Bulletin of the American Physical Society
APS March Meeting 2020
Volume 65, Number 1
Monday–Friday, March 2–6, 2020; Denver, Colorado
Session W43: Computational Design and Discovery of Novel Materials VI: Machine Learning and High Throughput ComputingFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP DMP Chair: Yuanxi Wang, Pennsylvania State University Room: 702 |
Friday, March 6, 2020 8:00AM - 8:36AM |
W43.00001: New tools for detecting strong correlation in automated transition metal complex screening Invited Speaker: Fang Liu There are several prominent challenges in conducting quantum chemistry studies of transition metal complexes: generation of molecule structure, selection of electronic structure method, workflow control, and post-processing of simulation results. molSimplify is a user-friendly open-source toolkit that enables the seamless generation of candidate inorganic molecule structures, preparation, and execution of Density Functional Theory (DFT) calculations, and post-processing DFT results. Here we focus on extending the capability of molSimplify, enabling automated selection of electronic structure methods including both DFT and wavefunction based methods and automated control of simulations. For automatic method selection, we develop a python module, MultirefPredict, which is a high-level cross-platform workflow that calculates widely used multi-reference diagnostics for a given molecular system, without users handling the input and output of quantum chemistry packages. The backend of MutlrefPredict supports several quantum chemistry packages based on both DFT and wavefunction theory methods, with calculation performed on both CPU and GPU. MultirefPredict guides a user to choose between single-reference and multi-reference based methods, which is essential for obtaining accurate results for transition metal-containing systems. For in-situ simulation control, we developed an automated workflow that checks the current status of geometry optimization using a “dynamic classifier” that predicts the probability of failure based on geometric and electronic structure information collected on the fly. This workflow terminates the simulation when it is predicted to fail, saving numerous computational resources. These open-source tools, which are accessible to the whole transition metal chemistry community, are anticipated to make transition metal chemistry simulation more automated and transparent. |
Friday, March 6, 2020 8:36AM - 8:48AM |
W43.00002: High-throughput Design of Lead-free Organic-inorganic Lead Halide Semiconductors Beyond Perovskites Kesong Yang, Yuheng Li Organic-inorganic lead halide perovskites (including 2D hybrid perovskites) show great promise in optoelectronic applications such as light-emitting diodes and solar energy conversion. However, the poor stability and toxicity of lead halide perovskites severely limit their large-scale applications. In this talk, I am going to present our high-throughput design strategy of lead- free hybrid halide semiconductors with robust materials stability and desired material properties beyond perovskites. On the basis of 24 prototype structures that include perovskite and non-perovskite structures and several typical organic cations, a comprehensive quantum materials repository that contains 4507 hypothetical hybrid compounds was built using large-scale first-principles calculations. After a high-throughput screening of this repository, we have rapidly identified 23 candidates for light-emitting diodes and 13 candidates for solar energy conversion. |
Friday, March 6, 2020 8:48AM - 9:00AM |
W43.00003: Perovskite Genomics: Optimizing the performance of large sets of perovskite materials with atomistic simulations Gianaurelio Cuniberti, Hagen Eckert, Florian Pump, Josua Vieten, Martin Roeb, Christian Sattler Perovskites are ideal candidates to be applied in two-step redox cycles to convert, store and utilize energy from concentrated solar radiation by heating the perovskite materials to high temperatures (up to 1,500 °C) and thus transferring them to an energy-rich state. In a second step at lower temperatures, this energy can be used for a variety of chemical processes. As the overall performance of such redox materials is dominated by the diffusion rate of oxide ions through the constituting lattice, tuning the redox thermodynamics of such materials through composition adjustment allows the design of ideal perovskite materials. For an enhanced insight into the diffusion properties of a large set of perovskite materials, Molecular Dynamics simulations are applied for an efficient funneling of up to 58,000 candidate oxides to a few hundred for subsequent intensive studies. A major challenge lies in the generalization of the force field generation for datasets including large numbers of different oxides involving the characteristic electrostatic properties to be known for each material. To this end, we use data obtained by DFT stored in the database of the Materials Project. In this way, we are able to suggest materials with optimized and tailored dynamic properties. |
Friday, March 6, 2020 9:00AM - 9:12AM |
W43.00004: Computational discovery of semiconducting high-entropy chalcogenide alloys Zihao Deng, Logan Williams, Guangsha Shi, Emmanouil Kioupakis High-entropy materials are formed by mixing typically five or more principal components into a single crystal structure. While significant progress has been made to synthesize entropy-stabilized metals and ceramics for structural applications, little attention has been paid to the discovery of new entropy-stabilized semiconductors. Here, we present a new class of entropy-stabilized semiconducting alloys based on the IV-VI binary chalcogenides, namely GexSnyPb1–x–ySzSetTe1–z–t high-entropy chalcogenides (HECs). By utilizing high-throughput first-principles calculations, we investigate the thermodynamic stability of HECs over their entire composition space, and show that more than 50% of the investigated compositions are stable with respect to phase segregation. We further studied the enthalpic effect of the individual elements via machine learning on the high-throughput data. Our work demonstrates the potential of entropy stabilization in the discovery of novel multicomponent semiconductor alloys. |
