Consistency of Perturbative Yang–Mills Theory

Perturbative Yang–Mills consistency has two linked parts. An anomaly-free Slavnov–Taylor identity constrains ultraviolet counterterms and ties gauge, ghost, and matter vertices to one gauge coupling. Positive physical scattering additionally requires a suitable BRST state space and scattering operator: the physical quotient must have a positive inner product, and scattering must preserve that product. This page explains the counterterm argument and proves the conditional passage from BRST-invariant scattering to physical unitarity. The setting is order-by-order perturbation theory around a specified vacuum; it does not construct the nonperturbative theory.

Required background. Gauge-fixed Yang–Mills action and ghost sector supplies the nilpotent BRST differential and auxiliary field. Symmetry constraints and the space of counterterms supplies locality, power counting, and symmetry restoration by counterterms.

Helpful background. BRST cohomology and physical observables develops the general cohomological meaning of physical operators and states.

The Slavnov functional identity

For clarity, first display the pure gauge, ghost, and auxiliary sector. Let $\Sigma$ be the gauge-fixed classical action with sources $K^{a\mu}$ and $L^a$ coupled to the nonlinear BRST variations:

$$\Sigma =S_{\rm inv}+s\Psi +\int\mathrm d^4x\left[ K^{a\mu}sA_\mu^a+L^asc^a \right].$$

The sources are needed because the composite operators $sA$ and $sc$ also renormalize. Matter fields require corresponding sources for $s\psi$ and $s\bar\psi$ and add their paired derivative terms to the identity. With a consistent choice of left and right Grassmann derivatives, the displayed sector of the renormalized 1PI functional obeys

$$\mathcal S(\Gamma) =\int\mathrm d^4x\left[ \frac{\delta\Gamma}{\delta K^{a\mu}} \frac{\delta\Gamma}{\delta A_\mu^a} +\frac{\delta\Gamma}{\delta L^a} \frac{\delta\Gamma}{\delta c^a} +B^a\frac{\delta\Gamma}{\delta\bar c^a} \right]=0.$$

Differentiating this functional equation with respect to fields produces the Slavnov–Taylor identities among propagators and vertices. They replace the simple Abelian Ward identity because a gauge transformation of $A_\mu$ is nonlinear and the Faddeev–Popov operator depends on $A_\mu$. The original generalized identities were obtained independently in Slavnov 1972, pp. 99–104 and Taylor 1971, pp. 436–444; their BRST organization was developed in Becchi, Rouet, and Stora 1976, pp. 287–321.

At tree level $\Gamma=\Sigma$ and $\mathcal S(\Sigma)=0$ is equivalent to BRST invariance. Linearizing around $\Sigma$ defines an operator $\mathcal B_\Sigma$ with $\mathcal B_\Sigma^2=0$. This nilpotency is the algebraic device that classifies both admissible counterterms and possible symmetry breakings.

Why the counterterm space closes

The all-orders argument is inductive:

1. Suppose the theory has been renormalized through order $\hbar^{n-1}$ and the Slavnov identity holds to that order.
2. Locality of ultraviolet subtractions implies that the order-$n$ divergent part is an integrated local polynomial. In four-dimensional power-counting-renormalizable Yang–Mills theory it has ghost number zero and dimension at most four.
3. The order-$n$ Slavnov identity requires the divergence to be closed under $\mathcal B_\Sigma$.
4. Ghost-number-zero cohomology consists of gauge-invariant parameter deformations, while $\mathcal B_\Sigma$-exact terms are field, source, and gauge-fixing redefinitions. Subtracting those terms restores the identity at order $n$.

For pure Yang–Mills theory the CP-even invariant deformation is the $F^2$ action; $F\widetilde F$ is the topological CP-odd possibility. With matter, gauge-invariant masses and couplings allowed by the declared model join the list. Arbitrary mass terms for an unbroken gauge field, unmatched three- and four-gauge couplings, or a ghost mass in linear covariant gauge are not allowed counterterms.

