5
$\begingroup$

Let $(X,d)$ be a metric space and let $q(x,y)={d(x,y)\over 1+d(x,y)}$ be another metric(which we know it is). Show that $d(x,y)$ and ${d(x,y)\over d(x,y)+1}$ are topologically equivalent.

I used the definition that metrics $f,g$ on metric spaces $(X,d/ q)$ are called equivalent if $U\subset X$ is open with respect to $f$ if and only if it is open with respect to $g$.

I managed one direction: $q(x,y)\le d(x,y)$. Let $U\subset X$ be open with respect to $q$ and let $x\in U$. So there exists $r>0$ such that $B_q(x,r)\subset U$. For any $y\in X$, such that $d(x,y)<r$, $q(x,y)\le d(x,y)<r$. That implies that $y\in B_d(x,r)\Rightarrow y\in B_q(x,r)\Rightarrow B_{d}(x,r)\subset B_q(x,r)\subset U$ (So I hope. All those transitions seem a little counterintuitive.). With that uncertainty in mind, I am a little lost in the other direction.

I would appreciate any help on the issue.

$\endgroup$
2
  • 1
    $\begingroup$ Probably will be easier to work with this, more explicit, definition of equivalency: $d$ and $q$ are equivalent metrics in $X$ if for any $x\in X$ and any $\epsilon>0$ exists $r>0$ that $$\Bbb B_q(x,r)\subseteq\Bbb B_d(x,\epsilon)$$ and for any $\gamma>0$ exists some $s>0$ such that $$\Bbb B_d(x,s)\subseteq\Bbb B_q(x,\gamma)$$ $\endgroup$
    – user173262
    Commented Nov 20, 2016 at 2:19
  • $\begingroup$ This is what I would arrive at as soon as I find those parameters.. Arriving at that is not what I struggle in, but finding the right parameters and the right directions. I am a little baffled there... $\endgroup$
    – Meitar
    Commented Nov 20, 2016 at 2:22

2 Answers 2

4
$\begingroup$

Hint: use

$0 \le \frac{x}{2} \le \frac{x}{1+x} \le x$ for $0\le x \le 1$. The last inequality holds for all $x\ge0$

So $\frac{1}{2}d(x,y)\le q(x,y)\le d(x,y)$

More detail:

Take $B_d(x,\epsilon)$, take $\epsilon_1=\min\{\epsilon,1\}$, take $r=\frac{\epsilon_1}{2}$

Then $B_q(x,r)\subseteq B_d(x,\epsilon_1)\subseteq B_d(x,\epsilon)$

Conversely, take $B_q(x,\epsilon)$ and $r=\epsilon$

Then $B_d(x,r)\subseteq B_q(x,r)=B_q(x,\epsilon)$

$\endgroup$
7
  • $\begingroup$ I am trying to use it but I am still clueless as for what it legitimate and what isn't. I want to arrive at: $a\cdot d(x,y)\le q(x,y)$. Taking $d(x,y)<r$, I am looking for $b$ satisfying $\q(x,y)<b$, so that $ad(x,y)\le q(x,y)<ra$, which requires me to solve: $d(x,y)\le {q(x,y)\over a}<{b\over a}=r$. Am I in the right direction? $\endgroup$
    – Meitar
    Commented Nov 20, 2016 at 2:44
  • $\begingroup$ I modified it to a better inequality that would do what you want. $\endgroup$
    – Momo
    Commented Nov 20, 2016 at 2:57
  • $\begingroup$ before using $d(x,y)\leq 2d’(x,y)$ fact, how to show $d(x,y)\leq 1$? Because $x=d(x,y)$. Link: math.stackexchange.com/q/4370772/861687 $\endgroup$
    – user264745
    Commented Jan 31, 2022 at 21:06
  • $\begingroup$ @user264745 $d(x,y)\le 1$ can be assumed to be WLOG. As in my answer, if the ball centered in $x$ of radius $\epsilon$ is a subset of $U$, then the ball centered in $x$ of radius $\min\{\epsilon, 1\}$ is also a subset of $U$. And the later ball has the property $d(x,y)\le 1$ for $y$ in ball. $\endgroup$
    – Momo
    Commented Jan 31, 2022 at 22:31
  • $\begingroup$ But when we work with $B_{d’}(x,r)$, and try to show $y\in B_{d’}(x,r) \Rightarrow y\in B_d (x, \epsilon_{1})$. We don’t know yet $y\in B_d (x, \epsilon_{1})$, which implies $d(x,y)\lt \epsilon_{1} \leq 1$. Let $y\in B_{d’}(x,r)$. Then $d’(x,y)\lt r$. From this information, how can we say $d(x,y)\leq 1$? $\endgroup$
    – user264745
    Commented Feb 1, 2022 at 7:36
1
$\begingroup$

For one direction: we have that for any $x\in X$ and any $\epsilon>0$ exists $r>0$ such that

$$y\in\Bbb B_d(x,r)\implies y\in\Bbb B_q(x,\epsilon)$$

this mean that

$$d_d(x,y)<r\implies d_q(x,y)=\frac{d_d(x,y)}{1+d_d(x,y)}<\epsilon$$

If $\frac{r}{1+r}<\epsilon$ and when $r$ decreases then the LHS decreases (i.e. the function $\frac{r}{1+r}$ is increasing) we are done. Then we can check that choosing $r<\frac{\epsilon}{1-\epsilon}$ works whenever $\epsilon<1$.

If $\epsilon\ge 1$ then using any $\epsilon^*<1$ for the bound $r<\frac{\epsilon^*}{1-\epsilon^*}$ would work (in particular $r=1$ works whenever $\epsilon\ge1$).

In general $r=\epsilon$ is an easy solution too.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .