Fractal deconfinement and confinement in Sierpinski ice

Technical Summary: Fractal Deconfinement and Confinement in Sierpinski Ice

Problem Statement
The six-vertex model is a cornerstone of exactly solvable constrained statistical mechanics, typically studied on square lattices where it exhibits Berezinskii-Kosterlitz-Thouless (BKT) transitions and Coulomb phases with emergent gauge fields. However, the role of dimensionality and geometry in the long-wavelength behavior of such constrained models remains a critical area of inquiry. This paper investigates the six-vertex model on the Sierpinski gasket, a hierarchical fractal with Hausdorff dimension $d_f = \log_2 3 \approx 1.58$ and finite ramification. The authors aim to determine how the fractal geometry and finite ramification alter the confinement properties of charges and the nature of correlations, specifically comparing the standard "ice" point (equal vertex weights) against other parameter regimes.

Methodology
The authors utilize the recursive hierarchical structure of the Sierpinski gasket to derive exact recursion relations for the partition function and observables.

• Model Definition: The model is defined on the edges of the gasket with Ising spins ($\sigma_i = \pm 1$) representing arrows. The ice rule (two spins in, two out) restricts vertices to six configurations. Under Ising symmetry ($\sigma_j \to -\sigma_j$), these reduce to three unique vertex types with Boltzmann weights $(x, y, z)$.
• Recursive Formulation: The partition function is calculated by considering the $n$-th generation of the gasket as three copies of the $(n-1)$-th generation. The authors define recursion variables based on the boundary conditions of the sub-triangles:
  • $Z_n$: The partition function for neutral boundary conditions (no charges at corners).
  • $Q_n$: The constrained partition function for configurations with a pair of opposite charges at the corner vertices.
• Recursion Equations: For the symmetric case $y=z \equiv \lambda$, the authors derive coupled recursion relations:
  $$Z_{n+1} = \lambda^3 Z_n^3 + 2x^3 Q_n^3$$
  $$Q_{n+1} = \lambda x^2 Q_n^3 + \lambda^2 x Q_n^2 Z_n$$
  These relations are generalized for $y \neq z$ using additional variables to track spin chirality.
• Observables: Key parameters are defined as ratios of these partition functions:
  • $P_n = D_n/F_n$: A measure of spin correlations (chirality ordering).
  • $R_n = Z_n/Q_n$: The relative weight of a neutral system versus a charged system, directly related to the free energy cost of a string connecting charges.
  • String tension $\tau$ and fractal dimension of the string $d_l$ are extracted from the asymptotic behavior of $R_n$ and the scaling of string lengths.

Key Contributions and Results

1. Entropic Confinement at the Ice Point:
   At the degenerate ice point ($x=y=z$), the authors analytically demonstrate that charges are entropically confined. The ratio $R_n$ grows asymptotically as $R_n \sim \alpha^{2n} = \alpha^L$ (where $\alpha \approx 1.7$), leading to a linear potential $\beta V(L) = L \ln(\alpha)$. This implies a non-zero string tension $\tau = \ln(\alpha)$. Consequently, the correlation length is short ($\xi \approx 0.75$ edge lengths), and correlations decay exponentially. The entropy of the ice point is extensive, scaling with the fractal dimension $d_f$.

2. Fractal Deconfinement:
   The paper identifies a specific regime where confinement vanishes. In the four-vertex limit where $y=0$ and $x \geq z$, the system exhibits fractal deconfinement.

   • Vanishing Tension: In this regime, the recursion variable $R_n$ flows to a finite fixed point $R^* = 2x/z$. This results in a vanishing string tension ($\tau \to 0$).
   • Fractal String Dimension: Unlike the 1D scaling of strings in confined phases, the string joining deconfined charges in this regime is a statistical fractal. The average length of the string scales as $\langle l \rangle_{n+1} = \frac{5}{2} \langle l \rangle_n$. This yields a fractal dimension for the string of $d_l = \log_2(5/2) \approx 1.3$, which is strictly less than the fractal dimension of the lattice ($d_f \approx 1.58$).
   • Emergent Superspin Model: The deconfined phase is mapped to an emergent Ising superspin model. The low-energy states are described by "superspins" ($S_p = \pm 1$) associated with dual plaquettes, representing the circulation of directed loops. The charges act as defects that prevent superspin flipping, reducing entropy but not creating an energetic barrier, thus maintaining $\tau=0$.

3. Phase Diagram:
   The analysis reveals a rich phase diagram where, for generic finite weights, the system flows to a confining fixed point ($P^*=1, R^* \to \infty$). However, specific lines in the parameter space (specifically $y=0, x \geq z$ and the point $x=y, z=0$) host the deconfined phase.

Significance and Claims
The paper claims to provide the first analytical demonstration of fractal deconfinement, a phenomenon where the string connecting charges is itself a statistical fractal with a dimension distinct from the underlying lattice. This contrasts with standard confinement in square lattices, where strings typically stay close to the shortest path, and highlights how finite ramification and fractal geometry can fundamentally alter the nature of emergent gauge theories.

Furthermore, the authors propose a concrete experimental realization of this system using Artificial Spin Ice (ASI). They argue that by tuning the relative distances of the tips of nanomagnetic islands at the vertices, one can engineer the specific Boltzmann weights required to access the degenerate ice point or the fractal deconfined regime. This addresses a long-standing experimental obstacle in creating vertex-degenerate configurations in ASI, offering a platform to study the rich phenomenology of Sierpinski ice, including the dynamics of fractal strings and charge confinement.

The work also suggests that these techniques can be extended to other fractals (e.g., Sierpinski carpet) and 3D structures (e.g., Sierpinski pyramid), potentially offering insights into corrections for loop effects in pyrochlore spin ice and quantum generalizations on NISQ platforms.