Perovskite LEDs have spent a decade getting brighter and more efficient while failing the same test. Efficiencies now rival commercial displays — red devices above 30%, green close behind — but the devices kept dying young at the brightness a display actually runs at. A paper published November 6 in Nature Communications changes the terms of that failure. A team led by Yucai Yuan, Baodan Zhao, and Dawei Di at Zhejiang University, with collaborators at the University of Science and Technology of China, reports green perovskite LEDs with a measured half-life — T50, the time for luminance to fall to half its starting value — of 1,350 hours at an initial luminance of 1,000 cd/m², the brightness unit the display industry calls a nit. The devices pair that lifetime with a maximum energy-conversion efficiency of 28.0% and a peak external quantum efficiency of 24.5%.

The number that matters most is the first one, and the word that matters most is “measured.” The paper’s abstract draws a line the field has been blurring for years: the half-lives actually measured for visible perovskite LEDs had in most cases stayed below 1,000 hours at any brightness, “in contrast to the much longer extrapolated lifetimes reported.” This result is the opposite kind of number: 1,350 hours of a device actually running at 1,000 nits until it faded to half brightness — the first measured crossing of the threshold, achieved not with a new emitter chemistry but with a change in how the device is driven.

The threshold: why 1,000 hours at 1,000 nits

The two numbers are not arbitrary. A thousand nits is roughly the brightness of a phone or television display running outdoors or in a bright room — the brightness at which a display is stressed, not the brightness at which it idles. A thousand hours is not a commercial lifetime — the minimum lifetime required for an OLED display is generally put at 10,000 hours — but it is the order of magnitude at which a laboratory result stops being a curiosity and starts being a durability claim a display engineer can take seriously.

The extrapolation problem is what makes the crossing meaningful rather than merely large. A 2024 green device claimed an “equivalent” T50 above 326,000 hours at 100 nits, and a 2022 near-infrared device from the same Zhejiang group an estimated 32,675 hours — both model projections, not hours on a clock. The 1,350-hour figure is smaller than those projections and worth more than them, because it was watched happening.

Against the prior display-brightness record — a Nano Letters buried-interface paper reporting a T90 of 726 minutes, about 12 hours, at 1,000 nits with a 29.2% EQE — the new device runs more than a hundred times longer at the same brightness before losing half its light. T90 and T50 are different milestones on the decay curve, so the comparison is directional rather than exact — but the gap is large enough that the direction is unambiguous.

The mechanism: a fluorinated monolayer, and a voltage below the bandgap

The structural change is a single molecular layer, and the mechanism it enables is the paper’s real contribution. Between the metal-oxide layer and the polymeric hole-transport layer, the team inserts a fluorinated self-assembled monolayer — a one-molecule-thick film that chemically anchors to the oxide surface. The abstract reports “Ohmic-like characteristics for charge injection”: holes enter with little resistance, without the voltage drop a mismatched interface normally imposes.

That low-resistance injection is what lowers the operating voltage, and the voltage is where the lifetime comes from. At roughly 1,000 nits, the devices run at about 2.1 volts — below the bandgap voltage. A green photon carries roughly 2.3 electron-volts, so a device emitting green light while driven at 2.1 V is producing photons more energetic than the electrical energy supplied per electron, with the difference made up thermally. Sub-bandgap operation is physically permitted but unusual, and it is the hinge of the paper’s argument: at that voltage, the abstract states, ion migration is suppressed.

Ion migration is the degradation engine in halide perovskites. Under an electric field, halide ions drift through the crystal, creating nonradiative defects and shifting the emitter’s composition — which is why perovskite LEDs have historically aged fastest exactly when driven hardest. The conventional response has been to fight migration with chemistry: passivate defects, pin ions, stiffen the lattice. The Zhejiang team’s response is to remove the field instead.

