In-Sensor Computing for MEMS Sensors: Toward an Electromechanical Neural Network
The rise of machine learning models for processing sensor data has led to the development of Edge-AI, which aims to perform these data processing tasks locally, directly at the sensor level. This approach reduces the amount of data transmitted and eases the load on centralized computing centers, providing a solution to decrease the overall energy consumption of systems. In this context, the concept of in-sensor computing has emerged, integrating data acquisition and processing within the sensor itself. By leveraging the physical properties of sensors and alternative computing paradigms, such as reservoir computing and neuromorphic computing, in-sensor computing eliminates the energy-intensive steps of signal conversion and processing.
Applying this concept to MEMS sensors enables the processing of signals such as acceleration, strain, or acoustic signals, with a significant reduction, or even elimination, of traditional electronic components. This has rekindled interest in mechanical computing devices and their integration into MEMS sensors like microphones and accelerometers. Recent research explores innovative MEMS devices integrating recurrent neural networks or reservoir computing, showing promising potential for energy efficiency. However, these advancements are still limited to proof-of-concept demonstrations for simple classification tasks with a very low number of neurons.
Building on our expertise in MEMS-based computing, this doctoral work aims to push these concepts further by developing a MEMS device that integrates a reprogrammable neural network with learning capabilities. The objective is to design an intelligent sensor that combines detection and preprocessing on a single chip, optimized to operate with extremely low energy consumption, in the femtoJoule range per activation. This thesis will focus on the design, fabrication, and validation of this new device, targeting low-frequency signal processing applications in high-temperature environments, paving the way for a new generation of intelligent and autonomous sensors.
New silicon-based alloys and composites for all-solid-state batteries: from combinatorial synthesis by magnetron sputtering to mechanosynthesis
All-solid state lithium-ion batteries using sulphide-based electrolytes are among the most studied at present in order to improve energy density, safety and fast charging. Although lithium metal was initially the preferred choice for the anode, the difficulties encountered in its implementation and the performance achieved suggest that alternatives should be proposed. Silicon offers an interesting compromise in terms of energy density and lifetime. However, it is necessary to look at anode materials developed for all-solid state batteries. To this end, we propose to collaborate with CEA Tech Nouvelle-Aquitaine, which has set up a combinatorial synthesis methodology using magnetron sputtering, in order to accelerate the identification of new compositions of silicon-based materials. Libraries of materials with compositions gradient in thin films will be prepared at CEA Tech Nouvelle-Aquitaine and then studied at CEA Grenoble. The most promising compositions will then be prepared by mechanosynthesis and characterised at CEA Grenoble. Significant work will be carried out on milling processes in order to optimise particle size and homogeneity, as well as structure and microstructure. Attention will also be paid to integration in all-solid state cells, drawing on the laboratory's expertise.
Design and reliability of modular architecture for reconfigurable and repairable PV panels
The integration of photovoltaic modules has become a challenge for adaptation to climate change, notably with the installation of specific PV modules in urban spaces, on vehicles or on agricultural farms. These modules are required to operate in more complex situations presenting high temporal variability and changing exposure to the sun. The scientific challenges of the project are to determine the conditions needed for optimizing the performance of PV modules regarding these external disturbances by the study of reconfigurable electrical module architectures. A reliability model will be developed to integrate the impact of the system architecture, in order to guarantee an improved level of reliability. In-depth work will be carried out on the entire PV module, from cell technologies to the final electrical characteristics requested, including electrical switching technologies. In a second phase, we will develop a design methodology in conjunction with a precise state of the art of available power switching technologies. The method will be applied to a use case responding primarily to the problem of shading and/or localised failure of the PV module. Finally, the proposed architectures will be evaluated by life cycle analysis. The designs authorizing maintenance or replacement of certain elements will be detailed and compared to the performance of usual modules.
Fast parameter inference of gravitational waves for the LISA space mission
Context
In 2016, the announcement of the first direct detection of gravitational waves ushered in an era in which the universe will be probed in an unprecedented way. At the same time, the complete success of the LISA Pathfinder mission validated certain technologies selected for the LISA (Laser Interferometer Space Antenna) project. The year 2024 started with the adoption of the LISA mission by the European Space Agency (ESA) and NASA. This unprecedented gravitational wave space observatory will consist of three satellites 2.5 million kilometres apart and will enable the direct detection of gravitational waves at undetectable frequencies by terrestrial interferometers. ESA plans a launch in 2035.
