---
title: Thermal Processes
module: Thermodynamics
moduleNumber: 8
lessonNumber: 4
order: 804
summary: >
  Heat rarely sits still: it stretches solids, pushes real gases off their ideal
  isotherms, and leaks across walls by conduction, convection, and radiation. Each
  behavior becomes a number a designer can use. Thermal expansion sets the gaps in a
  bridge and the stress in a clamped rod; the van der Waals equation and a phase
  diagram fix when $pV=nRT$ or a latent-heat term applies; Fourier's law, Newton
  cooling, and Stefan–Boltzmann radiation give the rate of heat flow. We assemble these
  into thermal-resistance networks and transient time constants, then mark where
  contact resistance, phase change, or a hidden thermal bridge breaks the simple model.
topics: [Thermodynamics]
draft: false
sources:
  - book: Tipler & Mosca
    ref: "Ch. 20 — Thermal Properties and Processes; §§20-1–20-4"
---

## Thermal expansion and constrained materials

Most solids expand when their temperature rises because the average separation of
their atoms increases. Over a limited temperature interval, the linear change of a
uniform rod is described by $\Delta L=\alpha L_0\Delta T$, where $\alpha$ is the
coefficient of linear expansion, $L_0$ is the initial length, and $\Delta T$ is the
temperature change. The formula concerns a length difference, not the final length;
the final value is $L_f=L_0+\Delta L$. Celsius-degree and kelvin differences have
equal size, so either unit can be used for $\Delta T$.

A free rod experiences little mechanical resistance while it expands. A constrained
rod requires a stress calculation: supports supply forces
and internal stress develops. Bridges, rail joints, piping, electrical lines, and
precision instruments therefore include expansion gaps, flexible couplings, or
carefully selected material pairs. The relevant design question is often the
differential expansion of two connected parts. A bimetallic strip bends because
the material with larger expansion coefficient attempts a greater length change
over the same temperature interval.

An isotropic solid has area change approximately $2\alpha A_0\Delta T$ and volume
change approximately $3\alpha V_0\Delta T$ when the fractional change is
small. The volume expansion coefficient is commonly denoted by $\beta$, with
$\beta\simeq3\alpha$ for an isotropic solid. Liquids and gases do not retain a
fixed shape, so volume expansion is the appropriate measure. A liquid in a filled rigid
container can develop large pressure changes when heated because the container and
liquid usually have different volume-expansion responses.

Thermal-expansion formulas are local approximations. Coefficients can vary with
temperature, materials can change phase, and an object can have an uneven
temperature field. A metal tube heated at one end does not have one single
temperature or one single expansion. In such a case, the local strain must be
integrated along the temperature profile. The simple uniform-temperature expression
remains valuable when thermal equilibration is fast compared with the mechanical
measurement and the range is small enough that a constant coefficient is accurate.

## Nonideal gases and real process paths

The ideal-gas equation treats molecules as point particles with no sustained
intermolecular forces. This approximation is effective at low density and away
from condensation. At higher density, a molecule occupies a non-negligible volume
and attractive forces alter the momentum delivered to a container wall. The
van der Waals equation introduces both corrections:

$$
\left(p+\frac{an^2}{V^2}\right)(V-nb)=nRT.
$$

The parameter $b$ reduces the volume available to molecular centres, while $a$
accounts for attraction that lowers the observed wall pressure. Both parameters
depend on the gas species. The equation is a model rather than an exact universal
law, but it explains why ideal-gas behavior fails most strongly at high pressure
and low temperature.

Real-gas isotherms can pass through a liquid--vapor coexistence region. Below the
critical temperature, compression at suitable temperature can produce a large
volume decrease while pressure changes little because gas condenses into liquid.
Above the critical temperature, no distinct liquid--vapor boundary remains and the
fluid can be compressed continuously from gas-like to liquid-like density. The
critical point terminates the coexistence curve and defines a characteristic
critical temperature and pressure for each substance.

A process calculation must identify whether the material remains a single phase.
Applying $pV=nRT$ through a condensation interval can predict an impossible smooth
volume path because the actual system changes phase. Conversely, using a latent-heat
term where the sample remains a supercritical fluid is equally inappropriate. A
pressure, volume, and temperature record specifies which material model and energy
terms belong in the calculation.

$$
% caption: Real-gas isotherms differ from ideal hyperbolas near condensation. Below the critical temperature, a pressure interval can contain liquid and vapor together; above it, the density changes continuously.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0,0)--(6.2,0) node[right] {volume};
\draw[->,black] (0,0)--(0,3.6) node[above] {pressure};
\draw[acc,very thick] (.55,3.2) .. controls (.95,2.45) and (1.35,2.02) .. (1.7,1.9) -- (3.5,1.9) .. controls (4.25,1.72) and (4.9,1.2) .. (5.85,.78);
\draw[black,thick,dashed] (.55,2.55) .. controls (1.9,1.72) and (3.6,1.2) .. (5.85,.72);
\draw[black,dotted] (1.7,0)--(1.7,1.9);
\draw[black,dotted] (3.5,0)--(3.5,1.9);
\node[acc] at (1.35,3.4) {below critical};
\node[black] at (4.7,2.0) {above critical};
\node[black,below] at (2.6,0) {coexistence};
\end{tikzpicture}
$$

## Phase diagrams and process selection

A phase diagram maps the equilibrium phase of a substance as a function of pressure
and temperature. Lines on the diagram represent coexistence of two phases: solid
and liquid, liquid and vapor, or solid and vapor. The triple point is the unique
pressure and temperature at which all three phases coexist. The liquid--vapor line
ends at the critical point, beyond which the distinction between liquid and gas is
lost. A phase diagram is therefore a guide to process selection before an energy
calculation begins.

