Radiation Heat Transfer Explained for Engineers

Of the three modes of heat transfer, radiation is the one engineers most often underestimate — right up until it dominates the energy balance and nothing else explains where the heat is going. Conduction needs a solid path and convection needs a moving fluid, but thermal radiation needs nothing at all: every surface above absolute zero emits electromagnetic energy, and that energy crosses a vacuum just as happily as it crosses air. It is how the Sun heats the Earth across 150 million kilometres of empty space, how a furnace cooks its load without touching it, and how a spacecraft sheds its waste heat with no air to convect into. This article works through the physics an engineer actually needs: the Stefan-Boltzmann law, why the temperature appears to the fourth power, what emissivity really measures, the absolute-temperature trap that wrecks more calculations than any other single mistake, and when radiation overtakes its two rivals. A fully worked example at the end ties it together with numbers you can reproduce in the radiation heat transfer calculator.

The Stefan-Boltzmann law

The governing equation of thermal radiation is the Stefan-Boltzmann law. For an ideal radiator — a black body — the total power emitted per unit surface area is

• E_b = σ·T⁴

where E_b is the emissive power in W/m², T is the absolutetemperature in kelvin, and σ is the Stefan-Boltzmann constant. The value used throughout this article and in the site’s calculator is

• σ = 5.67×10⁻⁸ W/(m²·K⁴)

(The 2018 CODATA value is 5.670374419×10⁻⁸; rounding to 5.67×10⁻⁸ is standard engineering practice and changes nothing at three significant figures.) A real surface is not a perfect black body — it emits some fraction ε of the black-body value, called the emissivity. So the emissive power of a real (gray) surface is ε·σ·T⁴, and the net radiative exchange between a surface at temperature T₁ and large surroundings at temperature T₂ is the workhorse form most engineers reach for:

• Q = ε·σ·A·(T₁⁴ − T₂⁴)

Here Q is the net heat-transfer rate in watts, A is the radiating surface area in m², ε is the surface emissivity, and T₁ and T₂ are the absolute temperatures of the surface and its surroundings. This is the equation behind almost every back-of-the-envelope radiation estimate, and it is the one the radiation calculator evaluates. The expression assumes the surroundings are large enough (or black enough) to act as an ideal sink, which is an excellent approximation for a small object in a big room, a pipe in open air, or a wall facing a much larger enclosure.

Why the fourth power, not a linear law?

The single most important feature of radiation — the one that makes it behave completely differently from conduction and convection — is that it scales with the fourth power of absolute temperature. Conduction (Q = kA·ΔT/L) and convection (Q = hA·ΔT) are both linear in the temperature difference. Radiation is not. The driving term is the difference of fourth powers, T₁⁴ − T₂⁴, and that has dramatic consequences.

The T⁴ dependence is not an empirical fudge factor; it falls directly out of the quantum statistics of black-body radiation. Planck’s law describes how much energy a black body emits at each wavelength, and integrating that spectral distribution over all wavelengths yields the compact σT⁴ result. Physically, two effects compound as a body gets hotter: it emits more photons, and each of those photons carries higher average energy. Both scale with temperature, and their combination produces the fourth-power law.

The practical upshot is that radiation is extraordinarily temperature sensitive. Double the absolute temperature and you multiply the emitted power by 2⁴ = 16. A surface at 600 K radiates sixteen times more energy per unit area than the same surface at 300 K, even though the Celsius temperatures (327 °C vs 27 °C) differ by only a factor of about twelve. This is exactly why radiation is negligible near room temperature but unavoidable in furnaces, flames, and incandescent surfaces: the curve turns up so steeply that a few hundred extra kelvin changes everything.

The absolute-temperature trap (use Kelvin)

Because the law uses T⁴, the temperatures in the Stefan-Boltzmann equation must be absolute — kelvin or rankine, never Celsius or Fahrenheit. This is the single most common error in radiation calculations, and it is not a small one. Consider a 600 °C surface.

• Correct: T = 600 + 273.15 = 873.15 K, so T⁴ = 5.81×10¹¹ K⁴.
• Wrong: using 600 directly gives 600⁴ = 1.30×10¹¹, which is less than a quarter of the correct value.

Using Celsius does not introduce a small percentage error — it produces an answer that is wrong by more than a factor of four, and the error gets worse the lower the temperature. There is no shortcut and no exception: convert every temperature to kelvin before raising it to any power. The Celsius-to-Kelvin converter does the arithmetic, but the rule is simple enough to internalise: add 273.15.

A closely related mistake is subtracting the temperatures before raising to the fourth power — computing (T₁ − T₂)⁴ instead of T₁⁴ − T₂⁴. These are entirely different quantities. The law uses the difference of the fourth powers, not the fourth power of the difference. Raise each temperature to the fourth power first, then subtract.

Emissivity: what it is and typical values

Emissivity ε is the ratio of the radiation a real surface emits to the radiation a perfect black body would emit at the same temperature. It ranges from 0 (a perfect reflector that emits nothing) to 1 (an ideal black body). Emissivity is a property of the surface, not the bulk material — it depends heavily on finish, oxidation, and coating, which is why two pieces of the same steel can have wildly different emissivities depending on whether they are polished or rusted.

