We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

physics.comp-ph

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Physics > Computational Physics

Title: Learning physical properties of liquid crystals with deep convolutional neural networks

Abstract: Machine learning algorithms have been available since the 1990s, but it is much more recently that they have come into use also in the physical sciences. While these algorithms have already proven to be useful in uncovering new properties of materials and in simplifying experimental protocols, their usage in liquid crystals research is still limited. This is surprising because optical imaging techniques are often applied in this line of research, and it is precisely with images that machine learning algorithms have achieved major breakthroughs in recent years. Here we use convolutional neural networks to probe several properties of liquid crystals directly from their optical images and without using manual feature engineering. By optimizing simple architectures, we find that convolutional neural networks can predict physical properties of liquid crystals with exceptional accuracy. We show that these deep neural networks identify liquid crystal phases and predict the order parameter of simulated nematic liquid crystals almost perfectly. We also show that convolutional neural networks identify the pitch length of simulated samples of cholesteric liquid crystals and the sample temperature of an experimental liquid crystal with very high precision.
Comments: 11 pages, 4 figures, supplementary information; accepted for publication in Scientific Reports
Subjects: Computational Physics (physics.comp-ph); Soft Condensed Matter (cond-mat.soft); Image and Video Processing (eess.IV)
Journal reference: Sci. Rep. 10, 7664 (2020)
DOI: 10.1038/s41598-020-63662-9
Cite as: arXiv:2004.01691 [physics.comp-ph]
  (or arXiv:2004.01691v1 [physics.comp-ph] for this version)

Submission history

From: Haroldo Ribeiro [view email]
[v1] Fri, 3 Apr 2020 17:41:54 GMT (3950kb,D)

Link back to: arXiv, form interface, contact.