Crossing a coexistence line changes phase and ordinarily requires latent heat.
Moving within a one-phase region changes temperature, pressure, or volume without
changing phase. A path at constant pressure can cross a boiling line as heat is
added; a path at constant temperature can cross a condensation line as pressure is
raised. The direction of a path matters because heating, cooling, compression, and
expansion can traverse the same boundary in opposite directions and reverse the
sign of the associated heat transfer.

Water is unusual because its solid--liquid coexistence line slopes in the opposite
direction from that of many substances. Liquid water is denser than ice, so raising
pressure near the melting point favors the liquid phase. Most materials have a
solid phase denser than their liquid phase, giving the more familiar slope. The
diagram records equilibrium boundaries; metastable supercooled or superheated
states can persist temporarily when nucleation is delayed, but they are not the
equilibrium phase predicted by the plotted regions.

Process descriptions should state whether pressure is controlled, volume is fixed,
or the sample is in contact with a reservoir. A sealed rigid vessel can cross a
phase boundary during heating while pressure changes strongly. An open pot at
nearly atmospheric pressure can boil at almost fixed pressure while volume changes
and vapor escapes. These are physically different paths through the same phase
diagram and require different system boundaries for a first-law account.

$$
% caption: The pressure-temperature state fixes the equilibrium phase before a process is calculated. Heating or compression paths cross coexistence lines only when the relevant pressure and temperature conditions reach the boundary.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0,0)--(6.2,0) node[right] {temperature};
\draw[->,black] (0,0)--(0,3.6) node[above] {pressure};
\draw[acc,thick] (1.9,1.15)--(.6,.35);
\draw[acc,thick] (1.9,1.15)--(2.3,3.15);
\draw[acc,thick] (1.9,1.15) .. controls (3.0,1.55) and (4.2,2.1) .. (5.2,2.75);
\draw[fill=white,draw=black,line width=.4pt] (1.9,1.15) circle (2pt);
\draw[fill=white,draw=black,line width=.4pt] (5.2,2.75) circle (2pt);
\node[black] at (1.0,2.35) {solid};
\node[black] at (3.3,2.6) {liquid};
\node[black] at (4.75,1.05) {vapor};
\node[black] at (2.15,.72) {triple};
\node[black] at (4.55,2.98) {critical};
\draw[black,->,thick] (.7,1.62)--(4.6,1.62);
\node[black,below] at (1.05,1.62) {heating};
\end{tikzpicture}
$$

## Conduction, convection, and radiation

The three rate models for a steady one-dimensional wall with a convection boundary
and radiative surroundings are

$$
\dot Q_{\rm cond}=\frac{kA(T_H-T_C)}{L},
\qquad
\dot Q_{\rm conv}=hA(T_s-T_\infty),
\qquad
\dot Q_{\rm rad}=\epsilon\sigma A(T_s^4-T_{\rm env}^4).
$$

| Mode | Transport path | Quantity that controls the model | Common limitation |
| :--- | :--- | :--- | :--- |
| conduction | through a material | $kA/L$ | temperature field must be close to one-dimensional |
| convection | surface to moving fluid | $hA$ | $h$ changes with geometry, flow, and fluid state |
| radiation | between radiating surfaces | $\epsilon\sigma A$ | absolute temperature and view geometry are required |

For layers in series, thermal resistances add:

$$
R_{\rm th}=\sum_i\frac{L_i}{k_iA},
\qquad
\dot Q_{\rm cond}=\frac{T_H-T_C}{R_{\rm th}}.
$$

The convection coefficient is a flow-dependent correlation rather than a material
constant. Natural convection follows buoyancy-driven motion; a fan, pump, or wind
changes the boundary layer and usually changes $h$. Radiation crosses a vacuum and
becomes important as $T_s^4-T_{\rm env}^4$ grows. A hot pipe commonly has all three
paths in parallel. Its steady surface temperature satisfies the full rate balance;
a changing surface temperature also requires the first law and the pipe heat capacity.

$$
% caption: Conduction passes energy through a solid wall, convection carries it away in moving fluid, and radiation transfers energy by electromagnetic emission. Real equipment often has all three paths in parallel.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\fill[acc!12] (.2,.55) rectangle (1.5,2.55);
\draw[acc,thick] (.2,.55) rectangle (1.5,2.55);
\node[acc] at (.85,1.55) {hot side};
\draw[black,thick] (1.5,.55) rectangle (2.7,2.55);
\node[black] at (2.1,.95) {wall};
\draw[acc,->,thick] (1.6,1.95)--(2.6,1.95);
\node[acc] at (2.1,2.85) {conduction};
\draw[black,->,thick] (2.9,1.35) .. controls (3.7,2.35) and (4.15,2.2) .. (4.55,1.7);
\node[black] at (3.75,2.6) {convection};
\draw[acc,->,thick] (2.85,.85)--(4.75,.85);
\node[acc] at (3.8,1.12) {radiation};
\draw[black,thick] (4.95,.55)--(4.95,2.55);
\node[black] at (4.95,2.85) {surroundings};
\end{tikzpicture}
$$

### Expansion measurements and composite devices

Thermal expansion becomes a measurement problem when two materials are joined or
when a small dimensional change must be detected reliably. A bimetallic strip has
two bonded layers with different linear expansion coefficients. Heating gives each
layer the same temperature change, but the layer with larger coefficient attempts a
larger free length change. Because the layers remain bonded, neither reaches its
free length. The composite bends with the larger-expansion material on the outside
of the curve. Thermostats use this mechanical response to convert temperature
change into motion of an electrical contact.

