Perovskite solar cells have a stability problem that has survived a decade of defect passivation, and a team at The Hong Kong Polytechnic University (PolyU) argues the field has been looking at the wrong scale. The defects that matter most, they report, are not the ones spread across a film’s surface or buried at its interfaces. They are the ones concentrated at the boundaries between neighboring grains — and, more precisely, at the boundaries between grains that expose different crystal facets. In a mixed-facet film, the two sides of a single grain boundary lose charge at different rates, and that imbalance turns the more stable facet into the first thing to fail. A single additive, 3-fluorophenylethylammonium bromide (m-F-PEABr), designed to sit only at those boundaries, evens out the losses. Treated cells reach 26.55% efficiency and retain more than 90% of it after 2,200 hours of continuous one-sun operation at room temperature — and 1,250 hours at 85 °C. Led by Yang Guang of PolyU’s Department of Electrical and Electronic Engineering with Zhejiang University collaborators, the paper appeared in Watt in November; this article treats it as analysis, not breaking news.
The paradox: the stable facet ages first
A perovskite film is a mosaic of tiny crystals, and each crystal exposes a particular face to the world — its facet. In the formamidinium lead iodide (FAPbI₃) films this team studies, two facets dominate: (100) and (111), and solution-processed films normally contain both, mixed together. Using phenylammonium iodide in the precursor to bias crystallization, the team made three kinds of films — (100)-dominant, (111)-dominant, and mixed (100)–(111) — and watched them age under light.
The mixed films degraded fastest, but the X-ray diffraction told a stranger story: in the mixed films, the (111) facet lost intensity fastest — faster than the (100) facet in the same film, and faster than the (111) facet in a (111)-dominant film. The (111) facet, on its own, is the more photostable of the two. Put it next to a (100) grain, and it becomes the first thing to fail.
The team confirmed the anomaly is a property of mixing by sweeping the (111)/(100) ratio from 0.07 to 3.33. The dependence is non-monotonic: films with ratios between 0.77 and 2.25 degraded progressively faster after 300 hours of illumination, but when the film became (111)-dominant at a ratio of 3.33, photostability was markedly restored.
What the boundaries are doing under light
The mechanism rests on a difference in defect identity between the two facets, established by density functional theory, surface photovoltage mapping, and positron annihilation spectroscopy. The (111) facet is richer in formamidinium vacancies — missing FA⁺ cations, denoted V_FA⁻ — because that defect forms more readily on the (111) surface. The (100) facet, by contrast, carries more iodine vacancies (V_I⁺) and iodine interstitials, so the two facets trap opposite charges: the (100) side is dominated by electron-trapping defects, the (111) side by hole-trapping ones.
That asymmetry is what makes a hetero-facet boundary dangerous. At the boundary between a (100) grain and a (111) grain, the two defect populations set up a microscale electric field pointing from the (100) grain toward the (111) grain. Under illumination, the field drives holes from the (100) side into the (111) side and electrons the other way. The (111) side accumulates holes — a non-equilibrium pile-up of positive charge — and the trapped holes then drive chemistry. Interstitial iodide is oxidized stepwise, from I_i⁻ through I_i⁰ to I_i⁺, disrupting the coordination of the [PbI₆]⁴⁻ octahedral framework. Electrostatic stress follows, and the boundary amorphizes or transforms, precipitating lead iodide (PbI₂) phases strictly along the hetero-facet boundaries. The (111) grain does not fail because it is fragile. It fails because its neighbor hands it holes it cannot shed.
The team quantified the imbalance directly, using a confocal time-resolved photoluminescence technique that measures the carrier recombination velocity, S, at a hidden grain boundary from each side. At an untreated (100)–(111) boundary, S is about 300 ± 80 cm/s on the (100) side and about 800 ± 200 cm/s on the (111) side — a nearly threefold asymmetry across a boundary a few hundred nanometres wide. Three-dimensional photoluminescence mapping before and after 100 and 500 hours of illumination shows the degradation pathway starting at the boundary inside the (111) grain and spreading from there.
