Department of Mathematics, University of California, Berkeley

Size: px

Start display at page:

Download "Department of Mathematics, University of California, Berkeley"

Tabitha Delphia Powers
5 years ago
Views:

1 ALGORITHMIC GALOIS THEORY Hendrik W. Lenstra jr. Mathematisch Instituut, Universiteit Leiden Department of Mathematics, University of California, Berkeley

2 K = field of characteristic zero, Ω = algebraically closed field containing K. K = {α Ω : n Z >0 : α n K } K K( K ) K = (algebraic) number field = finite extension of Q. K = K 0 K 1 K 2... K i+1 = K i ( Ki ) K = i 0 K i

3 K is the solvable closure of K inside Ω. Its elements can be written as nested radicals over K. The nesting depth d K (α) of α K over K is the smallest i with α K i. Example: ( ) 0 < d Q Exercise: compute d Q ( 1 + 2).

4 For any (possibly infinite) Galois extension K L, the following three algorithmic questions are equivalent: decide whether a given finite extension K M satisfies M L, for a given finite extension K M, compute the subfield L M of M, factor a given polynomial in K[X] into irreducible factors in L[X].

5 Theorem (Carl Cotner, 1995). There is an algorithm that, given algebraic number fields K M and α M, decides whether α K, and if so computes d K (α). Equivalently: Theorem. There is an algorithm that, given algebraic number fields K M, computes all subfields K 0 M, K 1 M, K 2 M,... of M. Can one do this in polynomial time?

6 Polynomial time algorithms An algorithm runs in polynomial time if there is a constant c, such that for every input, the run time of the algorithm is at most (2 + length of input) c.

7 Specifying number fields A number field is numerically specified by a system (a ijk ) of n 3 rational numbers a ijk. The additive group of the field is Q n, and the multiplication is given by (r i ) n i=1 (s j) n j=1 = = ( i,j a ) n ijkr i s j k=1. Elements of a number field are specified as vectors of rational numbers. Field homomorphisms (such as embeddings and automorphisms) are specified as matrices over Q.

8 There are polynomial time algorithms for: deciding whether a given system (a ijk ) Q n3 specifies an algebraic number field, finding all homomorphisms between two given number fields, finding a primitive element for a given extension K M, determining the irreducible polynomial of a given element in a given extension K M, intersecting subfields, factoring polynomials.

9 Fact. There is no polynomial time algorithm for computing the Galois closure of a given field extension K M. Proof. Take K = Q and M = Q(α), where α n α 1 = 0. Length of input: O(n c ). The degree of the Galois closure equals n!, so: run time length of output n! What about Galois groups?

10 Main techniques for determining Galois groups: reduction modulo primes Not much can be proved about the run time or about the correctness of such methods. results on permutation groups.

11 Theorem. There is an algorithm that for some c does the following: given an extension K M of number fields, and b Z >0, it decides in time at most (b + length of input) c whether the Galois group G of the normal closure F of K M satisfies #G b, and if so computes F and G.

12 Corollary. For each n there is a polynomial time algorithm that computes the Galois group of the Galois closure of a given extension K M of degree n. Open problem. Given finite extensions K M, K M, compute in polynomial time the intersection of M with the Galois closure of M.

13 Theorem. There is a polynomial time algorithm that, given an extension K M of number fields, decides whether the Galois group of its Galois closure equals the full symmetric group S [M:K]. Same for the alternating group A [M:K]. The proof uses that for n 8, the only sixfold transitive permutation groups of degree n are S n and A n.

14 K ab = maximal abelian extension of K inside Ω, K = solvable closure of K inside Ω. Theorem. There is a polynomial time algorithm for deciding whether a given extension K M of number fields satisfies M K ab. Theorem (Susan Landau & Gary Miller, 1985). There is a polynomial time algorithm for deciding whether a given extension K M of number fields satisfies M K.

15 For any (possibly infinite) Galois extension K L, the following three algorithmic questions are equivalent: decide whether a given finite extension K M satisfies M L, for a given finite extension K M, compute the subfield L M of M, factor a given polynomial in K[X] into irreducible factors in L[X].

16 The last two are equivalent under polynomial time reductions. Open problem: is there a polynomial time reduction of the second to the first? We can decide whether M K ab and whether M K in polynomial time. Can we compute K ab M and K M in polynomial time? Equivalently, can one factor in K ab [X] and K [X] in polynomial time?

17 µ = {ζ Ω : ζ n = 1 for some positive integer n}, K(µ) = maximal cyclotomic extension of K inside Ω, K ab = maximal abelian extension of K inside Ω, K = solvable closure of K inside Ω. Theorem. There are polynomial time algorithms that for a given extension K M of number fields compute the subfields K(µ) M, K ab M and K M of M.

18 Computing K ab M. Step 1: compute the largest Galois extension E of K inside M. 1a: write M = K(α) and factor the irreducible polynomial of α over K in M[X]. 1b: if there is an irreducible factor g of degree > 1, form M(β) = M[X]/(g), replace M = K(α) by K(α) K(β), and start again at 1a. 1c: otherwise E = M. Step 2: compute G = Aut K E and H = [G, G]; now K ab M equals {γ E : σ H : σγ = γ}.

19 Computing K(µ) M: Kronecker-Weber: Q ab = Q(µ), K(µ) M = K (Q ab M ).

20 Computing K M is done using fields K(α, β, γ, δ), where M = K(α) and β, γ, δ are conjugates of α over K. It uses the following result. Theorem (Ákos Seress, 1996). For each primitive solvable permutation group G of a finite set X there is a subset Y X with #Y 4 such that each σ G, σ 1, moves at least one element of Y.

21 Conclusion. For every number field K we can, in polynomial time, factor polynomials into irreducible factors in each of the rings K(µ)[X], K ab [X], K [X]. Problem. Do the same for the rings K i [X], i 0, where K 0 = K, K i+1 = K i ( Ki ), so that denesting of radicals will become feasible.

22 First result in the right direction: Theorem. There is a polynomial time algorithm that, given an extension K M of number fields, computes a set of representatives for the torsion subgroup of M /K.

Profinite Groups. Hendrik Lenstra. 1. Introduction

Profinite Groups. Hendrik Lenstra. 1. Introduction Profinite Groups Hendrik Lenstra 1. Introduction We begin informally with a motivation, relating profinite groups to the p-adic numbers. Let p be a prime number, and let Z p denote the ring of p-adic integers,