Introduction
The generation of abelian extensions of imaginary quadratic fields by special values of modular functions is a classical problem in complex multiplication. We study the generation of ray class fields by a single Weber value and of their intermediate fields by relative norms.
Let K K K be an imaginary quadratic field of discriminant d K d_K d K , and let O K \mathcal{O}_K O K be its ring of integers. For a nonzero ideal m \mathfrak{m} m of O K \mathcal{O}_K O K , let C ( m ) C(\mathfrak{m}) C ( m ) denote the ray class group of K K K modulo m \mathfrak{m} m . By class field
2020 Mathematics Subject Classification. Primary 11R37; Secondary 11F03, 11G05.
Key words and phrases. Class field theory, elliptic curves, modular functions. theory, there is a unique finite abelian extension K m K_{\mathfrak m} K m of K K K , unramified outside the prime ideals of O K \mathcal{O}_K O K dividing m \mathfrak m m , for which the Artin map modulo m \mathfrak m m induces an isomorphism
C ( m ) → ∼ Gal ( K m / K ) . C(\mathfrak m) \xrightarrow{\sim} \operatorname{Gal}(K_{\mathfrak m}/K). C ( m ) ∼ Gal ( K m / K ) .
This extension is called the ray class field of K K K modulo m \mathfrak m m ; see [7 Chapters III–IV] . When m = O K \mathfrak m=\mathcal{O}_K m = O K , it is the Hilbert class field H K H_K H K of $K.
Let E / C E/\mathbb{C} E / C be an elliptic curve associated with the lattice O K \mathcal{O}_K O K . Then E E E has complex multiplication by O K \mathcal{O}_K O K , and scalar multiplication gives an isomorphism
O K → ∼ End ( E ) , a ↦ [ a ] E . \mathcal{O}_K \xrightarrow{\sim} \operatorname{End}(E),\qquad a\mapsto[a]_E. O K ∼ End ( E ) , a ↦ [ a ] E .
In particular, the group Aut ( E ) \operatorname{Aut}(E) Aut ( E ) of automorphisms is isomorphic to O K × \mathcal{O}_K^\times O K × . We use the Weierstrass model
y 2 = 4 x 3 − g 2 x − g 3 , where g 2 = g 2 ( O K ) and g 3 = g 3 ( O K ) (1) y^2=4x^3-g_2x-g_3,\qquad\text{where }g_2=g_2(\mathcal{O}_K)\text{ and }g_3=g_3(\mathcal{O}_K) \tag*{(1)} y 2 = 4 x 3 − g 2 x − g 3 , where g 2 = g 2 ( O K ) and g 3 = g 3 ( O K ) (1)
for E E E . The associated complex analytic isomorphism is
ϕ E : C / O K → ∼ E ( C ) ( ⊂ P 2 ( C ) ) , z + O K ↦ { [ ℘ ( z ; O K ) : ℘ ′ ( z ; O K ) : 1 ] if z ∈ C ∖ O K , [ 0 : 1 : 0 ] if z ∈ O K . (2) \phi_E:\mathbb{C}/\mathcal{O}_K\xrightarrow{\sim}E(\mathbb{C})\left(\subset\mathbb{P}^2(\mathbb{C})\right),\qquad z+\mathcal{O}_K\mapsto
\begin{cases}
[\wp(z;\mathcal{O}_K):\wp'(z;\mathcal{O}_K):1] & \text{if }z\in\mathbb{C}\setminus\mathcal{O}_K,\\
[0:1:0] & \text{if }z\in\mathcal{O}_K.
\end{cases} \tag*{(2)} ϕ E : C / O K ∼ E ( C ) ( ⊂ P 2 ( C ) ) , z + O K ↦ { [ ℘ ( z ; O K ) : ℘ ′ ( z ; O K ) : 1 ] [ 0 : 1 : 0 ] if z ∈ C ∖ O K , if z ∈ O K . (2)
Here ℘ = ℘ ( ⋅ ; O K ) : C → C ∪ { ∞ } \wp=\wp(\mathord{\cdot};\mathcal{O}_K):\mathbb{C}\to\mathbb{C}\cup\{\infty\} ℘ = ℘ ( ⋅ ; O K ) : C → C ∪ { ∞ } is the Weierstrass function for the lattice O K \mathcal{O}_K O K , and ℘ ′ \wp' ℘ ′ is its derivative; see [5 Theorem 10.14] and [23 Proposition 3.6 in Chapter VI] . Let O E = [ 0 : 1 : 0 ] O_E=[0:1:0] O E = [ 0 : 1 : 0 ] denote the origin of E E E .
We consider the Weber function
h E : E → P 1 ( C ) = C ∪ { ∞ } h_E:E\to\mathbb{P}^1(\mathbb{C})=\mathbb{C}\cup\{\infty\} h E : E → P 1 ( C ) = C ∪ { ∞ }
defined by h E ( O E ) = ∞ h_E(O_E)=\infty h E ( O E ) = ∞ and
h E ( ϕ E ( z + O K ) ) = { g 2 g 3 Δ ℘ ( z ; O K ) if j ( E ) ≠ 1728 , 0 , g 2 2 Δ ℘ ( z ; O K ) 2 if j ( E ) = 1728 , g 3 Δ ℘ ( z ; O K ) 3 if j ( E ) = 0 ( z ∈ C ∖ O K ) , (3) h_E(\phi_E(z+\mathcal{O}_K))=
\begin{cases}
\frac{g_2g_3}{\Delta}\wp(z;\mathcal{O}_K) & \text{if }j(E)\ne1728,0,\\
\frac{g_2^2}{\Delta}\wp(z;\mathcal{O}_K)^2 & \text{if }j(E)=1728,\\
\frac{g_3}{\Delta}\wp(z;\mathcal{O}_K)^3 & \text{if }j(E)=0
\end{cases}
\qquad(z\in\mathbb{C}\setminus\mathcal{O}_K), \tag*{(3)} h E ( ϕ E ( z + O K )) = ⎩ ⎨ ⎧ Δ g 2 g 3 ℘ ( z ; O K ) Δ g 2 2 ℘ ( z ; O K ) 2 Δ g 3 ℘ ( z ; O K ) 3 if j ( E ) = 1728 , 0 , if j ( E ) = 1728 , if j ( E ) = 0 ( z ∈ C ∖ O K ) , (3)
where Δ = g 2 3 − 27 g 3 2 \Delta=g_2^3-27g_3^2 Δ = g 2 3 − 27 g 3 2 . It realizes the quotient map E → E / Aut ( E ) ≃ P 1 ( C ) E\to E/\operatorname{Aut}(E)\simeq\mathbb{P}^1(\mathbb{C}) E → E / Aut ( E ) ≃ P 1 ( C ) . See [22 Example 5.5.2] .
For a nonzero proper ideal m \mathfrak m m of O K \mathcal{O}_K O K , a point P ∈ E ( C ) P\in E(\mathbb{C}) P ∈ E ( C ) is called a primitive m \mathfrak m m -division point if
Ann O K ( P ) = { a ∈ O K ∣ [ a ] E P = O E } = m . \operatorname{Ann}_{\mathcal{O}_K}(P)=\{a\in\mathcal{O}_K\mid[a]_EP=O_E\}=\mathfrak m. Ann O K ( P ) = { a ∈ O K ∣ [ a ] E P = O E } = m .
The classical theory of complex multiplication, developed by Kronecker, Weber, Hasse, Deuring, and others, gives
H K = K ( j ( E ) ) H_K=K(j(E)) H K = K ( j ( E ))
and
K m = H K ( h E ( P ) ) for any primitive m -division point P ∈ E ( C ) . (4) K_{\mathfrak m}=H_K(h_E(P))\quad\text{for any primitive }\mathfrak m\text{-division point }P\in E(\mathbb{C}). \tag*{(4)} K m = H K ( h E ( P )) for any primitive m -division point P ∈ E ( C ) . (4)
As P P P varies over these points, the values h E ( P ) h_E(P) h E ( P ) form the full set of Galois conjugates over H K H_K H K of any one of them; see [13 §10.3] . Hasse further asked whether H K H_K H K in (4) can be replaced by K K K , that is, whether
K m a t h b f m = K ( h E ( P ) ) for every primitive m -division point P ∈ E ( C ) . (5) K_{mathbf{m}} = K(h_E(P)) \quad\text{for every primitive }\mathbf{m}\text{-division point }P \in E(\mathbb{C}). \tag*{(5)} K ma t hb f m = K ( h E ( P )) for every primitive m -division point P ∈ E ( C ) . (5)
For a nonzero ideal a \mathfrak{a} a of O K \mathcal{O}_K O K , let N ( a ) = ∣ O K / a ∣ N(\mathfrak{a}) = |\mathcal{O}_K/\mathfrak{a}| N ( a ) = ∣ O K / a ∣ be its absolute norm. Unless otherwise specified, we denote by p \mathfrak{p} p a nonzero prime ideal of O K \mathcal{O}_K O K . Put
φ ( a ) = N ( a ) ∏ p ∣ a ( 1 − 1 N ( p ) ) and Ψ ( a ) = N ( a ) ∏ p ∣ a ( 1 − 2 N ( p ) ) . \varphi(\mathfrak{a}) = N(\mathfrak{a}) \prod_{\mathfrak{p}\mid\mathfrak{a}} \left(1 - \frac{1}{N(\mathfrak{p})}\right)
\quad\text{and}\quad
\Psi(\mathfrak{a}) = N(\mathfrak{a}) \prod_{\mathfrak{p}\mid\mathfrak{a}} \left(1 - \frac{2}{N(\mathfrak{p})}\right). φ ( a ) = N ( a ) p ∣ a ∏ ( 1 − N ( p ) 1 ) and Ψ ( a ) = N ( a ) p ∣ a ∏ ( 1 − N ( p ) 2 ) .
In 1936 Sugawara [25 , 26 ] gave the first partial answers to Hasse’s question, proving (5) whenever at least one of the following conditions holds:
(S1) 4 O K ∣ m 4\mathcal{O}_K \mid\mathbf{m} 4 O K ∣ m and m ≠ 4 O K \mathbf{m} \ne4\mathcal{O}_K m = 4 O K ;
(S2) p 2 ∣ m \mathfrak{p}^{2} \mid\mathbf{m} p 2 ∣ m for some prime ideal p \mathfrak{p} p of O K \mathcal{O}_K O K such that p ∣ 2 O K \mathfrak{p} \mid2\mathcal{O}_K p ∣ 2 O K , and φ ( m ) ≥ 6 \varphi(\mathbf{m}) \ge6 φ ( m ) ≥ 6 ;
(S3) Ψ ( m ) ≥ 5 \Psi(\mathbf{m}) \ge5 Ψ ( m ) ≥ 5 .
His method uses addition formulas for the Weierstrass ℘ \wp ℘ -function to recover j ( E ) j(E) j ( E ) from suitable Galois conjugates of h E ( P ) h_E(P) h E ( P ) . Subsequently, Jung-Koo-Shin [8 ] and Koo-Shin-Yoon [11 ] proved (5) for m = N O K \mathbf{m} = N\mathcal{O}_K m = N O K , where N ≥ 2 N \ge2 N ≥ 2 is an integer, under the hypothesis that m \mathbf{m} m is the exact conductor of K m / K K_{\mathbf{m}}/K K m / K , using modular functions and the second Kronecker limit formula.
Let f K m / K \mathfrak{f}_{K_{\mathbf{m}}/K} f K m / K denote the exact conductor of K m / K K_{\mathbf{m}}/K K m / K . Since
f K m / K ∣ m and K m = K f K m / K , \mathfrak{f}_{K_{\mathbf{m}}/K} \mid\mathbf{m}
\quad\text{and}\quad
K_{\mathbf{m}} = K_{\mathfrak{f}_{K_{\mathbf{m}}/K}}, f K m / K ∣ m and K m = K f K m / K ,
it is natural to consider (5) under the hypothesis m = f K m / K \mathbf{m} = \mathfrak{f}_{K_{\mathbf{m}}/K} m = f K m / K . Morton [16 ] recently presented a proof of (5) under this hypothesis. Starting from Sugawara’s criteria and the conductor condition, he reduced the problem to seven remaining families of moduli and treated them by algebraic and arithmetic arguments involving explicit calculations on suitable elliptic curves.
In this paper, we give an alternative proof of (5) . We follow Morton’s reduction to the seven remaining families, recalled in Proposition 2.1, and treat them by two methods. Four families are handled by a character argument involving Siegel functions, the second Kronecker limit formula and the nonvanishing of Hecke L L L -values at s = 1 s = 1 s = 1 (Theorem 6.2). The other three are handled by explicit estimates for the Weierstrass ℘ \wp ℘ -function, the elliptic modular function j j j and the modular discriminant Δ \Delta Δ (Theorem 7.3). Our first main result is the following.
Theorem A (Theorem 7.4). Let K K K be an imaginary quadratic field, and let E E E and h E h_E h E be as in (1)–(3). If m \mathbf{m} m is a nonzero proper ideal of O K \mathcal{O}_K O K that is the exact conductor of K m / K K_{\mathbf{m}}/K K m / K , then
K m = K ( h E ( P ) ) K_{\mathbf{m}} = K(h_E(P)) K m = K ( h E ( P )) for every primitive m \mathbf{m} m -division point P ∈ E ( C ) P \in E(\mathbb{C}) P ∈ E ( C ) .
Ramachandra [17 Theorems 9 and 10] constructed ray class units whose relative norms, as well as all positive integral powers of these norms, generate every finite abelian intermediate field over K K K . He also observed that this norm-generation property can fail for simpler invariants [17 pp. 132–133] . Schertz [18 Conjecture 6.8.3] formulated related conjectures on the generation of intermediate fields by relative norms of simpler elliptic invariants. In this paper, however, we obtain the same norm-generation property by subtracting an explicit integer from a Weber value (Theorem 8.4). This gives the following explicit formulation of the solution to Hilbert’s twelfth problem for imaginary quadratic fields.
Theorem B (Theorem 9.4). Let K K K be an imaginary quadratic field of discriminant d K d_K d K , and let L L L be a finite extension of K K K . Then L / K L/K L / K is abelian if and only if there exist a positive integer N N N divisible by 4 4 4 and a subgroup H H H of the extended form class group C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) such that
L = K ( ∏ [ Q ] Γ 1 ( N ) ∈ H ( f v ( ∣ O K × ∣ / 2 ) ( [ 0 a Q ′ 1 N ] ( − ω ‾ Q ) − C N N O K ) ) ) . L=K\left(\prod_{[Q]_{\Gamma_1(N)}\in H}\left(f_{\mathfrak v}^{(|\mathcal{O}_K^\times|/2)}\left(\left[\begin{smallmatrix}0&a'_Q\\1&N\end{smallmatrix}\right](-\overline{\omega}_Q)-C_NN\mathcal{O}_K\right)\right)\right). L = K [ Q ] Γ 1 ( N ) ∈ H ∏ ( f v ( ∣ O K × ∣/2 ) ( [ 0 1 a Q ′ N ] ( − ω Q ) − C N N O K ) ) . Here, ω Q ∈ H \omega_Q\in\mathbb{H} ω Q ∈ H is the zero of Q ( x , 1 ) Q(x,1) Q ( x , 1 ) , and a Q ′ a'_Q a Q ′ is an inverse modulo N N N of the leading coefficient of Q Q Q . The functions f v ( k ) f_{\mathfrak v}^{(k)} f v ( k ) , the integers C m C_m C m and the extended form class groups C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) are defined in § 3 , § 8 and § 9 .
Throughout this paper, we use the following notation unless otherwise specified:
K K K : an imaginary quadratic field of discriminant d K d_K d K .
O K \mathcal{O}_K O K : the ring of integers of K K K .
m \mathfrak m m : a nonzero proper ideal of O K \mathcal{O}_K O K .
p \mathfrak p p : a nonzero prime ideal of O K \mathcal{O}_K O K .
E E E : the elliptic curve defined by the Weierstrass model in (1) with the parametrization ϕ E : C / O K ⟶ ∼ E ( C ) \phi_E:\mathbb{C}/\mathcal{O}_K\overset{\sim}{\longrightarrow}E(\mathbb{C}) ϕ E : C / O K ⟶ ∼ E ( C ) in (2) .
Furthermore, we write Galois actions on the right, denoting the action of σ \sigma σ on a a a by a σ a^\sigma a σ .
Remaining families of moduli
Following Morton [16 Section 1] , we reduce the proof of Theorem A to seven families of moduli not covered by the earlier results. For the sake of completeness, we give a proof of the reduction.
