Spin, Charges, and Extended Modular Sectors

A single scalar partition function is often too coarse for modular consistency. Spin structures, symmetry twists, charge insertions, topological defects, and extended chiral modules usually form a finite or integral vector under $S$ and $T$. The bootstrap object is that closed vector together with its phases, measures, and positivity cones.

Required background. Modular crossing and spectral bounds supplies the functional logic, and cosets and orbifolds supplies twisted Hilbert spaces and extended characters. Helpful background. Gauging generalized symmetries clarifies when a collection of defects or twists can be consistently summed.

Sector-valued traces under S and T

Let a finite symmetry group $G$ act on the CFT. Choose the convention that $g$ twists the spatial cycle and $h$ is inserted along Euclidean time. For commuting $g,h$,

$$Z_{g,h}(\tau,\bar\tau) =\operatorname{Tr}_{\mathcal H_g} \left( h\,q^{L_0-c_L/24} \bar q^{\bar L_0-c_R/24} \right).$$

Here $\mathcal H_g$ is the $g$-twisted Hilbert space. With this cycle convention, an anomaly-free theory obeys

$$S:\quad Z_{g,h}\longmapsto Z_{h,g^{-1}}, \qquad T:\quad Z_{g,h}\longmapsto Z_{g,hg}.$$

For nonabelian $G$, the pair must commute and is identified under simultaneous conjugation. An anomaly or discrete-torsion choice can multiply these maps by phases. Those phases are part of the representation; omitting them can turn a consistent projective transformation law into a false failure of invariance. Twisted traces and their modular closure are developed in Dijkgraaf, Vafa, Verlinde, and Verlinde 1989, § 4, pp. 495–507.

The orbifold torus amplitude has the schematic form

$$Z_{G} =\frac{1}{\lvert G\rvert} \sum_{\substack{g,h\in G\\gh=hg}} \epsilon(g,h)Z_{g,h},$$

where $\epsilon(g,h)$ is a consistent discrete-torsion phase. Modular invariance of this sum is a consequence of closure of the full set and compatibility of $\epsilon$ with the cycle transformations. Summing only twined untwisted traces $Z_{1,h}$ misses the $Z_{h,1}$ sectors generated by $S$.

A four-component check for a Z₂ symmetry

For $G=\mathbb Z_2=\{1,g\}$ and no anomaly, order the components as

$$\boldsymbol Z= \left(Z_{1,1},Z_{1,g},Z_{g,1},Z_{g,g}\right)^{\mathsf T}.$$

Since $g^{-1}=g$, the generators act by the permutation matrices

$$\rho(S)= \begin{pmatrix} 1&0&0&0\\ 0&0&1&0\\ 0&1&0&0\\ 0&0&0&1 \end{pmatrix}, \qquad \rho(T)= \begin{pmatrix} 1&0&0&0\\ 0&1&0&0\\ 0&0&0&1\\ 0&0&1&0 \end{pmatrix}.$$

Thus $S$ exchanges the twined and spatially twisted components, while $T$ exchanges $Z_{g,1}$ and $Z_{g,g}$. The equal-weight sum of all four components is invariant. This elementary example already shows why a one-component test cannot establish orbifold consistency.

Flavored traces and anomaly phases

For a holomorphic $U(1)$ current normalized by

$$J(z)J(0)\sim\frac{k_L}{z^2},$$

and an antiholomorphic current of level $k_R$, introduce chemical potentials $z,\bar z$:

$$Z(\tau,\bar\tau;z,\bar z) =\operatorname{Tr}_{\mathcal H} q^{L_0-c_L/24}\bar q^{\bar L_0-c_R/24} e^{2\pi izQ_L}e^{-2\pi i\bar zQ_R}.$$

For

$$\gamma= \begin{pmatrix}a&b\\r&d\end{pmatrix} \in SL(2,\mathbb Z),$$

the flavored trace transforms as a Jacobi-like object,

$$\begin{aligned} &Z\!\left( \frac{a\tau+b}{r\tau+d}, \frac{a\bar\tau+b}{r\bar\tau+d}; \frac{z}{r\tau+d}, \frac{\bar z}{r\bar\tau+d} \right)\\ &\qquad= \exp\!\left[ \frac{\pi ik_L r z^2}{r\tau+d} -\frac{\pi ik_R r\bar z^2}{r\bar\tau+d} \right] Z(\tau,\bar\tau;z,\bar z), \end{aligned}$$

up to any additional multiplier already present at $z=0$. The Gaussian factor is fixed by the current anomaly and charge normalization Dyer, Fitzpatrick, and Xin 2018, §§ 1–2. Dividing it away without tracking the resulting nonholomorphic completion changes the functional equation.

