Fig. 2: Benchmarking of the foundation model. | Nature Communications

Fig. 2: Benchmarking of the foundation model.

From: A foundation machine learning potential with polarizable long-range interactions for materials modelling

Fig. 2: Benchmarking of the foundation model.

a The interaction energies for Na-Na (green) and Na-Cl (purple) dimers predicted by our model, the model without polarizable long-range interactions (w/o-lr model), and reference calculations using density functional theory (DFT)39. b Evaluation of polarizable interactions in water molecules. The water molecule model is oriented in the yz-plane, with oxygen and hydrogen atoms shown in red and white, respectively. The energy response curves as a function of external electric field (\({{{\mathbf{\epsilon }}}}\)) strength applied along the x-axis. Results are compared among our model (red), the Charge Equilibration (QEq)-based model25 (blue), and the reference DFT calculations (gray) serving as the reference values. c–e Comparison of bulk modulus from DFT calculations (reference values) and our model (prediction values): (c) Voigt approach (BV), (d) Reuss method (BR), and (e) Hill average (BVRH). The R-squared values (R2) and mean absolute errors (MAE) of the bulk modulus are also shown.

Back to article page