A surface can sit in full midday sun and be colder than the air around it. The trick is a window in the sky

Put a small panel on a roof at noon in April, with the sun directly on it, and measure its temperature. If the panel is made correctly, it will read a few degrees below the surrounding air — not below the roof, below the air itself — while absorbing full sunlight, drawing no power, containing no refrigerant and having no moving parts.

The first reaction is usually that something has been misread, because it sounds like heat is flowing the wrong way. It is not. The panel is simply in thermal contact with something extremely cold, and that something is deep space, reachable through a gap in the atmosphere that happens to sit exactly where warm objects on Earth do most of their radiating.

Everything radiates, and what matters is at which wavelength

Any object above absolute zero emits thermal radiation, and the hotter it is, the shorter the wavelengths it emits. This is why a heated iron bar glows dull red and then orange: as its temperature climbs, its emission moves into the visible.

Objects at ordinary terrestrial temperatures — a roof, a road, a person, around 300 kelvin — radiate in the mid-infrared, with the peak of that emission near 10 micrometres. It is invisible to the eye but substantial in energy, which is what a thermal camera sees.

The Sun, at about 5,800 kelvin, radiates mostly between 0.3 and 2.5 micrometres: ultraviolet, visible light and near-infrared.

Those two bands scarcely overlap. That separation is the entire basis of what follows, because it means a surface can in principle be built to treat sunlight and its own thermal emission as two independent problems — reflecting almost all of one while radiating freely in the other.

The window in the sky

The atmosphere is not transparent to infrared. Water vapour and carbon dioxide absorb most of it strongly, which is why the planet retains as much heat as it does.

But between roughly 8 and 13 micrometres there is a band where the clear atmosphere is remarkably transparent. This is the atmospheric window, and it is the reason thermal cameras and weather satellites operate where they do.

A surface radiating into that window is not warming the air above it. Its photons travel straight out through the atmosphere and into space, which as a thermal reservoir sits at about 3 kelvin. The sky, at those wavelengths on a clear night, is genuinely and measurably cold.

This is not a new discovery so much as a newly engineered one. Making ice on clear desert nights when the air temperature never fell to freezing was practised for centuries in Persia and in parts of India — shallow trays of water, insulated from the ground, left open to a dry night sky, radiating away enough heat to freeze. The physics was in use long before anybody wrote down an emissivity spectrum.

Why daytime is the hard problem

Night-time radiative cooling is comparatively easy: there is no sunlight to reject, so a surface with high mid-infrared emissivity cools below air temperature on any clear, dry night.

Daytime is a different arithmetic, and the numbers are brutal. Direct sunlight delivers on the order of 1,000 watts per square metre. The most a surface can radiate through the atmospheric window at ambient temperature is on the order of 100 watts per square metre. So a daytime cooler that absorbs even five per cent of the sunlight falling on it has already taken in half of everything it can possibly throw away.

That sets two simultaneous and quite demanding requirements:

• Reflect the solar band almost perfectly — 95% or better across 0.3 to 2.5 micrometres, because every remaining per cent is a large fraction of the cooling budget.
• Emit strongly in 8 to 13 micrometres, ideally while emitting less outside that band, since outside the window the surface exchanges radiation with the atmosphere rather than with space.

A surface that does both is spectrally selective in a way no ordinary material is by accident, and engineering that selectivity is where the materials work lies. The first convincing demonstrations of sub-ambient daytime cooling, about a decade ago, used precisely engineered multilayer photonic films — effective, and far too expensive to put on a roof.

This is not refrigeration and it does not break any rule. The surface is not making heat flow from cold to hot; it is exchanging radiation with a reservoir at 3 kelvin, and losing that exchange badly. Space is simply a very large, very cold place that the atmosphere leaves a door open to.

How materials deliver it, and where the oxides come in

The mid-infrared half is easier than it sounds. Many ordinary materials emit strongly between 8 and 13 micrometres because their molecular bonds vibrate at those frequencies — silicon–oxygen bonds in silica and silicates, and carbon–oxygen bonds in many polymers, sit right in the window. Common polymers and silica-based materials are therefore good thermal emitters more or less for free.

The difficult half is rejecting sunlight, and it is solved by scattering rather than by reflection in the mirror sense.

A white coating is white because it contains particles that scatter visible light strongly. Scattering is most efficient when the particle size is comparable to the wavelength being scattered — a few hundred nanometres for visible light — so a coating engineered for maximum solar reflectance is a carefully sized particle dispersion in a transparent matrix. Some designs use porous polymers, where the scatterers are air voids rather than particles, achieved by phase separation during drying.

This is the same optics that appeared in our article on mineral sunscreen, pointed in the opposite direction. There the goal is to make particles small enough that visible scattering collapses and the film turns invisible. Here the goal is to size them for maximum visible scattering and make the film as white as physically possible. One body of physics, two opposite specifications.

The choice of pigment carries a subtlety worth knowing, and it is the same band-gap argument. Titanium dioxide is the world's dominant white pigment and scatters superbly — but its band gap means it absorbs ultraviolet, and in this application that absorption is a direct loss from a very tight budget. That is why barium sulphate and other UV-transparent pigments have attracted serious attention specifically for radiative cooling paints, where an ordinary excellent white pigment is not quite good enough.

The practical result is that high-performance radiative cooling is increasingly achievable as a paint or a film rather than a photonic stack — which is what moves it from a laboratory result towards something that can go on a warehouse roof.

