Fixed-Face-Area Closure Reconstruction in Coarse-Grained Loop Quantum Gravity

Technical Summary: Fixed-Face-Area Closure Reconstruction in Coarse-Grained Loop Quantum Gravity

Problem Statement
In Loop Quantum Gravity (LQG), coarse-graining a spin-network region often results in transported boundary fluxes that fail to satisfy the Gauss constraint (closure), leaving a non-zero closure defect. While a closed polyhedral comparison geometry is desirable for interpreting these states, standard reconstruction methods present a critical flaw: the usual $SL(2, \mathbb{C})$ boost used to enforce closure alters the norms of individual face vectors. Since the spin labels in LQG fix the face areas, changing these norms breaks the area-matching conditions required when coarse-grained regions share boundaries. The paper addresses the specific problem of constructing a closed polyhedral geometry that preserves every prescribed face area exactly, thereby maintaining compatibility with neighboring regions without altering the canonical boundary flux data.

Methodology
The author formulates a weighted nearest-closure reconstruction prescription. The approach treats the problem as a constrained optimization on the product of fixed-area spheres ($S_A = \prod S^2_{A_f}$).

1. Objective Function: The method minimizes a weighted squared-distance cost function, $E_X(Y) = \frac{1}{2} \sum w_f^{-1} \|Y_f - X_f\|^2$, where $X_f$ are the transported (non-closed) fluxes and $Y_f$ are the reconstructed closed fluxes. The weights $w_f$ allow for face-dependent resolution or uncertainty.
2. Constraints: The reconstruction is constrained to the closure surface $\sum Y_f = 0$ while strictly maintaining the norm constraint $\|Y_f\| = A_f$ for all faces.
3. Mathematical Framework:
   • The problem is analyzed using the Karush-Kuhn-Tucker (KKT) conditions.
   • Theorem 4 establishes that near any regular closed configuration (where face normals are not all collinear), there exists a unique, smooth, locally minimizing branch that is $SO(3)$-equivariant.
   • Corollary 5 demonstrates that the $3N$ components of the reconstructed vectors are determined by a single 3-dimensional Lagrange multiplier $\lambda \in \mathbb{R}^3$, reducing the problem to solving a 3D root-finding equation: $\sum \tilde{X}_f(\lambda) = 0$.
4. Graph Extension: For finite graphs with shared faces, the method is extended to a coupled reconstruction (Proposition 8). This minimizes the sum of node costs while enforcing that the transported area normals of shared faces match exactly ($Y^A_s + Y^B_s = 0$), ensuring consistency across the graph.

Key Contributions

• Fixed-Area Preservation: Unlike $SL(2, \mathbb{C})$ boosts, this reconstruction preserves individual face areas, preventing mismatches at shared boundaries.
• Dimensional Reduction: The reconstruction of an $N$-valent node is reduced to solving a 3-dimensional multiplier equation, independent of the valence $N$.
• Analytic Response: The paper provides a small-defect expansion (Proposition 6) yielding an analytic expression for the linear response of the normals and a quadratic formula for the reconstruction cost ($R^2_w$).
• Derived Observable Status: The paper clarifies that the reconstructed closed geometry is a "derived observable." It does not replace the canonical fluxes (which retain the closure defect and internal holonomy information) but provides a reproducible geometric representative for comparison.
• Coupled Solver: A coupled solver is developed for finite graphs that simultaneously enforces closure at nodes and area-normal matching on shared edges.

Results

• Numerical Verification: In 200 random six-face tests, the reconstruction achieved closure, area, and covariance residuals below $7 \times 10^{-16}$.
• Scaling Laws: Numerical scans confirmed the analytic predictions: the weighted face-resolved error scales quadratically with the defect norm ($\|D\|^2$), while the relative cost error scales linearly ($\|D\|$).
• Graph Consistency: Tests on a three-node cycle with independent perturbations showed that the coupled solver successfully enforced all closure and shared-face constraints simultaneously, reducing residuals to the numerical noise floor ($< 5.1 \times 10^{-16}$).
• Regularity: The method remains valid within a "regular branch" defined by non-collinear normals and non-vanishing denominators in the multiplier equation. The paper identifies the loss of rank in the response matrix $M$ as the boundary where this branch fails (e.g., as normals become collinear).

Significance and Claims
The paper claims to provide a reproducible fixed-area boundary readout that keeps canonical fluxes separate. Its primary significance lies in resolving the tension between enforcing geometric closure and preserving the area-matching constraints essential for gluing coarse-grained regions in LQG.

• The reconstruction serves as a diagnostic tool: the scalar cost $R^2_w$ and shape tensor invariants quantify the displacement and anisotropy of the transported data relative to the closed sector.
• The author explicitly states that this construction does not provide complete curvature information; full curvature still requires internal holonomies and transport data.
• The work is modest in scope, defining a specific geometric selection rule ($\pi_w$) rather than a new dynamical evolution or a complete quantum realization. It offers a consistent way to interpret coarse-grained states as closed polyhedra without altering the underlying spin-network data.