Abstract
To meet the challenge of energy demand for the total decarbonization of the society, greener and higher energy density technology must be developed. The Li-O2 battery (LOB), considered as one of the most promising post Li-ion technologies, has been investigated. More specifically, 3D imaging technique using X-ray tomography has been developed to study the electrochemical reaction and transport phenomena in the cathode of this battery. To overcome the transparency of the light elements in the cathode of LOB under X-ray, we deployed the in-line Zernike Phase Contrast during the acquisition. With this technique, the pore networks in the cathode at different state of discharge have been extracted. The nondissolution of discharged products can also be observed in 3D. A simple synthesis of a binder-free self-standing was developed. This material is selfstanding and easily upscalable. A study of recyclability of this material was conducted. We showed that this material can be fully recovered by inexpensive solvent after the cycling. We pushed forward our 3D investigation into 4D with time steps to understand the dynamics within the Li-O2 battery. An in-house coin-cell like in situ cell was designed. The timeresolved volumes of the new material have been analyzed by a particle tracking algorithm. Massive data has been collected during this work. The segmentation has become the most time-consuming in our data processing workflow. We have employed the deep learning to tackle this problem. The hyperparameters optimization problem has been discussed and some reflections on the ground truth have been brought out. We attempted to further generalize our neural network to a broader range of material. For this, the technique transfer learning has been employed