Defect Relative Entropy Overview

• Defect relative entropy is a quantitative measure that captures the distinguishability of defect-induced states across classical, quantum, and conformal field theory systems.
• It utilizes generalized metrics to overcome the limitations of traditional KL divergence, ensuring finite, symmetric analysis for defects in various models.
• The approach enables practical applications in clustering, error correction, and phase transition analysis by unifying statistical, quantum, and field-theoretic contexts.

Defect relative entropy is a quantitative measure of distinguishability between states, operators, or configurations that are associated with defects—objects breaking translational invariance, such as conformal defects, topological interfaces, or impurity perturbations—in classical probability, quantum information, and field-theoretic systems. The concept encompasses direct generalizations of classical and quantum relative entropy, as well as rigorous metric constructions designed to overcome the drawbacks of the traditional Kullback–Leibler divergence when applied to defect spaces. Recent advances provide closed formulas for defect relative entropy in two-dimensional conformal field theory (CFT), spin chains with defects, and metric generalizations for arbitrary probability vectors, facilitating unified analysis across statistical, quantum, and field-theoretic contexts (Liu et al., 2017, Berta et al., 2014, Ghasemi, 29 Jan 2026, Arias, 2019).

1. Classical and Quantum Relative Entropy: Defects and Shortcomings

Relative entropy, denoted $D(P\|Q) = \sum_i P(i)\log(P(i)/Q(i))$ in the classical case and $$D(\rho\|\sigma) = \operatorname{Tr}[\rho(\log\rho-\log\sigma)]$$ in the quantum case, quantifies the statistical or quantum distinguishability between distributions (P,Q) or density matrices ($\rho,\sigma$). When these distributions are induced by inserting defects or interfaces in a system—e.g., reduced density matrices on either side of a conformal/topological defect—relative entropy provides an operational measure of defect-induced change.

However, classical KL divergence is not a true distance: it is asymmetric ($D(P\|Q)\neq D(Q\|P)$), fails the triangle inequality, and can diverge if $Q(i)\to0$ for some $i$ with $P(i)>0$. This is problematic for quantitative analyses requiring a finite, symmetric, and well-behaved metric, especially when comparing defect states or probability profiles in high-dimensional or singular spaces (Liu et al., 2017).

2. Generalized Relative Entropy: Metric Construction and Properties

To address these deficiencies, Liu et al. (Liu et al., 2017) introduce a generalized relative entropy metric for nonnegative vectors $X = (x_1,\ldots,x_s)$ and $Y = (y_1,\ldots,y_s)$, representing defect-induced probability distributions. With normalized frequencies $p_x(i) = x_i/\sum_j x_j$, $p_y(i) = y_i/\sum_j y_j$ and a parameter $k>1$, the generalized distance is: $$d(X,Y) = \sum_{i=1}^s\Bigl[\,p_x(i)\log\frac{k p_x(i)}{(k-1)p_x(i)+p_y(i)} + p_y(i)\log\frac{k p_y(i)}{p_x(i)+(k-1)p_y(i)}\,\Bigr] + r\log\Bigl(1+\frac{1}{k-1}\Bigr)$$ where $r = 0$ if $X=Y$, else $r=1$. This metric satisfies:

• Non-negativity and identity: $d(X,Y)\ge0$ with equality iff $X=Y$.
• Symmetry: $d(X,Y)=d(Y,X)$.
• Triangle inequality: $d(X,Z)\le d(X,Y)+d(Y,Z)$.
• Finite range: For $X\neq Y$, $2\log(k/(k-1)) \le d(X,Y)\le 4\log(k/(k-1))$.

By careful choice of denominators, this metric remains finite even for defect states with singular support, correcting for the divergence of KL in the presence of vanishing entries. The structure ensures a compact metric space on defect distributions, suitable for clustering, pattern recognition, and metric-based algorithms (Liu et al., 2017).

3. Defect Relative Entropy in Conformal Field Theory

Defect relative entropy in CFT is defined as the quantum relative entropy between reduced density matrices associated with two defect operators $\mathcal{I}_A$, $\mathcal{I}_B$: $$D(\mathcal{I}_A\|\mathcal{I}_B) = \mathrm{Tr}[\rho_A \log \rho_A - \rho_A \log \rho_B]$$ In a universal replica partition function approach on the circle, this gives (Ghasemi, 29 Jan 2026): $$D(\mathcal{I}_A\|\mathcal{I}_B) = \sum_{(i,\bar j)}\, p^A_{i,\bar j} \log\frac{p^A_{i,\bar j}}{p^B_{i,\bar j}}$$ with probabilities derived from modular S-matrix data and defect coefficients. For diagonal theories,

$$p^a_i = |\mathcal{S}_{ai}|^2, \qquad D(\mathcal{I}_a\|\mathcal{I}_{a'}) = \sum_i |\mathcal{S}_{ai}|^2 \log\left(\frac{|\mathcal{S}_{ai}|^2}{|\mathcal{S}_{a'i}|^2}\right)$$

This reduces exactly to the classical KL divergence over the modular S-matrix probability vectors: $$D(\mathcal{I}_a\|\mathcal{I}_{a'}) = \sum_i p^a_i \ln \left(\frac{p^a_i}{p^{a'}_i}\right)$$

4. Quantum Relative Entropy, Defects, and Remainder Terms

In quantum information, the monotonicity of relative entropy under quantum channels is refined via a defect or remainder term quantifying recoverability, utilizing the Petz recovery map (Berta et al., 2014): $$D(\rho\|\sigma) - D(\mathcal{N}(\rho)\|\mathcal{N}(\sigma)) \ge -\log F\left(\rho, (\mathcal{V}\circ \mathcal{R}_{\sigma,\mathcal{N}}\circ\mathcal{U})(\mathcal{N}(\rho))\right)$$ where $$F(\alpha,\beta) = \|\sqrt{\alpha}\sqrt{\beta}\|_1^2$$ is the fidelity and $\mathcal{R}_{\sigma,\mathcal{N}}$ is the Petz map with unitary rotations. This quantifies loss due to the defect (quantum channel), establishing equivalence among several entropy inequalities with defect-based remainder terms. The major open challenge is proving the bound for the bare Petz map without unitary corrections.

5. Sandwiched Rényi Defect Relative Entropy and Fidelity

Defect relative entropy admits a Rényi generalization (Ghasemi, 29 Jan 2026, Arias, 2019): $$D_n(\rho_A\|\rho_B) = \frac{1}{n-1}\ln \sum_i (p^A_i)^n (p^B_i)^{1-n}$$ For $n\to1$ this recovers the standard relative entropy, while $n=1/2$ relates to the Uhlmann fidelity: $$F(\mathcal{I}_A\|\mathcal{I}_B) = \sum_i \sqrt{p^A_i p^B_i}$$ In conformal settings and spin chains, defect fidelity and Rényi entropies provide a finer graded separation of defect states. In infinite fermionic chains with defect strength $\lambda$, quantum Rényi relative entropy scales as (Arias, 2019): $$S(\rho_1\|\rho_0) = \frac{c_{\text{eff}}(\lambda)}{3}\ln\ell + \mathrm{const}$$ where $c_{\text{eff}}(\lambda)$ is the defect-dependent effective central charge.

6. Explicit Examples and Defect Relative Sectors

Worked examples in CFT include:

• Ising model: DRE between identity and $\epsilon$ defects vanishes; between $\sigma$ and identity gives $\ln2$; defect relative sectors comprise sets where $D=0$ (e.g., $\{\mathbb{I},\epsilon\}$).
• Tricritical Ising: Sectors with $D=0$ reflect symmetry relations among defects.
• $\widehat{su}(2)_k$ WZW: DRE vanishes only between $\mathbb{Z}_2$-paired defects.

In spin chains, the leading scaling of DRE is proportional to the effective central charge, which measures defect transparency. Zero DRE identifies “defect relative sectors”—sets of defects indistinguishable by relative entropy—in both topological and conformal settings (Ghasemi, 29 Jan 2026).

7. Physical Significance and Applications

Defect relative entropy enables operational classification of defects by their induced distinguishability, offers UV-finite and modular-invariant metrics for identifying symmetry and topological relations, and is applicable in diverse areas:

• Metric-based clustering and learning algorithms in classical and quantum domains (Liu et al., 2017).
• Probing RG interfaces and non-invertible symmetries in CFT (Ghasemi, 29 Jan 2026).
• Quantitative analysis of phase transitions and fidelity markers in many-body systems (Arias, 2019).
• Quantum information recoverability and error correction via Petz-type maps and remainder bounds (Berta et al., 2014).

A plausible implication is that defect relative entropy affords a unified framework bridging classical, quantum, and field-theoretical notions of defect and interface distinguishability, grounded in metric theory, modular data, and entropic invariants.