Find the mechanism before the formula
A fluid can remain apparently uniform after crossing coexistence because a new phase still has to create an interface. Move farther into the metastable region and the barrier shrinks; at a thermodynamic spinodal the uniform branch loses local stability. These are different boundaries with different evidence.
We will connect free-energy curvature to compressibility, distinguish binodal, metastable and spinodal states, calculate a classical spherical nucleation barrier, and show why the calculation must not be extrapolated all the way to the spinodal.
This guide belongs to the Gold Standard of Thermodynamics, the Gold Standard of Education and the How X Works hub.
Choose a reading route
Start with the question closest to the decision you need to make. Every route returns to the same worked evidence and transfer checks.
| Route | Chapters | Use it when |
|---|---|---|
| 1 | 1–3 | You need free-energy curvature, stability and phase-boundary definitions. |
| 2 | 4–5 | You want EOS branches and a classical nucleation calculation. |
| 3 | 6–8 | You need spinodal limits and a changed-driving-force test. |
| 4 | 9–10 | You must control kinetics, heterogeneity and model limits. |
| 5 | 11–12 | You want FAQs and a reproducible evidence checklist. |
Chapter contents
Open the group you need, then use the chapter links. Teaching sections remain open on the page.
Build the model
Test the mechanism
1. Local stability is a curvature question
For a one-component fluid at fixed temperature, local mechanical stability requires positive isothermal compressibility κT=−(1/v)(∂v/∂p)T>0. Equivalently, (∂p/∂v)T<0 or (∂p/∂ρ)T>0 on a stable or metastable single-phase branch.
At a mean-field spinodal, the relevant derivative vanishes and κT diverges. In a composition description, the analogous boundary is where the second derivative of the mixing free energy with respect to composition is zero.
The derivative is local. Phase coexistence is a global common-tangent or equal-chemical-potential condition, so the spinodal and coexistence curve should not be treated as the same line.
2. Separate binodal, metastable and unstable regions
The binodal marks equilibrium coexistence. Between the binodal and spinodal, a uniform phase is metastable: small fluctuations decay, but a sufficiently large nucleus of the stable phase can grow. Beyond the spinodal in a mean-field description, infinitesimal fluctuations lower free energy and separation begins without a classical nucleation barrier.
A metastable state is not an equilibrium phase selected by global free-energy minimisation. It is a local minimum whose lifetime depends on nucleation kinetics, impurities, surfaces, volume and observation time.
| Region | Curvature | Transformation route |
|---|---|---|
| Stable single phase | Locally and globally favoured | No phase change |
| Metastable | Local minimum, not global | Activated nucleation |
| Spinodal boundary | Zero restoring curvature | Limit of local stability |
| Unstable | Negative curvature direction | Spontaneous amplification |
3. An equation of state can continue onto metastable branches
A multiparameter or cubic equation of state can be evaluated beyond saturation on a metastable liquid or vapour branch until a spinodal derivative condition is reached. Numerical availability does not prove the extrapolated branch is accurate: metastable data are sparse and higher derivatives are sensitive.
The van der Waals loop is a teaching model. Equilibrium replaces the unstable loop by a coexistence construction, while the analytic EOS still exposes local extrema that act as mean-field spinodals. Modern fluids require a documented EOS and phase-selection rule.
- Record the EOS name and version.
- Identify the root or phase branch selected by the solver.
- Check κT, heat capacities and sound-speed stability.
- Do not use a two-phase equilibrium state as a homogeneous metastable state.
4. Classical nucleation balances volume gain and surface cost
For a spherical nucleus under the capillarity approximation, write ΔG(r)=4πr²σ−(4π/3)r³Δp, where σ is a constant interfacial tension and Δp is the positive bulk driving pressure in the chosen formulation. The surface term penalises small nuclei; the volume term favours growth.
Setting dΔG/dr=0 gives r*=2σ/Δp and ΔG*=16πσ³/(3Δp²). Nuclei smaller than r* tend to shrink in the idealised model; larger ones tend to grow.
Equivalent formulas use a chemical-potential or volumetric free-energy difference. Mixing conventions changes dimensions and is a serious error. State exactly which driving force appears in the equation.
5. Worked barrier: a nanometre-scale critical radius
Take σ=0.020 N m⁻¹, Δp=10 MPa and T=300 K. Then r*=2(0.020)/(1.0×10⁷)=4.0×10⁻⁹ m, or 4 nm. The barrier is ΔG*=16π(0.020)³/[3(1.0×10⁷)²]≈1.34×10⁻¹⁸ J.
At 300 K, kBT≈4.14×10⁻²¹ J, so this idealised barrier is about 324 kBT. A homogeneous event would be extremely rare unless a prefactor, large volume or heterogeneous site changes the practical rate.
The numbers illustrate scaling, not a fluid prediction. At four nanometres, curvature dependence of σ and molecular structure may already challenge the capillarity approximation.
