Selenium is experiencing renewed interest as a elemental semiconductor for a range of optoelectronic and energy applications due to its irresistibly simple composition and favorable wide bandgap. However, its high volatility and low radiative efficiency make it challenging to assess structural and optoelectronic quality, calling for advanced, non-destructive characterization methods. In this work, we employ a closed-space encapsulation strategy to prevent degradation during measurement and enable sensitive probing of vibrational and optoelectronic properties. Using temperature-dependent Raman and photoluminescence spectroscopy, we investigate grown-in stress, vibrational dynamics, and electron-phonon interactions in selenium thin films synthesized under nominally identical conditions across different laboratories. Our results reveal that short-range structural disorder is not intrinsic to the material, but highly sensitive to subtle processing variations, which strongly influence electron-phonon coupling and non-radiative recombination. We find that such structural disorder and grown-in stress likely promote the formation of extended defects, which act as dominant non-radiative recombination centers limiting carrier lifetime and open-circuit voltage in photovoltaic devices. These findings demonstrate that the optoelectronic quality of selenium thin films can be significantly improved through precise control of synthesis and post-deposition treatments, outlining a clear pathway toward optimizing selenium-based thin film technologies through targeted control of crystallization dynamics and microstructural disorder.

Foundation machine learning interatomic potentials (MLIPs) have emerged as powerful tools for atomistic simulation, yet different models encode chemical environments in incompatible latent spaces, limiting direct comparison and interoperability. The platonic representation hypothesis suggests that sufficiently capable models converge towards a shared statistical representation of reality. Here, motivated by this hypothesis, we show that independently developed MLIPs exhibit statistically consistent geometric organization of atomic environments. By projecting embeddings relative to a set of atomic anchors, we unify the latent spaces of seven MLIPs-spanning equivariant, non-equivariant, conservative and non-conservative architectures-into a common latent space that preserves chemical periodicity and structural invariants. This unified framework enables cross-model optimal transport, interpretable embedding arithmetic and the detection of representational biases. Furthermore, we show that deviation in this space provides a ground-truth-free measure for atypical structures, and signals physical prediction failures. Our results suggest that the platonic representation offers a practical route towards interoperable, comparable and interpretable foundation models for materials science.

Metal-organic frameworks (MOFs) are versatile materials with tunable crystal structures, morphologies, and chemistries, offering diverse physical and chemical properties. Although typically electrically insulating, specific combinations of organic and inorganic components can impart electrical conductivity to MOFs. The virtually limitless chemical space of MOFs, however, presents a significant challenge in identifying optimal candidates for various applications. Although density functional theory (DFT) can probe their electronic structure, its high computational cost hinders the discovery of novel electroactive MOFs using machine learning due to limited data. To tackle these challenges, a semiempirical extended tight-binding approach (GFN1-xTB) is employed to compute the electronic properties of a dataset of MOFs, and it is shown that GFN1-xTB approximates MOF band gaps well, as compared to semilocal DFT. These data are used to train an interpretable Δ-learning model that predicts the difference between low- and high-fidelity band gaps, given by xTB and DFT data at the hybrid level, respectively. This model outperforms direct models trained using only the DFT values. The Δ-learning model also outperforms models with deep-learning architecture, fine-tuned on our custom dataset to predict the band gaps of MOFs. With limited high-quality DFT band gaps, taking advantage of Δ-learning using low-cost GFN1-xTB leads to better predictions than relying on DFT data alone.

The discovery of materials has long been a fundamental building block for technological advancement, yet traditional trial-and-error methods are slow and costly to meet the growing demand for novel functionality, particularly in green energy technologies, energy storage, and electronics. In response to this challenge, high-throughput screening and data-driven workflows that combine computational simulations, machine learning models, and materials databases have emerged as powerful tools for accelerating materials discovery. Spinels (AB2X4) stand out as a versatile class of materials with applications ranging from energy storage to catalysis. However, their full compositional space remains largely unexplored. In this work, we present a data-driven framework to identify potentially synthesisable spinel compounds composed of the first 83 elements in the periodic table with oxygen (O2-) and three chalcogen anions (S2-, Se2-, Te2-). Over 30,000 charge-balanced and chemically plausible candidates, including inverse spinels, were sequentially filtered based on the stability, structural feasibility, and electronic properties criteria. Our workflow integrates materials databases, empirical heuristic rules, and machine learning predictions to efficiently reduce the candidate pool. As a result, 2,303 novel spinel candidates were identified from this workflow, offering a diverse subset of target compounds for further investigation.

