3
$\begingroup$

Consider a manifold $ N $ defined as follows $$ N=\{n\otimes n-m\otimes m:n,m\in S^2,\quad(n,m)=0\}\subset M^{3\times 3}, $$ where $ S^2 $ denotes the two dimensional sphere, $ (\cdot,\cdot) $ represents the inner product of the space $ \mathbb{R}^3 $ and $ M^{3\times 3} $ denotes the set of $ 3\times 3 $ matrices. For a closed curve $ \alpha:S^1\to N $, we can consider the homotopy class of it. Define maps $$ p_1:N\to\mathbb{R}P^2,\quad p_2:N\to\mathbb{R}P^2 $$ as follows. For any $ A\in N $, we can represent $ A=n\otimes n-m\otimes m $ with $ (n,m)\in S^2\times S^2 $ and $ (n,m)=0 $, we define $ p_1(A)=n\otimes n $ and $ p_2(A)=m\otimes m $. I guess that if $ [\alpha] $ corresponds to the element of order $ 4 $ in the fundamental group, then either of $ [p_i(\alpha)] $, $ i=1,2 $ is non-trivial. However I do not know how to prove this or give a conterexample. I have already know the covering space and fundamental group of $ N $ but I do not know how to go on. Can you give me some hints or references?

$\endgroup$
8
  • 1
    $\begingroup$ $\pi_1$ is perhaps a poor choice of name for a projection in a context when you are also talking about the fundamental group... Perhaps $p_1, p_2$? $\endgroup$ Commented Dec 2, 2023 at 2:34
  • $\begingroup$ How do you know $N$ that is a manifold? Also, surely $p_1(A)$ should be the line through $n$ and $p_2(A)$ the line through $m$. $\endgroup$
    – Mark Grant
    Commented Dec 2, 2023 at 8:02
  • $\begingroup$ @MarkGrant Define $M=\{(n,m)\colon n,m\in S^2, n\cdot m=0\}$. It is easy to see that $M$ is the total space of the unit circle bundle of $TS^2$, so $M\cong \Bbb RP^3\cong SO(3)$. Note that $\Bbb Z_2\times\Bbb Z_2$ acts on $M$ freely, where the action is generated by $(n,m)\mapsto (-n,m)$ and $(n,m)\mapsto(n,-m)$. This implies that $M/(\Bbb Z_2\times\Bbb Z_2)\cong N$ by sending the class of $(n,m)$ to $n\otimes n-m\otimes m$. Then $ N $ is a manifold. $\endgroup$ Commented Dec 2, 2023 at 14:31
  • $\begingroup$ Why is it clear that $n \otimes n$ and $m \otimes m$ can be recovered from their difference? $\endgroup$
    – LSpice
    Commented Dec 2, 2023 at 18:14
  • 1
    $\begingroup$ @LSpice You recover {n, -n} as the set of unit norm +1 eigenvectors and {m, -m} as the set of unit norm -1 eigenvectors. This uses the orthogonality assumption. $\endgroup$
    – mme
    Commented Dec 2, 2023 at 19:43

2 Answers 2

7
$\begingroup$

I'm not sure exactly what you were claiming, but here is a correct claim.

$\pi_1(N)$ has six order four elements $\{\pm i, \pm j, \pm k\}$, and there are three natural maps $p_1, p_2, p_3: N \to \Bbb{RP}^2$. Each order-four element of $\pi_1(N)$ survives under two of these maps, and is annihalated under one.

In particular, given any order-four $x \in \pi_1(N)$, at least one of $(p_1)_*x$ or $(p_2)_* x$ are nonzero, which is one way to interpret the question you posed. (It may survive under only one, or it may survive under both. If you use all of $p_1, p_2, p_3$, then all elements survive under exactly two.)


It seems from your comments you understand that your manifold $N$ is diffeomorphic to the quotient of $SO(3)$ by the subgroup $D = \Bbb Z/2 \times \Bbb Z/2$, where this is the subgroup of diagonal matrices in $SO(3)$. I will just think of $N$ as this quotient. In particular, there are three projection maps $p_i: N \to \Bbb{RP}^2$, sending a matrix in $SO(3)$ (up to multiplication by a diagonal matrix in $SO(3)$) to its $i$'th column (up to multiplication by $\pm 1$). The first two of these are exactly what you described; in your formulation, the third is $p_3(n \otimes n^T - m \otimes m^T) = \{o, -o\}$, where $o$ is a unit vector in the kernel of $n \otimes n^T - m \otimes m^T$; equivalently, $o$ is a unit vector perpendicular to both $m$ and $n$.

Consider the homomorphism $p: S^3 \to SO(3)$, given by sending a unit quaternion to the rotation it induces on $\text{span}(i,j,k)$. This is a double cover, and it's easy to verify that $p^{-1}(D) = Q = \{\pm 1, \pm i, \pm j, \pm k\}$ is the order-8 group of unit quaternions; all elements other than $\pm 1$ have order $4$. Thus we have $$N \cong SO(3)/D \cong S^3/p^{-1}(D) = S^3/Q.$$ This diffeomorphism provides an isomorphism $\pi_1(N) = Q$, justifying the claim that $\pi_1(N)$ has six elements of order four.

