Theta Terms, Periodicity, and Vacuum Sectors

A theta parameter is a character of the actual topological-charge lattice of a fully specified theory. If every admitted configuration has integer charge, its sector weight is $e^{i\theta Q}$ and $\theta\sim\theta+2\pi$. That familiar conclusion can fail, or can hold only after another discrete datum is shifted, when the global gauge group admits fractional-charge bundles, when a boundary is present, or when background counterterms are part of the definition. Orientation reversal sends $Q$ to $-Q$, so it sends $\theta$ to $-\theta$; this kinematic statement alone does not decide whether CP is unbroken or how many vacua the theory has.

This page develops that global definition and its checks. Instanton calculus, vacuum-energy branches, tunneling, strong-CP phenomenology, and axion dynamics require additional dynamical input and are left to their dedicated treatments.

Required background. When Is a Topological Term Well Defined? supplies the exponentiated-action and period tests. Large Gauge Transformations and Topological Sectors supplies the character of the disconnected transformation group and distinguishes a transformation component from a field sector.

Helpful background. Characteristic Classes and Chern–Weil Theory fixes the four-dimensional charge normalization. Global Form, Matter Representations, and the Faithful Gauge Group explains why the Lie algebra alone does not determine the admitted bundles or line operators.

Theta is a character of the admitted charge lattice

Work first on a closed, smooth, oriented Euclidean manifold $X$. Assume that the admissible configurations have already been partitioned into components $\mathcal C_\gamma$, and let $Q(\gamma)\in\mathbb R$ be constant on each component. The theta-weighted partition function is

$$Z_X(\theta) = \sum_{\gamma\in\Gamma_X} e^{i\theta Q(\gamma)}Z_{X,\gamma},$$

where $Z_{X,\gamma}$ is the functional integral restricted to $\mathcal C_\gamma$. Equivalently, the Euclidean action is written so that the path-integral weight is $e^{-S_{E,0}+i\theta Q}$. This formula presupposes a measure and a complete list of bundles, boundary conditions, defects, and background fields; it does not manufacture that global data.

Let $\Lambda_Q\subset\mathbb R$ be the additive subgroup generated by every admitted value of $Q$. The exact period group is its annihilator,

$$\operatorname{Per}(Q) = \left\{ \Delta\theta\in\mathbb R \;\middle|\; e^{i\Delta\theta q}=1 \text{ for every }q\in\Lambda_Q \right\}.$$

Thus, if $\Lambda_Q=q_0\mathbb Z$ and $q_0>0$ is the smallest positive charge unit, the fundamental period is $2\pi/q_0$. Integer charge gives $2\pi$; half-integer charge gives $4\pi$; and charges restricted to $m\mathbb Z$ would give the smaller period $2\pi/m$. The statement is about all admitted configurations, not only classical solutions or the saddles retained in an approximation.

There is one further global qualification. A shift may return every dynamical amplitude while multiplying the partition function by a phase that depends on a fixed background or tangential structure. Then the shift relates theories differing by an invertible counterterm; it is not a literal equality of the complete background-dependent functional until that counterterm is included in the identification.

Parameter, sector, character, and vacuum are different data

Four uses of the word “theta” must be kept separate.

• $\theta$ is a coupling coordinate on a family of theories.
• $\gamma$ labels a component of the configuration space used in the functional integral.
• A canonical theta sector is a one-dimensional character of an allowed disconnected transformation group.
• A vacuum is a lowest-energy state of the theory at fixed parameters and boundary conditions.

The first three statements are largely kinematic. The number of vacua, their energies, possible branch crossings, and domain walls are dynamical questions. A periodic partition function can have one smooth branch, several branches permuted by a shift, or a phase transition at a special angle; periodicity alone chooses none of them.

