Two weeks ago, a perovskite paper landed in Science that deserves to be read for what it does to the flexible side of the scaling problem, not for its headline number alone. On October 22, a team led by Yeyong Wu and Tiankai Zhang at Southeast University, with Yaowen Li and Guiying Xu at Soochow University, published a study that attacks the reason flexible perovskite modules have lagged their rigid cousins: the two layers that have to be printed together onto a plastic foil were chemically fighting each other before either could settle into place. Their fix is a single additive molecule, added to the same printing solution, that acts as a referee between them. The results are the strongest flexible-module numbers reported to date — a third-party certified 26.86% for a small-area flexible cell, and certified module efficiencies of 22.31% at 25 cm², 20.63% at 100 cm², and 20.15% at 900 cm², with a 25 cm² device holding 92.5% of its efficiency after 1,044 hours of maximum-power-point tracking at 85 °C and 65% relative humidity.
This article frames it the way the calendar demands: as an October paper surfacing in early November, not as breaking news. The numbers are certified and the stability test is a recognized industry protocol — but no outdoor operation has been demonstrated, no roll-to-roll line has run the process, and the largest module’s area accounting differs between the paper and the trade coverage, a discrepancy this article flags rather than papers over.
The problem: a kinetic mismatch on plastic
Flexible perovskite cells are built on plastic foils — PET coated with indium tin oxide — rather than glass, and the foil is where the trouble starts. The hole-selective contact in these inverted devices is a self-assembled monolayer (SAM), in this work Me-4PACz, a carbazole-phosphonic-acid molecule that anchors itself to the substrate through its phosphonic acid group binding to surface hydroxyl groups. On glass, that anchoring is straightforward. On a flexible foil, the ITO surface carries fewer hydroxyl groups, and distributes them less evenly — so the SAM anchors patchily, and the polar perovskite precursor wets the surface poorly.
The obvious workaround is to print the SAM and the perovskite together, in one co-deposition step, so the wetting problem disappears. But that creates a second problem, and it is the one this paper names precisely. In the mixed precursor solution, the hydroxyl groups on the substrate surface are rapidly lost to the polar solvent, while the perovskite colloids in solution bind strongly to the phosphonic acid groups of the SAM molecules. The two processes run on different clocks: the colloid–SAM interaction retards SAM adsorption, so the hydroxyl groups are depleted before the SAM can anchor efficiently. The result is a SAM layer with poor coverage and a perovskite film that crystallizes badly on top of it — uneven, defect-ridden, and unstable under heat and humidity. As the paper’s editor’s summary puts it, surface hydroxyls desorbing on flexible substrates disrupt both SAM assembly and perovskite crystallization, and that disruption is the reason flexible devices have fallen so far behind rigid ones as they scale.
The numbers tell the size of the gap. Small-area flexible cells had already reached 26.22% before this work — close to rigid-cell performance. But scaling the same material to module size cost far more efficiency on plastic than on glass, and the stability penalty was worse.
The fix: one molecule that binds the other side first
The team’s answer is what they call decoupled co-deposition, and the decoupling agent is a purpose-built molecule: 3,5-bis(methylsulfonyl)phenylammonium iodide, abbreviated BSAI. It carries an ammonium group and two sulfonyl groups, and it is added to the mixed solution containing the SAM and the perovskite precursors. Its job is to bind preferentially to the perovskite precursor species — the lead–iodide colloids — so that those colloids stop binding to the SAM’s phosphonic acid groups.
That single intervention does two things at once, and the paper is careful to separate them. First, with the colloid–SAM coordination disrupted, the Me-4PACz molecules can anchor quickly to the substrate’s hydroxyl groups before those groups desorb, producing a compact, uniform SAM across the large-area foil. Second, BSAI’s binding to the perovskite colloids retards nucleation, slowing crystal formation enough to promote oriented, even growth rather than the rushed, defective crystallization the interference used to cause. The additive also seals defects in the perovskite layer as it forms. One molecule, added to the printing ink, fixes the interface and the crystal at the same time.
For a manufacturer, an additive that rides in the existing printing ink is cheaper to adopt than a new deposition step — though no cost or yield data accompany the paper, and adoption is not demonstrated here.
