Let $V$ be a finite dimensional $\mathbb{K}$-vector space, $\phi : V \to W$ a linear mapping and $U$ a sub-vector space of $W$. Show that the following equation applies: $$\dim(\phi^{-1}(U))=\dim(U\cap\phi(V))+\dim(\ker(\phi))$$


Let $\phi:V\to W$ and $v\mapsto \phi(v)$ and $\Phi :\phi^{-1}(u)\mapsto u$ with $v\in V,\;\color{red}{\phi(v)\in W},\;u\in U$ and $U\subseteq W$

We insert this into the Rank–nullity theorem: $$\dim(\phi^{-1}(U))=\dim(\operatorname{im}(\Phi))+\dim(\ker(\Phi))$$

$\operatorname{im}\phi = \phi(v)$
$\operatorname{im} (\Phi)=\operatorname{im}\phi \cap U$
$\ker(\phi):=\{v\in \phi^{-1}(u):\phi(v)=0\}$
$0\in U\implies \ker \phi =\ker \phi^{-1} \implies \dim(\phi^{-1}(U))=\dim(U\cap\phi(V))+\dim(\ker(\phi))$

Any opinions on this "proof" (edited)?

  • $\begingroup$ It is not correct because \phi^{-1} could be not defined $\endgroup$ – Federico Fallucca Jun 24 at 22:03
  • 1
    $\begingroup$ The first line already fails. What is "$v$" in $v\mapsto\phi(V)$? $\phi^{-1}(U)$ is not an element of $V$, and $U$ is not an element of $W$, so "$\phi^{-1}(U)\mapsto U$" doesn't make sense. $\endgroup$ – user10354138 Jun 24 at 22:08
  • $\begingroup$ $v\in V,\; w\in \phi(v),\;u\in U;\; \phi^{-1}(u)= u$. Does that make more sense? $\endgroup$ – Doesbaddel Jun 24 at 22:11
  • $\begingroup$ Let $\phi:V\to W$ and $v\mapsto \phi(v)$ and $\Phi :\phi^{-1}(u)\mapsto u$ with $v\in V,\; w\in W,\;\phi(v)\in w,\;u\in U$ and $U\subseteq W$ Is this better? $\endgroup$ – Doesbaddel Jun 24 at 22:20
  • $\begingroup$ No it does not make sense. See $\phi(v)$ is an element, so how is $w$ an element of an element $\phi(v)$? An element is not a set. $\endgroup$ – TheLast Cipher Jun 25 at 7:33

You can use the nullity-rank theorem with the map $\Phi: \phi^{-1}(U)\to U$ :




because if $v\in Ker(\Phi)$ then $\Phi(v)=\phi(v)=0$ so $v\in Ker(\phi) $while if $v\in Ker(\phi)$ then $\phi(v)=0\in U$ so $v\in \phi^{-1}(U)$ and you have that $\Phi(v)=\phi(v)=0$ then $v\in Ker(\Phi)$


$im(\Phi)=U\cap im(\phi)$


$\dim(\phi^{-1}(U))=\dim(ker(\phi))+\dim(U\cap im(\phi))$

  • $\begingroup$ @Frederico Falluca: Why do you think $\operatorname{ker}(\Phi)=\operatorname{ker}(\phi)$? I don't think this is true, since $\Phi$ is a restriction of $\phi$ to the subspace $\phi^{-1}(U)$ of $V$. So I can't see a reason why we can say that all elements of $V$ mapped to the identity of $U$ is necessarily an element of $\phi^{-1}(U)$ when we consider that $\phi^{-1}(U) \subseteq V$. $\endgroup$ – TheLast Cipher Jun 25 at 8:19
  • 2
    $\begingroup$ Have I convinced you now? $\endgroup$ – Federico Fallucca Jun 25 at 9:36
  • 1
    $\begingroup$ @Frederico: thanks! $\endgroup$ – TheLast Cipher Jun 25 at 9:56
  • $\begingroup$ Thank you for the additions! $\endgroup$ – Doesbaddel Jun 26 at 8:36

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.