Four days ago, a perovskite paper landed that is worth reading for its diagnosis before its device. On October 29, a team led by Buyi Yan, Jizhong Yao, and Dongchen Lan — spanning Microquanta Semiconductor and Zhejiang University in Hangzhou, with China Three Gorges Renewables in Beijing and the Australian Centre for Advanced Photovoltaics in Sydney — published in Nature Communications a study that asks a narrower question than most scale-up papers bother to ask: when a perovskite device grows from a laboratory cell to a module, which of its three performance numbers actually does the leaking? Their answer is fill factor. Their fix is disarmingly small: cool the wet film for 15 seconds, to between 0 and 10 °C, before vacuum drying. The headline device — an eight-cell mini-module over a 19.48 cm² aperture — reached a certified steady-state efficiency of 23.9% at 83.3% fill factor, which the paper’s results text places among the highest-performing perovskite modules reported.
That framing deserves a note up front, because the paper’s own abstract is bolder than its results text. The abstract calls the 23.9% “a record certified steady-state efficiency” and the fill factor “the highest reported for perovskite modules.” The results section, as trade coverage of the paper notes, speaks more carefully: among the highest-performing. This article uses the careful wording throughout, for a reason the site’s own recent coverage supplies. A peer-reviewed August paper from Nanjing University and Renshine Solar reports a certified 24.0% at an 810 cm² aperture, and the UNSW–UtmoLight submodule covered here on October 31 reports 23.5% at 676 cm². Both are larger devices at comparable or higher efficiencies. The 23.9% figure is a strong result in the mini-module class, not an outright record across module sizes.
The diagnosis: fill factor is the leak
Perovskite’s scale-up problem is familiar by now. Small laboratory cells routinely exceed 26% — the Northwestern–DGIST inverted cell covered by this desk days ago reached 27.4% on a sweep — and then efficiency falls as the same material is coated over larger areas, cut into series-connected cells, and asked to behave as one device. The usual account of that fall is a list: coating non-uniformity, pinholes, shunt paths, interconnect losses, more interfaces at which things can go wrong. The Yan team’s contribution is to turn the list into an accounting.
Working across diverse published perovskite technologies, the authors quantify how much of the efficiency lost during scale-up belongs to each device parameter: open-circuit voltage, short-circuit current, and fill factor. Some current loss is unavoidable and should be set aside first — the laser scribes that interconnect cells in series consume area that generates no power. Once that dead-area loss is accounted for, fill factor emerges as the major contributor to the remaining loss, and it falls steadily as devices grow. The paper further resolves the fill-factor deficit into voltage-, diode-non-ideality-, and transport-associated parts — a decomposition that makes the result actionable, because it says which mechanism the fix must address.
Fill factor is the squareness of the current–voltage curve: the ratio of the device’s actual maximum power to the product of its open-circuit voltage and short-circuit current. A device can have excellent voltage and current and still deliver mediocre power if its curve sags in the middle — if charges recombine instead of reaching the contacts as the operating point moves toward maximum power. That sag is what grows with area in perovskite modules, and what the chill is designed to prevent.
The fix: fifteen seconds at 0–10 °C
Perovskite films made by solution coating start as a wet layer of precursor. In the usual factory sequence, a vacuum step then flashes the solvent away, and during drying a temporary intermediate phase forms before the final perovskite crystal. If that intermediate phase vanishes too quickly, crystals nucleate and grow unevenly, leaving a film with more defects, rougher topography, and weak spots distributed across the sheet — precisely the kind of disorder that scales badly, because a module is only as uniform as its worst region.
The team’s intervention, which they call cryo-controlled processing, inserts one step before the vacuum stage: the wet film is cooled for 15 seconds to 0–10 °C. The chill extends the lifetime of the intermediate phase, giving crystals more time to settle into even grains before the solvent leaves. Nothing about the chemistry changes — no new additive, no new layer, no new material. The only change is thermal history: a brief cold pause in the drying sequence.
The film-level measurements show the pause doing what the mechanism predicts. Surface roughness falls from 35.0 to 21.2 nm. Electron mobility more than doubles. Photoluminescence brightens, the standard signature of fewer defects trapping charges before collection. And the diode ideality factor — a number that flags wasteful recombination, with 1.0 as the ideal — falls from 1.29 to 1.15. Fill factor at the device level rises from about 80.0% to 84.5%, and the paper’s decomposition attributes most of that gain to the reduced diode non-ideality rather than to resistance or transport improvements. In other words, the chill mainly stops charges from recombining inside the film, which is the same mechanism the diagnosis identified as the dominant leak.
There is a second, commercially suggestive property: the fill-factor advantage grows with device size. A treatment whose benefit shrinks as devices get larger is a laboratory curiosity. A treatment whose benefit grows with area is a scale-up method, because it attacks a loss that itself grows with area. That trend, more than any single efficiency figure, is the reason this paper belongs in the scaling conversation rather than the record-chasing one.