The historical diagrammatic proof that massless Yang–Mills fields are perturbatively renormalizable and unitary under its stated assumptions is ’t Hooft 1971, pp. 173–199. The modern local-cohomology classification, including hypotheses, counterterms, and anomaly classes, is reviewed in Barnich, Brandt, and Henneaux 2000, §§ 2.3 and 12.2–12.3.

The anomaly condition

A regulator or subtraction can break the Slavnov identity at loop order $n$:

$$\mathcal S(\Gamma)=\hbar^n\Delta+O(\hbar^{n+1}).$$

Consistency implies

$$\mathcal B_\Sigma\Delta=0.$$

If $\Delta=\mathcal B_\Sigma\widehat\Delta$, a local counterterm $-\widehat\Delta$ removes the breaking. If instead $\Delta$ represents a nontrivial ghost-number-one cohomology class, it is a gauge anomaly and the Slavnov identity cannot be restored within the model. Pure Yang–Mills theory has no chiral matter measure and no perturbative gauge anomaly; vectorlike matter is also safe in this respect. A chiral gauge theory requires its representation-dependent anomaly coefficients to cancel. Power counting alone never proves that cancellation.

A single renormalized coupling

Write

$$A_0=Z_A^{1/2}A, \qquad c_0=Z_c^{1/2}c, \qquad g_0=\mu^\epsilon Z_g g.$$

If $Z_{3A}$, $Z_{4A}$, and $Z_{\bar c cA}$ denote the complete multiplicative factors of the three-gauge, four-gauge, and ghost–gauge vertices, the Slavnov identity requires

$$Z_{3A}=Z_gZ_A^{3/2}, \qquad Z_{4A}=Z_g^2Z_A^2, \qquad Z_{\bar c cA}=Z_gZ_cZ_A^{1/2}.$$

Equivalently,

$$Z_g =\frac{Z_{3A}}{Z_A^{3/2}} =\frac{Z_{4A}^{1/2}}{Z_A} =\frac{Z_{\bar c cA}}{Z_cZ_A^{1/2}}.$$

Matter vertices supply the same $Z_g$ after their wave-function factors are removed. These are not accidental equalities among a few one-loop graphs: they are consequences of the functional identity and must persist order by order. The background-field relation $Z_gZ_{\bar A}^{1/2}=1$ is a particularly economical realization, developed on Background-field Yang–Mills effective action.

As a practical Slavnov check, compute $Z_g$ from two independent vertices. A discrepancy after including all diagrams and counterterms signals a missing ghost graph, an inconsistent color normalization, or a symmetry-breaking subtraction that has not been repaired. The equalities expressed through the bare coupling appear in Srednicki 2006, author manuscript, § 73, p. 427, Eq. (73.2), PDF. His $Z_3,Z'_2,Z_{3g},Z_{4g},Z'_1$ correspond respectively to $Z_A,Z_c,Z_{3A},Z_{4A},Z_{\bar c cA}$ here. Their BRST origin is explained in Srednicki 2006, § 74, pp. 438–439, author-manuscript PDF; the full counterterm argument uses the local-cohomology hypotheses stated above.

Physical unitarity on cohomology

Covariant gauge quantization uses an indefinite-metric state space. The ordinary cutting relation for a gauge-fixed Green function therefore includes timelike and longitudinal gauge modes and ghost cuts; positivity is not expected term by term in that enlarged space. BRST supplies the physical quotient.

Let $Q$ be a conserved BRST charge on a common ghost-graded domain $\mathcal D$ in the auxiliary state space. Assume the quantum identities and asymptotic construction provide a scattering operator $S$, with $S$ and $S^{-1}$ preserving that domain and ghost number, such that

$$Q^2=0, \qquad [Q,S]=0.$$

These relations first give an algebraic statement. If $Q\psi=0$, then $QS\psi=SQ\psi=0$; if $\psi=Q\chi$, then $S\psi=QS\chi$. Thus scattering preserves both closed and exact states and descends to their quotient. No norm or probability has entered that argument.

