift by less than 1.5 ± 0.5 cm^-1 across all samples.

We also perform in situ Raman measurements as a function of carrier density in the same back-gated WS₂ structure. We find that the 2LA(M) peak red-shifts by 0.44 ± 0.06 cm^-1/V and the A’ peak red-shifts by 0.24 ± 0.03 cm^-1/V, while the E’ peak does not shift substantially. With a ~350 nF/cm² Al₂O₃ gate oxide capacitance, these shift rates correlate to ~0.19 ± 0.01 cm^-1 per 10¹² cm^-2 carriers for the 2LA(M) peak and ~0.1 ± 0.05 cm^-1 per 10¹² cm^-2 carriers for the A’ peak. This study is the first measurement of the carrier density-dependent Raman spectra of WS₂ without ionic liquid gating, which provides a better benchmark relative to the intrinsic carrier density of WS₂ compared with previous work [6].

Our results advance Raman spectroscopy for 2D materials in two ways: first, they extend previous studies on spectroscopic carrier density measurements to near-intrinsic carrier densities, and second, they showcase the utility of Raman spectroscopy for characterizing interactions between 2D materials and high-k dielectrics for industry-relevant integration. This work was supported in part by a NSF Graduate Fellowship (J.A.Y.), by the Stanford SystemX Alliance, and by the SRC-SUPREME Center.

[1] G. Lee et al., ACS Appl. Mater. Interfaces, 12, 23127 (2020).
[2] A. Michail et al., 2D Mater., 8, 015023 (2020).
[3] K. Schauble, E. Pop et al., ACS Nano, 14, 14798 (2020).
[4] J. A. Yang, E. Pop et al., arXiv:2309.10939 (2023).
[5] H. Zhang et al., Chem. Mater., 29, 6772 (2017).
[6] T. Sohier et al., Phys. Rev. X, 9, 031019 (2019).