2.1 Homogeneous nucleation theory

Suppose a sample of liquid is genuinely free of any defects — no dissolved gas, no particulate impurities, no surface-trapped gas pockets. The only way a vapour bubble can appear is by thermal fluctuation — a chance assemblage of molecules whose kinetic energy and momenta conspire to open a small vapour-filled hole. The rate at which this happens is set by a barrier in the free energy landscape, and the height of that barrier sets the tension a defect-free sample can hold.

The theory of barrier-crossing in this context is the classical nucleation theory of Volmer and Doring (1926), refined for liquid–vapour systems by Becker and Doring (1935). It produces the homogeneous tensile-strength estimate of $\sim -10^3$ atm that matches the molecular-cohesion arguments of Lesson 1.2.

ambient pressure p_∞ = -100 atm surface tension σ = 72 mN/m

temperature = 20 °C vapour pressure p_v = 0.023 atm

A vapour bubble of radius R in a liquid under tension Δp = p_v − p_∞ has Gibbs free energy ΔG(R) = −(4/3)πR³Δp + 4πR²σ. The first term (volume × pressure difference) drives growth; the second (surface area × surface tension) opposes it. ΔG peaks at the critical radius R* = 2σ/Δp with barrier height ΔG* = 16πσ³/(3Δp²). Above R* the bubble grows spontaneously; below R* it collapses. Thermal fluctuations cross the barrier at a rate J = J₀ exp(−ΔG*/kT) — exponentially sensitive to the barrier. For pure water at room temperature, the barrier is below 100 kT only when Δp exceeds ~1000 atm, recovering the homogeneous tensile-strength estimate of Lesson 1.2. The barrier is far too high at modest tensions to explain why real water tears at 0.1 atm — the resolution is heterogeneous nucleation, next lesson.

The Gibbs free energy of a vapour bubble

Consider a single spherical vapour bubble of radius $R$ appearing in a liquid at pressure $p_\infty$ and temperature $T$ . The Gibbs free energy of the system, relative to the same system with no bubble, has two contributions:

$\Delta G(R) \;=\; -\frac{4}{3} \pi R^3 \, (p_v - p_\infty) \;+\; 4 \pi R^2 \, \sigma.$

The first term is a volume contribution: $(p_v - p_\infty)$ is the pressure difference between the vapour and the surrounding liquid; multiplying by the bubble volume $\tfrac{4}{3} \pi R^3$ gives the work done by the vapour against the liquid. With the sign convention $p_v > p_\infty$ (the liquid is below the vapour pressure — under tension — so a vapour bubble can do work against the surroundings), this term is negative and favours growth.
The second term is a surface contribution: $4 \pi R^2$ is the bubble’s surface area; $\sigma$ is the interfacial tension between liquid and vapour (water at 20 °C: $\sigma = 72$ mN/m). The surface always opposes nucleation: creating a vapour-liquid interface costs energy.

The driving pressure is $\Delta p \equiv p_v - p_\infty$ . When $\Delta p > 0$ the liquid is under enough tension to favour vapour; when $\Delta p \le 0$ it is not and nucleation is impossible.

The critical radius and the barrier height

The two terms scale differently with $R$ : the volume term as $R^3$ and the surface term as $R^2$ . For small bubbles the $R^2$ surface term dominates and $\Delta G$ increases with $R$ ; for large bubbles the $R^3$ volume term dominates and $\Delta G$ decreases with $R$ . Between the two regimes $\Delta G$ has a maximum at the critical radius $R^*$ where the bubble is unstable: smaller bubbles collapse, larger bubbles grow.

Taking $\partial \Delta G / \partial R = 0$ :

$R^* \;=\; \frac{2 \sigma}{\Delta p}.$

This is the Young–Laplace radius — the radius at which surface tension just balances the driving pressure across a static spherical interface. The barrier height at $R^*$ is

$\Delta G^* \;=\; \frac{16 \pi \sigma^3}{3 (\Delta p)^2}.$

The barrier is exponentially sensitive to $\Delta p$ : doubling the tension reduces the barrier by a factor of four.

The nucleation rate

The rate of homogeneous nucleation per unit volume is set by the Boltzmann factor of the barrier:

$J \;=\; J_0 \, \exp\left(-\frac{\Delta G^*}{k_B T}\right),$

with a pre-exponential $J_0$ that contains the attempt frequency and a geometric statistical factor. For water near room temperature, the Volmer–Doring prefactor is $J_0 \approx 10^{30}$ events per cubic metre per second; the precise value matters surprisingly little, because the exponential dominates.

