Summary: Engineers have reached a major milestone in bio-electronics by printing artificial neurons that can directly communicate with living brain tissue. The team developed flexible, low-cost printed devices that produce electrical signals closely matching biological spikes, and these signals successfully drove responses in mouse cerebellar tissue.
Unlike conventional silicon chips, these printed neurons reproduce key features of brain signaling and energy efficiency, opening paths to advanced neuroprosthetics and brain-inspired computing that use far less power than current AI data centers.
Key Findings
- Biocompatible signaling: In collaboration with the lab of Indira M. Raman, the printed devices delivered voltage spikes to mouse cerebellar slices that elicited responses from living neurons, demonstrating physiological compatibility.
- High energy efficiency: By emulating neuronal signaling, the devices point toward computing hardware that could handle large-scale data and AI workloads with orders-of-magnitude lower power consumption than conventional digital systems.
- Structural mimicry: Instead of uniform, fixed transistors, these printed neurons are heterogeneous and dynamic, better reflecting the soft, three-dimensional architecture and adaptable connectivity of biological networks.
- Additive manufacturing: The aerosol jet printing approach deposits material only where needed, cutting waste and enabling low-cost, scalable production of flexible circuitry.
Source: Northwestern University
Northwestern engineers printed artificial neurons that can interact directly with the brain.
A new study from Northwestern University demonstrates flexible, printed devices that generate neuron-like electrical waveforms capable of activating living brain cells. When connected to slices of mouse cerebellum, the artificial neurons produced spikes matching biological timing and shape, reliably triggering activity in real neurons and neural circuits.
The work advances electronics that can communicate directly with the nervous system and supports applications such as hearing and visual prosthetics, motor neuroprostheses, and brain–machine interfaces. It also lays the foundation for neuromorphic hardware that replicates how biological neurons signal, allowing complex computation with much lower energy demands.
The study appears in Nature Nanotechnology.
“Artificial intelligence is driving rising demand for compute and data,” said Mark C. Hersam of Northwestern, who led the project. “Training modern AI consumes enormous power. The brain operates with vastly greater energy efficiency, so designing hardware inspired by its mechanisms can dramatically reduce energy and cooling needs while enabling more capable systems.”
Hersam, a materials science and engineering professor at Northwestern’s McCormick School, co-led the work with Vinod K. Sangwan, a research associate professor at McCormick.
From rigid silicon to dynamic brain-like systems
Traditional computing scales complexity by packing billions of identical transistors onto rigid, two-dimensional silicon chips. Each device behaves the same and the system is fixed after fabrication. Biological brains, by contrast, rely on diverse neuron types organized in soft, three-dimensional networks that constantly form and reshape connections.
Hersam explained that moving toward brain-like hardware requires new materials and manufacturing strategies that create heterogeneous, adaptive devices rather than uniform, static components.
Existing artificial neurons often generate simplified signals, so engineers must combine many devices to approximate biological behavior. To overcome this, the Northwestern team developed printed neurons capable of producing a rich repertoire of spike patterns, reducing the number of components needed for complex signaling.
Turning an imperfection into an advantage
The printed neurons are made from specially formulated electronic inks: nanoscale flakes of molybdenum disulfide (MoS2) for semiconducting elements and graphene for conduction, printed onto flexible polymer substrates using aerosol jet printing. A stabilizing polymer in the ink, previously seen as an unwanted residue, was intentionally partially decomposed to introduce useful, brain-like electrical behavior.
Rather than completely removing the polymer, the researchers allowed controlled partial decomposition. Passing current through the device promotes further localized decomposition, creating narrow conductive filaments. Those filaments concentrate current in space and time, producing sudden, neuron-like responses that mimic biological spiking.
These printed memristive networks generate diverse signaling patterns—single spikes, sustained firing, and bursting—closely resembling the temporal dynamics of living neurons. That richer signaling lets each device encode more information and perform more complex operations, improving efficiency and reducing system size.
Testing with living brain tissue
To validate biointeraction, the team partnered with Indira M. Raman’s lab to apply artificial neuron output to mouse cerebellar slices. The printed spikes matched key biological characteristics, including timing and waveform shape, and reliably elicited activity in Purkinje neurons and related circuits.
Hersam noted that previous approaches either produced spikes that were too slow or devices that switched too quickly. The printed devices operate within a physiological timescale and waveform shape that allows direct interaction with living neurons.
Manufacturing benefits include low cost, additive production that minimizes material waste, and compatibility with flexible substrates—advantages for both implantable neuroprosthetics and scalable neuromorphic hardware.
Hersam emphasized the urgency of energy-efficient hardware as AI scales: current trends in data-center power and water use are unsustainable, and brain-inspired electronics offer a promising alternative.
Funding: The research, titled “Multi-order complexity spiking neurons enabled by printed MoS2 memristive nanosheet networks,” was supported by the National Science Foundation.
Key Questions Answered:
A: Not a whole brain. These printed devices act as electrical translators that can interface with biological tissue. They could improve cochlear implants, visual prosthetics, and brain–computer interfaces that help paralyzed patients control devices with neural signals.
A: Current AI workloads require massive energy and cooling resources. Brain-inspired hardware that performs complex tasks while using far less power would reduce the environmental and infrastructure burdens of large-scale AI systems.
A: Software neural networks are algorithms running on conventional hardware. These printed neurons are physical, flexible circuits with built-in memory and dynamics, closer to biological “wetware” and able to interact directly with living tissue.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full.
- Additional context was added by editorial staff.
About this neurotech research news
Author: Amanda Morris
Source: Northwestern University
Contact: Amanda Morris – Northwestern University
Image: Image credit: Neuroscience News
Original Research: Closed access. “Printed MoS2 memristive nanosheet networks for spiking neurons with multi-order complexity” by Shreyash S. Hadke et al., Nature Nanotechnology. DOI: 10.1038/s41565-026-02149-6
Abstract
Printed MoS2 memristive nanosheet networks for spiking neurons with multi-order complexity
Reproducing the rich dynamical behavior of biological spiking neurons is critical for neuromorphic hardware and biohybrid interfaces, but scalable printed devices with physiologically relevant spiking characteristics have been elusive. This work demonstrates aerosol-jet-printed memristive networks of MoS2 nanosheets that exhibit thermally activated filamentary switching and snap-back negative differential resistance. These effects produce volatile threshold switching in fully printed graphene/MoS2/graphene devices on flexible substrates.
Thermal imaging and circuit modeling show that current-constricted filaments formed by Joule heating drive the nonlinear switching dynamics. The printed memristors enable oscillatory and spiking neuron circuits with tunable frequencies up to 20 kHz and stable operation beyond 106 cycles. Simple neuristor circuits realize first-, second-, and third-order spiking behaviors, including integrate-and-fire dynamics, spike latency, tonic firing, class 1 excitability, tonic bursting, and phasic responses. The generated spike waveforms match physiological timescales and stimulate Purkinje neurons in mouse cerebellar slices. These results position printed nanosheet memristive networks as a scalable platform for bio-realistic neuromorphic hardware and flexible brain–machine interfaces.