First- and Second-Order Phase Transitions
CSIR NET Part C: Ehrenfest and modern classification, order parameters, critical exponents, scaling laws, Landau theory, van der Waals, nucleation.
- 20 min
On this page · 16 sections
- Prerequisites
- The Hook: Water That Will Not Boil, Iron That Forgets
- Two Classifications
- The Order Parameter
- Critical Exponents
- Landau Theory of Continuous Transitions
- Landau Theory of First-Order Transitions
- The van der Waals Fluid: a Worked Mean-Field Model
- Nucleation: Why First-Order Transitions Need a Seed
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Thermodynamic Potentials, Maxwell Relations and Phase Equilibria: Gibbs free energy, Clausius–Clapeyron and the phase rule. This page starts where that one stops.
- Canonical and Grand Canonical Ensembles: \( F = -k_BT\ln Z \) and response functions as derivatives of a free energy.
- Taylor expansion and the minimum of a polynomial: Landau theory is nothing more than that.
The Hook: Water That Will Not Boil, Iron That Forgets
Two Classifications
Ehrenfest: count the derivatives
A phase transition is a point where the Gibbs free energy \( G(T, P) \) (or \( G(T, h) \) for a magnet) is non-analytic. Ehrenfest classified transitions by the lowest derivative of \( G \) that is discontinuous:
| Order | Discontinuous quantity | Consequence |
|---|---|---|
| First | \( S = -\partial G/\partial T \), \( V = \partial G/\partial P \) (or \( M = -\partial G/\partial h \)) | latent heat \( L = T\Delta S \), volume jump, coexistence |
| Second | \( C_P = -T\,\partial^2G/\partial T^2 \), \( \alpha \), \( \kappa_T \) (or \( \chi \)) | no latent heat; response functions jump |
| \( n \)-th | \( n \)-th derivative | rarely useful beyond 2 |
For a first-order line, equality of \( G \) on both sides gives Clausius–Clapeyron:
$$\boxed{\frac{dP}{dT} = \frac{L}{T\,\Delta v}}$$For a second-order line, \( \Delta s = \Delta v = 0 \) and Clausius–Clapeyron becomes 0/0. Equating \( ds \) and \( dv \) on both sides instead gives the Ehrenfest equations:
$$\boxed{\frac{dP}{dT} = \frac{\Delta C_P}{TV\,\Delta\alpha} = \frac{\Delta\alpha}{\Delta\kappa_T}}$$where \( \alpha = V^{-1}(\partial V/\partial T)_P \) and \( \kappa_T = -V^{-1}(\partial V/\partial P)_T \).
The modern view: first-order versus continuous
Ehrenfest’s scheme assumes finite jumps. Real continuous transitions usually have divergent response functions (the \( \lambda \)-shaped heat capacity of helium-4 at about 2.17 K, the susceptibility of a ferromagnet at \( T_c \)), which the scheme does not describe well. The working classification today is:
| First-order | Continuous (“second-order”) | |
|---|---|---|
| Order parameter at the transition | jumps | goes to zero continuously |
| Latent heat | yes | no |
| Coexistence of phases | yes | no |
| Metastability, hysteresis, supercooling | yes | no |
| Correlation length \( \xi \) | finite | diverges |
| Fluctuations | modest | large (critical opalescence) |
| Typical examples | melting, boiling below \( T_c \), ferromagnet below \( T_c \) as \( h \) changes sign, superconductor in a field | liquid–gas critical point, Curie point in zero field, superconductor at \( T_c \) in zero field, superfluid \( \lambda \) point |
The Order Parameter
Landau’s idea: a continuous transition breaks a symmetry, and an order parameter measures how much. It is zero in the symmetric (usually high-temperature) phase and non-zero in the ordered phase.
