Search

This Methods/Protocols article is intended for materials scientists interested in performing machine learning-centered research. We cover broad guidelines and best practices regarding the obtaining and treatment of data, feature engineering, model training, validation, evaluation and comparison, popular repositories for materials data and benchmarking data sets, model and architecture sharing, and finally publication. In addition, we include interactive Jupyter notebooks with example Python code to demonstrate some of the concepts, workflows, and best practices discussed. Overall, the data-driven methods and machine learning workflows and considerations are presented in a simple way, allowing interested readers to more intelligently guide their machine learning research using the suggested references, best practices, and their own materials domain expertise.

The prediction of new compounds via crystal structure prediction may transform how the materials chemistry community discovers new compounds. In the prediction of inorganic crystal structures there are three distinct classes of prediction: performing crystal structure prediction via heuristic algorithms, using a range of established crystal structure prediction codes, an emerging community using generative machine learning models to predict crystal structures directly and the use of mathematical optimisation to solve crystal structures exactly. In this work, we demonstrate the combination of heuristic and generative machine learning, the use of a generative machine learning model to produce the starting population of crystal structures for a heuristic algorithm and discuss the benefits, demonstrating the method on eight known compounds with reported crystal structures and three hypothetical compounds. We show that the integration of machine learning structure generation with heuristic structure prediction results in both faster compute times per structure and lower energies. This work provides to the community a set of eleven compounds with varying chemistry and complexity that can be used as a benchmark for new crystal structure prediction methods as they emerge.

In scientific disciplines, benchmarks play a vital role in driving progress forward. For a benchmark to be effective, it must closely resemble real-world tasks. If the level of difficulty or relevance is inadequate, it can impede progress in the field. Moreover, benchmarks should have low computational overhead to ensure accessibility and repeatability. The objective is to achieve a kind of ``Turing test'' by creating a surrogate model that is practically indistinguishable from the ground truth observation, at least within the dataset's explored boundaries. This objective necessitates a large quantity of data. This data encompasses numerous features that are characteristic of chemistry and materials science optimization tasks that are relevant to industry. These features include high levels of noise, multiple fidelities, multiple objectives, linear constraints, non-linear correlations, and failure regions. We performed 494498 random hard-sphere packing simulations representing 206 CPU days' worth of computational overhead. Simulations required nine input parameters with linear constraints and two discrete fidelities each with continuous fidelity parameters. The data was logged in a free-tier shared MongoDB Atlas database, producing two core tabular datasets: a failure probability dataset and a regression dataset. The failure probability dataset maps unique input parameter sets to the estimated probabilities that the simulation will fail. The regression dataset maps input parameter sets (including repeats) to particle packing fractions and computational runtimes for each of the two steps. These two datasets were used to create a surrogate model as close as possible to running the actual simulations by incorporating simulation failure and heteroskedastic noise. In the regression dataset, percentile ranks were calculated for each group of identical parameter sets to account for heteroskedastic noise, thereby ensuring reliable and accurate data. This differs from the conventional approach that imposes a-priori assumptions, such as Gaussian noise, by specifying mean and standard deviation. This technique can be extended to other benchmark datasets to bridge the gap between optimization benchmarks with low computational overhead and the complex optimization scenarios encountered in the real world., (© 2023 The Author(s).)

