Four-Dimensional a-Type Flow Constraints

For a unitary local relativistic RG flow between four-dimensional CFTs, the Euler-anomaly coefficient satisfies $a_{\rm UV}-a_{\rm IR}\ge0$. Spherical entanglement identifies $a$ at each endpoint, while dilaton scattering and a null-cone entropy construction prove the ordering under different explicit hypotheses. Neither proof turns an arbitrary finite term in a massive sphere entropy into a universal $a(R)$.

Required background. Anomaly coefficients and central charges fixes the $a$ and $c$ convention; monotonicity and flow constraints states the endpoint theorem; information measures along RG flows fixes the regional comparison and its limits.

Helpful background. Universal terms and geometry explains how smooth spherical logarithms isolate anomaly data.

Use the chapter’s scale-separation map, comparison table, and independent validity gates to distinguish an endpoint anomaly, a finite crossover prescription, and a proof-specific positive quantity.

The spherical logarithm selects the Euler anomaly

At a four-dimensional CFT, use the convention

$$\langle T^\mu{}_{\mu}\rangle =\frac{1}{16\pi^2} \left[ c\,W_{\mu\nu\rho\sigma}W^{\mu\nu\rho\sigma} -a\,E_4 +b\,\Box\mathcal R \right],$$

where

$$E_4 =R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma} -4R_{\mu\nu}R^{\mu\nu} +\mathcal R^2.$$

The coefficient $b$ is shifted by a local $\mathcal R^2$ counterterm, whereas $a$ and $c$ are fixed-point data. Away from a fixed point the trace also contains beta-function terms such as $\beta^i\mathcal O_i$, so the displayed equation should not be read as the full trace identity along the flow.

For a round spherical entangling surface of radius $R$ in the flat-space CFT vacuum,

$$S_{\rm sphere}(R) =\alpha_2\frac{R^2}{\epsilon^2} -4a\log\frac{R}{\epsilon} +S_{\rm finite}.$$

The sphere is conformally flat, so its universal logarithm selects the Euler coefficient rather than $c$. The conformal map and anomaly calculation are given in Casini, Huerta, and Myers 2011, § 4.3, eqs. (4.12)–(4.16); the separation of intrinsic and extrinsic terms for a general smooth surface appears in Solodukhin 2008, pp. 306–308, eqs. (4)–(6).

In this convention the standard free-field checks are:

Four-dimensional CFT field	$a$	Coefficient of $\log(R/\epsilon)$ on a sphere
One real conformal scalar	$1/360$	$-1/90$
One Weyl fermion	$11/720$	$-11/180$
One gauge vector	$31/180$	$-31/45$

For gauge fields the entanglement calculation must include a prescription that reproduces the anomaly, including the appropriate contact or edge contribution. The table fixes endpoint normalization; it does not assert that all three fields admit the same isolated mass deformation.

Dilaton scattering makes the endpoint difference positive

Komargodski and Schwimmer introduce a compensating dilaton with decay constant $f$. Anomaly matching fixes its four-derivative Wess–Zumino interaction. Their paper uses the positive coefficient

$$\widehat a=\frac{a}{16\pi^2},$$

where $a$ on the right is the dimensionless coefficient used on this page. In forward kinematics their low-energy amplitude is

$$\mathcal A(s,0) =\frac{2(\widehat a_{\rm UV}-\widehat a_{\rm IR})}{f^4}s^2 +O(s^4) =\frac{a_{\rm UV}-a_{\rm IR}}{8\pi^2f^4}s^2 +O(s^4).$$

Analyticity, crossing, unitarity, and the optical theorem give

$$\widehat a_{\rm UV}-\widehat a_{\rm IR} =\frac{f^4}{\pi} \int_0^\infty\frac{ds}{s^2}\,\sigma_{\tau\tau}(s) \ge0,$$

or equivalently

$$a_{\rm UV}-a_{\rm IR} =16\pi f^4 \int_0^\infty\frac{ds}{s^2}\,\sigma_{\tau\tau}(s) \ge0.$$

Here $\sigma_{\tau\tau}$ is the total dilaton–dilaton cross section in the amplitude convention above. The amplitude and sum rule are Komargodski and Schwimmer 2011, § 4, eqs. (4.6)–(4.8). Strict inequality follows when the absorptive cross section is nonzero on a set of nonzero measure; the positivity argument itself first establishes the non-strict inequality. The proof also needs the usual locality and high-energy behavior required for the dispersion relation and a limit in which the spectator dilaton does not alter the matter RG flow.

