Sign up ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Given a polynomial $p(x)=a_nx^n+\dots+a_1x+a_0$, can every root of the polynomial be represented as $\sum_{k=0}^\infty b_k$ with the $b_k$'s being a function of $a_0,\dots,a_n$ using only elementary operations of arithmetic and taking roots?

share|cite|improve this question
What field are the $a_i$ over, and what coefficients are the $b_i$ allowed to have? – Qiaochu Yuan Jun 13 '11 at 14:58
I don't want to restrict that (if different fields yield different answers that's very interesting) but one can assume we talk about the field of rational numbers, for a concrete case. – Gadi A Jun 13 '11 at 15:51
Sorry if this is a dumb question, but: Would this not be equivalent of asking about finding a sequence $c_k$ (instead of a series) that converges to the root, each term being an 'elemental' function of $a_i$? IF so, couldn't we think $c_k$ as the values provided by some (complex) root-finding iterative algorithm (say, Durand–Kerner)? – leonbloy Jun 13 '11 at 16:28

3 Answers 3

up vote 2 down vote accepted

I think this is true at least formally if you allow the $b_i$ to have coefficients in $\overline{\mathbb{Q}}$. This is because, if $K$ is an algebraically closed field of characteristic $0$, then the field of Puiseux series with coefficients in $K$ is also algebraically closed, and by iterating this construction for each coefficient $a_i$ I think we get the desired result abstractly, although I am not sure what one can say about actual (as opposed to formal) convergence.

share|cite|improve this answer
According to there is actual convergence in $\mathbb{C}$. – Akhil Mathew Jun 13 '11 at 16:14
@Akhil: this is true in sufficiently small neighborhoods, but I don't know if one can get global convergence even for quintic polynomials. See, for example, the explicit series roots given at… . – Qiaochu Yuan Jun 13 '11 at 16:16
Sorry, I only meant convergence near the origin. I see that is not quite what the OP wants though. – Akhil Mathew Jun 13 '11 at 16:27

The difficult part is to get a good a priori estimate $\Omega\subset{\mathbb C}$ of the set $S$ of roots. Starting with any $z_0\in\Omega$, e.g., with rational coordinates, Newton's rule $$z_{n+1}:=z_n-{p(z_n)\over p'(z_n)}\qquad(n\geq 0)\ ,$$ i.e., $$b_0=z_0, \qquad b_{n+1}:=-{p(z_n)\over p'(z_n)}\qquad(n\geq 0),$$ provides a series $\sum_{k\geq 0} b_k$ converging to a point $\zeta\in S$ where the $b_k$ depend rationally on the coefficients of $p$ (and the chosen point $z_0$).

There is a famous paper by Smale on this: "The fundamental theorem of algebra and complexity theory", Bulletin of the American Mathematical Society ${\bf 4}$ (1981), 1–36.

share|cite|improve this answer
In this approach it seems the problem is to find a radical function of the coefficients that makes for a good choice of $z_0$. No obvious candidates come to mind... – Qiaochu Yuan Jun 13 '11 at 18:54
Here's the direct link to the article. – t.b. Jun 13 '11 at 18:55
which begs the question: is there some iterative algorithm for finding all the complex roots of a polynomial (including the choosing of the starting point/s) that is guaranteed to find all roots? (disregarding numerical errors) – leonbloy Jun 13 '11 at 21:19

Isn't this theorem related? As I understand the question, the OP asks wether it can be solvable by radicals, and Abel's theorem states that for polynomial equations of degree $n \geq 5$ there is no general solution.

share|cite|improve this answer
The OP asks about a series, this implies infinite terms - Abel's theorem has nothing to say about that. – leonbloy Jun 13 '11 at 16:33
Ok. Sorry then.. – Beni Bogosel Jun 13 '11 at 16:36
Hi Beni. The theorem is very interesting, in the sense that it is what motivates me to ask about infinite series... – Gadi A Jun 13 '11 at 16:59

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.