08.08.04 · stat-mech / lattice-gauge

The roughening transition and the confining string

shipped3 tiersLean: none

Anchor (Master): Lüscher, Nucl. Phys. B 180 (1980) 317 (the universal string correction); Lüscher-Weisz, JHEP 0407 (2004) 014; Chui-Weeks, Phys. Rev. B 14 (1976) 4978 (SOS roughening via Coulomb gas); Itzykson-Drouffe, Statistical Field Theory Vol. 2, Ch. 6

Intuition Beginner

Heat the surface of a crystal and watch what its boundary does. At low temperature the surface is smooth: it sits flat on a crystal plane, with only a few isolated bumps where an atom has hopped up or a gap has opened. Raise the temperature and at some point the surface suddenly becomes rough: steps and terraces wander freely, the height of the surface fluctuates more and more the wider a patch you look at, and the crystal plane stops pinning it down. The temperature where this switch happens is the roughening temperature.

The same story shows up in a place that looks completely different: the force between a quark and an antiquark. In a confining theory the energy stored between the pair grows in proportion to their separation, as if a taut tube of field — a flux tube, or confining string — joins them. That tube is not a rigid rod. It is a flexible sheet that wobbles, and the way it wobbles is exactly the way a crystal surface wobbles.

So a single idea — a fluctuating interface that can be smooth or rough — describes both a melting crystal face and the string holding quarks together. The roughening transition is the temperature where the interface goes from gently rippled to wildly wandering.

Visual Beginner

On the left, a crystal surface at low temperature: a nearly flat staircase of atoms, hugging one horizontal plane. On the right, the same surface above the roughening point: the staircase has dissolved into broad terraces and steps that meander, and the typical height spread keeps growing as you widen your view.

The picture makes the key contrast visible. Smooth means the height spread saturates to a small value no matter how wide the window. Rough means the height spread grows — slowly, like the logarithm of the width — without ever settling down. The flux tube of a confining gauge theory is the same surface seen edge-on.

Worked example Beginner

Model a one-dimensional surface as a row of columns, each with an integer height. Neighbouring columns pay an energy cost for every unit of height difference between them: a step of size one costs energy $J$ , a step of size two costs $2 J$ , and so on. This is a baby version of the solid-on-solid model.

Start flat, with every column at height zero. The cheapest excitation raises one neighbouring boundary by one unit, costing energy $J$ . At low temperature these steps are rare and expensive, so the surface stays nearly flat: it is smooth.

Now ask what happens as temperature rises. Each possible step location can be up, down, or level, so steps become common once the temperature is comparable to $J$ . The surface starts to drift: heights of distant columns lose track of one another. The spread in heights, instead of staying bounded, begins to grow with the size of the patch you measure.

What this tells us: a competition between the step energy $J$ (which wants the surface flat) and the entropy of many possible step patterns (which wants it to wander) decides whether the interface is smooth or rough. The roughening temperature is where entropy wins.

Check your understanding Beginner

Formal definition Intermediate+

Fix a two-dimensional base lattice $Λ \subset Z^{2}$ . The solid-on-solid (SOS) model assigns to each site $x$ an integer height $h_{x} \in Z$ and weights configurations by the Boltzmann factor $e^{- β H}$ with the nearest-neighbour interface energy $$ H_{\mathrm{SOS}} = J \sum_{\langle x,y\rangle} \lvert h_x - h_y \rvert , $$ the sum running over nearest-neighbour bonds. Two standard relatives replace the absolute value by a quadratic cost: the discrete-Gaussian model, $H = \frac{J}{2} \sum_{⟨ x, y ⟩} (h_{x} - h_{y})^{2}$ with $h_{x} \in Z$ , and the Villain model, a periodic-Gaussian form chosen so that its duality is exact. All three share a single qualitative phase diagram and the same long-distance physics.

The order parameter for roughness is the height-difference variance $$ G(R) = \big\langle (h_x - h_y)^2 \big\rangle , \qquad R = \lvert x - y \rvert . $$ The interface is smooth if $G (R)$ saturates to a finite constant as $R \to \infty$ , and rough if it diverges. The roughening temperature $T_{R}$ (inverse $β_{R}$ ) separates the two: for $T < T_{R}$ the surface is smooth, for $T > T_{R}$ it is rough with the logarithmic law $$ G(R) ;\sim; \frac{1}{\pi K(T)},\ln R \qquad (T>T_R), $$ where $K (T)$ is an effective stiffness. The prerequisites used here are 08.08.01 and 08.15.01.

