This figure summarises our research direction. Our approach is indicated by the bubbles at the top, our main material targets are shown in the the boxes in the middle, and our goals indicated by the text the bottom. Blue and green bubbles indicate approaches from applied mathematics and computational physics, respectively.

Target 1. Surfaces and interfaces

Many of the forefront areas of materials science, including nanomaterials assembly, organic electronics, and spintronics, involve the deposition of thin films and submonolayers of organic molecules onto inorganic surfaces. Innovations in these areas require precise control over how the molecules arrange themselves within the thin film or submonolayer, as this arrangement has a decisive influence on device performance and other figures of merit. Unfortunately, it is extremely difficult to predict how molecules behave once deposited on surfaces, and strategies for controlling their arrangement are limited to a few poorly understood empirical trends. In order to improve this situation, we are working on the following topics. 

1.1. First-principles prediction for surface-assisted molecular self-assembly processes

‘Surface-assisted molecular self-assembly’ refers to the process by which molecules on a surface arrange themselves into a thin film or submonolayer. Surface-assisted molecular self-assembly is notoriously difficult to study in silico, mainly due to the high computational cost of modelling surface-molecule interactions.  Nonetheless, by making use of some techniques from applied mathematics, particularly machine learning and Markov chain Monte Carlo, we have succeeded at predicting the outcome of surface-assisted molecular self-assembly under conditions of low molecular coverage. We are currently working away at cases involving higher molecule coverage and full monolayers, and hope to report good news soon. 

1.2. Predicting how to control surface-assisted molecular self-assembly

Suppose you wish for the molecules inside of a thin film to be arranged in a specific way. What kind of molecule should you deposit onto the surface? What level of molecular coverage should you aim for? At what temperature should the surface be held at? These are open questions that materials science has long demanded answers for. Our present approach towards these questions is to analyse simulation data (generated via the methods mentioned above) using unsupervised machine learning. Under conditions of low molecular coverage, we have shown that this approach can predict the kinds of properties that molecules ought to have in order to arrange as desired on a surface. We hope to extend our approach to the cases of higher molecule coverage and full monolayers in the near future. 

Representative papers

Daniel Packwood, Patrick Han, and Taro Hitosugi. Chemical and entropic control of the molecular self-assembly process. Nature Communications 8, 2017, 14463. 

Daniel Packwood and Taro Hitosugi. Materials informatics for self-assembly of functionalized organic precursors on metal surfaces. Nature Communications 9, 2018, 2469.

Key words
Metal surfaces, organic submonolayers and thin films, density functional theory, equilibrium statistical mechanics, Markov chain Monte Carlo, kernelized machine learning, unsupervised machine learning, Bayesian machine learning.

Target 2. Novel semiconductors

[Clever words coming soon!]

Key words
Organic semiconductor, coordination polymer semiconductor, density functional theory, non-equilibrium statistical mechanics, band structure.