08.15.01 · stat-mech / topological-defects

The Kosterlitz-Thouless transition (2D XY model)

shipped3 tiersLean: none

Anchor (Master): Kosterlitz 1974 (J. Phys. C 7:1046); Nelson-Kosterlitz 1977 (Phys. Rev. Lett. 39:1201); Kardar Ch. 8 §8.4

Intuition Beginner

Picture a flat sheet covered with tiny compass needles, one at each point of a grid. Each needle can swing freely in the plane of the sheet, pointing in any direction like a clock hand. Neighboring needles prefer to point the same way: it costs energy when two adjacent needles disagree. This is the two-dimensional XY model, and the question is whether, at a given temperature, the whole sheet settles into one shared direction.

The surprising answer is no. In two dimensions a continuous direction can never lock in across the whole sheet at any positive temperature, because gentle long-wavelength waves of the direction cost almost nothing and pile up to wash out any global alignment. This is the content of the companion result on the absence of long-range order 08.02.03. So one might expect just a single hot, disordered phase all the way down to absolute zero.

Yet experiments and exact arguments show two distinct phases with a sharp transition between them. The resolution is a special kind of disturbance called a vortex: a point around which the needle direction winds once through a full turn, like water spiraling down a drain. At low temperature vortices come glued in tight pairs, one spinning clockwise, one counterclockwise. At high temperature these pairs rip apart and roam free.

The transition is the moment the pairs unbind. Below it, correlations between distant needles fade slowly, as a gentle power of distance; above it, they die off fast. Kosterlitz and Thouless found this in 1973, and it won the 2016 Nobel Prize in Physics.

Visual Beginner

The left panel shows a single vortex: the small arrows representing needle directions wind once around the center, turning through a full circle as you walk a loop around the core. The right panel shows a bound vortex-antivortex pair, one winding clockwise and one counterclockwise, with their distortions canceling at long range so the pair is cheap to make and stays local.

The picture captures the whole story in one image. A lone vortex stirs the direction field all the way out to the edge of the sheet, which costs an energy that grows with the size of the sheet. A bound pair only disturbs the region between the two cores, so it costs a fixed, finite amount. The transition is the temperature at which lone vortices become cheap enough to appear on their own.

Worked example Beginner

Walk around a single vortex and count how much the needle direction turns, then estimate the energy.

Step 1. Place a vortex at the center of the sheet. Stand far from the core and walk one full loop around it, staying on a circle of radius . As you go all the way around, the needle direction turns by exactly one full circle, which is degrees. That one full turn is the signature of a vortex; a loop with no vortex inside would bring the direction back with zero net turning.

Step 2. The amount the direction changes as you step a small distance is largest near the core and smallest far away. On a circle of radius the full degrees is spread evenly around a path of length , so the local twisting is gentle when is large and steep when is small.

Step 3. The energy stored in the twist on a thin ring at radius turns out to be the same for every ring spaced by equal factors of distance: the ring from to stores the same energy as the ring from to , and so on. Adding up equal contributions from each doubling of distance gives a total that grows like the number of doublings, which is the logarithm of the sheet size.

Step 4. Put in numbers. If the sheet is lattice spacings across, the number of doublings from one spacing out to the edge is about . So a lone vortex costs roughly ten times the energy of a single ring, while a bound pair separated by, say, spacings costs only about ring-units, because beyond the pair the twists cancel.

What this tells us: a single vortex costs an energy that grows with the size of the sheet, so lone vortices are heavily penalized at low temperature, while bound pairs cost only a fixed amount. The competition between this growing energy and the freedom a vortex has to sit anywhere on the sheet is what sets the transition temperature.

Check your understanding Beginner

Formal definition Intermediate+

The two-dimensional XY model assigns to each site of a square lattice a planar unit spin with . The reduced Hamiltonian (energy in units of , with the dimensionless coupling or stiffness) is

summed over nearest-neighbour bonds. In the continuum limit the slowly varying part of the field has the spin-wave action of a free massless boson,

which is the Gaussian field theory of 08.06.01. This quadratic theory alone produces the quasi-long-range order of the low-temperature phase: the correlation function decays as a power law,

with no nonzero limit at infinity, consistent with the absence of conventional long-range order 08.02.03.

