Engineers at Northwestern University have created printed artificial neurons that generate realistic electrical signals and can communicate directly with living brain cells. The flexible, low-cost devices were tested on slices of mouse brain tissue, where they successfully triggered responses from real neurons. This advance brings closer the development of brain-machine interfaces and neuroprosthetics for applications such as restoring hearing, vision, or movement. It also supports more energy-efficient computing inspired by the brain, which uses far less power than current digital systems for complex tasks.
Artificial intelligence (AI) relies on processing large amounts of data, which demands huge amounts of electricity. The brain performs similar work with much greater efficiency, using soft, three-dimensional networks of diverse neurons that constantly adapt and form new connections. In contrast, traditional computers use rigid silicon chips with billions of identical transistors that do not change after manufacturing.
From rigid silicon to dynamic brains
The engineers developed the artificial neurons using printable electronic inks made from nanoscale flakes of molybdenum disulfide, a semiconductor material that conducts electricity under certain conditions, and graphene, a highly conductive carbon material. These inks were deposited onto flexible polymer sheets through aerosol jet printing, a precise technique that sprays fine droplets of material. Instead of removing a stabilizing polymer completely, the process partially decomposes it to create narrow conductive pathways. This results in devices that produce varied neuron-like signals, including single spikes, continuous firing, and bursting patterns, rather than simple pulses.
When electrical signals from the printed devices were applied to mouse brain tissue, the timing and shape of the spikes closely matched those of biological neurons, activating real neural circuits. The approach is also low-waste and environmentally friendly because material is added only where needed.
Such brain-like hardware could reduce the massive energy and cooling demands of data centers that support AI. The work, published in Nature Nanotechnology, demonstrates a practical step toward electronics that interface seamlessly with the nervous system and toward computing systems that mimic the brain’s efficiency.