Search

Despite recent breakthroughs in deep learning for materials informatics, there exists a disparity between their popularity in academic research and their limited adoption in the industry. A significant contributor to this “interpretability-adoption gap” is the prevalence of black-box models and the lack of built-in methods for model interpretation. While established methods for evaluating model performance exist, an intuitive understanding of the modeling and decision-making processes in models is nonetheless desired in many cases. In this work, we demonstrate several ways of incorporating model interpretability to the structure-agnostic Compositionally Restricted Attention-Based network, CrabNet. We show that CrabNet learns meaningful, material property-specific element representations based solely on the data with no additional supervision. These element representations can then be used to explore element identity, similarity, behavior, and interactions within different chemical environments. Chemical compounds can also be uniquely represented and examined to reveal clear structures and trends within the chemical space. Additionally, visualizations of the attention mechanism can be used in conjunction to further understand the modeling process, identify potential modeling or dataset errors, and hint at further chemical insights leading to a better understanding of the phenomena governing material properties. We feel confident that the interpretability methods introduced in this work for CrabNet will be of keen interest to materials informatics researchers as well as industrial practitioners alike.

New featurization schemes for describing materials as composition vectors in order to predict their properties using machine learning are common in the field of Materials Informatics. However, little is known about the comparative efficacy of these methods. This work sets out to make clear which featurization methods should be used across various circumstances. Our findings include, surprisingly, that simple fractional and random-noise representations of elements can be as effective as traditional and new descriptors when using large amounts of data. However, in the absence of large datasets or for data that is not fully representative, we show that the integration of domain knowledge offers advantages in predictive ability.

In this paper, we demonstrate an application of the Transformer self-attention mechanism in the context of materials science. Our network, the Compositionally Restricted Attention-Based network (), explores the area of structure-agnostic materials property predictions when only a chemical formula is provided. Our results show that ’s performance matches or exceeds current best-practice methods on nearly all of 28 total benchmark datasets. We also demonstrate how ’s architecture lends itself towards model interpretability by showing different visualization approaches that are made possible by its design. We feel confident that and its attention-based framework will be of keen interest to future materials informatics researchers.

Over the past decade, raw material price spikes have called attention to the supply security of a variety of critical materials, including rhenium, rare earth elements, and helium. While market forces play an important role in creating and resolving these situations, transitions in technology also create step-changes in demand that increase or decrease the criticality of different materials. With an appropriate understanding of how materials are used in various applications, it is possible to explore the critical materials implications associated with the introduction of new technologies. Work is already underway to investigate and mitigate the materials impacts of emerging clean energy technologies related to solar power and energy storage. Rapid technological change is also being enabled by information technologies and the Internet of Things (IOT). Here, less work has been done around materials trends and their implications. This paper presents a case study around emerging technologies for data storage and what their implementation at mass scale (zettabyte, ZB) might mean for existing supply chains and market dynamics for certain critical materials.

As the technologies we use as a society have advanced, so have the materials used in these technologies. Some of these materials are exotic and highly specialized, making them particularly vulnerable to supply disruptions and supply disruptions particularly impactful. Such materials are designated as “critical” materials. Their level of criticality can be identified by accounting for a number of factors related to their supply risk and the extent to which a supply disruption would impact business operations or society at large. We highlight current methodologies used to assess materials criticality, how these assessments are used to reduce materials-related risk and to what extent there is room for improvement. Particularly, this paper reviews critical materials designations from the United States Department of Energy, the European Union, and the General Electric Company, and how they have changed over the period from 2008 to 2014. The changes suggest that the factors considered in criticality ratings have different natural time scales, and that criticality changes occur both due to supply-side risk mitigation as well as demand-side responses. Response options, whether on the supply or demand side, also span a range of time scales and the interaction between factors with different time scales can play a significant role in the dynamics. To date, many published analyses are snapshots in time. A detailed understanding of how risk profiles evolve remains an open question. The importance and impact of demand-side responses such as recycling, substitution and new technological development are discussed.