The scale of differential expansion can be estimated before a detailed stress
calculation. Two $0.50\ \mathrm m$ strips are heated through
$80\ \mathrm K$, one with $\alpha=23\times10^{-6}\ \mathrm{K^{-1}}$ and one with
$\alpha=12\times10^{-6}\ \mathrm{K^{-1}}$. Their free length changes differ by
$(23-12)\times10^{-6}(0.50)(80)=0.44\ \mathrm{mm}$. That small mismatch is enough
to create visible curvature in a thin bonded strip. A precise curvature prediction
also requires layer thicknesses, elastic moduli, and the position of the neutral
axis; the simple difference gives the physical origin and an order-of-magnitude
check.

Expansion coefficients are measured by comparing a calibrated length change with a
known temperature interval. The instrument must distinguish the sample expansion
from expansion of its supports, sensor housing, and reference ruler. Differential
methods reduce this error by placing a reference material with known coefficient
beside the sample in the same temperature environment. If temperature is not
uniform, a single sensor reading may not represent the average strain-producing
temperature, and the resulting coefficient can be biased.

Design margins should use the largest expected temperature excursion rather than a
nominal operating point. A pipe with an expansion joint is not "loosely assembled";
the joint permits controlled displacement so that thermal strain does not
become damaging force. Conversely, a deliberately constrained component may use
thermal stress for clamping or actuation, but its allowable temperature range must
be set by yield strength and fatigue limits as well as by its expansion coefficient.

$$
% caption: A bonded bimetallic strip bends because its layers attempt unequal free expansions. The larger-expansion layer occupies the outside of the curve, converting a temperature change into a measurable displacement.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[acc,very thick] (.7,1.15) .. controls (2.2,2.35) and (3.9,2.35) .. (5.5,1.15);
\draw[black,very thick] (.7,.88) .. controls (2.2,2.08) and (3.9,2.08) .. (5.5,.88);
\draw[acc,->,thick] (1.2,2.95)--(1.2,1.78);
\draw[acc,->,thick] (4.8,2.95)--(4.8,1.78);
\node[acc] at (1.2,3.25) {warm};
\node[acc] at (4.8,3.25) {warm};
\node[acc] at (3.1,2.72) {larger expansion layer};
\node[black] at (3.1,.3) {smaller expansion layer};
\end{tikzpicture}
$$

**Calorimetry as a thermal-property measurement.**

Calorimetry determines an unknown heat capacity by placing it in a controlled
energy balance with materials whose heat capacities are known. The system boundary is
the entire insulated calorimeter: sample, water, cup, thermometer, and any stirrer
that changes temperature. Heat exchange among those components is internal to the
chosen system. If external heat transfer and boundary work are negligible, the sum
of all internal-energy changes is zero. The final equilibrium temperature is common
to the parts that remain in thermal contact.

A warm sample dropped into cooler water has the signed energy balance
$C_s(T_f-T_{si})+m_wc_w(T_f-T_{wi})+C_c(T_f-T_{ci})=0$. The sample heat capacity
$C_s$ can be solved after $T_f$ is measured. The signs emerge from temperature
differences: a warm sample has $T_f-T_{si}<0$, while initially cooler water and cup
have positive differences. Writing every term with its signed temperature change
is safer than assigning separate verbal labels such as "heat lost" and "heat
gained" after the fact.

> **Worked example.** A $0.120\ \mathrm{kg}$ copper sample with
> $c=385\ \mathrm{J\,kg^{-1}\,K^{-1}}$ at $90\ \mathrm{degree}$ is dropped into
> $0.200\ \mathrm{kg}$ of water at $20\ \mathrm{degree}$ in a cup of heat capacity
> $60\ \mathrm{J\,K^{-1}}$. The sample heat capacity is $C_s=46.2\ \mathrm{J\,K^{-1}}$
> and the water-plus-cup capacity is $896.8\ \mathrm{J\,K^{-1}}$. The balance
> $46.2(T_f-90)+896.8(T_f-20)=0$ gives $T_f\approx23.4\ \mathrm{degree}$, close to the
> water temperature because the water and cup carry far more heat capacity than the
> copper.

The calculation assumes no phase change, no chemical reaction, negligible external
leakage, and heat capacities that are nearly constant over the temperature range.
A result outside the interval bracketed by the initial sample and water temperatures
signals a missed component or sign error. In precision work, evaporation, probe
immersion depth, stirring work, and heat loss during transfer from a heating bath
can be comparable with the desired signal and must be measured or corrected.

### Phase-change energy along a heating path

Heating a material through a phase change requires a staged energy account. Each
single-phase interval uses a sensible-heat term $mc\Delta T$. Each phase transition
uses a latent-heat term $mL$ at nearly constant temperature and pressure. A heating
curve therefore contains sloped sections where temperature changes and plateaus
where energy changes molecular arrangement without changing temperature. The total
energy is the sum of the terms in their physical order along the path.

