I think no. And I am looking for examples. I would like a sequence $y_n$ in $Y$ such that $||y_n-x||\rightarrow d(x,Y)$ while $y_n$ do not converge.

Can anyone give a proof or an counterexample to this question?

  • $\begingroup$ It's difficult to give a counterexample to a question. Make a statement and I'll try to find a counterexample. $\endgroup$ – Fly by Night Feb 6 '13 at 16:14
  • $\begingroup$ The statement is "given a point $x$ and a closed subspace $Y$ of a normed space, the distance from $x$ to $Y$ is achieved by some $y\in Y$" $\endgroup$ – Spook Feb 6 '13 at 16:16
  • $\begingroup$ I think that this is true for all closed $Y$ in a $T_4$ space. $\endgroup$ – Fly by Night Feb 6 '13 at 17:33
  • $\begingroup$ @DavidMitra Can you think of a counterexample when $X=l_\infty$ and $Y=c_0$? I tried to build one but I was not successful. One cannot use the construction in your link, as $c_0$ cannot be the kernel of any linear functional on $l_\infty$. $\endgroup$ – Theo Feb 6 '13 at 20:20
  • $\begingroup$ @DavidMitraThank you for the link. $\endgroup$ – Spook Feb 6 '13 at 21:23

This is a slight adaptation of a fairly standard example.

Let $\phi: C[0,1]\to \mathbb{R}$ be given by $\phi(f)=\int_0^{\frac{1}{2}} f(t)dt - \int_{\frac{1}{2}}^1 f(t)dt$. Let $Y_\alpha = \phi^{-1}\{\alpha\}$. Since $\phi$ is continuous, $Y_\alpha$ is closed for any $\alpha$.

Now let $\hat{f}(t) = 4t$ and notice that $\phi(\hat{f}) = -1$ (in fact, any $\hat{f}$ such that $\phi(\hat{f}) = -1$ will do). Then $$\inf_{f \in Y_0} \|\hat{f}-f\| = \inf \{ \|g\|\, | \, g+\hat{f} \in Y_0 \} = \inf \{ \|g\|\, | \, \phi(g) =1 \} = \inf_{g \in Y_1} \|g\|$$ It is clear that $g_n$ is an infimizing sequence for the latter problem iff $g_n+\hat{f}$ is an infimizing sequence for the initial problem.

It is well known that $Y_1$ has no element of minimum norm, consequently there is no $f \in Y_0$ that mnimizes $\|f-\hat{f}\|$.

  • $\begingroup$ We seem to have come to different conclusions. Have a look at my "proof" and see if you can find an error. $\endgroup$ – Fly by Night Feb 6 '13 at 17:42

There cannot be an example in $c_0 \subset \ell^\infty$. In fact $c_0$ is a closed ideal in the $C^*$-algebra $\ell^\infty$. For such pairs distance to ideal always attained be an element in the ideal.

  • $\begingroup$ It helps reability to format answers using Mathjax $\endgroup$ – Moo Oct 5 '16 at 3:59

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.