IN-PERSON EVENT - REGISTER HERE \;to attend
\nBy \;Prof. Tejs Vegge - \;Pioneer Cent er for Accelerating P2X Materials Discovery (CAPeX)\, Technical University of Denmark\, DTU Energy\, Lyngby\, De nmark
\nABSTRACT: Est ablishing a distributed infrastructure for autonomous materials discovery and synthesis plays a critical role in accelerating the development of adv anced energy materials in areas like sustainable batteries and electrocata lysts for the green transition. A central element in this process is the d evelopment of a closed-loop infrastructure or materials acceleration platf orm (MAP) [1\,2]\, where different nodes\, methods\, and even geographical ly distributed laboratory equipment can work jointly using autonomous work flows [3] to co-optimize materials and device-level properties. Here\, we show an example using the Fast INtention-Agnostic LEarning Server (FINALES ) framework to orchestrate a two-pronged optimization task\, where both op timization tasks vary the composition of a battery electrolyte composed of ethylene carbonate (EC)\, ethyl methyl carbonate (EMC)\, and lithium hexa fluorophosphate (LiPF6). One targets the optimization of ionic conductivity\, while the other aims to maximize the end-of-life (EOL) of c oin cells [4].
\nAnother key challenge is the controlled synthesis o f materials with a specific atomic structure. This underpins many technolo gical advances\, yet remains reliant on iterative\, trial-and-error approa ches. Nanoparticles (NPs)\, whose atomic arrangement dictates their emerge nt properties\, are particularly challenging to synthesize due to numerous tunable parameters. Here\, we introduce an autonomous approach explicitly targeting the synthesis of atomic-scale structures\, which autonomously d esigns synthesis protocols by matching real-time experimental total scatte ring (TS) and pair distribution function (PDF) data to simulated target pa tterns\, without requiring prior synthesis knowledge. We demonstrate this capability at a synchrotron\, successfully synthesizing two structurally d istinct gold NPs: 5 nm decahedral and 10 nm face-centred cubic structures. Ultimately\, specifying a simulated target scattering pattern\, thus repr esenting a bespoke atomic structure\, and obtaining both the synthesized m aterial and its reproducible synthesis protocol on demand may revolutioniz e materials design. Hence\, ScatterLab provides a generalizable blueprint for autonomous\, atomic structure-targeted synthesis across diverse system s and applications [5]. We also show how the application of a fine-tuned v ersion of the MACE-MP-0 foundation model [6] enables us to simulate the NP -synthesis conditions and guide the synthesis by reaching spatio-
\ntemporal scales that are outside the realm of traditional ab initio molecu lar dynamics simulations.
\nAs a third example\, we present FastCat - an AI-orchestrated self-driving closed-loop materials discovery system f or the autonomous discovery of platinum group metal-free multi-metal catal yst for alkaline OER (Oxygen Evolution Reaction) [7]. With FastCat\, we ha ve synthesized\, characterized\, and tested more than 500 Ni-based multiel ement layered double hydroxide (LDH) catalysts in one of the most extensiv e AI-orchestrated catalyst discovery campaigns to date. Our metaheuristic Bayesian Optimization identified known high-performance compositions and s everal novel multielement Ni-Fe-Cr-Co alloys with unprecedented overpotent ials at higher current densities.
\nFi nally\, we discuss how recent advancements in ML have demonstrated that si mple material representations like chemical formulas without any structura l information can sometimes achieve competitive property prediction perfor mance in common tasks. Our physics-based intuition would suggest that such representations are “incomplete\,” which indicates a gap in our under standing. A tomographic interpretation of structure-property rela tions is used to bridge that gap by defining what a material representatio n\, material properties\, the material\, and the relationships between the se are [8]. We apply concepts from information theory to verify this frame work by performing an exhaustive comparison of property-augmented represen tations on a range of materials’ property prediction objectives. Thus\, as scientists\, we might not know a priori which experiments or s imulations will ultimately provide the most valuable information to captur e the “ghost of the material\,” i.e.\, what is the fastest path to obt ain a complex material’s property
\n\;
\nSHORT BI O: \;Professor Tejs Vegge is Professor in Autonomous Material s Discovery at DTU Energy\, Technical University of Denmark\, and Director of the 13-year\, $45 M Danish Pioneer Center “CAPeX – Accelerating P2 X Materials Discovery.” He is a global leader in computational and AI-dr iven discovery of advanced batteries\, electrocatalysts\, and energy stora ge solutions\, with more than 250 scientific publications. His pioneering work integrates density functional theory\, uncertainty-aware machine lear ning\, and advanced in-situ characterization to enable predictive design o f nanostructured catalysts and electrochemical interfaces under real opera ting conditions. Professor Vegge has established and leads two major inter national consortia for “Materials Acceleration Platforms” – CAPeX an d BIG-MAP under BATTERY 2030+ – driving the global development of self-d riving laboratories for clean energy. He is also a member of the European Commission’s Technology Council for Advanced Materials and the high-leve l group on AI in Science\, and co-founder of the deep-tech startup PhaseTr ee.
\nH
osted by ICREA Prof. Maria Escudero Escribano\, NanoElectrocatalys
is and Sustainable Chemistry \;Group Leader.