Perovskite–silicon tandems have a coverage problem that sits exactly where the two halves of the device meet. The silicon bottom cell is textured with microscopic pyramids that trap light, but the hole-selective layer that must sit on those pyramids is a single layer of molecules, one molecule thick, and it has to coat every facet, ridge, and valley evenly. A paper published November 4 in Nature reports a molecular fix — and it is a fix of a kind this desk likes: it changes the molecule, not the process. A team led by Soochow University with LONGi’s central R&D institute designed a family of self-assembled molecules, TC0 through TC3, all derivatives of the standard carbazole-phosphonic-acid molecule 4PACz, in which a tailored twist suppresses the clumping that has ruined conformal coverage without sacrificing the charge transport that conjugation provides. Their best molecule, TC2, carried a two-terminal monolithic tandem to a champion 35.31%, with a certified stabilized value of 35.13%.

The paper, “Conjugation-regulated self-assembled molecules for tandem solar cells,” comes from Zheng Fang, Zhijie Rao, Chongtao Liu, Lei Ding, and Anqi Lv as equal first authors, led by Jiang Liu and Xiaohong Zhang of Soochow University and the Suzhou Laboratory, with groups at Northwestern Polytechnical University, Hong Kong Polytechnic University, and LONGi Central R&D Institute, among others. It was received April 21, 2026, accepted September 29, and published November 4; five authors are LONGi employees, disclosed as a competing interest.

The trade-off: conjugation versus coverage

In a monolithic perovskite–silicon tandem, the perovskite top cell is deposited directly onto the silicon bottom cell, and the junction needs a hole-selective contact: a layer that pulls holes out of the perovskite and passes them toward the silicon while blocking electrons. Self-assembled monolayers, or SAMs, have become the standard solution — a phosphonic-acid anchoring group bonds chemically to the transparent conductive oxide on the silicon, and a carbazole head group sets the energy level for hole extraction. The layer is a single molecule thick, deposited from solution, with almost no parasitic light absorption — a combination that has carried tandems past 33% and, in LONGi’s hands, to a 35.5% record certified by the European Solar Test Installation in July 2026.

But SAM design on textured silicon has carried a built-in contradiction, and the Nature paper states it plainly in its opening: increasing molecular conjugation enhances conductivity but often leads to unfavourable aggregation that impairs conformal coverage.

Conjugation is the sharing of electrons across a flat, connected network of chemical bonds. A more conjugated molecule conducts charge better, because its electrons are delocalized and mobile. But flat, conjugated molecules also like to stack: the same planar surfaces that let electrons move along a molecule let neighbouring molecules pile onto each other through π–π stacking, like plates in a cupboard. On a flat substrate, some stacking is tolerable. On a pyramid-textured surface, it is fatal to uniformity: aggregates form, the monolayer goes down patchy — thick in the valleys, bare on the ridges — and wherever coverage fails, the perovskite meets the oxide directly, creating recombination sites that eat voltage and fill factor. The field has been stuck choosing between molecules conductive enough to extract holes efficiently and molecules disordered enough to coat the pyramids evenly. The usual workarounds — diluting the SAM, mixing molecules, adding co-adsorbents — treat the symptom by complicating the process.

The new paper treats the cause by editing the molecule’s geometry. The authors call the approach rationally engineered molecular spatial conjugation: keep the conjugated backbone that carries charge, but build a tailored intramolecular distortion into the structure — a twist that breaks the flatness just enough to suppress excess π–π stacking, while leaving the through-molecule conduction path intact. The molecule is conjugated along its length and twisted across its width, so it conducts like a flat molecule and packs like a crooked one.

The design: four molecules, one sweet spot

The team synthesized four SAM derivatives, TC0 through TC3, all built on the 4PACz scaffold, differing in the degree and placement of the twist — letting them map the trade-off rather than merely assert it. TC2 emerged as the optimum: its structure combines a conjugated backbone with a moderately twisted phenyl side group, in the words of LONGi researcher Bo He, speaking to pv magazine. Too little twist, and the molecules still aggregate; too much, and the packing and adhesion suffer. TC2 sits in the middle, and the numbers show it.