> **Worked example.** Converting $0.250\ \mathrm{kg}$ of ice at $-10\ \mathrm{degree}$
> to liquid water at $20\ \mathrm{degree}$ takes three terms. Warming the ice to the
> melting point: $(0.250)(2050)(10)=5.13\ \mathrm{kJ}$. Melting it:
> $(0.250)(333.5\ \mathrm{kJ\,kg^{-1}})=83.4\ \mathrm{kJ}$. Warming the meltwater:
> $(0.250)(4184)(20)=20.9\ \mathrm{kJ}$. The total is about $109\ \mathrm{kJ}$, and the
> latent term dominates even though it changes no temperature.

The phase inventory must be tested before a temperature equation is solved. If a
warm liquid does not supply enough energy to melt all added ice, the final state
contains both ice and liquid at the melting temperature. Continuing the calculation
as though all ice melted can produce a final temperature below the phase-change
temperature, which contradicts the coexistence condition. The correct result is
found by stopping at the plateau and calculating how much mass has changed phase.

Real processes can depart from an ideal heating curve through superheating,
supercooling, pressure variation, or heat loss to the surroundings. Those effects
change the measured path but do not remove the distinction between sensible and
latent energy. A process model should state the pressure, the initial phase, and
whether the sample is allowed to equilibrate at the transition. These conditions
select the appropriate latent heat and transition temperature from the phase diagram.

$$
% caption: A multistage heating path alternates sensible-heat intervals with latent-heat plateaus. The plateau can carry more energy than a large temperature rise because energy changes phase rather than temperature.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0,0)--(6.25,0) node[right] {energy input};
\draw[->,black] (0,0)--(0,3.5) node[above] {temperature};
\draw[acc,very thick] (.35,.4)--(1.2,1.0)--(2.7,1.0)--(3.5,1.85)--(5.0,1.85)--(5.95,2.5);
\node[black] at (.75,1.3) {solid warms};
\node[black] at (1.95,.75) {melting};
\node[black] at (2.95,2.2) {liquid warms};
\node[black] at (4.25,1.58) {boiling};
\node[black] at (5.5,2.78) {vapor};
\end{tikzpicture}
$$

**Temperature-scale measurement and calibration.**

A thermometer measures some reproducible physical property and maps it to a
temperature scale. In a constant-volume gas thermometer, the gas pressure is
proportional to absolute temperature when the amount of gas and volume are fixed:
$p/T=\text{constant}$. A calibration at a reference temperature fixes the ratio.

> **Worked example.** A constant-volume gas thermometer reads $50.0\ \mathrm{torr}$ at
> the triple point $273.16\ \mathrm K$. The ideal-gas prediction at $300\ \mathrm K$ is
> $50.0(300/273.16)=54.9\ \mathrm{torr}$. Kelvin temperature is required, because the
> pressure is not proportional to a shifted Celsius scale.

Other thermometers use electrical resistance, thermoelectric voltage, liquid
column length, or thermal radiation. Each requires calibration because the measured
property may be nonlinear, sensitive to pressure, or affected by aging. A resistance
thermometer can be highly sensitive over a narrow range but needs a calibration
curve. A gas thermometer realizes a fundamental absolute-temperature reference at
low density, but it is slower and less convenient for routine measurements.

Calibration uncertainty has two parts: uncertainty in the reference point and
uncertainty in the property measurement. A sensor can be precise yet inaccurate if
its calibration is offset; it can be accurate on average yet noisy if its reading
fluctuates. Thermal contact also matters. A probe placed against a hot wall may
measure the wall temperature, while a poorly shielded probe in a flowing gas can be
affected by radiation or conduction along its leads. The indicated value represents
the target only after the sensor and target have reached the relevant thermal condition.

Temperature measurement is part of process selection. A phase boundary, expansion
calculation, radiation estimate, and calorimetry balance all depend on temperature,
but not always on the same temperature: surface temperature, reservoir temperature,
and average bulk temperature can differ. Recording sensor location, calibration,
and response time is therefore as important as recording the numerical reading.

### Heat exchangers and coupled energy streams

A heat exchanger transfers energy between flowing streams while keeping their
materials separated. In a steady exchanger with negligible external loss, the heat
rate lost by the hot stream equals the heat rate gained by the cold stream. For a
stream that remains in one phase, the rate is
$\dot Q=\dot m c_p(T_{\rm out}-T_{\rm in})$, where $\dot m$ is mass flow rate and
$c_p$ is the appropriate specific heat. The signs are opposite for the two streams,
but the magnitude match is an energy-conservation check before any detailed heat-
transfer coefficient is calculated.

Parallel-flow exchangers send both streams in the same direction. Their temperature
difference is largest near the inlet and falls rapidly along the device. Counterflow
exchangers send streams in opposite directions and can maintain a larger driving
temperature difference over more of their length. For suitable flow-rate ratios, a
cold stream leaving a counterflow exchanger can approach the hot-stream inlet
temperature more closely than is possible in simple parallel flow. This is a
geometric consequence of how the local temperature differences are distributed,
not a violation of energy conservation.

The transferred rate also depends on wall conduction and fluid-side convection.
Engineers combine these effects into an overall conductance $UA$ and use a
temperature-difference average appropriate to the flow arrangement. A large wall
conductivity alone does not guarantee a strong exchanger if one fluid has a thick
boundary layer. Conversely, increasing flow speed can raise convection but also
increases pumping power and pressure loss. Exchanger design therefore balances
thermal performance against the mechanical work required to move the fluids.