The fix: a molecule that only goes where it is needed
Rather than engineering the film toward a single facet orientation — hard to control precisely across compositions and at large area — the team chose to homogenize the recombination at the boundaries that already exist, using m-F-PEABr, a molecule whose architecture does two things at once.
First, it goes only to the boundaries. The molecule’s backbone is too bulky to enter the three-dimensional perovskite lattice, so it is sterically excluded from grain interiors and segregates into the defective grain boundaries. Atomic force microscopy infrared spectroscopy confirms the placement: the characteristic m-F-PEA⁺ signal at 1,243 cm⁻¹, the C–F vibration, is intensely localized at grain boundaries with negligible presence inside grains.
Second, once there, it performs a dual-anchoring repair. The terminal ammonium group (–NH₃⁺) mimics the missing FA⁺ cation, filling the formamidinium vacancy and neutralizing the local charge imbalance through hydrogen bonding. That primary coordination is locked in place by the meta-fluorinated benzene ring: the fluorine’s electronegativity creates a strong local dipole, angled to engage in weak C–F···I coupling with exposed iodide ions on the (111) facet. The meta isomer gives the optimal spatial distance for this secondary interaction, and the paper frames the dual-anchoring design as a principle that generalizes to other fluorinated aromatic ammonium salts.
Measured the same way as the problem, the result is the paper’s central number: after treatment, the recombination velocities on the two sides of a (100)–(111) boundary are approximately 300 ± 80 cm/s on the (100) side and 320 ± 100 cm/s on the (111) side — statistically indistinguishable. The asymmetry is gone, and trap-density measurements show a clear decrease in the treated devices that does not appear when m-F-PEABr is used only as a surface passivator, confirming a grain-boundary effect rather than a surface one.
One qualification belongs here, because it changes how the device numbers below should be read. Both the control and the treated films are additionally surface-passivated with piperazine hydroiodide (PipI), a proven surface-defect treatment applied to minimize the surface recombination gap. The 25.19%-to-26.55% comparison that follows is therefore the incremental benefit of grain-boundary homogenization on top of an already surface-passivated device — a cleaner test of the mechanism, and a reminder that the additive is one layer in a stack of treatments, not a standalone recipe.
The device numbers, handled in order
The devices are inverted single-junction cells — ITO/2PACz/perovskite/C₆₀/BCP/Ag — built on a FA₀.₉₃MA₀.₀₇PbI₃ absorber, measured through a 2.5 × 2.4 mm aperture, an active area of approximately 0.06 cm². This is a small laboratory cell, not a module, and every number below belongs to that scale. The paper reports no module-area result of any kind.
The champion treated cell reaches 26.55% power conversion efficiency, against 25.19% for the champion control, with an open-circuit voltage of 1.195 V, a fill factor of 84.5%, and a short-circuit current density of 26.30 mA/cm². Statistically, the picture is consistent: across the device population, the average open-circuit voltage rose by 0.035 V, the average fill factor by 2 percentage points, and the average short-circuit current by 0.2 mA/cm².
Two cautions travel with the efficiency figure. First, no independent certification is stated for the 26.55% value in the paper text consulted — it is the team’s reported champion, and should be quoted as such, not as a certified record. Second, the gain is real but modest: 1.36 percentage points over a control that was already at 25.19% with the same surface passivation. The paper’s claim is not that homogenizing recombination unlocks a new efficiency tier. It is that the same treatment that buys a point of efficiency buys far more in lifetime.
Stability: two temperatures, one protocol
The stability tests follow the ISOS-L-1 and ISOS-L-2 protocols: unencapsulated devices under continuous one-sun-equivalent illumination from a white LED, held at maximum power point, in nitrogen — at room temperature for ISOS-L-1 and at 85 °C for ISOS-L-2. The headline figure is the room-temperature T90, the time for efficiency to fall to 90% of its initial value: 2,200 hours for the treated devices, against 800 hours for the controls. At 85 °C the treated devices reach a T90 of 1,250 hours, against 270 hours for the controls — a more than fourfold improvement under the hotter, harsher condition.
