Landau-Ginzburg theory

Landau-Ginzburg theory supplements the local Landau potential with gradient terms and treats the order parameter as a spatial field.

For a real statistical source convention $Z[h]=\int \mathcal{D}\varphi e^{-H_0[\varphi ]+\int h\varphi }$, define $F=-\log Z$ and the displayed Legendre transform. On a differentiable stable branch, $\delta \Gamma /\delta m=h$. In the Landau approximation, the branch expression is ${\Gamma }_{\mathrm{L}}[m]=H_0[m]$. The exact Legendre effective free energy is convex; a bare nonconvex Landau free energy is a local approximation and requires branch or coexistence interpretation.

The Landau approximation evaluates a statistical field's partition function at a stable stationary configuration, omitting fluctuation-loop corrections to the free energy. Functional differentiation of that stationary value still gives a connected linear response: the inverse of the saddle's quadratic fluctuation operator. Freezing fields before differentiating would lose this response. A Gaussian field theory, in contrast, permits exact integration of its quadratic fluctuations.

With positive $g$, the quadratic coefficients select the component with the smaller negative coefficient. Equal negative coefficients give a circle of minima in the mean-field approximation; the negative diagonal is a spin-flop transition, ending at a bicritical point at the origin.

A Landau free energy can be expanded in local powers of the order parameter and its spatial derivatives. Internal and spatial symmetries constrain the allowed terms; truncation assumes slowly varying fields.

For a statistical scalar theory with gradient coefficient $K/2>0$, this rescaling gives a canonical $(\nabla \psi {)}^2/2$ term. A mass coefficient $r$ becomes $r/K$, a quartic coefficient $g$ becomes $g/K^2$, and a source $h$ becomes $h/\sqrt{K}$. Thus the simple propagator denominator $p^2+R$ and loop-coupling formulas assume these normalized coefficients. The associated measure Jacobian is field independent and belongs to the free-energy normalization.

A gradient energy penalizes spatial variation of an order parameter. Its leading isotropic form is usually proportional to the squared spatial gradient.

A gradient expansion orders a local effective free energy by the number of spatial derivatives of slowly varying fields. Symmetry determines the allowed derivative contractions.

Interfacial tension is the excess free energy per unit area of an equilibrium interface relative to its coexisting bulk phases.

For $F=\int [a{\varphi }^2/2+b{\varphi }^4/4+\kappa |\nabla \varphi {|}^2/2]$ with $a<0<b,\kappa$, the planar coexistence interface is

The excess free energy per area of the phi-four diffuse interface is

A Gaussian field theory has a free energy quadratic in its field modes. If $H_0={\sum }_{\mathbf{q}}^{+}J(q)|{\varphi }_{\mathbf{q}}{|}^2$ for a real field, then $⟨|{\varphi }_{\mathbf{q}}{|}^2⟩=J(q{)}^{-1}$ and its partition function is a product of elementary Gaussian integrals.

The uniform mode of a pure Gaussian field theory with negative squared mass has an unbounded quadratic energy. Its partition function diverges even with a finite-volume ultraviolet regulator, so it supplies no stable ordered branch. At positive squared mass the zero-source average is zero. Thus the scaling dimension of the field can yield a formal order-parameter critical exponent without establishing a negative-reduced-temperature magnetization law. A stabilizing quartic interaction is relevant below four dimensions and ordinarily changes the critical theory to the Wilson-Fisher fixed point; above four its coefficient is a dangerously irrelevant coupling for the ordered phase.

For the positive quadratic kernel $\tilde{\Delta }(p)={\kappa }^{-1}{p}^2+m^2$, $\kappa ,m>0$, the connected correlation function is its inverse. Fourier inversion gives $G(r)=\kappa (\sqrt{\kappa }m/r{)}^{D/2-1}{K}_{D/2-1}(\sqrt{\kappa }mr)/(2\pi {)}^{D/2}$. The Modified Bessel function of the second kind has an exponentially decaying large-argument tail, so the correlation length is $1/(\sqrt{\kappa }m)$ at fixed positive gradient coefficient. In three dimensions this becomes $\kappa e^{-\sqrt{\kappa }mr}/(4\pi r)$. The statement concerns the continuum long-distance theory; regulator-dependent contact terms do not define the physical correlation length.

For a real field, ${\varphi }_{-\mathbf{q}}={\varphi }_{\mathbf{q}}^{\ast }$. A sum ${\sum }_{\mathbf{q}}^{+}$ takes one representative from each nonzero pair $\{\mathbf{q},-\mathbf{q}\}$ so that every independent complex Fourier amplitude is counted once.

A Gaussian variational approximation chooses a quadratic trial kernel and minimizes a variational upper bound on the interacting free energy. The optimized kernel often equals the bare kernel plus a self-consistent, wavevector-independent mass shift.