The positive object in this proof is a scattering spectral integral. It is not an ordinary spatial partial trace. One may define a proof-specific spectral interpolant by changing the lower integration limit, but that construction is not automatically equal to a finite sphere-entropy subtraction.

The null-cone entropy route uses a Markov reference

Let $S_1(r)$ be the sphere entropy in the flowing vacuum and $S_0(r)$ the entropy in the UV CFT vacuum. For regions whose boundaries lie on one null cone, the UV CFT vacuum is Markov: it saturates the strong-subadditivity combination. Consequently

$$\Delta S(X)=S_1(X)-S_0(X)$$

inherits the useful inequality without the UV-local terms that would otherwise survive on wiggly unions and intersections. The flowing vacuum itself is not assumed to be Markov.

In four dimensions the resulting sphere inequality can be written

$$rS_1''(r)-S_1'(r) \le \frac{8a_{\rm UV}}{r}.$$

At an infrared CFT,

$$S_1(r)=\mu_2^{\rm IR}r^2 -4a_{\rm IR}\log(r/\epsilon)+\cdots,$$

so the left-hand side tends to $8a_{\rm IR}/r$ and the inequality gives $a_{\rm IR}\le a_{\rm UV}$. The Markov equality, four-dimensional sphere inequality, and endpoint evaluation are in Casini, Testé, and Torroba 2017, § III, eq. (8), and § IV, eqs. (17)–(19).

This proof requires the null construction. Concentric equal-time spheres alone leave curvature-dependent terms uncontrolled. It also proves an endpoint ordering, not a unique positive locally monotone $a(R)$ for every massive theory.

A massive real scalar tracks the normalization and threshold

Use one conformally coupled real scalar at the ultraviolet fixed point and perturb it by $m^2\phi^2/2$. Its endpoint data are

$$a_{\rm UV}=\frac1{360}, \qquad S_{\rm UV}^{\rm univ} =-\frac1{90}\log\frac{R}{\epsilon}, \qquad a_{\rm IR}=0$$

for the trivial gapped infrared theory. Komargodski and Schwimmer’s explicit compensated scalar calculation gives

$$\widehat a_{\rm UV}-\widehat a_{\rm IR} =\frac{1}{5760\pi^2} =\frac{1}{16\pi^2}\frac1{360},$$

and the dispersive cut begins at the two-particle threshold $s=4m^2$ Komargodski and Schwimmer 2011, appendix B, eqs. (B.1)–(B.13). Thus the integral accumulates precisely $\Delta a=1/360$ as its lower limit passes the massive spectral support.

Benchmark item	Quantitative outcome	Domain and error control
UV sphere logarithm	$-4a_{\rm UV}=-1/90$	$mR\ll1$ after separating the area term; exact endpoint coefficient
IR endpoint	$a_{\rm IR}=0$	$mR\gg1$ and a trivial gapped endpoint
Dilaton spectral support	$s\ge4m^2$	Two-scalar threshold
Integrated anomaly difference	$\Delta a=1/360$	Exact at leading spectator order; KS convention gives $1/(5760\pi^2)$
Effective-theory controls	$p^2\ll\Lambda^2$ and $f\gg m$	Corrections organized in $p^2/\Lambda^2$ and $m/f$

The exact anomaly values have no statistical uncertainty. A numerical sphere-entropy extraction has cutoff, fit-window, and finite-$mR$ errors, while the compensated scattering calculation has effective-theory truncation errors controlled by the two ratios in the table. A single Weyl fermion has no Lorentz-invariant mass by itself, and a gauge vector mass requires a Higgs or Stückelberg completion with extra fields; neither should be substituted silently for this scalar flow.

Local terms and altered endpoints test the boundary of the claim

Consider the finite surface counterterm

$$\delta S_\lambda =\lambda\int_{S^2_R}\!d^2y\,\sqrt h\, K_{ab}K^{ab}.$$

For a round sphere, $K_{ab}K^{ab}=2/R^2$ and the area is $4\pi R^2$, so

$$\delta S_\lambda=8\pi\lambda.$$

It shifts a finite constant but cannot change the coefficient of $\log(R/\epsilon)$. The adversarial result is therefore exact: the endpoint identification $S_{\rm univ}=-4a\log(R/\epsilon)$ survives, while any finite constant proposed as an $a(R)$ fails without a matching convention.