Learn how to build a Closed-loop Spectroscopy Lab: Light-mixing demo (CLSLab:Light) to perform color matching via RGB LEDs and a light sensor for under 100 USD and less than an hour of setup. Our tutorial covers ordering parts, verifying prerequisites, software setup, sensor mounting, testing, and an optimization algorithm comparison tutorial. We use secure IoT-style communication via MQTT, MicroPython firmware on a pre-soldered Pico W microcontroller, and the self-driving-lab-demo Python package. A video tutorial is available at https://youtu.be/D54yfxRSY6s. For complete details on the use and execution of this protocol, please refer to Baird et al. 1 ., Competing Interests: Declaration of interests The authors have been exploring selling at-cost kits via a crowdfunding platform called GroupGets. As of April 23, 2023, the cost breakdown is as follows: List price: 80 USD + shipping (depends on location). GroupGets fee: 15.57 USD (10% + 2.9% + 5 USD). Hardware cost: 51.19 USD. Labor/testing: 20 min (est. 6 USD). Profit: 2.24 USD, offset somewhat by bulk pricing discounts. Note that this assumes purchasing a Pico W with presoldered headers. See round 1 and round 2 and a discussion of packaging open-source hardware as commercial kits., (Copyright © 2023 The Author(s). Published by Elsevier Inc. All rights reserved.)

This work presents a new method for processing single-crystal semiconductors designed by a computational method to lower the process temperature. This research study is based on a CALPHAD approach (ThermoCalc) to theoretically design processing parameters by utilizing theoretical phase diagrams. The targeted material composition consists of Bi-Se 2 -Te-Sb (BSTS). The semiconductor alloy contains three phases, hexagonal, rhombohedral-1, and rhombohedral-2 crystal structures, that are presented in the phase field of the theoretical pseudo-binary phase diagram. The semiconductor is also evaluated by applying Hume-Rothery rules along with the CALPHAD approach. Thermodynamic modelling suggests that single-crystals of BSTS can be grown at significantly lower temperatures and this is experimentally validated by low-temperature growth of single crystalline samples followed by exfoliation, compositional analysis, and diffraction., (© 2023. The Author(s).)

As the thickness of a three-dimensional (3D) topological insulator (TI) becomes comparable to the penetration depth of surface states, quantum tunneling between surfaces turns their gapless Dirac electronic structure into a gapped spectrum. Whether the surface hybridization gap can host topological edge states is still an open question. Herein, we provide transport evidence of 2D topological states in the quantum tunneling regime of a bulk insulating 3D TI BiSbTeSe 2 . Different from its trivial insulating phase, this 2D topological state exhibits a finite longitudinal conductance at ~2e 2 /h when the Fermi level is aligned within the surface gap, indicating an emergent quantum spin Hall (QSH) state. The transition from the QSH to quantum Hall (QH) state in a transverse magnetic field further supports the existence of this distinguished 2D topological phase. In addition, we demonstrate a second route to realize the 2D topological state via surface gap-closing and topological phase transition mechanism mediated by a transverse electric field. The experimental realization of the 2D topological phase in a 3D TI enriches its phase diagram and marks an important step toward functionalized topological quantum devices., (© 2022. The Author(s).)

We demonstrate advantages of samples made by mechanical stacking of exfoliated van der Waals materials for controlling the topological surface state of a three-dimensional topological insulator (TI) via interaction with an adjacent magnet layer. We assemble bilayers with pristine interfaces using exfoliated flakes of the TI BiSbTeSe 2 and the magnet Cr 2 Ge 2 Te 6 , thereby avoiding problems caused by interdiffusion that can affect interfaces made by top-down deposition methods. The samples exhibit an anomalous Hall effect (AHE) with abrupt hysteretic switching. For the first time in samples composed of a TI and a separate ferromagnetic layer, we demonstrate that the amplitude of the AHE can be tuned via gate voltage with a strong peak near the Dirac point. This is the signature expected for the AHE due to Berry curvature associated with an exchange gap induced by interaction between the topological surface state and an out-of-plane-oriented magnet.

Human analysis of research data is slow and inefficient. In recent years, machine learning tools have advanced our capability to perform tasks normally carried out by humans, such as image segmentation and classification. In this work, we seek to further improve binary classification models for high-throughput identification of different microstructural morphologies. We utilize a dataset with limited observations (133 dendritic structures, 444 non-dendritic) and employ data augmentation via rotation and translation to enhance the dataset six-fold. Then, transfer learning is carried out using pre-trained networks VGG16, InceptionV3, and Xception achieving only moderate F1 scores (0.801-0.822). We hypothesize that feature engineering could yield better results than transfer learning alone. To test this, we employ a new nature-inspired feature optimization algorithm, the Binary Red Deer Algorithm (BRDA), to carry out binary classification and observe F1 scores in the range of 0.96., Competing Interests: All the authors declare that they have no conflict of interest or competing interests., (© 2022 The Authors.)

