A tiny stack of printed nanomaterials, thinner than a human hair and flexible enough to bend with living tissue, just did something no fabricated device has done before: it fired electrical spikes so lifelike that real brain cells responded as though they were hearing from a neighbor.
Researchers at Northwestern University built the devices by layering graphene and molybdenum disulfide (MoS₂) into a sandwich-like structure, then deposited them onto flexible surfaces using an aerosol-jet printer, a tool that works a bit like a highly precise spray gun for electronics. When voltage was applied, the printed circuits produced sharp electrical pulses that closely mimicked the signals biological neurons use to communicate. The team then aimed those pulses at Purkinje neurons, large cells in the cerebellum that help coordinate movement, in slices of living mouse brain. The cells fired back, confirming the artificial signals carried enough fidelity to be recognized by real neural tissue.
The results, published in Nature Nanotechnology in spring 2026, mark the first time printed neuromorphic hardware has directly stimulated identified neurons in biological tissue. The university’s press materials describe the achievement as a foundational step toward future neuroprosthetics and brain-machine interfaces, while cautioning that clinical applications remain distant.
How the printed neurons work
The key component is a memristor, a device whose electrical resistance changes depending on the voltage history applied to it. That property mirrors how biological synapses strengthen or weaken with repeated activity. In the Northwestern design, the graphene/MoS₂/graphene stack exploits natural defects and grain boundaries in the MoS₂ layer to create the switching behavior that makes memristive spiking possible. Sangwan and Hersam’s earlier research on these material properties, published in Nature, laid the groundwork for the current advance.
By tuning the geometry and layer thickness, the team engineered devices that accumulate incoming electrical pulses and then release a sharp spike once a threshold is crossed, much like a biological neuron integrating signals before firing. Arrays of these units reproduced patterns familiar to neuroscientists: regular firing, bursting, and oscillations. According to the paper, the devices achieved tunable spiking frequencies up to 20 kHz and remained stable through more than one million switching cycles.
What sets this apart from earlier neuromorphic chips etched into rigid silicon is the combination of printing, flexibility, and biological validation. Rigid implants tend to provoke scarring and inflammation when placed against the brain’s soft, convoluted surface. Aerosol-jet printing allows the circuitry to be deposited on bendable substrates that could, in principle, conform to the curves of the brain or spinal cord, reducing that mechanical mismatch.
The biological proof
Purkinje neurons were a deliberate choice. They are among the largest and most electrically active cells in the brain, and their behavior in cerebellar slices is extremely well characterized, making them a reliable test bed. When the printed devices delivered their spike trains, the Purkinje cells responded with measurable activity, confirming that the waveforms’ shape, timing, and amplitude fell within the range biological tissue treats as genuine input.
The concept of using memristors to bridge organic and inorganic systems is not entirely new. A 2020 study demonstrated that memristive synapses could connect biological and silicon spiking neurons in a hybrid circuit. What the Northwestern team added is the manufacturing method (printing rather than cleanroom fabrication), the flexible form factor, and direct stimulation of identified cell types in intact brain tissue rather than cultured neurons on a dish.
What this does not yet prove
Every biological test so far has taken place in mouse brain slices on a lab bench, not inside a living animal. Cerebellar slices are a standard tool in electrophysiology, but they strip the tissue of its blood supply, immune surveillance, and the network-level feedback loops that define a working brain. Whether the printed devices can maintain stable communication inside an intact nervous system, where immune cells attack foreign materials and electrical noise is far more complex, remains an open question.
Long-term biocompatibility is similarly unproven. Surviving a million switching cycles on the bench measures electrical endurance, not biological tolerance. No chronic implantation data have been reported, and the researchers’ own institutional materials frame neuroprosthetic applications as potential implications, not demonstrated outcomes.
There is also the question of direction. The published results confirm the printed devices can stimulate living neurons. Whether the system supports genuine two-way signaling, where biological activity modifies the memristor’s state in a meaningful feedback loop, has not been clearly established for this architecture. Without evidence of closed-loop interaction, the printed neurons are better described as sophisticated stimulators than as full partners in a neural circuit.
Competing approaches add further context. Parallel work on printed spiking neurons built from organic electrochemical transistors relies on ion-mediated mechanisms closer to the brain’s own signaling chemistry and can operate at very low voltages in wet, salty environments. How the MoS₂-based strategy compares in biocompatibility, energy efficiency, and signal fidelity has not been settled by direct comparison, and the best approach may ultimately depend on the specific medical application.
Scaling presents its own challenges. A handful of printed neurons driving identified cells in a brain slice is a crucial proof of principle, but practical neuroprosthetics would demand thousands or millions of channels, each with controllable timing and strength. Aerosol-jet printing is well suited to patterning complex layouts, yet integrating power delivery, data communication, and protective encapsulation at that density remains unexplored territory.
Where printed neurons fit in the larger race
The brain-computer interface field has accelerated sharply in recent years, with companies like Neuralink implanting electrode arrays in human patients and academic labs exploring everything from ultrasound-based stimulation to genetically encoded sensors. Most of those efforts rely on rigid or semi-rigid hardware that records or stimulates but does not itself behave like a neuron. The Northwestern work occupies a different niche: rather than simply listening to or poking the brain, it aims to speak the brain’s own electrical language through devices that can be printed, bent, and potentially scaled.
For people living with paralysis, movement disorders, or neurodegenerative disease, the practical question is how quickly any of this translates to the clinic. The honest answer, based on the published evidence, is that it will not be quick. Animal implantation studies, biocompatibility testing, regulatory review, and human trials each represent years of work. What the Nature Nanotechnology paper establishes is that the foundational engineering barrier, getting a printed device to produce signals a living neuron actually accepts, has been cleared.
Follow-up studies will need to probe chronic implantation in living animals, test whether feedback-driven learning between tissue and device is achievable, and push into higher-order brain regions where neural coding is more complex. Those experiments will determine whether printed neurons become a practical tool for medicine or remain an elegant laboratory demonstration. For now, the gap between a mouse brain slice and a human neuroprosthetic is wide, but for the first time, a printed device has stepped onto the biological side of that divide.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
