Large language models are increasingly applied to materials science, yet fundamental questions remain about their reliability and knowledge encoding. Evaluating 25 LLMs across four materials science tasks – over 200 base and fine-tuned configurations – we find that output modality fundamentally determines model behavior. For symbolic tasks, fine-tuning converges to consistent, verifiable answers with reduced response entropy, while for numerical tasks, fine-tuning improves prediction accuracy but models remain inconsistent across repeated inference runs, limiting their reliability as quantitative predictors. For numerical regression, we find that better performance can be obtained by extracting embeddings directly from intermediate transformer layers than from model text output, revealing an “LLM head bottleneck,” though this effect is property- and dataset-dependent. Finally, we present a longitudinal study of GPT model performance in materials science, tracking four models over 18 months and observing 9–43% performance variation that poses reproducibility challenges for scientific applications.

The synthesis of crystalline materials, such as zeolites, remains a notable challenge owing to a high-dimensional synthesis space, intricate structure-synthesis relationships and time-consuming experiments. Here, considering the 'one-to-many' relationship between structure and synthesis, we propose DiffSyn, a generative diffusion model trained on over 23,000 synthesis recipes that span 50 years of literature. DiffSyn generates probable synthesis routes conditioned on a desired zeolite structure and an organic template. DiffSyn achieves state-of-the-art performance by capturing the multi-modal nature of structure-synthesis relationships. We apply DiffSyn to differentiate among competing phases and generate optimal synthesis routes. As a proof of concept, we synthesize a UFI material using DiffSyn-generated synthesis routes. These routes, rationalized by density functional theory binding energies, resulted in the successful synthesis of a UFI material with a high Si/AlICP of 19.0, which is expected to improve thermal stability.

Organic molecular crystals underpin technologies ranging from pharmaceuticals to organic electronics, yet predicting solid-state packing of molecules remains challenging because candidate generation is combinatorial and stability is only resolved after costly energy evaluations. Here we introduce PackFlow, a flow matching framework for molecular crystal structure prediction (CSP) that generates heavy-atom crystal proposals by jointly sampling Cartesian coordinates and unit-cell lattice parameters given a molecular graph. This lattice-aware generation interfaces directly with downstream relaxation and lattice-energy ranking, positioning PackFlow as a scalable proposal engine within standard CSP pipelines. To explicitly steer generation toward physically favourable regions, we propose physics alignment, a reinforcement learning post-training stage that uses machine-learned interatomic potential energies and forces as stability proxies. Physics alignment improves physical validity without altering inference-time sampling. We validate PackFlow's performance against heuristic baselines through two distinct evaluations. First, on a broad unseen set of molecular systems, we demonstrate superior candidate generation capability, with proposals exhibiting greater structural similarity to experimental polymorphs. Second, we assess the full end-to-end workflow on two unseen CSP blind-test case studies, including relaxation and lattice-energy analysis. In both settings, PackFlow outperforms heuristics-based methods by concentrating probability mass in low-energy basins, yielding candidates that relax into lower-energy minima and offering a practical route to amortize the relax-and-rank bottleneck.

Increased use of supplementary cementitious materials (SCMs) is limited by variability in their compositions and reactivities, since these factors significantly affect concrete properties and durability. Traditional durability testing is time-consuming, making it infeasible to apply widely to emerging SCMs. This paper develops and applies a computational method to screen many blended cements containing coal fly ash and granulated blast furnace slag to identify those that have higher potential to produce high freeze-thaw resistance and low CO2 emissions concrete, varying SCM chemical composition, reactivity, and replacement levels. This is achieved using the Panoramix algorithm, which integrates thermodynamic modelling, random sampling, grid search algorithm, a freeze-thaw prediction model, and life cycle assessment (LCA) to probabilistically explore the design space, thereby rigorously accounting for uncertainties in raw material properties. A novel SCM reactivity database is developed and used here, enabling application of Panoramix to blended cement concretes. The modelling results show that at constant air content (4%), increasing the replacement levels of coal fly ash and granulated blast furnace slag decrease the mean value of the freeze-thaw resistance indicator but increase its variance, due to their broader chemical composition ranges than Portland clinker. At constant air void system quality and content, higher values of the freeze-thaw resistance indicator are calculated at increased SCM reactivity levels, and reduced climate change impacts are calculated at increased SCM substitution levels. The results indicate that a balance between higher freeze-thaw resistance and lower CO2 emissions can be approached using high reactivity SCMs and low clinker-to-cement ratios. The methodology used here facilitates multi-objective concrete selection considering environmental impacts and durability performance.

