Convection and Heat Transfer Coefficients
Newton’s law of cooling, Reynolds, Prandtl and Nusselt numbers, forced flow in pipes and over surfaces, and natural convection.
Newton’s law of cooling
When a fluid moves past a surface at a different temperature, heat flows between them at a rate set by the convection coefficient h. It lumps together all the complicated physics of the thin boundary layer where the fluid’s speed and temperature change.
Q̇ = h A (Ts − T∞)
h is not a material property. It depends on the fluid, its speed, the shape and size of the surface and whether the flow is laminar or turbulent. For a real problem you calculate it from a correlation for the Nusselt number, or take it from a table of typical values.
Reynolds, Prandtl and Nusselt numbers
- Reynolds number Re = ρ v L ÷ μ: the ratio of inertia to viscosity. It decides whether the flow is laminar or turbulent (the same Re as in Continuity, Bernoulli and the Energy Equation).
- Prandtl number Pr = μ cp ÷ k: how quickly momentum spreads compared with heat, a property of the fluid.
- Nusselt number Nu = h L ÷ k: the dimensionless coefficient. It says how much better convection is than pure conduction through the same fluid layer.
| Fluid | Prandtl number |
|---|---|
| Air (20 °C) | 0.71 |
| Water (20 °C) | 7.0 |
| Water (80 °C) | 2.2 |
| Engine oil (40 °C) | about 1 000 |
| Liquid metals | 0.01 to 0.03 |
Correlations have the form Nu = C Rem Prn: find Re and Pr, calculate Nu, and then h = Nu k ÷ L. The fluid properties are taken at the film temperature (the average of surface and fluid) or the bulk temperature, as the correlation states.
Forced convection in a pipe
For fully developed turbulent flow in a smooth pipe (Re above 10 000, 0.6 < Pr < 160, length over 10 diameters) the Dittus–Boelter equation is the standard:
Nu = 0.023 Re0.8 Prn
Worked example: water in the DN80 line
- The 10 L/s in the 77.9 mm pipe (from Pipe Friction) has Re = 162 784; water at 20 °C has Pr = 7.01 and k = 0.597 W/m·K.
- Heating the water: Nu = 0.023 × 162 7840.8 × 7.010.4 = 740, so h = Nu k ÷ D = 5 673 W/m²·K.
- With the pipe wall 20 K hotter than the water, each meter transfers h π D ΔT = 27.77 kW. At half the flow, h falls with v0.8 to about 3 258 W/m²·K.
Worked example: air in a duct
- 5 m/s of air at 20 °C in a 200 mm duct: Re = 66 412, Pr = 0.71, k = 0.0258 W/m·K.
- Nu = 0.023 × Re0.8 × Pr0.4 = 144, so h = 18.6 W/m²·K: 304 times lower than water in a pipe of similar size, which is why air-cooled heat exchangers are so much bigger than water-cooled ones.
Forced convection over surfaces
- Flat plate, laminar (ReL below 5 × 10⁵): NuL = 0.664 ReL½ Pr⅓ (average over the plate length L). Turbulent: NuL = 0.037 ReL0.8 Pr⅓.
- Cylinder in crossflow (Churchill–Bernstein): NuD = 0.3 + 0.62 Re½ Pr⅓ ÷ [1 + (0.4 ÷ Pr)⅔]¼ × [1 + (Re ÷ 282 000)⅝]⅘, valid for Re Pr above 0.2.
- Sphere, tube banks, impinging jets have their own correlations in the heat transfer texts; use the one for your geometry.
Worked example: a hot plate and a hot rod in a fan’s air stream
- Plate: air at 3 m/s over a 0.5 m × 0.5 m plate (flow along 0.5 m) at 60 °C in 20 °C air. ReL = 99 619 (laminar), Nu = 0.664 × 316 × 0.89 = 187, h = 9.6 W/m²·K. One side loses h A ΔT = 96 W.
- Cylinder: a 25 mm rod in air at 5 m/s: Re = 8 302, Nu = 48.3, h = 49.9 W/m²·K.
- In both cases h is about 15 to 30 W/m²·K: typical for forced air at a few m/s. To raise it, increase the speed (h grows roughly as v0.5 to v0.8) or use a better fluid.
Natural convection
With no fan, warm fluid next to a hot surface is lighter and rises, drawing cooler fluid in: natural (free) convection. The driving number is the Rayleigh number, Ra = g β ΔT L³ ÷ (ν α), with β = 1 ÷ T for a gas. For a vertical plate over the whole range the Churchill–Chu equation works:
NuL = { 0.825 + 0.387 Ra1/6 ÷ [1 + (0.492 ÷ Pr)9/16]8/27 }²
Worked example: a vertical panel in still air
- A 0.3 m high panel at 50 °C in 20 °C air (film temperature 35 °C; ν = 16.45 mm²/s, α = 23.29 mm²/s, Pr = 0.71).
- Ra = g β ΔT L³ ÷ (ν α) = 9.807 × (1 ÷ 308.15) × 30 × 0.3³ ÷ (0.00001645 × 0.00002329) = 67 297 071.
- Nu = 54.2 and h = Nu k ÷ L = 4.9 W/m²·K: the usual 3 to 8 W/m²·K for air at modest temperature differences. Radiation adds about the same again (see Thermal Radiation), so total h is about 10 to 12.
Typical values and boiling and condensation
Boiling and condensation involve a change of phase, which moves a great deal of energy (the latent heat) at nearly constant temperature, so h reaches 5 000 to 100 000 W/m²·K. Nucleate boiling on a heated surface follows a curve with a peak (the critical heat flux, about 1.1 MW/m² for water at atmospheric pressure) beyond which a vapor film forms and the surface temperature jumps: the “burnout” that destroys boilers and heaters. Condensing steam on a clean tube is dominated by the film of condensate.
Related
- Previous: Conduction: Walls, Pipes and Insulation. Next: Thermal Radiation.
- Fluid Mechanics and Pipe Flow: Reynolds number, velocity and pressure drop (the price of a higher h).
- Heat Sink and Thermal Resistance Calculator.
- Heat Transfer: the whole guide.
FAQ
Why is h for air so much lower than for water?
Air has a low conductivity (0.026 against 0.6 W/m·K), a low density and a low specific heat, so it carries less heat away for the same speed. Water in a pipe typically reaches 1 000 to 10 000 W/m²·K, air 10 to 100.
Does a faster flow always help?
Up to a point: h rises roughly as v⁰·⁸ in turbulent pipe flow but the pressure drop and pumping power rise as v² to v³. For electronics cooling, the fan noise limits the speed.
Where do I take the fluid properties?
At the film temperature for external flow and natural convection, and at the bulk (mean) temperature for flow in a pipe, as the correlation states. For most air and water work the variation is small.
How accurate are these correlations?
Typically ±20 percent, sometimes ±30. That is enough for sizing, but add margin, and test or measure where a temperature limit is critical.
References
- Incropera, F. P., et al., Fundamentals of Heat and Mass Transfer, Wiley (Dittus–Boelter, flat plate, Churchill–Bernstein, Churchill–Chu).
- Churchill, S. W. and Chu, H. H. S., “Correlating equations for laminar and turbulent free convection from a vertical plate”, Int. J. Heat Mass Transfer, 1975; Churchill, S. W. and Bernstein, M., J. Heat Transfer, 1977.
- Dittus, F. W. and Boelter, L. M. K., Univ. of California Publications in Engineering, 1930.
- Çengel, Y. A. and Ghajar, A. J., Heat and Mass Transfer, McGraw-Hill; VDI Heat Atlas.
Last reviewed: October 2026. Correlations are accurate to about 20 percent.
