I have a problem which I don't seem to be able to solve without help. It says:

Given the differential equation: $$x\cos\left(\dfrac yx\right)y^\prime=y\cos{\left(\dfrac yx\right)-x}$$ Find:

  1. An integrating factor of the form: $\mu_1=\mu_1(x)$
  2. An integrating factor of the form: $\mu_2=\mu_2(\frac yx)$

I am able to solve question number 1, but I don't have any means of answering question number 2, methodically. To solve question number 1, I first demonstrate that the differential equation is not exact: $$\underbrace{\left[y\cos{\left(\dfrac yx\right)-x}\right]}_{\equiv M(x,y)}dx+\underbrace{\left[-x\cos{\left(\dfrac yx\right)}\right]}_{\equiv N(x,y)}dy = 0$$

And, since $\partial M/\partial y \neq \partial N/\partial x$, it is not exact.

Therefore, I can try to find an integrating factor $\mu_1$ which:

  1. Is continuous, and
  2. Depends solely on x

using the formula: $$\mu_1(x)=\dfrac{\dfrac{\partial M}{\partial y}-\dfrac{\partial N}{\partial x}}{N}=\ldots=-\dfrac{2}{x}$$

Which (albeit discontinuous on $x=0$) is an integrating factor of the oringinal differential equation.

My two questions are:

  1. How can I find an integrating factor which depends exclusively on $\frac yx$? (Note: I've tried performing the change of variable: $v=y/x$, but I reach a plateau).
  2. Is $(-2/x)$ an integrating factor, even if it has a discontinuity at $x=0$? (it was taught at school that it has to be continuous for all $x$).

Thank you and apologies, I am kind of new to this area of mathematics.

Edit 1

Following @B. Goddard's answer below, I reach the following equation: $$\left(\dfrac yx \cos{\dfrac yx}-1\right)\dfrac {\partial \mu}{\partial y}-\left(\dfrac yx\cos{\dfrac yx}\right)\dfrac{\partial \mu}{\partial x}+2\mu\cos{\dfrac yx}=0$$

The problem is that this exercise is from an elementary course on differential equations, and we didn't study PDE's yet. (Maybe the above equation simplifies in a way I don't see?)

Maybe I should have attempted the change of variable $(v\equiv y/x)$ before multipliying the whole equation by $\mu(y/x)$?


Since the original DE is not defined at $x=0$, it doesn't matter that the integrating factor is discontinuous there. Everything is continuous on the domains of $M$ and $N$.

The answer to the first question is a bit tedious. So I'll just sketch the steps. First, you used a formula for the $\mu(x)$ integrating factor. Look at the derivation of that formula and follow the same steps.

Assume there is a $\mu(y/x)$ and multiply your DE by it. The new equation needs to be exact, so the (new) $\partial M/\partial y$ needs to equal the (new) $\partial N/\partial x.$

So that's two ugly product rules you have to compute. When you're taking the derivative of $\mu$ with respect to $y$, the chain rule makes you multiply by $1/x$. When you take the derivative with respect to $x$, the chain rule makes you multiply by $-y/x^2.$ Set $\partial M/\partial y = \partial N/\partial x.$

Put this last equation in the form $\mu' = \mu \times $ crud and simplify. The simplified crud should be a function of $y/x.$ If not, no integrating factor exists. (In the case of $\mu(x)$, your formula is the "crud".) If so, you can replace $y/x = z$ and solve the DE $\mu'(z) = \mu(z)\mbox{crud}(z)$ for $\mu$. Now you have the integrating factor.

Good luck.

  • $\begingroup$ Thank you for the answer. Following your procedure, I obtained a Partial Differential Equation (contains both $\partial \mu/\partial y$ and $\partial \mu/\partial x$), and we didn't study PDE's yet. I will update the answer to give you the equation that $\mu=\mu(y/x)$ would need to satisfy. Again thank you for your time. $\endgroup$ – Jose Lopez Garcia Jun 14 '18 at 14:07
  • $\begingroup$ This won't lead to a PDE. You should get $\mu' = \mu x $junk. It's a separable ODE. $\endgroup$ – B. Goddard Jun 14 '18 at 15:33

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.