Thesis supervisor: Péter Nagy
Location of studies (in Hungarian): BME Department of Physical Chemistry and Materials Science Abbreviation of location of studies: BME
Description of the research topic:
While the accuracy and reliable predictive power of the gold standard model of quantum chemistry, the CCSD(T) method, have been repeatedly corroborated against experiments, the reach of this method has only been extended most recently to molecules larger than a single amino acid. The reduced-cost and local CCSD(T) methods developed in our group for molecular energy computations were shown to offer outstanding efficiency reaching 100s or even up to a 1000 atoms, while retaining the inherent accuracy of CCSD(T) [1,2].
While we have parallel ongoing developments to implement similarly efficient CCSD(T) level methods for molecular properties other than the energy, the modeling of challenging thermodynamic and kinetic properties will still be only available for small molecules if we rely only on direct CCSD(T) property computations.
However, utilizing extensive recent progress in the field of machine learning force field (MLFF) developments, the demanding sampling of molecular configurations can be greatly accelerated. Here we will:
1) take advantage of both well established and state-of-the-art MLFF approaches to keep the number of configurations needed for the training MLFF training as small as possible and
2) more importantly, we will leverage our current record-fast, accurate, and continuously improving electronic structure models and codes to also reduce the time for a single data point generation
This competitive combination will allow to generate accurate and efficient MLFFs, which will be utilized to study dynamic, thermodynamic and kinetic properties of both gas and condensed phase molecular systems. These studies will be carried out as part of competitive research grants (such as our ERC Starting grant) using our MRCC [3] and other quantum chemistry program suites, in collaboration with the MRCC developer team and leading MLFF experts at ELTE, Budapest and the University of Luxembourg.
Requirements: introductory skills/experience in theoretical/computational chemistry and/or programming. Possible participation in additional Ph.D. talent support or scholarship programs are encouraged and supported.
[1] Journal of Chemical Theory and Computation 15, 5275 (2019) & 17, 860 (2021)
[2] Nature Communications 12, 3927 (2021), J. Am. Chem. Soc. 139, 17052 (2017)
[3] Journal of Chemical Physics 152, 074107 (2020), http://www.mrcc.hu
Number of students who can be accepted: 2
Deadline for application: 2024-05-30
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Login
