Vibrational spectroscopy of complex systems from neural network force fields

PhD n°8


EU mobility rules apply. In principle, applicants can have any nationality and any current residence (although immigration rules apply, favoring
EU applicants). Candidates who have already been awarded a PhD degree are not eligible. In addition, candidates who have already spent more than 12 months in the Germany within the last 3 years are not eligible (unless as part of a procedure for obtaining refugee status under the Geneva Convention).


This project will develop a machine-learning software to simulate spectroscopic properties of interfaces with neural-network force fields using analytical second-derivatives. This will allow for a numerically precise method (much more than a frozen-phonon approach) capable of delivering the
vibrational frequencies of systems with thousands of atoms with DFT quality. This tool will then be used to investigate the localized vibrational modes of several interfaces. We will then extend our software to provide the analytical third-order derivatives that open the door for the investigation of thermal transport properties and of Raman spectra. The energetically favoured interface reconstructions will be efficiently identified by crystal structure prediction, using machine learning force fields, together with a constrained variant of the minima hopping method. Electronic band diagrams across the interface will also be calculated using density functionals for interfaces, developed and tested in Jena. Researcher8 will contribute to WP2 and WP3.

The specific objectives are:

  • Software development: vibrational and phonon spectrum from analytical differentiation of the neural network.
  • Software development: third derivatives and calculation of thermal transport coefficients.
  • Training of neural-network force fields for the study of few paradigmatic systems also studied by Researcher7.
  • Apply neural-network force fields, to further speed up calculations of realistic atomic structures of interfaces. A library of force fields, that can provide accurate energies and forces, covering the whole periodic table is necessary for this task.
  • Study of phonon scattering by interfaces.

Expected Results:

  • Software tool to calculate vibrational and thermal properties from neural-network force-fields. This tool will be open-sourced and integrated with existing software.
  • Raman spectrum of point defects that can be directly used for the understanding of experimental data.
  • Understanding and quantitative evaluation of the phonon scattering at point defects and interfaces.

Planned secondment(s): 6 months

  • Intersectoral: TME; M18-M20; training on ML force fields for structural prediction of vibrational spectra of interfaces.
  • Academic: DTU, K.S. Thygesen; M30-M32; learning on physics and simulation of defects in bulk materials, aiming to extend the study to defects at interfaces and interface functionalization through control of defects.

Enrolment in Doctoral degree(s):

Martin Luther University Halle


Miguel Marques

PhD n°: PhD n°8
Country: Germany
This job is no longer accepting applications.