The devices: 19.48 cm², eight cells, 23.9%
The headline mini-module links eight cells in series over a 19.48 cm² aperture — smaller than a playing card — with about 97% of the aperture generating power and the remainder consumed by interconnects. The best device, which also carries an anti-reflection coating, reached the certified steady-state 23.9% at 83.3% fill factor. The anti-reflection coating is worth naming rather than burying: the headline figure is a coated champion device, not a bare-stack average, and comparisons with uncoated modules should carry that qualifier.
Reproducibility data accompany the champion, which is not always the case in module papers. Across 15 devices per group, the chilled modules show a narrower efficiency spread than the controls — steadier results, not just a better best device. For a manufacturer, that spread matters as much as the peak: a process that produces one hero module and fourteen mediocre ones is not a process.
The team also reports that the same process produced a much larger module: 829.92 cm², 56 cells, on glass measuring 30 by 40 cm. The detailed data for that module sit in the paper’s supplementary files, and no efficiency figure for it is stated in the coverage consulted — so none is stated here. The large module should be read as proof that the process runs at that size, not as a result in its own right. The same restraint applies to the certification: the certifying laboratory is named in the supplementary files rather than the main text consulted, so this article reports the 23.9% as the paper’s certified steady-state figure without naming a certifier it has not verified.
Stability: the same leak governs decay
The paper’s most interesting claim may be its second one: that the same fill-factor limitation governs not only efficiency loss during scale-up but also performance decay during operation. If the defects that sag the current–voltage curve at manufacture are also the sites where degradation proceeds, then preserving fill factor at fabrication should slow aging as well — and the aging data point that way.
Under indoor light and ultraviolet testing, chilled devices age more slowly than controls, and the improvement is concentrated where the diagnosis predicts: fill-factor losses during aging shrink more than threefold, with the ideality-linked portion falling from 11.8 to 5.3 percentage points. Under ultraviolet light at 60 °C, chilled devices hold 80% of their starting efficiency until a cumulative ultraviolet dose of nearly 150 kWh, against under 100 kWh for controls.
The caveats here are the ones that always attach to indoor aging data, and they should travel with the numbers. These are controlled light and heat exposures, reported as ultraviolet doses rather than operating hours, and no outdoor exposure data accompany the paper. Rain, frost, humidity, and years of daily thermal cycling are untested for this design, and no damp-heat or IEC reliability-sequence result is stated in the material consulted. A dose of 150 kWh is not a lifetime, and it cannot be converted into one without an outdoor irradiance assumption this article will not invent. What the data do show is a mechanism-consistent aging benefit under the tested conditions — a reason to run the outdoor test, not a substitute for it.
Why a process step matters more than a new material
Place this result beside the other two scaling answers this desk has covered in the last week, and its particular value comes into focus. The UNSW–UtmoLight result of October 31 was a process-simplification answer: remove the nickel oxide layer, form the hole-selective contact during fabrication, and prove it at 676 cm². The Northwestern–DGIST result of October 27 was a chemistry answer: gate two defect passivators geometrically so they can treat both defect polarities at once, demonstrated on a 0.049 cm² cell. The Yan team’s answer is a thermal-history answer, demonstrated on a mini-module, using slot-die coating and laser scribing — both standard thin-film factory tools.
That “in principle” carries the usual weight. The paper reports no cost data, and a chill step that is trivial in a laboratory coater can be nontrivial at meter scale, where uniform cooling of a moving wet film in seconds is an engineering problem of its own. The team behind the work is unusually well placed to answer that objection — two of the three lead contacts work at Microquanta Semiconductor, a Hangzhou manufacturer with a perovskite module business, rather than at a university alone — but placement is not data, and no factory-line demonstration is claimed.
There is also a useful tension inside the paper’s own framing that the field should notice. The abstract’s record language and the results text’s “among the highest-performing” language are both in the same publication, and the careful version is the one that survives contact with the August Nanjing–Renshine 24.0% at 810 cm² and the October UNSW 23.5% at 676 cm². None of those three figures is directly comparable with the others — different aperture areas, different measurement protocols, different coating routes — and the field’s efficiency tables will have to decide how to class them. What can be said without qualification is that the mini-module class now has a 23.9% certified steady-state data point at 83.3% fill factor, achieved by preserving fill factor rather than by pushing voltage or current.
What to watch
Three things, in order of how quickly they could change the assessment.
First, documentation: a published certificate naming the certifying laboratory and protocol, and the large-module data from the supplements brought into the open, would remove the two qualifiers this article has had to carry. The 829.92 cm² module’s efficiency, if disclosed, converts a proof-of-scaling into a scaling result.
Second, outdoor and standardized stability data. An IEC 61215 sequence or a season of outdoor exposure would test whether the indoor ultraviolet-dose advantage survives the combined stressors — heat, humidity, and light acting together — that have historically decided perovskite commercialization. The Nanjing–Renshine paper’s IEC pass is the standard this result will implicitly be measured against.
Third, replication on other stacks. The chill regulates intermediate-phase evolution during crystallization, a mechanism that should in principle generalize across perovskite compositions — but “should” is doing work there. The first independent demonstration on a different composition, or in a different laboratory’s coater, will show whether cryo-controlled processing is a design rule or one team’s recipe.
