Wearers of prototype eyeglasses built at the Beijing Institute of Technology (BIT) can detect infrared light at wavelengths beyond 2 μm with a luminance of over 700 cd m−2 — while retaining normal vision through the semi-transparent optics. The device, described in Science Advances (Fu et al., Sci. Adv. 12, eaed0245, 2026), is the first full-colour infrared-to-visible upconverter based on colloidal quantum dots (CQDs), and it extends spectral coverage from the near-infrared into the short-wave infrared, a range previous wearable upconversion lenses could not reach.

The work matters for R&D managers tracking photonics and imaging portfolios for a simple reason: incumbent infrared visualization — image intensifiers, InGaAs sensors, thermal cameras — renders infrared scenes as monochrome intensity maps. The BIT device preserves spectral information, converting different infrared wavelengths into distinct visible colours. That distinction opens application space in multispectral inspection, night vision and, per the group's stated roadmap, retinal prosthetics.

How the stack works

The team, led by Ge Mu and Xin Tang, stacked semiconducting mercury telluride (HgTe) CQDs on top of an OLED containing two emissive layers: one red, one cyan. The physics hinges on quantum confinement. Because the CQDs are small enough to confine electrons in all three dimensions, their carriers occupy discrete, atom-like energy subbands rather than a continuous band. The researchers engineered the OLED's electronic bands to align with those subbands, so photogenerated electrons and holes transfer seamlessly from the quantum dot layer into the diode.

The colour-preserving behaviour comes from hole-trapping barriers built into the emissive layers. When the intensity or wavelength of incident infrared light changes, the excited carriers recombine preferentially in different emissive layers, shifting the visible output between red and cyan. In effect, the device maps infrared spectral content onto visible hue — the property the authors call infrared "colour vision."

Measured results, and their limits

The reported performance figures come from a proof-of-concept integration into lightweight eyeglasses. The measured capability: projection of multispectral infrared light onto the retina without blocking ordinary vision, and detection of wavelengths beyond 2 μm at luminance above 700 cd m−2. The team has not yet published device-level metrics such as external quantum efficiency, response time or long-term stability under the stack, and the retinal-implant application remains a projection, not a demonstrated result.

The biological motivation is well grounded in the source literature: mammalian photoreceptors rely on opsins with covalently linked retinals, which cannot absorb photons at wavelengths beyond roughly 700 nm, so infrared light reaching the retina generates no neural signal. Prior approaches — photoreceptor-binding upconversion nanoparticles injected into mouse eyes, and wearable lenses using nonlinear upconversion projected onto the cornea — worked only within a narrow near-infrared window. The CQD approach widens that window substantially.

A staged research programme, not a one-off

Tang frames the result as the culmination of a deliberate sequence. "This full colour upconverter has been a long-term project in my group," he told Physics World. "Starting from an efficient single-colour upconverter, we then successfully demonstrated colour-tuneable OLEDs and Si-/Ge-/CMOS-integrated upconverters. And with our recent progress on the understanding of interfacial carrier transport between quantum dots and colour-tuneable OLEDs, we have now demonstrated this new full-colour upconverter."

That progression — from single-colour device, through colour-tuneable emitters and CMOS-integrated upconverters, to the current full-colour stack — signals that the group has been building a transferable process platform around quantum-dot/OLED interfaces, not just individual demonstrators.

The implant ambition

The next milestone Tang identifies is binding the upconverter to light-sensitive retinal proteins, creating an implantable retinal bionic photoreceptor. In that architecture, the device would transform incoming infrared photons into visible emissions that stimulate opsins on retinal neurons directly, bypassing damaged photoreceptor cells — potentially restoring visual function spanning both the visible and infrared ranges. This remains a design concept; no in vivo data accompany the current paper.

For groups budgeting infrared imaging R&D, the near-term deliverable is the wearable form factor already demonstrated; the retinal application is a longer-horizon bet contingent on biocompatibility and interface engineering the team has yet to publish on. Tang says the group will continue developing the implantable photoreceptor concept on the strength of its improved understanding of interfacial carrier transport.