The Kosterlitz-Thouless Transition: Vortex Unbinding in Two Dimensions

What the transition is: order without order in flatland

The natural stage for the KT transition is the two-dimensional XY model: a plane of classical spins, each a unit vector free to point in any direction in the plane, coupled to neighbors by an energy E = −J·Σ cos(θ_i − θ_j). The coupling J favors alignment, but in two dimensions thermal fluctuations of the phase angle θ are so effective that they destroy any true long-range order at every finite temperature.

This is the content of the Mermin-Wagner theorem (1966): a continuous symmetry cannot be spontaneously broken in 2D at T > 0. So there is no ordered phase in the usual sense — the average magnetization <ψ> is zero at all temperatures. Yet experiments on thin helium films clearly showed a superfluid transition. Kosterlitz and Thouless resolved the paradox: below a critical temperature the system enters a quasi-ordered phase where correlations decay only as a slow power law, g(r) ~ r^(−η), rather than exponentially. This is quasi-long-range order — enough coherence to carry a supercurrent, but not true order.

The mechanism: vortices and an energy-entropy balance

The key objects are topological defects called vortices: points around which the phase θ winds by ±2π. A single vortex costs an energy that grows with system size, E ≈ π·J·ln(R/a), where R is the sample size and a the lattice spacing. That logarithm is decisive.

The entropy of placing one vortex is S ≈ k_B·ln(R/a)² = 2k_B·ln(R/a), since there are about (R/a)² sites to put it on. The free energy is therefore:

• F = E − T·S ≈ (π·J − 2·k_B·T)·ln(R/a)

Below T_KT ≈ π·J/(2·k_B) the coefficient is positive: a free vortex is infinitely costly, so vortices survive only as tightly bound vortex–antivortex pairs whose energy is finite. Above T_KT the entropy wins, F turns negative, and it becomes favorable to unbind the pairs and proliferate free vortices. The free vortices screen the interaction and destroy quasi-order. This energy-entropy argument by Kosterlitz and Thouless (1973) is the heart of the transition. A full renormalization-group treatment of the vortex fugacity confirms it and yields the famous flow equations near the critical point.

Key quantities: the universal jump and a worked estimate

The transition's sharpest prediction is the universal jump in the superfluid density, derived by David Nelson and Kosterlitz in 1977. Right at the transition the superfluid areal mass density ρ_s drops discontinuously from a universal value to zero, fixed by:

• k_B·T_KT = (π/2)·(ħ²/m²)·ρ_s(T_KT⁻)

where m is the mass of the superfluid particle (a ⁴He atom) and ħ is the reduced Planck constant. Equivalently the phase-space density n_s·λ² at the jump equals 4 for an ideal 2D Bose gas. Rearranged, ρ_s(T_KT)/T_KT = 2·m²·k_B/(π·ħ²), a pure combination of constants — no material parameters.

Worked number: for a ⁴He film with m = 6.65×10⁻²⁷ kg at T_KT ≈ 1 K, the predicted jump is ρ_s ≈ 2·m²·k_B·T/(π·ħ²) ≈ 3.5×10⁻⁸ kg/m², corresponding to roughly a monolayer areal coverage — exactly the scale seen in torsional-oscillator experiments. The correlation length above T_KT diverges as ξ ~ exp(b/√(T−T_KT)) with b of order 1, an essential singularity far fiercer than any power law.

How it is observed and measured

Because no quantity diverges at T_KT, detecting the transition means hunting for its signatures rather than a peak:

• Superfluid ⁴He films: Bishop and Reppy (Phys. Rev. Lett. 40, 1727, 1978) used a torsional oscillator wound with Mylar to measure the superfluid mass and dissipation of adsorbed helium films, confirming the universal jump and its predicted magnitude — the landmark experimental verification.
• 2D superconductors and Josephson-junction arrays: the current–voltage relation becomes nonlinear, V ~ I^a, with the exponent a jumping from 1 (ohmic, above T_KT) to a ≥ 3 exactly at the transition, a hallmark used in thin films of NbN, InO, and cuprates.
• Ultracold atoms: Hadzibabic and Dalibard (Nature 441, 1118, 2006) imaged interference fringes and their proliferating dislocations — direct pictures of free vortices — and later experiments (Chin group, 2013) measured the universal jump in a 2D Bose gas directly.

In each case the tell-tale is the combination of a nonzero but algebraically decaying stiffness below T_KT and its universal discontinuity at the transition.

How it differs from its cousins

The KT transition is easy to confuse with more familiar phase changes; the distinctions are physical:

• vs. ordinary second-order transitions: those break a symmetry and have a power-law-divergent specific heat (a critical exponent α). The KT specific heat has only a smooth, non-singular bump near T_KT — the entropy released by vortex unbinding — with the actual singularity hidden in the essential-singularity correlation length.
• vs. 3D superfluidity/superconductivity: in three dimensions <ψ> is genuinely nonzero and vortices are line defects with finite energy per length; there is no vortex-unbinding transition of the KT type.
• vs. the Mermin-Wagner "no order" statement: Mermin-Wagner forbids true long-range order in 2D, and KT does not violate it — the low-T phase has only quasi-order, consistent with the theorem.

The transition also generalizes: two-dimensional melting (the KTHNY theory of Halperin, Nelson, and Young, 1978-79) proceeds through two successive KT-like steps, unbinding dislocations then disclinations, passing through an intermediate hexatic phase.

Significance, famous cases, and open questions

The KT transition was the first clear example of a phase change governed by topology rather than symmetry breaking, and it opened the entire field of topological phases — the same conceptual thread that runs through the quantum Hall effect and topological insulators. That lineage is precisely why the 2016 Nobel Prize in Physics went to Kosterlitz, Thouless, and Haldane; Vadim Berezinskii, who published the vortex idea in 1971–72, had died in 1980 and could not share it, though his name survives in "BKT."

• Famous realizations: ⁴He films, thin-film superconductors, ⁴He/³He layers, liquid-crystal films, Josephson arrays, cold-atom 2D gases, and exciton-polariton condensates.
• Regime of validity: strictly a 2D, U(1)-symmetric phenomenon; real 3D-coupled or finite-size systems show only broadened, rounded signatures.

Open and active questions include the exact nature of the BKT transition in strongly disordered superconductors (where it competes with a superconductor-insulator quantum transition), non-equilibrium BKT physics in driven-dissipative polariton condensates, and precise numerical determination of the logarithmic finite-size corrections that make the universal jump so hard to pin down experimentally.

The KT transition contrasted with a conventional second-order (Ising/mean-field) transition and with a true 3D long-range-ordered phase.

Property	KT transition (2D XY)	Second-order transition (3D)
Order parameter <ψ>	Zero on both sides (no symmetry breaking)	Nonzero below T_c (symmetry broken)
Low-T correlations	Quasi-long-range: g(r) ~ r^(−η), power law	True long-range order: g(r) → const
Order of transition	Infinite order (all derivatives of F continuous)	Second order (specific heat diverges)
Correlation length ξ above T_c	exp(b/√(T−T_KT)) — essential singularity	|T−T_c|^(−ν), power law, ν ≈ 0.63–0.71
Driving mechanism	Unbinding of bound vortex-antivortex pairs	Growth of ordered domains / spin flips
Exponent η at T_c	Universal value η = 1/4	Non-universal / different universality class