The chemical potentials initially define an analytic generating function near $z=\bar z=0$. A real thermodynamic chemical potential corresponds to an imaginary Euclidean insertion and can make the trace diverge when charged degeneracies outrun the Boltzmann suppression. Every bound must therefore state the domain in $(\tau,z)$ on which the trace converges and the contour on which derivatives are taken.

Charge-resolved multiplicities are nonnegative:

$$Z(\tau,z) =\sum_{\Delta,J,Q}d_{\Delta,J,Q}, q^{h-c_L/24}\bar q^{\bar h-c_R/24}e^{2\pi izQ}, \qquad d_{\Delta,J,Q}\ge0.$$

But the value of a twined or flavored trace need not be positive because the insertion supplies phases. Functional positivity should be imposed on the underlying charge or representation multiplicities, not on a complex numerical value of $Z(\tau,z)$.

Spin structures, defects, and extended characters

For fermions, periodic or antiperiodic boundary conditions around each torus cycle give a four-component spin-structure vector. The modular generators permute these components and can add phases determined by the gravitational and fermionic anomalies. A modular invariant obtained after a spin-structure sum is not evidence that each component is invariant.

A topological defect line wrapped on a cycle behaves similarly: exchanging the cycles converts a defect insertion in the trace into a defect-twisted Hilbert space. If defects are noninvertible, their fusion coefficients can enlarge the sector vector beyond group-labeled twists. Positivity is channel-dependent because a trace with a defect eigenvalue is generally signed or complex.

An extended chiral algebra reorganizes Virasoro modules into larger modules with characters $\widehat\chi_A$. In a rational theory,

$$\widehat{\boldsymbol\chi}(-1/\tau) =\widehat S\widehat{\boldsymbol\chi}(\tau), \qquad \widehat{\boldsymbol\chi}(\tau+1) =\widehat T\widehat{\boldsymbol\chi}(\tau).$$

The pairing matrix must commute with these extended transformations. This can strengthen a bootstrap problem because descendant towers are treated exactly, but it also changes the vacuum character and the definition of a primary gap. The relation between modular transformations and rational chiral data is classical Verlinde 1988, §§ 2–4; the rigorous categorical realization is beyond this page’s physical trace construction.

A sector consistency table

The table below is the working record for a sector-refined calculation. “Closure” means that every displayed $S$ or $T$ image is present in the declared vector, including its phase. Tail estimates apply only when a character or spectral series has actually been truncated.

Object	Definition and normalization	Modular image to include	Positivity statement	Closure and omitted-tail check
Untwisted trace	$Z_{1,1}$ with vacuum multiplicity one	fixed by $S,T$ only in the nonanomalous scalar case	state multiplicities are nonnegative in a unitary theory	verify vacuum shifts and multiplier before treating it as one component
Twined trace	$Z_{1,h}=\operatorname{Tr}_{\mathcal H}(h\,q^{\cdots}\bar q^{\cdots})$	$S$ requires $Z_{h,1}$	the trace value may be signed or complex	include every conjugacy class generated by $S,T$; do not impose scalar positivity
Twisted sector	$Z_{g,h}$ on $\mathcal H_g$ for $gh=hg$	$(g,h)\mapsto(h,g^{-1})$ and $(g,hg)$	multiplicities in $\mathcal H_g$ are nonnegative, insertion eigenvalues need not be	check simultaneous-conjugation classes and anomaly phases
Flavored trace	charge normalization $J(z)J(0)\sim k/z^2$	Jacobi phase and rescaled $z$	$d_{\Delta,J,Q}\ge0$ in charge-resolved coefficients	specify analytic domain; derivatives at $z=0$ do not control a divergent real-$\mu$ trace
Extended character	chosen algebra and module order	finite matrices $\widehat S,\widehat T$ or a declared kernel	multiplicities of full modules are nonnegative	verify $\widehat S^\dagger M\widehat S=M$ and $\widehat T^\dagger M\widehat T=M$
Truncated $q$ series	retain levels $n\le N$ in each declared component	transform the full character, not the truncated polynomial	coefficient positivity alone does not bound a transformed tail	if $a_n\le Ae^{\kappa n}$ and $\lvert q\rvert e^\kappa<1$, then $\lvert R_N\rvert\le A(\lvert q\rvert e^\kappa)^{N+1}/(1-\lvert q\rvert e^\kappa)$

The last row gives a genuine bound only after $A$ and $\kappa$ have been proved or conservatively estimated for every component. Modular transformation can move $\tau$ to a point where $\lvert q\rvert$ is larger, so the bound must be reevaluated after the map.

