(272g) Combining Coarse-Grained Modeling and Enhanced Sampling to Understand the Self-Assembly of Porous Crystals

Porous crystals are ordered materials featuring three-dimensional porous networks, whose chemistry and architecture can be manipulated to control how chemical species of interest adsorb, diffuse, and react. Such control can be exploited in areas such as gas and energy storage, molecular separations, catalysis, drug delivery, and chemical sensing. A well-known class of porous crystals are metal-organic frameworks (MOFs), whose structures are based in the interconnection of metal ions or metal-based cluster by organic ligands through coordination bonds. By virtue of their modularity and building block combinatorics, MOFs are exceptionally tunable, giving rise to a “design space” of at least a trillion structures. Exploring such vast design space has largely been aided by computational screening, but once a MOF design with desired properties is identified a remaining obstacle is not being able to predict how the desired MOF design can be synthesized. Thus, the development of MOFs would benefit immensely from a mechanistic and predictive understanding of how porous crystals form in general.

As porous crystals in general, MOFs are known to form through the spontaneous organization of precursor building blocks into a specific structure or pattern, in a process known as self-assembly. But our mechanistic understanding of self-assembly in porous crystals is still limited. As the time- and length-scale of crystal nucleation are too short and small, respectively, for experiments, simulations methods such as molecular dynamics (MD) are better poised to shed light into the porous crystal self-assembly pathways. The challenge is that the time- and length-scale of crystal nucleation may be too large for conventional atomistic MD (CAMD) simulations. For instance, in some reported CAMD simulations on a MOF system, the thermodynamically favored phase was never accessed during the simulation. Thus, it is apparent more advanced simulation approaches and/or protocols are needed to identify the rare events, kinetically arrested states and kinetic pathways that are bound to show during the self-assembly of porous crystals.

The first part of this presentation discusses how we approach introduction of directionality in interactions through the introduction of dual dummy sites in nodes and linkers, and how we overcome the time- and length-scale problem by studying the self-assembly of a coarse-grained, alchemical cubic crystal (inspired by the well-known porous crystal MOF-5). The alchemical nature of these crystals allows us to systematically study how modulating the strength of node-linker interactions varies the thermodynamics and kinetics of self-assembly. Next, we discuss the mapping of the underlying free energy surface of disordered and ordered phase transitions. The simulation of metastable state transitions using conventional molecular dynamics is impractical if large energetic barriers exist throughout the free energy surface. Instead, we adopted the replica exchange method to sample solution, amorphous, and crystalline states previously observed in our simulations, which we delineate through a series of different collective variables, e.g., Steinhardt bond order parameters, cluster sizes, and coordination numbers. Finally, by computing two-dimensional potentials of mean force surfaces, we identified minimum free energy pathways that describe transitions between phases, elucidating how initially seeded structures would evolve in time as a function of interaction parameters.