News: Microelectronics
16 February 2021
NeuroSys funded by BMBF initiative Clusters4Future
Germany’s Federal Ministry of Education and Research (BMBF) has announced the winners of the Clusters4Future ideas competition. The funded clusters include the future clusters ‘Hydrogen’ and ‘NeuroSys - Neuromorphic Hardware for Autonomous Artificial Intelligence Systems’ involving RWTH Aachen University and partners from the region.
As a supplier of deposition equipment, Aixtron SE of Herzogenrath, near Aachen, Germany is a partner in NeuroSys, developing, building and optimizing the necessary metal-organic chemical vapor deposition (MOCVD) tool for the new application. Work on NeuroSys is expected to run for several years.
NeuroSys researches adaptive and energy-efficient neuromorphic AI chips, with the aim of allowing intelligent and resource-saving on-site data processing and thus creating an essential prerequisite for artificial intelligence (AI) applications. However, the use of AI poses challenges such as high CO2 emissions, further exacerbating climate issues. They arise when training large neural networks based on modern graphics processing units (GPUs) using deep learning methods. GPU-based neural networks are therefore ecologically unsustainable.
Neuromorphic systems for resource-efficient on-site data processing
Resource-efficient neuromorphic hardware that makes neural networks more efficient and provides for data security as a design component is therefore becoming key to the widespread use of AI. This is especially true for application areas in autonomous vehicles, medical technology, and sensor networks for intelligent production or urban regions.
Neuromorphic systems are modeled on the two basic building blocks of the human brain: neurons and synapses. By integrating new materials with specific properties, they can ideally perform on-site processing of data in a way that conserves resources. This is summarized under the keyword memristive (from ‘memory’ and ‘resistor’ for electrical resistance).
RWTH Aachen University and Research Center Jülich have already been able to demonstrate the functionality of neuromorphic devices made of memristive materials. However, there are no pilot lines or production capacities worldwide for the manufacture or integration of neuromorphic chips on an industrial scale.
Chips based on novel 2D materials
Also, the system of hardware, design, algorithms and application-driven software must work together to take advantage of the great benefits of neuromorphic hardware. What is required, therefore, is a paradigm shift with the opportunity to take a leading position in this new technology, says Aixtron. NeuroSys aims to work out the crucial prerequisites.
Aixtron is at the beginning of the value chain in the project. Very thin layers of novel 2D materials, just a few atomic layers thick, are required as the material base for the neuromorphic chips. The production of these materials on the largest possible wafers is Aixtron’s core competence.
“Creating GPU-based neural networks that are environmentally sustainable is a major challenge. We are pleased to be able to make our contribution in the area of development and provision of the material base,” says Dr Michael Heuken, VP corporate research & development at Aixtron and professor at RWTH Aachen University. “As part of this and other projects, we will develop the world’s most powerful technology platform for the industrial production of these materials and operate it in our laboratory in Herzogenrath.”
Key element for structural change in Rheinischen Revier
“The Future Cluster is a great opportunity for the Aachen-Jülich region, especially in connection with the structural change in the ‘Rheinischen Revier’,” says research work coordinator professor Max Lemme, chair of Electronic Components at the RWTH Aachen University and managing director of AMO GmbH. “We are stepping up to transfer excellent science to companies and startups in the region. Our vision is to set up a production line in the Aachen region. There, the co-integration of neuromorphic functions using new materials into conventional silicon technology will then take place.”
In addition to RWTH Aachen University, the Research Center Jülich, AMO GmbH, IHK Aachen, the companies AixACCT Systems GmbH, Aixtron SE, AppTek GmbH, ELMOS Semiconductor SE, RWTH Innovation GmbH and STAR Healthcare Management are involved in NeuroSys. In addition, the start-ups AiXscale Photonics UG, Black Semiconductor GmbH, Clinomic GmbH and Gremse-IT GmbH are also involved.
Clusters4Future is part of the German government’s High-Tech Strategy 2025. The open-topic competition focuses on regional innovation networks that combine the strengths of the players, tap into emerging fields of innovation and develop solutions for the challenges of the future.
