This post is mainly to ask good reference that explains all details needed to solve a (solvable by separation of variables) non homogeneous PDE by using separation of variables. Take, for example
$ \left\{\begin{matrix} u_{tt}(x,t) + u_{xx}(x,t) =f(x,t) & 0<x <L, t>0 \\ u(0,t)=u(L,t)=0 & t>0 \\ u(x,0)=u(x,K)=0 & 0<x<L \end{matrix}\right. $
Every notes I have been given (including the ones I link above) say something like:
"We look for solutions u in the form u(x, t) = T(t)X(x). As before we look at the eigenvalue problem $\left\{\begin{matrix} X'' +\lambda X =0 \\ X(0) =0=X(L) & \end{matrix}\right.$"
Why? I understand this is valid for an homogeneous PDE where the basic principle is to asume $u(x,t)=X(x)T(t)$, so then $u_{tt} +u_{xx}=0=XT''+X''T$, and then
$$ \frac{X''}{X}=-\frac{T''}{T}= - \lambda \quad \quad (^*)$$
But this does not (seem to) work when the RHS of the equation is an arbitrary function, as $ (^*)$ would be something like
$$\frac{X''}{X}=-\frac{T''}{T}-\frac{f}{XT}$$
and I see no way to conclude anything similar from this last equation.
Note this was already asked here, but the answer was not an actual answer.