Fouling adds resistance over time. Scale, corrosion products, biological films, or
oil deposits create a low-conductivity layer between the fluid and wall. The same
inlet temperatures and flow rates then produce a smaller heat rate than a clean
device. Comparing measured outlet temperatures with the energy balance can expose
fouling, bypass flow, leaks, or an incorrect flow-meter calibration. The
temperature data alone do not identify the cause; mass-flow measurements and
pressure-drop data complete the diagnostic picture.

Phase change can make an exchanger especially effective because condensation or
boiling transfers large energy at nearly constant temperature. A condenser rejects
latent heat from a vapor stream, while an evaporator absorbs it into a refrigerant.
The single-phase formula using one constant specific heat then applies only to the
sections before or after the phase-change region. A real process model partitions
the exchanger into sensible-heat and latent-heat zones rather than assigning one
average temperature change to the entire stream.

$$
% caption: Counterflow heat exchange maintains a driving temperature difference along more of the device than parallel flow. Each stream's energy-rate change is determined from its mass flow, heat capacity, and inlet-to-outlet temperature change.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[black,thick] (.45,.55) rectangle (5.95,2.7);
\draw[black,thick] (.7,1.6)--(5.7,1.6);
\draw[acc,->,thick] (.75,2.2)--(5.5,2.2);
\node[acc] at (3.1,2.45) {hot stream};
\draw[black,->,thick] (5.65,1.0)--(.9,1.0);
\node[black] at (3.1,.78) {cold stream};
\draw[black,->,thick] (2.0,2.02)--(2.0,1.2);
\draw[black,->,thick] (3.2,2.02)--(3.2,1.2);
\draw[black,->,thick] (4.4,2.02)--(4.4,1.2);
\node[black] at (3.2,3.0) {heat exchanger};
\end{tikzpicture}
$$

**Thermal stress and constrained expansion.**

Free thermal expansion produces strain $\epsilon_{\rm th}=\alpha\Delta T$ with
little stress. If a uniform rod is held at fixed length by rigid supports, the total
axial strain must be zero. The mechanical strain from stress then cancels the free
thermal strain: $\sigma/E+\alpha\Delta T=0$. The resulting stress is
$\sigma=-E\alpha\Delta T$. Heating a fully constrained rod produces compressive
stress; cooling it produces tensile stress. The sign depends on the chosen tension-
positive convention, but the physical tendency is unambiguous.

> **Worked example.** A steel member with $E=200\ \mathrm{GPa}$ and
> $\alpha=12\times10^{-6}\ \mathrm{K^{-1}}$ undergoes a uniform $50\ \mathrm K$ rise.
> Fully constrained, its stress magnitude is
> $|\sigma|=E\alpha\Delta T=(200\ \mathrm{GPa})(12\times10^{-6})(50)=120\ \mathrm{MPa}$
> (compressive). A free $2.0\ \mathrm m$ member would instead lengthen by
> $\alpha L_0\Delta T=1.2\ \mathrm{mm}$. A small blocked expansion produces a large
> stress; allowing a few millimetres of movement at a joint removes most of the load.

The ideal formula assumes linear elasticity, uniform temperature, and perfectly
rigid supports. Actual frames have finite stiffness, so the thermal strain is
shared among member deformation, support deformation, and connection slip. A rod
with one flexible support develops less stress than a fully fixed rod. Welds,
bolts, and adhesive joints can concentrate stress near changes in cross section or
material. A uniform average stress may therefore underestimate local failure risk.

Temperature gradients introduce bending as well as axial stress. Heating one face
of a plate more than the other produces different expansion through its thickness,
causing curvature. Thermal shock occurs when surface temperature changes faster
than heat can diffuse to the core, so the surface and interior attempt incompatible
strains. Brittle materials are particularly sensitive because they tolerate little
tensile strain and may crack during rapid heating or cooling even when the average
temperature change is modest.

Stress calculations must be coupled to the thermal time history. A slow furnace
cycle can allow a component to remain nearly uniform and use the simple expansion
formula. A flame, quench, or laser pulse needs a transient temperature field before
stress can be evaluated. Measurements of surface temperature alone do not establish
the core temperature or the thermal gradient that drives bending. This is why
thermal design and structural design cannot be separated in constrained assemblies.

$$
% caption: A constrained heated rod cannot take its free thermal strain. Rigid supports convert the blocked expansion into compressive stress; a sliding joint would allow the predicted length change instead.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[black,thick] (.2,.3) rectangle (.55,2.4);
\draw[black,thick] (5.75,.3) rectangle (6.1,2.4);
\fill[acc!12] (.55,1.02) rectangle (5.75,1.68);
\draw[acc,thick] (.55,1.02) rectangle (5.75,1.68);
\draw[acc,->,thick] (2.2,2.55)--(2.2,1.85);
\draw[acc,->,thick] (4.1,2.55)--(4.1,1.85);
\node[acc] at (3.15,2.8) {heating};
\draw[black,->,thick] (1.35,1.35)--(.7,1.35);
\draw[black,->,thick] (4.95,1.35)--(5.6,1.35);
\node[black] at (.37,2.7) {support};
\node[black] at (5.93,2.7) {support};
\node[black] at (3.15,.6) {compressive stress};
\end{tikzpicture}
$$

**Insulation design and thermal bridges.**

Insulation design begins with a heat-loss target, not with material thickness alone.
A planar surface has overall resistance from interior convection, each solid layer,
exterior convection, and sometimes radiation across an air gap. The
overall heat-transfer coefficient is $U=1/R_{\rm total}$, and the steady loss is
$P=UA(T_{\rm in}-T_{\rm out})$. Adding a low-conductivity layer raises total
resistance, but the improvement becomes smaller once that layer dominates all other
resistances. Surface convection can then limit further gains.

