Film morphology and substrate strain contributions to ramp reversal memory in VO2

The ramp reversal memory (RRM) effect that appears in thin films with temperature-driven insulator-metal transitions (IMTs) is a nonvolatile memory effect induced by a simple reversal of temperature ramping from heating to cooling during the phase-coexistence state of the IMT (when both metallic and insulating domains coexist). The memory of specific temperatures can be recorded by this ramp reversal, which appears as a resistance increase around the reversal temperatures. Previous studies showed RRM in $\mathrm{V}{\mathrm{O}}_{2}, {\mathrm{V}}_{2}{\mathrm{O}}_{3}$, and $\mathrm{NdNi}{\mathrm{O}}_{3}$, indicating it is a general effect in relevant systems. These studies indicate the RRM originates from an increase in the critical temperature around phase boundaries of the coexisting metallic and insulating domains during the temperature ramp reversal. However, the physical mechanism responsible for the ${T}_{\mathrm{C}}$ increase remains elusive. To enhance our understanding of the effect and provide clues to the underlying physics, it is crucial to understand the role of materials' properties, such as thickness, grain size, and choice of substrate, which have yet to be explored. We report the RRM properties in $\mathrm{V}{\mathrm{O}}_{2}$ thin films as a function of these parameters, namely choice of substrate, crystallographic properties, film thickness, and morphology. We find that films' grain size correlates with the RRM magnitude. The film thickness has a positive effect on the RRM, but at a much lesser extent. Interestingly, thinning films by a wet-etching process had almost no effect on RRM properties, indicating that the grain structure and interaction with substrate defined during deposition determines the RRM features. This was further corroborated by comparing of films grown epitaxially on sapphire with those grown nonepitaxially on $\mathrm{Si}{\mathrm{O}}_{2}/\mathrm{Si}$ substrates, where the latter show a fivefold increase in RRM magnitude. These findings support the hypothesis that strain (or strain gradients) and strain relaxation develop at phase-separated grain boundaries during the ramp reversal process, which controls the magnitude of the RRM effect.