Adversarial change	What fails	Strongest surviving statement
Add $\lambda\int K_{ab}K^{ab}$	A finite sphere constant shifts by $8\pi\lambda$	The smooth CFT logarithmic coefficient still equals $-4a$
Introduce a physical boundary	Boundary anomaly coefficients can also enter logarithms	Bulk $a$ can be isolated only after specifying geometry and boundary terms
End at a nonconformal theory	There is no CFT coefficient $a_{\rm IR}$ to insert	A regulated finite-scale inequality may remain, but not the two-CFT endpoint theorem
Replace $a$ by $c$	The sphere Euler term is misidentified	No general ordering theorem for $c$ follows

Common pitfalls

Mixing anomaly normalizations. The dimensionless $a$ used in the sphere formula differs by $16\pi^2$ from the coefficient $\widehat a$ in the Komargodski–Schwimmer amplitude convention. Translate before comparing a loop calculation with $1/360$.

Massing every free-field entry in the same way. The scalar benchmark is a valid isolated relevant deformation. Fermion and vector masses require additional field content or a different theory.

Promoting a finite subtraction to the theorem. The protected information is the endpoint anomaly and the proof-specific positive integral or null-cone inequality. A finite sphere curve needs its own definition and scheme analysis.

Exercises

1. Compute the spherical logarithmic coefficient for each free field in the table.

Solution

The coefficient is $-4a$. Therefore

$$-4a_{\rm scalar}=-\frac{4}{360}=-\frac1{90},$$ $$-4a_{\rm Weyl}=-\frac{44}{720}=-\frac{11}{180}, \qquad -4a_{\rm vector}=-\frac{124}{180}=-\frac{31}{45}.$$

These are endpoint anomaly coefficients; reproducing the vector result in an entropy calculation requires the complete gauge-field prescription.

2. Translate the dilaton sum rule from $\widehat a$ to the page convention and state when it is strict.

Solution

Since $a=16\pi^2\widehat a$,

$$\Delta a =16\pi^2\frac{f^4}{\pi} \int_0^\infty\frac{ds}{s^2}\sigma_{\tau\tau}(s) =16\pi f^4 \int_0^\infty\frac{ds}{s^2}\sigma_{\tau\tau}(s).$$

Unitarity makes the total cross section nonnegative, so $\Delta a\ge0$. The inequality is strict when the cross section is nonzero on a set of nonzero measure and the dispersion relation has the stated convergence properties.

3. Evaluate the finite extrinsic-curvature counterterm on a round sphere and decide which $a$-identification it can change.

Solution

For $S^2_R$ embedded in a spatial slice, $K_{ab}=h_{ab}/R$, hence $K_{ab}K^{ab}=2/R^2$. Thus

$$\lambda\int_{S^2_R}\sqrt h\,K_{ab}K^{ab} =\lambda(4\pi R^2)\frac{2}{R^2} =8\pi\lambda.$$

This is independent of $R$. It changes a finite constant but not the logarithmic coefficient, so it cannot alter the endpoint $a$ extracted from $-4a\log(R/\epsilon)$. It does invalidate any claim that an unmatched finite constant is a universal running $a$.

References

• Casini, Horacio, Marina Huerta, and Robert C. Myers. “Towards a Derivation of Holographic Entanglement Entropy.” Journal of High Energy Physics 2011, no. 5 (2011): 036. DOI. Open PDF.
• Casini, Horacio, Eduardo Testé, and Gonzalo Torroba. “Markov Property of the Conformal Field Theory Vacuum and the a Theorem.” Physical Review Letters 118 (2017): 261602. DOI. Open PDF.
• Komargodski, Zohar, and Adam Schwimmer. “On Renormalization Group Flows in Four Dimensions.” Journal of High Energy Physics 2011, no. 12 (2011): 099. DOI. Open PDF.
• Solodukhin, Sergey N. “Entanglement Entropy, Conformal Invariance and Extrinsic Geometry.” Physics Letters B 665 (2008): 305–309. DOI. Open PDF.