Building a vector-valued crossing equation

Let $\boldsymbol Z$ contain all chosen sectors and let $\rho(S),\rho(T)$ be their transformation matrices, including multipliers. A modular crossing equation under $S$ is

$$\boldsymbol F(\tau,\bar\tau) =\boldsymbol Z(\tau,\bar\tau) -\rho(S)^{-1} \boldsymbol Z(-1/\tau,-1/\bar\tau)=0.$$

Decompose each component into charge, spin, or representation sectors. A functional is now vector-valued: it pairs with components and derivatives. Its positivity cone is determined by the allowed multiplicity vectors, not by component-wise positivity of $\boldsymbol Z$. Before optimizing a bound, perform three exact tests:

1. check the modular group relations satisfied by $\rho(S),\rho(T)$, allowing the declared central phase;
2. apply $S$ and $T$ to every sector label and confirm that the image is present; and
3. verify that the vacuum occurs only in the sectors and with the multiplicity allowed by the boundary conditions.

Only then is it meaningful to test a gap in a charged or twisted subsector. The resulting bound is conditional on the chosen sector completeness and anomaly class. Flavored modular methods can bound charged spectra precisely because they retain this extra structure Benjamin et al. 2016, §§ 2–4.

Common failure modes

Keeping twining but dropping twisting. $S$ exchanges a symmetry insertion with a spatial twist. The shortened vector is not modularly closed.

Using the wrong charge normalization. Rescaling $Q$ changes the current level and the Jacobi phase together. A charge bound quoted without $k$ has no invariant meaning.

Demanding positivity of a complex trace. Positivity belongs to multiplicities in a Hilbert-space or representation decomposition. Insertions can weight them by phases.

Assuming anomaly phases cancel. A projective modular action may be the correct boundary behavior in the presence of inflow. State the anomaly and test the corresponding projective relations.

Exercises

Using the displayed $\mathbb Z_2$ permutation matrices, verify that $\rho(S)^2=I$ and that the vector $(1,1,1,1)^{\mathsf T}$ is fixed by both generators.

Solution

$\rho(S)$ swaps components two and three, so applying it twice is the identity. $\rho(T)$ swaps components three and four. A vector with all entries equal is unchanged by either permutation; therefore the equal-weight orbifold sum is invariant in the anomaly-free convention.

Suppose an extended character has coefficients $a_n\le3e^{n/10}$ and is evaluated at $\lvert q\rvert=e^{-1}$. Bound the tail after level $N$.

Solution

Here $\lvert q\rvert e^\kappa=e^{-9/10}<1$, so

$$\lvert R_N\rvert \le \frac{3e^{-\frac9{10}(N+1)}}{1-e^{-9/10}}.$$

The same numerical bound cannot be reused after a modular transformation without recomputing $\lvert q'\rvert$.

Continue to higher genus

Higher-Genus and Mapping-Class Constraints replaces the torus pair of cycles by a genus-$g$ homology basis and introduces three-point coefficients through sewing.

References

• Benjamin, Nathan, Ethan Dyer, A. Liam Fitzpatrick, and Shamit Kachru. “Universal Bounds on Charged States in 2d CFT and 3d Gravity.” Journal of High Energy Physics 2016, no. 8 (2016): 041. doi:10.1007/JHEP08(2016)041. Open preprint.
• Dijkgraaf, Robbert, Cumrun Vafa, Erik Verlinde, and Herman Verlinde. “The Operator Algebra of Orbifold Models.” Communications in Mathematical Physics 123, no. 3 (1989): 485–526. doi:10.1007/BF01238812.
• Dyer, Ethan, A. Liam Fitzpatrick, and Yuan Xin. “Constraints on Flavored 2d CFT Partition Functions.” Journal of High Energy Physics 2018, no. 2 (2018): 148. doi:10.1007/JHEP02(2018)148. Open preprint.
• Verlinde, Erik. “Fusion Rules and Modular Transformations in 2D Conformal Field Theory.” Nuclear Physics B 300 (1988): 360–376. doi:10.1016/0550-3213(88)90603-7.