I know that the following exercise you can find on internet maybe with solution too, but I want to know, if my "solutions" are correct.

Let $X,Y\subset \mathbb{R}^n$, $X+Y=\{x+y;x\in X, y\in Y\}$. Prove it or find a counterexample.

1) If X, Y open, then X+Y is open

2) If X, Y closed, then X+Y is closed

3) If X, Y compact, then X+Y is compact

My ideas:
1) is true. Could you have a look if my solution is correct? My try: We know that for all $x\in X \exists \epsilon >0: B_{\epsilon}(x)=\{b\in X; \|b-x\|<\epsilon\}\subset X$, and for all $y\in Y \exists \epsilon' >0: B_{\epsilon'}(x)=\{b'\in X; \|b'-x\|<\epsilon'\}\subset Y$, because $X$ and $Y$ are open. Now: $\|z-(x+y)\|=\|z-x-y\|=\|z'-z'+z-x-y\|=\|z'-x+z-z'-y\|\le\|z'-x\|+\|z-z'-y\|<\epsilon+\epsilon'$. Define $\eta =\epsilon +\epsilon'$. We have: for $x+y\in X+Y \exists \eta >0, \eta =\epsilon +\epsilon': B_{\eta}(x+y)=\{z\in X+Y; \|z-(x+y)\|<\eta\}\subset X+Y$.

I'm not sure, if this is correct. If my "solution" is false, could you help to correct?

2) First I said that this is true, but after googeling I found out that you can find a counterexample. Could you help me to find the mistake of my "solution"? My try: For every sequence $(x_n)\subseteq X$ such that $x_n\to x_0$, $x_0\in \mathbb{R}^n$, it is $x_0\in X$, because X is closed. For every sequence $(y_n)\subseteq Y$, such that $y_n\to y_0$, $y_0\in \mathbb{R}^n$, it is $y_0\in Y$, because Y is closed.

Let $(z_n)\subseteq X+Y$ be a sequence, $z_n=x_n+y_n$ for every $n\in\mathbb{N}$ and let $z_n\to z_0 \in \mathbb{R}^n$. But it is $z_n=x_n+y_n\to x_0+y_0$ and by the uniqueness of limits $z_0=x_0+y_0\in X+Y$, because X and Y are closed.

But my try has to be wrong I think, because you find counterexamples for 2). And I don't find the mistake :(, could you help me?

3) Is it correct? I would say yes. Maybe I can prove it with a continuous function and $X+Y$ as it's compact image.


  • 2
    $\begingroup$ 2) The worry is that there's some sequence $(x_n+y_n)$ that converges, but not to a point in $X+Y$. As you've shown, if $(x_n)$ and $(y_n)$ converge, then this sequence must converge to a point in $X+Y$. But what if neither $(x_n)$ nor $(y_n)$ converged at all? Their sum still might. $\endgroup$ – user98602 May 12 '15 at 5:53
  • $\begingroup$ oh thank you, I find a counterexample that if the sum of two sequences $x_n+y_n$ if convergent, then $x_n$ and $y_n$ don't have to converge $\endgroup$ – topos May 12 '15 at 5:58
  • $\begingroup$ In the proof of "(1) sum of two open sets is open", how can you say that $B_\eta(x+y)$ is in X+Y $\endgroup$ – Infinite Mar 21 '17 at 6:07

Something really useful to learn for these type of exercises is the following piece of information.

Let $y\in \mathbb{R}^{n}$ be fixed and define $\varphi_{y}:\mathbb{R}^{n}\to \mathbb{R}^{n}$ by setting $\varphi_{y}(x)=x+y$. Now $\varphi_{-y}\circ \varphi_{y}(x)=(x+y)-y=x$ and similarly $\varphi_{y}\circ \varphi_{-y}(x)=x$. Hence $\varphi_{y}$ is a bijection and $\varphi_{y}^{-1}=\varphi_{-y}$. Both $\varphi_{y}$ and $\varphi_{-y}$ are continuous functions so $\varphi_{y}$ is a homeomorphism.

