6
$\begingroup$

$\DeclareMathOperator\SU{SU}$Let $G$ be a simple and simply-connected Lie group and $H\neq 1$ be a simple and simply connected subgroup, is it true that $\pi_3(G/H)=0$? If not, what is a counter-example?

To provide some background, I know that for a "block-diagonal" embedding $\SU(M)\rightarrow \SU(N)$ ($M<N$) the induced map $\pi_3(\SU(M))=\mathbb{Z}\rightarrow \pi_3(\SU(N))=\mathbb{Z}$ is multiplication by $1$ and therefore, from the sequence:

$$\pi_3(\SU(M))=\mathbb{Z}\rightarrow \pi_3(\SU(N))=\mathbb{Z}\rightarrow \pi_3(\SU(N)/\SU(M))\rightarrow 0$$

one concludes that $\pi_3(\SU(N)/\SU(M))=0$. For this example, it is also the case that $\pi_1(H)=\pi_1(\SU(M))=0$. So I am interested in knowing if this condition implies the conclusion more generally.

$\endgroup$
8
  • 6
    $\begingroup$ $G=S^3$, $H=1$ is a counterexample. $\endgroup$ Commented Dec 12, 2023 at 19:17
  • $\begingroup$ Thanks for the counterexample, but in fact, I am considering $H\neq 1$. $\endgroup$
    – Davi Costa
    Commented Dec 12, 2023 at 19:24
  • $\begingroup$ Per Igor's comment, I guess you want $H$ nontrivial so that $\pi_3(H) = \mathbb Z$. $\endgroup$ Commented Dec 12, 2023 at 19:24
  • 1
    $\begingroup$ Specifically, Mimura-Toda construct for every root of $G$ a certain Lie group monomorphism $S^3\to G$ and show that it induces a $\pi_3$-isomorphism if the root can be taken to the dominant root by an element of the Weyl group. $\endgroup$ Commented Dec 12, 2023 at 21:03
  • 3
    $\begingroup$ Have you considered the diagonal embedding of, say, $SU(2)$ in $SU(4)$? $\endgroup$
    – Tyrone
    Commented Dec 13, 2023 at 10:11

1 Answer 1

8
$\begingroup$

There are embeddings $i_1,i_2$ of $SU(2)\cong S^3$ into $SU(4)$ as the top-left and respectively bottom-right $2\times2$ blocks. As you point out, both $i_1$ and $i_2$ induce isomorphisms on $\pi_3$. Since $i_1,i_2$ are conjugate embeddings, they induce the same isomorphisms.

Now let $j:SU(2)\rightarrow SU(4)$ be the diagonal subgroup inclusion. Notice that this inclusion can be factored $$j:SU(2)\xrightarrow{\Delta}SU(2)\times SU(2)\xrightarrow{i_1\times i_2}SU(4)\times SU(4)\xrightarrow{\mu}SU(4)$$ where $\mu$ is the multiplication on $SU(4)$. To understand the action of $j$ on $\pi_3$ it is enough by the Hurewicz Theorem to understand its action on $H_3$, and hence on $H^3$ by the Universal Coefficient Theorem.

To this end let $x$ generate $H^3(SU(4))$. Then $i_1^*x=i_2^*x$ are generators of $H^3(SU(2))$. Call this image $y$. We have $$\mu^*x=x\otimes 1+1\otimes x$$ and hence $$j^*x=\Delta^*(i_1^*x\otimes 1+1\otimes i_2^*x)=i_1^*x+i_2^*x=2\cdot y.$$ That is, $j^*$ induces multiplication by $2$ on $H^3$.

We conclude that $j:\pi_3(SU(2))\rightarrow \pi_3(SU(4))$ is multiplication by $2$ (I'm being sloppy with signs because I want to use $i_1$ to identify each group with $\mathbb{Z}$). Thus if $X$ is the quotient $SU(4)/j(SU(2))$, then $$\pi_3(X)\cong \mathbb{Z}/2.$$

Clearly this construction can be generalised by considering the diagonal embeddings of $SU(2)$ into $SU(2\cdot n)$ for any $n\geq2$. Another generalisation would consider the diagonal embeddings of $Sp(1)$ into $Sp(n)$.

To tie the room together, notice that under the exceptional isomorphism $$Spin(4)\cong Spin(3)\times Spin(3),$$ the natural inclusion $Spin(3)\hookrightarrow Spin(4)$ corresponds to the diagonal subgroup of $Spin(3)\times Spin(3)$. The group $Spin(4)$ is not simple, but the method above explains why $$\pi_3(Spin(n)/Spin(3))\cong\mathbb{Z}/2$$ for any $n\geq5$. Of course, $$Spin(n)/Spin(3)\cong V_{n-3}(\mathbb{R}^n)$$ for $n\geq5$ is just the Stiefel manifold of $(n-3)$-frames in $\mathbb{R}^n$.

$\endgroup$
2
  • $\begingroup$ Quick check: where you write $\mathbb{Z}/2$, do you want $\mathbb{Z}/2\mathbb{Z}$? Or is this a standard shortening that I'm not familiar with? $\endgroup$ Commented Dec 16, 2023 at 20:17
  • 2
    $\begingroup$ @StevenStadnicki $\mathbb{Z}/2$ is the group you think it is. I believe this is fairly standard notation (at least in the area of mathematics in which I work). $\endgroup$
    – Tyrone
    Commented Dec 16, 2023 at 20:57

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