Documentation

TauCeti.NumberTheory.NumberField.Frobenius.Tower

Raising the base field: the tower formula for arithmetic Frobenius elements #

Let K โІ M โІ L be number fields with L / K and L / M Galois, let Q be a prime of ๐“ž L unramified over ๐“ž K, and write ๐”“ = Q โˆฉ ๐“ž M and ๐”ญ = Q โˆฉ ๐“ž K. An arithmetic Frobenius ฯƒ โˆˆ Gal(L/K) at Q acts on the residue field by x โ†ฆ x ^ ๐”‘๐”ญ, while an arithmetic Frobenius ฯ„ โˆˆ Gal(L/M) at Q acts by x โ†ฆ x ^ ๐”‘๐”“. Since ๐”‘๐”“ = ๐”‘๐”ญ ^ f(๐”“/๐”ญ), the two are related by

AlgEquiv.restrictScalars K ฯ„ = ฯƒ ^ f(๐”“/๐”ญ).

The exponent is a power, the residue degree of the intermediate prime over the base, and never an inverse. In particular ฯƒ ^ f(๐”“/๐”ญ) fixes M pointwise even though M / K need not be normal and ฯƒ itself need not preserve M.

This is the second of the two tower laws for Frobenius elements. The first, restriction along a normal subextension, takes no power at all and is IsArithFrobAt.restrictNormal in TauCeti.NumberTheory.NumberField.Frobenius.Restriction.

The two hypotheses that are not decoration #

The statement is relative to one prime Q of L. Replacing ฯƒ by an arbitrary conjugate โ€” an arbitrary representative of the Artin class of ๐”ญ โ€” makes it false when M / K is not normal: a conjugate need not stabilize Q, so its f-th power need not fix M pointwise, and it is then the restriction of no element of Gal(L/M) at all.

Unramifiedness of Q over ๐“ž K is likewise essential. At a ramified prime a Frobenius lift is determined only modulo inertia, so there is no equality of automorphisms to prove; the statement would have to be made in the quotient by the inertia subgroup, or about a coset. Here unramifiedness enters through Ideal.stabilizerHom_injective_of_isUnramifiedAt: the decomposition group of Q embeds into the automorphism group of the residue extension, so an element of Gal(L/K) stabilizing Q is pinned down by its residue action, and both ฯƒ ^ f(๐”“/๐”ญ) and AlgEquiv.restrictScalars K ฯ„ act by x โ†ฆ x ^ ๐”‘๐”“.

Main results #

References #

theorem NumberField.isArithFrobAt_int_of_absNorm_eq {K : Type u_1} {L : Type u_2} [Field K] [NumberField K] [Field L] [NumberField L] [Algebra K L] {p : โ„•} (P : Ideal (RingOfIntegers K)) [P.LiesOver (Ideal.span {โ†‘p})] (hnorm : Ideal.absNorm P = p) (Q : Ideal (RingOfIntegers L)) [Q.LiesOver P] {ฯƒ : Gal(L/K)} (hฯƒ : IsArithFrobAt (RingOfIntegers K) ฯƒ Q) :

A relative Frobenius above an ideal of norm p is a Frobenius over (p). Let P be an ideal of a number field K above (p) with absolute norm p. If Q lies above P, then a Frobenius at Q relative to K is also a Frobenius relative to โ„ค after restricting its automorphism from K-linearity to โ„š-linearity.

The norm equality makes the two residue cardinalities in the definitions equal; no primality hypothesis on p, P, or Q is required.

theorem NumberField.restrictScalars_eq_pow_inertiaDeg {K : Type u_1} {L : Type u_2} [Field K] [NumberField K] [Field L] [NumberField L] [Algebra K L] [IsGalois K L] {M : Type u_3} [Field M] [NumberField M] [Algebra K M] [Algebra M L] [IsScalarTower K M L] {Q : Ideal (RingOfIntegers L)} [Q.IsPrime] [Algebra.IsUnramifiedAt (RingOfIntegers K) Q] {ฯƒ : Gal(L/K)} (hฯƒ : IsArithFrobAt (RingOfIntegers K) ฯƒ Q) {ฯ„ : Gal(L/M)} (hฯ„ : IsArithFrobAt (RingOfIntegers M) ฯ„ Q) :

Raising the base field raises the Frobenius to the residue degree. For number fields K โІ M โІ L with L / K and L / M Galois and Q a prime of ๐“ž L unramified over ๐“ž K, the restriction to Gal(L/K) of an arithmetic Frobenius ฯ„ โˆˆ Gal(L/M) at Q is the f(๐”“/๐”ญ)-th power of an arithmetic Frobenius ฯƒ โˆˆ Gal(L/K) at Q, where ๐”“ = Q โˆฉ ๐“ž M and ๐”ญ = Q โˆฉ ๐“ž K.

