7.2. The cyclotomic-unit subgroups
There are two closely related real cyclotomic-unit subgroups. The logarithmic
saturation argument is most convenient for the squared family \varepsilon_a. The
index theorem is naturally stated for the normalised family
\eta_a = \zeta_p^{(1-a)/2}\frac{1-\zeta_p^a}{1-\zeta_p}, \qquad \eta_a^2=\varepsilon_a .
Here (1-a)/2 is understood modulo p, using that 2 is invertible modulo the
odd prime p.
The squared cyclotomic-unit subgroup. Let g=(p-3)/2. The subgroup
C^+_{\mathrm{sq}}\le E^+ is generated by -1 and by the units
\varepsilon_2,\varepsilon_3,\ldots,\varepsilon_{(p-1)/2}. Equivalently,
C^+_{\mathrm{sq}} = \left\langle -1,\ \varepsilon_a\ (2\le a\le (p-1)/2) \right\rangle .
Lean code for Definition7.2.1●1 definition
Associated Lean declarations
-
BernoulliRegular.CPlus[complete]
-
BernoulliRegular.CPlus[complete]
-
defdefined in BernoulliRegular/CyclotomicUnits/Subgroup.leancomplete
def BernoulliRegular.CPlus.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_three : 3 ≤ p) : Subgroup (NumberField.RingOfIntegers ↥(NumberField.maximalRealSubfield K))ˣ
def BernoulliRegular.CPlus.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_three : 3 ≤ p) : Subgroup (NumberField.RingOfIntegers ↥(NumberField.maximalRealSubfield K))ˣ
The real cyclotomic-unit subgroup `C⁺`, generated by `-1` and the standard real cyclotomic units.
The finite index set may be written as i\in \operatorname{Fin}((p-3)/2),
a=i+2. Thus the listed units are precisely the nontrivial real cyclotomic
generators in the prime-conductor range, together with the torsion unit -1.
The normalised cyclotomic-unit subgroup. The subgroup
C^+_{\mathrm{norm}}\le E^+ is generated by -1 and by the normalised real
cyclotomic units \eta_2,\eta_3,\ldots,\eta_{(p-1)/2}.
Lean code for Definition7.2.2●1 definition
Associated Lean declarations
-
BernoulliRegular.normalizedCPlus[complete]
-
BernoulliRegular.normalizedCPlus[complete]
-
defdefined in BernoulliRegular/CyclotomicUnits/NormalizedSubgroup.leancomplete
def BernoulliRegular.normalizedCPlus.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_odd : p ≠ 2) (hp_three : 3 ≤ p) : Subgroup (NumberField.RingOfIntegers ↥(NumberField.maximalRealSubfield K))ˣ
def BernoulliRegular.normalizedCPlus.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_odd : p ≠ 2) (hp_three : 3 ≤ p) : Subgroup (NumberField.RingOfIntegers ↥(NumberField.maximalRealSubfield K))ˣ
The TeX normalized subgroup `C⁺ = <-1, ε₂, ..., ε_g> ≤ (𝓞 K⁺)ˣ`.
The relation between the two families is exactly the relation \eta_a^2=\varepsilon_a.
Squares of normalised generators. For each standard generator index a,
\eta_a^2=\varepsilon_a.
Lean code for Lemma7.2.3●1 theorem
Associated Lean declarations
-
theoremdefined in BernoulliRegular/CyclotomicUnits/NormalizedSubgroup.leancomplete
theorem BernoulliRegular.normalizedCPlusGenerator_sq_eq_CPlusGenerator.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_odd : p ≠ 2) (hp_three : 3 ≤ p) (i : Fin ((p - 3) / 2)) : BernoulliRegular.normalizedCPlusGenerator hp_odd hp_three i ^ 2 = BernoulliRegular.CPlusGenerator hp_three i
theorem BernoulliRegular.normalizedCPlusGenerator_sq_eq_CPlusGenerator.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_odd : p ≠ 2) (hp_three : 3 ≤ p) (i : Fin ((p - 3) / 2)) : BernoulliRegular.normalizedCPlusGenerator hp_odd hp_three i ^ 2 = BernoulliRegular.CPlusGenerator hp_three i
This is the elementary identity
\left( \zeta_p^{(1-a)/2}\frac{1-\zeta_p^a}{1-\zeta_p} \right)^2 = \zeta_p^{1-a} \left(\frac{1-\zeta_p^a}{1-\zeta_p}\right)^2 = u_a\overline{u_a}.