The angle field is not globally single-valued: it is defined modulo , and this is where topology enters. A vortex of integer charge centered at is a configuration for which

for every loop encircling once. The integer is the winding number, an element of ; it cannot change under any smooth deformation of the field, which is what makes the vortex a topological defect. Decomposing into a smooth spin-wave part plus a singular vortex part (a planar Hodge decomposition) separates the partition function into a Gaussian factor and a sum over vortex configurations.

The energy of a configuration of vortices with charges at positions , computed by integrating with the singular field, is the two-dimensional Coulomb gas energy

where is the 2D Coulomb (logarithmic) Green's function, the lattice spacing, and a core energy. Charge neutrality is enforced by the finiteness of the energy. The fugacity controls the density of vortices.

Counterexamples to common slips

  • The transition is not a symmetry-breaking transition in the Landau sense: by 08.02.03 the continuous symmetry is never broken at , and the order parameter vanishes on both sides. What changes is the decay law of correlations, power-law below and exponential above, not the value of an order parameter.
  • Quasi-long-range order is genuine criticality, not a single critical point: the entire low-temperature phase is a line of fixed points with a continuously varying exponent . Treating as an isolated critical temperature with a single set of exponents misses this.
  • The vortices, not the spin waves, drive the transition. The spin-wave (Gaussian) theory is exactly soluble and has no transition by itself; omitting vortices, as a naive continuum treatment does, removes the physics entirely.

Key theorem with proof Intermediate+

The central quantitative claim is the energy-entropy estimate for the transition temperature, the argument by which Kosterlitz and Thouless first located the transition [Kosterlitz-Thouless 1973].

Theorem (energy-entropy estimate of ). Consider a single vortex of charge on a square sheet of linear size (area , in units of the lattice spacing ). Its free energy is

so a free vortex is thermodynamically favourable, , once . The marginal coupling gives the mean-field estimate of the transition.

Proof. The energy of one vortex is the self-energy from the logarithmic Coulomb gas with a single charge. Integrating for the field of a unit vortex over the sheet,

With this is . The energy grows with system size, the hallmark of an isolated topological defect in two dimensions.

The entropy comes from the number of places the vortex core can sit. There are of order lattice sites available, so the entropy (in units of ) is

The free energy, measured in units of so that , is therefore

As the sign of is the sign of . For (low temperature) the free energy diverges to : isolated vortices are forbidden and survive only as bound neutral pairs. For (high temperature) the free energy diverges to : the system gains free energy by proliferating free vortices, which destroy quasi-long-range order. The balance point , equivalently , is the estimate of the transition.

Bridge. This energy-entropy balance builds toward the renormalization-group treatment that makes it exact. The foundational reason a single number controls the transition is that both the vortex energy and the vortex entropy scale logarithmically with system size, so their competition is a pure comparison of coefficients of . This is exactly the structure that the RG exploits: the logarithm signals a marginal coupling, and marginal couplings are precisely the ones whose fate is decided by flow rather than by power counting, the lesson of 08.04.05. The crude estimate ignores the screening of vortex pairs, which renormalizes the stiffness downward as one coarse-grains; putting these together, the true transition is the point where the renormalized stiffness, not the bare one, hits . The energy-entropy argument identifies the transition with the balance , and the bridge is that this same balance reappears as the fixed-point condition of the Kosterlitz recursion relations, where it acquires the universal jump that the naive estimate cannot see. The two-dimensional Coulomb gas of vortices generalises the single-defect picture to the full interacting problem, and its dielectric screening is the foundational reason the bare and renormalized stiffness differ.

Exercises Intermediate+

Advanced results Master

The energy-entropy estimate locates the transition but cannot describe its character, because it ignores the polarizability of the vortex pairs. A dilute gas of bound pairs screens the interaction between any test pair, reducing the effective stiffness; the screening grows as one coarse-grains, and whether it runs away (free vortices) or saturates (bound phase) is decided by a renormalization-group flow in the two-dimensional space of couplings . This is the Kosterlitz analysis of 1974 [Kosterlitz 1974].

