string(13) "dissertations"
  • Python / C++ programming; Object oriented programming basics; Probabilities; Linear algebra
  • Machine learning toolkits (e.g. Scikit learn, PyTorch); Data analysis; Understanding of basic molecular dynamics concepts
  • George Giannakopoulos
  • ggianna [at] iit.demokritos.gr
  • The AI Lab

Polymeric systems are complex materials of great technological importance, in manufacturing, packaging, and for environmental applications. They are particularly challenging to study with molecular simulations, which is a powerful technique to investigate at the molecular level the properties of complex chemical systems, but it is extremely computationally intensive. Multiscale simulation strategies are thus required for polymer systems, that are usually elaborate and system-specific. These schemes could be generalized and streamlined by the application of Machine Learning (ML) methods.

A main advantage of ML models in the context of molecular simulations is that they are not constrained to a predefined mathematical function, therefore they are endowed with higher flexibility and expressive character compared to traditional molecular models. Recently, neural networks have shown great promise in the development of improved atomistic force fields.

In the overarching project into which the work will be integrated, we investigate the use of Machine Learning methods (including deep learning alternatives) to create optimized molecular force fields, using suitable descriptors for the local environment. The obtained potential will be integrated into a Molecular Dynamics open-source software (LAMMPS) to perform large-scale molecular simulations. Finally, we aim to create an automated pipeline for the evaluation of the Machine Learned force field against structural and thermodynamic properties of the systems.

The project is conducted at the National Centre of Scientific Research Demokritos and involves the synergy of Software and Knowledge Engineering Lab (SKEL | The AI lab) of the Institute of Informatics Telecommunications and the Molecular Thermodynamics and Modeling of Materials Laboratory (MTMML) of the Institute of Nanoscience and Nanotechnology.

Since the project can be either an internship (at least 2 months full time), or a BSc/MSc thesis, the exact assignment will be decided on a case-by-case basis with the interested student.

Supervisors

Dr Georgios Giannakopoulos, ggianna@iit.demokritos.gr

Dr Niki Vergadou, n.vergadou@inn.demokritos.gr

Relevant literature

  1. Husic, B. E.; Charron, N. E.; Lemm, D.; Wang, J.; Pérez, A.; Majewski, M.; Krämer, A.; Chen, Y.; Olsson, S.; De Fabritiis, G.; Noé, F.; Clementi, C. Coarse Graining Molecular Dynamics with Graph Neural Networks. J. Chem. Phys. 2020, 153 (19). https://doi.org/10.1063/5.0026133
  2. Doerr, S.; Majewski, M.; Pérez, A.; Krämer, A.; Clementi, C.; Noe, F.; Giorgino, T.; De Fabritiis, G. TorchMD: A Deep Learning Framework for Molecular Simulations. J. Chem. Theory Comput. 2021, 17 (4), 2355–2363. https://doi.org/10.1021/acs.jctc.0c01343
Figure from: Pinheiro, Max, et al. “Choosing the right molecular machine learning potential.” Chemical Science 12.43 (2021): 14396-14413.
https://doi.org/10.1039/D1SC03564A Licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/)
Skip to content