nLab twisted de Rham cohomology



Differential cohomology



Special and general types

Special notions


Extra structure





Twisted de Rham cohomology is the twisted cohomology-version of de Rham cohomology, a simple example of twisted differential cohomology.

For degree-3 twists this is the codomain of the twisted Chern character on twisted K-theory, and in its orbifold cohomology-generalization it is the codomain of the twisted equivariant Chern character on twisted equivariant K-theory.




(1-twisted de Rham cohomology)


(3-twisted de Rham cohomology)
For XX a smooth manifold and HΩ 3(X)H \in \Omega^3(X) a closed differential 3-form, the HH twisted de Rham complex is the 2\mathbb{Z}_2-graded vector space Ω even(X)Ω odd(X)\Omega^{even}(X) \oplus \Omega^{odd}(X) equipped with the HH-twisted de Rham differential

d+H():Ω even/odd(X)Ω odd/even(X), d + H \wedge(-) \;\colon\; \Omega^{even/odd}(X) \to \Omega^{odd/even}(X) \,,

Notice that this is nilpotent, due to the odd degree of HH, such that HH=0H \wedge H = 0, and the closure of HH, dH=0d H = 0.


There is also the cohomology of the chain complex whose maps are just multiplication by HH.

H():Ω even/odd(X)Ω odd/even(X), H \wedge(-) \;\colon\; \Omega^{even/odd}(X) \to \Omega^{odd/even}(X) \,,

This is also called H-cohomology (Cavalcanti 03, p. 19).


We discuss notions of twisted de Rham cohomology on (global quotient) orbifolds, as they are used for the codomain of the twisted equivariant Chern character on twisted equivariant K-theory.

This combines the above twistings in degrees 1 and 3, the latter induced from the curvature 3-form on a twisting 3-class, the former (an “inner local system”) induced from the flat connection on the circle principal bundle (over the inertia orbifold) which is classified by the transgression of the 3-twist to a 2-class. This requires that this connection be flat, hence that the transgressed 2-class is torsion, which is guaranteed by a Lemma that we discuss first, in Transgression of the 3-twist to a 1-twist on Inertia.

In all of the the following:

All twisting classes, in the following are in ordinary Borel-equivariant cohomology, hence in the ordinary cohomology of a Borel construction. The “3-twist” and “torsion 2-twist” are in integral cohomology (of the Borel construction) and the “1-twist” has coefficients a finite cyclic group.

Transgression of 3-twist to 1-twist on inertia

We discuss the transgression of a 3-twist on a (global quotient) orbifold to a 1-twist on its inertia orbifold, namely to a torsion 2-class:

  1. via the Künneth theorem

    as argued in Becerra & Uribe 2009, Section 3.2;

  2. via a Serre spectral sequence

    as sketched in Freed, Hopkins & Teleman 07, (3.5).

There is also a corresponding argument in terms of bundle gerbes, given in Tu & Xu 2006, Prop. 2.6 & 3.6.

Via Künneth theorem

Throughout, fix an element gGg \in G.

Consider the right group action of the direct product group C G(g)×C_G(g) \times \mathbb{Z} (of the centralizer subgroup with the integers) on the fixed locus X gX^g (2) given by

(3)X g×C G(g)× X g (x,(h,n)) xhg n. \array{ X^g \times C_G(g) \times \mathbb{Z} &\xrightarrow{\;\;\;\;}& X^g \\ \big( x, \, (h, n) \big) &\mapsto& x \cdot h \cdot g^n \mathrlap{\,.} }

and notice that the following function is a group homomorphism (by the fact that all elements of C G(g)C_G(g) commute with gg in GG):

(4)C G(g)× ϕ g G (h,n) hg n. \array{ C_G(g) \times \mathbb{Z} &\xrightarrow{\;\;\phi_g\;\;}& G \\ (h,n) &\mapsto& h \cdot g^n \mathrlap{\,.} }

In view of the action (3), the homomorphism (4) induces a map of Borel constructions

