Nonrenormalization Theorems: Wilsonian, 1PI, and Infrared Scope

The perturbative nonrenormalization theorem is a statement about local F-terms in a Wilsonian action written in holomorphic variables. It does not say that the Kähler potential, physical Yukawa couplings, or a massless theory’s nonlocal 1PI action receive no corrections. Keeping the functional, cutoff, field normalization, and infrared domain explicit resolves most apparent exceptions.

Required background. Holomorphic couplings and background superfields supplies the source-holomorphy argument. Supergraphs and quantum effective actions supplies superspace propagators and $D$-algebra.

Helpful background. The 1PI effective action defines the Legendre transform and its vertex functions.

The Wilsonian theorem

Let $S_\mu$ be obtained by integrating out modes between a UV scale $M$ and a nonzero Wilsonian scale $\mu$. For external momenta much smaller than $\mu$, it has a local derivative expansion,

$$S_\mu=\int d^4x\,d^4\theta\;K_\mu +\left[\int d^4x\,d^2\theta\; \left(W_\mu(\Phi)+\frac14 f_{ab,\mu}(\Phi)W^{a\alpha}W^b_\alpha\right) +\text{h.c.}\right]+\cdots.$$

Assume four-dimensional $\mathcal N=1$ supersymmetry, a supersymmetric regulator, and perturbation theory about a nonsingular background. Then:

• the local Wilsonian superpotential $W_\mu$ receives no perturbative loop corrections in holomorphic field variables;
• the two-derivative holomorphic gauge coupling has one-loop-exact perturbative Wilsonian running when the regulator and coupling convention preserve holomorphy; and
• $D$-terms, including $K_\mu$, generally do renormalize.

The theorem concerns perturbative loops. Nonperturbative sectors can generate a superpotential when their zero modes and symmetries permit it. The classic superspace proof and its hypotheses are developed in Grisaru, Siegel, and Roček 1979, pp. 429–450; the distinction between perturbative and nonperturbative holomorphic terms is emphasized in Seiberg 1993, pp. 469–475.

Why supergraphs produce full superspace integrals

A chiral vertex carries $\int d^2\theta$ and an antichiral vertex carries $\int d^2\bar\theta$. Superspace propagators contain spinor derivatives and superspace delta functions. For a loop diagram, $D$-algebra uses these delta functions to identify the vertex coordinates and leaves an integral over all four Grassmann coordinates,

$$\Delta S_{\mathrm{loop}}\sim \int d^4x\,d^4\theta\;\mathcal K_{\mathrm{loop}}.$$

To rewrite a full-superspace term as a chiral one, one uses

$$\int d^4\theta\;U =\int d^2\theta\left(-\frac14\bar D^2U\right).$$

For a derivative-free local expression $U=H(\Phi)$ built only from chiral external fields, $\bar D_{\dot\alpha}\Phi=0$ makes the projection vanish. A nonzero chiral-looking loop contribution instead needs spinor derivatives combined with a nonlocal factor such as $1/\Box$; its coefficient can also contain antichiral couplings supplied by conjugate vertices. Such a factor is excluded from the local Wilsonian derivative expansion at fixed $\mu$, but it can occur after massless modes are integrated all the way to zero momentum. Chiral integrals containing superspace or spacetime derivatives are separate higher-derivative F-terms and require their own classification.

The proof is not simply “the $\theta$ integrals do not match.” The decisive combination is superspace $D$-algebra and locality. Nor does it constrain an arbitrary higher-derivative chiral integral; special F-terms with derivatives or additional fields require their own analysis.

A complementary component and superspace treatment, including the separation of Wilsonian superpotentials from wavefunction renormalization, is given in Weinberg 2000, §§ 27.6 and 30.3.

Wilsonian and 1PI domains

The two functionals answer different questions.