> **Worked example.** One square metre of wall has interior convection resistance
> $0.12\ \mathrm{K\,W^{-1}}$, plaster $0.029\ \mathrm{K\,W^{-1}}$, $0.10\ \mathrm m$ of
> insulation $2.50\ \mathrm{K\,W^{-1}}$, and exterior $0.03\ \mathrm{K\,W^{-1}}$ in
> series. The total is $R_{\rm total}=2.68\ \mathrm{K\,W^{-1}}$, so
> $U=1/R_{\rm total}=0.37\ \mathrm{W\,m^{-2}\,K^{-1}}$. A $20\ \mathrm K$
> indoor–outdoor difference gives a loss of about $7.5\ \mathrm W$ through that square
> metre. This is a rate; the energy lost over a day follows by multiplying by time.

Thermal bridges create parallel low-resistance paths. A metal stud, concrete beam,
fastener, or uninsulated window frame can conduct around a high-resistance
insulation layer. The total transfer is the sum through the parallel paths, not the
transfer predicted from the insulation alone. A small bridge area can dominate if
its conductivity is orders of magnitude larger. Infrared images often display
these paths as warm exterior regions or cool interior regions under steady heating.

Air gaps are not automatically insulating. A narrow sealed gap can suppress
convection and add resistance, but a wide gap can develop natural convection loops.
Radiation across the gap can remain significant unless reflective surfaces reduce
emissivity. Moisture can also change conductivity and create condensation risk on
cold surfaces. Insulation design therefore includes vapor control, air sealing,
mechanical support, fire requirements, and long-term aging in addition to a nominal
conductivity number.

The design calculation should state boundary temperatures and surface coefficients.
Indoor air temperature is not necessarily the same as interior wall-surface
temperature, particularly near a cold bridge. Comfort, condensation, and material
durability depend on surface conditions, while energy cost depends on total heat
loss. A good model distinguishes those outcomes instead of treating one calculated
heat rate as the sole performance measure.

**Phase-boundary pressure dependence and uncertainty.**

The slope of an equilibrium phase boundary expresses how transition temperature
changes with pressure. The Clapeyron relation is
$\d P/\d T=L/[T(v_2-v_1)]$, where $L$ is latent heat per unit mass or mole in a
consistent convention and $v_2-v_1$ is the specific- or molar-volume change across
the transition. For vaporization, the vapor volume is much larger than liquid
volume, so the liquid--vapor boundary generally slopes upward: higher pressure
requires higher boiling temperature. Pressure cookers use this dependence to raise
the boiling temperature of water and thereby permit hotter liquid cooking.

The solid--liquid boundary depends on which phase has larger volume. Most materials
expand on melting, so increasing pressure favors the denser solid and raises melting
temperature. Water is unusual because ice has larger volume than liquid water.
Increasing pressure near the melting point favors the liquid and lowers the melting
temperature. The sign is not a diagram convention; it follows from the volume
change and is a measurable material property.

Phase-boundary measurements require careful uncertainty handling. The transition
temperature can be broadened by temperature gradients, dissolved impurities,
finite heating rate, and pressure-sensor calibration. A reported boiling point is
meaningful only with its pressure, sample composition, and temperature-sensor
location. During a slow equilibrium measurement, repeated heating and cooling
runs can estimate hysteresis and calibration drift. During rapid heating, a sensor
may lag the sample, making an apparent transition temperature differ from the true
equilibrium boundary.

Uncertainty propagates into derived quantities. The fractional uncertainty of a
conduction rate $P=kA\Delta T/L$ receives contributions from $k$, area,
length, and the measured temperature difference. When $\Delta T$ is small, a fixed
temperature-sensor error becomes a large fractional error in the heat rate. A
phase-boundary slope estimated from two nearby points has the same problem: a small
denominator magnifies pressure and temperature uncertainty. Collecting data across
a wider controlled range can improve the slope estimate, provided the same phase
boundary and material state remain relevant.

Measurement uncertainty sets the model distinctions that the experiment can
resolve. If the predicted thermal-bridge loss is smaller than the heat-flux
measurement uncertainty, the experiment cannot confirm the
bridge model. If two phase-boundary slopes differ by less than sensor drift, their
difference is not established. A complete thermal-process report therefore records
calibration, time response, geometry, boundary conditions, and uncertainty along
with the calculated transfer rate or transition point.

**Integrated validation of a thermal-process model.**

A thermal-process calculation should be checked at three levels: energy balance,
transport model, and measurement consistency. The energy balance compares stored
energy change with heat and work transfers over the selected interval. The transport
model identifies whether conduction, convection, radiation, phase change, or mass
flow sets each transfer path. The measurement check compares sensor outputs
with the temperatures, pressures, dimensions, and rates assumed by the
model. Agreement at only one level can be misleading. A heat rate may satisfy an
energy balance while being assigned to the wrong physical path.

Steady and transient conditions must be distinguished before data are averaged.
Equal inlet and outlet energy rates over a long exchanger run support a steady
approximation, whereas changing wall temperature or stored energy requires a
transient term. Likewise, a phase boundary inferred during rapid heating can be
shifted by sensor lag even when the final equilibrium temperature is correct. Time
stamps, sensor response times, and the duration used for energy integration should
be recorded with every reported thermal rate.