Let h K h_K h K denote the class number of K K K . By the results of Heegner [6 ] , Baker [1 , 2 , 3 ] and Stark [24 ] , we have
h K = 1 ⟺ d K ∈ { − 3 , − 4 , − 7 , − 8 , − 11 , − 19 , − 43 , − 67 , − 163 } (6) h_K=1\Longleftrightarrow d_K\in\{-3,-4,-7,-8,-11,-19,-43,-67,-163\} \tag*{(6)} h K = 1 ⟺ d K ∈ { − 3 , − 4 , − 7 , − 8 , − 11 , − 19 , − 43 , − 67 , − 163 } (6)
(see also [5 Theorem 12.34] ). For a positive integer N N N , let ζ N = e 2 π i / N \zeta_N = e^{2\pi i/N} ζ N = e 2 π i / N . We note that
O K = [ τ K , 1 ] where τ K = { d K 2 if d K ≡ 0 ( m o d 4 ) , − 1 + d K 2 if d K ≡ 1 ( m o d 4 ) . (7) \mathcal{O}_K = [\tau_K,1] \quad\text{where } \tau_K =
\begin{cases}
\frac{\sqrt{d_K}}{2} & \text{if } d_K \equiv0 \pmod{4},\\
\frac{-1+\sqrt{d_K}}{2} & \text{if } d_K \equiv1 \pmod{4}.
\end{cases} \tag*{(7)} O K = [ τ K , 1 ] where τ K = { 2 d K 2 − 1 + d K if d K ≡ 0 ( mod 4 ) , if d K ≡ 1 ( mod 4 ) . (7)
and
O K × = { ⟨ ζ 4 ⟩ if K = Q ( − 1 ) (equivalently, d K = − 4 ) , ⟨ ζ 6 ⟩ if K = Q ( − 3 ) (equivalently, d K = − 3 ) , ⟨ ζ 2 ⟩ otherwise . (8) \mathcal{O}_K^\times=
\begin{cases}
\langle\zeta_4\rangle& \text{if } K = \mathbb{Q}(\sqrt{-1}) \text{ (equivalently, } d_K=-4\text{)},\\
\langle\zeta_6\rangle& \text{if } K = \mathbb{Q}(\sqrt{-3}) \text{ (equivalently, } d_K=-3\text{)},\\
\langle\zeta_2\rangle& \text{otherwise}.
\end{cases} \tag*{(8)} O K × = ⎩ ⎨ ⎧ ⟨ ζ 4 ⟩ ⟨ ζ 6 ⟩ ⟨ ζ 2 ⟩ if K = Q ( − 1 ) (equivalently, d K = − 4 ) , if K = Q ( − 3 ) (equivalently, d K = − 3 ) , otherwise . (8)
(see [5 (5.13) and p. 92] ).
For a nonzero ideal m \mathfrak{m} m of O K \mathcal{O}_K O K , the ray class group C ( m ) C(\mathfrak{m}) C ( m ) is defined by
C ( m ) = I K ( m ) / P K , 1 ( m ) , C(\mathfrak{m}) = I_K(\mathfrak{m})/P_{K,1}(\mathfrak{m}), C ( m ) = I K ( m ) / P K , 1 ( m ) ,
where I K ( m ) I_K(\mathfrak{m}) I K ( m ) is the group of fractional ideals of K K K relatively prime to m \mathfrak{m} m and P K , 1 ( m ) P_{K,1}(\mathfrak{m}) P K , 1 ( m ) is its subgroup given by
P K , 1 ( m ) = ⟨ ν O K ∣ ν ∈ O K ∖ { 0 } , ν ≡ 1 ( m o d m ) ⟩ . P_{K,1}(\mathfrak{m}) = \langle\nu\mathcal{O}_K \mid\nu\in\mathcal{O}_K \setminus\{0\},\ \nu\equiv1 \pmod{\mathfrak{m}}\rangle. P K , 1 ( m ) = ⟨ ν O K ∣ ν ∈ O K ∖ { 0 } , ν ≡ 1 ( mod m )⟩ .
Let K m K_{\mathfrak{m}} K m denote the corresponding ray class field so that C ( m ) ≃ Gal ( K m / K ) C(\mathfrak{m}) \simeq\operatorname{Gal}(K_{\mathfrak{m}}/K) C ( m ) ≃ Gal ( K m / K ) via the Artin map modulo m \mathfrak{m} m . Note that if a \mathfrak{a} a and b \mathfrak{b} b are nonzero ideals of O K \mathcal{O}_K O K such that a ∣ b \mathfrak{a} \mid\mathfrak{b} a ∣ b , then K a ⊆ K b K_{\mathfrak{a}} \subseteq K_{\mathfrak{b}} K a ⊆ K b (see [7 Theorem 9.9 in Chapter V] ). The conductor of the extension K m / K K_{\mathfrak{m}}/K K m / K is defined as the smallest (with respect to divisibility) nonzero ideal f = f K m / K \mathfrak{f}=\mathfrak{f}_{K_{\mathfrak{m}}/K} f = f K m / K of O K \mathcal{O}_K O K such that K m ⊆ K f K_{\mathfrak{m}}\subseteq K_{\mathfrak{f}} K m ⊆ K f , that is, there is no nonzero ideal a \mathfrak{a} a of O K \mathcal{O}_K O K such that
a ∣ f , a ≠ f , K m ⊆ K a . \mathfrak{a}\mid\mathfrak{f},\qquad\mathfrak{a}\ne\mathfrak{f},\qquad K_{\mathfrak{m}}\subseteq K_{\mathfrak{a}}. a ∣ f , a = f , K m ⊆ K a .
For a rational prime p p p , we write the prime ideal factorization of p O K p\mathcal{O}_K p O K as follows:
p O K = { p p , s p l i t p p , s p l i t ′ if p splits in K , p p , i n e r t if p is inert in K , p p , r a m i f i e d 2 if p ramifies in K . p\mathcal{O}_K =
\begin{cases}
\mathfrak{p}_{p,\mathrm{split}}\mathfrak{p}'_{p,\mathrm{split}} & \text{if } p \text{ splits in } K,\\
\mathfrak{p}_{p,\mathrm{inert}} & \text{if } p \text{ is inert in } K,\\
\mathfrak{p}_{p,\mathrm{ramified}}^2 & \text{if } p \text{ ramifies in } K.
\end{cases} p O K = ⎩ ⎨ ⎧ p p , split p p , split ′ p p , inert p p , ramified 2 if p splits in K , if p is inert in K , if p ramifies in K .
When p p p is not inert in K K K , we write p p , n o n i n e r t \mathfrak{p}_{p,\mathrm{noninert}} p p , noninert for a prime ideal of O K \mathcal{O}_K O K lying above p p p .
Proposition 2.1. Let K K K be an imaginary quadratic field such that h K > 1 h_K>1 h K > 1 , and let m \mathfrak{m} m be a nonzero proper ideal of O K \mathcal{O}_K O K such that f K m / K = m \mathfrak{f}_{K_{\mathfrak{m}}/K}=\mathfrak{m} f K m / K = m . Assume that m \mathfrak{m} m satisfies none of (S1), (S2), (S3) and that
m ≠ N O K for every integer N > 1. (9) \mathfrak{m}\ne N\mathcal{O}_K \quad\text{for every integer } N>1. \tag*{(9)} m = N O K for every integer N > 1. (9) Then, up to interchanging conjugate prime ideals, m \mathfrak{m} m has one of the following seven forms:
p 2 , s p l i t 2 p 3 , n o n i n e r t , p 3 , s p l i t 2 , p 3 , s p l i t p 2 , s p l i t 2 , p 3 , s p l i t p 3 , s p l i t ′ , \mathfrak{p}_{2,\mathrm{split}}^2\mathfrak{p}_{3,\mathrm{noninert}},\qquad
\mathfrak{p}_{3,\mathrm{split}}^2,\qquad
\mathfrak{p}_{3,\mathrm{split}}\mathfrak{p}_{2,\mathrm{split}}^2,\qquad
\mathfrak{p}_{3,\mathrm{split}}\mathfrak{p}'_{3,\mathrm{split}}, p 2 , split 2 p 3 , noninert , p 3 , split 2 , p 3 , split p 2 , split 2 , p 3 , split p 3 , split ′ , p 5 , n o n i n e r t , p 3 , n o n i n e r t p 5 , n o n i n e r t , p 3 , s p l i t p 3 , s p l i t ′ p 5 , n o n i n e r t . (10) \mathfrak{p}_{5,\mathrm{noninert}},\qquad
\mathfrak{p}_{3,\mathrm{noninert}}\mathfrak{p}_{5,\mathrm{noninert}},\qquad
\mathfrak{p}_{3,\mathrm{split}}\mathfrak{p}'_{3,\mathrm{split}}\mathfrak{p}_{5,\mathrm{noninert}}. \tag*{(10)} p 5 , noninert , p 3 , noninert p 5 , noninert , p 3 , split p 3 , split ′ p 5 , noninert . (10)
Proof. Since h K > 1 h_K>1 h K > 1 , we have K ≠ Q ( − 1 ) , Q ( − 3 ) K\ne\mathbb{Q}(\sqrt{-1}),\mathbb{Q}(\sqrt{-3}) K = Q ( − 1 ) , Q ( − 3 ) by (6), and hence O K × = { ± 1 } O_K^\times=\{\pm1\} O K × = { ± 1 } by (8). For a nonzero ideal a \mathfrak{a} a of O K O_K O K , let
U 1 ( a ) : = { u ∈ O K × ∣ u ≡ 1 ( m o d a ) } , U_1(\mathfrak{a}):=\{u\in O_K^\times\mid u\equiv1\pmod{\mathfrak{a}}\}, U 1 ( a ) := { u ∈ O K × ∣ u ≡ 1 ( mod a )} ,
and then
∣ U 1 ( a ) ∣ = { 2 if a ∣ 2 O K , 1 if a ∤ 2 O K . (11) \lvert U_1(\mathfrak{a})\rvert=
\begin{cases}
2 & \text{if }\mathfrak{a}\mid2O_K,\\
1 & \text{if }\mathfrak{a}\nmid2O_K.
\end{cases} \tag*{(11)} ∣ U 1 ( a )∣ = { 2 1 if a ∣ 2 O K , if a ∤ 2 O K . (11)
The exact sequence
1 ⟶ U 1 ( a ) ⟶ O K × ⟶ ( O K / a ) × ⟶ C ( a ) ⟶ C ( O K ) ⟶ 1 (12) 1\longrightarrow U_1(\mathfrak{a})\longrightarrow O_K^\times\longrightarrow(O_K/\mathfrak{a})^\times\longrightarrow C(\mathfrak{a})\longrightarrow C(O_K)\longrightarrow1 \tag*{(12)} 1 ⟶ U 1 ( a ) ⟶ O K × ⟶ ( O K / a ) × ⟶ C ( a ) ⟶ C ( O K ) ⟶ 1 (12)
gives the extension degree formula
[ K a : H K ] = φ ( a ) ∣ U 1 ( a ) ∣ 2 [K_{\mathfrak{a}}:H_K]=\frac{\varphi(\mathfrak{a})\lvert U_1(\mathfrak{a})\rvert}{2} [ K a : H K ] = 2 φ ( a ) ∣ U 1 ( a )∣
(see [15 Theorem 1.7 in Chapter V] ). It follows that if p ∣ a \mathfrak{p}\mid\mathfrak{a} p ∣ a and p e ∥ a \mathfrak{p}^e\Vert\mathfrak{a} p e ∥ a , then
[ K a : K a p − 1 ] = λ p ( e ) [ U 1 ( a p − 1 ) : U 1 ( a ) ] [K_{\mathfrak{a}}:K_{\mathfrak{a}\mathfrak{p}^{-1}}]=\frac{\lambda_{\mathfrak{p}}(e)}{[U_1(\mathfrak{a}\mathfrak{p}^{-1}):U_1(\mathfrak{a})]} [ K a : K a p − 1 ] = [ U 1 ( a p − 1 ) : U 1 ( a )] λ p ( e )
where λ p ( e ) : = { N ( p ) − 1 if e = 1 , N ( p ) if e ≥ 2. \lambda_{\mathfrak{p}}(e):= \begin{cases} N(\mathfrak{p})-1 & \text{if }e=1,\\ N(\mathfrak{p}) & \text{if }e\ge2. \end{cases} λ p ( e ) := { N ( p ) − 1 N ( p ) if e = 1 , if e ≥ 2. {#eq:13}
Thus the hypothesis f K m / K = m f_{K_{\mathfrak{m}}/K}=\mathfrak{m} f K m / K = m is equivalent to
λ p ( e p ) [ U 1 ( m p − 1 ) : U 1 ( m ) ] = [ K m : K m p − 1 ] > 1 (13) \frac{\lambda_{\mathfrak{p}}(e_{\mathfrak{p}})}{[U_1(\mathfrak{m}\mathfrak{p}^{-1}):U_1(\mathfrak{m})]}=[K_{\mathfrak{m}}:K_{\mathfrak{m}\mathfrak{p}^{-1}}]>1 \tag*{(13)} [ U 1 ( m p − 1 ) : U 1 ( m )] λ p ( e p ) = [ K m : K m p − 1 ] > 1 (13)
for every p ∣ m \mathfrak{p}\mid\mathfrak{m} p ∣ m such that p e p ∥ m \mathfrak{p}^{e_{\mathfrak{p}}}\Vert\mathfrak{m} p e p ∥ m .
There are three possible cases:
{ gcd ( m , 2 O K ) ≠ O K and 2 is not inert in K , gcd ( m , 2 O K ) ≠ O K and 2 is inert in K , gcd ( m , 2 O K ) = O K . \begin{cases}
\gcd(\mathfrak{m},2O_K)\ne O_K\text{ and }2\text{ is not inert in }K,\\
\gcd(\mathfrak{m},2O_K)\ne O_K\text{ and }2\text{ is inert in }K,\\
\gcd(\mathfrak{m},2O_K)=O_K.
\end{cases} ⎩ ⎨ ⎧ g cd( m , 2 O K ) = O K and 2 is not inert in K , g cd( m , 2 O K ) = O K and 2 is inert in K , g cd( m , 2 O K ) = O K .
Case 1. Suppose that gcd ( m , 2 O K ) ≠ O K \gcd(\mathfrak{m},2O_K)\ne O_K g cd( m , 2 O K ) = O K and 2 2 2 is not inert in K K K . Let p ∣ gcd ( m , 2 O K ) \mathfrak{p}\mid\gcd(\mathfrak{m},2O_K) p ∣ g cd( m , 2 O K ) , and write m = p e b \mathfrak{m}=\mathfrak{p}^e\mathfrak{b} m = p e b for an ideal b \mathfrak{b} b of O K O_K O K such that gcd ( b , p ) = O K \gcd(\mathfrak{b},\mathfrak{p})=O_K g cd( b , p ) = O K . Since N ( p ) = 2 N(\mathfrak{p})=2 N ( p ) = 2 and f K m / K = m f_{K_{\mathfrak{m}}/K}=\mathfrak{m} f K m / K = m , we deduce by (14) that e ≥ 2 e\ge2 e ≥ 2 . Since (S2) fails, we must have
6 > φ ( m ) = 2 e − 1 φ ( b ) . (14) 6>\varphi(\mathfrak{m})=2^{e-1}\varphi(\mathfrak{b}). \tag*{(14)} 6 > φ ( m ) = 2 e − 1 φ ( b ) . (14)
Hence we get e ∈ { 2 , 3 } e\in\{2,3\} e ∈ { 2 , 3 } .
Case 1-1. Let e = 2 e=2 e = 2 , so φ ( b ) < 3 \varphi(\mathfrak{b})<3 φ ( b ) < 3 . The possibilities for b \mathfrak{b} b are
b = { O K , p 2 , s p l i t ′ , p 2 , s p l i t ′ 2 , p 2 , s p l i t ′ 2 p 3 , n o n i n e r t , p 3 , n o n i n e r t if p = p 2 , s p l i t , O K , p 3 , n o n i n e r t if p = p 2 , r a m i f i e d . \mathfrak{b}=
\begin{cases}
O_K,\ \mathfrak{p}'_{2,\mathrm{split}},\ \mathfrak{p}_{2,\mathrm{split}}'^{\,2},\ \mathfrak{p}_{2,\mathrm{split}}'^{\,2}\mathfrak{p}_{3,\mathrm{noninert}},\ \mathfrak{p}_{3,\mathrm{noninert}} & \text{if }\mathfrak{p}=\mathfrak{p}_{2,\mathrm{split}},\\
O_K,\ \mathfrak{p}_{3,\mathrm{noninert}} & \text{if }\mathfrak{p}=\mathfrak{p}_{2,\mathrm{ramified}}.
\end{cases} b = { O K , p 2 , split ′ , p 2 , split ′ 2 , p 2 , split ′ 2 p 3 , noninert , p 3 , noninert O K , p 3 , noninert if p = p 2 , split , if p = p 2 , ramified .
Furthermore, using (9) and (14), we find that
m = p 2 , s p l i t 2 p 3 , n o n i n e r t . (15) \mathfrak{m}=\mathfrak{p}_{2,\mathrm{split}}^{\,2}\mathfrak{p}_{3,\mathrm{noninert}}. \tag*{(15)} m = p 2 , split 2 p 3 , noninert . (15)
Case 1-2. Let e = 3 e=3 e = 3 . Then (15) yields φ ( b ) < 3 2 \varphi(\mathfrak b)<\frac{3}{2} φ ( b ) < 2 3 , and hence
b = { O K , p 2 , s p l i t ′ if p = p 2 , s p l i t , O K if p = p 2 , r a m i f i e d . \mathfrak b=
\begin{cases}
\mathcal O_K, & \mathfrak p'_{2,\mathrm{split}}\quad\text{if }\mathfrak p=\mathfrak p_{2,\mathrm{split}},\\
\mathcal O_K & \text{if }\mathfrak p=\mathfrak p_{2,\mathrm{ramified}}.