In lattice gauge theory 08.08.01, the relevant interface is the confining string: the surface spanning a large Wilson loop, on which the flux that produces the area law lives. Its transverse displacement field plays the role of $h_{x}$ , and its roughening controls the breakdown of the strong-coupling expansion.

Key theorem with proof Intermediate+

Theorem (SOS roughening is dual to a Coulomb-gas / Kosterlitz-Thouless transition). The discrete-Gaussian SOS model on a two-dimensional lattice is dual, via Poisson summation on the integer heights, to a two-dimensional lattice Coulomb gas of integer charges interacting logarithmically. Its roughening transition is therefore in the Kosterlitz-Thouless universality class of the 2D XY model 08.15.01: smooth $\leftrightarrow$ rough corresponds to charges bound $\leftrightarrow$ unbound, and the effective stiffness $K$ jumps to the universal value $K (T_{R}^{-}) = 2/ π$ at roughening.

Proof. Write the discrete-Gaussian partition function as a sum over integer height fields and insert the Poisson summation identity $$ \sum_{h\in\mathbb{Z}} f(h) ;=; \sum_{m\in\mathbb{Z}} \int_{-\infty}^{\infty}! d\phi; f(\phi), e^{2\pi i m\phi}, $$ applied at every site. This trades each integer height $h_{x}$ for a real field $ϕ_{x}$ together with a dual integer $m_{x}$ . Performing the now-Gaussian integral over $ϕ$ leaves a weight depending only on the duals $m_{x}$ , of the form $$ Z ;\propto; \sum_{{m_x}} \exp!\Big[ -,\pi,\tilde\beta \sum_{x,y} m_x, V(x-y), m_y \Big], $$ where $V (x - y)$ is the lattice Green's function of the Laplacian, growing as $- ln ∣ x - y ∣$ at large separation, and $\tilde{β} \propto 1/ β$ is the dual temperature. Charge neutrality $\sum_{x} m_{x} = 0$ is enforced by the zero mode. This is exactly the two-dimensional Coulomb gas of integer charges $m_{x}$ with logarithmic interactions, the model whose unbinding transition is the Kosterlitz-Thouless transition analysed in 08.15.01. The smooth phase is the dielectric phase of bound neutral dipoles, where $G (R)$ stays finite; the rough phase is the conducting plasma of free charges, where screening softens the surface and $G (R) \sim ln R$ . The Kosterlitz recursion relations for this gas have the same fixed-line structure as the XY model, and the stiffness reaches the universal value $K = 2/ π$ at the transition. $□$

Bridge. This duality builds toward the effective-string description developed in the Master tier and appears again in 12.18.16, where the confining flux tube is the physical interface being roughened. The central insight is that this is exactly the Kosterlitz-Thouless mechanism of 08.15.01 wearing a different costume: heights and charges are dual variables, so the roughening of a surface is dual to the unbinding of vortices, and the foundational reason both share a universal stiffness jump is that both reduce to the same logarithmic 2D Coulomb gas. Putting these together, the smooth/rough split of the lattice string and the bound/free split of XY vortices are one transition seen through two windows.

Exercises Intermediate+

Exercise 6 (hard, short-answer).

Explain why the height variance in the rough phase grows only logarithmically rather than linearly with $R$ , and why this is special to two interface dimensions.

Hint

The relevant Green's function is that of the 2D Laplacian.

Answer

In the rough phase the interface is an effectively free 2D Gaussian field, whose propagator is the inverse Laplacian. In two dimensions that Green's function grows as $ln R$ , so the height variance inherits the same logarithm. In one interface dimension the propagator grows linearly (random-walk wandering), and in three or more it saturates, so the marginal logarithmic case — and the resulting Kosterlitz-Thouless-type transition — is special to a two-dimensional base.

Rubric: full credit for tying the logarithm to the 2D Laplacian Green's function and contrasting other dimensions.

Exercise 7 (hard, short-answer).

The strong-coupling expansion of a Wilson loop is a sum over surfaces tiling the loop. Argue heuristically why roughening signals the convergence boundary of that expansion.