| Transition | Order parameter | Symmetry broken |
|---|---|---|
| Ferromagnet (Curie point) | magnetization \( M \) | spin reversal (Ising) or rotation (Heisenberg) |
| Antiferromagnet (Néel point) | staggered magnetization | sublattice exchange |
| Liquid–gas critical point | \( \rho_l - \rho_g \) | none exactly; effective particle–hole symmetry |
| Order–disorder alloy | sublattice occupation difference | sublattice exchange |
| Superfluid helium-4, BEC | condensate wavefunction \( \psi \) | global phase |
| Superconductor | gap \( \Delta \) (pair amplitude) | gauge (phase) |
| Nematic liquid crystal | orientational order \( S \) | rotation |
Critical Exponents
Near a continuous transition, define the reduced temperature \( t = (T - T_c)/T_c \). Thermodynamic quantities follow power laws:
| Exponent | Definition | Mean-field | 2D Ising (exact) |
|---|---|---|---|
| \( \alpha \) | \( C \sim \lvert t\rvert^{-\alpha} \) | 0 (jump) | 0 (logarithm) |
| \( \beta \) | \( m \sim (-t)^{\beta} \), \( t < 0 \) | 1/2 | 1/8 |
| \( \gamma \) | \( \chi \sim \lvert t\rvert^{-\gamma} \) | 1 | 7/4 |
| \( \delta \) | \( h \sim m^{\delta} \) at \( t = 0 \) | 3 | 15 |
| \( \nu \) | \( \xi \sim \lvert t\rvert^{-\nu} \) | 1/2 | 1 |
| \( \eta \) | \( G(r) \sim r^{-(d - 2 + \eta)} \) at \( t = 0 \) | 0 | 1/4 |
For the three-dimensional Ising class (which includes the liquid–gas critical point and uniaxial magnets) the exponents are known numerically, roughly \( \beta \approx 0.33 \), \( \gamma \approx 1.24 \), \( \nu \approx 0.63 \), \( \alpha \approx 0.11 \). They are neither the mean-field nor the 2D values.
Scaling relations
Only two exponents are independent. The others follow from the scaling hypothesis:
$$\boxed{\alpha + 2\beta + \gamma = 2 \ \ \text{(Rushbrooke)}} \qquad \boxed{\gamma = \beta(\delta - 1) \ \ \text{(Widom)}}$$$$\boxed{\gamma = \nu(2 - \eta) \ \ \text{(Fisher)}} \qquad \boxed{\nu d = 2 - \alpha \ \ \text{(Josephson, hyperscaling)}}$$Check with 2D Ising: \( 0 + \frac{1}{4} + \frac{7}{4} = 2 \); \( \frac{1}{8}\times14 = \frac{7}{4} \); \( 1\times\frac{7}{4} = \frac{7}{4} \); \( 1\times2 = 2 - 0 \). Mean-field exponents satisfy the first three in any dimension, but hyperscaling only at \( d = 4 \) (\( \frac{1}{2}\times4 = 2 \)). That is a signal that mean-field theory is exact only for \( d \geq 4 \), the upper critical dimension for Ising-like systems.
Universality
Exponents depend only on
- the spatial dimension \( d \),
- the symmetry (number of components \( n \)) of the order parameter,
- whether the interactions are short-range.
They do not depend on the lattice, the strength of the coupling or the chemistry. That is why a fluid’s critical point and a uniaxial magnet share exponents. The critical temperature, by contrast, is non-universal: it depends on every microscopic detail.
Landau Theory of Continuous Transitions
Near \( T_c \) the order parameter \( m \) is small. Write the free-energy density as a power series respecting the symmetry \( m \to -m \) (no odd powers when \( h = 0 \)):
$$\boxed{f(m, T) = f_0(T) + a(T - T_c)\,m^2 + b\,m^4 - h\,m}, \qquad a > 0,\ b > 0$$The equilibrium \( m \) minimises \( f \).
Zero field. \( \partial f/\partial m = 2a(T - T_c)m + 4bm^3 = 0 \):
- \( T > T_c \): only \( m = 0 \).
- \( T < T_c \): \( m^2 = \dfrac{a(T_c - T)}{2b} \), so \( m \propto (T_c - T)^{1/2} \): \( \beta = \frac{1}{2} \).