Molecular simulations put TC2’s surface coverage at 83.1%, against 75.0% for TC1 and 73.3% for TC3. Experimental measurements confirmed more than 80% coverage with strong anchoring stability — the molecule not only lands evenly but stays bonded. Eighty-three percent is not complete coverage, and the paper’s framing is that the trade-off has been managed, not eliminated. But on a surface of micron-scale pyramids, where the incumbent molecule leaves roughly a quarter of the surface effectively bare, an eight-point gain — achieved without thinning the layer or adding a process step — is the difference between a leaky interface and a working one.

The improved arrangement changes the electronics of the interface as well as its geometry. TC2 modified the work function of the indium tin oxide (ITO) recombination layer and provided favourable energy-level alignment for hole extraction, strengthening the internal electric field, while the better coverage reduced the defect population at the buried interface and suppressed recombination. Two effects, one molecule: the twist fixes the packing, and the fixed packing fixes both the energetics and the defect density.

There is a third effect, one layer up. TC2 promoted slower, more controlled perovskite crystallization on top of the SAM, producing larger, more uniform grains with higher crystallinity and reduced nonradiative recombination. Photoluminescence measurements put numbers on it — a quantum yield of 1.46% and a quasi-Fermi level splitting of 1.292 eV, the highest values among the SAMs tested. The quasi-Fermi level splitting is the ceiling on the voltage the device can produce, so the highest value among the candidates is a direct preview of the voltage result that follows.

The device numbers, handled in order

The devices are two-terminal monolithic tandems: a perovskite top cell on a heterojunction silicon bottom cell. The stack, as described in trade coverage, runs from a textured silicon bottom cell coated with an ITO recombination layer, then the TC2-based SAM as the hole-selective contact, then the perovskite absorber, a lithium fluoride interlayer, a fullerene (C₆₀) electron-transport layer, and a tin oxide buffer, completed with a transparent indium zinc oxide electrode, a silver contact, and a magnesium fluoride coating. Nothing in that stack is exotic except the SAM itself — the rest is the standard tandem architecture, which is precisely the point. The new molecule drops into the existing process; it does not demand a new one.

Under standard test conditions, TC2-based devices averaged an open-circuit voltage of 2.007 V and a fill factor of 84.30%, outperforming cells based on 4PACz, TC1, and TC3. The champion device reached 35.31% power conversion efficiency, with a short-circuit current density of 20.82 mA/cm², an open-circuit voltage of 2.007 V, and a fill factor of 84.53%.

Three efficiency figures circulate for this device, and they are different measurements, not contradictions. The 35.31% is the team’s own champion measurement. Independent testing by China’s National Center of Inspection on Solar Photovoltaic Products Quality (CPVT) certified a reverse-scan efficiency of 35.16% and a stabilized efficiency of 35.13%; trade coverage also reports a forward-scan certified value of 35.25%. The stabilized figure is the one to quote: it is measured by holding the device at its maximum power point until the output settles, rather than sweeping the voltage quickly past it, and it is the hardest of the three to flatter.

The certified cell had an active area of 1 cm². That is the first honesty constraint on the headline: this is a small laboratory cell, not a module, and every number above belongs to that scale. It is also worth stating plainly that this is not a record. LONGi’s own July 2026 tandem record stands at 35.5%, certified by the European Solar Test Installation. The TC2 cell’s certified stabilized 35.13% sits just below it. What it does, that a record chase does not, is demonstrate the number on a textured substrate through a molecular design rule rather than through device optimization alone.