The last equality is the standard computation
\overline{u_a} = \frac{1-\zeta_p^{-a}}{1-\zeta_p^{-1}} = \zeta_p^{1-a}\frac{1-\zeta_p^a}{1-\zeta_p}.
Odd-primary comparison of the two indices. For odd p,
p \mid [E^+ : C^+_{\mathrm{sq}}] \quad\Longleftrightarrow\quad p \mid [E^+ : C^+_{\mathrm{norm}}].
Lean code for Lemma7.2.4●2 theorems
Associated Lean declarations
-
theoremdefined in BernoulliRegular/CyclotomicUnits/NormalizedSubgroup.leancomplete
theorem BernoulliRegular.CPlus_index_prime_dvd_iff_normalizedCPlus_index_prime_dvd.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_odd : p ≠ 2) (hp_three : 3 ≤ p) : p ∣ (BernoulliRegular.CPlus hp_three).index ↔ p ∣ (BernoulliRegular.normalizedCPlus hp_odd hp_three).index
theorem BernoulliRegular.CPlus_index_prime_dvd_iff_normalizedCPlus_index_prime_dvd.{u_1} {p : ℕ} [Fact (Nat.Prime p)] {K : Type u_1} [Field K] [NumberField K] [IsCyclotomicExtension {p} ℚ K] [NumberField.IsCMField K] (hp_odd : p ≠ 2) (hp_three : 3 ≤ p) : p ∣ (BernoulliRegular.CPlus hp_three).index ↔ p ∣ (BernoulliRegular.normalizedCPlus hp_odd hp_three).index
-
theoremdefined in BernoulliRegular/CyclotomicUnits/IndexComparison.leancomplete
theorem BernoulliRegular.subgroup_index_prime_dvd_iff_of_relIndex_dvd_two_pow.{u_1} {G : Type u_1} [Group G] {H K : Subgroup G} (hHK : H ≤ K) {p r : ℕ} (hp : Nat.Prime p) (hp_odd : p ≠ 2) (hrel : H.relIndex K ∣ 2 ^ r) : p ∣ H.index ↔ p ∣ K.index
theorem BernoulliRegular.subgroup_index_prime_dvd_iff_of_relIndex_dvd_two_pow.{u_1} {G : Type u_1} [Group G] {H K : Subgroup G} (hHK : H ≤ K) {p r : ℕ} (hp : Nat.Prime p) (hp_odd : p ≠ 2) (hrel : H.relIndex K ∣ 2 ^ r) : p ∣ H.index ↔ p ∣ K.index
The subgroup C^+_{\mathrm{sq}} is contained in C^+_{\mathrm{norm}}, because
-1 is a normalised generator and each squared generator \varepsilon_a is the
square of the normalised generator \eta_a. Conversely, every class in
C^+_{\mathrm{norm}}/C^+_{\mathrm{sq}} has square one: it is enough to check this on
the generators, and it follows from \eta_a^2=\varepsilon_a and (-1)^2=1. Hence
the relative index [C^+_{\mathrm{norm}}:C^+_{\mathrm{sq}}] divides a power of 2.
The two ambient indices differ by this relative index. Since p is odd, p does
not divide any power of 2. Therefore the p-primary divisibility of
[E^+ : C^+_{\mathrm{sq}}] and [E^+ : C^+_{\mathrm{norm}}] is the same.