Integer charge gives a 2π identification only after the theory is fixed

The standard four-dimensional check uses a principal $SU(N)$ bundle on a closed oriented $X$. In the site’s Hermitian-generator convention, set

$$\mathcal A=g_{\mathrm{YM}}A, \qquad \mathcal F = \mathrm d\mathcal A-i\mathcal A\wedge\mathcal A = g_{\mathrm{YM}}F,$$

and normalize the fundamental trace by $\operatorname{tr}_F(T^aT^b)=\tfrac12\delta^{ab}$. Then

$$Q[A] = \frac{1}{8\pi^2} \int_X\operatorname{tr}_F(\mathcal F\wedge\mathcal F) = \frac{g_{\mathrm{YM}}^2}{8\pi^2} \int_X\operatorname{tr}_F(F\wedge F) \in\mathbb Z.$$

With all $SU(N)$ bundles included,

$$Z_X(\theta) = \sum_{n\in\mathbb Z}e^{in\theta}Z_{X,n}, \qquad Z_X(\theta+2\pi)=Z_X(\theta).$$

When the Fourier manipulations are justified, the fixed-charge contribution can be recovered by

$$Z_{X,n} = \frac{1}{2\pi} \int_0^{2\pi} \mathrm d\theta\, e^{-in\theta}Z_X(\theta).$$

This is a decomposition of one functional integral into admitted sectors; it is not an instanton approximation. The finite-action $\mathbb R^4$ route to the same integer and the distinction between kinematics and semiclassical dynamics are developed in Mariño 2015, § 4.3, printed pp. 112–123, especially eqs. (4.3.1)–(4.3.9), (4.3.22), and (4.3.36)–(4.3.74).

Orientation and CP reverse the topological charge

Reversing the orientation of $X$ changes the sign of the four-form integral,

$$Q_{\overline X}[A]=-Q_X[A].$$

For the standard Yang–Mills theta term, CP likewise maps the theory at $\theta$ to the theory at $-\theta$. A CP-invariant parameter value must therefore obey

$$2\theta\in\operatorname{Per}(Q),$$

and must also return every discrete counterterm and line-operator choice to itself. For a genuinely $2\pi$-periodic theory, the kinematic fixed points are $\theta=0$ and $\theta=\pi$ modulo $2\pi$. For a fixed theory with period $4\pi$, they are instead $0$ and $2\pi$ modulo $4\pi$, unless CP is combined with a permutation of additional data.

Being a fixed point of the parameter map is only the first question. It does not prove that the vacuum preserves CP, that CP is anomaly-free, or that the infrared is gapped. In pure $SU(N)$ Yang–Mills, coupling the center symmetry to its background sharpens the question at $\theta=\pi$ and can produce an anomaly or global inconsistency whose precise form depends on $N$ and the allowed counterterms. That case-specific constraint is derived in Gaiotto, Kapustin, Komargodski, and Seiberg 2017, §§ 2.2–2.3, arXiv v3 printed pp. 8–12, eqs. (2.7)–(2.13), Open PDF; it is not a universal conclusion about every theta term.

Global form and fractional sectors can change the identification

Changing the global gauge group changes which bundles and line operators belong to the theory. The cleanest example holds the algebra $\mathfrak{su}(2)$ fixed and compares $SU(2)$ with $SO(3)=SU(2)/\mathbb Z_2$.

For an $SO(3)$ bundle $E$ on a spin four-manifold, nonliftable sectors with $w_2(E)\ne0$ can have half-integral instanton number. In a standard orientation convention its fractional part is controlled by the Pontryagin square,

$$Q(E) \equiv +\frac14 \left\langle \mathcal P\bigl(w_2(E)\bigr),[X] \right\rangle \pmod 1.$$

