Ab initio and classical molecular dynamic simulations of supercooled liquid iron alloys at conditions of Earth’s core. This dataset includes molecular dynamic data of supercooled liquid iron at high temperature and pressure used to describe the phenomenon of liquid to solid nucleation. All outputs are in text form and either represent the time taken for a simulation to freeze spontaneously, or the size-frequency distribution of solid-like clusters of atoms withing supercooled liquids. In limited cases, the complete outputs and inputs are included. Due to the large volume that these occupy, most outputs are simply the histograms of size-frequency and the scripts used to perform the calculations. Large numbers of these molecular dynamic simulations were performed at each target temperature and volume (to achieve constant pressure) with the only difference being the initial trajectories of atoms. Also included are the ab initio molecular dynamic simulations used to train the embedded atom models which define the behaviour of the classical simulations. These are only included for FeSi, FeS and FeC alloys because Fe and FeO are borrowed from the studies Alfè et al., 2002 and Davies et al., 2021. This is described in Wilson et al., 2023. This data covers the conditions at the centre of the Earth for a range of supercooled temperatures. Volumes vary to maintain 360 GPa whilst temperatures range from 3800 K to 6000 K. Compositions between 100% Fe, 3% S, Si and C, and 10% O are explored. Higher concentrations of S, Si and C are not explored due to instability of the defining embedded atom models. Simulations were conducted using the VASP (Kresse and Hafner, 1993) and LAMMPS (Thompson et al., 2022) packages, using the ARCHER2 HPC system. Simulations were carried out between June 2020 and February 2023. This data was collected to understand the process of nucleation within the Earth’s liquid core, in hopes of resolving the Inner Core Nucleation Paradox. Those interested in the process of nucleation and sub-critical nuclei will find the data particularly relevant. This data was collected by Alfred Wilson-Spencer, Monica Pozzo, Dario Alfè and Chris Davies of the University of Leeds and University College London. This was a part of the NERC grant Resolving the Inner Core Nucleation Paradox (NE/T000228/1).