Plasmonic Neuronal Architectures at the Nanometer Scale

Researchers at the RPTU University Kaiserslautern-Landau and the University of Augsburg have unveiled a novel plasmonic architecture that incorporates the core functions of an artificial neuron on a nanoscale. By utilising light and surface plasmons, they have achieved the processing of multiple input signals in parallel, the selective weighing of these signals, and their subsequent evaluation — all within a single component. Innovative aspects include the use of the orbital angular momentum of light as an efficient encoding scheme. These results represent a significant step towards the development of ultrafast, energy-efficient neuromorphic networks. In the long term, such concepts could lay the basis for ultrafast, purely optical information processing.

New Architecture for “Plasmonic Neurons” Unveiled
An international research team involving the RPTU UniversityKaiserslautern-Landau has taken an important step toward ultrafast, light-based computers. In a recently published paper, the scientists present a novel concept for so-called plasmonic neural architectures—tiny building blocks that function like artificial nerve cells but operate using light instead of electrons.

The research focuses on developing a 'plasmonic neuron' at the nanoscale. The aim is to incorporate the core functions of biological neurons — receiving, weighing and processing signals — directly into a physical structure.

Light as a computing unit
Rather than using electrical signals, the system uses surface plasmons, which are light waves that travel along metal surfaces. These allow information to be transported extremely quickly and in the smallest of spaces. One notable feature is the use of the orbital angular momentum (OAM) of light, which enables multiple signals to be encoded and processed simultaneously.

For the first time, researchers have demonstrated an integrated platform combining all three key steps of an artificial neuron. Multiple input signals are fed in simultaneously and processed separately, with the strength of each signal precisely adjusted via nanoscale gaps in the waveguides. The result is then evaluated via a nonlinear photoelectric effect, similar to activation functions in neural networks. This integration into a single nanoscale component is considered a significant breakthrough, as previous approaches often required several separate components.

Outlook: Artificial neural networks at the speed of light
In the long term, this new architecture could pave the way for light-based artificial neural networks. Such systems could offer significantly higher speeds and energy efficiency than today's computers, which are limited by the so-called von Neumann bottleneck. The authors see great potential in this fundamental study. In the future, complex, scalable networks could emerge that process information entirely optically — and thus at nearly the speed of light. Plasmonic neural architectures could therefore be pivotal in the development of the next generation of ultra-fast, energy-efficient computers.

The study:
Christopher G. O. Weiß, Tobias Eul, Emily Kruel, Mario F. Pfeiffer, Bert Lägel, Benjamin Stadtmüller, Martin Aeschlimann
“Toward Plasmonic Neuronal Architectures at the Nanometer Scale,”
Nanophotonics: e70066. (2026) https://doi.org/10.1002/nap2.70066.