Sodium solid-state batteries (NaSSBs) with alloy anodes can increase cell-level energy density, but they suffer from chemo-mechanical degradation due to volume changes during sodiation and desodiation. Here, we develop a Sn hard carbon composite anode (SnHC), in which HC serves as a mechanically compliant scaffold that buffers alloying-induced stress and helps maintain electrode integrity. Meanwhile, Sn and Na-Sn phases provide coupled ionic and electronic transport pathways for Na+ diffusion and electron percolation. Morphology and pressure monitoring show that SnHC exhibits a 77% reduction in volume change and a 46% reduction in pressure fluctuations compared with Sn upon sodiation. This stress-buffering effect reduces structural degradation during cycling. As a result, SnHC delivers a reversible capacity of 588 mAh g-1 and retains 80% of its capacity over 1000 cycles when paired with NaNi1/3Fe1/3Mn1/3O2. Techno-economic analysis indicates that SnHC can reduce the projected manufacturing cost of NaSSBs by approximately 21% relative to HC counterparts.

The global physical economy—the system of material, energy and emission flows from extraction to final use—has not been tabulated and visualized, but such a visual can guide improved resource efficiency, inform limits on alternative energy scale up and coordinate system-wide emissions reductions. We develop a structure and conventions to integrate dispersed data to visualize the dominant (> 50Mt/yr) global material, energy and emission flows from extraction to end-use through a Sankey diagram. This provides a quantitative map to comprehend scale and connections for resources and sectors typically studied in isolation—making clear, for example, that the annual mass of CO2 emissions in 2019 were greater than all solid material produced globally. Fossil fuels are not only the primary source of emissions, but also constitute some of the largest mineral flows (> 15Gt/yr, greater than all metallic minerals combined). This structure can be used to explore the resource implications of scaling up low-carbon electricity generation and biomass feedstocks. The transition from fossil fuels to a solar and wind-dominant energy system can reduce the throughput of minerals and energy for the energy system. Demand for biomass may increase by 1.5-2X to 2050, requiring coordination to ensure its best use. The Sankey diagrams presented here provide a coherent understanding of the current physical flows underpinning economic activity and support collective visioning for using resources to provide final goods and services with greater efficacy.

Successful upcycling of industrial residues requires integrated valorization strategies. This work presents case studies on copperCopper ore flotation tailings and copper slag flotation tails, highlighting potential processing approaches for recoveryRecovery supported by experimental data and economic estimations. Although extracting critical metalsCritical metals yields products of comparatively higher market value relative to construction use or CO2 mineralizationCO mineralization, the extractionExtraction and recoveryRecovery processes remain cost-intensive at the current stage and often generate large volumes of residual waste, motivating the use of leached residues for construction applications. In parallel, directly using tailings as supplementary cementitious materials or in other civil engineering applications can consume larger volumes but may sacrifice potential metal value. CO2 mineralizationCO mineralization offers another pathway for tailings valorization aligned with decarbonization goals. Fundamental characterization and early calorimetry analyses of three tailings are presented, as well as tests mixing slag tails before and after NaHCO3-based aqueous carbonation in cement hydration systems. Preliminary insights into sustainable copperCopper tailings management are provided, with economic estimations under assumed scenarios in this study for valorizing 4.5 million tonnes of tailings annually, highlighting the potential for construction applications.

