Hydrogen-Assisted Asymmetric and Nonlinear Memristor Array for Reconfigurable Olfactory Graph Networks

Abstract

Memristor-based olfactory systems have attracted significant interest. However, a multifunctional memristor array capable of sensing, memory, and computation has not been realized. This study develops a selector-less crossbar array (CBA) composed of Pt/HfO2 nanorods/TiN memristors, termed “chemo-memristive” devices, that exhibits asymmetric current–voltage (I–V) characteristics under a hydrogen (H2) atmosphere. H2 exposure creates oxygen vacancies (VO) in the nanogap, corresponding to the ruptured filament region. The VO-H complexes form shallow traps that enable trap-assisted conduction under the TiN-injection polarity, thereby switching the I–V response from symmetric to a polarity-dependent, asymmetric one. This yields an H2‑assisted intermediate‑resistance state and enables analog resistance tuning via NG widening. Hence, precise conductance modulation and cell-selective readout were achieved by exploiting the forward-reverse current asymmetry, as validated in selector-free operation of a 3 × 3 CBA. Modified National Institute of Standards and Technology digit pattern-recognition simulations demonstrate high inference accuracy (>94%) with highly linear and symmetrical conductance modulation, suitable for large-scale arrays. The adjustable I–V properties allow an electrically reconfigurable olfactory network that can process H2 flow patterns using high-dimensional graph features. A single H2‑assisted CBA integrates selective sensing, which reinforces intended paths, with analog in‑memory computation, enabling combined neuromorphic and electronic‑olfaction functionality.