\end{cases} b = { O K , O K p 2 , split ′ if p = p 2 , split , if p = p 2 , ramified .
Applying (14), we obtain
m = p 2 , s p l i t 3 . (16) \mathfrak m=\mathfrak p_{2,\mathrm{split}}^3. \tag*{(16)} m = p 2 , split 3 . (16)
Case 2. Suppose that gcd ( m , 2 O K ) ≠ O K \gcd(\mathfrak m,2\mathcal O_K)\ne\mathcal O_K g cd( m , 2 O K ) = O K and 2 2 2 is inert in K K K . Let p = p 2 , i n e r t \mathfrak p=\mathfrak p_{2,\mathrm{inert}} p = p 2 , inert and so N ( p ) = 4 N(\mathfrak p)=4 N ( p ) = 4 . Write m = p e b \mathfrak m=\mathfrak p^e\mathfrak b m = p e b for an ideal b \mathfrak b b of O K \mathcal O_K O K such that gcd ( b , p ) = O K \gcd(\mathfrak b,\mathfrak p)=\mathcal O_K g cd( b , p ) = O K . If e ≥ 2 e\ge2 e ≥ 2 , then 4 O K ∣ m 4\mathcal O_K\mid\mathfrak m 4 O K ∣ m and then m = 4 O K \mathfrak m=4\mathcal O_K m = 4 O K since (S1) fails. But this contradicts (9), and hence e = 1 e=1 e = 1 . Since (S3) fails, we have
5 > Ψ ( m ) = Ψ ( p b ) = 2 Ψ ( b ) . (17) 5>\Psi(\mathfrak m)=\Psi(\mathfrak p\mathfrak b)=2\Psi(\mathfrak b). \tag*{(17)} 5 > Ψ ( m ) = Ψ ( pb ) = 2Ψ ( b ) . (17)
For a prime ideal q \mathfrak q q of O K \mathcal O_K O K of odd norm and an integer r ≥ 1 r\ge1 r ≥ 1 , we see that
Ψ ( q r ) = N ( q ) r − 1 ( N ( q ) − 2 ) and N ( q ) ≥ 3. \Psi(\mathfrak q^r)=N(\mathfrak q)^{r-1}(N(\mathfrak q)-2)
\quad\text{and}\quad N(\mathfrak q)\ge3. Ψ ( q r ) = N ( q ) r − 1 ( N ( q ) − 2 ) and N ( q ) ≥ 3.
And it follows from (18) that
b = O K , p 3 , n o n i n e r t , p 3 , s p l i t p 3 , s p l i t ′ , \mathfrak b=\mathcal O_K,\quad\mathfrak p_{3,\mathrm{noninert}},\quad\mathfrak p_{3,\mathrm{split}}\mathfrak p'_{3,\mathrm{split}}, b = O K , p 3 , noninert , p 3 , split p 3 , split ′ ,
and the assumption (9) forces us to get
m = p 2 , i n e r t p 3 , n o n i n e r t . \mathfrak m=\mathfrak p_{2,\mathrm{inert}}\mathfrak p_{3,\mathrm{noninert}}. m = p 2 , inert p 3 , noninert .
However, we observe by (11) that
λ p 3 , n o n i n e r t ( 1 ) [ U 1 ( p 2 , i n e r t ) : U 1 ( m ) ] = 3 − 1 2 = 1 , \frac{\lambda_{\mathfrak p_{3,\mathrm{noninert}}}(1)}
{[U_1(\mathfrak p_{2,\mathrm{inert}}):U_1(\mathfrak m)]}
=\frac{3-1}{2}=1, [ U 1 ( p 2 , inert ) : U 1 ( m )] λ p 3 , noninert ( 1 ) = 2 3 − 1 = 1 ,
contradicting (14). Thus Case 2 yields no remaining form.
Case 3. Suppose that gcd ( m , 2 O K ) = O K \gcd(\mathfrak m,2\mathcal O_K)=\mathcal O_K g cd( m , 2 O K ) = O K . Let m = ∏ p ∣ m p e p \mathfrak m=\prod_{\mathfrak p\mid\mathfrak m}\mathfrak p^{e_{\mathfrak p}} m = ∏ p ∣ m p e p . Since (S3) fails, we attain
5 > Ψ ( m ) = ∏ p ∣ m Ψ ( p e p ) = ∏ p ∣ m N ( p ) e p − 1 ( N ( p ) − 2 ) and N ( p ) ≥ 3 ( p ∣ m ) . 5>\Psi(\mathfrak m)=\prod_{\mathfrak p\mid\mathfrak m}\Psi(\mathfrak p^{e_{\mathfrak p}})
=\prod_{\mathfrak p\mid\mathfrak m}N(\mathfrak p)^{e_{\mathfrak p}-1}(N(\mathfrak p)-2)
\quad\text{and}\quad N(\mathfrak p)\ge3\;(\mathfrak p\mid\mathfrak m). 5 > Ψ ( m ) = p ∣ m ∏ Ψ ( p e p ) = p ∣ m ∏ N ( p ) e p − 1 ( N ( p ) − 2 ) and N ( p ) ≥ 3 ( p ∣ m ) .
Therefore we find that
m = { p 3 , n o n i n e r t , p 3 , n o n i n e r t 2 , p 3 , s p l i t p 3 , s p l i t ′ , p 3 , s p l i t 2 p 3 , s p l i t ′ , p 5 , n o n i n e r t , p 3 , n o n i n e r t p 5 , n o n i n e r t , p 3 , s p l i t p 3 , s p l i t ′ p 5 , n o n i n e r t . \mathfrak m=
\begin{cases}
\mathfrak p_{3,\mathrm{noninert}},\quad
\mathfrak p_{3,\mathrm{noninert}}^2,\quad
\mathfrak p_{3,\mathrm{split}}\mathfrak p'_{3,\mathrm{split}},\quad
\mathfrak p_{3,\mathrm{split}}^2\mathfrak p'_{3,\mathrm{split}},\\
\mathfrak p_{5,\mathrm{noninert}},\quad
\mathfrak p_{3,\mathrm{noninert}}\mathfrak p_{5,\mathrm{noninert}},\quad
\mathfrak p_{3,\mathrm{split}}\mathfrak p'_{3,\mathrm{split}}\mathfrak p_{5,\mathrm{noninert}}.
\end{cases} m = { p 3 , noninert , p 3 , noninert 2 , p 3 , split p 3 , split ′ , p 3 , split 2 p 3 , split ′ , p 5 , noninert , p 3 , noninert p 5 , noninert , p 3 , split p 3 , split ′ p 5 , noninert .
Applying (9) and (14), we derive the following five forms:
m = p 3 , s p l i t 2 , p 3 , s p l i t 2 p 3 , s p l i t ′ , p 5 , n o n i n e r t , p 3 , n o n i n e r t p 5 , n o n i n e r t , p 3 , s p l i t p 3 , s p l i t ′ p 5 , n o n i n e r t . (18) \mathfrak m=\mathfrak p_{3,\mathrm{split}}^2,\quad
\mathfrak p_{3,\mathrm{split}}^2\mathfrak p'_{3,\mathrm{split}},\quad
\mathfrak p_{5,\mathrm{noninert}},\quad
\mathfrak p_{3,\mathrm{noninert}}\mathfrak p_{5,\mathrm{noninert}},\quad
\mathfrak p_{3,\mathrm{split}}\mathfrak p'_{3,\mathrm{split}}\mathfrak p_{5,\mathrm{noninert}}. \tag*{(18)} m = p 3 , split 2 , p 3 , split 2 p 3 , split ′ , p 5 , noninert , p 3 , noninert p 5 , noninert , p 3 , split p 3 , split ′ p 5 , noninert . (18)
Combining (16), (17) and (19) completes the proof.
Fields of modular functions
We recall the modular functions in order to express Weber values and the Galois action on them. Their Fourier expansions and the identity relating differences of Fricke functions to Siegel functions provide the analytic tools for the proofs below.
Let H = { τ ∈ C ∣ Im ( τ ) > 0 } \mathbb{H}=\{\tau\in\mathbb{C}\mid\operatorname{Im}(\tau)>0\} H = { τ ∈ C ∣ Im ( τ ) > 0 } be the complex upper half-plane. For z 1 , z 2 ∈ C z_1,z_2\in\mathbb{C} z 1 , z 2 ∈ C such that z 1 / z 2 ∈ H z_1/z_2\in\mathbb{H} z 1 / z 2 ∈ H , we denote by [ z 1 , z 2 ] [z_1,z_2] [ z 1 , z 2 ] the lattice Z z 1 + Z z 2 \mathbb{Z}z_1+\mathbb{Z}z_2 Z z 1 + Z z 2 . Define the functions g 2 g_2 g 2 , g 3 g_3 g 3 and Δ \Delta Δ on H \mathbb{H} H by
g 2 ( τ ) = g 2 ( [ τ , 1 ] ) , g 3 ( τ ) = g 3 ( [ τ , 1 ] ) , Δ ( τ ) = Δ ( [ τ , 1 ] ) = g 2 ( τ ) 3 − 27 g 3 ( τ ) 2 ( τ ∈ H ) , (19) g_2(\tau)=g_2([\tau,1]),\qquad g_3(\tau)=g_3([\tau,1]),\qquad\Delta(\tau)=\Delta([\tau,1])=g_2(\tau)^3-27g_3(\tau)^2\qquad(\tau\in\mathbb{H}), \tag*{(19)} g 2 ( τ ) = g 2 ([ τ , 1 ]) , g 3 ( τ ) = g 3 ([ τ , 1 ]) , Δ ( τ ) = Δ ([ τ , 1 ]) = g 2 ( τ ) 3 − 27 g 3 ( τ ) 2 ( τ ∈ H ) , (19)
which are modular forms for SL 2 ( Z ) \operatorname{SL}_2(\mathbb{Z}) SL 2 ( Z ) of weight 4, 6, 12, respectively (see [13 pp. 36–38] ). Put
q τ = e 2 π i τ ( τ ∈ H ) . q_\tau=e^{2\pi i\tau}\qquad(\tau\in\mathbb{H}). q τ = e 2 π i τ ( τ ∈ H ) .
Lemma 3.1. The functions g 2 g_2 g 2 , g 3 g_3 g 3 , Δ \Delta Δ satisfy the following expansion formulas:
12 ( 2 π i ) 4 g 2 ( τ ) = 1 + 240 ∑ n = 1 ∞ n 3 q τ n 1 − q τ n ( τ ∈ H ) , (20) \frac{12}{(2\pi i)^4}g_2(\tau)=1+240\sum_{n=1}^{\infty}\frac{n^3q_\tau^n}{1-q_\tau^n}\qquad(\tau\in\mathbb{H}), \tag*{(20)} ( 2 π i ) 4 12 g 2 ( τ ) = 1 + 240 n = 1 ∑ ∞ 1 − q τ n n 3 q τ n ( τ ∈ H ) , (20) 6 3 ( 2 π i ) 6 g 3 ( τ ) = − 1 + 504 ∑ n = 1 ∞ n 5 q τ n 1 − q τ n ( τ ∈ H ) , (21) \frac{6^3}{(2\pi i)^6}g_3(\tau)=-1+504\sum_{n=1}^{\infty}\frac{n^5q_\tau^n}{1-q_\tau^n}\qquad(\tau\in\mathbb{H}), \tag*{(21)} ( 2 π i ) 6 6 3 g 3 ( τ ) = − 1 + 504 n = 1 ∑ ∞ 1 − q τ n n 5 q τ n ( τ ∈ H ) , (21) 1 ( 2 π i ) 12 Δ ( τ ) = q τ ∏ n = 1 ∞ ( 1 − q τ n ) 24 ( τ ∈ H ) . (22) \frac{1}{(2\pi i)^{12}}\Delta(\tau)=q_\tau\prod_{n=1}^{\infty}(1-q_\tau^n)^{24}\qquad(\tau\in\mathbb{H}). \tag*{(22)} ( 2 π i ) 12 1 Δ ( τ ) = q τ n = 1 ∏ ∞ ( 1 − q τ n ) 24 ( τ ∈ H ) . (22)
Proof. See [13 Proposition 4 in Chapter 4 and Theorem 5 in Chapter 18] .
For later use, we also introduce the well-known expansion formula of the Weierstrass function. For z ∈ C z\in\mathbb{C} z ∈ C , we also use the notation q z = e 2 π i z q_z=e^{2\pi iz} q z = e 2 π i z .
Lemma 3.2. We have
1 ( 2 π i ) 2 ℘ ( z ; [ τ , 1 ] ) = 1 12 + q z ( 1 − q z ) 2 + ∑ n = 1 ∞ ( q τ n q z ( 1 − q τ n q z ) 2 + q τ n q z − 1 ( 1 − q τ n q z − 1 ) 2 ) − 2 ∑ n = 1 ∞ n q τ n 1 − q τ n ( z ∈ C , τ ∈ H ) . \begin{aligned}
\frac{1}{(2\pi i)^2}\wp(z;[\tau,1])
={}&\frac{1}{12}+\frac{q_z}{(1-q_z)^2}\\
&+\sum_{n=1}^{\infty}\left(\frac{q_\tau^nq_z}{(1-q_\tau^nq_z)^2}+\frac{q_\tau^nq_z^{-1}}{(1-q_\tau^nq_z^{-1})^2}\right)-2\sum_{n=1}^{\infty}\frac{nq_\tau^n}{1-q_\tau^n}\qquad(z\in\mathbb{C},\ \tau\in\mathbb{H}).
\end{aligned} ( 2 π i ) 2 1 ℘ ( z ; [ τ , 1 ]) = 12 1 + ( 1 − q z ) 2 q z + n = 1 ∑ ∞ ( ( 1 − q τ n q z ) 2 q τ n q z + ( 1 − q τ n q z − 1 ) 2 q τ n q z − 1 ) − 2 n = 1 ∑ ∞ 1 − q τ n n q τ n ( z ∈ C , τ ∈ H ) .
Proof. See [13 Proposition 3 in Chapter 4] .
The elliptic modular function j j j on H \mathbb{H} H is defined by
j ( τ ) = 1728 g 2 ( τ ) 3 Δ ( τ ) = 1728 + 46656 g 3 ( τ ) 2 Δ ( τ ) ( τ ∈ H ) , (23) j(\tau)=1728\frac{g_2(\tau)^3}{\Delta(\tau)}=1728+46656\frac{g_3(\tau)^2}{\Delta(\tau)}\qquad(\tau\in\mathbb{H}), \tag*{(23)} j ( τ ) = 1728 Δ ( τ ) g 2 ( τ ) 3 = 1728 + 46656 Δ ( τ ) g 3 ( τ ) 2 ( τ ∈ H ) , (23)
which is a weakly holomorphic modular function for SL 2 ( Z ) \operatorname{SL}_2(\mathbb{Z}) SL 2 ( Z ) ([13 p. 39] ). Furthermore, its Fourier expansion with respect to q τ q_\tau q τ is given by
j ( τ ) = q τ − 1 + 744 + 196884 q τ + 21493760 q τ 2 + ⋯ ∈ q τ − 1 Z [ [ q τ ] ] (24) j(\tau)=q_\tau^{-1}+744+196884q_\tau+21493760q_\tau^2+\cdots\in q_\tau^{-1}\mathbb{Z}[[q_\tau]] \tag*{(24)} j ( τ ) = q τ − 1 + 744 + 196884 q τ + 21493760 q τ 2 + ⋯ ∈ q τ − 1 Z [[ q τ ]] (24)
with integer Fourier coefficients (see [5 Theorem 11.8] ).
