Let $V$ be a vector space.

Suppose we are given some upper semi-continuous map $\pi:V\rightarrow \bigcup_{k\le d} Gr(V,d)$, i.e, for any $x\in V$ we specify some subspace $\pi(x) \subseteq V$ of dimension no more than $d$, and these vary in a upper semi-continuous manner.

Must there exist a function $\eta$ from $V$ to the grassmannian $Gr(V,d)$ which is continuous/smooth, so that $\eta(x)$ always contains $\pi(x)$ as a subset?



Take $V=\mathbb{R}^2$, $d=1$ and define $\pi$ by $\pi(v)=\text{span}(v)$ if $\|v\|=1$, $\pi(v)=\{0\}$ otherwise. Then $\pi$ is usc and a continuous $\eta$ doesn't exist, as the image of $S^1$ is a nontrivial cocycle in $\text{Gr}(V,1)$.

Note also that $\pi(v)=\text{span}(v)$ for all $v$ is an example of a lsc $\pi$ with no $\eta$.

However, if $\pi:V\to\text{Gr}(V,k)$ is actually continuous for some $k<d$, then such $\eta$ does exist. To see this it is enough to assume $d=k+1$. Then use the contractibility of $V$ to find a no trivial section to the vector bundle $E$ over $V$ given by $E_v=\pi(v)^\perp$ and add it to $\pi$.


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.