OHMEGA/MONSTER

Researchers report vacuum-made perovskite-on-silicon solar cells hitting 30% at wafer size

A formamidinium eutectic lowers evaporation temperature by about 36C to enable 31.5% at 1 sq cm and 30.0% on a half-cut G12 wafer with unusually low scaling loss, but the efficiencies lack independent certification and the durability data so far do not prove a 20-year lifetime.

What happenedA Nature paper published 5 Aug 2026 (Luo, He, Ran et al.) reports a formamidinium eutectic that lowers FAI evaporation temperature by about 36C, enabling thermally evaporated perovskite-on-silicon tandems to reach 31.5% at 1 sq cm and 30.0% at 200 sq cm on a half-cut G12 wafer.

Why it mattersVacuum evaporation is solvent-free and compatible with existing factory lines, and unusually low scaling loss at commercial wafer size is a necessary step toward manufacturable tandems that can beat single-silicon efficiency limits.

Still openThe 31.5% and 30.0% figures are not independently certified and the 95% retention after 2000 hours damp-heat plus two months outdoors has not yet been shown to translate to a full IEC 61215 sequence pass or 20-year outdoor lifetime.

a silicon wafer inside a vacuum deposition chamber
AI-generated illustration — not a photograph of this story

A paper in Nature published 5 August 2026 by Luo, He, Ran and colleagues reports a formamidinium-based eutectic that lets a fragile perovskite ingredient evaporate about 36C cooler than its breakdown point, enabling a two-layer perovskite-on-silicon solar cell �� a stack that puts a thin perovskite crystal on top of conventional silicon to capture more of the solar spectrum �� to reach 31.5% steady-state efficiency at 1 cm2 and 30.0% at 200 cm2 on a commercial half-cut G12 wafer. The result is notable because it is the first thermally evaporated tandem at that wafer size to hit 30% with only a 3.99% loss on scaling, but the efficiencies are not independently certified and the durability data so far �� 95% retention after 2000 hours in damp-heat and negligible loss after two months outdoors �� do not yet amount to a 20-year warranty.

Why stacking matters

A single silicon cell is capped by physics at about 33.7% efficiency. Stacking a wide-bandgap perovskite on silicon lifts the theoretical ceiling to about 43% by splitting the spectrum: the perovskite takes high-energy photons, silicon takes the rest. That headroom is why tandems are pursued, and why 30% at commercial size matters more than a lab record alone. The current certified lab record is 35.5%, set by the Chinese manufacturer LONGi Green Energy and certified by the European Solar Test Installation (ESTI) in July 2026, improving on 34.85% certified by the US National Renewable Energy Laboratory (NREL) in April 2025. Those records are small-area, solution-processed devices. LONGi also reported 33% at 260.9 cm2 in June 2025, and a recent solution-processed tandem with homogeneous passivation reached 28.9% at 60 cm2 after exceeding 31% at 1 cm2 �� illustrating how much is typically lost when going large.

Why factories want vacuum

Solution processing �� spin-coating, blade- or slot-die coating �� dominates records because it easily forms high-quality films at low temperature. At module scale it brings solvent handling, drying dynamics and coating non-uniformity that worsen on the micron-scale textured silicon used to trap light.

Thermal evaporation �� heating source materials in vacuum to deposit controlled vapors �� is solvent-free, offers atomic-scale thickness and composition control, and is already scaled in OLED (organic light-emitting diode) displays and CIGS (copper indium gallium selenide) thin-film solar manufacturing. It is topography-tolerant and compatible with existing vacuum lines. That is the basis for the Nature paper's description of evaporation as "more industrially viable." A review in Joule cautions that evaporation brings different scale-up constraints �� deposition rate, source stability, lateral flux uniformity, material utilization and equipment capex/throughput �� and should be seen as complementary to solution processing, not a universal replacement.

The 36-degree fix

Formamidinium-based perovskites, made from formamidinium iodide (FAI), are preferred for tandems for their near-ideal bandgap around 1.5 eV (electron volts, a measure of photon energy), but FAI is thermally fragile: it must be heated to evaporate yet decomposes if too hot, wrecking stoichiometry and crystallinity.