Free energy and specific heat. Below \( T_c \), substituting the minimum gives \( f - f_0 = -\dfrac{a^2(T_c - T)^2}{4b} \). Then \( c = -T\,\partial^2 f/\partial T^2 \) gains an extra \( Ta^2/2b \) below \( T_c \):
$$\boxed{\Delta c = \frac{a^2T_c}{2b}} \quad \text{(a finite jump: } \alpha = 0\text{)}$$The entropy \( s = -\partial f/\partial T \) is continuous at \( T_c \): no latent heat, as a second-order transition requires.
Susceptibility. With a small field, \( h = 2a(T - T_c)m + 4bm^3 \). Then \( \chi^{-1} = \partial h/\partial m \):
$$\chi = \frac{1}{2a(T - T_c)}\ (T > T_c), \qquad \chi = \frac{1}{4a(T_c - T)}\ (T < T_c)$$So \( \gamma = 1 \), and the amplitude above is twice the amplitude below at equal \( \lvert T - T_c\rvert \). The first form is the Curie–Weiss law.
Critical isotherm. At \( T = T_c \), \( h = 4bm^3 \): \( \delta = 3 \).
Landau Theory of First-Order Transitions
Two ways make Landau’s free energy produce a jump.
A negative quartic term
$$f = \alpha(T - T_0)m^2 - c\,m^4 + e\,m^6, \qquad c > 0,\ e > 0$$Now \( m = 0 \) is locally stable for every \( T > T_0 \), but a second minimum at finite \( m \) can become lower first. At the transition \( T_1 \) the ordered minimum has \( f = 0 \) and \( \partial f/\partial m = 0 \). Writing \( A = \alpha(T_1 - T_0) \):
$$A - cm^2 + em^4 = 0, \qquad A - 2cm^2 + 3em^4 = 0$$Subtracting gives \( m^2 = c/2e \), and then \( A = c^2/4e \):
$$\boxed{T_1 = T_0 + \frac{c^2}{4\alpha e}, \qquad m_1^2 = \frac{c}{2e}}$$The order parameter jumps from 0 to \( m_1 \). The entropy \( s = -\partial f/\partial T = -\alpha m^2 \) also jumps by \( \alpha m_1^2 \), giving a latent heat \( L = T_1\alpha m_1^2 \) per unit volume. Between \( T_0 \) and the upper spinodal the two minima coexist: superheating and supercooling.
A cubic term
If symmetry allows an \( m^3 \) term (it does for the nematic–isotropic transition of liquid crystals, whose order parameter has no \( m \to -m \) symmetry), \( f = a(T - T^*)m^2 - wm^3 + bm^4 \) always gives a first-order jump. The rule: if the symmetry permits a cubic invariant, Landau theory predicts a first-order transition.
graph TD
A["Write f(m) allowed by symmetry"] --> B{"Cubic term allowed?"}
B -- "yes" --> C["First order: m jumps, latent heat"]
B -- "no" --> D{"Sign of quartic coefficient"}
D -- "positive" --> E["Continuous: m ~ (Tc - T)^1/2"]
D -- "negative (need m^6 > 0)" --> C
D -- "zero" --> F["Tricritical point: beta = 1/4"]At the tricritical point (\( c = 0 \)) the free energy is \( \alpha(T - T_0)m^2 + em^6 \), giving \( m^4 \propto (T_0 - T) \), so \( \beta = \frac{1}{4} \).
The van der Waals Fluid: a Worked Mean-Field Model
$$\left(P + \frac{a}{v^2}\right)(v - b) = RT$$per mole. The critical point is where the isotherm has an inflection with zero slope, \( \partial P/\partial v = \partial^2P/\partial v^2 = 0 \):
$$\boxed{v_c = 3b, \qquad P_c = \frac{a}{27b^2}, \qquad T_c = \frac{8a}{27Rb}, \qquad Z_c = \frac{P_cv_c}{RT_c} = \frac{3}{8}}$$In reduced variables (\( \tilde P = P/P_c \) and so on) the equation becomes universal: \( (\tilde P + 3/\tilde v^2)(3\tilde v - 1) = 8\tilde T \). This is the law of corresponding states.
Below \( T_c \) the isotherm has a wiggle with \( \partial P/\partial v > 0 \), which is mechanically unstable. The Maxwell equal-area construction replaces the wiggle by a horizontal line at the vapour pressure, chosen so that the areas above and below are equal (equivalently, \( \mu_l = \mu_g \)). The endpoints of that line are the coexisting liquid and vapour; the stretches between the line endpoints and the extrema (spinodals) are metastable (superheated liquid, supercooled vapour).