Independent measurements are valuable because they test different assumptions. A
calculated conductive loss can be compared with electrical heater input, a fluid
enthalpy-rate change, or a surface heat-flux sensor. A disagreement can indicate a
thermal bridge, radiation path, unmeasured convection, or calibration offset rather
than simple random error. Repeating a test with a changed insulation thickness,
flow speed, or boundary temperature tests which resistance dominates the observed
response.

The final report should state the system boundary, phase state, geometry, material
properties, boundary temperatures, reference pressure, and uncertainty range. Together,
these items permit independent reproduction and identify the assumptions under test.
A numerical answer without those conditions cannot establish whether a different
operating point, material batch, mounting method, or measurement interval would
produce the same result.

### Conduction resistance and layered barriers

Steady conduction through a plane wall is often clearer when written as a thermal
resistance. The thermal resistance of a layer with thickness $L$, conductivity $k$,
and cross-sectional area $A$ is $R_{\rm th}=L/(kA)$. The heat-transfer rate is then
$P=(T_H-T_C)/R_{\rm th}$. A large thickness or small conductivity increases the
resistance; a large cross-sectional area decreases it. This form has the same
structure as a simple series resistance calculation, but it represents a different
physical transport mechanism: energy transfer by microscopic interactions through
the material.

Layers in series carry the same steady heat rate, while their temperature drops add,
and the total resistance is the sum of the individual resistances.

> **Worked example.** A $10\ \mathrm{m^2}$ wall combines $0.10\ \mathrm m$ of brick
> ($k=0.70\ \mathrm{W\,m^{-1}\,K^{-1}}$) with $0.050\ \mathrm m$ of insulation
> ($k=0.040\ \mathrm{W\,m^{-1}\,K^{-1}}$). The layer resistances $L/(kA)$ are
> $0.014\ \mathrm{K\,W^{-1}}$ for the brick and $0.125\ \mathrm{K\,W^{-1}}$ for the
> insulation. Across a $25\ \mathrm K$ difference,
> $P=25/(0.014+0.125)\approx180\ \mathrm W$. The thin insulation dominates because its
> conductivity is far smaller than the brick's.

The temperature drop across each layer is proportional to its resistance. In the
example, most of the $25\ \mathrm K$ drop occurs across the insulation, not across
the brick. A surface thermometer therefore cannot determine the heat rate unless
the material stack, area, and boundary temperatures are known. Contact resistance
at gaps, fasteners, or poorly joined layers can also be important. It adds another
temperature drop at an interface and can invalidate a calculation based only on
bulk material thicknesses.

The steady formula assumes one-dimensional transfer, constant conductivity, and no
internal heat generation. Thermal bridges such as metal studs bypass insulation by
providing lower-resistance paths in parallel. In a real wall, conduction through
solid layers, convection at surfaces, and radiation across gaps can all contribute.
A resistance network applies only after each physical path is represented
with the correct series or parallel connection.

$$
% caption: Layered conduction is a series resistance problem. The same heat rate passes through every layer, while the largest temperature drop occurs across the largest thermal resistance.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\fill[acc!12] (.2,.5) rectangle (1.3,2.55);
\draw[acc,thick] (.2,.5) rectangle (1.3,2.55);
\draw[black,thick] (1.3,.5) rectangle (2.9,2.55);
\fill[black] (2.9,.5) rectangle (5.4,2.55);
\draw[black,thick] (2.9,.5) rectangle (5.4,2.55);
\fill[acc!12] (5.4,.5) rectangle (6.3,2.55);
\draw[acc,thick] (5.4,.5) rectangle (6.3,2.55);
\draw[acc,->,very thick] (.7,1.5)--(5.85,1.5);
\node[acc] at (3.05,1.85) {heat rate};
\node[acc] at (.75,2.85) {warm side};
\node[acc] at (5.85,2.85) {cool side};
\node[black] at (2.1,.92) {brick};
\node[black] at (4.15,.92) {insulation};
\draw[<->,black] (1.3,.22)--(2.9,.22);
\draw[<->,black] (2.9,.22)--(5.4,.22);
\end{tikzpicture}
$$

**Convective transfer and boundary layers.**

Convection transfers energy between a surface and a moving fluid. A common model is
$P_{\rm conv}=hA(T_s-T_f)$, where $T_s$ is surface temperature, $T_f$ is a
representative fluid temperature, $A$ is exposed area, and $h$ is the convective
heat-transfer coefficient. The coefficient combines the effect of fluid speed,
viscosity, density, thermal conductivity, geometry, and boundary-layer structure.
It is not a material constant belonging to the solid surface alone.

Near a stationary wall, fluid velocity falls to zero at the surface and rises across
a boundary layer. If this layer is thick and nearly motionless, energy must conduct
through it slowly. Faster external flow thins the boundary layer and replenishes
fluid near the surface, increasing the temperature gradient and the transfer rate.
Fans, pumps, and vehicle motion therefore increase convection. The same surface can
have very different heat-transfer rates in still air, a gentle breeze, and a forced
air duct.

Natural convection results when heating changes fluid density. Warm fluid near a
vertical surface becomes less dense and rises; cooler fluid replaces it from below.
The resulting circulation carries energy away without a fan. Its rate depends on
orientation and gravity as well as temperature difference. A horizontal heated
surface facing upward and one facing downward can have different convection
coefficients because buoyant motion develops differently around them.

