1

Given ordered groups $H:=(A,+)$ and $H':=(A, \boxplus)$ as per the same total order $ \leqslant$ of A,

can $H$ and $H'$ be group isomorphic without being order isomorphic?


I understand that there can be a group isomorphism between $H$ and $H'$ which does not preserve the order. For example, if $H:=H':=( \mathbb Z, +)$, $f(n):=-n$ is such a group isomorphism which does not preserve the order. Still, the two groups are order isomorphic, i.e. there exists some other group isomorphism which also preserves the order.

Shaun
  • 44,997
Crispost
  • 169
  • Question is related to a comment by @GEdgar in post

    https://math.stackexchange.com/questions/4810780/is-an-archimedean-topological-group-of-the-reals-isomorphic-to-mathbb-r

    – Crispost Nov 21 '23 at 12:59

1 Answers1

2

Here's an example of ordered groups which are isomorphic as groups, and which are isomorphic as linear orders, but which are not isomorphic as ordered groups. I think this answers the body of your question. In particular, after relabelling, you can assume they have the same underlying set and the ordering on the groups is the same.

Let $H$ be the subgroup $\Bbb Q(\sqrt 2) = \{a + b \sqrt 2 : a, b \in \Bbb Q\}$ of $\Bbb R$ with the inherited order and let $H'$ be the subgroup $\Bbb Q(\sqrt 3)$. As groups, we have $H \cong H' \cong \Bbb Q^2$. Both are dense countable linear orders, so as linear orders, we have $H \cong H' \cong \Bbb Q$. But they aren't ordered-group-isomorphic, which is morally because the orderings have some "information" about the algebraic properties of $\sqrt 2$ and $\sqrt 3$.

Indeed suppose $f: \Bbb Q(\sqrt 2) \to \Bbb Q(\sqrt 3)$ is an ordered-group isomorphism. Note that $f(1) > f(0) = 0$, so we can assume without loss of generality that $f(1) = 1$ (because $x \mapsto x/f(1)$ is an ordered-group isomorphism of $\Bbb Q(\sqrt 3)$, since $\Bbb Q(\sqrt 3)$ is actually an ordered field). Now it follows that $f$ fixes every rational. In particular if $y = f(\sqrt 2)$, then for all rationals $q$ with $q < \sqrt 2$ we have $q = f(q) < f(\sqrt 2) = y$ and for all $q > \sqrt 2$ we have $q = f(q) > f(\sqrt 2) = y$. But there is no such element $y$ in $\Bbb Q(\sqrt 3)$ (remember for example that it's an ordered subgroup of $\Bbb R$ and the unique such element of $\Bbb R$ is $\sqrt 2$), so there can be no such $f$.


Addendum..

Suppose $(H_1, +_1, \le_1)$ and $(H_2, +_2, \le_2)$ are ordered groups, and there is some ordered set $(A, \le_A)$ and there are isomorphisms of ordered sets $f_1: (H_1, \le_1) \to (A, \le_A)$ and $f_2: (H_2, \le_2) \to (A, \le_A)$, and $g: (H_1, +_1) \to (H_2, +_2)$ is an isomorphism of groups, but we know that $(H_1, +_1, \le_1)$ and $(H_2, +_2, \le_2)$ are not isomorphic as ordered groups.

Define an operation $+_1^A$ on $A$ by $a +_1^A b = f_1(f_1^{-1}(a) +_1 f_1^{-1}(b))$ and similarly define an operation $+_2^A$ on $A$ by $a +_2^A b = f_2(f_2^{-1}(a) +_2 f_2^{-1}(b))$.

Then it's not hard to check that $(A, +_1^A, \le_A)$ and $(A, +_2^A, \le_A)$ are ordered groups, $f_1: (H_1, +_1, \le_1) \to (A, +_1^A, \le_A)$ and $f_2: (H_2, +_2, \le_2) \to (A, +_2^A, \le_A)$ are isomorphisms of ordered groups, $f_2 g f_1^{-1} : (A, +_1^A) \to (A, +_2^A)$ is an isomorphism of groups.

But since $(H_1, +_1, \le_1)$ and $(H_2, +_2, \le_2)$ are not isomorphic as ordered groups, it follows that $(A, +_1^A, \le_A)$ and $(A, +_2^A, \le_A)$ are not isomorphic as ordered groups.

Hence, your question is equivalent to "if $H$ and $H'$ are ordered groups, and are isomorphic as groups and as linear orders, must they be isomorphic as ordered groups?". This idea is a standard one when you're proving things about isomorphism or non-isomorphism, called "transport of structure". I don't think the above symbols are a terribly helpful way to think about it - the point is that given any bijection between $H$ and $A$, we can give $A$ the same structure as $H$ by "pulling the structure through the bijection". In the special case where $A$ already has an ordering and the bijection is an order isomorphism, the order this gives is the usual order on $A$, so the addition this gives must make $A$ into an ordered group, since we know the addition and ordering are compatible in $H$.

  • Although this is a useful answer for a more general question, it does not answer the question itself, which assumes two ordered groups over the same totally ordered set $(A, \leqslant )$. – Crispost Nov 21 '23 at 15:47
  • 1
    @Crispost, it does answer that, I think. I addressed it briefly in the sentence "after relabelling, you can assume they have the same underlying set and the ordering on the groups is the same". In more detail: since they are both isomorphic to $\Bbb Q$ as totally ordered sets, you can take $A = \Bbb Q$ and define two different operations on $A$ by transporting the additive structure of $\Bbb Q(\sqrt 2)$ and $\Bbb Q(\sqrt 3)$ along the respective order-isomorphisms. Do let me know if you still think my answer is wrong - I may have misunderstood what you're asking. – Izaak van Dongen Nov 21 '23 at 16:40
  • 1
    It follows from this observation that your question is really equivalent to "if $H$ and $H'$ are ordered groups, and are isomorphic as groups and as linear orders, must they be isomorphic as ordered groups?" which is why that's what I answered. But certainly I could have given some more detail about that! Apologies for the confusion. – Izaak van Dongen Nov 21 '23 at 16:43
  • 1
    Thanks for your answer @Izaak van Dongen and for your effort to clarify it. I understand it better now and has been very helpful. – Crispost Nov 21 '23 at 18:03