Leveraging transfer learning for accurate estimation of ionic migration barriers in solids

Reshma Devi, Keith T. Butler, Gopalakrishnan Sai Gautam · 2025

Read Leveraging transfer learning for accurate estimation of ionic migration barriers in solids by Reshma Devi, Keith T. Butler, Gopalakrishnan Sai Gautam (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-06436.txt.

Introduction

Reshma Devi and co-authors use transfer learning in 2025 to forecast how easily ions move through solid materials. Ion speed sets battery charge rates, but the barrier behind it is hard to measure or compute. The team started from a network pre-trained on seven material properties at once. They reshaped it to tell apart the different paths an ion can take inside a crystal.

Fine-tuning used a hand-built set of about six hundred barrier values from the literature. The best model beat older methods and a general-purpose potential on held-out data. It also sorted good conductors from poor ones with fair accuracy. Faster screening could speed the hunt for better battery materials.

Catalog record

CREATORReshma Devi, Keith T. Butler, Gopalakrishnan Sai Gautam
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-06436

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Leveraging transfer learning for accurate estimation of ionic migration barriers in solids Reshma Devi, Keith T. Butler, Gopalakrishnan Sai Gautam Abstract Ionic mobility, which determines the rate performance of several applications, such as batteries, fuel cells, and electrochemical sensors, is exponentially dependent on the migration barrier ( E m E_{m} ) for ionic motion within solids, a quantity that is difficult to measure experimentally or estimate computationally. Previous approaches to identify materials with high ionic mobility have often relied on imprecise descriptors or rules-of-thumb given the lack of generalizable models to predict E m E_{m} swiftly and accurately. Here, we present a graph neural network based architecture that leverages principles of transfer learning to efficiently and accurately predict E m E_{m} across a diverse set of materials. We use a model (labeled MPT) that has been simultaneously pre-trained on seven distinct bulk properties, introduce modifications into the MPT model to build inductive bias on different migration pathways in a structure, and subsequently fine-tune (FT) on a manually-curated literature-derived dataset of 619 E m E_{m} data points calculated with density functional theory. Specifically, we propose four different architectural modifications of the MPT model that differ on the number of instances of the MPT model used, the structural images or bands taken as an input, and the presence/absence of attention layers. Importantly, our best-performing FT model (labeled MODEL-3) demonstrates substantial improvements in prediction accuracy compared to classical machine learning methods, graph models trained from scratch, and a universal machine learned interatomic potential, with a R 2 score of 0.703 and a mean absolute error of 0.261 eV on the test set. Notably, MODEL-3 is able to distinguish different migration pathways within a given structure and also demonstrates excellent ability to generalize across intercalant compositions and chemistries. As a classifier, MODEL-3 exhibits 80% accuracy and 82.8% precision in identifying materials that are ‘good’ ionic conductors (i.e., structures with E m < E_{m}< 0.65 eV). Thus, our work demonstrates the effective use of FT strategies, architectural modifications necessary to learn a target property, and the adaptability of our MPT model, which can be extended to make predictions on other data-scarce material properties. Finally, the demonstrated abilities of MODEL-3 highlights its potential use in identifying materials with high ionic mobility with applications in batteries and beyond. 1 Introduction Ionic conductivity ( \sigma ) or diffusivity ( D ) or mobility ( \mu ) in a crystalline solid is an important material property that governs the rate performance of several applications, such as rechargeable batteries, fuel cells, and electrochemical sensors. In the case of secondary batteries, the core operating principle involves the reversible movement of mobile ions (such as Li + , Na + , K + , Mg 2+ , or Ca 2+ ) between electrode materials that are typically intercalation compounds and across an electronically-insulating electrolyte, which facilitates electron exchange via the external circuit. [ 1 ] Thus, the rate performance of a given rechargeable battery is often limited