Expanding about the critical point gives mean-field exponents: \( \rho_l - \rho_g \propto (T_c - T)^{1/2} \), \( \kappa_T \propto \lvert t\rvert^{-1} \), \( P - P_c \propto (\rho - \rho_c)^3 \) on the critical isotherm. Real fluids have \( \beta \approx 0.33 \): van der Waals is qualitatively right, quantitatively mean-field.
Nucleation: Why First-Order Transitions Need a Seed
Below a first-order transition, the new phase has lower bulk free energy per volume, \( \Delta g > 0 \), but creating an interface costs \( \sigma \) per area. For a spherical droplet of radius \( r \):
$$\Delta G(r) = 4\pi r^2\sigma - \frac{4}{3}\pi r^3\Delta g$$The maximum is at
$$\boxed{r^* = \frac{2\sigma}{\Delta g}, \qquad \Delta G^* = \frac{16\pi\sigma^3}{3\Delta g^2}}$$Droplets smaller than \( r^* \) shrink; larger ones grow. Close to the transition \( \Delta g \propto \lvert T - T_1\rvert \) is small, so \( \Delta G^* \) is huge and the system can stay metastable for a long time: supercooled water, superheated liquid in a bubble chamber. Dust and walls lower the barrier (heterogeneous nucleation). Continuous transitions have no barrier and no metastability.
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| First order: \( S \), \( V \), \( M \) jump; latent heat | Continuous: order parameter continuous; response functions diverge or jump | |
| \( \beta \): order-parameter exponent | \( \beta = 1/k_BT \) in the partition function | |
| Critical exponents: universal | \( T_c \): non-universal, depends on couplings and lattice | |
| Spinodal: limit of metastability, \( \partial P/\partial v = 0 \) | Binodal (coexistence): equal chemical potentials, Maxwell construction | |
| Tricritical \( \beta = \frac{1}{4} \) (quartic coefficient zero) | Ordinary mean-field \( \beta = \frac{1}{2} \) | |
| Clausius–Clapeyron (first order, \( L/T\Delta v \)) | Ehrenfest equations (second order, \( \Delta C_P/TV\Delta\alpha \)) | |
| Superconductor at \( T_c \) in zero field: second order | Superconductor in an applied field: first order with latent heat |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. Take \( R = 8.314 \) J mol\( ^{-1} \) K\( ^{-1} \).
Level 1: Foundation
- At a first-order transition, which derivatives of the Gibbs free energy are discontinuous?
- What is the mean-field value of the exponent \( \delta \)?
- Name the order parameter of the liquid–gas critical point.
Answers
- The first derivatives: entropy and volume (or magnetization). Their jumps give the latent heat \( L = T\Delta S \) and the volume change.
- \( \delta = 3 \): at \( T = T_c \), \( h = 4bm^3 \).
- The density difference \( \rho_l - \rho_g \), which vanishes at the critical point.
Level 2: Part B Style
Practice Q1. At a certain transition, the entropy and volume are continuous but the isobaric heat capacity and the thermal expansion coefficient change discontinuously. In Ehrenfest’s scheme the transition is:
(a) first order, with a latent heat (b) second order (c) third order (d) not a phase transition
Answer: (b) second order
The first derivatives of \( G \) (entropy, volume) are continuous, so there is no latent heat. The lowest discontinuous derivatives are the second ones (\( C_P = -T\partial^2G/\partial T^2 \), \( \alpha \propto \partial^2G/\partial T\partial P \)). Ehrenfest second order.
Practice Q2. A liquid boils at 350 K under 1 atm and has a molar enthalpy of vaporisation of 40 kJ mol\( ^{-1} \), independent of temperature. Treating the vapour as ideal and neglecting the liquid volume, its boiling point under 0.5 atm is closest to:
(a) 318 K (b) 333 K (c) 342 K (d) 175 K
Answer: (b) 333 K
Integrated Clausius–Clapeyron: \( \ln(P_2/P_1) = -\dfrac{L}{R}\left(\dfrac{1}{T_2} - \dfrac{1}{T_1}\right) \).