Materials discovery via machine learning has become an increasingly popular method due to its ability to rapidly predict materials properties in a time-efficient and low-cost manner. However, one limitation in this field is the lack of benchmark datasets, particularly those that encompass the size, tasks, material systems, and data modalities present in the materials informatics literature. This makes it difficult to identify optimal machine learning model choices including algorithm, model architecture, data splitting, and data featurization for a given task. Here, we attempt to address this lack of benchmark datasets by assembling a unique repository of 50 different datasets for materials properties. The data contains both experimental and computational data, data suited for regression as well as classification, sizes ranging from 12 to 6354 samples, and materials systems spanning the diversity of materials research. Data were extracted from 16 publications. In addition to cleaning the data where necessary, each dataset was split into train, validation, and test splits. For datasets with more than 100 values, train-val-test splits were created, either with a 5-fold or 10-fold cross-validation method, depending on what each respective paper did in their studies. Datasets with less than 100 values had train-test splits created using the Leave-One-Out cross-validation method. These benchmark data can serve as a basis for a more diverse benchmark dataset in the future to further improve their effectiveness in the comparison of machine learning models., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2021 The Authors. Published by Elsevier Inc.)

Plastic pollution is entering the world's oceans at alarming rates and is expected to outweigh fish populations by 2050. This plastic waste originates from land-based applications, like consumer product packaging, and is composed of high-durability polyolefins. These conventional plastics possess desirable properties, including high chemical stability, moisture barrier, and thermoplastic characteristics. Unfortunately, if these materials reach marine environments, they fragment into microplastics that cannot be biologically assimilated. The aim of this review is to investigate commercial polymers that are biodegradable in marine environments but have comparable product stability and moisture barrier properties to polyolefins. Among commercially available biopolymers, thermoplastic starches (TPS) and polyhydroxyalkanoates (PHAs) have been shown to biodegrade in marine environments. Moreover, these biopolymers are thermoplastics and possess similar thermoforming properties to polyolefins. At present, TPS and PHAs have limitations, including chemical instability, limited moisture barrier properties, and high production costs. To replace conventional polymers with PHAs and TPS, these properties must be improved., Competing Interests: Declarations of Interests The authors declare no competing interests., (Copyright © 2020. Published by Elsevier Inc.)

Spectroscopy of discrete Landau levels (LLs) in bulk-insulating three-dimensional topological insulators (3D TIs) in perpendicular magnetic field characterizes the Dirac nature of their surface states. Despite a number of studies demonstrating the quantum Hall effect (QHE) of topological surface states, quantitative evaluation of the LL energies, which serve as fundamental electronic quantities for study of the quantum states, is still limited. In this work, we explore the density of states of LLs by measuring quantum capacitance ( C Q ) in a truly bulk insulating 3D TI via a van der Waals heterostructure configuration. By applying dual-gate voltages, we access the individual surface states' LLs and extract their chemical potentials to quantify the LL spacings of each surface. We evaluate the LLs' energies at two distinguished QH states, namely, dissipationless (ν = ±1) and dissipative (ν = 0) states in the 3D TI.

