Let $X = \{ 0, 1 \}$ and $X^{\mathbb N_0} = \{ x_0 x_1 x_2 \ldots : x_i \in X \}$ be the space of all infinite sequences, then a metric could be defined on it $$ d(u,v) := \frac{1}{2^r} \mbox{ with } r := \min\{ n : u_n \ne v_n \}. $$ Here $d(u,v) < \frac{1}{2^n}$ iff $u,v$ coincide in their first $n$ symbols. Regarding to this metric we have:

  • $U$ is open iff $U = W\cdot X^{\mathbb N_0}$ where $W$ is any collection of finite sequences (i.e. $W \subseteq X^* := X^0 \cup X^1 \cup X^2 \cup \ldots$) and $w \cdot x$ denotes the concatenation of a finite sequence and an infinite sequence $x$, yielding an infinite sequence
  • $U$ is closed iff $U$ has the property: if every prefix of a word $u$ is prefix of some word from $U$ then $u \in U$ (because $u \in \overline U$ iff every prefix of $u$ is a prefix of some word of $U$)
  • $U$ is clopen iff $U = W \cdot X^{\mathbb N_0}$ where $W$ is any finite collection of finite sequences

And the open balls are the sets of the form $v \cdot X^{\mathbb N_0}$ for some finite sequence $v$. Now this metric could be extended, for this define $I_n(u,v)$ to be true iff every infix up to a length $n$ of $u$ is contained as infix in $v$ and vice verse, namely $$ \begin{array}{lcl} I_n(u,v) & \Leftrightarrow & \forall i \exists j : u_{i+k} = v_{j+k}, k = 0,\ldots,n-1\\ & & \forall i \exists j : v_{i+k} = u_{j+k}, k = 0,\ldots,n-1 . \end{array} $$ and define $$ d'(u,v) := \max\left\{ d(u,v), \frac{1}{2^r} \right\} \mbox{ with } r := \max\{ n : I_n(u,v) \}. $$ Here $d'(u,v) < \frac{1}{2^n}$ iff $u,v$ coincide in all their infixes (including the first $n$ symbols) up to a length of $n$. Now $d(u,v) \le d'(u,v)$, so that convergence in $d'$ implies convergence in $d$, and every set that is open with regard to $d'$ is open with regard to $d$ (and also for closed sets), so that the topology induced by $d'$ is a refinement of the topology induced by $d$. Now I want to get results about the open sets and the topology induced by $d'$, and their relations. I already know that $d'$ is not complete.

In particular what closed sets with regard to $d$ are open with regard to $d'$ (I already know that not all $d$-closed sets are $d'$-open)? Any suggestions or hints how to attack this problem?

  • $\begingroup$ By your definition, if $u\neq v$, then $d'(u,v)=1$ since some $r_i$ will be $0$ (namely, any $i$ such that $u_i\neq v_i$). $\endgroup$ – Eric Wofsey Nov 26 '13 at 12:29
  • $\begingroup$ Yes, my first definition was totally wrong. I corrected it now. $\endgroup$ – StefanH Nov 26 '13 at 13:00
  • $\begingroup$ Isn't this the usual product topology? The $1/2^r$-ball around a sequence $a$ consists of sequences that have the same first $r$ coordinates as the sequence $a$. $\endgroup$ – Martin Sleziak Nov 26 '13 at 13:45
  • $\begingroup$ The first metric induces the usual product topology, but the second induces a refinement thereof which properties I am interested in. $\endgroup$ – StefanH Nov 26 '13 at 13:47
  • $\begingroup$ Normally infixes are used when considering bi-infinite sequences. $\endgroup$ – Benjamin Steinberg Nov 26 '13 at 14:42

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.

Browse other questions tagged or ask your own question.