Argon-Ion Plasma Defect Engineering of GaN Memristors for Multilevel Conductance and Quantized Synapses in Leaky-Integrate & Fire Networks

Overview

Artificial intelligence hardware is increasingly constrained by the energy cost of data movement: a single DRAM access can consume 200× the energy of an on-chip compute operation. Neuromorphic computing reduces this bottleneck by integrating memory and computation. However, a key challenge is low precision, as learning is often demonstrated using idealized synaptic weights rather than actual physical states of the devices. This study investigated whether defect-engineered gallium nitride memristors could support hardware-constrained neuromorphic learning by fabricating and characterizing GaN memristors, constraining a Spiking Neural Network to measured conductance states. Memristive behavior was induced in GaN by argon-ion plasma treatment, introducing lattice defects and tuning charge trapping behavior. Dynamic I-V sweeps and pulse programming were used to quantify resistive switching, analog programmability, and state stability. Pulse-train measurements were converted into discrete conductance states and implemented as differential conductance pairs (G+ - G-). Treated devices showed reproducible resistive switching, pulse-dependent conductance modulation, and 27-29 stable monotonic conductance levels; when integrated into a logic circuit, they also enabled adaptive learning. The device-constrained network achieved 86.8% test accuracy on the MNIST handwritten-digit task. Spike-driven inference reduced the estimated synaptic workload by ~6×, while weight storage decreased ~3.2× relative to float32 parameters. This study demonstrates that defect-engineered GaN memristors can support hardware-constrained neuromorphic learning using experimentally measured conductance states, advancing toward low-power edge AI, neuromorphic sensors, and embedded intelligent systems.

Awards (4)

• Fourth Award of $600 $600
• King Abdulaziz & his Companions Foundation for Giftedness and Creativity: Full Scholarship from King Fahd University of Petroleum and Minerals(KFUPM) (and a $400 cash prize) $400
• King Abdulaziz & his Companions Foundation for Giftedness and Creativity: NOT TO BE READ -- $400 cash prize for each Full Scholarship from King Fahd University award recipient $400
• Society Alumni Special Award: Society for Science Alumni Special Award

Competition history

• ISEF 2026 Materials Science · Entry MATS014 Affiliated fair in Republic of Moldova

Resources

• ISEF abstract page ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Materials Science and Engineering projects
• ISEF Materials Science and Engineering projects
• Every ISEF 2026 project