Municipal solid waste incineration (MSWI) ash, a byproduct of incineration with energy recovery from municipal solid waste (MSW) materials, is often disposed of in landfills despite containing valuable metals. Previous efforts to recover metals from MSW have faced processing challenges in achieving qualities high enough to retain market value. This work evaluates the economic feasibility of a metal recovery process capable of extracting high-purity metals and metal hydroxides via a combination of electrolytic refining, chemical precipitation, and onsite regeneration of reagents. We estimate cash flows and capital investment for a representative facility processing 40,000 metric tons of MSWI ash residues per year. Operating cash flows yield net revenues of up to $28/tonne ash, contingent on generous electricity pricing and the presence of copper, zinc, and magnesium in regional MSW. Discounted cash flow analysis demonstrates initial investment may be recovered within a payback period of 13 years despite a high cost high relative to contemporary resource recovery projects from MSWI ash. Findings show resource recovery from waste to be profitable over annual operation, yet in need of financial incentives such as credits or subsidies to reduce the burden of initial investment.

The increasing demands of sustainable energy, electronics, and biomedical applications call for next-generation functional materials with unprecedented properties. Of particular interest are emerging materials that display exceptional physical properties, making them promising candidates in energy-efficient microelectronic devices. As the conventional Edisonian approach becomes significantly outpaced by growing societal needs, emerging computational modeling and machine learning (ML) methods are employed for the rational design of materials. However, the complex physical mechanisms, cost of first-principles calculations, and the dispersity and scarcity of data pose challenges to both physics-based and data-driven materials modeling. Moreover, the combinatorial composition-structure design space is high-dimensional and often disjoint, making design optimization nontrivial. In this Account, we review a team effort toward establishing a framework that integrates data-driven and physics-based methods to address these challenges and accelerate materials design. We begin by presenting our integrated materials design framework and its three components in a general context. We then provide an example of applying this materials design framework to metal-insulator transition (MIT) materials, a specific type of emerging materials with practical importance in next-generation memory technologies. We identify multiple new materials which may display this property and propose pathways for their synthesis. Finally, we identify some outstanding challenges in data-driven materials design, such as materials data quality issues and property-performance mismatch. We seek to raise awareness of these overlooked issues hindering materials design, thus stimulating efforts toward developing methods to mitigate the gaps.

The success of sodium-ion batteries (SIBs) hinges on mitigating underperformance in ways that are cost effective, manufacturable, and scalable. This work investigates interfacial, morphological, and bulk interventions to enhance the performance of layered metal oxide cathode active materials (CAMs) for SIBs. We mapped the full space of literature-reported SIB CAM challenges and their mitigations. We then estimated the manufacturing costs for a diverse and representative set of mitigation approaches. Adding sacrificial salts can be cost effective, given low materials costs and minimal process changes. By contrast, many methods are reported to tune CAM morphology. Several are likely challenging at scale due to process throughput and yield limitations. Finally, bulk modifications can mitigate the moisture sensitivity of some CAMs, a likely less costly route than expanding stringent atmosphere controls during manufacturing. We end by discussing the limits and promise of process cost analysis, given the current state of battery reporting in the literature.

Retrosynthesis strategically plans the synthesis of a chemical target compound from simpler, readily available precursor compounds. This process is critical for synthesizing novel inorganic materials, yet traditional methods in inorganic chemistry continue to rely on trial-and-error experimentation. Emerging machine-learning approaches struggle to generalize to entirely new reactions due to their reliance on known precursors, as they frame retrosynthesis as a multi-label classification task. To address these limitations, we propose Retro-Rank-In, a novel framework that reformulates the retrosynthesis problem by embedding target and precursor materials into a shared latent space and learning a pairwise ranker on a bipartite graph of inorganic compounds. We evaluate Retro-Rank-In's generalizability on challenging retrosynthesis dataset splits designed to mitigate data duplicates and overlaps. For instance, for Cr2AlB2, it correctly predicts the verified precursor pair CrB + Al despite never seeing them in training, a capability absent in prior work. Extensive experiments show that Retro-Rank-In sets a new state-of-the-art, particularly in out-of-distribution generalization and candidate set ranking, offering a powerful tool for accelerating inorganic material synthesis.

