5
$\begingroup$

Let $\Omega$ be a linearly (i.e. fully) ordered set, and let $\Lambda_{\Omega}$ be the ordered abelian group consisting of those $(\lambda_\omega)_{\omega\in\Omega}\in\mathbb{R}^{\Omega}$ with well-ordered support. Note that $\mathbb{R}^{\Omega}$ itself has incomparable elements, but restricting to elements of $\Lambda_{\Omega}$ ensures that all pairs of elements are comparable: $(\lambda_\omega)_{\omega\in\Omega}<(\lambda'_\omega)_{\omega\in\Omega}$ if $\lambda_{\omega_0}<\lambda'_{\omega_0}$ where $\omega_0$ is the least element of $\Omega$ for which $\lambda_{\omega_0}\neq\lambda'_{\omega_0}$. A group of the form $\Lambda_{\Omega}$ is called a Hahn group.

For $\lambda,\lambda'\in\Lambda_{\Omega}$ with $\lambda'>0$, write $\lambda\ll\lambda'$ if $\lambda=0$ or the first non-zero entry of $\lambda'$ precedes that of $\lambda$. (This means that $\lambda$ is infinitely smaller than $\lambda'$: that is, $n\lambda<\lambda'$ for all $n\in\mathbb{Z}$.)

Let $\Lambda_0$ be a subgroup of $\Lambda_{\Omega}$ and $\zeta:\Lambda_0\to\Lambda_{\Omega}$ a (not necessarily order-preserving) homomorphism for which $\zeta(\lambda)\ll\lambda$ for all positive $\lambda\in\Lambda_0$. Can $\zeta$ be extended to a homomorphism $\Lambda_{\Omega}\to\Lambda_{\Omega}$ that also satisfies $\zeta(\lambda)\ll\lambda$ for all positive $\lambda\in\Lambda_{\Omega}$?

A positive answer would help answer another question.

$\endgroup$
1
$\begingroup$

I don't see why such an extension should exist in such generality.

For a heuristic explanation, let's consider a related question. Let $\Omega$ be a finitely generated (abelian) group generated by a set $S$. Let $\Lambda^\prime_\Omega$ be the ordered preserving group consisting of those order preserving group homomorphisms $(\lambda_\omega)_{\omega\in\Omega}$ with well ordered support (this is a subgroup of $\Lambda_\Omega$). I claim that a homomorphism $\zeta:\Lambda_S\to\Lambda_\Omega$ such that $\zeta(\lambda)\ll\lambda$ for positive $\lambda$, and such that if $\lambda\ll\lambda^\prime$, then $\zeta(\lambda)\ll\zeta(\lambda^\prime)$, cannot necessarily be extended to a homomorphism $\Lambda_\Omega\to\Lambda_\Omega$ that satisfies $\zeta(\lambda)\ll\lambda$ for those $\lambda$. Note that $\Lambda_S$ is a subgroup of $\Lambda_\Omega^\prime$, which in turn is a subgroup of $\Lambda_\Omega$. I'm therefore unable to see a reason for why this should be true for a general subgroup of a Hahn group.

