3
$\begingroup$

Let $F = G\cup H$ where $G = \{(x,y) \in \mathbb{R}^2 : x=0, y = [0,1]\cap \mathbb{Q}\}$ and $H = \{(x,y) \in \mathbb{R}^2 : x > 0, y \in [0,1]\backslash \mathbb{Q}$. Find the connected components and path components of $F$.

I know that connected components are equivalence classes under the "is connected to" relation (i.e. $a\cong b$ iff there is a connected set containing $a$ and $b$). Also, path components are equivalence classes under the "is joined by a path to" relation, where a path between points $a,b \in A$ is a continuous function $\alpha : [0,1]\to A, \alpha(0)=a, \alpha(1)=b.$ I know that singleton sets in $\mathbb{Q}$ are connected and the union of connected sets with nonempty intersection is connected. I also know that if $A$ is connected, then $\overline{A}$ is connected. I think every subset $S$ of $G$ with two or more points is disconnected; for an irrational number $r$ in $(0,1),$ the sets $\{(0, y) : y \in (-\infty, r)\}$ and $\{(0,y) : y \in (r,\infty)\}$ are open sets that separate $S$. The fact that the rationals are dense in $\mathbb{R}$ may be useful. I think the connected components of $G$ are singleton sets. Are the connected components of $H$ not also singleton sets?

As for path components, I think points in $G$ and $H$ belong to different path components, and using the density of the rationals might be useful for this. But I'm not sure exactly what the path components should be.

Edit: I think the components of $H$ are $\{(x,y) : x > 0, y = r\}$ for irrational points $r$. But I'm not sure how to show that the singleton sets in $G$ are the connected components in $F$ and that the components in $H$ are components in $F$.

$\endgroup$
1
  • $\begingroup$ The set $F$ is connected (as I argue in my answer) but removing any point $(0,y)$ with rational $y$, with $0<y<1$ makes it disconnected. There is a more extreme example of a connected set, such that removing one particular point makes it totally disconnected (i.e. every component is a singleton). This more complicated example is called the Knaster–Kuratowski fan, see en.wikipedia.org/wiki/Knaster%E2%80%93Kuratowski_fan and math.stackexchange.com/questions/3167685/… $\endgroup$
    – Mirko
    Jun 30, 2021 at 1:40

3 Answers 3

2
+50
$\begingroup$

The path-components of $F$ are the singletons in $G$, as well as horizontal lines (more precisely positive open half-rays) in $H$. On the other hand, $F=G\cup H$ is connected (and hence is its own component, just one.)

To prove the statement about path-components, say there was a path from a point $h\in H$ to a point $g\in G$, say $h=(z,p)$ and $g=(0,q)$, where $z>0$, $p$ is irrational, $q$ is rational, and both $p,q\in[0,1]$. We may assume without loss of generality that $\gamma:[0,1]\to F$ is a path with $\gamma(0)=h$ and $\gamma(1)=g$. Let $L_p:=\{(x,p):x>0\}$ and define $T=\{t\ge0:\gamma([0,t])\subseteq L_p\}$ and $s=\sup T.$ Left to the reader (ask for details if needed): the set $T$ is both open and closed, and $0\in T$, hence $T=[0,1]$ and $s=1$, hence (by continuity) $\gamma(1)\in L_p$, contradicting $\gamma(1)=g$.

The above shows more generally that if $\gamma$ is any path containing at least one point in the set $L_p:=\{(x,p):x>0\}$ (for some fixed irrational $p\in[0,1]$) then $\gamma([0,1])\subset L_p$. It remains to show that if $q\in[0,1]$ is rational, then singleton $\{(0,q)\}$ is a path-component of $F$. The above argument shows that no path starting at $(0,q)$ could contain any point in $H$. But it is quite obvious that if the path has to stay in $G$ all the time, then it has to stay at $(0,q)$ all the time.

