Perovskite–silicon tandems have a stability problem that no amount of interface engineering has fully solved, and the field has mostly attacked it by trying things: new passivation layers, new additives, new encapsulation schemes — each tested by ageing cells for a thousand hours and seeing what survives. A paper published October 29 in Nature Energy offers a different kind of answer: a measurable property of the perovskite itself that predicts, before any ageing test begins, whether the absorber will hold up under light and heat. The property is lattice compliance — the crystal lattice’s capacity to elastically accommodate stress — and the team that identified it, led by Fuzong Xu, Jiang Liu, and Xuechun Zhang at King Abdullah University of Science and Technology (KAUST), with Stefaan De Wolf among the co-authors, used it to build bifacial monolithic perovskite–silicon tandems that delivered 36.2 mW/cm² of power-generation density under a 0.20 ground albedo and retained 90% of their initial efficiency after more than 1,000 hours of ISOS-L-2 ageing.

The paper, “Perovskite lattice compliance enables efficient and stable bifacial perovskite/silicon tandem photovoltaics,” matters less for its device numbers than for what it replaces. Trial-and-error stability screening — build a composition, age it, keep it or discard it — is slow, expensive, and tells you nothing about the next composition. A descriptor you can measure on a film, before a device exists, is a screening criterion the whole tandem lane can use.

The bromide trade: fixing one failure mode exposes another

The absorbers in this study sit at roughly 1.55 eV — a deliberately bromide-lean choice. Wide-bandgap perovskites suffer light-induced halide segregation, the tendency of iodide and bromide to unmix under illumination into domains that drag the voltage down. Cutting the bromide largely removes that failure mode. But it does not remove failure. It shifts the dominant stability challenge, in the paper’s own framing, to the intrinsic photothermal resilience of the absorber — how the crystal itself holds up when light and heat arrive together, day after day.

That is a harder problem to screen for, because it lives inside the crystal rather than at an interface. A passivation layer can be swapped in an afternoon; a crystal’s response to heat is a property of its composition. And heat is not a minor axis for a tandem in the field: a module on a rooftop or in a desert array runs hot — cell temperatures of 65 °C and above are routine, not abuse cases — and a perovskite film bonded to a silicon wafer is under stress from the moment the sun comes up.

Lattice compliance, in plain terms

Lattice compliance is the crystal’s willingness to flex. A perovskite is a soft ionic crystal: a cage of lead and halide ions with a cation — caesium, methylammonium, formamidinium, or a mixture — sitting inside each cage. When the film heats up, the lattice wants to expand; clamped to a substrate that expands differently, or with light driving ions around, the lattice is stressed. A stiff lattice has nowhere for that stress to go except into defects: the crystal distorts locally, atoms shift out of place, and the strain accumulates as microstrain — small, inhomogeneous distortions of the lattice spacing visible as broadened X-ray diffraction peaks. A compliant lattice accommodates the same stress elastically, the way a spring takes a load and returns to shape. The distortion stays reversible instead of freezing in as damage.

The team made this concrete by comparing four compositions that differ only in their cation and halide mix: (Cs-MA-FA)(I), (Cs-MA-FA)(I-Br), (Cs-FA)(I), and (Cs-FA)(I-Br). Temperature-dependent X-ray diffraction, which watches the crystal structure as the film is heated from 25 °C to 85 °C, ranked them, and density-functional-theory calculations of a two-mode stiffness descriptor — the lattice’s resistance to shear and to tetragonal distortion — agreed: 246.9 GPa for (Cs-MA-FA)(I), 255.4 for (Cs-MA-FA)(I-Br), 312.4 for (Cs-FA)(I), and 331.8 for (Cs-FA)(I-Br), with lower meaning softer and more compliant. The methylammonium-containing, bromide-free composition is the most compliant; the caesium–formamidinium bromide-containing composition is the stiffest. Nanomechanical atomic-force microscopy, mapping Young’s modulus across the films, corroborated the ranking at the grain level.

