Universal, Multipartite, and Shape-Dependent Entanglement

Entanglement in quantum field theory is not one number. A raw subregion entropy records local ultraviolet structure; a universal coefficient records what survives an allowed change of regulator; mutual information measures total correlation; negativity tests a partial transpose; reflected entropy studies a canonical purification; and an entropy cone constrains a whole vector of subsystem entropies. The central skill of this chapter is to choose the construction that answers the physical question and to carry its hypotheses all the way to the conclusion.

Helpful background. Review UV divergences and the area law, mutual information, anomaly coefficients, and topological entanglement diagnostics as needed.

Geometry, state, and measure

Before writing an entropy symbol, fix four pieces of data.

1. Subsystem. Name the region and its algebra. A lattice factor, a split inclusion, and a sharp type-III local algebra do not supply the same density matrix.
2. State. Distinguish a vacuum, thermal state, excited state, or restricted state. A conformal map that works for the vacuum need not work for an excitation.
3. Ultraviolet prescription. State the cutoff, subtraction, counterterm class, and limiting family. A constant is universal only relative to the changes that are allowed.
4. Information construction. Decide whether the question concerns entropy, total correlation, mixed-state entanglement, a canonical purification, spectral data, or multipartite compatibility.

The chapter map begins with these data and then separates four questions that are often conflated.

Matched subtraction can expose universal geometric data; positive separation leads to mutual information; a partial transpose or canonical purification defines a mixed-state measure; and a declared state class defines an entropy cone. These branches answer different questions. Schematic.

Read the arrows as operations, not as implications. Knowing a universal corner coefficient does not determine negativity. Knowing mutual information does not certify that the correlation is quantum. Satisfying a holographic entropy inequality does not prove that a state has a gravitational dual.

What the chapter establishes

The geometric route starts with universal terms, develops smooth shape variations, resolves corners and cusps, and checks exact interval, sphere, and cylinder geometries. The correlation route moves from disjoint mutual information to negativity and reflected entropy. The final route treats entropy cones, entanglement spectra, topological subtraction, thermal and excited states, and measure selection.

The comparison below is the chapter’s common reference. “Regulator independent” never means “assumption free”: the state, algebra, geometry, and limiting procedure still belong to the definition.

Entanglement constructions, their domains, and the strongest conclusion each supports

Construction	State class	Region geometry	Regulator dependence	Physical meaning	Computability	Main ambiguity
Raw subregion entropy	Any regulated state	One chosen region and algebra	Power divergences and many finite terms depend on the prescription	Entropy of the regulated restriction	Replica methods, covariance matrices, tensor networks, or numerics	No intrinsic sharp-region density matrix in a type-III algebra
Universal log, sphere, or corner term	Usually vacuum CFT or a specified gapped phase	Fixed smooth shape, angle, or subtraction geometry	Invariant only under the declared counterterm and regulator class	Anomaly, stress-tensor, defect, or long-range data	Replica geometry, conformal maps, shape response, or continuum fits	A cutoff-independent-looking constant can still be scheme dependent
Mutual information	Any state on two declared algebras	Preferably regions at positive separation	Local boundary divergences cancel in a matched construction	Total classical and quantum correlation	Entropy differences, relative entropy, replica correlators, or bounds	It does not isolate mixed-state entanglement
Logarithmic negativity	Bipartite mixed or pure state with a transpose convention	Adjacent and disjoint regions have different ultraviolet behavior	Requires a type-I regulator, split, or algebraic replacement	Non-PPT entanglement and an entanglement monotone in its domain	Trace norm, even-replica continuation, or covariance methods	Fermionic transpose choices and zero-negativity bound entanglement
Reflected entropy	Mixed state in a fixed algebraic realization	Bipartite regions, with support and split data stated	The continuum split or cutoff limit must be controlled	Total correlation in the canonical doubled purification	Square-root purification, replicas, or Gaussian covariance data	The auxiliary copies are not a physical preparation
Entropy vector and cone	Generic quantum, stabilizer, holographic, or another declared class	All nonempty unions of the same parties	Each component needs one common subsystem prescription	Compatibility with homogeneous entropy inequalities	Entropy evaluation followed by linear or convex-cone tests	Special-class inequalities are not generic QFT laws
Entanglement spectrum	Regulated state with symmetry sectors declared	One cut, boundary condition, and physical size	High levels and additive modular-energy shifts are cutoff sensitive	Spectrum of a regulated reduced state; possible low-energy diagnostics	Correlation matrices, diagonalization, or tensor-network transfer data	Degeneracy alone need not diagnose a phase
Topological subtraction	Gapped 2+1-dimensional ground state under standard subtraction assumptions	Matched overlapping regions far from physical boundaries	Local boundary terms cancel; finite-size corrections remain	A long-range constant; γ = log D for intrinsic topological order in the standard domain	Balanced entropy combinations and size extrapolation	Interfaces, corners, gaplessness, or small regions can mimic a constant
Excess or thermal entropy	Matched vacuum, thermal, excited, or eigenstate family	Fixed region, volume, and ensemble	Ultraviolet cancellation requires identical short-distance structure	State-dependent entropy, including thermal and classical contributions	Conformal maps, relative entropy, free fields, or many-body numerics	A volume law is not automatically bipartite quantum entanglement

The table also explains why no single “best entanglement measure” exists. A measure is strong when its invariances and operation class match the question, and misleading when they do not. Representative primary constructions include the CFT interval formulas Calabrese and Cardy 2004, §§2–3, the disjoint-region expansion Cardy 2013, §§2–3, the entanglement-spectrum diagnostic Li and Haldane 2008, pp. 010504-1–010504-4, and topological subtraction Kitaev and Preskill 2006, pp. 110404-1–110404-4.

