Digital tranformation

Digital tranformation

Digital tranformation

With cutting-edge expertise in electronics, robotics and cybersecurity and unique capabilities for designing, building and managing innovative technology platforms, the CEA is a leader in the field of digital transformation. Its work is carried out within the Fundamental Research division (DRF) and the Technological Research division (DRT).

More information on the DRF    More information on the DRT

The CEA has a unique range of expertise to address the various issues involved in the digital transformation.»

Digital transformation

3,000

The number of employees working in the field of digital transformation within the CEA.

We live in a time when the role of smart systems and objects is growing (for mobility, energy, manufacturing, communications, etc.) and data volumes are increasing (both from industry and the academic world) - even more so during the pandemic. The CEA intends to contribute to the digital transformation in a way that maintains France’s strategic, economic and sovereignty interests, addresses society’s concerns and meets the needs of science and industry.

The industrial revolution of the nineteenth century changed society profoundly. The digital transformation, and the resulting changes in habits, has been radically changing our society since the early 2000s. This transformation speeded up during the pandemic and the various lockdowns, when there was massive use of digital technologies, and this is destined to become permanent. Some aspects of this, such as the accessibility of networks and equipment, energy consumption and environmental impact associated with digital practices, and more generally system vulnerability, have become major issues. High-power digital technology (high performance computing, processing big data) has been an essential “tool” for an increasing number of the CEA’s programmes for many years.

In its mission to provide technological support for industrial innovation, the CEA has a unique range of expertise to address the various issues involved in this transformation. Underpinned by outstanding fundamental research and innovative technology platforms, together with a strong entrepreneurial and industrial culture, its work, carried out by over 3,000 employees, covers the entire digital value chain: from the design of the most advanced components, including nanometre scale components, right through to the development of global digital solutions (hardware and software) incorporating a cybersecurity approach. To increase its ability to meet society’s requirements ever more efficiently, the CEA launched a foresight exercise across its activities in March 2020, which involved over 200 experts from all its operational divisions. Its conclusions are expected in 2021.

The work of the CEA’s teams, mainly at the Technological Research Division and the Fundamental Research Division, concerns both R&D on digital technologies (both hardware and software) and more fundamental upstream research. Three main topics are covered: microelectronics (hardware), digital systems (software) and cybersecurity to ensure the integrity and protection of both components and systems.


Digital transformation

Microelectronics

In microelectronics, a field in which the CEA has been involved from the outset via its research on the atom, work is being carried out on the design of integrated circuits and devices (connected objects, imaging devices, etc.). Work in this field also covers the nanosciences and quantum engineering, focusing on two main areas: nano-physics with the aim of controlling and manipulating quantum phenomena to obtain new functional properties, and the synthesis, modelling and characterisation of nanomaterials, nano-objects and nanodevices. The expertise of the CEA’s teams in quantum technologies has led to the organisation being one of the joint leaders, together with the CNRS, Inria, and the Universities, of the French government’s plan focusing on these technologies.

Digital transformation

Digital systems

The CEA is investigating several areas in digital systems: artificial intelligence technologies; cyber-physical systems, designed as computer networks with physical inputs and outputs (instead of standalone devices); tools and engineering for the design and validation of digital systems, for example with the Papyrus platform (used for “step by step” modelling of the information system to be developed) and the Frama-C platform (programme analysis and testing); digital instrumentation for which the CEA is a world leader in the field of non-destructive testing with its CIVA simulation software.

With the increasing development of digital technology, cybersecurity has become a major issue for governments and economic stakeholders. It is a fast-growing sector with several competing French industrial companies. The CEA has to deal with issues specific to cybersecurity both as the operator of its own information systems (IS) and as an operator of industrial control systems (ICS). It therefore carries out both operational activities and R&D activities. This is particularly the case in the areas of energy generation and management, as well as overall security.

In this field, this involves meeting its own needs and those of the French Defence Procurement Agency (DGA) and the French defence industry (Thalès, MBDA, Naval Group, etc.), and also developing the technologies (software and hardware) for the future trusted solutions required by industrial companies seeking to ensure their products, systems and services are secure. It is this expertise that led the public authorities, in February 2021, to task the CEA, together with the CNRS and Inria, with the leadership of the “Cybersecurity” priority research programme, which is part of the French economic recovery plan.

