I am looking for an explicit example for a Hamel basis for $\ell^{p}$?. As we know that for a Banach space a Hamel basis has either finite or uncountably infinite cardinality and for such a basis one can express any element of the vector space as a finite linear combination of these. After some trying I could not write one explicitly. A quick google search did not reveal anything useful except for the proof of uncountability of a an infinite Hamel basis. Maybe I am being a bit silly but I don't think the answer is as obvious as for a Schauder basis for the same case.

So, what is an explicit example for a Hamel basis for $\ell^{p}$??

  • 9
    $\begingroup$ You are not going to find any. This is not something one can do explicitly. $\endgroup$ – Mariano Suárez-Álvarez Sep 11 '12 at 15:18
  • 3
    $\begingroup$ @Mariano Interesting! Can you please provide some reference. $\endgroup$ – Abhishek Gupta Sep 11 '12 at 15:20
  • 2
    $\begingroup$ The fact that every vector space has a (Hamel) basis requires the Axiom of Choice (AC). Even though $\ell^p$ is a very specific space, I guess one can at least deduce a weak form of AC from an explicit basis. Choice is always an abstacle for explicitness. $\endgroup$ – Hagen von Eitzen Sep 11 '12 at 15:52
  • 1
    $\begingroup$ Asaf's answer in this thread on $\mathbb{R^N}$ is closely related. The first version works if you read $\ell^p$ instead of $\mathbb{R^N}$. $\endgroup$ – t.b. Sep 11 '12 at 16:02
  • 1
    $\begingroup$ @Jose: I don't think we need a tag for Hamel bases related questions. If you disagree feel free to raise the issue on meta. In the meantime, I ask that you stop adding the tag to other questions and let other people voice their opinion as well. $\endgroup$ – Asaf Karagila Feb 13 '18 at 16:08

The existence of a Hamel basis for $\ell^p$ cannot be proved without some of the axiom of choice, which in modern terms usually means that we cannot write it explicitly.

It is consistent with ZF+DC (a weak form of the axiom of choice which is sufficient to do a lot of the usual mathematics) that all sets of real numbers have Baire property, and in such model we have that every linear function from $\ell^p$ to itself is continuous.

It is also true (in ZF) that $\ell^p$ is separable for $1\leq p<\infty$. It is a known fact that continuous endomorphisms are determined completely by the countable dense set.

If there exists a Hamel basis then its cardinality is at least $\frak c$ (or rather exactly that), and therefore it has $2^\frak c$ many permutations, each extends uniquely to a linear automorphism, which is continuous.

Now, note that $\ell^p$ has size $\leq\frak c$ itself, since it is a separable metric space (and again, this is in fact $\frak c$) and therefore it has only $\frak c$ many continuous endomorphisms.

Cantor's theorem tells us that $2^\frak c\neq c$, and therefore in Shelah's model where every set of real numbers have the Baire property there is no Hamel basis for $\ell^p$.

  • 5
    $\begingroup$ It would probably be a useful idea to explain what is the relation between what you wrote and the question of whether one can make explicit a basis. Also, assuming we all know what «ZF+DC» means, or the relevance of something being true there is possibly not great. In fact, I am pretty sure that anyone who can understand what you wrote already knew the answer to the question :-) $\endgroup$ – Mariano Suárez-Álvarez Sep 11 '12 at 16:55
  • 2
    $\begingroup$ Mariano, due to lack of proper keyboard access this will wait for tomorrow. I agree that some improvement may be good. As for the last part I disagree completely. I knew what is ZF+DC long before I knew this answer. $\endgroup$ – Asaf Karagila Sep 11 '12 at 17:04
  • 1
    $\begingroup$ If you're interested, the argument can be simplified a bit: given a Hamel basis we can construct a discontinuous linear map more explicitly. Theorem ($\mathsf{ZF}+\mathsf{AC}_\omega$): If $X$ is an infinite-dimensional normed space and $B$ is a Hamel basis for $X$, then there exists a discontinuous linear functional $f : X \to \mathbb{R}$. Proof: without loss of generality suppose $B$ consists of unit vectors. Pick a countable subset $C = \{x_1, x_2, \dots\} \subset B$ (with no repetitions). Define $f$ on $B$ by $f(x_n) = n$ and $f(x) = 0$ for $x \notin B$... $\endgroup$ – Nate Eldredge Jan 21 '16 at 19:24
  • 1
    $\begingroup$ Then $f$ extends to a linear functional $f : X \to \mathbb{R}$. It is discontinuous since $x_n/n \to 0$ and yet $f(x_n/n) = 1$ for all $n$. QED. If you want a map $T : X \to X$, pick your favorite nonzero $y \in X$ and let $T(x) = f(x)y$. Maybe there is a way to do it strictly in $\mathsf{ZF}$, but analysis there tends to be so weird that I don't care to try. $\endgroup$ – Nate Eldredge Jan 21 '16 at 19:26
  • 1
    $\begingroup$ @Nate: That is somewhat optimal too, since Brunner proved that if $D$ is a set which cannot be mapped onto the integers, then $\ell_1(D)$ has a Hamel basis and every linear functional on it is continuous. $\endgroup$ – Asaf Karagila Jan 21 '16 at 19:33

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.