To see this, choose any element $(\lambda_\omega)_{\omega\in\Omega}\in\Lambda^\prime_\Omega$. Since $S$ generates $\Omega$, we can write $\omega = \sum_i n_is_i$ for some finite $\{s_i\}\in S$, and since every $\lambda$ is a group homomorphism, we find that $(\lambda_\omega)_{\omega\in\Omega} = \sum_in_i(\lambda_{s_i})_{s_i\in S}$ where $(\lambda_{\omega_0})_{\omega_0\in\Omega}+(\lambda_{\omega_1})_{\omega_1\in\Omega} = (\lambda_{\omega_0+\omega_1})_{\omega_0,\omega_1\in\Omega}$ i.e. the expected definition. Now, let $\overline{T}_\lambda$ consist of those $s_i$ for which $\lambda_i<0$ (note the strict inequality), and let $\overline{(\lambda_{s_i})_{s_i\in\overline{T}_\lambda}}$ consist of $\lambda$ restricted to $\overline{T}_\lambda$ and then extended to $S$ by defining it to be zero everywhere else. Then $(\lambda_\omega)_{\omega\in\Omega} - \sum_in_i\overline{(\lambda_{s_i})_{s_i\in\overline{T}_\lambda}}$ is positive. Because $\zeta$ is a group homomorphism, we find that $\zeta(\lambda_\omega)_{\omega\in\Omega} + \sum_in_i\zeta(-\overline{(\lambda_{s_i})_{s_i\in\overline{T}_\lambda}}) \ll (\lambda_\omega)_{\omega\in\Omega} + \sum_in_i(-\overline{(\lambda_{s_i})_{s_i\in\overline{T}_\lambda}})$. If the first nonzero element of $(\lambda_\omega)_{\omega\in\Omega}$ is positive, so that $(\lambda_\omega)_{\omega\in\Omega}$ is positive, then $\sum_in_i(-\overline{(\lambda_{s_i})_{s_i\in\overline{T}_\lambda}})\ll (\lambda_\omega)_{\omega\in\Omega}$. This means that $\sum_in_i\zeta(-\overline{(\lambda_{s_i})_{s_i\in\overline{T}_\lambda}})\ll \zeta(\lambda_\omega)_{\omega\in\Omega}$, so it is not necessarily true, even in this ideal scenario, that $\zeta(\lambda_\omega)_{\omega\in\Omega}\ll(\lambda_\omega)_{\omega\in\Omega}$.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Many thanks for your answer. There are some things I don't follow in it though, possibly because I'm missing something obvious. For one thing, 'we find that $(\lambda_\omega)_{\omega\in\Omega} = \sum_in_i(\lambda_{s_i})_{s_i\in S}$' -- should the left-hand side here be just $\lambda_\omega$, that is, the image of $\omega$ under the homomorphism $\lambda\in\Lambda'_{\Omega}$? More fundamentally, you claim that a homomorphism ( + extra properties) with domain $\Lambda_S$ can't necessarily be extended to a homomorphism with domain $\Lambda_{\Omega}$... $\endgroup$ – shane.orourke Aug 15 '16 at 22:18
  • $\begingroup$ ...what's wrong with just putting the homomorphism equal to zero on elements supported in $\Omega\backslash S$? The only structure on $\Omega$ that is required for my purposes is the ordered one, not any additive structure that it may also be endowed with. $\endgroup$ – shane.orourke Aug 15 '16 at 22:19
  • $\begingroup$ @shane.orourke Since $\omega = \sum n_i s_i$ and $\lambda$ is a homomorphism, for every $\omega\in\Omega$, $\lambda_\omega = n_i\lambda_{s_i}$, so that $\lambda$ itself is specified by $(\lambda_\omega)_{\omega\in\Omega} = \sum_i n_i(\lambda_{s_i})_{s_i\in S}$. As for the second question (which I have hopefully understood correctly), if $\zeta$ could be extended by just letting it be zero on $\Lambda_\Omega$ except on $\Lambda_S$, then $\zeta(\lambda)$ isn't even positive for positive $\lambda$. $\endgroup$ – user97187 Aug 15 '16 at 23:36
  • $\begingroup$ But $\zeta(\lambda)$ isn't required to be positive for positive $\lambda$; if it were, then $\zeta$ would be an order-preserving homomorphism. $\endgroup$ – shane.orourke Aug 15 '16 at 23:52
  • $\begingroup$ @shane.orourke In the question, you defined what $\lambda\ll\lambda^\prime$ means for positive $\lambda,\lambda^\prime$. Of course, this can be defined for nonnegative $\lambda,\lambda^\prime$, but I thought the symbol $\ll$ was only defined for positive maps $\Omega\to\mathbf{R}$. (Also, if $\zeta(\lambda)$ is positive for positive $\lambda$, why should it be order-preserving?) $\endgroup$ – user97187 Aug 16 '16 at 0:21

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.