Let $\mathcal L$ be the set of all group topologies on $\Bbb Z$. It is known that $(\mathcal L,\subseteq)$ is a modular complete lattice [1].

Is $(\mathcal L,\subseteq)$ distributive?


[1] Lamper, Milan. Complements in the lattice of all topologies of topological groups. Arch. Math. (Brno) 10 (1974), no. 4, 221--230 (1975).

  • $\begingroup$ What is a group topology on Z? Can it be isomorphic to a group topology on Q? On Z_p? $\endgroup$ – The Masked Avenger Sep 12 '14 at 15:42
  • $\begingroup$ A group topology on a group $G$ is a topology $\mathcal T$ on the set $G$ with which $(G,\mathcal T)$ is a topological group, that is, the function $(x,y)\mapsto xy^{-1}$ is continuous. $\endgroup$ – Minimus Heximus Sep 12 '14 at 15:45
  • 2
    $\begingroup$ Note that $(\mathcal{L}, \subseteq)$ is distributive if and only if the lattice $M_3$ ( en.wikipedia.org/wiki/Distributive_lattice#mediaviewer/… ) cannot be embedded in $(\mathcal{L}, \subseteq)$. So one idea might be to find three group topologies $\tau_1,\tau_2,\tau_3$ with identical infimum and identical supremum, or show that this cannot be done. $\endgroup$ – Dominic van der Zypen Sep 17 '14 at 7:41
  • 1
    $\begingroup$ There is also this result of Lukacs-Palfy: a group $G$ is abelian iff the subgroups lattice of $G \times G$ is modular (see a proof here). $\endgroup$ – Sebastien Palcoux Sep 25 '14 at 15:37
  • 3
    $\begingroup$ "something in the structure of $\mathbb{Z}$ other than abelianness, must be used to prove distributivity": Yes, perhaps the distributivity of its subgroups lattice. More generally (Ore's theorem) a group is locally cyclic iff its subgroups lattice is distributive (see a proof here) $\endgroup$ – Sebastien Palcoux Sep 26 '14 at 16:23

For each $\alpha \in S^1$, the map $\varphi_{\alpha} \colon \mathbb Z \to S^1$, given by $n \mapsto \alpha^n$, induces a topology $\tau_{\alpha}$ on $\mathbb Z$.

A basis of neighborhoods of $0$ for $\tau_{\alpha}$ is given by the sets $$U_{n,\alpha} := \left\{k \in \mathbb Z \mid |\alpha^k-1| < \frac1n \right\}, \quad n \in \mathbb N.$$

I denote by $\tau_{\alpha} \wedge \tau_{\beta}$ the largest group topology that is contained in $\tau_{\alpha} \cap \tau_{\beta}$ - which is in general different from $\tau_{\alpha} \cap \tau_{\beta}$.

Claim 1: If $\alpha,\beta \in S^1$ are irrational and such that $\alpha/\beta \in S^1$ is also an irrational angle, then $\tau_{\alpha} \wedge \tau_{\beta}$ is trivial.

Proof: Let $U \in \tau_{\alpha} \wedge \tau_{\beta} \subset \tau_{\alpha} \cap \tau_{\beta}$ and $0 \in U$. There exists some $V \in \tau_{\alpha} \wedge \tau_{\beta}$ such that $V-V \subset U$ and $0 \in V$. Then, there exists $n$, such that $U_{n,\alpha} - U_{n,\beta} \subset V - V \subset U$. However, $\mathbb Z=U_{n,\alpha} - U_{n,\beta}$. Indeed, for any $k \in \mathbb Z$, there exists some $m$, such that $|\alpha^m-1|<\frac1n$ and $|\beta^{k+m}-1|=|\beta^{m} - \beta^{-k}|<\frac1n$, because $(\alpha,\beta)$ generates a dense subgroup of $S^1 \times S^1$ and thus can approximate the point $(1,\beta^{-k})$ arbitrarily well. q.e.d.

Claim 2: $\tau_{\alpha} \vee \tau_{\beta}$ is the topology induced from the map $\mathbb Z \to S^1 \times S^1$, $n \mapsto (\alpha^n,\beta^n)$.

In particular, applying various automorphisms of $S^1 \times S^1$, we see that $$\tau_{\alpha} \vee \tau_{\beta} = \tau_{\alpha} \vee \tau_{\alpha\beta} = \tau_{\beta} \vee \tau_{\alpha\beta}.$$

Let $\alpha,\beta \in S^1$ be irrational angles, such that $\alpha/\beta$ is also irrational and consider the topologies $\tau_{\alpha},\tau_{\beta},\tau_{\alpha \beta}$. Any pairwise meet (in the lattice of group topologies) is the trivial topology (by Claim 1) and any pairwise join yields the same topology (by Claim 2). Hence,

enter image description here

embeds as a sub-lattice in the lattice of group topologies. Hence, the lattice of group topologies cannot be distributive.

| cite | improve this answer | |
  • $\begingroup$ By $V+V \subset U$ do you mean $V-V \subset U$? By an irrational angle $\alpha$ do you mean the angle between positive reals and the line passing through point $\alpha$ and 0 is irrational? $\endgroup$ – Minimus Heximus Nov 8 '14 at 20:08
  • $\begingroup$ I corrected $V+V$ to $V-V$. An angle $\alpha \in S^1$ is irrational, if $\alpha = \exp(2\pi i t)$, for some irrational real number $t$. $\endgroup$ – Andreas Thom Nov 8 '14 at 20:10
  • $\begingroup$ Thank you. I'll examine the answer more carefully and I'll award the bounty after it disappeared from featured questions. $\endgroup$ – Minimus Heximus Nov 8 '14 at 20:13
  • $\begingroup$ It seems I can prove your last double equations directly but I do not understand your proof using Claim2. Will you explain more? $\endgroup$ – Minimus Heximus Nov 10 '14 at 9:53
  • $\begingroup$ Yes, there are many proofs and it is elementary to see it directly. Using the automorphisms (for example $(z,w) \mapsto (z,zw)$) of $S^1 \times S^1$ just seemed more natural to me. Claim 2 is just the observation that the join of topologies is related to the diagonal embedding into the the product of topological spaces - again almost a triviality. $\endgroup$ – Andreas Thom Nov 10 '14 at 14:12

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.