(X gC G(g)×B)X g(C G(g)×)i gϕ gXG, \big( X^g \sslash C_G(g) \times B \mathbb{Z} \big) \;\simeq\; X^g \sslash \big( C_G(g) \times \mathbb{Z} \big) \xrightarrow{ \;\; i_g \sslash \phi_g \;\; } X \sslash G \,,

(where the homotopy equivalence shown on the left follows since the group action of \mathbb{Z} on X gX^g is the trivial action, by definition (3), and using that the Borel construction on a point is the classifying space *BS 1 \ast \sslash \mathbb{Z} \,\simeq\, B \mathbb{Z} \simeq S^1, hence the circle, in the present case)

and hence the corresponding pullback in integral cohomology:

(5)H (XG) (i gϕ g) * H ((X gC G(g))×B) H ((X gC G(g))) H (B;){ | {0,1} 0 | else H ((X gC G(g)))H 1((X gC G(g))), \begin{array}{rrl} H^\bullet \big( X \sslash G \; \, \mathbb{Z} \big) & \xrightarrow{ \; (i_g \sslash \phi_g)^\ast \; } & H^\bullet \big( ( X^g \sslash C_G(g) ) \times B \mathbb{Z} \; \, \mathbb{Z} \big) \\ & \,\simeq\, & H^\bullet \big( ( X^g \sslash C_G(g) ) \; \, \mathbb{Z} \big) \otimes_{\mathbb{Z}} \underset{ \mathclap{ \simeq \left\{ \array{ \mathbb{Z} &\vert& \bullet \in \{0,1\} \\ 0 &\vert& else } \right. } }{ \underbrace{ H^\bullet \big( B \mathbb{Z} ;\, \mathbb{Z} \big) } } \\ &\simeq & H^\bullet \big( ( X^g \sslash C_G(g) ) \; \, \mathbb{Z} \big) \,\oplus\, H^{\bullet-1} \big( ( X^g \sslash C_G(g) ) \; \, \mathbb{Z} \big) \,, \end{array}

where the second line uses the Künneth theorem.

(The following definition becomes a proposition if one uses conceptualization of transgression as laid out in transgression in group cohomology.)


For gGg \in G write

τ g:H (XG;)H 1((X gC G(g));) \tau_g \;\colon\; H^\bullet \big( X \sslash G ;\, \mathbb{Z} \big) \xrightarrow{\;\;} H^{\bullet-1} \big( ( X^g \sslash C_G(g) ) ;\, \mathbb{Z} \big)

for the composite of (5) with projection onto the second direct summand.


The image of the transgression map τ g\tau_g (Def. ) is in the torsion subgroup.

(Becerra & Uribe 2009, Lemma 3.2)

The group homomorphism (4) factors through the cyclic group gG\mathbb{Z} \to \langle g\rangle \to G which is generated by gGg \in G. By the assumption that GG is a finite group, so that its integral cohomology is entirely torsion. Since the transgression map factors through a tensor product with pullback along this factorization, by definition, the claim follows.

Via the Serre spectral sequence


For each gGg \in G and each point in the Borel construction X gC G(g)gX^g \!\sslash\! \frac{C_G(g)}{\langle{g}\rangle}, there is a homotopy fiber sequence of the form

(6)Bg X gg (pb) * X gC G(g)g \array{ B \langle{g}\rangle &\longrightarrow& X^g \!\sslash\! \langle{g}\rangle \\ \big\downarrow &{}^{{}_{(pb)}}& \big\downarrow \\ \ast &\longrightarrow& X^g \!\sslash\! \frac{C_G(g)}{\langle{g}\rangle} }


First notice that the short exact sequence

1gC G(g)G/g1 1 \to \langle{g}\rangle \hookrightarrow C_G(g) \twoheadrightarrow G/\langle{g}\rangle \to 1

deloops to a homotopy fiber-sequence of the form

(7)Bg BC G(g) BC G(g)g \array{ B \langle{g}\rangle &\longrightarrow& B C_G(g) \\ && \big\downarrow \\ && B \frac{C_G(g)}{\langle{g}\rangle} }

(using this Prop.).

