I am reading a paper where I find the following definition:

Definition: Let $L$ be an orthocomplemented lattice. A function $\mu: L \rightarrow [0,1]$ is called a probability measure on $L$ if

  1. $\mu(0) = 0$, $\mu(1) = 1$,
  2. $\mu(a_1 \lor a_2 \lor \ldots) = \sum_{k=1}^\infty \mu(a_k)$ for mutually disjoint $a_i$.

A function $\nu: B(\mathbb{R}) \rightarrow L$, where $B(\mathbb{R})$ is the family of Borel sets on $\mathbb{R}$, is called a $L$-valued measure on $B(\mathbb{R})$ if

  1. $E \cap F = \emptyset$ implies $\nu(E) \perp \nu(F)$ for any $E, F \in B(\mathbb{R})$,
  2. $\nu(E_1 \cup E_2 \cup \ldots) = \nu(E_1) \lor \nu(E_2) \lor \ldots$ for any family $E_i \in B(\mathbb{R})$ such that $E_i \cap E_j = \emptyset$ for $i \neq j$,
  3. $\nu(\emptyset) = 0, \nu(\mathbb{R}) = 1$.

There is one thing that I don't understand in this definition. On both lists, in item 2, there appears a join of potentially infinitely many elements of $L$. Is this well-defined? I thought that for such a thing to be well-defined, the lattice needs to be complete. Is an orthocomplemented lattice automatically complete?


An orthocomplemented lattice need not be complete.

Consider the system of all subsets of $\mathbb N$ that are either finite or with finite complement. This is a sub-Boolean algebra of the powerset of $\mathbb N$, hence a orthocomplemented lattice; a non-complete one, since the system of all finite sets containing only the even numbers has no supremum in it.

The Definition in question needs at least "Let $L$ be a $\aleph_0$-complete orthocomplemented lattice" to make sense.

Perhaps there is a small sentence, something like "unless otherwise stated, all lattices are complete in this paper" in the preliminaries section of the paper.

| cite | improve this answer | |
  • 1
    $\begingroup$ Its the small sentences that always get you in the end... $\endgroup$ – goblin GONE Sep 18 '13 at 2:05

I'd assume they meant that the identity in item 2 should hold when $a_1 \vee a_2 \vee ...$ is defined.

I haven't really studied orthocomplemented lattices before, so I'm just working from the Wikipedia definition, but how about the following for a non-complete orthocomplemented lattice (apologies for the strange notation):

$0<1<2<...< \infty -2 < \infty - 1 < \infty$,

$0<1'<2'<...< (\infty -2)' < (\infty - 1)' < \infty$,

with anything exclusively from line 1 neither bigger nor smaller than anything exclusively from line 2.

This is a bounded lattice (both strands are totally ordered), with lower bound $0$ and upper bound $\infty$, and for $n \geq 1$ $(\infty - n)'$ can be chosen as the orthocomplement for $n$, $n'$ as the orthocomplement for $(\infty - n)$, $(\infty - n)$ as the orthocomplement for $n'$, and $n$ as the orthocomplement for $(\infty - n)'$. $\{0,1,2...\}$ has no least upper bound in this lattice.

| cite | improve this answer | |
  • $\begingroup$ Perhaps it would be better to reverse the second chain and then just say that $x\mapsto x'$ is the orthocomplementation. $\endgroup$ – Gejza Jenča Apr 10 '14 at 21:09

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.