Dilute Instanton Ensembles and Theta Dependence

A dilute gas of charge $+1$ instantons and charge $-1$ anti-instantons turns the sector phase $e^{i\theta Q}$ into a periodic vacuum energy. The familiar cosine is not a universal strong-coupling theorem: it follows when individual event weights are finite and the ensemble is dilute enough that connected interactions can be neglected.

Required background. Instanton measures, zero modes, and determinants supplies the one-event weight. Theta parameters, theta states, and sector sums supplies the phase $e^{i\theta Q}$ and its periodicity.

Helpful background. Topological susceptibility and vacuum response defines the response observable extracted below.

Independent topological events

Let $\mathcal K$ be the fully integrated one-instanton density per Euclidean spacetime volume after every bosonic and fermionic zero mode relevant to the vacuum amplitude has been saturated. Assume:

• the instanton and anti-instanton weights are equal at $\theta=0$;
• their charges are $+1$ and $-1$;
• the mean separation is large compared with the event core;
• interactions and overlap corrections are negligible at the desired order.

In a Euclidean volume $\mathcal V$, a configuration with $n_+$ instantons and $n_-$ anti-instantons has $Q=n_+-n_-$. Independent center integrations and indistinguishability give

$$\frac{Z(\theta)}{Z_{\rm pert}} = \sum_{n_+,n_-=0}^{\infty} \frac{(\mathcal K\mathcal V)^{n_++n_-}}{n_+!\,n_-!} e^{i\theta(n_+-n_-)}.$$

The sums factorize:

$$\frac{Z(\theta)}{Z_{\rm pert}} = \exp\!\left(\mathcal K\mathcal V e^{i\theta}\right) \exp\!\left(\mathcal K\mathcal V e^{-i\theta}\right) = \exp\!\left(2\mathcal K\mathcal V\cos\theta\right).$$

The product in the middle line is meant literally; combining its exponents gives the final expression. Dividing by the $\theta=0$ partition function removes the common normalization:

$$\frac{Z(\theta)}{Z(0)} = \exp\!\left[2\mathcal K\mathcal V(\cos\theta-1)\right].$$

Thus the vacuum-energy density obeys

$$\mathcal E(\theta)-\mathcal E(0) = 2\mathcal K(1-\cos\theta),$$

and the dilute-gas susceptibility is

$$\chi = \left.\frac{\partial^2\mathcal E}{\partial\theta^2}\right|_{\theta=0} =2\mathcal K.$$

The derivation is exact under the independent-event assumptions. Whether those assumptions hold is a separate dynamical question. Mariño 2015, § 4.5, pp. 139–146 derives the one-instanton density and its leading charge-$\pm1$ theta contribution; the Poisson exponentiation above additionally assumes an ideal gas of those events.

Probability and cumulants

At $\theta=0$, $n_+$ and $n_-$ are independent Poisson variables with mean $\mathcal K\mathcal V$. Their difference $Q$ therefore follows a Skellam distribution,

$$P(Q) = e^{-2\mathcal K\mathcal V} I_{|Q|}(2\mathcal K\mathcal V),$$

where $I_n$ is a modified Bessel function. The cumulant-generating function is

$$\log\left\langle e^{i\theta Q}\right\rangle = 2\mathcal K\mathcal V(\cos\theta-1).$$

Consequently,

$$\frac{\langle Q^2\rangle_c}{\mathcal V}=2\mathcal K, \qquad \frac{\langle Q^4\rangle_c}{\mathcal V}=2\mathcal K,$$

while derivatives with respect to $i\theta$ alternate signs in the Taylor coefficients of $\mathcal E(\theta)$. Equality of all even charge cumulants per volume is a diagnostic of the ideal unit-charge gas, not a general property of gauge theory.

Where the one-event density comes from

In a weakly coupled regime, $\mathcal K$ is obtained by integrating the one-instanton measure over its non-position moduli. The construction is displayed in the moduli-to-measure chain. The underlying one-instanton weight and its fermion factors were derived in ‘t Hooft 1976, §§ III–VII, pp. 3436–3449. For four-dimensional Yang–Mills theory on $\mathbb R^4$, the size integration typically reaches $\rho\Lambda\sim1$, so $\mathcal K$ is not calculable by simply extending the one-loop BPST measure through the infrared.

