Mean-Field Theory and the Upper Critical Dimension

Welcome to the final lesson of our module on the Renormalization Group. In our previous discussion, we saw how the RG framework provides a powerful explanation for universality, showing that microscopically distinct systems can flow to the same fixed point and thus share identical critical behavior. This entire picture hinges on the stability of fixed points and the relevance of operators.

Today, we will bring this line of reasoning to a satisfying conclusion by connecting it back to the mean-field theories we studied in an earlier module. We will investigate why these simple, elegant theories, which completely neglect fluctuations, sometimes give the correct critical exponents and sometimes fail spectacularly. The answer lies in the concept of the upper critical dimension, $d_{c}$ . You will learn how to determine this dimension for a given theory and understand, from two different but complementary perspectives, why it marks the boundary for the validity of mean-field descriptions.

The Limits of Mean-Field Theory

Let's begin by recalling the central approximation of mean-field theory: it replaces the complex, fluctuating environment of a given degree of freedom (like a single spin) with a single, static, average value—the "mean field." This simplification linearizes an otherwise intractable many-body problem, but it comes at the cost of ignoring fluctuations. Near a critical point, however, fluctuations become correlated over vast distances and can no longer be ignored.

The question then becomes: under what conditions is it actually safe to neglect these fluctuations? As it turns out, the answer depends crucially on the spatial dimensionality, $d$ , of the system.

Statistical Field Theory

This excerpt from David Tong's lecture notes on Statistical Field Theory concisely summarizes the established facts regarding the validity of mean-field theory (MFT) in different dimensions.

Please read the introductory paragraphs of the section titled 1.2.3 Validity of Mean Field Theory. Pay close attention to the bullet points that outline the success and failure of MFT as a function of dimension d. This reading introduces the key concepts of a lower critical dimension (d_l), below which ordered phases are destroyed by fluctuations, and an upper critical dimension (d_c), above which MFT becomes exact. For the Ising-like models we've been considering, d_c=4.

Our goal for this lesson is to understand the origin of this upper critical dimension, $d_{c} = 4$ . We will do this first using a powerful physical argument known as the Ginzburg criterion, and then show how the RG framework provides a deeper, more fundamental explanation.

The Ginzburg Criterion: When Do Fluctuations Matter?

The Ginzburg criterion provides a self-consistency check for mean-field theory. It compares the magnitude of the thermal fluctuations of the order parameter, which MFT neglects, to the value of the order parameter itself, as predicted by MFT. Mean-field theory can be considered valid only if the fluctuations are small compared to the mean value.

To make this quantitative, we work within the Landau-Ginzburg framework. The free energy functional for a spatially varying order parameter $m (r)$ is:

F [m (r)] = \int d^{d} r [f_{0} + a (T - T_{c}) m^{2} + b m^{4} + c ∣\nabla m ∣^{2}]

Here, the gradient term $c ∣\nabla m ∣^{2}$ penalizes spatial variations, or fluctuations. We want to compare the average size of a fluctuation, $⟨ δ m^{2} ⟩$ , within a characteristic volume, to the squared value of the mean-field order parameter, $m_{0}^{2}$ . The characteristic volume for fluctuations is set by the correlation length, $ξ$ , so we consider a volume $ξ^{d}$ .

Mean-field theory is valid if:

⟨(δ m)^{2} ⟩_{ξ^{d}} ≪ m_{0}^{2}

This is the Ginzburg criterion. Evaluating this condition leads directly to a condition on the dimensionality $d$ .

The following video provides a clear, step-by-step derivation of this result.

Landau Ginzburg theory of Phase Transitions

In this section of his lecture on Landau-Ginzburg theory, Professor Jos Thijssen presents a detailed derivation of the Ginzburg criterion.

Please watch the segment from the Ginzburg criterion. The key steps are: Formulating the criterion as a comparison between fluctuations and the mean-field solution. Relating the fluctuation of the global magnetization, \langle \Delta M^2 \rangle, to an integral over the correlation function G(r, r'). Using scaling arguments based on the correlation volume \xi^d to find the size of local fluctuations, \langle \delta m(r)^2 \rangle \sim \xi^{2-d}. Recalling the mean-field scaling for the order parameter, m_0^2 \sim |T-T_c|, and the correlation length, \xi \sim |T-T_c|^{-1/2}. Combining these results to show that the ratio of fluctuations to the mean value scales as |T-T_c|^{(d-4)/2}. For this ratio to be small as T \to T_c, we must have d>4.

For a complementary perspective, the following text derives the same result by analyzing the contribution of fluctuations to the specific heat.

1 Unit 4-8: Fluctuations and the Ginzburg Criterion We ...

This document from the University of Rochester provides a written derivation of the Ginzburg criterion by calculating the fluctuation correction to the specific heat.

Please read the section titled The Ginzburg Criterion. The logic here is to calculate the specific heat correction, \delta c, arising from fluctuations and compare it to the finite jump, c_{MF}, predicted by mean-field theory. The calculation shows that the most singular part of the fluctuation correction scales as \delta c \sim |T-T_c|^{(d-4)/2}. This correction diverges for d<4, signaling the breakdown of MFT, while it vanishes for d>4, confirming that MFT is valid there.

Both of these analyses arrive at the same powerful conclusion: for $d > 4$ , fluctuations become negligible near the critical point, and mean-field theory provides a correct description of the critical exponents. For $d < 4$ , fluctuations dominate, and mean-field theory breaks down. The dimension $d_{c} = 4$ is the upper critical dimension for this $ϕ^{4}$ -type theory.