It remains to show that $F$ is connected. Suppose there was a non-empty closed-and-open set $U\subseteq F,$ we will show that $U=F.$ As before, if $p$ is any irrational in $[0,1]$, let $L_p:=\{(x,p):x>0\}$. Observe that for any irrational $p$ in $[0,1]$, if $\varnothing\not= L_p\cap U$ then $L_p\subset U$. We cannot have $U\subseteq G$ (since then the non-empty $U$ could not be open in $F$), hence it follows that $\varnothing\not= H\cap U$. Fix any $z>0$ and irrational $p$ in $[0,1]$ with $h:=(z,p)\in H\cap U.$ Let $r=\sup Y$ where $Y=\{y\ge p: y \in [0,1]\setminus\mathbb{Q},\ L_y\subset U\}.$ If $r$ is irrational, then using that $U$ is closed, we must have $r\in Y$, but then using that $U$ is open one easily shows that there must be $\varepsilon>0$ with $r+\varepsilon\in Y$, contradicting that $r=\sup Y.$ Hence $r$ is rational. We will show that $r=1.$ Suppose to the contrary that $p<r<1.$ Since $U$ is closed (and since $L_y\subset U$ for every irrational $y$ with $p\le y<r$) we must have $(0,r)\in U.$ But, since $U$ is open, some neighborhood $N$ of $(0,r)$ must be contained in $U$, and, using that $r<1$, it follows that $N$ must contain the points $(\delta, r+\delta)$ for all small enough $\delta>0$ (and with $\delta$ irrational). Hence $U$ must contain $L_{r+\delta}$ for all small enough irrational $\delta>0$, contradicting the definition of $r=\sup Y.$ This proves $r=1$ and it follows that every point $(a,b)$ in $F$ with $a\ge0$ and $p\le b\le1$ must belong to $U.$ In a similar way one proves that every point $(a,b)$ in $F$ with $a\ge0$ and $p\ge b\ge0$ must belong to $U,$ hence $U=F.$ This completes the proof that $F$ is connected, a nice problem, thank you!

Edit 12 July, 2021. Addressing the following comments.

Comment (to be clarified). Could you provide more details on why $T$ is both open and closed (by using the definitions and giving formal proofs, ideally)?

More details. So, $T=\{t\ge0:\gamma([0,t])\subseteq L_p\}$ (where, as above, $\gamma(0)=h=(z,p)$ and $L_p:=\{(x,p):x>0\}$). We first show that $T$ is closed. Clearly $0\in T$ and if $t\in T$ and $0\le t'<t$ then also $t'\in T$. It follows that if $s=\sup T$, then $T$ is equal to either $[0,s)$ or $[0,s]$. (Note that if $s=0$ then clearly $s\in T=\{0\}$ which is closed, so we may assume below that $s>0.$) We will show that $T=[0,s]$ (which is clearly closed). We definitely have that $[0,s)\subseteq T$ (since if $0\le t'<s$ then there is some $t\in T$ with $t'<t<s$, hence $t'\in[0,t]\subseteq T$). By continuity, $\gamma(s)=\lim\limits_{t\nearrow s}\gamma(t).$ Let $\gamma(t)=(\xi(t),\eta(t))$, that is $\xi(t)$ and $\eta(t)$ denote the $x$- and the $y$-coordinates of $\gamma(t)$, respectively. We have $\eta(s)=\lim\limits_{t\nearrow s}\eta(t)=\lim\limits_{t\nearrow s} p=p.$ That is, $\gamma(s)=(\xi(s),p)$ and it has to be the case that $\xi(s)>0$ and $\gamma(s)\in L_p.$ Hence $s\in T$ and $T=[0,s]$ which completes the proof that $T$ is closed. Next we prove that $T$ is open in the relative topology of $[0,1].$ If $s=1$ there is nothing to prove since $T=[0,1]$ which is open in $[0,1].$ So, assume it were the case that $0\le s<1.$ Since $\xi(s)>0$, by continuity there is some $\varepsilon>0$ such that $\xi(s+\delta)>0$ whenever $0\le\delta\le\varepsilon.$ We will show that $(s,s+\varepsilon]\subseteq T.$ Assume to the contrary that there is some $t'$ with $s<t'\le s+\varepsilon$ such that $\gamma(t')\not\in L_p.$ This means that if $p'=\eta(t')$ then $p'\not=p.$ But then the restriction $\gamma:[s,t']\to H$ would be a continuous path remaining in $H$ all the time and connecting a point in $L_p$ to a point in $L_{p'}$ which is a contradiction since $L_p$ and $L_{p'}$ are different components of $H.$ This completes the proof that $T$ is open (and, hence, that it is both closed and open).

Comment (to be clarified). Also, could you formalize the statement "but it is quite obvious that if the path has to stay in $G$ all the time, then it has to stay at $(0,q)$ all the time."?

