Hamiltonian Constraints and Radial Canonical Transformations

Holographic counterterms act as canonical transformations on radial phase space. They shift canonical momenta by local functions of the induced fields while preserving the symplectic form and transforming every Hamiltonian, momentum, and gauge constraint consistently. This viewpoint separates a scheme change, which is a local canonical reparametrization, from a new boundary condition, which selects a different Lagrangian submanifold and may define a different theory.

Required background. Radial Hamilton–Jacobi flow supplies the constrained phase space. Helpful background. Presymplectic systems explains boundary ambiguities, and local field redefinitions gives the QFT comparison.

First application. Treat a scalar counterterm as a generating functional and compute the induced shift of radial momentum while preserving the symplectic form.

Counterterms as generating functionals

Let $(q^A,\pi_A)$ denote induced fields and radial momenta on a cutoff surface. Adding a local counterterm functional $F[q]$ changes the principal function to $S'=S+F$. Hence

$$q'^A=q^A, \qquad \pi'_A=\pi_A+\frac{\delta F}{\delta q^A}.$$

The radial symplectic form is preserved:

$$\Omega'=\int\mathrm d^d x\, \delta\pi'_A\wedge\delta q^A =\Omega+\int\mathrm d^d x\, \delta^2F\wedge\delta q=\Omega,$$

because the Hessian of $F$ is symmetric while the field-space wedge product is antisymmetric. Boundaries of the cutoff surface require the edge contribution to be retained before this conclusion is used.

For gravity, $q^A$ includes $\gamma_{ij}$ and the shift must be applied to the complete tensor momentum. A curvature counterterm produces derivatives of $\delta\gamma_{ij}$; integration by parts and any corner terms are part of the transformation.

Constraints in renormalized variables

Suppose the original constraints are $\mathcal H_\perp(q,\pi)=0$, $\mathcal H_i(q,\pi)=0$, and $\mathcal G(q,\pi)=0$. The transformed expressions are

$$\mathcal H'_\alpha(q,\pi') =\mathcal H_\alpha\!\left(q, \pi'-\frac{\delta F}{\delta q}\right)=0.$$

Thus counterterms do not erase the constraints. Their divergent parts cancel divergent momenta; their finite parts move local contacts among renormalized responses. The transformed momentum and gauge constraints become the Ward identities in the new scheme. Failing to transform a constraint when shifting a momentum is the canonical version of mixing two renormalization conventions.

An anomaly appears when no covariant local $F$ removes a logarithmic scale dependence while preserving all desired symmetries. The finite canonical variables still exist, but their scale transformation contains the anomaly.

Scalar momentum-shift calculation

Take

$$F[\phi,\gamma] =\frac12\int\mathrm d^d x\sqrt\gamma \left(c_0\phi^2+c_2\phi\Box_\gamma\phi\right).$$

Ignoring an actual boundary of the cutoff surface, variation gives

$$\pi'_\phi =\pi_\phi+\sqrt\gamma \left(c_0\phi+c_2\Box_\gamma\phi\right).$$

The leading asymptotic pieces of $\pi_\phi$ are precisely of this local form, so a divergent choice of $c_0,c_2,\ldots$ defines a finite renormalized momentum. A finite change $c_2\mapsto c_2+\delta c_2$ shifts the one-point function by a local derivative of the source and the two-point function by a polynomial in momentum. It cannot alter the nonanalytic separated-point kernel.

This calculation supplies an invariant check: compute the Klein–Gordon symplectic flux before and after the shift. It agrees because $F$ changes the symplectic potential by an exact field-space variation.

Scheme change versus boundary condition

A local canonical transformation retains the same independent coordinate $q$ at the boundary. A Legendre transform instead exchanges coordinate and momentum,

$$\widetilde S[\pi]=S[q]-\int\pi_Aq^A,$$

and changes which datum is fixed. A mixed condition $\pi_A=\delta W[q]/\delta q^A$ selects a different boundary submanifold. These may implement alternate quantization, multi-trace deformation, or an ensemble change. They should not be reported as mere contact-term schemes.

The adversarial test is simple: add $F$ and verify that the spectrum and nonlocal kernel remain unchanged; then impose a mixed condition and observe that the allowed modes or poles change. If both operations give the same reported conclusion, the calculation has probably failed to distinguish scheme from theory data.

Practical checklist

• fix normal orientation and the radial symplectic potential;
• include lapse, shift, gauge multipliers, and their constraints;
• generate the momentum shift from one covariant local functional;
• transform constraints and Ward identities together;
• retain corner and edge terms when the cutoff surface has a boundary;
• test a nonlocal correlator, charge, or symplectic flux for invariance;
• label Legendre and mixed transformations as changes of boundary data.

Canonical language organizes renormalization and scheme dependence. It does not imply that radial evolution is a unitary time evolution or a literal Wilsonian integration over momentum shells.

The statement that holographic counterterms act as a radial canonical transformation is established in Papadimitriou 2010; it does not identify every finite canonical transformation with Wilsonian coarse graining.

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

• Papadimitriou, I. “Holographic Renormalization as a Canonical Transformation.” Journal of High Energy Physics 2010, 014 (2010). DOI. arXiv.
• Papadimitriou, I., and Skenderis, K. “AdS/CFT Correspondence and Geometry.” In IRMA Lectures in Mathematics and Theoretical Physics 8, 73–101 (2005). DOI. arXiv.
• Witten, E. “Multi-Trace Operators, Boundary Conditions, and AdS/CFT Correspondence.” arXiv:hep-th/0112258 (2001). arXiv.