Here, we investigate a sodium-aluminosilicate-based geopolymer refractory insulation (GRI) with an optimized mixture of closed-cell and open-cell porosities as an innovative internal thermal insulation and containment material for high-temperature molten salt storage applications. The closed-cell porosities are achieved via the addition of aluminosilicate cenospheres, a byproduct of coal-fired power plants. The open-cell porosities are minimized by the addition of lightweight aggregates as well as via high-temperature heat treatment. The mixed closed- and open-cell pore structure allows partial permeation of the molten salt into the GRI while maintaining most of its thermal insulation performance. This thermal insulation and containment design allows the molten salt to freeze inside the insulation layer and effectively form a self-containing and self-healing barrier for molten salts. Extensive immersion tests in molten nitrate and molten chloride salts were conducted for >50 days. X-ray computed tomography characterizations and thermal conductivity measurements before and after the immersion tests were performed to understand the stability of the geopolymer matrix and the pore structure during molten salt immersion. The results indicated stable performance and promising applicability of the proposed insulation/ containment concept for high-temperature molten salt storage applications.

As electrification trends and clean energy deployment drive up copper demand, there will be pressure on copper supply chains. With annual copper demand expected to grow by 50% and reach 49 Mt by 2035, the world will continue to need additional sources of copper supply. While expanding mining projects could increase copper production, given the significant stock of material, secondary copper can play a vital role in meeting demand. We analyze the opportunity to meet growing copper demand via increased scrap collection and improved technical recycling efficiencies. We use an economic model of the global copper system—with China analyzed separately from the rest of the world—to quantify supply evolution by incorporating price feedback between demand and supply. The model quantifies the impact of the increased collection on the displacement of mining production and demonstrates how increasing recycling can modulate supply risks and copper prices. Aligned with recent literature on future copper flows, we find that there is an opportunity to increase scrap supply in 2040 by 46% (6.3 Mt) compared with the baseline.

Demand for many of the metals used in the energy transition is expected to grow rapidly. Many of these are by-products, often considered critical because their production responds weakly to prices and is instead tied to the economics of the host mineral. We present a model of prices and production for jointly produced commodities that accounts for interconnectivity between host and by-product markets at the mine level. We demonstrate this method using the copper-cobalt-nickel system, in which approximately 99% of cobalt is a by-product of copper or nickel mining. Our results show that the model more accurately captures the economic benefits of diversified mine outputs than previous approaches. Furthermore, changes in demand drivers for any two commodities produce non-linear effects on production and price. We challenge the prior best-practice assumption that cobalt cannot impact the copper or nickel markets. Recognizing the importance of both copper and cobalt for future electrification, we emphasize that incentivizing the copper industry to reduce cobalt supply risks could inadvertently undermine copper supply.

Datacenters have become the backbone of modern digital infrastructure, powering the rapid rise of artificial intelligence and promising economic growth and technological progress. However, this expansion has brought growing tensions in the local communities where datacenters are already situated or being proposed. While the mainstream discourse often focuses on energy usage and carbon footprint of the computing sector at a global scale, the local socio-environmental consequences—such as health impacts, water usage, noise pollution, infrastructural strain, and economic burden—remain largely underexplored and poorly addressed. In this work1, we surface these community-level consequences through a mixed-methods study that combines quantitative data with qualitative insights. Focusing on Northern Virginia’s “Data Center Alley,” we highlight how datacenter growth reshapes local environments and everyday life, and examine the power dynamics that determine who benefits and who bears the costs. Our goal is to bring visibility to these impacts and prompt more equitable and informed decisions about the future of digital infrastructure.

