MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Suppose A and B are compact connected sets in the XY plane and XZ plane respectively in R^3. Suppose further that the the range of x-values taken by A and B are the same (i.e, projections of A and B onto the x-axis are the same closed interval). Is there always a connected set in R^3 whose projections onto XY and XZ planes are A and B respectively?

share|cite|improve this question
up vote 6 down vote accepted

The answer is yes.

Let $\alpha,\beta:[0,1]\to[0,1]\times \mathbb R$ be two paths; $\alpha(t)=\left(\alpha_1(t),\alpha_2(t)\right)$ and $\beta(t)=\left(\beta_1(t),\beta_2(t)\right)$. Assume that $\alpha_1(0)=\beta_1(0)=0$, $\alpha_1(1)=\beta_1(1)=1$.

Claim. The points $a=\left(0,\alpha_2(0),\beta_2(0)\right)$ and $b=\left(1,\alpha_2(1),\beta_2(1)\right)$ lie in the same connected coponent of the set $\Sigma\subset \mathbb R^3$ formed by points of the following type $ \left(\alpha_1(t),\alpha_2(t),\beta_2(\tau)\right)$ such that $\alpha_1(t)=\beta_1(\tau)$.

Proof. Note that for generic smooth choice of $\alpha$ and $\beta$ the set $\Sigma$ is a smooth 1-dimensional manifold which might be not connected, but it has only two boundary points in $a$ and $b$. Thus, in this case one can connect these points by a curve. The general case can be done by approximation. $\square$

The rest is easy: one can approximate $A$ and $B$ by path-connected sets $A'$ and $B'$ with the same property. Thus one can present $A'$ and $B'$ as a union of curves $\alpha$ and $\beta$ with the above property. Moreover we can assume that the ends of $\alpha$ and $\beta$ are fixed (i.e. $a$ and $b$ are fixed). For each pair $(\alpha,\beta)$, choose the connected component of $a$ in $\Sigma$ and take the union of all of them.

share|cite|improve this answer
Nice answer!. Could you point me to a reference to the fact: "a connected set in the plane can be approximated by path connected set". – arun s Mar 8 '10 at 7:53
A connected open set in the plane has to be path-connected (that is standard fact and it is easy to prove). So you may pass to $\epsilon$-neighborhood of your set (and take its closure if you want to keep it compact). – Anton Petrunin Mar 8 '10 at 15:44

Edit: This answer is wrong. But, since Anton based his answer on this idea, I am leaving it up.

No. Let $A$, resp. $B$, be the graph $x=f(y)$, resp. $x=g(z)$, for some continuous functions $f$ and $g$. Pick numbers $0<d<b<a<c<1$, and arrange things so that $f(0)=g(0)=0$, $f(1)=g(1)=1$, and $f(y)$ grows from 0 to $a$, then decreases to $b$, then finally increases to 1, whereas $g(z)$ grows from $0$ to $c$, decreases to $d$, and increases to 1. (Here, $a$, $b$, $c$, $d$ are function values, i.e., values of $x$, not of the arguments $y$ or $z$.)

Now if $C$ projects to $A$ and $B$, respectively, it is not hard to see that $(0,0,0)\in C$ and $(1,1,1)\in C$. Trying to move $(x,y,z)$ continuously from the former to the latter, a contemplation of how $x$ must grow and shrink in order to maintain $x=f(y)$ and $x=g(z)$ leads to a contradiction. (Hard to put in words, but a pair of graphs reveals it I think.)

share|cite|improve this answer
No, it is connected. – Anton Petrunin Mar 7 '10 at 18:27
Rats. You're right. I had mislabeled one of my pictures. – Harald Hanche-Olsen Mar 7 '10 at 19:07
It was nice mistake, I made my answer based on it :) – Anton Petrunin Mar 7 '10 at 19:12
To the "edit": No, you can not make an example this way; it follows directly from my answer. (It also contains an answers to the original question, but you need some extra technical steps.) – Anton Petrunin Mar 7 '10 at 20:09
Right; I see it now. I edited the “edit”. – Harald Hanche-Olsen Mar 7 '10 at 22:54

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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