Perovskite solar research has a headline problem: the numbers that travel furthest belong to the architectures least suited to cheap manufacturing. The record-chasing inverted cells now sit above 27% — this desk covered a 27.4% result only yesterday. But the architecture that can actually be printed like a poster, with no hole-transport layer and no noble-metal electrode at all, has been stuck in the mid-teens, held back by a single bad interface. A paper published online October 27 in ACS Applied Materials & Interfaces takes a molecule from the record-chasers’ own playbook and applies it to that neglected interface — and the printable cell improves to 14.7%.
The result comes from Nanyang Technological University in Singapore — Qinjie Wu, Natalia Yantara, Yeow Boon Tay, Hongkang Fu, Xuechao Xing, Wei Lin Leong, and Nripan Mathews, spanning materials science, the Energy Research Institute @ NTU, and electrical engineering. The paper, “Extending Self-Assembled Monolayer Interface Engineering to Carbon-Based Perovskite Solar Cells,” was received August 21, 2026, accepted October 14, and published online October 27. It does not pretend that 14.7% competes with record cells. It shows that a known, cheap interface treatment transfers to the architecture where cheapness matters most.
The platform the tandem beat ignores
Start with the device. A carbon-based perovskite solar cell, or C-PSC, is built the way a printed sign is built. Three mesoporous layers — titanium dioxide, zirconium dioxide, and carbon — are screen-printed in sequence onto conductive glass, and the perovskite precursor is then infiltrated into the porous scaffold, crystallizing inside it. There is no separate hole-transport layer and no gold or silver electrode to evaporate: the thick carbon layer is the electrode. The whole stack is printable, uses inexpensive materials, and avoids the vacuum steps that dominate conventional cell cost.
That combination is why the architecture has long been described as one of the most plausible routes to perovskite commercialization, with stability demonstrations stretching beyond a year under continuous illumination in the earlier literature. It is also why the architecture tends to disappear from the news cycle: a carbon cell at 14 or 15% does not photograph well next to a 27% inverted cell. But the manufacturing question — what can actually be coated, dried, and shipped at scale without vacuum equipment and precious metals — is a different question from the efficiency question, and on it the carbon platform has always had the stronger hand.
The trade-off has been performance, and the performance gap has a specific address. In a conventional cell, dedicated transport layers pull one carrier out and block the other. In a carbon cell, the perovskite meets the carbon electrode directly, across a rough, porous, high-surface-area junction inside the scaffold. That junction is where the new paper operates.
The bottleneck: the carbon junction
The abstract states the bottleneck plainly: carbon-based cells “remain limited by poor interfacial charge extraction and substantial nonradiative recombination at the perovskite/carbon junction suppressing open-circuit voltage.” Both halves of that sentence deserve unpacking, because they describe two distinct losses that compound each other.
Charge extraction is the first problem. When light creates electron-hole pairs in the perovskite inside the scaffold, the holes must cross into the carbon electrode to be collected. If the energy levels at the boundary are misaligned, that transfer is sluggish: carriers linger at the interface instead of leaving, and lingering carriers recombine.
Nonradiative recombination is the second problem, and it is the one that eats voltage. The perovskite surface inside the scaffold is imperfectly terminated, and undercoordinated lead ions — Pb²⁺ sites missing a bonding partner — act as trap states. A carrier that falls into a trap recombines without emitting a photon, its energy lost as heat, dragging down the open-circuit voltage. A junction that both extracts slowly and traps readily is doubly penalized: carriers that wait around are carriers that get trapped.
The treatment: a SAM borrowed from the p–i–n playbook
The treatment is MeO-2PACz, full name [2-(3,6-dimethoxy-9H-carbazol-9-yl)ethyl]phosphonic acid: a carbazole-based molecule carrying a phosphonic acid anchoring group. In the inverted p–i–n cells that dominate the record tables, phosphonic-acid SAMs like MeO-2PACz are standard equipment: deposited as a single molecular layer on a metal-oxide contact, the phosphonic acid anchors to the surface while the carbazole head group sets the energy level for hole extraction and passivates defects beneath it. The molecule is, in effect, a molecular carpet — one molecule thick, chemically bonded at one end, electronically tuned at the other.
The new paper’s move is to take that carpet out of the flat, spin-coated world of p–i–n cells and lay it inside a mesoscopic carbon scaffold instead. The abstract is explicit that this transfer had not been done: while phosphonic-acid SAMs “are widely used in p–i–n devices, their application in mesoscopic carbon architectures remains unexplored.” The SAM is applied as a post-treatment — after scaffold and perovskite are in place — modifying the interface from within rather than being deposited as a discrete layer during stack construction.
