Technology · Neuromorphic Computing

Self-Heating Nanopores Let Salt Compute Like Brain Cells

A new device that uses salt and heat — not solid-state electronics — to mimic the brain's memory could reshape how we build energy-efficient computers.

The human brain runs on roughly 20 watts and processes information using ions moving through fluid channels. Conventional computers, by contrast, switch electrons through solid silicon and consume orders of magnitude more power. A team of researchers has now built a device that brings computing one step closer to biology: a fluidic memristor that uses self-generated heat to precipitate and dissolve salt inside nanoscale pores, creating a memory effect that behaves much like a synapse.

How salt and heat become memory

The key mechanism: self-heating-induced blocking

When an electric current passes through a nanopore filled with salt solution, Joule heating raises the temperature enough to cause salt to precipitate, physically blocking the pore. This blocked state represents a high-resistance "off" memory. When the current drops, the salt dissolves again and the pore reopens — a low-resistance "on" state. The device remembers which state it was in, just as a biological synapse strengthens or weakens its connection based on past activity. The work, published in Nature Communications by Fan and colleagues, is the first demonstration of a memristor that achieves resistive switching purely through thermally triggered salt precipitation and dissolution.

Three things worth knowing

Ions instead of electrons: Unlike conventional memristors that use solid materials and charge carriers, this device uses ions in solution — the same type of signal carriers the brain uses. This makes it inherently more compatible with biological systems for future brain-computer interfaces.
Self-heating is the key: The device generates its own heat internally through the current flowing through the nanopore, so no external heater is needed. The precipitation and dissolution cycle is fully reversible and repeatable, giving it reliable memory behavior.
Multiple neuromorphic functions: The nanopore memristor can emulate short-term plasticity, long-term potentiation, and spike-timing-dependent plasticity — three fundamental learning rules observed in biological neurons — all within a single fluidic channel.

Why it matters

Neuromorphic computing — building hardware that mimics the brain — has largely relied on solid-state materials that approximate neural behavior. But the brain is fundamentally a fluidic, ionic system, not an electronic one. A fluidic memristor that genuinely uses ions and heat to create memory bridges the gap between artificial and biological computing more closely than any solid-state device can. Future applications could include low-power edge computing, adaptive sensors, and implantable devices that interface directly with neural tissue without the need for signal conversion between electronic and ionic domains.

The discovery also opens a new materials paradigm: instead of engineering ever-more-complex solid-state structures, researchers can now design computational devices using simple salt solutions and nanoscale channels — a reminder that sometimes the most elegant computing substrate is already inside our heads.