Now for each map $p_i: N \to \Bbb{RP}^2$, using the relationship between the fundamental group and deck transformations of the universal covering space, we can determine the corresponding map on fundamental groups $\epsilon_i: Q \to \Bbb Z/2$ as follows. Write $P_i: S^3 \to S^2$ for the lift of $p_i$ to the universal covers. For each $q \in Q$, we have $$P_i(q) = P_i(1)(-1)^{\epsilon_i(q)}.$$

Because $i$ acts by conjugation as $\begin{pmatrix} 1 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & -1 \end{pmatrix}$ and similarly for $j, k$, it is then straightforward to compute that $$\epsilon_1(\pm i) = 0, \quad \epsilon_2(\pm i) = \epsilon_3(\pm i) = 1,$$ $$\epsilon_2(\pm j) = 0, \quad \epsilon_1(\pm j) = \epsilon_3(\pm j) = 1,$$ $$\epsilon_3(\pm k) = 0, \quad \epsilon_1(\pm k) = \epsilon_2(\pm k) = 1.$$

$\endgroup$
7
  • $\begingroup$ For what it's worth, the three maps $\epsilon_i: Q \to \Bbb Z/2$ are the three distinct surjective homomorphisms from $Q$ to $\Bbb Z/2$. $\endgroup$
    – mme
    Commented Dec 4, 2023 at 0:01
  • $\begingroup$ I want to ask how to obtain the formula $ P_i(q)=P_i(1)(-1)^{\epsilon_i(q)} $. I and confused about the notation here. I am also have trouble in understanding the explicit loop represented by $ \{i,j,k\} $ and find it difficult to do the computations at the end of your answer. Can you give me more details? $\endgroup$ Commented Dec 7, 2023 at 8:25
  • $\begingroup$ @LuisYankaAnnalisc I could be wrong, but it sounds like you might not have enough familiarity with covering spaces. The claim here is that if $f: X \to Y$ is any map and $\tilde f: \tilde X \to \tilde Y$ is a lift between the universal covers, this lifted map is equivariant under the group of deck transformations: if $\gamma \in \pi_1(X)$, we have $$\tilde f(\gamma \cdot \tilde x) = (f_* \gamma) \cdot \tilde f(\tilde x).$$ That is the content of the formula $P_i(q) = P_i(1) (-1)^{\epsilon_i(q)}.$ $\endgroup$
    – mme
    Commented Dec 7, 2023 at 14:38
  • $\begingroup$ Here, the action of $Q$ on $S^3$ is by left-multiplication by a quaternion --- sending $1$ to $q$ --- and the action of $\Bbb Z/2$ on $S^2$ is $n \cdot \vec v = (-1)^n \cdot \vec v$; that is, we either leave the point alone or negate it. $\endgroup$
    – mme
    Commented Dec 7, 2023 at 14:39
  • $\begingroup$ Unfortunately, I don't know a textbook reference for this formula, but it's not hard to prove directly using the definitions. Next, if $1 \in S^3$ is our chosen basepoint, loops in the quotient $S^3/Q$ corresponding to $q$ arise as the image of paths in $S^3$ running from $1$ to $q$. So, the element $i$ arises from a path from $1$ to $i$ in $S^3$. This is maybe not desirable for your purposes. However, since $i$ generates $\ker(\epsilon_1)$, and represents the fiber of the projection map $p_1: SO(3) \to \Bbb{RP}^2$, as in Neil's answer. Similarly with $j, k$. $\endgroup$
    – mme
    Commented Dec 7, 2023 at 14:43
3
$\begingroup$

I'm not too comfortable with the description of $N$ in terms of matrix tensor products, but the manifold $M/(\mathbb{Z}_2\times\mathbb{Z}_2)$ you describe in your comment is the manifold of pairs of orthogonal lines in $\mathbb{R}^3$, which can be viewed as the projectivized tangent bundle of $\mathbb{R}P^2$. This makes it clear that there is a fibre bundle $$ \mathbb{R}P^1 \to N\stackrel{p_1}{\to} \mathbb{R}P^2, $$ whose long exact sequence in homotopy groups looks like $$ \cdots \to \pi_2(\mathbb{R}P^2) \to \pi_1(\mathbb{R}P^1) \to \pi_1(N) \xrightarrow{(p_1)_\sharp}\pi_1(\mathbb{R}P^2)\to \pi_0(\mathbb{R}P^1) \to \cdots . $$ Since $\mathbb{R}P^1$ is connected the homomorphism induced by $p_1$ from $\pi_1(N)$ to $\pi_1(\mathbb{R}P^2)\cong\mathbb{Z}_2$ is surjective. If you know $\pi_1(N)\cong\mathbb{Z}_4$ by other means, this should imply that the generator is mapped non-trivially.

$\endgroup$
6
  • 3
    $\begingroup$ We have $\pi_1(N) \cong Q$, the 8-element quaternion group. N is sometimes known as the full flag manifold of R^3. The map from pi_1(RP^1) sends a generator to one of the order 4 elements. $\endgroup$
    – mme
    Commented Dec 2, 2023 at 19:42
  • $\begingroup$ @mme Interesting. Then doesn't this answer the OP's question in the negative? $\endgroup$
    – Mark Grant
    Commented Dec 2, 2023 at 20:29
  • $\begingroup$ Good point! ${}$ $\endgroup$
    – mme
    Commented Dec 2, 2023 at 22:48
  • $\begingroup$ @MarkGrant So what is the final answer for this question. By the way, is my explanation that $ N $ is a manifold right? $\endgroup$ Commented Dec 3, 2023 at 3:13
  • 1
    $\begingroup$ @MarkGrant Actually I am considering some subsets of the symmetric matrices. For any symmetric matrix $ A $, we have the representation $ A=\lambda_1 n\otimes n+\lambda_2 m\otimes m+\lambda_3 p\otimes p $, where $ (n,m,p) $ forms an orthonormal basis of $ \mathbb{R}^3 $. Here $ N $ is a spectial case that the three eigenvalues are $ 1,-1 $ and $ 0 $. $\endgroup$ Commented Dec 3, 2023 at 8:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .