(739c) Combining Biased Sampling and Markov State Models to Characterise the Assembly and Exchange Dynamics of Molecular Materials in Solution

Achieving an efficient and scalable production of advanced molecular materials is a central issue in the development of novel chemical processes. Leveraging self-assembly to efficiently produce molecular materials such as crystals, porous matrices or non-covalent polymers has the potential to impact a variety of applications, including pharmaceutical manufacturing, separations, sensing and catalysis. In this context, understanding the molecular-level fundamentals of self-assembly, growth, and dissolution processes is key.

A representative example is provided by molecular crystals. It is known that both the particle morphology and bulk structure of molecular crystals depend on molecular-level kinetics, which in turn are impacted by the composition of the parent phase and by the conformational complexity of their constitutive units [1]. In this case obtaining insight into molecular-level thermodynamics, mechanisms, and kinetics of crystal nucleation and growth, while explicitly accounting for the composition of the liquid phase and the structural complexity of the building units is essential to understand, model and design efficient crystallization processes.

Molecular dynamics (MD) simulations naturally provide the spatial resolution necessary to achieve this level of detail. However in order to apply MD to investigate assembly processes we need to bridge the gap between computationally accessible timescales - of the order of ~μs-ms - and the characteristic timescale of relevant events – often exceeding the seconds. A viable approach at tackling this challenge is the application of biased sampling methods such as Well-Tempered Metadynamics (WTMetaD) [2] to achieve an extensive exploration of the configuration space and allow for an efficient calculation of transition rates.

In this contribution we provide an overview of the application of WTMetaD and Markov State Models (MSM) [3] to the investigation of self-assembled molecular materials in solution. We discuss three classes of systems of increasing complexity.

We begin by analysing the conformational isomerism of ibuprofen in solution and at the crystal/solution interface. In this case we employ WTMetaD to obtain free energy surfaces associated to ibuprofen conformational transitions. Free energy basins allow to identify a set of states used build a MSM. Transition rates between states are computed from extensive unbiased sampling or by reconstructing the transition time distribution from state to state infrequent WTMetaD simulations [4, 5]. This analysis uncovers how the interplay between solute, solvent and interface structure leads to significant variations in the conformers equilibrium probability, in the relaxation rates of the conformer population, and in the dominant mechanism of conformational rearrangement of ibuprofen.

We continue by discussing the construction of a thermodynamic model for describing the conformational isomerism of structural building units (SBUs) involved in the early stages on the nucleation of MIL101(Cr) metal-organic framework [6]. In this case, by running extensive WTMetaD simulations of the formation of SBUs from smaller precursors we unveil a large number of potential structural isomers of SBUs. We analyse the ensemble of sampled structures by identifying a subset of these SBUs that retain a configuration compatible with that adopted in the MIL101(Cr) bulk. By systematically building equilibrium models for SBUs formation at increasing ionic strength we show that ions tend to screen ligand-metal interactions and, above a certain threshold, ions favour the formation of crystal-like SBUs by establishing salt bridges.

Finally, we discuss the application of infrequent WTMetaD to compute rates and identify molecular mechanisms associated with monomer exchange in structural variants of 1,3,5-benzenetricarboxamides (BTA), a family of molecules capable of assembling into water-soluble supramolecular polymers [7]. In this case molecular simulations capture supramolecular dynamics consistent with experimental observations, providing the information necessary to link structural and dynamical features of BTA variants. These findings lead to the construction of a general kinetic model capable of reproducing key features of experimental FRET mixing essays for non-covalent BTA-like fibres.

To conclude, in this contribution we highlight how the systematic application of enhanced sampling methods to discover relevant intermediate states, compute transition rates, and build kinetic models, opens up the possibility of systematically broadening our understanding of mechanisms, kinetics, and thermodynamics of a variety of molecular assemblies in the liquid phase ranging from molecular crystals to supramolecular fibres.

References

1. M. A. Lovette, A. B. Browning, D.W. Griffin, J. P. Sizemore, R. C. Snyder, M. F. Doherty, Ind. Eng. Chem. Res, 2008, 47, 9812
2. A. Barducci, G. Bussi, M. Parrinello, Phys, Rev. Lett., 2008, 100, 020603
3. G. R. Bowman, V. S. Pande, Springer, USA
4. P. Tiwary and M. Parrinello, Phys. Rev. Lett., 2013, 111, 230602
5. M. Salvalaglio, P. Tiwary, M. Parrinello, J Chem Theory and Comput, 2014, 10 (4), 1420-1425
6. D. C. Cantu, B. P. McGrail, and V. A. Glezakou, Chem. Mater. 26, 6401, 2014
7. D. Bochicchio, M. Salvalaglio, G.M. Pavan, 2017. Nat. Commun., 8.