Berend Smit

Copper nanoparticles (Cu NPs) have a broad applicability, yet their synthesis is sensitive to subtle changes in reaction parameters. This sensitivity, combined with the time- and resource-intensive nature of experimental optimization, poses a major challenge in achieving reproducible and size-controlled synthesis. While machine learning (ML) shows promise in materials research, its application is often limited by scarcity of large high-quality experimental data sets. This study explores ML to predict the DLS-derived hydrodynamic diameter of Cu NPs from microwave-assisted polyol synthesis using a small data set of 25 in-house performed syntheses. Latin hypercube sampling is used to efficiently cover the parameter space while creating the experimental data set. Ensemble regression models successfully predict hydrodynamic diameters with good predictive performance given the limited data set. Since quantitative regression requires a unique DLS-derived hydrodynamic diameter, the regression model is restricted to monomodal DLS distributions, while a complementary classification model identifies synthesis conditions for which quantitative prediction is applicable. Using equivalent out-of-sample validation (OOB and LOOCV), the ML and DoE models showed comparable generalization (MAE = 40.92 and 40.46 nm, respectively). The final ensemble model achieved an R2 of 0.74, an MAE of 23.81 nm, compared to an R2 of 0.60, an MAE of 33.54 nm for the DoE model, while retaining the complete synthesis parameter space, making it better suited for synthesis guidance. Additionally, classification models using both random forests and large language models (LLMs) are evaluated to distinguish between large and small particles. These classification models exhibited only modest predictive performance, indicating that this small data set is insufficient to fully exploit the capabilities of complex LLMs. Overall, this study demonstrates that carefully curated small data sets, paired with robust classical ML, can effectively support the synthesis of Cu NPs and highlights that for lab-scale studies, complex models like LLMs may offer limited benefits. The validation experiments further indicate the potential of ML guided synthesis to improve the efficiency of experimental optimization and reduce resource consumption.

Metal–organic frameworks (MOFs) are promising materials for adsorption and separation, but accurately predicting water adsorption remains a major challenge in molecular simulation. Classical force fields, such as UFF, often fail to capture the strong, directional hydrogen-bonding interactions between water and MOFs, while high-throughput density functional theory (DFT) calculations are computationally prohibitive. Here, we develop a transferable machine-learning potential (MLP) for water adsorption in Al-based MOFs by fine-tuning the pretrained MACE foundation model. Using a data set of 420 Al-MOFs with diverse chemical environments, we show that accurate prediction of water adsorption thermodynamics requires not only an improved description of water–framework interactions but also explicit treatment of framework flexibility. The resulting model reproduces experimental heats of adsorption and Henry coefficients for a series of benchmark Al-MOFs and achieves DFT-level accuracy across more than 400 Al-MOFs. Analysis of MIL-160 establishes practical accuracy requirements of approximately 10 kJ mol–1 in total energies and 40 meV Å–1 in atomic forces for reliable adsorption thermodynamics. Compared with DFT-level predictions, UFF underestimates water adsorption enthalpies by more than 10 kJ mol–1 for 74% of the materials studied. In addition, the MLP identifies lower-energy framework configurations than previously reported DFT-optimized structures, highlighting the importance of enhanced configurational sampling. These results demonstrate that foundation-model-based machine-learning potentials can enable DFT-accurate, high-throughput screening of water adsorption in flexible MOFs.

Metal-organic frameworks (MOFs) have potential for light-harvesting applications owing to their tunable optoelectronic properties via building block selection. The presence of bound solvent in MOFs is known to impact gas adsorption properties, but little is known about their effect on optical properties. Here, we investigate how the presence of bound water can modulate the optical properties of a reported MOF photocatalyst, SU-101. Using the GW approximation and the ab initio Bethe-Salpeter equation (BSE) and treating exciton-phonon coupling with a finite difference method, we show that exciton-phonon interactions in SU-101 are strong and highly sensitive to the presence of bound water molecules. Specifically, we find that the presence of bound water is associated with a stronger localization of excitons due to phonons. Guided by our calculations, we synthesize SU-101 and study this MOF under different water loadings. We measure photoluminescence (PL) emission and UV-Vis absorption spectra of the MOFs, the latter in good agreement with our computational predictions. By considering the interplay of water loading, phonons, and excitons, our calculations rationalize measured UV-Vis absorption and PL spectra. Our work also highlights the pivotal role of exciton-phonon interactions and their sensitivity to hydration on the optoelectronic properties of MOFs with large pores and flexible organic linkers.

Aromatic volatile organic compounds (VOCs) are toxic air pollutants that pose serious health risks even at trace concentrations. Their nonpolar character makes the design of efficient sorbents particularly challenging, as adsorption is governed mainly by weak dispersion forces. Here, we identify pore geometry as an effective structural descriptor for discovering metal-organic frameworks (MOFs) capable of efficient trace-level VOC capture. Screening a diverse set of MOFs revealed that one-dimensional channels with rhombic or square cross sections enhance host-guest interactions and promote strong affinity for aromatic molecules. Guided by this principle and sustainability-by-design criteria, we identify MIP-211(Al) (Materials from the Institute of Porous Materials of Paris) as a top performer with among the highest toluene uptake of 4.7 mmol g^-1 at 0.0008 P/P₀ (31 ppm). This material combines excellent cycling stability, facile regeneration under vacuum, and scalable green synthesis. Synchrotron powder X-ray diffraction (SPXRD) and density functional theory (DFT) calculations confirm that rhombic pore geometry governs strong toluene affinity at trace amounts. Attenuated Total Reflectance Fourier Transform Infrared (ATR-FTIR) spectroscopy showed that MIP-211(Al) retains 60% of its performance up to 15% relative humidity. Applying pore geometry as a descriptor in a secondary screening identified a candidate with a toluene uptake of 3.6 mmol g^-1 at 0.0008 P/P₀ (31 ppm) at dry conditions, with potential applicability up to 70% relative humidity. This study demonstrates the importance of pore geometry as a nanoscale parameter for efficient trace aromatic VOC adsorption.