Property	Wilsonian action $S_\mu$	1PI action $\Gamma$
modes integrated out	momenta above $\mu>0$	all quantum modes
locality	derivative expansion for external momenta below $\mu$	may be nonlocal in a massless theory
perturbative superpotential	unchanged in local holomorphic variables	no unconditional transfer at massless exceptional kinematics
nonperturbative F-terms	allowed when zero modes and dynamics permit them	inherited in vertices, with additional infrared qualifications
$D$-terms and running	$K_\mu$ and wavefunction factors generally run	full momentum-dependent vertices generally run
infrared-sensitive chiral form	excluded from the local derivative expansion	possible through nonlocal factors such as $1/\Box$
natural use	RG flow and local operator coefficients	exact vertex functions and quantum field equations
field normalization	conveniently holomorphic	often expressed in canonical or renormalized fields

The decision map below puts the infrared test in the right place. Follow the Wilsonian branch only when a positive separation scale and a local derivative expansion have been specified; follow the 1PI branch through the mass-gap question before transferring any zero-momentum exactness claim.

Domain classifier for four-dimensional rigid $\mathcal N=1$ effective actions. At Wilsonian cutoff $\mu>0$ and external momentum $|p|\ll\mu$, shell contributions admit a local derivative expansion: the local superpotential has no perturbative loop correction in holomorphic variables, although Kähler data and canonical couplings run. A full 1PI action needs an infrared test. A complete mass gap permits a low-momentum local expansion, whereas massless propagation produces logarithms or inverse boxes. In massless cubic Wess–Zumino theory the one-loop correction is a nonlocal Kähler $D$-term; the first chiral 1PI term occurs at two loops and is not a local Wilsonian superpotential correction. The layout is schematic; the structured description gives the complete domain and status table.

In a massless theory, a 1PI contribution can schematically take the form

$$\int d^4\theta\; \frac{D^2}{\Box}\,H(\Phi)+\text{h.c.}$$

and, after superspace projection, resemble a chiral integral at exceptional momentum. It is still a nonlocal infrared effect, not a generated local Wilsonian superpotential. This distinction and its implications for vacuum analysis are made explicit in Poppitz and Randall 1996, pp. 281–284. At a generic background where all internal fields acquire a mass, the infrared singularity is cut off and the distinction often becomes less dramatic; that is a domain statement, not a new theorem.

The 1PI effective potential also depends on the Kähler metric. Even when the Wilsonian $W$ is exact, solving for canonically normalized masses or scattering amplitudes requires $K$ and wavefunction renormalization.

When protection claims apply places the perturbative Wilsonian theorem and the massless-1PI locality caveat beside branch-metric and nonperturbative claims without transferring assumptions between rows.

A massless chiral theory at one and two loops

Consider

$$W_{\mathrm{tree}}=\frac{y}{3}\Phi^3$$

with no mass. Classify possible one-loop terms:

1. A local correction $\delta y\,\Phi^3$ to the holomorphic Wilsonian superpotential is forbidden perturbatively.

2. A wavefunction term

   $$\int d^4\theta\;Z(\mu)\Phi^\dagger\Phi$$

   is allowed and is logarithmically renormalized.

3. After the canonical change of variables $\Phi_c=Z^{1/2}\Phi$, the physical Yukawa parameter is

   $$y_c(\mu)=\frac{y_h}{Z(\mu)^{3/2}}.$$

   Thus $y_c$ runs although the holomorphic coefficient $y_h$ does not.

4. A nonlocal 1PI structure containing $D^2/\Box$ is not ruled out at zero mass and exceptional external momentum. It must not be relabeled as a Wilsonian threshold correction.

At two loops there is a sharper test of an overstrong 1PI reading. In the massless theory, the full 1PI functional contains an infrared contribution whose zero-momentum projection has the chiral form

$$\Delta\Gamma_{\mathrm{1PI}}^{(2)} =c\int d^4x\,d^2\theta\; y^3(y^\dagger)^2\Phi^3+ \text{h.c.},$$

where $c\neq0$ depends on the normalization used for the cubic coupling. Before the exceptional zero-momentum limit, the same contribution is represented by a nonlocal full-superspace structure with inverse powers of $\Box$. Its dependence on $y^\dagger$ is a warning that it is not a holomorphic local Wilsonian coefficient. The first such chiral 1PI correction occurs at two loops, not at one loop; it is calculated in Buchbinder et al. 1995, § 4, Eqs. (34)–(43) and interpreted as an infrared holomorphic anomaly in Poppitz and Randall 1996, pp. 281–284.

