Navigating Columnar Order: An integrated steering and analysis pipeline for discotic liquid crystals

In this contribution, we present a tool for implementing and steering molecular dynamics simulations with a goal of achieving a high throughput of different molecular systems and a minimum of direct user interaction. The systems we are studying consist of small rigid molecules which are known to self-assemble into long quasi-onedimensional columns in a columnar liquid crystal mesophase. Unlike the vast majority of similar discotic liquid crystals, our materials of interest lack the long flexible highly entropic tails which are typically thought to stabilize the columns. To go about this, we are focused on implementing a series of molecular dynamics simulations of a group of eight fluorinated triphenylenes known to demonstrate liquid crystalline behavior and eventually expand to include several dozen related compounds. We do this using TinSPy, a python package developed in house that monitors actively running simulations in order to steer simulations in a semi-autonomous fashion based on real time analysis of the systems as they are running in order to efficiently instantiate and equilibrate new molecular systems into a simulation pipeline. This method of steering simulations has already proven itself useful to us, and we expect that in time it will allow us to scale up our system to simulations of even broader groups of molecular structures in order to aid in a machine-learning approach to molecular design.