The devices: 26.86% certified, and above 20% at every scale
The headline small-area result is a third-party certified 26.86% for a flexible cell — the paper’s structured abstract rounds it to 26.9%, and the university release gives the precise 26.86% figure this article uses throughout. Against the previous small-area flexible benchmark of 26.22% cited in the paper, that is a real step, not a rounding artifact. But the small-area figure is the least interesting number in the paper, for the usual reason: small-area flexible cells were already close to rigid performance. The test of this strategy is what survives the jump to module size.
The module results, as reported by the university release and the trade coverage, are 22.31% for a 5 × 5 cm² module, 20.63% for a 10 × 10 cm² module, and 20.15% for a 30 × 30 cm² module — the last with a 750 cm² active device area. The release describes all three as record efficiencies reported to date for flexible modules, and the Science structured abstract gives the same certified values, rounded: 22.3%, 20.6%, and 20.2%, on device areas of 14.55, 80, and 750 cm² respectively. The 20.15% figure at 750 cm² device area is the number that changes the conversation: the first time a flexible perovskite module has held above 20% at that scale, a threshold that matters commercially because a module that cannot hold 20% at size is difficult to sell against silicon on any surface, curved or flat.
One area-basis caveat belongs in the open. The university release and trade coverage describe the largest module as a 30 × 30 cm² sheet — 900 cm² total, 750 cm² active — while the Science structured abstract lists the largest module’s substrate area as 1,050 cm² with the same 750 cm² device area. The certified efficiency, 20.15% (20.2% rounded), is the same in both accounts; what differs is the stated total substrate area. This article follows the release framing of 900 cm² in its headline and body, and flags the discrepancy here and in the sourcing note rather than silently choosing one.
A second qualifier travels with the certified figures. The Science abstract describes the efficiencies as certified and the stability pass as certified by TÜV SÜD, but the material consulted does not name the certifying laboratory for the 26.86% cell or the module efficiencies themselves. That is a thinner certification trail than the Nanjing University–Renshine rigid-module paper of August, which names TÜV SÜD explicitly for its 24.0% at 810 cm². The responsible formulation is that the figures are the paper’s reported certified values, with the certifier for the efficiency figures themselves not named in the sources consulted.
Stability: a real protocol, under real stress
The stability data are where this paper separates itself from most flexible-module reports, and they deserve to be read in full. A 25 cm² flexible device — the one with a 14.55 cm² device area, using 1,3-diaminopropane dihydroiodide as the passivation molecule — retained 92.5% of its initial 22% efficiency after 1,044 hours of maximum-power-point tracking at 85 °C and 65% relative humidity. That is the ISOS-L-3 protocol: continuous light, elevated temperature, and high humidity together, the combination that has historically killed perovskite devices fastest. The paper reports that the modules retained more than 92% of their efficiency after roughly 1,000 hours under the same protocol, and that the devices passed all sequences of IEC TS 63163:2021, the international standard written specifically for flexible photovoltaic modules, as certified by TÜV SÜD.
Two things make that combination worth taking seriously. First, the stress is combined — heat and humidity and light at once — rather than the single-stressor indoor aging that much of the field still reports. The Microquanta cryo paper covered by this desk on November 2, for instance, reported its aging benefit as an ultraviolet dose under controlled light, with no damp-heat or IEC sequence stated. An IEC TS 63163 pass, certified by TÜV SÜD, is a different class of evidence: it is the standard a buyer or insurer would ask about, and passing all of its sequences is the closest a laboratory result comes to a commercial durability claim. Second, the stability result is mechanism-consistent with the chemistry: if the failure mode was a patchy SAM and a defective buried interface, then fixing that interface should slow degradation under heat and humidity, and the retention figures point that way.
The limits, stated plainly by the authors themselves, are that long-term outdoor operation and mass-production readiness have not yet been demonstrated. A thousand hours under ISOS-L-3 is a severe test, but it is not a year on a roof or a car roof, and it is not a roll-to-roll line running at speed. Thermal shock cycles from cold to extreme heat are reported as survived, but the paper’s own framing is that this work provides a foundation for industrial development, not a demonstration of it. The next steps the team names — refining the decoupling molecules, integrating tandem cell designs, and scaling toward roll-to-roll manufacturing — are research directions, not announcements.
