The Ginzburg Criterion: Balancing Fluctuations and Mean-Field Theory

Welcome to the next lesson in our course on Advanced Statistical Mechanics.

In our previous session, we made a significant leap beyond mean-field theory by incorporating Gaussian fluctuations. By expanding the Landau-Ginzburg functional to second order, we derived the Ornstein-Zernike form for the two-point correlation function, $G (q) \propto 1/ (q^{2} + ξ^{- 2})$ , and identified the correlation length $ξ$ . This framework gave us a quantitative handle on the spatial structure of fluctuations near a critical point, although it still yielded the classical mean-field exponents $ν = 1/2$ and $η = 0$ .

Today, we will turn this framework back on itself to ask a crucial question: under what conditions is the mean-field approximation, which neglects fluctuations, self-consistent? We will derive the celebrated Ginzburg criterion, which compares the magnitude of the very fluctuations we calculated to the mean-field order parameter. This criterion will not only delineate the domain of validity for mean-field theory but also introduce the pivotal concept of the upper critical dimension, explaining why mean-field theory provides an excellent description for some physical systems (like conventional superconductors) yet fails for others (like magnets or fluids near their critical point).

The Physical Argument: When Do Fluctuations Matter?

The mean-field approximation, at its core, replaces the influence of all other degrees of freedom on a given one with their average value, thereby neglecting fluctuations. This approximation is justifiable only if the fluctuations are, in some sense, "small" compared to the average value itself. But how do we quantify this?

The key insight, developed by Ginzburg and others, is to compare the thermal fluctuations of the order parameter within a correlation volume to the mean-field value of the order parameter. A correlation volume is a region of size $ξ^{d}$ , where $ξ$ is the correlation length. Inside this volume, fluctuations are strongly correlated, acting in concert; outside, they are essentially independent. The mean-field theory should be reliable if the typical deviation from the mean within one such coherently fluctuating block is much smaller than the mean itself.

The following video provides an excellent and intuitive walk-through of this argument. It formalizes the criterion and then uses a powerful heuristic argument based on "patches" of correlated spins to arrive at the final result.

Landau Ginzburg theory of Phase Transitions

In this video, Prof. Jos Thijssen explains the physical basis of the Ginzburg criterion. He begins by identifying the neglected second-order fluctuation terms in the saddle-point approximation and then develops a brilliant argument to estimate their magnitude.

Please watch from this segment. The core logic unfolds as follows: The Criterion Stated: The saddle-point approximation is justified if the expectation value of the squared fluctuation, \langle (\delta m)^2 \rangle, is much smaller than the squared mean-field order parameter, m_0^2. Relating Fluctuations to Correlations: He shows how the total fluctuation of the global magnetization is related to the integral of the two-point correlation function. The "Patch" Argument: This is the crucial intuitive step. The system is pictured as being composed of V/\xi^d independent patches, each of volume \xi^d. He uses this to relate the fluctuation of the local order parameter, \langle (\delta m(\mathbf{r}))^2 \rangle, to the correlation length \xi. Final Comparison: Finally, he compares the calculated fluctuation magnitude to the mean-field order parameter m_0^2 \propto |t|, leading to a condition on the dimensionality d.

A More Formal Derivation

Prof. Thijssen's argument provides the essential physical picture. Now, let's formalize it with a more direct calculation, following the clear and rigorous approach found in modern statistical field theory texts. The Ginzburg criterion, in its most common form, states that mean-field theory is valid if the thermal fluctuation of the order parameter, averaged over a correlation volume, is small compared to the squared mean-field order parameter.

Let's work below $T_{c}$ , where we have a non-zero order parameter $m_{0}$ . We require:

⟨ ϕ^{2} ⟩_{ξ^{d}} ≪ m_{0}^{2}

Here, $⟨ ϕ^{2} ⟩_{ξ^{d}}$ is the mean-squared fluctuation within a volume $ξ^{d}$ . A simpler, but equivalent, condition is to compare the total fluctuation in a correlation volume with the order parameter itself. The fluctuation contribution to the free energy in a correlation volume should be smaller than the mean field value of the free energy in that volume. A very direct way to check for self-consistency is to demand that the mean square fluctuation of the order parameter at a single point, $⟨ ϕ^{2} (r)⟩$ , is small compared to the mean-field order parameter squared, $m_{0}^{2}$ . However, since $⟨ ϕ^{2} ⟩$ is an average over the entire volume, a more robust condition compares the average fluctuation inside a correlation-sized region to the order parameter.

Let's adopt the crisp argument laid out in the lecture notes from Oxford University.

Chapter 7 Landau theory

This section of the lecture notes provides a direct and elegant derivation of the Ginzburg criterion by explicitly comparing the size of the fluctuation term to the mean-field term within a correlation volume.

Please read the entirety of Section 7.7. The argument proceeds as follows: It starts by formulating the criterion physically: the sum of correlations over a volume \xi^d must be smaller than the mean-field order parameter contribution in that same volume. This is stated in Eq. (7.61) and the subsequent paragraph. The left-hand side, representing the integrated fluctuation, is evaluated. As you know from our last lesson, this integral is simply the Fourier transform of the correlator at zero momentum, \tilde{G}(k=0), which is equal to \xi^2/\kappa. This is shown in the text leading up to the equation after (7.62). The right-hand side, the mean-field contribution, is expressed in terms of the Landau-Ginzburg parameters and the correlation length. Since m_0^2 = \langle M \rangle^2 = -b/(2c) and \xi^2 = \kappa/(-2b) (for T<T_c), this term becomes \xi^d m_0^2 = (\kappa/4c)\xi^{d-2}. Putting these together yields the Ginzburg criterion in its final form: \xi^{4-d} \ll \kappa^2/(4c).

