Periodic Reporting for period 1 - XYMOF (Artificial Intelligence meets Material Genome: intelligent design of new MOFs for xylene separation challenge.)
Reporting period: 2022-09-01 to 2024-08-31
• What is the problem being addressed?
This project addresses the major industrial challenge of efficient chemical separation: responsible for 10–15% of global energy use. In particular, separating xylene isomers (p-, o-, and m-xylene) is difficult due to their similar properties. The core issue is understanding how to achieve efficient separation through computational modelling and the development of new porous materials.
• Why is it important for society?
Improving xylene separation can reduce industrial energy consumption and carbon emissions, contributing to sustainability goals. Para-xylene, a key isomer, is widely used to produce plastic bottles and other everyday items. Enhancing this process supports cleaner production, less environmental harm, and improved societal well-being.
• What are the overall objectives?
The main goal is to develop computer-based design tools, drawing on the Material Genome concept, molecular modeling, and AI, to create new metal-organic frameworks (MOFs) capable of energy-efficient separation of liquid xylene isomers. The project aims to achieve 99.9% recovery of para-xylene.
Scientific Objective 1: Evaluating MOFs for Xylene Separation
Using high-throughput simulations, the project screens thousands of real and hypothetical MOFs from databases like CoRE MOF (~15,000 entries). The goal is to assess liquid-phase xylene separation performance and create a training dataset linking structural features to separation ability.
Scientific Objective 2: Understanding What Makes MOFs Effective
Multivariate data analysis (MDA) is applied to identify patterns between MOF structure and separation efficiency. These insights lead to design principles that guide the creation of improved MOFs. Top-performing MOFs are then re-tested to validate predictions.
Scientific Objective 3: Designing Better MOFs with AI
The final step uses AI and machine learning to generate novel MOFs based on structure-performance data from earlier stages. These next-generation materials are optimized specifically for highly efficient xylene separation, expanding the frontiers of AI-driven materials design.
These models are being used to generate novel MOFs that go beyond what currently exists, pushing the boundaries of how AI can contribute to materials design. The ultimate aim is to create next-generation materials specifically tailored for efficient and effective xylene separation.
This project addresses the major industrial challenge of efficient chemical separation: responsible for 10–15% of global energy use. In particular, separating xylene isomers (p-, o-, and m-xylene) is difficult due to their similar properties. The core issue is understanding how to achieve efficient separation through computational modelling and the development of new porous materials.
• Why is it important for society?
Improving xylene separation can reduce industrial energy consumption and carbon emissions, contributing to sustainability goals. Para-xylene, a key isomer, is widely used to produce plastic bottles and other everyday items. Enhancing this process supports cleaner production, less environmental harm, and improved societal well-being.
• What are the overall objectives?
The main goal is to develop computer-based design tools, drawing on the Material Genome concept, molecular modeling, and AI, to create new metal-organic frameworks (MOFs) capable of energy-efficient separation of liquid xylene isomers. The project aims to achieve 99.9% recovery of para-xylene.
Scientific Objective 1: Evaluating MOFs for Xylene Separation
Using high-throughput simulations, the project screens thousands of real and hypothetical MOFs from databases like CoRE MOF (~15,000 entries). The goal is to assess liquid-phase xylene separation performance and create a training dataset linking structural features to separation ability.
Scientific Objective 2: Understanding What Makes MOFs Effective
Multivariate data analysis (MDA) is applied to identify patterns between MOF structure and separation efficiency. These insights lead to design principles that guide the creation of improved MOFs. Top-performing MOFs are then re-tested to validate predictions.
Scientific Objective 3: Designing Better MOFs with AI
The final step uses AI and machine learning to generate novel MOFs based on structure-performance data from earlier stages. These next-generation materials are optimized specifically for highly efficient xylene separation, expanding the frontiers of AI-driven materials design.
These models are being used to generate novel MOFs that go beyond what currently exists, pushing the boundaries of how AI can contribute to materials design. The ultimate aim is to create next-generation materials specifically tailored for efficient and effective xylene separation.
