4

I am new to this topic so I appreciate any help on this.

a)$$\lim_{(x,y) \rightarrow (0,0)} f(x,y) = \frac{x^2y}{x^4 +y^2}$$

For $x=0 \lor y=0 : f(x,y) \longrightarrow 0$ but for $y = x^2$ the limit is $\frac{1}{2}$ so the limit does not exist.

b) $$\lim_{(x,y) \rightarrow (0,0)} f(x,y) = \frac{e^{xy^3}-1}{x^2 +y^4}$$ I don't have any ideas here

c) $$\lim_{(x,y) \rightarrow (0,0)} f(x,y) = (x^2 +y^2)^{xy}$$ Let $x \geq y$ then $|x^2 +y^2|^{xy} \leq |2y^2|^{y^2} \longrightarrow 1 \quad $for $y \longrightarrow 0$ but for $x = \frac{1}{y} \Longrightarrow f(x,y) = (\frac{1}{y^2} + y^2)^1 \longrightarrow \infty \quad$ for $y \longrightarrow 0 $ so the limit does not exist.

3 Answers3

5

For b):

we have that $f(0,y) = f(x, 0) = 0$ for every $x,y\neq 0$. If $x$ and $y$ are both non zero, we have that $$ \frac{e^{xy^3} - 1}{x^2+y^4} = \frac{e^{xy^3} - 1}{xy^3}\cdot \frac{xy^3}{x^2+y^4}. $$ The first factor at the r.h.s. is bounded (for $x,y$ bounded), since $(e^t-1)/t \to 1$ as $t\to 0$.

The second factor can be estimated by $$ \left|\frac{xy^3}{x^2+y^4}\right| = |y| \cdot \frac{|x| y^2}{x^2+y^4}\leq \frac{|y|}{2}. $$

Hence the limit exists and is equal to $0$.

Rigel
  • 14,434
2

Another way to process is to set $y=mx^\alpha$ with some well chosen $\alpha>0$ so that the various sums $x^a+y^b$ become homogeneous.

Of course $m=m(x,y)$ is not a constant, just the ratio of the rate of convergence of $x$ and $y$.

Part a:

With $y=mx^2$

$\displaystyle f(x,y)=\frac{x^2y}{x^4+y^2}=\frac{mx^4}{x^4+m^2x^4}=\frac{m}{1+m^2}$

This is bounded but can take multiple values (e.g $m=1,\ f(x,x^2)=\frac 12$ and $m=0,\ f(x,0)=0$), so there is no limit in $(0,0)$.

Part b:

With $x=my^2$

$\displaystyle f(x,y)=\frac{e^{x^3y}-1}{x^2+y^4}\sim\frac {x^3y}{x^2+y^4}=\frac{m^3y^7}{m^2y^4+y^4}=\frac{m^3y^3}{1+m^2}=\underbrace{\bigg(\frac{m^2}{1+m^2}\bigg)}_\text{bounded}xy\to 0$

So there is a limit and it is $0$.

Part c:

The previous method do not work well here, but polar coordinates do

$f(x,y)=(x^2+y^2)^{xy}=\exp(xy\ln(x^2+y^2))=\exp(\underbrace{\cos(\theta)\sin(\theta)}_\text{bounded}\underbrace{r^2\ln(r^2)}_{\to 0})\to e^0=1$

zwim
  • 28,563
  • 1
    Be careful. Looking along families of curves will only prove a limit does not exist, not that it does. – Ted Shifrin May 22 '17 at 15:58
  • Please read, m is not a constant, no matter how many time I write this, people don't see it. These are not curves. – zwim May 22 '17 at 16:12
  • It's still very misleading. How do you handle $f(x,y) = \begin{cases} 1, & |y|>|x|^3 \text{ or } y=0 \ 0, & \text{otherwise} \end{cases}$ or $f(x,y) = \begin{cases} 1, & |y|>e^{-1/x^2} \text{ or } y=0 \ 0, & \text{otherwise}\end{cases}$? – Ted Shifrin May 22 '17 at 16:18
1

You've made a mistake in c): $(1/y,y)$ does not approach $(0,0)$ as $y\to 0.$

To do it, we can apply $\ln$ to get $xy\ln (x^2+y^2).$ Recall $|xy|\le (x^2+y^2)/2.$ Thus

$$\tag 1 |xy \ln (x^2+y^2)| \le \frac{(x^2+y^2)\ln (x^2+y^2)}{2}.$$

Now $u\ln u \to 0$ as $u\to 0^+.$ Thus the limit of $(1)$ as $(x,y) \to (0,0)$ is $0.$ This implies the orginal limit is $1.$

zhw.
  • 105,693