Gauge-Field Entanglement and Edge Contributions

Gauge-field entropy can contain classical boundary-flux uncertainty, conditional bulk entanglement, representation-index terms from an extension, and continuum contact or edge determinants. These contributions belong to different prescriptions and have different operational meanings. They may be compared only after the algebra, boundary condition, flux measure, zero modes, and ultraviolet regulator have been matched.

Required background. Gauge subregions and centers fixes the algebra, flux sectors, and extended-space dictionary.

Helpful background. Lattice gauge Hamiltonians and Gauss law supplies a controlled finite regulator; UV divergences and the area law supplies the local divergence structure.

Chapter map. The overview gives the task-to-resource map, the comparison table, and the three gates for an operational claim.

Three decompositions that must be separated

For a finite lattice algebra with discrete electric-center sectors $E$, use a direct sum:

$$\rho_{\mathcal A_A}=\bigoplus_Ep_E\rho_{A,E}, \qquad S_{\mathcal A_A}=H(p_E)+\sum_Ep_ES(\rho_{A,E}).$$

This formula uses a sectorwise-normalized trace. It contains no separate representation-index term.

For a non-Abelian extended-Hilbert-space prescription, let $R=(r_\ell)_{\ell\in\partial A}$ be the complete constrained tuple of cut-link irreducible representations and $d_R=\prod_{\ell\in\partial A}d_{r_\ell}$. Then Donnelly 2012, Eqs. (28)–(32) gives

$$S_{\rm ext} =H(p_R)+\sum_Rp_RS(\rho_{A,R}) +\sum_Rp_R\log d_R.$$

The last term comes from maximally mixed representation indices created by the cut-link extension. It vanishes for Abelian groups and is not part of the continuum Maxwell formula below.

For continuum Maxwell theory, normal electric flux $E_\perp(x)$ is a continuous boundary field. The formal electric-center expression is the functional integral

$$S_A =\int\mathcal D E_\perp\,p[E_\perp] \left[-\log p[E_\perp]+S(\rho_{A,E_\perp})\right].$$

This is Donnelly and Wall 2015, Eq. (3). Both the density $p[E_\perp]$ and its first term are defined relative to the chosen measure $\mathcal D E_\perp$; they are not regulator-independent numbers by themselves.

Maxwell edge modes and the contact term

Place a brick wall a proper distance $\epsilon$ from a flat entangling surface and impose magnetic-conductor boundary conditions on the bulk modes. The omitted normal electric flux must then be summed explicitly. With transverse Laplacian eigenfunctions $\psi_n$ and eigenvalues $\lambda_n>0$,

$$E_\perp(x)=\sum_{n>0}E_n\psi_n(x),$$

the on-shell action in the product geometry used by Donnelly and Wall is

$$I(E_\perp) =\sum_{n>0}\frac{\beta E_n^2} {2\lambda_n\log(\epsilon^{-1})}.$$

This is Donnelly and Wall 2015, Eqs. (4)–(7). Their lattice-derived flux measure, Eq. (8), depends on the minimal electric charge, transverse area, and number of boundary cells. Gaussian integration then gives the edge determinant

$$Z_{\rm edge} =\det{}'\!\left[ \frac{\log(\epsilon^{-1})}{\beta} \frac{2\pi\operatorname{Area}(T)}{q^2} (-\nabla_T^2) \right]^{1/2},$$

up to the explicitly described local rescaling anomaly; see Donnelly and Wall 2015, Eq. (9). The prime excludes the transverse constant mode in that flat setup, where constant $E_\perp$ has infinite energy. On a compact space, global $U(1)$ flux can instead be quantized and must be summed or fixed according to the boundary ensemble. Gauss law alone does not prescribe one universal zero-mode deletion.

Kabat 1995, §3, Eqs. (3.1)–(3.2) and pp. 13–15 isolated the gauge-field contact contribution and emphasized that it was not the entropy of ordinary transverse photons. Donnelly and Wall 2015, Eqs. (9)–(10) subsequently matched that determinant to normal-flux edge modes. The chronology matters: the contact determinant is a gauge-fixed path-integral component, while its edge interpretation emerges only in the matched brick-wall construction.

The full curved-space treatment includes bulk scalars and vectors, ghosts, flat connections, zero modes, global electric and magnetic sectors, and the edge factor. Donnelly and Wall 2016, Eqs. (39)–(46) and (53)–(56) gives the brick-wall factorization and its lattice-to-continuum entropy measure. An isolated determinant can have a negative universal coefficient without being a negative probability or a negative total entropy.

