Upper Critical Dimension of φ⁴ Theory

Recommended resources

The freely available resources this lesson will direct you to.

Hello! Welcome to your next lesson on the statistical mechanics of phase transitions.

In our last session, we derived the Ginzburg criterion, which provided a powerful physical argument for the breakdown of mean-field theory. We found that for spatial dimensions $d < 4$ , fluctuations in the order parameter overwhelm the mean-field value as we approach the critical point, rendering the mean-field approximation invalid. This led us to identify $d_{c} = 4$ as a special, marginal dimension.

Today, we will solidify this crucial result. Our learning objective is to calculate the upper critical dimension for $ϕ^{4}$ theory, above which mean-field exponents are exact. We will move from the intuitive "patch" argument of the last lesson to a more formal calculation, which should resonate with your background in quantum field theory. We'll analyze how fluctuations provide a correction to a macroscopic thermodynamic quantity—the specific heat—and by examining the nature of this correction, we'll pinpoint the dimension where its character changes.

The Ginzburg Criterion Revisited

Let's begin by briefly revisiting the core argument from our previous lesson. The Ginzburg criterion states that mean-field theory is valid only if the mean-squared fluctuation of the order parameter is much smaller than the square of the mean-field order parameter itself:

⟨(δ ϕ)^{2} ⟩ ≪ ϕ_{0}^{2}

By estimating the size of both terms as a function of the reduced temperature $t = (T - T_{c}) / T_{c}$ , we arrived at the condition:

∣ t ∣^{(d - 4) /2} ≪ 1

This inequality shows that as we approach the critical point ( $t \to 0$ ), fluctuations can only be neglected if the exponent is positive, which requires $d > 4$ .

To reinforce this fundamental idea, let's look at it from two excellent resources.

Landau Ginzburg theory of Phase Transitions

First, let's re-watch the key segment from Jos Thijssen's lecture. This time, focus specifically on the final steps where the dependencies on temperature and dimensionality are combined to produce the final criterion. This will serve as a quick but powerful recap of the physical argument.

Watch the video from 42:50 to 47:17. Pay close attention to how the expressions for the fluctuation (\langle\delta m_r^2 angle \propto t^{(d-2)/2}) and the mean-field value m_0^2 \propto t are brought together to form the ratio t^{(d-4)/2}, leading to the conclusion about the upper critical dimension.

[PDF] Chapter 7 Landau theory

Next, for a concise and well-written textual version of the same argument, please read the following section from a set of notes on renormalization.

Read Section 7.7, "How mean-field theory goes wrong: the Ginzburg criterion". This section provides a clear, step-by-step derivation that mirrors the video's logic. Note how it translates the physical requirement into the language of the Landau-Ginzburg field theory, relating the fluctuations to the correlator G(r) and arriving at the same condition involving \xi^{4-d}.

Both approaches confirm that mean-field theory is only self-consistent for $d > 4$ . Now, let's prove this from a different angle.

Calculating Fluctuation Corrections to the Specific Heat

A more formal way to test the validity of a theory is to calculate the first-order correction and see if it's small compared to the leading-order result. In our case, the leading-order result is the mean-field prediction, and the correction comes from including Gaussian fluctuations around the saddle-point.

We'll calculate the correction to the specific heat, $C_{V}$ . In mean-field theory, the specific heat exhibits a simple jump discontinuity at $T_{c}$ , which corresponds to a critical exponent $α_{MFT} = 0$ . If the fluctuation correction gives rise to a stronger singularity (i.e., a divergence with $α > 0$ ), then mean-field theory has clearly failed to capture the true critical behavior.

The following resource from Mehran Kardar's renowned textbook provides the complete field-theoretic calculation.

[PDF] 8.334 Statistical Mechanics II: Statistical Physics of Fields

Please study the following sections from Kardar's lecture notes. This is a dense but rewarding read that formalizes our investigation. It shows how to compute the free energy including Gaussian fluctuations and from it, the correction to the specific heat.

Read the text from equation (II.70) to just before the heading "The Ginzburg Criterion". This covers two key parts: Fluctuation Free Energy (Eq. II.70 - II.72): Understand how the partition function is approximated by integrating over Gaussian fluctuations \phi(x) around the mean-field solution ar{m}. The resulting free energy has a new term originating from this functional integral (analogous to a one-loop correction in QFT). Specific Heat Correction (Eq. II.73 - II.75): Follow the derivation of the singular part of the heat capacity, C_{singular}, by taking the second derivative of the fluctuation free energy with respect to temperature t. Focus on the resulting integral, C_F, and how its dependence on dimensionality is analyzed.

Let's break down the central calculation from the notes.

The fluctuation correction to the specific heat (just above $T_{c}$ ) is proportional to the integral:

C_{f l u c} \propto \frac{\partial ^{2}}{\partial t ^{2}} \int \frac{d ^{d} q}{( 2 π ) ^{d}} ln (K (q^{2} + ξ^{- 2}))

where $t \propto (T - T_{c})$ and $ξ^{- 2} \propto t$ . After taking the derivatives, we arrive at an integral of the form:

C_{f l u c} \propto \int \frac{d ^{d} q}{( 2 π ) ^{d}} \frac{1}{( q ^{2} + ξ ^{- 2} ) ^{2}}

The fate of mean-field theory hinges on the behavior of this integral as $t \to 0$ (or equivalently, $ξ \to \infty$ ).

Analyzing the Integral

Let's analyze this integral by changing to $d$ -dimensional spherical coordinates. The volume element is $d^{d} q = S_{d - 1} q^{d - 1} d q$ , where $S_{d - 1} = 2 π^{d /2} /Γ (d /2)$ is the surface area of a $(d - 1)$ -sphere.

