Periodic Reporting for period 1 - MAGNIFY (Decoding the Mechanisms Underlying Metal-Organic Frameworks Self-Assembly)
Reporting period: 2022-12-01 to 2025-05-31
Metal-Organic Frameworks (MOFs) are porous materials with many societally relevant potential applications, such as carbon capture, removal of environmental toxins and drug-delivery. Despite the progress in the field, synthesizing a MOF currently requires tens to hundreds costly and time-consuming trial-and-error synthesis experiments because our ability to correlate the synthesis conditions with the desired MOF structure is very limited. To overcome this, we need to decode the mechanisms underlying MOF self-assembly, a highly complex non-equilibrium process covering a wide range of time- and length-scales, from the formation of the building units to nucleation and growth.
The MAGNIFY project is devoted to developing a multi-scale computational methodology that decodes the mechanisms underlying MOF self-assembly and enables predicting synthesis conditions-structure relationships. This ambitious interdisciplinary project combines state-of-the-art multi-scale modelling techniques with machine-learning approaches for optimizing the simulations and analyzing the data produced. We develop and validate our models in tandem with synthesis experiments. We further apply our methodology to three central problems in MOF rational design: (i) determining how synthesis conditions (temperature, solvent, reactants, metal-to-ligand ratio, additives) drive the resulting MOF material's topology and point defects, (ii) studying phase diagrams and phase transitions of MOFs, and (iii) tackling the very challenging task of predicting the synthesis conditions for producing brand new MOFs.
MAGNIFY is inscribed within a global effort of the worldwide research community to accelerate the quest for new materials to solve pressing societal problems of the 21st century. With our novel, interdisciplinary approach, we hope to consolidate the role of simulation and data science in a field that has been predominantly driven by direct experiments up to now and bring added value to push the boundaries of MOF synthesis.
The MAGNIFY project is devoted to developing a multi-scale computational methodology that decodes the mechanisms underlying MOF self-assembly and enables predicting synthesis conditions-structure relationships. This ambitious interdisciplinary project combines state-of-the-art multi-scale modelling techniques with machine-learning approaches for optimizing the simulations and analyzing the data produced. We develop and validate our models in tandem with synthesis experiments. We further apply our methodology to three central problems in MOF rational design: (i) determining how synthesis conditions (temperature, solvent, reactants, metal-to-ligand ratio, additives) drive the resulting MOF material's topology and point defects, (ii) studying phase diagrams and phase transitions of MOFs, and (iii) tackling the very challenging task of predicting the synthesis conditions for producing brand new MOFs.
MAGNIFY is inscribed within a global effort of the worldwide research community to accelerate the quest for new materials to solve pressing societal problems of the 21st century. With our novel, interdisciplinary approach, we hope to consolidate the role of simulation and data science in a field that has been predominantly driven by direct experiments up to now and bring added value to push the boundaries of MOF synthesis.
Within the first half of MAGNIFY, we have set up tools to study both synthesis and phase transitions of MOFs from the Zeolitic-Imidazolate Framework (ZIF) family. ZIFs resemble zeolites in that their metal-ligand-metal angle is similar to the T-O-T zeolite angle, most of them are very stable and promising for many industrial and environmental applications. In order to model both the early polymerization events of self-assembly as well as nucleation and growth as a whole, we needed to go from the atom scale to a scale where the minimal unit are molecules or molecular fragments. We already had a partially reactive force field for the atomistic simulations,[J. Chem. Phys., 157, 184502 (2022)] so we developed coarse grained (CG) models[J. Chem. Phys., 158, 194107 (2023); J. Chem. Phys., 160, 094115 (2024)]. The development of CG models for ZIFs was an important result by itself, as this was the first time that a systematic exploration of CG strategies was done for MOFs. Moreover, it was also the first time that the three strategies we tested were applied to model porous materials, so we had to tackle new challenges that had never been encountered by other researchers before.
