Response Functions: Heat Capacity, Compressibility, and Expansivity

Every thermodynamic measurement you have ever made is a response function: how much a system's temperature rises when you pour heat in (heat capacity), how much its volume shrinks when you squeeze it (compressibility), how much it swells when you warm it (thermal expansion). These are the only thermodynamic quantities you can actually put a meter on — you cannot measure entropy directly, but you can measure the heat capacity and integrate. Response functions are the bridge between the abstract potentials of thermodynamics and the laboratory. What makes them more than a bookkeeping list is that they are all second derivatives of a single thermodynamic potential, and that shared parentage forces exact relations among them. Because a mixed second derivative does not care about the order of differentiation, the response functions are locked together by Maxwell relations and by a small set of identities that hold for any substance whatsoever. The most famous is $C_p - C_V = TV\alpha^2/\kappa_T$ : the gap between the heat capacity at constant pressure and at constant volume is not a property of some particular gas — it is fixed, universally, by the thermal expansion $\alpha$ and the compressibility $\kappa_T$ . Measure those two and you predict the difference, for water or for xenon. There is a second reason these quantities are central: every response function is a fluctuation in disguise. The heat capacity is the variance of the energy, the compressibility is the variance of the density, the magnetic susceptibility is the variance of the magnetization. This fluctuation–response duality — which the canonical ensemble will make exact — means the response functions are always non-negative in a stable system, and their divergence is the signature of a critical point. This article defines the five workhorse responses, derives the identities that bind them, and shows the numbers for real materials.

Key takeaways

Response functions — heat capacities $C_V,C_p$ , compressibilities $\kappa_T,\kappa_S$ , thermal expansion $\alpha$ — are the measurable second derivatives of a thermodynamic potential.
The universal identity $C_p - C_V = TV\alpha^2/\kappa_T \ge 0$ fixes the heat-capacity gap for any substance from its expansion and compressibility; for an ideal gas it equals $Nk_B$ .
The ratio identity $C_p/C_V = \kappa_T/\kappa_S = \gamma$ ties the two heat capacities to the two compressibilities.
Every response is a fluctuation: $C_V=\langle\Delta E^2\rangle/k_BT^2$ , $\kappa_T\propto\langle\Delta N^2\rangle$ — so stability forces them non-negative, and they diverge at critical points.
Adiabatic responses are stiffer than isothermal ones: $\kappa_S=\kappa_T/\gamma<\kappa_T$ ; this is why the speed of sound uses $\kappa_S$ , the fix Laplace made to Newton.
Key idea: response functions are second derivatives of one potential, which locks them into exact, substance-independent identities — the most useful being $C_p-C_V=TV\alpha^2/\kappa_T$ — and identifies each with a thermal fluctuation.

The core insight

Response functions are the second derivatives of a thermodynamic potential, so they are not independent: mixed-derivative equality forces exact, substance-independent relations among them — chiefly $C_p-C_V=TV\alpha^2/\kappa_T$ and $C_p/C_V=\kappa_T/\kappa_S$ . Each response is also a fluctuation ( $C_V=\langle\Delta E^2\rangle/k_BT^2$ ), so stability makes them non-negative and criticality makes them diverge.

Where this lives today

Response functions are the currency of experimental condensed-matter physics. A heat-capacity measurement is often the first probe of a new material: a sharp peak in $C_p$ flags a phase transition, a linear-in- $T$ term reveals conduction electrons, an exponential fall-off exposes an energy gap (a superconductor or a magnet). Because $C_V=\langle\Delta E^2\rangle/k_BT^2$ , calorimetry is literally fluctuation spectroscopy — it counts how the energy jitters. Compressibility and its magnetic cousin, the susceptibility, play the same role for density and spin fluctuations; the divergence of the susceptibility is how one detects a magnetic critical point. The identities themselves are used constantly to reconcile data with theory. Statistical-mechanical models compute $C_V$ ; experiments at ambient pressure measure $C_p$ ; the bridge is $C_p-C_V=TV\alpha^2/\kappa_T$ . In geophysics the adiabatic compressibility $\kappa_S$ inferred from seismic wave speeds is converted to the isothermal $\kappa_T$ to build equations of state for the Earth's interior. And the entire modern theory of critical phenomena is organized around critical exponents — the power laws by which $C_p$ , $\kappa_T$ , and $\chi$ diverge at $T_c$ — which the renormalization group predicts and experiments confirm to remarkable precision.

