Heat Transfer & Insulation · Guide · glidepathscience.org
GuideHeat Transfer & Insulation
Contents
This Guide explains the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round. Part 1 prepares the Intro track and Part 2 the Advanced track; each takes about twelve minutes. The worked examples use glass panes, attics, kitchens and spacecraft, not the sim’s own problems.
Part 1 · The essentials
Math you’ll need: algebra, unit conversions and fourth powers.
Heat flows downhill in temperature
Heat is energy moving because of a temperature difference, and it always moves from hot to cold. A room cools on a winter night because energy leaves it, not because “cold” comes in; the flow stops only when the temperatures are equal. A furnace or a stove does not fight cold, it replaces energy as fast as it leaves.
Heat moves three ways. Conduction passes energy through matter that stays put: the wall of a mug, the glass of a window. Convection carries it with a moving fluid: air warmed at a radiator rises and is replaced by cooler air. Radiation carries it as infrared light and needs no matter at all, which is how the Sun’s energy reaches us. Most real objects lose heat all three ways at once.
Conduction and Fourier’s law
In a solid, a particle jiggling hard bumps its neighbors and sets them jiggling harder; energy passes along while every particle stays in its place. In metals free electrons carry most of it, which is why metals conduct so well. The rate through a slab of area $A$ and thickness $L$, with its faces $ΔT$ apart, is Fourier’s law:
$P = kAΔT/L$
The thermal conductivity $k$ is a property of the material, in W/(m·K). Typical values: copper 401, steel 50, glass 1.0, brick 0.72, wood 0.12, fiberglass 0.040, still air 0.025. Good insulators are mostly trapped, still air. Twice the area doubles the flow; twice the thickness halves it.
A steel railing feels colder than a wooden bench on the same cold morning, though both are at the air’s temperature: steel draws heat out of your hand hundreds of times faster, and what you feel is your skin cooling.
R-values, layers in series and the temperature profile
Insulation is rated by its R-value, the thermal resistance of one square meter: $R = L/k$, in m²·K/W. (In the United States the label uses ft²·°F·h/BTU; one U.S. unit is 0.1761 m²·K/W, so an R-13 batt is about 2.3 m²·K/W.) The heat flux through one square meter is then $q = ΔT/R$.
Heat leaving a room crosses every layer of the wall one after another, and the thin films of still air clinging to each face too: about 0.13 m²·K/W inside and 0.04 outside, where wind thins the film. In series, resistances add:
$R_{tot} = R_{in} + ΣL/k + R_{out}$ $q = ΔT/R_{tot}$
The same $q$ crosses every layer, so each layer’s temperature drop is $qR$. The layer with the largest R takes most of the drop: the profile is steep across the insulation and nearly flat across brick or drywall. Mass and thickness alone do not insulate; a low conductivity does.
Windows and air films
A single pane is 4 mm of glass, $R$ = 0.004 m²·K/W, almost nothing. Its two air films, 0.17 m²·K/W together, are nearly all of its resistance, so its U-value, $U = 1/R_{tot}$, is about 5.7 W/(m²·K). A double pane traps a layer of gas between the panes and roughly halves the loss.
Convection
Air touching a warm surface warms, becomes less dense and rises, and cooler air takes its place. The rate is written with a convection coefficient $h$:
$P = hAΔT$
$h$ is about 5–10 W/(m²·K) in still air and several times more in a breeze. Wind does not make air colder; it keeps replacing the warmed layer at your skin, which is why a windy day feels colder at the same temperature.
Radiation and the vacuum flask
Every surface radiates infrared, $εσAT^4$, with $T$ in kelvins, $σ$ = 5.67 × 10⁻⁸ W/(m²·K⁴) and an emissivity ε between 0 and 1. It also absorbs radiation from its surroundings at $T_s$, so its net loss is
$P = εσA(T^4 − T_s^4)$
Dark and painted surfaces have ε near 0.9 in the infrared, even white paint; polished metal has only a few percent. A space blanket and a flask’s silvering work by having a small ε.
A vacuum flask blocks every path: the vacuum between its two walls stops conduction and convection across the gap (radiation crosses a vacuum easily), the silvered faces cut radiation to a few percent, and the stopper stops warm moist air leaving the top. What remains leaks mainly through the stopper and the glass neck.
Part 2 · Going further
Math you’ll need: algebra and exponentials; derivatives, and one partial differential equation in the last two sections.
The gradient form, thermal circuits and U-values
At a point, Fourier’s law reads $q = −k\,dT/dx$. In steady state with nothing stored or generated, energy conservation gives $dq/dx = 0$: the same flux crosses every layer, and the gradient $−q/k$ is steepest where $k$ is smallest. For an area $A$, each layer is a resistance $R/A$ in K/W and the flow is $P = ΔT/Σ(R_i/A)$: Ohm’s law with temperature for voltage. Codes rate assemblies by $U = 1/R_{tot}$.
Thermal bridging and its bounds
Studs interrupt a wall’s insulation, and paths side by side add their conductances, not their resistances. The parallel-path estimate treats the stud and the insulated cavity as separate strips, $U = fU_{stud} + (1 − f)U_{cav}$, with $f$ the framing fraction (often 0.15–0.25 once plates and headers are counted). It forbids heat to cross between the strips, so it is an upper bound on R. The isothermal-planes estimate combines the cavity layer’s two materials in parallel and puts everything else in series, letting heat cross sideways freely: a lower bound. The real wall, solved in two dimensions as in the Wall Lab, lies between: heat in the drywall and sheathing turns toward the stud. A steel stud is a far stronger bridge: in the Wall Lab, steel studs at 406 mm cut a 140 mm fiberglass wall from 3.84 to about 1.9 m²·K/W, and continuous foam outside the sheathing wins much of that back.
