Effects of Interactions and Defect Motion on Ramp Reversal Memory in Locally Phase Separated Materials

TL;DR

This work tackles ramp reversal memory (RRM) in metal–insulator transition materials by extending a defect-motion framework to include interactions between metallic and insulating domains via a Correlated Random Field Ising dynamics. The authors integrate defect diffusion–segregation with a 2D RFIM on a lattice, linking local defect density to $T_c^{(i)}$ and simulating a ramp-reversal protocol to reproduce both hysteresis and the RRM effect, with memory strength controlled by the nearest-neighbor interaction $J$ and subloop turnaround temperature $T_H$. Key findings show that (i) RRM is amplified by domain interactions and avalanche dynamics, (ii) the maximum RRM occurs near the inflection point of the warming branch, and (iii) higher $J$ shifts this optimum and broadens memory, aligning with experiments on VO$_2$ and suggesting broad applicability to other phase-separated oxides. The model provides a minimal, predictive framework for designing memory and neuromorphic devices in materials with patterned electronic phase coexistence and supports the generality of RRM beyond VO$_2$ to other correlated electron systems.

Abstract

The ramp-reversal memory (RRM) effect in metal-insulator transition metal oxides (TMOs), a non-volatile resistance change induced by repeated temperature cycling, has attracted considerable interest in neuromorphic computing and non-volatile memory devices. Our previously introduced defect motion model successfully explained RRM in vanadium dioxide (VO$_2$), capturing observed critical temperature shifts and memory accumulation throughout the sample. However, this approach lacked interactions between metallic and insulating domains, whereas the RRM only appears when TMOs are brought into the metal-insulator coexistence regime. Here, we extend our model by combining the Random Field Ising Model with defect diffusion-segregation, thereby enabling accurate hysteresis modeling while predicting the relationship between RRM and domain interactions. Our simulations demonstrate that maximum RRM occurs when the turnaround temperature approaches the warming branch inflection point, consistent with experimental observations on VO$_2$. Most significantly, we find that increasing nearest-neighbor interactions enhances the maximum memory effect, thus providing a clear mechanism for optimizing RRM performance. Since our model employs minimal assumptions, we predict that RRM should be a widespread phenomenon in materials exhibiting patterned phase coexistence of electronic domains. This work not only advances fundamental understanding of memory behavior in TMOs but also establishes a much-needed theoretical framework for optimizing device applications.

Figures (10)

• Figure 1: (a) Experimental $T_{c}$ map measured during the first warming process. (b) Theoretical $T_{c}$ map using correlated random field (CRF) with Gaussian blurring effect to facilitate comparison with the experimental resolution of the $T_c$ map in Panel (a).
• Figure 2: Schematic representation of a portion of the lattice used in the simulation. Arrows indicate the defect current driven by the gradient of chemical potential. Labels with Ising variable $+1$ and $-1$ represent metallic (yellow) and insulating (brown) regions, respectively. In the mixed phase, defects concentrate preferentially in metallic regions.
• Figure 3: (a) Temperature protocol used in our simulation.
• Figure 4: Simulated hysteresis loops of the model under different conditions. The black curve represents the initial major hysteresis loop (ML1), while the yellow curve corresponds to the final major loop (ML4) after six subloops. The progression from dark red to orange to yellow indicates the sequence of the loops. (a) Results with defect motion but without interactions ($J=0$). (b) Results with interactions ($J=0.5$) but without defect motion. Loops stack on top of each other since there is no RRM effect. (c) Results with both defect motion and interactions ($J=0.5$) included. Subloops have counter-clockwise rotations.
• Figure 5: Simulated curves of fractional area change versus temperature during the warming process. The difference of two heating curves $\Delta A(T) = A_{\text{ML3}}(T) - A_{\text{ML2}}(T)$ is plotted, where $A$ is the fraction of insulating area from Figure \ref{['fig:loops']}. There are six temperature reversal subloops between Major Loop 2 and Major Loop 3, as shown in Figure \ref{['fig:5-phases']}. (a) Without interactions ($J = 0$), resulting in smoother curves. (b) With interactions ($J = 0.8$), showing jaggedness due to avalanches, as well as enhanced memory due to interactions.