The paired top and bottom Dirac surface states, each associated with a half-integer quantum Hall (QH) effect, and a resultant integer QH conductance (νe^{2}/h), are hallmarks of a three-dimensional topological insulator (TI). In a dual-gated system, chemical potentials of the paired surface states are controlled through separate gates. In this work, we establish tunable capacitive coupling between the surface states of a bulk-insulating TI BiSbTeSe_{2} and study the effect of this coupling on QH plateaus and Landau level (LL) fan diagram via dual-gate control. We observe nonlinear QH transitions at low charge density in strongly coupled surface states, which are related to the charge-density-dependent coupling strength. A splitting of the N=0 LL at the charge neutrality point for thin devices (but thicker than the 2D limit) indicates intersurface hybridization possibly beyond single-particle effects. By applying capacitor charging models to the surface states, we explore their chemical potential as a function of charge density and extract the fundamental electronic quantity of LL energy gaps from dual-gated transport measurements. These studies are essential for the realization of exotic quantum effects such as topological exciton condensation.

Three dimensional (3D) topological insulators (TIs) are an important class of materials with applications in electronics, spintronics and quantum computing. With the recent development of truly bulk insulating 3D TIs, it has become possible to realize surface dominated phenomena in electrical transport measurements e.g. the quantum Hall (QH) effect of massless Dirac fermions in topological surface states (TSS). However, to realize more advanced devices and phenomena, there is a need for a platform to tune the TSS or modify them e.g. gap them by proximity with magnetic insulators, in a clean manner. Here we introduce van der Waals (vdW) heterostructures in the form of topological insulator/insulator/graphite to effectively control chemical potential of the TSS. Two types of gate dielectrics, normal insulator hexagonal boron nitride (hBN) and ferromagnetic insulator Cr 2 Ge 2 Te 6 (CGT) are utilized to tune charge density of TSS in the quaternary TI BiSbTeSe 2 . hBN/graphite gating in the QH regime shows improved quantization of TSS by suppression of magnetoconductivity of massless Dirac fermions. CGT/graphite gating of massive Dirac fermions in the QH regime yields half-quantized Hall conductance steps and a measure of the Dirac gap. Our work shows the promise of the vdW platform in creating advanced high-quality TI-based devices.

Despite numerous studies on three-dimensional topological insulators (3D TIs), the controlled growth of high quality (bulk-insulating and high mobility) TIs remains a challenging subject. This study investigates the role of growth methods on the synthesis of single crystal stoichiometric BiSbTeSe 2 (BSTS). Three types of BSTS samples are prepared using three different methods, namely melting growth (MG), Bridgman growth (BG) and two-step melting-Bridgman growth (MBG). Our results show that the crystal quality of the BSTS depend strongly on the growth method. Crystal structure and composition analyses suggest a better homogeneity and highly-ordered crystal structure in BSTS grown by MBG method. This correlates well to sample electrical transport properties, where a substantial improvement in surface mobility is observed in MBG BSTS devices. The enhancement in crystal quality and mobility allow the observation of well-developed quantum Hall effect at low magnetic field.

A total of 73 new quaternary rare-earth germanides RE 4 M 2 XGe 4 ( RE = rare-earth metal; M = Mn-Ni; X = Ag, Cd) were prepared through reactions of the elements. The solid solution Nd 4 Mn 2 Cd(Ge 1- y Si y ) 4 was also prepared under the same conditions and found to be complete over the entire range. All of these compounds adopt the monoclinic Ho 4 Ni 2 InGe 4 -type structure (space group C2/ m, a = 14.2-16.7 Å, b = 4.0-4.6 Å, c = 6.8-7.5 Å, β = 106-109°), as revealed by powder X-ray diffraction analysis and single-crystal X-ray diffraction analysis on selected members. The structure determination of Nd 4 (Mn 0.78(1) Ag 0.22(1) ) 2 Ag 0.83(1) Ge 4 disclosed disorder of Mn and Ag atoms within the tetrahedral site and Ag deficiencies within the square planar site. Within the solid solution Nd 4 Mn 2 Cd(Ge 1- y Si y ) 4 , the end-members and two intermediate members were structurally characterized; as the Si content increases, the Cd sites become less deficient and the individual [Mn 2 Tt 2 ] layers contract but become further apart from each other. Electronic band structure calculations confirm that the Ag-Ge or Cd-Ge bonds are the weakest in the structure and thus prone to distortion. Thermal property measurements confirm expectations from machine-learning predictions that these quaternary germanides should exhibit low thermal conductivity, which was found to be <10 W m -1 K -1 for Nd 4 Mn 2 AgGe 4 .