Films, linearized radiation and combined surfaces
A surface film is convection and radiation together. Radiation can be written as a coefficient when temperatures are close: $εσ(T^4 − T_s^4) ≈ h_r(T − T_s)$ with $h_r = 4εσT_m^3$, about 5 W/(m²·K) near room temperature, accurate to about $(ΔT/2T_m)^2$. Convection and radiation from one surface act in parallel, so their coefficients add, $h_c + h_r$. Because a single pane’s inside film is three quarters of its resistance, its inside face sits close to the outdoor temperature: $T_{si} = T_{in} − qR_{in}$, below freezing on a cold night, so frost forms on the inside.
Radiation between plates, and shields
Two large parallel gray plates exchange $q = σ(T_1^4 − T_2^4)/(1/ε_1 + 1/ε_2 − 1)$. A thin shield between them adds another gap, and with it another term $1/ε_a + 1/ε_b − 1$. With every emissivity equal, $N$ shields divide the flux by $N + 1$; a shiny shield adds $2/ε_s − 1$, a large number, and cuts it far more. Spacecraft multilayer blankets stack many such foils in vacuum.
Newton’s law of cooling and the Biot number
If a body stays at one temperature throughout, $ρcV\,dT/dt = −hA(T − T_∞)$, and its excess temperature decays as $e^{−t/τ}$ with $τ = ρcV/(hA)$; for a sphere $V/A = D/6$. The body stays uniform when its inside conducts heat far faster than its surface sheds it: the Biot number $Bi = hL_c/k$, with $L_c = V/A$, should be under about 0.1. Small metal parts in air qualify; a potato in boiling water does not.
Transient conduction and thermal mass
Inside a solid, $∂T/∂t = α\,∂^2T/∂x^2$ with diffusivity $α = k/(ρc)$. A change at one face soaks a distance $L$ in about $t ≈ L^2/α$: twice as thick takes four times as long. A daily swing at the face travels in as a damped wave: its amplitude falls as $e^{−x/δ}$ and it arrives late by $(x/δ)P/(2π)$, with $δ = \sqrt{αP/π}$ for period $P$. Heavy walls therefore shave the peaks off a hot day and deliver them, reduced, in the evening.
Convection in a window’s gap
Across a narrow gas gap heat goes by conduction, $k/L$, and by radiation, which does not depend on the width. As the gap widens, conduction falls, until the gas starts to rise along the warm pane and sink along the cold one. The Rayleigh number measures that drive; the Nusselt number $Nu$, the factor by which motion multiplies conduction, stays near 1 below a few thousand and then grows (the sim uses Wright’s 1996 correlation). So an air gap is best near 18 mm, argon near 15 mm, and a low-e coating, which cuts the radiation about twentyfold, helps more than any width.
Reference
Common mistakes
Part 1
- Saying “cold flows in”: energy flows out, from hot to cold.
- Leaving the thickness in centimeters or millimeters in $kAΔT/L$ or $L/k$.
- Thinking a heavy or thick layer must insulate: R is $L/k$.
- Forgetting the air films, which are most of a single pane’s resistance.
- Using degrees Celsius in $T^4$, or forgetting the surroundings’ $T_s^4$.
- Thinking a vacuum blocks radiation.
Part 2
- Averaging R-values of paths side by side: average the U-values.
- Using the lumped τ when $Bi$ is large.
- Taking $V/A$ of a sphere as $D$ or $D/2$; it is $D/6$.
- Expecting a wider window gap always to help.
The equation sheet
| Part 1 | |
|---|---|
| Fourier’s law | $P = kAΔT/L$ |
| R-value | $R = L/k$ (per m²) |
| in series | $R_{tot} = R_{in} + ΣL/k + R_{out}$ |
| flux, power | $q = ΔT/R_{tot}$, $P = qA$ |
| convection | $P = hAΔT$ |
| radiation | $P = εσA(T^4 − T_s^4)$ |
| Part 2 | |
|---|---|
| U-value | $U = 1/R_{tot}$ |
| parallel paths | $U = fU_{stud} + (1 − f)U_{cav}$ |
| inside surface | $T_{si} = T_{in} − qR_{in}$ |
| linear radiation | $h_r = 4εσT_m^3$ |
| plates, shields | $q = σ(T_1^4 − T_2^4)/Σ(1/ε_a + 1/ε_b − 1)$ |
| lumped, Biot | $τ = ρcV/(hA)$, $Bi = hL_c/k$ |
| diffusion | $t ≈ L^2/α$, $α = k/(ρc)$ |
$P$ heat flow (W), $q$ heat flux (W/m²), $k$ thermal conductivity (W/(m·K)), $A$ area, $L$ thickness, $R$ R-value per square meter (m²·K/W), $U$ U-value, $h$ convection coefficient, $ε$ emissivity, $σ$ = 5.670 × 10⁻⁸ W/(m²·K⁴), $T_s$ surroundings’ temperature, $f$ framing fraction, $T_m$ mean temperature (K), $ρ$ density, $c$ specific heat, $V$ volume, $L_c = V/A$, $α$ diffusivity, $τ$ time constant.
Further reading
- OpenStax, University Physics Volume 2, Chapter 1, §1.6 “Mechanisms of Heat Transfer” (CC BY-NC-SA 4.0).
- J. L. Wright, “A correlation to quantify convective heat transfer between vertical window glazings,” ASHRAE Transactions 102(1) (1996).
- This sim’s Field Manual (H): the model and what it leaves out.
Heat Transfer & Insulation v1.0 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org