The CEA is also involved in the digital transformation through its long-established expertise in the area of high performance computing (HPC), in particular under the impetus of the “Simulation programme” led by its Military Applications Division, and by the exponential requirements of projects and research tools for data generation and processing. The CEA has one of the largest scientific computing complexes in Europe at its Bruyères-le-Châtel site. It provides services for defence, industry and academic research. The CEA also leads or is involved in a number of major digital research infrastructures, in particular in the field of health with France Génomique, CATI and N4HCloud.

High-power digital technology now plays a central role in a great many fields. Digital simulation and HPC have been massively used for many years, for example in climatology to gain a better understanding of meteorological phenomena or to assess natural hazards; in chemistry and biology to investigate the mechanisms of living things; in materials physics to qualify new concepts and check their robustness; in the field of energy to design future facilities or make the maintenance and safety of current facilities easier.

The use of data is therefore crucial in the understanding and analysis of physical, chemical and biological phenomena. However, obtaining reliable, robust results requires high quality big data to be available. Here too, the CEA is in a special position, as it has recognised expertise in the production and management of such data: observational, spatial, environmental, health and fundamental physics data, as well as results of digital simulations, etc.

This expertise also covers the reconstruction of information on fragmented data, or even more importantly, by correlating and integrating heterogeneous data. This is essential for medicine: it will be possible to develop personalised diagnostics and treatments through the cross-analysis of numerous sources of health data (genomic, clinical, imaging, etc.), rendering it coherent and ready for interpretation. This approach relies on having multidisciplinary teams, involving R&D engineers and data scientists, and platforms focussing on both industrial innovation and exploratory research (quantum solutions, ultimate electronics).

With this wealth of expertise in all the operational divisions, in 2020 the CEA, with its German partner Forschungzentrum Jülich (FZJ), initiated the creation of a virtual Franco-German laboratory, AIDAS (AI – Data Analytics – Scalable Simulation), working in the field of high-power digital technology. Its aim is to consolidate the partnerships on these topics and establish itself as a major European-scale player.





digital

Key events 2020

Quantum supremacy” on a laptop computer?
Quantum technologies

“Quantum supremacy” on a laptop computer?

— A researcher and his partners at the Irig have demonstrated that a real (and therefore imperfect) quantum computer is several billion times easier to simulate than a perfect quantum computer! In 2019, Google announced that its own device had achieved quantum supremacy, having taken just a few minutes to complete a task that would take the most powerful supercomputers around 10,000 years. The researchers have shown that simulating a quantum computer on a laptop gives results similar to those from the Google experiment, at least for some tasks - and with an algorithm that is a few billion times faster than Google’s. These results suggest that current quantum computers only have a tiny fraction of the computing power that a perfect quantum computer would have. It would be pointless to increase the number of qubits to try to reach this level of power - it would be better to improve their fidelity. This is, however, an extremely difficult task for which there is no systematic method.

Irig: Interdisciplinary Research Institute of Grenoble (CEA/Grenoble-Alpes University)
Qubit: Quantum analogue of a bit, the most simple information storage unit

Adding noise to ensure totally secure communications
Quantum technologies

Adding noise to ensure totally secure communications

— How can you protect the confidentiality of communications if you cannot totally trust the devices used to communicate? Researchers at the IPhT, the University of Basel and ETH Zurich University have provided some answers to this question, which lies at the heart of quantum cryptography research. Although existing protocols use encryption methods based on keys produced using quantum principles, they cannot detect interception by a third party. Researchers have therefore deliberately added noise to the information on the encryption key so that, even if there is an undetected intrusion, a “spy” receives so little real information about the encryption key that the security of the protocol remains guaranteed.

