A team at Northwestern University has fabricated artificial neurons by printing them onto flexible plastic, and the tiny devices fire electrical spikes so similar to those of real brain cells that researchers believe they could eventually form a direct link between electronics and living neural tissue. The work, published in Nature Nanotechnology in early 2026, represents one of the first times printed circuits have replicated the complex firing patterns of biological neurons rather than producing simple on-off pulses.
The devices are built from sheets of molybdenum disulfide (MoS₂), each only a few atoms thick, sandwiched between layers of graphene. Using a technique called aerosol-jet printing, the Northwestern group deposited these stacks onto bendable substrates at room temperature, bypassing the rigid silicon wafers and high-heat processes that conventional chip fabrication demands. The result: flexible memristive circuits that generate voltage spikes and oscillations at tunable frequencies up to 20 kHz, covering and exceeding the range at which most biological neurons operate.
Why the firing patterns matter
Real neurons do not simply switch on and off. They encode information through the timing, rhythm, and bursting patterns of their electrical spikes. Earlier artificial neurons built from memristors, devices whose resistance changes depending on the voltage history applied to them, could produce basic oscillations but struggled to match that complexity.
The Northwestern devices go further. By adjusting voltage levels, device geometry, and how individual elements are connected, the team led by materials scientists Vinod K. Sangwan and Mark C. Hersam coaxed the printed circuits into tonic spiking, bursting, and other multi-pattern behaviors that neuroscientists associate with real neural computation. The underlying mechanism is the formation and dissolution of tiny conductive filaments inside the MoS₂ layer, a process that creates the volatile threshold switching needed for repetitive firing.
Because the materials are atomically thin, the circuits can be packed densely without the bulk of traditional silicon hardware. And because the substrate bends, a sheet of artificial neurons could, in principle, conform to the curved, soft surface of a brain or spinal cord, a basic requirement for any future implantable interface.
The research behind the result
This paper did not appear out of nowhere. Sangwan and Hersam have spent years building the case for two-dimensional semiconductors as a platform for brain-inspired electronics. An earlier study by the pair, cataloged through the U.S. Department of Energy’s Office of Scientific and Technical Information, argued that materials like MoS₂ are uniquely suited to mimicking neurons because of their tunable conductivity, strong electrostatic control, and compatibility with unconventional substrates. The new paper delivers on that argument with a working printed device.
The project drew on infrastructure funded by the National Science Foundation’s Division of Materials Research, which supports Materials Research Science and Engineering Centers (MRSECs) at universities across the country. Northwestern hosts one of those centers, and its shared fabrication and characterization tools, including electron microscopy, Raman spectroscopy, and electrical probing stations, were essential for verifying filament formation, threshold voltages, and device endurance. The NSF’s public access portal allows anyone to trace the award identifiers and linked publications behind the work, connecting long-term federal investment in basic materials science to applied results like this one.
What the paper does not show
The headline claim, that these artificial neurons communicate with living brain cells, deserves careful reading. What the Nature Nanotechnology paper demonstrates is that the printed devices produce electrical spikes with biological-grade timing and complexity. That is a necessary condition for interfacing with real neurons, but it is not the same thing as a proven connection.
No data in the paper confirms that the team placed these devices against cultured brain tissue or living neural circuits and recorded signals passing in both directions. The gap matters. Biological neurons respond to a mix of electrical and chemical cues, and any interface would need to survive in a warm, salty, chemically reactive environment. The flexible substrate is a promising start, but biocompatibility, long-term stability in bodily fluids, and faithful signal translation between silicon-free electronics and wet tissue remain open challenges.
Scaling is another question mark. A handful of printed neurons with tunable spiking is not the same as a large-scale neuromorphic network capable of useful computation. Device-to-device variability, power draw, and electrical cross-talk in dense arrays are engineering problems that grow harder as the number of elements increases. The aerosol-jet process is attractive for patterning, but maintaining uniform memristive behavior across thousands or millions of printed elements has not yet been shown.
Where this fits in the neuromorphic landscape
Northwestern’s printed neurons enter a field that already includes several high-profile hardware approaches to brain-inspired computing. Intel’s Loihi 2 chip and IBM’s earlier TrueNorth processor both mimic neural spiking in silicon, but they are rigid, fabricated in advanced semiconductor foundries, and designed primarily for AI workloads rather than biological interfacing. Organic electrochemical transistors, pursued by groups in Europe and Asia, offer biocompatibility but have struggled with speed and density.
What sets the Northwestern work apart is the combination of flexibility, printability, and spiking complexity. No other group has reported a room-temperature-printed device on a bendable substrate that can reproduce multiple distinct neural firing patterns across such a wide frequency range. That combination is what makes the leap to brain-machine interfaces at least plausible, even if it remains years away.
For researchers tracking neuromorphic hardware, the practical takeaway as of spring 2026 is that aerosol-jet printing now offers a viable route to flexible, neuron-like circuits from two-dimensional materials, no cleanroom required. Turning that capability into functioning brain interfaces will demand parallel progress in biocompatible encapsulation, signal transduction between living and printed cells, and large-scale circuit design. The present work does not solve those problems, but it removes one of the barriers that made them feel distant: the lack of a printed device that actually fires like a neuron.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
