Let $G$ be a finite group and let $D(g)$ be a projective representation of $G$ i.e. \begin{equation} D(g) D(h) = e^{i \omega(g,h)} D(gh) \end{equation} These can be classified by the equivalence relation $\omega(g,h) \sim \omega(g,h)+\theta(g)+ \theta(h) - \theta(gh)$ subject to the condition $\omega(g,h)+\omega(gh,l)-\omega(h,l)-\omega(g,hl) = 0$. The distinct equivalent classes of projective representations are labeled by elements of $H^2(G,U(1))$. I know that for every such finite group, there is atleast one finite covering group C with the property that every projective representation of G can be lifted to an ordinary representation of C [1].

My question is about the relationship between the irreducible representations (irreps) of C and the group $H^2(G,U(1))$. Specifically, is the following statement true?:

Every irrep $\Gamma_i$ of C can be associated an element of $\nu \in H^2(G,U(1))$ like $\Gamma^\nu_i$. The group property of $H^2(G,U(1))$ is reflected in the Clebsch-Gordan decomposition of tensor product of irreps of C: \begin{equation} \Gamma^\nu_i \otimes \Gamma^\mu_j \cong \bigoplus_k \Gamma^{\nu+\mu}_k \end{equation}

I have noticed that this is true for all cases I have seen when $ H^2(G,U(1)) \cong \mathbb{Z}_2$ like $G = \mathbb{Z}_2 \times \mathbb{Z}_2$, $C = D_8$ but I am unsure if this is true in general.

[1] https://en.wikipedia.org/wiki/Schur_multiplier#Relation_to_projective_representations

PS: I am a physicist and my interest in the above question comes from its use in condensed matter physics.

  • 1
    $\begingroup$ I do not understand your question at all. Could you explain it in the case $H^2(G,U(1))=0$? There are a number of finite groups with that property. $\endgroup$
    – abx
    Jul 22 '18 at 18:50
  • $\begingroup$ Consider $G= \mathbb{Z}_2$ which has $H^2(G,U(1)) =0$. In this case, the covering group is $\mathbb{Z}_2$ itself. All irreducible representations of $\mathbb{Z}_2$ are labeled by the trivial element of $H^2(G,U(1))$. $\endgroup$
    – sawd
    Jul 22 '18 at 19:10

The answer to your question is Yes. Consider your covering group $C$ as a central extension: $$1 \to N \to C \to G \to 1$$ and suppose it is given by a 2-cocycle $\alpha \in H^2(G, N)$. Then for any homomorphism $f: N \to \mathbb{C}^\times$, we get a 2-cocycle $f \circ \alpha \in H^2(G,\mathbb{C}^\times)$ by composition of maps, and this yields a homomorphism $$\operatorname{Hom}(N,\mathbb{C}^\times) \to H^2(G,\mathbb{C}^\times).$$ You are assuming that $C$ is a "representation group", which amounts to the property that this homomorphism is an isomorphism. Thus, any projective representation of $G$ lands in a commutative diagram $$\require{AMScd} \begin{CD} 1 @>>> N @>>> C @>>> G @>>> 1 \\ @. @VVV @VVV @VVV \\ 1 @>>> \mathbb{C}^\times @>>> GL(V) @>>> PGL(V) @>>> 1 \end{CD}$$ The 2-cocycle for your projective representation corresponds to the 1-dimensional representation of $N$ given by the leftmost vertical arrow. Any tensor product of representations of $C$ induces a tensor product of 1-dimensional representations of $N$. Tensor product of 1-dimensional representations of $N$ yields the group law on $\operatorname{Hom}(N,\mathbb{C}^\times)$, and by our assumption, this is the group law on $H^2(G, \mathbb{C}^\times)$.

I found all the statements I needed in this exposition by Mendonca.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.