---
title: Generalised Discrete Torsion in Orbifolds
url: https://www.emergentmind.com/topics/generalised-discrete-torsion
type: topic
---

# Generalised Discrete Torsion in Orbifolds

Searching arXiv for recent and foundational papers on generalized discrete torsion and related discrete torsion frameworks.
{"query":"ti:\"On Generalised Discrete Torsion\" OR generalised discrete torsion orbifold equivariant cohomology", "max_results": 10}
{"query":"generalized discrete torsion equivariant cohomology orbifold B-field arXiv", "max_results": 10}
arXiv search: "On Generalised Discrete Torsion" [2604.01225], related papers on discrete torsion, Chen–Ruan cohomology, F-theory, ABJM, GLSM, and non-invertible gauging.
arXiv search: "discrete torsion non-invertible symmetries lazy cohomology B-field" and "F-theory torsion homology discrete gauge symmetries".
Generalised discrete torsion is a refinement of ordinary orbifold discrete torsion in a \(2\)d gauged sigma model with target space \(M\) and discrete gauge group \(G\). Ordinary discrete torsion is classified by \(H^2(BG;U(1))\), whereas the generalised form is encoded by the equivariant group
\[
H^2_G(M;U(1)) \;=\; H^2\!\big((EG\times M)/G;U(1)\big),
\]
so it can assign different local discrete torsion phases to different singular loci of the orbifold \(M/G\) while still arising from a single global flat \(B\)-field class [2604.01225]. In this formulation it unifies ordinary discrete torsion and ordinary flat \(B\)-fields: if \(M=\mathrm{pt}\) one recovers \(H^2(BG;U(1))\), while if \(G=1\) one obtains \(H^2(M;U(1))\) [2604.01225].

## 1. Cohomological definition and relation to ordinary discrete torsion

In the ordinary orbifold construction, a discrete torsion class \(\omega\in H^2(BG;U(1))\) assigns phases depending only on commuting twists \(g,h\in G\) around the worldsheet cycles. The standard cocycle relation is
\[
\omega(g,h)+\omega(gh,k)=\omega(g,hk)+\omega(h,k),
\]
with the usual coboundary equivalence
\[
\tilde\omega(g,h)=\omega(g,h)+\psi(g)+\psi(h)-\psi(gh),
\]
and on a torus the phase is
\[
\langle g,h\rangle_\omega=\omega(g,h)-\omega(h,g).
\]
For \(G=\mathbb Z_2\times \mathbb Z_2\), this leaves a single nontrivial choice since \(H^2(B(\mathbb Z_2\times \mathbb Z_2);U(1))\cong \mathbb Z_2\) [2604.01225].

Generalised discrete torsion replaces the classifying space \(BG\) by the Borel construction \((EG\times M)/G\). The gauged sigma model together with a \(G\)-bundle on the worldsheet defines a map \(\Sigma\to (EG\times M)/G\), and a class in \(H^2_G(M;U(1))\) can be pulled back and integrated to produce the exponentiated phase
\[
\exp\!\Big(2\pi i\int_\Sigma f^*(\omega)\Big).
\]
The continuous identity component of \(H^2_G(M;U(1))\) is the part of the \(B\)-field that survives the gauging, while the finite quotient
\[
[H^2_G(M;U(1))] \;=\; H^2_G(M;U(1))/H^2_G(M;U(1))_0
\]
is the genuinely discrete generalised discrete torsion [2604.01225].

## 2. Local phases on singular strata

The distinctive feature of the generalised theory is locality on the orbifold singular set. If \(N\subset M\) is a singular locus fixed by a subgroup \(H\le G\), the equivariant class restricts to
\[
\iota_N^*(\omega)\in H^2_H(N;U(1)),
\]
and at a fixed point \(x\) this gives an element of
\[
H^2(BH;U(1)).
\]
Different singular loci of \(M/G\) can therefore carry different local phases [2604.01225].

These local phases are not arbitrary decorations. They arise as restrictions of one global class in \(H^2_G(M;U(1))\), so they must satisfy compatibility relations inherited from the global cohomology. This is precisely the sense in which generalised discrete torsion gives a consistent implementation of Gaberdiel and Kaste’s prescription for inserting local discrete torsion phases by hand at higher genus. The same structure is identified with the equivariant gerbe formulation: for torus orbifolds, the equivariant gerbe phases reproduce the commutator phase from the cocycle together with the additional shift terms [2604.01225].

A plausible implication is that generalised discrete torsion should be viewed less as a collection of independent sector-by-sector signs than as a globally constrained equivariant \(B\)-field datum. That viewpoint is also consistent with the physical interpretation developed for ordinary orbifolds, where discrete torsion measures differences in gauge actions on \(B\)-fields [2406.02676].

## 3. Torus orbifolds and explicit classifications

For torus orbifolds \(T^n/G\), the quotient can be rewritten as
\[
T^n/G \;=\; \mathbb R^n/\hat G,
\]
where \(\hat G\) is a crystallographic extension of \(G\) by the lattice \(\mathbb Z^n\). In that setting,
\[
H^2_G(T^n;U(1)) = H^2(B\hat G;U(1)),
\]
so generalised discrete torsion becomes ordinary discrete torsion for a larger discrete group \(\hat G\) [2604.01225].

