A solar panel on a rooftop in Stanford, California, kept producing electricity well after sunset. No battery was involved. Instead, the panel itself was the power source, harvesting the invisible stream of heat that the Earth sends skyward every night. The device, built by engineers at Stanford and reported in a peer-reviewed Applied Physics Letters paper, generated 50 milliwatts per square meter in the dark, a tiny but real trickle of power drawn from nothing more than the temperature difference between the cooling panel surface and the warmer air around it.
That number is vanishingly small next to what the same panel produces at noon. A conventional solar module delivers roughly 200 watts per square meter in direct sunlight, meaning the nighttime output is about four thousand times weaker. But the point, as the researchers behind these experiments have argued, is not to replace daytime solar. It is to prove that the hardware already sitting on millions of rooftops could, with modifications, keep working around the clock.
As of mid-2026, several research groups have demonstrated working prototypes, and the power densities are climbing. The question is whether the physics can scale from lighting a single LED to doing something the grid actually notices.
How a solar panel generates power in the dark
The underlying physics is straightforward, even if the engineering is not. Every object above absolute zero radiates heat as infrared light. At night, the Earth’s surface is warmer than the upper atmosphere and outer space, so heat flows upward. A panel pointed at the sky acts as a radiator, shedding infrared energy and cooling below the temperature of the surrounding air.
In the Stanford experiment, led by Tristan Deppe and Jeremy Munday, a standard photovoltaic cell served as that radiator. A thermoelectric generator (TEG), sandwiched between the cooling cell and the ambient environment, converted the temperature gap into voltage. The team measured an open-circuit voltage of about 100 millivolts and a power density of 50 mW/m², all from a physical device rather than a simulation, according to the Applied Physics Letters paper archived by the U.S. Department of Energy.
An earlier experiment took a slightly different approach. Aaswath Raman’s group, then split between Stanford and UCLA, built a sky-facing radiative cooler paired with a thermoelectric module and showed it could generate enough electricity to light an LED. That work, published in the journal Joule and available through ScienceDirect, was one of the first clear demonstrations that radiative cooling could do useful electrical work after dark.
Pushing past the first benchmarks
Researchers have since more than doubled the Stanford figure. An open-access study published in iScience and archived on PubMed Central reported a nighttime power density exceeding 100 mW/m² by optimizing the size of the radiative surface and the thermal resistance between it and the thermoelectric module. That result suggests 50 mW/m² is not a hard ceiling and that careful engineering can push output higher.
A separate line of research skips the thermoelectric module entirely. So-called thermoradiative diodes work like solar cells running in reverse: instead of absorbing incoming photons, they emit infrared photons and generate a voltage in the process. A team at UNSW Sydney, led by researchers including Phoebe Pearce and Nicholas Ekins-Daukes, built a thermoradiative diode from mercury cadmium telluride (HgCdTe), the same semiconductor used in night-vision cameras. The device produced measurable electrical power from infrared thermal emission, though at roughly the scale needed to run a wristwatch, according to a UNSW Sydney release and a related Nature feature.
The thermoradiative approach is appealing because it could, in theory, be integrated directly into a photovoltaic cell without bolting on a separate thermoelectric module. But the UNSW team’s output has not yet been quantified in mW/m² in published work, making direct comparisons with the thermoelectric results difficult.
What the technology cannot do yet
The gap between laboratory proof and practical deployment is wide, and the published research is candid about it.
No long-term outdoor data exists. Every verified result comes from controlled lab or short-duration rooftop tests. None of the primary studies document how performance holds up over months of exposure to dust, humidity, rain, or seasonal temperature swings.
Cost is uncharted. Neither the Applied Physics Letters paper nor the iScience study provides estimates of manufacturing cost per square meter. Thermoelectric modules are commercially available but adding them to every solar panel on a roof is an engineering and economic challenge no one has publicly costed out. HgCdTe, the material behind the thermoradiative diodes, is expensive and currently produced in small volumes for defense and imaging applications.
Grid connection is unexplored. The experiments produce milliwatt-scale DC power at around 100 millivolts. Feeding that into a home inverter or grid-tied system would require power electronics that do not yet exist for this application. The sources do not describe any integration with inverters, storage, or grid controls.
Independent replication is thin. The 100+ mW/m² result from the iScience paper stands as a single experimental demonstration. No third-party group has publicly reported reproducing that specific figure.
How nighttime solar fits into the bigger energy picture
To put the numbers in perspective: a rooftop solar array covering 20 square meters produces roughly 4,000 watts at peak sun. The same area equipped with the best demonstrated nighttime technology would produce about 2 watts after dark. That is enough to run a few sensors, keep a microcontroller alive, or trickle-charge a small battery, but it is not going to power a refrigerator, let alone a household.
Battery storage, by contrast, is already solving the nighttime problem at grid scale. Lithium-ion pack prices have fallen below $140 per kilowatt-hour in many markets, and large battery installations routinely shift gigawatt-hours of solar energy from afternoon to evening peaks. Nighttime radiative harvesting is not competing with batteries on those terms, at least not in its current form.
Where the technology could find a nearer-term role is in off-grid and low-power applications. Remote weather stations, agricultural sensors, wildlife monitors, and IoT devices in places without reliable grid access often need only milliwatts to operate. A panel that produces a small but continuous trickle of power day and night could reduce or eliminate the need for battery replacements in those settings.
The researchers behind these prototypes have framed them as complements to daytime solar, not replacements. The UNSW team described thermoradiative diodes as a way to harvest energy from the warm Earth during hours when photovoltaics sit idle. The Stanford work showed that a standard PV module could be adapted with a thermoelectric layer for nocturnal output. Both visions point toward hybrid panels that work around the clock, even if the nighttime contribution remains a fraction of daytime production.
Where the research goes from here
The path from 50 mW/m² to something grid-relevant is steep but not necessarily impossible. Radiative cooling research has accelerated in recent years, driven partly by interest in passive building cooling, and the materials science overlaps significantly. If thermoradiative diodes can be fabricated from cheaper semiconductors, or if thermoelectric module costs drop with scale, the economics could shift.
For now, the honest summary is this: engineers have proven that solar-style panels can generate real, measurable electricity after dark by tapping the same infrared radiation the Earth has been sending into space since long before anyone thought to capture it. The power is tiny. The potential is not. And the next few years of outdoor testing, cost analysis, and independent replication will determine whether nighttime solar becomes a footnote in energy research or a quiet, steady contributor to a grid that never fully sleeps.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
