[CCoE Notice] CHE ENG: Karun Kumar Rao PhD Defense

[Knudsen, Rachel W]

NAME Karun Kumar Rao



COMMITTEE CO-CHAIRS: Dr. Lars Grabow and Dr. Yan Yao



DATE: Tuesday, December 01, 2020



TIME: 2:00 PM (CST)



LOCATION:  Microsoft Teams Join Microsoft Teams Meeting<https://urldefense.com/v3/__https:/teams.microsoft.com/l/meetup-join/19*3ameeting_ZWI1NmZjYmYtNDE5Ni00MGI1LTk0MGYtYWUwYTYxZDIxOTcz*40thread.v2/0?context=*7b*22Tid*22*3a*22170bbabd-a2f0-4c90-ad4b-0e8f0f0c4259*22*2c*22Oid*22*3a*2234b1c3d8-bc56-4e62-ad8b-3b4c7c61052f*22*7d__;JSUlJSUlJSUlJSUlJSUl!!LkSTlj0I!QSlB0pFA3QCw9DTUpKiF1PDWJKe2tlLSUGw4LvSmS2RLahBoPtmpx7dRnOWssFd5jKxQ$>



TITLE:

Machine Learning Guided Discovery of Advanced Functional Energy Materials





Abstract

     All solid-state batteries provide many safety advantages over traditional lithium-ion batteries by replacing the combustible organic liquid electrolyte with a ceramic solid-state electrolyte (SSE). However, the ionic conductivity in these SSEs is often several orders of magnitude lower than in their liquid counterparts. First-principles (i.e., with density functional theory, or DFT) molecular dynamics (MD) is an established approach to calculate and study ionic conductivity, but is limited in the number and type of materials that can be simulated due to the high computational cost. To this end, we leverage advanced machine learning (ML) algorithms to more efficiently calculate ionic conductivity and optimize material composition.

     To accelerate the calculation of forces and energies in MD, we train an artificial neural network force field, which scales linearly and enables the calculation of ionic conductivity at experimentally relevant scales. However, predicting a material’s ionic conductivity directly from its crystal structure is limited by data availability, incomplete material descriptors, and the inability of models to extrapolate to physically relevant conditions or new materials. By using a partial least squares algorithm with valence electronic density as an input, we identify and quantify the BCC anion substructure and interstitial density as effective physical descriptors. Additional machine learning models trained to predict the lithium probability density circumvent training limitations and highlight the importance of property representation in model performance. Our novel 3d material segmentation network provides both quantitative and qualitative insight on the topology of diffusion pathways to accelerate SSE design. Using these models, we identified several new promising classes of solid-state electrolyte candidates to have conductivities greater than 16 mS/cm as verified by DFT-MD simulations.

     The methods and outcomes described in this thesis generalize to other solid-state systems including single atom alloys for heterogeneous catalysis. The proposed combinations of first principles simulation data with ML models will greatly accelerate the rate of materials design and discovery, and can automate calculating structure-property relationships for other applications with high accuracy and without sacrificing interpretability.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://Bug.EGR.UH.EDU/pipermail/engi-dist/attachments/20201124/6d8879e7/attachment.html