Regensburg 2022 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

O: Fachverband Oberflächenphysik

O 45: Focus Session: Catalysis at Liquid Interfaces

O 45.5: Topical Talk

Wednesday, September 7, 2022, 16:15–16:45, H4

Understanding liquid metal catalysts for graphene synthesis using machine learning interatomic potentials — •Hendrik H. Heenen — Fritz-Haber-Institut der MPG, Berlin, Germany

High-quality, near defect-free graphene can be synthesized on the levelled and uniform surfaces of liquid metal catalysts. This smoothness on the microscale is sometimes accompanied by seemingly different catalytic properties, the determination of which is, however, ambiguous. Assessing distinct catalytic properties of liquid metal catalysts by first principles atomistic simulations has so far been challenged due to the intractable long length and time scales necessary to model the liquid phase. Using computationally efficient machine learning interatomic potentials (MLIPs) trained to first principles data allows to extrapolate predictability to necessary scales and opens an avenue for obtaining the desired microscopic insight.

In this talk I will present strategies to train and employ MLIPs for the simulation of graphene synthesis on liquid metal catalysts. I will introduce the data-efficient training of MLIPs via fairly automatic workflows as a tool to extend the predictive accuracy of e.g. density functional theory to larger scales. On basis of these potentials, large-scale simulations can be performed to compute experimental observables and elucidate microscopic processes relevant to graphene synthesis. Further, one can identify trends between different metals and directly compare between the solid and liquid states of a catalyst. Findings based on these simulation approaches shed new light on the role of the liquid state of liquid metal catalysts.