Then notice that we have a pasting diagram of homotopy pullbacks as follows:

Bg * (pb) X gC G(g) X gC G(g)g (pb) BG BC G(g)g \array{ B \langle{g}\rangle &\longrightarrow& \ast \\ \big\downarrow &{}^{{}_{(pb)}}& \big\downarrow \\ X^g \!\sslash\! C_G(g) &\xrightarrow{\;}& X^g \!\sslash\! \frac{C_G(g)}{\langle{g}\rangle} \\ \big\downarrow &{}^{{}_{(pb)}}& \big\downarrow \\ B G &\xrightarrow{\;}& B \frac{C_G(g)}{\langle{g}\rangle} }


  • the bottom square being a homotopy pullback expresses that the group action of g\langle{g}\rangle on X gX^g is trivial, so that the group action of C G(g)C_G(g) on X gX^g is induced by that of C G(g)/gC_G(g)/\langle{g}\rangle;

  • the total rectangle is a homotopy pullback by (7);

  • hence the top square is a homotopy pullback by the pasting law.

It follows – once we know (do we?) that the action of the fundamental group of X gC G(g)gX^g \sslash \frac{C_G(g)}{\langle{g}\rangle} on the integral cohomology of BgB\langle{g}\rangle is trivial – that we have a Serre spectral sequence

(8)H p(X gC G(g)g;H q(Bgg;))H p+q(X gC G(g);). H^p \left( X^g \!\sslash\! \frac{C_G(g)}{\langle{g}\rangle} ;\, H^q \big( B\langle{g}\rangle{g} ;\, \mathbb{Z} \big) \right) \;\; \overset{\;\;\;\;}{\Rightarrow} \;\; H^{p+q} \big( X^g \!\sslash\! C_G(g) ;\, \mathbb{Z} \big) \,.

But since g/(ord(g))\langle{g}\rangle \simeq \mathbb{Z}/(ord(g)) is a cyclic group (by the assumption that GG is a finite group) it follows (by this Prop. or Lem. 4.51 in BSST 07) that its group cohomology is concentrated in even degrees:

H n(Bg;)=H grp n(g;){ | n=0 /ord(g) | n=positive multiple of2 0 | otherwise H^n \big( B \langle{g}\rangle ;\, \mathbb{Z} \big) \;=\; H^n_{grp} \big( \langle{g}\rangle ;\, \mathbb{Z} \big) \;\simeq\; \left\{ \begin{array}{ccl} \mathbb{Z} &\vert& n = 0 \\ \mathbb{Z}/ord(g) &\vert& n = \,\text{positive multiple of}\, 2 \\ 0 &\vert& \text{otherwise} \end{array} \right.

Therefore the spectral sequence (8) associates with any 3-twist αα |X gH 3(X gC G(g);)\alpha \mapsto \alpha_{\vert X^g} \in H^3\big( X^g \sslash C_G(g);\, \mathbb{Z} \big) a “transgressed” degree-1 class

(9)τ g(α)H 1(X g;/ord(g))βH 2(X g;), \tau_g(\alpha) \;\in\; H^1 \left( X^g ;\, \mathbb{Z}/ord(g) \right) \xrightarrow{\;\;\beta\;\;} H^2 \left( X^g ;\, \mathbb{Z} \right) \,,

which the Bockstein homomorphism identifies with a torsion 2-class in integral cohomology.

Essentially this conclusion is claimed as FHT 07, (3.5).


Now we can indicate the definition of the twisted equivariant de Rham cohomology. In outline:

Write Λ((XG))\Lambda (\prec (X \sslash G)) for the inertia orbifold of the global quotient orbifold of XX (1).

