The heat transfer coefficient calculator above solves for h, the single number that stands between a surface and the fluid moving past it. Everything about the boundary layer — its thickness, its velocity profile, whether it is laminar or turbulent, whether the flow is driven by a pump or by buoyancy — is compressed into that one coefficient, and Newton's law of cooling then reads simply Q = hAΔT.
Arb Digital builds free physics calculators that each own one step of a chain. This page is the step that produces h. The live heat transfer calculator consumes it: in convection mode you type h in and it returns the heat flow. The Nusselt number calculator and the Biot number calculator both consume it too, one to express it as a dimensionless ratio against conduction in the fluid and the other against conduction inside the solid. And the thermal conductivity calculator owns a different quantity entirely: k is a genuine material property, while h is not.
What This Heat Transfer Coefficient Calculator Does
It works in two directions, because h is arrived at in two quite different ways in practice.
The experimental route takes a measured heat flow, the surface area it crossed and the temperature difference that drove it, and divides. This is the honest way to get h for a real piece of equipment, and it is what a commissioning test or a laboratory rig produces. The answer describes that apparatus at that condition and nothing else.
The predictive route takes a Nusselt number from a published correlation and converts it into h using the fluid's thermal conductivity and the characteristic length that the correlation was defined on. This is how a designer gets h before anything is built. The tool then reports the heat flux, the convective thermal resistance, and h converted into the imperial units still used in much of the HVAC and process world.
How to Use It
- Pick the direction. From a measured heat flow if you have test data, from a Nusselt number if you are designing.
- Enter the wetted area, not the footprint. A finned surface has an area several times its projected one, and using the wrong figure changes h by exactly that factor.
- Use the bulk fluid temperature. Not the film temperature and not a temperature taken inside the boundary layer. The definition of h is tied to the bulk value.
- In Nusselt mode, match the characteristic length to the correlation. A correlation defined on diameter and a length taken as radius differ by a factor of two in the answer.
- Read the convective resistance if you are building a network. It is 1 ÷ hA, and it adds in series with the conduction resistances either side.
The Formula: How h Is Calculated
Newton's law of cooling defines the coefficient rather than deriving it: Q = hA(Ts − T∞). Rearranged,
h = Q ÷ [A(Ts − T∞)]
with h in watts per square metre per kelvin. Because the expression uses a temperature difference, Celsius and kelvin give the same number and no conversion is needed. From a Nusselt number the relation is h = Nu × k ÷ L, where k is the conductivity of the fluid and L is the characteristic length of the geometry. The convective thermal resistance is R = 1 ÷ (hA), which is what lets a convection step sit in the same series network as a conduction step.
MIT OpenCourseWare's Intermediate Heat and Mass Transfer, from the Mechanical Engineering department, is a good open reference for how correlations for Nusselt number are built and where each one is valid. OpenStax University Physics Volume 2, section 1.6 on mechanisms of heat transfer, sets out conduction, convection and radiation as distinct mechanisms and is explicit that convection is treated qualitatively at that level precisely because the coefficient is not a simple material constant. Georgia State University's HyperPhysics page on heat transfer gives the same three-mechanism framing with the conduction relation worked out.
Work the defaults by hand. A heat flow of 2,500 W crossing 1.5 m² with a surface at 90 °C and the fluid at 20 °C gives a temperature difference of 70 K, so h = 2,500 ÷ (1.5 × 70) = 2,500 ÷ 105 = 23.81 W/m²·K. The heat flux is 2,500 ÷ 1.5 = 1,666.7 W/m². The convective resistance is 1 ÷ (23.81 × 1.5) = 0.0280 K/W, which cross-checks against ΔT ÷ Q = 70 ÷ 2,500 = 0.0280 K/W exactly. Converting to imperial units with 1 BTU/h·ft²·°F = 5.678263 W/m²·K gives 4.193 BTU/h·ft²·°F.
h Is Not A Material Property, And That Changes Everything
Thermal conductivity belongs to a material. Look up the conductivity of copper and you have a number that is true wherever the copper is. The heat transfer coefficient has no such standing. It belongs to a configuration: this surface, this fluid, this speed, this orientation, this roughness, at this temperature difference.
The range is extraordinary. Still air against a vertical panel gives roughly 2 to 10 W/m²·K. Air blown hard across the same panel gives 25 to 250. Water in a tube gives hundreds to many thousands. Boiling and condensing run into the tens of thousands. That is four orders of magnitude for the same physical quantity, which is why no table of h values is trustworthy without its full context attached.
The practical consequence is that h is the dominant uncertainty in most thermal calculations. Conductivities are known to a per cent or two; a convection coefficient from a published correlation carries fifteen to twenty-five per cent uncertainty in the best cases and more when the geometry is not quite what the correlation assumed. Designing to a single value of h without a margin is the standard way that thermal designs go wrong.
Natural Convection Breaks Newton's Law
The relation Q = hAΔT looks linear, and in forced convection it very nearly is: at fixed fan speed, h barely moves as the temperature difference changes, so heat flow really is proportional to ΔT. In natural convection it is not, and this catches people out constantly.
Buoyancy is what drives natural convection, and buoyancy comes from the temperature difference itself. Raise ΔT and the flow speeds up, so h rises too. For a vertical plate in laminar natural convection h scales roughly as ΔT to the power of a quarter, which makes the heat flow go as ΔT to the power of five quarters rather than the first power.
So an h measured at one temperature difference does not transfer to another. A panel tested at 70 K above ambient will show a different coefficient at 20 K above ambient, and using the first figure for the second case underestimates nothing and overestimates the cooling. The tool reports h at the condition you enter, and that is the only condition it describes.