For k ∈ { 1 , 2 , 3 } k \in\{1,2,3\} k ∈ { 1 , 2 , 3 } and v = [ v 1 v 2 ] ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) \mathbf{v}=[v_1\ \ v_2]\in M_{1,2}(\mathbb{Q})\setminus M_{1,2}(\mathbb{Z}) v = [ v 1 v 2 ] ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) , the Fricke function f v ( k ) f_{\mathbf{v}}^{(k)} f v ( k ) on H \mathbb{H} H is defined by
f v ( k ) ( τ ) = { g 2 ( τ ) g 3 ( τ ) Δ ( τ ) ℘ ( v 1 τ + v 2 ; [ τ , 1 ] ) if k = 1 , g 2 ( τ ) 2 Δ ( τ ) ℘ ( v 1 τ + v 2 ; [ τ , 1 ] ) 2 if k = 2 , g 3 ( τ ) Δ ( τ ) ℘ ( v 1 τ + v 2 ; [ τ , 1 ] ) 3 if k = 3 ( τ ∈ H ) . (25) f_{\mathbf{v}}^{(k)}(\tau)=
\begin{cases}
\dfrac{g_2(\tau)g_3(\tau)}{\Delta(\tau)}\wp(v_1\tau+v_2;[\tau,1]) & \text{if } k=1,\\
\dfrac{g_2(\tau)^2}{\Delta(\tau)}\wp(v_1\tau+v_2;[\tau,1])^2 & \text{if } k=2,\\
\dfrac{g_3(\tau)}{\Delta(\tau)}\wp(v_1\tau+v_2;[\tau,1])^3 & \text{if } k=3
\end{cases}
\qquad(\tau\in\mathbb{H}). \tag*{(25)} f v ( k ) ( τ ) = ⎩ ⎨ ⎧ Δ ( τ ) g 2 ( τ ) g 3 ( τ ) ℘ ( v 1 τ + v 2 ; [ τ , 1 ]) Δ ( τ ) g 2 ( τ ) 2 ℘ ( v 1 τ + v 2 ; [ τ , 1 ] ) 2 Δ ( τ ) g 3 ( τ ) ℘ ( v 1 τ + v 2 ; [ τ , 1 ] ) 3 if k = 1 , if k = 2 , if k = 3 ( τ ∈ H ) . (25)
Observe that f v ( k ) f_{\mathbf{v}}^{(k)} f v ( k ) is holomorphic on H \mathbb{H} H . We simply write f v f_{\mathbf{v}} f v for f v ( 1 ) f_{\mathbf{v}}^{(1)} f v ( 1 ) . By some basic properties of the Weierstrass ℘ \wp ℘ -function (see [5 Lemma 10.4] ), we have
f u = f v ⟺ u ≡ v or − v ( m o d M 1 , 2 ( Z ) ) . (26) f_{\mathbf{u}}=f_{\mathbf{v}}\Longleftrightarrow\mathbf{u}\equiv\mathbf{v}\text{ or }-\mathbf{v}\pmod{M_{1,2}(\mathbb{Z})}. \tag*{(26)} f u = f v ⟺ u ≡ v or − v ( mod M 1 , 2 ( Z )) . (26)
For a positive integer N N N , consider the field F N \mathcal{F}_N F N defined by
F N = { Q ( j ) if N = 1 , Q ( j , f v ∣ v ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) ) if N ≥ 2. \mathcal{F}_N=
\begin{cases}
\mathbb{Q}(j) & \text{if } N=1,\\
\mathbb{Q}\left(j,f_{\mathbf{v}}\mid\mathbf{v}\in\dfrac{1}{N}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z})\right) & \text{if } N\geq2.
\end{cases} F N = ⎩ ⎨ ⎧ Q ( j ) Q ( j , f v ∣ v ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) ) if N = 1 , if N ≥ 2.
Proposition 3.3. Let N ≥ 2 N\geq2 N ≥ 2 be an integer.
(i) The extension F N / F 1 \mathcal{F}_N/\mathcal{F}_1 F N / F 1 is Galois and
Gal ( F N / F 1 ) ≃ GL 2 ( Z / N Z ) / ⟨ − I 2 ⟩ . \operatorname{Gal}(\mathcal{F}_N/\mathcal{F}_1)\simeq\operatorname{GL}_2(\mathbb{Z}/N\mathbb{Z})/\langle-I_2\rangle. Gal ( F N / F 1 ) ≃ GL 2 ( Z / N Z ) / ⟨ − I 2 ⟩ . (ii) More precisely, if α ∈ GL 2 ( Z / N Z ) / ⟨ − I 2 ⟩ \alpha\in\operatorname{GL}_2(\mathbb{Z}/N\mathbb{Z})/\langle-I_2\rangle α ∈ GL 2 ( Z / N Z ) / ⟨ − I 2 ⟩ , then
f v α = f v α ~ ( v ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) ) . (27) f_{\mathbf{v}}^\alpha=f_{\mathbf{v}\widetilde{\alpha}}\qquad\left(\mathbf{v}\in\dfrac{1}{N}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z})\right). \tag*{(27)} f v α = f v α ( v ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) ) . (27) where α ~ ∈ M 2 ( Z ) \widetilde{\alpha}\in M_2(\mathbb{Z}) α ∈ M 2 ( Z ) is any lift of α \alpha α .
(iii) Specifically, if γ ∈ SL 2 ( Z ) \gamma\in\operatorname{SL}_2(\mathbb{Z}) γ ∈ SL 2 ( Z ) , then
f v ∘ γ = f v [ γ ] = f v γ ( v ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) ) , f_{\mathbf{v}}\circ\gamma=f_{\mathbf{v}^{[\gamma]}}=f_{\mathbf{v}\gamma}\qquad\left(\mathbf{v}\in\dfrac{1}{N}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z})\right), f v ∘ γ = f v [ γ ] = f v γ ( v ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) ) , where [ γ ] [\gamma] [ γ ] is the reduction of γ \gamma γ to GL 2 ( Z / N Z ) / ⟨ − I 2 ⟩ \operatorname{GL}_2(\mathbb{Z}/N\mathbb{Z})/\langle-I_2\rangle GL 2 ( Z / N Z ) / ⟨ − I 2 ⟩ .
Proof. See [20 Theorem 6.6 (1), (2) and Proposition 6.21 (3)] .
We further note that F N \mathcal{F}_N F N is the field of meromorphic modular functions of level N N N whose Fourier expansions with respect to q τ 1 / N q_\tau^{1/N} q τ 1/ N have coefficients in the N N N th cyclotomic field Q ( ζ N ) \mathbb{Q}(\zeta_N) Q ( ζ N ) ; see [20 Proposition 6.9 (1)] .
For v = [ v 1 v 2 ] ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) \mathbf{v}=[v_1\ \ v_2]\in M_{1,2}(\mathbb{Q})\setminus M_{1,2}(\mathbb{Z}) v = [ v 1 v 2 ] ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) , the Siegel function g v g_{\mathbf{v}} g v on H \mathbb{H} H is given by the following infinite product formula
g v ( τ ) = − e π i v 2 ( v 1 − 1 ) q τ 1 2 ( v 1 2 − v 1 + 1 6 ) ( 1 − q τ v 1 e 2 π i v 2 ) ∏ n = 1 ∞ ( 1 − q τ n + v 1 e 2 π i v 2 ) ( 1 − q τ n − v 1 e − 2 π i v 2 ) ( τ ∈ H ) . g_{\mathbf{v}}(\tau)=-e^{\pi i v_2(v_1-1)}q_\tau^{\frac{1}{2}(v_1^2-v_1+\frac{1}{6})}(1-q_\tau^{v_1}e^{2\pi i v_2})\prod_{n=1}^{\infty}(1-q_\tau^{n+v_1}e^{2\pi i v_2})(1-q_\tau^{n-v_1}e^{-2\pi i v_2})\qquad(\tau\in\mathbb{H}). g v ( τ ) = − e π i v 2 ( v 1 − 1 ) q τ 2 1 ( v 1 2 − v 1 + 6 1 ) ( 1 − q τ v 1 e 2 π i v 2 ) n = 1 ∏ ∞ ( 1 − q τ n + v 1 e 2 π i v 2 ) ( 1 − q τ n − v 1 e − 2 π i v 2 ) ( τ ∈ H ) .
For the original definition of g v g_{\mathbf{v}} g v in terms of a Klein form and the Dedekind eta function, one may refer to [12 §2.1] . Note that g v g_{\mathbf{v}} g v has no zeros or poles on H \mathbb{H} H . Furthermore, if v ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{v} \in\frac{1}{N}M_{1,2}(\mathbb{Z}) \setminus M_{1,2}(\mathbb{Z}) v ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) , then g v 12 N g_{\mathbf{v}}^{12N} g v 12 N belongs to F N \mathcal{F}_N F N and satisfies the properties (27) and (28), with f v f_{\mathbf{v}} f v replaced by g v 12 N g_{\mathbf{v}}^{12N} g v 12 N (see [12 Proposition 1.3 in Chapter 2] ).
Lemma 3.4. If u , v ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) \mathbf{u},\mathbf{v} \in M_{1,2}(\mathbb{Q}) \setminus M_{1,2}(\mathbb{Z}) u , v ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) satisfy neither u ≡ v ( m o d M 1 , 2 ( Z ) ) \mathbf{u} \equiv\mathbf{v} \pmod{M_{1,2}(\mathbb{Z})} u ≡ v ( mod M 1 , 2 ( Z )) nor u ≡ − v ( m o d M 1 , 2 ( Z ) ) \mathbf{u} \equiv-\mathbf{v} \pmod{M_{1,2}(\mathbb{Z})} u ≡ − v ( mod M 1 , 2 ( Z )) , then
( f u − f v ) 6 = j 2 ( j − 1728 ) 3 2 30 3 24 g u + v 6 g u − v 6 g u 12 g v 12 . (f_{\mathbf{u}}-f_{\mathbf{v}})^6=\frac{j^2(j-1728)^3}{2^{30}3^{24}}\frac{g_{\mathbf{u}+\mathbf{v}}^6g_{\mathbf{u}-\mathbf{v}}^6}{g_{\mathbf{u}}^{12}g_{\mathbf{v}}^{12}}. ( f u − f v ) 6 = 2 30 3 24 j 2 ( j − 1728 ) 3 g u 12 g v 12 g u + v 6 g u − v 6 .
Proof. See [13 Theorem 2 in Chapter 18] and [12 p. 29 and p. 51] .
Primitive division points and Galois conjugates of their Weber values
We describe primitive division points in terms of their annihilators and relate the Galois conjugates of their Weber values to Fricke invariants. We then express these conjugates as values of Fricke functions at CM points associated with reduced quadratic forms.
For a ∈ O K ≃ End ( E ) a \in\mathcal{O}_K \simeq\operatorname{End}(E) a ∈ O K ≃ End ( E ) and P = ϕ E ( z + O K ) ∈ E ( C ) P=\phi_E(z+\mathcal{O}_K) \in E(\mathbb{C}) P = ϕ E ( z + O K ) ∈ E ( C ) , we simply write a P aP a P for [ a ] E ( P ) = ϕ E ( a z + O K ) [a]_E(P)=\phi_E(az+\mathcal{O}_K) [ a ] E ( P ) = ϕ E ( a z + O K ) .
Lemma 4.1. Write m = ∏ p p e p \mathfrak{m}=\prod_{\mathfrak{p}}\mathfrak{p}^{e_{\mathfrak{p}}} m = ∏ p p e p (e p ≥ 0 e_{\mathfrak{p}}\geq0 e p ≥ 0 ), and let P = ϕ E ( z + O K ) ∈ E ( C ) P=\phi_E(z+\mathcal{O}_K)\in E(\mathbb{C}) P = ϕ E ( z + O K ) ∈ E ( C ) with z ∈ K z\in K z ∈ K .
(i) We have
Ann O K ( P ) = ∏ p p max { 0 , − v p ( z ) } , (28) \operatorname{Ann}_{\mathcal{O}_K}(P)=\prod_{\mathfrak{p}}\mathfrak{p}^{\max\{0,-v_{\mathfrak{p}}(z)\}}, \tag*{(28)} Ann O K ( P ) = p ∏ p m a x { 0 , − v p ( z )} , (28) where v p v_{\mathfrak{p}} v p is the normalized p \mathfrak{p} p -adic valuation on K K K with v p ( 0 ) = ∞ v_{\mathfrak{p}}(0)=\infty v p ( 0 ) = ∞ . Thus we obtain
P is a primitive m -division point ⟺ max { 0 , − v p ( z ) } = e p for all p . P\text{ is a primitive }\mathfrak{m}\text{-division point}\quad\Longleftrightarrow\quad\max\{0,-v_{\mathfrak{p}}(z)\}=e_{\mathfrak{p}}\text{ for all }\mathfrak{p}. P is a primitive m -division point ⟺ max { 0 , − v p ( z )} = e p for all p . (ii) Let a ∈ O K a\in\mathcal{O}_K a ∈ O K . If P P P is a primitive m \mathfrak{m} m -division point, then
a P is a primitive m -division point ⟺ a + m ∈ ( O K / m ) × . aP\text{ is a primitive }\mathfrak{m}\text{-division point}\quad\Longleftrightarrow\quad a+\mathfrak{m}\in(\mathcal{O}_K/\mathfrak{m})^\times. a P is a primitive m -division point ⟺ a + m ∈ ( O K / m ) × .
Proof. (i) We deduce that for a ∈ O K a\in\mathcal{O}_K a ∈ O K
a ∈ ∏ p p max { 0 , − v p ( z ) } ⟺ v p ( a ) ≥ max { 0 , − v p ( z ) } for all p a\in\prod_{\mathfrak{p}}\mathfrak{p}^{\max\{0,-v_{\mathfrak{p}}(z)\}}
\quad\Longleftrightarrow\quad
v_{\mathfrak{p}}(a)\geq\max\{0,-v_{\mathfrak{p}}(z)\}\text{ for all }\mathfrak{p} a ∈ p ∏ p m a x { 0 , − v p ( z )} ⟺ v p ( a ) ≥ max { 0 , − v p ( z )} for all p
⟺ a ∈ O K and a z ∈ O K \Longleftrightarrow\quad a\in\mathcal{O}_K\text{ and }az\in\mathcal{O}_K ⟺ a ∈ O K and a z ∈ O K
⟺ a P = O E \Longleftrightarrow\quad aP=O_E ⟺ a P = O E
⟺ a ∈ Ann O K ( P ) . \Longleftrightarrow\quad a\in\operatorname{Ann}_{\mathcal{O}_K}(P). ⟺ a ∈ Ann O K ( P ) .
This proves (29). The second assertion follows from the definition of a primitive m \mathfrak{m} m -division point.
(ii) Consider the multiplication map
μ a : O K / m → O K / m , b + m ↦ a b + m , \mu_a : \mathcal{O}_K/\mathfrak{m} \to\mathcal{O}_K/\mathfrak{m}, \qquad b+\mathfrak{m} \mapsto ab+\mathfrak{m}, μ a : O K / m → O K / m , b + m ↦ ab + m ,
which is an O K \mathcal{O}_K O K -module homomorphism. We then derive by the assumption Ann O K ( P ) = m \operatorname{Ann}_{\mathcal{O}_K}(P)=\mathfrak{m} Ann O K ( P ) = m that for b ∈ O K b\in\mathcal{O}_K b ∈ O K
b + m ∈ ker μ a ⟺ a b ∈ m ⟺ ( a b ) P = O E ⟺ b ( a P ) = O E ⟺ b ∈ Ann O K ( a P ) , b+\mathfrak{m}\in\ker\mu_a \Longleftrightarrow ab\in\mathfrak{m} \Longleftrightarrow(ab)P=O_E \Longleftrightarrow b(aP)=O_E \Longleftrightarrow b\in\operatorname{Ann}_{\mathcal{O}_K}(aP), b + m ∈ ker μ a ⟺ ab ∈ m ⟺ ( ab ) P = O E ⟺ b ( a P ) = O E ⟺ b ∈ Ann O K ( a P ) ,
and hence
Ann O K ( a P ) ⊇ m and ker μ a = Ann O K ( a P ) / m . (29) \operatorname{Ann}_{\mathcal{O}_K}(aP)\supseteq\mathfrak{m}\quad\text{and}\quad\ker\mu_a=\operatorname{Ann}_{\mathcal{O}_K}(aP)/\mathfrak{m}. \tag*{(29)} Ann O K ( a P ) ⊇ m and ker μ a = Ann O K ( a P ) / m . (29)
Therefore we achieve that
a P is a primitive m -division point ⟺ Ann O K ( a P ) = m aP\text{ is a primitive }\mathfrak{m}\text{-division point}\Longleftrightarrow\operatorname{Ann}_{\mathcal{O}_K}(aP)=\mathfrak{m} a P is a primitive m -division point ⟺ Ann O K ( a P ) = m
⟺ ker μ a = { 0 + m } by (30) \Longleftrightarrow\ker\mu_a=\{0+\mathfrak{m}\}\quad\text{by (30)} ⟺ ker μ a = { 0 + m } by (30)
⟺ μ a is injective \Longleftrightarrow\mu_a\text{ is injective} ⟺ μ a is injective
⟺ μ a is bijective since ∣ O K / m ∣ = N ( m ) < ∞ \Longleftrightarrow\mu_a\text{ is bijective}\quad\text{since }|\mathcal{O}_K/\mathfrak{m}|=N(\mathfrak{m})<\infty ⟺ μ a is bijective since ∣ O K / m ∣ = N ( m ) < ∞
⟺ a + m is a unit in O K / m . \Longleftrightarrow a+\mathfrak{m}\text{ is a unit in }\mathcal{O}_K/\mathfrak{m}. ⟺ a + m is a unit in O K / m .
Remark 4.2. Lemma 4.1 (ii) also follows directly from (i).
