Revealing Hidden Patterns through Chemical Intuition and Interpretable Machine Learning: A Case Study of Binary Rare-Earth Intermetallics <i>RX</i>

Gvozdetskyi, Volodymyr; Selvaratnam, Balaranjan; Oliynyk, Anton O.; Mar, Arthur

doi:10.1021/acs.chemmater.2c02425

Cited by 3 publications

(5 citation statements)

References 76 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We note that although the primary features are physically motivated, the models generated by SISSO by combining primary features do not necessarily have a physical interpretation. An important advantage of SISSO is that it can work well with a relatively small amount of data. − This is critical in this case because of the high computational cost of acquiring GW+BSE training data for molecular crystals. To train SISSO, we generated a data set containing GW+BSE SF driving force values of 101 PAH molecular crystals with up to ∼500 atoms in the unit cell (the PAH101 set).…”

Section: Resultsmentioning

confidence: 99%

Computational Discovery of Intermolecular Singlet Fission Materials Using Many-Body Perturbation Theory

Wang,

Gao,

Luo

et al. 2024

J. Phys. Chem. C

View full text Add to dashboard Cite

Intermolecular singlet fission (SF) is the conversion of a photogenerated singlet exciton into two triplet excitons residing on different molecules. SF has the potential to enhance the conversion efficiency of solar cells by harvesting two charge carriers from one high-energy photon, whose surplus energy would otherwise be lost to heat. The development of commercial SFaugmented modules is hindered by the limited selection of molecular crystals that exhibit intermolecular SF in the solid state. Computational exploration may accelerate the discovery of new SF materials. The GW approximation and Bethe−Salpeter equation (GW+BSE) within the framework of many-body perturbation theory is the current state-of-the-art method for calculating the excited-state properties of molecular crystals with periodic boundary conditions. In this Review, we discuss the usage of GW+BSE to assess candidate SF materials as well as its combination with lowcost physical or machine learned models in materials discovery workflows. We demonstrate three successful strategies for the discovery of new SF materials: (i) functionalization of known materials to tune their properties, (ii) finding potential polymorphs with improved crystal packing, and (iii) exploring new classes of materials. In addition, three new candidate SF materials are proposed here, which have not been published previously.

show abstract

Section: Resultsmentioning

confidence: 99%

Computational Discovery of Intermolecular Singlet Fission Materials Using Many-Body Perturbation Theory

Wang,

Gao,

Luo

et al. 2024

J. Phys. Chem. C

View full text Add to dashboard Cite

show abstract

“…In total, there were 23 features: three to express the composition ( x , y , and z ), 10 for the elemental properties of M, and 10 for the elemental properties of X. These features are similar to those that were successfully used to tackle a related problem in the structural classification of binary rare-earth intermetallics . It has been hypothesized that many properties of an element can be derived merely from its position in the periodic table .…”

Section: Resultsmentioning

confidence: 99%

Discovery of Ternary Antimonides A–Al–Sb (A = Rb or Cs) with Desired Structural Motifs Guided by Machine Learning

Owens-Baird,

Gvozdetskyi,

Sarkar

et al. 2024

Chem. Mater.

Self Cite

View full text Add to dashboard Cite

Specific structural motifs in inorganic solids are often related to their targeted physical properties. For many classes of solids, such as Zintl phases and polar intermetallics, the crystal structures are diverse and not easy to predict. Various antimonides that are potential thermoelectric materials were proposed to be synthesizable on the basis of their estimated formation energies. Their structures were broadly classified as clathrate, channel, layered, or network through a machine learning model trained on existing ternary phases and features based on elemental properties using the sure independence screening and sparsifying operator algorithm. Through experimental validation, three new ternary antimonides were synthesized and confirmed to form layered structures: tetragonal RbAlSb 2 and CsAlSb 2 , which are isopointal but not isotypic to LiBSi 2 ; and monoclinic Rb 2 Al 2 Sb 3 , which adopts the Na 2 Al 2 Sb 3 -type structure. Reinvestigation of the related compound Cs 2 In 2 Sb 3 revealed a low thermal conductivity and p-type semiconducting behavior.