In parallel with the technical aspects, the LISA mission introduces several data analysis challenges that need to be addressed for the mission’s success. The mission needs to prove that with simulations, the scientific community will be able to identify and characterise the detected gravitational wave signals. Data analysis involves various stages, one of which is the rapid analysis pipeline, whose role is to detect new events and characterise the detected events. The last point concerns the rapid estimation of the position in the sky of the source of gravitational wave emission and their characteristic time, such as the coalescence time for a black hole merger.
These analysis tools form the low-latency analysis pipeline. As well as being of interest to LISA, this pipeline also plays a vital role in enabling multi-messenger astronomy, consisting of rapidly monitoring events detected by electromagnetic observations (ground-based or space-based observatories, from radio waves to Gamma rays).
PhD project
The PhD project focuses on the development of event detection and identification tools for the low-latency alert pipeline (LLAP) of LISA. This pipeline will be an essential part of the LISA analysis workflow, providing a rapid detection of massive black hole binaries, as well as a fast and accurate estimation of the sources’ sky localizations as well as coalescence time. These are key information for multi-messager follow-ups as well as for the global analysis of the LISA data.
While rapid analysis methods have been developed for ground-based interferometers, the case of space-based interferometers such as LISA remains a field to be explored. Adapted data processing will have to consider how data is transmitted in packets, making it necessary to detect events from incomplete data. Using data marred by artefacts such as glitches or missing data packages, these methods should enable the detection, discrimination and analysis of various sources: black hole mergers, EMRIs (spiral binaries with extreme mass ratios), bursts and binaries from compact objects. A final and crucial element of complexity is the speed of analysis, which constitutes a strong constraint on the methods to be developed.
To this end, the problems we will be tackling during this thesis will be:
1. The fast parameter inference of the gravitational waves, noticeably, the sky position, and the coalescence time. Two of the main difficulties reside in the multimodality of the posterior probability distribution of the target parameters and the stringent computing time requirements. To that end, we will consider different advanced inference strategies including:
(a) Using gradient-based sampling algorithms like Langevin diffusions or Hamiltonian Monte Carlo methods adapted to LISA’s gravitational wave problem,
(b) Using machine learning-assisted methods to accelerate the sampling (e.g. normalising flows),
(c) Using variational inference techniques.
2. The early detection of black hole mergers.
3. The increasing complexity of LISA data, including, among others, realistic noise, realistic instrument response, glitches, data gaps, and overlapping sources.
4. The online handling of the incoming 5-minute data packages with the developed fast inference framework.
This thesis will be based on applying Bayesian and statistical methods for data analysis and machine learning. However, an effort on the physics part is necessary, both to understand the simulations and the different waveforms considered (with their underlying hypotheses) and to interpret the results regarding the detectability of black hole merger signals in the context of the rapid analysis of LISA data.
Spectrometry and Artificial Intelligence: development of explainable, sober and reliable AI models for materials analysis
The discovery of new materials is crucial to meeting many current societal challenges. One of the pillars of this discovery capacity is to have means of characterizing these materials which are rapid, reliable and whose measurement uncertainties are qualified, even quantified.
This PhD project is part of this approach and aims to significantly improve the different ion beam induced spectrometry (IBA) techniques using advanced artificial intelligence (AI) methods. This project aims to develop explainable, sober and reliable AI models for materials analysis.
The PhD project proposed here has three main objectives:
- Develop an uncertainty model using probabilistic machine learning techniques in order to quantify the uncertainties associated with a prediction.
- Due to the very large number of possible combinatory-generated configurations, it is important to understand the intrinsic dimensionality of the problem. We wish to implement means of massive dimensionality reduction, in particular non-linear methods such as autoencoders, as well as PIML (Physics Informed Machine Learning) concepts.
- Evaluate the possibility of generalization of this methodology to other spectroscopic techniques.
Development of a ML-based analysis framework for fast characterization of nuclear waste containers by muon tomography
This PhD thesis focuses on developing an advanced analysis framework for inspecting nuclear waste containers using muon tomography, particularly the scattering method. Muon tomography, which leverages naturally occurring muons from cosmic rays to scan dense structures, has proven valuable in areas where traditional imaging methods fail. CEA/Irfu, with expertise in muon detectors, seeks to harness AI and Machine Learning (ML) to optimize muon data analysis, particularly to reduce long exposure times and improve image reliability.
