A Berlin research team has published what may be the most consequential interface-engineering result the perovskite solar field has seen in years: a triple-junction cell built entirely from perovskite semiconductors that achieves 27.3% certified power conversion efficiency and holds more than 90% of that output after 770 consecutive hours of operation. The paper, published July 9 in Joule by researchers at the Helmholtz-Zentrum Berlin (HZB) and collaborators at the University of Potsdam, Technical University Berlin, and Switzerland’s federal materials lab Empa, names the specific buried interface that has blocked this class of cell from advancing — and shows how a two-layer graphene-oxide sandwich fixes it.
The result matters because the stability problem it solves is not incidental to perovskite multi-junction technology — it has been the organizing obstacle to it. Prior all-perovskite triple-junction devices built with the conventional polymer hole-transport layer typically lost 10% of their efficiency after about 380 hours of continuous operation. The HZB device more than doubles that lifetime, while simultaneously pushing efficiency to one of the highest figures recorded for a fully perovskite architecture.
What Makes All-Perovskite Cells Different — and Why They’ve Been Fragile
Perovskite semiconductors absorb light exceptionally well, are lightweight, and can be deposited on plastic, fabric, and other flexible substrates by low-cost coating methods — properties that silicon cannot match. Multi-junction designs stack two or three perovskite layers, each engineered to capture a different slice of the solar spectrum, which lets the combined device harvest far more light than any single absorber can. For a triple-junction configuration, the theoretical efficiency ceiling reaches approximately 52%, compared to the 33.7% Shockley-Queisser limit that governs any single-junction cell regardless of the material.
That makes all-perovskite multi-junction architecture scientifically compelling — and it makes its persistent fragility a genuine problem for the field. The device category can in principle enable not just higher efficiency at lower cost, but entirely different product geometries: lightweight flexible modules for building-integrated photovoltaics, curved aerospace surfaces, and IoT power applications where silicon panels are physically impossible. But realizing those possibilities has required a level of durability the technology has repeatedly failed to deliver.
The root cause has been consistently traced to a single location: the interface between the tin-lead (Sn-Pb) perovskite bottom subcell and the hole transport layer sitting above it. The Sn-Pb absorber captures near-infrared light that the other two subcells miss, making it essential to the device’s light-harvesting architecture. But it is also the most chemically vulnerable component in the stack, and the material historically used to extract charge from it has been making things worse.
Why PEDOT:PSS Was the Weakest Layer
That material is PEDOT:PSS — poly(3,4-ethylenedioxythiophene):polystyrene sulfonate — a conducting polymer that has served as the standard hole transport layer in inverted perovskite devices for years. It has two well-documented failure modes when placed next to a tin-lead absorber. First, at thicknesses typically between 30 and 50 nanometers, it absorbs parasitic light that should be reaching the active semiconductor — robbing the system of an estimated 0.5 to 1.4 milliamps per square centimeter of current. Second, PEDOT:PSS is both acidic and hygroscopic: it draws moisture and creates a chemically hostile environment at the interface with the tin-containing layer, accelerating the oxidation of tin ions (Sn²⁺ to Sn⁴⁴⁺) that degrades the absorber’s performance. The combination of optical and chemical losses made PEDOT:PSS-based all-perovskite triple-junction devices reliable only to the 380-hour threshold before efficiency fell below 90%.
Prof. Steve Albrecht, who heads HZB’s Department of Perovskite Tandem Solar Cells and led the research team, described the architecture using an analogy that has circulated widely since publication: the three perovskite absorber layers are like the buns of a Big Mac, he explained, separated by different functional layers serving the role of fillings. The problem the team set out to solve was in exactly one of those fillings — the interface between the middle and bottom buns — and the solution replaced the problematic layer entirely.
How the GO/SAM Bilayer Works
Self-assembled monolayers (SAMs) — ultra-thin films of large organic molecules that spontaneously organize into single-molecule sheets on a surface — had already proven effective as hole-transport contacts in lead-based perovskite cells. Kevin Prince, co-first author of the study and a postdoctoral researcher at HZB, explained the team’s starting logic: “We had already established self-assembled monolayers as hole-conducting contact layers in lead-based perovskite solar cells with great success, so it made sense to use these in tin-lead perovskites and triple-junctions as well.”