The physical picture

Picture poking a material and watching it flinch. Add a joule of heat: the temperature ticks up by $1/C$ , so a large heat capacity means a sluggish thermometer — a big system, or one with many places to hide energy (vibrations, rotations, phase changes). Press on it: the volume gives by $\kappa$ , so a large compressibility means a soft material. Warm it: it lengthens by $\alpha$ . Each response is the slope of a state variable against the knob you turn, and the second derivative of a potential because the state variable was itself already a first derivative. The subtle and beautiful fact is that which knob you hold fixed matters. Heating a gas at constant volume dumps all the energy into raising the temperature; heating it at constant pressure lets it expand, and some of your heat is spent doing work pushing the atmosphere back — so $C_p>C_V$ . The difference is exactly the work of expansion per degree, and it is pinned down universally.

Figure

Heat at constant V versus constant p: at fixed pressure some heat does expansion work, so C_p exceeds C_V — the gap is TVα²/κ_T

The second picture worth carrying is the fluctuation view. Left alone at temperature $T$ , a system's energy is not razor-sharp — it jitters around its mean, and the size of that jitter is exactly $k_BT^2 C_V$ . A material with a large heat capacity is one whose energy fluctuates a lot; a soft, highly compressible material is one whose density fluctuates a lot. This turns thermodynamic responses into statistical quantities and, crucially, makes them positive: a variance cannot be negative, so $C_V\ge0$ and $\kappa_T\ge0$ in any stable phase. The punchline plot below shows what happens when the fluctuations run away — at a critical point the compressibility diverges, the material becomes infinitely soft, and light scatters off the giant density fluctuations (critical opalescence).

Figure

Response functions diverge at a critical point: κ_T (and C_p) shoot up as T → T_c — the runaway of fluctuations

The core equations

The five workhorse response functions are defined as first derivatives of a state variable — hence second derivatives of a potential. Two heat capacities:

C_V = \left(\frac{\partial U}{\partial T}\right)_V = T\left(\frac{\partial S}{\partial T}\right)_V, \qquad C_p = \left(\frac{\partial H}{\partial T}\right)_p = T\left(\frac{\partial S}{\partial T}\right)_p \qquad (1)

where: $U$ — internal energy; $H=U+pV$ — enthalpy; $S$ — entropy. $C_V$ is measured at fixed volume, $C_p$ at fixed pressure; both are $T\,\partial S/\partial T$ , so they are $-T\,\partial^2 F/\partial T^2$ and $-T\,\partial^2 G/\partial T^2$ respectively. Two compressibilities and the thermal expansion coefficient:

\kappa_T = -\frac1V\left(\frac{\partial V}{\partial p}\right)_T,\quad \kappa_S = -\frac1V\left(\frac{\partial V}{\partial p}\right)_S,\quad \alpha = \frac1V\left(\frac{\partial V}{\partial T}\right)_p \qquad (2)

where: $\kappa_T$ — isothermal compressibility; $\kappa_S$ — adiabatic (isentropic) compressibility; $\alpha$ — coefficient of volume thermal expansion. All are intensive; the $1/V$ makes them per-unit-volume responses. The central identity relates the two heat capacities through the mechanical responses. It holds for every substance:

C_p - C_V = \frac{TV\alpha^2}{\kappa_T} \ge 0 \qquad (3)

where: the right side is manifestly non-negative ( $\alpha^2\ge0$ , $\kappa_T>0$ , $T,V>0$ ), so $C_p\ge C_V$ always. The gap grows with thermal expansion (more work done pushing back the surroundings) and with rigidity (small $\kappa_T$ ). A companion identity relates the two heat capacities to the two compressibilities:

