Artificial intelligence and machine learning are revolutionizing the way that we do materials science. In this project, we propose to take full advantage of the new available computational infrastructure to take our understanding of interatomic interactions in complex materials to the next level, maintaining an accurate representation of these atomic interactions throughout longer length and time scales than is currently possible. We will achieve this ambitious objective by creating hierarchical reactive interatomic force fields, where the different interatomic interactions are broken down into manageable pieces and "glued" together with machine learning.
With this methodology, we will create a reactive force field for cellulose, a renewable material which is a strategic asset for the Finnish economy. This force field will be used to realize different computational experiments on cellulose, to better understand its chemistry and the properties of novel cellulose-based materials.