Is the Running Ginzburg–Landau Parameter of Confining Flux Tubes a Fluctuation Artifact?

When a quark and an antiquark are pulled apart, the gluon field between them collapses into a narrow tube of energy, called a flux tube. Physicists often describe this tube with a borrowed picture: the vacuum acts like a superconductor with electric and magnetic roles swapped, and the tube is the analog of a magnetic vortex. That picture comes with a closed-form profile, the Clem ansatz, which has two length scales. One is the core radius $\xi$. The other is the penetration depth $\lambda$, the distance over which the field dies off outside the core. Their ratio, the Ginzburg–Landau parameter $\kappa=\lambda/\xi$, should be a fixed property of the vacuum. High-precision lattice simulations of two-color gauge theory in two space dimensions show that it is not. When the Clem profile is fitted to measured tubes, the fitted $\kappa$ grows with the quark separation $R$. The same simulations find that the exponential tail length $\lambda$ disagrees with the inverse of the lightest glueball mass, with $\lambda M_0 = 1.16(3)$ rather than 1 (Intrinsic Width of the Flux Tube as a tool to explore confining mechanisms in Lattice Gauge Theories; Intrinsic Width of the flux tube in 2+1 dimensional Yang-Mills theories). Both findings are currently read as evidence that the dual-superconductor picture is incomplete.

This proposal tests a different reading. Fitting a profile derived for a static, classical vortex to a tube that fluctuates quantum mechanically may by itself produce a running $\kappa$ and a shifted $\lambda M$, even in a system that truly is a superconductor. The test has two parts. First, we build a controlled testbed: lattice simulations of the Abelian Higgs model, a genuine superconductor whose true $\kappa$ can be measured independently from particle masses. Second, we compute the fluctuation-averaged vortex profile with the functional renormalization group (a method that integrates out fluctuations scale by scale) in its local potential approximation. That truncation keeps the full nonlinear effective potential of the fields but simplifies how they depend on momentum. The outcome is binary and useful either way. If a real superconductor reproduces the "$\kappa$ problem", then the SU(2) observation cannot be used against dual superconductivity. If it does not, the SU(2) result becomes a much sharper statement.

Core Claim

The fitted $\kappa(R)$ and the ratio $\lambda M_0 \neq 1$ can be predicted from two effects that the Clem ansatz omits. The first is the smearing of the tube's position by its long-wavelength transverse wobbles. These are the modes described by effective string theory, and they grow with $R$. The second is the difference between three definitions of a mass: the curvature of the effective potential, the rate of spatial exponential decay, and the true particle (pole) mass. The local potential approximation equates these three by construction. Its extension with field-strength renormalization, known as LPA′, separates them.

We claim that a profile built from a fluctuation-dressed vortex convolved with a string-smearing kernel will reproduce the $R$-dependence of the fitted $\kappa$ in exact Abelian Higgs simulations. We also claim it will account for any gap between the fitted tail length and the measured pole mass. Both are checked against ground truth that the SU(2) problem lacks. The dressed profile is

$$\rho(r;R)=\int d^{2}u\; G_{R}(u)\,\rho_{\rm vortex}\big(|r-u|;\,\xi_{\rm eff},\lambda_{\rm eff}\big),$$

where $\rho$ is the energy density at transverse distance $r$ from the tube axis and $G_R$ is a Gaussian kernel. Its variance $s^2(R)$ grows like $\log R/(2\pi\sigma)$, with $\sigma$ the string tension, which is the leading-order effective-string broadening. The vortex profile $\rho_{\rm vortex}$ uses effective scales $\xi_{\rm eff}$ and $\lambda_{\rm eff}$ taken from the renormalization-group-dressed Higgs and vector screening masses, not from their potential curvatures.

The novelty lies in separating a vacuum property, the dressed masses, from a geometry-dependent smearing, $G_R$, and in calibrating that separation in a model where the answer is known. No work in the file applies renormalization-group flows to flux-tube profiles in position space, or tests the Clem-fit inconsistency on a system with a known constant $\kappa$.

