Computational materials science / molecular simulation / soft matter
I study materials with computer simulations, from single atoms up to whole films. In my doctoral work at Aalto University, I used all-atom molecular dynamics (MD), coarse-grained models, and dissipative particle dynamics (DPD) to design block copolymer coatings for paper, partly together with industry. Before that, I worked on polymer/graphene composites, and my first degree was in metallurgy. I enjoy teaching these methods, and I am now bringing machine learning into how I simulate.
7 papers, 1 in review
DPD and all-atom models of coatings that assemble on paper-like surfaces. I studied how the block ratio and the chemistry decide the order in the film, when the film cracks, and how much water gets in.
All-atom models of alkylsilane monolayers on silica, and how water and chain length shape their structure. This is my newest paper, which is now in review.
Polymer/graphene nanocomposites with treated interfaces, for better adhesion and shape memory. This was my M.Sc. work, which came after a B.Sc. in metallurgy.
Each paper is explained on the research page, with a figure that runs in your browser. My research-related code is listed below, and my full background is in the CV.
established and emerging
GROMACS and LAMMPS, for polymers, monolayers, and interfaces, dry and in water.
Several atoms per bead, for assemblies that all-atom runs cannot reach.
Soft beads for films and phase separation at the mesoscale.
ORCA scans to fit and check force field terms, such as those at the Si-C bond of alkylsilanes.
Order parameters, density profiles, hydrogen bonds, and the water next to a surface.
MACE trained on quantum chemistry data and tested in MD.
A potential that picks its own DFT calculations while it runs MD.
Bead mappings and coarse-grained force fields learned from all-atom forces.
I am adding these to my work one project at a time. The ideas behind them are explained on the machine learning in MD page. I am also learning plane-wave DFT for crystals and surfaces, enhanced sampling with PLUMED, and Bayesian optimization for choosing the next experiment.
nine industries
courses and explainers
I am building a set of projects in machine learning and simulation, and each one has its own repository with a notebook that runs on Colab. There are thirty-five numbered projects, thirteen small simulations, thirteen analysis tools from my doctoral work, and a few extras. Machine learning on molecules and simulations is the largest group. The others cover simulation engines and force fields, continuum models, Bash, C++, and GPU code, and the standard tools of my field.
on GitHubbuilt, not on GitHub yetin progressplanned
44 demos in five rooms