Where it sits in the scaling arc
Place this result beside the other perovskite scaling stories of the last two weeks, and the pattern sharpens. October has produced four distinct answers to the same question — can perovskite hold its efficiency as it grows? — and each attacks a different part of the problem. The Northwestern–DGIST small-cell result of October 27 was a defect-passivation answer at 0.049 cm², reaching 27.4%. The UNSW–UtmoLight submodule of October 31 was a process-simplification answer at 676 cm², reaching 23.5% without nickel oxide. The Microquanta–Zhejiang mini-module of November 2 was a thermal-history answer at 19.48 cm², reaching a certified 23.9% by preserving fill factor. The Southeast–Soochow result is the flexible answer: the same scaling question asked on plastic, answered with interface chemistry, holding above 20% to a 750 cm² device area with a certified stability pass.
The comparisons that matter are the honest ones. Against rigid modules, the flexible numbers still lag: the Nanjing–Renshine rigid module holds a certified 24.0% at an 810 cm² aperture with its own IEC 61215 pass, against 20.15% flexible at 750 cm² device area. That gap is the price of the plastic substrate, and this paper narrows it without closing it. Against the flexible field itself, the advance is in the combination — efficiency at scale plus a recognized stability standard — rather than in any single figure.
There is also a generality claim worth noting, because it determines whether BSAI is a recipe or a design rule. The paper reports that the strategy also works on rigid substrates and with different SAM molecules, not only Me-4PACz on PET. If that holds up under independent replication, the decoupling idea — bind the colloids first, let the SAM anchor second — becomes a principle other groups can apply to their own stacks, including the tandem designs the team names as its next direction. Replication is the operative word: a single laboratory’s demonstration, however well certified, is not yet a design rule.
What to watch
Three things, in order of how quickly they could change the assessment.
First, documentation. A published certificate naming the certifying laboratory for the 26.86% cell and the module efficiencies would remove the qualifier this article has had to carry, and a clarification of the largest module’s substrate area — 900 cm² as the release states, or 1,050 cm² as the Science abstract states — would settle the area accounting. Neither changes the certified efficiency on the 750 cm² device area, but both would let the figures be quoted without footnotes.
Second, outdoor and manufacturing data. A season of outdoor exposure would test whether the ISOS-L-3 retention survives the combined stressors of real deployment, and a roll-to-roll demonstration — the team’s own named next step — would test whether a decoupling molecule that works in a laboratory coater works in a moving web, where uniform dosing of an additive across meters of foil is its own engineering problem. No cost data accompany the paper, so the economic case for the additive is untested as well.
Third, replication on other stacks and in other laboratories. The paper’s generality claim — rigid substrates, different SAM molecules — decides whether decoupled co-deposition is a platform or one team’s molecule. The first independent demonstration on a different composition, or a tandem integration, will show which.
The bottom line
A Southeast University–Soochow University team has given flexible perovskite’s scaling problem a chemical diagnosis precise enough to act on — perovskite colloids bind the SAM’s phosphonic acid groups faster than the SAM can anchor to a hydroxyl-poor plastic surface — and a fix small enough to ride in the printing ink: a decoupling molecule, BSAI, that binds the colloids first, lets the SAM anchor, and slows crystallization into oriented, even grains. The small-area flexible cell reaches a certified 26.86%, the modules hold 22.31%, 20.63%, and 20.15% at 25, 100, and 900 cm², and a 25 cm² device retains 92.5% of its efficiency after 1,044 hours at 85 °C and 65% relative humidity, with a TÜV SÜD-certified pass of all IEC TS 63163 sequences.
It is an October paper, not a breaking story, and its limits are its authors’ own: no outdoor operation demonstrated, no mass production shown, no cost data, and a certification trail for the efficiency figures that names no laboratory in the material consulted. Frame it as the strongest flexible-module result reported to date, achieved by stopping two layers from fighting each other — and let the certificate, the outdoor test, and the roll-to-roll line decide whether a decoupling molecule becomes a manufacturing step.
No investment recommendation is made or implied anywhere in this article.
