•Procedure for using machine learning models with atomistic calculation data.•Many examples of how machine learning can accelerate atomistic calculations.•Discussion of future challenges and ...opportunities in this field.
In recent years, there has been a large effort in the materials science community to employ materials informatics to accelerate materials discovery or to develop new understanding of materials behavior. Materials informatics methods utilize machine learning techniques to extract new knowledge or predictive models out of existing materials data. In this review, we discuss major advances in the intersection between data science and atom-scale calculations with a particular focus on studies of solid-state, inorganic materials. The examples discussed in this review cover methods for accelerating the calculation of computationally-expensive properties, identifying promising regions for materials discovery based on existing data, and extracting chemical intuition automatically from datasets. We also identify key issues in this field, such as limited distribution of software necessary to utilize these techniques, and opportunities for areas of research that would help lead to the wider adoption of materials informatics in the atomistic calculations community.
Constructed to satisfy 17 known exact constraints for a semilocal density functional, the strongly constrained and appropriately normed (SCAN) meta-generalized-gradient-approximation functional has ...shown early promise for accurately describing the electronic structure of molecules and solids. One open question is how well SCAN predicts the formation energy, a key quantity for describing the thermodynamic stability of solid-state compounds. To answer this question, we perform an extensive benchmark of SCAN by computing the formation energies for a diverse group of nearly 1000 crystalline compounds for which experimental values are known. Due to an enhanced exchange interaction in the covalent bonding regime, SCAN substantially decreases the formation energy errors for strongly bound compounds, by approximately 50% to 110 meV/atom, as compared to the generalized gradient approximation of Perdew, Burke, and Ernzerhof (PBE). However, for intermetallic compounds, SCAN performs moderately worse than PBE with an increase in formation energy error of approximately 20%, stemming from SCAN's distinct behavior in the weak bonding regime. The formation energy errors can be further reduced via elemental chemical potential fitting. We find that SCAN leads to significantly more accurate predicted crystal volumes, moderately enhanced magnetism, and mildly improved band gaps as compared to PBE. Altogether, SCAN represents a significant improvement in accurately describing the thermodynamics of strongly bound compounds.
We use density functional theory (DFT) in conjunction with grand canonical linear programming (GCLP), a powerful automated tool for analyzing ground state thermodynamics, to exhaustively enumerate ...the 515 thermodynamically stable lithiation reactions of transition metal silicides, stannides and phosphides, and compute cell potential, volume expansion, and capacity for each. These reactions comprise an exhaustive list of all possible thermodynamically stable ternary conversion reactions for these transition metal compounds. The reactions are calculated based on a library DFT energies of 291 compounds, including all transition metal silicides, phosphides and stannides found in the Inorganic Crystal Structure Database (ICSD). We screen our computational database for the most appealing anode properties based on gravimetric capacity, volumetric capacity, cell potential, and volume expansion when compared with graphitic carbon anodes. This high‐throughput computational approach points towards several promising anode compositions with properties significantly superior to graphitic carbon, including CoSi2, TiP and NiSi2.
A first principles based, high‐throughput screening for lithium ion battery anode materials in the categories of transition metal silicides, stannides and phosphides. Among these materials, DFT in conjunction with GCLP is used to predict reaction capacities, voltages and volume expansion. Based on performance in these metrics several candidate materials for investigation are proposed.
Conventional machine learning approaches for predicting material properties from elemental compositions have emphasized the importance of leveraging domain knowledge when designing model inputs. ...Here, we demonstrate that by using a deep learning approach, we can bypass such manual feature engineering requiring domain knowledge and achieve much better results, even with only a few thousand training samples. We present the design and implementation of a deep neural network model referred to as ElemNet; it automatically captures the physical and chemical interactions and similarities between different elements using artificial intelligence which allows it to predict the materials properties with better accuracy and speed. The speed and best-in-class accuracy of ElemNet enable us to perform a fast and robust screening for new material candidates in a huge combinatorial space; where we predict hundreds of thousands of chemical systems that could contain yet-undiscovered compounds.
Thermoelectric technology, harvesting electric power directly from heat, is a promising environmentally friendly means of energy savings and power generation. The thermoelectric efficiency is ...determined by the device dimensionless figure of merit ZTdev, and optimizing this efficiency requires maximizing ZT values over a broad temperature range. Here, we report a record high ZTdev ~1.34, with ZT ranging from 0.7 to 2.0 at 300 to 773 kelvin, realized in hole-doped tin selenide (SnSe) crystals. The exceptional performance arises from the ultrahigh power factor, which comes from a high electrical conductivity and a strongly enhanced Seebeck coefficient enabled by the contribution of multiple electronic valence bands present in SnSe. SnSe is a robust thermoelectric candidate for energy conversion applications in the low and moderate temperature range.
