Periodic Reporting for period 1 - HydroMOF (Hydrogen Storage in Electric Field Responsive Metal Organic Frameworks Studied by Machine Learning Potentials)
Berichtszeitraum: 2022-09-01 bis 2024-08-31
Let’s imagine that, in a near-future city, only hydrogen-powered cars move quietly along streets. The air is fresh, thanks to nearby hydrogen refueling stations that allow drivers to fill up quickly. Factories have changed too, using hydrogen as a primary energy source to power their machinery. In parks, families enjoy picnics, while children fly hydrogen-powered drones. The imagination of this future city is bright and inspiring, but there's a big challenge: efficient hydrogen storage systems haven't been developed yet. The HydroMOF project, supported by the EU’s HORIZON Marie Skłodowska-Curie Actions, aims to tackle this issue by developing smart and controllable systems for storing hydrogen using nanoporous metal-organic frameworks (MOFs). Recent studies have shown that external electric fields can induce controlled structural changes in MOFs, allowing them to undergo reversible transformations, which opens up new possibilities for hydrogen storage. Computational chemistry and machine learning methods have been applied in this project to investigate the potential of switchable MOFs. By combining machine learning potentials with molecular dynamics simulations, the HydroMOF project aimed to overcome the limitations of quantum chemistry methods in terms of simulation time and length scales, as well as the accuracy and reliability issues associated with classical molecular dynamics simulations. The results of the HydroMOF project can be a door-opener for fast and accurate modeling of switchable MOFs, significantly boosting their applications in hydrogen storage.
The goal of the HydroMOF project was to predict the structural changes of nanoporous materials, like metal-organic frameworks (MOFs), under applied electric fields. Here, the electric field is considered as an external stimulus that can lead to pore openings and closings by ligand rotations. This means we could potentially control access to pores in MOFs by switching the external electric field on and off. Once achieved, this technology could significantly improve hydrogen storage capabilities.
We aimed to be fast and accurate in our research, and the answer lay in using cutting-edge machine learning methodologies. However, none of the existing machine learning potentials in the literature could account for electric field effects. This meant we needed to develop a solution that enables modeling MOFs, and indeed any other material, under electric fields of various strengths and directions.
The main achievement of the HydroMOF project is our introduction of a novel electric field-dependent machine learning methodology. With this approach, we can determine how a MOF responds to electric fields at different strengths and directions by training on a dataset that includes energies and forces of MOF configurations obtained under various electric field conditions. This represents a significant improvement, as machine learning potentials can be generated once and then used for any future simulations quickly and accurately, regardless of the electric field settings.
We aimed to be fast and accurate in our research, and the answer lay in using cutting-edge machine learning methodologies. However, none of the existing machine learning potentials in the literature could account for electric field effects. This meant we needed to develop a solution that enables modeling MOFs, and indeed any other material, under electric fields of various strengths and directions.
The main achievement of the HydroMOF project is our introduction of a novel electric field-dependent machine learning methodology. With this approach, we can determine how a MOF responds to electric fields at different strengths and directions by training on a dataset that includes energies and forces of MOF configurations obtained under various electric field conditions. This represents a significant improvement, as machine learning potentials can be generated once and then used for any future simulations quickly and accurately, regardless of the electric field settings.
Through the HydroMOF project, we developed a new electric field modified machine learning methodology that can provide quantum-accurate trajectories of MOFs under an external electric field in a shorter time. To model switchable MOFs, electric field dependent machine learning potentials have been constructed and used in molecular dynamics simulations to monitor the structural changes of MOFs exposed to the electric field. With this novel and simple method introduced to the literature, we expect improvements in the design of smart and switchable MOFs, as well as their applications as hydrogen storage mediums. In the long term, our results will enhance current hydrogen storage technologies, helping to reduce dependency on fossil fuels, decarbonize industrial processes, and decrease greenhouse gas emissions.
To ensure the success and wide usability of our novel methodology, we are continuing our research by applying it to different MOFs that respond uniquely to electric fields, such as those exhibiting pore opening/closing, breathing behavior, or linker swinging. We will also investigate the hydrogen uptake of the MOFs responding differently to electric field.
To ensure the success and wide usability of our novel methodology, we are continuing our research by applying it to different MOFs that respond uniquely to electric fields, such as those exhibiting pore opening/closing, breathing behavior, or linker swinging. We will also investigate the hydrogen uptake of the MOFs responding differently to electric field.