Periodic Reporting for period 1 - M4MID (Toward Desirable Metal Organic Framework Mixed Matrix Materials through Machine learning-guided Interface Design)
Berichtszeitraum: 2023-06-01 bis 2025-05-31
Metal-organic frameworks (MOFs) are a class of porous crystalline materials that have gained significant attention for their potential in clean energy, gas separation, and storage technologies. Their tuneable structures, exceptionally high surface areas, and chemical flexibility make them ideal candidates for next-generation composite materials, especially when integrated with polymers. Mixed Matrix Materials based on Metal-Organic Frameworks (commonly referred to as M4s) are an emerging class of hybrid materials that combine porous crystalline fillers (MOFs) with synthetic polymers. These nanocomposites are being developed for a wide range of applications, including energy-efficient gas separation membranes, biomedical drug delivery systems, and sustainable, fire-resistant plastics. By blending the structural tunability of MOFs with the mechanical flexibility of polymers, M4s offer the potential to meet demanding performance requirements across multiple sectors.
Despite their promise, designing high-performance M4s remains a significant challenge. The interface between the MOF and polymer a zone where crystalline and amorphous chemistries meet is notoriously difficult to model and optimize. Poor compatibility at this interface can lead to mechanical failure, poor transport properties, and limited scalability. Addressing these interfacial issues through experimental trial-and-error is time-consuming and often cost-prohibitive.
The M4MID (Toward Mechanically-sound Metal Organic Framework Mixed Matrix Materials through Machine learning-guided Interface Design) project was launched to tackle this problem using computational tools. Its goal was to develop a multiscale simulation workflow capable of predicting the mechanical properties of M4s such as stiffness, elasticity, and thermal conductivity starting from atomistic descriptions of the materials and their interfaces. The project successfully implemented a modular workflow combining quantum-level calculations (density functional theory), classical molecular dynamics, and continuum mechanics (via FFT-based simulations), enabling property prediction across scales.
A key innovation of the project was the automated generation of realistic MOF-polymer interface models, including the structural optimization of MOF surfaces and equilibration with polymer chains. These simulations were validated against experimental data and used to quantify how interface structure affects the thermal transport and mechanical performance of the composite. This work lays the foundation for predictive material design, where simulation can guide the selection of components and structures before fabrication.
A significant progress was made in identifying the structural and chemical descriptors needed to support data-driven screening. Key features such as interfacial density gradients, pore size distribution, and van der Waals interaction energies were extracted from simulations and organized into descriptor libraries. These insights set the stage for future ML models that can predict mechanical properties and thermal conductivity with reduced reliance on computationally intensive simulations.
The expected impact of M4MID is far-reaching. By enabling faster, more accurate prediction of material behaviour, it contributes to EU priorities in sustainable materials development, green technologies, and digital transformation in science and engineering. The project’s simulation tools and methodologies support more efficient design of advanced composites for clean energy, filtration, and manufacturing applications. Moreover, the open and extensible nature of the workflow facilitates broader uptake by academic and industrial users alike.
In summary, M4MID advances the ability to model, understand, and ultimately design MOF-polymer composites with tailored mechanical and thermal properties. Its results mark an important step toward replacing empirical development cycles with simulation-guided innovation.
Despite their promise, designing high-performance M4s remains a significant challenge. The interface between the MOF and polymer a zone where crystalline and amorphous chemistries meet is notoriously difficult to model and optimize. Poor compatibility at this interface can lead to mechanical failure, poor transport properties, and limited scalability. Addressing these interfacial issues through experimental trial-and-error is time-consuming and often cost-prohibitive.
The M4MID (Toward Mechanically-sound Metal Organic Framework Mixed Matrix Materials through Machine learning-guided Interface Design) project was launched to tackle this problem using computational tools. Its goal was to develop a multiscale simulation workflow capable of predicting the mechanical properties of M4s such as stiffness, elasticity, and thermal conductivity starting from atomistic descriptions of the materials and their interfaces. The project successfully implemented a modular workflow combining quantum-level calculations (density functional theory), classical molecular dynamics, and continuum mechanics (via FFT-based simulations), enabling property prediction across scales.
