Ion Transport at Hybrid Inorganic-Organic Interfaces with Machine-Learning Molecular Dynamics
Title: Ion Transport at Hybrid Inorganic-Organic Interfaces with Machine-Learning Molecular Dynamics
DNr: Berzelius-2026-318
Project Type: LiU Berzelius
Principal Investigator: Waldemar Kaiser <waldemar.kaiser@liu.se>
Affiliation: Linköpings universitet
Duration: 2026-11-16 – 2027-06-01
Classification: 10407
Keywords:

Abstract

Ion transport plays a central role in the operation, performance, and degradation of a wide range of energy and electronic devices, from batteries and solar cells to memristors. Ionic motion is increasingly recognized as an important factor in the operation and stability of halide perovskite solar cells. However, while ion migration within the perovskite absorber is well studied, much less is known about the fate of mobile ions when they reach interfaces with organic charge-transport layers. Whether ions accumulate at these interfaces, are blocked by them, or migrate into the organic material remains poorly understood. Atomistic simulations can provide direct insight into these processes, but the heterogeneous nature of perovskite/organic interfaces and the long time and length scales associated with ion migration make them prohibitively expensive for conventional ab initio molecular dynamics (AIMD). Here, we will overcome this limitation using machine-learning interatomic potentials (MLIPs) to enable large-scale machine-learning molecular dynamics (MLMD) simulations of ion transport at perovskite/organic interfaces. We will fine-tune MLIPs starting from foundation models using training data from density functional theory calculations to accurately describe the energies and forces of heterogeneous interfaces. Particular emphasis will be placed on assessing the transferability of the MLIPs across different interface sizes, atomic configurations, and organic transport layers. Their accuracy will be validated against AIMD for representative model systems by comparing structural and vibrational properties as well as ion migration barriers. The MLIPs will then be used to simulate the ion dynamics at realistic perovskite/organic interfaces. We will determine how the organic layer interacts with the perovskite at pristine and defective surfaces, and we will probe the migration of hydrogen and halide ions from the perovskite into the organic layers. These simulations will be complemented by nudge elastic band calculations to quantify migration barriers and to identify favorable interfacial migration pathways. The project will establish an atomistic picture of ion migration across hybrid inorganic-organic interfaces. Beyond perovskite solar cells, the methodology developed here will provide a transferable framework for studying ionic dynamics at molecular interfaces relevant to batteries, mixed ionic-electronic conductors, and memristive devices. The PI has extensive experience with MLMD, including the accurate prediction of ion migration in solid electrolytes [1], ML-accelerated simulations of ion dynamics and vibrational properties [2], and MLIPs for dynamic halide perovskite surfaces [3] and electrolyte grain boundaries [4]. The project is supported by the Wallenberg Initiative Materials Science for Sustainability (WISE) and forms part of our broader effort to understand and control materials interfaces for sustainable technologies. [1] T. Miyagawa, W. Kaiser, et al., J. Mater. Chem. A 12, 11344–11361 (2024). [2] M. Grumet, W. Kaiser, et al., AI Sci. 2, 11001 (2026). [3] F. Delgado, W. Kaiser, et al., ACS Energy Lett. 10, 3367–3374 (2025), [4] H. Chu, W. Kaiser, et al. Nat. Nanotechnol. 21, 1174-1183 (2026)