Session H16: Computational Materials Design and Discovery -- Structure Prediction and Optimization

Solving Unknown Crystal Structures in a High-Throughput Manner

Databases of density functional theory (DFT) calculations, such as the Open Quantum Materials Database (OQMD), have paved the way for accelerated materials discovery. DFT requires a crystal structure as input; however, due to inherent challenges in solving a compound’s structure from powder diffraction data alone, there are thousands of known compounds whose structures remain unsolved. Although structure solution methods involving DFT-based algorithmic optimization have been demonstrated, their computational cost limits their applicability. We present a more rapid DFT-based structure solution method in which we search the OQMD for structure types that match the cell geometry and stoichiometry of the compound, and evaluate them as possible solutions using DFT and match to diffraction pattern. As this approach is straightforward and inexpensive, we employed it in a high-throughput manner to solve hundreds of previously-undetermined structures from the Powder Diffraction File, including many new elpasolites, transition metal oxides, and mixed anion structures, as well as wide range of scarcely explored structure types.   

Determining nanoparticle structures using FANTASTX

Determining the atomistic structures of nanoparticles is a fundamental problem. Their structures determine their functionality, and therefore their effectiveness in applications. Although there are both experimental and computational methods to determine these nanoscale structures, they both possess limitations. We develop FANTASTX (Fully Automated Nanoscale To Atomistic Structure from Theory and eXperiment) to overcome the limitations of either by combining both experimental and computational data. We demonstrate the effectiveness of FANTASTX by determining the structures of Au and IrO_x nanoparticles from x-ray pair distribution function (PDF) data and density functional theory (DFT) calculations, using multi-objective optimization and a variety of canonical and grand canonical sampling algorithms.   

Evolutionary Niching in the GAtor Genetic Algorithm for Molecular Crystal Structure Prediction

The goal of molecular crystal structure prediction is to find all plausible polymorphs for a given molecule. This requires performing global optimization over a high dimensional search space. Genetic algorithms (GAs) perform global optimization by mimicking evolution. New structures are generated by breeding the fittest structures in the population. Typically, the fitness function is based on relative lattice energies, such that structures with lower energies have a higher probability of being selected for mating. GAs may be adapted to perform multi-modal optimization by using evolutionary niching methods that support the formation of several stable subpopulations and suppress the over-sampling of densely populated regions. Evolutionary niching is implemented in the GAtor molecular crystal structure prediction code by using machine learning to dynamically cluster the population by structural similarity. A cluster-based fitness function is constructed such that structures in less populated clusters have a higher probability of being selected for breeding. Using evolutionary niching increases the success rate of generating the experimental structure of 1,3-dibromo-2-chloro-5-fluorobenzene and additional low-energy structures with similar packing motifs.   

First-principles study on electrode – solid electrolyte interfaces in solid-state battery via efficient structure prediction method

Understanding the origin of high interfacial resistances between electrodes and solid-electrolytes is crucial for commercial application of solid-state batteries (SSBs). [Chem. Mater. 26, 4248-4255] [ACS Appl. Mater. Interfaces 9, 286-292]. Here we utilized the CALYPSO structure prediction method [Physical Review B 82, 094116] to investigate the interface structures between LiCoO₂ cathode and β-Li₃PS₄ sulfide electrolyte. About 20000 configurations are sampled to search the stable interface structures. The results show that the interfacial reaction layer is formed accompanied by the cation mixing and anion mixing. It is confirmed that the preferential Li depletion can occur at interface upon charging, even in very distorted region. Furthermore, CALYPSO interface structure prediction method can be applied to the solid-solid interface systems beyond the SSBs.   

An efficient algorithm for novel two-dimensional crystal structure prediction

Two-dimensional (2D) materials are promising for their intriguing standalone properties and a number of combinatorial heterostructures. In this regard, crystal structure prediction (CSP) can enhance both material and property spaces significantly, accelerating innovative materials discovery. However, conventional approaches based on global optimization may be inefficient for 2D CSP due to enormously enhanced search space. Here, we will discuss an efficient algorithm for predicting novel 2D materials based on spatial symmetry of the atomic arrangements. We show the method was used to predict a number of novel 2D silicon crystals [1] and group IV and group VI compounds (namely, TXene) [2]. In addition, we will show that the method is efficient and transferable, and can be further applied to propose novel 2D materials.   