This article provides a critical review of the socioeconomic impacts of mining critical raw materials (CRMs), whose global demand is rising at an unprecedented pace due to the ongoing energy transition. Our review focuses on five key socioeconomic aspects: health, employment, human development, public services, and culture. These areas are significantly affected by mining in general, and by CRM extraction in particular, each with distinct features. By critically surveying the scientific literature on the spillovers of extractive industry, this review investigates and highlights the complexity of the interconnected impacts on the communities and regions that host mining projects. Thus, we call for a holistic, interdisciplinary approach to better understand these effects and their interactions. This perspective can help the international community pursue not only the energy transition, but also a just energy transition—one that internalizes the negative socioeconomic externalities and considers social justice, while mitigating environmental harms.

Effective strategies for waste valorization from industrial residues include metal recovery, use in construction materials, and CO2 mineralization, all of which extend the value of these materials and reduce direct disposal. This study reviews valorization potential across an array of industrial residues, including pyrometallurgical slags, hydrometallurgical residues, power-plant combustion ashes, and mine wastes. Critical metal recovery is prioritized over construction material use driven by metal supply needs, starting with the extraction of rare earth elements (REEs) followed by other metals (Li, Co, Cu, Ni, etc.) depending on the profitability of the extraction and recovery processes. By-products generated from post-metal recovery processes are considered for construction materials. When metal recovery is limited, valorization instead focuses on use in construction materials, first based on their suitability as supplementary cementitious materials (SCMs), followed by structural aggregates, and other civil applications. The study discusses the CO2 mineralization potential and CO2 uptake of varied industrial residues, summarizing their effects on cement hydration and the mechanical performance of carbonated industrial residues used in construction applications, as well as the economic implications and environmental impacts of scaled use. We present two integrated valorization examples for red mud and steel slags. REE concentrations in red mud vary geographically and do not typically provide economic benefit (revenue-to-cost ratio at 1:10). Variability in steel slag composition coupled with operational complexity prevents CO2 mineralization potential, which could be up to about -100 kg CO2 eq./ton.

Inorganic synthesis planning currently relies primarily on heuristic approaches or machine learning models trained on limited data sets, which constrains its generality. We demonstrate that language models (LMs) without task-specific fine-tuning can recall synthesis conditions reported in the scientific literature. Off-the-shelf models, such as GPT-4.1, Gemini 2.0 Flash, and Llama 4 Maverick achieve a Top-1 precursor prediction accuracy of up to 53.8% and a Top-5 performance of 66.8% on a held-out set of 1000 reactions. They also predict calcination and sintering temperatures with mean absolute errors of <126 °C, matching or surpassing specialized regression models. Ensembling these LMs further enhances predictive accuracy and reduces inference cost per prediction by up to 70%. Given the broad, cross-domain knowledge of LMs, we evaluate whether they enable knowledge transfer by training a transformer, SyntMTE, on 28,548 LM-generated reaction recipes. Compared to a model trained on literature-reported data, we find that a model trained solely on LM-generated data exhibits competitive performance (only 6% worse). Conversely, a model trained on both the LM-generated and literature-reported data improves performance by up to 4%. In a case study on Li7La3Zr2O12 solid-state electrolytes, we demonstrate that SyntMTE reproduces the experimentally observed dopant-dependent sintering trends. Our hybrid workflow enables scalable and data-efficient inorganic synthesis planning.

Cement production contributes to >6

Artificial intelligence (AI) is increasingly used for the inverse design of materials, such as crystals and molecules. Existing AI research on molecules has integrated chemical structures of molecules with textual knowledge to adapt to complex instructions. However, this approach has been unattainable for crystals due to data scarcity from the biased distribution of investigated crystals and the lack of semantic supervision in peer-reviewed literature. In this work, we introduce a contrastive language-crystals model (CLaC) pre-trained on a newly synthesized dataset of 126k crystal structure-text pairs. To demonstrate the advantage of using synthetic data to overcome data scarcity, we constructed a comparable dataset extracted from academic papers. We evaluate CLaC's generalization ability through various zero-shot cross-modal tasks and downstream applications. In experiments, CLaC achieves state-of-the-art zero-shot generalization performance in understanding crystal structures, surpassing latest large language models.