Research Program

Prior work supplies the components but not the connection. Towards quantitative precision in the chiral crossover: Masses and fluctuation scales shows the key mechanism in a quark–meson model. The three mass definitions agree to within 1% at zero temperature, but equating curvature and onset masses at finite density produces errors of up to 38%. That paper also shows that LPA′ stays within about 3% of fully momentum-dependent results. We reuse its three-way mass comparison and move it from homogeneous mesons to a vortex profile, which that paper never considers.

Polyakov-quark–meson–diquark model for two-color QCD solves local-potential flows for two-color QCD, but only for a spatially uniform Polyakov-loop background. The Yang–Mills vertex-expansion calculations do something similar in momentum space. Landau gauge Yang-Mills correlation functions and Nonperturbative finite-temperature Yang-Mills theory generate the gluon mass gap dynamically, but only for homogeneous vacua. These works motivate the proposal: they establish that the truncation handles mass gaps well. They also constrain it, because none has faced an inhomogeneous background, so we use the simplest truncation that separates the mass definitions rather than a full vertex expansion.

The first component is lattice Monte Carlo for the U(1) Abelian Higgs model in 3+1 dimensions. External Dirac monopole–antimonopole pairs at separation $R$ create a single vortex string of length $R$. We scan the bare couplings so that the true $\kappa$ covers type-I, near-critical and type-II values. The true $\kappa$ is fixed independently from the Higgs and vector pole masses, which we measure from zero-momentum temporal correlators. The screening masses come from spatial correlators, and runs at several temporal extents give finite-temperature points. We then measure the energy-density profile at several $R$ and fit it with exactly the Clem procedure used for SU(2). The resulting "fitted $\kappa(R)$" is compared against the constant true $\kappa$.

This testbed has one honest limitation. A 2+1-dimensional Abelian superconductor confines magnetic flux, which is a pseudoscalar in two space dimensions and so cannot form a line between two point sources. The direct geometric analog of the SU(2)$_{2+1}$ setup therefore does not exist. What carries over between the two systems is the fitting pathology, not the numerical values.

The second component is an LPA′ flow for the same model. The effective potential $V_k$ and the field-strength factors $Z_k$ are integrated from the lattice cutoff down to scale $k$. This yields curvature, screening and approximate pole masses for the Higgs and vector fields. Those masses fix $\xi_{\rm eff}$ and $\lambda_{\rm eff}$, and the classical vortex equations are then solved with the dressed potential. The string modes are handled separately. Their spectrum is gapped by $\sim\pi/R$, and it enters only through the smearing kernel $G_R$. Separating them this way avoids double-counting the translational zero mode.

The third component transfers the pipeline to SU(2)$_{2+1}$. The lattice data sets there have no public repository, so we would regenerate a subset with the multilevel algorithm across Cellini et al.'s range, $\beta \in [8.768, 14.011]$ and $T/T_c\in[0.11,0.85]$. We fix the dressed-vortex parameters at one $R$ and predict the fitted $\kappa$ at all other $R$ with no further tuning.

Baselines

The naive reference point is the undressed Clem ansatz with a constant $\kappa$, fitted separately at each $R$. This is the procedure that produces the $\kappa$ problem in the first place. It serves both as the thing we explain and as the zero-parameter-change reference. In the Abelian Higgs testbed we score it by the gap between its fitted $\kappa(R)$ and the pole-mass $\kappa$.

The strongest existing competitor is effective-string smearing of a classical vortex: the convolution above with tree-level masses in place of dressed ones. It is the most direct extension of the intrinsic-width analysis of Cellini et al., and it is exactly what a rigid or Nambu–Gotō string treatment would add. The string-width studies with stochastic normalizing flows (Numerical determination of the width and shape of the effective string using Stochastic Normalizing Flows; Studying Effective String Theory using deep generative models) validate the Nambu–Gotō prediction of $\pi/3$ for the width coefficient, reporting $1.03\pm0.02$ and $1.01(9)$ respectively. Those results justify taking the kernel variance from string theory rather than fitting it.

