Let $V$ be a finite-dimensional vector space over $K$, and let $V^* := \{ f : V \to K : \mbox{ f is linear } \}$ be the space of linear functionals, the so called dual space of $V$. Also let $U \le V^*$ be a subspace of the dual space, set $$ W = \{ v \in V : f(v) = 0 \mbox{ for all } f \in U \}. $$ Is there an easy way to show that $W = 0$ implies $U = V^*$?


First , dual space $V^*$ is also a finite linear space , so is subspace $U$.

To be precise , $V^* \cong V$ . Let's assume that $U \not= V^*$

Lemma: For every basis $\{f_1 , ...,f_n\}$ of dual space $V^*$ , there is a corresponding basis $\{v_1 , ...,v_n\} $ in $V$ satisfies $f_i({v_j})=\delta_{ij} \text{ where } i,j\in \{1,..,n\}$

Proof: Consider $V^{**}$ , the dual space of $V^*$. We can find the dual basis of $\{f_1 , ...,f_n\}$ in $V^{**}$, namely , $\{\bar v_1 , ...,\bar v_n\} $ with $ \bar v_i (f_j)=\delta_{ij}$ . By the equivalence $V^{**} \cong V^* \cong V$ , there is a isomorphism $\pi : V^{**} \rightarrow V$ . Let $v_i = \pi (\bar v_i)$ , the set $\{v_1 , ...,v_n\} $ is actually the basis of $V$ and satisfies the condition $f_i (v_j)=\delta_{ij}$

$U$ have a basis $\{f_1 , ...,f_n\}$ corresponding to the elements $\{v_1 , ...,v_n\} $ in $V$

$A=Span\{v_1 , ...,v_n\}$ is a subspace of $V$

Extent $\{f_1 , ...,f_n\}$ to get the basis of $V^*$ , namely , $\{f_1,...,f_n,f_{n+1},...,f_{n+m}\}$

By lemma, we know that the corresponding elements $\{v_1,...,v_n,v_{n+1},...v_{n+m}\}$ is the basis of $V$ and $f_i({v_j})=\delta_{ij} \text{ where } i,j\in \{1,..,n+m\}$

Therefore, $span\{v_{n+1},...v_{n+m}\}\subset Ker(f_i) $ for $i\in \{1,...,n\}$

This result induces that $span\{v_{n+1},...v_{n+m}\}\subset W$ , contradicting to the condition $W=0$

  • $\begingroup$ "It is obvious that $p \in W$" - sorry but I do not see that. $p \in W$ is equivalent to $f_i(p) = 0$ for $i = 1, \ldots, n$, but all we know is $p \notin A$ and I do not see in what sense the $u_i$'s and the $f_i$'s are related to enforce that this implies $p \in W$? Surely as $V \cong V^{\ast}$ there exists a correspondence, but this is not canonical and depends on the choice of a basis, so maybe you should state your correspondence more explicitly so I can see that it is indeed obvious? Otherwise it seems to me like a lot of "hand-waving"... $\endgroup$ – StefanH Oct 11 '15 at 14:52
  • $\begingroup$ @Stefan I edit my answer. The key is using the corresponding condition. $\endgroup$ – Yiran Oct 11 '15 at 15:14
  • $\begingroup$ But I do not see why they should hold? If I have some $f : V \to K$ linear, why should there exists some vector $u_1 \in V$ and a possible extension to a basis $u_1, \ldots, u_m$ such that $f(u_1) = 1, f(u_i) = 0$, $i = 2,\ldots, m$? $\endgroup$ – StefanH Oct 11 '15 at 15:25
  • $\begingroup$ Oh , indeed we can firstly extent $\{f_1 , ...,f_n\}$ to the basis $\{f_1 , ...,f_n,f_{n+1},...,f_{n+m}\}$ of $V^*$. Then $f_i(v_j) = \delta_{ij}$ holds for $ i,j\in {1,...,n+m}$. The corresponding elements $\{v_1 , ...,v_n,v_{n+1},...,v_{n+m}\}$ is exactly basis of $V$ $\endgroup$ – Yiran Oct 11 '15 at 15:30
  • $\begingroup$ I am not sure that I get your points. What is the basis of dual space? How the basis element look like? Think about it. Then I believe you can find the way. $\endgroup$ – Yiran Oct 11 '15 at 15:40

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.