果冻传媒

The total chain of energy research covers many length and time scales, starting with primary processes at the atomic scale, like chemical or nuclear reactions, thermal vibrations, light absorption and emission by molecules or crystalline solids, continuing all the way to macroscopic energy production or conversion systems, like reaction vessels, nuclear or fusion plants, solar cells, (heat) batteries, (photo-)electrochemical devices.

From atoms
At the atomic level, use is made of quantum mechanical ab-initio electronic structure theories, often based on Density Functional Theory (DFT). Because the calculation of forces between atoms by DFT is very expensive, force fields (FFs) are used to speed up Molecular Dynamics (MD) simulations of the time evolution of a system at the atomic level. The traditional construction of FFs is laborious, using complex parameterizations, with parameters determined from comparison with experiments and/or DFT calculations. Such FFs often only work in the specific context for which they were constructed.

To reactions
A recent breakthrough is the use of on-the-fly Machine-Learned FFs (MLFFs). This leads to a speed-up of many orders of magnitude, making simulations possible with ab-initio precision of realistic, large systems, including complex chemical reactions.

Next to this, microkinetic modeling provides the framework for the analysis of reaction mechanisms and is used to identify critical reaction intermediates and rate-determining reaction steps. By this, the limiting processes at electrochemical surfaces and interfaces can be identified, providing important information to design improved catalysts and structures.

Identifying molecules and materials
Researchers at CCER also employ AI techniques to assist in identifying molecules and materials that are potentially effective for various energy conversion and storage purposes. The primary task here is to sift through the vast space of chemically feasible molecules and materials to pinpoint appropriate candidates for further studies. In a process akin to a funnel, where potential candidates are progressively evaluated using more precise AI and computational methods, only a group of promising ones are left for experimental verification, thereby accelerating the discovery process while reducing associated costs.

Current and planned collaborations

• The CCER is nationally connected to Computational Science NL and is co-founder of the Dutch Computational Science conference (DUCOMS)
• The experience in AI and ML within CCER is broad and can be used in areas presently outside the scope of CCER. For example, MLFFs can be used to replace reactive force fields presently used in the simulation of the combustion of iron powder particles.