Read the conditions alongside the numbers. The devices are unencapsulated and tested in nitrogen, so the test isolates photo-thermal degradation from the moisture-driven failure modes that have historically been perovskites’ hardest problem; no damp-heat data (85 °C at 85% relative humidity) appears in the paper text consulted. The illumination is a white LED rather than a full solar spectrum, and the test is continuous indoor aging, not outdoor exposure. A T90 of 2,200 hours is a strong laboratory durability signal, not a warranty claim.
Where it sits in the stability arc
The PolyU result slots into this desk’s perovskite stability arc as the heterogeneity answer. The Northwestern–DGIST paper of October 27 treated both defect polarities on the surface at once, reaching 27.4% with a 1,030-hour T90 at 85 °C under ISOS-L-3. The Southeast–Soochow flexible-module paper of October 22 fixed a buried-interface problem on plastic and carried it to 900 cm² modules with a certified IEC TS 63163 pass. The PolyU paper differs from both in what it names as the failure: not a surface, not an interface, not a process step, but a within-film heterogeneity — some grains aging faster than others because of who their neighbors are.
That framing explains the interest at a modest efficiency gain. Most passivation strategies reduce the total number of defects; this one reduces the variance of defect behavior across the film. A film in which every boundary loses charge at the same rate has no focal point for degradation. The paper’s own generality claim is narrower — the mechanism is demonstrated for mixed (100)/(111) boundaries in FAPbI₃-based perovskites, as Nam-Gyu Park notes in his accompanying Research Highlight. The honest comparison: PolyU’s 2,200-hour room-temperature T90 is the longer clock, but it was measured under ISOS-L-1/L-2 in nitrogen, not the ISOS-L-3 ambient-air protocol of the Northwestern–DGIST work, so the numbers are not directly comparable.
What to watch
Modules come first. The first demonstration that m-F-PEABr, or any grain-boundary homogenization treatment, holds its stability benefit at module area — and that the additive distributes uniformly through a coated film rather than pooling or depleting — will show whether this is a process or a laboratory observation.
A damp-heat result would convert the stability claim from a single-axis one into a commercial one: the 85 °C result in nitrogen is demanding, but perovskite commercialization has always been decided substantially by moisture, and a treatment that holds up under heat and light still has to prove itself under humidity, ideally encapsulated.
Generalization is the third watch item. The mechanism should, in principle, apply to other compositions — mixed-halide wide-bandgap perovskites for tandems — and to other facet pairs beyond (100)/(111). The first demonstration on a different composition will show whether homogenizing inter-grain recombination is a general stability strategy or a property of this particular FAPbI₃ system. And the thermal-ceiling question — whether a guanidinium or amidinium analogue of m-F-PEABr can push past the 120 °C deprotonation limit the authors flag for primary ammonium cations — will decide whether the chemistry has headroom or a hard stop.
The bottom line
A PolyU team, with collaborators at Zhejiang University, has identified a degradation pathway that earlier passivation strategies could not see, because it lives not in the average defect density of a film but in the difference between neighbors. In mixed-facet FAPbI₃ films, (100) and (111) grains carry different defects — iodine vacancies on one side, formamidinium vacancies on the other — and the resulting microscale electric field piles holes up on the (111) side of their shared boundary, oxidizing interstitial iodide and precipitating lead iodide along the boundary. One bulky additive, m-F-PEABr, parks at the boundary and evens out the recombination — from 300 and 800 cm/s on the two sides to 300 and 320 cm/s — and the treated cell reaches 26.55% and holds 90% of its efficiency for 2,200 hours at room temperature and 1,250 hours at 85 °C, against 800 and 270 hours for the control.
It is a small-area laboratory result — a 0.06 cm² cell, unencapsulated, tested in nitrogen, with no module data, no humidity data, and no independent certification — and it should be quoted with all of those qualifications attached. It is also a new way of thinking about why perovskite films age: not grain by grain, but boundary by boundary, and not by how many defects a film has, but by how unevenly they are shared. The number to watch is not 26.55%. It is the first module that ages evenly.