Hint

Count how many surfaces contribute once the interface delocalises.

Answer

The strong-coupling series weights the minimal spanning surface plus deformations, each suppressed by powers of the inverse coupling. In the smooth phase the surface is pinned, deformations are rare, and the series converges geometrically. At roughening the interface delocalises: the number of contributing surfaces of a given area proliferates and their entropy competes with the per-plaquette suppression, so the series ceases to converge term-by-term. Roughening is therefore the radius of convergence of the strong-coupling expansion seen physically.

Rubric: full credit for the entropy-of-surfaces-versus-suppression argument identifying roughening with the convergence boundary.

Advanced results Master

Above roughening the long-distance physics of the confining string is captured by an effective bosonic-string action: the leading term is the area (Nambu-Goto) functional, and its small-fluctuation expansion is a free 2D Gaussian theory for the $d - 2$ transverse displacement fields living on the worldsheet. Quantising this free string on a cylinder of length $r$ (the interquark separation) and computing its ground-state Casimir energy yields the celebrated Lüscher term: the static potential between a quark and antiquark acquires a universal, parameter-free correction $$ V(r) ;=; \sigma, r ;-; \frac{\pi (d-2)}{24},\frac{1}{r} ;+; \cdots , $$ which in $d = 4$ reads $V (r) = σ r - \frac{π}{12 r} + \dots$ ^{[Lüscher 1980]}. The coefficient $- π /12$ is fixed entirely by the central charge of the free-boson worldsheet (the conformal anomaly), independent of the gauge group, the lattice details, or the string tension $σ$ ; it is the interface analogue of a Casimir energy ^{[Lüscher-Symanzik-Weisz 1980]}. The same string fluctuations make the flux tube broaden logarithmically in $r$ , $w^{2} (r) \sim ln (r / r_{0})$ , which is precisely the rough-phase height law of the SOS model carried over to the gauge string ^{[Lüscher-Münster-Weisz 1981]}.

Roughening matters because it draws the boundary of validity for the strong-coupling expansion of lattice gauge theory 08.08.01. Below the roughening coupling the spanning surface is smooth, the series converges, and the lattice string is rigid; above it the surface roughens, the series loses analyticity, and the long-distance potential develops its universal $- π / (12 r)$ tail ^{[Drouffe-Zuber 1983]}. The roughening point is not the deconfinement transition: the theory still confines on both sides. It is an analyticity boundary internal to the confined phase, the place where the lattice string stops feeling the underlying grid and starts behaving like a continuum string.

Synthesis. The roughening transition is the bridge that unifies three apparently distinct phenomena: the melting of a crystal facet, the Kosterlitz-Thouless unbinding of the 2D Coulomb gas 08.15.01, and the onset of universal string behaviour in confining gauge theory 08.08.01. This is exactly the Kosterlitz-Thouless mechanism: the SOS height model is dual to the logarithmic Coulomb gas, so the smooth-to-rough transition is dual to the dielectric-to-plasma transition, and the central insight is that one universal stiffness jump $K = 2/ π$ governs both. Putting these together, the foundational reason the Lüscher $- π / (12 r)$ term is universal is that, above roughening, the flux tube is a free 2D Gaussian interface whose conformal anomaly fixes the Casimir coefficient; this generalises the crystal-surface logarithm $w^{2} \sim ln r$ to the gauge string and is dual, through Poisson summation, to the very vortex physics that controls the XY model. The transition thus packages interface statistics, 2D criticality, and confining-string dynamics into a single critical point, appears again in 12.18.16 as the continuum limit of the lattice flux tube, and marks where strong-coupling analyticity ends without confinement itself ending.

Full proof set Master

Proposition (Casimir energy of the free interface and the Lüscher coefficient). A free Gaussian string with $d - 2$ transverse modes, quantised on a cylinder of circumference $r$ , has a ground-state energy whose finite part contributes a $- \frac{π ( d - 2 )}{24 r}$ correction to the linear confining potential; in $d = 4$ this is $- π / (12 r)$ .