Formalization (and justification). By "the path has to stay in $G$ all the time" we mean that $\gamma(t)\in G$ for all $t\in[0,1]$ or in other words $\gamma([0,1])\subseteq G$, or we could also say $\gamma:[0,1]\to G$ (rather than the correct but not so specific $\gamma:[0,1]\to F$). Similarly, by "It has to stay at $(0,q)$ all the time" we mean that it has to be the constant path $\gamma(t)=g=(0,q)$ for all $t\in[0,1].$ Indeed, if $\gamma:[0,1]\to G$ were not constant, then there is some $t'$ such that $\gamma(t')=g'=(0,q')\not=(0,q).$ But then the restriction $\gamma:[0,t']\to G$ would be a path in $G$ connecting $(0,q)$ to $(0,q')$ which is a contradiction (since the components, and hence the path-components of $G$ are singletons).

Comment (to be clarified). Also, could you further justify the statement "If $r$ is irrational, then using that $U$ is closed, we must have $r\in Y$, but then using that $U$ is open one easily shows that there must be $\varepsilon>0$ with $r+\varepsilon\in Y$, contradicting that $r=\sup Y.$"?

Further justification. By construction, $r\ge p.$ If $r$ were irrational and if $r\not\in Y,$ then $r>p$ (since $p\in Y$). Consider the point $(1,r)$ (where $1$ is picked somewhat arbitrary, any positive number in its place would do). Pick $y_n\in Y$ with $p<y_0<...<y_n<y_{n+1}<...<r$ and with $\lim\limits_{n\to\infty}y_n=r.$ Then $(1,y_n)\in U$ for all $n$, and the sequence of points $(1,y_n)$ converges to the point $(1,r)$, and since $U$ is closed it follows that $(1,r)\in U$. This in turn implies that $L_r\subseteq U$ and hence $r\in Y.$ Next we clarify that part of the statement that says: "but then using that $U$ is open one easily shows that there must be $\varepsilon>0$ with $r+\varepsilon\in Y.$" Indeed, since $U$ is open, there must be some open ball $B$ centered at $(1,r)$ such that $B\cap F\subseteq U$. Then $(1,r+\varepsilon)\in U$ whenever $r+\varepsilon$ is irrational and $\varepsilon$ is smaller than the radius of $B$. For such $\varepsilon$ we must have $r+\varepsilon\in Y.$

$\endgroup$
4
  • $\begingroup$ Could you provide more details on why $T$ is both open and closed (by using the definitions and giving formal proofs, ideally)? Also, could you formalize the statement "but it is quite obvious that if the path has to stay in $G$ all the time, then it has to stay at $(0,q)$ ..."? $\endgroup$
    – Alfred
    Jul 11, 2021 at 17:51
  • $\begingroup$ Also, could you further justify the statement "if r is irrational, then using that $U$ is closed, we must have $r\in Y$, but then using that $U$ is open one easily shows that ..."? $\endgroup$
    – Alfred
    Jul 11, 2021 at 17:55
  • $\begingroup$ "ask for details if needed": Someone has done it in math.stackexchange.com/q/4196133 $\endgroup$
    – Paul Frost
    Jul 12, 2021 at 23:16
  • $\begingroup$ @Paul The question there seems suspiciously similar to the question here, and seems to have copied and used (with some modification) the notation in my answer here. I wonder if these are really different users that asked these two questions. I voted your answer there up (but will read it later ... want to post a slightly different answer here first, before I read all the details of that one ... though I suspect it would be the same answer with different wording :) $\endgroup$
    – Mirko
    Jul 13, 2021 at 0:30
0
$\begingroup$

As you mentioned in your question, the set $\mathbb{Q}$ is totally disconnected, therefore each component of $G$ is singleton. On the other hand, in set $H$, for each irrational point $y$ in $[0,1]$, the set $S_y=\{(a,b) \in \mathbb{R} ^2 : a > 0, b=y \}$ lies in the same component because $S_y$ is homomorphism with $(0, \infty)$. In fact, $S_y$ is the component of $H$ since if we adjoint any point $(a,b)$ with $b \neq y$ (assume $b < y$) to the set $S_y$, then the there exist a number $q$ such that $b < q <y$, the set $A=\{(x,y) \in \mathbb{R} ^2 : y > q \} \cap S_y$ and set $\{(a,b)\}$ form an separation of $S_y$, therefore $S_y$ is no longer connected, hence, $S_y$ must be the component.

The path component of $G$ and $H$ are the same as the component. You can easily verify this by the fact that path component lies in component.