\( \dfrac{1}{T_2} = \dfrac{1}{350} + \dfrac{R\ln2}{L} = 2.857\times10^{-3} + \dfrac{8.314\times0.693}{40000} = 2.857\times10^{-3} + 1.441\times10^{-4} = 3.001\times10^{-3} \) K\( ^{-1} \).
\( T_2 \approx 333.2 \) K. Option (d) assumes \( T \propto P \), which is not how vapour pressure works.
Practice Q3. In Landau theory with \( f = a(T - T_c)m^2 + bm^4 - hm \) (\( a, b > 0 \)), the ratio of the zero-field susceptibility at \( T_c + \Delta T \) to that at \( T_c - \Delta T \) is:
(a) 1/2 (b) 1 (c) 2 (d) 4
Answer: (c) 2
Above: \( \chi = 1/[2a\Delta T] \). Below, with \( m^2 = a\Delta T/2b \): \( \chi^{-1} = -2a\Delta T + 12bm^2 = -2a\Delta T + 6a\Delta T = 4a\Delta T \). Ratio \( 4a\Delta T/2a\Delta T = 2 \).
Practice Q4. For a magnetic system the measured exponents are \( \alpha = 0.11 \) and \( \beta = 0.33 \). Assuming the scaling relations hold, \( \gamma \) and \( \delta \) are closest to:
(a) 1.23 and 4.7 (b) 1.00 and 3.0 (c) 1.56 and 5.7 (d) 1.23 and 3.7
Answer: (a) 1.23 and 4.7
Rushbrooke: \( \gamma = 2 - \alpha - 2\beta = 2 - 0.11 - 0.66 = 1.23 \). Widom: \( \delta = 1 + \gamma/\beta = 1 + 1.23/0.33 = 4.73 \). (b) is the mean-field set; (d) uses \( \delta = \gamma/\beta \) without the 1.
Level 3: Part C Style
Practice Q5. A system has the Landau free-energy density (in consistent units)
$$f(m) = (T - 100)\,m^2 - 2m^4 + m^6$$Consider the statements:
- For \( T \) slightly above 100, \( m = 0 \) is a local minimum of \( f \).
- The equilibrium transition occurs at \( T = 101 \).
- At the transition the order parameter jumps from 0 to \( \lvert m\rvert = 1 \).
- The latent heat per unit volume at the transition is 101 (in the same units).
Which are correct?
(a) 1 and 2 only (b) 2 and 3 only (c) 1, 2 and 3 only (d) 1, 2, 3 and 4
Answer: (d) 1, 2, 3 and 4
Here \( \alpha = 1 \), \( T_0 = 100 \), \( c = 2 \), \( e = 1 \).
- Near \( m = 0 \), \( f \approx (T - 100)m^2 \), which has a minimum for \( T > 100 \). True.
- \( T_1 = T_0 + c^2/4\alpha e = 100 + 4/4 = 101 \). True. (Check: at \( T = 101 \), \( f = m^2 - 2m^4 + m^6 = m^2(1 - m^2)^2 \geq 0 \), with a second zero at \( m^2 = 1 \).)
- \( m_1^2 = c/2e = 1 \). True.
- \( s = -\partial f/\partial T = -m^2 \), so \( \Delta s = m_1^2 = 1 \) and \( L = T_1\Delta s = 101 \). True.
Practice Q6. The vapour pressure of a liquid is measured:
| \( T \) (K) | 300 | 320 | 340 |
|---|---|---|---|
| \( P \) (kPa) | 10.0 | 21.2 | 41.2 |
Assuming ideal vapour, negligible liquid volume and constant latent heat, the molar enthalpy of vaporisation is closest to:
(a) 15 kJ mol\( ^{-1} \) (b) 30 kJ mol\( ^{-1} \) (c) 45 kJ mol\( ^{-1} \) (d) 60 kJ mol\( ^{-1} \)
Answer: (b) 30 kJ/mol
\( \ln P \) against \( 1/T \) has slope \( -L/R \). Using the end points:
\( L = R\,\dfrac{\ln(41.2/10.0)}{1/300 - 1/340} = 8.314\times\dfrac{1.416}{3.922\times10^{-4}} \approx 3.00\times10^4 \) J mol\( ^{-1} \).
