Endow the set $\mathbb N$ of positive integers with the topology $\tau$ generated by the base consisting of arithmetic progressions $a+b\mathbb N_0$ where $\mathbb N_0=\{0\}\cup\mathbb N$, where $a,b\in\mathbb N$. This topology is often referred to as the Furstenberg topology or the profinite topology. The space $\mathbb N_\tau:=(\mathbb N,\tau)$ is homeomorphic to the space $\mathbb Q$ of rational numbers (being a second-countable regular countable space without isolated points). So $\mathbb N_\tau$ has many non-trivial homeomorphisms. But I know no non-trivial homeomorphism of $\mathbb N_\tau$ which would be multiplicative is the sense that $f(x\cdot y)=f(x)\cdot f(y)$ for any $x,y\in\mathbb N$.

Problem 1. Is the identity function a unique multiplicative homeomorphism of $\mathbb N_\tau$?

An affirmative answer to this problem follows from an affirmative answer to

Problem 2. Are there prime numbers $a,b,c$ such that for any $d\in\mathbb N$ there exists $n\in\mathbb N$ such that $a^n\equiv 1\!\!\!\mod\! d\;\;$ but $\;\;b^n\not\equiv 1\!\!\!\mod\! c$?

  • $\begingroup$ I'm always confused when someone calls $\mathbb{N}$ the set of positive integers... (+1 to the question, anyway!) $\endgroup$ – Qfwfq Sep 9 '18 at 10:50
  • $\begingroup$ @Qfwfq Thank you for your +1. Concerning the notations, I arrived to the following compromise: $\omega=\{0,1,2,\dots\}$ is the set of finite ordinals but $\mathbb N$ of positive integers. Otherwise, it will be not faire: two notations for the set $\{0,1,2,3,\dots,\}$ and none for $\{1,2,3,\dots\}$. Moreover, according to the Eastern Ortodox traditions, we start counting from 1, not from 0. So, the lowest floor in our country is the 1-st floor, not the ground floor. So, for us (at least) the set $\{1,2,3,\dots\}$ for natural numbers is more natural than $\{0,1,2,3,\dots\}$. $\endgroup$ – Taras Banakh Sep 9 '18 at 13:21
  • $\begingroup$ @Qfwfq The history of mathematics witnesses that for long time zero was not considered to be a (natural) number. It simply did not exist. $\endgroup$ – Taras Banakh Sep 9 '18 at 13:24
  • $\begingroup$ I see your pont, ad it makes sense... It's just a habit and psychological think for me ;) $\endgroup$ – Qfwfq Sep 9 '18 at 15:38


First observe that the automorphisms of the semigroup $\mathbf{N}^*$ (which you denote $\mathbb{N}$) are induced by permutations of primes.

Consider the automorphism $f$ induced by the transposition $(2,3)$ (thus, mapping $2^a.3^b.c$ to $2^b.3^a.c$, $c$ coprime to 6).

I claim that $f$ is continuous. Indeed, consider any convergent sequence $m_i\to m$. Thus, $m_i=m+r_i$ with $r_i$ tending to 0 in the profinite completion of $\mathbf{Z}$ (that is, for every $n\ge 1$ there exists $i_0$ such that $n$ divides $r_i$ for all large $i$).

Write $m=2^a.3^b.c$ with $c$ coprime to 6. There exists $i_0$ such that $2^{a+1}3^{b+1}$ divides $r_i$ for all $i\ge i_0$. So, for $i\ge i_0$, $m_i=2^a.3^b.c+2^{a+1}3^{b+1}.t_i$ for some $t_i$. So $m_i=2^a3^b(c+6t_i)$ with $c$ coprime to 6. Thus $f(m_i)=2^b3^a(c+6t_i)=f(m)+r'_i$, with $r'_i=2^{b-a}3^{a-b}r_i$. Thus $r'_i$ also tends to 0 in the profinite completion of $\mathbf{Z}$. Thus $f(m_i)$ tends to $f(m)$. Hence, $f$ is continuous.

The argument adapts immediately to any finitely supported permutation of the set of primes.

  • 2
    $\begingroup$ Remark: there exists a permutation of primes that induces a non-continuous monoid automorphism. Indeed, from Dirichlet's theorem, there exist two sequences of primes $(p_n)$, $(q_n)$ such that $p_n$ tends to 1 in the profinite completion, and $q_n$ equals 2 modulo 3. We can suppose they are injective with disjoint image. Let $t$ exchange $p_n$ and $q_n$ for all $n$, and induce $f$. Then $p_n\to 1$ while $f(p_n)=q_n$ does not tend to $1=f(1)$. $\endgroup$ – YCor Sep 8 '18 at 23:46
  • $\begingroup$ Additional remark: there also exist infinitely supported permutations $\sigma$ (actually $2^{\aleph_0}$ many) of primes inducing a continuous monoid automorphism $f_\sigma$. Namely, consider a sequence of primes $(p_n)_{n\ge 0}$ tending to 1 in the profinite topology (as in my previous comment), and any permutation $\sigma$ supported on $\{p_n:n\ge 0\}$. Then $f_\sigma$ is continuous. This is not hard to check (one checks that in general, $f_\sigma$ continuous at 1 implies $f_\sigma$ continuous, and for such a choice of $\sigma$, one checks continuity at 1). I can give details upon request. $\endgroup$ – YCor Sep 9 '18 at 8:22
  • $\begingroup$ But $\mathbb{N}^*=\beta\mathbb{N}\setminus\mathbb{N}$. $\endgroup$ – Ramiro de la Vega Sep 9 '18 at 12:41
  • $\begingroup$ @RamirodelaVega I explicitly used $\mathbf{N}^*$ to mean $\{n\in\mathbf{Z}:n>0\}$; I'm aware that notation varies. $\endgroup$ – YCor Sep 9 '18 at 13:19
  • $\begingroup$ It is interesting to remark that such "finitely supported" multiplicative biejctions of $\mathbb N$ are discontinuous in the Golomb topology (generated by the arithmetic sequences $a+b\mathbb N_0$) because of the density of primes in this topology. $\endgroup$ – Taras Banakh Sep 9 '18 at 13:34

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.