Existing benchmarks for computational materials discovery primarily evaluate static predictive tasks or isolated computational sub-tasks. While valuable, these evaluations neglect the inherently iterative and adaptive nature of scientific discovery. We introduce MAterials Discovery Environments (MADE), a novel framework for benchmarking end-to-end autonomous materials discovery pipelines. MADE simulates closed-loop discovery campaigns in which an agent or algorithm proposes, evaluates, and refines candidate materials under a constrained oracle budget, capturing the sequential and resource-limited nature of real discovery workflows. We formalize discovery as a search for thermodynamically stable compounds relative to a given convex hull, and evaluate efficacy and efficiency via comparison to baseline algorithms. The framework is flexible; users can compose discovery agents from interchangeable components such as generative models, filters, and planners, enabling the study of arbitrary workflows ranging from fixed pipelines to fully agentic systems with tool use and adaptive decision making. We demonstrate this by conducting systematic experiments across a family of systems, enabling ablation of components in discovery pipelines, and comparison of how methods scale with system complexity.

Understanding the electrode/electrolyte interface is essential for tuning electrocatalyst activity. Here, we combine operando optical spectroscopy, laser-induced current transient (LICT) measurements, and surface-enhanced infrared absorption spectroscopy (SEIRAS) to investigate the origin of cation-dependent oxygen evolution reaction (OER) activity on electrodeposited iridium oxide in 0.1 M MOH (M = TMA(+), K+, Na+, and Li+). We find that OER activity increases with increasing cation size (TMAOH > KOH > NaOH > LiOH). Operando optical spectroscopy reveals that the energetics of the redox transitions and the population of the redox-active species are independent of the electrolyte. Instead, the intrinsic turnover frequency varies strongly with the nature of the cation. LICT, SEIRAS, and quantum mechanics/molecular mechanics (QM/MM) simulations suggest that the interfacial solvent structure is the origin of this difference. With increasing cation size, the fraction of isolated water molecules and cation-coordinated water molecules increases, producing a more disordered interfacial environment. LICT measurements confirm that the potential of maximum entropy shifts closer to the water oxidation potential in the presence of larger cations in the electrolyte. We propose that a more disordered interface results in more isolated and reactive OH- ions and faster reorganization of the interfacial solvent structure during the rate-determining O-O bond formation step, thereby accelerating the OER kinetics. Through our work, using multimodal operando spectroscopy and molecular simulations, we highlight how interfacial solvent structure, controlled by electrolyte cations, governs reactivity at complex electrochemical interfaces.

To address pressing scientific challenges such as climate change, increasingly sophisticated generative models are being developed to efficiently sample the large chemical space of potential functional materials. The proliferation of these models has necessitated the establishment of rigorous evaluation metrics. While uniqueness (U), novelty (N), and stability (S) of samples serve as standard metrics, their current formulations show several limitations. U and N rely on binary comparisons of crystals, rendering them dependent on heuristic thresholds, incapable of quantifying the degree of similarity, sensitive to atomic coordinate perturbations, and not invariant to sample permutation. Similarly, the binary assessment of S risks a premature exclusion of marginally unstable yet potentially novel candidates. These limitations are addressed by making the aforementioned metrics continuous. Furthermore, we integrate them into a unified metric ``continuous SUN" (cSUN), which offers a smoother score distribution and greater tunability than the conventional binary SUN metric. Experimental results demonstrate that our continuous metrics provide granular insights into sample distributions and secondarily serve as a soft screening score for selecting samples for further analysis. Finally, the use of cSUN as a reward signal in reinforcement learning is explored, showing that its adjustable weighting scheme effectively enhances sample diversity and convergence to a better optima in a Chemeleon2 case study.