The authors synthesize a formamidinium-based eutectic, or Eu �� a mixture where components interact to depress the melting/evaporation point below that of any component alone �� that lowers FAI's effective evaporation temperature by an average of 36C below its degradation threshold. That window enables stable vapor flux without breakdown, yielding evaporated perovskite films with enhanced crystallinity and atomic-scale compositional homogeneity. The accessible abstract and reference PDF do not disclose the eutectic's co-former, molar ratio, phase diagram, absolute evaporation or degradation temperatures, or detailed thermal-analysis (DSC/TGA, which track melting and weight loss on heating) and surface-composition mapping (ToF-SIMS) data; those are cited to Supplementary Notes 1-19, which remain paywalled. The mechanism is analogous to deep-eutectic melting-point depression, but beyond the 36C figure the chemistry is not verifiable from the open text.

Sequential evaporation �� likely lead iodide followed by FAI/Eu �� then builds the perovskite top cell on the silicon bottom cell.

30% at wafer size �� and the asterisk

The sequentially evaporated tandems achieve 31.5% steady-state efficiency at 1 cm2 �� steady-state meaning power held at the maximum power point over time, not a brief flash �� and 30.0% at 200 cm2 on a half-cut G12 wafer. A G12 wafer is the industry-standard 210 mm silicon wafer; a half-cut is about 200 cm2, the commercial module building block, versus 1 cm2 lab cells.

Scaling from 1 to 200 cm2 incurs only 3.99% relative loss, claimed as the lowest scaling penalty reported for perovskite-based tandems. For context, the 60 cm2 solution device fell about 7% from >31% to 28.9%, and LONGi's 33% at 260.9 cm2 versus 35.5% at small area is also about 7% loss. If the figure holds under independent certification, it suggests evaporation's uniformity translates to better area scaling, where defects and series resistance usually dominate.

The caveats are load-bearing. The open Nature text discloses no certifying lab �� NREL, ESTI or Fraunhofer ISE CalLab �� no aperture versus active area, no hysteresis or MPP tracking duration, and no independent verification. The 3.99% is a claimed lowest, not an independently verified ranking, and 200 cm2 is a half-cut G12, not a full 400 cm2 wafer or module. Direct comparison to ~1 cm2 records without noting area mismatch misreads the advance: the significance is low loss at commercial size, not beating 35.5% outright.

Double the damp-heat test is not a 20-year warranty

The Eu-based tandem retains 95% of its initial efficiency after 2000 hours of damp-heat aging at 85C and 85% relative humidity, and shows negligible loss after two months of real-world outdoor operation.

International qualification standard IEC 61215 requires damp-heat at 85C/85% RH for 1000 hours and thermal cycling between -40C and 85C for 200 cycles, with a pass criterion of �25% power loss. The 95% retention after 2000 hours therefore meets the 5% threshold at double the required duration, matching the best reported encapsulated single-junction perovskites �� a Nature Energy 2025 self-assembled bilayer device showed <4% loss after 2000 hours damp-heat and 3% loss after 1200 thermal cycles.

Damp-heat alone does not prove lifetime. The ISOS consensus for perovskite stability defines five stress categories �� dark storage (ISOS-D), outdoor (ISOS-O), light soaking (ISOS-L), thermal cycling (ISOS-T) and light-humidity-thermal cycling (ISOS-LT) �� plus extensions for light-dark cycling (ISOS-LC) and electrical bias (ISOS-V), noting that light soaking drives ion and defect migration and phase segregation not captured by 85C/85% RH. A Communications Materials review finds perovskites have passed parts of IEC 61215 but often not in sequence �� UV precondition followed by thermal cycling followed by humidity-freeze �� or not in air, and that artificial aging is not equivalent to real operation.