The project will involve familiarizing with muography (muon tomography image) principles, simulating muon interactions with waste containers, and developing ML-based data augmentation and image processing techniques. The outcome should yield efficient tools to interpret muon images, enhance analysis speed, and classify container contents reliably. The thesis aims to improve nuclear waste inspection’s safety and reliability by producing cleaner, faster, and more interpretable muon tomography data through innovative analysis methods.
On-line monitoring of bioproduction processes using 3D holographic imaging
The culture of adherent cells on microcarriers (MCs) is a promising approach for various bioproduction applications, such as drug manufacturing and delivery, regenerative medicine, and tracking of cellular differentiation. However, the analysis of single cell morphology and behavior without affecting the substrate integrity remains a major challenge. Lens-free holographic imaging is emerging as a promising solution for real-time, non-invasive monitoring of cellular processes. This technique captures wide field of view images without requiring exogenous labeling or sample manipulation, thus preserving the integrity of the cellular environment.
This thesis proposes the development of a 3D lens-free imaging system to monitor cells on MCs in near real-time. The microscope will be coupled with advanced algorithms for data reconstruction and analysis in on-line bioreactors. The use of deep learning techniques will allow for real-time segmentation and analysis of single cells, facilitating the tracking of cellular dynamics. This innovative project paves the way to a non-invasive monitoring of 3D multicellular samples, with potential applications on organ-on-chip and more complex organoids systems.
Plasma Etching development for the advanced nodes using SADP techniques
The miniaturization of the electronics components involves the development of new processes. Indeed, the 193nm immersion lithography alone does not permit anymore to achieve the dimensional requirements of the most advanced technological nodes (=10nm). Since the last 10 years, multi-patterning techniques have been developed to overcome the i193nm lithography limitations. Herein, we will study the « Self-Aligned Double Patterning » (SADP) technique that divides by two the initial pitch of the lithographical patterns. This technology relies on a conformal deposition of a dielectric film (spacer) over the initial patterns (mandrel). The spacers will be then used as a mask during the pattern transfer by plasma etching. The small targeted dimensions require a perfect control of the etching processes. However, the etching steps can damage the materials used herein leading to a dimension loss. One of the main challenge will be to control the etching steps and so the plasma-induced modification in order to satisfy the specifications (dimension, profile, material consumption, etch rate, uniformity…). Besides, the goal will be also to propose new SADP approaches allowing us to generate different type of patterns in order to produce planar FDSOI transistors, which is currently little reported in literature.
The challenges of this PhD ?
To develop innovative etching processes
To explore new couple of material (spacer/mandrel) and to propose an industrial integration flow that will be validated by electrical tests
To identify the technological obstacles and to propose solutions for overcoming them
To put in place a reliable characterization protocol in order to detect the physical and chemical modifications of the materials used and to accurately measure the final patterns’ dimensions
Sperm 3D - Male infertility diagnostic tool using holography for imaging and 3D tracking
Infertility is a growing problem in all developed countries. The standard methods for the diagnostic of male infertility examine the concentration, motility and morphological anomalies of individual sperm cells. However, one in five male infertility cases remain unexplained with the standard diagnostic tools.
In this thesis, we will explore the possibility to determine the male infertility causes from the detailed analysis of 3D trajectories and morphology of sperms swimming freely in the environment mimicking the conditions in the female reproductive tract. For this challenging task, we will develop a dedicated microscope based on holography for fast imaging and tracking of individual sperm cells. Along with classical numerical methods, we will use up-to date artificial intelligence algorithms for improving the imaging quality as well as for analysis of multi-dimensional data.
Throughout the project we will closely collaborate with medical research institute (CHU/IAB) specialized in Assisted Reproductive Technologies (ART). We will be examining real patient samples in order to develop a new tool for male infertility diagnosis.
X-ray diffusion assisted by Artificial Intelligence: the problem of the representativeness of synthetic databases and the indistinguishability of predictions.
The advent of artificial intelligence makes it possible to accelerate and democratize the processing of small-angle X-ray scattering (SAXS) data, an expert technique for characterizing nanomaterials that allows to determine the specific surface area, volume fraction and characteristic sizes of structures between 0.5 to 200 nm.
However, there is a double problem around SAXS assisted by Artificial Intelligence: 1) the scarcity of data requires training the models on synthetic data, which poses the problem of their representativeness of real data, and 2) the laws of physics stipulate that several candidate nanostructures can correspond to a SAXS measurement, which poses the problem of the indistinguishability of predictions. This thesis therefore aims to build an artificial intelligence model adapted to SAXS trained on experimentally validated synthetic data, and on the expert response which weights the categorization of predictions by their indistinguishability.