Mimicking Silent Synapse Recruitment: A SiO x /Cu‐Pancake Memristive Device For Analog Neuromorphic Computing

Analog memristive devices are essential for continuous synaptic weight updates in energy‐efficient neuromorphic computing. Achieving robust and predictable multi‐level conductance modulation requires precise physical control over the active switching volume to avoid stochastic instabilities. While various non‐filamentary switching concepts exist, structurally engineering and stabilizing a distributed nanoscale morphology to ensure reliable, area‐wide analog switching remains a fundamental materials challenge. To address this gap, we introduce an electroforming‐free TiN//Cu//TiN device that leverages the controlled thermal evolution of embedded copper into laterally extended, pancake‐like nanoparticles (Cu‐PCs). By combining in situ transmission electron microscopy with potential‐ and time‐resolved impedance spectroscopy, as well as area‐scaling analyses, we demonstrate that this unique morphology creates a highly heterogeneous Schottky barrier landscape. Applying an electrical bias drives local oxygen vacancy redistribution, dynamically lowering these interfacial barriers and progressively recruiting previously inactive, high‐resistance Cu‐PC pathways into a parallel conduction ensemble. This progressive “recruitment” provides an intrinsic physical basis for continuous analog tunability, functionally mimicking the activation of biological silent synapses and establishing targeted metal nanoparticle morphology engineering as a powerful paradigm for robust neuromorphic hardware.