Instanton Measures, Zero Modes, and Determinants

A classical instanton is not yet a contribution to an observable. One must identify inequivalent collective coordinates, remove gauge redundancy, replace zero eigenvalues by collective-coordinate Jacobians, evaluate the remaining determinants, and combine the ultraviolet scale dependence with the running coupling. For a charge-one $SU(N)$ instanton, this procedure explains both the factor $g^{-4N}$ and the scale-invariant geometric factor $d^4x_0\,d\rho/\rho^5$.

Required background. Gauge instantons, topological charge, and moduli fixes the BPST normalization and the $4N$ bosonic moduli. Zero modes, collective coordinates, and moduli measures supplies the general change of variables.

Helpful background. The Faddeev–Popov construction supplies the gauge quotient and ghost determinant. Running couplings and dimensional transmutation supplies the renormalization-group interface.

From a saddle to a one-instanton measure

Let $\mathcal A_\mu(x;a)$ be a family of charge-one solutions labeled by moduli $a^\alpha$. A tangent must be projected into background gauge:

$$Z_{\mu,\alpha} = \frac{\partial\mathcal A_\mu}{\partial a^\alpha} +D_\mu\Omega_\alpha, \qquad D_\mu Z_{\mu,\alpha}=0.$$

The moduli-space metric is the zero-mode norm

$$G_{\alpha\beta} = \frac{1}{g^2}\int d^4x\, Z^a_{\mu,\alpha}Z^a_{\mu,\beta} = \frac{2}{g^2}\int d^4x\, \operatorname{tr}\!\left( Z_{\mu,\alpha}Z_{\mu,\beta} \right).$$

Changing from normalized Gaussian amplitudes to $a^\alpha$ produces

$$d\mu_{\rm coll} = \frac{\sqrt{\det G}}{(2\pi)^{n_B/2}} \prod_{\alpha=1}^{n_B}da^\alpha.$$

For a charge-one $SU(N)$ instanton, $n_B=4N$. Because every entry of $G$ carries $g^{-2}$, its square root contributes

$$\sqrt{\det G}\propto g^{-n_B}=g^{-4N} \propto \left(\frac{8\pi^2}{g^2}\right)^{2N},$$

where the omitted numerical factor is absorbed into the normalization constant. This is the origin of the coupling power; it is not an additional loop effect.

Gauge transformations that approach the identity at infinity are not moduli. Background gauge removes them, and the Faddeev–Popov operator is given boundary conditions that exclude constant transformations. Constants at infinity instead rotate the $SU(2)$ embedding: their orbit is divided by $H_2=\mathbb Z_2$ for $SU(2)$ and by the stabilizer $H_N$ defined on the gauge-instanton page for $N\ge3$. Thus “gauge orientation” and “gauge redundancy” are processed at different stages.

The full sequence is summarized below. Follow the lower checks as carefully as the central chain: a correct classical solution can still produce an uncontrolled size integral.

Construction of a one-instanton contribution (schematic). Gauge directions are removed before the physical modes are classified. Bosonic zero modes supply the collective-coordinate Jacobian and orientation measure, fermion zero modes must be saturated, negative modes require a contour, and the remaining determinants are normalized to the vacuum. A correct saddle exponent is not enough if the size integral or event ensemble is uncontrolled.

Determinants and the scale factor

After gauge fixing, a dimensionless one-loop factor must be normalized to the vacuum background. With the framed boundary condition just stated, its schematic form is

$$\frac{\det \Delta_{\rm gh}^{(I)}} {\det \Delta_{\rm gh}^{(0)}} \left[ \frac{\det \Delta_{\rm vec}^{(0)}} {\det{}'\Delta_{\rm vec}^{(I)}} \right]^{1/2} \prod_{\text{Dirac fields}} \frac{\det{}'\mathcal D_I}{\det\mathcal D_0}.$$

The prime omits genuine physical zero modes, not gauge transformations already excluded by the ghost boundary condition. Bosonic zero modes have become collective-coordinate integrals; fermionic zero modes become Grassmann integrals and must be saturated by insertions or masses. Negative modes, when present, require a contour prescription rather than omission. A self-dual BPST instanton has no physical negative mode, in contrast with a false-vacuum bounce; see the instanton–bounce boundary and mode comparison. Other gauge-group conventions are equivalent only after their residual group volumes, ghost zero modes, and orientation volumes are transformed together.

In pure $SU(N)$ Yang–Mills theory, the one-loop one-instanton density may be written

$$dn_I = C_N\, d^4x_0\,\frac{d\rho}{\rho^5} \left[\frac{8\pi^2}{g^2(\mu)}\right]^{2N} \exp\!\left[-\frac{8\pi^2}{g^2(\mu)}\right] (\mu\rho)^{b_0}, \qquad b_0=\frac{11N}{3}.$$

The constant $C_N$ depends on the renormalization scheme and on the normalization of the orientation volume. The powers have separate origins:

• $d^4x_0\,d\rho/\rho^5$ is scale invariant: under $x_0,\rho\mapsto\lambda x_0,\lambda\rho$, the five differentials supply $\lambda^5$, canceled by $\rho^{-5}$. Orientation coordinates are dimensionless.
• $[8\pi^2/g^2]^{2N}$ is the $g^{-4N}$ zero-mode Jacobian just derived.
• $e^{-8\pi^2/g^2}$ is the classical BPST weight.
• $(\mu\rho)^{b_0}$ is the net scale dependence of the regulated nonzero-mode and ghost determinants.