In this study, a series of Ge1–x Mn x Te (x = 0–0.21) compounds were prepared by a melting–quenching–annealing process combined with spark plasma sintering (SPS). The effect of alloying MnTe into ...GeTe on the structure and thermoelectric properties of Ge1–x Mn x Te is profound. With increasing content of MnTe, the structure of the Ge1–x Mn x Te compounds gradually changes from rhombohedral to cubic, and the known R3m to Fm-3m phase transition temperature of GeTe moves from 700 K closer to room temperature. First-principles density functional theory calculations show that alloying MnTe into GeTe decreases the energy difference between the light and heavy valence bands in both the R3m and Fm-3m structures, enhancing a multiband character of the valence band edge that increases the hole carrier effective mass. The effect of this band convergence is a significant enhancement in the carrier effective mass from 1.44 m 0 (GeTe) to 6.15 m 0 (Ge0.85Mn0.15Te). In addition, alloying with MnTe decreases the phonon relaxation time by enhancing alloy scattering, reduces the phonon velocity, and increases Ge vacancies all of which result in an ultralow lattice thermal conductivity of 0.13 W m–1 K–1 at 823 K. Subsequent doping of the Ge0.9Mn0.1Te compositions with Sb lowers the typical very high hole carrier concentration and brings it closer to its optimal value enhancing the power factor, which combined with the ultralow thermal conductivity yields a maximum ZT value of 1.61 at 823 K (for Ge0.86Mn0.10Sb0.04Te). The average ZT value of the compound over the temperature range 400–800 K is 1.09, making it the best GeTe-based thermoelectric material.
The use of hydrogen as fuel is a promising avenue to aid in the reduction of greenhouse effect gases released in the atmosphere. In this work, we present a high-throughput density functional theory ...(HT-DFT) study of 5,329 cubic and distorted perovskite ABO3 compounds to screen for thermodynamically favorable two-step thermochemical water splitting (TWS) materials. From a data set of more than 11,000 calculations, we screened materials based on the following: (a) thermodynamic stability and (b) oxygen vacancy formation energy that allow favorable TWS. From our screening strategy, we identify 139 materials as potential new candidates for TWS application. Several of these compounds, such as CeCoO3 and BiVO3, have not been experimentally explored yet for TWS and present promising avenues for further research. We show that taking into consideration all phases present in the A–B–O ternary phase, as opposed to only calculating the formation energy of a compound, is crucial to assess correctly the stability of a compound as it reduces the number of potential candidates from 5,329 to 383. Finally, our large data set of compounds containing stabilites, oxidation states, and ionic sizes allowed us to revisit the structural maps for perovskites by showing stable and unstable compounds simultaneously.
The broad-based implementation of thermoelectric materials in converting heat to electricity hinges on the achievement of high conversion efficiency. Here we demonstrate a thermoelectric figure of ...merit ZT of 2.5 at 923 K by the cumulative integration of several performance-enhancing concepts in a single material system. Using non-equilibrium processing we show that hole-doped samples of PbTe can be heavily alloyed with SrTe well beyond its thermodynamic solubility limit of <1 mol%. The much higher levels of Sr alloyed into the PbTe matrix widen the bandgap and create convergence of the two valence bands of PbTe, greatly boosting the power factors with maximal values over 30 μW cm(-1) K(-2). Exceeding the 5 mol% solubility limit leads to endotaxial SrTe nanostructures which produce extremely low lattice thermal conductivity of 0.5 W m(-1) K(-1) but preserve high hole mobilities because of the matrix/precipitate valence band alignment. The best composition is hole-doped PbTe-8%SrTe.
SnTe is a potentially attractive thermoelectric because it is the lead-free rock-salt analogue of PbTe. However, SnTe is a poor thermoelectric material because of its high hole concentration arising ...from inherent Sn vacancies in the lattice and its very high electrical and thermal conductivity. In this study, we demonstrate that SnTe-based materials can be controlled to become excellent thermoelectrics for power generation via the successful application of several key concepts that obviate the well-known disadvantages of SnTe. First, we show that Sn self-compensation can effectively reduce the Sn vacancies and decrease the hole carrier density. For example, a 3 mol % self-compensation of Sn results in a 50% improvement in the figure of merit ZT. In addition, we reveal that Cd, nominally isoelectronic with Sn, favorably impacts the electronic band structure by (a) diminishing the energy separation between the light-hole and heavy-hole valence bands in the material, leading to an enhanced Seebeck coefficient, and (b) enlarging the energy band gap. Thus, alloying with Cd atoms enables a form of valence band engineering that improves the high-temperature thermoelectric performance, where p-type samples of SnCd0.03Te exhibit ZT values of ∼0.96 at 823 K, a 60% improvement over the Cd-free sample. Finally, we introduce endotaxial CdS or ZnS nanoscale precipitates that reduce the lattice thermal conductivity of SnCd0.03Te with no effect on the power factor. We report that SnCd0.03Te that are endotaxially nanostructured with CdS and ZnS have a maximum ZTs of ∼1.3 and ∼1.1 at 873 K, respectively. Therefore, SnTe-based materials could be ideal alternatives for p-type lead chalcogenides for high temperature thermoelectric power generation.