From the beginning, the project has focused on analyzing MOFs for separating xylene isomers—a critical task for the chemical industry.
In the first phase (WP1), an existing MOF database was screened using a newly developed Python-based software package. This automated workflow, designed for HPC systems, enabled large-scale simulations to evaluate MOFs’ interactions with xylene mixtures by measuring adsorption, diffusion, and selectivity.
The simulations combined classical molecular dynamics, grand-canonical Monte Carlo, and Density Functional Theory to explore MOF flexibility during separation. This helped identify key structural features, such as pore size, diffusion paths, and adsorption energies, that influence performance. These insights were used to build a training dataset for the next phase (WP2).
In WP2, statistical methods including Principal Component Analysis (PCA), Cluster Analysis (CA), and Multiple Regression (MR) were applied to reveal complex structure-performance relationships. These techniques enabled the design of better MOFs using faster, simpler models, reducing reliance on expensive simulations.
Results have been shared through numerous invited talks and presentations, including:
- Invited Lecture: Suyetin M. (2025), Prediction of Adsorption and Separation Properties of Metal-Organic Frameworks: A Decade of Advancements from Molecular Simulations to Machine Learning Approaches. / Friedrich Schiller University Jena (Friedrich-Schiller-Universität Jena), Germany.
- Invited Lecture: Suyetin M. (2025), Multi-scale simulations of material properties. / Martin Luther University Halle-Wittenberg. (Martin-Luther-Universität Halle-Wittenberg), Germany.
- Suyetin M. (2024), Exploring the Limits of Zr-MOFs in Adsorption Pumps: A computational Study on their Potential for Cooling and Heating Applications / MOF2024, Singapore.
- Invited Lecture: Suyetin M. (2024), Current atomistic approaches for simulating material properties. / Helmholtz Association - Jülich Research Center, Germany.
- Invited Lecture: Suyetin M. (2024), Atomistic methods for modelling material properties. / Swansea University, UK.
- Suyetin M. (2023), Small Molecules Separation via Molecular Dynamics Simulations in Metal - Organic Frameworks / 9th bwHPC Symposium: Computational Chemistry and Materials Science, The University of Mannheim, Germany.
- Invited Lecture: Suyetin M. (2023) Investigating the separation of small molecules by utilizing Molecular Dynamics Simulations on Model of Flexible Metal-Organic Frameworks. / CECAM Flagship workshop: Fluids in porous materials: from fundamental physics to engineering applications, EPFL, Switzerland.
- Invited Lecture: Suyetin M. (2023) Simulation approaches in simulations of molecules separation / CECAM Flagship Workshop: Computational methods for modelling bionano interactions and nanomaterials functionality/The effective implicit surface model (EISM) for predicting and understanding peptide-surface interactions. University College Dublin, Ireland.
- Invited Lecture: Suyetin M. (2023) Application of simulation approaches for MOFs’ research. / School of Chemistry, the University of Manchester, UK.
In the first phase (WP1), an existing MOF database was screened using a newly developed Python-based software package. This automated workflow, designed for HPC systems, enabled large-scale simulations to evaluate MOFs’ interactions with xylene mixtures by measuring adsorption, diffusion, and selectivity.
The simulations combined classical molecular dynamics, grand-canonical Monte Carlo, and Density Functional Theory to explore MOF flexibility during separation. This helped identify key structural features, such as pore size, diffusion paths, and adsorption energies, that influence performance. These insights were used to build a training dataset for the next phase (WP2).
In WP2, statistical methods including Principal Component Analysis (PCA), Cluster Analysis (CA), and Multiple Regression (MR) were applied to reveal complex structure-performance relationships. These techniques enabled the design of better MOFs using faster, simpler models, reducing reliance on expensive simulations.
Results have been shared through numerous invited talks and presentations, including:
- Invited Lecture: Suyetin M. (2025), Prediction of Adsorption and Separation Properties of Metal-Organic Frameworks: A Decade of Advancements from Molecular Simulations to Machine Learning Approaches. / Friedrich Schiller University Jena (Friedrich-Schiller-Universität Jena), Germany.