\frac{C_p}{C_V} = \frac{\kappa_T}{\kappa_S} \equiv \gamma \qquad (4)

where: $\gamma$ — the adiabatic index. Since $\gamma\ge1$ , this says $\kappa_T\ge\kappa_S$ : a material is stiffer under fast (adiabatic) compression than slow (isothermal) compression, because the heat of compression cannot escape and adds pressure. For the ideal gas the identities collapse to famous numbers. With $\alpha=1/T$ and $\kappa_T=1/p$ :

C_p - C_V = \frac{TV(1/T)^2}{1/p} = \frac{Vp}{T} = Nk_B \quad (=nR \text{ per mole}) \qquad (5)

where: the last step uses $pV=Nk_BT$ . This is Mayer's relation $C_p-C_V=nR$ — the heat-capacity gap of any ideal gas is exactly the gas constant, independent of whether it is monatomic, diatomic, or polyatomic.

Derivation AThe universal identity C_p − C_V = TVα²/κ_T

The goal is to prove that the difference of the two heat capacities is fixed, for any substance, by the thermal expansion and the isothermal compressibility — a relation that lets you predict $C_p$ from $C_V$ (or vice versa) using only mechanical measurements. The strategy is to write the entropy as a function of $(T,V)$ , differentiate at constant pressure, and use a Maxwell relation plus the triple product rule to convert the awkward derivative $(\partial V/\partial T)_p$ and $(\partial p/\partial T)_V$ into the standard response functions.

Step 1. Start from $C_p - C_V = T[(\partial S/\partial T)_p - (\partial S/\partial T)_V]$ . Treat $S=S(T,V)$ and expand the constant-pressure derivative: $(\partial S/\partial T)_p = (\partial S/\partial T)_V + (\partial S/\partial V)_T(\partial V/\partial T)_p$ . Why: the chain rule accounts for the fact that at constant pressure the volume also changes with temperature, contributing through $(\partial S/\partial V)_T$ .

Step 2. So $C_p - C_V = T(\partial S/\partial V)_T(\partial V/\partial T)_p$ . Apply the Maxwell relation from the Helmholtz free energy, $(\partial S/\partial V)_T = (\partial p/\partial T)_V$ . Why: $F(T,V)$ has $dF=-S\,dT-p\,dV$ , and equality of mixed second derivatives $\partial^2 F/\partial V\partial T$ gives exactly this Maxwell relation.

Step 3. Now $C_p - C_V = T(\partial p/\partial T)_V(\partial V/\partial T)_p$ . Use the triple product rule $(\partial p/\partial T)_V = -(\partial p/\partial V)_T(\partial V/\partial T)_p$ to express the pressure derivative through mechanical responses. Why: the cyclic relation $(\partial p/\partial T)_V(\partial T/\partial V)_p(\partial V/\partial p)_T=-1$ lets us trade the hard-to-measure constant-volume pressure slope for the compressibility and expansion.

Step 4. Substitute and identify the response functions. With $(\partial V/\partial T)_p = V\alpha$ and $-(\partial p/\partial V)_T = 1/(V\kappa_T)$ :

C_p - C_V = T\left[\frac{1}{V\kappa_T}(V\alpha)\right](V\alpha) = \frac{TV\alpha^2}{\kappa_T}.

Why: assembling the pieces, the volume factors combine to a single $V$ , the two expansion factors give $\alpha^2$ , and the compressibility appears in the denominator — a manifestly non-negative result. *Verified: symbolic derivation gives $C_p-C_V=TV\alpha^2/\kappa_T$ ; substituting the ideal-gas responses $\alpha=1/T$ , $\kappa_T=1/p$ yields $Nk_B$ (Mayer's relation), confirmed in verify_aq-sm-response-functions.py.*

Go deeper · Derivation B · 1 minEvery response is a fluctuation — and hence non-negative (the owned result)

Derivation BEvery response is a fluctuation — and hence non-negative (the owned result)

This is the article's owned structural result: the response functions are not merely second derivatives of a potential, they are variances of the underlying microscopic quantities, which makes their non-negativity a consequence of statistics rather than a separate postulate. The goal is to show $C_V=\langle\Delta E^2\rangle/k_BT^2$ (and, by the same logic, $\kappa_T\propto\langle\Delta N^2\rangle$ ), so that stability of a phase is the same statement as boundedness of its fluctuations. The strategy is to differentiate the canonical mean energy with respect to temperature and recognize the variance.

