Section 4.4 Issues and Next Directions
After the initial examples of series solutions, two of questions seem natural. First, I might observe that \(P\) and \(Q\) were, so far, only rational functions, which means that I only dealt with polynomials in the calculation. What if \(P\) and \(Q\) were other analytic functions? In this cases, we would have to expand \(P\) and \(Q\) as series about the ordinary point and then multiply them by the series for \(y\) in the calculation. This is possible, but miserable.
I might also wonder why I restricted myself to homogeneous cases. What happens if I add a forcing term \(f(t)\text{?}\) I can certainly do this is \(f\) is analytic on the same domain as \(P\) and \(Q\text{.}\) In all our examples so far, I compared the coefficients to \(0\text{.}\) If there is a forcing terms, I would instead compare the coefficients to the coefficients of \(f\) instead of \(0\text{.}\) This leads to more complications, since the relationship between the coefficients may no longer be a linear recurrence relation.
In both of these situations (non-polynomial \(P\) and \(Q\) or non-homogeneous DE with forcing), I can easily end up in a situation were the challenges of computation prevent me from finding a nice form for the coefficients of \(y\text{.}\) Quite frequently, I will only decide to calculate the first few terms. The result is a Taylor polynomial approximation to the solution, instead of a complete Taylor series solution. However, approximate solutions (if they have sufficient precision) are often sufficient. One of the advantages of working with Taylor series solutions is the easy availability of Taylor polynomial approximations.