That distinction is why this desk tracks the result in the manufacturing lane rather than the record lane. A post-treatment applied to an already-printed scaffold is compatible, in principle, with the screen-printing line that makes carbon cells attractive; a treatment requiring a new layer, a new tool, or a vacuum step would erase the platform’s advantage. Whether it slots in at production speed and uniformity is unproven, but the form factor of the fix is the right shape for the platform.
The mechanism has two parts, mapping onto the two losses above. First, the treatment “modulates the perovskite/carbon interface, improving energy-level alignment,” so holes cross into the carbon more readily. Second, it passivates “undercoordinated Pb²⁺ sites” — the phosphonic acid coordinates to electron-deficient lead centers that would otherwise act as traps. One molecule, two jobs, at the one interface where both jobs were undone.
The evidence: faster extraction, longer-lived carriers
A device curve alone would not establish that this mechanism is operating. The paper’s evidence, as reported in the abstract, rests on two transient measurements, and each addresses a different half of the claim.
Transient photocurrent analysis measures how quickly photogenerated charge leaves the device after a light pulse; faster decay means carriers are extracted more promptly rather than lingering at the interface — the signature of improved alignment. Transient photovoltage analysis holds the device at open circuit and watches how long the voltage persists after a perturbation; a longer-lived signal means carriers survive longer before recombining — the signature of suppressed trap-assisted recombination. The abstract reports both: “accelerated charge extraction and extended recombination lifetimes.” The two measurements corroborate the two halves of the mechanism rather than merely restating the efficiency gain.
Two further observations travel with the headline. Hysteresis — the difference between forward and reverse current-voltage sweeps, a classic symptom of mobile ions and interfacial trapping — is reported as suppressed. And stability, the carbon platform’s traditional strong suit, improves rather than being traded away: 98% of initial efficiency retained after 2,400 hours under controlled humidity, without encapsulation. Both fit the same story, but both carry the caveats detailed below.
The device numbers, handled in order
The numbers that can be stated from the paper’s abstract are these: a champion power conversion efficiency of 14.7%, an open-circuit voltage of 0.90 V, and a fill factor of 70.7%, with suppressed hysteresis and 98% of initial efficiency retained after 2,400 hours under controlled humidity without encapsulation.
The honesty constraints start with what is not stated. The abstract does not report a short-circuit current density, a control-device efficiency, a device area, or the humidity level of the stability test. The full text is paywalled, so this article works from the abstract and claims none of those figures. That absence matters: without a control value, the size of the improvement over untreated cells cannot be quantified from the abstract alone; without an area, the result cannot be placed on the laboratory-versus-module spectrum; and without a humidity level, “controlled humidity” cannot be compared with any standard stress protocol.
The voltage is where the mechanism shows up most legibly. A 0.90 V open-circuit voltage is modest by record-cell standards — yesterday’s Northwestern–DGIST champion reached 1.187 V — but for a carbon cell, where the junction has historically suppressed voltage through nonradiative recombination, 0.90 V alongside a 70.7% fill factor represents a junction doing its job rather than throttling the device. The fill factor, which measures how square the current-voltage curve is, is the second legible signature: a leaky, trap-dominated interface flattens the curve, and 70.7% is respectable, if unremarkable, for this architecture.
The efficiency should be placed against the right baseline. At 14.7%, the champion sits near the historical range for the classic triple-mesoscopic architecture with a zirconia spacer — earlier literature put that configuration around 15% — and well below the best carbon-cell results with modified scaffolds and additives, which have reached into the high teens and beyond 20% in some reports. The authors are not claiming a carbon-cell record. They are demonstrating that SAM interface engineering transfers to carbon cells and moves the numbers in the expected direction through the expected mechanism. The value is the transfer itself: it opens the entire SAM library — molecules already optimized, commercialized, and understood — to an architecture that had not used them.
What this is not: a record, a module, or a weather test
The limits of the result are as instructive as the result, and they should travel with the headline wherever it goes.
First, 14.7% is far below record perovskite cells, and no one should quote it in the same breath as 27.4% without the architecture attached. The comparison that matters is carbon cell to carbon cell. Judged as a process fix for the cheapest architecture in the field, it is a meaningful step; judged against the record tables, it loses by a wide margin.
Second, the stability figure is a controlled-storage result, not a weather test. “Retaining 98% of initial efficiency after 2,400 hours under controlled humidity without encapsulation” describes laboratory storage — the humidity level is not stated in the abstract — not damp-heat stress at 85 °C and 85% relative humidity, not thermal cycling, and not outdoor exposure. The unencapsulated qualifier is genuinely encouraging, since carbon’s hydrophobicity has always been part of the platform’s durability story, but a storage test answers a different question from a stress test. The stability claim should be quoted with “controlled humidity” attached, every time.