A guided route

For geometry and universality, read the first four pages in order. Track which coefficient is being extracted, which local terms are subtracted, and whether the deformation is smooth or singular. The interval and sphere examples then show how exact conformal maps encode both the power and the limits of the result.

For correlation and mixed-state entanglement, begin with mutual information and keep one finite-dimensional compass in mind. With natural logarithms, a Bell pair has

$$I(A{:}B)=2\log 2, \qquad \mathcal E(A{:}B)=\log 2, \qquad S_R(A{:}B)=2\log 2.$$

A perfectly correlated classical bit has $I=S_R=\log 2$ but $\mathcal E=0$, while a product state gives zero for all three. Thus equalities in special states do not make the constructions interchangeable.

For multipartite, spectral, and long-range questions, first make every subsystem choice consistent. Tripartite information can have either sign in ordinary quantum states; a finite entanglement spectrum belongs to a regulator; and topological subtraction requires a gapped scaling regime. Read the final measure-selection page whenever the desired conclusion is operational rather than purely geometric.

Universality has a domain

The second map separates a potentially universal remainder from the choices that still delimit it.

A matched construction can remove local cutoff terms, but it does not remove the need to state geometry, counterterm class, algebra, state, replica continuation, partial transpose, purification, ensemble, or entropy-cone class. Schematic.

Three failure tests recur throughout the chapter.

• Vary an allowed scheme choice. If the proposed coefficient shifts under an allowed local counterterm, it was not universal in that comparison.
• Approach the edge of the domain. Send a smooth deformation toward a cusp, separated regions toward contact, or a gapped subtraction toward $L/\xi\sim1$. A new singular term or failed scaling law marks a change of problem.
• Replace the construction. Exchange even for odd negativity replicas, change a fermionic transpose, choose another purification, or test a holographic cone as though it were the generic quantum cone. A changed answer is expected because the observable has changed.

The primary literature makes these boundaries concrete: smooth CFT shape response is controlled by the nonlocal second variation Faulkner, Leigh, and Parrikar 2016, §§2–5; negativity requires the even-replica continuation Calabrese, Cardy, and Tonni 2013, §§3–4; reflected entropy starts from the canonical square-root purification Dutta and Faulkner 2021, §§2–3; and holographic monogamy belongs to the holographic entropy cone Bao et al. 2015, §§2–4.

Check your preparation

1. Why can the coefficient of a logarithm be universal when the entropy itself diverges?

Answer

Local ultraviolet physics produces regulator-dependent power divergences. In a fixed dimension and geometry, an allowed local counterterm may be unable to change a particular logarithmic coefficient, so that coefficient can survive regulator changes even though the full entropy does not. One must still state the state, algebra, shape, and counterterm class.

2. Which of mutual information, negativity, and reflected entropy vanishes for a perfectly correlated classical bit?

Answer

Negativity vanishes because the state is separable and its partial transpose is positive. Mutual information and reflected entropy are both $\log 2$ because both can detect total correlation, including classical correlation. This example prevents the three measures from being interpreted as interchangeable entanglement scales.

3. Does observing $I_3(A{:}B{:}C)<0$ prove that a state is holographic?

Answer

No. The inequality $I_3\leq0$ is necessary for the standard holographic entropy cone, but many nonholographic states also satisfy it. Conversely, generic quantum states can have either sign, so a positive value rules out that holographic inequality without violating a generic quantum law.

4. What must be scaled in a topological-entanglement calculation?

Answer

Use the same regulator and matched subtraction geometry at several region sizes. Require the region widths, junction-smoothing scales, and distances to physical boundaries to be large compared with the correlation length, control corners, and fit the residual finite-size dependence. A stable constant without these controls is not yet a topological diagnosis.

References

• Bao, Ning, Sepehr Nezami, Hirosi Ooguri, Bogdan Stoica, James Sully, and Michael Walter. “The Holographic Entropy Cone.” Journal of High Energy Physics 09 (2015): 130. DOI. Open PDF.
• Calabrese, Pasquale, and John Cardy. “Entanglement Entropy and Quantum Field Theory.” Journal of Statistical Mechanics (2004): P06002. DOI. Open PDF.
• Calabrese, Pasquale, John Cardy, and Erik Tonni. “Entanglement Negativity in Extended Systems: A Field Theoretical Approach.” Journal of Statistical Mechanics (2013): P02008. DOI. Open PDF.
• Cardy, John. “Some Results on the Mutual Information of Disjoint Regions in Higher Dimensions.” Journal of Physics A 46 (2013): 285402. DOI. Open PDF.
• Dutta, Souvik, and Thomas Faulkner. “A Canonical Purification for the Entanglement Wedge Cross-Section.” Journal of High Energy Physics 03 (2021): 178. DOI. Open PDF.
• Faulkner, Thomas, Robert G. Leigh, and Onkar Parrikar. “Shape Dependence of Entanglement Entropy in Conformal Field Theories.” Journal of High Energy Physics 04 (2016): 088. DOI. Open PDF.
• Kitaev, Alexei, and John Preskill. “Topological Entanglement Entropy.” Physical Review Letters 96 (2006): 110404. DOI. Open PDF.
• Li, Hui, and F. Duncan M. Haldane. “Entanglement Spectrum as a Generalization of Entanglement Entropy: Identification of Topological Order in Non-Abelian Fractional Quantum Hall Effect States.” Physical Review Letters 101 (2008): 010504. DOI. Open PDF.