Artificial neurons successfully communicate with living brain cells

Northwestern University engineers have achieved a groundbreaking feat in bioelectronics by developing printed artificial neurons that not only mimic the electrical signals of their biological counterparts but can also directly interact with and activate real brain cells. This significant advancement, detailed in a forthcoming publication in the journal Nature Nanotechnology on April 15, marks a pivotal step towards creating sophisticated brain-machine interfaces and radically more energy-efficient artificial intelligence.

The core innovation lies in the creation of flexible, low-cost devices that produce electrical signals remarkably similar to those generated by living neurons. In rigorous experimental settings involving slices of mouse brain tissue, these synthetic neurons successfully elicited responses from biological neurons, demonstrating an unprecedented level of seamless integration between electronic components and living neural systems. This capability moves beyond mere imitation, suggesting a future where artificial systems can actively participate in and influence neural processes.

A New Era for Brain Interfaces and Neuroprosthetics

The implications of this breakthrough are far-reaching, particularly for the development of advanced brain-machine interfaces (BMIs) and neuroprosthetics. Such technologies could revolutionize the treatment of neurological disorders and injuries, offering renewed hope for individuals experiencing sensory or motor impairments. Imagine implants that could restore lost hearing by directly stimulating auditory nerves with precisely timed artificial signals, or prosthetic limbs that respond to neural commands with natural fluidity, driven by an artificial neuron network that understands and translates intent. Similarly, visual prosthetics could be enhanced to provide more nuanced and comprehensive visual information by interacting more directly with the visual cortex.

The research team, led by Mark C. Hersam, a distinguished professor at Northwestern University, envisions these artificial neurons as building blocks for a new generation of computing systems. The human brain, renowned for its extraordinary energy efficiency, operates on principles that modern digital computers struggle to replicate. While conventional silicon-based computing relies on vast networks of identical transistors, the brain’s complexity arises from a heterogeneous, dynamic, and three-dimensional arrangement of diverse neuron types with constantly adapting connections. This inherent difference makes the brain a compelling model for future computing architectures.

The Energy Imperative: Brain-Inspired Computing for AI

The escalating demand for artificial intelligence (AI) has brought into sharp focus a critical bottleneck: power consumption. Current AI models, particularly those leveraging deep learning, require immense datasets for training, a process that translates into colossal energy expenditures. Hersam articulated this challenge, stating, "The way you make AI smarter is by training it on more and more data. This data-intensive training leads to a massive power-consumption problem. Therefore, we have to come up with more efficient hardware to handle big data and AI."

The stark contrast in energy efficiency between the brain and digital computers underscores the need for a paradigm shift. The brain is estimated to be five orders of magnitude more energy-efficient than a digital computer, a remarkable disparity that fuels the pursuit of brain-inspired computing. "Because the brain is five orders of magnitude more energy efficient than a digital computer, it makes sense to look to the brain for inspiration for next-generation computing," Hersam explained.

This pursuit is driven by the current limitations of traditional silicon-based computing. Modern computers achieve their processing power by densely packing billions of identical transistors onto rigid, two-dimensional chips. Each component functions in a uniform manner, and once manufactured, the system’s architecture is largely fixed. This contrasts sharply with the brain’s biological architecture, which is characterized by a vast array of specialized neurons interconnected in a flexible, three-dimensional network that continuously rewires itself through learning and adaptation.

"Silicon achieves complexity by having billions of identical devices," Hersam elaborated. "Everything is the same, rigid and fixed once it’s fabricated. The brain is the opposite. It’s heterogeneous, dynamic and three-dimensional. To move in that direction, we need new materials and new ways to build electronics."

Previous attempts at creating artificial neurons often resulted in devices that produced overly simplistic electrical signals. To achieve more complex behaviors akin to biological neurons, these systems typically required large and energy-intensive networks of individual components.

