Dual-Functional SiO x Cross-Point Devices Enabled by Top Electrode Engineering for Ultralow-Current Resistive Memory, Ultrasensitive Sarcosine Sensing, and Artificial Intelligence-Assisted Biomarker Classification

In the pursuit of ultra-low-power neuromorphic and bioelectronic systems, this work investigates two complementary SiOx-based resistive-switching architectures with engineered top electrodes: a Cu/CuAl structure employing metallic-filament CBRAM and an IrOx/SiOx/W cross-point architecture employing vacancy-mediated switching. Introducing a Cu–Al alloy interfacial layer into a Cu/SiOx/W structure enables ultralow-current operation with an energy consumption of ∼0.03 fJ/bit, endurance exceeding 109 program/erase (P/E) cycles, and a memory window (MW) greater than 2500. Further advancement to a three-dimensional IrOx/SiOx/W cross-point architecture significantly improves device reliability, achieving endurance beyond 1010 P/E cycles at operating currents of 10 and 50 μA while maintaining stable switching for more than 103 cycles at an ultralow current of 200 nA with an MW exceeding 3 × 104. Beyond memory functionality, the porous IrOx top electrode enables label-free biomarker sensing through its stable redox states and super-Nernstian pH response. TEM, HRTEM, and FESEM analyses reveal a porous nanostructured IrOx architecture that facilitates efficient oxygen exchange, proton transport, and redox activity. The resulting platform achieves ultrasensitive sarcosine detection with a limit of detection as low as ∼10 fM. To examine the feasibility of automated interpretation of sensing signals, current–time responses were analyzed using a Convolutional Spiking Neural Network (CSNN). The model achieved an overall classification accuracy of 95.8%, together with a sensitivity of 96.02% and specificity of 100%. These results demonstrate that complementary top-electrode-engineered SiOx architectures can provide ultralow-power memory operation, while the IrOx/SiOx/W cross-point platform further enables ultrasensitive biomarker detection and AI-assisted signal analysis, highlighting their potential for future bioelectronic and neuromorphic systems.