No investment recommendation is made or implied anywhere in this article.
Sources and further reading
- Zhang, H., Yu, B., Zhang, D., et al. “Stabilizing high-efficiency perovskite solar cells by homogenizing inter-grain recombination.” Watt 1:11 (2026). DOI: 10.1007/s44503-026-00013-1 — the primary source for the mechanism (facet-dependent defect identities V_I⁺ and V_FA⁻; asymmetric recombination velocities of ~300 ± 80 and ~800 ± 200 cm/s at untreated (100)–(111) boundaries, converging to ~300 ± 80 and ~320 ± 100 cm/s after treatment; the hole-accumulation and iodide-oxidation degradation pathway; the non-monotonic (111)/(100) ratio dependence, 0.77–2.25 degrading faster, 3.33 restored), the m-F-PEABr dual-anchoring design (ammonium filling of V_FA⁻, meta-fluorine C–F···I coupling, AFM-IR localization at 1,243 cm⁻¹, steric exclusion from the lattice), the device figures (26.55% champion vs. 25.19% control, 1.195 V, 84.5% fill factor, 26.30 mA/cm², average gains of 0.035 V, 2 points of fill factor, 0.2 mA/cm²; EQE-integrated 25.62 mA/cm²), the device stack (ITO/2PACz/perovskite/C₆₀/BCP/Ag; FA₀.₉₃MA₀.₀₇PbI₃; ~0.06 cm² active area, 2.5 × 2.4 mm aperture), the PipI surface passivation applied to both control and target, the stability results (unencapsulated, ISOS-L-1/L-2, white-LED one-sun, MPP tracking in N₂; T90 2,200 h at room temperature vs. 800 h control; T90 1,250 h at 85 °C vs. 270 h control), and the thermal ceiling above ~120 °C with the guanidinium/amidinium proposal. Team led by Yang Guang (PolyU Department of Electrical and Electronic Engineering) with Zhejiang University collaborators. No module-area result, damp-heat data, or independent efficiency certification is stated in the paper text consulted, and none is claimed here.
- Park, N.-G. “Homogenizing inter-grain recombination stabilizes perovskite solar cells.” Watt (2026). DOI: 10.1007/s44503-026-00015-z — the Research Highlight in the same journal, the source for the framing that the mechanism is demonstrated primarily for mixed (100)/(111) grain boundaries in FAPbI₃-based perovskites, and for the reported 26.55% efficiency, 1.20 V open-circuit voltage, and retention above 90% after more than 2,200 hours.
- Tech Xplore (November 4, 2026), “Researchers uncover hidden cause of aging in perovskite solar cells” — trade coverage of the same result, the source for the November 4 announcement date, the PolyU–Zhejiang collaboration, and Yang Guang’s quoted framing that uneven charge losses between neighboring grains are a key cause of aging; it carries the 26.55% and 2,200-hour figures but no device-area, control-curve, or module data, which is why this article works from the paper directly.
- Excitonic.com, “Two Defect-Fixers That Don’t Fight Each Other Take Perovskite Cells to 27.4%” (October 27, 2026) — the paired stability story: Northwestern–DGIST orthogonal defect passivation, 27.4% champion, 1,030-hour T90 at 85 °C under ISOS-L-3; and “Flexible Perovskites Stop Fighting Themselves: 26.86% Certified, Modules Hold Above 20% at 900 cm²” (November 5, 2026) — the scaling counterpoint, with certified module efficiencies and a TÜV SÜD-certified IEC TS 63163 pass that the PolyU result does not yet have.
A note on sourcing: the full body text of the Zhang paper was readable through the journal’s online version, and all device, mechanism, and stability figures above are taken from it directly, with trade coverage used only for the announcement date and the team framing. The complete author list beyond the first author (Hengyu Zhang) and the paper’s exact publication date are not stated in the material consulted, and neither is claimed here. No damp-heat, outdoor-exposure, cost, yield, or module-area data are stated, because none appears in the paper text consulted. Background on perovskite defect chemistry and ISOS stability protocols is drawn from the standard perovskite literature.