| About this Abstract |
| Meeting |
MS&T23: Materials Science & Technology
|
| Symposium
|
Energy Materials for Sustainable Development
|
| Presentation Title |
Deep Learning Tomography Reconstruction of Lithium Ion Battery Defects: Providing High Resolution at Large Length Scales |
| Author(s) |
Nathan S. Johnson, Yulia Trenikhina, Hrishi Bale, William Harris, Steve Kelly |
| On-Site Speaker (Planned) |
Nathan S. Johnson |
| Abstract Scope |
Successfully manufacturing lithium-ion batteries necessitates tailoring material properties across various length scales. Defects in lithium-ion particle fabrication occur at nanometer scales, porosity between battery layers can emerge at micron scales, and cracks in the device can span up to millimeters. Effectively imaging all defects in a device demands the collection of multiple datasets at different length scales and often the use of multiple imaging techniques. In this study, we demonstrate the application of deep learning to generate X-ray tomography datasets that cover large volumes (from millimeters to microns) while providing high resolution (from microns to nanometers). This technique is employed to detect porosity and discontinuities in the Cu charge-collecting layer of a lithium-ion battery device. Although these defects are too small to observe at the millimeter scale, they are easily discerned at the micron scale. We showcase that deep learning reconstruction can enhance the resolution of datasets, enabling the detection of these defects at the millimeter scale. The presence, size, and morphology of the defects are verified using separate independent X-ray tomography scans at high resolution, as well as through scanning electron microscopy. Furthermore, we discuss the limitations of this technique and provide a preview of its application on other sustainable energy materials, such as solid oxide fuel cells and additional solid-state battery devices. |