Northwestern Engineers Print Artificial Neurons That Communicate With Living Brain Cells

Engineers at Northwestern University have successfully developed 3D-printed artificial neurons capable of communicating directly with living biological brain cells. In a milestone for both bioelectronics and computing architecture, the research team created soft, flexible devices that generate electrical signals realistic enough to trigger measurable responses in real neural tissue. The breakthrough, published in the journal Nature Nanotechnology, demonstrates a new level of biocompatibility between synthetic hardware and biological systems. By bridging the gap between rigid electronics and the soft, dynamic environment of the nervous system, the technology opens new pathways for advanced medical implants and highly energy-efficient computing systems.

The development arrives at a critical moment for the technology industry, which is currently grappling with the massive power consumption required to run modern artificial intelligence. Today's computing infrastructure relies on billions of nearly identical transistors packed tightly onto rigid silicon chips. While these systems are capable of performing monumental calculations, they require vast amounts of electricity and generate immense heat, leading to unsustainable energy demands for AI data centers. Mark C. Hersam, the Northwestern professor who co-led the study, noted that because the human brain is roughly five orders of magnitude—or 100,000 times—more energy-efficient than a digital computer, it provides the ultimate blueprint for next-generation hardware.

Unlike the rigid, uniform, and fixed nature of silicon chips, the human brain operates through diverse, specialized neurons arranged in three-dimensional, constantly adapting networks. Silicon achieves its complexity through sheer volume, utilizing billions of identical devices that remain static once manufactured. The brain, conversely, is heterogeneous and dynamic. To replicate this biological efficiency, researchers have long attempted to build "neuromorphic" chips that mimic neural spiking behavior. However, previous attempts using organic materials often fired too slowly, while those utilizing metal oxides fired too quickly to interact naturally with living tissue.

To overcome these limitations, the Northwestern team turned to soft, printable nanomaterials. They engineered specialized electronic inks formulated from nanoscale flakes of two distinct materials: molybdenum disulfide, which acts as a semiconductor, and graphene, which serves as a highly efficient electrical conductor. Using a high-precision manufacturing technique known as aerosol jet printing, the researchers deposited these inks onto flexible polymer substrates. This additive manufacturing approach not only reduces waste by placing material exactly where it is needed, but it also results in soft electronic devices that can bend and flex, making them far more compatible with the physical properties of biological tissue than traditional rigid silicon.[2]

The breakthrough in the neurons' signaling capability actually stemmed from an unconventional manufacturing decision. Typically, when researchers use electronic inks, they attempt to completely burn off or remove the stabilizing polymers after printing, as these materials can obstruct the flow of electrical current. However, the Northwestern team chose a different approach, deliberately leaving the stabilizing polymer partially intact. By passing a current through the device to drive further decomposition, they created specific structural imperfections. Rather than ruining the device, these deliberate flaws allowed the artificial neurons to produce far more advanced and varied signaling behaviors.

Because of this unique structural composition, the printed devices can generate a wide variety of firing patterns that closely mirror the complexity of a real nervous system. Instead of merely producing simple, uniform electrical pulses, the artificial neurons can emit single spikes, continuous firing sequences, and rhythmic bursts of activity. This signaling diversity is crucial because real brain cells do not all behave in the exact same way. By capturing this complexity, each artificial neuron can encode significantly more information, potentially reducing the total number of components required in a computing system and drastically improving overall efficiency. The devices also proved highly durable, remaining stable for more than one million firing cycles.[1]

To verify that these synthetic signals could truly interface with biological systems, Hersam's engineering team collaborated with Indira M. Raman, a professor of neurobiology at Northwestern. The researchers connected the artificial neurons to slices of mouse cerebellum—a region of the brain that helps coordinate movement and contains neurons with well-documented electrical behaviors. When the team applied the artificial voltage spikes to the biological tissue, they found that the synthetic signals perfectly matched the timing, duration, and shape of natural neuron activity.[1]

The results of the biological testing were unequivocal: the living neural tissue responded to the artificial spikes exactly as if they were originating from a biological peer. The synthetic signals successfully activated Purkinje neurons, a major type of cerebellar brain cell, triggering real neural circuits in a manner indistinguishable from natural brain function. The researchers noted that they were operating within a temporal range that had never been previously demonstrated for artificial neurons, proving that they had achieved not just the correct speed, but the precise spike shape required to communicate directly with living cells.

The medical implications of this biocompatibility are profound. For decades, researchers have worked to develop brain-machine interfaces and neuroprosthetics to help patients with severe neurological damage, spinal cord injuries, or sensory loss. However, traditional rigid implants often struggle to communicate naturally with the nervous system and can cause tissue scarring over time. By utilizing flexible, printed electronics that speak the exact electrical language of the brain, future medical devices could integrate seamlessly into the body. This could lead to vastly improved implants for restoring hearing, vision, or motor control, providing patients with smoother, more natural sensory feedback and movement.[1]

Beyond medicine, the breakthrough offers a tangible path forward for the future of artificial intelligence hardware. If engineers can scale up the production of these printed, heterogeneous neurons, they could build computing architectures that process massive datasets without the crippling energy requirements of today's server farms. By moving away from rigid silicon and embracing the dynamic, low-power principles of biological brains, the technology industry could eventually develop AI systems that are not only vastly more efficient but also capable of more nuanced, brain-like reasoning.[2]