Two months outdoors is short against those benchmarks. An all-vacuum tandem from the Hong Kong University of Science and Technology (HKUST) team reported in Nature Materials retained about 80% of temperature-corrected performance ratio after eight months outdoors in Bolzano, Italy, laminated and measured against a silicon heterojunction reference, with negligible loss only in the first three months. Silicon warranties are 25 years; Oxford PV, which shipped first commercial 72-cell tandem panels in 2024, says 20-year lifetime by 2027 is critical for levelized cost of energy, targeting 25% efficiency with 15-year life in 2026 and 30% with 30-year life by 2030. The Nature paper's outdoor location, climate, mounting, encapsulation type and performance-ratio methodology are not detailed in the accessible text.

Vacuum vs vacuum: where this lead still has gaps

On efficiency and area, the Eu route leads competing vapor approaches. Close-space sublimation (CSS) reported in Nature Energy in May 2026 achieved 18.5% for a fully vacuum-processed single-junction and 24.3% for a monolithic perovskite/silicon tandem on micro-textured silicon at about 1 cm2, with 10-minute conversion, an effective rate around 47 nm per minute �� about ten times conventional point-source co-evaporation of a few nm per minute for FA-based compositions �� a reusable mixed-halide organic source stable for more than 28 depositions and a 96 cm2 large-area source demonstration. The HKUST all-vacuum co-evaporation with 5 mol% PbCl2 co-source achieved certified 18.35% at 0.25 cm2 and 27.2% tandem with a solution-processed SAM (self-assembled monolayer, a single-molecule contact layer) hole layer (24.3% all-vacuum stack), retaining 80% of peak power after 1,080 hours under ISOS-L-2 at 75C and about 80% after eight months outdoors.

The Eu tandem exceeds those on efficiency and demonstrates the first thermally evaporated large-area at G12, but publishes no throughput, material utilization or lateral flux uniformity at G12, no cycle time for sequential evaporation, and no long ISOS-L/LC/V or thermal-cycling data. A meta-analysis in Advanced Energy Materials finds vacuum evaporation reaches cost parity with solution only by raising evaporation rate to about 720 nm per minute and number of linear sources to about 10 �� industrially achievable but not yet demonstrated at scale. The Joule review and CSS paper both note that lateral flux uniformity and utilization remain unquantified for the Eu route at G12, while CSS demonstrated conformal coverage on planar, nano- and micro-textured silicon without re-tuning and low precursor consumption per cycle.

Silicon dominates tandem cost, so the vacuum-versus-solution decision hinges on yield, throughput and use of existing vacuum lines rather than material cost alone. No LONGi or Oxford PV public commentary endorses thermal evaporation as lower-cost or higher-yield than solution at gigawatt scale; LONGi's 35.5% devices remain solution-processed with no mass-production plan disclosed.

What would change the picture is independent certification of 31.5% and 30.0% with aperture area and MPP protocol stated, full IEC 61215 sequence data including thermal cycling, humidity-freeze and UV, longer multi-climate outdoor testing under ISOS-O with encapsulation disclosed, and published deposition rate, utilization and uniformity at full G12 in an in-line process. Until then, the eutectic clears a genuine materials bottleneck for the fab-friendly vacuum route and shows unusually low scaling loss at wafer size �� a necessary step toward manufacturability, not yet proof of a bankable module.

Source recordSources / claims / limits

How this piece is framed: Bottleneck cleared at wafer scale — but bankability not yet proven: a materials fix (eutectic) unlocks the fab-friendly vacuum route to 30% at commercial size with record-low scaling loss, now facing the harder tests of certification, throughput, and 20-year lifetime.