> For more information

IPhT: Institute of Theoretical Physics, Saclay (CEA/CNRS)

SamurAI is making artificial intelligence accessible
The Internet of things

SamurAI is making artificial intelligence accessible

— Artificial intelligence (AI) and low consumption are not necessarily incompatible! The proof of this is the SamurAI system, developed by the CEA, in which these two previously incompatible concepts go hand in hand for the first time. The system is based on an architecture which combines high-energy efficiency AI accelerators and low consumption versatile IoT nodes. An idle mode mechanism saves energy, but remains “vigilant” and immediately activates the computing part when complex processing is needed.
The versatility of SamurAI means that its computing performance is up to four times faster, 3.5 times more energy efficient and uses half the power of similar IoT nodes. It has been successfully tested on a people counting and scene classification scenario, and keyword recognition in natural language.

IoT: Internet of Things. Term used for all connected physical things that have their own digital identity and can communicate with one another.

Displays
Displays

Aledia is enjoying a great deal of success

— The 3D nanowire microLED technology developed by Aledia is destined to replace today’s liquid crystal and OLED display technology for all types of products, whether they are laptops, tablets, smartphones, augmented-reality glasses or even televisions. Protected by 197 patents, this technology will enable the production of high-definition, energy-efficient displays that are 1,000 times brighter than existing displays and provide better image quality.
This start-up, which is a CEA spin-off, has also successfully produced the first microLED chips on 300 mm silicon wafers. It will be possible to make 60 to 100 smartphone displays on a single 300 mm wafer, as against four to six on the sapphire substrate currently used by the LED industry.

OLED: Organic Light-Emitting Diode. Electronic component that emits light. It is made of several superimposed layers of organic (crystal or polymer) semiconductors. OLED is derived from LED.
LED: Light-Emitting Diode. Device that emits light when an electric current passes through it. It is in particular used in flat screens.

Preprocessing MRI images
The brain

Preprocessing MRI images

— Researchers at CEA-Joliot have developed an algorithm for pre-processing raw images of the brain obtained using fMRI in primates. The aim is to help to harmonise the methods used in different laboratories, improve the procedures, make it easier to compare and interpret images and limit the use of animals for scientific purposes.
To assess the robustness of their module, named Pypreclin, the researchers pre-processed the fMRI data acquired at 3 teslas under various conditions and compared their results with those obtained using an existing pre-processing method. They also used Pypreclin on fMRI images obtained in other laboratories in the PRIME-DE consortium with other image acquisition protocols.
Their results show that Pypreclin is a robust image pre-processing tool that is suitable for various fMRI experimental and technical situations. The tool even manages to deal with the metal artefact of deep brain stimulation implants, paving the way for its transposition to the world of clinical imaging, for example for patients with Parkinson’s disease.

CEA-Joliot: CEA Frédéric Joliot Institute
fMRI: Functional magnetic resonance imaging used to study brain function
PRIME-DE: International database of non-human primate brain imaging which is open to the entire scientific community

Protect your personal data!
Social networks

Protect your personal data!

— As you probably know, the photos you post on social networks may cause you problems at some point in your life.
With the support of the MAIF foundation, the CEA has developed Your Data Stay Yours (YDSYO), a mobile app to raise users’ awareness of the effects of sharing personal data. Based on image recognition and deep learning using neural networks, YDSYO simulates, for each photo posted, their potential impact in key situations such as looking for a bank loan, a house or a job. The app gives an impact score that is calculated based on notes made by volunteers via free crowdsourcing.

Traffic in the Paris metro will soon run more smoothly
Transport

Traffic in the Paris metro will soon run more smoothly

Delays and line outages on the Paris metro can now be prevented using artificial intelligence! Each line has several thousand devices sending all types of data to the control room. This data is analysed by a supervision system that controls the whole network. Software tools designed by CEA-List can characterise the status of the system and predict any changes. By identifying situations that may lead to breakdowns, operators can take action as early as possible.
The initial results show that these tools detect 90% of the breakdowns recorded by the Paris public transport system, between two hours and just a few minutes before they happen. The rate of false alarms is four times lower, due to the improved triggering thresholds in 99% of cases.

CEA-List: One of the institutes in the Technological Research Division. It focuses on smart digital systems.

