Fig. 7: Performance of CEGANN for dynamical classification during the nucleation and growth of a zeolite crystal from an amorphous parent phase. | npj Computational Materials

Fig. 7: Performance of CEGANN for dynamical classification during the nucleation and growth of a zeolite crystal from an amorphous parent phase.

From: CEGANN: Crystal Edge Graph Attention Neural Network for multiscale classification of materials environment

Fig. 7

a, b The radial distribution function between silica nodes (gT-T) and the number of silica neighbors (Qn) is very similar between the amorphous and zeolite phases. c CEGANN predicts the fraction of silica sites that are part of the zeolite, as it nucleates and grows from the synthesis mixture. The snapshots of the simulation box corresponding to points A–D are shown in the lower panel. Silica nodes of the amorphous phase are shown in orange, whereas the crystalline silica detected by CEGANN is shown in green. For clarity, the organic cations and water molecules are not shown.

Back to article page