Definition 4.3. Let N m > 1 N_{\mathfrak{m}}>1 N m > 1 be the least positive integer in m \mathfrak{m} m . For C ∈ C ( m ) C\in C(\mathfrak{m}) C ∈ C ( m ) , take any ideal c \mathfrak{c} c of O K \mathcal{O}_K O K in C C C , and express
m c − 1 = [ ω 1 , ω 2 ] with ω 1 , ω 2 ∈ C such that ω = ω 1 ω 2 ∈ H , \mathfrak{m}\mathfrak{c}^{-1}=[\omega_1,\omega_2]\quad\text{with }\omega_1,\omega_2\in\mathbb{C}\text{ such that }\omega=\frac{\omega_1}{\omega_2}\in\mathbb{H}, m c − 1 = [ ω 1 , ω 2 ] with ω 1 , ω 2 ∈ C such that ω = ω 2 ω 1 ∈ H , 1 = r 1 ω 1 + r 2 ω 2 with r 1 , r 2 ∈ 1 N m Z . 1=r_1\omega_1+r_2\omega_2\quad\text{with }r_1,r_2\in\frac{1}{N_{\mathfrak{m}}}\mathbb{Z}. 1 = r 1 ω 1 + r 2 ω 2 with r 1 , r 2 ∈ N m 1 Z . We define the Fricke invariant f m ( C ) f_{\mathfrak{m}}(C) f m ( C ) and the Siegel-Ramachandra invariant g m ( C ) g_{\mathfrak{m}}(C) g m ( C ) by
f m ( C ) = f [ r 1 r 2 ] ( ∣ O K × ∣ / 2 ) ( ω ) and g m ( C ) = g [ r 1 r 2 ] ( ω ) 12 N m , f_{\mathfrak{m}}(C)=f_{\left[\begin{smallmatrix}r_1\\r_2\end{smallmatrix}\right]}^{(|\mathcal{O}_K^\times|/2)}(\omega)\quad\text{and}\quad g_{\mathfrak{m}}(C)=g_{\left[\begin{smallmatrix}r_1\\r_2\end{smallmatrix}\right]}(\omega)^{12N_{\mathfrak{m}}}, f m ( C ) = f [ r 1 r 2 ] ( ∣ O K × ∣/2 ) ( ω ) and g m ( C ) = g [ r 1 r 2 ] ( ω ) 12 N m , respectively. These values depend only on the ray class C C C , not on the choices of c \mathfrak{c} c , ω 1 \omega_1 ω 1 and ω 2 \omega_2 ω 2 (see [12 §2.1 and §11.1] and [13 §6.2 and §6.3] ).
Let σ m : C ( m ) → Gal ( K m / K ) \sigma_{\mathfrak{m}}:C(\mathfrak{m})\to\operatorname{Gal}(K_{\mathfrak{m}}/K) σ m : C ( m ) → Gal ( K m / K ) denote the Artin map modulo m \mathfrak{m} m . By the Shimura reciprocity law, we have the following result.
Proposition 4.4. With the notation of Definition 4.3, the invariants f m ( C ) f_{\mathfrak{m}}(C) f m ( C ) and g m ( C ) g_{\mathfrak{m}}(C) g m ( C ) belong to K m K_{\mathfrak{m}} K m . For every C ′ ∈ C ( m ) C'\in C(\mathfrak{m}) C ′ ∈ C ( m ) , they satisfy
f m ( C ) σ m ( C ′ ) = f m ( C C ′ ) and g m ( C ) σ m ( C ′ ) = g m ( C C ′ ) . f_{\mathfrak{m}}(C)^{\sigma_{\mathfrak{m}}(C')}=f_{\mathfrak{m}}(CC')\quad\text{and}\quad g_{\mathfrak{m}}(C)^{\sigma_{\mathfrak{m}}(C')}=g_{\mathfrak{m}}(CC'). f m ( C ) σ m ( C ′ ) = f m ( C C ′ ) and g m ( C ) σ m ( C ′ ) = g m ( C C ′ ) .
Proof. See [12 Theorem 1.1 in Chapter 11] .
For a fractional ideal a \mathfrak{a} a relatively prime to m \mathfrak{m} m , let [ a ] m [\mathfrak{a}]_{\mathfrak{m}} [ a ] m denote its class in C ( m ) C(\mathfrak{m}) C ( m ) .
Lemma 4.5. Let P = ϕ E ( z + O K ) ∈ E P=\phi_E(z+\mathcal{O}_K)\in E P = ϕ E ( z + O K ) ∈ E with z ∈ K z\in K z ∈ K be a primitive m \mathfrak{m} m -division point, and let a ∈ O K a\in\mathcal{O}_K a ∈ O K . If a + m ∈ ( O K / m ) × a+\mathfrak{m}\in(\mathcal{O}_K/\mathfrak{m})^\times a + m ∈ ( O K / m ) × , then
h E ( P ) σ m ( [ a O K ] m ) = h E ( a P ) . h_E(P)^{\sigma_{\mathfrak{m}}([a\mathcal{O}_K]_{\mathfrak{m}})}=h_E(aP). h E ( P ) σ m ([ a O K ] m ) = h E ( a P ) . Proof. Note that c = z m \mathfrak{c}=z\mathfrak{m} c = z m is a nonzero ideal of O K \mathcal{O}_K O K which is relatively prime to m \mathfrak{m} m by the assumption Ann O K ( P ) = m \operatorname{Ann}_{\mathcal{O}_K}(P)=\mathfrak{m} Ann O K ( P ) = m and Lemma 4.1 (i). Set
C = [ c ] m = [ z m ] m and D = [ a z m ] m . C=[\mathfrak{c}]_{\mathfrak{m}}=[z\mathfrak{m}]_{\mathfrak{m}}\quad\text{and}\quad D=[az\mathfrak{m}]_{\mathfrak{m}}. C = [ c ] m = [ z m ] m and D = [ a z m ] m . Then we see by (3), (26) and Definition 4.3 that
h E ( P ) = f m ( C ) , (30) h_E(P)=f_{\mathfrak{m}}(C), \tag*{(30)} h E ( P ) = f m ( C ) , (30) h E ( a P ) = f m ( D ) . (31) h_E(aP)=f_{\mathfrak{m}}(D). \tag*{(31)} h E ( a P ) = f m ( D ) . (31) And, we find that
h E ( P ) σ m ( [ a O K ] m ) = f m ( [ z m ] m ) σ m ( [ a O K ] m ) by (31) = f m ( [ a z m ] m ) by Proposition 4.4 = h E ( a P ) by (32). \begin{aligned}
h_E(P)^{\sigma_{\mathfrak{m}}([a\mathcal{O}_K]_{\mathfrak{m}})}
&=f_{\mathfrak{m}}([z\mathfrak{m}]_{\mathfrak{m}})^{\sigma_{\mathfrak{m}}([a\mathcal{O}_K]_{\mathfrak{m}})} &&\text{by (31)}\\
&=f_{\mathfrak{m}}([az\mathfrak{m}]_{\mathfrak{m}}) &&\text{by Proposition 4.4}\\
&=h_E(aP) &&\text{by (32).}
\end{aligned} h E ( P ) σ m ([ a O K ] m ) = f m ([ z m ] m ) σ m ([ a O K ] m ) = f m ([ a z m ] m ) = h E ( a P ) by (31) by Proposition 4.4 by (32).
Let Q ( d K ) \mathcal{Q}(d_K) Q ( d K ) be the set of primitive positive definite binary quadratic forms over Z \mathbb{Z} Z of discriminant d K d_K d K , that is,
Q ( d K ) = { a x 2 + b x y + c y 2 ∈ Z [ x , y ] ∣ gcd ( a , b , c ) = 1 , b 2 − 4 a c = d K , a > 0 } . \mathcal{Q}(d_K)=\{ax^2+bxy+cy^2\in\mathbb{Z}[x,y]\mid\gcd(a,b,c)=1,\ b^2-4ac=d_K,\ a>0\}. Q ( d K ) = { a x 2 + b x y + c y 2 ∈ Z [ x , y ] ∣ g cd( a , b , c ) = 1 , b 2 − 4 a c = d K , a > 0 } .
For each Q = a x 2 + b x y + c y 2 ∈ Q ( d K ) Q=ax^2+bxy+cy^2\in\mathcal{Q}(d_K) Q = a x 2 + b x y + c y 2 ∈ Q ( d K ) , define ω Q \omega_Q ω Q as the zero of the polynomial Q ( x , 1 ) Q(x,1) Q ( x , 1 ) lying in H \mathbb{H} H , namely,
ω Q = − b + d K 2 a . \omega_Q=\frac{-b+\sqrt{d_K}}{2a}. ω Q = 2 a − b + d K .
We say that a form Q = a x 2 + b x y + c y 2 ∈ Q ( d K ) Q=ax^2+bxy+cy^2\in\mathcal{Q}(d_K) Q = a x 2 + b x y + c y 2 ∈ Q ( d K ) is reduced if
∣ b ∣ ≤ a ≤ c and b ≥ 0 if either ∣ b ∣ = a or a = c . (35) |b|\leq a\leq c\quad\text{and}\quad b\geq0\text{ if either }|b|=a\text{ or }a=c. \tag*{(35)} ∣ b ∣ ≤ a ≤ c and b ≥ 0 if either ∣ b ∣ = a or a = c . (35)
In this case, one sees directly from (35) that
a ≤ ∣ d K ∣ 3 . (36) a\leq\sqrt{\frac{|d_K|}{3}}. \tag*{(36)} a ≤ 3 ∣ d K ∣ . (36)
Moreover, we call Q Q Q the principal reduced form if a = 1 a=1 a = 1 , in which case
Q = { x 2 − d K 4 y 2 if d K ≡ 0 ( m o d 4 ) , x 2 + x y + 1 − d K 4 y 2 if d K ≡ 1 ( m o d 4 ) . Q=
\begin{cases}
x^2-\dfrac{d_K}{4}y^2 & \text{if }d_K\equiv0\pmod{4},\\
x^2+xy+\dfrac{1-d_K}{4}y^2 & \text{if }d_K\equiv1\pmod{4}.
\end{cases} Q = ⎩ ⎨ ⎧ x 2 − 4 d K y 2 x 2 + x y + 4 1 − d K y 2 if d K ≡ 0 ( mod 4 ) , if d K ≡ 1 ( mod 4 ) .
Let Q r e d ( d K ) \mathcal{Q}_{\mathrm{red}}(d_K) Q red ( d K ) denote the set of reduced forms in Q ( d K ) \mathcal{Q}(d_K) Q ( d K ) . As is well known,
Q r e d ( d K ) ⟶ C ( O K ) \mathcal{Q}_{\mathrm{red}}(d_K)\longrightarrow C(\mathcal{O}_K) Q red ( d K ) ⟶ C ( O K )
sending Q Q Q to [ [ ω Q , 1 ] ] O K [\![\omega_Q,1]\!]_{\mathcal{O}_K} [ [ ω Q , 1 ] ] O K is bijective (38) (see [5 Theorems 2.8 and 7.7] ).
Lemma 4.7. With the notation of Definition 4.3 , we have
f m ( C ) = f w ( ∣ O K × ∣ / 2 ) ( ω Q ) and g m ( C ) = g w ( ω Q ) 12 N m f_m(C)=f_w^{(|\mathcal{O}_K^\times|/2)}(\omega_Q)\quad\text{and}\quad g_m(C)=g_w(\omega_Q)^{12N_m} f m ( C ) = f w ( ∣ O K × ∣/2 ) ( ω Q ) and g m ( C ) = g w ( ω Q ) 12 N m for some Q ∈ Q r e d ( d K ) Q\in\mathcal{Q}_{\mathrm{red}}(d_K) Q ∈ Q red ( d K ) and w ∈ 1 N m M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{w}\in\frac{1}{N_m}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z}) w ∈ N m 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) .
Proof. We get by (38) that
m c − 1 = ν [ ω Q , 1 ] \mathfrak{m}\mathfrak{c}^{-1}=\nu[\omega_Q,1] m c − 1 = ν [ ω Q , 1 ]
for some ν ∈ K × \nu\in K^\times ν ∈ K × and Q ∈ Q r e d ( d K ) Q\in\mathcal{Q}_{\mathrm{red}}(d_K) Q ∈ Q red ( d K ) .
Hence we attain by Definition 4.3 that
f m ( C ) = f w ( ∣ O K × ∣ / 2 ) ( ω Q ) and g m ( C ) = g w ( ω Q ) 12 N m f_m(C)=f_w^{(|\mathcal{O}_K^\times|/2)}(\omega_Q)\quad\text{and}\quad g_m(C)=g_w(\omega_Q)^{12N_m} f m ( C ) = f w ( ∣ O K × ∣/2 ) ( ω Q ) and g m ( C ) = g w ( ω Q ) 12 N m
for some Q ∈ Q r e d ( d K ) Q\in\mathcal{Q}_{\mathrm{red}}(d_K) Q ∈ Q red ( d K ) and w ∈ 1 N m M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{w}\in\frac{1}{N_m}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z}) w ∈ N m 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) .
The second Kronecker limit formula
We first recall the second Kronecker limit formula, which relates character sums of logarithms of Siegel-Ramachandra invariants to Hecke L L L -values at s = 1 s=1 s = 1 . We then derive an identity for differences of Weber values used in the character argument of § 6 .
Let χ \chi χ be a nonprincipal character of the ray class group C ( m ) C(\mathfrak{m}) C ( m ) . We define the Stickelberger element S m ( χ ) S_{\mathfrak{m}}(\chi) S m ( χ ) by
S m ( χ ) = ∑ C ∈ C ( m ) χ ( C ) ln ∣ g m ( C ) ∣ . (37) S_{\mathfrak{m}}(\chi)=\sum_{C\in C(\mathfrak{m})}\chi(C)\ln\lvert g_{\mathfrak{m}}(C)\rvert. \tag*{(37)} S m ( χ ) = C ∈ C ( m ) ∑ χ ( C ) ln ∣ g m ( C )∣ . (37)
Furthermore, we define the L L L -function L m ( ⋅ , χ ) L_{\mathfrak{m}}(\cdot,\chi) L m ( ⋅ , χ ) by
L m ( s , χ ) = ∑ a χ ( [ a ] m ) N ( a ) s ( s ∈ C with Re ( s ) > 1 ) , L_{\mathfrak{m}}(s,\chi)=\sum_{\mathfrak{a}}\frac{\chi([\mathfrak{a}]_{\mathfrak{m}})}{\mathrm{N}(\mathfrak{a})^s}\qquad(s\in\mathbb{C}\text{ with }\operatorname{Re}(s)>1), L m ( s , χ ) = a ∑ N ( a ) s χ ([ a ] m ) ( s ∈ C with Re ( s ) > 1 ) ,
where a \mathfrak{a} a runs over all nonzero ideals of O K \mathcal{O}_K O K relatively prime to m \mathfrak{m} m . Then it is well known that L m ( ⋅ , χ ) L_{\mathfrak{m}}(\cdot,\chi) L m ( ⋅ , χ ) extends holomorphically to a neighborhood of s = 1 s=1 s = 1 , and
L m ( 1 , χ ) ≠ 0 (38) L_{\mathfrak{m}}(1,\chi)\ne0 \tag*{(38)} L m ( 1 , χ ) = 0 (38)
(see [7 Theorem 2.14, Proposition 4.1 in Chapter IV and Theorem 10.2 in Chapter V] ).
Denote by f χ \mathfrak{f}_{\chi} f χ the conductor of χ \chi χ , namely, the smallest divisor of m \mathfrak{m} m such that χ \chi χ factors through the natural projection C ( m ) ↠ C ( f χ ) C(\mathfrak{m})\twoheadrightarrow C(\mathfrak{f}_{\chi}) C ( m ) ↠ C ( f χ ) .