The relation $p_1(E)\equiv\mathcal P(w_2(E))\pmod 4$ and the $SO(3)$ instanton normalization are given in Aharony, Seiberg, and Tachikawa 2013, § 6.1, arXiv v5 printed pp. 37–38, especially eqs. (6.1) and (6.5), Open PDF. In the site’s characteristic-class convention $Q(E)=\tfrac14\langle p_1(E),[X]\rangle$, giving the displayed positive sign. The two consistent $SO(3)$ line-operator theories, conventionally denoted $SO(3)_+$ and $SO(3)_-$, are exchanged by a $2\pi$ shift,

$$SO(3)_+^{\,\theta} \simeq SO(3)_-^{\,\theta+2\pi}.$$

Each fixed choice returns after $4\pi$ on spin manifolds. On general oriented nonspin manifolds, quarter-instanton sectors can occur and an $8\pi$ range, together with the relevant tangential and counterterm data, may be required. These statements and their line-operator meaning are established in Aharony, Seiberg, and Tachikawa 2013, § 1.2, arXiv v5 printed p. 5, eq. (1.5) and n. 4, Open PDF.

The global form and tangential structure determine what a theta shift actually returns

Theory and manifold	Admitted charge	Effect of a 2π shift	Fixed-theory period
SU(2), closed oriented four-manifold	Integer	Returns the same theory	2π
SO(3)±, spin four-manifold	May be half-integer	Exchanges the + and − line/counterterm choices	4π for either fixed choice
SO(3), general oriented nonspin four-manifold	May include quarter-instanton sectors	Need not return the same global theory	Can require 8π and additional tangential data

The lesson is not that quotient groups always multiply a period by the order of the quotient. One must compute the admitted charge subgroup and track how a shift acts on every discrete theta angle, line spectrum, and background counterterm. For $SU(N)/\mathbb Z_k$, that calculation is carried out in Aharony, Seiberg, and Tachikawa 2013, § 2.1, arXiv v5 printed pp. 12–13, eqs. (2.1)–(2.5), and § 2.3, printed p. 16, eqs. (2.8)–(2.9), Open PDF.

Boundaries turn an absolute phase into relative Chern–Simons data

On a closed manifold, an allowed theta-period shift can leave the exponentiated bulk action unchanged. On a manifold with boundary, the same calculation exposes relative boundary data. In the illustrative trivial-bundle case where a compact Abelian connection $a$ is globally defined on $X$ and $F=\mathrm da$,

$$\begin{aligned} \Delta S_\theta &= \frac{\Delta\theta}{8\pi^2} \int_X F\wedge F \\ &= \frac{\Delta\theta}{8\pi^2} \int_{\partial X}a\wedge\mathrm da = \frac{k}{4\pi} \int_{\partial X}a\wedge\mathrm da, \qquad k=\frac{\Delta\theta}{2\pi}. \end{aligned}$$

Here $\Delta$ denotes a parameter shift, not a field variation. Separately, varying the connection gives

$$\delta S_\theta = \frac{\theta}{4\pi^2} \int_{\partial X}\delta a\wedge F,$$

so a differentiable variational problem needs a boundary condition or boundary action. The bulk density $F\wedge F$ remains gauge invariant because $F$ is gauge invariant; possible gauge variation belongs to a chosen Chern–Simons boundary representative or completion. These are three distinct questions.

For a nontrivial bulk bundle, $a\wedge\mathrm da$ exists only patchwise and a relative characteristic class can remain; the displayed equality is then replaced by a differential-cohomological or filling construction. In the trivial-bundle subcase, it shows why a $2\pi$ shift induces a level-one Abelian Chern–Simons term at the boundary. That term is available as a spin theory; the standard bosonic oriented diagonal theory requires even level, matching the $4\pi$ period of compact $U(1)$ gauge theory on arbitrary oriented nonspin four-manifolds. Witten derives this spin/nonspin distinction for the four-dimensional Abelian theta term in Witten 1995, § 2, arXiv v1 printed pp. 3–4, eq. (2.4) and n. 1, Open PDF.

For a non-Abelian connection, $\operatorname{tr}(\mathcal F\wedge\mathcal F)$ is locally the exterior derivative of the corresponding Chern–Simons three-form. The primitive changes under patching and large transformations. Consequently, a boundary or a theta interface must specify boundary conditions, edge degrees of freedom, an inflow theory, or a relative counterterm. “Theta is periodic” on closed manifolds does not by itself identify those boundary completions.

