Real-time 3D Coherent Diffraction Data Inversion Through Deep Learning

Mathew J. Cherukara

Assistant Scientist, Center for Nanoscale Materials

‚ÄčArgonne National Lab

I am Mathew Cherukara, I received my Ph.D in materials science and engineering from Purdue University with an emphasis on computational materials science, and a bachelors in materials engineering from the Indian Institute of Technology (IIT) Madras. I am now an assistant scientist at the Center for Nanoscale Materials at Argonne National Laboratory where I perform X-ray coherent diffraction imaging (CDI) experiments and X-ray fluorescence mapping to study dynamic processes at slow and ultra-fast (sub-ns) timescales. I use inputs from our X-ray imaging techniques to build experimentally informed models that can in turn be used to make predictions at spatio-temporal scales the experiment cannot access. Underlying both the analysis of data and model development are machine learning techniques that accelerate the process of data abstraction and model development. In particular, I build AI models based on deep convolutional neural network (CNN) encoders to rapidly translate X-ray imaging data to real-space structure and lattice strain information. I am the recipient of research awards from the Materials Research Society (MRS), Defense Threat Reduction Agency (DTRA) and the College of Engineering at Purdue. ‚Äč