Mutual Information and Correlation Spreading
Mutual information measures all correlations accessible to two chosen subregion algebras, while a connected correlator tests one pair of observables. Tracking both after a localized excitation distinguishes a blind operator from a genuinely absent correlation signal. Neither quantity alone is a universal measure of causal influence.
Required background. Entanglement growth supplies the evolving reduced states, and mutual information in QFT supplies the finite relative-entropy definition.
Helpful background. Front extraction supplies threshold and window controls.
Two correlation diagnostics
Section titled “Two correlation diagnostics”For disjoint regions and ,
Leading local boundary divergences cancel at nonzero separation. For bounded operators , Pinsker-type reasoning bounds the connected correlator by mutual information:
Thus a nonzero connected correlator certifies nonzero mutual information, but a zero correlator for one operator pair does not imply . For unbounded field operators, smear them and establish an energy or norm cutoff before using this bound.
Wolf et al. 2008, Theorem 1 and Eqs. (4)–(6) use mutual information to connect total correlations and area-law structure.
Localized Gaussian excitation
Section titled “Localized Gaussian excitation”Prepare a finite-energy wavepacket near in a Gaussian field and evolve it toward a separated region . At each time compute:
- from covariance-matrix symplectic spectra;
- connected field and momentum correlators with declared smearing;
- the change relative to the unexcited state;
- an intervention contrast if signaling is the question.
Extract arrival contours using the same threshold protocol. If the field correlator is blind because the excitation is momentum-like, the momentum correlator or mutual information may still respond. Conversely, pre-existing vacuum mutual information may be nonzero before the excitation; use or a matched control rather than calling the baseline a transmitted signal.
Mutual information and connected correlators occupy the direct-diagnostic layer. Their comparison helps identify operator blindness and pre-existing correlations without assigning either a universal transport velocity. The map is schematic.
Long-range initial correlations
Section titled “Long-range initial correlations”Critical and long-range-entangled initial states can have algebraically decaying . A quench changes that correlation pattern, but nonzero spacelike is compatible with microcausality. To claim transported information, specify a local encoding choice and show that receiver statistics depend on it only inside the causal future.
An apparent early front can be a nonzero initial correlation, a threshold artifact, or a response of only one chosen operator. Matched initial-state and observable controls distinguish these cases. The map is schematic.
Common pitfalls
Section titled “Common pitfalls”Equating correlation with signaling. Mutual information can be nonzero in the vacuum. Signaling requires dependence on a localized intervention choice.
Using unsmeared fields in a norm bound. Local fields are unbounded distributions. Smear and regulate them before applying finite-operator inequalities.
References
Section titled “References”- Wolf, Michael M., Frank Verstraete, Matthew B. Hastings, and J. Ignacio Cirac. “Area Laws in Quantum Systems: Mutual Information and Correlations.” Physical Review Letters 100 (2008): 070502. DOI.