"The field of microelectronics will write a new page"

“The turning point is coming,” CEA-Leti CEO Sébastien Dauvé told EE Times Europe. “Just four years ago, semiconductors were considered just a commodity. The semiconductor industry is now highly politicized, and what we’re seeing between the U.S. and China The situation in between is likely to reshape the hyper-optimized world we have had until now and the already internationalized semiconductor value chain is being restructured by the massive investments we are seeing."

The EU CHIP Act, the US CHIP Act, and other similar efforts around the world aim to unleash innovation and accelerate the transition of semiconductor products from research (labs) to design and manufacturing (fabs).

The recent death of Gordon Moore, co-founder of Intel and creator of Moore's Law, comes at a critical moment in the history of microelectronics. “Gordon Moore defined the laws of evolution for microelectronics that have never been denied for more than 60 years,” said Dauvé. "Even as efforts continue to pursue Moore's Law, no one can ignore physical limitations and ever-increasing costs. "The next few years are likely to be a new chapter in the field of microelectronics," Dauvé said. "We can see that the traditional focus on More Moore's complements Beyond Moore are becoming increasingly important as embedded systems don't necessarily require the most advanced nodes. "

The More Moore concept is about cutting-edge electronics and producing smaller, higher-performance transistors for supercomputers (CPUs and GPUs) to make them more powerful and energy-efficient. The Beyond Moore concept refers to power electronics technologies involving imagers, sensors, LEDs, 5G telecoms at high frequencies or less advanced nodes. For more than 20 years, CEA-Leti has taken a stance beyond Moore, like other European microelectronics groups such as STMicroelectronics, NXP and Infineon.

"Today's technology risks are likely to be well balanced in the next few years between these two main areas (More Moore and More than Moore)," Dauvé said.

Quantum computing can solve problems that today's supercomputers cannot. A variety of materials and methods have been explored so far, and CEA-Leti is developing a scalable quantum processor based on state-of-the-art silicon technology.

"We are not as advanced as other technologies, especially in the field of entanglement," Dauvé said. "However, we believe that with the advantages of the entire microelectronics toolbox, if we succeed in creating the basic unit of quantum computing tomorrow, we will be able to easily replicate and scale to thousands of qubits."

CEA-Leti is collaborating with the CEA Department of Fundamental Research (DRF) and the French National Center for Scientific Research (CNRS). “Several teams work together in fundamental and technical research, and in this context Maud Vinet, who is responsible for most of the internal projects, wanted to create a start-up to accelerate her work, and we encouraged her Do it so.”

In late November 2022, Sequance was spun off from CEA-Leti and CNRS to develop and commercialize the first million-qubit quantum computer. The startup leverages proven semiconductor technology to bring this operational quantum computer to market.

A week later, at IEDM 2022, Vinet, then Director of Quantum Hardware Projects at CEA-Leti and CEO of Siquance, demonstrated fully depleted silicon-on-insulator (FD-SOI) technology to CEA-Leti and CNRS, enabling fully fault-tolerant quantum Compute using very large scale integration (VLSI) manufacturing and design technologies.

In her talk, Vinet explained that fully fault-tolerant computing allows for the implementation of algorithms that assume qubits are perfect, and suggested that these "perfect qubits, also known as logical qubits, would be composed of many more physical qubits to achieve Quantum error correction.”

The newspaper said, “FDSOI technology and its back gate provide a way to move charge away from the interface in the qubit on the one hand and to center the Vt of the transistor in the control electronics at low temperatures. It is therefore A unique option for designing and manufacturing high-performance quantum systems on a chip, (and) CEA-Leti, CEA-IRIG, CNRS Institut Néel and their spin-off Siquance are leveraging these FD-SOI capabilities to advance the most advanced quantum technologies in VLSI technology calculate."

As embedded AI algorithms become more complex, they also become more data-intensive. CEA-Leti is currently researching new computing methods for secure, low-latency, low-power IoT solutions.

“We are currently working on completely new computing architectures,” Dauvé said. “With in-memory computing, we will interweave memory and transistors in a more advanced way, greatly reducing the time required to move between memory blocks and computing blocks, which can bring significant gains from an energy consumption perspective. ”

CEA-Leti developed gate-all-around (GAA) nanosheet fabrication devices as an alternative to FinFET technology for high-performance computing (HPC) applications. Dauvé said GAA technology is suitable for the most advanced nodes and can support CEA-Leti's in-memory computing method.

At IEDM 2020, CEA-Leti published two papers outlining the advantages of combining 3D architectures with resistive random access memory (RRAM) for in-memory computing, as well as their applications in edge AI and neural networks.

"Today, storage-class memories like high-density 3D crossbar RRAM are promising for applications requiring large amounts of on-chip memory," says the first paper, "3D RRAMs with Gate-All-Around (GAA) Stacked Nanosheet Transistors for In-Memory" explains. RRAM is a leading candidate due to its high density, good scalability, low operating voltage, and ease of integration with CMOS devices. Another attractive aspect of RRAMs is their ability to perform primitive Boolean logic operations for in-memory and neuromorphic computing. However, if the 1T1R design is the most reliable architecture for IMC, the cell size is still limited by traditional access transistors. "

TSMC, Samsung and Intel have announced that they will use GAA transistors in the next few years.

CEA-Leti takes a clear-eyed approach aimed at maximizing performance within a given budget or limited resources. It also encourages researchers to think differently, considering lifecycle analysis from design to application rather than just performance/cost analysis.

“For many of our projects, we conduct life cycle analyzes to quickly identify key materials that we will try to replace or technology options that ultimately prove to have a much lower carbon impact, both in terms of production and usage,” Dauvé explain. "We have started publishing on this topic and have come up with some interesting results on the design of beamforming antennas. We have proposed an architecture based on a technology that is slightly different from what is typically employed and can Impacting the entire life cycle of these antennas really creates momentum for our internal teams.”

CEA-Leti also implements the concept of collaborative optimization of applied technologies. It aims to enhance end-of-life management of electronic equipment and related databases to ensure life cycle analysis and improve optimization of environmental impacts at system and usage levels.

When asked about employees’ intrinsic green motivation, Dauvé said: “We can sense that engineers and researchers are very keen to take concrete action, and I would say this is starting to become a habit in our innovation practices.”

CEA-Leti has also joined SEMI's Semiconductor Climate Alliance to address environmental challenges from the lab to the fab. “The digital industry faces serious challenges not only in terms of its own energy consumption, but also in its ability to provide more sustainable solutions, with the foundries themselves in terms of consumption of water, energy, gas, etc. Facing serious challenges.”