Insights into MOCVD of TMDC Thin Films by In Situ Spectral Reflectance Measurements

Metal-organic chemical vapor deposition (MOCVD) has developed into a mainstream technique for 2D transition metal dichalcogenide (TMDC) thin films and heterostructures. It enables large-area and multi-wafer scale growth with excellent controllability and reproducibility. Nevertheless, the development of a dedicated in-situ monitoring technique to study film formation and support large-scale fabrication of 2D TMDC is still missing. There are a few reports on different in-situ characterization methods during the TMDC (MO)CVD, which mainly focus on confirming the formation of monolayer (ML) domains. One recent work is employing in-situ spectroscopic ellipsometry to study the coverage and thickness of MoS₂ and WS₂ thin films (up to trilayers), which is a pioneering work correlating optical properties of TMDC thin films with the thickness deposited [1]. However, ellipsometry has some inherent disadvantages. For instance, its integration in commercial MOCVD reactors is difficult, and real-time data processing is relatively complex.
In comparison, in-situ reflectance spectroscopy is a mature technique widely used in the MOCVD of III-V semiconductors. In this work, a commercial AIXTRON CCS multi-wafer reactor equipped with LayTec in-situ monitoring systems (3 single wavelengths + one broadband white light source and wavelength selective detection) was employed to deposit and monitor TMDC thin-film (MoS₂, WS₂, and WSe₂) growth with varied thickness (from ML to multilayer) on 2” sapphire and SiO₂/Si wafers. Spectroscopic reflectance data were collected (through the same viewport as used for normal incidence) and analyzed by a CCD detector (from 275 to 810 nm, every 10 s for each wafer individually).
Firstly, single-step MOCVD processes were designed to deposit WSe₂ ML, bilayer (BL), and trilayer (TL) films. Using the transient reflectance spectra recorded, dynamic differential reflectance spectra can be obtained. Upon inspecting the features (i.e., plateaus and valleys) in a differential reflectance vs. wavelength (x-axis) and time (y-axis) colormap, nucleation, lateral growth, and coalescence of WSe₂ ML can be identified. Furthermore, varying material deposition rates in these stages can be detected and correlated to microscopic film formation processes. Similar observations can be made for MoS₂ and WS₂ ML. Following the guidance of differential reflectance, a precise termination of ML growth can be reproducibly realized with less than 10% bilayer coverage even for our simple single-step processes. The drastic change of differential reflectance is probably caused by the direct-to-indirect band gap transition as growth proceeds from ML to BL. The coalescence of BL and TL however cannot be easily identified, which can be explained by the similar optical properties of BL and TL TMDC.
Furthermore, reflectance spectra of WSe₂ thin films at different surface temperatures were recorded during cool-down. The positions of two fitted reflectance peaks are both thickness- and temperature-dependent and can be used to estimate the deposited thickness of WSe₂ at any sample temperature.
Finally, both WS₂ ML and multilayer (> 5 layers) samples were deposited on SiO₂(100 nm)/Si substrates. Due to their high reflectance, the differential reflectance transitions from negative to positive values as the ML coalesces. The thickness of multilayers can still be deduced from the in-situ spectral reflectance measurements.
In conclusion, white-light in-situ reflectance spectroscopy can determine the coalescence of TMDC ML and estimate the thickness of TMDC thin films at varied growth temperatures. It can be further engineered into a real-time data acquisition and processing technique to support the efficient development and optimization of MOCVD processes for large-scale fabrication.

Reference:
[1] E. Houser, T. V. Mc Knight, J.M. Redwing, and F.C. Peiris, J. Cryst. Growth 640, 127741 (2024).