Sources

Claims, and how far we tracked each down

  • [confirmed] A Nature article titled 'Thermally evaporated perovskite/silicon tandems via formamidinium eutectic' by Luo, He, Ran et al. was published 5 Aug 2026 with DOI 10.1038/s41586-026-10970-1 · read in full (as of 2026-08-06)
  • [confirmed] Solution processing remains the dominant route to high-performance perovskite/silicon tandems but struggles to simultaneously achieve industrial scalability and long-term reliability · read in full (as of 2026-08-06)
  • [confirmed] Thermal evaporation is more industrially viable than solution processing for perovskite/Si tandems but had not been successfully demonstrated for large-area devices before this work, largely due to thermal degradation of formamidinium iodide (FAI) during high-temperature evaporation · read in full (as of 2026-08-06)
  • [confirmed] The authors synthesized a formamidinium-based eutectic (Eu) that lowers the effective evaporation temperature of FAI by an average of 366C below its degradation threshold, enabling stable FAI evaporation without thermal degradation · read in full (as of 2026-08-06)
  • [confirmed] Evaporated perovskite films made with the Eu exhibit enhanced crystallinity and atomic-scale compositional homogeneity · read in full (as of 2026-08-06)
  • [confirmed] Sequentially evaporated perovskite/Si tandems achieved a steady-state efficiency of 31.5% at 1 cm8 · read in full (as of 2026-08-06)
  • [confirmed] The team demonstrated the first thermally evaporated large-area perovskite/Si tandem on a commercial half-cut G12 wafer, delivering 30.0% steady-state efficiency at 200 cm8 · read in full (as of 2026-08-06)
  • [confirmed] Scaling from 1 cm8 to 200 cm8 incurred only 3.99% relative efficiency loss, representing the lowest reported efficiency penalty for area scaling in perovskite-based tandems · read in full (as of 2026-08-06)
  • [confirmed] The Eu-based tandem retains 95% of its initial efficiency after 2000 hours of damp-heat aging at 856C and 85% relative humidity · read in full (as of 2026-08-06)
  • [confirmed] The Eu-based tandem exhibits negligible power loss after two months of real-world outdoor operation · read in full (as of 2026-08-06)
  • [confirmed] The current certified laboratory record for perovskite/silicon tandem efficiency is 35.5% (LONGi, certified by ESTI, July 2026), improving on 34.85% (NREL, April 2025), 34.6% (ESTI, June 2024) and 33.9% (2023) · read in full (as of 2026-08-06)
  • [confirmed] LONGi reported a 33% efficiency for a 260.9 cm8 two-terminal perovskite-silicon tandem cell certified by NREL in June 2025, illustrating the gap between small-cell records and larger-area devices · read in full (as of 2026-08-06)
  • [confirmed] A recent solution-processed perovskite-silicon tandem achieved certified 28.9% at 60 cm8 and >31% at 1 cm8 via homogeneous passivation with phosphonic acids and piperazinium chloride · read in full (as of 2026-08-06)
  • [confirmed] The theoretical efficiency limit for perovskite/silicon tandems is ~43%, far exceeding the Shockley-Queisser limit of ~33.7% for single-junction cells · read in full (as of 2026-08-06)
  • [confirmed] IEC 61215 requires damp-heat testing at 856C/85% RH for at least 1000 hours and thermal cycling for at least 200 cycles between -406C and 856C for design qualification of terrestrial PV modules · read in full (as of 2026-08-06)
  • [confirmed] Other perovskite devices have demonstrated <4% loss after 2000h damp-heat (856C/85% RH) and 3% loss after 1200 thermal cycles, showing that 95% retention after 2000h is at the leading edge but not unprecedented for encapsulated single-junction PSCs · read in full (as of 2026-08-06)
  • [confirmed] All-vacuum-deposited perovskite-on-silicon tandems have separately achieved 27.2% at 1 cm8 and retained ~80% of initial performance after 8 months outdoor operation in Italy, providing a benchmark for evaporated tandem stability · read in full (as of 2026-08-06)