Step 1. In the canonical ensemble the mean energy is $U=\langle E\rangle = -\partial_\beta\ln Z$ , and differentiating once more gives $\partial_\beta^2\ln Z = \langle E^2\rangle - \langle E\rangle^2 = \langle\Delta E^2\rangle$ , the energy variance. Why: each $-\partial_\beta$ acting on $\ln Z$ pulls down one factor of $E$ and subtracts the mean, exactly generating the cumulants of the energy distribution.

Step 2. Convert the $\beta$ -derivative to a temperature derivative using $\beta=1/k_BT$ , so $\partial_\beta = -k_BT^2\,\partial_T$ . Then $\langle\Delta E^2\rangle = -\partial_\beta U = k_BT^2(\partial U/\partial T)_V = k_BT^2 C_V$ . Why: the Jacobian $d\beta/dT=-1/k_BT^2$ carries the derivative from inverse-temperature to temperature and lands the heat capacity.

Step 3. Therefore $C_V = \langle\Delta E^2\rangle/(k_BT^2)$ . Since a variance is non-negative and $T^2>0$ , $C_V\ge0$ automatically. The same argument in the grand canonical ensemble gives $\kappa_T = V\langle\Delta N^2\rangle/(k_BT\langle N\rangle^2)\ge0$ . Why: fluctuation formulas are variances, and a variance can never be negative — stability is guaranteed by the structure of probability, not assumed.

Step 4. Read the divergence backward. A response function blows up precisely when the corresponding fluctuation becomes macroscopic — at a critical point, where $\langle\Delta N^2\rangle$ grows without bound, so $\kappa_T\to\infty$ and $C_p\to\infty$ . Why: criticality is the runaway of fluctuations; the diverging response is the thermodynamic shadow of correlations extending across the whole sample. *Verified: $C_V=\langle\Delta E^2\rangle/k_BT^2\ge0$ and the sign structure of the identities are checked; $C_p-C_V\ge0$ follows from $\alpha^2\ge0$ , $\kappa_T>0$ in the verify script.*

Limiting cases

The response identities specialize cleanly across regimes:

Case	Result	Meaning
Ideal gas	$C_p - C_V = Nk_B$ (Mayer)	The gap equals the gas constant per mole, independent of atomicity.
Incompressible / condensed	$\alpha\to0$	$C_p\to C_V$ : solids and liquids expand little, so their two heat capacities nearly coincide.
Adiabatic vs isothermal	$\kappa_S=\kappa_T/\gamma$	Adiabatic compression is stiffer; the ratio is the adiabatic index $\gamma$ .
Critical point	$\kappa_T,C_p\to\infty$	Fluctuations diverge; the material becomes infinitely soft (critical opalescence).
Water below 4 °C	$\alpha<0$	Thermal expansion goes negative, yet $C_p-C_V=TV\alpha^2/\kappa_T$ stays positive ( $\alpha^2$ ).

Common trap“C_p equals C_V for solids and liquids, so the distinction only matters for gases”

Tempting, but wrong in detail. For condensed matter $\alpha$ is small, so $C_p-C_V=TV\alpha^2/\kappa_T$ is small — but it is not zero, and it grows with temperature. For copper at room temperature $C_p-C_V$ is a few percent of $C_V$ , and near the melting point it can reach ten percent or more; ignoring it corrupts high-temperature heat-capacity data. The deeper trap is which heat capacity your experiment measures: a calorimeter at atmospheric pressure gives $C_p$ , but the statistical-mechanical formulas (equipartition, Debye) most naturally give $C_V$ — you must convert with $C_p-C_V=TV\alpha^2/\kappa_T$ before comparing theory to experiment. And note $\alpha$ can be negative (water below 4 °C, some polymers, zirconium tungstate), but $C_p>C_V$ still holds because only $\alpha^2$ enters.

