Theta Terms and Topological Effects in Sigma Models

A theta term weights topological sectors by a phase. This exact bookkeeping fixes periodicity and special symmetry points, but it does not select a unique infrared phase. In two-dimensional sigma models, a dilute instanton gas, a large-$N$ branch expansion, an anomaly, and an exact result at a special value of $N$ are different forms of information with different regimes of validity.

Required background. The CP(N−1) model fixes the flux normalization, while theta dependence, CP, and branches supplies the vacuum-sector framework. Helpful background. Dilute instanton ensembles and theta dependence states the separation and fugacity conditions needed for a semiclassical gas.

Sector sum and exact periodicity

For a model whose smooth finite-action configurations on a closed oriented Euclidean spacetime have

$$Q =\frac{1}{2\pi}\int F \in\mathbb Z,$$

write

$$Z(\theta) =\sum_{Q\in\mathbb Z}e^{-i\theta Q}Z_Q .$$

The sign in the phase follows the Euclidean convention $S_E=S_{E,0}+i\theta Q$; reversing the orientation or the definition of $Q$ reverses that sign without changing the physics. Integer charge gives

$$Z(\theta+2\pi)=Z(\theta).$$

This is exact under the stated global conditions. On a manifold with boundary, in a background $PSU(N)$ bundle, or with twisted compactification data, the bulk integral can be fractional. The complete system—including boundary terms or background-field counterterms—must still transform consistently. One should not infer the periodicity of the complete partition function from a fractional saddle in isolation.

The vacuum-energy density and topological susceptibility are

$$\mathcal E(\theta) =-\lim_{V\to\infty}\frac{1}{V}\ln Z(\theta), \qquad \chi_t =\left. \frac{\partial^2\mathcal E}{\partial\theta^2} \right|_{\theta=0} =\lim_{V\to\infty}\frac{\langle Q^2\rangle_c}{V},$$

where the last equality uses the same Euclidean sign convention. Contact terms and the definition of the renormalized topological density must be fixed before comparing $\chi_t$ between regulators.

Integer charge and theta periodicity do not guarantee that $\chi_t$ has a finite continuum limit. In the exceptional $\mathrm{CP}^1\simeq O(3)$ model, the conventional topological susceptibility has a logarithmic ultraviolet divergence associated with the small-size instanton or dislocation tail. Periodicity and anomaly constraints remain meaningful, but a numerical susceptibility requires a declared regulator, contact-term prescription, and continuum observable Bonanno et al. 2023, §§ II–IV.

What a dilute gas would predict

Suppose unit-charge instantons and anti-instantons have fugacity $\kappa$ per unit volume, are individually semiclassical, and their separations are much larger than their cores. Summing independent events gives

$$Z(\theta) \simeq \exp\!\left[ 2\kappa V\cos\theta \right],$$

and hence

$$\mathcal E(\theta)-\mathcal E(0) \simeq 2\kappa(1-\cos\theta), \qquad \chi_t\simeq2\kappa.$$

The cosine is not a universal theta potential. On $\mathbb R^2$, sigma-model instantons have size moduli, and the size measure can fail at either end. Large lumps can overlap and destroy dilution; in $\mathrm{CP}^1\simeq O(3)$, the small-size tail obstructs a finite continuum topological susceptibility. A controlled compactification with specified twists can instead produce fractional events and a multi-branch potential; its result belongs to that compactified regime until adiabatic continuity has been independently justified.

Large-N branches

For $\mathrm{CP}^{N-1}$ at large $N$, the natural scaling is

$$\mathcal E(\theta) =N\min_{k\in\mathbb Z} f\!\left(\frac{\theta+2\pi k}{N}\right).$$

Expanding a single branch near its minimum gives

$$\mathcal E_k(\theta) -\mathcal E(0) =\frac{\chi_t}{2} (\theta+2\pi k)^2 +O\!\left(\frac{(\theta+2\pi k)^4}{N^3}\right),$$

with the $N$ scaling of $\chi_t$ fixed by the chosen action normalization. The envelope over $k$ restores exact $2\pi$ periodicity even though one analytic branch is not periodic by itself. Neighboring branches cross at $\theta=\pi$ at leading large $N$, producing a cusp and two charge-conjugate vacua in that limit Witten 1979, §§ 3–4.

