Department of Chemistry, and
College of Chemistry and Chemical Engineering, Xiamen University, Xiamen, China
Assoc. Prof. Pavlo O. Dral
June 9, 2022 (9:30-10:30 AM, Bangkok Time)
Accelerating and Improving Computational Chemistry with Artificial Intelligence/Machine Learning
Machine learning (ML) is an essential tool for moving forward the general field of artificial intelligence and ML has also made its way into computational chemistry, by accelerating and improving the accuracy of quantum chemical (QC) simulations [1-2]. After initial explorations resulting in general concepts such as Δ-learning , our research has yielded general-purpose, artificial intelligence-enhanced quantum mechanical method 1 (AIQM1),  which approaches the accuracy of golden-standard, traditional CCSD(T)/CBS approach for closed-shell, neutral organic molecules in their ground state at the speed of semiempirical quantum mechanical methods while retaining good accuracy for charged systems and excited states. AIQM1 can be used out-of-the-box, i.e., it does not need retraining. This method enables us to perform simulations we have not been able to do with either traditional quantum chemical approaches or with experimental techniques. AIQM1 and many other methods are implemented in our MLatom program package, which is a user-friendly package for atomistic machine learning simulations .
 P. O. Dral. Quantum Chemistry in the Age of Machine Learning. J. Phys. Chem. Lett. 2020, 11, 2336–2347. [Link]
 P. O. Dral, M. Barbatti. Molecular Excited States Through a Machine Learning Lens. Nat. Rev. Chem. 2021, 5, 388–405. [Link]
 R. Ramakrishnan, P. O. Dral, M. Rupp, O. A. von Lilienfeld. Big Data Meets Quantum Chemistry Approximations: The Δ-Machine Learning Approach. J. Chem. Theory Comput. 2015, 11, 2087–2096. [Link]
 P. Zheng, R. Zubatyuk, W. Wu, O. Isayev, P. O. Dral. Artificial Intelligence-Enhanced Quantum Chemical Method with Broad Applicability. Nat. Commun. 2021, 12, 7022. [Link]
 P. O. Dral, F. Ge, B.-X. Xue, Y.-F. Hou, M. Pinheiro Jr, J. Huang, M. Barbatti. MLatom 2: An Integrative Platform for Atomistic Machine Learning. Top. Curr. Chem. 2021, 379, 27. MLatom.com
Pavlo O. Dral is an Associate Professor at Xiamen University Pavlo O. Dral is an Associate Professor at Xiamen University since 2019. His research is focused on accelerating and improving quantum chemistry with artificial intelligence/machine learning since 2013, the topic on which he published ca. 20 peer-reviewed publications including articles in Nat. Commun., Nat. Rev. Chem., and Chem. Sci., Overall, Pavlo Dral has published over 40 articles, and his work has been cited over 2500 times with h-index of 21 (Google Scholar). Pavlo Dral is also the main developer of the MLatom package for atomistic machine learning simulations. His background is theoretical and computational chemistry, he obtained PhD in 2013 from University of Erlangen-Nuremberg and did post-doc with Prof. Walter Thiel from 2013 till 2019 in Max Planck Institute for Coal Research. More information is available on Dral’s group website dr-dral.com.
The Abdus Salam International Centre for Theoretical Physics (ICTP), Trieste, Italy
Prof. Sandro Scandolo
June 10, 2022 (9:00-09:45 AM, Bangkok Time)
First-Principles Molecular Dynamics: A Window into Planetary Cores
The interiors of planets are characterized by extreme conditions of pressure and temperature and are among the most inaccessible places in the universe. Matter at those conditions transforms in ways that challenge chemical intuition: water becomes a metal, methane dissociates into diamond, and carbon dioxide becomes a ultrahard solid. Computational modelling is in most cases the only tool we have to investigate matter at conditions that are difficult to reproduce even in the laboratory. In order to be predictive, simulations must be based on chemically accurate first-principle methods and at the same time they must be able to sample a large number of potentially relevant atomic configurations. In this talk I will illustrate how atomistic simulations based on molecular dynamics and density-functional theory have contributed to the understanding of the interiors of planets and also how they have led to the discovery of new materials. I will discuss challenges and limitations of the various approaches and highlight the role of machine learning in accelerating the simulations.
As Head of Research, ICTP, Prof. Sandro Scandolo coordinates the work of the various scientific sections, chair the Academic Committee, and advise the Director on strategic research priorities and emerging directions. He is a condensed matter physicist and currently work on the atomistic simulation of high-pressure phenomena. He is the author of about 140 scientific publications in peer-reviewed international journals (24,000 citations, H-index: 47 (GoogleScholar)) and he contributed to the birth of Quantum Espresso, a popular software suite for the quantum modeling of materials. He organized more than 40 Workshops and Schools in Trieste as well as in many other countries including Ghana, Ethiopia, Kenya, Sudan, Nigeria, Iran, Vietnam, India, Philippines, and Colombia.