Boundary and Defect Weyl Anomalies

A boundary or conformal defect adds local geometric data to the Weyl anomaly. Intrinsic curvature, extrinsic curvature, the pullback of the ambient Weyl tensor, and the normal-bundle connection can all appear, subject to dimension, codimension, parity, and Wess–Zumino consistency. Their coefficients characterize the boundary or defect, but relations to displacement or stress-tensor observables are specific to the geometry and normalization in which they are proved.

Required background. Anomaly Coefficients and Central Charges supplies the bulk classification. Displacement Operators and Ward Identities fixes the broken-translation operator. Helpful background. Boundaries, Flux, and Boundary Ward Identities supplies the distributional boundary terms.

Local geometry at the support

Let a $p$-dimensional conformal defect $\Sigma$ sit in a $d$-dimensional CFT, with codimension $q=d-p$. Split ambient coordinates into tangential indices $a,b$ and normal indices $i,j$. The geometric building blocks include

• the induced metric $\gamma_{ab}$ and its intrinsic curvature $\widehat R_{abcd}$;
• the second fundamental form $K^i_{ab}$ and its traceless part $\mathring K^i_{ab}$;
• pullbacks of the ambient Weyl tensor;
• the curvature of the normal-bundle connection when $q>1$.

The trace Ward identity has distributional form

$$\langle T^\mu{}_{\mu}(x)\rangle =\mathcal A_{\rm bulk}(x) +\delta_\Sigma(x)\,\mathcal A_\Sigma(x),$$

where $\delta_\Sigma$ is normalized by

$$\int_M d^dx\sqrt g\,\delta_\Sigma f =\int_\Sigma d^py\sqrt\gamma\,f|_\Sigma.$$

This definition prevents codimension-dependent Jacobians from being hidden in the anomaly coefficients.

Only defect densities of Weyl weight $p$ can contribute without dimensionful couplings. Hence local parity-even defect Weyl anomalies occur naturally when the defect dimension $p$ is even. Odd-dimensional supports can instead have parity-odd terms, global anomalies, or universal finite observables; they should not be forced into an even-dimensional Euler classification.

A two-dimensional support

For a two-dimensional defect, a schematic parity-even basis is

$$\mathcal A_\Sigma =\frac{1}{4\pi}\left[ b_E\widehat R +b_K\,\mathring K^i_{ab}\mathring K_i^{ab} +b_W\,\gamma^{ac}\gamma^{bd}W_{abcd} +\cdots\right].$$

The ellipsis allows codimension-specific contractions and normal-bundle terms. The intrinsic Euler coefficient $b_E$ is type A on the defect. The extrinsic and pullback-Weyl terms are type B. Exact signs and numerical factors differ across the literature, so a quoted “defect central charge” must include this basis equation.

For a boundary, the bulk Euler density also requires a boundary completion so that its integral gives the Euler characteristic. In a four-dimensional BCFT, additional cubic extrinsic-curvature and ambient-Weyl contractions can appear on the three-dimensional boundary. These are not obtained by simply restricting the bulk anomaly to $\partial M$.

Displacement data

Broken normal translations define the displacement operator through

$$\nabla_\mu T^{\mu i}(x) =\delta_\Sigma(x)D^i(y)+\text{geometric contact terms}.$$

Conformal symmetry fixes its dimension to $\Delta_D=p+1$ and its flat-defect two-point function to

$$\langle D^i(y)D^j(0)\rangle =\frac{C_D\,\delta^{ij}}{|y|^{2p+2}}.$$

Reflection positivity gives $C_D\ge0$ in a unitary defect sector. Shape variations of $W$ relate $C_D$ to coefficients of particular extrinsic-curvature anomaly terms in cases where the tensor basis and contact terms have been fully controlled. There is no dimension-independent equation “defect anomaly equals $C_D$.” The proportionality constant depends on $p$, $q$, the delta-function convention, and the normalization of $D^i$ Billò et al. 2016, §§5–6.

A reliable classification procedure

For a proposed anomaly term:

1. enumerate all scalar densities of the required Weyl weight, including parity-odd terms if allowed;
2. impose tangential diffeomorphisms, normal-frame covariance, and any orientation symmetry;
3. compute the Weyl variation and impose Wess–Zumino consistency;
4. quotient by Weyl variations of finite local bulk, boundary, and defect counterterms;
5. choose a basis and record integration-by-parts relations;
6. match remaining coefficients to independently normalized observables.

The higher-codimension four-dimensional-defect classification illustrates why every step is necessary: many candidate intrinsic, extrinsic, ambient, and normal-bundle scalars collapse or mix after consistency and counterterms are imposed Chalabi et al. 2022, §§2–4.

Datum	Universal when…	Required check
Defect Euler coefficient	its density is normalized and boundary completions are included	topological integral on a closed support
Extrinsic type-B coefficient	the invariant is Weyl covariant and not counterterm-exact	shape variation and basis independence
$C_D$	the displacement Ward identity and two-point normalization are fixed	flat-support reflection positivity
Stress-tensor one-point coefficient	ambient $T$ convention and defect normalization are fixed	Ward identities and tensor structure
Parity-odd coefficient	orientation and regulator preserve the stated symmetries	allowed finite Chern–Simons-like shifts

For RG flows localized on a two-dimensional defect, the defect Euler coefficient obeys an endpoint inequality under the unitarity and locality hypotheses of Jensen and O’Bannon 2016. That theorem is dimension-specific; it does not establish a codimension-independent monotone for every defect coefficient in the table.

Common pitfalls

Using intrinsic curvature alone. Extrinsic curvature, ambient Weyl curvature, and the normal bundle supply independent invariants. Their availability changes with codimension.

Importing a $C_D$ relation from another dimension. Shape-response coefficients and displacement normalization are convention- and dimension-dependent. Restate the Ward identity before using the relation.

Ignoring boundary completion of the Euler term. On a manifold with boundary, the bulk Euler integral alone is not topological. Missing surface terms corrupt any extraction of the type-A coefficient.

Exercises

Why must $\Delta_D=p+1$ for a $p$-dimensional defect?

Solution

The divergence $\nabla_\mu T^{\mu i}$ has dimension $d+1$. The defect delta function has dimension $q=d-p$. Matching dimensions in $\nabla_\mu T^{\mu i}=\delta_\Sigma D^i$ gives $\Delta_D=d+1-q=p+1$.

References

• Billò, M., Gonçalves, V., Lauria, E., and Meineri, M. “Defects in Conformal Field Theory.” Journal of High Energy Physics 2016, 091 (2016). arXiv. DOI.
• Chalabi, A., Herzog, C. P., O’Bannon, A., Robinson, B., and Sisti, J. “Weyl Anomalies of Four Dimensional Conformal Boundaries and Defects.” Journal of High Energy Physics 2022, 166 (2022). arXiv. DOI.
• Jensen, K., and O’Bannon, A. “Constraint on Defect and Boundary Renormalization Group Flows.” Physical Review Letters 116, 091601 (2016). arXiv. DOI.