Friday, March 6, 2020 9:12AM - 9:24AM |
W43.00005: High-throughput discovery of metal-organic frameworks for cooperative CO2 adsorption Eric Taw, Jeffrey R Long, Jeffrey B Neaton, Maciej Haranczyk Recently, a new class of post-synthetically modified metal-organic frameworks (MOFs) with non-Langmiur stepped isotherms have been discovered and tuned to reversibly and selectively adsorb CO2 under common flue gas conditions. However, very few MOFs are known to exhibit step-like isotherms, a result of a cooperative adsorption phenomenon. Here, we present a computational screening procedure to discover new CO2 adsorbent MOFs with the potential for step-like isotherms and cooperative adsorption. Our workflow is based on the hypothesis that the distance between accessible, undercoordinated metal sites is a key indicator for whether a MOF modified with ethylenediamine will exhibit cooperative adsorption. We screen the Computational-Ready Experimental MOF (CoRE-MOF) database using the fast marching algorithm to assess metal site distances given arbitrary pore geometries, and discuss candidate materials for experimental validation. |
Friday, March 6, 2020 9:24AM - 9:36AM |
W43.00006: Machine Learning Accelerated Discovery of Mixed Anion Materials Jiahong Shen, Cheol Woo Park, Jiangang He, Christopher Mark Wolverton Mixed-anion materials are interesting counterparts to their more widely studied single-anion compounds due to the increased flexibility and tunability of properties afforded by the presence of multiple anions. Here, we demonstrate how computational approaches, based on DFT datasets can be combined with materials informatics and machine learning (ML) models to accelerate materials discovery. We utilize and compare a variety of materials representations, including a recently proposed improved crystal graph convolutional neural network (iCGCNN) model, and the Voronoi tessellation approach incorporated in the Materials-Agnostic Platform for Informatics and Exploration (MAGPIE). ML models are trained on a set of 450,000 DFT data prior calculated in the Open Quantum Materials Database (OQMD) and evaluated on a separate test set of 3,000 mixed-anion compounds where iCGCNN outperforms random forest models in predictive accuracy by ~300%. By introducing more mixed-anion compounds into the training set, the performance of the iCGCNN model is further improved, and it allows us to make predictions of a large number of stable (and hence, likely synthesizeable) new ternary oxychalcogenides, which are subsequently validated by DFT calculations. |
Friday, March 6, 2020 9:36AM - 9:48AM |
W43.00007: High-throughput Computational Study of Double Perovskite Oxides Jiangang He, Christopher Mark Wolverton Perovskite oxides have been intensively studied for decades due to the extraordinary variability of compositions and structures and their attractive applications in superconductivity, magnetoresistance, multiferroicity, catalysis, solid oxide fuel cells, and etc. Extending from single perovskite $AB$O$_3$ to double perovskite $A_2BB'$O$_6$ significantly increases the tunability towards the targeted physical and chemical properties. However, the number of possible compositions of double perovskite is prohibitively large to explore entirely experimentally. In this talk, we will present how to use a multi-step high-throughput computational method to screen $\sim$ 5000 compositions of double perovskite $A_2BB'$O$_6$ ($A$=Ca, Sr, Ba, and La; $B$ and $B'$ are metal elements). With the $\sim$ 2000 stable/metastable (and so likely synthesizable) compounds predicted by our calculations, we will show how statistical learning of the large dataset can capture the correlation among composition, stability, and crystal structure (i.e., cubic perovskite, distorted perovskite, and non-perovskite). The results will accelerate new double perovskite oxides prediction and discovery. |
Friday, March 6, 2020 9:48AM - 10:00AM |
W43.00008: Automated phase mapping of high throughput X-ray diffraction data Yizhou Zhu, Christopher Mark Wolverton Combinatorial synthesis and high-throughput characterization have become powerful tools to accelerate the discovery and design of novel materials. However, one key question in the high-throughput workflow is the phase mapping problem. Correct, automatic identification of the number, identity, and fraction of phases in XRD data is a crucial step to perform autonomous high-throughput characterization and establish further understanding on the composition-structure-property relationships. Traditional analysis performed manually could take days for a single material system, while an automatic solver, if lacks the domain-specific knowledge, is likely to give unphysical results. In this work, we demonstrate how we use phase diagrams generated based on DFT calculations databases, such as the Open Quantum Materials Database (OQMD), to provide domain-specific knowledge and enforce reasonable constraints. We show how DFT provides a good initial guess to phase mapping algorithms, we can address the “peak-shifting” caused by alloying behavior using machine learning techniques, and the method is capable of uncovering minor phases present in the XRD data. By combining first-principles calculations and machine learning techniques, our approach enables rapid phase identification and mapping. |
Friday, March 6, 2020 10:00AM - 10:12AM |