The stability results are the second half of the paper’s claim, and they are reported on encapsulated devices. Tandem cells using TC2 retained 94% of their initial efficiency after 1,500 hours of continuous operation under the ISOS-L-1 protocol — indoor light soaking at room temperature, tracked at maximum power — and 82% after 1,000 hours at 65 °C under the ISOS-L-2 protocol. The paper also reports improved resistance to light-induced degradation and phase segregation in TC2-based perovskite films, consistent with the more uniform crystallization described above. Both protocols are laboratory stress tests, not outdoor exposure, and the ISOS-L-2 retention of 82% is the weaker of the two numbers — a reminder that heat remains the harder axis for this stack. But 94% retention after 1,500 hours is a serious stability result for a tandem whose headline is an interface molecule, because interface molecules are exactly the component that usually fails first.

Why a molecule beats a process change

The field has attacked SAM coverage on textured silicon from several directions in the past year — a symmetry-breaking molecular strategy, a chemical-polishing route to a certified 31.71% tandem, and a dual-anchored zirconium oxide interface that reached 34.0%. Each adds something to the process: a second molecule, a polishing step, a nanoparticle interlayer. The TC2 result adds nothing to the process. The SAM is still deposited the same way; only the molecule is different. That is the manufacturing argument. A process addition has to be qualified, tooled, and costed at every step of a production line; a molecule substitution has to be synthesized at scale and slotted into the existing deposition step. The latter is not free — SAM synthesis at volume, purity control, and solution stability are real industrial questions — but it is a shorter road than re-engineering the line.

There is also a design-rule argument that travels further than this one molecule. The paper demonstrates a tunable parameter — intramolecular distortion — that trades against π–π stacking in a controlled way, with TC0 through TC3 mapping the curve and TC2 marking the optimum. That is a rule other groups can apply to other SAM scaffolds, other anchor chemistries, and other textured substrates — and the field’s central manufacturing question for tandems is now exactly this: how to coat a rough surface with a single molecular layer, uniformly, at scale.

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

First, it is not a record: 35.13% certified stabilized sits below LONGi’s own 35.5% July record. The value is the design rule and the textured-substrate coverage, not the leaderboard position.

Second, the demonstration is a 1 cm² laboratory cell. Nothing in the material consulted addresses large-area uniformity, coating throughput, or yield — the quantities that decide whether a molecule survives translation from a laboratory spin-coater to a production line.

Third, the stability evidence covers two laboratory stress axes — continuous illumination at room temperature and at 65 °C — and not damp heat at 85 °C and 85% relative humidity, thermal cycling, or outdoor exposure. The 82% retention after 1,000 hours at 65 °C is respectable but not yet commercial-grade durability, and moisture-driven degradation, historically perovskites’ hardest failure mode, is not addressed in the material consulted.

Fourth, the mechanism evidence available here is second-hand. The full text of the Nature paper sits behind a paywall; this article works from the paper’s abstract and metadata as published by Nature, supplemented by pv magazine’s trade coverage and two independent summaries that corroborate the headline figures. One figure deserves a flag: trade coverage of the paper’s supplementary data reports that reference 4PACz devices averaged 24.34% across ten cells against 35.11% for TC2 — a strikingly large gap that, if it reflects the textured-substrate penalty on the incumbent molecule, would make the coverage story even more dramatic. This desk has not inspected the supplementary data directly, so that comparison is reported here as trade coverage’s reading of it, not as a verified figure.

What to watch

The most informative follow-up would be a second scaffold. If intramolecular distortion is genuinely a design rule rather than a TC2-specific trick, then a twisted derivative of a different SAM family should show the same pattern: an intermediate twist outperforming both the flat and the over-twisted variants. A second molecule family would convert a demonstration into a rule. A damp-heat result would convert the stability question from a laboratory claim into a commercial one, and scale-up data on a larger-area textured substrate would show whether a laboratory SAM deposition becomes a production step.

The bottom line

A Soochow–LONGi team has shown that the SAM trade-off on textured silicon — conjugation for conductivity versus aggregation against coverage — can be broken inside the molecule itself. A moderately twisted phenyl side group on a 4PACz-derived scaffold, the molecule called TC2, suppresses excess π–π stacking without giving up charge transport, lifting simulated surface coverage to 83.1% and measured coverage above 80% on pyramid-textured silicon. The resulting tandem reaches 35.31% in the laboratory and a certified stabilized 35.13% on a 1 cm² cell, with 94% of initial efficiency retained after 1,500 hours.

