Report 026 · Lab Science
In March 2026 the headlines said solar cells had done the impossible: 130% efficiency. That number sounds like free energy, a machine that gives back more than you put in. It isn't, and it also isn't wrong. It's a real, clever result wearing a figure that means something very different from what most readers hear. A measurement scientist on the two efficiencies hiding behind one percent sign.
"Solar cells just did the impossible." "Breaks a physical barrier deemed impossible." The coverage of a March 2026 result out of Kyushu University and Johannes Gutenberg University Mainz reached, understandably, for the language of a broken rule, because the number at the center of it was 130% efficiency. A device that is more than 100% efficient reads like a perpetual-motion machine, something that hands back more energy than it takes in.
The underlying science is genuinely good. The problem is only the number, or rather the single word attached to it. There are two completely different things a scientist can mean by the efficiency of anything that turns light into energy, and one of them is allowed to pass 100% while the other can never come close. Almost all of the confusion in that headline lives in the gap between the two. Sorting out which efficiency a paper is quoting is, in my day-to-day work with spectra and detectors, the first question you ask of any reported number, and it is the whole story here.
The team paired a class of materials that undergo singlet fission with a specially designed molybdenum-based "spin-flip" emitter. Singlet fission is the phenomenon at the heart of the result: in certain materials, a single absorbed photon can split its energy into two separate excitations rather than the usual one. In a solar-cell context, that means one incoming photon can, in principle, generate two electron-hole pairs instead of one.
Working in solution, the researchers measured an effective quantum yield of about 130%. In their own description, roughly 1.3 of the emitter molecules were activated for every photon the system absorbed. That is the 130%. It is a real, carefully measured figure, and the authors are clear that it is a proof of concept, a demonstration of a photophysical pathway, not a working panel.
Now the distinction that the headline flattens. When you count "1.3 molecules activated per photon absorbed," you are measuring a quantum yield, sometimes called quantum efficiency: a ratio of output events to input photons. It is a headcount. How many things came out for each photon that went in? And a headcount is perfectly allowed to exceed 100%, because one photon really can produce two excitations. Singlet fission is precisely the trick for doing it. The theoretical ceiling for a pure singlet-fission process is 200%, two excitations per photon, and 130% sits sensibly below that.
Crucially, nothing about a quantum yield over 100% violates the conservation of energy, because you are counting particles, not energy. When one photon splits into two excitations, those two share the original photon's energy. Each carries about half. You get more quanta, each individually weaker. Two half-loaves, not two loaves.
The other efficiency is the one people picture when they read the word, and it is a different animal entirely. Power conversion efficiency is a ratio of energies: electrical power out divided by solar power in. That number can never exceed 100%, because more energy out than in is exactly the perpetual-motion machine physics forbids. For an ordinary single-junction silicon cell, it saturates around 26%. The entire promise of singlet fission is to lift the theoretical ceiling on that energy efficiency, not to send it past 100%. A silicon cell paired with singlet fission has a theoretical power-conversion limit of roughly 45%. Impressive, valuable, and nowhere near the triple digits.
So "130% efficiency" is a quantum-yield number quietly relabeled as though it were a power number. The headcount went above 100%, as it is allowed to. The energy efficiency did not, and could not.
To be clear, external quantum efficiency above 100% is not a mistake or a one-off. It is a documented, reproducible thing. A decade ago, researchers built an actual silicon-and-pentacene tandem solar cell that was reported to exceed 100% external quantum efficiency at pentacene's main absorption peak, by exactly this photon-splitting mechanism. The same paper puts the design's theoretical power conversion efficiency at 45%. That single pairing is the whole lesson in miniature: the quantum-efficiency count can pass 100 while the energy efficiency stays capped at 45. They are different axes. A result can be high on one and modest on the other at the same time, and reporting one under the other's name is where the "impossible" story comes from.
There is a second translation the headline skips, and the authors themselves flag it. The 130% was measured in solution, in a beaker, as a demonstration of the energy-harvesting pathway. It was not a solar cell on a roof or even a solid device on a bench. As the Kyushu group's Nobuo Kimizuka put it, applying the work in solar cells "will require integrating singlet-fission materials with spin-flip emitters in solid-state systems." That is a real and hard road, not a formality. Plenty of beautiful solution-phase photochemistry never survives the move into a working, durable, manufacturable device.
So the honest summary carries two asterisks, neither of which makes the science less interesting. Quantum yield is not power efficiency, and a solution demonstration is not a device. Strip both away and what remains is still a legitimately clever piece of work: a new handle on steering the energy that singlet fission produces. It just isn't a 130%-efficient solar panel, because there is no such thing and the researchers never claimed there was.
When you see a light-to-energy efficiency quoted above 100%, the useful reflex is a single question: efficiency of what? If it is quantum efficiency or quantum yield, output events per input photon, then over 100% is not only possible but often the entire goal, and the right response is interest, not disbelief. If it is power conversion efficiency, energy out per energy in, then anything over 100% is a red flag that two different measurements got poured into the same word. And a second question close behind it: is this a device, or a demonstration? The same discipline runs through everything I do with a spectrometer. A measurement is an answer to a specific question, and it is only as meaningful as your grip on which question it actually answered. The percent sign is the same; the thing being counted is not.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.