For a given experimental geometry, set a criterion for observable nucleation — say, one bubble per cubic millimetre per second — and invert to solve for the $\Delta p$ at which this rate is achieved:

$\Delta G^* \;\lesssim\; 50 \, k_B T \implies \Delta p \;\gtrsim\; 1000 \text{ atm at room temperature}.$

The barrier crosses the $\sim 50 k_B T$ threshold (the practical onset of homogeneous nucleation) at tensions near $-1000$ atm. This matches the theoretical cohesive estimate of Lesson 1.2 to within a factor of two — a remarkable consistency between molecular-mechanics and statistical-mechanics estimates.

▶ Computing the critical radius and barrier height

Take $\Delta G(R) = -\tfrac{4}{3} \pi R^3 \Delta p + 4 \pi R^2 \sigma$ and differentiate:

$\frac{d \Delta G}{dR} = -4 \pi R^2 \Delta p + 8 \pi R \sigma.$

Setting this to zero:

$-4 \pi R^2 \Delta p + 8 \pi R \sigma = 0 \implies R = \frac{2 \sigma}{\Delta p} \equiv R^*.$

The second derivative is $-8 \pi R \Delta p + 8 \pi \sigma$ , which at $R^*$ evaluates to $-8 \pi R^* \Delta p + 8 \pi \sigma = -16 \pi \sigma + 8 \pi \sigma = -8 \pi \sigma < 0$ , confirming $R^*$ is a maximum of $\Delta G$ .

Plugging $R^*$ back into $\Delta G$ :

$\Delta G^* = -\frac{4}{3} \pi (R^*)^3 \Delta p + 4 \pi (R^*)^2 \sigma = -\frac{4}{3} \pi \frac{8 \sigma^3}{(\Delta p)^2} + 4 \pi \frac{4 \sigma^3}{(\Delta p)^2} \cdot \frac{1}{\Delta p} \cdot \Delta p.$

Cleaning up:

$\Delta G^* = \frac{-32 \pi \sigma^3 + 48 \pi \sigma^3}{3 (\Delta p)^2} = \frac{16 \pi \sigma^3}{3 (\Delta p)^2}.$

This is also equal to $\tfrac{4}{3} \pi (R^*)^2 \sigma = \tfrac{1}{3} \cdot 4 \pi (R^*)^2 \sigma$ — one third of the bubble’s surface energy at the critical radius. The barrier is the energy cost of building a critical-radius interface, less the volume energy gain at that radius.

For water at 20 °C, $\sigma = 72$ mN/m and $\Delta p = 1000$ atm $= 10^8$ Pa:

$R^* = \frac{2 \cdot 0.072}{10^8} = 1.4 \times 10^{-9}\ \text{m} = 1.4\ \text{nm},$

a few molecular diameters. The barrier is

$\Delta G^* = \frac{16 \pi (0.072)^3}{3 (10^8)^2} \approx 6.3 \times 10^{-19}\ \text{J} \approx 154\ k_B T,$

just above the practical $\sim 50 k_B T$ rate threshold. Doubling the tension to $\Delta p = 2000$ atm collapses $\Delta G^*$ to $\sim 38 k_B T$ and produces effectively instantaneous nucleation.

What homogeneous theory does and does not explain

The homogeneous nucleation theory succeeds in recovering the theoretical tensile strength of an idealised defect-free liquid. The agreement with the Briggs and Zheng experimental data — both of which used preparation techniques that approached the genuinely defect-free limit — is excellent.

It fails completely to explain why a tap-water sample tears at $-0.1$ atm. At that tension the barrier $\Delta G^*$ is roughly $10^{10}\ k_B T$ , and the homogeneous nucleation rate is approximately

$J \sim 10^{30} \cdot e^{-10^{10}} \approx 10^{30} \cdot e^{-2.3 \times 10^{10}} \approx 10^{30 - 10^{10}} \approx 10^{-10^{10}}$

events per cubic metre per second. The age of the universe is roughly $10^{17}$ seconds. Even integrated over the volume of the Pacific Ocean ( $10^{18}$ m³) for the age of the universe, we expect zero homogeneous nucleation events at this tension. And yet tap water tears at fractions of an atmosphere of tension all the time.

The discrepancy tells us that all observed cavitation at modest tensions must be heterogeneous — driven by preexisting defects in the sample that lower the barrier locally below the homogeneous threshold. The next lesson develops the most important heterogeneous mechanism: gas pockets trapped in surface crevices.