Proof. Each transverse displacement field is a free massless 2D boson on the worldsheet, a strip (for an open string between static quarks) of width $r$ with appropriate boundary conditions. Expanding in transverse Fourier modes labelled by $n = 1, 2, \dots$ , the worldsheet field is a collection of harmonic oscillators with frequencies $ω_{n} = π n / r$ . The zero-point energy is the (regularised) sum $$ E_0 = \frac{d-2}{2}\sum_{n=1}^\infty \omega_n = \frac{(d-2)\pi}{2r}\sum_{n=1}^\infty n . $$ The divergent sum is assigned its finite part by zeta-function regularisation, $\sum_{n \geq 1} n \to ζ (- 1) = - \frac{1}{12}$ , the standard Casimir prescription. Hence $$ E_0 \to \frac{(d-2)\pi}{2r}\cdot\Big(-\frac{1}{12}\Big) = -\frac{\pi(d-2)}{24,r}, $$ and for $d = 4$ , $E_{0} = - π / (12 r)$ . Adding this to the classical string energy $σ r$ gives the static potential $V (r) = σ r - π / (12 r) + \dots$ . The coefficient depends only on the number of transverse modes and the free-boson central charge, not on any microscopic coupling, which is the statement that the Lüscher term is universal. $□$

Proposition (logarithmic broadening of the rough interface). A free 2D Gaussian interface of linear size $r$ has transverse mean-square width $w^{2} (r) = ⟨ h^{2} ⟩ - ⟨ h ⟩^{2}$ growing as $ln r$ .

Proof. The free interface Hamiltonian is $H = \frac{σ}{2} \int (\nabla h)^{2}$ , so in Fourier space each mode $h_{q}$ has variance $⟨ ∣ h_{q} ∣^{2} ⟩ = 1/ (β σ q^{2})$ by equipartition. The local width is $$ w^2 = \langle h(x)^2\rangle = \int \frac{d^2 q}{(2\pi)^2},\frac{1}{\beta\sigma, q^2}, $$ with the integral running from an infrared cutoff $\sim 1/ r$ to an ultraviolet cutoff $\sim 1/ a$ . The measure $d^{2} q / q^{2} = d q / q \times (angle)$ produces a logarithm: $$ w^2 = \frac{1}{2\pi\beta\sigma},\ln\frac{r}{a}. $$ Thus the width diverges logarithmically with the system size $r$ , the hallmark of the rough phase, and the same logarithm that defines roughening in the SOS height variance $G (R) \sim ln R$ . $□$

Connections Master

The roughening transition is dual, by Poisson summation on integer heights, to the unbinding transition of the 2D Coulomb gas, placing it in the same universality class as the Kosterlitz-Thouless transition of the 2D XY model 08.15.01; the universal stiffness jump $K = 2/ π$ is shared between the two.
It controls the long-distance behaviour of the confining string in Wilson's lattice gauge theory 08.08.01: below roughening the spanning surface is smooth and the strong-coupling expansion converges, while above it the flux tube becomes a fluctuating continuum string with logarithmic broadening.
The continuum confinement story — the area law, the effective string, and the static quark-antiquark potential — is developed in the quantum-field-theory pointer 12.18.16, where the Lüscher $- π / (12 r)$ correction appears as the universal signature that the flux tube has roughened into a genuine string.

Historical & philosophical context Master

The roughening transition entered statistical mechanics through the theory of crystal growth and equilibrium crystal shapes: Burton, Cabrera, and Frank had argued in 1951 that a crystal facet loses its sharpness above a characteristic temperature, and the modern understanding as a thermodynamic phase transition crystallised in the 1970s when Chui and Weeks identified the SOS model's roughening as dual to the two-dimensional Coulomb-gas unbinding, the same mechanism Kosterlitz and Thouless had found for the XY model ^{[Chui-Weeks 1976]}. The transplant of these ideas into gauge theory was Michael Lüscher's: in 1980 he showed that, above roughening, the confining string fluctuates as a free bosonic surface and that its Casimir energy produces a universal correction $- π / (12 r)$ to the interquark potential, fixed by the conformal anomaly alone and independent of all microscopic data ^{[Lüscher 1980]}. The philosophical payoff is a striking unity: a melting crystal face, the superfluid-film transition, and the string holding quarks together are governed by one critical point, and the appearance of the pure number $- π /12$ in the strong-interaction potential is a direct, measurable fingerprint of two-dimensional conformal field theory living on the worldsheet of the flux tube.