6. A barrier is not yet a nucleation rate
Classical rate expressions take a form such as J=J₀ exp(−ΔG*/kBT). The prefactor J₀ contains attachment kinetics, density of candidate sites and dynamical factors. Two systems with the same barrier can nucleate at different rates.
The probability of observing at least one event also grows with volume and time. A long-lived microlitre and a rapidly observed nanolitre need not show the same apparent metastability limit.
Heterogeneous nucleation at walls, particles, scratches or dissolved gases usually reduces the barrier through geometry and surface chemistry. A measured onset is therefore often a kinetic and apparatus-dependent boundary, not the thermodynamic spinodal.
7. Classical droplets lose meaning near the spinodal
As the spinodal is approached, the restoring curvature tends to zero, susceptibility and correlation length increase, and the critical fluctuation becomes diffuse. Cahn and Hilliard showed that the work of formation approaches zero while the nucleus no longer resembles a sharp, homogeneous droplet.
This is precisely where extrapolating constant-σ classical nucleation formulas is least defensible. The formula can suggest a shrinking radius as driving force grows, while a diffuse-field treatment predicts a broad fluctuation structure near the spinodal.
A spinodal derived from an approximate EOS is model-dependent. Finite-range fluctuations can blur or shift the mean-field picture, and some real-fluid metastable limits are reached by nucleation before a thermodynamic spinodal is experimentally accessible.
8. Transfer check: double the driving pressure
Keep σ=0.020 N m⁻¹ and T=300 K but double Δp from 10 to 20 MPa. The classical radius halves from 4 nm to 2 nm. The barrier falls by a factor of four to about 3.35×10⁻¹⁹ J, or roughly 81 kBT.
The exponential rate can therefore change by many orders of magnitude even though the driving pressure changes only by a factor of two. This sensitivity explains why modest temperature, pressure or impurity shifts can move observed onset sharply.
The transfer also exposes the limit: at 2 nm, treating σ as a planar, size-independent interfacial tension is more questionable. Recalculate the numbers, then tighten the model rather than presenting the smaller radius as more precise.
9. Observed limits mix thermodynamics with protocol
Cavitation, boiling, condensation and crystallisation experiments report onset under a finite ramp rate, sample volume, surface condition and detection threshold. Faster ramps can carry a system deeper into metastability before an event is observed.
To compare data with an EOS spinodal, retain the entire protocol: temperature and pressure calibration, dwell time, volume, nucleation-site preparation, decompression or heating rate, and uncertainty. Repeated trials should be treated as a distribution of onset states, not one exact point.
Independent acoustic or response-function measurements can test the approach to instability, but they also require frequency and nonequilibrium interpretation.
10. Limits and misconceptions
- The binodal is not the spinodal.
- A metastable state can persist even though it is not the global free-energy minimum.
- A nucleation onset is not automatically a thermodynamic stability limit.
- The classical barrier assumes a sharp spherical interface and bulk properties.
- Heterogeneous sites can dominate a nominally homogeneous experiment.
- An EOS root is not trustworthy merely because a solver returns it.
Keep equilibrium, local stability and kinetics on separate lines of the argument. Confusing them can turn a protocol-dependent observation into a false claim about a universal phase boundary.
11. Frequently asked questions
Does κT literally become infinite in a real experiment?
The mean-field derivative diverges at the model spinodal. Finite size, dynamics, impurities, fluctuations and premature nucleation can prevent an experiment from reaching that mathematical limit.
Is spinodal decomposition the same as nucleation?
No. Nucleation crosses a finite barrier from a metastable state. Spinodal decomposition amplifies infinitesimal fluctuations in an unstable state without that classical barrier.
Can REFPROP return metastable states?
Its documented controls can evaluate metastable branches for supported fluids and inputs. Users must still identify the branch, validity range and proximity to the spinodal.
12. A reproducible metastability checklist
| Stage | Evidence to retain |
|---|---|
| State | T, p or ρ, composition and phase branch |
| EOS | Model, version and derivative method |
| Stability | κT or free-energy curvature and spinodal criterion |
| Nucleation | Driving-force convention, σ, r* and ΔG* |
| Protocol | Volume, time, ramp and surface preparation |
| Transfer | Sensitivity to σ, driving force and model choice |
The gold standard is a boundary-aware explanation: coexistence sets the equilibrium reference, curvature locates the model stability limit, and a separately stated kinetic model explains why transformation is observed earlier.
Reading routes · Previous chapter · Contents · Continue reading
Continue through the Gold Standard ecosystem
Return to the Gold Standard of Thermodynamics, the Gold Standard of Education parent, or the How X Works hub.
Read next in this subject
- Helmholtz-Energy Equations of State, Residual Properties and Consistency Checks
- Fundamental Derivative of Gas Dynamics, EOS Convexity and Nonclassical Waves
- Same Endpoints, Different Pressure–Volume Paths
- Polymer Science: Binodal Curves, Spinodal Limits and Quench Paths
- How Science Works: Spinodal Decomposition