The difficulty was that SAMs alone did not work in the tin-lead system. Carbazole-based SAMs — the most common variety used in perovskite PV — produced non-uniform buried interfaces on the Sn-Pb absorber that suppressed hole transport and raised non-radiative recombination losses. Positive charge carriers simply could not move efficiently enough out of the absorber and into the electrode circuit.
The breakthrough came when co-first author Yeonghun Yun and colleagues placed a thin layer of graphene oxide (GO) directly beneath the SAM. “We therefore experimented with additional layers beneath the SAM layer to act as a kind of substrate,” Yun said. That addition proved decisive. The GO layer provided an electronically and morphologically compatible foundation for the SAM, improving surface coverage across the buried interface, promoting more uniform perovskite crystal formation, and improving the orientation of crystallites at the critical contact zone. Both electronic carrier losses and ionic losses dropped sharply.
The chemistry of the fix is straightforward. Where PEDOT:PSS was acidic and moisture-attracting, graphene oxide is chemically benign — it neither acidifies the interface nor draws in water that would attack the tin ions. The GO/SAM bilayer forms a protective barrier that shields the Sn-Pb absorber from the degradation mechanisms that PEDOT:PSS had been enabling.
What the Numbers Show
In single-junction Sn-Pb test devices, the GO/SAM approach delivered a power conversion efficiency of 22.1% with an open-circuit voltage of 0.88 V — a strong baseline result for this class of cell that validated the interface approach before the team moved to the full triple-junction architecture.
Integrated into the triple-junction stack in place of PEDOT:PSS, the bilayer’s optical improvement alone raised photogeneration in the Sn-Pb subcell to 10.3 milliamps per square centimeter — recovering roughly 0.5 milliamps per square centimeter from the parasitic absorption the polymer had been causing. The full device achieved a certified power conversion efficiency of 27.3%, stabilizing at 27.0% under continuous operation.
Stability — historically the more demanding metric — proved equally striking. Devices retained over 90% of their initial efficiency (T90) after 770 hours of continuous operation at 25°C. PEDOT:PSS-based equivalents hit the same T90 threshold at approximately 380 hours. The GO/SAM architecture more than doubled the operational lifetime of this device class.
“Our analyses suggest that, with improvements to the quality of the individual perovskite layers and interlayer films, the efficiency of this architecture could be increased to over 30%,” Albrecht said. An HZB program targeting exactly that outcome — all-perovskite triple-junction cells above 30% efficiency and five-year operational stability — is already underway.
How This Positions All-Perovskite Cells Against Perovskite-Silicon Hybrids
The broader context matters for interpreting the 27.3% figure. The current efficiency leader for triple-junction cells is EPFL and CSEM’s perovskite-silicon device, which reached 30.02% in March 2026 — but that architecture combines perovskite absorbers with a silicon bottom cell, which requires the complex bonding and processing steps that silicon wafer technology demands. Oxford PV’s perovskite-silicon tandem modules, the first commercially shipped, carry 24.5% module efficiency but require silicon manufacturing infrastructure. These hybrids are closer to market and hold higher certified efficiency records.
All-perovskite designs like the HZB device occupy a different lane. Because every absorber layer is a perovskite semiconductor, the entire device can in principle be manufactured using solution-based roll-to-roll coating methods at low cost — and crucially, on flexible substrates that silicon wafers cannot match. The practical value of that flexibility is not only manufacturing economy; it is access to application markets where silicon panels are physically incompatible. Curved building facades, lightweight aerospace power sources, IoT sensor panels, and integration into textiles and consumer electronics are product categories where high-efficiency flexible solar cells have no silicon competitor.
The HZB result does not deliver commercial readiness. Seven hundred and seventy hours is a meaningful laboratory milestone — more than double prior state-of-the-art — but it falls far short of the 20,000 to 25,000 annual operating hours a panel must endure to reach the 25-to-30-year commercial warranties that the solar industry requires, and it does not yet address the three outdoor degradation mechanisms (halide phase segregation, copper corrosion, and edge pattern formation) that a separate HZB study identified as the key barriers to real-world perovskite durability. The cell also uses a lead-containing absorber in the Sn-Pb layer, keeping lead toxicity as an open environmental question for commercialization.