Write the auxiliary indefinite inner product as $[\psi,\phi]_K$, conjugate-linear in its first entry, and define the adjoint $Q^\times$ by $[\psi,Q\chi]_K=[Q^\times\psi,\chi]_K$. Assume $Q^\times=Q$ on the compatible domain. For closed $\psi$,

$$\begin{aligned} [\psi,Q\chi]_K&=[Q\psi,\chi]_K=0,\\ [Q\chi,Q\eta]_K&=[\chi,Q^2\eta]_K=0. \end{aligned}$$

Exact states are therefore null and orthogonal to every closed state, so changing a representative by an exact state does not change the induced inner product. The adjoint is the indefinite one: a nilpotent self-adjoint operator on a positive Hilbert space would instead obey $\lVert Q\chi\rVert^2=\langle\chi,Q^2\chi\rangle=0$ and hence vanish.

At ghost number zero, let $\mathcal C=\ker Q\cap\mathcal D^0$ and $\mathcal E=Q\mathcal D^{-1}$. To obtain a physical Hilbert space, assume that the form on $\mathcal C$ is nonnegative and its radical—the vectors orthogonal to all of $\mathcal C$—is precisely $\mathcal E$. If the exact-state range is not closed, a topology and quotient-by-closure prescription must first be specified, and both $S$ and $S^{-1}$ must preserve that chosen closed null subspace. With this prescription and Hilbert completion understood, the usual shorthand is

$$\mathcal H_{\rm phys} =\ker Q/\operatorname{im}Q.$$

Positivity is a separate input. In the free or perturbative asymptotic Yang–Mills construction, the quartet mechanism pairs unphysical gauge and ghost sectors; identifying the remaining classes with the positive transverse-polarization sector requires the corresponding completeness and state-space assumptions. It is not a consequence of anomaly cancellation alone. The canonical BRST state-space argument explains the distinction between exact states, the radical and positivity.

Finally assume scattering preserves the auxiliary form, $S^\times S=SS^\times=\mathbf1$, on the same compatible scattering domain. For closed $\psi,\phi$,

$$[S\psi,S\phi]_K =[\psi,S^\times S\phi]_K =[\psi,\phi]_K.$$

The induced operator on physical classes is an isometry. Because $S^{-1}$ also preserves closed and exact states, that isometry is onto and has an isometric inverse. Both extend to the Hilbert completion, so physical scattering is unitary. This proof separates three requirements: BRST commutation gives quotient descent, the state-space construction gives positivity, and preservation of the form together with the inverse gives unitarity. Existence of the required asymptotic scattering problem and control of its infrared limits remain assumptions.

Gauge-parameter independence uses a related BRST argument, not just the quotient-isometry proof. Since changing $\xi$ changes the gauge fermion,

$$\frac{\partial S_{\rm gf+gh}}{\partial\xi} =s\left(\frac{\partial\Psi}{\partial\xi}\right),$$

its insertion between BRST-closed physical states vanishes, provided the measure and renormalization preserve BRST and no boundary or infrared term invalidates the argument. This does not say that off-shell propagators or the effective action are independent of $\xi$.

The perturbative ceiling

The result established here is perturbative renormalizability plus unitarity of the physical, anomaly-free BRST cohomology under the assumptions needed to define the scattering problem. It does not supply:

• a nonperturbative continuum construction of four-dimensional Yang–Mills theory;
• a proof of a mass gap, confinement, or the infrared particle spectrum;
• a global gauge slice free of Gribov copies;
• an infrared-finite colored-particle $S$-matrix in a confining theory; or
• anomaly cancellation for an undeclared chiral matter representation.

Even perturbatively, massless amplitudes may require inclusive, dressed, or factorized observables before an infrared-finite prediction exists. For QCD, external colored partons are useful short-distance ingredients, not nonperturbative asymptotic states. Renormalizability controls ultraviolet counterterms; it does not settle those infrared and global questions.