Using the above property, many exercises of this sort become much easier.

  1. In particular, for any open set $X\subseteq \mathbb{R}^{n}$ the set $X+y$ is an open set for all $y\in\mathbb{R}^{n}$ (since $\varphi_{y}$ is a homeomorphism). In particular, \begin{align*} X+Y=\bigcup_{y\in Y}(X+y) \end{align*} is an open set as the union of open sets.

  2. Since arbitrary unions of closed sets aren't closed, we can't apply the above reasoning. But there is a standard counter example for this part. Take $X=\{-n+\frac{1}{n}:n\in\mathbb{N}\}$ and $Y=\mathbb{N}$. Both $X$ and $Y$ are closed but $X+Y=\left\{(m-n)+\frac1n\:;\:m,n\in\mathbb{N}\right\}$ has a subset $A:=\{\frac{1}{n}:n\in\mathbb{N}\}$, which has a limit point at zero but $0\notin X+Y$. Hence $X+Y$ is not closed.

  3. Since $X$ and $Y$ are compact then $X\times Y\subseteq\mathbb{R}^{2n}$ is compact, and use the continuity of the function $(x,y)\mapsto x+y$ and the fact that continuous images of compact sets are compact, to conclude that $X+Y$ is compact.

  • 2
    $\begingroup$ It's not true that $X+Y = \{\frac 1 n: n\in \mathbb N\}$, because, for example, $1+\frac 1 n = 2 + (-1+\frac 1 n) \in X+Y$. But $\{\frac 1 n: n\in \mathbb N\}$ is a subset of $X+Y$, which still yields a counterexample. $\endgroup$ – Jack Lee May 12 '15 at 14:41
  • $\begingroup$ @JackLee: Right, good call. I totally overlooked that one. Thanks for pointing it out, I'll edit it. $\endgroup$ – T. Eskin May 12 '15 at 14:43
  • $\begingroup$ Can these be applied in any NLS?@T.Eskin $\endgroup$ – sani Apr 9 '18 at 6:56

3)The sum is a continuous operation. The image $X+Y$ of the compact set $X×Y$ is therefore compact

  • $\begingroup$ Can these be applied in any NLS?@Arpit Kansal $\endgroup$ – sani Apr 9 '18 at 6:56

A famous counterexample is the sum of the ice-cream cone $$ A=\{x\in \mathbb R^3: x_2^2 + (x_3-x_1)^2 \le x_1^2, \ x_1\ge 0\} $$ and the line $$ B=\{x\in \mathbb R^3: x_2=x_3=0\}. $$ Then $$ A+B = \{ x\in \mathbb R^3: \ x_3>0 \text{ or } x_2=x_3=0\}, $$ which is not closed: $x\in A+B$ means there is $t\in \mathbb R$ such that $x- \pmatrix{t\\0\\0}\in A$, that is, $x_1-t\ge0$ and $$ x_2^2 + (x_3-(x_1-t))^2 \le (x_1-t)^2, $$ which is equivalent to $$ 0\le (x_1-t)^2 - (x_3-(x_1-t))^2- x_2^2 = 2x_3(x_1-t)- x_3^2 - x_2^2 . $$ Hence the point $(0,1,0)$ is not in $A+B$, but $$ \pmatrix{ 0\\ 1 \\ 1/n} = \pmatrix{n+1\\1\\1/n} + \pmatrix{-(n+1)\\0\\0} \in A+B. $$

As you might already know, the sume $A+B$ can be proven to be closed if $A$ is compact and $B$ is closed.


The set of integers,Z and qZ where q is a rational number, are closed subset of R but Z+qZ is not closed in R as it is dense in R.

  • $\begingroup$ I think that $q$ should be irrational. $\endgroup$ – Watson Apr 3 '16 at 16:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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