by the ionic mobility of the electroactive ion in the intercalation electrode and/or in the solid electrolyte (in case of all-solid-state batteries). [ 2 , 3 ] Thus, significant improvements in rate performance of batteries (and other electrochemical applications) can be effected by the use of novel materials [ 4 ] that exhibit high \mu of the electroactive ion, indicating that computational or experimental screening strategies [ 5 ] need to predict \mu/\sigma/D swiftly and accurately for the identification of candidate materials. Mathematically, \mu and \sigma are directly proportional to one another, with \sigma usually described by the Nernst-Einstein equation (\sigma=\frac{q^{2}xD(x)}{k_{B}T}) , where q , x , k_{B} , and T are the charge and concentration of the intercalant, the Boltzmann constant, and temperature, respectively. D(x) relates the diffusive flux to the concentration gradient via Fick’s first law, [ 6 ] and can be expressed as D(x)=D_{J}(x)\theta(x) . D_{J} is the jump diffusion coefficient that captures the rate of atomic migrations and the correlations among atomic hops, and \theta is the thermodynamic factor that accounts for any non-ideal interactions between the intercalant and the diffusion carrier (typically vacancies). Thus, macroscopic diffusion of ions in a solid ( D ) is directly related to microscopic atomic hops ( D_{J} ) that happen within the crystal structure. In the case of ideal interactions between the intercalant and vacancies, where each atomic hop exhibits an identical frequency (or probability) of occurrence, D simplifies to Equation 1 . D=f.g.a^{2}.\nu.\exp\left(-\frac{E_{m}}{k_{B}T}\right) (1) f is the correlation factor, g is the geometric factor that describes the diffusion channel connectivity, a is the hop distance, \nu is the pre-factor that depends on vibrational frequencies of atoms, and E_{m} is the activation barrier associated with the atomic migration. Note that the x dependence of D can arise from x dependence of f , g , a , \nu , and/or E_{m} . Importantly, E_{m} is the most dominant factor that determines D (and by extension \sigma/\mu ), given that it has an exponential influence on D , and hence becomes the most crucial quantity to calculate/measure for rate performance estimation in any application. Considering reasonable battery performance metrics, such as charge/discharge at a rate of C/2 and an operating temperature of 300 K, the tolerable limit for E_{m} in electrodes lies between 525 meV for micron-sized particles and 650 meV for nano-sized particles. [ 7 ] Minimizing E_{m} is essential for high \mu , prompting efforts to identify factors that lower E_{m} . Previously proposed design principles for selecting structures with high \mu include avoiding preferred coordination environments, minimizing coordination number changes during migration, and maximizing volume per anion with non-close-packed structures. [ 7 , 8 ] However, these principles are not universally applicable, especially for large intercalants, prompting Lu et al. [ 9 ] to refine the strategies for Ca 2+ that emphasized optimal transition state geometry not containing face-sharing polyhedra and having higher degrees of freedom. Moreover, close-packed structures (i.e., structures without a high volume per anion) have also demonstrated high \sigma , as with the case of Mg 2+ in spinel chalcogenides. [ 10 , 11 , 12 , 13 ] Other structural descriptors that have been identified to correlate with E_{m} in solids, such as spinels, garnets, and olivines, include migrating ion-anion distances, [ 14 ] and the ‘migration number’ [ 15 ] that encompasses electronegativities, oxidation states, and ionic radii. Nevertheless, despite advancements in understanding factors that influence \mu (or E_{m} ) in specific systems, a generalized rule or model that is applicable across a wide variety of solid systems [ 16 ] remains elusive so far. For models predicting E_{m} to be practically useful, the models have to make predictions that are as accurate as and are significantly faster than experimental measurements or computational techniques. Additionally, for constructing generalizable models for E_{m} predictions, a reliable dataset of measured or calculated E_{m} across a wide variety of systems is necessary. In terms of measuring E_{m} , direct experimental techniques such as electrochemical impedance spectroscopy (EIS [ 17 , 18 ] ), galvanostatic intermittent titration technique, [ 19 , 20 , 21 , 22 ] and nuclear magnetic resonance based methods [ 23 , 24 , 25 ] are typically

Read the complete text