What compliance buys: shallower defects, quieter band edges

The mechanism the paper traces runs from that mechanical ranking through the defect physics to the device. Under combined light and heat, the stiffer compositions accumulate thermally induced microstrain; the compliant ones suppress it. That matters because microstrain is the physical environment defects live in. In the compliant films, deep-level defects — the trap states deep inside the bandgap that capture charge carriers and kill them non-radiatively — shift toward shallower states, closer to the band edges, where they do far less damage. And the band-edge disorder, measured as the Urbach energy, the smearing of the absorption edge that quantifies how messy the electronic states near the bandgap are, is reduced.

The team tested the causal chain in both directions. Time-dependent photoluminescence under one sun, under five suns, and under combined light and heat at 85 °C showed the compliant compositions holding their emission while the stiffer ones faded. Adding 10 mol% methylammonium chloride to the stiff (Cs-FA)(I) composition improved its photoluminescence stability under the same stress, consistent with the methylammonium cation being part of what makes the lattice compliant.

The single-junction devices carry the same ranking into working cells. The compliant (Cs-MA-FA)(I) composition reached a certified 26.45% power conversion efficiency as a single-junction cell, and the initial efficiencies in the stability tests ran 25.3% for the compliant composition against 24.1% for the stiff (Cs-FA)(I-Br) under the 65 °C ageing condition. Thermal cycling under the ISOS-T-2 protocol was run across all four compositions plus two 1.68 eV absorbers — the compliance idea is being tested against the wider-bandgap compositions the field actually wants for tandems, not only the 1.55 eV system where it was developed.

Why bifacial changes the stability question

The device demonstration is deliberately bifacial, and that choice is part of the argument. A bifacial tandem collects light from both faces: direct sun on the front, ground-reflected light on the rear. For a perovskite–silicon tandem, the rear light matters disproportionately, because the silicon bottom cell is the half that sees it — the perovskite top cell is opaque to the visible light the rear face receives, so the reflected photons pass through to the silicon. That changes the current-matching problem at the heart of every monolithic tandem. In a two-terminal device the two subcells are wired in series, so the device current is set by whichever subcell produces less; under front-only illumination, the silicon bottom cell is usually the limiting one. Rear illumination feeds the silicon extra current, pushing the device toward current matching — and, if the rear contribution is large enough, past it, into a regime where the perovskite top cell becomes the limiter instead.

The supplementary data make that evolution explicit. With no rear illumination, the champion tandem measured a short-circuit current density of 15.2 mA/cm² and a power-generation density of 23.8 mW/cm² — silicon-limited. At an albedo-equivalent rear illumination of 0.20, the current rose to 23.9 mA/cm² and the power-generation density to 36.2 mW/cm², approaching current matching. At 0.25, the current reached 24.7 mA/cm² and the power density 37.3 mW/cm², and the device had crossed into perovskite-limited operation. The albedo figure is not arbitrary: 0.20 is the default ground albedo in PVsyst, the industry’s standard yield-modelling software, in the middle of measured ranges for ordinary ground. The 0.25 case represents bright grass, light soil, or concrete, not a whitened or mirrored ground.

This is also why the paper reports power-generation density rather than a single efficiency figure, and why this article does not quote one for the tandem. A bifacial device under 1,000 W/m² of front illumination plus 200 W/m² from the rear is receiving 1,200 W/m² in total; an efficiency number would depend on which irradiance you divide by, and the authors, correctly, do not pretend otherwise. Power-generation density — watts out per square centimetre of device — is the honest metric for a bifacial cell, and the one that matters commercially. The supplementary material is explicit that power-generation-density values should only be compared together with their rear-illumination condition: the 36.2 mW/cm² figure belongs to albedo 0.20, and it should always be quoted with that attached.

The stability result is the other half of the demonstration, and it is where the descriptor earns its keep. Under ISOS-L-2 ageing — one-sun illumination at 65 °C, the consensus laboratory protocol for light soaking at elevated temperature — the bifacial tandems retained 90% of their initial efficiency after more than 1,000 hours. That is a serious number for a perovskite top cell, because heat plus light is the axis on which perovskites have historically failed first, and a bifacial device in the field runs hotter than a monofacial one: it absorbs light from both sides and sits above a reflective ground that returns heat as well as light. The single-junction outdoor data agree: devices based on the compliant (Cs-MA-FA)(I) composition, exposed outdoors at KAUST from July 4 to August 10, 2025, retained 89% of their initial efficiency, against 79% for the stiff (Cs-FA)(I-Br) composition. Laboratory ageing and a Saudi summer agree on the ranking. For scale, the 36.2 mW/cm² figure sits ahead of the same group’s 2023 Matter bifacial benchmark of above 27 mW/cm² — three years of progress, increasingly from the absorber itself rather than the interfaces around it.

