I want to find $\displaystyle\lim_{(x,y)\to(0,0)} \frac{xy}{x+y} = 0$. After trying different ways to approach $(0,0)$, I am fairly convinced the limit is $0$, but I need to prove it by definition, and I seem to be stuck. I want to prove that $\forall \epsilon >0, \exists \delta > 0 $ such that $0<\parallel(x,y)\parallel<\delta \implies |\frac{xy}{x+y}| < \epsilon$.

I'm having trouble with the denominator. I know that to get $|\frac{xy}{x+y}|$ to be less than something, I have to show that $|x+y|$ can be made greater than something, but I don't know what. Any suggestions?

  • $\begingroup$ What if you (can you?) approach along the line $y = -x$? $\endgroup$ – Dylan Moreland May 4 '12 at 21:03
  • $\begingroup$ @Dylan: Well, the function is not even defined along that line. Does that matter? $\endgroup$ – Javier May 4 '12 at 21:05
  • $\begingroup$ @Javier: Well, it matters in-so-far as that means that your definition cannot be satisfied, since you can find points $(x,y)$ with $0\lt\lVert (x,y)\rVert \lt\delta$ but for which $\left|\frac{xy}{x+y}\right|\lt\epsilon$ does not hold... $\endgroup$ – Arturo Magidin May 4 '12 at 21:08
  • $\begingroup$ consider approaching on this curve $y=-x+x^3$, when $x$ is relatively small. $\endgroup$ – Yimin May 4 '12 at 21:10
  • $\begingroup$ @ArturoMagidin: Oh, I thought you could just sort of ignore it. This means that the limit doesn't exist, then? $\endgroup$ – Javier May 4 '12 at 21:13

If you approach $(0,0)$ along the line $x=0$ the function has constant value $0$ and the limit is $0$.

But now suppose that you approach along a curve like $y=x^2-x$. Then $$\frac{xy}{x+y} = \frac{x^3-x^2}{x^2} =x - 1$$ and the limit as $x\to 0$ is...

  • $\begingroup$ I guess this is a simpler way to know. How did you think of approaching along $y=x^2-x$? $\endgroup$ – Javier May 4 '12 at 21:17
  • 2
    $\begingroup$ @Javier: Experience, I'm afraid, not much more. $\endgroup$ – Arturo Magidin May 4 '12 at 21:23
  • 7
    $\begingroup$ @JavierBadia Clearly approach along a linear function won't do the job. So lets try the quadratic $y=ax^2 + bx$. Then we get that $\frac{ax^3 + bx^2}{ax^2 + (b+1)x} = \frac{ax^2 + bx}{ax +(b+1)}$. Note that if $b \neq -1$, then the limit will be zero. If $b=-1$, then we get $\frac{ax^2 - x}{ax} = \frac{ax-1}{a}$ and the limit is $-\frac1{a}$. So moving along the curve $y = ax^2 -1$ gives the limit $-\frac1a$ $\endgroup$ – user17762 May 4 '12 at 21:25

The sum and product of two numbers are independent, you can assign them arbitrary values at will.

So while you make them both converge to zero, you can follow any $(s,p)$ curve, with arbitrary ratios.

For example, following an hyperbolic spiral $\rho=s/\theta$,


This expression alternates between minus and plus infinity on every turn.


Put $x=rcos\theta$ and $y=rsin\theta$ s.t. $$lim_{(r,\theta)\rightarrow (0,0)}\frac{rcos\theta.sin\theta}{cos\theta+sin\theta}=0$$ if and only if the denominator $cos\theta+sin\theta\ne 0$. However, approaching $r=0$ along $\theta=-\pi/4$ the given limit doesn't exist.

  • 1
    $\begingroup$ This is misleading, for reasons explained on the page in at least two posts, for almost 5 years when this was posted. $\endgroup$ – Did Mar 19 '18 at 12:34

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.