Herein, we report two heterometallic ultramicroporous metal–organic frameworks, MIP-212(Al/Cu) and MIP-212(Al/Zn) (MIP stands for Materials from Institute of Porous Materials of Paris), synthesized via a hard–soft acid–base design strategy. In these robust pyrazolate–carboxylate architectures, pyrazolates selectively coordinate Cu²⁺ or Zn²⁺, while carboxylates bind Al³⁺, generating chain-based inorganic building units built up from connected M²⁺-pyrazolate polyhedra and μ₂–OH-corner-shared AlO₆ octahedra, respectively. The resulting structures feature dual ultranarrow tunnel-like pores, one decorated with μ₂–OH groups. MIP-212(Al/Cu) combines pore confinement with Cu²⁺ open metal sites (OMS) to deliver benchmark-level CO₂ uptake at low pressure (2.30 mmol g^–1 at 0.15 bar, 298 K) and a CO₂/N₂ Ideal Adsorbed Solution Theory (IAST) selectivity of ∼30. However, the OMS also imparts marked hydrophilicity, diminishing CO₂ uptake under humid conditions. Markedly, replacing octahedral Cu²⁺ with tetrahedral Zn²⁺ centers in MIP-212(Al/Zn) suppresses OMS while preserving framework topology, resulting in significantly lower water affinity (up to ca. 4-fold reduction at 0.2 bar of H₂O) and superior CO₂ breakthrough performance at 50% RH. These findings demonstrate that metal coordination geometry is a powerful lever to modulate hydrophilicity and sorption behavior in MOFs, enabling the rational design of sorbents for efficient CO₂ capture under realistic, moisture-rich environments.

Metal–organic frameworks (MOFs) are promising materials for adsorption and separation, but accurately predicting water adsorption remains a major challenge in molecular simulation. Classical force fields, such as UFF, often fail to capture the strong, directional hydrogen-bonding interactions between water and MOFs, while high-throughput density functional theory (DFT) calculations are computationally prohibitive. Here, we develop a transferable machine-learning potential (MLP) for water adsorption in Al-based MOFs by fine-tuning the pretrained MACE foundation model. Using a dataset of 420 Al-MOFs with diverse chemical environments, we show that accurate prediction of water adsorption thermodynamics requires not only an improved description of water–framework interactions but also explicit treatment of framework flexibility. The resulting model reproduces experimental heats of adsorption and Henry coefficients for a series of benchmark Al-MOFs and achieves DFT-level accuracy across more than 400 Al-MOFs. Analysis of MIL-160 establishes practical accuracy requirements of approximately 10 kJ/mol in total energies and 40 meV/Å in atomic forces for reliable adsorption thermodynamics. Compared with DFT-level predictions, UFF underestimates water adsorption enthalpies by more than 3.5 kJ/mol for 91% of the materials studied. In addition, the MLP identifies lower-energy framework configurations than previously reported DFT-optimized structures, highlighting the importance of enhanced configurational sampling. These results demonstrate that foundation-model-based machine-learning potentials can enable DFT-accurate, high-throughput screening of water adsorption in flexible MOFs.

Metal-organic frameworks (MOFs) are porous materials with the potential for gas adsorption and separation technologies due to their tunable structural and chemical characteristics. However, simulating gas adsorption isotherms in MOFs with DFT-level accuracy remains a key challenge. In this work, we use MIL-120 as a case study and propose a comprehensive computational workflow to fine-tune a pre-trained MACE potential, enabling the creation of accurate machine-learning interatomic potentials tailored for frameworks with dynamic structural behavior. Using this ML potential to accelerate sampling, we uncover a strong coupling between CO₂ adsorption and the complex dynamic hydrogen-bond network on the MIL-120 pore surface. The presence of CO₂ induces local configurational rearrangements that reshape the pore environment. This insight provides a generalizable strategy for simulating adsorption in other flexible MOF systems.

Metal-organic frameworks (MOFs) have potential for light-harvesting applications owing to their tunable optoelectronic properties via building block selection. The presence of bound solvent in MOFs is known to impact gas adsorption properties, but little is known about their effect on optical properties. Here, we investigate how the presence of bound water can modulate the optical properties of a reported MOF photocatalyst, SU-101. With ab initio GW and Bethe-Salpether equation and treating exciton-phonon coupling with a finite difference method, we show that exciton-phonon interactions in SU-101 are strong and highly sensitive to the presence of bound water molecules. Specifically, we find that the presence of bound water is associated with a stronger localization of excitons due to phonons. Guided by our calculations, we synthesize SU-101 with different water loadings, and measure its photoluminescence emission and UV-Vis absorption spectra, the latter in excellent agreement with our computational predictions. By considering the interplay of water loading, phonons, and excitons, our calculations rationalize measured UV-Vis absorption and photoluminescence spectra. Our work also highlights the pivotal role of exciton-phonon interactions and their sensitivity to hydration on the optoelectronic properties of MOFs with large pores and flexible organic linkers.