Software support for sparse computation
The performance of computers has become limited by data movement in the fields of AI, HPC and embedded computing. Hardware accelerators do exist to handle data movement in an energy-efficient way, but there is no programming language that allows them to be implemented in the code supporting the calculations.
It's up to the programmer to explicitly configure DMAs and use function calls for data transfers and do program analysis to identify memory bottleneck
In addition, compilers were designed in the 80s, when memories worked at the same frequency as computing cores.
The aim of this thesis will be to integrate into a compiler the ability to perform optimizations based on data transfers.
Numerical optimisation of internal safety devices of batterry cells depending on chemistry
Thermal runaway (TR) of a battery pack's elementary accumulator is a key factor that can lead to various safety issues, such as fires or explosions, involving both property and people. Several safety devices can prevent and/or mitigate the consequences of thermal runaway, including the PTC (Positive Temperature Coefficient) to limit short-circuit current, the CID (Current Interrupt Device) to disconnect the external electrical terminals from the internal active elements, and the Safety Vent for cell depressurization. Internal gas pressure is the main triggering factor. However, since the gas quantity strongly depends on the chemistry involved, these safety devices should be optimized for future battery generations.
In this PhD thesis, we will develop a methodology for sizing these safety devices through numerical simulations, incorporating all characterizations from the material scale to abusive cell testing. This research will therefore focus on both numerical and experimental aspects in parallel, in collaboration with other laboratories in our department
EM Signature Modeling in Multi-path Scenario for Object Recognition and Semantic Radio SLAM
Context:
The vision for future communication networks includes providing highly accurate positioning and localization in both indoor and outdoor environments, alongside communication services (JCAS). With the widespread adoption of radar technologies, the concept of Simultaneous Localization and Mapping (SLAM) has recently been adapted for radiofrequency applications. Initial proof-of-concept demonstrations have been conducted in indoor environments, producing 2D maps based on mmWave/THz monostatic backscattered signals. These measurements enable the development of complex state models that detail the precise location, size, and orientation of target objects, as well as their electromagnetic properties and material composition.
Beyond simply reproducing maps, incorporating object recognition and positioning within the environment adds a semantic layer to these applications. While semantic SLAM has been explored with video-based technologies, its application to radiofrequency is still an emerging area of research. This approach requires precise electromagnetic models of object signatures and their interactions with the surrounding environment. Recent studies have developed iterative physical optics and equivalent current-based models to simulate the free-space multistatic signature of nearby objects.
PhD Thesis:
The objective of this thesis is to study and model object backscattering in a multi-path scenario for precise imaging and object recognition (including material properties). The work will involve developing a mathematical model for the backscattering of sensed objects in the environment, applying it to 3D SLAM, and achieving object recognition/classification. The model should capture both near- and far-field effects while accounting for the impact of the antenna on the overall radio channel. The study will support the joint design of antenna systems and the associated processing techniques (e.g., filtering and imaging) required for the application.
The PhD student will be hosted in the Antenna and Propagation Laboratory at CEA LETI in Grenoble, France. The research will be conducted in partnership with the University of Bologna.
Application:
The position is open to outstanding students with a Master of Science degree, “école d’ingénieur” diploma, or equivalent. The student should have a specialization in telecommunications, microwaves, and/or signal processing. The application must include a CV, cover letter, and academic transcripts for the last two years of study.
Advanced RF circuit design in a system and technology co-optimization approach
This thesis addresses the two major challenges facing Europe today in terms of integrating the communication systems of the future. The aim is to design RF integrated circuits using 22nm FDSOI technology in the frequency bands dedicated to 6G, which will not only increase data rates but also reduce the carbon footprint of telecoms networks. At the same time, it is essential to consider the evolution of silicon technologies that could improve the energy efficiency and effectiveness of these circuits. This work will be carried out with an eye to the design methodology of radio frequency systems.
Within the framework of the thesis, the objective will be broken down into three phases. Firstly, simulation tools will be developed to predict the performance of Leti's future 10nm FDSOI technology. The second stage will involve identifying the most relevant architectures available in the literature for the application areas envisaged for the technology. A link with upstream telecoms projects will be systematically established to ensure that the candidate understands the systems' challenges.
