Deep learning techniques for the localization and classification of liquid crystal phase transitions

Deep Learning techniques such as supervised learning with convolutional neural networks and inception models were applied to phase transitions of liquid crystals identify transition temperatures the respective phases involved. In this context achiral well chiral systems studied involving isotropic liquid, nematic solely orientational order, fluid smectic one-dimensional positional order hexatic local two-dimensional positional, so-called bond-orientational order. Discontinuous 1st continuous 2nd investigated. It is demonstrated that simpler transitions, namely Iso-N, Iso-N*, N-SmA can accurately be identified for all unseen test movies studied. For more subtle SmA*-SmC*, SmC*-SmI*, SmI*-SmF*, proof-of-principle evidence provided, demonstrating capability deep even those despite some incorrectly characterized movies. Overall, we demonstrate provision a substantial varied dataset textures there no principal reason why one could not develop generalizable automate identification crystal sequences novel compounds.