Let $H$ be a finite group. We write ${{\mathbb{C}}}^{*n}$ for the $n$-dimensional complex torus $({{\mathbb{C}}}^*)^n$. We have a short exact sequence $$ 0\to {{\mathbb{Z}}}^n\to {{\mathbb{C}}}^n\to{{\mathbb{C}}}^{*n}\to 1,$$ which gives a connecting isomorphism $$ \Delta\colon {\rm Hom}(H,{{\mathbb{C}}}^{*n})\to H^2(H,{{\mathbb{Z}}}^n). $$

Let $\varphi\colon H\to {{\mathbb{C}}}^{*n}$ be a homomorphism. We obtain a central group extension $$ 0\to {{\mathbb{Z}}}^n\to E(\varphi)\to H\to 1$$ corresponding to $\Delta(\varphi)\in H^2(H,{{\mathbb{Z}}}^n)$. Namely, $$ E(\varphi)=H\times_{{{\mathbb{C}}}^{*n},\varphi} {{\mathbb{C}}}^n.$$

Now let $\sigma\in {\rm Aut}({{\mathbb{C}}})$ be an automorphism (not necessarily continuous). We define an action of $\sigma$ on ${{\mathbb{C}}}^{*n}$ by $\sigma(z_1,\dots,z_n)=(\sigma(z_1),\dots,\sigma(z_n))$. We obtain a new homomorphism $$ \sigma\varphi:=\sigma\circ\varphi\colon\ H\to {{\mathbb{C}}}^{*n}$$ and a new extension $$ 0\to {{\mathbb{Z}}}^n\to E(\sigma\varphi)\to H\to 1,$$ which is not isomorphic to $E(\varphi)$ when $\sigma\varphi\neq\varphi$.

Question. Can it happen that $E(\sigma\varphi)$ is not isomorphic to $E(\varphi)$ as groups (and not only as group extensions)?

Motivation: I am trying to construct a homogeneous space $G/H$ of a connected linear ${{\mathbb{C}}}$-group $G$, such that the topological fundamental groups of $\tau(G/H)$ and of $G/H$ are not isomorphic.


1 Answer 1


Here's an example where they are not isomorphic, with $n=1$.

Let $k$ be the order of the image of $\varphi$, which is also the image of $\sigma\phi$ (since $\mathbf{C}^*$ has a single subgroup of each given order). Then $\varphi$ can be viewed, after some fixed identification, as the fibre product $H\times_{(\mathbf{Z}/k\mathbf{Z})^n,\varphi}\mathbf{Z}^n$. An element $\sigma$ is just the choice of some element in $\ell\in (\mathbf{Z}/k\mathbf{Z})^*$, and this changes $\varphi$ to $\varphi^\ell:h\mapsto\varphi(h)^\ell$, .

Now consider the semidirect product $H=(\mathbf{Z}/5\mathbf{Z})\ltimes_4 (\mathbf{Z}/11\mathbf{Z})$, here the notation means that the generator of order 5 acts by multiplication by 4 on the right (which is indeed of order 5 modulo 11), and $\varphi$ the canonical projection to $\mathbf{Z}/5\mathbf{Z}$, and $\sigma_c$ the automorphism of $\mathbf{Z}/5\mathbf{Z}$ given by multiplication by $c\in(\mathbf{Z}/5\mathbf{Z})^*$. Then $E(\sigma_c\varphi)=\mathbf{Z}\ltimes_{4^{c^{-1}}}(\mathbf{Z}/11\mathbf{Z})$, where the generator of $\mathbf{Z}$ acts by multiplication by $4^{c^{-1}}$ (beware that $c^{-1}$ is modulo 5 but $4^{c^{-1}}$ is modulo 11). Then $E(\sigma_1\varphi)=E(\varphi)$ and $E(\sigma_2\varphi)$ are not isomorphic ($4^{2^{-1}}$ is then equal to $-2$). The argument if that these groups both have a unique normal subgroup of order 11, and a unique homomorphism $f$ onto $\mathbf{Z}$ up to sign, and that elements of $f^{-1}(\{-1,1\})$ act on the normal subgroup by multiplication by $\pm 4$ in $E(\varphi)$, by multiplication by $\pm 2$ in $E(\sigma_2\varphi)$, and the automorphisms of $\mathbf{Z}/11\mathbf{Z}$ of multiplication by $\pm 2$ and by $\pm 4$ are not conjugate.

In your original language, $\varphi$ is the homomorphism to $\mathbf{C}^*$ mapping the generator of order 5 to $\omega_5=\exp(2i\pi/5)$ and mapping the generator of order 11 to 1, and $\sigma$ is any field automorphism of $\mathbf{C}$ mapping $\omega_5$ to $\omega_5^2$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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