I posted this on mathstack and it is not a homework problem, I am doing a modeling problem and this ODE comes up. I don't know whether it could be considered to be a research problem yet. Please help me to close it if it is not placed at the right place.

I have a problem with finding the closed form solution of the following ODEs. I am NOT sure that a closed form solution exists. Closed form here means that the solution can be presented as integrals/ power series. Here is the ODE : I only consider $x\in (0,1)$ and $c_i$ are known non-zero real numbers.

$y''(x) + [\frac{c_1}{x^2}+\frac{c_2}{(1-x)^2} +c_3(\frac{1}{x}+\frac{1}{1-x})]y'(x)+[c_4(\frac{1}{x}+\frac{1}{1-x})+\frac{c_5}{x^2}+\frac{c_6}{(1-x)^2}]y(x)=0 $

I can find the solution inform of power series which is very ugly. I would like to ask you all that whether you can give a method that can be used to find the solution in a nicer way. Thanks so much for your time. I really appreciate it.

  • $\begingroup$ This equation, with arbitrary $c_j$ does not have explicit solutions, or a useful power series solution. You can only do qualitative study, or try to compute solutions numerically. $\endgroup$ – Alexandre Eremenko Feb 18 '13 at 1:54
  • $\begingroup$ @Steven: You are asking for the solutions of a second order linear equation with singular points at $x=0$, $x=1$, and $x=\infty$. When $c_1=0$ and $c_2$=0, these singular points are *regular singular points* (in which case, this is known as a *Fuchsian* equation), otherwise, they are not. Generally, for nearly all values of the constants $c_i$, these are not expressible in terms of elementary functions. $\endgroup$ – Robert Bryant Feb 18 '13 at 1:56

This is a second degree linear ODE with rational function coefficients, and this problem has been completely solved algorithmically by Kovacic in the mid-eighties:

Kovacic, Jerald J. "An algorithm for solving second order linear homogeneous differential equations." Journal of Symbolic Computation 2, no. 1 (1986): 3-43.

All the major computer algebra systems implement this, and (not surprisingly) Mathematica does not find a solution in closed form in the generality you give. However, since your $c_i$ are ``known real numbers'', perhaps you are lucky, and a closed for solution is known in your special case. The relevant Mathematica incantataion is "DSolve", as in:

DSolve[y''[x] + (c1/x^2 + c2/(1 - x)^2 + c3 (1/x + 1/(1 - x))) y'[ x] + (c4 (1/x + 1/(1 - x)) + c5/x^2 + c6/(1 - x)^2) y[x] == 0, y[x], x]


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.