Fig. 8: A separately trained denoiser effectively removes thermal perturbations (simulated at 1000 K or roughly 0.5Tm) in SiO2 β-quartz, α-cristobalite, and β-cristobalite polymorphs. | npj Computational Materials

Fig. 8: A separately trained denoiser effectively removes thermal perturbations (simulated at 1000 K or roughly 0.5Tm) in SiO2 β-quartz, α-cristobalite, and β-cristobalite polymorphs.

From: Score-based denoising for atomic structure identification

Fig. 8

The denoised structures match the ideal references used during training. 16 iterations were used for denoising. The silicon atoms are colored in beige and the oxygens in red.

Back to article page