If $F$ is $\mathbb{R}, \mathbb{C}$, or $ \mathbb{H}$, the Grassmannian manifold $G_k(\textbf F^n)$ is the space of all $k$ dimensional subspaces of the $n$ dimensional vector space $F^n$. The Stiefel manifold $V_k(\textbf F^n)$ is the set of $k$-tuples representing $k$ orthonormal vectors in $F^n$. In other words $$ V_k(\textbf F^n) = \{A\in\text{Mat}_{n\times k}(\textbf F^n)|A^\ast A = I_{k\times k}\}.$$

There is a natural projection $V_k(\textbf F^n)\longrightarrow G_k(\textbf F^n)$ sending a $k$-tuple to the $k$-dimensional subspace that it spans. The fiber of this projection over each point is all $k$-tuples that live in a fixed $k$-dimensional subspace of $\textbf F^n$, which can be thought of $V_k(\textbf F^k) = O(k,\textbf F)$.

So we have the fibrations $$ O(k,\textbf F)\rightarrow V_k(\textbf F^n)\longrightarrow G_k(\textbf F^n).$$

When $k=1$, these reduce to the Hopf fibrations \begin{eqnarray*}S^0&\rightarrow& S^{n-1}\longrightarrow \mathbb{R}P^{n-1}\\ S^1&\rightarrow& S^{2n-1}\longrightarrow \mathbb{C}P^{n-1}\\ S^3&\rightarrow& S^{4n-1}\longrightarrow \mathbb{H}P^{n-1}\end{eqnarray*}

If each of the spheres $S^{n-1}, S^{2n-1},$ and $S^{4n-1}$ are given the round metric, there are "natural" metrics on $\mathbb{R}P^{n}, \mathbb{C}P^{n},$ and $\mathbb{H}P^{n}$, respectively, defined as the metrics that make these submersions Riemannian submersions. When $\textbf{F} = \mathbb{R}$, this is the constant curvature 1 metric on $\mathbb{R}P^n$ and when $\textbf F=\mathbb{C}$, this is the Fubini-Study metric on $\textbf{C}P^n$.

Question: Are there "natural" metrics on $V_k(\textbf{F}^n)$ which give generalization of this? That is, is there a generalization of the Fubini-study metric to $G_k(\textbf{F}^n)$. If so, (where) can I find out more information about these metrics? If not, why won't it work?


There is no need to proceed via the Stiefel-Manifolds. You can directly realize the Grassmannians as homogeneous spaces. In the real case, this takes the form $G(k,n)=SO(n)/S(O(k)\times O(n-k))$. This corresponds to a so-called symmetric decomposition of the Lie algebra $\mathfrak{so}(n)$, thus making $G(k,n)$ into a compact symmetric space. (The Riemannian metric is induced from the restriction of the Killing form of $\mathfrak{so}(n)$ to the orthocomplement of $\mathfrak{o}(k)\times\mathfrak{o}(n-k)$). The resulting metric is relatively simple, but does not have constant curvature. There is a lot known about such spaces, "Riemannian symmetric spaces" is the right key-word to start looking for. The picture for the complex and quaternionic fields is quite similar, you just have to replace $SO$ by $SU$ respectively $Sp$ (quaternionically unitary group).

  • $\begingroup$ Does the cone over a Grassmannian $G(k,n) = SO(n)/S(O(k)\times O(n-k))$ manifold admit an Einstein metric? $\endgroup$ – Bilateral Jul 31 '15 at 22:12
  • 1
    $\begingroup$ @Bilateral : I don't know the answer to this quesiton. Let me just remark that there is a big difference between question on the $SO(n)$-invariant geometry of $G(k,n)$ (which can esssentially be studied by linear algebra respectively finite dimensional representation theory), and questions about existence of other metrics (or geometric structures) on $G(k,n)$. For the latter, it is much more difficult to exploit the fact that you are dealing with a homgeneous space. $\endgroup$ – Andreas Cap Aug 2 '15 at 13:11
  • $\begingroup$ Thanks, I understand. Do you know if someone has checked if if the $SO(n)$-invariant metric on $G(k,n)$ is Einstein on the corresponding cone over $G(k,n)$? (Since as you said the $SO(n)$-invariant metric on $G(k,n)$ is not Einstein). $\endgroup$ – Bilateral Aug 2 '15 at 18:45
  • $\begingroup$ @Bilateral: l am pretty sure that the $SO(n)$-invariant metric on $G(k,n)$ is Einstein. What I said is that it does not have constant curvature (which refers to sectional curvature). The argument that it is Einstein should be as follows: The Ricci tensor of the $SO(n)$-invariant metric itself defines a (possibly degenerate) bundle metric on $TG(k,n)$. Such a metric corresponds to a bilinear form on the space of $k\times (n-k)$-Matrices which is invariant under the natural action of $O(k)\times O(n-k)$ (by multiplication from both sides). But there is only one such form up to constant factors. $\endgroup$ – Andreas Cap Aug 3 '15 at 7:56

It is an standard fact that the space of real grassmanian $G(k,n)$ is homeomorphic to all projections in $M_{n}(\mathbb{R})$ with trace $k$.(A projection is a matrix with $A=A^{tr}=A^{2}$. The inner product of $M_{n}(\mathbb{R})\simeq \mathbb{R}^{n^{2}}$ is $trace(AB^{tr})$

This inner product is invariant under the action of $O(n)$. So actually O(n) is acting on G(k,n) isometricaaly and transitively. The action is transitive since every two projections with the same rank are unitary equivalent, a standard fact in linear algebra. It is also mentioned in "k theory and C* algebra, A freilndly approach" By Wegg Olsen. this would implies that the metric has constant curvature.


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.