Rare-earth (RE) elements are enablers for a wide range of technologies, including high-strength permanent magnets, energy-efficient lighting, high-temperature thermal barrier coatings, and catalysts. While direct material substitution is difficult in many of these applications because of the specific electronic, optical, or electrochemical properties imparted by the individual rare-earth elements, we describe an example from the aviation industry where supply chain optimization may be an option. Ceramic matrix composite engine components require environmental barrier coatings (EBCs) to protect them from extreme temperatures and adverse reactions with water vapor in the hot gas path. EBC systems based on rare-earth silicates offer a unique combination of environmental resistance, thermal expansion matching, thermal conductivity, and thermal stability across the service temperature window. Several pure rare-earth silicates and solid solutions have been demonstrated in EBC applications. However, all rely on heavy rare-earth elements (HREEs) for phase stability. This article considers the possibility of using separation tailings containing a mixture of HREEs as a source material in lieu of using the high-purity HREE oxides. This option arises because the desired properties of RE-silicate EBCs derive from the average cation size rather than the electronic properties of the individual rare-earth cations. Because separation tailings have not incurred the costs associated with the final stages of separation, they offer an economical alternative to high-purity oxides for this emerging application.

Over the next few decades, the challenge of water scarcity is expected to grow more acute as water demands from the power generation, agriculture, industrial, and municipal sectors all increase. Energy production requires copious amounts of water, with the volume of water used by power generation ranking second only to that used for agriculture. This article reviews options for managing the water requirements associated with power generation. Although the effects of both existing and emerging modes of power generation on water use trends are explored, the primary focus is on thermal systems, which account for the majority of existing capacity.

Over the next twenty-five years, global energy consumption is projected to grow by almost half, and electricity generation is expected to nearly double. The massive investment in infrastructure required to satisfy this demand presents a major opportunity for innovation in how energy is produced, stored, transmitted, and used. In particular, there is keen interest in sustainable energy technologies capable of improving efficiency and reducing environmental footprint. Membranes have the potential to play a significant role in a number of relevant separations applications, including CO2 capture, energy storage, and water production for energy production. This article seeks to highlight opportunities for membranes-related R&D relevant to sustainable energy, of which the broader membrane community may not be fully aware. We pose some specific technical questions related to two topics – CO2 capture in power generation, and water treatment for energy production – with the twin goals of stimulating thought and discussion within the broader membrane community, and sparking increased engagement between researchers and industrial end-users. This article is based upon a plenary talk on opportunities for membrane technologies in the context of sustainable energy, given at the 2010 NAMS/ICIM conference.

Zeolite membranes have been studied for decades, but are not sufficiently robust for widespread practical application. We examine an unusual natural clinoptilolite material which has been compacted over time into crystalline blocks containing essentially no macroporosity. When sectioned, this material behaves as a solid, continuous molecular sieve membrane. Untreated membranes sliced from this dense material were found to have as much as two times higher ideal selectivity for H2 over CO2 and C1/C2 hydrocarbons than would be predicted by Knudsen diffusion. 1.2 mm-thick membrane sections modified by simple hydrothermal treatments and applied to the separation of hydrogen from CO2, CH4, C2H4, and C2H6 demonstrated H2 permeance as high as 5.2 × 10−7 mol m−2 s−1 Pa−1 combined with ideal selectivities of 57 (H2/CO2), 22 (H2/CH4), 98 (H2/C2H4) and 78 (H2/C2H6) at 25 °C; and 13 (H2/CO2), 6 (H2/CH4), 26 (H2/C2H4) and 12 (H2/C2H6) at 500 °C. These modified natural zeolite membranes were thermally and chemically stable, and their hydrogen permeance was reproducible after multiple temperature cycles. These unique natural zeolite membranes have the potential to be engineered for high-temperature, energy-efficient industrial separation and purification applications including hydrogen separation, and to serve as a model for the development of robust synthetic zeolite membranes with superior separation characteristics.

