4

The problem statement is,

Show that $f:\mathbb{R}^2\to\mathbb{R}$ defined by $f(x,y)=x+y$ is continuous using open sets.

I know that to show $f$ to be continuous I take an arbitrary open set $O\subset\mathbb{R}$ and show that $f^{-1}(O)$ is open in $\mathbb{R}^2$. We can simplify this by first showing that for any $(a,b)\in\mathbb{R},$ $f^{-1}(a,b)$ is open in $\mathbb{R}^2.$ Now, we have $$f^{-1}(a,b)=\{(x,y):f(x,y)\in(a,b)\}=\{(x,y):a<x+y<b\}$$ To show that $f^{-1}(a,b)$ is open, we need to show that for every $u\in f^{-1}(a,b)$, there exists a $\delta>0$ such that $B_\delta(u)\subset f^{-1}(a,b).$

One way I thought about showing this was to first fix $y$ in $(x,y)\in f^{-1}(a,b)$ and range over all $x$. Do the same for $y$ and then try to find an appropriate $\delta$. I did find this question (which is exactly mine)

Using the open set definition of continuity to directly prove a function is continuous

Yet, the I didn't completely understand the hints or answers given.

Thanks for any help or feedback.

S.D.
  • 1,561
  • a point is closed, so its counterimage is a close set.. – Exodd Oct 12 '14 at 02:55
  • From you're hint I was able to conclude that $f^{-1}(a,b)$ is a plane with two lines with slope -1 intersecting the y and x axes at a and b. This is exactly what the answer/hint was for the question I linked. So do these lines continue on forever, but have holes at x=y=a and x=y=b? – S.D. Oct 12 '14 at 03:07
  • you should remove the topological vector spaces tag –  Oct 12 '14 at 03:08

5 Answers5

5

You'll have to check the details, but I think this would be the outline:

Let $(x,y) \in f^{-1}(a,b)$, then $a < x + y < b$ hence we can find an $\epsilon > 0$ (check this!) such that $$ a + \epsilon < x + y < b - \epsilon. $$ Then (for $u = (x, y)$) we have $B_{\epsilon/2}(u) \subset f^{-1}(a,b)$, for let $(x', y') \in B_{\epsilon/2}(u)$, then $|x - x'| < \epsilon/2$ and $|y - y'| < \epsilon/2$, therefore (check this one too!) $$ a < x' + y' < b. $$ In conclusion, for every point in $f^{-1}(a,b)$ we can find an open neighbourhood $B_{\epsilon/2}(u)$ such that $B_{\epsilon/2}(u) \subset f^{-1}(a,b)$, proving that $f^{-1}(a,b)$ is open.

2

A picture is helpful:

enter image description here

The diagonal lines are the lines $x+y=a$ and $x+y=b$, so $f^{-1}[(a,b)]$ is the strip lying strictly between them. The point $p$ is any point in that region; you want to find an open ball with centre at $p$ that lies wholly within the strip, as the one in the sketch just barely does. If you take the radius of the ball to be the minimum of the distances from $p$ to the two edges of the strip, you’ll be in business, and calculating that minimum in terms of the coordinates of $p$ is just a bit of easy analytic geometry.

Brian M. Scott
  • 616,228
  • Thank you so much Brian. This was what I was trying to get at in the comments. I thought that this was how $f^{-1}(a,b)$ looked like. Next time I'll make sure to just draw a picture and include it in my question. – S.D. Oct 12 '14 at 03:17
  • @Shant: You’re very welcome. – Brian M. Scott Oct 12 '14 at 03:18
1

You could prove that the squares $(x,x+\frac{b-a}{2})\times(a-x,a-x+\frac{b-a}{2})$ cover $f^{-1}(a,b)$ as $x$ ranges over all real values.

  • But if we were to cover $f^{-1}(a,b)$, couldn't the open ball we find for an element in this covering be outside of $f^{-1}(a,b)$? – S.D. Oct 12 '14 at 03:11
  • Ah, sorry, I was using the product topology! –  Oct 12 '14 at 03:12
  • No problem at all. I came across some things regarding the product topology, but didn't know how it connected into the problem. I did come across the projection function, which felt like it could be useful, but couldn't find a way to tie it in. – S.D. Oct 12 '14 at 03:15
1

Exodd is right! In addition, the pre-image of the open interval in $\mathbb{R}$ is a band region in $\mathbb{R}^2$, which is also open.

TJH
  • 129
  • Is the banded region in $\mathbb{R}^2$ infinitely long, with 4 holes at $x=y=b$ and $x=y=a$? Or are the boundaries from (a,b) on the y axes and $(a,b)$ on the x axes with slope -1 lines connecting the ends? – S.D. Oct 12 '14 at 03:13
  • Yes, this is an infinitely long band. However, it has no holes. – TJH Oct 12 '14 at 03:24
1

$$ \{(x,y)\mid a<x+y<b\}=\{(x,y)\mid x+y<b\}\cap \{(x,y)\mid x+y> a\} $$

Intersection of two open half-planes.

ir7
  • 6,249