Machine Learning Prediction of the Critical Cooling Rate for Metallic Glasses from Expanded Datasets and Elemental Features

Abstract

We use a random forest (RF) model to predict the critical cooling rate (R C) for glass formation of various alloys from features of their constituent elements. The RF model was trained on a database that integrates multiple sources of direct and indirect R C data for metallic glasses to expand the directly measured R C database of less than 100 values to a training set of over 2000 values. The model error on 5-fold cross-validation (CV) is 0.66 orders of magnitude in K/s. The error on leave-out-one-group CV on alloy system groups is 0.59 log units in K/s when the target alloy constituents appear more than 500 times in training data. Using this model, we make predictions for the set of compositions with melt-spun glasses in the database and for the full set of quaternary alloys that have constituents which appear more than 500 times in training data. These predictions identify a number of potential new bulk metallic glass systems for future study, but the model is most useful for the identification of alloy systems likely to contain good glass formers rather than detailed discovery of bulk glass composition regions within known glassy systems.

Extracting accurate materials data from research papers with conversational language models and prompt engineering

“…Out of the 309 unique datapoints, there were 222 unique material compositions (some compositions had multiple values originating from different research papers) in the standardized database, and after removing non-metallic systems standardized-MG database contained 298 unique datapoints for 217 unique material compositions. This size of 217 unique compositions is significantly larger than the previous largest hand-curated database published by Afflerbach et al 40 , which had just 77 entries. This result shows that, at least in this case, ChatExtract can generate more quality data with much less time than human extraction efforts.…”

Flexible, model-agnostic method for materials data extraction from text using general purpose language models

“…Within these, 129 are unique systems (multiple values are reported for some systems and we kept these to allow the user to manage them as they wish). The database is larger than the size of a recently published manually curated database of critical cooling rates, 59 which is the most state-of-the-art and complete such database of which we are aware, and consists of only 77 unique compound datapoints. To provide comparison to other existing methods, we used ChemDataExtractor2 (CDE2), 15 a state-of-the-art named entity recognition (NER) based data extraction tool.…”

Machine Learning-Guided Exploration of Glass-Forming Ability in Multicomponent Alloys

“…39 In particular, a variety of ML algorithms have been applied to correlate the glass structure and chemistry with elasticity, 29 the propensity for plastic deformation, 16,17,23,33 and thermal and physical properties. [18][19][20][21][22][24][25][26][27][28][29][30][31][32] In regard to GFA, ML models demonstrate success in predicting the glass transition temperature, 35,36 the critical cooling rate, 37 and the critical casting diameter of MGs, 15,21,26,28 and further identify new glass-forming systems, 22,31 using random forest, 25,32 support vector machine (SVM), 28 and neural network 18,31,34 algorithms. Until now, the ML studies of GFA have been mainly focused on binary and ternary alloy systems, and further efforts are required to explore multicomponent alloys using ML algorithms.…”