Scalar Counterterms and the Renormalized On-Shell Action

Local covariant counterterms cancel every power and logarithmic divergence of a free scalar’s regulated AdS action. The remaining functional depends on the normalizable response, while finite local counterterms change contact terms and a Legendre or multi-trace boundary term can change the quantization itself. These three operations—removing divergences, choosing a scheme, and choosing a boundary theory—must not be conflated.

Required background. Fefferman–Graham expansions provide the source and response coefficients. Local counterterms and subdivergences supply the locality principle. Helpful background. Renormalization schemes and finite parts and regulator removal distinguish finite choices from the existence of the limit.

First application. Regulate a free massive scalar at radial coordinate epsilon, cancel each power and logarithmic divergence, and obtain a finite two-point generating functional.

Regulated scalar action

Use Euclidean Poincaré AdS$_{d+1}$ with boundary at $z=0$, region $z\ge\epsilon$, and action

$$S[\phi]=\frac{\mathcal N}{2} \int\mathrm d^{d+1}x\sqrt G \left[(\nabla\phi)^2+m^2\phi^2\right].$$

The normalization $\mathcal N$ is retained because it fixes the boundary two-point coefficient. With outward normal $n=-(z/L)\partial_z$, the on-shell action is

$$S_{\mathrm{reg}}[\phi_\epsilon] =\frac{\mathcal N}{2} \int_{z=\epsilon}\mathrm d^d x\sqrt\gamma\, \phi_\epsilon n^M\partial_M\phi.$$

Substituting

$$\phi=z^{d-\Delta} \bigl(\phi_{(0)}+z^2\phi_{(2)}+\cdots\bigr) +z^\Delta\bigl(\phi_{(2\Delta-d)}+\cdots\bigr)$$

produces powers of $\epsilon$ whose coefficients are local in $\phi_{(0)}$. Covariance rewrites them in terms of the induced field and metric at the cutoff.

Counterterm recursion and logarithms

For a nonresonant scalar, the first terms have the schematic covariant form

$$S_{\mathrm{ct}} =\frac{\mathcal N}{2} \int_{z=\epsilon}\mathrm d^d x\sqrt\gamma \left[ \frac{d-\Delta}{L}\phi^2 +\frac{L}{2\Delta-d-2}\,\phi\Box_\gamma\phi +\cdots \right],$$

with coefficients translated consistently if the normal or Laplacian convention changes. The derivative term can be integrated by parts on a boundary without boundary; otherwise the resulting edge term must be kept.

When $\nu=\Delta-d/2$ is an integer, one denominator vanishes. The corresponding counterterm becomes logarithmic,

$$S_{\log} =-\frac12\log(\epsilon\mu) \int\mathrm d^d x\sqrt\gamma\,\mathcal A[\phi,\gamma],$$

where the local density $\mathcal A$ is fixed by the resonant recursion. Changing the renormalization scale gives

$$\mu\frac{\mathrm d}{\mathrm d\mu}S_{\mathrm{ren}} =-\frac12\int\mathrm d^d x\sqrt{g_{(0)}}\,\mathcal A,$$

up to the displayed normalization convention. The logarithm is therefore physical scale dependence or an anomaly; it is not removed by pretending the resonance is absent Skenderis 2002, §§3–4.

Finite generating functional and two-point data

Define

$$S_{\mathrm{ren}} =\lim_{\epsilon\to0} \left(S_{\mathrm{reg}}+S_{\mathrm{ct}}\right).$$

Its first variation at fixed standard-quantization source is

$$\delta S_{\mathrm{ren}} =\mathcal N\int\mathrm d^d x\sqrt{g_{(0)}} \left[(2\Delta-d)\phi_{(2\Delta-d)} +\mathcal F_{\mathrm{local}}\right]\delta\phi_{(0)}.$$

For a regular Euclidean solution in momentum space, the ratio of response to source is proportional to $p^{2\nu}$ for noninteger $\nu$ and to $p^{2\nu}\log(p^2/\mu^2)$ at integer $\nu$. Local polynomials in $p^2$ can be shifted by finite counterterms; the nonanalytic term and its normalization are invariant separated-point data.

This gives a useful independent check: differentiate the renormalized action twice and compare the nonlocal kernel with the boundary limit of the regular bulk-to-boundary propagator. Agreement tests the factors $\mathcal N$, $2\Delta-d$, the Fourier convention, and the selected branch simultaneously.

Alternate quantization and multi-trace data

Let $\nu=\sqrt{d^2/4+m^2L^2}$. In the window $0<\nu<1$, both radial modes can satisfy the appropriate normalizability condition. Standard quantization assigns dimension $\Delta_+=d/2+\nu$; alternate quantization assigns $\Delta_-=d/2-\nu$. Passing between them is a Legendre transform of the renormalized functional, not a finite scheme change Klebanov and Witten 1999, §2.

A mixed condition such as

$$\phi_{(2\Delta-d)}=f\,\phi_{(0)}$$

implements multi-trace data after normalization and local terms are fixed. Varying $f$ changes the boundary condition and generally the spectrum. By contrast, adding a finite local term $c\int\sqrt{g_{(0)}}\,\phi_{(0)}\Box\phi_{(0)}$ shifts only contact data at this order.

Worked regulator-removal check

For a massless scalar in AdS$_5$, $d=\Delta=4$ and

$$\phi=\phi_{(0)}+\frac{z^2}{4}\Box\phi_{(0)} +z^4\left(\phi_{(4)}+\log z^2\,\psi_{(4)}\right)+\cdots.$$

The $z^4$ resonance produces the logarithmic momentum-space kernel $p^4\log(p^2/\mu^2)$. A subtraction that leaves a residual $\log\epsilon$ fails the regulator-removal test. A subtraction that removes the $p^4\log p^2$ term as well has altered nonlocal physics and is not a local renormalization scheme.

The strongest result is a finite generating functional with a declared branch and scheme. It is not a proof that the chosen boundary condition belongs to a complete holographic theory.

Common pitfalls

Setting the cutoff field equal to the source. They differ by a power of $\epsilon$ and derivative corrections. Translate before taking the limit.

Calling every finite term a scheme. A Legendre transform or mixed boundary condition changes the variational problem and can change the spectrum.

Dropping logarithms. Resonant logarithms encode scale dependence. Omitting them violates the radial equation and the anomaly identity.

The chapter overview contains the structure diagram and validity and failure diagram. They are embedded there once so that their shared chapter-level context is not repeated on every article.

For the chapter-wide comparison of assumptions, counterevidence, falsifiers, and claim ceilings, see the claim-domain table.

References

• Bianchi, M., Freedman, D. Z., and Skenderis, K. “Holographic Renormalization.” Nuclear Physics B 631 (2002): 159–194. DOI. arXiv.
• de Haro, S., Solodukhin, S. N., and Skenderis, K. “Holographic Reconstruction of Spacetime and Renormalization in the AdS/CFT Correspondence.” Communications in Mathematical Physics 217 (2001): 595–622. DOI. arXiv.
• Klebanov, I. R., and Witten, E. “AdS/CFT Correspondence and Symmetry Breaking.” Nuclear Physics B 556 (1999): 89–114. DOI. arXiv.
• Skenderis, K. “Lecture Notes on Holographic Renormalization.” Classical and Quantum Gravity 19 (2002): 5849–5876. DOI. arXiv.