The case \(T^6/\mathbb Z_2^2\) exhibits both the richness and the rigidity of the construction. The relevant generalised torsion group is
\[
H^2_{\mathbb Z_2^\alpha\times \mathbb Z_2^\beta}(T^6;U(1)) = U(1)^3\times \mathbb Z_2^{19}.
\]
Here the \(U(1)^3\) factor is the continuous \(B\)-field on the three invariant \(2\)-tori, while the \(19\) discrete \(\mathbb Z_2\)-valued combinations are the genuine generalised discrete torsion classes. Local torsion on an \(\alpha\)-fixed \(T^2\), for example, is determined by
\[
\big[\iota_{T^2_\alpha}^*(\omega)\big] = \big(\langle \tilde\alpha,S_1\rangle,\langle \tilde\alpha,S_2\rangle,\langle \tilde\alpha,\tilde\beta\rangle\big).
\]
Although these local data govern whether a singular \(T^2\)-locus is resolved into a \(2\)-cycle or deformed into \(3\)-cycles in the worldsheet CFT, complete smoothness forces strong compatibility conditions because the singular loci intersect. The result is that all generalised phases must vanish except the ordinary torsion \(\langle \alpha,\beta\rangle\), so the only fully smooth Calabi–Yau outcomes are the familiar two with
\[
(h^{1,1},h^{2,1})=(51,3)\quad\text{or}\quad(3,51)
\]
[2604.01225].

The \(T^7/\mathbb Z_2^3\) example is more flexible but still globally constrained. In this case,
\[
H^2_{\mathbb Z_2^3}(T^7;U(1)) = \mathbb Z_2^{21}.
\]
The local torsion on an \(\alpha\)-fixed \(T^3\) is
\[
\big[\iota_{T^3_\alpha}^*(\omega)\big]=\langle \tilde\alpha,S_{5,6,7}\rangle,
\]
with analogous expressions for \(\beta\) and \(\gamma\). The local phases determine the Betti numbers of smooth \(G_2\) resolutions, but they are correlated by global relations such as
\[
\langle\widetilde{\alpha\beta},\tilde\gamma\rangle = \langle \alpha\beta,\gamma\rangle +a_3\langle \alpha\gamma,S_3\rangle +a_5\langle \beta\gamma,S_5\rangle.
\]
Consequently only
\[
(b_2,b_3)=(8+\ell,\,47-\ell), \qquad \ell\in\{0,4,8\},
\]
occur, so the orbifold CFT realizes only \(3\) out of the \(9\) possible Betti numbers of Joyce \(G_2\) resolutions [2604.01225].

## 4. Geometric realizations in string compactification

Generalised discrete torsion is closely tied to the geometry of compactification spaces, especially when discrete gauge symmetries are related to torsional homology or to distinct realizations of the same low-energy vacuum.

In F-theory, the simplest \(\mathbb Z_2\) case already shows that discrete symmetry and discrete torsion need not coincide geometrically. A genus-one fibred Calabi–Yau threefold \(P_Q\) with a bi-section yields a six-dimensional F-theory vacuum with a \(\mathbb Z_2\) gauge symmetry, but in five-dimensional M-theory on \(P_Q\) there is no genuine discrete gauge symmetry and no torsional homology in \(H_\ast(P_Q,\mathbb Z)\). The relevant relation is only modulo the elliptic fiber class,
\[
2[A_C]=[T^2].
\]
By contrast, the associated Jacobian fibration \(P_W\) does support a genuine \(\mathbb Z_2\) gauge symmetry already in five dimensions, together with
\[
{\rm Tor}\,H_3(P_W,\mathbb Z)=\mathbb Z_2,
\qquad
{\rm Tor}\,H_2(P_W,\mathbb Z)=\mathbb Z_2
\]
on a smooth manifold [1410.7814]. This sharpens the distinction between a discrete symmetry that appears only in the F-theory limit and one realized by actual torsional cycles in M-theory.

In heterotic \(T^6/(\mathbb Z_2\times \mathbb Z_2)\) orbifolds, discrete torsion enters as an additional cocycle phase in the one-loop partition function,
\[
\epsilon^{\,t_1,t_2}_{t_1',t_2'} = e^{\pi i\, \epsilon\,(t_1 t_2' - t_2 t_1')},
\qquad \epsilon=0,1.
\]
Turning on torsion changes the surviving twisted states to charge conjugates of those in the no-torsion model. In the GLSM resolution, this reverses the exceptional gauge charges of the Fermi multiplets,
\[
E_{r\,xy}^{(\times)} = -\,E_{r\,xy},
\]
so the torsion resolution is genuinely \((0,2)\) and requires logarithmic FI terms whose singularities signal NS5-branes on the exceptional cycles [2311.08498].