  • [likely] G12 wafers are 210 mm format; a half-cut G12 device at 200 cm8 corresponds to commercial wafer scale, versus laboratory 1 cm8 cells · read in full (as of 2026-08-06)
  • [confirmed] Thermal evaporation offers solvent-free, vacuum-based deposition with excellent thickness uniformity and precise compositional control, compatible with textured silicon and existing semiconductor/display manufacturing infrastructure · read in full (as of 2026-08-06)
  • [confirmed] The specific chemical composition of the formamidinium eutectic, its melting-point depression mechanism, and the certification laboratory for the 31.5% and 30.0% steady-state efficiencies were not disclosed in the accessible abstract · read in full (as of 2026-08-06)
  • [confirmed] IEC 61215:2021 requires damp-heat MQT 13 at 85C/85% RH for 1000h and thermal cycling MQT 11 between -40C and 85C for 200 cycles, with pass criterion ≤5% power loss (industry benchmark). · read in full (as of 2026-08-06)
  • [confirmed] Nature Energy 2025 self-assembled bilayer (SAB) inverted PSCs: champion <4% loss after 2000h damp-heat (85C/85% RH) and 3% loss after 1200 thermal cycles (-40 to 85C); average 94% retention after 2000h (n=7, avg 24.3% PCE) and 94% after 1288 cycles, meeting IEC 61215:2021 at 2x duration and >6x cycles. · read in full (as of 2026-08-06)
  • [confirmed] HKUST Nature Materials 2026 all-vacuum tandems: 27.2% at 1cm2 with solution-processed SAM HTL (24.3% all-vacuum stack), laminated devices tested outdoors in Bolzano, Italy retained ~80% of initial temperature-corrected performance ratio (PRtc) after 8 months; negligible PRtc loss first 3 months vs HJT Si reference. · read in full (as of 2026-08-06)
  • [confirmed] Luo et al. Eu tandem 95% retention after 2000h damp-heat (85C/85% RH) = 5% loss at double IEC 61215:2021 1000h duration, thus meets 5% threshold at extended stress and matches best reported encapsulated PSCs (<4% loss after 2000h, Nature Energy SAB). · read in full (as of 2026-08-06)
  • [confirmed] Accessible Nature abstract/reference PDF confirms verbatim: 31.5% steady-state at 1cm2, 30.0% at 200cm2 on half-cut G12, 3.99% relative scaling loss claimed as lowest reported, and 'first thermally evaporated large-area' claim; no certifying lab (NREL/ESTI/Fraunhofer), aperture vs active area, hysteresis, or MPP tracking duration disclosed; SI Notes 1-19 paywalled. · read in full (as of 2026-08-06)
  • [confirmed] Eu lowers FAI effective evaporation temperature by average 36C below degradation threshold enabling stable evaporation, yielding enhanced crystallinity and atomic-scale homogeneity; co-former identity, molar ratio, phase diagram, absolute T_evap/T_degradation, DSC/TGA, vapor pressure, ToF-SIMS residue not in accessible text, cited to SI Notes 1-19 (paywalled). · read in full (as of 2026-08-06)
  • [confirmed] Joule review (ns3) explicitly frames thermal evaporation scale-up constraints as deposition-rate limitations, precursor/source stability, lateral flux uniformity, material utilization, and equipment capex/throughput, stating evaporation should be evaluated as complementary to solution processing, not a universal replacement · read in full (as of 2026-08-06)
  • [confirmed] Advanced Energy Materials meta-analysis (ns7) finds vacuum evaporation's most significant limitation is high cost, mitigated only by raising evaporation rate to 20720 nm min221 and number of linear sources to 20210 to reach cost parity with solution processing, which is considered industrially achievable but not yet demonstrated at scale · read in full (as of 2026-08-06)
  • [confirmed] Conventional thermal co-evaporation effective perovskite deposition rates are typically few nm min221, particularly for FA-based compositions, because limited thermal stability of FAI fundamentally constrains maximum attainable organic deposition rate, making high-throughput deposition challenging (ns1) · read in full (as of 2026-08-06)
