arXiv:2610.19286
Mutual Information of Measurement in Conformal Field Theories
Israel Klich, Ravisankar Rajagopal, Diana Vaman
Abstract
The mutual information of measurement of a local operator between two spatial regions characterizes the reduction in the uncertainty of measurement outcomes in one region given the outcome in the other. We study the Rényi-Shannon mutual information of measurement in conformal field theories (CFTs). Explicit calculations for scalar primaries and descendants in large $N$ CFTs are performed in the two extreme limits of the intervals being adjacent or largely separated. The results depend on only the cross ratio and the scaling dimension $Δ$ of the operator being measured, with striking similarities to the mutual information of entanglement. We find a surprising quadratic dependence as $(Δ-1/2)^2$ for adjacent intervals. This dependence superficially seems to violate the information processing inequality, which we discuss and resolve. We also offer a holographic interpretation of the mutual information of measurement.
AI Summary
The paper studies classical correlations between outcomes of measuring a local observable in two spatial regions of a CFT, using Rényi–Shannon mutual information. It focuses on scalar single-trace primaries and descendants in two-dimensional large-$N$ CFTs, where suppressed higher connected correlators make the measurement distribution approximately Gaussian. A characteristic-functional expression for the second Rényi–Shannon entropy reduces the Gaussian calculation to a determinant of the equal-time covariance restricted to the measured region. Toeplitz/Fisher–Hartwig asymptotics give the leading adjacent-interval result $I_n=\tfrac12(\Delta-\tfrac12)^2\log[\ell_1\ell_2/(\epsilon(\ell_1+\ell_2))]$, independent of $n\geq1$ at Gaussian order. A second derivation uses the Casini–Huerta–Myers conformal map to thermal correlators. At $\Delta=1/2$, the logarithmic coefficient vanishes, leaving an order-one contribution. The apparent conflict between increasing mutual information under differentiation and data processing is resolved by noting that discrete derivatives crossing the partition boundary are not local processing within either region. For widely separated intervals, the transformed covariance motivates $I_2\propto\eta^{2\Delta}$; an explicit $\Delta=1$ calculation establishes cancellation of the linear term, without computing the leading coefficient. Odd connected cumulants cancel exactly from the second Rényi–Shannon entropy, so leading non-Gaussian corrections begin at order $1/N^2$ under the stated counting. The adjacent logarithm is established within a scaling-preserving regulator family; generic ultraviolet regularizations require contact-term subtraction. General interacting CFTs, finite-cross-ratio formulas, higher dimensions, and explicit holographic corrections are left open.
Central claim may not hold: The adjacent-interval logarithmic coefficient is established only for a specially imposed family of contact-subtracted symbols, not for general positive short-distance measurement regulators. For a Hermitian scalar primary at Δ=1, the contact-subtracted kernel has negative spectral density, while the proposed positive replacement reverses the separated-point correlation tail and therefore does not represent the claimed measurement covariance. Appendix B's symbol (sin²θ)^(Δ−1/2) has two Fisher–Hartwig singularities and produces twice the logarithmic determinant coefficient stated there.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.