Atomistic simulations of matter, especially those that leverage first-principles (ab initio) electronic structure theory, provide a microscopic view of the world, underpinning much of our understanding of chemistry and materials science. Over the last decade or so, machine-learned force fields have transformed atomistic modeling by enabling simulations of ab initio quality over unprecedented time and length scales. However, early machine-learning (ML) force fields have largely been limited by (i) the substantial computational and human effort required to develop and validate potentials for each particular system of interest and (ii) a general lack of transferability from one chemical system to the next. Here, we show that it is possible to create a general-purpose atomistic ML model, trained on a public dataset of moderate size, that is capable of running stable molecular dynamics for a wide range of molecules and materials. We demonstrate the power of the MACE-MP-0 model—and its qualitative and at times quantitative accuracy—on a diverse set of problems in the physical sciences, including properties of solids, liquids, gases, chemical reactions, interfaces, and even the dynamics of a small protein. The model can be applied out of the box as a starting or “foundation” model for any atomistic system of interest and, when desired, can be fine-tuned on just a handful of application-specific data points to reach ab initio accuracy. Establishing that a stable force-field model can cover almost all materials changes atomistic modeling in a fundamental way: experienced users obtain reliable results much faster, and beginners face a lower barrier to entry. Foundation models thus represent a step toward democratizing the revolution in atomic-scale modeling that has been brought about by ML force fields.

Palladium-catalysed cross-coupling reactions, central tools in fine-chemical synthesis, predominantly employ soluble metal complexes despite recognized challenges with product purification and catalyst reusability^1,2,3. Attempts to tether these homogeneous catalysts on insoluble carriers have been thwarted by suboptimal stability, which leads to a progressively worsening performance due to metal leaching or clustering⁴. The alternative application of supported Pd nanoparticles has faced limitations because of insufficient activity under the mild conditions required to avoid thermal degradation of the substrates or products. Single-atom heterogeneous catalysts lie at the frontier^{5,6,7,8,9,10,11,12,13,14,15,16,17,18}. Here, we show that the Pd atoms anchored on exfoliated graphitic carbon nitride (Pd-ECN) capture the advantages of both worlds, as they comprise a solid catalyst that matches the high chemoselectivity and broad functional group tolerance of state-of-the-art homogeneous catalysts for Suzuki couplings, and also demonstrate a robust stability in flow. The adaptive coordination environment within the macroheterocycles of ECN facilitates each catalytic step. The findings illustrate the exciting opportunities presented by nanostructuring single atoms in solid hosts for catalytic processes that remain difficult to heterogenize.

Carbon nitrides integrating macroheterocycles offer unique potential as hosts for stabilizing metal atoms due to their rich electronic structure. To date, only graphitic heptazine-based polymers have been studied. Here, we demonstrate that palladium atoms can be effectively isolated on other carbon nitride scaffolds including linear melem oligomers and poly(triazine/heptazine imides). Increased metal uptake was linked to the larger cavity size and the presence of chloride ions in the polyimide structures. Changing the host structure leads to significant variation in the average oxidation state of the metal, which can be tuned by exchange of the ionic species as evidenced by X-ray photoelectron spectroscopy and supported by density functional theory. Evaluation in the semi-hydrogenation of 2-methyl-3-butyn-2-ol reveals an inverse correlation between the activity and the degree of oxidation of palladium, with oligomers exhibiting the highest activity. These findings provide new mechanistic insights into the influence of the carbon nitride structure on metal stabilization.

Single‐atom heterogeneous catalysts with well‐defined architectures are promising for deriving structure–performance relationships, but the challenge lies in finely tuning the structural and electronic properties of the metal. To tackle this point, a new approach based on the surface diffusion of gold atoms on different cavities of N‐doped carbon is presented. By controlling the activation temperature, the coordination neighbors (Cl, O, N) and the oxidation state of the metal can be tailored. Semi‐hydrogenation of various alkynes on the single‐atom gold catalysts displays substrate‐dependent catalytic responses; structure insensitive for alkynols with γ‐OH and unfunctionalized alkynes, and sensitive for alkynols with α‐OH. Density functional theory links the sensitivity for alkynols to the strong interaction between the substrate and specific gold‐cavity ensembles, mimicking a molecular recognition pattern that allows to identify the cavity site and to enhance the catalytic activity.