Sources and further reading
- Wu, Y., Zhang, X., Zhang, T., Xu, G., Shen, Y., Guo, C., Miao, Q., Ding, J., Yin, Y., Cheng, Q., Liu, W., Wu, X., Chen, K., Kou, W., Cao, Z., Chen, H., Chen, W., Chen, Y., Gao, F., Li, Y., & Li, Y. “Decoupled self-assembled monolayer and perovskite deposition for efficient flexible solar modules.” Science (published October 22, 2026). DOI: 10.1126/science.aef9745 — the primary source for the paper title and author list, the decoupled co-deposition strategy, the BSAI decoupling agent (3,5-bis(methylsulfonyl)phenylammonium iodide; ammonium and dual sulfonyl groups binding Pb–I colloids), the Me-4PACz SAM, the kinetic-mismatch mechanism (hydroxyl desorption vs. SAM anchoring; colloid–phosphonate interaction retarding adsorption), the certified efficiencies (26.9% small-area flexible, rounded; 22.3%, 20.6%, 20.2% on device areas of 14.55, 80, and 750 cm²), the stability results (>92% retention after ~1,000 hours under ISOS-L-3 at 85 °C and 65% relative humidity; all IEC TS 63163:2021 sequences passed, certified by TÜV SÜD), the passivation molecules (phenethylammonium iodide for the 750 cm² device; 1,3-diaminopropane dihydroiodide for the 22.0% device), the prior small-area flexible benchmark of 26.22%, and the generality claim (rigid substrates, different SAM molecules).
- Southeast University, School of Materials Science and Engineering (October 23, 2026), “Decoupled SAM-Perovskite Co-deposition for Efficient and Stable Flexible Solar Modules” — the university release, and the source for the precise 26.86% figure, the module figures as 22.31%, 20.63%, and 20.15% at 5 × 5, 10 × 10, and 30 × 30 cm², the “record efficiencies reported to date” framing, and the team leadership (Tiankai Zhang, Southeast; Yaowen Li and Guiying Xu, Soochow; Zhang a co-first author, Li and Xu co-corresponding).
- Interesting Engineering (October 26, 2026), “Flexible solar cells hit 26.86% efficiency in chemical breakthrough” — trade coverage of the same paper, and the source for the 92.5% retention figure as quoted (“retained 92.5 percent of its initial PCE of 22 percent after 1044 hours of maximum power point tracking at 85 °C and 65 percent relative humidity”), the decoupling molecule described as ammonium and iodide ions plus two sulfonate groups, the thermal-shock survival, and the IEC 63163 framing.
- Troy Technical (November 1, 2026), “Perovskite Solar Cells: Flexible Modules with 26.9% Efficiency” — confirming the module figures and the 750 cm² active area at 900 cm², and the source for the authors’ stated limits (long-term outdoor operation and mass-production readiness not yet demonstrated; future work on roll-to-roll manufacturing and tandem compatibility).
- Excitonic.com, “A 15-Second Chill Holds the Fill Factor: Perovskite Mini-Modules Hit 23.9% Certified” (November 2, 2026) — the paired scaling story: cryo-controlled processing, certified steady-state 23.9% at 19.48 cm², and the competing rigid-module context (Nanjing University–Renshine Solar, certified 24.0% at 810 cm² aperture, TÜV SÜD, IEC 61215 pass); and “Perovskite’s Scaling Test, Passed at 30 cm: UNSW and UtmoLight Hit 23.5% Without Nickel Oxide” (October 31, 2026) — 23.5% certified stabilized at 676 cm² aperture without nickel oxide.
A note on sourcing: the full body text of the Wu paper was not directly readable through the browser fetch used for this article, which returned the editor’s summary, structured abstract, abstract, and references only. Detailed device figures are therefore reported from the Science structured abstract and the university release, with trade coverage used for the quoted stability figure and the authors’ stated limits. The largest module’s substrate area is stated as 900 cm² (30 × 30 cm²) in the release and trade coverage but as 1,050 cm² in the Science structured abstract; both agree on the 750 cm² device area and the 20.15% (20.2% rounded) certified efficiency, and this article follows the release framing while flagging the difference. No certifying laboratory is named here for the 26.86% cell or the module efficiencies, because none is named in the material consulted; the TÜV SÜD attribution applies to the IEC TS 63163 pass, as stated in the paper’s structured abstract. No outdoor-exposure, cost, yield, or roll-to-roll data are stated, because none appears in the material consulted.