SLUSCHI-UP predictions are continuously validated against experimentally measured melting temperatures through MeltBench, our open benchmark platform for inorganic materials.
In SLUSCHI-UP, the simulation cell size is automatically chosen to ensure a minimum radius of 11 Å and at least 100 atoms in the system.
MeltBench-10
A compact benchmark suite spanning metals, semiconductors, covalent solids, ionic solids, carbides, alloys, and difficult entropy-driven cases. Melting Temperatures are in Kelvin. Mo19Ru35Ta2W8 is based on DFT study.
MeltBench-100
A comprehensive benchmark platform. Coming up soon.
Highest melting point leaderboard
Top 10 Allegro entries ranked by PBE-corrected melting temperature.
Data
Column Definitions
Material: Reduced chemical formula of the simulated system.
MLP: Universal machine-learning interatomic potential (uMLIP) used in the SLUSCHI-UP simulation workflow.
GNN MT (K): Melting temperature predicted directly by the graph neural network model.
GNN NN Exp MT (K): Experimental melting temperature of the nearest-neighbor material identified by the GNN model.
GNN NN Material: Nearest-neighbor material selected by the GNN model.
GNN_Exp_MT: GNN-derived estimate of experimental melting temperature inferred from chemically similar materials.
experimental_MT_K: Experimental melting temperature collected from literature or benchmark databases.
Reconciled_MT_K: Final reference melting temperature used for benchmarking.
SLUSCHI-UP MP: Melting temperature predicted by the SLUSCHI-UP coexistence simulation workflow.
STD: Estimated statistical uncertainty of SLUSCHI-UP melting temperature prediction.
PBE_CORR: DFT correction to melting temperature estimated using PBE calculations.
R2SCAN_CORR: DFT correction to melting temperature estimated using R2SCAN calculations.
N_atoms: Total number of atoms in the simulation cell.
Rad: Effective simulation-cell radius used during structure generation.