The Renormalization Group Viewpoint

The Ginzburg criterion provides a compelling physical argument, but the RG framework offers a deeper explanation rooted in the very structure of the theory space. From the RG perspective, the upper critical dimension is the dimension at which the crucial interaction term transitions from being relevant to irrelevant.

Recall the RG flow for our $ϕ^{4}$ theory. The theory is described by two key parameters: a mass-like term $r$ (related to temperature) and an interaction coupling $u$ . The non-interacting theory ( $u = 0$ ) corresponds to the Gaussian fixed point. The mean-field exponents are precisely the exponents of this Gaussian fixed point.

The critical behavior of the full, interacting theory is determined by the stability of this fixed point. We've seen that the engineering dimension of the coupling $u$ is $[u] = 4 - d$ . This immediately tells us its relevance:

For $d > 4$ : $[u] < 0$ . The coupling $u$ is irrelevant. Under RG flow, it shrinks to zero. Any interacting theory will flow towards the non-interacting Gaussian fixed point. Therefore, the long-distance physics is described by mean-field theory, which becomes exact.
For $d < 4$ : $[u] > 0$ . The coupling $u$ is relevant. It grows under RG flow, driving the system away from the unstable Gaussian fixed point towards a new, non-trivial, interacting fixed point—the Wilson-Fisher fixed point. The critical behavior is governed by this new fixed point, leading to non-mean-field exponents.
For $d = 4$ : $[u] = 0$ . The coupling $u$ is marginal. This is the borderline case where the interaction's importance doesn't change under simple scaling. Perturbation theory at this dimension reveals logarithmic corrections to the mean-field behavior.

This logic is perfectly visualized in the following RG flow diagram.

RG flow in the parameter space of the coupling $g$ (analogous to our $u$) and a parameter $\mu$ related to temperature ($r$). For $d>4$, flows approach the Gaussian fixed point at $g=0$. For $d<4$, the Gaussian fixed point is unstable in the $g$ direction, and flows are driven towards the stable Wilson-Fisher fixed point.

This picture reveals that the breakdown of mean-field theory for $d < 4$ is not just a failure of an approximation, but a fundamental feature of the theory: the system is irresistibly drawn to a different, more complex universality class.

To see this in a calculation, we can examine what happens when we try to apply perturbation theory. The next video does exactly this, showing how the attempt to calculate corrections to the Gaussian theory fails for $d < 4$ .

9. Perturbative Renormalization Group Part 1

In this MIT OpenCourseWare lecture, Professor Mehran Kardar calculates the first-order perturbative correction to the susceptibility and analyzes its dimensional dependence.

Please watch the final part of the lecture, from analyzing the susceptibility correction. The key insight is in the analysis of the integral that appears in the correction term. At large k behavior, he shows that the integral's behavior depends on d-4. For d>4, the integral is finite and just provides a constant shift. The leading singular behavior as T \to T_c remains linear, so \gamma=1 (the MFT value). For d<4, the integral itself depends on T-T_c. As he shows around the divergent correction, the correction term diverges more strongly than the original term. This is a "divergent perturbation theory" and signals the breakdown of the entire expansion around the Gaussian theory. Finally, from dimensional analysis, he presents a beautiful heuristic argument, showing that the dimensionless expansion parameter is effectively u \xi^{4-d}, which clearly blows up as \xi \to \infty for d<4.

The failure of perturbation theory for $d < 4$ is the computational signal of what the RG flow diagram tells us visually: the Gaussian fixed point is unstable, and we cannot describe the critical physics by making small corrections around it. The theory is fundamentally different.

Conclusion

In this lesson, we have unified our understanding of mean-field theory and the renormalization group by introducing the concept of the upper critical dimension. This provides a crucial piece of the methodology of theoretical physics: knowing when your simplest approximations are justified and when you must turn to more sophisticated tools.

Key Takeaways:

The upper critical dimension, $d_{c}$ , is the spatial dimension above which the critical exponents of a system take their simple mean-field values.
The Ginzburg criterion provides a physical argument for $d_{c}$ . It states that mean-field theory is valid when the thermal fluctuations of the order parameter within a correlation volume are small compared to the mean order parameter itself. For $ϕ^{4}$ theory, this leads to $d_{c} = 4$ .
The renormalization group provides a more fundamental explanation. At the Gaussian (mean-field) fixed point, the $ϕ^{4}$ interaction coupling is irrelevant for $d > 4$ , relevant for $d < 4$ , and marginal at $d = 4$ .
For $d > d_{c}$ , the RG flow drives the system to the Gaussian fixed point, making mean-field theory exact for describing critical behavior.
For $d < d_{c}$ , the flow is towards a non-trivial interacting fixed point (like the Wilson-Fisher fixed point), resulting in non-mean-field exponents and the breakdown of simple perturbation theory.

This concludes our module on the Renormalization Group. You have seen how it explains scale invariance, universality, and the validity of mean-field theory. In our next module, we will explore the Effective Field Theory framework, which formalizes the RG idea of separating scales into a powerful, predictive tool for constructing theories without needing to know all the details of the underlying microscopic physics.

Recommended resources

The Limits of Mean-Field Theory

The Ginzburg Criterion: When Do Fluctuations Matter?

The Renormalization Group Viewpoint

Conclusion