Davis, Figiel, Johnson and Pełczyński's Factorization Theorem states that each weakly compact operator $T:X \to Y$ between Banach spaces $X$ and $Y$ factors through a reflexive Banach space $Z$. In addition, they prove that $Z$ can be constructed to have a Schauder basis if $Y$ has a shrinking basis or an unconditional basis. It was later shown that if $Y=C[0,1]$ then $Z$ can be constructed with a basis.

I am interested generally in positive and negative results regarding when $Z$ can be constructed to have a basis. For example:

Q1: If $Y=L_1[0,1]$, can $Z$ be constructed to have a basis for any weakly compact $T:X \to Y$?

Q2: Is there an example of a space $Y$ with basis and weakly compact $T:X \to Y$ so that every reflexive $Z$ which $T$ factors through does not have a basis?

I'm also interested to know if anyone is aware of other results similar to the those stated above.

  • 1
    $\begingroup$ I never thought about this nice question. My guess is that Q2 has an affirmative answer when $Y^*$ has the bounded approximation property but a negative answer in general. $\endgroup$ Jun 5, 2016 at 2:25
  • $\begingroup$ Who proved the positive result when $Y=C[0,1]$? $\endgroup$ Jun 5, 2016 at 2:26
  • 3
    $\begingroup$ The $Y=C[0,1]$ theorem was proved by Ghoussoub, Maurey and Schachermayer in "Slicings, Selections and their applications." (Can. J. Math). They use it to give an alternative proof of Zippin's Theorem (i.e. a space with a separable dual embedds into a space with a shrinking basis). $\endgroup$ Jun 5, 2016 at 16:34

1 Answer 1


I think Q1 has a positive answer. To see this, review how the results are proved in DFJP. You have a weakly compact subset $W$ of $Y$ (the closure of the image of the unit ball of $X$ under a weakly compact operator) and you apply the factorization technique (which we now know is just a real interpolation method applied to the pair $(W, B_Y)$) to $W$ to obtain a larger weakly compact symmetric subset $C$ of $Y$. Consider the normed space that has $C$ as its unit ball. It is proved that $C$ is weakly compact in in this normed space and hence this normed space is reflexive.

When $Y$ has a (always Schauder) basis, instead of applying the interpolation technique to the original weakly compact set $W$, we enlarged $W$ to a bigger weakly compact set $W_1$ that is invariant for the partial sum projections associated with the basis for $Y$. Then the interpolation technique, applied to $W_1$, produces a still larger weakly compact set $C$ that is invariant for the partial sum projections associated with the basis for $Y$. From this it is easy to see that the basis for $Y$ is also a basis for the space that has $C$ as its unit ball.

For $W_1$ we used the closed convex hull of the union over $n$ of $P_n W$, where $(P_n)$ are the partial sum projections for the basis for $Y$. When the basis is shrinking, we proved that $W_1$ is weakly compact. Suppose that $Y=L_1$ and you use the Haar basis for $Y$. Even though the Haar basis is far from being shrinking, the same construction produces a weakly compact set $W_1$. To see this, recall that a weakly closed set $D$ in $L_1$ is weakly compact iff for all $\epsilon > 0$ there is $M=M_\epsilon$ s.t. $D\subset M B_{L_2} + \epsilon B_{L_1}$ (``uniform integrability") and use the fact that both $B_{L_2}$ and $B_{L_1}$ are invariant under the Haar basis projections.

  • 1
    $\begingroup$ Thanks for the answer. Combing this with what I proved with Dan a few years ago it follows that if A is an analytic (with respect to the strong operator topology) collection of weakly compact operators into $L_1$ there is a separable reflexive space $X$ that every operator in A factors through $X$. $\endgroup$ Jun 6, 2016 at 14:30
  • $\begingroup$ To clarify: for Q2 you guess that if $Y$ has a basis, $Y^*$ has the BAP and $T:X \to Y$ is weakly compact then there is a $Z$ with a basis so that $T$ factors though $Z$. $\endgroup$ Jun 6, 2016 at 14:32
  • 1
    $\begingroup$ Yes, that was one of my guesses. But thinking more about why I guessed that, I now conjecture that not both of my guesses are correct. In other words, I haven't a clue about the general case. $\endgroup$ Jun 6, 2016 at 19:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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