Proposition 5.1. Let χ 0 \chi_{0} χ 0 be the primitive character on C ( f χ ) C(\mathfrak{f}_{\chi}) C ( f χ ) associated with χ \chi χ . If f χ ≠ O K \mathfrak{f}_{\chi}\ne\mathcal{O}_K f χ = O K , then we have
L f χ ( 1 , χ 0 ) ( ∏ p ∣ m p ∤ f χ ( 1 − χ 0 ‾ ( [ p ] ) ) ) = − π χ 0 ( [ γ d K f χ ] ) 3 N m ∣ d K ∣ ω ( m ) T γ ( χ 0 ‾ ) S m ( χ ‾ ) , L_{\mathfrak{f}_{\chi}}(1,\chi_{0})\left(\prod_{\substack{\mathfrak{p}\mid\mathfrak{m}\\ \mathfrak{p}\nmid\mathfrak{f}_{\chi}}}\left(1-\overline{\chi_{0}}([\mathfrak{p}])\right)\right)=-\frac{\pi\chi_{0}([\gamma\sqrt{d_K}\mathfrak{f}_{\chi}])}{3N_{\mathfrak{m}}\sqrt{\lvert d_K\rvert}\omega(\mathfrak{m})T_{\gamma}(\overline{\chi_{0}})}S_{\mathfrak{m}}(\overline{\chi}), L f χ ( 1 , χ 0 ) p ∣ m p ∤ f χ ∏ ( 1 − χ 0 ([ p ]) ) = − 3 N m ∣ d K ∣ ω ( m ) T γ ( χ 0 ) π χ 0 ([ γ d K f χ ]) S m ( χ ) , where γ \gamma γ is an element of K K K so that γ d K f χ \gamma\sqrt{d_K}\mathfrak{f}_{\chi} γ d K f χ is a nonzero ideal of O K \mathcal{O}_K O K relatively prime to f χ \mathfrak{f}_{\chi} f χ , N m N_{\mathfrak{m}} N m is the least positive integer in m \mathfrak{m} m ,
ω ( m ) = { ν ∈ O K × ∣ ν ≡ 1 ( m o d m ) } , T γ ( χ 0 ‾ ) = ∑ μ + f χ ∈ ( O K / f χ ) × χ 0 ‾ ( [ μ O K ] ) e 2 π i Tr K / Q ( μ γ ) . \omega(\mathfrak{m})=\left\{\nu\in\mathcal{O}_K^\times\mid\nu\equiv1\pmod{\mathfrak{m}}\right\},\qquad T_{\gamma}(\overline{\chi_{0}})=\sum_{\mu+\mathfrak{f}_{\chi}\in(\mathcal{O}_K/\mathfrak{f}_{\chi})^\times}\overline{\chi_{0}}([\mu\mathcal{O}_K])e^{2\pi i\operatorname{Tr}_{K/\mathbb{Q}}(\mu\gamma)}. ω ( m ) = { ν ∈ O K × ∣ ν ≡ 1 ( mod m ) } , T γ ( χ 0 ) = μ + f χ ∈ ( O K / f χ ) × ∑ χ 0 ([ μ O K ]) e 2 π i Tr K / Q ( μ γ ) . Proof. See [12 Theorem 2.1 in Chapter 11] or [21 Theorem 9 in Chapter II] .
Remark 5.2. By the definition (39), we see that S m ( χ ) ‾ = S m ( χ ) ‾ \overline{S_{\mathfrak{m}(\chi)}}=\overline{S_{\mathfrak{m}}(\chi)} S m ( χ ) = S m ( χ ) .
Lemma 5.3. Let m \mathfrak{m} m be a nonzero proper ideal of O K \mathcal{O}_K O K which is one of the forms
p 3 2 , s p l i t , p 3 2 , s p l i t p 3 ′ , s p l i t , p 3 , n o n i n e r t p 5 , n o n i n e r t , p 3 , s p l i t p 3 ′ , s p l i t p 5 , n o n i n e r t , (39) \mathfrak{p}_3^2,\mathrm{split},\quad\mathfrak{p}_3^2,\mathrm{split}\mathfrak{p}'_3,\mathrm{split},\quad\mathfrak{p}_3,\mathrm{noninert}\mathfrak{p}_5,\mathrm{noninert},\quad\mathfrak{p}_3,\mathrm{split}\mathfrak{p}'_3,\mathrm{split}\mathfrak{p}_5,\mathrm{noninert}, \tag*{(39)} p 3 2 , split , p 3 2 , split p 3 ′ , split , p 3 , noninert p 5 , noninert , p 3 , split p 3 ′ , split p 5 , noninert , (39) so 2 + m ∈ ( O K / m ) × 2+\mathfrak{m}\in(\mathcal{O}_K/\mathfrak{m})^\times 2 + m ∈ ( O K / m ) × . Let P ∈ E ( C ) P\in E(\mathbb{C}) P ∈ E ( C ) be a primitive m \mathfrak{m} m -division point, and let δ = h E ( 2 P ) − h E ( P ) \delta=h_E(2P)-h_E(P) δ = h E ( 2 P ) − h E ( P ) .
Let χ \chi χ be a character of C ( m ) C(\mathfrak{m}) C ( m ) . If h K > 1 h_K>1 h K > 1 and f χ = m \mathfrak{f}_{\chi}=\mathfrak{m} f χ = m , then we have δ ≠ 0 \delta\ne0 δ = 0 and
\sum_{C\in C(\mathfrak{m})}\chi(C)\log\left\lvert\frac{\delta^{12N_{\mathfrak{m}}}{\sigma_{\mathfrak{m}}(C)}\right\rvert=-\overline{\chi}(C_P)\left(1+2\overline{\chi}([\mathcal{O}_K]_{\mathfrak{m}})\right)S_{\mathfrak{m}}(\chi),}
where C P C_P C P is the ray class described in Remark 4.6 (i).
Proof. Since h K > 1 h_K>1 h K > 1 , we get ∣ O K × ∣ / 2 = 1 |\mathcal{O}_K^\times|/2=1 ∣ O K × ∣/2 = 1 and
j ( τ K ) ≠ 0 , 1728 ( i f h K > 1 ) (40) j(\tau_K)\ne0,1728\qquad(if\ h_K>1) \tag*{(40)} j ( τ K ) = 0 , 1728 ( i f h K > 1 ) (40)
(see [5 p. 167, Theorems 11.2 and 12.34] ). By (33), we achieve
h E ( P ) = f v ( τ K ) for some v ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) . (41) h_E(P)=f_{\mathbf{v}}(\tau_K)\quad\text{for some }\mathbf{v}\in M_{1,2}(\mathbb{Q})\setminus M_{1,2}(\mathbb{Z}). \tag*{(41)} h E ( P ) = f v ( τ K ) for some v ∈ M 1 , 2 ( Q ) ∖ M 1 , 2 ( Z ) . (41)
Here v = [ v 1 v 2 ] \mathbf{v}=\begin{bmatrix}v_1&v_2\end{bmatrix} v = [ v 1 v 2 ] is chosen so that P = ϕ E ( z + O K ) P=\phi_E(z+\mathcal{O}_K) P = ϕ E ( z + O K ) , where z = v 1 τ K + v 2 z=v_1\tau_K+v_2 z = v 1 τ K + v 2 . Note that 3 v ∉ M 1 , 2 ( Z ) 3\mathbf{v}\notin M_{1,2}(\mathbb{Z}) 3 v ∈ / M 1 , 2 ( Z ) because m \mathfrak{m} m is one of the ideals listed in (41). Hence we find that
δ 6 = ( f 2 v ( τ K ) − f v ( τ K ) ) 6 by (43) = J K g 3 v ( τ K ) 6 g 2 v ( τ K ) 12 g v ( τ K ) 6 by Lemma 3.4, (42) \begin{aligned}
\delta^6=(f_{2\mathbf{v}}(\tau_K)-f_{\mathbf{v}}(\tau_K))^6\quad\text{by (43)} \\
&=J_K\frac{g_{3\mathbf{v}}(\tau_K)^6}{g_{2\mathbf{v}}(\tau_K)^{12}g_{\mathbf{v}}(\tau_K)^6}\quad\text{by Lemma 3.4,} \tag*{(42)}
\end{aligned} δ 6 = ( f 2 v ( τ K ) − f v ( τ K ) ) 6 by (43) = J K g 2 v ( τ K ) 12 g v ( τ K ) 6 g 3 v ( τ K ) 6 by Lemma 3.4, (42)
where
J K = j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 2 30 3 24 ∈ H K J_K=\frac{j(\tau_K)^2(j(\tau_K)-1728)^3}{2^{30}3^{24}}\in H_K J K = 2 30 3 24 j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 ∈ H K
≠ 0 \ne0 = 0 by (42).
Let m ′ = Ann O K ( 3 P ) \mathfrak{m}'=\operatorname{Ann}_{\mathcal{O}_K}(3P) m ′ = Ann O K ( 3 P ) . Then we have by Lemma 4.1 (i) that
m ′ = m gcd ( m , 3 O K ) , \mathfrak{m}'=\frac{\mathfrak{m}}{\gcd(\mathfrak{m},3\mathcal{O}_K)}, m ′ = g cd( m , 3 O K ) m ,
and so
m ′ ∣ m , m ′ ≠ m , m ′ ≠ O K . \mathfrak{m}'\mid\mathfrak{m},\qquad\mathfrak{m}'\ne\mathfrak{m},\qquad\mathfrak{m}'\ne\mathcal{O}_K. m ′ ∣ m , m ′ = m , m ′ = O K .
Observe that
∑ C ∈ C ( m ) χ ( C ) log ∣ ν σ m ( C ) ∣ for any ν ∈ K m ′ × = ∑ A ∈ ker ( C ( m ) → C ( m ′ ) ) ∑ B ∈ R χ ( A B ) log ∣ ν σ m ( A B ) ∣ \begin{aligned}
\sum_{C\in\mathcal{C}(\mathfrak{m})}\chi(C)\log\left|\nu^{\sigma_{\mathfrak{m}}(C)}\right|
\quad\text{for any }\nu\in K_{\mathfrak{m}'}^\times\\
&=\sum_{A\in\ker(\mathcal{C}(\mathfrak{m})\to\mathcal{C}(\mathfrak{m}'))}\sum_{B\in\mathcal{R}}\chi(AB)\log\left|\nu^{\sigma_{\mathfrak{m}}(AB)}\right|
\end{aligned} C ∈ C ( m ) ∑ χ ( C ) log ν σ m ( C ) for any ν ∈ K m ′ × = A ∈ k e r ( C ( m ) → C ( m ′ )) ∑ B ∈ R ∑ χ ( A B ) log ν σ m ( A B )
where R \mathcal{R} R is a complete set of representatives of C ( m ) / ker ( C ( m ) → C ( m ′ ) ) \mathcal{C}(\mathfrak{m})/\ker(\mathcal{C}(\mathfrak{m})\to\mathcal{C}(\mathfrak{m}')) C ( m ) / ker ( C ( m ) → C ( m ′ ))
= ( ∑ A ∈ ker ( C ( m ) → C ( m ′ ) ) χ ( A ) ) ⋅ ( ∑ B ∈ R χ ( B ) log ∣ ν σ m ( B ) ∣ ) (43) =\left(\sum_{A\in\ker(\mathcal{C}(\mathfrak{m})\to\mathcal{C}(\mathfrak{m}'))}\chi(A)\right)\cdot\left(\sum_{B\in\mathcal{R}}\chi(B)\log\left|\nu^{\sigma_{\mathfrak{m}}(B)}\right|\right) \tag*{(43)} = A ∈ k e r ( C ( m ) → C ( m ′ )) ∑ χ ( A ) ⋅ ( B ∈ R ∑ χ ( B ) log ν σ m ( B ) ) (43)
since ν ∈ K m ′ \nu\in K_{\mathfrak{m}'} ν ∈ K m ′ and so ν σ m ( A ) = ν \nu^{\sigma_{\mathfrak{m}}(A)}=\nu ν σ m ( A ) = ν
= 0 ⋅ ( ∑ B ∈ R χ ( B ) log ∣ ν σ m ( B ) ∣ ) =0\cdot\left(\sum_{B\in\mathcal{R}}\chi(B)\log\left|\nu^{\sigma_{\mathfrak{m}}(B)}\right|\right) = 0 ⋅ ( B ∈ R ∑ χ ( B ) log ν σ m ( B ) )
by [19 Proposition 4 in Chapter VI] , because f χ = m f_\chi=\mathfrak{m} f χ = m implies that χ \chi χ is nontrivial on ker ( C ( m ) → C ( m ′ ) ) \ker(\mathcal{C}(\mathfrak{m})\to\mathcal{C}(\mathfrak{m}')) ker ( C ( m ) → C ( m ′ ))
Now, we derive that
∑ C ∈ C ( m ) χ ( C ) log ∣ ( δ 12 N m ) σ m ( C ) ∣ = 2 N m ∑ C ∈ C ( m ) χ ( C ) log ∣ J K σ m ( C ) ∣ + ∑ C ∈ C ( m ) χ ( C ) log ∣ ( g 3 v ( τ K ) 12 N m ) σ m ( C ) ∣ \begin{aligned}
\sum_{C\in\mathcal{C}(\mathfrak{m})}\chi(C)\log\left|(\delta^{12N_{\mathfrak{m}}})^{\sigma_{\mathfrak{m}}(C)}\right| \\
&=2N_{\mathfrak{m}}\sum_{C\in\mathcal{C}(\mathfrak{m})}\chi(C)\log\left|J_K^{\sigma_{\mathfrak{m}}(C)}\right|+\sum_{C\in\mathcal{C}(\mathfrak{m})}\chi(C)\log\left|(g_{3\mathbf{v}}(\tau_K)^{12N_{\mathfrak{m}}})^{\sigma_{\mathfrak{m}}(C)}\right|
\end{aligned} C ∈ C ( m ) ∑ χ ( C ) log ( δ 12 N m ) σ m ( C ) = 2 N m C ∈ C ( m ) ∑ χ ( C ) log J K σ m ( C ) + C ∈ C ( m ) ∑ χ ( C ) log ( g 3 v ( τ K ) 12 N m ) σ m ( C )
− 2 ∑ C ∈ C ( m ) χ ( C ) log ∣ ( g 2 v ( τ K ) 12 N m ) σ m ( C ) ∣ − ∑ C ∈ C ( m ) χ ( C ) log ∣ ( g v ( τ K ) 12 N m ) σ m ( C ) ∣ = 2 N m ∑ C ∈ C ( m ) χ ( C ) log ∣ J K σ m ( C ) ∣ + ∑ C ∈ C ( m ) χ ( C ) log ∣ ( g m ′ ( [ 3 z m ′ ] m ′ ) N m / N m ′ ) σ m ( C ) ∣ \begin{aligned}
-2\sum_{C\in C(\mathfrak m)}\chi(C)\log\left|(g_{2v}(\tau_K)^{12N_{\mathfrak m}})^{\sigma_{\mathfrak m}(C)}\right|-\sum_{C\in C(\mathfrak m)}\chi(C)\log\left|(g_v(\tau_K)^{12N_{\mathfrak m}})^{\sigma_{\mathfrak m}(C)}\right| \\
&=2N_{\mathfrak m}\sum_{C\in C(\mathfrak m)}\chi(C)\log\left|J_K^{\sigma_{\mathfrak m}(C)}\right|+\sum_{C\in C(\mathfrak m)}\chi(C)\log\left|\left(g_{\mathfrak m'}([3\mathfrak z\mathfrak m']_{\mathfrak m'})^{N_{\mathfrak m}/N_{\mathfrak m'}}\right)^{\sigma_{\mathfrak m}(C)}\right|
\end{aligned} − 2 C ∈ C ( m ) ∑ χ ( C ) log ( g 2 v ( τ K ) 12 N m ) σ m ( C ) − C ∈ C ( m ) ∑ χ ( C ) log ( g v ( τ K ) 12 N m ) σ m ( C ) = 2 N m C ∈ C ( m ) ∑ χ ( C ) log J K σ m ( C ) + C ∈ C ( m ) ∑ χ ( C ) log ( g m ′ ([ 3 z m ′ ] m ′ ) N m / N m ′ ) σ m ( C )
− 2 ∑ C ∈ C ( m ) χ ( C ) log ∣ ( g m ( C P ) σ m ( [ 2 O K ] m ) ) σ m ( C ) ∣ − ∑ C ∈ C ( m ) χ ( C ) log ∣ g m ( C P ) σ m ( C ) ∣ -2\sum_{C\in C(\mathfrak m)}\chi(C)\log\left|\left(g_{\mathfrak m}(C_P)^{\sigma_{\mathfrak m}([2O_K]_{\mathfrak m})}\right)^{\sigma_{\mathfrak m}(C)}\right|-\sum_{C\in C(\mathfrak m)}\chi(C)\log\left|g_{\mathfrak m}(C_P)^{\sigma_{\mathfrak m}(C)}\right| − 2 C ∈ C ( m ) ∑ χ ( C ) log ( g m ( C P ) σ m ([ 2 O K ] m ) ) σ m ( C ) − C ∈ C ( m ) ∑ χ ( C ) log g m ( C P ) σ m ( C )
by (34) and (35)
= 0 + 0 − 2 χ ‾ ( C P [ 2 O K ] m ) S m ( χ ) − χ ‾ ( C P ) S m ( χ ) =0+0-2\overline{\chi}(C_P[2O_K]_{\mathfrak m})S_{\mathfrak m}(\chi)-\overline{\chi}(C_P)S_{\mathfrak m}(\chi) = 0 + 0 − 2 χ ( C P [ 2 O K ] m ) S m ( χ ) − χ ( C P ) S m ( χ )
by (45) due to the fact J K J_K J K , g m ′ ( [ 3 z m ′ ] m ′ ) ∈ K m ′ × g_{\mathfrak m'}([3\mathfrak z\mathfrak m']_{\mathfrak m'})\in K_{\mathfrak m'}^\times g m ′ ([ 3 z m ′ ] m ′ ) ∈ K m ′ ×
= − χ ‾ ( C P ) ( 1 + 2 χ ‾ ( [ 2 O K ] m ) ) S m ( χ ) . =-\overline{\chi}(C_P)(1+2\overline{\chi}([2O_K]_{\mathfrak m}))S_{\mathfrak m}(\chi). = − χ ( C P ) ( 1 + 2 χ ([ 2 O K ] m )) S m ( χ ) .