From Pentagonal Geometries to Two-Dimensional Materials

Most of the most popular two-dimensional (2D) materials, such as semiconducting MoS₂ and magnetic CrI₃, adopt hexagonal structures. Inspired by the geometries of the existing 15 types of convex pentagons that can tessellate a plane without creating a gap or overlap, we combined these pentagonal geometries and density functional theory (DFT) calculations to predict novel 2D materials. We showed that this combination leads to a new direction in the field of 2D materials. In particular, we discovered a hidden pattern of pentagons called the Cairo tessellation in a group of bulk materials with the pyrite structure. We predicted single-layer PtP₂ to exhibit a completely planar, pentagonal structure and a direct band gap. Our work shows that encoding quantum mechanics into pentagonal geometries and with the help of DFT calculations open up a novel route for accelerating discovery of new 2D materials.   

Dominant in-plane cleavage direction of CrPS4

In-plane cleavage directions of two dimensional (2D) crystals are displayed and often well-defined in their flakes exfoliated by the mechanical exfoliation method. Here, we investigate the correlation between dominant in-plane cleavage direction and elastic properties in different directions. CrPS₄ flakes show a preferential in-plane cleavage direction of 67.5°, corresponding to <110> direction. To explain it, we calculated the directional dependence of Young’s modulus and fracture energy using first-principles density functional theory calculations. We found that fracture energy is directly relevant to the in-plane cleavage direction of CrPS₄. Our study can provide a facile approach to figure out the direction of 2D crystals without complex characterization process, which is valuable for material processing of 2D materials.   

Conformations, electronic, and mechanical properties of monolayer diamene

Experiments show that diamond-like (diamene) films can be obtained from multilayer graphene, either by compression or surface passivation. Based on DFT calculations, we predict that diamene assumes various conformations, requiring either one or both surfaces being passivated. Conformations with surfaces fully passivated exhibit insulating properties, whereas diamene films presenting a clean surface show variable electronic properties, from insulating to semiconducting and metallic. Interestingly, regardless being monolayers or having a clean surface, all conformers of diamene exhibit elastic moduli comparable to those of the corresponding sp³ bulk phase, retaining Young's moduli along their transverse direction that are between 50% and 80% of bulk phases. Moreover we suggest that diamene film can be obtained from multilayer graphene by means of compression and/or wear without the need of chemical passivants, and the energies and pressures involved in the transformation depend critically on the structure and chemistry of the interface with the substrate.   

Inverse Design for Self-assembly of Materials with Targeted Mechanical Properties

Inverse design is a promising yet challenging approach to develop new materials. To create materials with new properties, several inverse approaches have been proposed to design building blocks for target structures. However, the relationship between a material's structure and its properties is often unknown for novel materials. This calls for the development of inverse approaches that can directly target materials properties without having to target structures as the intermediary. Here, we present an inverse-design approach for particles with interacting potentials to self-assemble crystal structures with targeted mechanical properties. We do so through a novel molecular dynamics implementation of the ‘digital alchemy’ inverse-design approach. We give examples in which model particles interact via isotropic pair potentials that are designed to yield structures with a desired bulk modulus. Our results demonstrate that we can directly target mechanical properties via inverse materials design, and that our algorithm can be generalized to other properties.   

