Iontronic devices use ions as signal carriers, making them promising candidates for interfaces between biological nerves and artificial systems. Many ionic neuromorphic devices rely on nanoconfinement to control ion transport, bu the limited range of interactions at the interfaces and ionic crosstalk during integration restrict their scalability and flexibility.

In a study published in Nature Electronics, a team led by ZHAO Ziguang from the Hangzhou Institute of Medicine of the Chinese Academy of Sciences developed a multiphasic-gel-based ion transporter (GIT) system that uses ions to reproduce features of neural signal processing. This system combines low-energy operation with synapse-like learning and memory, and has been integrated with a rat nerve to demonstrate biologically driven signal processing.

Researchers developed the GIT system with cascaded heterointerface ion traps. Electrostatic interactions and steric hindrance across these interfaces regulated ion accumulation and dissipation, allowing the GIT system to switch between high- and low-conductance states. The system exhibited bipolar ionic rectification, with rectification ratios exceeding 10³, and supported neuromorphic responses under millivolt-scale electrical stimulation.

By varying the strength and timing of input spikes, researchers demonstrated synapse-like potentiation and depression, together with short- and long-term memory effects. Different GIT channels reproduced Hebbian and anti-Hebbian learning behaviours, while a dual-channel configuration of the GIT system implemented the Bienenstock–Cooper–Munro learning rule. The energy consumption was as low as 0.61 pJ per spike.

Besides, researchers built a neuro-iontronic processor by integrating two parallel GIT channels and connected it to proximal and distal segments of a transected sciatic nerve in a rat. Driven by endogenous neural potentials, the processor operated without an external power source or supplementary electrical signals. It bridged the interrupted signal pathway and modulated downstream neural activity through potentiation or depression. Electromyographic recordings showed corresponding changes in neuromuscular activation.

The findings of this study demonstrate an approach for ionic neuromorphic processing that the GIT system can interact directly with biological neural signals. The GIT system could support the development of neural interfaces, smart sensors and biohybrid computing systems.