show abstract

“…These compounds were examined because the distribution of the three structure types adopted (44 FeB, 33 TlI, and 29 CsCl) appears to be puzzlingly haphazard. Similar to how structure maps for other classes of compounds were historically developed through a combination of trial-and-error and chemical intuition, the elements R and X were rearranged in different sequences until a reasonable separation of these structure types was achieved . A good descriptor for the x axis of the RX structure map was found to be a simple combination of R-element features, x = Z r m , where Z is the atomic number and r m is the metallic radius.…”

Section: Resultsmentioning

confidence: 99%

“…For the main test cases, to compare the two approaches (DO and DT) for descriptor scoring, three data sets were examined. Crystal structure data were gathered for ABX 3 compounds (perovskites vs nonperovskites) from previous literature, for oxides AB 2 O 4 (spinels vs nonspinels) from Pearson’s Crystal Data , and for rare-earth intermetallics RX (CsCl, TlI, and FeB structure types) from our own recent work . The number of samples in these data sets, a schematic of the workflow, and the classification accuracies for the test sets are summarized in Figure .…”

Section: Methodsmentioning

confidence: 99%

“…Similar to how structure maps for other classes of compounds were historically developed through a combination of trial-and-error and chemical intuition, the elements R and X were rearranged in different sequences until a reasonable separation of these structure types was achieved. 29 A good descriptor for the x axis of the RX structure map was found to be a simple combination of R-element features, x = Zr m , where Z is the atomic number and r m is the metallic radius. However, the functional form of the descriptor for the y axis was not obvious.…”

Section: Rare-earth Intermetallics Rxmentioning

confidence: 99%

See 1 more Smart Citation

Interpretable Machine Learning in Solid-State Chemistry, with Applications to Perovskites, Spinels, and Rare-Earth Intermetallics: Finding Descriptors Using Decision Trees

2023

Self Cite

View full text Add to dashboard Cite

Machine-learning methods have exciting potential to aid materials discovery, but their wider adoption can be hindered by the opaqueness of many models. Even if these models are accurate, the inability to understand the basis for the predictions breeds skepticism. Thus, it is imperative to develop machine-learning models that are explainable and interpretable so that researchers can judge for themselves if the predictions are consistent with their own scientific understanding and chemical insight. In this spirit, the sure independence screening and sparsifying operator (SISSO) method was recently proposed as an effective way to identify the simplest combination of chemical descriptors needed to solve classification and regression problems in materials science. This approach uses domain overlap (DO) as the criterion to find the most informative descriptors in classification problems, but sometimes a low score can be assigned to useful descriptors when there are outliers or when samples belonging to a class are clustered in different regions of the feature space. Here, we present a hypothesis that the performance can be improved by implementing decision trees (DT) instead of DO as the scoring function to find the best descriptors. This modified approach was tested on three important structural classification problems in solid-state chemistry: perovskites, spinels, and rare-earth intermetallics. In all cases, the DT scoring gave better features and significantly improved accuracies of ≥0.91 for the training sets and ≥0.86 for the test sets.

show abstract

Revealing Hidden Patterns through Chemical Intuition and Interpretable Machine Learning: A Case Study of Binary Rare-Earth Intermetallics RX

Cited by 3 publications

References 76 publications

Computational Discovery of Intermolecular Singlet Fission Materials Using Many-Body Perturbation Theory

Computational Discovery of Intermolecular Singlet Fission Materials Using Many-Body Perturbation Theory

Discovery of Ternary Antimonides A–Al–Sb (A = Rb or Cs) with Desired Structural Motifs Guided by Machine Learning

Interpretable Machine Learning in Solid-State Chemistry, with Applications to Perovskites, Spinels, and Rare-Earth Intermetallics: Finding Descriptors Using Decision Trees

Contact Info

Product

Resources

About