Worked examples

1 · Predicting C_p from C_V for a real gas (a genuine prediction)

Nitrogen near room temperature behaves nearly ideally. Using measured mechanical responses $\alpha=1/T$ and $\kappa_T=1/p$ for the ideal-gas approximation, predict $C_p-C_V$ per mole at $T=300\,\mathrm{K}$ , and compare with the tabulated $C_V=\tfrac52 R$ (diatomic) to get $\gamma$ .

Step 1. Assemble the identity for one mole: $C_p-C_V = \frac{TV\alpha^2}{\kappa_T}$ with $\alpha=1/T$ , $\kappa_T=1/p$ , and $V=RT/p$ . Substitute: $\frac{T(RT/p)(1/T)^2}{1/p} = \frac{RT/p}{1/p}\cdot\frac1T\cdot\frac{T}{1}$ … collecting, $= R$ .

Step 2. So $C_p-C_V = R = \mathbf{8.314\ J\,mol^{-1}K^{-1}}$ — Mayer's relation, no matter the gas. This is a prediction from mechanical data alone: measure how the gas expands and compresses, and you know the heat-capacity gap without any calorimetry.

Step 3. For diatomic nitrogen the equipartition value is $C_V=\tfrac52 R = 20.79\ \mathrm{J\,mol^{-1}K^{-1}}$ (three translational + two rotational quadratic modes). Then $C_p = C_V + R = \tfrac72 R = \mathbf{29.10\ J\,mol^{-1}K^{-1}}$ .

Step 4. Adiabatic index: $\gamma = C_p/C_V = \frac{7/2}{5/2} = \mathbf{1.40}$ — the measured $\gamma$ of air, and exactly the ratio $\kappa_T/\kappa_S$ . The whole thermodynamic response of air at room temperature falls out of counting modes and applying one identity.

2 · Compressing water: isothermal versus adiabatic (exam level)

Water at $25^\circ$ C has $\kappa_T=4.5\times10^{-10}\,\mathrm{Pa^{-1}}$ , $C_p=4.18\,\mathrm{J\,g^{-1}K^{-1}}$ , $C_V=4.13\,\mathrm{J\,g^{-1}K^{-1}}$ . (a) Find $\gamma$ and the adiabatic compressibility $\kappa_S$ . (b) Which would you use to compute the speed of sound, and why?

Step 1. Adiabatic index: $\gamma = C_p/C_V = 4.18/4.13 = \mathbf{1.012}$ — very close to 1, as expected for a nearly incompressible liquid where the two heat capacities almost coincide.

Step 2. Adiabatic compressibility from the ratio identity $\kappa_S = \kappa_T/\gamma = \frac{4.5\times10^{-10}}{1.012} = \mathbf{4.45\times10^{-10}\ Pa^{-1}}$ — slightly smaller than $\kappa_T$ : water is a touch stiffer when squeezed too fast for heat to escape.

Step 3. The speed of sound is $c=1/\sqrt{\rho\kappa_S}$ , not $\kappa_T$ , because a sound wave compresses the medium far faster than heat can diffuse away — the compressions are adiabatic, not isothermal.

Step 4. Numerically $c = 1/\sqrt{(1000)(4.45\times10^{-10})} = 1/\sqrt{4.45\times10^{-7}} = \mathbf{1499\ m/s}$ — the measured speed of sound in water is about $1480$ – $1500$ m/s. Using $\kappa_T$ instead would give the wrong answer; this is the liquid analogue of Newton's isothermal blunder that Laplace fixed for air.