Precombustion CO2 capture is an important option for the management of greenhouse gas emissions from power generation, because the higher concentrations can lead to improved separation economics. Membranes can provide a significant boost in the performance by separating CO2 at high temperature and at high pressure, provided they are properly integrated into a power plant. In this study, membrane performance targets were derived for CO2 capture using H2-selective membranes in integrated gasification combined cycle (IGCC) power plant systems. Three key differences were found relative to plants configured for the production of industrial-grade H2. First, the H2/CO2 selectivity requirement is lower when the permeate is combusted in a gas turbine, as opposed to purified for industrial use. Second, the membrane is subject to additional selectivity requirements. For example, CO2 product purity specs impose a H2/N2 selectivity requirement. Third, the plant design must account for the energy associated with unshifted CO, CH4 and other fuel components that are not separated by the membrane. In contrast to H2 production systems where the high H2/CO2 selectivity requirement favors ultra-high selectivity metal-based membranes, the requirements for IGCC indicate a wider range of materials, including ceramic, zeolite, and polymeric membranes, should be considered.

A simple method to mask hydrophobically porous templates for patterned nanorod growth is described. Patterning was achieved by treating the template with an alkylsilane to render the pores hydrophobic, followed by irradiation with UV light to selectively degrade the alkyl siloxyl coating in the illuminated regions to produce a template with both hydrophobic and hydrophilic pores. Patterned arrays comprising a single metallic composition and multiple metallic compositions were electrodeposited in the hydrophobically masked template.

Alkylsilane-modified nanoporous ceramic membranes exclude water from their pores yet exhibit transmembrane electrical conductivity in aqueous electrolyte solutions. That effect was studied using impedance spectroscopy and (29)Si NMR. Anodic aluminum oxide membranes with alkylsilane-functionalized pores exhibited a transmembrane electrical resistance that increased with the length of the hydrocarbon chain. Microstructural studies revealed that the conduction was due primarily to a small number of "hydrophilically defective" pores in membranes modified by long-chain alkylsilanes and both hydrophilic defects and surface conduction in pores modified by short-chain alkylsilanes. Hydroxyl groups in short-chain alkylsilane layers act as "water wires" to enable surface ion transport. The local concentration of hydroxyl groups decreased with alkylsilane chain length, explaining the resistance trend. This constitutes the first direct evidence that alkylsilane functionalization affects electrical as well as wetting properties.

An evaporation-assisted templating method was developed for the synthesis of nanostructured europium-doped yttrium oxide (Y2O3:Eu3+, YEO) thin films. The method involves spin-coating a solution containing yttrium salt and block copolymer followed by thermal oxidation. Uniform films with controllable thickness ranging from 100 to 500 nm were obtained by tuning the composition of the precursor solution and the processing conditions. However, the method was not completely general as extremes in salt or polymer concentration triggered uncontrolled cracking and texturing of the film. The photoluminescence was proportional to both the thickness and surface roughness. Nanocrystalline, porous YEO particles produced using the evaporation method exhibited an intrinsic quantum efficiency approximately 25% of that of particles with micron-sized crystallites.

Heterogeneous mesoporous oxide structures consisting of adjacent domains of mesoporous titania and silica (TiO 2 /SiO 2 ) were synthesized by sequential deposition into the pores of a porous anodic aluminum oxide (AAO) support. Each anodic alumina macropore contained distinct mesoporous structural domains corresponding to the two oxide structures. Transmission electron microscopy imaging showed that the titania and silica regions contained organized pores with diameters of ∼8–10 nm and ∼13–16 nm, respectively. Nitrogen adsorption measurements showed a typical Type IV behavior characteristic of mesoporous materials. The surface area for the heterogeneous composites ranged between ∼75 and ∼100 m 2 /g. Gas permeability measurements of the samples indicated complete filling of the AAO support by the mesoporous TiO 2 /SiO 2 regions. Preliminary results regarding the possibility of extending this method to mesoporous ZrO 2 /SiO 2 compositions are also presented.

While exhibiting a well-defined nanometer-level structure, surfactant-templated nanoscopic silicas produced via self-assembly do not always possess long-range order. We demonstrate that long-range order can be controlled by guiding the self-assembly of nanostructured silica-surfactant hybrids with low-strength electric fields (E approximately 200 V/m) to produce nanoscopic silica with both the micrometer- and nanometer-level structures oriented parallel to the applied field. Under the influence of the electric field, nanoscopic silica particles migrate, elongate, and merge into fibers with a rate of migration proportional to the applied field strength. The linear dependence with the field strength indicates that the process is governed by electroosmotic flow but not by polarization effects. Realignment of the short-range ordered surfactant nanochannels along the fiber axis accompanies the migration.