The Upper Critical Dimension and Physical Consequences

The result $ξ^{4 - d} ≪ const.$ is one of the most important in the theory of critical phenomena. Let's unpack its consequences. The right-hand side involves the microscopic parameters of our model ( $κ, c$ ) and is essentially a constant. The left-hand side contains the correlation length $ξ$ , which diverges as we approach the critical point ( $T \to T_{c}$ ). The behavior of the inequality depends critically on the sign of the exponent $4 - d$ .

Case 1: $d > 4$ (High Dimensions)
The exponent $4 - d$ is negative. As $T \to T_{c}$ , $ξ \to \infty$ , and therefore $ξ^{4 - d} \to 0$ . This means the Ginzburg criterion becomes more satisfied as we get closer to the critical point. Fluctuations become progressively irrelevant. In this regime, mean-field theory is not just an approximation; it becomes exact at the critical point. The dimension $d_{u} = 4$ is called the upper critical dimension for this class of models (described by a scalar order parameter and a $ϕ^{4}$ interaction).
Case 2: $d < 4$ (Low Dimensions, including our own)
The exponent $4 - d$ is positive. As $T \to T_{c}$ , $ξ \to \infty$ , and the left side of the inequality, $ξ^{4 - d}$ , also diverges. No matter how small the fluctuations are far from the transition, the inequality will inevitably be violated once we get close enough to $T_{c}$ . In this regime, fluctuations dominate near the critical point, and mean-field theory breaks down, yielding incorrect critical exponents.

This distinction resolves a major puzzle. Why does mean-field theory work so well for some systems but not others? The answer often lies in the effective microscopic length scale $ξ_{0}$ that sets the scale for the correlation length $ξ = ξ_{0} ∣ t ∣^{- ν}$ . The Ginzburg criterion defines a reduced temperature, $t_{G}$ , below which fluctuations dominate. A rough estimate gives $∣ t_{G} ∣ \sim (ξ_{0} / a)^{- 2 d / (4 - d)}$ , where $a$ is the lattice spacing.

For systems like the liquid-gas transition or ferromagnets, the microscopic length scale $ξ_{0}$ is on the order of the interatomic spacing ( $ξ_{0} \sim a$ ). This makes $t_{G}$ of order 1, meaning the fluctuation-dominated regime is wide and experimentally accessible.
For a conventional (BCS) superconductor, the fundamental entities are Cooper pairs. The size of a Cooper pair, which sets the microscopic length scale $ξ_{0}$ , can be hundreds or thousands of Angstroms ( $ξ_{0} ≫ a$ ). This enormous value of $ξ_{0}$ makes $t_{G}$ incredibly small (e.g., $1 0^{- 14} - 1 0^{- 16}$ ). The fluctuation-dominated regime is experimentally inaccessible, and mean-field theory (in the form of Ginzburg-Landau or BCS theory) works essentially perfectly.

You can consolidate this understanding with another excellent resource.

[PDF] Chapter 2 Ginzburg-Landau Phenomenology

This text from Cambridge University provides a concise summary of the Ginzburg criterion and its physical implications, perfectly complementing the previous derivation.

Please read section 2.9 Ginzburg Criterion. It poses the central question about superconductors and derives the criterion by comparing fluctuation corrections to the mean-field specific heat jump. While the method is slightly different, the conclusion is identical. Pay close attention to the final discussion which explicitly contrasts the cases of superfluids and superconductors, explaining the role of the microscopic length scale \xi_0.

Conclusion

In this lesson, we have established the boundary for the validity of mean-field theory by deriving the Ginzburg criterion. This criterion is not just a mathematical curiosity; it is a profound statement about the physics of collective behavior.

Here are the key takeaways:

The Ginzburg criterion provides a self-consistency check for mean-field theory by comparing the magnitude of thermal fluctuations to the mean-field order parameter.
Mean-field theory breaks down when $ξ^{4 - d}$ becomes large, where $ξ$ is the correlation length and $d$ is the spatial dimension.
This defines an upper critical dimension, $d_{u} = 4$ . Above this dimension, fluctuations are irrelevant at the critical point, and mean-field theory is exact. Below it, fluctuations dominate, and mean-field theory fails.
The width of the critical region where fluctuations dominate is determined by the Ginzburg temperature, $t_{G}$ . For systems with a large microscopic length scale $ξ_{0}$ , like superconductors, this region is experimentally inaccessible, explaining the success of mean-field descriptions.

We started with a mean-field theory, used it to calculate fluctuations, and then used those fluctuations to find the limit of the original theory. This powerful idea marks the end of our analysis using the simple Gaussian approximation. We now understand why and when we must go further. To get the correct critical exponents in $d < 4$ , we must tackle the non-Gaussian $ϕ^{4}$ term in the Landau-Ginzburg functional, which we have so far neglected.

In the next lesson, we will begin developing the tools for this task. We will introduce a source term into the functional integral to define a generating functional, a powerful object that allows us to systematically calculate all correlation functions. This will be our first step towards a perturbative treatment of the interacting theory.

Recommended resources

The Physical Argument: When Do Fluctuations Matter?

A More Formal Derivation

The Upper Critical Dimension and Physical Consequences

Conclusion