MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have two questions after reading the Hahn-Banach theorem from Conway's book ( I have googled to know the answer but I have not found any result yet. Also I am not sure that whether my questions have been asked here somewhere on this forum - so please feel free to delete them if they are not appropriate )

Here are my questions:

  1. We know that if $M$ is a linear subspace of $X$ and $f :M\to\mathbb{F}$ and $f$ is linear,bounded by a seminorm $p$ then $f$ can be extended onto $X$ by some functional $F$. Can $F$ be unique ? Under what condition $F$ will be an unique extension? It would be appreciate if you could give me one example that $F$ could not be unique.

  2. If the above $\mathbb{F}$ is replaced a Banach space $Y$, i.e, let $M$ be a closed subspace of a Banach space $X$, and $f :M\to Y$ be a bounded linear operator, can we extend $f$ by a bounded operator $F :X\to Y$ ? if not, what condition should be put on $Y$ to have a such extension?

thanks so much

share|cite|improve this question
I think that the correct spelling is "bounded".... "bonded" reminds me of bondage. – André Henriques Jun 25 '11 at 22:44
Steven, for your 2nd question one needs to put conditions on $M$, $X$ or $Y$. The classic example is to take $M=c_0$ sitting inside $X=\ell^\infty$, take $Y=c_0$, and let $f$ be the identity mapping; then there is no bounded lin. extension of $f$ to a map $X\to Y$, because that would give a bounded lin. projection of $\ell^\infty$ onto $c_0$, which is impossible by a theorem of Phillips. I recommend that you look into a text on Banach space theory: I haven't read Albiac and Kalton's book but my guess is that they might discuss issues like this, and the text itself is at an introductory level. – Yemon Choi Jun 25 '11 at 23:19
In question 1, if $X=\mathbb R^2$ with norm $\|(x_1,x_2)\| = \max(x_1,x_2)$, then extension is not unique. – Gerald Edgar Jun 25 '11 at 23:29
For question 2 it is enough for $M$ to be a complemented subspace, i.e., that it has a closed complement in $X$, since this is equivalent to having a continuous projection from $X$ onto $M$. If the extension property has to hold for every $Y$ and every $f$, then this condition is also necessary, as can be seen by taking $f$ to be the identity map on $M$. – godelian Jun 26 '11 at 0:14
Steven: I forget the exact wording of the "Phillips lemma", but it is almost surely mentioned in the Kalton-Albiac book (or indeed others) – Yemon Choi Jun 26 '11 at 1:15

$Y$ is called an injective Banach space if the extension exists for all $X$, $M$, and $f$. An example is $Y = l^\infty$. (Should be in Banach space text books. Here's a paper: )

share|cite|improve this answer

Note quite what you asked, but related:

Continuous extensions of (continuous) functionals from $M$ are unique if and only if $M$ is a dense subspace of $X$. Otherwise its closure is a proper closed subspace and therefore there exists a nonzero bounded functional $\phi$ vanishing on the closure, which implies that $F+\phi$ is bounded and also extends $f$.

For the second question, it is easier to put conditions on $M$ so that for every $Y$, every map from $M$ to $Y$ can be extended. As mentioned in the comments, a necessary and sufficient condition is that $M$ is a complemented subspace of $X$.

share|cite|improve this answer
Sorry, I only realized after this answer was posted that the question is about extensions bounded by seminorms, not just continuous extensions (the most commonly used corollary of Hahn-Banach theorem). This should have been a comment. – godelian Jun 26 '11 at 15:28

When $X^*$ is strictly convex and $M$ is a subspace of $X$, every norm one $f\in M^*$ admits a unique norm one extension $F\in X^*$, because given two of such extensions $F_1$ and $F_2$, $(F_1+F_2)/2$ is norm one.

In fact this property characterizes $X^*$ strictly convex. See Theorem 7.11 in B.V. Limaye "Functional analysis", 2nd. ed.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.