It is not a record, not a module, and not a weather test. It is a molecular design rule for the hardest coating problem in tandem manufacturing, demonstrated by a team that includes the manufacturer who would have to use it. The number to watch is not 35.13%. It is the second scaffold: the first twisted derivative of a different SAM family that shows the same optimum in the middle, proving that conjugation control is a rule and not a single molecule.

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

Sources and further reading

  • Fang, Z., Rao, Z., Liu, C., Ding, L., Lv, A., et al. “Conjugation-regulated self-assembled molecules for tandem solar cells.” Nature (2026), published November 4, 2026. DOI: 10.1038/s41586-026-11226-8 — the primary source for the paper’s framing (the SAM trade-off on rough textured substrates; molecular spatial conjugation with tailored intramolecular distortion suppressing excess π–π stacking without sacrificing charge transport), the author list and affiliations (Soochow University, Northwestern Polytechnical University, Hong Kong Polytechnic University, LONGi Central R&D Institute, KAUST, Chinese University of Hong Kong), the publication history (received April 21, 2026; accepted September 29, 2026), and the competing-interest disclosure (five LONGi employees). The full text is paywalled; this article works from the abstract and article metadata and does not claim any figure that appears only in the full text or supplementary information.
  • pv magazine Global (November 6, 2026), “Longi achieves 35.31% efficiency in perovskite-silicon tandem solar cell based on self-assembled molecules” — trade coverage of the same result, the source for the TC0–TC3 molecule family and the TC2 structure (conjugated backbone with a moderately twisted phenyl side group), the coverage figures (83.1% simulated for TC2 vs. 75.0% TC1 and 73.3% TC3; >80% measured), the device figures (champion 35.31%, 20.82 mA/cm², 2.007 V, 84.53% fill factor; averages 2.007 V and 84.30% fill factor), the CPVT certification (reverse-scan 35.25%, stabilized 35.13%), the stack description (textured silicon HJT bottom cell, ITO recombination layer, TC2 SAM, perovskite, LiF, C₆₀, SnOₓ, IZO, Ag, MgFₓ), the photoluminescence figures (1.46% quantum yield, 1.292 eV quasi-Fermi level splitting), the stability results (94% after 1,500 hours ISOS-L-1; 82% after 1,000 hours at 65 °C ISOS-L-2), the Bo He quotation, and the context on prior SAM work and LONGi’s July 2026 35.5% record (ESTI-certified).
  • CiVL (November 5, 2026), “Stacked solar cell turns more than a third of sunlight into electricity” — independent summary corroborating the certified stabilized 35.13%, the 1 cm² active area, the 94% retention after 1,500 hours, and the author leadership (Jiang Liu, Xiaohong Zhang; corresponding authors Long Jiang, Bo He, Jun Yin, Kai Wang). It reports the certified reverse-scan value as 35.16% with an 84.72% fill factor, against pv magazine’s 35.25%; this article quotes the stabilized 35.13% as the headline certified figure and notes the scan-value discrepancy rather than resolving it.
  • list.solar (November 6, 2026), “Perovskite-Silicon Tandem Cell: 35.1% Stabilized in Nature” — trade summary corroborating the 35.13% certified stabilized value, the 1.007 cm² aperture, and the forward/reverse scan split (35.25% forward, 35.16% reverse), and the source for the supplementary-data comparison of reference 4PACz devices (24.34% average across ten cells) against TC2 (35.11%), reported here as that outlet’s reading of the supplementary data, not as a directly verified figure.

A note on sourcing: the molecular structures of TC0–TC3, the detailed coverage measurements, and the control-device comparisons sit in the paper’s full text and supplementary information, which this desk has not inspected. 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 4PACz chemistry, SAM hole-selective contacts, 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 process result.