With these tools at hand, we embarked in the study of several important reactive processes of ZIFs, including the mechanisms of the early stages of the nucleation of ZIF-8,[ J. Chem. Phys., 157, 184502 (2022)] the phase transformation between ZIF-4 and its amorphous ambient pressure polymorph,[ J. Mater. Chem. A, 12, 4572 - 4582 (2024)], and the first stages of its self-assembly.[E. Méndez, R. Semino; Chem. Sci.; DOI: 10.1039/D5SC00807G (2025)] These works implied both important methodological developments, including the creation of a neural network capable of identifying a polymorph at the atomic level and the implementation of metadynamics schemes, and have also helped us answering important questions such as: what are the microscopic degrees of freedom that drive phase transformations of ZIFs? What is the mechanism of such phase transformations? Is early nucleation more or less favorable than late growth? Are there differences in the growth mechanisms of different polymorphs at the molecular level? These questions, among others, cannot be fully addressed experimentally. Another very interesting result we had, was that we could predict the place of ZIF-4-cp into the ZIFs phase diagram, [J. Mater. Chem. A, 12, 31108 - 31115 (2024)] which has been elusive to experimental measurement up to the moment of this writing.
With these tools at hand, we embarked in the study of several important reactive processes of ZIFs, including the mechanisms of the early stages of the nucleation of ZIF-8,[ J. Chem. Phys., 157, 184502 (2022)] the phase transformation between ZIF-4 and its amorphous ambient pressure polymorph,[ J. Mater. Chem. A, 12, 4572 - 4582 (2024)], and the first stages of its self-assembly.[E. Méndez, R. Semino; Chem. Sci.; DOI: 10.1039/D5SC00807G (2025)] These works implied both important methodological developments, including the creation of a neural network capable of identifying a polymorph at the atomic level and the implementation of metadynamics schemes, and have also helped us answering important questions such as: what are the microscopic degrees of freedom that drive phase transformations of ZIFs? What is the mechanism of such phase transformations? Is early nucleation more or less favorable than late growth? Are there differences in the growth mechanisms of different polymorphs at the molecular level? These questions, among others, cannot be fully addressed experimentally. Another very interesting result we had, was that we could predict the place of ZIF-4-cp into the ZIFs phase diagram, [J. Mater. Chem. A, 12, 31108 - 31115 (2024)] which has been elusive to experimental measurement up to the moment of this writing.
Countless systems in the field of porous materials are too big to be considered at the all-atom level, including studying the structure and guest transport properties in MOF-containing composites at the nanoparticle level, or in MOFs containing long-scale defects, or in mixed-metal or mixed-ligand or mesoporous MOFs to cite a few examples. Our in-depth study of the merits and setbacks of CG strategies applied to MOFs, together with our explanation of how to deal with the problems intrinsic to 3D connected porous solids, set solid basis for the use of CG models to study all these problems and to tackle similar problems associated to other porous solids besides MOFs.
We have developed other novel ideas that can also be extended to studying other materials or other kinds of reactivities, such as the use of neural networks to classify polymorphs even at the very beginning of their appearance in a reactive process, when only few atoms have started to exhibit subtle changes in their local structure.
Furthermore, our works constitute examples of how computer simulation studies can be applied to solving long-standing questions in the synthesis field. In particular, we have been able to answer questions that could not have been answered through direct experimental measurements, including placing a MOF in its phase diagram and shedding light into the free energy of early nucleation and late growth of MOFs. More generally, these studies prove the enormous potential of computer simulation in bringing added value to fields that are traditionally almost exclusively experimental.
We have developed other novel ideas that can also be extended to studying other materials or other kinds of reactivities, such as the use of neural networks to classify polymorphs even at the very beginning of their appearance in a reactive process, when only few atoms have started to exhibit subtle changes in their local structure.
Furthermore, our works constitute examples of how computer simulation studies can be applied to solving long-standing questions in the synthesis field. In particular, we have been able to answer questions that could not have been answered through direct experimental measurements, including placing a MOF in its phase diagram and shedding light into the free energy of early nucleation and late growth of MOFs. More generally, these studies prove the enormous potential of computer simulation in bringing added value to fields that are traditionally almost exclusively experimental.