The paper’s evidence for that chain, as stated in the abstract, is a striking correlation: the accumulated photon output of the devices over their T50 lifespan “is found to sharply increase as the driving voltage drops to sub-bandgap values.” Total light emitted over a lifetime is the right metric for this claim, because it folds efficiency and durability into one number. The lifetime gain tracks the voltage, not the emitter, which is what makes this a device-architecture result rather than a materials result.

That framing matters for how far the result travels. A new emitter chemistry has to be re-proven for every color and every stack. A voltage-reduction strategy built on an interfacial monolayer is, in principle, a design rule: fix the injection, drop the voltage, starve the migration. Whether it generalizes is the open question — the abstract does not say whether the same trick has been demonstrated in red or blue devices.

The devices: 28.0% efficient, 24.5% EQE, and a flexible display

The efficiency figures travel with the lifetime rather than ahead of it, which is itself a change of posture for this field. A maximum energy-conversion efficiency of 28.0% and a peak EQE of 24.5% are strong numbers for a green perovskite LED, but they are not records — pure-red devices published days earlier reached 31.3% EQE, and green devices have been reported at 27.1% EQE and above. The point is that efficiency and lifetime arrived together, at display brightness, in the same device. A 24.5% EQE that survives 1,350 hours at 1,000 nits is a different kind of achievement from a 30% EQE measured for minutes at low brightness.

The team also built a flexible active-matrix perovskite LED display prototype, presented as evidence of the devices’ potential for wearable optoelectronics. An active-matrix display — pixels switched by a transistor backplane — is the architecture a real display uses, and a flexible one is the form factor a wearable needs. But the abstract gives no specifications: no size, resolution, pixel count, brightness, or lifetime measured on the display itself. Read it as a demonstration that the stack can be built into a display, not as a display result.

The collaboration spans Zhejiang University and the University of Science and Technology of China (Bo Li and Fengjia Fan), plus Linrun Feng of LinkZill Technology in Hangzhou. The paper was received July 27, 2026, accepted October 16, and published November 6 — a fast transit: the pure-red 31.3% result appeared in the same journal five days earlier, on November 1.

The lane: efficiency was solved; stability was the asterisk

In the desk’s perovskite LED lane, the recent story has been rising efficiencies with stability always the asterisk. The pure-red result of November 1 — 31.3% EQE from a lattice reinforced with rigid-ring ammonium and an electrostatic polymer, with a 343.3-hour lifetime — pushed efficiency past 30% while improving durability, but its lifetime is still measured in hundreds of hours, and the abstract does not state the brightness at which it was measured. Green, the color where the display industry needs both brightness and longevity, was the color where the measured lifetime had stayed under 1,000 hours. The new result closes that specific gap, at the specific brightness, by measurement.

The comparison with the red result clarifies what each paper actually did. The red paper attacked the emitter: stiffen the lattice, suppress field-induced distortion and ion migration inside the perovskite itself, and the device lasts longer because its material is tougher. The green paper attacked the drive: leave the emitter alone, fix the injection interface, run the device at a voltage where migration is starved. Both suppress ion migration; one does it by hardening the material, the other by removing the force. The two approaches are complementary, and the field’s next step is likely to combine them.

There is also a quieter contrast with the blue problem. The most recent review of the field notes that red and green perovskite LEDs have both passed 30% EQE, while blue has not — the best blue devices sit in the mid-20s, with lifetimes measured in minutes to hours at display brightness. A full-color perovskite display needs all three, and green crossing 1,000 hours does not move blue. But it does change the question blue has to answer: the stability bar is now a measured number at 1,000 nits, set by a green device, and blue’s deficit is measured against a real device rather than an extrapolated one.

What this is not: a product, a red device, or a blue device

Several limits are limits of the material consulted rather than of the work itself. The full text was not directly readable through the browser fetch used for this article, which returned the abstract, author list, affiliations, and publication metadata only — so the figures here are the abstract’s, and several questions remain open, stated as open rather than guessed at.