Many physical properties of functional materials are governed by their impurities rather than their bulk characteristics. Defects in crystals can activate electronic and ionic conductivity, create active centres for catalysis or store information through localized spin configurations. Accurate modelling of defect behaviour is therefore essential for predicting material performance and optimizing functionality across a vast application space. However, defect simulations are sensitive to choices made during setup, execution and analysis. In this Perspective, we highlight best practices for calculating and reporting point defect properties through computational methods, with a focus on the widely adopted supercell approach. Key considerations include accurate representation of the structural and electronic properties of the host material, appropriate choice of charge states, sufficient optimization of defect geometries and reproducible calculation of defect formation energies. Adhering to these practices will facilitate robust comparisons between studies and improve the integration of computational predictions with experimental results. We emphasize the importance of reporting computational parameters and correction schemes. Ultimately, an open approach to point defect simulations will strengthen the impact of computational studies and accelerate materials engineering. Point defects critically influence material properties and require accurate computational modelling for reliable predictions. This Perspective outlines best practices for defect simulations using supercell approaches, emphasizing methodological transparency, reproducibility and robust reporting to improve consistency and integration with experimental studies.

Discovering functional crystalline materials entails navigating an immense combinatorial design space. Although recent advances in generative artificial intelligence have enabled the sampling of chemically plausible compositions and structures, a fundamental challenge remains: the objective misalignment between the likelihood-based sampling in generative modelling and the targeted focus on underexplored regions where novel compounds reside. Here we introduce a reinforcement learning framework that guides latent denoising diffusion models in finding diverse and novel, yet thermodynamically viable, crystalline compounds. Our approach integrates group-relative policy optimization with verifiable, multi-objective rewards that jointly balance creativity, stability and diversity. Beyond de novo generation, we demonstrate enhanced property-guided design that preserves chemical validity while targeting desired functional properties. This approach establishes a modular foundation for controllable AI-driven inverse design that addresses the novelty-validity trade-off across the scientific discovery applications of generative models.

Data-driven strategies are reshaping computational materials design by accelerating the prediction of compounds with targeted functionalities. Beyond high-throughput screening, the integration of generative artificial intelligence enables exploration across vast chemical spaces comprising millions of known and hypothetical materials. This abundance of candidates presents a challenge: identifying which candidate compounds are not only low in energy but also synthetically accessible. Here we assess advances towards closing this synthesis gap for inorganic crystals. These include the incorporation of thermodynamic potentials (from internal energies at 0 K to Gibbs free energies at reaction conditions)—crucial for evaluating phase stability and reaction driving forces—chemical heuristics (from charge neutrality to electronegativity rules) and machine learning models (from positive unlabelled learning to large language models) to guide compound selection and prioritization. Looking forward, the development of more robust synthesizability metrics, synthesis planning tools, and agentic workflows integrating experimental feedback will narrow the divide between virtual screening and real-world materials realization. Digital workflows can calculate millions of hypothetical compounds but the divide between what is calculable and what is synthesizable is a challenge in computational materials discovery. This Perspective examines strategies towards closing the synthesis gap for inorganic crystals, including the incorporation of thermodynamic potentials, chemical heuristics and machine learning models.

Oxidation states underpin the understanding of active states, reaction mechanisms and catalytic performance of electrocatalysts. However, determining them at complex solid-liquid interfaces is challenging. Here we use multimodal spectroscopy to investigate polarized iridium oxide (IrOx) electrodes, a model water oxidation catalyst, to identify potential-dependent iridium and oxygen oxidation states. By integrating multiple operando spectroscopies (optical (ultraviolet-visible), Ir L-edge and O K-edge X-ray absorption spectroscopy) with electrochemistry mass spectrometry and density functional theory calculations, we identify the sequential depletion of electron densities from the Ir5d band (corresponding to Ir3+→Ir4+→Ir5+), followed by electron removal from the O2p band, forming electrophilic oxygen species (O-1) due to enhanced Ir-O covalency and electronic state overlap. Time-resolved measurements reveal distinct lifetimes for Ir5+ and O-1 states under water oxidation conditions, Ir5+ remains unreactive whereas O-1 is consumed at a time constant commensurate with the reaction rate, indicating that O-1 drives the oxygen evolution reaction. These findings demonstrate the necessity of using multiple operando techniques to gain a unified understanding of the evolution of oxidation states and active sites with potential for water oxidation on oxide catalysts.