Finally, in order to validate the concepts developed, the design of an LNA and a VCO as part of an ongoing project in the laboratory will be proposed.
The candidate will join a large team that works on new communication systems and addresses aspects of architectural study, modeling and design of integrated circuits. The candidate must have serious skills in the design of integrated circuits and radio frequency systems as well as good ability to work in a team.
Scalable thermodynamic computing architectures
Large-scale optimisation problems are increasingly prevalent in industries such as finance, materials development, logistics and artificial intelligence. These algorithms are typically realised on hardware solutions comprising clusters of CPUs and GPUs. However, at scale, this can quickly translate into latencies, energies and financial costs that are not sustainable. Thermodynamic computing is a new computing paradigm in which analogue components are coupled together in a physical network. It promises extremely efficient implementations of algorithms such as simulated annealing, stochastic gradient descent and Markov chain Monte Carlo using the intrinsic physics of the system. However, no clear vision of how a realistic programmable and scalable thermodynamic computer exists. It is this ambitious challenge that will be addressed in this PhD topic. Aspects ranging from the development computing macroblocks, their partitioning and interfacing to a digital system to the adaptation and compilation of algorithms to thermodynamic hardware may be considered. Particular emphasis will be put on understanding the trade-offs required to maximise the scalability and programmability of thermodynamic computers on large-scale optimisation benchmarks and their comparison to implementations on conventional digital hardware.
In situ 3D visualization and modeling of grain growth during solidification of 316L steel in welding and additive manufacturing processes
CEA is currently carrying out R&D studies to assess the potential of Additive Manufacturing (AM) processes using wire deposition (WAAM and WLAM) for 316L steel, a material used in the manufacture of a large number of components. These processes are similar to the welding techniques currently used in the manufacture and repair of parts for the nuclear industry. Microstructures with a strong crystallographic texture are often obtained after welding or additive manufacturing, leading to highly anisotropic mechanical behaviors, and the prediction of these microstructures is also a key element in ensuring the reliability of non-destructive testing of parts manufactured in this way.
The aim of the thesis, which will be based on a coupled experimental/simulation approach, is to gain a better understanding of the main physical phenomena involved in solidification, in particular grain growth.
To this end, an original approach to characterizing these phenomena will be conducted on the basis of an innovative instrumented test, with the aim of obtaining a high-resolution quasi-3D view of the molten zone during solidification. The results of the experimental approach will enrich the physical models of solidification, already implemented in a 3D CA-FE (Cellular Automaton-Finite Element) model, combining a Cellular Automata (CA) approach and thermal or multiphysics modeling (FE) of the molten bath, to simulate the solidification microstructures resulting from additive manufacturing and welding processes.
Effect of water radiolysis on the hydrogen absorption flux by austenitic stainless steels in the core of a nuclear pressurized water reactor
In pressurized water nuclear reactors, the core components are exposed to both corrosion in the primary medium, pressurized water at around 150 bar and 300°C, and to neutron flux. The stainless steels in the core are damaged by a combination of neutron bombardment and corrosion. In addition, radiolysis of the water can have an impact on the mechanisms and kinetics of corrosion, the reactivity of the medium and, a priori, the mechanisms and kinetics of hydrogen absorption by these materials. This last point, which has not yet been studied, may prove problematic, as hydrogen in solid solution in steel can lead to changes in (and degradation of) the mechanical properties of the steel and induce premature cracking of the part. This highly experimental thesis will focus on the study of the impact of radiolysis phenomena on the corrosion and hydrogen uptake mechanisms of a 316L stainless steel exposed to the primary medium under irradiation. Hydrogen will be traced by deuterium, and neutron irradiation simulated by electron irradiation on particle accelerators. An existing permeation cell will be modified to allow in operando measurement by mass spectrometry of the deuterium permeation flux through a sample exposed to the simulated primary water under radiolysis conditions. The distribution of hydrogen in the material, as well as the nature of the oxide layers formed, will be analysed in detail using state-of-the-art techniques available at the CEA and in partner laboratories. The doctoral student will ultimately be required to (i) identify the mechanisms involved (corrosion and hydrogen entry), (ii) estimate their kinetics and (iii) model the evolution of hydrogen flux in the steel in connection with radiolysis activity.
