OPEN
Fix $G\in \mathbb{Q}(z, y_0, y_1, \cdots, y_{n-1}), N\in \mathbb{Z}_{>0}$ and suppose $f(z)\in \mathbb{Z}[\frac{1}{N}][[z]]$ satisfies $$f^{(n)}(z)=G(z, f(z), f'(z), \cdots, f^{(n-1)}(z)).$$ Suppose further that $G(0, f(0), \cdots, f^{(n-1)}(0))$ is defined. Is it necessarily the case that $f(z)$ is the Taylor expansion at $0$ of an algebraic function?
The open status of this problem reflects the current belief of the owner of this website. There may be literature on this problem that I am unaware of, which may partially or completely solve the stated problem. Please do your own literature search before expending significant effort on solving this problem. If you find any relevant literature not mentioned here, please add this in a comment.
General remarks This is a version of Conjecture 1.1.1 of
[LL25], though surely it has been considered earlier--perhaps as early as Eisenstein's final paper
[E52], where he proved that Taylor expansions of algebraic functions lie in $\mathbb{Z}[\frac{1}{N}][[z]]$--that is, only finitely many primes appear in the denominators of their Taylor coefficients. For linear ODE, some version of this conjecture has been considered by Andr\'e and Christol, see e.g. Remarque 5.3.2 of
[A04].
[LL25] proves a number of special cases for ODE (and initial conditions) with some algebro-geometric meaning.
Note that the condition that $G(0, f(0), \cdots, f^{(n-1)}(0))$ is defined is necessary; otherwise (for example) the power series $$f(z)=\sum_n {2n \choose n}^2\left(\frac{z}{16}\right)^n$$ is a counterexample. In other words, one needs to be taking the Taylor expansion at a non-singular point of our differential equation. The case of singular points (and linear ODE) is considered by Christol in e.g.
[C84]; it would be interesting to find a reasonable conjecture in the non-linear case.
Formalizability The statement has now been formalized, see
here.
Some speculation on the problem's difficulty My expectation is that substantial new ideas are needed to resolve this question, assuming the answer is, as I expect, "yes." One natural case to consider would be the case of Picard-Fuchs equations (or Gauss-Manin connections) where already the conjecture is open in general, though Theorem 3.1.1 of
[LL25] resolves it in the case of initial conditions which are a linear combination of cycle classes, for example. I expect a number of other special cases of this sort to be doable with existing techniques.
If the answer is "no," it's quite possible that there is a simple counterexample (perhaps already implicit in the literature) -- in this case it would be good to try to repair the question appropriately.
Comments on interest I think of this and related questions as being among the most basic open problems in the theory of algebraic differential equations. It is closely related to--though it seems neither to follow from nor imply--the Grothendieck-Katz $p$-curvature conjecture
[K72] (for linear ODE) and the Ekedahl--Shepherd-Barron--Taylor conjecture
[ESBT99] (for non-linear ODE). See
[2] for a special case of the $p$-curvature conjecture.
View the LaTeX source
This page was last edited 29 May 2026. View history