Classic optimization problems – such as the famous “Traveling Salesman Problem,” the optimal partitioning of networks (Maxcut), or channel assignment in 5G networks – belong to the class of so-called NP-complete problems. This means that as the problem size increases, the computational effort required for classical computers rises exponentially, so that even high-performance computers eventually reach their limits. To make such problems more tangible, they can be represented as graphs and formulated using the so-called Ising model – a physical model that can be used to determine the energetically most favorable, i.e., “best,” state of a system. “Ising machines” offer a promising way to quickly find this minimum: these are special analog computers in which the phase of electronic oscillators maps the individual states of a graph, while weighted couplings between the oscillators represent its edges.

The challenge: With today’s chip technology, the necessary dense yet extensive interconnections are virtually impossible to implement. This is precisely where the German-Taiwanese research project “MesMerIsing” from TU Darmstadt and National Cheng Kung University comes in – “Memristor-based Machine for Rapidly Solving Optimization Problems in the Ising Form.” The researchers are relying on memristors – electronic components that, unlike classical transistors, can “remember” past states analogously and permanently. Applied directly on the chip, they could serve as fast, space- and energy-efficient coupling elements that also function at room temperature – without the need for complex cooling systems, such as those required by quantum computers.

The results of the research project – in which TU professors Klaus Hofmann and Christian Hochberger from the Department of Electrical Engineering and Information Technology, as well as Lambert Alff from the Department of Materials- and Geosciences (all part of the “Matter and Materials” research field), are participating – are relevant in the long term for numerous application areas: financial applications, drug research, supply chains and logistics, as well as AI and machine learning methods or classical image recognition. The chosen approach – specialized hardware instead of traditional processors – promises speeds several orders of magnitude higher while consuming significantly less energy.

The project strengthens the joint International Joint Research Lab (IJRL) “Memristor Technology” between TU Darmstadt and NCKU. The collaboration with Taiwan, one of the world’s leading semiconductor hubs, also bolsters the development of new microelectronics as a key technology in line with the German federal government’s high-tech agenda.

“MesMerIsing” is funded by the Federal Ministry of Research, Technology and Space (BMFTR) as part of research and innovation cooperation with Taiwan in the field of artificial intelligence. TU Darmstadt will receive approximately 1.195 million euros over a three-year period for this project; the Taiwanese partners will be funded in parallel and separately by the National Science and Technology Council (NSTC).