Kirchhoff’s law of thermal radiation ties emissivity to absorptivity: at a given wavelength and temperature, a good emitter is an equally good absorber. A surface with ε = 0.9 also absorbs 90% of the radiation that lands on it; a polished surface with ε = 0.05 both emits and absorbs poorly, which is precisely why it makes a good radiation shield. The following values are worth carrying in your head to within a factor of two.

Two intuitions deserve emphasis. First, polished metals are dramatically different from every other surface — an order of magnitude lower in emissivity — which is why a bright aluminum foil radiates so little and a black-anodized heat sink radiates so much. Second, visible color is a poor guide: snow and water look bright or transparent to the eye but are nearly black (ε ≈ 0.95) in the infrared band where room-temperature objects actually radiate. Emissivity is an infrared-band property, and assuming ε = 1 for a polished metal can overstate its radiation by a factor of ten to fifty.

Black body vs gray body

A black body is an idealization: a surface that absorbs all incident radiation and emits the theoretical maximum (ε = 1) at every wavelength. No real surface is perfectly black, though a small hole into a large cavity comes very close, which is why furnace ports and calibration sources are built that way.

A gray body is the working engineering model: a surface whose emissivity is constant with wavelength but less than one. The gray approximation is what lets us pull a single ε out front and write Q = εσA(T₁⁴ − T₂⁴). It is not exact — real emissivities vary with wavelength and viewing angle, and selective surfaces (like the spectrally tuned coatings on solar collectors) deliberately break the gray assumption by absorbing strongly in the solar band while emitting weakly in the infrared. For most engineering estimates, however, treating a surface as gray with a representative ε is accurate enough and enormously convenient.

View factors (a brief word)

The simple Q = εσA(T₁⁴ − T₂⁴) form quietly assumes that essentially all of the surface’s radiation reaches the surroundings — true when a small object is enclosed by much larger surroundings. When two finite surfaces exchange heat directly, only a fraction of the radiation leaving one actually strikes the other. That fraction is the view factor (also called the shape, configuration, or angle factor) F₁₂, defined as the fraction of radiation leaving surface 1 that arrives at surface 2.

For two black surfaces, the net exchange becomes Q = σ·A₁·F₁₂·(T₁⁴ − T₂⁴). View factors obey two useful rules: the reciprocity relation A₁F₁₂ = A₂F₂₁, and the summation rule that all view factors from one surface to a complete enclosure add to one (ΣF_1j = 1). Tabulated charts and closed-form expressions exist for common geometries — parallel plates, coaxial cylinders, perpendicular rectangles — and for gray surfaces the analysis extends to a radiation-network (radiosity) method. Full view-factor and enclosure analysis is beyond a single calculator, but it is worth knowing the term exists: when your simple estimate seems too high, a view factor well below one is often the reason.

The radiative heat-transfer coefficient h_rad

The T⁴ nonlinearity makes radiation awkward to combine with convection, which is linear in ΔT. The standard trick is to linearize the radiation term into an effective coefficient h_rad that looks and behaves like a convective coefficient, so the two can be added in a resistance network. Starting from Q = εσA(T₁⁴ − T₂⁴) and factoring the difference of fourth powers:

• h_rad = ε·σ·(T₁ + T₂)·(T₁² + T₂²)

With this coefficient, the radiation rate takes the familiar linear form Q = h_rad·A·(T₁ − T₂), directly comparable to the convective Q = h_conv·A·(T₁ − T₂). A surface losing heat by both convection and radiation in parallel then has a combined coefficient h = h_conv + h_rad, and you can build a single thermal resistance R = 1/(h·A). The catch is that h_rad depends on the very temperatures you are solving for, so problems where radiation matters are usually iterative. The convection calculator handles the parallel convective branch, and comparing the two coefficients is the cleanest way to see which mode is carrying the load.

When does radiation dominate?

Because radiation grows as T⁴ while conduction and convection grow linearly in ΔT, the crossover is all about temperature. As a rough field guide for surfaces losing heat to a room-temperature environment:

• Near room temperature (up to ~100 °C): radiation and natural convection are comparable, with radiation often slightly smaller. A bare radiator panel or a person in a room loses heat by both in similar measure — which is why radiant comfort matters even when air temperature is fixed.
• Moderate temperatures (~100–400 °C): radiation grows quickly and becomes the equal or larger partner for surfaces in still air. Insulated-pipe and equipment heat-loss estimates that ignore radiation start to err noticeably here.
• High temperatures (above ~500–600 °C): radiation dominates. In furnaces, boilers, fired heaters, and around flames, radiative transfer carries the overwhelming majority of the heat, and the design is built around it. Natural convection becomes a rounding error by comparison.

Forced convection changes the picture: a strong fan or fast liquid flow can keep convection competitive to much higher temperatures, because a large h_conv raises the linear term. But in still air or a vacuum, radiation always wins eventually, and in a vacuum it is the only mode available — the reason spacecraft thermal control is entirely a radiation problem. The honest engineering move is to estimate all three modes (conduction through the conduction calculator, convection, and radiation) and keep whichever ones are within an order of magnitude of the largest.