In the pursuit of materials with exceptional mechanical properties, a machine-learning model is developed to direct the synthetic efforts toward compounds with high hardness by predicting the elastic moduli as a proxy. This approach screens 118 287 compounds compiled in crystal structure databases for the materials with the highest bulk and shear moduli determined by support vector machine regression. Following these models, a ternary rhenium tungsten carbide and a quaternary molybdenum tungsten borocarbide are selected and synthesized at ambient pressure. High-pressure diamond anvil cell measurements corroborate the machine-learning prediction of the bulk modulus with less than 10% error, as well as confirm the ultraincompressible nature of both compounds. Subsequent Vickers microhardness measurements reveal that each compound also has an extremely high hardness exceeding the superhard threshold of 40 GPa at low loads (0.49 N). These results show the effectiveness of materials development through state-of-the-art machine-learning techniques by identifying functional inorganic materials.

Solid-state lighting using laser diodes is an exciting new development that requires new phosphor geometries to handle the greater light fluxes involved. The greater flux from the source results in more conversion and therefore more conversion loss in the phosphor, which generates self-heating, surpassing the stability of current encapsulation strategies used for light-emitting diodes, usually based on silicones. Here, we present a rapid method using spark plasma sintering (SPS) for preparing ceramic phosphor composites of the canonical yellow-emitting phosphor Ce-doped yttrium aluminum garnet (Ce:YAG) combined with a chemically compatible and thermally stable oxide, α-Al 2 O 3 . SPS allows for compositional modulation, and phase fraction, microstructure, and luminescent properties of ceramic composites with varying compositions are studied here in detail. The relationship between density, thermal conductivity, and temperature rise during laser-driven phosphor conversion is elucidated, showing that only modest densities are required to mitigate thermal quenching in phosphor composites. Additionally, the scattering nature of the ceramic composites makes them ideal candidates for laser-driven white lighting in reflection mode, where Lambertian scattering of blue light offers great color uniformity, and a luminous flux >1000 lm is generated using a single commercial laser diode coupled to a single phosphor element.

Attempts to prepare Gd12Co5Bi, a member of the rare-earth (RE) intermetallics RE12Co5Bi, which were identified by a machine-learning recommendation engine as potential candidates for thermoelectric materials, led instead to formation of the new compound Gd12Co5.3Bi with a very similar composition. Phase equilibria near the Gd-rich corner of the Gd-Co-Bi phase diagram were elucidated by both lab-based and variable-temperature synchrotron powder X-ray diffraction, suggesting that Gd12Co5.3Bi and Gd12Co5Bi are distinct phases. The higher symmetry structure of Gd12Co5.3Bi (cubic, space group Im3̅, Z = 2, a = 9.713(6) Å), as determined from single-crystal X-ray diffraction, is closely related to that of Gd12Co5Bi (tetragonal, space group Immm). Single Co atoms and Co-Co dumbbells are disordered with occupancies of 0.78(2) and 0.22(2), respectively, in Gd12Co5.3Bi, but they are ordered in Gd12Co5Bi. Consistent with this disorder, the electrical resistivity shows less dependence on temperature for Gd12Co5.3Bi than for Gd12Co5Bi. The thermal conductivity is low and reaches 2.8 W m(-1) K(-1) at 600 °C for both compounds; however, the temperature dependence of the thermal conductivity differs, decreasing for Gd12Co5.3Bi and increasing for Gd12Co5Bi as the temperature increases. The unusual trends in thermal properties persist in the heat capacity, which decreases below 2R, and in the thermal diffusivity, which increases at higher temperatures.