The last two factors form a one-loop renormalization-group invariant combination. Since

$$\mu\frac{dg}{d\mu} =-\frac{b_0}{16\pi^2}g^3+O(g^5),$$

one finds

$$\frac{d}{d\log\mu} \left[ -\frac{8\pi^2}{g^2(\mu)} +b_0\log(\mu\rho) \right] =O(g^2).$$

Thus, to one-loop accuracy,

$$e^{-8\pi^2/g^2(\mu)}(\mu\rho)^{b_0} =e^{-8\pi^2/g^2(1/\rho)}.$$

This standard measure was obtained by evaluating the determinants in the instanton background; see ‘t Hooft 1976, §§ III–V, pp. 3436–3448 and the corrected normalization analysis in Bernard 1979, pp. 3013–3019. Mariño 2015, § 4.5, pp. 129–146 derives the same coupling and size powers and makes the infrared limitation explicit.

Matter zero modes and observable dependence

With $N_f$ active Dirac fermions in the fundamental representation,

$$b_0=\frac{11N}{3}-\frac{2N_f}{3}.$$

For $Q=+1$, each Dirac flavor supplies the chiral zero modes required by the index. In a vacuum amplitude the Grassmann integral vanishes in the massless limit. For a small nonzero mass satisfying $m_f\rho\ll1$, one mass insertion saturates each flavor pair and supplies the dimensionless factor

$$\prod_{f=1}^{N_f}(m_f\rho).$$

External fermion fields can saturate the same modes instead, producing a correlation-function selection rule. The detailed chirality and flavor structure is derived on Fermion Zero Modes, Index Data, and Selection Rules.

More generally, let $n_m$ denote the number of active fundamental flavors saturated by small masses, and let an observable insertion contribute a further power $\rho^{p_{\mathcal O}}$ after positions and tensor structures are accounted for. The size integral therefore contains the local power

$$\int d\rho\,\rho^\alpha, \qquad \alpha=b_0-5+n_m+p_{\mathcal O}$$

at one loop, before logarithmic corrections. A flavor with $m_f\rho\gtrsim1$ is not described by the small-mass factor and requires threshold matching instead. The ultraviolet endpoint converges only if $\alpha>-1$, while the formal infrared endpoint converges only if $\alpha<-1$. These mathematical tests do not extend weak coupling: the semiclassical expression must already be abandoned when $\rho\Lambda\sim1$. The Instanton Size Modulus and Infrared Limitations performs this check for explicit matter content.

Normalization checks

A usable measure should pass four independent tests:

1. Dimensions: $d^4x_0\,d\rho/\rho^5$, $(\mu\rho)^{b_0}$, and $m_f\rho$ are dimensionless.
2. Renormalization scale: explicit $\mu$-dependence cancels the running of the classical weight to the stated loop order.
3. Zero-mode accounting: every omitted eigenvalue reappears as a bosonic collective coordinate or a Grassmann integral.
4. Domain of integration: orientation stabilizers are quotiented once, and the $\rho$ range stays inside $\rho\Lambda\ll1$ if the result is called semiclassical.

Failure of any one test invalidates the claimed prefactor even if the exponent is correct.

Common pitfalls

Writing $\det{}'$ without saying what was removed. Translation, scale, orientation, fermion, gauge, and negative modes have different treatments. A prime is meaningful only together with that classification.

Treating $C_N$ as universal. The complete density is scheme dependent; physical observables become scheme independent only after all ingredients are combined consistently.

Integrating through strong coupling. A formally divergent large-$\rho$ integral signals loss of control. Cutting it off by hand is a model assumption, not a first-principles instanton prediction.

Exercises

1. Show directly that the combination of the classical weight and $(\mu\rho)^{b_0}$ is $\mu$-independent at one loop.

Solution

The beta function gives $d(1/g^2)/d\log\mu=b_0/(8\pi^2)+O(g^2)$. Therefore

$$\frac{d}{d\log\mu} \left(-\frac{8\pi^2}{g^2}\right) =-b_0+O(g^2),$$

which cancels $d[b_0\log(\mu\rho)]/d\log\mu=b_0$.

2. For pure $SU(2)$, determine the one-loop small- and large-$\rho$ behavior of the vacuum size integral.

Solution

Here $b_0=22/3$, so the explicit power is

$$\rho^{b_0-5}=\rho^{7/3}.$$

Because $7/3>-1$, the integral converges at $\rho=0$. It grows toward large $\rho$, where $g(1/\rho)$ becomes strong. The one-loop semiclassical formula therefore loses validity before its formal upper endpoint can be interpreted.

References

• Bernard, Claude. “Gauge Zero Modes, Instanton Determinants, and Quantum-Chromodynamic Calculations.” Physical Review D 19 (1979): 3013–3019. DOI.
• ‘t Hooft, Gerard. “Computation of the Quantum Effects Due to a Four-Dimensional Pseudoparticle.” Physical Review D 14 (1976): 3432–3450; erratum 18 (1978): 2199. DOI.
• Mariño, Marcos. Instantons and Large N: An Introduction to Non-Perturbative Methods in Quantum Field Theory. Cambridge: Cambridge University Press, 2015. DOI.