1
$\begingroup$

Let $f:V\to W$ be a linear transformation from a vector space $V$ to a vector space $W$. Suppose that $b\in W$ satisfies $\phi(b)=0$ for all $\phi\in\ker(f^*)$. Show that there exists $x\in V$ such that $f(x)=b$.

I have to confess that this is part of a homework problem that was handed to me before christmas. I spent the last two weeks think on this problem and I am desperate.. I learned that if $f$ is injective, $f^*$ is surjective and I expect this to be the dual statement to this but when I try to mimic the proof, I fail. Any help would be nice..

$\endgroup$
  • 3
    $\begingroup$ HINT: Look up "Fredholm alternative", the finite dimensional version. $\endgroup$ – Giuseppe Negro Jan 8 at 13:45
  • $\begingroup$ You saved my day! Thank you SO much!! $\endgroup$ – user625722 Jan 8 at 13:56
2
$\begingroup$

This answer does not assume finite dimensionality of $V$ or $W$ (so it is a generalized version of the Fredholm alternative). That is, I claim that for any linear map $f:V\to W$, $$\bigcap_{\phi\in\ker (f^*)}\ker\phi=\operatorname{im}f.$$

We have the following exact sequence of vector spaces: $$\{0\}\to \ker f \hookrightarrow V \overset{f}{\longrightarrow} W\twoheadrightarrow \operatorname{coim}f\to \{0\},$$ where $\operatorname{coim}f$ is the coimage $W/\operatorname{im} f$. Dualizing this exact sequence, we obtain the exact sequence $$\{0\}\to (\operatorname{coim}f)^*\hookrightarrow W^*\overset{f^*}{\longrightarrow} V^*\twoheadrightarrow (\ker f)^*\to\{0\},$$ where $(\operatorname{coim}f)^*$ is considered a subspace of $W^*$ via the identification that sends ${\psi} \in (\operatorname{coim}f)^*$ to $\hat\psi\in W^*$ such that $$\hat\psi(w)={\psi}(w+\operatorname{im}f)\ \forall w\in W.$$ Similarly, $(\ker f)^*$ is a quotient of $V^*$ via the restriction map sending $\sigma\in V^*$ to $\sigma|_{\ker f}\in (\ker f)^*$ (precisely, $(\ker f)^*\cong V^*/(\operatorname{cokr} f)^*$, where $\operatorname{cokr}f$ is the cokernel $V/\ker f$). By the exactness, we get $$\ker (f^*)=(\operatorname{coim}f)^*.$$ That is, if $w\notin \operatorname{im}f$, then $w+\operatorname{im}f$ is non-zero in $\operatorname{coim}W$. Therefore, there exists ${\phi}_w\in (\operatorname{coim} W)^*$ that does not vanish at $w+\operatorname{im} f$. That is, $\hat\phi_w(w)\ne 0$. Therefore, $w\notin \bigcap_{\phi\in\ker(f^*)}\ker\phi$. This shows that $\bigcap_{\phi\in\ker(f^*)}\ker\phi\subseteq\operatorname{im}f$. It is also obvious that $\bigcap_{\phi\in\ker(f^*)}\ker\phi\supseteq\operatorname{im}f$. The claim is now evident.

$\endgroup$
  • $\begingroup$ This is a little beyond the scope of my course but a very nice answer. Thank you! $\endgroup$ – user625722 Jan 8 at 14:58

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.