Abelian BF theory and linking observables

Use oriented D equal to four with path-integral weight exp(iS). Take a to be a compact U(1) one-form connection and b a compact U(1) two-form connection, both normalized by angular holonomies. Their local redundancies are derivatives of a zero-form and a one-form respectively. The BF integrand is a four-form, so it can be integrated on spacetime without a metric. Choose the overall minus sign below to fix the positive linking phase used later.

The coefficient N is a nonzero positive integer for the compact untwisted theory considered here. Its quantization follows from large-gauge consistency, not from dimensional analysis alone. A noncompact Gaussian theory would have different global sectors and cannot be assumed to reproduce the same finite discrete gauge theory. On manifolds with boundary, integrations by parts leave boundary terms and require an additional boundary condition or boundary theory; the closed-manifold formula does not supply those choices.

Insert a charge-m surface operator formed by integrating b over a closed oriented surface. Integrating over b enforces a curvature singularity for a supported on that surface. A charge-q Wilson line then measures the resulting holonomy. When the line bounds an oriented two-chain, intersect that chain with the source surface to define the linking number. Stokes' theorem turns the singular curvature into a root-of-unity phase.

The worldline has dimension one and the surface has dimension two, whose sum is D minus one as linking requires. In a dynamical interpretation, the surface can represent a string worldsheet while the line represents a particle worldline. Reversing the selected orientation changes the linking sign and conjugates the phase. On a manifold where the cycles do not admit the specified linking construction, one must use the appropriate global pairing instead of inventing a unique integer from a drawing.

In the compact level-N untwisted theory, the relevant line and surface charges are residues modulo N. On a closed spatial three-manifold, the topological ground-state space can be represented by flat discrete gauge sectors. For a three-torus there are three independent one-cycles, each with N possible holonomies, giving N cubed states. On a three-sphere there are no such nontrivial cycles and this simple theory has one state.

The general count involves first cohomology with finite coefficients and can contain torsion information, so replacing it by N to the first Betti number is not always correct. Twisted discrete gauge theories can add action data beyond this untwisted BF model. A microscopic finite sample can also split topological sectors by exponentially small processes, whereas the ideal topological action records the limiting sector structure. Exact linking phases, low-energy degeneracy, and an experimentally established gapped phase are related statements with different evidential requirements.