Background
This project presents a comprehensive framework for the predictive design and discovery of sustainable magnetic materials, bridging first-principles calculations, advanced machine learning methodology, including brain-like neural networks (BLNNs), and explainable artificial intelligence approaches.
Research question
This project aims to establish an advanced framework for predictive materials discovery, leveraging machine learning (ML) and brain-like neural networks (BLNNs) to target sustainable and functional magnetic materials. The focus is on identifying and characterizing novel topological magnets for neuromorphic computing paradigms that promote sustainability. The WISE-WASP connection is bringing a double benefit in addressing the research question, both by applying AI methods to materials design, and by designing materials enabling future computing paradigms.
Aim
By uniting expertise in density functional theory, atomistic spin dynamics, ML-based classification, and BLNNs for exploratory data analysis, this work will produce a highly adaptable toolset. It is designed initially for topological magnetic materials but will also be applicable for any class of functional magnets, providing predictive insights for both fundamental research and sustainable technology applications.
The objective is to demonstrate the use of magnetic neuromorphic computing for complex problem solving. The project’s defining feature is its commitment to explainability at each stage, making complex magnetic interactions and computational processes scientifically interpretable and enabling robust material design and discovery.
Synergy and Team
This project brings together researchers from Computer Science and Materials Science from KTH, Uppsala University, and Örebro University, with extensive expertise in their respective research areas, and strong networks within both Swedish and international research communities. The NEST project will extend and strengthen the existing collaborations between these researchers: four of the PIs involved in the project (M. Pankratova, S. Lowry, A. Edström, P. Herman) were leading two WASP-WISE pilot projects. These pilot projects provided an initial platform for collaboration, which will be further developed during the NEST project. The integration of two teams, combined with the addition of A. Bergman will enhance the versatility of the NEST environment, creating an ideal setting for the development and mentorship of PhD students.
Sustainability aspects
The soaring energy consumption of information technology urgently calls for the development of more energy-efficient information processing technology. One possible source of inspiration in addressing this crucial problem is the human brain, which is estimated to process information 1 million times more energy efficiently than currently available computers. The final goal of this project is to use efficient combinations of atomistic spin models and AI-based computational methods to optimize the properties of topological magnetic materials for neuromorphic or other unconventional computing. In particular, the project will investigate stabilization mechanisms and responses to external stimuli, including electrical currents (in metals) or applied electric fields (in insulators), magnetic fields, spin-orbit torques, mechanical deformations/strain fields, temperature, optical pulses, as well as gradients in previously mentioned fields. This will allow for identifying energy-efficient means of stabilizing and controlling the behavior of topological spin textures for next-generation computing applications
Contact Main PI
Pawel Herman, paherman@kth.se
Maryna Pankratova, maryna.pankratova@physics.uu.se