For the target data, the reported SU(2)$_{2+1}$ numbers are $\lambda\sqrt\sigma=0.244(4)$ and $\lambda M_0=1.16(3)$. We found no reported result on Lacuna for the numerical values of the fitted $\kappa$ as a function of $R$. Our reproduction of that curve therefore becomes the baseline. Both models get identical profile data, the same fitting windows in transverse distance, and the same number of free parameters, which is two per model: string tension and kernel offset $R_0$ for the smearing baseline, and the corresponding pair for ours.

Beating the smearing baseline means one of two things. Either our model matches the fitted $\kappa(R)$ in the Abelian Higgs testbed within statistical errors where the baseline does not, or it predicts the out-of-sample $R$ values in SU(2) with lower $\chi^2$ per degree of freedom.

The ablation that isolates the key ingredient keeps the smearing kernel but replaces screening masses with curvature masses, which is pure LPA. If this version performs as well as LPA′, the mass splitting is irrelevant and our central mechanism is wrong.

Why not just…?

Why not just fit a better asymptotic form for the tail? For a string in three space dimensions the field decays like a Bessel function, roughly $e^{-mr}/\sqrt r$, not as a pure exponential. Fitting a pure exponential over a finite window biases $\lambda$ upward. This is a real threat to the glueball-ratio half of our claim, because it may explain $\lambda M \neq 1$ on its own. We address it directly: every fit is repeated with the correct asymptotic prefactor and a varied window. If the corrected fits give $\lambda\,m_{\rm pole}=1$ in the Abelian Higgs testbed, we drop the mass-splitting explanation for $\lambda M_0$ and keep only the $\kappa(R)$ question. A prefactor correction does not depend on $R$, so it cannot by itself make $\kappa$ run.

Why not just add rigidity terms to the effective string? The normalizing-flow study of rigid strings finds that curvature terms narrow the string but leave the profile nearly Gaussian (Numerical determination of the width and shape of the effective string using Stochastic Normalizing Flows). Cellini et al. attribute the exponential tails to bulk massive modes. A pure string theory therefore cannot produce $\lambda$ at all. It can only supply the kernel $G_R$, and that is how we use it.

Why not just run Monte Carlo on the Abelian Higgs model and stop there? That experiment alone answers the first question, whether a genuine superconductor shows a running fitted $\kappa$. It is cheap, and we perform it first. It does not tell us which fluctuation produces the running, and it gives no predictive rule to carry over to SU(2), where no independent $\kappa$ exists. The renormalization-group component earns its place only if it predicts the SU(2) curve out of sample. If it does not, the proposal reduces to the Monte Carlo diagnostic, which would still be worth publishing.

Why not just use full momentum-dependent vertex expansions? They are more accurate in homogeneous vacua (Landau gauge Yang-Mills correlation functions), but they have never been solved around an inhomogeneous background. LPA′ captures the curvature–screening split to within the reported 3% in the quark–meson case. We treat that level of accuracy as sufficient for an effect of order 16%.

Decisive Checks

The first check is in the Abelian Higgs model. We fit the Clem ansatz as a function of $R$ at fixed pole-mass $\kappa$, across type-I and type-II couplings. A clear, reproducible running would show that the SU(2) $\kappa$ problem does not falsify dual superconductivity. A flat fitted $\kappa$ would make the SU(2) inconsistency a genuine sign of non-Abelian dynamics.

The second check asks whether the smearing baseline gets even the direction of the running right. Naive broadening enlarges the apparent core, which lowers the fitted $\kappa$, whereas the lattice SU(2) data show it rising. If the sign is wrong, the dressed-mass term must supply the rising trend, and the LPA-versus-LPA′ ablation shows whether it does.

The third check compares $\lambda_{\rm fit}\,m_{\rm pole}$ with $\lambda_{\rm fit}\,m_{\rm scr}$ in the testbed, both at low temperature and near the transition. If a genuine superconductor shows a ratio near 1.16 when the fit uses the pole mass, and near 1 when it uses the screening mass, then the SU(2) value is best read as a mismatch between mass definitions and not as new physics.

The fourth check is the out-of-sample SU(2)$_{2+1}$ prediction of the fitted $\kappa$ at held-out $R$ and temperatures. The proposal succeeds only if it beats the smearing baseline there.