Why a descriptor travels further than a recipe

The field’s stability work has mostly produced recipes: add this molecule, anneal at that temperature, encapsulate this way. Recipes are useful and they are also fragile — they work on one stack, in one lab, until they do not, and they give the next group nothing to calculate. A descriptor is different. Lattice compliance can be computed from first principles before a film is ever deposited, and it can be measured on a film by temperature-dependent diffraction or nanomechanical mapping before a device is built. A group screening twenty candidate compositions can rank them by compliance and age only the top three. That is the practical meaning of “replacing trial-and-error stability screening with a measurable materials descriptor”: it converts a thousand-hour test into a calculation and a measurement. It also reframes the cation and halide choices the field argues about most — methylammonium’s volatility, bromide’s segregation — as mechanical questions: the methylammonium-containing composition is the most compliant and the most stable under light and heat, which complicates the simple story that methylammonium must be eliminated, though volatility and compliance are different failure modes.

What this is not: a module, a weather test, or a record

First, it is not a module. The devices are laboratory cells, and nothing in the material consulted addresses large-area uniformity, coating throughput, yield, or cost — the quantities that decide whether a design rule survives translation from a spin-coater to a production line.

Second, the stability evidence is laboratory ageing plus a five-week outdoor single-junction run. ISOS-L-2 at 65 °C is demanding, and 90% retention after 1,000 hours is a real result, but it is not damp heat at 85 °C and 85% relative humidity, not thermal cycling under illumination, and not a year of outdoor exposure of a tandem module. The outdoor data are single-junction devices, not the bifacial tandem itself, and moisture-driven degradation — historically perovskites’ hardest failure mode — is not addressed in the material consulted.

Third, there is no headline efficiency figure for the tandem in the material consulted, and the absence is informative: bifacial devices are properly reported in power-generation density at a stated albedo, so the usual leaderboard comparison against the 35%-plus monofacial records cannot be made directly. Fourth, the mechanism evidence, while multi-modal, comes from one laboratory and one materials system, across four compositions that share a cation/halide family. Whether it generalizes to other perovskite families, to tin-containing absorbers, or to the wide-bandgap compositions that monofacial tandems actually use, is the open question the descriptor now has to answer.

What to watch

The most informative follow-up would be a second materials family. If lattice compliance is genuinely a descriptor rather than a four-composition correlation, then a different perovskite system — a tin–lead absorber, a wide-bandgap bromide composition, a two-dimensional/three-dimensional heterostructure — should show the same pattern: the more compliant variant outlasting the stiffer one under light and heat. A second family would convert a demonstration into a rule. A damp-heat result on the compliant composition would convert the stability question from a laboratory claim into a commercial one, and a larger-area demonstration would show whether a lattice-level design rule survives the coating and contacting steps that dominate manufacturing. For compliance to become a routine screening tool, the field will also need a fast, standard way to measure it — a single number from a single technique, reported the same way by every lab.

The bottom line

A KAUST team has shown that the photothermal stability of a bromide-lean perovskite absorber is governed by something you can measure before you age it: lattice compliance, the crystal’s capacity to elastically accommodate stress. Across four compositions, the most compliant — (Cs-MA-FA)(I), at 246.9 GPa on the team’s two-mode stiffness scale — suppresses thermally induced microstrain, shifts deep-level defects toward shallower states, and reduces band-edge disorder, while the stiffest — (Cs-FA)(I-Br), at 331.8 GPa — accumulates the damage. Built into bifacial monolithic tandems, the compliant absorber delivers 36.2 mW/cm² at a 0.20 ground albedo and holds 90% of its efficiency after more than 1,000 hours at one sun and 65 °C.

It is not a record, not a module, and not a weather test, and it carries no headline efficiency figure — deliberately, because bifacial devices are reported in power-generation density at a stated albedo. It is a design rule for the hardest problem in tandem manufacturing: knowing, before you build and age a cell, whether the absorber inside it will survive the heat and light of the field. The number to watch is not 36.2 mW/cm². It is the second materials family.