What it can and cannot do, especially in India

The honest assessment matters more here than usual, because the gap between the physics and the brochure is wide.

Humidity, not technology, sets the limit. Water vapour absorbs in the atmospheric window and partially closes it. On a dry, clear day the sky is a cold sink; on a humid overcast day it is nearly a mirror returning the radiation. This means the achievable cooling in Jodhpur in April and in Kochi in July are entirely different propositions, and any claim that does not state the humidity conditions is not a claim you can act on. For much of India, the monsoon months are exactly when cooling is most wanted and radiative cooling works least well.

The cooling power is modest. Net figures of a few tens of watts per square metre are realistic. That is not an air conditioner and cannot become one. What it is very good at is stopping a surface from heating up in the first place.

The real prize is the "does not get hot" case, not the "goes below ambient" case. A conventional dark roof in Indian summer sun reaches 30 to 40 degrees above air temperature, and that difference is what drives heat into the building all afternoon. A good cool roof sits near or a little below air temperature. Eliminating that gradient reduces the heat entering the building enormously, and it does so passively, permanently and at the cost of paint. Sub-ambient performance is scientifically the more striking result; near-ambient performance across a whole roof is the one that changes an electricity bill.

Other applications are genuinely attractive. Cooling the condenser of an air conditioner or a refrigeration plant raises its efficiency, because a condenser rejecting heat to a cooler surface works better. Urban heat island mitigation is a direct fit. So is off-grid preservation and any equipment enclosure standing in the sun.

And there is a failure mode nobody photographs: dirt. A radiative cooler's performance rests on reflecting 95% or more of sunlight, and a layer of dust does not reflect 95% of anything. In dusty conditions — much of northern India for much of the year — soiling can remove most of the benefit within a season, which makes cleanability and soiling resistance more important to real performance than another per cent of laboratory reflectance. The same applies to weathering and to ultraviolet degradation of polymer binders over a roof's expected life. It is also worth noting that in a heating-dominated climate a cool roof is a winter penalty, which is why this is a climate-specific technology rather than a universal good.

Why it matters for students and researchers

Radiative cooling is an unusually clean demonstration that a material's spectrum matters more than any single number describing it. "Reflectivity" and "emissivity" quoted as bare figures mean very little here; what decides performance is reflectance across one band and emittance across another, and a material can be excellent at one and useless at the other. Anyone who learns to ask "at which wavelengths?" before accepting an optical claim has acquired a habit that transfers to solar absorbers, low-emissivity glazing, thermal camouflage and greenhouse films alike.

It is also a field where measurement is harder than it looks, and where many published claims are difficult to compare. Sub-ambient results depend heavily on how well the test surface was shielded from convective heat gain — a well-insulated sample under a wind shield in dry desert air and a bare panel on a humid rooftop are not measuring the same thing. Reported cooling powers should always be read together with the humidity, the sky conditions, the wind and the enclosure.

The open problems are practical and pressing. Soiling-resistant and easily cleanable coatings would deliver more real-world cooling than further spectral optimisation. Switchable or temperature-responsive surfaces, which cool in summer and stop cooling in winter, would remove the climate penalty. Field data from Indian conditions — humid coastal, dusty inland, high-altitude dry — is thin relative to how often this technology is proposed for India. And low-cost manufacturing of durable high-reflectance coatings is the difference between an interesting result and a deployed one.

Frequently asked questions

Does this break the second law of thermodynamics?

No. Heat is flowing from a warmer object to a colder one exactly as it should — it is just that the colder object is deep space at about 3 kelvin, reached through a band of wavelengths the atmosphere does not absorb. The surface ends up below the local air temperature because it is coupled more strongly to the cold sky than to the warm air touching it, and that coupling is radiative rather than conductive.

Can radiative cooling actually make ice?

Historically, yes — it is documented practice in dry regions of Persia and India, where shallow trays of water left open to a clear night sky froze even though the air never reached freezing. It needs a dry, cloudless sky, good insulation from the ground and shelter from wind, since convective gain from the air will overwhelm a few tens of watts per square metre very easily. Modern demonstrations reproduce it, and it remains a demonstration rather than a refrigeration method.

Is this just painting the roof white?

Partly, and painting a roof white is genuinely worthwhile. The difference is that an ordinary white paint is chosen to look white — that is, to reflect visible light — while a radiative cooling coating is engineered for two things at once: very high reflectance across the whole solar band including the near-infrared, and high emittance specifically in the 8 to 13 micrometre window. Cool-roof standards measure both properties for precisely this reason, and a paint can score well on one and poorly on the other.

Will it work during the monsoon?

Much less well, and this is the honest limitation. Cloud cover and high humidity close the atmospheric window, so the sky stops behaving like a cold sink. A cool roof still helps in those conditions by reflecting sunlight, which is a large benefit in itself — but the distinctive sub-ambient behaviour largely disappears when the sky is humid or overcast.

Could this cool a phone, a car or a person?

Only where the surface has a clear view of the sky and is not simultaneously being heated from behind. A car roof parked in the open is a reasonable candidate and coatings for this are being developed; a phone in a pocket or a hand has no sky view and cannot use the mechanism at all. Radiative-cooling textiles that are transparent in the mid-infrared, letting body heat radiate out, are a real research area with promising results and considerable practical caveats.