Origins · 1 minPierre-Simon Laplace and how this was discovered

Pierre-Simon Laplace

1749–1827 · Paris

Laplace's fingerprint on response functions is the celebrated correction to the speed of sound. Newton had computed it assuming the compressions in a sound wave were isothermal, using $\kappa_T$ , and got an answer about 18% too low. In 1816 Laplace recognized that a sound wave oscillates far too fast for heat to diffuse — the compressions are adiabatic — so the stiffer $\kappa_S=\kappa_T/\gamma$ is the right response function, restoring the missing factor $\sqrt\gamma$ and matching experiment. This was one of the first triumphs of distinguishing adiabatic from isothermal responses, the very distinction at the heart of $C_p$ versus $C_V$ . Laplace, the towering figure of celestial mechanics and probability, worked with Lavoisier on early calorimetry (they measured heat capacities with an ice calorimeter) and did much to establish heat as a quantitative science. The adiabatic-versus-isothermal insight he pioneered is the reason every acoustics and every equation-of-state calculation must pick the correct response function.

History

1842

— Mayer states the relation C_p − C_V = R for gases from the mechanical equivalent of heat.

1816

— Laplace corrects Newton's speed-of-sound formula by using the adiabatic (not isothermal) compressibility.

1871

— Maxwell's Theory of Heat systematizes the relations among thermodynamic derivatives.

1876

— Gibbs's work provides the potential-based foundation from which all response identities follow.

1907

— Einstein's specific-heat model, and later Debye (1912), compute C_V microscopically, forcing the C_p−C_V conversion.

Going further

References. Pathria & Beale, Statistical Mechanics, 3rd ed., §3.6 and Ch. 1 · Callen, Thermodynamics and an Introduction to Thermostatistics, 2nd ed., Chs. 3, 7 · Reif, Fundamentals of Statistical and Thermal Physics, Ch. 5 · Kardar, Statistical Physics of Particles, §1.9 · Landau & Lifshitz, Statistical Physics Part 1, §§16, 20 · Blundell & Blundell, Concepts in Thermal Physics, Chs. 16–17.

Provenance. The definitions and identities are standard (Callen Ch. 3; Pathria §3.6; Reif Ch. 5). A verifier should reproduce: $C_p-C_V=TV\alpha^2/\kappa_T$ from the Maxwell relation and triple-product rule; the ideal-gas reduction to $Nk_B$ (Mayer) using $\alpha=1/T$ , $\kappa_T=1/p$ ; the ratio $C_p/C_V=\kappa_T/\kappa_S=\gamma$ ; and $C_V=\langle\Delta E^2\rangle/k_BT^2\ge0$ . Numerical instances (diatomic $\gamma=1.40$ , $C_p=\tfrac72 R=29.10$ ; water $\gamma=1.012$ , $\kappa_S=4.45\times10^{-10}\,\mathrm{Pa^{-1}}$ , $c\approx1499$ m/s) are recomputed in verify_aq-sm-response-functions.py. $R=8.314\ \mathrm{J\,mol^{-1}K^{-1}}$ (CODATA).

Builds on: Thermodynamic Potentials and the Fundamental Relation · Maxwell Relations · Helmholtz Free Energy, Enthalpy, and Gibbs Free Energy · Second Law of Thermodynamics and Entropy · Legendre Transforms. Leads to: Euler Relation and Gibbs–Duhem Equation · Third Law of Thermodynamics · Carnot Cycle and Thermodynamic Temperature · Energy Fluctuations in the Canonical Ensemble.

Why does heating at constant pressure need more heat than at constant volume?

Because at constant pressure the system is allowed to expand as it warms, and expansion does work against the surroundings (pushing back the atmosphere or a piston). That work has to be paid for out of the heat you supply, so some of your heat goes into pushing rather than into raising the temperature. To get the same temperature rise you therefore need to add more heat at constant pressure — hence $C_p>C_V$ . At constant volume no expansion work is done, so every joule of heat goes straight into the internal energy and the temperature. The exact size of the surplus is the expansion work per degree, which the derivation shows is $TV\alpha^2/\kappa_T$ .

