Autonomous Discovery of Materials for Intercalation Electrodes

Abstract

The development of automated computational tools is required to accelerate the discovery of novel battery materials. In this work, we design and implement a workflow, in the framework of Density Functional Theory, which autonomously identifies materials to be used as intercalation electrodes in batteries, based on descriptors like adsorption energies and diffusion barriers. A substantial acceleration for the calculations of the kinetic properties is obtained due to a recent implementation of the Nudged Elastic Bands (NEB) method, which takes into consideration the symmetries of the system to reduce the number of images to calculate. We have applied this workflow to discover new cathode materials for Mg batteries, where two of these materials display a threefold increase in the potential of the Chevrel phase, the state‐of‐the‐art cathode in commercial prototype Mg batteries.

Toward Better and Smarter Batteries by Combining AI with Multisensory and Self‐Healing Approaches

“…By building upon the experience from existing computational software packages like the Atomic Simulation Environment (ASE) [34] and workflow management tools like FireWorks, [35] AiiDA, [36] or MyQueue, [34] it can become possible to include experimental data with full provenance to accelerate the search for new battery materials. [7] These tools will be made available to the battery community through the BIG-MAP App Store [37] and GitHub registry, [38] and should ultimately facilitate experiments by autonomously launching simulations and simulations initiating and running experiments through the BIG-MAP orchestration infrastructure (Figure 2).…”

“…[4] Developing a versatile and chemistry neutral infrastructure that is capable of monitoring, predicting, and controlling the dynamic properties and evolution of these interfaces and interphases like the solidelectrolyte interphase (SEI), is a cornerstone of the long-term roadmap of the large-scale European initiative BATTERY 2030+ [5] and the BIG-MAP project in particular. While the battery development process has historically been limited by a sequential and time-consuming discovery pipeline, the maturation of automated computer simulations at the atomic [6,7] to micro [8] and mesostructural [9] scale, data-based [10] and data-driven [11] approaches to predict materials properties yield new opportunities to accelerate the discovery of battery materials and cell designs. [12] The development of generic procedures for automated translation of information across scales from the atomic and microstructural to the cell level is, however, non-trivial.…”

Autonomous Intelligent Agents for Accelerated Materials Discovery

“…9,10 This new wave of AI in materials has a theme of acceleration, enabled either by bypassing tools or methods via surrogate models, [18][19][20][21][22][23] or by identication of new materials via adaptive schemes that combine models with decision-making approaches. [24][25][26][27][28][29][30][31][32] A number of autonomous frameworks for materials discovery have been designed and demonstrated, such as optimization of carbon nanotube syntheses via a robotics interface; 27 organic molecule synthesis robots 33,34 for autonomously navigating complex chemical reaction networks with reagentbased decision-making; and composition-based autonomous search for low thermal hysteresis shape-memory alloys. 35 Among other efforts, [36][37][38] these demonstrate proof-of-concepts for autonomous discovery of materials for specic target properties or applications.…”

Autonomous Discovery of Materials for Intercalation Electrodes

“…During the last decades, quantum mechanical calculations, mostly within the framework of Density Functional Theory (DFT), have been efficiently used to predict properties and design novel materials for multiple applications. Recently, to accelerate the materials discovery, autonomous workflow schemes have been implemented, [5,6,7,8,9] and more specifically for the discovery of battery materials, where a special focus was put on the kinetic properties, i.e. cation diffusivity.…”

“…The Mg 2+ ion occupies a tetrahedral site in structures (a-d).In the Garnet structures (e,f) Mg 2+ has eight closest oxygen neighbors. The structures are visualized using VESTA [48]. .…”