Aggressively optimizing validation statistics can degrade interpretability of data-driven materials models.

One of the key factors in enabling trust in artificial intelligence within the materials science community is the interpretability (or explainability) of the underlying models used. By understanding what features were used to generate predictions, scientists are then able to critically evaluate the credibility of the predictions and gain new insights. Here, we demonstrate that ignoring hyperparameters viewed as less impactful to the overall model performance can deprecate model explainability. Specifically, we demonstrate that random forest models trained using unconstrained maximum depths, in accordance with accepted best practices, often can report a randomly generated feature as being one of the most important features in generated predictions for classifying an alloy as being a high entropy alloy. We demonstrate that this is the case for impurity, permutation, and Shapley importance rankings, and the latter two showed no strong structure in terms of optimal hyperparameters. Furthermore, we demonstrate that, for the case of impurity importance rankings, only optimizing the validation accuracy, as is also considered standard in the random forest community, yields models that prefer the random feature in generating their predictions. We show that by adopting a Pareto optimization strategy to model performance that balances validation statistics with the differences between the training and validation statistics, one obtains models that reject random features and thus balance model predictive power and explainability.

A High-Throughput Structural and Electrochemical Study of Metallic Glass Formation in Ni-Ti-Al.

On the basis of a set of machine learning predictions of glass formation in the Ni-Ti-Al system, we have undertaken a high-throughput experimental study of that system. We utilized rapid synthesis followed by high-throughput structural and electrochemical characterization. Using this dual-modality approach, we are able to better classify the amorphous portion of the library, which we found to be the portion with a full width at half maximum (fwhm) of >0.42 Å-1 for the first sharp X-ray diffraction peak. Proper phase labeling is important for future machine learning efforts. We demonstrate that the fwhm and corrosion resistance are correlated but that, while chemistry still plays a role in corrosion resistance, a large fwhm, attributed to a glassy phase, is necessary for the highest corrosion resistance.

An Inter-Laboratory Study of Zn-Sn-Ti-O Thin Films using High-Throughput Experimental Methods.

High-throughput experimental (HTE) techniques are an increasingly important way to accelerate the rate of materials research and development for many technological applications. However, there are very few publications on the reproducibility of the HTE results obtained across different laboratories for the same materials system, and on the associated sample and data exchange standards. Here, we report a comparative study of Zn-Sn-Ti-O thin films materials using high-throughput experimental methods at National Institute of Standards and Technology (NIST) and National Renewable Energy Laboratory (NREL). The thin film sample libraries were synthesized by combinatorial physical vapor deposition (cosputtering and pulsed laser deposition) and characterized by spatially resolved techniques for composition, structure, thickness, optical, and electrical properties. The results of this study indicate that all these measurement techniques performed at two different laboratories show excellent qualitative agreement. The quantitative similarities and differences vary by measurement type, with 95% confidence interval of 0.1-0.2 eV for the band gap, 24-29 nm for film thickness, and 0.08 to 0.37 orders of magnitude for sheet resistance. Overall, this work serves as a case study for the feasibility of a High-Throughput Experimental Materials Collaboratory (HTE-MC) by demonstrating the exchange of high-throughput sample libraries, workflows, and data.

Automated Phase Segmentation for Large-Scale X-ray Diffraction Data Using a Graph-Based Phase Segmentation (GPhase) Algorithm.

The creation of composition-processing-structure relationships currently represents a key bottleneck for data analysis for high-throughput experimental (HTE) material studies. Here we propose an automated phase diagram attribution algorithm for HTE data analysis that uses a graph-based segmentation algorithm and Delaunay tessellation to create a crystal phase diagram from high throughput libraries of X-ray diffraction (XRD) patterns. We also propose the sample-pair based objective evaluation measures for the phase diagram prediction problem. Our approach was validated using 278 diffraction patterns from a Fe-Ga-Pd composition spread sample with a prediction precision of 0.934 and a Matthews Correlation Coefficient score of 0.823. The algorithm was then applied to the open Ni-Mn-Al thin-film composition spread sample to obtain the first predicted phase diagram mapping for that sample.

A high-throughput investigation of Fe-Cr-Al as a novel high-temperature coating for nuclear cladding materials.

High-temperature alloy coatings that can resist oxidation are urgently needed as nuclear cladding materials to mitigate the danger of hydrogen explosions during meltdown. Here we apply a combination of computationally guided materials synthesis, high-throughput structural characterization and data analysis tools to investigate the feasibility of coatings from the Fe­Cr­Al alloy system. Composition-spread samples were synthesized to cover the region of the phase diagram previous bulk studies have identified as forming protective oxides. The metallurgical and oxide phase evolution were studied via in situ synchrotron glancing incidence x-ray diffraction at temperatures up to 690 K. A composition region with an Al concentration greater than 3.08 at%, and between 20.0 at% and 32.9 at% Cr showed the least overall oxide growth. Subsequently, a series of samples were deposited on stubs and their oxidation behavior at 1373 K was observed. The continued presence of a passivating oxide was confirmed in this region over a period of 6 h.

Optical cell for combinatorial in situ Raman spectroscopic measurements of hydrogen storage materials at high pressures and temperatures.

An optical cell is described for high-throughput backscattering Raman spectroscopic measurements of hydrogen storage materials at pressures up to 10 MPa and temperatures up to 823 K. High throughput is obtained by employing a 60 mm diameter × 9 mm thick sapphire window, with a corresponding 50 mm diameter unobstructed optical aperture. To reproducibly seal this relatively large window to the cell body at elevated temperatures and pressures, a gold o-ring is employed. The sample holder-to-window distance is adjustable, making this cell design compatible with optical measurement systems incorporating lenses of significantly different focal lengths, e.g., microscope objectives and single element lenses. For combinatorial investigations, up to 19 individual powder samples can be loaded into the optical cell at one time. This cell design is also compatible with thin-film samples. To demonstrate the capabilities of the cell, in situ measurements of the Ca(BH(4))(2) and nano-LiBH(4)-LiNH(2)-MgH(2) hydrogen storage systems at elevated temperatures and pressures are reported.

A combinatorial characterization scheme for high-throughput investigations of hydrogen storage materials.

In order to increase measurement throughput, a characterization scheme has been developed that accurately measures the hydrogen storage properties of materials in quantities ranging from 10 ng to 1 g. Initial identification of promising materials is realized by rapidly screening thin-film composition spread and thickness wedge samples using normalized IR emissivity imaging. The hydrogen storage properties of promising samples are confirmed through measurements on single-composition films with high-sensitivity (resolution <0.3 µg) Sievert's-type apparatus. For selected samples, larger quantities of up to ∼100 mg may be prepared and their (de)hydrogenation and micro-structural properties probed via parallel in situ Raman spectroscopy. Final confirmation of the hydrogen storage properties is obtained on ∼1 g powder samples using a combined Raman spectroscopy/Sievert's apparatus.

Demonstration of magnetoelectric scanning probe microscopy.

A near-field room temperature scanning magnetic probe microscope has been developed using a laminated magnetoelectric sensor. The simple trilayer longitudinal-transverse mode sensor, fabricated using Metglas as the magnetostrictive layer and polyvinylidene fluoride as the piezoelectric layer, shows an ac field sensitivity of 467+/-3 microV/Oe in the measured frequency range of 200 Hz-8 kHz. The microscope was used to image a 2 mm diameter ring carrying an ac current as low as 10(-5) A. ac fields as small as 3 x 10(-10) T have been detected.