This example separates four statements that are often conflated: no perturbative correction to the local Wilsonian $W_h$, one-loop wavefunction renormalization, a two-loop infrared chiral term in the massless 1PI functional, and nontrivial momentum dependence of 1PI vertices.

Assumptions that must accompany an exactness claim

Before using “nonrenormalized,” record:

• the functional ($S_\mu$ or $\Gamma$);
• the operator (superpotential, gauge kinetic term, higher F-term, or $D$-term);
• the perturbative or nonperturbative scope;
• the regulator and preserved supersymmetry;
• the holomorphic or canonical field coordinates;
• the infrared condition (mass gap, generic background, or exceptional momentum excluded); and
• the field-space patch and singular loci.

Changing one entry can change the conclusion. In particular, the Wilsonian theorem cannot be moved to the 1PI action merely by taking $\mu\to0$: the derivative expansion can fail precisely in that limit.

Common pitfalls

“The superpotential never changes.” Perturbative Wilsonian loops do not change it in holomorphic variables. Instantons, gaugino condensation, or other strong dynamics can generate nonperturbative terms.

“A running Yukawa violates nonrenormalization.” Canonical normalization introduces wavefunction factors. The holomorphic and canonically normalized Wilsonian couplings are different coordinates on coupling space; neither replaces the momentum-dependent 1PI data needed for an observable amplitude.

“Any chiral-looking 1PI term is a local F-term.” A factor $1/\Box$ records infrared propagation. Locality must be checked before applying a Wilsonian theorem.

Exercises

Suppose $W_h=m_h\Phi^2/2+y_h\Phi^3/3$ and the Wilsonian Kähler term is $Z\Phi^\dagger\Phi$. Express the canonical mass and Yukawa coupling in terms of $m_h$, $y_h$, and $Z$.

Solution

With $\Phi_c=Z^{1/2}\Phi$,

$$m_c=\frac{m_h}{Z}, \qquad y_c=\frac{y_h}{Z^{3/2}}.$$

Their scale dependence can therefore be entirely due to $Z(\mu)$ even when $m_h$ and $y_h$ are perturbatively unrenormalized.

Why does adding a nonzero mass improve the transfer of a zero-momentum statement from the Wilsonian action to the 1PI action?

Solution

A mass cuts off the infrared region that could generate $1/\Box$ or other nonanalytic momentum dependence. The 1PI functional then admits a local expansion at momenta well below the mass. This removes the specific IR loophole, although canonical normalization and nonperturbative effects still have to be treated separately.

Why does the factor $(y^\dagger)^2$ in the two-loop chiral-looking term not contradict holomorphy of the Wilsonian superpotential?

Solution

The term belongs to the massless 1PI functional at exceptional momentum. Before taking that limit it comes from a nonlocal full-superspace expression, so it is neither a local Wilsonian coefficient nor constrained to be holomorphic in the chiral source $y$. The finite-cutoff Wilsonian superpotential remains $W_h=y\Phi^3/3$ in perturbation theory.

References

• I. L. Buchbinder, S. M. Kuzenko, A. Yu. Petrov, and J. V. Yarevskaya, “Superfield Effective Potential,” arXiv:hep-th/9501047 (1995), § 4, arXiv.
• Marcus T. Grisaru, Warren Siegel, and Martin Roček, “Improved Methods for Supergraphs,” Nuclear Physics B 159 (1979), 429–450, DOI.
• Erich Poppitz and Lisa Randall, “Holomorphic Anomalies and the Nonrenormalization Theorem,” Physics Letters B 389 (1996), 280–286, arXiv, DOI.
• Nathan Seiberg, “Naturalness versus Supersymmetric Non-renormalization Theorems,” Physics Letters B 318 (1993), 469–475, arXiv, DOI.
• Steven Weinberg, The Quantum Theory of Fields, Volume III: Supersymmetry, Cambridge University Press (2000), §§ 27.6 and 30.3, DOI.