Printable Materials: Mimicking Neural Complexity

The Northwestern team’s innovative approach addresses this limitation by employing soft, printable materials that more closely emulate the brain’s structural and functional characteristics. Their breakthrough hinges on the development of electronic inks formulated from nanoscale flakes of molybdenum disulfide (MoS₂), a semiconductor material, and graphene, an excellent electrical conductor. These inks were precisely deposited onto flexible polymer substrates using aerosol jet printing, a cost-effective and scalable manufacturing technique.

A key insight from the research was the novel utilization of the polymer binder within the inks. Traditionally, researchers viewed this polymer as an impediment to electrical performance and sought to remove it entirely after printing. However, the Northwestern team discovered a way to leverage this polymer to enhance the device’s functionality.

"Instead of fully removing the polymer, we partially decompose it," Hersam explained. "Then, when we pass current through the device, we drive further decomposition of the polymer. This decomposition occurs in a spatially inhomogeneous manner, leading to formation of a conductive filament, such that all the current is constricted into a narrow region in space."

This controlled decomposition process results in the formation of a narrow conductive path that generates a sudden electrical response, mirroring the characteristic "firing" of a biological neuron. Crucially, the resulting artificial neurons are capable of producing a rich spectrum of electrical signals, including single spikes, sustained firing patterns, and complex bursting behaviors. This sophisticated signaling repertoire closely approximates the diverse communication methods employed by real neurons. The ability of each artificial neuron to generate more complex signals means that fewer components are required to perform advanced computational tasks, leading to substantial improvements in overall computing efficiency.

Direct Validation: Testing on Real Brain Tissue

To rigorously assess the real-world applicability of their artificial neurons, the researchers collaborated with Indira M. Raman, a professor of neurobiology at Northwestern University. Her team conducted experiments applying the artificial neuron-generated signals to slices of mouse cerebellum, a brain region critical for motor control and coordination.

The results were highly encouraging. The electrical spikes produced by the artificial neurons exhibited key biological characteristics, including precise timing and duration, that were indistinguishable from natural neural activity. These signals effectively activated the living neurons in the brain slices, triggering neural circuits in a manner consistent with natural brain function.

"Other labs have tried to make artificial neurons with organic materials, and they spiked too slowly," Hersam noted. "Or they used metal oxides, which are too fast. We are within a temporal range that was not previously demonstrated for artificial neurons. You can see the living neurons respond to our artificial neuron. So, we’ve demonstrated signals that are not only the right timescale but also the right spike shape to interact directly with living neurons." This temporal precision and spike fidelity are critical for establishing a robust and meaningful interface with biological neural systems.

Sustainable Manufacturing and the Future of AI

Beyond their impressive performance, the new artificial neurons offer significant advantages in terms of manufacturing sustainability and cost-effectiveness. The printing process is inherently simple and inexpensive, and the additive manufacturing approach ensures that material is deposited only where it is needed, thereby minimizing waste. This is particularly important in the context of burgeoning AI development, where the environmental footprint of technology is becoming an increasingly critical concern.

The immense power demands of modern AI are already straining global energy resources. Large-scale data centers, the backbone of AI computation, consume vast amounts of electricity and require substantial water resources for cooling. Hersam highlighted the alarming scale of this issue: "To meet the energy demands of AI, tech companies are building gigawatt data centers powered by dedicated nuclear power plants. It is evident that this massive power consumption will limit further scaling of computing since it’s hard to imagine a next-generation data center requiring 100 nuclear power plants. The other issue is that when you’re dissipating gigawatts of power, there’s a lot of heat. Because data centers are cooled with water, AI is putting severe stress on the water supply. However you look at it, we need to come up with more energy-efficient hardware for AI."

The research, titled "Multi-order complexity spiking neurons enabled by printed MoS₂ memristive nanosheet networks," was supported by the National Science Foundation, underscoring the significant investment and interest in advancing this critical area of scientific inquiry. This development represents a significant stride towards realizing computing systems that are not only more powerful but also fundamentally more sustainable and in harmony with biological principles. The journey from imitation to direct interaction with living neural tissue signifies a profound leap forward, opening up a future where the lines between biological and artificial intelligence blur in ways that promise to redefine technology and human augmentation.