$\endgroup$
1
  • $\begingroup$ What you write (about the components and path components of $G$, and of $H$) is correct, but you do not address the question as to what are the components and the path-components of $F$ (which is $F=G\cup H$). I believe that for $F$, the path components are just as in your answer, but that on the other hand, $F$ is connected (and I posted an answer arguing this). Thank you $\endgroup$
    – Mirko
    Jun 30, 2021 at 0:19
0
$\begingroup$

I am posting a second answer, even if my first one is correct, and in a linked question there is also a relevant answer (by another user). $\def\la{\langle} \def\ra{\rangle}$

I will just prove here that $F$ is connected (hopefully in an easier to read way, than in my first answer). Notationally, $\la x,y\ra$ is a point with coordinates $x,y$, while $(a,b)$ is an interval with endpoints $a,b.$

Suppose that $U$ is a non-empty open and closed subset of $F$. We will show that $U=F.$ Let $V=\{y:\exists x\ge0, \la x,y\ra\in U\}.$ As in my other answer if $p\in[0,1]$ is any irrational, let $L_p:=\{\la x,p\ra:x>0\}.$ Since each $L_p$ is connected, it is clear that if $L_p\cap U\not=\varnothing$ for some $p$ then $L_p\subseteq U.$

Claim 1. $V$ is open and closed in $[0,1]$ (and hence $V=[0,1]$).
Proof. Take any $y\in V.$ If $y$ is irrational, then $L_y\subseteq U$ and in particular $\la 1,y\ra\in U.$ Since $U$ is open there is some $\varepsilon>0$ such that $\la 1,y+t\ra\in U$ whenever $y+t$ is irrational and $-\varepsilon<t<\varepsilon$. For such $t$ we have $L_{y+t}\subseteq U$.
If $q\in(y-\varepsilon,y+\varepsilon)$ is rational then $\la0,q\ra=\lim\limits_{n\to\infty}\la\frac1n,q+\frac{\sqrt2}n\ra$. Note that $\la\frac1n,q+\frac{\sqrt2}n\ra\in L_{q+\frac{\sqrt2}n}\subseteq U$ for all large enough $n$ : more precisely, whenever $q+\frac{\sqrt2}n\in(y-\varepsilon,y+\varepsilon),$ i.e. whenever $\frac{\sqrt2}n<y+\varepsilon-q$. Using that $U$ is closed it follows that $\la0,q\ra\in U$ and $q\in V$.
Thus $(y-\varepsilon,y+\varepsilon)\subseteq V$, showing that every irrational $y\in V$ is an interior point of $V.$
Now suppose that $y\in V$ is rational. Then $\la0,y\ra\in U$ and since $U$ is open there is $\varepsilon>0$ such that $\la |t|,y+t\ra\in U$ whenever $t$ is irrational with $|t|<\varepsilon$ (and with $y+t\in[0,1]$). It follows that $L_{y+t}\subseteq U$ for all such $t$.
If $q\in(y-\varepsilon,y+\varepsilon)\cap[0,1]$ is rational then (similar to an earlier argument) we have $q\in V,$ hence $(y-\varepsilon,y+\varepsilon)\cap[0,1]\subseteq V$ which shows that every rational $y\in V$ is an interior point of $V.$
This completes the proof that $V$ is open.

Next we will show that $V$ is closed. Suppose that $y_n$ is a sequence of points in $V$, and that $y=\lim\limits_{n\to\infty}y_n.$ We need to show that $y\in V.$ Using that $V$ is open, we may assume without loss of generality that each $y_n$ is irrational (otherwise replace $y_n$ with some irrational $z_n\in(y_n-\frac1n,y_n+\frac1n)\cap V$; notice that this will not change the limit value $y$).
If $y$ is rational then $(0,y)=\lim\limits_{n\to\infty}(\frac1n,y_n)$ and since $(\frac1n,y_n)\in L_{y_n}\subseteq U$ and $U$ is closed, it follows that $(0,y)\in U$ and $y\in V.$
If $y$ is irrational then $(1,y)=\lim\limits_{n\to\infty}(1,y_n)$ and since $(1,y_n)\in L_{y_n}\subseteq U$ and $U$ is closed, it follows that $(1,y)\in U$ and $y\in V.$ This completes the proof that $V$ is closed, and that Claim 1 holds.

Claim 2. $U=F.$ This easily follows from Claim 1 (namely that $V=[0,1] $), using that if $L_p\cap U\not=\varnothing$ for some irrational $p$ then $L_p\subseteq U.$

Since the only nonempty closed and open subset of $F$ is $F$ itself, it follows that $F$ is connected.

$\endgroup$

You must log in to answer this question.

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