Bayesian Neural Inference Using Ferroelectric Memory Transistors
An increasing number of safety-critical systems now rely on artificial intelligence functions that must operate under strict energy constraints and in environments characterized by data scarcity and high uncertainty. However, conventional deterministic AI approaches provide only point estimates and lack principled uncertainty quantification, which can lead to unreliable or unsafe decisions in real-world deployment.
This PhD is positioned within the emerging field of Bayesian electronics, which aims to implement probabilistic inference directly in hardware by leveraging the intrinsic stochasticity of nanoscale devices to represent and manipulate probability distributions. While memristive devices have previously been explored for Bayesian inference, their limited endurance and high programming energy remain critical bottlenecks for on-chip learning.
The objective of this thesis is to investigate ferroelectric field-effect memory transistors (FeMFETs) as building blocks for hardware Bayesian neural networks. The work will involve characterizing and modeling the exploitable ferroelectric randomness for sampling and probabilistic weight updates, designing Bayesian neuron and synapse architectures based on FeMFETs, and evaluating their robustness, energy efficiency, and system-level performance for safety-critical inference under uncertainty.
Adaptive and explainable Video Anomaly Detection
Video Anomaly Detection (VAD) aims to automatically identify unusual events in video that deviate from normal patterns. Existing methods often rely on One-Class or Weakly Supervised learning: the former uses only normal data for training, while the latter leverages video-level labels. Recent advances in Vision-Language Models (VLMs) and Large Language Models (LLMs) have improved both the performance and explainability of VAD systems. Despite progress on public benchmarks, challenges remain. Most methods are limited to a single domain, leading to performance drops when applied to new datasets with different anomaly definitions. Additionally, they assume all training data is available upfront, which is unrealistic for real-world deployment where models must adapt to new data over time. Few approaches explore multimodal adaptation using natural language rules to define normal and abnormal events, offering a more intuitive and flexible way to update VAD systems without needing new video samples.
This PhD research aims to develop adaptable Video Anomaly Detection methods capable of handling new domains or anomaly types using few video examples and/or textual rules.
The main lines of research will be the following:
• Cross-Domain Adaptation in VAD: improving robustness against domain gaps through Few-Shot adaptation;
• Continual Learning in VAD: continually enriching the model to deal with new types of anomalies;
• Multimodal Few-Shot Learning: facilitating the model adaptation process through rules in natural language.
Theoretical studies of orbital current and their conversion mechnism for leveraging spin-orbit torques based devices performances
The proposed PhD thesis aims at understanding and identifying the key parameters governing the conversion of orbital moments into spin currents, with the goal of enhancing the write efficiency of spin-orbit torque magnetic random-access memory (SOT-MRAM) devices. The work will employ a multiscale modeling approach comprising ab initio, tight-binding and atomistic calculations of the Orbital Hall Effect (OHE) and Orbital Rashba-Edelstein Effect (OREE). These phenomena exhibit larger magnitudes and diffusion lengths compared to their spin counterparts, Spin Hall Effect (SHE) and Rashba-Edelstein Effect (REE). Furthermore, they are present in a broader range of materials, including low-resistivity light metals. This opens very interesting opportunities for more efficient and conductive materials, potentially lifting the barriers limiting the technological deployment of SOT-MRAM.
This thesis will play a key role in a close collaboration between SPINTEC and LETI laboratories at CEA. The PhD student will conduct ab initio calculations at SPINTEC to unveil fundamental material characteristics to exploit the described orbitronic phenomena, and will construct multi-orbital Hamiltonians at LETI to calculate orbital and spin transport, in strong interaction/synergy with experimentalists working on SOT-MRAM development. The PhD will be co-supervised by M. Chshiev, K. Garello at Spintec and J. Li at LETI. This PhD project will be at the heart of collaborations with leading theoretical and experimental groups at national and international level.
Highly motivated candidates with a strong background in solid-state physics, condensed matter theory, and numerical simulations are encouraged to apply. The selected candidate will perform calculations using Spintec’s computational cluster, leveraging first-principles DFT-based packages and other simulation tools. Results will be rigorously analyzed, with opportunities for publication in international peer-reviewed journals.
