This course covers a theoretical introduction as well as a hands-on training session on machine learned potentials and their application in atomistic simulation with quantum mechanical accuracy. We will discuss the basic idea and theoretical foundation of the machine learning models, molecular representations, as well as the data generation process based on active learning. After that, we will guide you through a python based example, where you can do the whole workflow yourself, covering all steps from quantum simulations, data set generation, training of machine learning models and molecular dynamics simulation. In the first day of the workshop, we will cover the theory as well as set up jupyter notebooks, so we can use the second day for the hands-on session. To follow the hands-on exercise, a basic knowledge of python programming is required.
Patrick Reiser is a postdoctoral researcher at the Karlsruhe Institute of Technology, working on machine learning models for materials science. He is focusing on graph neural networks for property prediction of molecules and materials as well as on machine learning models for advanced atomistic simulations.
Pascal Friederich is tenure-track professor at the Karlsruhe Institute of Technology, leading the AiMat research group (https://aimat.science). His research focuses on the development and application of machine learning methods for molecular design, materials simulations and materials acceleration platforms.
This mini workshop, organized by WP5 of GCMAC, intends to provide an overview of different approaches and data management strategies that have recently been developed and adopted to enable the acceleration required in the discovery, design and integration of energy materials. A complex workflow leads from the innovative design of an advanced functional energy material or component to the integration and demonstration at the level of the fully operational energy device. An effective process along this design-to-demonstration path demands data standardization in view of typology and ontology. There is thus a need for the curation of large data sets to resolve and manage the heterogeneity of data representations, scale-bridging simulation techniques and methodologies. The workshop will provide state-of-the-art approaches to foster an end-to-end data management strategy that consists of data migration, data curation, and warehousing activities. The invited speakers will provide use-cases related to data management plans (DMP), development of autonomous data workflows, artificial intelligence techniques for self-learned procedures, and data science models.
16:05-16:35 Data management plan in project Big-Map
Speaker: Prof. Ivano E. Catelli (Technical University Denmark- DTU)
16:35-17:05 Interoperability and communication protocols for materials acceleration platforms.
Speaker: Dr. Michael Greenwood (CanmetMaterials, NRCAN, Canada)
17:05-17:35 Data management plan and workflow development in GCMAC
Speakers: Dr. Kourosh Malek, Dr. Jenia Jitsev (Forschungszentrum Juelich GmbH)
17:35-18:00 Q&A and round table discussions
This workshop is part of a series of scientific networking events that are organized by the German Canadian Materials Acceleration Centre (GC-MAC). The specific purpose of the workshop will be to ponder the question how theory and computation can help accelerate materials discovery, design and development in the field of electrocatalysis for electrochemical energy conversion and storage.
Three highly topical and interrelated aspects will be scrutinized:
understanding (deciphering) the local reaction environment that prevails at the surface (i.e., in the reaction plane) of an electrocatalytically active material under conditions (electrode potential, pH, ion concentrations, reactant concentrations) that exist in operating electrochemical cell when it is being operated (and not under conditions that are easy to handle in a simulation); devising a new set of “beyond-volcano” activity-stability descriptors that account for electronic as well as ionic effects (and if needed for transport phenomena at various scales); this set will be the basis for a highly effective comparative materials assessment (screening and selection for optimal functionality) and an optimized tailored design of novel materials and components; developing and using activity concepts that account for the multistep nature of the reaction of interest (beyond oversimplified single-step concepts such as rate-determining step or potential-determining step). Impulse presentations will set the stage for lively discussions about these topics.
Location: Lecture hall of the central library (building 04.7), Forschungszentrum JÜLICH GmbH, JÜLICH, Germany
Host: RWTH Aachen/FZ JÜLICH
Date And Time
2022-05-20 @ 09:00 AM to
2022-05-20 @ 04:00 PM
To register for this event email your details to firstname.lastname@example.org
Register using webmail: Gmail / AOL / Yahoo / Outlook
German Canadian Materials Acceleration Centre
Copyright © 2022 German Canadian Materials Acceleration Centre - All Rights Reserved.
GC-MAC is financially supported by the German Federal Ministry of Science and Education (BMBF) under grant number 01DM21001