  • [confirmed] Close-space sublimation (CSS) Nature Energy May 2026 (ns1/ns8) achieves 18.5% PCE (stabilized 18.2%) for fully vacuum-processed p-i-n single-junction MAPb(I0.79Br0.21)3 (1.64 eV) and 24.3% for monolithic perovskite/silicon tandem on micro-textured Si (23.5% planar, 23.7% nano-textured) at ~1 cm282, with 10 min conversion at 1 mbar, effective rate ~47 nm min221 (~10217 conventional point-source), reusable mixed-halide organic source stable >28 depositions and 540 min cumulative operation, and 96 cm282 large-area source demonstrated · read in full (as of 2026-08-06)
  • [confirmed] HKUST Nature Materials all-vacuum co-evaporation with 5 mol% PbCl2 co-source (ns2) achieves certified 18.35% (19.3% lab) at 0.25 cm282 and 18.5% at 1 cm282 for 1.67 eV WBG, and tandems at 27.2% (1 cm282 with solution SAM HTL) and 24.3% all-vacuum stack, retaining 80% peak MPPT after 1,080 h ISOS-L-2 at 752115212C (1,250 h at 65212C) and ~80% temperature-corrected performance ratio after 8 months outdoor in Bolzano, Italy · read in full (as of 2026-08-06)
  • [confirmed] Eu route efficiency (31.5% at 1 cm282, 30.0% at 200 cm282) exceeds CSS (24.3% at ~1 cm282) and HKUST all-vacuum (24.3% at 1 cm282) tandems, but CSS/HKUST provide longer stability datasets and explicit throughput metrics; direct area-normalized comparison is limited because Eu 200 cm282 is half-cut G12 while CSS/HKUST are ~1 cm282 lab cells · read in full (as of 2026-08-06)
  • [confirmed] Oxford PV CEO (ns4) states current modules 25% efficiency with 15-year lifetime target in 2026, 27% with 20-year lifetime in 2027, and 30% with 30-year lifetime by 2030, calling 20-year lifetime critical for commercialization and LCOE competitiveness, implying 2-month outdoor negligible loss is insufficient as bankability proxy · read in full (as of 2026-08-06)
  • [confirmed] Oxford PV notes perovskite materials and process steps are not inherently expensive and silicon cell dominates tandem cost, but company pursues licensing to tier-one manufacturers (Trina Solar) and evaluates TOPCon vs HJT bottom cells, indicating vacuum vs solution scale-up decision hinges on yield, throughput and existing vacuum lines rather than material cost alone · read in full (as of 2026-08-06)
  • [confirmed] ISOS consensus (ns5) defines five stress categories 212 dark storage (ISOS-D), outdoor (ISOS-O), light soaking (ISOS-L), thermal cycling (ISOS-T), light-humidity-thermal cycling (ISOS-LT) 212 plus extensions light-dark cycling (ISOS-LC) and electrical bias (ISOS-V) and intrinsic testing (ISOS-I), explicitly stating damp-heat alone cannot assess full lifetime and that light soaking promotes ion/defect migration and phase segregation not captured by 85C/85% RH · read in full (as of 2026-08-06)
  • [confirmed] Communications Materials review (ns6) finds PSCs have passed parts of IEC 61215 but often not in sequence (UV precondition 212 thermal cycling 212 humidity freeze) or not in air, and that accelerated tests at diverse temperatures (RT to 85212C) are not strictly comparable; outdoor demonstration is required beyond lab damp-heat because artificial aging is not equivalent to real operation and unexpected failure modes emerge · read in full (as of 2026-08-06)
  • [likely] No LONGi or Oxford PV public commentary endorses thermal evaporation as lower-cost/higher-yield than solution at GW scale; LONGi's record 35.5% devices remain solution-processed and company has no mass-production plan disclosed, while Oxford PV's pilot line in Brandenburg uses vapor-based perovskite but targets lifetime over throughput 212 indicating industry has not converged on vacuum as definitively more industrially viable · read in full (as of 2026-08-06)
  • [confirmed] Joule review and CSS paper both note lateral flux uniformity and material utilization remain unquantified for Eu route at G12 scale; CSS demonstrates conformal coverage on planar/nano/micro-textured Si without parameter adjustment and low precursor consumption per cycle with reusable sources, highlighting that Eu route's claim of topography tolerance and uniformity at 200 cm282 lacks published flux-uniformity or utilization data · read in full (as of 2026-08-06)

Where we hit a limit / what to double-check