A key innovation of the project was the automated generation of realistic MOF-polymer interface models, including the structural optimization of MOF surfaces and equilibration with polymer chains. These simulations were validated against experimental data and used to quantify how interface structure affects the thermal transport and mechanical performance of the composite. This work lays the foundation for predictive material design, where simulation can guide the selection of components and structures before fabrication.
A significant progress was made in identifying the structural and chemical descriptors needed to support data-driven screening. Key features such as interfacial density gradients, pore size distribution, and van der Waals interaction energies were extracted from simulations and organized into descriptor libraries. These insights set the stage for future ML models that can predict mechanical properties and thermal conductivity with reduced reliance on computationally intensive simulations.
The expected impact of M4MID is far-reaching. By enabling faster, more accurate prediction of material behaviour, it contributes to EU priorities in sustainable materials development, green technologies, and digital transformation in science and engineering. The project’s simulation tools and methodologies support more efficient design of advanced composites for clean energy, filtration, and manufacturing applications. Moreover, the open and extensible nature of the workflow facilitates broader uptake by academic and industrial users alike.
In summary, M4MID advances the ability to model, understand, and ultimately design MOF-polymer composites with tailored mechanical and thermal properties. Its results mark an important step toward replacing empirical development cycles with simulation-guided innovation.
The M4MID project successfully delivered a robust multiscale simulation framework to examine the mechanical performance of MOF-polymer composites (M4s). This included a series of integrated computational workflows that connect molecular-level models to continuum-scale mechanical property predictions, focusing on elastic behaviour and thermal conductivity.
Key technical and scientific activities:
1. Atomistic Interface Modeling:
More than 60 MOF-polymer interface systems were constructed, covering a broad matrix of MOF structures (e.g. UiO-66, ZIF-8, MIL-125, HKUST-1) and polymers (PEG, PVDF, PS, PMMA, PIM-1, PP). MOF surface slabs were modeled and relaxed using density functional theory (DFT) with CP2K, while polymers were generated and equilibrated using classical molecular dynamics (MD) in LAMMPS. The resulting interfaces were assembled and structurally optimized.
2. Force Field Validation and Simulation Accuracy:
Multiple force fields (UFF4MOF, DREIDING, AMBER, TraPPE) were tested and validated against available experimental data for individual polymers and MOFs. Careful tuning of non-bonded interactions ensured realistic interface behaviour during mechanical testing simulations.
3. Mechanical and Thermal Properties Examination:
Young’s modulus, bulk modulus, and shear modulus were computed for both neat materials and composites via tensile deformation simulations. Non-equilibrium molecular dynamics (NEMD) was used to determine thermal conductivity, and results were cross-compared with experimental and literature values to confirm reliability.
4. Continuum Mechanics Integration:
Atomistic simulation outputs were upscaled using the FFT-MAD framework, enabling continuum-level analysis of elastic response through Representative Volume Elements (RVEs). This multiscale integration allowed prediction of composite behaviour beyond the atomistic limit.
5. Automation of Simulation Pipelines:
A 21-step automated workflow was developed to manage the entire simulation process ranging from polymer chain generation and MOF surface cleaving to equilibration and mechanical testing. The pipeline was deployed across HPC systems and modularized for extensibility.
6. High-throughput Screening Capabilities:
Although full high-throughput coverage was constrained by resource availability, simulation prioritization strategies enabled processing of a representative set of MOF/polymer interfaces. The generated data now forms a structured database of interface structures and mechanical responses.
7. Descriptor Analysis and Machine Learning Foundations:
A key scientific achievement was the identification of interfacial descriptors that govern composite mechanical and thermal behaviour. These include interfacial density gradients, pore geometry, local stress profiles, and van der Waals interaction energies. While large-scale ML training was not completed, these insights establish the foundation for predictive screening models in follow-up work.
Outcomes:
• A validated multiscale simulation pipeline linking DFT, MD, and continuum modelling for M4 mechanical property prediction.
• A library of equilibrated MOF-polymer interface structures and their associated elastic and thermal properties.
• Descriptor sets suitable for ML model development, along with prototype models tested on limited data.
• Improved understanding of how MOF surface chemistry and geometry influence interfacial strength and composite performance.
• Transferable simulation workflows supporting further material screening and design beyond the original project scope.