The powers of $N$ are explicit if one writes, on a single branch,

$$\mathcal E(\theta)-\mathcal E(0) =\frac{\chi_t}{2}\theta^2 \left(1+b_2\theta^2+\cdots\right)$$

with the parenthesis understood as multiplying the quadratic term. Expanding $Nf(\theta/N)$ gives

$$\chi_t=\frac{f''(0)}{N}, \qquad b_2=\frac{f^{(4)}(0)}{12f''(0)N^2}.$$

Changing the action normalization changes the coefficients but not these large-$N$ powers under the declared convention. The branch periodicity and scaling checks are reproduced by the chapter’s benchmark.

This result is not the dilute-gas cosine: their higher theta derivatives have different scaling and shape. Nor does the leading large-$N$ crossing determine every finite-$N$ theory. Subleading effects, anomalies, exact equivalences, and the order of the infinite-volume limit must be considered.

Special angles and anomaly constraints

If charge conjugation $C$ sends $Q\mapsto-Q$, then it is a symmetry at $\theta=0$ and, using $2\pi$ periodicity, at $\theta=\pi$. A mixed anomaly can prevent the $\theta=\pi$ theory from having a unique, trivially gapped, symmetry-preserving vacuum. It does not by itself choose among:

• spontaneous $C$ breaking with degenerate vacua;
• a gapless infrared theory;
• a nontrivial gapped infrared sector with the required degeneracy or symmetry realization.

The precise constraint depends on $N$ and on background bundles. For the standard $\mathrm{CP}^{N-1}$ model, even $N$ has a mixed $PSU(N)$–$C$ anomaly at $\theta=\pi$; for odd $N$, the related obstruction is a global inconsistency between symmetry-preserving counterterm choices at $\theta=0$ and $\pi$. The background construction and its dynamical ceiling are explained in Gaiotto et al. 2017, § 1.1, pp. 3–5 and the even/odd refinement in Komargodski et al. 2019, § 2.

An anomaly is a nonperturbative constraint on possible endpoints. A proposed large-$N$, semiclassical, lattice, or exact description must match it, but matching does not make that description unique.

The special CP¹ / O(3) case

Because $\mathrm{CP}^1\simeq S^2$, the model is equivalent to the $O(3)$ sigma model after matching the kinetic and topological normalizations. The accepted continuum identification is that at $\theta=\pi$ it flows to the gapless $SU(2)_1$ Wess–Zumino–Witten fixed point with a marginally irrelevant perturbation. This conclusion combines the spin-chain mapping, non-Abelian bosonization, anomaly matching, and numerical evidence; it is not implied by anomaly matching alone. Haldane supplies the integer-versus-half-integer spin-chain distinction Haldane 1983, pp. 1153–1156, while the WZW endpoint and critical exponents are developed in Affleck 1986, §§ 3–5.

The conclusion uses the $N=2$ target equivalence and its symmetry realization. It does not imply that every $\mathrm{CP}^{N-1}$ theory at $\theta=\pi$ is gapless. Large $N$ instead favors a branch crossing and spontaneous $C$ breaking, consistent with the anomaly by a different infrared mechanism.

Comparing the five kinds of evidence

Before combining results, state what each controls:

• Sector quantization controls the allowed theta phase and periodicity.
• Dilute semiclassics controls a fugacity expansion only when cores are small and events well separated.
• Large N controls a saddle and branch expansion at fixed $\theta/N$, with an explicit order of limits.
• Anomaly matching excludes some infrared possibilities but usually leaves several.
• Exact equivalence or integrability can determine special models after its quantum conditions are independently established.

The dual-variable conditions show how flux, auxiliary, and bosonized descriptions retain their global assumptions. The strong-laboratory regime comparison prevents a theta conclusion from being transferred to a model with different topology or observables.

Common pitfalls

Calling every theta dependence instanton dominance. The Fourier expansion of a periodic function does not prove a dilute gas. A controlled saddle-size distribution and suppressed interactions are required.

Using an anomaly as a complete phase solution. An anomaly rules out specified symmetry and gap combinations. Degeneracy and gaplessness can both match the same obstruction.

Generalizing from CP¹ to all N. The $O(3)$ equivalence and its $\theta=\pi$ endpoint are special. The large-$N$ theory has different controlled infrared behavior.

