MOCVD of Monolayer WS₂ on Si/SiO₂ Substrates

2D transition metal dichalcogenides (TMDC) have triggered huge research interest in the field of (opto-)electronics since they possess unique characteristics such as appreciable carrier mobilities and direct bandgaps as monolayers (ML). For their wafer-scale synthesis, metal-organic chemical vapor deposition (MOCVD) stands out due to its superior scalability, reproducibility and uniformity. A widely employed growth substrate is sapphire considering its commercial availability as well as chemical and thermal stability. However, for device fabrication, the deposited TMDC layers or structures on sapphire commonly need to be transferred onto dielectrics, silicon or other conductive substrates. This transfer process usually has limited scalability, introduces contaminations and causes degradation of materials properties.
This work aims to realize a direct MOCVD process of TMDC (here WS₂) on substrates suited for transfer-less device processing. The growth of WS₂ is conducted in a commercial AIXTRON CCS multi-wafer MOCVD reactor with advanced ARGUS and LayTec in-situ monitoring systems. Tungsten hexacarbonyl (WCO) and ditert-butyl sulfide (DTBS) are used as precursors. N₂ is chosen as carrier gas. Deposition of WS₂ is firstly attempted on conductive Si substrates. However, the Si surface is attacked by DTBS with the formation of SiS₂ as product. In-situ passivation by forming a protective layer of SiO₂ on Si using water vapor proves to be unsuccessful due to limited oxide quality achievable with the parameters accessible in the MOCVD tool. Therefore, 2" Si/SiO₂ wafers (oxide thickness of 100 nm through ex-situ thermal oxidation) are chosen. Results are compared with those on 2" sapphire under the same conditions.
To study the impact of temperature and determine the optimal growth parameters, a series of 15 min nucleation tests (from 650 °C to 750 °C) on both sapphire and Si/SiO₂ are conducted. It is found that higher nucleation temperatures result in smaller nucleation densities and larger domain sizes. On the other hand, as more chalcogen adatoms tend to desorb from the surface or gas-phase pre-reactions are more likely to happen, a slower deposition rate and therefore less coverage are observed. Nuclei on both sapphire and Si/SiO₂ feature a higher photoluminescence (PL) intensity and a red-shift with increased nucleation temperature, which could result from the varied domain size and higher density of chalcogen vacancies. Compared to those on sapphire, samples grown on Si/SiO₂ have generally a higher intensity and exhibit a blue-shift.
For the coalescence of ML WS₂ on both types of substrates, a three-stage growth recipe is developed. WS₂ on Si/SiO₂ has ML (around 96%) and BL (around 10%) coverages comparable to those of WS₂ on sapphire under the same growth conditions. Similar to the nucleation-only samples, the ML WS₂ on Si/SiO₂ features a significantly higher intensity in both Raman and PL spectra, compared to samples on sapphire. In PL spectra, the blue-shift can be observed as well. In order to clarify if the signal differences result from either a superior WS₂ quality or enhanced substrate reflection, the effect of substrates has to be excluded. Therefore, both as-grown layers are each transferred onto the same and the other substrates type. The PL intensity difference observed in the as-grown samples disappears in the transferred samples. This confirms that the considerably higher PL intensity of as-grown WS₂ on Si/SiO₂ originates in the enhanced substrate reflection. Furthermore, by comparing samples before and after transfer, it is found that both transfer processes result in decreased PL intensity, no matter if transferring onto sapphire or Si/SiO₂. Interestingly, WS₂ grown on sapphire experiences a blue-shift after transfer probably due to released strain, while for WS₂ on Si/SiO₂ a red-shift can be observed.