For $H=\int [a{\varphi }^2/2+\kappa |\nabla \varphi {|}^2/2+\gamma ({\nabla }^2\varphi {)}^2/2+B{\varphi }^2|\nabla \varphi {|}^2]$ and an even positive trial kernel $J(q)$, define $S_0={\sum }_q^{+}1/J(q)$ and $S_2={\sum }_q^{+}{q}^2/J(q)$ using a positive-wavevector sum for a real field. Isserlis theorem gives $⟨H_4{⟩}_0=4B{S}_0{S}_2/V$. The Feynman-Bogoliubov inequality therefore has stationary kernelsThus both the mass and gradient coefficient shift. In the thermodynamic limit, their self-consistency equations areA coarse-graining ultraviolet cutoff $\Lambda$ is necessary for the first integral in three dimensions: its large-$k$ radial integrand tends to a constant. Physical solutions require $J(q)>0$.

For an exact dimensionless Hamiltonian $H$ and trial Hamiltonian $H_0$, the free energies obeywhere $F_0=-\log Z_0$ and the expectation uses the trial Gibbs distribution $Z_0^{-1}{e}^{-H_0}$.

The identity $e^{-F}=e^{-F_0}⟨e^{-(H-H_0)}{⟩}_0$ and the Jensen inequality for the convex exponential function give $\log ⟨e^{-(H-H_0)}{⟩}_0\ge -⟨H-H_0{⟩}_0$. Taking minus the logarithm proves the bound.

A Brazovskii model has a quadratic Fourier kernel minimized on a sphere of nonzero wavevector magnitude. The large phase space of soft modes can prevent the Gaussian mass from vanishing and turn a mean-field continuous transition into a fluctuation-induced first-order transition.

A gradient-quartic Brazovskii model stabilizes a negative quadratic gradient coefficient with $\gamma ({\nabla }^2\varphi {)}^2/2$ and uses the quartic interaction $B{\varphi }^2|\nabla \varphi {|}^2$. For $\gamma ,B>0$ and a zero-average compositional order parameter, its nonzero-wavevector instability is a candidate for smectic phase ordering. The Gaussian variational kernel for a gradient quartic interaction shifts both the quadratic coefficients.

For $\varphi =A\cos (q_{0}z)$, the gradient-quartic Brazovskii model has volume-averaged free-energy densityThe averages are $⟨{\cos }^2⟩=⟨{\sin }^2⟩=1/2$ and $⟨{\cos }^2{\sin }^2⟩=1/8$. For $B>0$, the stable amplitude is zero for $a\ge a_c$ and satisfies $A^2=(a_c-a)/(B{q}_0^2)$ below it. The mean-field approximation therefore predicts a continuous phase transition.

For $G(q)=a+\kappa q^2+\gamma q^4$ with $\kappa <0<\gamma$, the minimum occurs on the sphere $q=q_0=\sqrt{-\kappa /(2\gamma )}$. The minimum value is $a-{\kappa }^2/(4\gamma )$.

In three dimensions, let $J(q)=r+\gamma (q^2-q_{\ast }^2{)}^2$, with $q_{\ast }>0$ and small $r>0$. Near the soft shell, $J(q)\simeq r+4\gamma q_{\ast }^2(q-q_{\ast }{)}^2$, soThe radial integral uses ${\int }_{-\infty }^{\infty }ds/(r+A{s}^2)=\pi /\sqrt{Ar}$. The $r^{-1/2}$ divergence obstructs the zero-mass limit of a self-consistent Gaussian variational approximation at a fixed nonzero ordering wavevector. This is the mechanism behind the Brazovskii fluctuation-induced first-order transition; Gross, Ignatiev and Chakraborty's study of fluctuation-driven ordering analyzes the corresponding standard Brazovskii model.

In a self-consistent Gaussian treatment, fluctuations near the soft-mode sphere generate a positive mass shift that diverges as the renormalized mass approaches zero. The disordered phase therefore remains locally stable until its free energy crosses that of a finite-amplitude modulated phase, producing a first-order transition.

A smectic phase has one-dimensional translational order, represented at the simplest level by a periodically modulated field $\varphi (\mathbf{r})=A\cos ({\mathbf{q}}_0\cdot \mathbf{r})$.

A polar helical smectic has a vector order parameter of constant magnitude whose direction rotates at a nonzero wavevector. For an isotropic Brazovskii free energy, all constant rotations of the helical plane are degenerate until a longitudinal-gradient term such as $|\nabla \cdot \mathbf{p}{|}^2$ selects its orientation.