In ABJM orbifolds, discrete torsion is implemented by twisting the crossed product algebra with a cocycle \(\alpha\in Z^2(\Gamma,U(1))\). If the class \([\alpha]\) has order \(m\), then \(m\) enters the moduli space nontrivially and changes the effective M-theory fiber quotient to
\[
\mathbb Z_{\frac{k|\Gamma|}{m}}.
\]
In this framework discrete torsion modifies the dimensions of irreducible projective representations, rescales effective Chern–Simons levels, and alters monopole quantization [1011.4733].

A further geometric realization appears in intersecting D6-brane model building on
\[
T^6/(\mathbb Z_2\times \mathbb Z_6' \times \Omega\mathcal R).
\]
Here the non-trivial discrete torsion choice \(\eta=-1\) produces exceptional homology, rigid fractional three-cycles, reduced moduli, and exact discrete gauge symmetries from massive \(U(1)\) factors. The fractional cycle decomposition
\[
\Pi_a^{\mathrm{frac}}=\frac14\left(\Pi_a^{\mathrm{bulk}}+\sum_{k=1}^3 \Pi_a^{\mathbb Z_2^{(k)}}\right)
\]
is the basic geometric mechanism behind these effects [1303.6845].

## 5. Extensions beyond finite-group orbifolds

Once the symmetry is no longer an ordinary finite group, discrete torsion splits into two distinct notions. For a fusion category \(\mathcal C\), one notion consists of Morita equivalence classes of gaugeable algebra structures on a fixed object, while the other consists of cocycle-like twists of the tensor product. The latter are represented by normalized natural isomorphisms
\[
\omega:\otimes \Rightarrow \otimes
\]
satisfying
\[
\omega_{X,1}=\omega_{1,X}=\mathrm{id}_X,
\]
and
\[
\omega_{X,Y\otimes Z}\circ (\mathrm{id}_X\otimes \omega_{Y,Z}) = \omega_{X\otimes Y,Z}\circ (\omega_{X,Y}\otimes \mathrm{id}_Z).
\]
These lazy \(2\)-cocycles form a group \(Z^2_\ell(\mathcal C)\), and modulo lazy coboundaries one obtains the lazy cohomology group
\[
H^2_\ell(\mathcal C)=Z^2_\ell(\mathcal C)/B^2_\ell(\mathcal C).
\]
For \(\mathcal C=\mathrm{Vec}_G\), both notions reduce to \(H^2(G,U(1))\), but for general fusion categories they no longer coincide [2406.02676].

Only the twist notion generalizes the older picture of discrete torsion as a difference in gauge actions on \(B\)-fields. On a gaugeable algebra \((A,\mu,\Delta)\), a twist acts by
\[
\mu_\omega=\mu\circ \omega_{A,A},\qquad \Delta_\omega=\omega^{-1}_{A,A}\circ \Delta,
\]
and this preserves the structure of a symmetric special Frobenius algebra [2406.02676]. This suggests that the equivariant \(B\)-field interpretation of generalised discrete torsion survives in non-invertible settings, but the classification problem becomes categorical rather than purely group cohomological.

A different refinement arises in orbifolds with discrete torsion and localized singularities. There the generalized symmetry data can be extracted either from the asymptotic geometry using Chen–Ruan orbifold cohomology or from the fermionic adjacency matrix of the D-brane probe quiver. With discrete torsion \(\alpha\in H^2(\Gamma;U(1))\), the physically useful reformulation introduces
\[
\Gamma_\alpha \equiv \ker \alpha,
\qquad
X_\alpha = \mathbb R^6/\Gamma_\alpha \to X=\mathbb R^6/\Gamma,
\]
and defines
\[
H^n_{\mathrm{orb},\alpha}(X)\equiv H^n_{\mathrm{orb}}(X_\alpha).
\]
In this description, discrete torsion effectively reorganizes the background into a minimal covering space without that discrete torsion, and the free factors of the resulting orbifold cohomology count higher-dimensional branes localized at singular strata [2504.10484].

## 6. Terminological scope and distinction from generalized torsion

The expression “generalised discrete torsion” should be distinguished from the unrelated group-theoretic notion of generalized torsion. In group theory, a nontrivial element \(g\in G\) is called a generalized torsion element if some nonempty finite product of its conjugates is the identity,
\[
g^{x_1} g^{x_2}\cdots g^{x_n}=1.
\]
This notion is used in the study of knot groups, \(3\)-manifold groups, and orderability, where it functions as an obstruction to bi-orderability [1811.07532]. It is also central to Dehn filling constructions in \(3\)-manifold topology, where a meridian can become generalized torsion after surgery [2009.00152].

The two subjects share the word “torsion” but not the same underlying structure. Generalised discrete torsion concerns flat equivariant \(B\)-fields, orbifold phases, and their cohomological or categorical classification. Generalized torsion in group theory concerns finite products of conjugates of a nontrivial group element. The lexical overlap is therefore potentially misleading, and in current usage the two theories belong to different mathematical and physical contexts [1811.07532].

Source: https://www.emergentmind.com/topics/generalised-discrete-torsion