For αH 3(XG;)\alpha \in H^3\big( X \sslash G; \, \mathbb{Z} \big) a “3-twist” in the degree-3 integral cohomology of the homotopy quotient (Borel construction) of XX by GG (1)

Then equivariant twisted de Rham cohomology of XX is the de Rham cohomology of Λ((XG))\Lambda (\prec (X \sslash G)) which is both

  • 1-twisted by \nabla and

  • 3-twisted by H 3H_3

(Tu & Xu 2006, Def. 3.10, Freed, Hopkins & Teleman 2007, (3.19), Bunke, Spitzweck & Schick 08, Def. 3.15).


Of 1-Twisted de Rham cohomology

Equivalence with π 1\pi_1-invariant dR cohomology

In the following, let X\mathrm{X} be a smooth manifold, which we assume, without real restriction of generality, to be connected. Therefore we write π 1(X)\pi_1(\mathrm{X}) for its fundamental group, for any fixed choice of basepoint x 0Xx_0 \,\in\, \mathrm{X}.

We write X^X\widehat{\mathrm{X}} \xrightarrow{\;} \mathrm{X} for its universal cover (also canonically a smooth manifold). This is a π 1(X)\pi_1(\mathrm{X})-principal bundle, in particular it has a canonical π 1(X)\pi_1(\mathrm{X})-action by deck transformations.

For \mathcal{L} a complex line bundle over X\mathrm{X} with flat connection \nabla, write

(10)hol :π 1(X) × hol_\nabla \,\colon\, \pi_1(\mathrm{X}) \xrightarrow{\;\;} \mathbb{C}^\times

for the group homomorphism from the fundamental group to the multiplicative group of units ×={0}\mathbb{C}^\times \,=\, \mathbb{C} \setminus \{0\} which is given by sending a smooth curve λ:[0,1]X\lambda \colon [0,1] \to \mathrm{X} (with λ(0)=λ(1)\lambda(0) = \lambda(1)) to its holonomy under the parallel transport with respect to \nabla.


(1-Twisted dR cohomology equivalent to π 1\pi_1-invariant dR cohomology on universal cover)
For \mathcal{L} a complex line bundle over X\mathrm{X} with flat connection \nabla, there is a natural isomorphism between

  1. the \nabla-twisted de Rham cohomology on X\mathrm{X}:

    H (Ω dR (X;),) H^\bullet \Big( \Omega_{dR}^\bullet \big( \mathrm{X} ;\, \mathcal{L} \big) ,\, \nabla \Big)
  2. the untwisted but π 1\pi_1-invariant complex-valued de Rham cohomology on the universal cover X^\widehat{\mathrm{X}}

    H ((Ω dR (X^;)) π 1(X),d) H^\bullet \bigg( \Big( \Omega^\bullet_{dR} \big( \widehat{\mathrm{X}} ;\, \mathbb{C} \big) \Big)^{\pi_1(\mathrm{X})} ,\, \mathrm{d} \bigg)

    where π 1 ( X ) \pi_1(\mathrm{X}) acts on differential forms by pullback along deck transformations combined with multiplication by the holonomy (10) of \nabla:

    π 1(X)×Ω dR (X^;) Ω dR (X^;) ([λ],A) hol (λ)[λ] *(A) \array{ \pi_1(\mathrm{X}) \times \Omega^\bullet_{\mathrm{dR}} \big( \widehat{\mathrm{X}} ;\, \mathbb{C} \big) &\xrightarrow{\phantom{--}}& \Omega^\bullet_{\mathrm{dR}} \big( \widehat{\mathrm{X}} ;\, \mathbb{C} \big) \\ \big( [\lambda] ,\, A \big) &\mapsto& hol_\nabla(\lambda) \cdot [\lambda]^\ast(A) }

A proof is spelled out in this pdf. It proceeds by observing that the bundle and its connection trivialize after pullback to the universal cover X^\widehat{\mathrm{X}}, in fact that the local connection form ω^\widehat{\omega} on X^\widehat{\mathrm{X}} becomes exact (by the Poincaré lemma for simply connected domains, here):

(11)| |C (X^; ×)ω^=dlog,Ω dR 1(X^,), \underset{ \mathclap{\phantom{\vert^{\vert}}} { \ell \in } \atop { C^\infty\big( \widehat{\mathrm{X}} ;\, \mathbb{C}^\times \big) } }{\exists} \;\;\; \widehat{\omega} \;=\; - \mathrm{d} log \ell \,, \;\;\;\; \in \; \Omega^1_{dR} \big( \widehat{\mathrm{X}} ,\, \mathbb{C} \big) \,,

and checking that multiplication by this potential \ell (11) constitutes a isomorphic (bijective) chain map between the cochain complexes in question.


