Machine‐Learning and Data‐Intensive Methods for Accelerating the Development of Rechargeable Battery Chemistries: A Review

Sendek, Austin D.; Cubuk, Ekin D.; Ransom, Brandi; Nanda, Jagjit; Reed, Evan J.

doi:10.1002/9783527817252.ch16

Cited by 6 publications

(5 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Off late, materials screening approaches using high-throughput density functional theory (DFT , ) calculations with/without machine learning (ML), have resulted in a significant number of theory-predicted candidates in a variety of applications, including batteries. ,− Indeed, some of the theoretical predictions have also been validated by subsequent experiments. − However, modeling disordered rocksalt compositions, and subsequently performing a computational screening across various compositions is nontrivial, owing to the configurational complexity and length-scale of the system. Specifically, disordered rocksalts do not have significant long-range order, necessitating large supercells, which in turn results in multiple symmetrically distinct Li-TM arrangements to consider.…”

Section: Introductionmentioning

confidence: 99%

Constructing and Evaluating Machine-Learned Interatomic Potentials for Li-Based Disordered Rocksalts

Choyal,

Sagar,

Sai Gautam

2024

J. Chem. Theory Comput.

View full text Add to dashboard Cite

Lithium-based disordered rocksalts (LDRs), which are an important class of positive electrode materials that can increase the energy density of current Li-ion batteries, represent a significantly complex chemical and configurational space for conventional density functional theory (DFT)-based highthroughput screening approaches. Notably, atom-centered machine-learned interatomic potentials (MLIPs) are a promising pathway to accurately model the potential energy surface of highly disordered chemical spaces, such as LDRs, where the performance of such MLIPs has not been rigorously explored yet. Here, we represent a comprehensive evaluation of the accuracy, transferability, and ease of training of five atom-centered MLIPs, including the artificial neural network potentials developed by the atomic energy network (AENET), the Gaussian approximation potential (GAP), the spectral neighbor analysis potential (SNAP) and its quadratic extension (qSNAP), and the moment tensor potential (MTP), in modeling a 11-component LDR chemical space. Specifically, we generate a DFT-calculated data set of 10,842 configurations of disordered LiTMO 2 and TMO 2 compositions, where TM = Sc, Ti, V, Cr, Mn, Fe, Co, Ni, and/or Cu. To provide a point-of-comparison on the performance of atom-centered MLIPs, we also trained the neural equivariant interatomic potential (NequIP) on a subset of our data. Importantly, we find AENET to be the best potential in terms of accuracy and transferability for energy predictions, while MTP is the best for atomic forces. While AENET is the fastest to train among the MLIPs considered at low number of epochs (300), the training time increases significantly as epochs increase (3300), with a corresponding reduction in training errors (∼60%). Note that AENET and GAP tend to overfit in small data sets, with the extent of overfitting reducing with larger data sets. Finally, we observe AENET to provide reasonable predictions of average Li-intercalation voltages in layered, single-TM LiTMO 2 frameworks, compared to DFT (∼10% error on average). Our study should pave the way both for discovering novel disordered rocksalt electrodes and for modeling other configurationally complex systems, such as high-entropy ceramics and alloys.

show abstract

Section: Introductionmentioning

confidence: 99%

Constructing and Evaluating Machine-Learned Interatomic Potentials for Li-Based Disordered Rocksalts

Choyal,

Sagar,

Sai Gautam

2024

J. Chem. Theory Comput.

View full text Add to dashboard Cite

show abstract

“…As electrification becomes the norm, these various utilities will require tailored performance optimization of cycle life, operating voltage, and power. In many previous works, battery materials have been investigated for maximizing these properties among others and the methods developed therein could be repurposed to target specific values, not just the maxima. − The vast growth of the battery market would also be easier to sustain with a more diverse set of battery materials. The detrimental environmental and social effects from the dependence on cobalt for Li–Co–O cathodes have become apparent .…”

Section: Introductionmentioning

confidence: 99%

Electrolyte Coatings for High Adhesion Interfaces in Solid-State Batteries from First Principles

Ransom,

Ramdas,

Lomeli

et al. 2023

ACS Appl. Mater. Interfaces

Self Cite

View full text Add to dashboard Cite

We introduce an adhesion parameter that enables rapid screening for materials interfaces with high adhesion. This parameter is obtained by density functional theory calculations of individual single-material slabs rather than slabs consisting of combinations of two materials, eliminating the need to calculate all configurations of a prohibitively vast space of possible interface configurations. Cleavage energy calculations are used as an upper bound for electrolyte and coating energies and implemented in an adapted contact angle equation to derive the adhesion parameter. In addition to good adhesion, we impose further constraints in electrochemical stability window, abundance, bulk reactivity, and stability to screen for coating materials for next-generation solid-state batteries. Good adhesion is critical in combating delamination and resistance to lithium diffusivity in solid-state batteries. Here, we identify several promising coating candidates for the Li7La3Zr2O12 and sulfide electrolyte systems including the previously investigated electrode coating materials LiAlSiO4 and Li5AlO8, making them especially attractive for experimental optimization and commercialization.

show abstract

“…Compared to previous deployments of machine learning [15] in the field of battery electrolyte optimization, [11] we investigate whether an improvement in conductivity may already be achieved through a single iteration cycle. This approach is mostly analogous to the workflow of Attia et al [16] for fast charging protocol optimization, as herein we are using a highthroughput electrolyte formulation robot and a machine learning based optimizer, that were not integrated and in fact run at two different locations asynchronously.…”

Section: Introductionmentioning

confidence: 99%

One-shot active learning for globally optimal battery electrolyte conductivity

Rahmanian

Vogler

Wölke

et al. 2022

Preprint

View full text Add to dashboard Cite

Non-aqueous aprotic battery electrolytes, among other requirements, need to perform well over a wide range of temperatures in practical applications. Herein we present a one-shot active learning study to find all conductivity optima, confidence bounds, and relating trends in the temperature range from 30 °C to 60 °C. This optimization is enabled by a high-throughput formulation and characterization setup guided by one-shot active learning utilizing robust and heavily regularized polynomial regression. Whilst there is an initially good agreement for intermediate and low temperatures, there is a need for the active learning step to globally improve the model. Optimized electrolyte formulations likely correspond to the highest physically possible conductivities within this system when compared to literature data. A thorough error propagation analysis yields a fidelity assessment of conductivity measurements and electrolyte formulation.

show abstract

Machine‐Learning and Data‐Intensive Methods for Accelerating the Development of Rechargeable Battery Chemistries: A Review

Cited by 6 publications

References 59 publications

Constructing and Evaluating Machine-Learned Interatomic Potentials for Li-Based Disordered Rocksalts

Constructing and Evaluating Machine-Learned Interatomic Potentials for Li-Based Disordered Rocksalts

Electrolyte Coatings for High Adhesion Interfaces in Solid-State Batteries from First Principles

One-shot active learning for globally optimal battery electrolyte conductivity

Contact Info

Product

Resources

About