$$\lim_{(x,y)\to(0,0)}\frac{x^2+y^2+5xy}{x-y}$$
I've tried approximating using parabolas and even use polar coordinates but it seems to always result in $0$. The limit seems simple so I must be forgetting something basic. Hints? Thank you.
$$\lim_{(x,y)\to(0,0)}\frac{x^2+y^2+5xy}{x-y}$$
I've tried approximating using parabolas and even use polar coordinates but it seems to always result in $0$. The limit seems simple so I must be forgetting something basic. Hints? Thank you.
On
Write \begin{align} \dfrac{x^2+y^2+5xy}{x-y} &= x-y +\dfrac{7xy}{x-y} \\ &=x-y+y\left(\dfrac{7x-7y+7y}{x-y}\right) \\ &=x-y+7y+\dfrac{7y^2}{x-y}=x+6y+\dfrac{7y^2}{x-y}. \end{align} To see that the limit fails to exist, consider $2$ paths: the first one is $(x,y) = (2t,t)$,and the limit is $0$. And the second path is $(x,y) = (t^2+t,t)$, and the limit is $7$. Thus you have $2$ different values of the limit, hence it does not exist.
On
A coordinate transformation can work - but you just have to get a little bit more ingenious. The other answers have shown that the trick is the line $x = y$, so we should choose a transformation which makes analyzing it a bit more palatable. Consider the transformation to a tilted coordinate system in which the $y$-axis lies along this line. There are many ways to do this, one of which is the transformation matrix
$$\mathbf{T} := \begin{bmatrix} 1 && -1 \\ 1 && 1 \end{bmatrix}$$
Geometrically, this matrix is the rotation of the coordinate system around the origin counterclockwise by 45 degrees, thus turning the line $x = y$ counterclockwise just the right amount to align with the $y$-axis (as it lies at a 45-degree angle) multiplied by a scaling by $\sqrt{2}$ to make the entries integer and so easy to work with. Under this transformation, a coordinate vector $(x, y)$ transforms as
$$(x, y) \mapsto (x - y, x + y)$$
That is, define $u := x - y$ and $v := x + y$ as your new coordinates. Then the denominator becomes $u$, and the numerator is solved as follows: note that $v^2 = (x^2 + 2xy + y^2)$ so
$$x^2 + y^2 + 5xy = (x^2 + 2xy + y^2) + 3xy = v^2 + 3xy$$
and now, by recognizing the sum-and-difference-of-the-same-thing pattern, that you can take $u + v = 2x$ and $u - v = -2y$ so $(u + v)(u - v) = -4xy$ and $\frac{3}{4} (u + v)(u - v) = 3xy$. Thus
$$\begin{align}\frac{x^2 + y^2 + 5xy}{x - y} &= \frac{v^2 - \frac{3}{4}(u + v)(u - v)}{u} \\ &= \frac{v^2 - \frac{3}{4}(u^2 - v^2)}{u} \\ &= \frac{v^2 - \frac{3}{4}u^2 + \frac{3}{4}v^2}{u}\\ &= \frac{-\frac{3}{4}u^2 + \frac{7}{4}v^2}{u}\\ &= -\frac{3}{4} u + \frac{7}{4} \frac{v^2}{u} \\ &= \frac{7}{4} \frac{v^2}{u} - \frac{3}{4} u \end{align}$$
and now when you take the limit - note that the origin is unchanged, so the limit is still to $(0, 0)$ but now in $(u, v)$ coordinates instead of $(x, y)$ coordinates - the second term is zero, but the first has no limit: if $v$ goes to 0 first, then the limit as $u$ goes to 0 is 0, but if $u$ goes to 0 first, the limit becomes $\pm \infty$, and is such for every value of $v$ except 0 for such an approach. That is, the function is unboundedly large in an arbitrarily small circle around the origin, and thus the limit at that origin does not exist.
On
Although there is an already accepted answer, here is a more systematic method:
This reveals
$$\frac{x^2+y^2+5xy}{x-y} =\frac{x^2 + (x-h(x))^2 + 5x(x-h(x))}{h(x)}= \frac{7x^2 - 7xh(x) + (h(x))^2}{h(x)}$$ $$ = \color{blue}{\frac{7x^2}{h(x)}} - 7x + h(x)$$ The $\color{blue}{\mbox{blue}}$ term reveals of what order $h(x)$ should converge to $0$ to produce a path $(x,y(x)) \stackrel{x \to 0}{\longrightarrow}(0,0)$ such that the limit for $x \rightarrow 0$ does not exist.
For example $h(x) =x^4 \Rightarrow y=x-x^4$ produces a limit of $+\infty$.
The quotient blows up near the line $x=y$ no matter how close to $(0,0)$ you get, so the limit does not exist.
More precisely: Each ball of radius $\delta$ around $(0,0)$ contains points with arbitrarily large function values. Consider a sequence of points within the ball that converges towards $(\frac12\delta,\frac12\delta)$ from the $x>y$ side. At points in this sequence the numerator of the fraction goes towards $\frac74\delta^2$, while the denominator goes to $0$ from above, so the function values become abitrarily large.
(Conversely, considering a sequence that approaches $(\frac12\delta,\frac12\delta)$ from the $x<y$ side, will give you arbitrarily large negative function values).