Let $A\subset X$ (A is closed and linear subspace). And there is a point $x_0\in X$ such that $\inf_{y\in A}\|y-x_0\|= \epsilon$ ($\epsilon>0$). I want to prove that there is a functional $f\in X^*$ such that $f(x_0)=1$ and $\|f\|=1/\epsilon$.

I think I have to use Hahn-Banach for this. I'm going to re-write vector $x\in A_1=span(A,x_0)$ as $x=y+\lambda x_0$ ($y\in A, \lambda\in R$). I'm not sure how to go about using Hahn-Banach to prove the existence of the functional and that $f(x_0)=1$.

  • 1
    $\begingroup$ $x_o \in X$ and $y \in A$ I assume ? $\endgroup$ – Rebellos Feb 25 at 21:25
  • $\begingroup$ yes, edited the question $\endgroup$ – Jack Feb 25 at 21:30
  • $\begingroup$ no, want to show f(x_0)=1 and also f(y)=0. I think for y it would be obvious after showing it for x_0. $\endgroup$ – Jack Feb 25 at 21:38
  • 1
    $\begingroup$ I suppose that the infimum may be equal to $\varepsilon$. $\endgroup$ – Rebellos Feb 25 at 22:20

Since $A$ is a closed subspace of $X$, then it exists $y \in A$ such that : $\inf\{\|y-x_0\| : x_0 \in X\} = d(x_0,Y)= \varepsilon$.

Let $I = A \bigcup\langle x_0\rangle = \{ y + \lambda x_0 : y \in A, \lambda \in \mathbb R\}$. Consider the functional $f:I \to \mathbb R$ such that : $$f(y+\lambda x_0) = \lambda d(x_0,Y)=\lambda \varepsilon$$ Note that for $y=0$ and $\lambda =1$, it is : $$f(x_0) = d(x_0,Y)=\varepsilon$$ For $\lambda =0$ it is : $f(y) = 0$.

Now, note that : $$|f(y+\lambda x_0)| = |\lambda|d(x_0,Y) \leq |\lambda|\left\|x_0 +\frac{1}{\lambda}y \right\| = \|y + \lambda x_0\|$$ Using that and a small other argument, you can prove that $\|f\| = 1$. I'll leave this up to you.

After you are done with that, just use the Hahn-Banach Theorem to extend that function to the whole $X$.

Now, consider the functional : $$g = \frac{1}{\varepsilon}f$$ Then, you can clearly see that : $$g(x_0) = \frac{1}{\varepsilon}f(x_0) = 1 $$ $$g(y) = \frac{1}{\varepsilon} f(y) = 0$$ $$\|g\| = \left\| \frac{1}{\varepsilon}f\right\|= \frac{1}{\varepsilon}$$ The exercise is complete.


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.