(Invited) Enabling High Quality Dielectric Passivation on Monolayer WS2 Using a Sacrificial Graphene Oxide Template

Abstract

Two-dimensional transition metal dichalcogenides (2D TMDC’s) hold a wide variety of applications, among which microelectronic devices. 2D TMDC’s are promising alternatives for today’s silicon-based technology but suffer from various integration challenges. Among others, direct dielectric growth on the 2D surface occurs poorly due to their self-passivated surfaces resulting in island-like growth initiated at defect sites. This work focuses on enabling uniform high-k dielectric nucleation and growth on 2D TMDC’s via a sacrificial, graphene oxide-based buffer layer. Essentially, the role of the sacrificial graphene buffer layer is twofold: 1) serve as a passivation layer, protecting the underlying 2D TMDC (i.e., WS2) during processing and 2) act as a nucleation layer, enabling uniform atomic layer deposition (ALD) high-k dielectric (i.e,. HfO2) growth. A graphene layer is transferred on monolayer WS2, after which polymeric transfer residues are cleaned via a combination of wet treatments and a dry hydrogen downstream plasma exposure. The cleaned graphene is functionalized via a dry UV/O3 oxidative treatment, which enriches its basal plane with carbon-oxygen single bond functionalities by sacrificing its π-network. This way, the graphene carbon lattice itself remains intact and retains its passivation properties. Moreover, a substrate dependency is observed where graphene transferred on WS2 requires longer UV/O3 exposure to ensure functionalization compared to a SiO2 substrate and can be explained by UV-light induced, ultrafast charge transfer between the graphene and WS2 monolayer. The carbon-oxygen groups formed on graphene’s basal plane act as nucleation sites during a subsequent HfO2 ALD process. A constant roughness is noted for the HfO2 film grown on the graphene oxide capped WS2, indicating uniform material deposition even in the early stages of the ALD process. This is in sharp contrast when HfO2 is directly grown on the WS2 monolayer, whose roughness stabilizes only after many ALD cycles (up to 256), demonstrating the poor, island-like growth mechanism. In addition, a similar hafnium areal density is measured for the graphene oxide capped WS2 compared bare SiO2, which was selected to represent a well-functionalized surface and thus confirming the efficiency of the graphene oxide-based nucleation layer. For devices incorporating the graphene oxide buffer layer concept, a crucial question is to what extent the graphene oxide layer gives rise to leakage currents by providing a conductive path between source and drain. Electrical evaluation of a GFET by means of I-V measurements reveal a significant shift in Dirac point towards positive gate voltage for transferred graphene, which is typically associated with p-type doping originating from the polymeric transfer residues still present on the surface. The Dirac point shifts towards 0-point voltage after final hydrogen downstream plasma cleaning, confirming the effectiveness of the established cleaning protocol. Moreover, Graphene’s conductivity is suppressed after UV/O3 to the nA range, as caused by the formation of carbon-oxygen functionalities and associated sacrifice of its π-network. The residual current may be further suppressed by extending the UV/O3 treatment, but nevertheless demonstrates that the graphene oxide buffer layer will not give rise to significant leakage currents.