The Ginzburg criterion requires order-parameter fluctuations inside a correlation volume to be small compared with the squared mean-field order parameter. Its failure marks the fluctuation-dominated critical region.

For the dimensionless scalar Landau-Ginzburg theory with stiffness $c$, quadratic coefficient $r<0$ and quartic term $u{\varphi }^4/4$, the mean-field approximation gives ${\varphi }_0^2=|r|/u$ and correlation length $\xi \asymp (c/|r|{)}^{1/2}$. Restricting the Gaussian Fourier transform covariance $(c{q}^2+2|r|{)}^{-1}$ to $|q|\lesssim {\xi }^{-1}$ estimates the variance of the field averaged over a correlation volume as $c^{-d/2}|r{|}^{(d-2)/2}$, up to a dimension-dependent constant. Dividing by ${\varphi }_0^2$ proves the displayed Ginzburg criterion ratio. It diverges below four dimensions, vanishes above four and is marginal at four. Short-distance fluctuations renormalize microscopic coefficients and should not be confused with this long-distance critical test.

The critical region is the range of control parameters near a thermodynamic critical point where correlated order parameter fluctuations invalidate a simple mean-field approximation. For short-range scalar quartic theory below four dimensions, the dimensionless Ginzburg criterion ratio grows as $g{\xi }^{4-D}$ in fixed microscopic units. The range where this ratio is of order one or larger is fluctuation dominated. Its width depends on nonuniversal microscopic coefficients even though the limiting critical exponents can be universal.

For a scalar quartic interaction $g{\varphi }^4/4!$, the tadpole relation $R=r_{\Lambda }+(g/2)I_D(R)$ can be subtracted at the critical point. When $I_D(0)$ exists, $A(T-T_C)=R[1+(g/2)J_D(R)]$, with $J_D(R)=\int d^{D}p(2\pi {)}^{-D}/[p^2(p^2+R)]$. This integral is finite as $R\to 0$ only above four dimensions and is logarithmic at four. A self-consistent one-loop relation is not an exact interacting critical-exponent formula.

For the scalar quartic one-loop critical-mass subtraction, the subtracted tadpole is $I_D(0)-I_D(R)=R{J}_D(R)$. Above four dimensions $J_D(0)=K_D{\Lambda }^{D-4}/(D-4)$ is finite. At four dimensions $J_4=(K_4/2)\log [({\Lambda }^2+R)/R]$. For $2<D<4$, $J_D\sim K_D{R}^{(D-4)/2}{\int }_0^{\infty }{q}^{D-3}(1+q^2{)}^{-1}dq$, which diverges as $R\to 0$. Hence a smooth linear thermal mass survives this test only above the ordinary upper critical dimension four. The massless subtraction is infrared divergent for $D\le 2$ and cannot prove absence of an interacting transition there.

For $R>0$, a spherical ultraviolet cutoff gives the following exact examples of the one-loop critical-mass subtraction integral:The areas of the unit spheres in dimensions three, four and five are $4\pi$, $2{\pi }^2$ and $8{\pi }^2/3$. Dividing by $(2\pi {)}^D$ reduces the integrals to constants times ${\int }_0^{\Lambda }dp/(p^2+R)$, ${\int }_0^{\Lambda }pdp/(p^2+R)$ and ${\int }_0^{\Lambda }{p}^{2}dp/(p^2+R)$ respectively. Elementary integration proves the formulas. As $R\to 0$, the first has a power divergence, the second a logarithmic divergence, and the third a finite limit. These examples make the ordinary upper critical dimension four visible without treating a self-consistent one-loop equation as an exact interacting critical exponent prediction.

At an upper critical dimension, the leading Ginzburg criterion ratio has zero thermal power rather than a negative one. It does not vanish asymptotically, but its magnitude can depend on the coupling. The criterion alone does not establish a power-law divergence or new critical power exponents. Running marginal interactions generally produce logarithmic corrections; the ordinary quartic and tricritical sextic scalar theories are examples.

For a tuned multicritical even Landau potential with leading positive $M^{2n}$ term, $M^2\sim |t{|}^{1/(n-1)}$ and $\xi \sim |t{|}^{-1/2}$. Long-wavelength Gaussian fluctuations in a correlation volume have variance proportional to ${\xi }^{2-D}$. Their ratio to $M^2$ therefore scales as $|t{|}^{(D-2)/2-1/(n-1)}$, vanishing only for $D>2n/(n-1)$. At equality the criterion is marginal, requiring further fluctuation analysis.

For a quadratic kernel $K(t,k)\sim t+k^2$, the singular heat-capacity correction is proportional to

A Gaussian field with free-energy kernel $A+\kappa q^2$ has structure factor proportional to $(A+\kappa q^2{)}^{-1}$ and correlation length $\sqrt{\kappa /A}$.