A New X-Ray Score to Distinguish Success and Failure in Materials Synthesis

Researchers in the UC Davis Department of Physics and Astronomy have
developed a simple way to tell, quickly and quantitatively, whether a theoretically
predicted material is actually present in high-throughput powder X-ray diffraction
(PXRD) data, or if it likely never formed at all.

In a study published in Acta Materialia, Kyoka Nagashima, Peter Klavins and
Professor Valentin Taufour introduce a numerical “K-factor” that compares
experimental PXRD (X-ray diffraction) patterns with theoretically predicted ones. The new metric
combines how many diffraction peaks line up in angle with how well their
intensities match, turning a complicated pattern-matching problem into a single
score between 0 and 1. High K-values indicate strong evidence that the predicted
phase is present; low values suggest it is absent, even if other crystalline phases
are in the sample.

The team tested their approach on an intensively studied family of materials known as half-antiperovskites, with composition M₃Z₂X₂ (M = transition metal, Z = chalcogen, X =
main-group element). These compounds host stacked kagome lattices, triangular-based networks that can give rise to exotic electronic and magnetic behavior such as Dirac
and Weyl fermions. Recent high-throughput calculations predicted dozens of new half-antiperovskites to be thermodynamically stable, many of them magnetic.

Using a solid-state synthesis route, Nagashima and colleagues systematically attempted to make 41 of these predicted compounds. As a benchmark, they first synthesized
seven half-antiperovskites that are already known, including Co₃Sn₂S₂ and Ni₃In₂S₂, and showed that their PXRD patterns give K-factors above 0.92, essentially a perfect
match between experiment and theory within reasonable uncertainty in lattice parameters.

They then applied the same procedure to the remaining predictions. Despite being flagged as stable by theory, none of the new candidates reached similarly high K-values; all
fell below 0.69, clearly separated from the cluster of previously reported compounds. The result suggests that, at least under the tested synthesis conditions, the predicted
half-antiperovskites do not crystallize in the proposed structure.

A key strength of the method is that it also works when the desired phase is only a minority component. By mixing a known compound, Co₃Sn₂S₂, with its unreacted
precursors, the team showed that the K-factor still reliably identifies the correct phase even when it makes up as little as 10 percent of the sample by weight. The authors also
explored how noise and impurity phases can artificially inflate the score and outlined practical checks to avoid false positives.

Because the K-factor does not rely on large structural databases or machine-learning training, it is well suited for integrating with autonomous “self-driving” laboratories that
are starting to generate huge volumes of PXRD data. The metric gives experimenters a fast, transparent and quantitative way to report both successful and unsuccessful
synthesis attempts, crucial information for refining theoretical predictions and for teaching automated platforms how to search more intelligently.

The work highlights an often overlooked aspect of materials discovery: learning from the compounds that don’t appear. By turning negative results into a structured,
quantitative dataset, the UC Davis team’s method provides an important bridge between high-throughput computational predictions and real-world materials chemistry.

reference: K. Nagashima et al. “A quantitative criterion for evidencing predicted compounds in high-throughput powder X-ray diffraction data: Illustration with half-antiperovskite” Acta
Materialia 303 121525 (2026)