A Wireless Brain Implant Smaller Than a Grain of Salt Just Recorded Neurons for an Entire Year

Brain implants have a size problem. Current devices for recording neural activity are bulky enough to damage the tissue they’re trying to monitor, and the wires that connect them to external electronics tug against the brain with every movement, causing inflammation and signal degradation over weeks.

A new device called MOTE (microscale optoelectronic tetherless electrode) shatters these limitations. At less than one nanoliter in volume (smaller than a grain of salt), powered entirely by light, and communicating wirelessly via its own built-in LED, MOTE recorded neural activity in awake mice for 365 consecutive days without degradation.

Key Findings

1. MOTE’s total volume is less than 1 nanoliter (less than 0.001 mm³)—orders of magnitude smaller than any previous wireless neural recording device.
2. The device recorded neural signals (local field potentials and action potentials) continuously for 365 days in awake, freely moving mice.
3. Power delivery and data transmission both use light: a 623 nm external LED powers the device via a photovoltaic cell, while the device transmits data by emitting 825 nm pulses through the same diode.
4. The device uses pulse-position modulation (PPM) encoding, making it far more resistant to noise than amplitude-based systems.
5. Input-referred noise is 14.8 μV RMS with bandwidth from <10 Hz to >10 kHz—sufficient to capture both local field potentials and individual action potentials.
6. Total power consumption is just 1 μW, with 50% going to the low-noise amplifier.
7. Encapsulation uses atomic layer deposition of SiO₂, Si₃N₄, and Al₂O₃ (total thickness <1.5 μm), providing chronic biocompatibility in corrosive biological fluids.

Source: Nature Electronics (2025) | Lee, Ghajari et al.

How MOTE Works: A Single Diode Does Double Duty

MOTE’s core innovation is elegant in its simplicity. A single AlGaAs (aluminum gallium arsenide) diode serves as both the power receiver and the data transmitter. For 93.4% of the time, the diode operates as a photovoltaic cell, absorbing 623 nm red light from an external LED to generate electrical power. For 0.06% of the time, the same diode switches to LED mode, emitting 825 nm infrared pulses that encode the neural signal being recorded. The remaining time handles transitions between modes.

This time-division multiplexing eliminates the need for separate power and communication hardware—a major reason why other wireless implants are orders of magnitude larger. Radio-frequency devices need antennas sized to their operating wavelength. Ultrasound devices need piezoelectric transducers. MOTE needs only a single compound semiconductor diode smaller than the width of a human hair.

The CMOS brain: amplification and encoding on a speck

Beneath the diode sits a custom CMOS (complementary metal-oxide-semiconductor) circuit that does three things:

1. Low-noise amplification: Neural signals at the electrode surface are tiny—on the order of microvolts. The amplifier boosts these signals with a noise floor of 14.8 μV RMS across a bandwidth of <10 Hz to >10 kHz, sufficient to detect both slow local field potentials and fast action potentials. This consumes 500 nanowatts—half the device’s total power budget.
2. Pulse-position modulation encoding: Rather than encoding neural signals as voltage amplitudes (which are easily corrupted by noise), the circuit converts voltage values into the timing of optical pulses. The interval between a reference pulse and an encoding pulse corresponds directly to the measured voltage. This scheme is inherently more noise-resistant, borrowing a principle from satellite communications where high channel bandwidth compensates for low signal-to-noise ratios.
3. LED driving: A voltage step-up circuit stacks three capacitors in series to create pulses bright enough to drive the LED efficiently. Without this step-up, LED efficiency drops by orders of magnitude at the low voltages available from a nanoscale photovoltaic cell.

Why Size Matters: The Tissue Damage Problem

The brain is mechanically delicate. Current neural implants—even “miniaturized” ones—displace volumes on the scale of microliters (thousandths of a milliliter). In a mouse brain weighing about 0.4 grams, that’s a significant fraction of total volume. The implant compresses surrounding tissue, triggers immune responses, and the resulting gliosis (scarring) gradually degrades recording quality over weeks to months.

Tethered devices add another problem: wires connecting the implant to external electronics move relative to the brain tissue with every heartbeat, every breath, every head movement. This mechanical mismatch drives chronic inflammation at the tissue-electrode interface, progressively isolating the recording electrodes from the neurons they’re trying to monitor.

