Self-sustaining, Green Neuromorphic Interfaces

For bio-inspired neuromorphic interfaces to emulate biological signal processing and self-sustainability, the mismatch between sensing and computing signals must be addressed. Here, the authors report sensor-driven, integrated neuromorphic systems with signal matching at the biological level.

Like Comment
Read the paper

Biosystem is featured with the self-autonomous and intelligent response to environmental stimuli. Information processing behind consumes a good portion of energy in biosystem. In order to reduce the consumption and hence improve energy sustainability (e.g., for survival and adaptation), biosystem employs smart strategies in that a) ultralow electrical signal with the amplitude approaching the thermodynamic limit (e.g., tens of mV of action potential) is used, and b) information storage and processing is integrated at local site (e.g., neuron) for efficient response. These strategies enable a highly efficient functional integration in the sensing-decision-action interface. For example, a unitary signal (action potential) flow from the sensory organelle can readily drive bio-computation, saving both the time delay and energy waste for signal conversion that usually is involved in current electronic systems. Therefore, enabling both a signal match and functional emulation to the biosystem can be the key to developing intelligent microsystems that feature a high level of self-sustained autonomy.  

We have demonstrated an electronic microsystem (Fig. 1, left) that can emulate above functionalities and respond to environmental stimuli without external energy input, yielding feature of self-autonomous intelligence like a living organism. We have ‘borrowed’ from biosystem in both material and function to enable this. Protein nanowires, an exciting conductive biomaterial harvested from microbe Geobacter (Fig. 1, right), are used to construct the key components that are enabled with unique functions.

Figure 1. (Left) Schematic of the self-supporting microsystem made from (right) protein nanowires harvested from microbe Geobacter.

Specifically, protein nanowires are used to make memristor-based artificial neuron that can mimic signal processing in a biological neuron. Most importantly, the protein nanowires have the unique catalytic effect to enable the artificial neuron to work at the biological amplitude (e.g., <100 mV). Circuit constructed from the artificial neuron then can directly process (bio-amplitude) environmental signals without the need of amplification. Meanwhile, we previously have also discovered that the thin-film device (termed ‘Air-gen’) made from protein nanowires can harvest electricity from the ambient humidity. Since humidity is ubiquitous, this provides a means to continuously get electricity from almost all environments.

Now we can piece the two key components together to construct a microsystem, in which the electricity from Air-Gen is used to drive sensor and neuromorphic circuit constructed from protein nanowire memristor. Importantly, environmental energy signal can be usually small and may not be able to drive conventional electronics. But the neuromorphic circuit constructed can be directly powered by the environmental signals for information processing, since it functions at the biological signal level. As a result, the electronic microsystem can get energy from the environment to directly drive sensing and computation without the need of external energy source (e.g., a battery). This leads to a full energy self-sustainability, much like the self-autonomy in a living organism.

It is also worthwhile to note that the protein nanowires are an exciting “green” electronic material that is renewably produced from microbes without producing “e-waste.” This research therefore also heralds the potential of future green electronics made from sustainable biomaterials that are more amenable to interacting with human body and diverse environments. We are certainly looking forward to evolved versions of self-sustained, intelligent microsystems.

Jun Yao

Assistant Professor, University of Massachusetts, Amherst