We consider ordinary linear differential equations with polynomial coefficients. Each such equation has a finite dimensional vector space of solutions, but usually none of these solutions can be expressed in closed form. We discuss the problem of finding out for a given specific differential equation whether one (or some, or all) of its solutions admit a closed form representation. In the first part of the talk, we explain the classical algorithms for finding polynomial and rational solutions. After that, we turn to hyperexponential solutions. These are solutions that can be written in the form $\exp(u(x))v_1(x)^{e_1}\cdots v_k(x)^{e_k}$ for certain rational functions $u,v_1,\dots,v_k$ and constants $e_1,\dots,e_k$. There is a classical algorithm for finding solutions of this form, but it is very expensive. We present a more efficient algorithm, published a few weeks ago at ISSAC'13. (Joint work with F. Johansson and M. Mezzarobba)
Forschungs-