Exact Maxwell sphere benchmark

For the vacuum of four-dimensional Maxwell theory across a sphere of radius $R$, define $c_{\log}$ by

$$S_{\rm univ}=c_{\log}\log(R/\epsilon).$$

In the convention of Donnelly and Wall 2015, Eqs. (12)–(14), the matched coefficients are:

Contribution	$c_{\log}$	Meaning
Transverse bulk or thermal calculation	$-16/45$	local propagating modes with the stated boundary treatment
Normal-flux edge determinant	$-1/3=-15/45$	electric-center edge modes with matched measure
Bulk plus edge	$-31/45$	anomaly-matched universal coefficient
Four-dimensional conformal anomaly	$-31/45$	independent target value

The quantitative check is exact:

$$-\frac{16}{45}-\frac13 =-\frac{16}{45}-\frac{15}{45} =-\frac{31}{45}.$$

The edge term repairs the missing universal logarithm under the matched prescription. It is not a claim that $-\tfrac13\log(R/\epsilon)$ is a separately measurable Bell-pair count, nor does it make regulator-dependent area terms universal.

Exact finite-group benchmark

The discrete comparison avoids differential-entropy ambiguities. Use the topological $Z_2$ state and simply connected four-boundary-link region worked out on the gauge-subregions page. There are three independent electric boundary generators and eight equiprobable sectors. Every conditional state is pure and every $Z_2$ representation has dimension one.

Quantity	Electric-center algebra	Extended Hilbert space	Magnetic-center algebra
Center Shannon term	$3\log2$	$3\log2$	$0$
Representation-index term	$0$	$0$	$0$
Conditional quantum term	$0$	$0$	$0$
Total assigned entropy	$3\log2$	$3\log2$	$0$
Gauge-invariant distillable entanglement	$0$	$0$	$0$

The electric and extended spectra both consist of eight eigenvalues $1/8$, so they agree sector by sector and give

$$S_E=S_{\rm ext}=3\log2=2.0794415417.$$

Changing to the magnetic center leaves the global state fixed but gives a single supported, pure Wilson-loop sector and total entropy zero, as in Casini, Huerta, and Rosabal 2014, Eqs. (60)–(62). The total and classical terms are therefore center dependent. The conditional term remains zero and, for the task of local gauge-invariant operations, so does the distillable entanglement.

For a non-Abelian finite-group unit test, use the trivial and two-dimensional standard irreducible representations of $S_3$. Choose sector weights, representation dimensions, and conditional entropies

$$p_R=\left(\frac13,\frac23\right), \qquad d_R=(1,2), \qquad S(\rho_{A,R})=(0,\log2).$$

These data have the required conjugate representation indices across the cut and a one-ebit multiplicity state in the standard sector. They give the exact split

Term	Nats	Bits
Center $H(p_R)$	$0.6365141683$	$0.9182958341$
Conditional or distillable	$0.4620981204$	$2/3$
EHS representation indices	$0.4620981204$	$2/3$
Sectorwise algebraic total	$1.0986122887=\log3$	$1.5849625007$
Extended-space total	$1.5607104090$	$2.2516291674$

The EHS total exceeds the sectorwise algebraic total by the representation-index term, while the distillable value equals only the conditional contribution. The exact operational result is Van Acoleyen et al. 2016, Eq. (7) and the direct/converse proof on pp. 3–4:

$$E_D^{\rm gauge}=\sum_Rp_RS(\rho_{A,R}).$$

Adversarial control: change the flux measure

Consider one regulated Gaussian flux mode with density relative to $dE$,

$$p(E)=\frac1{\sqrt{2\pi\sigma^2}} e^{-E^2/(2\sigma^2)}.$$

Its differential entropy is

$$h_E=-\int dE\,p(E)\log p(E) =\frac12\log(2\pi e\sigma^2).$$

Relabel the same mode by $E'=2E$. Then $p'(E')=p(E'/2)/2$, and

$$h_{E'}=h_E+\log2.$$

No physical state or correlation changed, yet the isolated continuous center entropy shifted by $0.6931471806$. A lattice-derived measure such as Donnelly and Wall’s Eq. (8), together with its determinant normalization and local counterterms, is therefore part of the prescription. Under a mere relabeling, the regulated discrete entropy and the anomaly-matched coefficient $-31/45$ do not change; an isolated differential term does.

Combining this measure adversary with the $Z_2$ center adversary identifies the strongest invariant claims:

• changing the center can change the total and classical terms even at a fixed finite regulator;
• changing a continuum flux coordinate shifts the isolated differential entropy;
• the conditional distillable term is fixed only after the operation class and algebra are fixed; and
• the Maxwell spherical logarithm is universal only for the complete matched bulk-plus-edge calculation.

