Choosing an Entanglement Measure for the Physical Question
An entanglement or correlation measure is useful only when its construction matches the physical question. Geometry, mixed-state certification, renormalization, symmetry, and scrambling require different observables; comparing their numerical values without matching domains creates a ranking with no operational meaning.
Required background. Use the Information-Measure Domain and Comparison Atlas. Helpful background. Compare Entanglement Negativity in QFT, Reflected Entropy and Canonical Purifications, and Multipartite Information and Entropy Cones.
Begin with the task
Section titled “Begin with the task”For universal geometric data, use a logarithmic coefficient, sphere constant, corner function, or matched subtraction—not a raw cutoff-dependent entropy. For total correlation between separated algebras, use mutual information. For mixed-state entanglement certification, negativity can be appropriate after choosing a transpose convention. For canonical-purification correlation, use reflected entropy. For multipartite linear constraints, use entropy vectors and the cone belonging to the declared state class.
For renormalization-group comparisons, choose a quantity with a proven monotonicity theorem in the stated dimension and geometry. For symmetry-resolved questions, declare charge sectors and whether probabilities or within-sector entropies are included. For scrambling or channel questions, an entropy snapshot is usually insufficient; specify time-dependent algebras, reference states, or channel distances.
The diagram is a compact decision map.
Choose the construction from the question: geometric universality, total correlation, mixed-state entanglement, canonical purification, or multipartite constraints. Similar numerical scales do not make measures interchangeable. Schematic.
Decision matrix
Section titled “Decision matrix”| Question | Measure | Essential declaration | Strongest supported conclusion |
|---|---|---|---|
| Is a geometric coefficient universal? | log coefficient, sphere constant, or corner function | dimension, shape, state, counterterm class | scheme-independent data for that family |
| Are separated regions correlated? | mutual information | two algebras, separation, regulator | bound on total bounded-observable correlation |
| Is a mixed state non-PPT? | logarithmic negativity | partial transpose, grading, geometry | entanglement certification within that construction |
| How correlated is the canonical doubling? | reflected entropy | doubled algebra, support, split or cutoff | canonical-purification correlation |
| Which linear entropy constraints hold? | entropy vector and cone | parties, state class, purifier | membership or violation for that cone |
| How distinguishable are evolutions? | energy-constrained channel distance | Hamiltonian, energy, ancilla | operational bias on the constrained input set |
Apply the matrix to four examples. Entanglement harvesting needs detector algebras and a two-probe entanglement measure. An RG claim needs a dimension-specific monotone, not any decreasing entropy. A gauge-region comparison needs an algebra and center. Scrambling needs a time-dependent information-flow task, often conditional information or a channel quantity.
Adversarial domain check
Section titled “Adversarial domain check”The validity figure asks what changes if a hidden choice is varied.
Counterterms affect raw geometric constants; purification and transpose choices affect mixed-state measures; ensemble and state class affect entropy cones. A conclusion survives only variations that its definition is designed to ignore. Schematic.
Before reporting a result, ask whether a different algebra, regulator, purification, transpose, state class, or energy constraint could reverse the conclusion. If so, make that choice part of the claim rather than hiding it.
Further reading
Section titled “Further reading”- Calabrese, Pasquale, John Cardy, and Erik Tonni. “Entanglement Negativity in Quantum Field Theory.” Physical Review Letters 109 (2012): 130502. DOI.
- Dutta, Souvik, and Thomas Faulkner. “A Canonical Purification for the Entanglement Wedge Cross-Section.” Journal of High Energy Physics 03 (2021): 178. DOI.
- 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.