Hyunsoo Park
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Hyunsoo Park has a total of 70 co-authors. Here are the top co-authors by number of shared publications:
- Aron Walsh
Department of Materials, Imperial College London, London, United Kingdom
15 shared publications - Jihan Kim
Department of Chemical and Biomolecular Engineering Korea Advanced Institute of Science and Technology (KAIST) Daejeon Republic of Korea
14 shared publications - Anthony Onwuli
Imperial College London
5 shared publications - Kinga Oliwia Mastej
Imperial College London
5 shared publications - Yeonghun Kang
Vector Institute for Artificial Intelligence
4 shared publications - Berend Smit
École Polytechnique Fédérale de Lausanne (EPFL) Rue de l’Industrie 17
3 shared publications
Latest Publications
- Open Access
- 19 October 2026
- Published by The Open Journal in Journal of Open Source Software
- Vol. 11 (126) , 11015
SMACT v4: Enhanced Screening and Prediction Tools for Computational Materials Design
- 19 October 2026
- Published by The Open Journal in Journal of Open Source Software
- Vol. 11 (126) , 11015
Open AccessSMACT v4: Enhanced Screening and Prediction Tools for Computational Materials Design
- Open Access
- 16 July 2026
- Published by American Chemical Society (ACS) in ACS Energy Letters
- Vol. 11 (8) , 5311-5317
Data-Driven Design of Metal-Organic Frameworks for Photoelectrochemical Reactions
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.
- 16 July 2026
- Published by American Chemical Society (ACS) in ACS Energy Letters
- Vol. 11 (8) , 5311-5317
Open AccessData-Driven Design of Metal-Organic Frameworks for Photoelectrochemical Reactions
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.
- Open Access
- 6 July 2026
- Published by Springer Nature in Nature Machine Intelligence
- Vol. 8 (7) , 1087-1099
Guiding generative models to uncover diverse and novel crystals via reinforcement learning
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.
- 6 July 2026
- Published by Springer Nature in Nature Machine Intelligence
- Vol. 8 (7) , 1087-1099
Open AccessGuiding generative models to uncover diverse and novel crystals via reinforcement learning
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.
- Open Access
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (3) , 035064
Continuous SUN (stable, unique, and novel) metric for generative modeling of inorganic crystals
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.
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (3) , 035064
Open AccessContinuous SUN (stable, unique, and novel) metric for generative modeling of inorganic crystals
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.
- 6 May 2026
- Published by Springer Nature in Nature Synthesis
- Vol. 5 (5) , 651-659
Closing the synthesis gap in computational materials design
- 6 May 2026
- Published by Springer Nature in Nature Synthesis
- Vol. 5 (5) , 651-659
Closing the synthesis gap in computational materials design
- Open Access
Guiding Generative Models to Uncover Diverse and Novel Crystals via Reinforcement Learning
Discovering functional crystalline materials entails navigating an immense combinatorial design space. While recent advances in generative artificial intelligence have enabled the sampling of chemically plausible compositions and structures, a fundamental challenge remains: the objective misalignment between likelihood-based sampling in generative modelling and targeted focus on underexplored regions where novel compounds reside. Here, we introduce a reinforcement learning framework that guides latent denoising diffusion models toward diverse and novel, yet thermodynamically viable crystalline compounds. Our approach integrates group relative policy optimisation 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 scientific discovery applications of generative models.
Open AccessGuiding Generative Models to Uncover Diverse and Novel Crystals via Reinforcement Learning
Discovering functional crystalline materials entails navigating an immense combinatorial design space. While recent advances in generative artificial intelligence have enabled the sampling of chemically plausible compositions and structures, a fundamental challenge remains: the objective misalignment between likelihood-based sampling in generative modelling and targeted focus on underexplored regions where novel compounds reside. Here, we introduce a reinforcement learning framework that guides latent denoising diffusion models toward diverse and novel, yet thermodynamically viable crystalline compounds. Our approach integrates group relative policy optimisation 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 scientific discovery applications of generative models.
- Open Access
- 10 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Data-driven exploration of AB2X4 (X = O, S, Se, Te) spinel chemical space
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.
- 10 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessData-driven exploration of AB2X4 (X = O, S, Se, Te) spinel chemical space
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.
- Open Access
- 15 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Closing the synthesis gap in computational materials design
Data-driven strategies are reshaping computational materials design by accelerating the prediction of novel 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. 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 prioritisation. Looking forward, the development of more robust synthesisability metrics, synthesis planning tools, and agentic workflows integrating experimental feedback will narrow the divide between virtual screening and real-world materials realisation.
- 15 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessClosing the synthesis gap in computational materials design
Data-driven strategies are reshaping computational materials design by accelerating the prediction of novel 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. 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 prioritisation. Looking forward, the development of more robust synthesisability metrics, synthesis planning tools, and agentic workflows integrating experimental feedback will narrow the divide between virtual screening and real-world materials realisation.
- Open Access
- 6 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Unifying Design for Homogeneous and Heterogeneous Catalysis via d-Band Center Alignment
A longstanding challenge in catalysis is the absence of unifying principles that couple the molecular-level reactivity of homogeneous catalysts with the durability and separability inherent to heterogeneous catalysts. Here, we present a computation-guided framework for the rational design of heterogeneous Rh-P nanoparticles that emulate the catalytic properties of homogeneous catalysts. By employing the d-band center as a transferable electronic descriptor, we align the electronic structure of Rh–P nanoparticles with that of benchmark Rh–phosphine complexes, enabling predictive control over hydroformylation activity. The workflow integrates machine learning-accelerated molecular dynamics, density functional theory, and experimental validation to screen a compositional library of Rh–P phases. A strong quantitative correlation is established between the deviation in d-band center and catalytic activity (R² = 0.994). Experimental evaluation reveals that Rh₃P, identified as the optimal composition through electronic structure matching, exhibits superior catalytic activity than the previously reported most active phase. In particular, Rh₃P achieved a reaction rate of 13,357 h⁻¹, representing a 25% increase over the state-of-the-art RhP nanoparticle system. This work establishes a generalizable framework for electronically guided catalyst design at the molecular level.
- 6 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessUnifying Design for Homogeneous and Heterogeneous Catalysis via d-Band Center Alignment
A longstanding challenge in catalysis is the absence of unifying principles that couple the molecular-level reactivity of homogeneous catalysts with the durability and separability inherent to heterogeneous catalysts. Here, we present a computation-guided framework for the rational design of heterogeneous Rh-P nanoparticles that emulate the catalytic properties of homogeneous catalysts. By employing the d-band center as a transferable electronic descriptor, we align the electronic structure of Rh–P nanoparticles with that of benchmark Rh–phosphine complexes, enabling predictive control over hydroformylation activity. The workflow integrates machine learning-accelerated molecular dynamics, density functional theory, and experimental validation to screen a compositional library of Rh–P phases. A strong quantitative correlation is established between the deviation in d-band center and catalytic activity (R² = 0.994). Experimental evaluation reveals that Rh₃P, identified as the optimal composition through electronic structure matching, exhibits superior catalytic activity than the previously reported most active phase. In particular, Rh₃P achieved a reaction rate of 13,357 h⁻¹, representing a 25% increase over the state-of-the-art RhP nanoparticle system. This work establishes a generalizable framework for electronically guided catalyst design at the molecular level.
- Top Cited Papers
Accelerating CO2 direct air capture screening for metal-organic frameworks with a transferable machine learning force field