The middle point gives the same slope (\( \ln(21.2/10.0) = 0.751 \), \( 1/300 - 1/320 = 2.083\times10^{-4} \), \( L \approx 30.0 \) kJ mol\( ^{-1} \)), confirming that \( L \) is constant over the range.
Practice Q7. A gas obeys the van der Waals equation with \( a = 0.364 \) Pa m\( ^6 \) mol\( ^{-2} \) and \( b = 4.27\times10^{-5} \) m\( ^3 \) mol\( ^{-1} \). Consider:
- Its critical temperature is about 304 K.
- Its critical pressure is about 7.4 MPa.
- Its critical compressibility factor \( P_cv_c/RT_c \) is 0.375, whatever \( a \) and \( b \) are.
- Near its critical point the model predicts \( \rho_l - \rho_g \propto (T_c - T)^{1/3} \).
Which are correct?
(a) 1 and 2 only (b) 1, 2 and 3 only (c) 3 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
- \( T_c = \dfrac{8a}{27Rb} = \dfrac{8\times0.364}{27\times8.314\times4.27\times10^{-5}} \approx 303.8 \) K. True.
- \( P_c = \dfrac{a}{27b^2} = \dfrac{0.364}{27\times(4.27\times10^{-5})^2} \approx 7.39\times10^6 \) Pa. True.
- \( Z_c = 3/8 \) for every van der Waals gas. True.
- Van der Waals is a mean-field theory: \( \beta = \frac{1}{2} \), not \( \frac{1}{3} \). The value near \( \frac{1}{3} \) is what real fluids show. False.
Quick Revision Table
| Idea | Formula or fact |
|---|---|
| First order | \( S \), \( V \) jump; \( dP/dT = L/T\Delta v \) |
| Second order (Ehrenfest) | \( dP/dT = \Delta C_P/TV\Delta\alpha = \Delta\alpha/\Delta\kappa_T \) |
| Landau (continuous) | \( f = a(T - T_c)m^2 + bm^4 \); \( m^2 = a(T_c - T)/2b \) |
| Landau specific-heat jump | \( \Delta c = a^2T_c/2b \) |
| Landau susceptibility | \( 1/2a(T - T_c) \) above, \( 1/4a(T_c - T) \) below |
| Mean-field exponents | \( \alpha = 0 \), \( \beta = \frac{1}{2} \), \( \gamma = 1 \), \( \delta = 3 \), \( \nu = \frac{1}{2} \), \( \eta = 0 \) |
| 2D Ising exponents | \( 0 \) (log), \( \frac{1}{8} \), \( \frac{7}{4} \), 15, 1, \( \frac{1}{4} \) |
| Scaling | \( \alpha + 2\beta + \gamma = 2 \); \( \gamma = \beta(\delta - 1) \); \( \gamma = \nu(2 - \eta) \); \( \nu d = 2 - \alpha \) |
| First-order Landau (\( m^6 \)) | \( T_1 = T_0 + c^2/4\alpha e \), \( m_1^2 = c/2e \) |
| Tricritical | \( \beta = \frac{1}{4} \) |
| van der Waals | \( v_c = 3b \), \( P_c = a/27b^2 \), \( T_c = 8a/27Rb \), \( Z_c = 3/8 \) |
| Nucleation | \( r^* = 2\sigma/\Delta g \), \( \Delta G^* = 16\pi\sigma^3/3\Delta g^2 \) |
Key Takeaways
What’s Next
- Diamagnetism, Paramagnetism and Ferromagnetism: the Weiss mean-field ferromagnet as the textbook continuous transition.
- The Ising Model: where the 2D exponents in the table above come from, and why 1D has no transition.
- Bose–Einstein Condensation: a phase transition with no interactions at all.
- Phase diagrams, the phase rule and real-gas behaviour from the chemist’s side: Phase Equilibria and the Phase Rule and Ideal and Real Gases.