If $\theta$ jumps across an oriented interface, the same transgression gives a Chern–Simons coupling proportional to $\Delta\theta/2\pi$. Reversing the interface orientation reverses its sign. Whether the resulting interface exists as an absolute bosonic, spin, or relative theory is a separate level-quantization and global-definition test Gaiotto, Kapustin, Komargodski, and Seiberg 2017, § 2.6, arXiv v3 printed pp. 16–17, Open PDF.

First application: compare theta, Chern–Simons, BF, and finite twists

The chapter’s theta, compact Chern–Simons, BF, and finite-gauge thread provides a useful anti-conflation test. These topological terms live in different dimensions and their parameters have different mathematical types.

Four topological terms use distinct kinds of parameter

Model or datum	Allowed label	Identification and orientation	What the label means
Four-dimensional theta sector weight	Continuous character coordinate θ modulo the annihilator of the admitted charge group	A period shift returns every sector weight; orientation sends θ to −θ	A coupling in a family of theories, not the topological sector or a vacuum state
Compact U(1) Chern–Simons	Level k: integral for a spin theory; even in the standard bosonic diagonal convention	No theta circle for k; orientation sends k to −k	A quantized response or theory datum, depending on whether the connection is fixed or summed
Compact BF	Positive integer N	No periodic identification of N; an orientation sign can be compensated by reversing one field	The compact coupling and, when both fields are summed, the order of the untwisted finite gauge theory
ZN Dijkgraaf–Witten twist	r in H³(BZN, U(1)) ≅ ZN	r is defined modulo N; orientation sends r to −r	A discrete theta angle weighting finite-bundle sectors, hence a label of the theory rather than a state in it

The Abelian Chern–Simons row uses the standard compact lattice normalization. Integral spin levels and even diagonal levels for an ordinary oriented bosonic theory follow from Belov and Moore 2005, §§ 1 and 2.1.1, arXiv v1 printed pp. 3–4 and 8–10, eqs. (1.1)–(1.3) and (2.9)–(2.12b), Open PDF.

To make the last row explicit, choose a representative $\omega_r$ of the class $r\in H^3(B\mathbb Z_N,U(1))$. A fixed flat bundle $P\to M^3$, classified by $f_P:M\to B\mathbb Z_N$, receives the unit phase

$$\mathcal W_r[M;P] = \exp\!\left( 2\pi i \left\langle f_P^*\omega_r,[M]\right\rangle \right).$$

Summing the bundle makes the field dynamical,

$$Z_{N,r}(M) = \sum_{[P]\in\pi_0\operatorname{Bun}^{\mathrm{flat}}_{\mathbb Z_N}(M)} \frac{\mathcal W_r[M;P]} {\lvert\operatorname{Aut}(P)\rvert}.$$

The untwisted choice $r=0$ agrees with compact $BF_N$ only after the compact global data and normalization are matched. Nonzero $r$ changes the finite-gauge theory by a discrete sector weight. The finite cocycle construction is given in Dijkgraaf and Witten 1990, §§ 6.2 and 6.4, printed pp. 415–416 and 420–423, eqs. (6.8)–(6.10) and (6.22)–(6.26), Open PDF; the automorphism-weighted measure and gluing interpretation are given in Freed and Quinn 1993, §§ 1–2, printed pp. 438–445.

Kapustin and Seiberg’s continuum action uses a different diagonal coefficient $p_{\mathrm{KS}}\in\mathbb Z_{2N}$: odd $p_{\mathrm{KS}}$ requires spin structure, while the ordinary oriented bosonic Dijkgraaf–Witten class $r\in\mathbb Z_N$ corresponds to $p_{\mathrm{KS}}=2r$. With that bridge fixed, compact BF and its three-dimensional finite-gauge twists are compared in Kapustin and Seiberg 2014, §§ 3 and 5, arXiv v2 printed pp. 9–13 and 20–21, eqs. (3.1)–(3.10) and (5.1)–(5.3), Open PDF. The comparison illustrates the page’s central distinction: a continuous periodic theta coordinate, a quantized level, and a finite cohomology class are not interchangeable merely because all multiply topological terms.