- Invited Lecture: Suyetin M. (2025), Multi-scale simulations of material properties. / Martin Luther University Halle-Wittenberg. (Martin-Luther-Universität Halle-Wittenberg), Germany.
- Suyetin M. (2024), Exploring the Limits of Zr-MOFs in Adsorption Pumps: A computational Study on their Potential for Cooling and Heating Applications / MOF2024, Singapore.
- Invited Lecture: Suyetin M. (2024), Current atomistic approaches for simulating material properties. / Helmholtz Association - Jülich Research Center, Germany.
- Invited Lecture: Suyetin M. (2024), Atomistic methods for modelling material properties. / Swansea University, UK.
- Suyetin M. (2023), Small Molecules Separation via Molecular Dynamics Simulations in Metal - Organic Frameworks / 9th bwHPC Symposium: Computational Chemistry and Materials Science, The University of Mannheim, Germany.
- Invited Lecture: Suyetin M. (2023) Investigating the separation of small molecules by utilizing Molecular Dynamics Simulations on Model of Flexible Metal-Organic Frameworks. / CECAM Flagship workshop: Fluids in porous materials: from fundamental physics to engineering applications, EPFL, Switzerland.
- Invited Lecture: Suyetin M. (2023) Simulation approaches in simulations of molecules separation / CECAM Flagship Workshop: Computational methods for modelling bionano interactions and nanomaterials functionality/The effective implicit surface model (EISM) for predicting and understanding peptide-surface interactions. University College Dublin, Ireland.
- Invited Lecture: Suyetin M. (2023) Application of simulation approaches for MOFs’ research. / School of Chemistry, the University of Manchester, UK.
This project addresses a key challenge in chemical engineering: efficiently separating xylene isomers, which have nearly identical properties. Traditional industrial methods are highly energy-intensive and contribute significantly to carbon emissions.
What sets this project apart is its use of advanced computational techniques, including molecular dynamics, Monte Carlo simulations, and DFT, to explore how MOF flexibility affects xylene separation.
Three main goals have guided the project so far:
• MOF screening: Thousands of MOFs have been assessed through high-throughput simulations, identifying promising candidates for liquid-phase xylene separation.
• Design insights: Data analysis has revealed how MOF structure affects separation performance, resulting in design rules for improved materials.
• Expected outcome: By project end, the goal is to deliver MOFs that recover at least 99.9% of para-xylene—surpassing current benchmarks.
Potential impacts:
Scientifically, this work advances material design by replacing trial-and-error with powerful modelling tools, offering a new standard for developing porous materials. Environmentally, more efficient xylene separation could reduce industrial energy use and CO2 emissions, supporting climate goals. Since para-xylene is essential in everyday products like plastic bottles and clothing, cleaner production benefits both industry and society.
Long-term, the modelling approaches developed here could be applied to challenges in energy, environment, and health, amplifying the project’s broader impact.
What sets this project apart is its use of advanced computational techniques, including molecular dynamics, Monte Carlo simulations, and DFT, to explore how MOF flexibility affects xylene separation.
Three main goals have guided the project so far:
• MOF screening: Thousands of MOFs have been assessed through high-throughput simulations, identifying promising candidates for liquid-phase xylene separation.
• Design insights: Data analysis has revealed how MOF structure affects separation performance, resulting in design rules for improved materials.
• Expected outcome: By project end, the goal is to deliver MOFs that recover at least 99.9% of para-xylene—surpassing current benchmarks.
Potential impacts:
Scientifically, this work advances material design by replacing trial-and-error with powerful modelling tools, offering a new standard for developing porous materials. Environmentally, more efficient xylene separation could reduce industrial energy use and CO2 emissions, supporting climate goals. Since para-xylene is essential in everyday products like plastic bottles and clothing, cleaner production benefits both industry and society.
Long-term, the modelling approaches developed here could be applied to challenges in energy, environment, and health, amplifying the project’s broader impact.