While doing some exercises in Lie groups, I see that the Lorentz group $O(1,3)$ has four connected components and $\pi_0(O(1,3))$ is the Klein four-group $\mathbb{Z}/2 \times \mathbb{Z}/2$. Not only that, but I can find explicit elements representing each connected component $\{1,a,b,ab\}$ which form a subgroup isomorphic to $\pi_0$.

My question then: is it true that for any Lie group $G$, the group $\pi_0(G)$ embeds as a subgroup of $G$? I believe it would follow that any such embedding would take distinct elements of $\pi_0$ to distinct components of $G$.

  • 3
    $\begingroup$ Not sure about the general case , but $\{\pm I_2\}\subset\mathrm{GL}_2(\mathbb{R})$ lie in the same component. There are two distinct questions: the existence of sections of $G\to\pi_0(G)$ and the existence of subgoups of $G$ isomorphic to $\pi_0(G)$. $\endgroup$ Jan 30, 2020 at 2:48
  • 3
    $\begingroup$ In the case where $G$ is the normaliser of a maximal torus in a reductive Lie group, your question is asking about a lift of the normaliser of the Weyl group, which does not exist in general. (The easiest example is when the ambient group is $\operatorname{SL}_2$, in which case the normaliser is a non-split extension of $\operatorname{GL}_1$ by $\mathbb Z/2\mathbb Z$.) @MikeMiller's example is the compact form of this example. $\endgroup$
    – LSpice
    Jan 30, 2020 at 3:04
  • $\begingroup$ See also mathoverflow.net/questions/150949/… (already served up by MO as a related question), which offers a weaker version of this property. $\endgroup$
    – LSpice
    Jan 30, 2020 at 3:06
  • 1
    $\begingroup$ The claim "I believe..." is not true. Consider the quaternion group $Q$ (of order 8) and $S$ the 1-dimensional circle group; let $z_Q$ and $z_S$ be their unique elements of order $2$ and $z=(z_Q,z_S)\in Q\times S$. Define $G=(Q\times S)/\langle z\rangle$. Then the quotient $C_2\times C_2$ indeed embeds into $G$, but cannot be mapped as a section (i.e., for each embedding $C_2\times C_2\to G$ there exists two distinct elements mapping into the same component). $\endgroup$
    – YCor
    Jan 30, 2020 at 8:43

1 Answer 1


No. The first example that comes to mind is $\text{Pin}(2) \subset S^3$, given as $S^1 \cup jS^1$. This group has two components, but every element of $jS^1$ squares to $-1 \in S^1$. Thus every element of the non-identity component has order 4.

So there is no section of the map $\text{Pin}(2) \to \Bbb Z/2$ which sends the non-identity component to 1.


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.