Nor is any of them automatically a “vacuum label.” For example, $r$ fixes which Dijkgraaf–Witten theory is being quantized; ground states on a chosen spatial manifold are vectors in that theory’s state space. Changing $r$ changes the theory, whereas choosing a ground state does not.

What periodicity does not determine

The charge lattice and its annihilator settle a precise kinematic question, but several common conclusions require more.

Vacuum branches. A relation such as $E(\theta+2\pi)=E(\theta)$ does not determine whether $E$ is smooth, whether several branches are permuted, or whether a cusp appears. Those claims require dynamics and an order of limits. Large-$N$ Yang–Mills supplies an important multibranch regime, not a universal finite-$N$ theorem Witten 1998, § 1, arXiv v1 printed pp. 2–3, eqs. (1.3)–(1.7), Open PDF.

CP realization. A point fixed by $\theta\mapsto-\theta$ is a candidate CP-symmetric theory. The vacuum can preserve CP, break it spontaneously, or match an anomaly through other infrared data.

Instanton dominance. Integer $Q$ and a Fourier sector sum do not imply that dilute instantons control the path integral. The characteristic number exists independently of any semiclassical approximation.

Redundant theta angles. Matter can change which theta combination is physical. An anomalous chiral redefinition may move theta into fermion-mass phases, and a genuinely massless fermion can remove a parameter that was physical in pure gauge theory. The invariant combination and the global measure must be specified before declaring an observable theta dependence.

Absolute boundary periodicity. Equality on closed manifolds can become equality only up to a boundary invertible theory. A boundary condition is part of the theory, not an optional afterthought.

Common pitfalls

Reading the period from the local Lie algebra. The algebra fixes the local curvature, not the bundle sectors or line spectrum. Compute the charge lattice for the declared global form.

Calling every quantized topological coefficient a theta angle. A theta angle is a character coordinate and can be periodically identified. A Chern–Simons level or BF coefficient instead lies on a quantized lattice; a Dijkgraaf–Witten class is a finite discrete label.

Treating theta sectors as multiple vacua. A sector of field configurations, a character of large transformations, and a vacuum state are different objects. Their relations are model- and dynamics-dependent.

Inferring unbroken CP at a fixed angle. The parameter can be fixed while the state breaks CP or an anomaly obstructs a trivial symmetric vacuum.

Dropping the boundary transgression. The bulk density can be gauge invariant even though its local Chern–Simons primitive is not globally defined. Boundaries and interfaces retain that relative information.

Check your understanding

1. Read a period from a fractional charge lattice

Suppose every admitted charge lies in $\tfrac12\mathbb Z$ and both integer and half-integer sectors occur. Find the fundamental theta period.

Answer

A shift must obey $e^{i\Delta\theta/2}=1$, so $\Delta\theta\in4\pi\mathbb Z$. The fundamental period is $4\pi$. Checking only the integer subsectors would incorrectly give $2\pi$.

2. Distinguish a period from a permutation

Why is $\theta\to\theta+2\pi$ not a period of one fixed spin $SO(3)_+$ theory?

Answer

Half-instanton sectors can acquire a minus sign. Equivalently, the shift changes the allowed dyonic line/counterterm choice and maps $SO(3)_+$ to $SO(3)_-$. Applying the shift twice returns the fixed theory, giving period $4\pi$.

3. Locate the orientation-fixed angles

Find the fixed points of $\theta\mapsto-\theta$ for a theory whose fixed-theory period is $4\pi$.

Answer

The condition is $2\theta\in4\pi\mathbb Z$, so $\theta=0$ or $2\pi$ modulo $4\pi$. The familiar pair $0,\pi$ assumes a $2\pi$-periodic theory.

4. Recover the boundary level

For compact Abelian gauge theory, what Chern–Simons level is induced by a theta jump $\Delta\theta=2\pi$?

