2
$\begingroup$

Consider the normed linear space $\Bbb R^2$ with this norm $\|(x,y)\|=|x|+|y|$.

Define $f$ to be a bounded linear functional on $X=\{(x,y):x=y\}$ by $f(x,y)=3x$.

If $g$ is a Hahn-Banach extension of $f$ on $\Bbb R^2$ given by $g(x,y)=\alpha x+\beta y$ then find $\alpha-\beta$.

My try:

Since $g$ is an extension of $f$ so $g|_X=f\implies g(x,x)=(\alpha+\beta)x=3x\implies \alpha+\beta =3$

Now $\|f\|=\sup \{f(x,x):\|(x,x)\|=1,(x,x)\in \Bbb R\}=3\implies \|g\|=3$

So $\sup \{g(x,y):\|(x,y)\|=1,(x,y)\in \Bbb R\}=3\implies \alpha x+\beta y\le 3$ where $|x|+|y|=1$.

Take $x=0,y=1\implies \alpha x+\beta y\le 3\implies \beta\le 3$.

How to find the value of $\alpha,\beta $ from above.

Please give some hints.

$\endgroup$
2
$\begingroup$

I think you have $\|f\| = \sup \{|3x| : 2|x| = \|(x,x)\| = 1\} = 3/2$, not $\|f\| =3$.

Then, since $\alpha + \beta =3$ and since $\max \{|\alpha |, |\beta|\} = 3/2$, you must have $\alpha = \beta = 3/2$ (why?). Hence, $\alpha - \beta =0$.

Here, I used that the dual norm to the $\ell^1$ norm is the $\ell^\infty $ norm.

$\endgroup$
  • $\begingroup$ Can't this be done without using duality? $\endgroup$ – Learnmore Feb 28 '17 at 11:32
  • $\begingroup$ Since $||g||=\frac{3}{2}\implies$ $\sup\{\alpha x+\beta y\}=\frac{3}{2};|x|+|y|=1$ The extrema points of $|x|+|y|=1=(\pm 1,0),(0,\pm 1)$ $\endgroup$ – Learnmore Feb 28 '17 at 11:35
  • $\begingroup$ Hence at $(1,0)$ we have $\alpha x+\beta y\implies \alpha=\frac{3}{2}$;Similarly at $(0,1)$ we have $\alpha x+\beta y\implies \beta=\frac{3}{2}\implies \alpha=\beta$ $\endgroup$ – Learnmore Feb 28 '17 at 11:37
  • $\begingroup$ Is it correct?Thanks for pointing $||f||=\frac{3}{2}$ $\endgroup$ – Learnmore Feb 28 '17 at 11:38
  • $\begingroup$ @Ben: In the present case, the formula for the dual is very easy to see (this is essentially what you are doing yourself): It is easy to see $\|(x,y)\|_{(\ell^1)'} \leq \|(x,y)\|_{\ell^\infty } $. Conversely, $\|(x,y)\|_{(\ell^1)'} \geq |x|$, as can be seen by "testing" with $(1,0) $. Testing with $(0,1) $ completes the proof. $\endgroup$ – PhoemueX Feb 28 '17 at 11:47

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.