C_{f l u c} \propto \int_{0}^{Λ} d q \frac{q ^{d - 1}}{( q ^{2} + ξ ^{- 2} ) ^{2}}

where $Λ \sim 1/ a$ is a momentum cutoff related to the lattice spacing.

We are interested in the behavior as $ξ \to \infty$ .

Case 1: $d > 4$
The integrand at large $q$ behaves as $q^{d - 1} / q^{4} = q^{d - 5}$ . Since $d - 5 > - 1$ , the integral diverges at the upper limit $q \to Λ$ . This is a UV divergence. Crucially, this divergence is independent of $ξ$ (and thus of temperature). The fluctuations merely add a constant background to the specific heat. The leading singular behavior remains the jump discontinuity from mean-field theory ( $α = 0$ ). Thus, MFT is valid.
Case 2: $d < 4$
The integrand at large $q$ now behaves as $q^{d - 5}$ with $d - 5 < - 1$ , so the integral converges at the upper limit. The dominant contribution comes from small $q$ (an IR effect). For $q ≪ ξ^{- 1}$ , we can approximate the integral:
$C_{f l u c} \propto \int_{0}^{ξ^{- 1}} d q \frac{q ^{d - 1}}{( ξ ^{- 2} ) ^{2}} \propto ξ^{4} (ξ^{- 1})^{d} = ξ^{4 - d}$
Since $ξ \propto ∣ t ∣^{- ν_{MFT}} = ∣ t ∣^{- 1/2}$ , we find:
$C_{f l u c} \propto (∣ t ∣^{- 1/2})^{4 - d} = ∣ t ∣^{- (4 - d) /2}$
The specific heat itself is related to this fluctuation contribution, and its divergence is governed by the exponent $α = (4 - d) /2$ . Since $α > 0$ for $d < 4$ , the specific heat diverges at $T_{c}$ . This is a much stronger singularity than the mean-field jump ( $α = 0$ ). The fluctuation correction dominates, and mean-field theory fails.
Case 3: $d = 4$
The integrand behaves as $q^{- 1}$ for large $q$ . The integral diverges logarithmically, $C_{f l u c} \propto ln (Λ ξ) \propto - ln ∣ t ∣$ . This is the marginal case, which marks the boundary of validity for mean-field theory.

This analysis confirms our previous result: the upper critical dimension for $ϕ^{4}$ theory is $d_{c} = 4$ .

Test your understanding!

The Ginzburg criterion compared the fluctuation of the order parameter to its mean value. This lesson's calculation analyzed the fluctuation correction to the specific heat. Why do both approaches yield the same upper critical dimension?

Show answer

Both methods are fundamentally probing the same question: when do fluctuation effects become comparable to or larger than the mean-field predictions?

The Ginzburg criterion, $⟨(δ ϕ)^{2} ⟩ / ϕ_{0}^{2} ≪ 1$ , directly compares the scale of fluctuations to the scale of the order parameter. This ratio diverges for $d < 4$ .

The specific heat calculation compares the singular part of the mean-field specific heat ( $C_{MFT} \sim t^{0}$ , a jump) with the singular part of the fluctuation correction ( $C_{f l u c} \sim t^{- (4 - d) /2}$ ). The ratio $C_{f l u c} / C_{MFT}$ also diverges for $d < 4$ .

Both are different ways of quantifying the relative importance of the $ϕ^{4}$ interaction term (which is neglected in the Gaussian fluctuation calculation but responsible for the non-zero $ϕ_{0}$ ) versus the quadratic (Gaussian) fluctuation term. They both lead to the same condition on dimensionality because they are scaling the same underlying physical competition.

What Happens at $d_{c} = 4$ ?

At the upper critical dimension, a remarkable simplification occurs. The critical behavior becomes universal and is correctly described by mean-field theory, albeit with possible logarithmic corrections. This means that microscopic details become irrelevant, and different physical systems belonging to the same universality class will exhibit identical critical exponents and scaling functions.

Scaling Behavior at the Upper Critical Dimension — CAPTION: This figure illustrates the concept of universality at the upper critical dimension. Data from different models (CLG, CTTP, Manna) are plotted with rescaled axes. The fact that all data points collapse onto a single, universal curve demonstrates that the underlying critical behavior is identical and is described by mean-field theory, as predicted by the Ginzburg criterion and our calculation.

Conclusion

Key Takeaways:

We have rigorously calculated the upper critical dimension for $ϕ^{4}$ theory to be $d_{c} = 4$ .
We did this by analyzing the fluctuation contribution to the specific heat, finding that for $d < 4$ , this correction produces a stronger singularity than the mean-field prediction, thereby invalidating it.
For $d > 4$ , fluctuation corrections are not singular enough to alter the leading critical behavior described by mean-field theory. The mean-field exponents are therefore exact.
This result is consistent with the Ginzburg criterion, which compares the magnitude of order parameter fluctuations to the mean-field value. Both are different perspectives on the same physical competition.

Preview of the Next Lesson:

We have now firmly established when mean-field theory works (for $d > 4$ ) and when it fails (for $d \leq 4$ ). But this leaves a crucial question unanswered: what are the correct critical exponents in the physically relevant dimensions of $d = 2$ and $d = 3$ ? And why are they universal?

Our next topic, Scaling and Universality, addresses this head-on. We will introduce the scaling hypothesis, a powerful phenomenological framework that postulates relationships between different critical exponents and organizes different physical systems into universality classes. This will set the stage for our eventual study of the Renormalization Group, the theoretical machinery that finally explains why scaling and universality emerge.