Brain-Inspired Tech: Synaptic Transistors Mimic Learning with Light and Electricity
"New research explores how proton-gated transistors can replicate brain functions, opening doors to advanced AI and memory technologies."
Our brains are incredibly efficient at handling complex problems, a feat that still challenges modern computers. The secret lies in the vast network of neurons connected by biological synapses. Researchers are working hard to understand these synapses and replicate their functions in electronic devices. The goal? To create computers that are more like our brains – capable of parallelism, structural plasticity, and robustness.
Traditional electronic devices like memristors and transistors are being adapted to mimic synaptic behavior. Synaptic transistors, in particular, offer a promising approach. These three-terminal devices can transmit signals through a semiconductor channel and simultaneously learn by adjusting the interaction between the channel and an insulating layer. This is closer to how our brains work compared to two-terminal devices.
Now, scientists are exploring the use of UV light to influence these synaptic transistors, drawing inspiration from studies showing that moderate UV exposure can enhance learning and memory in the brain. A new study combines a nanogranular silicon dioxide (SiO2) layer with an indium-gallium-zinc oxide (IGZO) channel in a synaptic transistor, demonstrating the potential of this approach for creating more brain-like computing systems.
From Polymers to Neuromorphic Systems
A 2022 review describes polymer synaptic transistors as a developing route toward energy-efficient neuromorphic computing. It highlights designs spanning materials, device structures, and circuits, including chargeable dielectrics and semiconductors for synaptic functions. Organic transistors are also being explored for emulating synaptic and neuronal behaviors because of their low power consumption, flexibility, stretchability, and biocompatibility. More recent work surveys manufacturing pathways from device fabrication through advanced circuit integration.
How Does This Synaptic Transistor Work?
The key to this innovation is the use of amorphous indium-gallium-zinc oxide (a-IGZO) as the channel material, combined with nanogranular SiO2 as the gate oxide. This combination allows the transistor to mimic the short-term plasticity and short-term memory functions of a biological synapse. The device operates at low energy consumption, around 1.08 picojoules per pulse, and at voltages within 100 mV.
- Electrical Stimulus: The transistor exhibits both short-term potentiation (STP) and short-term depression (STD), meaning its response to a stimulus can either increase or decrease over time.
- Pulse Number: The amplitude of the excitatory postsynaptic current (EPSC) changes with the number of pulses, following a saturating exponential function.
- UV Light Influence: The UV light's frequency has a significant effect on the synapse's plasticity, promoting paired-pulse facilitation (PPF) and enhancing the EPSC gain. Other parameters like intensity and duration have less impact.
Light and Electrical Plasticity
Recent research has focused on amorphous oxide semiconductors, whose persistent photoconductivity is considered useful for optoelectronic synaptic devices. One reported device combines transparent IGZO with a proton-conductor gate to respond to electrical and ultraviolet-light stimuli. Another technique enables perception and memory from ultraviolet through visible light, with a maximum relaxation time of 129.41 seconds, a paired-pulse facilitation index of 216.40%, and an A20/A1 gain index of 3232.51%. Two-dimensional-material neuromorphic systems are likewise being investigated for data-intensive applications.
Promise Without a Complete Record
The supplied reviews present oxide and transistor-based devices as promising rather than documenting a specific failed prototype or definitive performance limit. Oxide transistors are described as offering low-temperature processability, large-area uniformity, low off-state leakage, and defect-mediated temporal dynamics. Another review says transistor-based devices are suitable for emulating synaptic functions because they provide synergistic control over synaptic-weight changes. These descriptions establish potential advantages, but the supplied material does not provide enough evidence to quantify failures or resolve remaining implementation challenges.
Synapses, Memristors, and Conventional Chips
A 2026 comparison report identifies synaptic transistors and memristors as two prominent device technologies designed to emulate biological synapses. It places both within neuromorphic computing's effort to address energy-efficiency and parallel-processing limitations associated with traditional von Neumann architecture. A separate comparison describes the synaptic-transistor-versus-SoC sector as nascent, with market potential estimated in the billions but still small relative to traditional semiconductor markets. The supplied biological study provides a different reference point: in rat SON neurons, 20 μmol·L−1 morphine reduced EPSC frequency by 65% and mEPSC frequency by 45%.
The Future of Brain-Like Computing
This research demonstrates the potential of proton-gated synaptic transistors for mimicking the brain's learning and memory functions. By combining electrical and light stimuli, these devices offer a new approach to creating more efficient and sophisticated AI systems.
Toward Brain-Level Efficiency
A May 8, 2026 report describes researchers developing organic synaptic transistors for neuromorphic computing. The stated aim is to match the human brain's 20-watt efficiency for artificial intelligence. The report frames these devices as part of an effort to improve the efficiency of future AI hardware.
While challenges remain in fabricating complex synaptic transistor networks, this work paves the way for future research in areas like multi-presynaptic inputs and organic wearable insulators. The nanogranular SiO2/IGZO structure offers a promising platform for advancing synaptic electronics.
Ultimately, this research brings us closer to a future where computers can learn, adapt, and solve problems more like our own brains. It's a significant step towards truly intelligent machines.