First, the device area is not stated in the abstract, so this article states none. A lifetime measured on a few-square-millimetre laboratory pixel does not automatically transfer to a large-area device, where current crowding, heating, and non-uniformity all worsen with size.

Second, the encapsulation is not stated. Whether the 1,350-hour devices were sealed under glass, coated with a barrier film, or tested bare determines how much of the lifetime belongs to the device and how much to the packaging — and an unencapsulated lifetime would be a stronger claim than an encapsulated one. The abstract does not say, so neither does this article.

Third, the identity of the fluorinated monolayer is not stated. The molecule’s name, anchoring chemistry, and deposition method sit in the full text, and transferability depends on them: a commercially available phosphonic-acid SAM would be adoptable by any group tomorrow, while a bespoke synthesis would be a longer road. Until the full text circulates, the monolayer is described here as the abstract describes it, and no further.

Fourth, green only. The sub-bandgap mechanism should, in principle, apply wherever injection can be made Ohmic-like, but blue devices face a harder version: a wider bandgap means a higher operating voltage, and chloride-containing blue perovskites suffer halide segregation that green bromide devices do not. The paper demonstrates the mechanism in green; whether it generalizes is a claim for future papers, not this one.

Fifth, a single laboratory, a single stack. The field has learned to wait for independent replication before treating a lifetime figure as a benchmark — particularly for a figure that depends on a measurement condition, constant-current driving at 1,000 nits, that other groups will implement slightly differently. The 1,350 hours is a reported measurement, not yet a reproduced one.

What to watch

Three things, in order of how quickly they could change the assessment.

First, the full text. When it circulates, the device area, the encapsulation, the monolayer’s identity, and the control-device numbers — what the same stack achieves without the fluorinated layer — will be the first figures this desk checks. The control comparison sizes the monolayer’s contribution: if the control dies in tens of hours, the interlayer is doing the work the abstract claims.

Second, red and blue. A demonstration of sub-bandgap operation in a red device — where the narrower bandgap makes the voltage target easier — would show the mechanism is a design rule rather than a green-specific recipe. A blue demonstration would be the harder test: blue blocks full-color perovskite displays, and its wide bandgap and halide segregation are exactly the problems a low-voltage strategy might or might not solve.

Third, replication and scale. An independent group reproducing 1,000-plus hours at 1,000 nits on a different stack, or the same group demonstrating the lifetime on a larger-area device, would convert a laboratory milestone into a field benchmark. Until then, the figure stands as the best measured visible-PeLED lifetime at display brightness — and as a challenge to every other group to measure rather than extrapolate.

The bottom line

A Zhejiang University team, with USTC collaborators, has reported green perovskite LEDs with a measured T50 of 1,350 hours at 1,000 nits, a maximum energy-conversion efficiency of 28.0%, and a peak EQE of 24.5% — the first measured crossing of the 1,000-hour, 1,000-nit threshold for a visible perovskite LED. The enabler is a fluorinated self-assembled monolayer between the metal-oxide and polymeric hole-transport layers, which gives Ohmic-like charge injection and drops the operating voltage at display brightness to about 2.1 V, below the bandgap, where ion migration is suppressed. Total photon output over the device lifetime rises sharply as voltage falls below the bandgap — the evidence that the lifetime comes from the drive rather than the emitter. A flexible active-matrix display prototype shows the stack can be built into a display.

It is green only, on a laboratory-scale device whose area and encapsulation the abstract does not state, with a monolayer whose identity awaits the full text. Frame it as the first measured crossing of the threshold the display industry actually uses — achieved by lowering the voltage rather than hardening the emitter — and let the full text, the red and blue tests, and independent replication decide whether sub-bandgap operation becomes the field’s design rule for stability.

No investment recommendation is made or implied anywhere in this article.

