Abstract
Switching between different levels of resolution is essential for multiscale modeling, but restoring details at higher resolution remains challenging. In our previous study, we have introduced deepBackmap, a deep neural-network-based approach to reverse-map equilibrated molecular structures for condensed-phase systems. Our method combines data-driven and physics-based aspects, leading to high-quality reconstructed structures. In this work, we expand the scope of our model and examine its chemical transferability. To this end, we train deepBackmap solely on homogeneous molecular liquids of small molecules and apply it to a more challenging polymer melt. We augment the generator’s objective with different force-field-based terms as a prior to regularize the results. The best performing physical prior depends on whether we train for a specific chemistry or transfer our model. Our local environment representation combined with the sequential reconstruction of fine-grained structures helps in reaching transferability of the learned correlations.
Highlights
The demand to further expand the accessible length scale and timescale in computer simulations for molecular systems remains consistently high
We argue that our sequential approach combined with the local-environment representation is well suited to achieve chemical transferability, as long as the generation of one atom only relies on short-range force-field related features
The temperature discrepancy for the test and training sets arises from the different boiling and melting points of the molecules, as we wish to probe the model’s chemical transferability in its liquid state
Summary
The demand to further expand the accessible length scale and timescale in computer simulations for molecular systems remains consistently high. Multiscale modeling aims at linking different levels of resolution. The reduced resolution in a coarse-grained (CG) model smooths the energy landscape and, thereby, effectively accelerates the simulation. Atomistic details are sometimes necessary for a thorough investigation of processes on smaller scales. The goal is, to use a CG model with a reduced number of degrees of freedom where it is possible and switch back to a higher resolution where it is needed.[6,7] the process of reintroducing lost degrees of freedom is challenging as it requires us to reinsert details with the correct statistical weight: Given the CG configuration, the generated atomistic structure should follow the Boltzmann distribution of atomistic microstates
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have