2
$\begingroup$

I want to determine the number of roots of $f(X) = X^3-5X+20$ in $\mathbb{Z}_p$ using Hensel's lemma (lemma is on the bottom). Unfortunately I am not very well trained to solve this. Take for example $p=2$. Then reducing the polynomial modulo 2 we get \begin{align*} f(X) &= X^3 - 5X+20 \equiv 0 \mod 2 \\ f'(X) &= 3X^2-5\equiv X + 1 \mod 2 \end{align*} That would mean that only $0\in\mathbb{Z}/2\mathbb{Z}$ gives a solution such that $f(0) \equiv 0$ and $f'(0)\not \equiv 0$ modulo 2, which then by Hensel's lemma gives a unique root of $f$ in $\mathbb{Z}_2$, which is $0$ modulo $2$.

The problem is that I don't know the relation between finding a root via Hensel's lemma and there really being one. I have found one root via Hensel's lemma, but what guarantees me there aren't more?

Any help on how to proceed is most welcome. Thank you in advance.


The lemma I am using:

Let $K$ be a complete field with discrete valuation $v$, valuation ring $O$ and residue field $k$. Let $f(X)\in O(X)$. Suppose that the reduction $\bar{f}(X)\in k[X]$ has a simple root $\bar{a}$, ie $\bar{f}(\bar{a}) = 0$ and $\bar{f}'(\bar{a})\neq0$. Then there exists a unique $x\in O$ such that $f(x) = 0$ and $\bar{x}=\bar{a}$.

$\endgroup$
4
  • $\begingroup$ Are you lifting roots mod $p$ to roots mod powers of $p$? That's what Hensel's Lemma is about. There are $2$ roots mod $2$, by inspection. $\endgroup$ Commented May 20, 2015 at 19:58
  • $\begingroup$ I added the lemma as I learned it. Could you explain how you determined there are two roots? $\endgroup$
    – Marc
    Commented May 20, 2015 at 20:04
  • $\begingroup$ Plugged in. But that may be because my meaning for $\mathbb{Z}_2$ is different from yours. $\endgroup$ Commented May 20, 2015 at 20:16
  • $\begingroup$ Ah my communication is so off.... I meant the $2$-adic number field. $\endgroup$
    – Marc
    Commented May 20, 2015 at 20:18

2 Answers 2

3
$\begingroup$

This polynomial has a double root at $1$ in $\mathbb{F}_2$, so the out of the box Hensel's lemma won't apply. However we can still apply a modified version by going one power of two up. In particular the double root at 1 splits into two simple roots mod 4 (at 1 and 3). Now a modified Hensel's lemma (with essentially the same proof as the usual version) gives us that these each have a unique 2-adic lift.

One thing that is different about this modified Hensel's lemma from what you are used to is that if we want to solve $f(x) \equiv 0 \bmod 2^n$, the lifts we get will only be unique mod $2^{n-1}$. In particular we get that $f(x) \equiv 0 \bmod 2^n$ has 5 solutions for all $n > 2$.

$\endgroup$
3
  • $\begingroup$ Thank you very much! I looked at the proof and was able to adjust the assumption to powers of $2$. One more question, is there a way to see whether a root splits in different ones in a higher power? Take $p=5$ for example. Then $f$ has a triple root at zero modulo 5, which (I believe) are no roots anymore when we look modulo 25. It would be exhausting (and impossible) to check all powers. Thank you. $\endgroup$
    – Marc
    Commented May 20, 2015 at 21:45
  • $\begingroup$ I think the strategy is to increase the prime power one step at a time. If you ever have a root such that the derivative does not vanish then it automatically lifts to all higher powers by this modified Hensel's lemma. If you reach a power such that there are no solutions lifting the one you are looking at then there are no p-adic solutions. One thing to potentially be worried about is maybe you keep going up a prime power finding lifts but having the derivatives keep vanishing, but I think one can show that implies the polynomial has a double root. $\endgroup$
    – Nate
    Commented May 21, 2015 at 3:51
  • $\begingroup$ For $p=5$, Eisenstein applies, since all coeffs are divisible by $5$ and the constant is not divisible by $25$. So irreducible, no roots in $\Bbb Z_5$. $\endgroup$
    – Lubin
    Commented May 23, 2015 at 5:19
3
$\begingroup$

Just for the number of roots in the $2$-adic integers $\Bbb Z_2$, I recommend forming $g(X)=f(X+5)=X^3+15X^2+70X+120$. Then I use something rather more powerful than plain old Hensel, namely the Newton polygon, in which you plot $(n,v_2(a_n))$ for every monomial $a_nX^n$ in your polynomial. The points you plot then are $(0,3)$, $(1,1)$, $(2,0)$, and $(3,0)$. You then take the “upper convex hull” of these points, which has segments of slope $-2$, $-1$, and $0$, each of width $1$. This says that in $\Bbb Z_2$, there are roots of $2$-adic valuation $2$, $1$, and $0$, three in all. You may be able to take $g$ and find the roots directly with your form of Hensel.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .