Skip to main navigation Skip to search Skip to main content

Dispersive CFT sum rules

  • McGill University
  • Stony Brook University
  • California Institute of Technology

Research output: Contribution to journalArticlepeer-review

68 Scopus citations

Abstract

We give a unified treatment of dispersive sum rules for four-point correlators in conformal field theory. We call a sum rule “dispersive” if it has double zeros at all double-twist operators above a fixed twist gap. Dispersive sum rules have their conceptual origin in Lorentzian kinematics and absorptive physics (the notion of double discontinuity). They have been discussed using three seemingly different methods: analytic functionals dual to double-twist operators, dispersion relations in position space, and dispersion relations in Mellin space. We show that these three approaches can be mapped into one another and lead to completely equivalent sum rules. A central idea of our discussion is a fully nonperturbative expansion of the correlator as a sum over Polyakov-Regge blocks. Unlike the usual OPE sum, the Polyakov-Regge expansion utilizes the data of two separate channels, while having (term by term) good Regge behavior in the third channel. We construct sum rules which are non-negative above the double-twist gap; they have the physical interpretation of a subtracted version of “superconvergence” sum rules. We expect dispersive sum rules to be a very useful tool to study expansions around mean-field theory, and to constrain the low-energy description of holographic CFTs with a large gap. We give examples of the first kind of applications, notably we exhibit a candidate extremal functional for the spin-two gap problem.

Original languageEnglish
Article number243
JournalJournal of High Energy Physics
Volume2021
Issue number5
DOIs
StatePublished - May 2021

Keywords

  • AdS-CFT Correspondence
  • Conformal and W Symmetry
  • Conformal Field Theory

Fingerprint

Dive into the research topics of 'Dispersive CFT sum rules'. Together they form a unique fingerprint.

Cite this