The exponent is the residue degree of ๐”“ over ๐“ž K, and it is a power: the residue field of ๐”“ has ๐”‘๐”ญ ^ f(๐”“/๐”ญ) elements, so the two Frobenius elements have residue actions x โ†ฆ x ^ ๐”‘๐”ญ and x โ†ฆ x ^ ๐”‘๐”ญ ^ f(๐”“/๐”ญ). Both sides are read inside Gal(L/K), where ฯ„ is placed by AlgEquiv.restrictScalars.

theorem NumberField.isArithFrobAt_iff_restrictScalars_eq_pow_inertiaDeg {K : Type u_1} {L : Type u_2} [Field K] [NumberField K] [Field L] [NumberField L] [Algebra K L] [IsGalois K L] {M : Type u_3} [Field M] [NumberField M] [Algebra K M] [Algebra M L] [IsScalarTower K M L] {Q : Ideal (RingOfIntegers L)} [Q.IsPrime] [Algebra.IsUnramifiedAt (RingOfIntegers K) Q] {ฯƒ : Gal(L/K)} (hฯƒ : IsArithFrobAt (RingOfIntegers K) ฯƒ Q) (ฯ„ : Gal(L/M)) :

The relative Frobenius restricts to exactly that power. An element of Gal(L/M) is an arithmetic Frobenius at Q precisely when its restriction to Gal(L/K) is the f(๐”“/๐”ญ)-th power of a fixed arithmetic Frobenius ฯƒ โˆˆ Gal(L/K) at Q.

The converse direction is uniqueness rather than a second computation: a Frobenius of Gal(L/M) at Q exists, is carried to ฯƒ ^ f(๐”“/๐”ญ) by restrictScalars_eq_pow_inertiaDeg, and AlgEquiv.restrictScalars is injective.

The tower formula for Mathlib's chosen Frobenius. Mathlib's arithFrobAt picks a Frobenius coherently across the fibre above a prime of the base; at an unramified Q the choices over ๐“ž K and over ๐“ž M are related by the tower formula, because at such a Q there is nothing left to choose.

This is the form the Artin symbol is built from, NumberField.artinSymbol being the conjugacy class of arithFrobAt (๐“ž K) (L โ‰ƒโ‚[K] L) at a prime above ๐”ญ.

theorem NumberField.exists_isArithFrobAt_pow_inertiaDeg {K : Type u_1} {L : Type u_2} [Field K] [NumberField K] [Field L] [NumberField L] [Algebra K L] [IsGalois K L] (M : Type u_4) [Field M] [NumberField M] [Algebra K M] [Algebra M L] [IsScalarTower K M L] [IsGalois M L] (Q : Ideal (RingOfIntegers L)) [Q.IsPrime] (๐”“ : Ideal (RingOfIntegers M)) (๐”ญ : Ideal (RingOfIntegers K)) (hQM : Ideal.under (RingOfIntegers M) Q = ๐”“) (hQK : Ideal.under (RingOfIntegers K) Q = ๐”ญ) (hur : โˆ€ (Q' : Ideal (RingOfIntegers L)) [inst : Q'.IsPrime] [Q'.LiesOver ๐”ญ], Algebra.IsUnramifiedAt (RingOfIntegers K) Q') (ฯƒ : Gal(L/K)) (hฯƒ : IsArithFrobAt (RingOfIntegers K) ฯƒ Q) :
โˆƒ (ฯ„ : Gal(L/M)), IsArithFrobAt (RingOfIntegers M) ฯ„ Q โˆง AlgEquiv.restrictScalars K ฯ„ = ฯƒ ^ ๐”“.inertiaDeg (RingOfIntegers K)

The tower formula, in the form the Artin symbol consumes. For number fields K โІ M โІ L with L / K and L / M Galois, a prime Q of ๐“ž L with ๐”“ = Q โˆฉ ๐“ž M and ๐”ญ = Q โˆฉ ๐“ž K, and ๐”ญ unramified in L, every arithmetic Frobenius ฯƒ โˆˆ Gal(L/K) at Q has ฯƒ ^ f(๐”“/๐”ญ) as the image in Gal(L/K) of an arithmetic Frobenius at Q in Gal(L/M).

The unramifiedness hypothesis is phrased for all primes above ๐”ญ, matching the argument that NumberField.artinSymbol takes; only its value at Q is used.

theorem NumberField.pow_inertiaDeg_apply_algebraMap {K : Type u_1} {L : Type u_2} [Field K] [NumberField K] [Field L] [NumberField L] [Algebra K L] [IsGalois K L] {M : Type u_3} [Field M] [NumberField M] [Algebra K M] [Algebra M L] [IsScalarTower K M L] {Q : Ideal (RingOfIntegers L)} [Q.IsPrime] [Algebra.IsUnramifiedAt (RingOfIntegers K) Q] {ฯƒ : Gal(L/K)} (hฯƒ : IsArithFrobAt (RingOfIntegers K) ฯƒ Q) (y : M) :

The power of the Frobenius fixes the intermediate field. Even though M / K need not be normal, and ฯƒ therefore need not preserve M, its f(๐”“/๐”ญ)-th power fixes M pointwise: it is the image of an element of Gal(L/M).

