Machine-Learning aided design of novel radiation detection materials

Hierarchical machine-learning framework to efficiently identify candidates for next-generation radiation detectors
Like

Scintillators are materials that emit light when exposed to high-energy radiation such as X-rays and γ-rays (Figure 1). Each photon of radiation that interacts with the scintillator material results in a distinct flash of light, meaning that in addition to being highly sensitive, scintillation detectors are able to capture spectroscopic profiles for the radioactive materials of interest.

Scintillator operation
Figure 1: Scintillator operation. Scintillation starts with the conversion of the high-energy incoming particles to a cascade of excited electrons and holes – more generally called carriers –within the material. This material, by definition, has a band gap – a gap in the electronic structure between occupied and unoccupied states – Eg. As these hot carriers migrate through the material, they may encounter traps that absorb them, while some carriers escape these traps and finally encounter a common recombination center, or activator, at which point they recombine, emitting visible or near-visible photons that are imaged with cameras.

 

Historically, new detector materials have been predominantly developed through resource intensive trial-and-error techniques, typically requiring about 10 years from discovery to deployment. While this approach has had success, it leaves a vast space of potentially revolutionary materials unexplored. Scintillators have diverse applications - ranging from medical imaging (for instance, they are used in CT scanners), to X-ray security, to space applications.

However, each application has different requirements - some may require high resolution while for another application the light yield may be paramount. While scintillator development has seen exponential advances over the years, there are still a limited number of commercially available scintillators today. Hence most applications use off-the-shelf scintillators and make do with whatever is available rather than using a custom designed, and thus more optimized, scintillator tailored to the application.

To design custom scintillators, we need to search the space of possible materials and identify chemically compatible compositions and then isolate candidates that should i) be thermodynamically stable,  ii) experimentally formable,  iii) have a suitable band gap, iv) havefavorable activator states to exhibit scintillation, and finally v) possess the requisite target property. In this work, we focus on steps (i)- (iii).

Figure 2:  Hierarchical down-selection framework for novel scintillator discovery. Starting with more than 5 million potential chemistries, through a series of ML models that each target a different aspect of the scintillator material, we identify about 300 double perovskites that are likely to exhibit a wide band gap. These are then assessed for position of activator states within the gap and other required and desired properties.

 

As a first step, we focus on the class of double oxide perovskites and implement a hierarchical screening process (Figure 2) in which cross-validated and predictive machine learning models for band gap classification and regression, trained using exhaustive datasets that span 68 elements of the periodic table, are applied sequentially. That is, we first classify compounds based on whether they have a small or large band gap and then, for those with suitably large band gaps, we predict the actual value of the band gap. This provides a more robust screening than trying to predict the band gap for both metallic and insulating compounds. We apply these models to investigate a previously identified [1] exhaustive chemical space of formable and stable cubic single and double oxide perovskites, ultimately identifying 310 candidates which are predicted to be experimentally formable, thermodynamically stable, are insulators, and possess a wide band gap. This set of materials may be further investigated using models to predict high-fidelity band gaps [2], defect states [3], and targeted scintillator properties [4].

Our multi-step hierarchical screening approach may be generalized to investigate other classes of materials in addition to the oxide perovskites examined here. The efficiency of these models provides further impetus to the application of physics-based ML models to the discovery of novel functional materials.

For more details, please refer to our paper at: https://www.nature.com/articles/s43246-023-00373-4

LA-UR-23-33741

References:

[1] Talapatra, A., Uberuaga, B. P., Stanek, C. R., & Pilania, G. (2021). A machine learning approach for the prediction of formability and thermodynamic stability of single and double perovskite oxides. Chemistry of Materials33(3), 845-858.

[2] Pilania, G., Gubernatis, J. E., & Lookman, T. (2017). Multi-fidelity machine learning models for accurate bandgap predictions of solids. Computational Materials Science129, 156-163.

[3] Pilania, G., McClellan, K. J., Stanek, C. R., & Uberuaga, B. P. (2018). Physics-informed machine learning for inorganic scintillator discovery. The Journal of chemical physics148(24).

[4] Pilania, G., Liu, X. Y., & Wang, Z. (2019). Data-enabled structure–property mappings for lanthanide-activated inorganic scintillators. Journal of Materials Science54, 8361-8380.

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Subscribe to the Topic

Optical Materials
Physical Sciences > Materials Science > Optical Materials
Computational Materials Science
Physical Sciences > Materials Science > Computational Materials Science
Computational Design Of Materials
Physical Sciences > Chemistry > Theoretical Chemistry > Computational Chemistry > Computational Design Of Materials

Related Collections

With collections, you can get published faster and increase your visibility.

Memristors and non-volatile memory devices

This Collection brings together the latest developments in the realization and optimization of memristive technologies for modern applications that take advantage of neural networks and neuromorphic computing.

Publishing Model: Open Access

Deadline: Apr 01, 2024

Chiral topological matter

This Collection brings together the latest advances in our understanding of quantum materials where the intertwining of chirality and topology gives rise to novel and unexpected phenomenology.

Publishing Model: Open Access

Deadline: Jan 01, 2024