I came to a question in James Stewart's "Calculus Early Transcendentals" about implicit differentiation saying that:

Find all points on the curve $x^2y^2+xy=2$ where the slope of the tangent line is $-1$

After implicit differentiation I came to $$y'x(2xy+1)+y(2xy+1)=0\tag{1}$$ $$(2xy+1)(y'x+y)=0$$ Then $xy=\frac{-1}{2}$ or $y'=\frac{-y}{x}$

But by pluging $xy=\frac{-1}{2}$ into the original equations we get $$x^2y^2+xy=\frac{1}{4}-\frac{1}{2}\ne2$$ So, we conclude that $xy=\frac{-1}{2}$ is an extraneous solution. Then what caused it to be so ?? Q1

Also, we could start from equation $(1)$ to reach that $$y'=\frac{-y(2xy+1)}{x(2xy+1)} \tag{2}$$ We need the tangent to be $-1$. So we equate $y'=-1$ $$\frac{y(2xy+1)}{x(2xy+1)}=1$$ We now multiply by $x(2xy+1)$ to get $$y(2xy+1)=x(2xy+1)$$ I am now afraid to divide by $2xy+1$ in order not to get a missing solution. [Is it wise not to divide ?? Q2 When to remove the common factor of the numerator and denominator and not afraid of missing solutions ?? Q3]

So, I subtracted RHS-LHS to get $$y(2xy+1)-x(2xy+1)=0$$ Factorizing by grouping $$(y-x)(2xy+1)=0$$

Now we reached our result $y=x$. Sadly along with the extraneous solution. Is multiplying by the denominator of the rational equation $(2)$ is the cause of the second appearance of the extraneous solution ?? Q4

  • 1
    $\begingroup$ Solve for $xy$ may be a method. $\endgroup$ – jonsno Apr 26 '18 at 12:14
  • $\begingroup$ I don't understand your question! You have yourself verified that $2xy+1$ is not satisfied by our original curve, so indeed we have $y'=-y/x$ and subsequently $y=x$. Put $y=x$ in original equation to get points of tangency.. $\endgroup$ – jonsno Apr 26 '18 at 12:32

After implicit differentiation I came to $$y'x(2xy+1)+y(2xy+1)=0\tag{1}$$ $$(2xy+1)(y'x+y)=0$$

This is correct. Now rather than solving this for $y'$, you could simply substitute $y'=-1$, since this is what you want, and add the equation of the curve to get the following system: $$\left\{\begin{array}{rcl} \left(\color{red}{2xy+1}\right)\left(\color{blue}{y-x}\right) & = & 0 \\ (xy)^2+xy & = & 2 \end{array}\right.$$ You need to consider both equations since you're looking for points:

  • on the curve (i.e. satisfying the second equation);
  • and with the correct slope (i.e. satisfying the first equation).

Now from the first equation you have either $\color{red}{xy=-\tfrac{1}{2}}$ or $\color{blue}{y=x}$. You say this first case corresponds to an "extraneous solution" but it's not a solution since no points lying on the curve satisfy this equation. Solutions are only those points satisfying both equations.

Substitution of $y=x$ into the second equation leads to $x=\pm 1$ which then leads to the two correct solutions, namely the points $(1,1)$ and $(-1,-1)$.


$$ x^2y^2+xy-2 = 0 $$

Calling $z = xy \Rightarrow z^2+z-2 = (z+2)(z-1) = (xy+2)(xy-1) = 0 \Rightarrow\left\{ \begin{array}{rcl} xy +2 & = & 0\\ xy - 1 & = & 0 \end{array}\right.$

Attached a plot of $(xy+2)(xy-1) = 0$ In red $xy+2=0$ and in blue $xy-1 = 0$

enter image description here

Hence the tangency should be determined over $xy-1=0$ with two solutions for $x = \pm 1$


You are over thinking it.

Start with

$$ (2xy+1)(y'x+y)=0 $$

Set $y'= -1$:

$$ (2xy+1)(-x+y)=0 $$

So $(2xy+1)=0$ and/or $(-x+y)=0$.

$x=y$ is always a solution whether $(2xy+1)=0$ or not. Since x,y are real, the and condition is never met.

If $(2xy+1)=0$ then $(y'x+y)$ can be any value and doesn't tell you anything about $y'$

Now plug $y=x$ into the original equation to solve for the points.

Hope this helps.



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.