Abstract Accurate theoretical predictions of desired properties of materials play an important role in materials research and development. Machine learning (ML) can accelerate the materials design by building a model from input data. For complex datasets, such as those of crystalline compounds, a vital issue is how to construct low-dimensional representations for input crystal structures with chemical insights. In this work, we introduce an algebraic topology-based method, called atom-specific persistent homology (ASPH), as a unique representation of crystal structures. The ASPH can capture both pairwise and many-body interactions and reveal the topology-property relationship of a group of atoms at various scales. Combined with composition-based attributes, ASPH-based ML model provides a highly accurate prediction of the formation energy calculated by density functional theory (DFT). After training with more than 30,000 different structure types and compositions, our model achieves a mean absolute error of 61 meV/atom in cross-validation, which outperforms previous work such as Voronoi tessellations and Coulomb matrix method using the same ML algorithm and datasets. Our results indicate that the proposed topology-based method provides a powerful computational tool for predicting materials properties compared to previous works.
Predicting synthesizability of crystalline materials via deep learning
Abstract Predicting the synthesizability of hypothetical crystals is challenging because of the wide range of parameters that govern materials synthesis. Yet, exploring the exponentially large space of novel crystals for any future application demands an accurate predictive capability for synthesis likelihood to avoid a haphazard trial-and-error. Typically, benchmarks of synthesizability are defined based on the energy of crystal structures. Here, we take an alternative approach to select features of synthesizability from the latent information embedded in crystalline materials. We represent the atomic structure of crystalline materials by three-dimensional pixel-wise images that are color-coded by their chemical attributes. The image representation of crystals enables the use of a convolutional encoder to learn the features of synthesizability hidden in structural and chemical arrangements of crystalline materials. Based on the presented model, we can accurately classify materials into synthesizable crystals versus crystal anomalies across a broad range of crystal structure types and chemical compositions. We illustrate the usefulness of the model by predicting the synthesizability of hypothetical crystals for battery electrode and thermoelectric applications.
- Publication Date:
- NSF-PAR ID:
- 10323624
- Journal Name:
- Communications Materials
- Volume:
- 2
- Issue:
- 1
- ISSN:
- 2662-4443
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Tuning the properties of atomic crystals in the two-dimensional (2D) limit is synthetically challenging, but critical to unlock their potential in fundamental research and nanotechnology alike. 2D crystals assembled using superatomic blocks could provide a route to encrypt desirable functionality, yet strategies to link the inorganic blocks together in predetermined dimensionality or symmetry are scarce. Here, we describe the synthesis of anisotropic van der Waals crystalline frameworks using the designer superatomic nanocluster Co 3 (py) 3 Co 6 Se 8 L 6 (py = pyridine, L = Ph 2 PN(Tol)), and ditopic linkers. Post-synthetically, the 3D crystals can be mechanically exfoliated into ultrathin flakes (8 to 60 nm), or intercalated with the redox-active guest tetracyanoethylene in a single-crystal-to-single-crystal transformation. Extensive characterization, including by single crystal X-ray diffraction, reveals how intrinsic features of the nanocluster, such as its structure, chirality, redox-activity and magnetic profile, predetermine key properties of the emerging 2D structures. Within the nanosheets, the strict and unusual stereoselectivity of the nanocluster's Co edges for the low symmetry (α,α,β) isomer gives rise to in-plane structural anisotropy, while the helically chiral nanoclusters self-organize into alternating Δ- and Λ-homochiral rows. The nanocluster's high-spin Co edges, and its rich redox profile make themore »
-
Two-dimensional (2D) materials have emerged as promising functional materials with many applications such as semiconductors and photovoltaics because of their unique optoelectronic properties. Although several thousand 2D materials have been screened in existing materials databases, discovering new 2D materials remains challenging. Herein, we propose a deep learning generative model for composition generation combined with a random forest-based 2D materials classifier to discover new hypothetical 2D materials. Furthermore, a template-based element substitution structure prediction approach is developed to predict the crystal structures of a subset of the newly predicted hypothetical formulas, which allows us to confirm their structure stability using DFT calculations. So far, we have discovered 267 489 new potential 2D materials compositions, where 1485 probability scores are more then 0.95. Among them, we have predicted 101 crystal structures and confirmed 92 2D/layered materials by DFT formation energy calculation. Our results show that generative machine learning models provide an effective way to explore the vast chemical design space for new 2D materials discovery.
-
Predictions of the structures of stoichiometric, fractional, or nonstoichiometric hydrates of organic molecular crystals are immensely challenging due to the extensive search space of different water contents, host molecular placements throughout the crystal, and internal molecular conformations. However, the dry frameworks of these hydrates, especially for nonstoichiometric or isostructural dehydrates, can often be predicted from a standard anhydrous crystal structure prediction (CSP) protocol. Inspired by developments in the field of drug binding, we introduce an efficient data-driven and topologically aware approach for predicting organic molecular crystal hydrate structures through a mapping of water positions within the crystal structure. The method does not require a priori specification of water content and can, therefore, predict stoichiometric, fractional, and nonstoichiometric hydrate structures. This approach, which we term a mapping approach for crystal hydrates (MACH), establishes a set of rules for systematic determination of favorable positions for water insertion within predicted or experimental crystal structures based on considerations of the chemical features of local environments and void regions. The proposed approach is tested on hydrates of three pharmaceutically relevant compounds that exhibit diverse crystal packing motifs and void environments characteristic of hydrate structures. Overall, we show that our mapping approach introduces an advance inmore »
-
ZrS2, ZrSe2 and mixed alloy ZrSxSe2−x materials were achieved through chemical vapor transport. The incongruent melting system of Zr-S-Se formed crystalline layered flakes as a transport product that grew up to 2 cm in lateral size with cm-scale flakes consistently obtained for the entire compositional range exhibiting visible hexagonal features. Bulk flakes of the series ZrSxSe2−x (x=0, 0.15, 0.3, 0.6, 1.05, 1.14, 1.51, 1.8 and 2) were analyzed through Raman spectroscopy revealing significant convolution of primary bonding modes and shifting of Raman features as a function of increasing sulfur composition. Additionally, activation of new modes not present in the pure compounds are observed as effects which result from disorder introduced into the crystal due to the random mixing of S-Se in the alloying process. Further structural characterization was performed via x-ray diffraction (XRD) on the layered flakes to evaluate the progression of layer spacing function of alloy composition which was found to range between 6.24 Å for ZrSe2 and 5.85 Å for ZrS2. Estimation of the compositional ratios of the alloy flakes through energy dispersive spectroscopy (EDS) large-area mapping verified the relation of the targeted source stoichiometry represented in the layered flakes. Atomic-resolution high angle annular dark field (HAADF)-scanning transmissionmore »
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Publisher's Version of Record
- https://doi.org/10.1038/s43246-021-00219-x