MOTE sidesteps both problems. At less than 1 nanoliter, it displaces essentially no tissue. With no wires, there’s no chronic mechanical stress. The result: stable neural recordings for a full year in awake, behaving mice—with no observable signal degradation.

Chronic Recording: 365 Days and Counting

The researchers implanted MOTEs into the brains of awake mice and recorded neural activity across multiple time points spanning an entire year. The device captured both local field potentials (the aggregate electrical activity of neuronal populations) and individual action potentials (spikes from single neurons or small clusters). Signal quality remained consistent from day 4 through day 365, demonstrating that the encapsulation—less than 1.5 μm of atomic layer-deposited oxides and nitrides—successfully protected the electronics from the corrosive salt-water environment of living tissue for at least a year.

Before in vivo testing, the researchers validated MOTE’s recording fidelity in vitro by culturing cardiomyocytes (heart muscle cells derived from induced pluripotent stem cells) directly on the device. The MOTE successfully recorded cardiac electrical spikes, with frequency matching what was expected for beating cardiomyocytes—confirming that the PPM encoding faithfully reproduced biological electrical signals.

Study Details

Fabrication

MOTE is manufactured using a parallel fabrication process that simultaneously produces nearly 100 devices per chip (scalable to thousands per square centimeter). The CMOS circuits are fabricated using standard semiconductor processes, then heterogeneously integrated with the AlGaAs photovoltaic LED. Vacuum annealing at 300°C under extremely low pressure (<10⁻⁶ torr) removes transfer residues and promotes adhesion. Encapsulation uses atomic layer deposition of SiO₂, Si₃N₄, and Al₂O₃ (total <1.5 μm). High-pressure platinum sputtering provides both favorable electrode impedance and a light shield to prevent incident light from generating photocurrents in the electronics. Recording electrodes are 28.5 × 30.5 μm and 12.5 × 23 μm, spaced 294 μm apart.

Optical Safety

The external LED illumination is limited to <70 mW/mm², well below the 250 mW/mm² threshold for heat-induced brain damage. Red (623 nm) and infrared (825 nm) wavelengths are used because they penetrate tissue with minimal absorption and scattering, enabling communication even when the device is implanted at depth.

Limitations

MOTE currently provides a single recording channel (two electrodes). Scaling to multi-channel arrays will require addressing challenges in power delivery and data multiplexing. The device requires an external LED source positioned near the head, which constrains some experimental setups. While signal quality was maintained for 365 days, longer-term studies are needed to determine the ultimate lifetime. The PPM encoding scheme has limited temporal resolution compared to high-speed digital transmission; improving encoding density could enable higher-bandwidth neural recordings. Translation to larger brains (primates, humans) will require adapting the optical communication for greater tissue depth.

From Lab Tool to Brain-Computer Interfaces: Where MOTE Goes Next

Year-long recordings unlock previously impossible experiments

Chronic neural recording—tracking the same neurons over months—is the holy grail for understanding learning, memory consolidation, aging, and neurodegeneration. Most current technologies degrade within weeks due to tissue damage. MOTE’s year-long stability opens the door to experiments that were previously impossible: watching how individual neural circuits change as an animal learns a complex task over months, tracking the progression of neurodegenerative disease at the single-neuron level, or monitoring how the brain adapts to chronic drug treatment.

Stable wireless interfaces for paralyzed patients

Current brain-computer interfaces (BCIs) like those used in paralyzed patients rely on large, tethered electrode arrays that degrade over time. A future version of MOTE—scaled to multiple channels and adapted for the human brain—could provide stable, long-term neural interfaces without the chronic inflammation that plagues current devices. The wireless, subnanoliter form factor means such devices could be distributed across multiple brain regions without displacing significant tissue.

“Neural dust” goes from theory to practice

Because MOTE is fabricated using scalable semiconductor processes, future versions could deploy dozens or hundreds of independent recording nodes throughout the brain. Each would operate independently, powered and read by light, creating a distributed neural monitoring network. This “neural dust” concept has been theorized for years—MOTE represents the first device small and durable enough to make it practical.