Worked example: a hot furnace wall radiating to a workshop

A 2.5 m² section of oxidized, painted furnace wall sits at T₁ = 600 °C and radiates to a workshop whose surfaces are at T₂ = 25 °C. The surface emissivity is ε = 0.85. Find the net radiative heat-transfer rate Q, the heat flux q, and the linearized radiation coefficient h_rad. These are the exact inputs the radiation calculator uses, so every figure below is reproducible.

Step 1 — convert to absolute temperature. This is the step you can never skip:

• T₁ = 600 + 273.15 = 873.15 K
• T₂ = 25 + 273.15 = 298.15 K

Step 2 — raise each to the fourth power. Note we raise first, then subtract:

• T₁⁴ = 873.15⁴ = 5.8124×10¹¹ K⁴
• T₂⁴ = 298.15⁴ = 7.9020×10⁹ K⁴
• T₁⁴ − T₂⁴ = 5.8124×10¹¹ − 7.9020×10⁹ = 5.7334×10¹¹ K⁴

Notice that T₂⁴ is barely more than 1% of T₁⁴. At high temperatures the cold side contributes almost nothing — the hot surface dominates the exchange, another face of the T⁴ sensitivity.

Step 3 — apply the Stefan-Boltzmann law with σ = 5.67×10⁻⁸ W/(m²·K⁴):

• ε·σ·A = 0.85 × 5.67×10⁻⁸ × 2.5 = 1.2049×10⁻⁷
• Q = 1.2049×10⁻⁷ × 5.7334×10¹¹ = 69,080 W ≈ 69.1 kW

Step 4 — heat flux and coefficient. The flux is just Q per unit area, and h_rad comes from the linearization formula:

• q = Q/A = 69,080 / 2.5 = 27,632 W/m²
• h_rad = ε·σ·(T₁ + T₂)·(T₁² + T₂²) = 0.85 × 5.67×10⁻⁸ × 1171.30 × 8.5128×10⁵ = 48.1 W/(m²·K)

That single 2.5 m² panel sheds roughly 69 kilowatts by radiation alone. For perspective, natural convection from a vertical surface in still air runs an h_conv of perhaps 5–15 W/(m²·K) — three to ten times smaller than the 48 W/(m²·K) radiation coefficient we just computed. The radiation term is carrying most of the heat, which is exactly why furnace and boiler heat balances are radiation-dominated and why you cannot comfortably stand near a furnace wall even when the air around you feels only warm. Had we mistakenly used 600 and 25 in Celsius, the same formula would have returned roughly 11 kW — under a sixth of the true value — a vivid illustration of the absolute-temperature trap.

Common mistakes to avoid

• Using Celsius or Fahrenheit. The T⁴ term demands absolute temperature. Convert to kelvin first, every time — this single error can throw the answer off by a factor of four or more.
• Computing (T₁ − T₂)⁴ instead of T₁⁴ − T₂⁴. Raise each temperature to the fourth power, then subtract. The two are not interchangeable and the difference is enormous.
• Assuming ε = 1 for everything. Only a true black body has unit emissivity. Polished metals sit at 0.02–0.1 and radiate ten to fifty times less than the black-body value — using ε = 1 for a shiny surface vastly overstates the heat transfer.
• Forgetting the sign of the exchange. If T₂ exceeds T₁, the term (T₁⁴ − T₂⁴) goes negative and the surface absorbs heat rather than emitting it. The sign carries the physics — a cool surface under a hot ceiling gains energy.
• Confusing h_rad with q. The linearized coefficient h_rad (W/m²·K) folds radiation into a resistance network alongside convection; the heat flux q (W/m²) is the actual energy crossing a unit area. They have different units and different jobs.
• Ignoring the view factor for finite geometries. The εσA(T₁⁴ − T₂⁴) form assumes the surroundings fully enclose the surface. When two finite surfaces face each other, a view factor below one reduces the exchange, sometimes dramatically.

The takeaway

Thermal radiation reduces to one durable equation, Q = εσA(T₁⁴ − T₂⁴), with σ = 5.67×10⁻⁸ W/(m²·K⁴) and temperatures in kelvin always. The fourth-power dependence is the whole story: it makes radiation negligible near room temperature and overwhelming in furnaces, where doubling the absolute temperature multiplies the emitted power sixteenfold. Emissivity sets how close a real surface comes to the black-body ideal — near 0.05 for polished metal, near 0.95 for paint, oxide, and most non-metals — and getting it wrong scales the answer linearly. Convert to absolute temperature, raise to the fourth power before subtracting, respect the emissivity of the actual finish, and reach for the view factor when surfaces are finite. Linearize into h_rad when you need to add radiation to convection, and compare all three modes before deciding which one governs. Run the worked example above through the radiation calculator and you should land on the same 69 kW, 27.6 kW/m², and 48 W/(m²·K) — a sign your method, and your unit handling, are sound.