Blended positive electrodes in solid-state batteries: Effect of the electrode fabrication process on electrochemistry
The development of cost-effective, high-energy-density solid-state batteries (SSBs) is essential for the large-scale adoption of next-generation energy storage technologies. Among various cathode candidates, LiFePO4 (LFP) and LiFe1??Mn?PO4 (LFMP) offer safety and cost advantages but suffer from low working voltages and limited kinetics compared to Ni-rich layered oxides such as LiNi0.85Mn0.05Co0.1O2 (NMC85). To balance energy density, rate capability, and stability, this PhD project aims to develop blended cathodes combining LFMP and NMC85 in optimized ratios for solid-state configurations employing sulfide electrolytes (Li6PS5Cl). We will investigate how fabrication methods- including slurry-based electrode processing and binder-solvent optimization- affect the electrochemical and structural performance. In-depth operando and in situ characterizations (XRD, Raman, and NMR) will be conducted to elucidate lithium diffusion, phase transition mechanisms, and redox behavior within the blended systems. Electrochemical impedance spectroscopy (EIS) and titration methods will quantify lithium kinetics across various states of charge. By correlating processing conditions, microstructure, and electrochemical behavior, this research seeks to identify optimal cathode compositions and manufacturing strategies for scalable, high-performance SSBs. Ultimately, the project aims to deliver a comprehensive understanding of structure–property relationships in blended cathodes, paving the way for practical solid-state battery technologies with enhanced safety, stability, and cost efficiency.
Growth and Characterization of AlScN: A New Promising Material for Piezoelectric and Ferroelectric Devices
III-nitride semiconductors — GaN, AlN, and InN — have revolutionized the lighting market and are rapidly entering the power electronics sector. Currently, new nitride compounds are being explored in the search for novel functionalities. In this context, aluminum scandium nitride (AlScN) has emerged as a particularly promising new member of the nitride family. Incorporating scandium into AlN leads to:
* Enhanced Piezoelectric Constants: Making AlScN highly attractive for the fabrication of piezoelectric generators and high-frequency SAW/BAW filters.
* Increased Spontaneous Polarization: The enhanced polarization can be exploited in designing high-electron-mobility transistors (HEMTs) with very high channel charge densities.
* Ferroelectricity: The recently discovered (2019) emergence of ferroelectric properties opens up possibilities for developing new non-volatile memory devices.
Over the past five years, AlScN has become a major focus of research, presenting numerous open questions and exciting opportunities to explore.
This PhD thesis will focus on the study of the growth and properties of AlScN and GaScN synthesized by molecular beam epitaxy (MBE). The student will receive training in the use of an MBE system for the synthesis of III-nitride semiconductors and in the structural characterization of materials using atomic force microscopy (AFM) and X-ray diffraction (XRD). The variation of the polarization properties of the materials will be investigated by analyzing the photoluminescence of quantum well structures. Finally, the student will be trained in the use of simulation software to model the electronic structure of the samples, aiding in the interpretation of the optical results.
Development and Characterization of Terahertz Source Matrices Co-integrated in Silicon and III-V Photonics Technology
The terahertz (THz) range (0.1–10 THz) is increasingly exploited for imaging and spectroscopy (e.g. security scanning, medical diagnostics, non-destructive testing) because many materials are transparent to THz radiation and have unique spectral signatures. However, existing sources struggle to offer both high power and wide tunability: electronic sources (diodes, QCLs) deliver milliwatts but over narrow bands, while photonic emitters (photomixers in III–V semiconductors) are tunable across broad bands but emit only microwatts. This thesis aims to overcome these limitations by developing an integrated matrix of THz sources. The approach is based on photomixing two 1.55 µm lasers in III–V photodiodes to generate a phase-coherent THz current coupled to THz antennas.
Initially, the PhD student will experimentally investigate an existing 16-element THz antenna array (STYX project) CEA-CTReg/DNAQ: setting up the test bench, measuring phase coherence, optical coupling, radiation lobes, and constructive interference. These experiments will provide a scientific foundation for the subsequent design of an integrated photonic array on silicon. The student will simulate the photonic architecture (couplers, waveguides, phase modulators, Si/III–V transitions) synchronizing multiple InGaAs photodiodes. Prototyping will include the fabrication of silicon photonic circuits (CEA-LETI) and THz photodiodes/antennas in InP (III-V Lab or, to be confirmed, Heinrich-Hertz-Institut of the Fraunhofer—HHI), followed by their hybrid integration (bonding, alignment).
This thesis will also rely on close collaboration with the IMS laboratory (Bordeaux), which is nationally and internationally recognized for its expertise in silicon photonics and THz systems. IMS will provide complementary expertise in optical modeling, electromagnetic simulation, and experimental characterization, reinforcing the multidisciplinary strength of the project.
Finally, the ultimate goal of this thesis is to develop a proof-of-concept demonstrator with a few phase-locked THz emitters (e.g. 4–16) will be produced and characterized, showing enhanced beam directivity and output power thanks to constructive interference. This demonstration will pave the way for large-scale THz source arrays with significantly improved range and penetration for advanced THz imaging systems.
Ultra-fast pathogenic bacteria detection in human blood
This project aims to develop a versatile and easy-to-use surface plasmon resonance imaging (SPRi) instrument for the rapid and broad-spectrum detection of low concentrations of pathogenic bacteria in complex samples, particularly blood. SPRi is a label-free technique that allows real-time probing of a sample (regardless of its optical transparency). Due to the high sensitivity of the plasmon phenomenon, the dynamic range of measurable index variation is limited by SPRi detection when reading is performed at a fixed angle, as is the case in commercially available devices. This reduces the use of such optical instruments to the study of environments whose index remains relatively stable during the experiment and whose molecular probes have molecular weights comparable to the targets (monitoring of bimolecular interactions).
This considerably limits the detection of growing bacteria in complex environments. Our laboratory has developed original solutions for the detection of very low levels of contamination in food matrices (creation of a start-up in 2012), but SPRi is unsuitable for the detection of bacteria in blood, partly due to the very high intrinsic variability of this matrix.
These limitations will be overcome by integrating five complementary components:
1. The design of an instrument optimized for real-time recording of SPR images over a defined range of illumination angles;
2. The development of dedicated SPR data analysis and processing to extract the most relevant information for each probe from the images in real time;
3. The functionalization of biochips through a combination of appropriate probes (series of peptides such as antimicrobial peptides (AMPs), antibodies, and even bacteriophages) to optimize the number of possible identifications with a reduced set of probes;
4. The learning of specific “4D-SPRi signatures” of model strains in blood matrices;
5. Validation of the performance of the new “4D-SPRi” instrument as a tool for detecting and characterizing bacteria from hospital strains compared to reference techniques.
AI-Driven Network Management with Large Language Models LLMs
The increasing complexity of heterogeneous networks (satellite, 5G, IoT, TSN) requires an evolution in network management. Intent-Based Networking (IBN), while advanced, still faces challenges in unambiguously translating high-level intentions into technical configurations. This work proposes to overcome this limitation by leveraging Large Language Models (LLMs) as a cognitive interface for complete and reliable automation.
This thesis aims to design and develop an IBN-LLM framework to create the cognitive brain of a closed control loop on the top of an SDN architecture. The work will focus on three major challenges: 1) developing a reliable semantic translator from natural language to network configurations; 2) designing a deterministic Verification Engine (via simulations or digital twins) to prevent LLM "hallucinations"; and 3) integrating real-time analysis capabilities (RAG) for Root Cause Analysis (RCA) and the proactive generation of optimization intents.
We anticipate the design of an IBN-LLM architecture integrated with SDN controllers, along with methodologies for the formal verification of configurations. The core contribution will be the creation of an LLM-based model capable of performing RCA and generating optimization intents in real-time. The validation of the approach will be ensured by a functional prototype (PoC), whose experimental evaluation will allow for the precise measurement of performance in terms of accuracy, latency, and resilience.
Multi-modal in situ nuclear magnetic resonance analysis of electrochemical phenomena in commercial battery prototypes
Advancing electrochemical energy storage technologies is impossible without a molecular-level understand-ing of processes as they occur in practical, commercial-type devices. Aspects of the battery design, such as the chemistry and thickness of electrodes, as well as configurations of current collectors and tabs, influence the electronic and ionic current density distributions and determine kinetic limitations of solid-state ion transport. These effects, in turn, modulate the overall battery performance and longevity. For these reasons, optimistic outcomes of conventional ‘coin’ cell tests often do not converge into high-performance commercial cells. Safety concerns associated with high energy density and flammable components of batteries are another subject paramount for conversion from fossil to green energy sources.
Nuclear magnetic resonance (NMR) spectroscopy and imaging (MRI) are exceptionally sensitive to the structural environment and dynamics of most elements in active battery materials.
Recently, plug-and-play NMR and surface-scan MRI methods have been introduced. In the context of fun-damental electrochemical research, merging two innovative complementary concepts within one multi-modal (NMR-MRI) device would enable diverse analytical solutions and reliable battery performance metrics for academia and the energy sector.
In this project, an advanced analytical framework for in situ analysis of fundamental phenomena such as sol-id-state ion transport, intercalation and associated phase transitions, metal plating dynamics, electrolyte deg-radation and mechanical defects in commercial Li- and Na-ion batteries under various operational conditions will be developed. A range of multi-modal (NMR-MRI) sensors will be developed and employed for deep analysis of fundamental electrochemical processes in commercial battery cells and small battery packs.
IO access scheduling on magnetic tapes using machine learning
Numerical simulations are used to obtain responses to physical phenomena that
cannot be reproduced, either because they are too dangerous or too expensive.
The models used for these simulations are increasingly complex, in terms of
size and precision, and require access to increasingly large computing and
data storage capacities. To this end, and in order to optimize costs, the use
of mass storage technologies such as magnetic tapes is critical. However, to
ensure good overall system performance, the development of algorithms and
mechanisms related to data placement and tape access scheduling is essential.
The objective of the thesis is to study the technology of magnetic tapes, as
well as existing mechanisms such as RAO (Recommended Access Order) or request
retention; and to implement new strategies for the optimization of magnetic
tape performance.