This proposal comprises three parts: generate an interatomic machine learning (ML) potentials to describe the interaction in (i) yttrium hydrides (YHx), (ii) W-He and (iii) the cobalt-phosphorus (Co-P) binary system. The purpose of the first part is to investigate the impact of hydride formation on the mechanical properties YHx by means of large-scale atomistic modelling. This aims to investigate the mechanical properties additively materials containing dispersed yttrium hydride precipitates, which are efficient high-density moderators and necessary for commercial realization of microreactor technology. To this end we will conduct molecular dynamics modelling to extract the critical (i.e. Peierls) stress required to enable dislocation glide past the obstacles. Previous attempts to generate empirical potentials for the Y-H binary system have been fruitless because of the difficulty to represent the H interaction. To overcome tis hurdle we have generated an ML for Y based on the atomic-cluster expansion (ACE) formalism. It will be expanded to include H interaction. For this part of the project we collaborate with partners at ICAMS at Ruhr University, Bochum. To realize the project requires significant amounts of DFT-data generation, based both on ab initio molecular dynamics (AIMD) and conventional self-consistent modelling. The second part aims to investigate the impact of He bubbles on the mechanical properties of tungsten GBs. To this end we will generate a binary ACE potential and model the phenomenon using molecular dynamics modelling. The parametrization of ACE potentials requires systematic fitting to large DFT databases meticulously designed to cover the full descriptor space. For the current project this poses a challenge that the interaction between He atoms is largely governed by van der Waals (vdW) interaction. To capture such effects, it is necessary to deviate from conventional DFT modelling and incorporate vdW interaction via non-local exchange-correlation functionals. The final part aims to investigate the formation of Co-P compounds on the surface of Co nanoparticles. This aims to investigate the prospect of introducing organic alloying elements to reduce the melting temperature to facilitate the sintering of cemented carbides. To this end we will generate a binary ACE potential and model the phenomenon using molecular dynamics modelling. The parametrization of ACE potentials requires systematic fitting to large DFT databases meticulously designed to cover the full descriptor space. For the current project this poses two challenges. The first is that P has several morphologies with only subtle energy differences, whose relative stability is largely governed by van der Waals (vdW) interaction. To capture such effects, it is necessary to deviate from conventional DFT modelling and incorporate vdW interaction via non-local exchange-correlation functionals. The second challenge is the fact that Co is ferromagnetic, which necessitates a spin-polarized description. By jointly incorporating both effects, which to the best of the applicants’ knowledge has not yet been attempted, the basis for modelling intermetallic compounds, and e.g. metal-organic frameworks, will be improved.