Proof using an L L L -function argument
In this section, we prove Theorem A for four of the seven families in Proposition 2.1 using the second Kronecker limit formula.
Lemma 6.1. Let m \mathfrak m m be a nonzero proper ideal of O K O_K O K which is one of the forms p 3 , s p l i t 2 \mathfrak p^2_{3,\mathrm{split}} p 3 , split 2 , p 3 , s p l i t 2 p 3 , s p l i t ′ \mathfrak p^2_{3,\mathrm{split}}\mathfrak p'_{3,\mathrm{split}} p 3 , split 2 p 3 , split ′ , p 3 , n o n i n e r t p 5 , n o n i n e r t \mathfrak p_{3,\mathrm{noninert}}\mathfrak p_{5,\mathrm{noninert}} p 3 , noninert p 5 , noninert , p 3 , s p l i t p 3 , s p l i t ′ p 5 , n o n i n e r t \mathfrak p_{3,\mathrm{split}}\mathfrak p'_{3,\mathrm{split}}\mathfrak p_{5,\mathrm{noninert}} p 3 , split p 3 , split ′ p 5 , noninert .
Let P ∈ E ( C ) P\in E(\mathbb C) P ∈ E ( C ) be a primitive m \mathfrak m m -division point, and let L = K ( h E ( P ) ) L=K(h_E(P)) L = K ( h E ( P )) . If L ≠ K m L\ne K_{\mathfrak m} L = K m , then there exists a character χ \chi χ of C ( m ) C(\mathfrak m) C ( m ) such that
f χ = m and χ ∣ σ m − 1 ( Gal ( K m / L ) ) ≠ 1. \mathfrak f_\chi=\mathfrak m\quad\text{and}\quad\chi|_{\sigma_{\mathfrak m}^{-1}(\operatorname{Gal}(K_{\mathfrak m}/L))}\ne1. f χ = m and χ ∣ σ m − 1 ( Gal ( K m / L )) = 1.
Proof. If h K = 1 h_K=1 h K = 1 , then (4) gives L = K m L=K_{\mathfrak m} L = K m . So we may assume that h K > 1 h_K>1 h K > 1 , and hence O K × = { ± 1 } O_K^\times=\{\pm1\} O K × = { ± 1 } . The assumption L ≠ K m L\ne K_{\mathfrak m} L = K m yields
σ m − 1 ( Gal ( K m / L ) ) ≠ { [ O K ] m } . \sigma_{\mathfrak m}^{-1}(\operatorname{Gal}(K_{\mathfrak m}/L))\ne\{[O_K]_{\mathfrak m}\}. σ m − 1 ( Gal ( K m / L )) = {[ O K ] m } .
Let
R m : = ker ( C ( m ) ↠ C ( O K ) ) . R_{\mathfrak m}:=\ker(C(\mathfrak m)\twoheadrightarrow C(O_K)). R m := ker ( C ( m ) ↠ C ( O K )) .
Since R m ≅ Gal ( K m / H K ) R_{\mathfrak m}\cong\operatorname{Gal}(K_{\mathfrak m}/H_K) R m ≅ Gal ( K m / H K ) and H K L = K m H_KL=K_{\mathfrak m} H K L = K m by (4), we observe that
R m ∩ σ m − 1 ( Gal ( K m / L ) ) = { [ O K ] m } . R_{\mathfrak m}\cap\sigma_{\mathfrak m}^{-1}(\operatorname{Gal}(K_{\mathfrak m}/L))=\{[O_K]_{\mathfrak m}\}. R m ∩ σ m − 1 ( Gal ( K m / L )) = {[ O K ] m } .
On the other hand, we have the natural isomorphisms
Z / 3 Z ≅ O K / p 3 , n o n i n e r t , Z / 9 Z ≅ O K / p 3 , s p l i t 2 , Z / 5 Z ≅ O K / p 5 , n o n i n e r t . \mathbb Z/3\mathbb Z\cong O_K/\mathfrak p_{3,\mathrm{noninert}},\qquad
\mathbb Z/9\mathbb Z\cong O_K/\mathfrak p^2_{3,\mathrm{split}},\qquad
\mathbb Z/5\mathbb Z\cong O_K/\mathfrak p_{5,\mathrm{noninert}}. Z /3 Z ≅ O K / p 3 , noninert , Z /9 Z ≅ O K / p 3 , split 2 , Z /5 Z ≅ O K / p 5 , noninert .
Note that the multiplicative groups ( Z / 3 Z ) × (\mathbb Z/3\mathbb Z)^\times ( Z /3 Z ) × , ( Z / 9 Z ) × (\mathbb Z/9\mathbb Z)^\times ( Z /9 Z ) × , ( Z / 5 Z ) × (\mathbb Z/5\mathbb Z)^\times ( Z /5 Z ) × are cyclic groups of order 2, 6, 4 generated by − 1 + 3 Z -1+3\mathbb Z − 1 + 3 Z , 2 + 9 Z 2+9\mathbb Z 2 + 9 Z , 2 + 5 Z 2+5\mathbb Z 2 + 5 Z , respectively. Thus we define characters ξ 1 \xi_1 ξ 1 , ξ 2 \xi_2 ξ 2 , ξ 3 \xi_3 ξ 3 on ( Z / 3 Z ) × (\mathbb Z/3\mathbb Z)^\times ( Z /3 Z ) × , ( Z / 9 Z ) × (\mathbb Z/9\mathbb Z)^\times ( Z /9 Z ) × , ( Z / 5 Z ) × (\mathbb Z/5\mathbb Z)^\times ( Z /5 Z ) × by letting
ξ 1 ( − 1 + 3 Z ) = − 1 , ξ 2 ( 2 + 9 Z ) = ζ 6 , ξ 3 ( 2 + 5 Z ) = i . \xi_1(-1+3\mathbb Z)=-1,\qquad\xi_2(2+9\mathbb Z)=\zeta_6,\qquad\xi_3(2+5\mathbb Z)=i. ξ 1 ( − 1 + 3 Z ) = − 1 , ξ 2 ( 2 + 9 Z ) = ζ 6 , ξ 3 ( 2 + 5 Z ) = i .
respectively. We then see that
ξ 1 ( − 1 + 3 Z ) = ξ 2 ( − 1 + 9 Z ) = ξ 3 ( − 1 + 5 Z ) = − 1. (44) \xi_1(-1+3\mathbb{Z})=\xi_2(-1+9\mathbb{Z})=\xi_3(-1+5\mathbb{Z})=-1. \tag*{(44)} ξ 1 ( − 1 + 3 Z ) = ξ 2 ( − 1 + 9 Z ) = ξ 3 ( − 1 + 5 Z ) = − 1. (44)
If we write m = ∏ p ∣ m p e p \mathfrak{m}=\prod_{\mathfrak{p}\mid\mathfrak{m}}\mathfrak{p}^{e_{\mathfrak{p}}} m = ∏ p ∣ m p e p , then we get by the Chinese remainder theorem that
( O K / m ) × ≃ ∏ p ∣ m ( O K / p e p ) × . (\mathcal{O}_K/\mathfrak{m})^\times\simeq\prod_{\mathfrak{p}\mid\mathfrak{m}}(\mathcal{O}_K/\mathfrak{p}^{e_{\mathfrak{p}}})^\times. ( O K / m ) × ≃ p ∣ m ∏ ( O K / p e p ) × .
Thus, by (49), we can identify ( O K / m ) × (\mathcal{O}_K/\mathfrak{m})^\times ( O K / m ) × with the group in the second column of Table 1 , and define a character θ \theta θ as follows:
Table 1. A character θ \theta θ of ( O K / m ) × (\mathcal{O}_K/\mathfrak{m})^\times ( O K / m ) ×
The exact sequence given in (12) yields the surjective homomorphism
q m : ( O K / m ) × ↠ R m ≃ Gal ( K m / H K ) , a + m ↦ [ a O K ] m q_{\mathfrak{m}}:(\mathcal{O}_K/\mathfrak{m})^\times\twoheadrightarrow R_{\mathfrak{m}}\simeq\operatorname{Gal}(K_{\mathfrak{m}}/H_K),\qquad a+\mathfrak{m}\mapsto[a\mathcal{O}_K]_{\mathfrak{m}} q m : ( O K / m ) × ↠ R m ≃ Gal ( K m / H K ) , a + m ↦ [ a O K ] m
with
ker ( q m ) = the image of O K × in ( O K / m ) × = { 1 + m , − 1 + m } . \ker(q_{\mathfrak{m}})=\text{the image of }\mathcal{O}_K^\times\text{ in }(\mathcal{O}_K/\mathfrak{m})^\times=\{1+\mathfrak{m},-1+\mathfrak{m}\}. ker ( q m ) = the image of O K × in ( O K / m ) × = { 1 + m , − 1 + m } .
Since the character θ \theta θ of ( O K / m ) × (\mathcal{O}_K/\mathfrak{m})^\times ( O K / m ) × is trivial on ker ( q m ) = { 1 + m , − 1 + m } \ker(q_{\mathfrak{m}})=\{1+\mathfrak{m},-1+\mathfrak{m}\} ker ( q m ) = { 1 + m , − 1 + m } by (44) , it induces a character θ ~ \widetilde{\theta} θ of ( O K / m ) × / ker ( q m ) (\mathcal{O}_K/\mathfrak{m})^\times/\ker(q_{\mathfrak{m}}) ( O K / m ) × / ker ( q m ) . Moreover, since ( O K / m ) × / ker ( q m ) (\mathcal{O}_K/\mathfrak{m})^\times/\ker(q_{\mathfrak{m}}) ( O K / m ) × / ker ( q m ) injects into the finite abelian group C ( m ) C(\mathfrak{m}) C ( m ) by (12) , θ ~ \widetilde{\theta} θ extends to a character η \eta η of C ( m ) C(\mathfrak{m}) C ( m ) (see [19 Proposition 1 in Chapter VI] ).
For a divisor a \mathfrak{a} a of m \mathfrak{m} m , let
π m , a : C ( m ) ⟶ C ( a ) \pi_{\mathfrak{m},\mathfrak{a}}:C(\mathfrak{m})\longrightarrow C(\mathfrak{a}) π m , a : C ( m ) ⟶ C ( a )
be the natural surjective homomorphism (cf. [7 Lemma 6.1 in Chapter V] ). Then we deduce that for a character ξ \xi ξ of C ( m ) C(\mathfrak{m}) C ( m )
f ξ = m f_\xi=\mathfrak{m} f ξ = m
⟺ ξ ∣ ker ( π m , m p − 1 ) ≠ 1 for every p ∣ m \Longleftrightarrow\xi|_{\ker(\pi_{\mathfrak{m},\mathfrak{m}\mathfrak{p}^{-1}})}\ne1\text{ for every }\mathfrak{p}\mid\mathfrak{m} ⟺ ξ ∣ k e r ( π m , m p − 1 ) = 1 for every p ∣ m
⟺ for each p ∣ m , there is an element C p ∈ ker ( π m , m p − 1 ) such that ξ ( C p ) ≠ 1. (45) \Longleftrightarrow\text{for each }\mathfrak{p}\mid\mathfrak{m},\text{ there is an element }C_{\mathfrak{p}}\in\ker(\pi_{\mathfrak{m},\mathfrak{m}\mathfrak{p}^{-1}})\text{ such that }\xi(C_{\mathfrak{p}})\ne1. \tag*{(45)} ⟺ for each p ∣ m , there is an element C p ∈ ker ( π m , m p − 1 ) such that ξ ( C p ) = 1. (45)
For each p ∣ m \mathfrak{p}\mid\mathfrak{m} p ∣ m , take an element u p u_{\mathfrak{p}} u p of ( O K / m ) × (\mathcal{O}_K/\mathfrak{m})^\times ( O K / m ) × which reduces to the identity of ( O K / m p − 1 ) × (\mathcal{O}_K/\mathfrak{m}\mathfrak{p}^{-1})^\times ( O K / m p − 1 ) × as in Table 2 . The last column of Table 2 follows directly from (50).
Table 2. An element u p ∈ ( O K / m ) × u_{\mathfrak p}\in(\mathcal O_K/m)^\times u p ∈ ( O K / m ) × reducing to the identity of ( O K / m p − 1 ) × (\mathcal O_K/m\mathfrak p^{-1})^\times ( O K / m p − 1 ) ×
For each p ∣ m \mathfrak p\mid m p ∣ m , choose a representative a p ∈ O K a_{\mathfrak p}\in\mathcal O_K a p ∈ O K of u p ∈ ( O K / m ) × u_{\mathfrak p}\in(\mathcal O_K/m)^\times u p ∈ ( O K / m ) × . Then we observe that a p O K a_{\mathfrak p}\mathcal O_K a p O K is relatively prime to m m m and
a p ≡ 1 ( m o d m p − 1 ) . a_{\mathfrak p}\equiv1\pmod{m\mathfrak p^{-1}}. a p ≡ 1 ( mod m p − 1 ) .
And, the class
A p : = q m ( u p ) = [ a p O K ] m ∈ R m ( = ker ( C ( m ) → C ( O K ) ) ) A_{\mathfrak p}:=q_m(u_{\mathfrak p})=[a_{\mathfrak p}\mathcal O_K]_m\in R_m(=\ker(C(m)\to C(\mathcal O_K))) A p := q m ( u p ) = [ a p O K ] m ∈ R m ( = ker ( C ( m ) → C ( O K )))
satisfies
π m , m p − 1 ( A p ) = [ a p O K ] m p − 1 = [ O K ] m p − 1 (46) \pi_{m,m\mathfrak p^{-1}}(A_{\mathfrak p})=[a_{\mathfrak p}\mathcal O_K]_{m\mathfrak p^{-1}}=[\mathcal O_K]_{m\mathfrak p^{-1}} \tag*{(46)} π m , m p − 1 ( A p ) = [ a p O K ] m p − 1 = [ O K ] m p − 1 (46)
and
η ( A p ) = θ ~ ( q m ( u p ) ) = θ ( u p ) ≠ 1. (47) \eta(A_{\mathfrak p})=\widetilde{\theta}(q_m(u_{\mathfrak p}))=\theta(u_{\mathfrak p})\ne1. \tag*{(47)} η ( A p ) = θ ( q m ( u p )) = θ ( u p ) = 1. (47)
Hence we obtain by (45) that
Here, if η ∣ σ m − 1 ( Gal ( K m / L ) ) ≠ 1 \eta|_{\sigma_m^{-1}(\operatorname{Gal}(K_m/L))}\ne1 η ∣ σ m − 1 ( Gal ( K m / L )) = 1 , then χ = η \chi=\eta χ = η satisfies the required property (46), and we are done. Otherwise, if
η ∣ σ m − 1 ( Gal ( K m / L ) ) = 1 , (48) \eta|_{\sigma_m^{-1}(\operatorname{Gal}(K_m/L))}=1, \tag*{(48)} η ∣ σ m − 1 ( Gal ( K m / L )) = 1 , (48)
choose an element
A ∈ σ m − 1 ( Gal ( K m / L ) ) ∖ { [ O K ] m } A\in\sigma_m^{-1}(\operatorname{Gal}(K_m/L))\setminus\{[\mathcal O_K]_m\} A ∈ σ m − 1 ( Gal ( K m / L )) ∖ {[ O K ] m }
by (47). Since A ∉ R m A\notin R_m A ∈ / R m by (48), A R m AR_m A R m in C ( m ) / R m C(m)/R_m C ( m ) / R m is nontrivial. So there is a character ψ \psi ψ of the abelian group C ( m ) / R m C(m)/R_m C ( m ) / R m such that
ψ ( A R m ) ≠ 1 (49) \psi(AR_m)\ne1 \tag*{(49)} ψ ( A R m ) = 1 (49)
(see [19 p. 61] ). Let ψ ~ \widetilde{\psi} ψ be its inflation to C ( m ) C(m) C ( m ) so that
ψ ~ ∣ R m = 1 (50) \widetilde{\psi}|_{R_m}=1 \tag*{(50)} ψ ∣ R m = 1 (50)
and
ψ ~ ( B ) = ψ ( B R m ) ( B ∈ C ( m ) ) . \widetilde{\psi}(B)=\psi(BR_m)\quad(B\in C(m)). ψ ( B ) = ψ ( B R m ) ( B ∈ C ( m )) .
Since C ( m ) / R m ≃ C ( O K ) C(\mathfrak m)/R_{\mathfrak m}\simeq C(O_K) C ( m ) / R m ≃ C ( O K ) , we attain f ~ ψ ~ = O K \widetilde{f}_{\widetilde{\psi}}=O_K f ψ = O K . Put
χ = η ψ ~ . \chi=\eta\widetilde{\psi}. χ = η ψ .