What the result does establish is that the chemical instability plaguing all-perovskite triple-junction cells is a solvable interface-engineering problem with a specific, identifiable fix — not an intrinsic property of the materials class. The buried interface between the Sn-Pb absorber and its hole transport layer is now a defined engineering target. The GO/SAM bilayer is a modular solution: it addresses exactly that interface and can be adopted without redesigning the rest of a triple-junction stack.
Where the Research Goes From Here
The authors have published the full device architecture, interface characterization protocols, and stability measurements in Joule, giving other laboratories a detailed blueprint. The paper’s author list — 23 researchers across four institutions — reflects the scale of collaboration the field requires to move from interface chemistry to certified device metrics.
Near-term priorities are likely to include demonstrating the GO/SAM bilayer’s performance on larger device areas, validating stability under real outdoor illumination rather than controlled laboratory aging at 25°C, and testing whether deposition methods compatible with high-throughput roll-to-roll manufacturing can maintain the interface quality achieved in laboratory conditions. The HZB HIG program’s stated targets — above 30% efficiency and five-year stability — give the field a specific benchmark to pursue with the interface architecture now established.
For the perovskite community, the July 9 Joule paper answers a question the field has been working on for years. It does not answer the commercial question — that depends on years of scale-up work still ahead. But it identifies where the problem was, names a solution, and publishes the data.
Frequently Asked QuestionsWhat exactly did the HZB team change to improve stability?
The team replaced PEDOT:PSS — the polymer hole transport layer that sits between the middle and bottom perovskite absorber subcells — with a two-layer combination of graphene oxide (GO) and a self-assembled monolayer (SAM). PEDOT:PSS is chemically acidic and attracts moisture, both of which accelerate the oxidation of the tin ions in the tin-lead absorber beneath it. Graphene oxide is chemically inert; it neither acidifies the interface nor draws moisture, which eliminates those degradation pathways. The SAM then handles hole extraction efficiently. Together, the bilayer removed the chemical attack on the absorber and also eliminated the parasitic light absorption that PEDOT:PSS had been causing.
Is 770 hours enough to consider all-perovskite solar cells commercially ready?
No. A silicon solar panel operating at a typical solar irradiance carries roughly 1,800 to 2,000 operating hours per year in most markets, and commercial warranties extend to 25 or 30 years — meaning a commercially viable panel must sustain performance for approximately 45,000 to 60,000 operational hours under real outdoor conditions. The HZB result was measured under controlled laboratory conditions (25°C, continuous illumination) and does not replicate the combined stresses of heat, UV, moisture, and temperature cycling that occur outdoors. A separate HZB study published in June 2026 identified three distinct outdoor degradation mechanisms in perovskite cells — halide phase segregation, copper corrosion, and edge pattern formation — none of which is yet fully addressed by the GO/SAM architecture alone. The 770-hour result establishes that the interface chemistry problem is solvable; the commercialization problem requires years of additional scale-up and outdoor validation work.
Can all-perovskite cells do things silicon panels cannot?
Yes, and this is the less-discussed part of why the all-perovskite architecture is worth pursuing in parallel with perovskite-silicon hybrids. Because every absorber layer in an all-perovskite device is a perovskite semiconductor, all layers can in principle be deposited by low-cost, roll-to-roll-compatible coating processes — and on flexible substrates including plastic and fabric. Silicon wafers are rigid and fragile; they cannot be shaped to curves or integrated into textiles or lightweight aerospace structures. An all-perovskite flexible cell achieving efficiency in the high twenties or above thirty percent, if one can be demonstrated at commercial scale, would be viable in markets — curved building facades, wearable electronics, satellite power, lightweight IoT sensors — where silicon is not a realistic option regardless of how inexpensive it becomes.
How does this result compare to current efficiency records in solar generally?
The 27.3% figure is certified specifically for an all-perovskite triple-junction cell — a category where the prior T90 lifetime record was approximately 380 hours. For context: the overall triple-junction efficiency record belongs to a perovskite-silicon device (EPFL/CSEM, 30.02%, March 2026), not an all-perovskite one. The highest single-junction perovskite record is around 27.3% (NREL-certified, 2025) — meaning the HZB triple-junction result matches or exceeds what single perovskite layers achieve, which is notable. Commercial silicon panels currently ship at 22–24% efficiency. The theoretical ceiling for a triple-junction all-perovskite cell, if the engineering can reach it, is approximately 52%. Full technical details of the HZB device architecture, interface characterization, and stability data are available in the HZB triple-junction perovskite paper published in Joule.