Fermion zero modes can eliminate the vacuum contribution altogether. If at least one Dirac fermion is exactly massless and no external fermion fields are inserted, the one-event Grassmann integrals are unsaturated, so $\mathcal K_{\rm vac}=0$. For active flavors with $m_f\rho\ll1$, small masses supply $\prod_f(m_f\rho)$; heavier thresholds require matching rather than extrapolating this factor. This is consistent with removing $\theta$ by an anomalous axial change of variables when a quark is massless; it is not captured by inserting a mass-independent $\mathcal K$ into the cosine.

The comparison of canonical saddle calculations places the loop parameter, diluteness parameter, and evidence status side by side. Here the two essential controls are $g^2(1/\rho)\ll1$ over the contributing sizes and a small packing fraction.

To make the second test quantitative when the isolated one-event vacuum activity is nonzero, let $D(\rho)$ be the fully saturated size density for one sign of the charge after the remaining non-position moduli have been integrated, so $\mathcal K=\int d\rho\,D(\rho)$. Approximating an event core by a four-ball of radius $\rho$ gives

$$\phi_{\rm pack} \sim 2v_4\int d\rho\,D(\rho)\rho^4, \qquad v_4=\frac{\pi^2}{2}.$$

The factor two counts instantons and anti-instantons at $\theta=0$. For a narrow distribution, $\phi_{\rm pack}\sim2\mathcal K v_4\bar\rho^4\ll1$. This geometric estimate is necessary but not sufficient: a long-range interaction can make the cluster integral large even when cores rarely overlap.

Interactions and higher harmonics

When events interact, the partition function is organized by connected clusters rather than independent Poisson factors. Let $\xi=(\rho,\Omega,\ldots)$ collect all non-position collective data, including size and gauge orientation. At $\theta=0$, write the one-event intensive measure as $w_a(\xi_a)d\xi_a$ and the full two-event measure at relative separation $R$ as $w^{(2)}_{ab}(R;\xi_a,\xi_b)d\xi_a d\xi_b$. Subtracting the disconnected limit gives

$$\frac{\delta\log Z}{\mathcal V} = \frac12\sum_{a,b} e^{i\theta(q_a+q_b)} \int d^4R\,d\xi_a\,d\xi_b\, \left[ w^{(2)}_{ab}(R;\xi_a,\xi_b) -w_a(\xi_a)w_b(\xi_b) \right],$$

where $q_a,q_b\in\{+1,-1\}$. In a bosonic theory, or when fermion saturation factorizes event by event, one may write $w^{(2)}_{ab}=w_aw_b e^{-U_{ab}}$ and recover the Mayer factor $e^{-U_{ab}}-1$. The dimensionless interaction $U_{ab}(R;\xi_a,\xi_b)$ must retain its dependence on relative gauge orientation and the other collective coordinates; integrations over the instanton–anti-instanton separation, sizes, and relative orientation are explicit in Khoze, Krauss, and Schott 2020, § 2.2.

Massless fermions require the more general connected-measure formula above. Even when each isolated vacuum activity vanishes, an instanton–anti-instanton pair can acquire a nonzero fermion determinant through overlap of their would-be zero modes. That pair contribution must therefore be computed from the full Dirac determinant or its quasi-zero-mode overlap matrix, not from a product of two separately saturated $D(\rho)$ factors; instanton–anti-instanton molecular weights provide a concrete realization in Schäfer, Shuryak, and Verbaarschot 1995, pp. 1267–1273.