W43.00009: High-throughput Exploration of Ternary Electrides John Lasseter, Mina Yoon A class of materials that has gained community attention recently is electrides. These rare materials stabilize “free” electrons, loosely bound to a low-dimensional cavity space, which act as anions to chemically stabilize the structure. This characteristic feature makes them highly desirable for applications such as battery anodes, superconductors, and topological electronics. However, very few electrides have been synthesized due to their chemical instability; they often undergo degradation in ambient conditions. Thus, a challenge in this field is to discover electrides with higher stability. In this work ternary electrides of the chemical species A-B-C were targeted; the choice of species and the stoichiometry between them allows a vast space within which to engineer compounds with enhanced stabilities and desirable properties. Properties of interest include magnetic, electronic carrier density, and band topology. A global structure search was performed with first-principles calculations to identify stable and metastable structures for a given composition. Highly parallelized GPU-accelerated DFT codes were used to explore the vast parameter space. Our database will guide experimental synthesis efforts of new ternary electrides with multidimensional functional properties. |
Friday, March 6, 2020 10:12AM - 10:24AM |
W43.00010: High-throughput screening and classification of layered di-metal chalcogenides Jinchen Wei, Chao Wang, Tao Zhang, Chenmin Dai, Shiyou Chen Through employing the layered-crystal determination program based on the topology-scaling algorithm, 450 MmNnXx (M, N = metal elements, X = S, Se) layered di-metal chalcogenides (LDCs) are identified from the 1602011 crystalline materials known in two Material Genome databases (Materials Project and OQMD). Their structures are classified into three types, standard structures, hetero-layered structures and large-cation intercalated layered structures. 312 cation-intercalated LDCs are hard to be exfoliated into few-layer 2D structures. In contrast, the other two types of structures may be exfoliated into stable few-layer 2D structures due to the weak inter-layer van der Waals interaction. The band structure screening identifies 24 semiconductors with small in-plane effective masses and thus high mobility of carriers. Furthermore, 83 found LDCs composed of the magnetic metal elements provide a new platform for search of 2D magnetic crystals. This work extends the search of layered materials from metal dichalcogenides to ternary chalcogenides and can serve as a map for the future discovery of novel 2D semiconductors and magnetic materials. |
Friday, March 6, 2020 10:24AM - 10:36AM |
W43.00011: Effect of a Parameter in a Descriptor on the Efficiency of a Crystal Structure Search Using Bayesian Optimization Nobuya Sato, Tomoki Yamashita, Tamio Oguchi, Koji Hukushima, Takashi Miyake A crystal structure search method using Bayesian optimization (BO) has been developed recently. BO is a machine learning technique for the global optimization. It has been reported that the method can search a stable structure efficiently. In this method, a crystal structure is represented by a numerical vector referred to as a descriptor. Descriptors often have a parameter which should be predetermined. We reveal by case studies for crystalline silicon, silicon oxide, and yttrium–cobalt alloy that the efficiency of a crystal structure search depends heavily on a parameter value. Our analysis indicates that the efficiency is related to the distribution of the descriptor. Therefore, we introduce an information measure based on the descriptor distribution, which is evaluated only from a set of crystal structures. The measure succeeds in estimating a parameter value where the crystal structure search works efficiently, and thus, can be used to predetermine the value of a parameter. |
Friday, March 6, 2020 10:36AM - 10:48AM |
W43.00012: An active-learning framework for the discovery of new crystalline materials Kirsten Winther, Raul Flores, Christopher Paolucci, Ankit Jain, Michal Bajdich, Thomas bligaard A challenging problem in materials science is that the space of possible crystal structures is so vast that it is impossible to sample with brute-force screening approaches. However, electronic structure methods combined with machine-learning (ML) techniques have a huge potential to speed up the search [1][2], which hold a great promise for the discovery of novel materials and/or catalysts for energy applications. |
Friday, March 6, 2020 10:48AM - 11:00AM |
W43.00013: Unraveling nanoscale features controlling the diffusion of multi-component alloys through machine learning methods S. Mohadeseh Taheri-Mousavi, S. Sina Moeini-Ardakani, Ryan W. Penny, Ju Li, John Hart The immense compositional breadth of non-dilute multi-component alloys has made their well-targeted design extremely challenging. Yet, with the emergence of additive manufacturing which allows pointwise specification of the alloying components, new methods to predict alloy compositions and properties are needed. Here, we present a newly-developed numerical framework whereby a machine-learning algorithm supervised by atomistic-scale simulations is used to explore the nanoscale features controlling the diffusivity of atomic components in these heavily alloyed compounds. Analysis of all possible atomic configurations within a model medium-entropy alloy reveals how the size and cohesive energy of alloying elements alter the tendency of alloying elements to exchange their sites. Our developed theoretical model provides a pathway to calculate a macroscopic diffusivity rate from the information obtained from the nanoscale mechanisms. In the future, this approach can guide the selection of composition and processing parameters for conventional as well as additive manufacturing methods, and it could enable design of metals with tailored gradient diffusivity. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700