First “all-electron” GW calculation on a thousand atoms
Theory

First “all-electron” GW calculation on a thousand atoms

Researchers at the Irig and the Néel Institute have developed a series of theoretical innovations that pave the way for the simulation of very large quantum systems – around a thousand atoms – in a complex electrostatic environment, such as that of an organic photovoltaic cell or light-emitting diode. Together with other methods introduced by the researchers, these new developments have been implemented in the “beDeft” massively parallel computing code. They are currently the subject of a major challenge on the “AMD Rome” extension of the Irène supercomputer at the CEA’s Very Large Computing Centre (TGCC) at Bruyères-le-Châtel, with a view to demonstrating the first “all-electron” GW calculation on a system of a thousand atoms.

Irig: Interdisciplinary Research Institute of Grenoble (CEA/Grenoble-Alpes University)
Néel Institute: Laboratory carrying out research in condensed matter physics (CNRS)
GW: Many-body perturbation theory, also called GW theory, used to calculate the electronic properties of organic materials

miRViz: open-access analysis of microRNAs
Data analysis

miRViz: open-access analysis of microRNAs

— Researchers at the Irig have built miRViz, an open-access website using the power of networks to visually analyse data from microRNAs, which are small RNAs (2500 in humans) that do not code for any proteins but are major regulators of gene expression. Experimental measurements of their level of expression and their functional role generate large amounts of data that are difficult to analyse in the absence of any dedicated bioinformatics tools. Using miRViz, the researchers obtained public data on patients with adrenocortical carcinoma and analysed it from new perspective. They visually retrieved the results that were initially published, without needing any programming expertise. They also demonstrated, for the first time, that the expression of miR-29, a microRNA expressed in these tumours, is a favourable prognostic marker for survival.

Irig: Interdisciplinary Research Institute of Grenoble (CEA/Grenoble Alpes University)
RNA: Ribonucleic acid, particularly used by cells as an intermediate medium for genes to synthesise the proteins they need.
Adrenocortical carcinoma: Cancer that develops from the cortex of the adrenal gland, which is located above the kidneys. The adrenal glands play an important role in regulating the metabolism of ions and salts and in producing androgens.

Simulating the behaviour of nuclei
Neural networks

Simulating the behaviour of nuclei

— Theoreticians at CEA-Irfu and the CEA’s Military Applications Division have developed an artificial intelligence that can predict the properties of more than 1800 atomic nuclei using an algorithm trained on just 210 nuclei. The accuracy achieved is comparable to that of the state of the art, with a computation time reduced by a factor of ten to a factor of a thousand. It uses a completely new approach: the algorithm does not learn one observable but several “intermediate” measurements calculated by EDF, such as the response of the nuclei to deformations or vibrations. In addition, as the training is limited to 10% of the nuclei, the neural network itself determines (by “active learning”) which nuclei provide it with most information to make its predictions. This approach is already being used by a Franco-Belgian collaboration for studying primordial nucleosynthesis, as well as being used to develop new, more complex EDF interactions. It has also become possible to use neural networks for nuclear dynamic calculations, including fission.

CEA-Irfu: CEA Institute of research into the fundamental laws of the universe
EDF: Energy density functionals, used to calculate the properties of an atom’s nuclei. They describe the interactions between neutrons and protons (which make up the nucleus) as a function of the energy density of the system.

Understanding the “dark” universe and the formation of primordial galaxies
High-performance computing

Understanding the “dark” universe and the formation of primordial galaxies

— ​The Extreme-Horizon collaboration, which involves teams from the CEA’s Fundamental Research Division (Irfu) and Military Applications Division, has produced a completely new simulation of the evolution of cosmic structures (galaxies, stars and supermassive black holes), from a few moments after the Big Bang right up to the present day. It describes the intergalactic regions, which represent 90% of the volume of the universe, at an unprecedented resolution. This simulation was able to refine the cosmological parameters used to explore the large structures of the distant universe, and to describe the processes involved in the formation of ultra-compact massive galaxies when the universe was only 2 to 3 billion years old. The simulation was run on the Joliot-Curie supercomputer, at the CEA’s Very Large Computing Centre (TGCC).

> For more information

Irfu: CEA Institute of research into the fundamental laws of the universe