(Hypergeometric integral solutions of KZ-equation)
For the special case that the complex line bundle \mathcal{L} is trivial (so that the flat connection \nabla is represented by a globally defined differential 1-form already on the base manifold X\mathrm{X}) the statement of Prop. (or rather its holomorphic version) plays a central role in the discussion of the “hypergeometric integral construction” of solutions to the Knizhnik-Zamolodchikov equation, where it is applied to the case that X\mathrm{X} is an nn-punctured Riemann sphere (e.g. a trinion). In fact it is so central to this construction that the function \ell (11) which trivializes the connection form on the universal cover and thereby induces the isomorphism in Prop. came to be called the “master function”, in this context (Slinkin & Varchenko 2019, §2.1).

On the other hand, none of the many references listed there really make the Proposition explicit.

Of 3-Twisted de Rham cohomology


If the de Rham complex (Ω (X),d dr)(\Omega^\bullet(X),d_{dr}) is formal, then for HΩ cl 3(X)H \in \Omega_{cl}^3(X) a closed differential 3-form, the HH-twisted de Rham cohomology (Def. ) of XX coincides with its H-cohomology, for any closed 3-form HH.

(Cavalcanti 03, theorem 1.6).


Twist in degree 1

The classical case of twisted de Rham cohomology with twists in degree 1, given by flat connections on flat line bundles and more generally on flat vector bundles), and its equivalence to sheaf cohomology with coefficients in abelian sheaves of flat sections (local systems):


For extensive application, see also the “hypergeometric integral construction” of solutions to the Knizhnik-Zamolodchikov equation.

In this context there is also:

See also at supersymmetric quantum mechanics and see

Twist in degree 3


The concept of H 3H_3-twisted de Rham cohomology was introduced (in discussion of the B-field in string theory), in:

Further discussion (often as the codomain of the twisted Chern character on twisted K-theory):


The generalization of 3-twisted de Rham cohomology to orbifolds (often as the codomain of the twisted equivariant Chern character on twisted equivariant K-theory):

See also

Twist in higher degrees

Discussion of higher-degree twisted de Rham cohomology (often as the Chern character-like image of higher twisted K-theory):

A spectral sequence for higher twisted de Rham cohomology (analgous to the Atiyah-Hirzebruch spectral sequence for twisted K-theory from Atiyah & Segal 2005):

In the broader context of twisted cohomology theory and the Chern-Dold character map:

Braid representations via twisted cohomology of configuration spaces

The “hypergeometric integral” construction of conformal blocks for affine Lie algebra/WZW model-2d CFTs and of more general solutions to the Knizhnik-Zamolodchikov equation, via twisted de Rham cohomology of configuration spaces of points, originates with:

following precursor observations due to:

The proof that for rational levels this construction indeed yields conformal blocks is due to:


See also:

This “hypergeometric” construction uses results on the twisted de Rham cohomology of configuration spaces of points due to:


reviewed in:

  • Yukihito Kawahara, The twisted de Rham cohomology for basic constructions of hyperplane arrangements and its applications, Hokkaido Math. J. 34 2 (2005) 489-505 [[doi:10.14492/hokmj/1285766233]]

Discussion for the special case of level=0=0 (cf. at logarithmic CFT – Examples):

Interpretation of the hypergeometric construction as happening in twisted equivariant differential K-theory, showing that the K-theory classification of D-brane charge and the K-theory classification of topological phases of matter both reflect braid group representations as expected for defect branes and for anyons/topological order, respectively:

Last revised on October 22, 2023 at 10:53:56. See the history of this page for a list of all contributions to it.