Trying to solve a non-homogeneous differential equation, whether it is linear, Bernoulli, Euler, you solve the related homogeneous equation and then you look for a particular solution depending on the "class" of the non-homogeneous term.

Just to make an example, let's say we have this equation

$$y'' - y' - 2y = 2x e^x$$

we solve $$y'' -2y' +y =0$$ and we get solutions of the form $y_{\text{hom}}(x) = C_1 e^{2x} + C_2 e^{-x}$.

Then, since the known term is of the form $p(x)(x) e^{\mu x}$, and $\mu$ is not a solution to the characteristic polynomial, we should find the solutions through those of the form $q(x)e^{\mu x}$. If, on the other hand, $\mu$ were a solution of the characteristic polynomial, we should have looked for solutions through those of the form $q(x) x^q e^{\mu x}$ where $q$ is the multiplicity of the root $\mu$. (where $q(x)$ is the general polynomial of the same degree as $p(x)$.

I have a whole list of these "classes" of known terms, like $p(x)$, $p(x)e^{\mu x}$, $e^{\alpha x}(a \sin(\beta x) + a' \cos(\beta x))$, and the corresponding general solutions (like I said before, also with the multiplicity involved).

My question is, where do these come from? I have no idea how such classes were found or anything; my book and my professors just said these were given and we should just learn them.

Edit: from the comment I see that something's going on, unfortunately I don't have enough background to know how to work with Laurent series. I would appreciate any explanation. Thanks.

  • $\begingroup$ the "classes of known terms" you mention are just linear combinations of $x^ne^{\mu x}$. By taking derivatives wrt. $\mu$ you can reduce it just to $e^{\mu x}$. Solving a constant linear ODE with $e^{\mu x}$ at the RHS is reasonably easy and unsurprising (in fact, solve it for $e^{\mu+\epsilon}$ and expand to a Laurent series in $\epsilon$ if you you want to easily include the case when $\mu$ is a root of the char. poynomial) $\endgroup$ – user8268 May 16 '11 at 10:45
  • $\begingroup$ @user8268: would you care to post your comment as an answer? (feel free to add something if you want) Even though this question didn't raise much attention, I still got some info ;) $\endgroup$ – Andy May 17 '11 at 19:51

The possible right-hand-sides for this method are, themselves, solutions of linear homogeneous differential equations with constant coefficients. User8268 provided another description of this class in a comment. If your equation is Ly=f, where L is a linear differential operator with constant coefficients, and Mf=0 where M is another such operator, then your desired solution is also a solution of MLy=0, and knowledge of solutions of such (homogeneous differential equations with constant coefficients) can then be used to show what the form of y must be.


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.