Operational interpretation

A regional agent restricted to gauge-invariant operations cannot coherently mix center sectors. The sector label is classical data for that task, while the conditional quantum term can be distilled asymptotically subject to the locality assumptions of the protocol. Representation indices introduced solely to split a cut link are not independent laboratories.

Different boundary physics can change this conclusion. If charged boundary matter is dynamical, operations that were forbidden in the fixed-boundary theory may become available. Operational language must therefore specify the boundary system and allowed reference resources rather than attaching a universal meaning to the phrase “edge mode.”

Common pitfalls

Appending a non-Abelian $\log d_R$ term to Maxwell theory. Abelian Maxwell representations are one-dimensional. Its edge contribution comes from the regulated flux measure and determinant.

Interpreting a contact determinant in isolation. Gauge fixing splits the path integral into pieces that are not separately physical. Compare the complete matched bulk-plus-edge result.

Treating differential flux entropy as coordinate invariant. It shifts under a rescaling of the continuous label. State the measure and regulator.

Calling every edge term distillable. Center uncertainty and extension multiplicity are not Bell-pair yield under local gauge-invariant operations.

Exercises

1. Derive the discrete block entropy

Let $\rho=\bigoplus_Ep_E\rho_E$ with normalized $\rho_E$. Derive its entropy and identify the part that survives the gauge-invariant distillation theorem.

Solution

The eigenvalues of the $E$ block are $p_E\lambda_{E,k}$, where $\lambda_{E,k}$ are the eigenvalues of $\rho_E$. Therefore

$$\begin{aligned} S(\rho) &=-\sum_{E,k}p_E\lambda_{E,k} \log(p_E\lambda_{E,k})\\ &=-\sum_Ep_E\log p_E +\sum_Ep_ES(\rho_E)\\ &=H(p_E)+\sum_Ep_ES(\rho_E). \end{aligned}$$

For pure global lattice states under local gauge-invariant operations, the distillable term is the conditional average $\sum_Ep_ES(\rho_E)$. The center Shannon term is not part of the Bell-pair rate.

2. Close the Maxwell anomaly mismatch

The transverse calculation gives $-16/45$ and the edge determinant gives $-1/3$. Show that their sum equals the anomaly result and explain why a negative coefficient is not a negative entropy.

Solution

Since $-1/3=-15/45$,

$$-\frac{16}{45}-\frac13 =-\frac{16}{45}-\frac{15}{45} =-\frac{31}{45},$$

which equals the four-dimensional conformal-anomaly coefficient. This number multiplies the universal logarithmic term inside an entropy that also contains positive regulator-dependent area contributions. Neither the coefficient nor the gauge-fixed edge determinant is a standalone probability entropy.

3. Transform a flux density

For a normalized density $p(E)$, let $E'=aE$ with $a\ne0$. Derive the transformation of differential entropy and evaluate it for $a=2$.

Solution

Probability conservation gives

$$p'(E')=\frac1{|a|}p\!\left(\frac{E'}a\right).$$

Changing variables back to $E$,

$$\begin{aligned} h[p'] &=-\int dE'\,p'(E')\log p'(E')\\ &=-\int dE\,p(E)\log\frac{p(E)}{|a|}\\ &=h[p]+\log|a|. \end{aligned}$$

For $a=2$, the shift is $\log2=0.6931471806$. This exact coordinate dependence is why a continuum center entropy requires a reference measure and cannot be compared in isolation.

References

• Casini, Horacio, Marina Huerta, and José Alejandro Rosabal. “Remarks on Entanglement Entropy for Gauge Fields.” Physical Review D 89 (2014): 085012. DOI. Open PDF.
• Donnelly, William. “Decomposition of Entanglement Entropy in Lattice Gauge Theory.” Physical Review D 85 (2012): 085004. DOI. Open PDF.
• Donnelly, William, and Aron C. Wall. “Entanglement Entropy of Electromagnetic Edge Modes.” Physical Review Letters 114 (2015): 111603. DOI. Open PDF.
• Donnelly, William, and Aron C. Wall. “Geometric Entropy and Edge Modes of the Electromagnetic Field.” Physical Review D 94 (2016): 104053. DOI. Open PDF.
• Kabat, Daniel N. “Black Hole Entropy and Entropy of Entanglement.” Nuclear Physics B 453 (1995): 281–299. DOI. Open PDF.
• Van Acoleyen, Karel, Nick Bultinck, Jutho Haegeman, Michael Marien, Volkher B. Scholz, and Frank Verstraete. “The Entanglement of Distillation for Gauge Theories.” Physical Review Letters 117 (2016): 131602. DOI. Open PDF.