The bottom line
A Microquanta–Zhejiang University team has given perovskite’s scaling problem a diagnosis precise enough to act on — fill factor, degraded mainly through diode non-ideality, is the dominant loss as devices grow — and a fix small enough to fit an existing line: chill the wet film for 15 seconds to 0–10 °C before vacuum drying, let the intermediate phase linger, and grow a smoother, less defective film. The film metrics move together (roughness 35.0 to 21.2 nm, mobility more than doubled, ideality factor 1.29 to 1.15), the device fill factor rises from about 80.0% to 84.5%, and the eight-cell mini-module reaches a certified steady-state 23.9% at 83.3% fill factor, with a narrower spread across 15 devices per group and an aging benefit concentrated in the same fill-factor channel.
It is a mini-module result, not a module result: 19.48 cm², a coated champion device, certification and large-module details in the supplements, and stability measured in indoor ultraviolet doses rather than outdoor years. Frame it as among the highest-performing mini-module results reported, achieved by holding the number that scale-up usually loses — and let the certificate, the 829.92 cm² data, and the outdoor test decide whether a fifteen-second chill becomes a manufacturing step.
No investment recommendation is made or implied anywhere in this article.
Sources and further reading
- Yan, B., Gong, R., Huang, Y., Gu, N., Li, R., Liu, D., Yan, K., Xu, X., Yang, Y. M., Yao, J., & Lan, D. “Retarding efficiency loss upon scaling in perovskite photovoltaics as fill factor becomes the bottleneck.” Nature Communications (published October 29, 2026; received June 29, 2026; accepted October 6, 2026). DOI: 10.1038/s41467-026-78867-1 — the primary source for the diagnosis (fill factor as the major scale-up contributor, resolved into voltage-, diode-non-ideality-, and transport-associated parts), the cryo-controlled process (15-second chill to 0–10 °C before vacuum drying, regulating intermediate-phase evolution), the abstract’s headline figures (19.48 cm² aperture, certified steady-state 23.9%, 83.3% fill factor), the author list and affiliations (Microquanta Semiconductor, Hangzhou; Zhejiang University, Hangzhou; China Three Gorges Renewables, Beijing; Australian Centre for Advanced Photovoltaics, Sydney), and the claim that the fill-factor advantage increases across device scales and governs operational decay. Equal-contribution authors: Buyi Yan, Ruihao Gong, Dongchen Lan; correspondence to Yan, Yao, and Lan. Note: the article is published as an Article in Press subject to further edits before the final Version of Record.
- ENTECH Online (November 2, 2026), “Perovskite Solar Mini-Modules Keep Efficiency as They Grow” — trade coverage of the same paper, and the source for the detailed figures the abstract does not carry: surface roughness 35.0 to 21.2 nm, electron mobility more than doubled, ideality factor 1.29 to 1.15, device fill factor about 80.0% to 84.5%, the eight-cell series layout, about 97% active aperture share, the anti-reflection coating on the best device, 15 devices per group with a narrower spread, the 829.92 cm² 56-cell module on 30 by 40 cm glass, the aging figures (fill-factor aging loss down more than threefold; ideality-linked loss 11.8 to 5.3 points; 80% retention to nearly 150 kWh ultraviolet dose at 60 °C vs. under 100 kWh for controls), the slot-die coating and laser scribing route, and the results text’s “among the highest-performing” wording. Its editor’s note states these numbers were checked against the Article in Press PDF.
- Excitonic.com, “Perovskite’s Scaling Test, Passed at 30 cm: UNSW and UtmoLight Hit 23.5% Without Nickel Oxide” (October 31, 2026) — the paired scaling story: 23.5% certified stabilized at 676 cm² aperture without nickel oxide, and the competing August 12, 2026 Nanjing University–Renshine Solar Nature paper reporting 24.0% certified at 810 cm² aperture and 22.0% at 0.72 m² with TÜV SÜD certification and an IEC 61215 pass, both cited here for the record-claim context.
- Excitonic.com, “Two Defect-Fixers That Don’t Fight Each Other Take Perovskite Cells to 27.4%” (October 27, 2026) — the small-area contrast: a 27.4% champion inverted cell on a 0.049 cm² aperture via sterically gated orthogonal defect passivation.
A note on sourcing: the full body text and supplementary files of the Yan paper were not directly readable through the browser fetch used for this article, which returned the abstract, author list, affiliations, and publication metadata only. Detailed device and aging figures are therefore reported via the November 2 trade coverage whose editor’s note states they were checked against the Article in Press PDF, and are attributed accordingly. No certifying laboratory is named here, because the certifier sits in the supplementary files and was not independently verified; no efficiency is stated for the 829.92 cm² module, because none appears in the material consulted; and no outdoor, damp-heat, or IEC reliability result is claimed, because none is stated in that material. The 23.9% figure is described as among the highest-performing mini-module results, not as an outright record, because the paper’s results text uses that wording and larger-area results at comparable or higher certified efficiencies have been published.