5
$\begingroup$

In most references about Grassmann manifold, we usually introduce the following map: suppose that $S$ is a fixed subspace of codimension $k$, and let $\Omega_S \subset Gr(k,\mathbb{K})$ be all subspaces $W$ such that $W \cap S = \{0\}$. $\Omega_S$ is an open set.

Fix some $U \in \Omega_S$. One can show that $W \in \Omega_S$ if and only if there exists some linear map $\beta_W \in Hom(U,S)$ such that $W = \{x + \beta_W x : x \in U \}$. From there, one can easily check that there exists a bijection $\Omega_S \simeq Hom(U,S)$.

I'm interested in show that this bijection is actually an homeomorphism of topological spaces!

So far, I haven't been able to make a lot about this claim. What bugs me is that the topologies at stake are not necessarily intuitive to work with: $Gr(r, \mathbb{K})$ is endowed with the quotient topology, and $Hom(U,S)$ is really just an euclidean space, so might just take some random norm/distance.

Any help greatly appreciated !

EDIT : Look here Lee's page 22 for more informations.

$\endgroup$
1
  • $\begingroup$ But Lee uses these bijections to define the topology on the Graßmannian, so you should be more explicit about which topology you choose on it. Since you said 'quotient topology' I assume you mean the construction as a homogeneous space en.m.wikipedia.org/wiki/… ? $\endgroup$ – Ben Oct 31 '18 at 9:55
2
$\begingroup$

This is becoming old, but anyway, I was able to find an answer myself and never posted it here, so here we go. Define $$ t : \Omega_S \to Hom(U, S). $$ as above. Let $\pi : V_0^k \to Gr(k, \mathbb{K})$ where $V_0^k := \{ (v_1, \dots, v_k) : v_i\text{'s linearly independent} \}$, with $$ \pi(v_1, \dots, v_k) = Vect(v_1, \dots, v_k), $$defines the quotient topology over $Gr(k, \mathbb{K})$, in which $\Omega_S$ is open. By definition, it suffices to show that $$ t' : \pi^{-1}(\Omega_S) \to Hom(U,S) $$ in continuous, where $t'(v_1, \dots, v_k) = -pr_{S/Vect(v_1, \dots, v_k)}$ ($pr_{A/B}$ is the projection on $A$ w.r.t. to $B$). Let $w_1, \dots, w_{n-k}$ be a base for $S$. Define the (clearly) continuous application $$A : \pi^{-1}(\Omega_S) \ni (v_1, \dots, v_k) \mapsto [v_1, \dots, v_k, w_1, \dots, w_{n-k} ] \in \mathcal{M}_{n}(\mathbb{K})$$ (i.e. the matrix with columns $v_i's$ and $w_i's$). Then $$ t'(v_1, \dots, v_k) = \left( x \mapsto \sum_{i=1}^{n-k} \left\langle x, (A(v_1, \dots, v_k)A^t(v_1, \dots, v_k))^{-1}w_i \right\rangle w_i\right) $$which is easily seen to be continuous.

$\endgroup$

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.