Study of the influence of the microstructure of a 316L steel produced by the L-PBF process on its mechanical properties: characterization and modeling of creep and fatigue behavior
Research into additive manufacturing for the nuclear industry shows that the production of 316L austenitic steel components using laser powder bed fusion (L-PBF) presents technical challenges, including process control, material properties, qualification and prediction of mechanical behaviour under service conditions. The final properties differ from traditional processes, often exhibiting anisotropy that challenges existing design standards.
These differences are linked to the unique microstructure resulting from the L-PBF process. Controlling the manufacturing chain, from consolidation to qualification, requires an understanding of the interactions between process parameters, microstructure and mechanical properties.
The aim of this thesis is to study the relationships between the microstructure, texture and mechanical properties of 316L steel manufactured by the L-PBF process, under static or cyclic loading. This includes the influence on creep and fatigue properties, and the development of a model to predict mechanical behaviour. Using samples of 316L steel with specific microstructures consolidated by L-PBF, the proposed study aims to establish links between microstructure and mechanical properties to better predict in-service behaviour.
Head-on Reflections of High-Speed Combustion Waves: Experimental and Numerical Investigation and Mitigation Measures.
This thesis focuses on the analysis of hydrogen safety in industries, particularly in cases of accidents where hydrogen is released or generated, such as in nuclear power plants. The interest in hydrogen safety has increased with the use of fuel cells for mobility. In compartmentalized buildings, flammable atmospheres can form, leading to explosions that compromise safety. Flame dynamics are influenced by boundary conditions, especially confined geometries that accelerate the flames. This phenomenon can result in a deflagration-to-detonation transition, causing significant damage to structures through shock waves and combustion waves. Research shows that certain geometric configurations and hydrogen mixtures produce higher pressures, even with low hydrogen concentrations. Three key questions are raised: the influence of geometry on pressure and impulse, the optimal hydrogen concentration, and the possibility of mitigating these effects with sound-absorbing coatings. To answer these questions, experiments and simulations will be conducted to understand and model these phenomena, providing practical tools for safety engineers.
Polycrystalline numerical simulations of the mechanical behavior of fuel rod cladding used in pressurized water reactors
The fuel rods of pressurized water nuclear reactors are made of uranium oxide pellets stacked in zirconium alloy tubes. In reactor, these materials undergo mechanical loading that lead to their irreversible deformation. In order to guarantee the safety and increase the performance of nuclear reactors, this deformation must be modeled and predicted as precisely as possible. In order to further improve the predictivity of the models, the polycrystalline nature of these materials as well as the physical deformation mechanisms must be taken into account. This is the objective of this study, which consists of developing a physically based multi-scale numerical model of the fuel rod cladding.
The mechanical behavior of metallic materials is usually modeled by considering the material as homogeneous. In fact metallic materials are made of many crystalline grains clustered together. The behavior of the material is therefore the result of the deformation of individual grains but also their interactions between each other. In order to take into account the polycrystalline nature of the material, mean-field self-consistent polycrystalline models have been developed for many years. These models are based on the theory of homogenization of heterogeneous materials. Recently, a polycrystalline model, developed in a linear and isothermal framework, has been coupled with an axisymmetric 1D finite element calculation to simulate the in-reactor deformation of cladding tubes. A complex mechanical loading history, mimicking the stresses and strains experienced by the cladding has been simulated.
The objective of this PhD work is to extend the field of application of this model in particular by applying it to a non-linear framework in order to simulate high stress loadings, to extend it to anisothermal conditions but also to carry out 3D finite element simulations with at each element and each time step a simulation using the polycrystalline model. These theoretical and numerical developments will finally be applied to the simulation of the behavior of fuel rods in a power ramp situation thanks to its integration into a software platform used for industrial applications. This approach will allow to better assess the margins available to operate the reactor in a more flexible manner, allowing it to adapt to changes in the energy mix in complete safety.