Sources and further reading

  • Yuan, Y., Xing, S., Zhang, G., Li, B., Zhang, S., Liu, S., Pan, Y., Gu, Z., Zhou, K., Tang, W., Gao, Y., Zhang, S., Ren, Z., Zhang, S.-Y., Fan, F., Zou, C., Feng, L., Zhao, B., & Di, D. “Addressing the 1000-h/1000-nit stability challenge in perovskite LEDs.” Nature Communications (published November 6, 2026; received July 27, 2026; accepted October 16, 2026). DOI: 10.1038/s41467-026-79298-8 — the primary source for every figure quoted: T50 of 1,350 h at an initial luminance of 1,000 cd/m², maximum energy-conversion efficiency of 28.0%, peak EQE of 24.5%, the fluorinated self-assembled monolayer between the metal-oxide and polymeric hole-transport layers, Ohmic-like charge injection, operating voltage of ~2.1 V at ~1,000 nits, suppression of ion migration at sub-bandgap voltage, the sharp rise in accumulated photon output as voltage drops below the bandgap, the measured-versus-extrapolated lifetime distinction, and the flexible active-matrix PeLED display prototype. Affiliations: Zhejiang University (State Key Laboratory of Extreme Photonics and Instrumentation, College of Optical Science and Engineering, ZJU-UIUC Institute; ZJU-Hangzhou Global Scientific and Technological Innovation Center; Zhejiang Key Laboratory of Optoelectronic Information Technology), University of Science and Technology of China, and LinkZill Technology, Hangzhou. Correspondence to Baodan Zhao and Dawei Di.
  • Zhou, X., Ren, Z., Wang, J., et al. “Robust perovskite lattice for durable and efficient pure-red light-emitting diodes.” Nature Communications (published November 1, 2026). DOI: 10.1038/s41467-026-79295-x — the paired red result cited for the lane context: 31.3% EQE and a 343.3-hour lifetime from a lattice reinforced with rigid-ring ammonium and an electrostatic polymer. The brightness at which the red lifetime was measured is not stated in the abstract consulted, so none is stated here.
  • Wu, X., et al. “Stabilizing the Buried Interface Phase for Perovskite LEDs with a Remarkable T90 Lifetime at 1000 nit.” Nano Letters 25, 18158–18166 (2025). DOI: 10.1021/acs.nanolett.5c05743 — the prior display-brightness stability comparator: 29.2% EQE with a T90 of 726 minutes at 1,000 cd/m². Cited with the caveat that T90 and T50 are different milestones on the decay curve.
  • Ding, S., et al. “Phase stabilization via A-site ion anchoring for ultra-stable perovskite light emitting diodes.” Materials Horizons 11, 5265–5273 (2024). DOI: 10.1039/d4mh00701h — the extrapolation example: green PeLEDs with 27.1% EQE and an “equivalent” T50 of over 326,252 hours at 100 cd/m², reported as an equivalent rather than a measured lifetime.
  • Guo, B., et al. “Ultrastable near-infrared perovskite light-emitting diodes.” Nature Photonics (2022). DOI: 10.1038/s41566-022-01046-3 — the earlier Zhejiang near-infrared result from the same group, with an estimated T50 of 32,675 hours at a radiance equivalent to a green OLED at 1,000 cd/m², and the source for the context that typical perovskite LED lifespans were “on the order of 10–100 hours” against a 10,000-hour minimum for an OLED display (via the Zhejiang University news report of August 2022).
  • Frontiers of Engineering / Higher Education Press, “Focus on performance of perovskite light-emitting diodes (2026)” — the field review cited for the state of play across colors: red and green PeLEDs above 30% EQE, blue yet to surpass 30%, with blue lifetimes at display brightness measured in minutes to hours.

A note on sourcing: the full body text and supplementary files of the Yuan paper were not directly readable through the browser fetch used for this article, which returned the abstract, author list, affiliations, funding, and publication metadata only. No device area, encapsulation method, monolayer identity, emission wavelength, control-device lifetime, or display-prototype specification is stated here, because none appears in the material consulted. No market-size figure is quoted, because none is needed for this device result.