When Do Vortex Knots in Real Magnets Stay Tied?

A knotted vortex line, meaning a closed loop of twisted order such as a whirlpool's core, normally unties itself. When two strands touch, they cut and re-splice ("reconnect"), and the knot falls apart into plain rings. Recent mathematical work shows this is not always allowed. In materials whose local order can be twisted in ways that do not commute (doing twist A then B differs from B then A), each strand carries a label from a non-commuting group. Crossings and reconnections must respect that algebra, and some knots provably cannot come undone. Every medium where this has been shown so far is either an ultracold atomic gas or a liquid crystal, and every proof assumes the ideal symmetry holds exactly at all length scales.

This proposal asks whether a crystalline solid can host such protected knots, and over what range of knot sizes. The core idea is that topology alone cannot answer this. In a real magnet, weak spin-orbit effects lock the spin pattern to the crystal. Beyond a computable length, the non-commuting twists stop being free, and the protection that holds for small knots fails for large ones. Density functional theory (DFT), the standard method for computing a material's electronic ground state from its atoms alone, is the right tool for measuring that length. The result would be a protection window with both a lower and an upper size limit. Its upper edge is set by first-principles numbers, and the window can be tested against direct field simulations.

Core Claim

Topological protection of a knot is a statement about a symmetry group. In a solid, the effective group depends on length scale. We claim that for a noncoplanar multi-sublattice magnet, a protected knot of radius $R$ survives only inside a window

$$\xi \ll R \ll \ell_a, \qquad \ell_a = \sqrt{\rho_s / K},$$

Here $\xi$ is the vortex core size, $\rho_s$ is the spin stiffness (energy cost of slowly twisting the spin pattern), and $K$ is the anisotropy energy density that pins the spin pattern to the lattice. The form of $\ell_a$ comes from balancing gradient energy against anisotropy energy, the same balance that sets a magnetic domain-wall width. Below $\ell_a$, strands carry labels from the full non-commuting group, and the invariants of Topologically Protected Vortex Knots in an Experimentally Realizable System apply. Above $\ell_a$, strands get tied to domain walls and the protection argument no longer governs the outcome.

DFT supplies $\rho_s$, $K$, and, critically, which spin-rotation symmetries of the ground state survive once spin-orbit coupling is included. That surviving symmetry determines whether the strand labels are non-commuting at all.

The test case is the tetrahedral "3Q" order reported by DFT in Discovery of van Hove singularities: electronic fingerprints of 3Q magnetic order in a van der Waals quantum magnet for intercalated Co$_x$TaS$_2$. That paper uses DFT to link a noncoplanar four-sublattice spin state to electronic signatures; it says nothing about line defects. Our hypothesis is that, in the limit where only spin exchange matters, the tetrahedral 3Q state has a residual symmetry of spin rotations combined with lattice operations. That would make its coarse-grained order parameter a spin frame modulo a tetrahedral group. Its line defects could then carry charges in the binary tetrahedral group $T^*$, the same group Rajamäki, Annala and Möttönen found in spin-2 condensates and bent-core nematics.

This is a hypothesis, not a result. It may fail in two ways. The defects may require accompanying lattice dislocations. Or the effective group may turn out to be commuting, for instance $\mathbb{Z}_2$. Settling it is the first experiment.

Relation to Prior Work

Three papers supply the topological machinery, and we reuse them as exact ground truth inside the protection window:

All three assume a fixed, exact symmetry and say nothing about energetics. Our proposal differs by treating that symmetry as scale-dependent and computing where it breaks.

The method for passing first-principles numbers to a continuum model comes from A computational picture of hydride formation and dissipation in Nb SRF cavities. That paper computes Nb and NbH electronic parameters in JDFTx and feeds them into three-dimensional time-dependent Ginzburg–Landau simulations, a standard continuum model of an evolving order parameter. It finds that hydrides lower the vortex entry field by about 50%. Its model has a single complex order parameter, whose vortices carry ordinary additive integer charges, so it can never represent protected knots.

That paper and Vortex nucleation in superconductors within time-dependent Ginzburg-Landau theory in two and three dimensions, which shows inhomogeneities localize vortex nucleation and cut the superheating field to about 60% of its ideal value, also act as a warning. Real disorder weakens the order parameter locally, and this matters for whether knots can escape by attaching to weak regions.

The DFT side uses Noncollinear density functional theory, which treats spin as a full 2×2 density matrix. This formulation is needed to rotate a noncoplanar spin frame relative to the lattice with spin-orbit coupling switched on.

False-positive and false-negative assignments of topological insulators in density functional theory and hybrids constrains us. It shows that approximate functionals can flip topological conclusions, so our inferred symmetry and anisotropy must be checked across functionals.

Research Program

Step 1: what symmetry survives. We take the 3Q and competing single-Q states of Co$_x$TaS$_2$ and run noncollinear DFT with and without spin-orbit coupling. We map the energy as the tetrahedral spin frame is rigidly rotated over all orientations relative to the crystal. Without spin-orbit coupling this surface should be flat. With it, the shape and depth of the surface give $K$ and identify which discrete orientations remain degenerate. Separately, we compute spin-spiral energies at long wavelength to extract $\rho_s$.