Third, the demonstration is a champion cell result, not a module result. Nothing in the abstract addresses large-area uniformity, coating throughput, or yield — the quantities that decide whether a post-treatment survives translation from a laboratory scaffold to a printed module. Screen-printing is inherently scalable, and a post-treatment is inherently compatible with a printing line, but compatibility is not demonstration. The gap between “works on a laboratory scaffold” and “works on a production line” is where most perovskite process results stall, and this one has not yet crossed it.
Fourth, the mechanism evidence available here is abstract-level. The transient results are reported as conclusions, not shown as data, in the material consulted, and the supporting information’s corroborating spectra — ultraviolet photoelectron spectroscopy, Kelvin probe force microscopy, photoluminescence, impedance — sit behind the paywall. Treat the mechanism as the authors’ reported interpretation, not as independently inspected evidence.
What to watch
The most informative follow-up would be a second SAM on the same architecture. If MeO-2PACz is genuinely transferable rather than a one-molecule curiosity, then other phosphonic-acid SAMs — Me-4PACz, 2PACz, or mixed monolayers — should produce the same directional improvement, with the magnitude tracking their energy levels and binding chemistry. A second molecule would convert a demonstration into a design rule.
A damp-heat result would convert the stability question from a storage claim into a commercial one. A treated carbon cell held at 85 °C and 85% relative humidity, tracked at maximum power, would answer the moisture question the storage test leaves open — on the platform whose durability story is its main selling point.
Scale-up data would answer the process question. A slot-die or blade-coated version of the post-treatment, or any demonstration on module-scale scaffolds with reported uniformity and yield, would show whether a laboratory wash becomes a production step. Until then, the manufacturing case rests on form-factor compatibility, not on manufacturing evidence.
And the control comparison deserves publication in full. The abstract omits an untreated baseline efficiency, so the improvement cannot yet be sized from public material. When the full text circulates, the treated-versus-control gap — and the device area — will be the first numbers this desk checks.
The bottom line
A Nanyang Technological University team has taken MeO-2PACz, a self-assembled monolayer from the inverted-cell playbook, and applied it as a post-treatment to the perovskite/carbon interface of a triple-mesoscopic TiO₂/ZrO₂/carbon cell — an architecture where SAMs had not previously been used. The champion device reaches 14.7% efficiency at 0.90 V open-circuit and 70.7% fill factor, with suppressed hysteresis and 98% of initial efficiency retained after 2,400 hours under controlled humidity without encapsulation.
It is not a record, not a module, and not a weather test. It is a transferable process fix for the cheapest, most printable perovskite architecture in the field. The number to watch is not 14.7%. It is the second SAM: the first additional monolayer that produces the same directional gain on a carbon scaffold, proving that the p–i–n playbook has genuinely arrived in the printable world.
No investment recommendation is made or implied anywhere in this article.
Sources and further reading
- Wu, Q., Yantara, N., Tay, Y. B., Fu, H., Xing, X., Leong, W. L., and Mathews, N. “Extending Self-Assembled Monolayer Interface Engineering to Carbon-Based Perovskite Solar Cells.” ACS Applied Materials & Interfaces (2026), published online October 27, 2026. DOI: 10.1021/acsami.6c17328 — the primary source for the mechanism (MeO-2PACz post-treatment, improved energy-level alignment, passivation of undercoordinated Pb²⁺ sites), the transient-measurement evidence (accelerated charge extraction, extended recombination lifetimes), and all device figures quoted (champion 14.7%, 0.90 V, 70.7% fill factor, suppressed hysteresis, 98% retained after 2,400 hours under controlled humidity without encapsulation), as well as the paper’s history (received August 21, 2026; accepted October 14, 2026) and the NTU affiliations. The full text is paywalled; this article works from the abstract and does not claim any figure the abstract does not state — no short-circuit current density, control-device efficiency, device area, or humidity level is quoted, because none appears in the abstract.
- Meroni, S. M. P., Worsley, C., Raptis, D., and Watson, T. M. “Triple-Mesoscopic Carbon Perovskite Solar Cells: Materials, Processing and Applications.” Energies (2021) — background on the triple-mesoscopic architecture (screen-printed TiO₂/ZrO₂/carbon scaffold, infiltrated perovskite, no hole-transport layer, no noble-metal electrode) and its low-cost, scalable, stable character.
- Excitonic.com, “Two Defect-Fixers That Don’t Fight Each Other Take Perovskite Cells to 27.4%” — the Northwestern–DGIST orthogonal-passivation result cited for the record-cell comparison (27.4% champion, 1.187 V) and for the shared theme of treating multiple defect populations at a perovskite interface.
Background on phosphonic-acid SAMs, MeO-2PACz chemistry ([2-(3,6-dimethoxy-9H-carbazol-9-yl)ethyl]phosphonic acid), and their established use as hole-selective contacts in p–i–n devices is drawn from the standard perovskite literature. No market-size figure is quoted, because none is needed for this process result, and no module, damp-heat, or outdoor figure is stated, because none appears in the sources consulted.