Theorem (Kosterlitz recursion relations). Let be the renormalized stiffness and the renormalized vortex fugacity at coarse-graining scale . To leading order in the vortex density,

The half-line is a line of fixed points (the quasi-ordered phase), and the point is the multicritical point governing the transition.

Writing for the deviation from criticality, the linearized flow near the fixed point is

whose trajectories satisfy the conserved quantity , i.e. the flow lines are hyperbolae . The separatrix , the straight lines through the origin, divides the plane: initial conditions with on the side flow into the fixed line (, low temperature); those with flow to large (vortex unbinding, high temperature). The physical temperature axis crosses the separatrix at one point, which is the true .

Theorem (essential singularity of the correlation length). Approaching the transition from above, the correlation length diverges not as a power law but with an essential singularity

for a nonuniversal constant .

This follows from the hyperbolic flow. On the high-temperature side the trajectory passes through the narrow neck between the separatrix branches; the RG "time" spent traversing the neck scales as , and exponentiates this. The exponent of the divergence is in the usual classification; the transition is of infinite order, with all derivatives of the free energy continuous, which is why it evaded the Landau classification entirely.

Theorem (universal jump of the stiffness; Nelson-Kosterlitz 1977). The renormalized stiffness is the macroscopic (superfluid) stiffness. As it approaches the universal value

and drops discontinuously to zero for . The dimensionless jump is independent of all microscopic detail.

The value is forced by the fixed-point structure: the stable fixed line terminates exactly at , so the last point of the quasi-ordered phase reached by the flow has regardless of the bare coupling. Nelson and Kosterlitz [Nelson-Kosterlitz 1977] recognized that for a two-dimensional superfluid the stiffness is the superfluid density (up to factors of and mass), so the prediction is a universal jump at the transition. This was confirmed in thin-film superfluid helium by Bishop and Reppy in 1978, one of the cleanest quantitative checks of a renormalization-group prediction in condensed-matter physics.

Synthesis. The Kosterlitz-Thouless transition is the foundational example of a phase transition driven by topology rather than by symmetry breaking. The central insight is that the two-dimensional XY model evades the Mermin-Wagner prohibition on long-range order 08.02.03 not by a loophole but by replacing the order parameter with quasi-long-range order, a line of critical points on which the exponent varies continuously, and the transition is the unbinding of vortex-antivortex pairs whose winding number lives in . Putting these together, the energy-entropy estimate , the mapping to a logarithmic Coulomb gas, the sine-Gordon dual, and the Kosterlitz recursion relations are four presentations of one structure, and the bridge between them is the marginal coupling: the logarithm common to the vortex energy and the Coulomb interaction is exactly what makes marginal at the fixed point, which is the foundational reason the renormalization-group treatment of 08.04.05 is mandatory rather than optional here. This is exactly the same marginal-operator analysis that governs the sine-Gordon and free-boson conformal field theories, and it generalises to every system with a order parameter in two dimensions: superfluid films, thin-film superconductors, two-dimensional melting through the KTHNY theory, and arrays of Josephson junctions. The universal jump identifies the macroscopic stiffness with the fixed-point value, and the bridge to experiment is that this dimensionless number is measurable as the discontinuity in the superfluid density of a helium film, a renormalization-group prediction with no free parameters.

Full proof set Master

Proposition (logarithmic self-energy of a vortex). A single charge- vortex on a disk of radius and core radius has energy .

Proof. Outside the core the field minimizing subject to is the rotationally symmetric solution , for which . Then

The integral is the source of the size-dependent energy; the additive constant absorbs the unknown short-distance physics inside the core.

Proposition (energy-entropy balance). The single-vortex free energy on a sheet of area changes sign at .

Proof. From the previous proposition with , , the energy is . The number of available core positions is , giving entropy . The free energy in units of is . The constant is dominated by the term as , so . The free energy is positive for (vortices suppressed) and negative for (vortices proliferate), changing sign at .