theorem NumberField.restrictScalars_eq_of_inertiaDeg_eq_one {K : Type u_1} {L : Type u_2} [Field K] [NumberField K] [Field L] [NumberField L] [Algebra K L] [IsGalois K L] {M : Type u_3} [Field M] [NumberField M] [Algebra K M] [Algebra M L] [IsScalarTower K M L] {Q : Ideal (RingOfIntegers L)} [Q.IsPrime] [Algebra.IsUnramifiedAt (RingOfIntegers K) Q] {ฯƒ : Gal(L/K)} (hฯƒ : IsArithFrobAt (RingOfIntegers K) ฯƒ Q) {ฯ„ : Gal(L/M)} (hฯ„ : IsArithFrobAt (RingOfIntegers M) ฯ„ Q) (hf : (Ideal.under (RingOfIntegers M) Q).inertiaDeg (RingOfIntegers K) = 1) :

At residue degree one the restricted Frobenius does not move. If ๐”“ = Q โˆฉ ๐“ž M has residue degree one over ๐“ž K, then the restriction to Gal(L/K) of an arithmetic Frobenius of Gal(L/M) at Q is the arithmetic Frobenius of Gal(L/K) at Q itself.

This is the case a fixed-field fibre count runs through: it contracts the primes of M whose residue degree over K is one, exactly so that the restriction of the relative Frobenius is the absolute one.

theorem NumberField.isArithFrobAt_restrictScalars_of_inertiaDeg_eq_one {K : Type u_1} {L : Type u_2} [Field K] [NumberField K] [Field L] [NumberField L] [Algebra K L] {M : Type u_3} [Field M] [NumberField M] [Algebra K M] [Algebra M L] [IsScalarTower K M L] {Q : Ideal (RingOfIntegers L)} [Q.IsPrime] {ฯ„ : Gal(L/M)} (hฯ„ : IsArithFrobAt (RingOfIntegers M) ฯ„ Q) (hf : (Ideal.under (RingOfIntegers M) Q).inertiaDeg (RingOfIntegers K) = 1) :

A relative Frobenius at residue degree one restricts to an absolute one. If Q โˆฉ ๐“ž M has residue degree one over ๐“ž K, then the restriction to Gal(L/K) of an arithmetic Frobenius of Gal(L/M) at Q is itself an arithmetic Frobenius at Q over ๐“ž K.

Residue degree one says the two residue fields have the same size, and restricting scalars does not move the automorphism, so the two Frobenius conditions are the same condition. Neither L / K Galois nor Q unramified is needed: this is a statement about Q alone.

Trivial relative Frobenius elements #

A prime that is already inert enough over โ„š has trivial relative Frobenius: if the absolute Frobenius ฯ at Q has ฯ ^ n = 1 and the residue field of Q โˆฉ ๐“ž M has p ^ n elements, then the residue action of the relative Frobenius, raising to the power p ^ n, is the identity.

theorem NumberField.isArithFrobAt_one_of_pow_eq_one {L : Type u_2} [Field L] [NumberField L] {M : Type u_3} [Field M] [Algebra M L] {Q : Ideal (RingOfIntegers L)} {p n : โ„•} {ฯ : Gal(L/โ„š)} [Q.LiesOver (Ideal.span {โ†‘p})] (hฯ : IsArithFrobAt โ„ค ฯ Q) (hฯn : ฯ ^ n = 1) (hcard : Nat.card (RingOfIntegers M โงธ Ideal.under (RingOfIntegers M) Q) = p ^ n) :

The identity is a relative Frobenius at a prime whose base residue field absorbs the absolute one. With M โІ L number fields, Q a prime of ๐“ž L above the rational prime p, and ฯ โˆˆ Gal(L/โ„š) an arithmetic Frobenius at Q over โ„ค with ฯ ^ n = 1, the residue field of Q โˆฉ ๐“ž M having p ^ n elements forces the relative residue action x โ†ฆ x ^ p ^ n to be the identity.

theorem NumberField.isArithFrobAt_eq_one_of_pow_eq_one {L : Type u_2} [Field L] [NumberField L] {M : Type u_3} [Field M] [Algebra M L] {Q : Ideal (RingOfIntegers L)} [Q.IsPrime] [Algebra.IsUnramifiedAt (RingOfIntegers M) Q] {p n : โ„•} {ฯ : Gal(L/โ„š)} [Q.LiesOver (Ideal.span {โ†‘p})] (hฯ : IsArithFrobAt โ„ค ฯ Q) (hฯn : ฯ ^ n = 1) (hcard : Nat.card (RingOfIntegers M โงธ Ideal.under (RingOfIntegers M) Q) = p ^ n) {ฯƒ : Gal(L/M)} (hฯƒ : IsArithFrobAt (RingOfIntegers M) ฯƒ Q) :
ฯƒ = 1

The relative Frobenius is trivial at such a prime. Under the hypotheses of NumberField.isArithFrobAt_one_of_pow_eq_one, and with L / M unramified at Q, every arithmetic Frobenius of Gal(L/M) at Q is the identity. No normality hypothesis on L/M is needed.