Scalable Kinetic Monte Carlo Platform Models Ion Transport Dynamics in Polymer Memristive Systems

Ion transport within polymers forms the basis of many crucial technologies, from energy storage to next-generation electronic devices, yet accurately modelling this process remains a significant challenge due to the complex interplay of multiple physical factors. Gerliz M. Gutiérrez-Finol, Kirill Zinovjev, Alejandro Gaita-Ariño, and Salvador Cardona-Serra present a new computational platform that overcomes these limitations, employing a kinetic Monte Carlo approach to simulate ion migration in solid-state systems. This scalable method efficiently models both the random movement and collective behaviour of ions, capturing critical phenomena like charge relaxation and device learning rates with unprecedented detail. By combining computational efficiency with a physically realistic representation of materials, the platform offers a versatile tool for accelerating the development of high-performance, low-carbon energy storage and advanced electronic devices, and promises to advance exploratory research in these vital fields.

Polymer Memristor Model Validation and Data

This document provides supplementary materials and validation data for a computational model of polymer-based memristive devices, a type of non-volatile memory that changes resistance based on applied voltage history. The research simulates these devices and validates the model against experimental data, demonstrating how different parameters affect simulated behavior and providing detailed results beyond the scope of the main research paper. The simulations accurately capture the reduction of noise with an increasing number of independent simulation paths, suggesting effective averaging of random fluctuations. The model also accurately simulates device relaxation after a polarizing voltage is removed, with the relaxation time parameter controlling current decay.

Furthermore, simulations demonstrate that increasing the voltage sweep rate reduces current and hysteresis loop amplitude, consistent with physical limitations. Supplementary figures visually demonstrate these points, showing noise reduction with increasing simulation paths, comparing simulated relaxation decay curves with experimental data, illustrating the effect of voltage sweep rate on hysteresis loops, and displaying the time evolution of conductance for different relaxation times and applied voltage waveforms. This platform utilizes a vectorized, rail-based representation of device geometry, enabling rapid simulation of lateral ion transport and space-charge effects while preserving the stochastic nature of ion hopping events. The simulator accommodates a wide range of materials and allows direct integration of experimental input parameters without code modification, increasing its versatility and ease of use.

To enhance computational efficiency and reduce environmental impact, the team implemented the model using highly energy-efficient graphics processing units, significantly improving performance while minimizing the carbon footprint of large-scale simulations. Validation involved testing the simulator’s ability to replicate key behaviors observed in polymer-based memristive devices. The simulations accurately captured relaxation decay, current-voltage hysteresis, and spike-timing-dependent plasticity, demonstrating the platform’s ability to model complex ion-driven phenomena. Furthermore, the simulator successfully predicted learning and forgetting rates, validating its potential for exploring synaptic plasticity in neuromorphic computing applications. This platform balances physical realism with computational speed, revealing qualitative trends in ion-driven phenomena without the overhead of complex differential equation solvers. The core of the simulation involves calculating site-specific hopping probabilities for ions, determined by local effective fields that incorporate applied bias and Coulombic interactions with neighboring ions.

Hopping probabilities are assigned using a Boltzmann factor, where forward and backward probabilities are normalized to preserve detailed balance and allow drift and diffusion to emerge naturally. The simulator efficiently reproduces the interplay of drift, diffusion, and Coulombic interactions without explicitly solving Poisson’s equation, a significant advancement in computational efficiency. Validation of the platform using polymer-based memristive devices demonstrates its ability to capture key behaviors, including relaxation decay and current-voltage hysteresis. The simulation accurately models spike-timing-dependent plasticity, a crucial element in neural networks, and learning/forgetting rates, demonstrating its potential for advanced device modeling. By employing a vectorized and parallelizable implementation, the platform efficiently reproduces complex interactions, offering a scalable solution for exploring ion-driven phenomena in energy storage and devices. This approach utilizes a vectorized, rail-based representation of device geometry, facilitating rapid simulations of lateral ion transport and space-charge effects, and importantly, allows for easy integration of experimental data without code modification. The simulator’s predictions align with experimental observations from polymer-based memristive materials, accurately capturing behaviours such as relaxation decay, current-voltage hysteresis, and the dynamics of potentiation and depression relevant to learning processes.

By adopting a probabilistic treatment of physical events, the researchers eliminated the need for complex differential equation solvers, creating a computationally lightweight and flexible tool. This allows for rapid exploration of how electric field strength, material morphology, and material properties influence ionic behaviour, offering a practical complement to more resource-intensive techniques. The authors acknowledge that the model currently focuses on polymeric ion migration devices, and future work could extend its capabilities by incorporating effects such as charge trapping caused by defects, further enhancing its versatility and predictive power.