A certified 35.5% perovskite-silicon tandem result shows why stacking absorbers can beat silicon's single-junction limit, while leaving durability and scale as the harder commercial tests.
By Lachlan Brown · Edited by Lachlan Brown
Published August 20, 2026 · How we edit
LONGi says a crystalline silicon-perovskite tandem solar cell developed by its research team has reached 35.5% power-conversion efficiency. According to the company’s July 15 announcement, the result was independently certified by the European Solar Test Installation, or ESTI.
That number is best read beside another: 3.8%. In 2009, Akihiro Kojima and colleagues reported a perovskite-sensitized solar cell at that efficiency in the Journal of the American Chemical Society. In 17 years, a material family that began as a fragile laboratory curiosity has become the top half of a tandem device capable of converting more than a third of incoming solar power into electricity.
This is a certified research-cell result, not a commercial solar panel. LONGi’s announcement does not specify the area of the 35.5% cell or provide an accompanying peer-reviewed paper, so some of the technical detail needed to assess the latest step is not yet public.
Power-conversion efficiency is the share of incident light power that a cell turns into usable electrical power under standardized test conditions. It is not the same as the energy a panel will deliver over a year on a roof or in a solar farm, where heat, shade, angle, dirt, wiring and local weather all matter.
The distinction between a cell and a module matters as well. A research cell is a carefully made device. A module connects and encapsulates many cells across a much larger area, adding inactive borders, interconnections and optical losses. LONGi says its larger tandem cells have reached 34.3% over 261 square centimeters and 32.2% over 274 square centimeters. Its tandem modules have reached 31.4% and 29.4%.
Those numbers are not a weakness hidden in the release. They show why cell records and product performance should not be treated as interchangeable.
The original 2009 paper used methylammonium lead iodide nanocrystals to sensitize a porous titanium-dioxide electrode. The iodide version converted 3.8% of the incoming solar energy. A bromide version produced a relatively high voltage but lower overall efficiency.
Those first devices also contained a liquid electrolyte that dissolved the perovskite. They lost performance within minutes. The starting point was scientifically interesting, but nowhere close to a roof-ready technology.
Perovskite is the name of a crystal structure, not one fixed substance. Researchers can change the ions occupying that structure and tune the material’s bandgap, meaning the range of photon energies it can absorb. Over the following years, teams replaced the liquid design with solid transport layers, improved crystal growth, reduced defects and learned to protect the thin boundaries where one material meets another.
That tunability is especially useful when a perovskite sits on silicon.
A conventional silicon cell has one light-absorbing junction. Photons with too little energy pass through without freeing an electron. Photons with more energy than the silicon bandgap can use still free a charge, but much of their surplus energy becomes heat. Recombination, in which an electron falls back before contributing useful current, removes more.
Detailed calculations put the practical theoretical ceiling for a single-junction crystalline silicon cell at roughly 29%, as a recent review of multijunction photovoltaics notes. A more general idealized limit for any single junction, assuming the best possible bandgap, is about 33.7%.
A tandem changes the optical division of labor. The wide-bandgap perovskite top cell absorbs higher-energy visible light. Lower-energy red and near-infrared light passes into the silicon bottom cell. Each junction handles a portion of the spectrum better suited to it, reducing the energy lost when one absorber has to accept every photon.
So 35.5% does not mean researchers somehow broke the physics governing silicon. The device exceeds silicon’s single-junction ceiling because it is no longer a single-junction silicon device.
Stacking absorbers creates its own losses. Light has to cross transparent electrodes. Charges have to move through selective layers and a connection between the two subcells. In a two-terminal tandem, the subcells are wired in series, so the lower current limits the whole device. Changes in the color balance of sunlight can shift that current matching.
LONGi’s published 34.58% predecessor offers a useful look at the sort of engineering behind recent records. In a 2025 Nature paper, Lingbo Jia and more than 50 coauthors reported an asymmetric self-assembled molecule used as a hole-selective layer. Its shape improved coverage over textured silicon and helped align energy levels at the buried interface.
In plainer language, the treatment made it easier for useful charge to leave the perovskite while giving fewer electrons and holes a place to recombine. ESTI certified that earlier device at 34.58%. LONGi has not yet published equivalent device-level detail for the new 35.5% result, so it would be premature to say exactly which change supplied the additional 0.92 percentage points.
A solar product has to be more than efficient on its best day. It must be coated uniformly over large areas, leave a factory at high yield, tolerate partial shade and electrical stress, and survive heat, moisture and ultraviolet exposure for decades. Most leading perovskite compositions also contain lead, increasing the importance of robust encapsulation, leakage testing and responsible end-of-life handling.
A 2025 Nature Photonics review organized the field’s remaining work around three connected problems: efficiency, long-term stability and scalability. Progress in one does not automatically solve the others. A delicate interface treatment that performs beautifully on a small research device may prove slow, expensive or inconsistent across thousands of full-size wafers.
I recently wrote about solar’s record contribution to global electricity in 2025 and the grid bottleneck that follows it. Higher-efficiency tandems will not replace transmission, storage or good project design. Their attraction is narrower: more watts from the same panel area could reduce the land, mounting, cabling and labor required for a given capacity, provided the added cell cost and lifetime hold up.
The next persuasive milestone would pair high module efficiency with transparent manufacturing yield and degradation data. That is the evidence needed to turn an exceptional measurement into dependable hardware.
Standards
Articles on Science Blog are edited and fact-checked before publication. We use AI tools in the newsroom. See our editorial standards and masthead.
Written by
Lachlan Brown is Editor-in-Chief of Science Blog and co-founder of Brown Brothers Media. He writes about psychology, human behavior, and the unexpected ways scientific findings reshape how we understand everyday life.
Clear science for curious people
A Brown Brothers Media publication. All rights reserved © 2026.
Clear science for curious people — free PDF, then our science email.
Also available via RSS.

source