Re-visitation of Two Models for Predicting Mechanically-Induced Disordering after Cryogenic Impact Milling

To compare the prediction accuracy of two models used to characterize the complete disordering potential of materials after extensive cryogenic milling.

Elastic shear moduli (μ_s) were simulated in silico. Comparison with available literature values confirmed that computations were reasonable. Complete disordering potential was predicted using the critical dislocation density (ρ_crit) and bivariate empirical models. To compare the prediction accuracy of the models, each material added for dataset expansion was cryomilled for up to 5 hr. Mechanical disordering after comminution was characterized using PXRD and DSC, and pooled with previously published results.

Simulated μ_s enabled predictions using the ρ_crit model for 29 materials. This model mischaracterized the complete disordering behavior for 13/29 materials, giving an overall prediction accuracy of 55%. The originally published bivariate empirical model classification boundary correctly grouped the disordering potential for 31/32 materials from the expanded dataset. Recalibration of this model retained a 94% prediction accuracy, with only 2 misclassifications.

Prediction accuracy of the ρ_crit model decreased with dataset expansion, relative to previously published results. Overall, the ρ_crit model was considerably less accurate relative to the bivariate empirical model, which retained very high prediction accuracy for the expanded dataset. Although the empirical model does not imply a mechanism, model robustness suggests the importance of glass transition temperature (T_g) and molar volume (M_v) on formation and persistence of amorphous materials following extensive cryomilling.