Effect of alloying concentration on the mechanical properties of B2Cr1-xMox

Transition metal borides as Mo₂FeB₂, Mo₂NiB₂, and WCoB have recently caught attention because they exhibit excellent mechanical properties for wear resistant applications. On the other hand, it is widely known that the addition of Cr and/or Mo as alloying elements improves the mechanical properties and corrosion resistance of some alloys. In this investigation, first principles calculations were performed to study the structural properties of the ternary B₂Cr_1-xMo_x within DFT formalism. The B₂Cr_1-xMo_x compounds were studied for a concentration x of 25%, 50%, and 75%. The Minima Hopping method was used to search over the stable and metastable structures for each concentration. B₂Cr_1-xMo_x structures were optimized to obtain the lattice parameters, total energies and formation enthalpies in each case. The elastic constants were calculated to verify the mechanical stability of the compounds, as well as the bulk modulus, shear modulus, Young’s modulus, and Poisson’s Ratio. The mechanical strength of the ternary was maximized as a function of the alloying concentration. Free energies were also calculated to study the thermal stability of the found crystal phases.   

Amorphous materials modeling and classification for low mechanical loss mirror coatings using machine learning methods

Instead of the well-defined atomic structures of crystals, amorphous materials are more complicated due to intrinsic randomness. Modeling and predicting the properties of amorphous materials (amorphous Ta₂O₅, doped Ta₂O₅) are important to understand experimental results and to find lower mechanical loss mirror coatings to reduce thermal noise in the next generation of LIGO laser interferometer gravitational wave detectors. In our work, thousands of atomic models of amorphous materials are generated using reverse Monte Carlo (RMC) and molecular dynamics (MD) simulations based on experimental data. Classifying them into different groups according to their properties and features with the help of machine learning, enables us to understand the differences between these models and use the information from these structures to find the best materials for low mechanical loss mirror coating.   

Discovery of Novel Layered Heteroanionic Materials from Pauling’s Rules

Heteroanionic materials, such as oxyfluorides and oxynitrides, display a myriad of functional properties derived from the presence of two anions of difference size, charge, and electronegativity. To aid synthetic efforts focused on compound discovery, we propose a workflow that can be used to predict new and stable n = 1 Ruddlesden-Poppper type heteroanionic materials using principles from inorganic chemistry and supported with density functional theory calculations. We show that a simple structural optimization scheme based on Pauling’s rules is an effective method to evaluate the stability of novel heteroanionic materials and predict new stable oxyfluorides. Last, we describe some of the properties exhibited by the newly identified compounds.   

Pressure-induced dimerization of the hyperkagome framework in Na3Ir3O8

The ambient-pressure cubic Na₃Ir₃O₈ phase has been observed to undergo a symmetry-breaking transformation around 10 GPa [1]. Structure determination proved to be a challenge due to the large system size and the low symmetry of the new ground state. We performed extensive ab initio evolutionary searches [2] without any input from the high-pressure experiment and identified a complex monoclinic phase with 56 atoms per unit cell that agreed well with the collected XRD patterns. According to our ab initio calculations, the monoclinic phase features a dimerized Ir hyperkagome framework and a lower bulk modulus compared to that of the starting cubic phase. Study of the compound's electronic structure revealed significant convergence problems in the DFT+U approach.

[1] F. Sun, H. Zheng, Y. Liu, E. D. Sandoval, C. Xu, J. Xu, C. Q. Jin, C. J. Sun, W. G. Yang, H. K. Mao, J. F. Mitchell, A. N. Kolmogorov, and D. Haskel, Phys. Rev. B 98, 085131 (2018)
[2] https://github.com/maise-guide/maise   

Numerical Studies of Thermal Conductivity in Functionalized Carbon Nanotubes

Although carbon nanotubes (CNTs) possess a large thermal conductivity, when they are incorporated in a polymer matrix their severe boundary resistance makes them ineffective at improving the thermal conductivity of the resulting composite. This resistance at CNT interfaces can be altered via chemical functionalization using mixed molecular chains to match the thermal impedance between CNTs and their environment. We explore the vast chemical space of possible configurations through meta-heuristics such as genetic algorithms and present candidate structures with optimal thermal conductance. We make comparisons of these CNT systems to certain harmonic lattices and look for important correlations among molecule parameters that contribute to the thermal conductivity, as is done in machine learning. We will summarize these results in general design rules for improving the thermal conductivity across molecular interfaces.