These technical accomplishments significantly enhance the ability to design MOF-based composite materials with tailored mechanical properties and provide a reproducible simulation foundation for future high-throughput and data-driven approaches.
Key technical and scientific activities:
1. Atomistic Interface Modeling:
More than 60 MOF-polymer interface systems were constructed, covering a broad matrix of MOF structures (e.g. UiO-66, ZIF-8, MIL-125, HKUST-1) and polymers (PEG, PVDF, PS, PMMA, PIM-1, PP). MOF surface slabs were modeled and relaxed using density functional theory (DFT) with CP2K, while polymers were generated and equilibrated using classical molecular dynamics (MD) in LAMMPS. The resulting interfaces were assembled and structurally optimized.
2. Force Field Validation and Simulation Accuracy:
Multiple force fields (UFF4MOF, DREIDING, AMBER, TraPPE) were tested and validated against available experimental data for individual polymers and MOFs. Careful tuning of non-bonded interactions ensured realistic interface behaviour during mechanical testing simulations.
3. Mechanical and Thermal Properties Examination:
Young’s modulus, bulk modulus, and shear modulus were computed for both neat materials and composites via tensile deformation simulations. Non-equilibrium molecular dynamics (NEMD) was used to determine thermal conductivity, and results were cross-compared with experimental and literature values to confirm reliability.
4. Continuum Mechanics Integration:
Atomistic simulation outputs were upscaled using the FFT-MAD framework, enabling continuum-level analysis of elastic response through Representative Volume Elements (RVEs). This multiscale integration allowed prediction of composite behaviour beyond the atomistic limit.
5. Automation of Simulation Pipelines:
A 21-step automated workflow was developed to manage the entire simulation process ranging from polymer chain generation and MOF surface cleaving to equilibration and mechanical testing. The pipeline was deployed across HPC systems and modularized for extensibility.
6. High-throughput Screening Capabilities:
Although full high-throughput coverage was constrained by resource availability, simulation prioritization strategies enabled processing of a representative set of MOF/polymer interfaces. The generated data now forms a structured database of interface structures and mechanical responses.
7. Descriptor Analysis and Machine Learning Foundations:
A key scientific achievement was the identification of interfacial descriptors that govern composite mechanical and thermal behaviour. These include interfacial density gradients, pore geometry, local stress profiles, and van der Waals interaction energies. While large-scale ML training was not completed, these insights establish the foundation for predictive screening models in follow-up work.
Outcomes:
• A validated multiscale simulation pipeline linking DFT, MD, and continuum modelling for M4 mechanical property prediction.
• A library of equilibrated MOF-polymer interface structures and their associated elastic and thermal properties.
• Descriptor sets suitable for ML model development, along with prototype models tested on limited data.
• Improved understanding of how MOF surface chemistry and geometry influence interfacial strength and composite performance.
• Transferable simulation workflows supporting further material screening and design beyond the original project scope.
These technical accomplishments significantly enhance the ability to design MOF-based composite materials with tailored mechanical properties and provide a reproducible simulation foundation for future high-throughput and data-driven approaches.
Results and Potential Impacts
The M4MID project has delivered a technically validated and transferable multiscale simulation workflow to predict the mechanical properties of MOF-polymer composites (M4s) starting from atomic-level models. This framework represents a significant advance in the in silico design of hybrid materials and addresses a core bottleneck in materials development: the ability to predict interfacial compatibility and mechanical performance without relying on costly experimental trial-and-error.
Overview of Results:
• A modular, automated pipeline for constructing, equilibrating, and simulating MOF-polymer interfaces has been developed and tested across a diverse matrix of 60+ MOF/polymer combinations.
• Atomistic mechanical testing via molecular dynamics simulations yielded reliable predictions of Young’s modulus, tensile strength, and thermal conductivity for neat polymers, MOFs, and composite interfaces.
• A representative subset of interface structures was upscaled using the FFT-MAD continuum framework to derive bulk mechanical properties.
• Critical interfacial descriptors—such as density gradients, pore geometry, van der Waals interaction energies, and local stress fields—were identified and compiled into a structured database suitable for ML model development.