Is C_p − C_V = TVα²/κ_T really true for every substance, or just ideal gases?

Every substance, with no exceptions and no approximations. The derivation uses only the definitions of the response functions, one Maxwell relation (from the equality of mixed partial derivatives of a potential), and the triple-product rule of calculus — none of which assume any particular equation of state. So it holds for gases, liquids, solids, magnets (with the magnetic analogue), rubber, anything in equilibrium. The ideal gas is just the case where $\alpha=1/T$ and $\kappa_T=1/p$ make the right side collapse to the simple $Nk_B$ . For water, copper, or liquid helium you plug in the measured $\alpha$ and $\kappa_T$ and get the correct gap. That universality is exactly why the identity is so useful: it lets you convert between theory ( $C_V$ ) and experiment ( $C_p$ ) for real materials.

How can a response function be a fluctuation? They feel like completely different ideas.

This is one of the deepest facts in statistical physics, the fluctuation–response (or fluctuation–dissipation) connection. A response function measures how much a system's average changes when you turn a knob — say, how much its energy rises per degree of temperature. A fluctuation measures how much that same quantity jitters around its average all on its own, with no knob turned. The canonical ensemble shows they are the same number: $C_V=\langle\Delta E^2\rangle/k_BT^2$ . The reason is that both are second derivatives of $\ln Z$ — the first derivative gives the mean, the second gives simultaneously the response (how the mean shifts) and the variance (the spread). Physically, a system that responds strongly to a nudge is one that already fluctuates strongly, because the same soft "restoring force" governs both.

Why is the adiabatic compressibility smaller than the isothermal one?

When you compress a material quickly (adiabatically), the work of compression heats it up, and the extra thermal pressure pushes back, so the material resists further compression more strongly — it is stiffer, i.e. less compressible. When you compress slowly (isothermally), you let that heat leak away to keep the temperature fixed, so there is no extra thermal pressure and the material gives more easily. Hence $\kappa_S<\kappa_T$ , with the exact ratio $\kappa_T/\kappa_S=C_p/C_V=\gamma$ . This is not a small technicality: sound waves are adiabatic (too fast for heat to diffuse), so the speed of sound uses $\kappa_S$ . Newton computed the speed of sound in air using $\kappa_T$ and got an answer about 18% too low; Laplace corrected it with $\kappa_S$ , restoring the factor $\sqrt\gamma$ .

Can the thermal expansion coefficient α be negative? Doesn't that break C_p ≥ C_V?

Yes, $\alpha$ can be negative, and no, it does not break the inequality. Water between $0$ and $4^\circ$ C contracts as it warms ( $\alpha<0$ ), which is why lakes freeze from the top down; some polymers, silicon at low temperature, and zirconium tungstate over a wide range also have negative expansion. But the identity $C_p-C_V=TV\alpha^2/\kappa_T$ contains $\alpha^2$ , which is non-negative regardless of the sign of $\alpha$ . So $C_p\ge C_V$ still holds — the sign of the expansion does not matter, only its magnitude. At exactly $\alpha=0$ (e.g. water at $4^\circ$ C, or an Invar alloy engineered for zero expansion), the two heat capacities coincide, $C_p=C_V$ , which is a real and measurable coincidence at that special temperature.

Why do response functions diverge at a critical point instead of just being large?

Because at a critical point the correlation length — the distance over which fluctuations are coordinated — grows without bound, so the fluctuations of density (or magnetization) become macroscopic. Since $\kappa_T\propto\langle\Delta N^2\rangle$ and $\chi\propto\langle\Delta M^2\rangle$ , an unbounded fluctuation means an unbounded response: the compressibility and susceptibility literally go to infinity as $T\to T_c$ . Physically the material becomes infinitely soft — an infinitesimal pressure change produces a large density change — which is why a fluid near its critical point scatters light so strongly it turns milky (critical opalescence). The rate of divergence is a power law with a critical exponent, and computing those exponents is the central achievement of the renormalization group.