From the residual group $H$ we compute the group of strand labels, $\pi_1(G/H)$. This is the group of distinct ways the order can twist around a loop, here the lift of $H$ into the spin double cover. We also determine by explicit construction whether each label needs a lattice dislocation alongside it. If, at the relevant scale, the labels form a commuting group or every non-trivial label requires a dislocation, the solid-state branch stops there. That outcome is itself a reportable negative result.

Step 2: continuum model. We build a coarse-grained spin-frame field model with stiffness $\rho_s$, a core-energy term setting $\xi$, and anisotropy $K$ taken from Step 1. It is integrated with dissipative relaxation dynamics on three-dimensional grids, following the finite-element practice of the Ginzburg–Landau papers above. Before any material-specific run, we validate the solver at $K = 0$:

  • Colored trefoils and figure-eights with $Q_b = 3$ must survive.
  • Configurations with $Q_b = 0$ must decay to trivial loops.

Together these checks reproduce Classification Theorem 13 of the $T^*$ paper numerically.

Step 3: map the window. We sweep knot radius $R$ across $\xi$ and $\ell_a$ and record survival, decay pathway and lifetime. A final arm inserts inclusions where the order parameter is suppressed, sized from Harbick-style parameters. This tests whether strands can end on a weak region and escape the protection argument entirely, an open question the file flags directly.

Baselines

Naive reference. The ordinary Gauss linking number, the standard commuting invariant, predicts that a knot survives only if pairwise linking is nonzero. It is the "do nothing" predictor any condensed-matter group would reach for first. It is run on exactly the same ensemble of initial configurations (trefoils, figure-eights, Borromean rings, unknots, at every $R$ and $K$). It is expected to fail on Borromean rings, following the quaternion-invariant paper.

Strongest existing method. This is the ideal non-commuting invariant ($Q_b$, $Q$, or $i$, as appropriate) from the three topology papers, applied as if the symmetry were exact. It is the strongest competitor because inside the window it is a theorem and should be essentially perfect. It is evaluated on the same simulated ensemble with no tuning. Our scale-aware predictor beats it only if it matches the ideal invariant for $R \ll \ell_a$ and is clearly more accurate for $R \gtrsim \ell_a$.

Abelian multiscale pipeline. The single-component DFT-to-Ginzburg–Landau approach of Harbick et al. is rebuilt with the same DFT inputs where they map. It predicts that every knot decays, so it bounds how much the non-commuting structure adds.

Key-ingredient ablation. The full model is run with $K$ set to zero, versus with $K$ drawn from a generic range rather than taken from DFT. This isolates the contribution of first-principles anisotropy.

Metric. We use balanced accuracy of predicting survival versus decay over the configuration ensemble, plus the error in the predicted critical radius. No reported result was found on Lacuna for knot-survival prediction in any medium, so every baseline number here will be produced in this study under identical grids, time budgets and initial conditions.

Why not just…?

Why not just do the symmetry analysis by hand from the magnetic space group? This is the most serious threat. Hand group theory tells which anisotropy terms are allowed, but not their size or sign, which determine $\ell_a$ and which orientations remain degenerate. It also cannot decide whether 3Q actually beats single-Q order in the material. The DFT-free ablation settles the question directly. If generic $K$ values reproduce the survival map as well as DFT-derived ones, DFT adds only bookkeeping, and we will say so.

Why not just simulate the field equations with generic parameters and watch? Simulations alone cannot say whether a real material lies inside the window. They are also sensitive to grid resolution near reconnection, which is the concern raised for the Gross–Pitaevskii and Ginzburg–Landau route. We use simulation as ground truth, but the invariant tells us which runs should come out which way. A mismatch at $K = 0$ would expose a solver error rather than physics.

Why not just use commuting invariants such as linking or winding numbers? In the quaternion-group setting, the quaternion-invariant paper shows that the linking number misses the protection of Borromean rings and that commuting charges cancel freely. The linking-number baseline quantifies this failure on our own ensemble.

Why not just stay with spin-2 condensates and liquid crystals? Those media already have $T^*$ protection, and Rajamäki et al. note that there is still no recipe for tying knots in them. A crystal would offer a pinning landscape, but also the disorder and anisotropy that might destroy protection. Only a calculation tied to a specific material can tell which effect wins, which is why the solid-state question is worth asking even though it may come out negative.

Decisive Checks

The program succeeds if four things hold:

  • DFT shows the 3Q residual symmetry gives non-commuting strand labels that need no lattice dislocation, and the result is stable across functionals.
  • The $K = 0$ simulations reproduce the $T^*$ classification exactly.
  • Survival falls off near the DFT-predicted $\ell_a$ and not at an arbitrary scale.
  • $\ell_a$ comfortably exceeds $\xi$.

Any one failure is informative. A commuting group rules out this material. A window narrower than a few core sizes rules out practical protection. Escape through inclusions would show that disorder, not topology, decides the fate of knots in solids.