Accelerated adsorption structure determination of lignocellulosic molecules

Joakim Jestilä*, Nian Wu, Fabio Priante, Adam Foster

*Corresponding author for this work

Research output: Contribution to conferencePosterScientific


Lignocellulosic biomass is a highly heterogeneous composite material that remains underutilized as feedstock for renewable materials. One of the main challenges hindering atom-efficient utilization is confident identification of the components and their three-dimensional atomic level structure. Here, we present a study of the adsorption of unbranched prototype hemicellulose
components, the xylooligosaccharides β-xylose and β-1,4-linked-xylotetraose. As flexible molecules, efficient and realistic traversal of the corresponding conformational phase space becomes critical for the success of the minimum structure search. The global optimization challenge is met with Bayesian optimization as implemented in the BOSS code. This enables
construction of a surrogate model of the corresponding density functional theory (DFT) potential energy surface, which can subsequently be traversed to yield predicted local and global minima configurations. In addition, the DFT data is used in training of the Neural Equivariant Interatomic Potential (NequIP), further accelerating structure determination, as the potential learns to equate a given atomistic configuration with the corresponding quantum chemical energies and forces, effectively bypassing the electronic degrees of freedom.
Original languageEnglish
Publication statusPublished - 10 Oct 2023
MoE publication typeNot Eligible
EventNanoscience Days - Jyväskylä, Finland
Duration: 10 Oct 202311 Oct 2023


ConferenceNanoscience Days


Dive into the research topics of 'Accelerated adsorption structure determination of lignocellulosic molecules'. Together they form a unique fingerprint.

Cite this