Bibliography Master

@article{Luscher1980Roughening,
  author  = {L\"uscher, Martin},
  title   = {Symmetry-breaking aspects of the roughening transition in gauge theories},
  journal = {Nuclear Physics B},
  volume  = {180},
  year    = {1980},
  pages   = {317--329}
}

@article{LuscherSymanzikWeisz1980,
  author  = {L\"uscher, Martin and Symanzik, Kurt and Weisz, Peter},
  title   = {Anomalies of the free loop wave equation in the WKB approximation},
  journal = {Nuclear Physics B},
  volume  = {173},
  year    = {1980},
  pages   = {365--396}
}

@article{LuscherMunsterWeisz1981,
  author  = {L\"uscher, Martin and M\"unster, Gerhard and Weisz, Peter},
  title   = {How thick are chromoelectric flux tubes?},
  journal = {Nuclear Physics B},
  volume  = {180},
  year    = {1981},
  pages   = {1--12}
}

@article{ChuiWeeks1976,
  author  = {Chui, S. T. and Weeks, John D.},
  title   = {Phase transition in the two-dimensional Coulomb gas, and the interfacial roughening transition},
  journal = {Physical Review B},
  volume  = {14},
  year    = {1976},
  pages   = {4978--4982}
}

@article{DrouffeZuber1983,
  author  = {Drouffe, Jean-Michel and Zuber, Jean-Bernard},
  title   = {Strong coupling and mean field methods in lattice gauge theories},
  journal = {Physics Reports},
  volume  = {102},
  year    = {1983},
  pages   = {1--119}
}

@article{LuscherWeisz2004,
  author  = {L\"uscher, Martin and Weisz, Peter},
  title   = {String excitation energies in SU(N) gauge theories beyond the free-string approximation},
  journal = {Journal of High Energy Physics},
  volume  = {0407},
  year    = {2004},
  pages   = {014}
}

Prerequisites

08.08.01
08.15.01

Tier anchors

beginner: Weeks, in Ordering in Strongly Fluctuating Condensed Matter Systems (1980), the crystal-surface picture; Kardar, Statistical Physics of Fields, Ch. 9 (interface fluctuations)
intermediate: Drouffe-Zuber, Phys. Rep. 102 (1983) 1, the strong-coupling expansion and roughening; Kogut, Rev. Mod. Phys. 51 (1979) 659, §VI; Lüscher-Münster-Weisz, Nucl. Phys. B 180 (1981) 1
master: Lüscher, Nucl. Phys. B 180 (1980) 317 (the universal string correction); Lüscher-Weisz, JHEP 0407 (2004) 014; Chui-Weeks, Phys. Rev. B 14 (1976) 4978 (SOS roughening via Coulomb gas); Itzykson-Drouffe, Statistical Field Theory Vol. 2, Ch. 6

References

Lüscher, M. — Symmetry-breaking aspects of the roughening transition in gauge theories · Nucl. Phys. B 180 (1980) 317; the effective string and the universal -pi/(12 r) correction
Lüscher, M., Symanzik, K., Weisz, P. — Anomalies of the free loop wave equation in the WKB approximation · Nucl. Phys. B 173 (1980) 365; the conformal anomaly behind the Lüscher term
Lüscher, M., Münster, G., Weisz, P. — How thick are chromoelectric flux tubes? · Nucl. Phys. B 180 (1981) 1; logarithmic broadening of the flux tube and roughening
Chui, S. T., Weeks, J. D. — Phase transition in the two-dimensional Coulomb gas, and the interfacial roughening transition · Phys. Rev. B 14 (1976) 4978; the SOS-to-Coulomb-gas duality
Drouffe, J.-M., Zuber, J.-B. — Strong coupling and mean field methods in lattice gauge theories · Phys. Rep. 102 (1983) 1; roughening as the breakdown radius of the strong-coupling series
Lüscher, M., Weisz, P. — String excitation energies in SU(N) gauge theories beyond the free-string approximation · JHEP 0407 (2004) 014; the open-string spectrum and Lorentz-invariance constraints
Weeks, J. D. — The roughening transition · in Ordering in Strongly Fluctuating Condensed Matter Systems, ed. Riste (Plenum 1980); the crystal-surface and SOS exposition

Estimated time

beginner: 18m
intermediate: 45m
master: 85m