Independent checks and limits

1. Slavnov check: extract $Z_g$ independently from the three-gauge and ghost–gauge vertices after dividing by wave-function factors. They must agree.
2. Cohomology check: replacing a physical external polarization by its momentum must give a BRST-trivial contribution to the complete amplitude, not necessarily to each diagram.
3. Physical norm check: identify the adjoint and domain, test positivity of the physical quotient, and verify that scattering and its inverse preserve the form. Commutation with $Q$ alone is insufficient.
4. Gauge-parameter check: a complete physical result has $\partial_\xi\mathcal M=0$; an off-shell Green function need not.
5. Anomaly check: the possible breaking has ghost number one. It must be shown cohomologically trivial or canceled by the matter content before the all-orders induction can proceed.

Common pitfalls

Equating power counting with renormalizability. Dimensionless $g$ makes only finitely many local operators possible; the Slavnov identity is what prevents unrelated gauge, ghost, and vertex counterterms.

Demanding positive cuts in the enlarged state space. Covariant gauges are indefinite. Positivity is a statement about the BRST quotient after unphysical and ghost sectors cancel.

Using perturbative BRST as a nonperturbative existence proof. The argument is an order-by-order formal construction around a chosen vacuum and gauge patch. Its validity ceiling must travel with the conclusion.

Exercise: descent is not unitarity

Consider two finite-dimensional algebraic examples, with all vectors at ghost number zero and $Q=0$.

1. On $\mathbb C$ with its positive inner product, take $S=2\mathbf1$. Does scattering descend to cohomology, and is the induced operator unitary?
2. On $\mathbb C^2$ with $[v,w]_K=v_1^*w_1-v_2^*w_2$, take $S=\mathbf1$. Which physical-state assumption fails, even though $S$ preserves the form?

Solution

Since $\ker Q$ is the whole space and $\operatorname{im}Q=0$, cohomology equals the original space in both examples. Both operators commute with $Q$ and have inverses.

In the first example, the induced map multiplies every norm squared by four: $\lVert S\psi\rVert^2=4\lVert\psi\rVert^2$. It descends, but $S^\times S=4\mathbf1$, so it is not unitary. BRST commutation does not imply preservation of probability.

In the second example, $S^\times S=\mathbf1$, but $[(0,1),(0,1)]_K=-1$. The physical candidate quotient is still indefinite and cannot be the positive Hilbert space required for probabilities. Form preservation does not establish positivity.

These are logical counterexamples to incomplete implications, not models of Yang–Mills scattering. The gauge theory must supply both missing ingredients through its state-space and scattering construction.

References

• Glenn Barnich, Friedemann Brandt, and Marc Henneaux, “Local BRST Cohomology in Gauge Theories,” Physics Reports 338 (2000), 439–569, DOI, Open PDF.
• C. Becchi, A. Rouet, and R. Stora, “Renormalization of Gauge Theories,” Annals of Physics 98 (1976), 287–321, DOI.
• Gerard ’t Hooft, “Renormalization of Massless Yang–Mills Fields,” Nuclear Physics B 33 (1971), 173–199, DOI, Open PDF.
• A. A. Slavnov, “Ward Identities in Gauge Theories,” Theoretical and Mathematical Physics 10 (1972), 99–104, DOI.
• Srednicki, Mark. Quantum Field Theory. Author’s prepublication manuscript, copyright 2006. Open PDF. Author’s draft and published-edition errata. The cited printed labels refer to this manuscript, not the 2007 book.
• J. C. Taylor, “Ward Identities and Charge Renormalization of the Yang–Mills Field,” Nuclear Physics B 33 (1971), 436–444, DOI.

Further reading

• Kugo, Taichiro, and Izumi Ojima. “Local Covariant Operator Formalism of Non-Abelian Gauge Theories and Quark Confinement Problem.” Progress of Theoretical Physics Supplement 66 (1979): 1–130. DOI.