The connected integral must be finite. When a nonzero one-event vacuum activity $\mathcal K$ exists, its magnitude must be parametrically smaller than the ideal contribution $2\mathcal K$; when $\mathcal K=0$, the correct comparison is instead between successive nonzero connected clusters. A neutral instanton–anti-instanton pair changes the $\theta$-independent normalization, while a connected cluster of net charge $k$ contributes a harmonic proportional to $\cos(k\theta)$ when CP is unbroken at $\theta=0$:

$$\mathcal E(\theta)-\mathcal E(0) = \sum_{k\ge1}c_k\,[1-\cos(k\theta)].$$

The ideal gas retains only $c_1=2\mathcal K$. Higher harmonics can arise from correlated clusters or elementary fractional events under different boundary conditions. Periodicity alone does not determine the coefficients, their signs, or the convergence of this expansion. The finite-temperature instanton gas and its breakdown regimes are reviewed from the semiclassical measure in Gross, Pisarski, and Yaffe 1981, pp. 43–80.

The order of limits matters. The thermodynamic limit $\mathcal V\to\infty$ is taken after constructing the intensive cluster expansion. At finite volume, $Z(\theta)$ is analytic. Nonanalytic branch changes can emerge only in an infinite-volume limit and cannot be inferred from the elementary cosine without additional dynamics.

Common pitfalls

Calling the cosine universal. It is the leading result of an ideal unit-charge gas. Strongly coupled Yang–Mills theory need not have a calculable BPST fugacity or a single cosine vacuum energy.

Ignoring fermion saturation. The bosonic action does not determine whether an instanton contributes to $Z$. Unsaturated Grassmann zero modes make the vacuum contribution vanish.

Using event number as the dilution test. The mean number grows with volume. Diluteness is a local packing condition comparing event size and mean separation.

Exercises

1. Derive the ideal-gas susceptibility directly from charge fluctuations at $\theta=0$.

Solution

For independent Poisson variables, $\operatorname{Var}(n_\pm)=\mathcal K\mathcal V$. Since $Q=n_+-n_-$,

$$\langle Q^2\rangle_c =\operatorname{Var}(n_+)+\operatorname{Var}(n_-) =2\mathcal K\mathcal V.$$

Dividing by $\mathcal V$ gives $\chi=2\mathcal K$, in agreement with two derivatives of the vacuum energy.

2. Suppose elementary events have charges $\pm1/N$, equal fugacity $\zeta$, and are independently summed within a compactified effective theory. What local harmonic do they generate, and why is this not by itself a $2\pi N$-periodic microscopic theory?

Solution

The independent sum gives $2\zeta[1-\cos(\theta/N)]$ on one branch. A microscopic $2\pi$-periodic theory must also include the branch relabeling or global constraint that restores $2\pi$ periodicity. The local fractional-event expression is valid only with its compactification and branch data specified.

3. A narrow size distribution has $\bar\rho=0.2\,\ell$ and total event density $n=2\mathcal K=0.1\,\ell^{-4}$. Estimate the four-dimensional packing fraction using spherical cores. Does this estimate alone prove that the gas is ideal?

Solution

The estimate is

$$\phi_{\rm pack} \sim n\,v_4\bar\rho^4 = 0.1\,\frac{\pi^2}{2}(0.2)^4 \simeq7.9\times10^{-4}.$$

Core overlap is therefore rare in this model. That does not prove ideality: the Mayer integral can still be important if interactions are long-ranged, singular, or enhanced by quasi-zero modes.

References

• Gross, David J., Robert D. Pisarski, and Laurence G. Yaffe. “QCD and Instantons at Finite Temperature.” Reviews of Modern Physics 53 (1981): 43–80. 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.
• Khoze, Valentin V., Frank Krauss, and Matthias Schott. “Large Effects from Small QCD Instantons: Making Soft Bombs at Hadron Colliders.” Journal of High Energy Physics 2020, no. 4 (2020): 201. DOI.
• Mariño, Marcos. Instantons and Large N: An Introduction to Non-Perturbative Methods in Quantum Field Theory. Cambridge: Cambridge University Press, 2015. DOI.
• Schäfer, Thomas, Edward V. Shuryak, and Jacobus J. M. Verbaarschot. “The Chiral Phase Transition and Instanton–Anti-Instanton Molecules.” Physical Review D 51 (1995): 1267–1281. DOI.