The Small Coefficient Controls The Whole Path
Convection resistances add in series with conduction ones, and series resistances are dominated by the largest term. Since resistance is 1 ÷ hA, the largest resistance sits where h is smallest — which is almost always on the air side.
Take a heat exchanger with water inside at h = 3,000 W/m²·K and air outside at h = 30 W/m²·K over equal areas. The overall coefficient is 1 ÷ (1/3,000 + 1/30) = 29.7 W/m²·K. Doubling the water-side coefficient to 6,000 lifts the overall figure to 29.85, a gain of half a per cent. Doubling the air-side coefficient to 60 lifts it to 58.8, very nearly double.
This is why air-side surfaces are finned and liquid-side ones are not. Fins multiply the area on the weak side where it buys something, and would be a waste of metal on the strong side. Use the heat transfer calculator to run the whole path once you have both coefficients, the LMTD calculator for the mean temperature difference across an exchanger, and the heat loss calculator for building-envelope work where the same logic applies to walls and windows.
Getting h From A Correlation Without Getting It Wrong
Every Nusselt correlation comes with three things attached, and dropping any of them invalidates the result. First, a characteristic length: diameter for a tube, plate length for a flat plate, diameter for cross-flow over a cylinder. Second, a temperature at which the fluid properties are to be evaluated, which is often the film temperature rather than the bulk. Third, a range of validity in Reynolds and Prandtl number.
The Reynolds number calculator tells you whether you are in the laminar or turbulent branch, and the Prandtl number calculator gives the fluid parameter that almost every correlation depends on. Feed those into the correlation, get a Nusselt number, and bring it here to turn it into an h. The Nusselt number calculator handles the dimensionless step directly.
Two further checks are worth making. The Nusselt number is the ratio of convective to conductive transport in the fluid, so a value below one is a warning that something is wrong — convection cannot be worse than pure conduction. And once you have h, the Biot number calculator tells you whether the solid can be treated as isothermal, which decides whether a lumped-capacitance transient analysis is legitimate at all. If you need to move the flux figure between unit systems, the heat flux converter does that conversion.
Arb Digital builds free tools like this one because useful pages earn attention. If you want tools, calculators or content built for your own audience, we can help.
Browse All Free Tools Talk to Arb DigitalCommon Mistakes to Avoid
- Treating h as a property of the fluid — it belongs to the whole arrangement. The same water can give 100 or 10,000 W/m²·K depending on how it is moving and what it is moving past.
- Reusing an h measured at a different temperature difference — in natural convection the coefficient itself depends on ΔT, so the relation is not linear and the value does not transfer.
- Using the projected area instead of the wetted area — a finned surface has several times the area of its footprint, and the error goes straight into h at full size.
- Mismatching the characteristic length to the correlation — a Nusselt number defined on diameter and converted using a radius is wrong by a factor of two, with nothing in the result to reveal it.
- Improving the wrong side of an exchanger — the overall coefficient is controlled by the smallest h, so effort spent on the already-strong liquid side buys almost nothing.
Related Free Tools From Arb Digital
Once you have h, the heat transfer calculator runs conduction, convection or radiation for a whole surface, and the heat loss calculator covers building envelopes. The Nusselt number and Biot number calculators are the two dimensionless groups that consume h, while the Reynolds number and Prandtl number calculators supply what the correlations need. For the material property rather than the surface coefficient use the thermal conductivity calculator, for exchanger driving temperature the LMTD calculator, and for unit changes the thermal conductivity converter. Everything Arb Digital publishes is on the free online tools hub.
Frequently Asked Questions
It is the constant in Newton's law of cooling that links heat flow to surface area and temperature difference, measured in watts per square metre per kelvin. It summarises everything about the boundary layer between a surface and the fluid moving past it into a single number.
No. Unlike thermal conductivity, h depends on the geometry, the flow speed, the orientation, the surface condition and the temperature difference as well as the fluid. The same fluid can give coefficients four orders of magnitude apart in different configurations.
Divide the measured heat flow by the product of the wetted surface area and the difference between surface and bulk fluid temperature. Because the expression uses a temperature difference, Celsius and kelvin give the same number and no unit conversion is needed.
Multiply the Nusselt number by the fluid's thermal conductivity and divide by the characteristic length the correlation was defined on. Using a different length than the correlation intended is the most common error in this direction and there is nothing in the result to flag it.
Roughly 2 to 10 watts per square metre per kelvin for still air, 25 to 250 for forced air, hundreds to thousands for liquids in tubes, and tens of thousands for boiling or condensing. Any quoted value is meaningless without the geometry and flow conditions that produced it.
Because convective resistances add in series and resistance is one over h times area, so the smallest coefficient makes the largest resistance and dominates the total. That is why air-side surfaces get fins and liquid-side surfaces usually do not.
In forced convection barely at all, so heat flow is close to linear in the temperature difference. In natural convection it does, because buoyancy is driven by that same difference, and for a vertical plate the coefficient rises roughly as the difference to the power of a quarter.
A coefficient from a published correlation typically carries fifteen to twenty-five per cent uncertainty even when the geometry matches the correlation well, and more when it does not. It is normally the largest single uncertainty in a thermal calculation, so designs need margin on it.
This tool is provided for educational and study use. It implements the definition of the convective heat transfer coefficient and the Nusselt conversion only; it does not select or evaluate correlations, account for radiation in parallel with convection, or handle phase change, so treat its output as a physics result rather than a design-grade figure.