• Preliminary ML frameworks (e.g. Random Forests, Neural Networks) were implemented using small datasets, demonstrating feasibility for property prediction and screening in future studies.
Potential Impacts:
The outcomes of M4MID are expected to have a high impact in both academic and industrial materials development communities. By providing predictive capabilities for mechanical performance at the design stage, the project enables:
• Accelerated discovery of M4 composites for applications such as hydrogen storage, gas separation membranes, flame-retardant plastics, and biomedical carriers.
• Reduction of prototyping cycles, material waste, and development time, contributing to EU goals in green and sustainable innovation.
• Increased reliability and scalability of hybrid materials in real-world applications, bridging the gap between fundamental science and industrial manufacturing.
The project also contributes to broader EU strategic priorities in the digitalization of research and the integration of data-driven and AI-enhanced tools in materials science.
Key Needs for Further Uptake and Success:
To maximize the impact and enable widespread use of M4MID’s results, the following enablers are identified:
• Further Research and Data Expansion: Continued simulation campaigns to expand the database of MOF/polymer interfaces and complete machine learning model training are essential for generalizability and robustness.
• Demonstration and Validation: Experimental validation of predicted mechanical and thermal properties will strengthen confidence and open the path toward industrial partnerships.
• Access to HPC Resources: Scaling up simulations to cover broader chemical and structural diversity will require sustained access to high-performance computing, especially GPU-based infrastructures.
• Community and Market Engagement: Collaboration with industrial stakeholders (e.g. membrane producers, packaging companies) and integration into broader materials platforms (e.g. Materials Cloud, OpenMSP) can drive adoption.
The M4MID project has delivered a technically validated and transferable multiscale simulation workflow to predict the mechanical properties of MOF-polymer composites (M4s) starting from atomic-level models. This framework represents a significant advance in the in silico design of hybrid materials and addresses a core bottleneck in materials development: the ability to predict interfacial compatibility and mechanical performance without relying on costly experimental trial-and-error.
Overview of Results:
• A modular, automated pipeline for constructing, equilibrating, and simulating MOF-polymer interfaces has been developed and tested across a diverse matrix of 60+ MOF/polymer combinations.
• Atomistic mechanical testing via molecular dynamics simulations yielded reliable predictions of Young’s modulus, tensile strength, and thermal conductivity for neat polymers, MOFs, and composite interfaces.
• A representative subset of interface structures was upscaled using the FFT-MAD continuum framework to derive bulk mechanical properties.
• Critical interfacial descriptors—such as density gradients, pore geometry, van der Waals interaction energies, and local stress fields—were identified and compiled into a structured database suitable for ML model development.
• Preliminary ML frameworks (e.g. Random Forests, Neural Networks) were implemented using small datasets, demonstrating feasibility for property prediction and screening in future studies.
Potential Impacts:
The outcomes of M4MID are expected to have a high impact in both academic and industrial materials development communities. By providing predictive capabilities for mechanical performance at the design stage, the project enables:
• Accelerated discovery of M4 composites for applications such as hydrogen storage, gas separation membranes, flame-retardant plastics, and biomedical carriers.
• Reduction of prototyping cycles, material waste, and development time, contributing to EU goals in green and sustainable innovation.
• Increased reliability and scalability of hybrid materials in real-world applications, bridging the gap between fundamental science and industrial manufacturing.
The project also contributes to broader EU strategic priorities in the digitalization of research and the integration of data-driven and AI-enhanced tools in materials science.
Key Needs for Further Uptake and Success:
To maximize the impact and enable widespread use of M4MID’s results, the following enablers are identified:
• Further Research and Data Expansion: Continued simulation campaigns to expand the database of MOF/polymer interfaces and complete machine learning model training are essential for generalizability and robustness.
• Demonstration and Validation: Experimental validation of predicted mechanical and thermal properties will strengthen confidence and open the path toward industrial partnerships.
• Access to HPC Resources: Scaling up simulations to cover broader chemical and structural diversity will require sustained access to high-performance computing, especially GPU-based infrastructures.
• Community and Market Engagement: Collaboration with industrial stakeholders (e.g. membrane producers, packaging companies) and integration into broader materials platforms (e.g. Materials Cloud, OpenMSP) can drive adoption.