Hybrid lead halide perovskites exhibit a delicate interplay between average crystallographic symmetry, local structural disorder and A-site orientational dynamics, giving rise to unusual vibrational and electronic behavior. Here, we combine large-scale molecular dynamics with a density-functional-theory-accurate machine learning force field to resolve the structural dynamics of perovskites across mesoscopic length scales. In formamidinium lead iodide (FAPbI3), we identify a high-temperature α phase with dynamic local order and correlated tilt nanodomains, an ordered γ phase with long-range a+a+a+ tilt coherence, and, below ∼100 K, a history-dependent γ' state consisting of locally γ-like nanoscale regions separated by sharp twin-like boundaries. This low-temperature disordered state is not a distinct bulk polymorph, but a kinetically arrested metastable twin-domain network selected by the interplay between shallow tilt energetics and slowing FA reorientation. This picture is supported by our low-temperature X-ray diffuse scattering measurements and accounts for the broadened low-energy vibrational response found in the simulations. Furthermore, this unique structural landscape imprints a spatially varying electronic disorder with implications for macroscopic optoelectronic properties, reflected in substantial band-edge broadening retained at low temperature. Our results reconcile the debated low-temperature behavior of FAPbI3 in terms of competition between ordered and arrested structural states, and more broadly identify molecular reorientation as a kinetic selector of metastable framework topology in soft molecular crystals, placing thermal history on equal footing with composition as a determinant of structural and optoelectronic properties.

This paper presents the third version of the Efficiency Tables compiling the record efficiencies of materials considered as emerging inorganic absorbers for photovoltaic (PV) technologies. The materials collected in these Tables are selected based on their progress in recent years, and their demonstrated potential as future PV absorbers. The first part of the paper consists of an overview of the current status of emerging PV materials. The second section details the inclusion criteria used for the different technologies presented in the paper, the verification means used by the authors, and recommendations for measurement best practices. The third part compiles the highest world-class certified solar cell efficiencies, and the highest non-certified cases (some independently confirmed). This section includes a brief state-of-the-art for the different classes of materials defined in the paper. The final part summarizes the main conclusions of this third version of Emerging Inorganic Solar Cell Efficiency Tables, highlighting the most relevant progress published in the last two years.

Pnictogen chalcohalides (MChX) have recently emerged as promising nontoxic and environmentally friendly photovoltaic absorbers, combining strong light absorption coefficients with favorable low-temperature synthesis conditions. Despite these advantages and reported optimized morphologies, device efficiencies remain below 10%, far from their ideal radiative limit. To uncover the origin of these performance losses, we present a systematic and fully consistent first-principles investigation of the defect chemistry across the Bi-based chalcohalide family. Our results reveal a complex defect landscape dominated by chalcogen vacancies of low formation energy, which act as deep nonradiative recombination centers. Despite their moderate charge-carrier capture coefficients, the high equilibrium concentrations of these defects reduce the theoretical maximum efficiencies by 6% in BiSeI and by 10% in BiSeBr. In contrast, sulfur vacancies in BiSI and BiSBr are comparatively benign, presenting smaller capture coefficients due to weaker electron-phonon coupling. Interestingly, despite its huge nonradiative charge-carrier recombination rate, BiSeI presents the best conversion efficiency among all four compounds owing to its most suitable bandgap for outdoor photovoltaic applications. Our findings identify defect chemistry as a critical bottleneck in MChX solar cells and propose chalcogen-rich synthesis conditions and targeted anion substitutions as effective strategies for mitigation of detrimental vacancies.

Understanding how ions interact with electrode surfaces at the molecular level is essential for improving the performance of energy storage devices and electrocatalysts. However, progress has been limited by the structural disorder and poorly defined surface chemistries of conventional carbon-based electrodes. In this work, we use layered metal-organic frameworks (MOFs) as model systems to investigate how different functional groups influence electric double-layer capacitance. We find that electrodes with deprotonated M-O and M-S groups exhibit significantly enhanced capacities with alkali metal cations, most notably Li+, compared to tetraethylammonium (TEA+), while no enhancement is observed for MOFs with protonated M-NH groups. The largest capacity increase is seen for MOF electrodes with metal-hydroxy linkages paired with Li+ electrolytes, which we attribute to strong Li-O interactions and improved charge screening. This mechanism is supported by solid-state nuclear magnetic resonance spectroscopy experiments and molecular simulations, which reveal specific Li+ binding at oxygen-rich sites, while operando X-ray techniques rule out cation intercalation as a contributing factor. Overall, these results highlight a chemically tunable strategy for enhancing charge storage in porous electrodes and offer new insights into how surface functionality impacts electric double-layer behavior.