Proposition (eigenvalue of the fugacity at the fixed point). At the fixed point , the vortex fugacity is exactly marginal; for it is irrelevant and for relevant.

Proof. The fugacity term in the dual sine-Gordon action is , whose scaling dimension is (computed from the free-boson two-point function: the operator has dimension -normalized to in these conventions). In two dimensions an operator of dimension has RG eigenvalue , so to linear order. The eigenvalue vanishes at (marginal), is negative for (irrelevant, ), and positive for (relevant, grows). This is the linearized recursion relation; the quadratic feedback from vortex screening closes the flow.

Proposition (hyperbolic flow and the conserved quantity). Near the fixed point, with , the trajectories of the linearized flow are hyperbolae for a constant .

Proof. Linearizing about : from and , with giving near . Forming and , one finds , so is constant along the flow. Setting gives , a hyperbola. The separatrix consists of the lines ; trajectories with and terminate on the fixed line, while trajectories escape to large .

Proposition (universal value of the renormalized stiffness at ). The renormalized stiffness at the transition is .

Proof. The renormalized stiffness is . On the low-temperature side the flow terminates on the fixed line , and the trajectory at the transition is the separatrix, which enters the fixed line exactly at its endpoint , since any endpoint with would correspond to a flow that did not reach criticality. Hence the last quasi-ordered state, reached as , has , i.e. . For the fugacity grows without bound, vortices proliferate, and . The limiting value is independent of the bare coupling because every subcritical trajectory is funneled to the same fixed-line endpoint, which is the content of universality.

The essential-singularity statement for is established in Kosterlitz 1974 [Kosterlitz 1974] from the time spent traversing the hyperbolic neck; the proof that the jump is exactly for superfluid density, identifying stiffness with , is Nelson-Kosterlitz 1977 [Nelson-Kosterlitz 1977].

Connections Master

  • Mermin-Wagner theorem 08.02.03. The Kosterlitz-Thouless transition is the resolution of the puzzle that Mermin-Wagner poses: a continuous symmetry cannot break in two dimensions, yet the XY model still has two distinct phases. The transition reconciles these by being a transition between two symmetric phases, distinguished by power-law versus exponential decay of correlations, with the order parameter vanishing throughout. The spin-wave contribution that destroys long-range order is exactly the Gaussian fluctuation Mermin-Wagner identifies; KT adds the vortices on top.

  • Momentum-shell renormalization group 08.04.05. The Kosterlitz recursion relations are a two-coupling RG flow of precisely the kind the momentum-shell method produces, and the marginality of the vortex fugacity at the fixed point is why an RG treatment is required rather than optional. The momentum-shell language of relevant, irrelevant, and marginal operators classifies the fugacity directly: is marginal at , the borderline case whose fate the flow alone decides, exactly the situation the general RG framework is built to handle.

  • Gaussian field theory and free boson 08.06.01. The low-temperature phase of the XY model is the two-dimensional free massless boson, and its quasi-long-range order with continuously varying exponent is computed entirely from the Gaussian propagator . The vortex sector is the non-Gaussian addition; via the sine-Gordon duality the full problem is the free boson perturbed by a cosine, so the Gaussian theory is the unperturbed backbone of the entire analysis.

  • Fundamental group 03.12.00. The vortex charge is the winding number of the angle field around the defect core, an element of , the fundamental group of the order-parameter space. This is the precise sense in which the vortex is a topological defect: its charge is a homotopy invariant that no smooth deformation can remove, and the classification of defects in ordered media is the classification of homotopy classes of maps from loops into the order-parameter manifold.