Answer

Since $k=\Delta\theta/(2\pi)$, the induced level is $k=1$. It is an admissible spin Chern–Simons term. A standard bosonic oriented diagonal term requires even level, so the corresponding universal nonspin shift is $\Delta\theta=4\pi$.

5. Classify the three labels

Classify $\theta$, the compact BF coefficient $N$, and a $\mathbb Z_N$ Dijkgraaf–Witten label $r$.

Answer

$\theta$ is a continuous character coordinate modulo its period. $N$ is an integer coupling and finite-group order, not a periodic coordinate. $r$ is a discrete cohomology class defined modulo $N$. None is automatically a vacuum-state label.

6. Diagnose an overclaim about vacuum branches

A calculation proves $Z(\theta+2\pi)=Z(\theta)$. Has it proved that two vacuum branches cross at $\theta=\pi$?

Answer

No. It has established the periodicity of the complete partition function under the stated global assumptions. Branch decomposition, crossing, CP breaking, and domain-wall physics require dynamical information and an order of limits.

Continue by the question you want to answer

• Theta Dependence in Yang–Mills and QCD will develop vacuum energy, CP-sensitive observables, and QCD-specific qualifications.
• Topological Sectors, Boundary Data, and Global Form will analyze when finite-action configuration spaces actually decompose into charge sectors.
• Theta Parameters, Theta States, and Sector Sums will develop the canonical theta-state and sector-superposition viewpoints.
• Chern–Simons Actions and Level Quantization will give the intrinsic three-dimensional level and spin/framing analysis.
• BF Couplings and Discrete Topological Data will develop the compact BF and discrete-twist construction.
• Background Responses and Invertible Phases will separate fixed-field theta responses from dynamical topological theories.

These destinations are the next specialized treatments; the definitions and checks needed here are complete without them.

References

• Aharony, Ofer, Nathan Seiberg, and Yuji Tachikawa. “Reading between the Lines of Four-Dimensional Gauge Theories.” Journal of High Energy Physics 2013, no. 8 (2013): 115. doi:10.1007/JHEP08(2013)115. Open PDF, arXiv:1305.0318v5.
• Belov, Dmitriy M., and Gregory W. Moore. “Classification of Abelian Spin Chern–Simons Theories.” arXiv:hep-th/0505235v1 (2005). Stable arXiv record.
• Dijkgraaf, Robbert, and Edward Witten. “Topological Gauge Theories and Group Cohomology.” Communications in Mathematical Physics 129, no. 2 (1990): 393–429. doi:10.1007/BF02096988. Open PDF.
• Freed, Daniel S., and Frank Quinn. “Chern–Simons Theory with Finite Gauge Group.” Communications in Mathematical Physics 156, no. 3 (1993): 435–472. doi:10.1007/BF02096860. Open PDF, arXiv:hep-th/9111004v3.
• 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:10.1007/JHEP05(2017)091. Open PDF, arXiv:1703.00501v3.
• Kapustin, Anton, and Nathan Seiberg. “Coupling a QFT to a TQFT and Duality.” Journal of High Energy Physics 2014, no. 4 (2014): 001. doi:10.1007/JHEP04(2014)001. Open PDF, arXiv:1401.0740v2.
• Mariño, Marcos. Instantons and Large N: An Introduction to Non-Perturbative Methods in Quantum Field Theory. Cambridge University Press, 2015. doi:10.1017/CBO9781107705968.
• Witten, Edward. “On S-Duality in Abelian Gauge Theory.” Selecta Mathematica, New Series 1 (1995): 383–410. doi:10.1007/BF01671570. Open PDF, arXiv:hep-th/9505186v1.
• Witten, Edward. “Theta Dependence in the Large $N$ Limit of Four-Dimensional Gauge Theories.” Physical Review Letters 81, no. 14 (1998): 2862–2865. doi:10.1103/PhysRevLett.81.2862. Open PDF, arXiv:hep-th/9807109v1.