Two-dimensional electrically conductive metal-organic frameworks (MOFs) are emerging as promising materials for electrochemical energy storage, electrocatalysis, and electrochemical separations, but their stability under extended operation remains poorly understood. Here, we systematically investigate three isostructural frameworks: Cu3(HHTP)2 (HHTP = 2,3,6,7,10,11hexahydroxytriphenylene), Ni3(HITP)2 (HITP = 2,3,6,7,10,11-hexaiminotriphenylene), and Ni3(HHTP)2, to disentangle the roles of metal centre and ligand chemistry in governing electrochemical degradation in electrochemical double-layer capacitors. Despite their nearly identical layered hexagonal structures, the three MOFs diverge in their electrochemical durability. We find that electrochemical stability follows the order Ni3(HITP)2 > Ni3(HHTP)2 > Cu 3 (HHTP) 2 , revealing that both metal identity and ligand chemistry govern durability. Threeelectrode measurements revealed that Cu₃(HHTP)₂ degraded predominantly at the positive electrode, whereas Ni 3 (HITP) 2 showed comparable degradation at both electrodes. Ex situ analyses reveal distinct degradation pathways: Cu3(HHTP)2 undergoes ligand oxidation coupled with metal redox, involving HHTP oxidation, Cu 2+ reduction to Cu⁺, and Cu⁺ dissolution, leading to substantial Cu loss at the electrode surface. Cu3(HHTP)2 also exhibits two configuration-dependent pathways: Cu⁺ dissolution in electrolyte-rich conditions and Cu-O coordination distortion in electrolyte-limited cells. In contrast, Ni3(HITP)2 degrades more gradually through ligand-based redox activity and accompanying ligand-field distortions without metal redox, consistent with negligible Ni leaching (<2%). Ni3(HHTP)2 provided an 2 essential intermediate case showing stable octahedral Ni 2+ sites with only ~0.4 eV pre-edge shifts in Ni K-edge X-ray absorption spectroscopy even at full degradation, verifying ligandonly deterioration while Ni 2+ remains redox-inert. Overall, these results reveal that degradation in conductive MOFs is governed by the coupled effects of metal identity and ligand chemistry. Cu-based frameworks are prone to multiple degradation pathways involving ligand and metal redox, whereas Ni analogues degrade via ligand-centred processes without metal redox. These mechanistic insights highlight routes to designing MOF structures with enhanced durability, advancing their prospects as next-generation materials for electrochemical applications.

Machine learning interatomic potentials (MLIPs) can now reproduce the energy, forces and stresses of bulk materials with high accuracy compared to first-principles calculations. The description of imperfections, where coordination environments and electron counts deviate from those found in pristine reference structures, remains a challenge. We find that the current generation of foundation MLIPs do not describe the defect physics of the semiconductor Sb2Se3. We introduce global defect charge embeddings that distinguish the bonding characteristics of different charge states. We further employ a multi-fidelity approach that combines low-cost (semi-local exchange-correlation functional) reference data with high-quality (non-local hybrid functional) energies and forces that describe well the subtleties of the defect energy landscape. The resulting defect-capable force fields can find stable structural configurations and predict charge-transition levels in quantitative agreement with direct quantum mechanical calculations, at a fraction of the computational cost.

Abstract Metal-organic frameworks (MOFs) are modular materials that are chemically versatile, offering vast design possibilities through the combination of metal clusters, organic ligands, and framework topologies. Traditionally utilized in gas storage and separation, the high porosity and tunable properties of MOFs position them as attractive candidates for (photo)electrochemical applications, including batteries, fuel cells, and solar-driven fuel conversion. We present a data-driven workflow integrating hybrid quantum chemical calculations with a multimodal AI model (MOFTransformer) to efficiently predict oxidation and reduction potentials for over 270,000 known and hypothetical MOFs. Through high-throughput screening, we identify MOFs with band-edge alignments that are thermodynamically compatible with coupled CO2-to-CO reduction and oxygen evolution, highlighting structural motifs and electronic properties for further investigation with significantly reduced computational effort compared to traditional approaches.