Let $L$ be a Dedekind-complete Banach lattice. Let $\mathcal{B}$ be the family of nonempty norm-compact subsets of $L$ that are bounded from below. Endow $\mathcal{B}$ with the topology induced by the Hausdorff metric.

Consider the function $\psi\colon \mathcal{B}\to L$ defined by $\psi(A)= \inf{(A)}$ for all $A\in \mathcal{B}$.

I have a number of questions on this, the most important of which is the following:

Q1. What conditions on $L$ guarantee that $\psi$ is a continuous function?

The other questions depend on the answer to Q1, in obvious ways.

Q2. What is an example of a Dedekind-complete Banach lattice for which $\psi$ is not continuous?

Q3 Suppose that $\psi$ is continuous and $E$ is a Banach lattice. Will the same hold for the the space of regular operators $\mathcal{L}^r(E,L)$ (the linear span of positive operators)? How about when $L$ has a strong order dual?

Now suppose that $X\subseteq L$ is a non-empty norm-compact sub-lattice of $L$. Take the restriction of $\psi$ to the space $\mathcal{B}(X)$ of nonempty compact subsets of $X$. My final question is the following:

Q4 What conditions on $X$ guarantee that $\psi$ restricted to $\mathcal{B}(X)$ is continuous?

This latter question Q4 is the same as asking, What conditions guarantee that $X$ as a topological lattice has a neighborhood base of $inf$-sub-semilattices?


I guess finite dimensionality is needed for Q2. For example, in $\ell_2$ consider the sets $X_n = \{n^{-1/2} e_j : 1 \le j \le n \}$. Then in the Hausdorff metric $X_n \to \{0\}$ but the infimum of each $X_n$ has norm one.

It looks like you need to restrict attention to compact sets that are in a common order interval.

| cite | improve this answer | |
  • $\begingroup$ Thanks Bill. I figured so much riding to uni. But is it always the case that if $X$ is a compact sublattice, the restriction of $\psi$ to $X$ is continuous. So Q4 is of interest to me. $\endgroup$ – Rabee Tourky Jun 18 '13 at 6:24
  • $\begingroup$ In $L_2(0,1)$ take '$X_n = \{1_{A_i} : 1 \le i \le 2^n \}$', where `$A_i = [(i-1)/2^n, i/2^n]$'. This gives an order bounded counterexample to Q2 which can be generalized to any order continuous non discrete Banach lattice. But the lattice generated by the $X_n$ is not totally bounded. $\endgroup$ – Bill Johnson Jun 18 '13 at 6:47
  • $\begingroup$ The only infinite dimensional example that I can think of where $X$ is a sublattice that is compact and where $\psi$ restricted to compact subsets of $X$ is continuous is monotone functions in $L_1(0,1)$. $\endgroup$ – Rabee Tourky Jun 18 '13 at 6:51
  • $\begingroup$ In the previous comment, I mean order continuous infinite dimensional space. $\endgroup$ – Rabee Tourky Jun 18 '13 at 8:15

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.