Historical & philosophical context Master

Vadim Berezinskii first analyzed the two-dimensional XY model with topological excitations in 1971, in two papers in Zhurnal Eksperimental'noi i Teoreticheskoi Fiziki (translated in Sov. Phys. JETP 32 (1971) 493) [Berezinskii 1971], identifying that vortices control the low-temperature behaviour and arguing for a transition distinct from any symmetry-breaking one. Working independently in the West, J. Michael Kosterlitz and David J. Thouless published the energy-entropy argument and the qualitative theory of vortex unbinding in 1973 in Journal of Physics C: Solid State Physics 6 (1973) 1181 [Kosterlitz-Thouless 1973], introducing the phrase "metastability and phase transitions in two-dimensional systems" and giving the estimate . The quantitative renormalization-group treatment, with the recursion relations for stiffness and fugacity and the hyperbolic flow that yields the essential singularity in the correlation length, is Kosterlitz's 1974 paper in the same journal, J. Phys. C 7 (1974) 1046 [Kosterlitz 1974].

The universal jump in the stiffness was derived by David Nelson and Kosterlitz in 1977 in Physical Review Letters 39 (1977) 1201 [Nelson-Kosterlitz 1977], who identified the renormalized stiffness with the superfluid density of a two-dimensional superfluid and predicted the dimensionless discontinuity . The prediction was confirmed in superfluid helium films by D. J. Bishop and J. D. Reppy in 1978. The same theoretical apparatus was extended by Halperin, Nelson, and Young in the late 1970s to two-dimensional melting, where dislocation and disclination unbinding produce a two-stage KTHNY transition through an intermediate hexatic phase. Kosterlitz and Thouless shared half of the 2016 Nobel Prize in Physics (with F. Duncan M. Haldane awarded the other half) for theoretical discoveries of topological phase transitions and topological phases of matter; Berezinskii, who died in 1980, was not eligible, and the transition is now commonly written BKT to record his priority.

Bibliography Master

@article{Berezinskii1971,
  author  = {Berezinskii, V. L.},
  title   = {Destruction of Long-range Order in One-dimensional and Two-dimensional Systems Having a Continuous Symmetry Group I. Classical Systems},
  journal = {Sov. Phys. JETP},
  volume  = {32},
  year    = {1971},
  pages   = {493--500}
}

@article{KosterlitzThouless1973,
  author  = {Kosterlitz, J. M. and Thouless, D. J.},
  title   = {Ordering, Metastability and Phase Transitions in Two-dimensional Systems},
  journal = {J. Phys. C: Solid State Phys.},
  volume  = {6},
  year    = {1973},
  pages   = {1181--1203}
}

@article{Kosterlitz1974,
  author  = {Kosterlitz, J. M.},
  title   = {The Critical Properties of the Two-dimensional $xy$ Model},
  journal = {J. Phys. C: Solid State Phys.},
  volume  = {7},
  year    = {1974},
  pages   = {1046--1060}
}

@article{NelsonKosterlitz1977,
  author  = {Nelson, David R. and Kosterlitz, J. M.},
  title   = {Universal Jump in the Superfluid Density of Two-dimensional Superfluids},
  journal = {Phys. Rev. Lett.},
  volume  = {39},
  year    = {1977},
  pages   = {1201--1205}
}

@article{BishopReppy1978,
  author  = {Bishop, D. J. and Reppy, J. D.},
  title   = {Study of the Superfluid Transition in Two-dimensional $^4$He Films},
  journal = {Phys. Rev. Lett.},
  volume  = {40},
  year    = {1978},
  pages   = {1727--1730}
}

@article{HalperinNelson1978,
  author  = {Halperin, B. I. and Nelson, David R.},
  title   = {Theory of Two-Dimensional Melting},
  journal = {Phys. Rev. Lett.},
  volume  = {41},
  year    = {1978},
  pages   = {121--124}
}

@book{KardarFields,
  author    = {Kardar, Mehran},
  title     = {Statistical Physics of Fields},
  publisher = {Cambridge University Press},
  year      = {2007}
}

@article{JoseKadanoff1977,
  author  = {Jos{\'e}, Jorge V. and Kadanoff, Leo P. and Kirkpatrick, Scott and Nelson, David R.},
  title   = {Renormalization, Vortices, and Symmetry-breaking Perturbations in the Two-dimensional Planar Model},
  journal = {Phys. Rev. B},
  volume  = {16},
  year    = {1977},
  pages   = {1217--1241}
}