Toward Plasmonic Neuronal Architectures at the Nanometer Scale.

We introduce a plasmonic computing platform that unites multiplexed optical inputs, static weighting, and nonlinear activation within a single nanoscale architecture-marking a critical step toward the realization of a plasmonic neural cell. In this approach, the orbital angular momentum (OAM) of light serves as an orthogonal and inherently parallel encoding scheme, enabling distinct optical channels to be directed into individual plasmonic waveguides. The weighting of input signals is governed by precisely engineered nanoscale gaps, whereas the system's nonlinear activation is revealed through two-photon photoemission, spatially resolved using photoemission electron microscopy (PEEM). By confining electromagnetic fields far below the diffraction limit and minimizing surface plasmon polariton (SPP) propagation lengths, the design mitigates optical absorption losses while fully exploiting the unique field enhancement capabilities of plasmonic nanostructures. This physically integrated platform is compatible with scalable nanofabrication defined through a two-step electron-beam lithography (EBL) process and provides a compact, energy-efficient building block for ultrafast neuromorphic photonic circuits. Together, these results outline a tangible route toward dense plasmonic neural networks capable of performing all-optical information processing at the native speed of light.