Convection correlations are empirical or derived from fluid-flow models over
restricted ranges. Using one tabulated coefficient outside its flow speed or
geometry range can be less accurate than a simple conduction estimate through a
known layer. Surface fins increase area and can improve transfer, but only when the
fin itself conducts energy effectively enough to maintain a significant temperature
difference along its length. More area is not automatically more cooling if the
added area is nearly at fluid temperature.

**Radiative exchange and emissivity.**

Every surface at nonzero absolute temperature emits thermal radiation. A diffuse
gray surface emits at rate $\epsilon\sigma AT^4$, where $\epsilon$ is emissivity,
$A$ is surface area, and $\sigma$ is the Stefan--Boltzmann constant. Emissivity
lies between zero and one and compares the surface emission with that of an ideal
blackbody at the same temperature. Dark, rough, or coated surfaces can have high
emissivity; polished metals often have lower values over some wavelength ranges.
The property depends on surface condition and spectral range; visible color alone
does not determine it.

Net radiation matters when a surface exchanges energy with surroundings. A
surface facing a large enclosure at temperature $T_{\rm env}$ is described by
$P_{\rm rad}=\epsilon\sigma A(T^4-T_{\rm env}^4)$, with the fourth powers in kelvins.

> **Worked example.** A $1.0\ \mathrm{m^2}$ surface of emissivity $0.80$ at
> $400\ \mathrm K$ faces surroundings at $300\ \mathrm K$. Its net radiative loss is
> $P_{\rm rad}=(0.80)(5.67\times10^{-8})(1.0)(400^4-300^4)\approx790\ \mathrm W$. A
> surface at $310\ \mathrm K$ over the same $300\ \mathrm K$ surroundings loses only
> about $51\ \mathrm W$: because emitted power scales as $T^4$, radiative loss climbs
> steeply with absolute temperature, not merely with the temperature difference.

Radiation travels through vacuum, unlike conduction and convection. A spacecraft
must reject internally generated energy by radiation because there is no surrounding
fluid for convection. On Earth, radiation often acts in parallel with convection.
A hot matte pipe can radiate strongly while air simultaneously carries energy away.
Reducing emissivity with a reflective finish can lower radiation but may have little
effect if convection dominates the total thermal resistance.

Geometric view also matters. A small surface does not radiate all of its emission
to one chosen target unless that target fills its field of view. Detailed enclosure
calculations use view factors and multiple reflections. The simple large-surroundings
formula is appropriate when the environment approximately surrounds the object at
one uniform temperature. It is a modeling choice that should be checked before
applying a single ambient-temperature value to a complex room or furnace.

### Thermal time constants and transient response

A body does not usually reach a new thermal environment temperature instantly.
When its internal temperature is nearly uniform, the lumped-capacitance model gives
$mc\,\d T/\d t=-hA(T-T_\infty)$ for convection to surroundings at $T_\infty$. The
solution is $T-T_\infty=(T_i-T_\infty)e^{-t/\tau}$ with thermal time constant
$\tau=mc/(hA)$. The mass and specific heat set thermal capacitance; the product
$hA$ sets the transfer conductance. A large, massive object with small exposed area
responds slowly, while a thin fin or small sensor in fast flow responds quickly.

After one time constant, the temperature difference from surroundings has fallen to
about $37\%$ of its initial value. After three time constants it is about $5\%$;
after five it is less than one percent. These percentages refer to the remaining
difference, not to an absolute temperature. A thermometer placed suddenly in a new
environment must be allowed several time constants before its reading represents
the target temperature to a chosen tolerance.

> **Worked example.** A $0.50\ \mathrm{kg}$ aluminum body with
> $c=900\ \mathrm{J\,kg^{-1}\,K^{-1}}$, exposed area $0.050\ \mathrm{m^2}$, and
> $h=20\ \mathrm{W\,m^{-2}\,K^{-1}}$ has thermal capacitance
> $mc=450\ \mathrm{J\,K^{-1}}$ and conductance $hA=1.0\ \mathrm{W\,K^{-1}}$, so
> $\tau=mc/(hA)=450\ \mathrm s$. A reading after one minute is far from equilibrium;
> even a fifteen-minute wait is only about two time constants and leaves a substantial
> fraction of the initial difference.

The lumped model requires internal conduction to be fast enough that the body has
no large internal temperature gradients. Thick, poorly conducting bodies violate
that assumption: the surface can cool rapidly while the core remains warm. A
temperature sensor attached to the surface then measures surface response, not
average stored energy. More detailed transient-conduction models divide the body
into spatial regions and solve the diffusion equation rather than using one
temperature and one time constant.

$$
% caption: A lumped body approaches surroundings exponentially. One thermal time constant reduces the initial temperature difference to about 37 percent; the response is slower for larger heat capacity and faster for larger transfer conductance.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0,0)--(6.2,0) node[right] {time};
\draw[->,black] (0,0)--(0,3.5) node[above] {temperature gap};
\draw[acc,very thick] plot[domain=.2:5.7,samples=180] (\x,{3.05*exp(-.5*(\x-.2))});
\draw[black,dashed] (2.2,0)--(2.2,1.12);
\draw[black,dashed] (0,1.12)--(2.2,1.12);
\node[black,below] at (2.2,0) {one time constant};
\node[acc] at (4.0,1.55) {exponential decay};
\end{tikzpicture}
$$