Bipartite fidelity of critical dense polymers

TL;DR

This paper derives exact formulas for bipartite fidelity in critical dense polymers, a logarithmic CFT model, and compares lattice results with conformal field theory predictions, revealing universal correction terms.

Contribution

It provides the first exact closed-form expression for bipartite fidelity in a logarithmic CFT lattice model and explores universal correction terms.

Findings

Exact expression for $ ext{F}_d$ matches CFT predictions for $c=-2$

Universal correction term $-2 ext{log}((1+x)/(2 ext{sqrt}x))$ for $ ilde{ ext{F}}_2$

Lattice results agree with conformal field theory derivations

Abstract

We investigate the bipartite fidelity $\mathcal F_d$ for a lattice model described by a logarithmic CFT: the model of critical dense polymers. We define this observable in terms of a partition function on the pants geometry, where $d$ defects enter at the top of the pants lattice and exit in one of the legs. Using the correspondence with the XX spin chain, we obtain an exact closed-form expression for $\mathcal F_d$ and compute the leading terms in its $1/N$ asymptotic expansion as a function of $x = N_A/N$, where $N$ is the lattice width at the top of the pants and $N_A$ is the width of the leg where the defects exit. We find an agreement with the results of St\'ephan and Dubail for rational CFTs, with the central charge and conformal weights specialised to $c=-2$ and $\Delta = \Delta_{1,d+1} = d(d-2)/8$. We compute a second instance $\mathcal {\tilde F}_2$ of the bipartite fidelity for $d=2$ by imposing a different rule for the connection of the defects. In the conformal setting, this choice corresponds to inserting two boundary condition changing fields of weight $\Delta = 0$ that are logarithmic instead of primary. We compute the asymptotic expansion in this case as well and find a simple additive correction compared to $\mathcal F_2$, of the form $-2\log((1+x)/(2\sqrt{x}))$. We confirm this lattice result with a CFT derivation and find that this correction term is identical for all logarithmic theories, independently of $c$ and $\Delta$.