40 micrometers. On its own the number says nothing. It starts to mean something next to this: the photovoltaic implant already sitting under the retinas of patients in Europe is built from pixels of 100.

On September 9, the National Institute of Genetics in Mishima, Shizuoka Prefecture, announced a retinal implant whose stimulating electrodes are made of carbon rather than metal. The Japanese side of the work came from the institute's retinal physiology group, of assistant professor Akihiro Matsumoto and professor Keisuke Yonehara, working with Aarhus University, the ophthalmology department at Aarhus University Hospital and the Technical University of Denmark. The paper ran in Advanced Healthcare Materials on August 25.

The phrase artificial photoreceptor suggests a swap: one dead cell out, one tiny replacement part in. That is not what this is.

What dies, and what survives

Light passes through the cornea and the lens and lands on the retina, the sheet of neural tissue lining the back of the eye. Photoreceptors there convert it into electrical signals, which are handed off through other retinal neurons and carried up the optic nerve into the brain.

In retinitis pigmentosa, an inherited disease, and in age-related macular degeneration, those photoreceptors degenerate and drop out. Nothing reaches the brain from that patch of retina anymore. That is why glasses are beside the point. A lens can sharpen the image landing on the retina, but there is nothing left there to read it.

What matters for engineers is the part that does not die. Bipolar cells and ganglion cells, the neurons downstream of the photoreceptors, often survive in usable condition. Stimulate them electrically and the brain still receives something. The entire field of retinal prostheses rests on that one fact.

The implant doesn't stand in for a photoreceptor

The group's own term for the technology is a photovoltaic retinal prosthesis. A grid of tiny solar cells is slipped under the retina. Each cell turns light into current, and that current stimulates the surviving neurons sitting directly above it.

So the device is not a stand-in for a single photoreceptor. It sits in the space the photoreceptors have vacated and stimulates the wiring one layer further downstream. Nor does ambient daylight power it. The light that drives these implants is near-infrared, and it comes from glasses with a camera on the front, which turn the scene into a pattern and project it into the eye. The implant is the receiver.

The pixel width sets the ceiling

The photovoltaic implant furthest along in humans is PRIMA, funded by Science Corporation. It is built from pixels 100 micrometers wide. Results from 38 patients with geographic atrophy, an advanced stage of age-related macular degeneration that the trial paper puts at more than 5 million people worldwide, appeared in the New England Journal of Medicine on October 20, 2025. Of the 32 assessed at twelve months, 26 (81%) had a clinically meaningful gain in visual acuity, averaging 0.51 logMAR, or roughly 25.5 letters on the eye chart. 84% said they used it at home to read letters, numbers or words.

Acuity tops out near 20/420, for a structural reason: that is the finest detail a 100-micrometer pixel can resolve. In practical terms it means standing 21 times closer than a healthy eye to make out the same thing. The same paper reports that participants using the zoom in the PRIMA glasses read fonts down to a Snellen equivalent of 20/42, below what the pixels alone allow. But zoom enlarges a scene onto the same coarse grid. The grid gets finer only if the pixels do.

Shrinking them is harder than it sounds. A Stanford group whose senior author, Daniel Palanker, also worked on the PRIMA trial laid the problem out in the Journal of Neural Engineering this year. With flat pixels the electric field stays too tightly constrained, and a layer of degenerated tissue roughly 40 micrometers thick separates the implant from the neurons it is trying to reach. Their answer was to build upward. Electrodes shaped as three-dimensional structures, they calculated, could bring the pitch down to 20 micrometers, which in a human eye would correspond to about 20/80, or five times what the current clinical device delivers.

The 40-micrometer pixels out of Mishima each carry a three-dimensional carbon pillar. That is the same bottleneck, approached from the materials side.

Why carbon, and why in three dimensions

Electrodes in earlier implants have generally been metal. Metal is awkward to machine into the fine three-dimensional shapes that the physics now asks for.

Pyrolytic carbon, made by baking a polymer structure at high temperature until it chars into pure carbon, is biocompatible, chemically stable and can be formed into pillars. Nobody had managed to build it directly onto individual microscopic solar cells, and the obvious worry was heat: pyrolysis needs temperatures semiconductors tend not to enjoy. The cells came through it. Under near-infrared light, single pixels still produced an open-circuit voltage of up to about 0.5 volts. The team made them in two sizes, 40 and 200 micrometers.

On excised mouse retina, stimulation from the implant drove a clear rise in spiking, with spike amplitudes comparable to the tissue's natural response to white light. Carbon electrodes produced slightly larger evoked spikes than gold ones. When the team blocked sodium channels with tetrodotoxin the spikes vanished, which confirms the recordings were neurons firing rather than electrical artifact. Pig retina, closer to a human eye in size and structure, responded to the three-dimensional carbon version as well.

Between an excised retina and a living eye

Excised means tissue in a dish. The implant was not placed in a living animal and left to work, nobody has been implanted with it, and the next steps the group lists are laboratory ones: safety and long-term stability, the shape and spacing of the three-dimensional electrodes, and solar cells that are smaller and packed more densely.

Safety and long-term stability are not a formality. A retinal implant of the previous generation, the Argus II from Second Sight Medical Products, reached more than 350 people using an array of 60 electrodes. The company wound down production in 2019 and stopped technical support in 2020. IEEE Spectrum later found patients whose devices were failing with nobody left to repair them. The engineering worked. The company did not. The carbon work sits at a different stage and on different money: Danish private foundations, the European Research Council and Japanese public research grants.

In a field where the width of a pixel sets the ceiling, the carbon route is at least not closed. Anything clinical comes after that, not alongside it.

In your country, is blindness talked about as something to live with, or as something medicine is expected to undo one day?

References