A two-dimensional system of coupled phases cannot order the ordinary way — thermal fluctuations forbid it. Yet it still has a transition, of a stranger kind: below a critical temperature, vortices are bound in pairs and correlations fall off gently as a power law; above it the pairs unbind, correlations decay exponentially, and the length over which phases stay aligned diverges without ever developing true long-range order. No symmetry breaks, no local order parameter moves. Cross the temperature below and watch the character of the correlations flip.
The BKT transition is defined by how the phase–phase correlation $C(r)=\langle\cos(\theta_r-\theta_0)\rangle$ decays. Below $T_{\mathrm{BKT}}$ it decays algebraically, $C(r)\sim r^{-\eta(T)}$ with an exponent that climbs to exactly $\tfrac14$ at the transition — a whole low-temperature phase of critical, scale-free correlations. Above $T_{\mathrm{BKT}}$ it decays exponentially, cut off at a correlation length $\xi$ that does not vanish smoothly but diverges with an essential singularity as the transition is approached from above.
The transition temperature is fixed by the coupling as $T_{\mathrm{BKT}}=\tfrac{\pi}{2}J$; the correlation-length parameter is $b\approx1.5$. The stiffness that survives up to the transition drops there by a universal amount — the Nelson–Kosterlitz jump $\rho_s(T_{\mathrm{BKT}}^-)=\tfrac{2}{\pi}T_{\mathrm{BKT}}$. All three numbers come straight from the library's BKT universals.
The left plot is $C(r)$ on log–log axes, where a power law is a straight line. Keep $T$ below $T_{\mathrm{BKT}}$ and the correlations are a clean straight line whose slope $-\eta$ steepens toward $-\tfrac14$ as you warm up — the whole phase is critical. Push $T$ past $T_{\mathrm{BKT}}$ and the line bends down and falls off a cliff: the exponential cutoff at $\xi$ has arrived. The right plot tracks $\xi(T)$; approach the transition from above and it shoots up faster than any power — the essential singularity — while below the transition it is effectively infinite, the signature that there is order of a kind without a length scale at all.
The coupled-oscillator networks the platform controls are close cousins of the 2D XY model, and the same vortex physics governs when a large array of them can hold a common phase. Knowing $T_{\mathrm{BKT}}$ for a given coupling graph tells you the noise budget within which synchronisation survives as topological order rather than fragile long-range alignment — a more robust thing to aim for on real, fluctuating hardware.
| Regime | Correlations | Vortices |
|---|---|---|
| $T < T_{\mathrm{BKT}}$ | algebraic $r^{-\eta}$, $\eta\le\tfrac14$ | bound in pairs |
| $T = T_{\mathrm{BKT}}$ | $r^{-1/4}$; stiffness jumps by $\tfrac{2}{\pi}T_{\mathrm{BKT}}$ | on the verge of unbinding |
| $T > T_{\mathrm{BKT}}$ | exponential $e^{-r/\xi}$ | free; $\xi$ shrinks as $T$ rises |
Evidence boundary: the panel plots the library's exact BKT universals — $T_{\mathrm{BKT}}=\tfrac\pi2 J$, $\eta(T_{\mathrm{BKT}})=\tfrac14$, $\xi\sim\exp(b/\sqrt{T-T_{\mathrm{BKT}}})$ with $b\approx1.5$ — live in your browser, verified against a reference to better than $10^{-12}$. The correlation curves are the standard analytic forms of each phase, not a Monte-Carlo simulation of a specific lattice; it is a teaching panel for the transition's shape, not a measurement.