> Academic opportunities > PHD positions

Deep learning for multi-modal and multi-resolution electron tomography reconstructions

Technological challenge: Advanced nano characterization (learn more)

Department: Département des Plateformes Technologiques (LETI)

Laboratory: Laboratoire Microscopie Mesures et Défectivité

Start Date: 01-09-2021

Location: Grenoble

CEA Code: SL-DRT-22-0630

Contact: zineb.saghi@cea.fr

With recent advances in instrumentation and numerical methods for inverse problems, electron tomography is becoming a key 3D characterization tool capable of addressing current challenges of miniaturisation of microelectronic devices. With ultra-fast spectrometers for electron energy loss spectroscopy (EELS) and multi-detector systems in energy dispersive X-ray spectroscopy (EDX), it is now possible to acquire several signals simultaneously to reconstruct in 3D the structure and morphology of an object with sub-nanometric resolution, as well as its chemical composition with a resolution of a few nanometres. In the framework of an interdisciplinary project, we have implemented compressed sensing approaches for EELS/EDX tomographic reconstruction from a very limited number of projections. The quality and resolution of the chemical reconstructions were greatly improved, but the volumes were reconstructed separately. The objective of this PhD thesis is to develop a deep learning based methodology to take advantage of the multi-modal and multi-resolution aspect of EELS/EDX tomography. This approach would allow: 1) A gain in execution time and signal-to-noise ratio, 2) Simultaneous reconstruction of the volumes from all signals, 3) An improvement in the resolution of chemical volumes by taking into account morphological information.

See all positions Download the offer (.zip)

Email Bookmark and Share