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

Sources and further reading

  • Xu, F., Liu, J., Zhang, X., et al. “Perovskite lattice compliance enables efficient and stable bifacial perovskite/silicon tandem photovoltaics.” Nature Energy (2026), published October 29, 2026. DOI: 10.1038/s41560-026-02162-y — the primary source for the paper’s framing (bromide-lean ~1.55 eV absorbers; lattice compliance as a descriptor governing photothermal stability), the author list and KAUST affiliation, and the headline device figures (36.2 mW/cm² at albedo 0.20; 90% retention after 1,000+ hours ISOS-L-2). The full text is paywalled; this article works from the abstract and article metadata, the openly available supplementary information, and the summaries below, and does not claim any figure that appears only in the full text.
  • Supplementary Information to Xu et al., Nature Energy (2026) — openly available; the source for the albedo-equivalent test definition (rear irradiance = albedo × front irradiance; 0.20 as PVsyst-default baseline, 0.25 as bright-ground scenario), the current-matching evolution (Jsc 15.2 mA/cm² and PGD 23.8 mW/cm² at albedo 0; 23.9 mA/cm² and 36.2 mW/cm² at 0.20; 24.7 mA/cm² and 37.3 mW/cm² at 0.25), the four-composition compliance ranking and DFT two-mode stiffness values (246.9, 255.4, 312.4, 331.8 GPa), the methylammonium-chloride experiments, the single-junction certified 26.45% figure, the ISOS-L-2 and ISOS-T-2 single-junction ageing conditions and initial efficiencies, and the outdoor stability data at KAUST (July 4–August 10, 2025; 89% vs 79% retention).
  • scienmag.com (October 2026), “Softer perovskite lattices yield stable bifacial tandem solar cells” — summary corroborating the 36.2 mW/cm² power-generation density at 0.20 albedo, the 90% retention after 1,000+ hours under ISOS-L-2, the ~1.55 eV absorber bandgap, the mechanism (microstrain suppression, deep-level defects shifting shallower, reduced band-edge disorder), and the author leadership.
  • nanoparalleldaily.com (October 2026), “Perovskite Lattice Compliance Boosts Tandem Solar Cell Stability” — independent summary corroborating the KAUST affiliation, the 36.2 mW/cm² figure, and the 90% retention after 1,000+ hours at 65 °C under one-sun illumination.
  • De Bastiani, M., et al. “Efficient bifacial monolithic perovskite/silicon tandem solar cells via bandgap engineering.” Nature Energy 6, 167–175 (2021) — the earlier KAUST bifacial tandem work cited in the paper’s references.
  • Ugur, E., et al. “Front-contact passivation through 2D/3D perovskite heterojunctions enables efficient bifacial perovskite/silicon tandem solar cells.” Matter 6, 2919–2934 (2023) — the prior bifacial benchmark from the same group, power-generation densities above 27 mW/cm².
  • Khenkin, M. V., et al. “Consensus statement for stability assessment and reporting for perovskite photovoltaics based on ISOS procedures.” Nature Energy 5, 35–49 (2020) — the source for the ISOS-L-2 and ISOS-T-2 protocol definitions used in this article.
  • Excitonic.com, “Twist the Molecule, Not the Process: Conjugation Control Solves the Tandem SAM Trade-Off” (November 10, 2026) — the paired tandem story: the Soochow–LONGi self-assembled monolayer and its certified stabilized 35.13% monofacial tandem, the front-only efficiency figure this bifacial result is deliberately not compared against.

A note on sourcing: the full text of the Xu paper sits behind a paywall; this article works from the paper’s abstract and metadata as published by Nature, the openly available supplementary information, and two independent summaries that corroborate the headline figures. No headline efficiency figure for the tandem is quoted, because none is supplied in the material consulted — the paper reports power-generation density at stated albedo conditions, and this article follows that convention. No damp-heat (85 °C/85% relative humidity) data, module-area result, or large-area uniformity figure is stated, because none appears in the sources consulted. Background on halide segregation, ISOS stability protocols, and perovskite–silicon tandem architecture is drawn from the standard perovskite literature. No market-size figure is quoted, because none is needed for this materials result.