MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

During my research I have recently stumbled upon the problem of finding the relative homotopy sets $\pi_1(U_n,U_n/O_n)$ and $\pi_1(U_{2n},U_{2n}/USp_{2n})$ for $n$ large enough to be in the stable regime where Bott periodicity applies.

The subgroup $U_n/O_n\subset U_n$ contains all elements $g\in U_n$ with $g=g^T$ and the Cartan embedding is \begin{align} U_n/O_n&\hookrightarrow U_n\\\ g\cdot O_n&\mapsto gg^T \end{align} Similarly, the subgroup $U_{2n}/USp_{2n}\subset U_{2n}$ contains all elements $g\in U_{2n}$ with $g=Jg^TJ^{-1}$ and the Cartan embedding is \begin{align} U_{2n}/USp_{2n}&\hookrightarrow U_{2n}\\\ g\cdot USp_{2n}&\mapsto gJg^TJ^{-1}, \end{align} with $J:=\begin{pmatrix}0&1_n\newline-1_n&0\end{pmatrix}$.

In order for my other results to be consistent, I expect the result to be \begin{align} X:=\pi_1(U_n,U_n/O_n)&=0\\\ Y:=\pi_1(U_{2n},U_{2n}/USp_{2n})&=\mathbb{Z}_2 \end{align} However, with the following argument I find that both are trivial: Using the exact sequence of homotopy groups/sets, \begin{align} \pi_1(U_n/O_n)&\stackrel{i_1}{\to}\pi_1(U_n)\to X\to\pi_0(U_n/O_n)\\\ \pi_1(U_{2n}/USp_{2n})&\stackrel{i_2}{\to}\pi_1(U_{2n})\to Y\to\pi_0(U_{2n}/USp_{2n}), \end{align} where $i_{1}$ and $i_2$ are induced by the Cartan embeddings above, there is the bijection \begin{align} X=\pi_1(U_n)/\text{img}(i_1)=\mathbb{Z}/\text{img}(i_1)\\\ Y=\pi_1(U_{2n})/\text{img}(i_2)=\mathbb{Z}/\text{img}(i_2) \end{align} since $\pi_0(U_n/O_n)=\pi_0(U_{2n}/USp_{2n})=0$.

The determinant induces a bijection $\pi_1(U_n)\to\pi_1(U_1)=\mathbb{Z}$ with inverse induced by $U_1\hookrightarrow U_n$ (adding ones to the diagonal), so a representative loop of the class $m\in\pi_1(U_n)=\mathbb{Z}$ is $$\phi:t\mapsto\text{diag}(e^{i2\pi m t},1,\dots,1).$$ This has a preimage under $i_1$, namely the loop $$\phi_1:t\mapsto\text{diag}(e^{i\pi m t},1,\dots,1)\cdot O_n.$$ This is indeed a loop since $\phi_1(0)=(1,1,\dots,1)\cdot O_n$ and $\phi_1(1)=(-1,1,\dots,1)\cdot O_n=(1,1,\dots,1)\cdot O_n$.

Similarly, there is a preimage under $i_2$, namely the loop $$\phi_2:t\mapsto\text{diag}(e^{i\pi m t},1,\dots,1)\cdot USp_{2n}.$$ Again, this is a loop since $\phi_2(0)=(1,1,\dots,1)\cdot USp_{2n}$ and $\phi_2(1)=(-1,1,\dots,1)\cdot USp_{2n}=(1,1,\dots,1)\cdot USp_{2n}$. Thus, both $i_1$ and $i_2$ are surjective and $X=Y=0$.

However, for the expected result to hold, $i_1$ should be surjective in the sequence with $X$ and $\text{img}(i_2)=2\mathbb{Z}$ (even integers only) in the sequence with $Y$. This does not seem to be the case - is my argument flawed or is it true that both relative homotopy sets are in fact trivial?

Thank you!

share|cite|improve this question
up vote 4 down vote accepted

There's a "famous" 24 term sequence of fibrations, involving the spaces of the real and complex K-theory spectra, coming from Bott periodicity. (But apparantly, not famous enough for google to cough up a good reference.) Part of this sequence are homotopy fibration sequences of the form $$ Sp/U \to U/O \xrightarrow{f} U $$ and $$ O/U\to U/Sp\xrightarrow{g} U.$$ ($O$, $U$, and $Sp$ are the infinite versions of the orthogonal, unitary, and unitary-symplectic groups

Assuming that $f$ and $g$ correspond to your Cartan embeddings, you should be able to identify your relative $\pi_1$s with the $\pi_0$s of the homotopy fibers of these fiber sequences. If so, then $$\pi_1(U_n,U_n/O_n) = \pi_1(U,U/O) = \pi_0(Sp/U)=0$$ and $$\pi_1(U_{2n},U_{2n}/USp_{2n}) = \pi_1(U,U/Sp) = \pi_0(O/U) = Z/2,$$ which is the result you were expecting.

If this is all true, then $i_2: \pi_1(U_{2n}/USp_{2n})\to \pi_1(U_{2n})$ isn't really surjective. Maybe $\phi_2$ isn't really a loop: is $diag(-1,1,...,1) \in USp$?

share|cite|improve this answer
You are right, it isn't. Somehow I had myself convinced of the opposite by showing that $i_2(\phi_2(1))\in USp_{2n}$, which is obviously not the right thing to check. In fact, even if it were a loop, then it would not even map to the loop $\phi$ in $U_{2n}$ for odd $m$... Thanks for the great answer! – Pierre Oct 14 '11 at 5:49
While your hint that $i_2$ may not really be surjective helped me to resolve my problem, I'm still puzzled about the way you derived the relative $\pi_1$s. Are you saying that $f$ and $g$ are the projections of a Serre fibration? Then I only know the statement that $\pi_n(E,F)=\pi_n(B)$ for a Serre fibration $p:E\to B$ with fiber $F=p^{-1}(b)$ over the basepoint $b$ of $B$, which does not give me your statement. – Pierre Sep 4 '12 at 12:51

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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