Exercises

1. For the dilute unit-charge gas, compute the connected second and fourth derivatives of the vacuum energy at $\theta=0$.

Solution

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

$$\mathcal E''(0)=2\kappa, \qquad \mathcal E^{(4)}(0)=-2\kappa.$$

Thus the normalized fourth cumulant has a fixed dilute-gas sign and magnitude. A different branch structure or interacting ensemble need not obey this relation.

2. Consider

$$\mathcal E(\theta) =\frac{\chi}{2} \min_{k\in\mathbb Z}(\theta+2\pi k)^2.$$

Show that it is $2\pi$-periodic and has a cusp at $\theta=\pi$.

Solution

Under $\theta\mapsto\theta+2\pi$, relabel $k\mapsto k-1$, leaving the minimum unchanged. For $-\pi<\theta<\pi$, the minimizing branch is $k=0$. Immediately above $\pi$, it is $k=-1$. The left and right derivatives at $\pi$ are $\chi\pi$ and $-\chi\pi$, so the derivative jumps. The two branches represent degenerate charge-conjugate vacua at the crossing.

3. Starting from $\mathcal E_k(\theta)=Nf((\theta+2\pi k)/N)$, derive the large-$N$ scaling of $\chi_t$ and $b_2$. Compare it with the dilute-gas value.

Solution

On the branch around $\theta=0$,

$$Nf(\theta/N) =Nf(0)+\frac{f''(0)}{2N}\theta^2 +\frac{f^{(4)}(0)}{24N^3}\theta^4+\cdots.$$

Comparing with

$$\mathcal E(\theta)-\mathcal E(0) =\frac{\chi_t}{2}\theta^2(1+b_2\theta^2+\cdots)$$

gives

$$\chi_t=\frac{f''(0)}N, \qquad b_2=\frac{f^{(4)}(0)}{12f''(0)N^2}.$$

For the dilute cosine, $1-\cos\theta=\theta^2/2-\theta^4/24+\cdots$, so $b_2=-1/12$, independent of $N$. The different scaling is a diagnostic that the two approximations describe different regimes.

4. An anomaly excludes a unique trivially gapped $C$-preserving vacuum at $\theta=\pi$. Give two inequivalent infrared endpoints that still satisfy the constraint, and state what extra evidence would distinguish them.

Solution

Two $C$-conjugate gapped vacua can match the anomaly through spontaneous $C$ breaking. A gapless conformal theory can match it through its massless degrees of freedom and symmetry realization. A controlled branch calculation or finite-volume vacuum splitting can support the first; operator dimensions, central charge, and symmetry action from an exact equivalence or continuum calculation can support the second. Anomaly matching is a necessary consistency test for either endpoint, not a selection rule between them.

Continue

Use Strong-Coupling Phases and Cross-Method Evidence to compare claims with nonoverlapping systematics.

References

• Affleck, Ian. “Exact Critical Exponents for Quantum Spin Chains, Nonlinear Sigma Models at Theta = π and the Quantum Hall Effect.” Nuclear Physics B 265 (1986): 409–447. DOI.
• Bonanno, Claudio, Massimo D’Elia, and Francesca Margari. “Topological Susceptibility of 2d CP¹ or O(3) Non-Linear Sigma-Model: Is It Divergent or Not?” Physical Review D 107 (2023): 014515. DOI.
• Gaiotto, Davide, Anton Kapustin, Zohar Komargodski, and Nathan Seiberg. “Theta, Time Reversal, and Temperature.” Journal of High Energy Physics 2017, no. 5 (2017): 091. DOI.
• Haldane, F. D. M. “Nonlinear Field Theory of Large-Spin Heisenberg Antiferromagnets: Semiclassically Quantized Solitons of the One-Dimensional Easy-Axis Néel State.” Physical Review Letters 50 (1983): 1153–1156. DOI.
• Komargodski, Zohar, Adar Sharon, Ryan Thorngren, and Xinan Zhou. “Comments on Abelian Higgs Models and Persistent Order.” SciPost Physics 6 (2019): 003. DOI.
• Witten, Edward. “Instantons, the Quark Model, and the 1/N Expansion.” Nuclear Physics B 149 (1979): 285–320. DOI.