We see from (48) and (49) that
χ ( A ) = η ( A ) ψ ~ ( A ) = ψ ( A R m ) ≠ 1 , \chi(A)=\eta(A)\widetilde{\psi}(A)=\psi(AR_{\mathfrak m})\ne1, χ ( A ) = η ( A ) ψ ( A ) = ψ ( A R m ) = 1 ,
and so
χ ∣ σ m − 1 ( Gal ( K m / L ) ) ≠ 1. \chi\big|_{\sigma_{\mathfrak m}^{-1}(\operatorname{Gal}(K_{\mathfrak m}/L))}\ne1. χ σ m − 1 ( Gal ( K m / L )) = 1.
For each p ∣ m \mathfrak p\mid\mathfrak m p ∣ m , recall that A p ∈ R m A_{\mathfrak p}\in R_{\mathfrak m} A p ∈ R m , and observe by (50) and (47) that
χ ( A p ) = η ( A p ) ψ ~ ( A p ) = η ( A p ) ≠ 1. \chi(A_{\mathfrak p})=\eta(A_{\mathfrak p})\widetilde{\psi}(A_{\mathfrak p})=\eta(A_{\mathfrak p})\ne1. χ ( A p ) = η ( A p ) ψ ( A p ) = η ( A p ) = 1.
Therefore (45) and (46) give f χ = m \mathfrak f_\chi=\mathfrak m f χ = m , and hence χ \chi χ is a required character of C ( m ) C(\mathfrak m) C ( m ) satisfying (46).
Theorem 6.2. Let K K K be an imaginary quadratic field, and let E E E and h E h_E h E be as in (1) –(3) . If m \mathfrak m m is a nonzero proper ideal of O K O_K O K that is the exact conductor of K m / K K_{\mathfrak m}/K K m / K and is one of four cases
p 3 2 , s p l i t , p 3 2 , s p l i t p 3 ′ , s p l i t , p 3 , n o n i n e r t p 5 ′ , n o n i n e r t , p 3 , s p l i t p 3 ′ , s p l i t p 5 ′ , n o n i n e r t , (51) \mathfrak p_3^2{}_{\mathrm{,split}},\quad\mathfrak p_3^2{}_{\mathrm{,split}}\mathfrak p'_3{}_{\mathrm{,split}},\quad\mathfrak p_3{}_{\mathrm{,noninert}}\mathfrak p'_5{}_{\mathrm{,noninert}},\quad\mathfrak p_3{}_{\mathrm{,split}}\mathfrak p'_3{}_{\mathrm{,split}}\mathfrak p'_5{}_{\mathrm{,noninert}}, \tag*{(51)} p 3 2 , split , p 3 2 , split p 3 ′ , split , p 3 , noninert p 5 ′ , noninert , p 3 , split p 3 ′ , split p 5 ′ , noninert , (51) then we have
K m = K ( h E ( P ) ) K_{\mathfrak m}=K(h_E(P)) K m = K ( h E ( P )) for every primitive m \mathfrak m m -division point P ∈ E ( C ) P\in E(\mathbb C) P ∈ E ( C ) .
Proof. If h K = 1 h_K=1 h K = 1 , then the result follows directly from (4) . Now, consider the case where h K > 1 h_K>1 h K > 1 . Let
L = K ( h E ( P ) ) . L=K(h_E(P)). L = K ( h E ( P )) .
Since K m / K K_{\mathfrak m}/K K m / K is an abelian extension, so is L / K L/K L / K . Suppose on the contrary that L L L is properly contained in K m K_{\mathfrak m} K m . Then one can choose a character χ \chi χ satisfying
f χ = m (52) \mathfrak f_\chi=\mathfrak m \tag*{(52)} f χ = m (52)
and
χ ∣ σ m − 1 ( Gal ( K m / L ) ) ≠ 1 (53) \chi\big|_{\sigma_{\mathfrak m}^{-1}(\operatorname{Gal}(K_{\mathfrak m}/L))}\ne1 \tag*{(53)} χ σ m − 1 ( Gal ( K m / L )) = 1 (53)
by Lemma 6.1 . Since 2 + m ∈ ( O K / m ) × 2+\mathfrak m\in(O_K/\mathfrak m)^\times 2 + m ∈ ( O K / m ) × with m \mathfrak m m listed in (51) , we get by Lemma 4.5 that
h E ( 2 P ) = h E ( P ) σ m ( [ 2 O K ] m ) ∈ L , h_E(2P)=h_E(P)^{\sigma_{\mathfrak m}([2O_K]_{\mathfrak m})}\in L, h E ( 2 P ) = h E ( P ) σ m ([ 2 O K ] m ) ∈ L ,
so
δ : = h E ( 2 P ) − h E ( P ) ∈ L . \delta:=h_E(2P)-h_E(P)\in L. δ := h E ( 2 P ) − h E ( P ) ∈ L .
By Lemma 5.3 , we have δ ≠ 0 \delta\ne0 δ = 0 . We then find that
∑ C ∈ C ( m ) χ ( C ) log ∣ ( δ 12 N m ) σ m ( C ) ∣ \sum_{C\in C(\mathfrak m)}\chi(C)\log\left|(\delta^{12N_{\mathfrak m}})^{\sigma_{\mathfrak m}(C)}\right| C ∈ C ( m ) ∑ χ ( C ) log ( δ 12 N m ) σ m ( C )
∑ A ∈ σ m − 1 ( Gal ( K m / L ) ) ∑ B ∈ R χ ( A B ) log ∣ ( δ 12 N m ) σ m ( A B ) ∣ \sum_{A\in\sigma_m^{-1}(\operatorname{Gal}(K_m/L))}\sum_{B\in\mathcal{R}}\chi(AB)\log\left|(\delta^{12N_m})^{\sigma_m(AB)}\right| A ∈ σ m − 1 ( Gal ( K m / L )) ∑ B ∈ R ∑ χ ( A B ) log ( δ 12 N m ) σ m ( A B )
where R \mathcal{R} R is a complete set of representatives of C ( m ) / σ m − 1 ( Gal ( K m / L ) ) C(\mathfrak{m})/\sigma_m^{-1}(\operatorname{Gal}(K_m/L)) C ( m ) / σ m − 1 ( Gal ( K m / L ))
( ∑ A ∈ σ m − 1 ( Gal ( K m / L ) ) χ ( A ) ) ⋅ ( ∑ B ∈ R χ ( B ) log ∣ ( δ 12 N m ) σ m ( B ) ∣ ) \left(\sum_{A\in\sigma_m^{-1}(\operatorname{Gal}(K_m/L))}\chi(A)\right)\cdot\left(\sum_{B\in\mathcal{R}}\chi(B)\log\left|(\delta^{12N_m})^{\sigma_m(B)}\right|\right) A ∈ σ m − 1 ( Gal ( K m / L )) ∑ χ ( A ) ⋅ ( B ∈ R ∑ χ ( B ) log ( δ 12 N m ) σ m ( B ) )
since ( δ 12 N m ) σ m ( A ) = δ 12 N m (\delta^{12N_m})^{\sigma_m(A)}=\delta^{12N_m} ( δ 12 N m ) σ m ( A ) = δ 12 N m
= 0 ⋅ ( ∑ B ∈ R χ ( B ) log ∣ ( δ 12 N m ) σ m ( B ) ∣ ) =0\cdot\left(\sum_{B\in\mathcal{R}}\chi(B)\log\left|(\delta^{12N_m})^{\sigma_m(B)}\right|\right) = 0 ⋅ ( B ∈ R ∑ χ ( B ) log ( δ 12 N m ) σ m ( B ) )
by (60) and [19 Proposition 4 in Chapter VI]
On the other hand, we deduce from (40), (59), Proposition 5.1 and Remark 5.2 that
S m ( χ ) ≠ 0. (54) S_{\mathfrak{m}}(\chi)\ne0. \tag*{(54)} S m ( χ ) = 0. (54)
We find that
∑ C ∈ C ( m ) χ ( C ) log ∣ ( δ 12 N m ) σ m ( C ) ∣ = − χ ‾ ( C P ) ( 1 + 2 χ ‾ ( [ 2 O K ] m ) ) S m ( χ ) \sum_{C\in C(\mathfrak{m})}\chi(C)\log\left|(\delta^{12N_m})^{\sigma_m(C)}\right|=-\overline{\chi}(C_P)(1+2\overline{\chi}([2\mathcal{O}_K]_{\mathfrak{m}}))S_{\mathfrak{m}}(\chi) C ∈ C ( m ) ∑ χ ( C ) log ( δ 12 N m ) σ m ( C ) = − χ ( C P ) ( 1 + 2 χ ([ 2 O K ] m )) S m ( χ )
by Lemma 5.3
by (61), since ∣ χ ‾ ( [ 2 O K ] m ) ∣ = 1 \left|\overline{\chi}([2\mathcal{O}_K]_{\mathfrak{m}})\right|=1 ∣ χ ([ 2 O K ] m ) ∣ = 1 ,
which yields a contradiction. Therefore we conclude that L = K m L=K_m L = K m , as desired.
Proof by inequality arguments
We treat the remaining three families by comparing the absolute value of a difference of Weber values with those of its Galois conjugates.
Lemma 7.1. If h K > 1 h_K>1 h K > 1 and ∣ d K ∣ ≥ 20 |d_K|\ge20 ∣ d K ∣ ≥ 20 , then we have
∣ j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 ∣ < 877383 ∣ q τ K ∣ 5 / 2 \left|\frac{j(\omega_Q)^2(j(\omega_Q)-1728)^3}{j(\tau_K)^2(j(\tau_K)-1728)^3}\right|<877383|q_{\tau_K}|^{5/2} j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 < 877383∣ q τ K ∣ 5/2 for all nonprincipal reduced forms Q ∈ Q r e d ( d K ) Q\in\mathcal{Q}_{\mathrm{red}}(d_K) Q ∈ Q red ( d K ) .
Proof. See [8 Lemma 6.3 (ii)] .
Lemma 7.2. Let ( N , t ) ∈ { ( 12 , 5 ) , ( 8 , 3 ) , ( 5 , 2 ) } (N,t)\in\{(12,5),(8,3),(5,2)\} ( N , t ) ∈ {( 12 , 5 ) , ( 8 , 3 ) , ( 5 , 2 )} and w = [ w 1 w 2 ] ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{w}=\begin{bmatrix}w_1&w_2\end{bmatrix}\in\frac{1}{N}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z}) w = [ w 1 w 2 ] ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) .
If τ ∈ H \tau\in\mathbb{H} τ ∈ H satisfies Im ( τ ) ≥ 3 / 2 \operatorname{Im}(\tau)\ge\sqrt{3}/2 Im ( τ ) ≥ 3 /2 , then
∣ ℘ ( t ( w 1 τ + w 2 ) ; [ τ , 1 ] ) − ℘ ( w 1 τ + w 2 ; [ τ , 1 ] ) ∣ < ( 2 π ) 2 N 2 10 . \left|\wp(t(w_1\tau+w_2);[\tau,1])-\wp(w_1\tau+w_2;[\tau,1])\right|<\frac{(2\pi)^2N^2}{10}. ∣ ℘ ( t ( w 1 τ + w 2 ) ; [ τ , 1 ]) − ℘ ( w 1 τ + w 2 ; [ τ , 1 ]) ∣ < 10 ( 2 π ) 2 N 2 . If ∣ d K ∣ ≥ 15 |d_K|\ge15 ∣ d K ∣ ≥ 15 , then for every integer s s s
∣ ℘ ( t ( τ K + s ) N ; [ τ K , 1 ] ) − ℘ ( τ K + s N ; [ τ K , 1 ] ) ∣ > ( 2 π ) 2 2 ∣ q τ K ∣ 1 / N . \left|\wp\left(\frac{t(\tau_K+s)}{N};[\tau_K,1]\right)-\wp\left(\frac{\tau_K+s}{N};[\tau_K,1]\right)\right|>\frac{(2\pi)^2}{2}|q_{\tau_K}|^{1/N}. ℘ ( N t ( τ K + s ) ; [ τ K , 1 ] ) − ℘ ( N τ K + s ; [ τ K , 1 ] ) > 2 ( 2 π ) 2 ∣ q τ K ∣ 1/ N .
Proof. By Lemma 3.2, we have the expansion formula
1 ( 2 π i ) 2 ℘ ( z ; [ τ , 1 ] ) = 1 12 + q z ( 1 − q z ) 2 + ∑ n = 1 ∞ ( q τ n q z ( 1 − q τ n q z ) 2 + q τ n q z − 1 ( 1 − q τ n q z − 1 ) 2 ) − 2 ∑ n = 1 ∞ n q τ n 1 − q τ n ( τ ∈ H , z ∈ C ) . (55) \frac{1}{(2\pi i)^2}\wp(z;[\tau,1])=\frac{1}{12}+\frac{q_z}{(1-q_z)^2}
+\sum_{n=1}^{\infty}\left(\frac{q_\tau^nq_z}{(1-q_\tau^nq_z)^2}+\frac{q_\tau^nq_z^{-1}}{(1-q_\tau^nq_z^{-1})^2}\right)-2\sum_{n=1}^{\infty}\frac{nq_\tau^n}{1-q_\tau^n}\quad(\tau\in\mathbb{H},\ z\in\mathbb{C}). \tag*{(55)} ( 2 π i ) 2 1 ℘ ( z ; [ τ , 1 ]) = 12 1 + ( 1 − q z ) 2 q z + n = 1 ∑ ∞ ( ( 1 − q τ n q z ) 2 q τ n q z + ( 1 − q τ n q z − 1 ) 2 q τ n q z − 1 ) − 2 n = 1 ∑ ∞ 1 − q τ n n q τ n ( τ ∈ H , z ∈ C ) . (55)
(i) Let τ ∈ H \tau\in\mathbb{H} τ ∈ H with Im ( τ ) ≥ 3 / 2 \operatorname{Im}(\tau)\ge\sqrt{3}/2 Im ( τ ) ≥ 3 /2 . Take u = [ u 1 u 2 ] , v = [ v 1 v 2 ] ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{u}=[u_1\ \ u_2],\ \mathbf{v}=[v_1\ \ v_2]\in\frac{1}{N}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z}) u = [ u 1 u 2 ] , v = [ v 1 v 2 ] ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) such that
{ 0 ≤ u 1 ≤ 1 2 and u = w or − w ( m o d M 1 , 2 ( Z ) ) , 0 ≤ v 1 ≤ 1 2 and v = t w or − t w ( m o d M 1 , 2 ( Z ) ) . (56) \begin{cases}
0\le u_1\le\frac{1}{2}\quad\text{and}\quad\mathbf{u}=\mathbf{w}\ \text{or}\ -\mathbf{w}\pmod{M_{1,2}(\mathbb{Z})},\\
0\le v_1\le\frac{1}{2}\quad\text{and}\quad\mathbf{v}=t\mathbf{w}\ \text{or}\ -t\mathbf{w}\pmod{M_{1,2}(\mathbb{Z})}. \tag*{(56)}
\end{cases} { 0 ≤ u 1 ≤ 2 1 and u = w or − w ( mod M 1 , 2 ( Z )) , 0 ≤ v 1 ≤ 2 1 and v = t w or − t w ( mod M 1 , 2 ( Z )) . (56)
Put ρ : = ∣ q τ ∣ \rho:=|q_\tau| ρ := ∣ q τ ∣ . Since τ ∈ H \tau\in\mathbb{H} τ ∈ H , we see that 0 < ρ < 1 0<\rho<1 0 < ρ < 1 . We then obtain by (55) and (56) that
1 ( 2 π ) 2 ∣ ℘ ( t ( w 1 τ + w 2 ) ; [ τ , 1 ] ) − ℘ ( w 1 τ + w 2 ; [ τ , 1 ] ) ∣ = 1 ( 2 π ) 2 ∣ ℘ ( v 1 τ + v 2 ; [ τ , 1 ] ) − ℘ ( u 1 τ + u 2 ; [ τ , 1 ] ) ∣ \begin{aligned}
\frac{1}{(2\pi)^2}\left|\wp(t(w_1\tau+w_2);[\tau,1])-\wp(w_1\tau+w_2;[\tau,1])\right| \\
&=\frac{1}{(2\pi)^2}\left|\wp(v_1\tau+v_2;[\tau,1])-\wp(u_1\tau+u_2;[\tau,1])\right|
\end{aligned} ( 2 π ) 2 1 ∣ ℘ ( t ( w 1 τ + w 2 ) ; [ τ , 1 ]) − ℘ ( w 1 τ + w 2 ; [ τ , 1 ]) ∣ = ( 2 π ) 2 1 ∣ ℘ ( v 1 τ + v 2 ; [ τ , 1 ]) − ℘ ( u 1 τ + u 2 ; [ τ , 1 ]) ∣
by evenness and periodicity of the Weierstrass ℘ \wp ℘ -function (see [5 Lemma 10.4] )
≤ ∣ q z ′ ( 1 − q z ′ ) 2 ∣ + ∣ q z ( 1 − q z ) 2 ∣ + ∑ n = 1 ∞ ( ρ n + v 1 ( 1 − ρ n + v 1 ) 2 + ρ n − v 1 ( 1 − ρ n − v 1 ) 2 ) + ∑ n = 1 ∞ ( ρ n + u 1 ( 1 − ρ n + u 1 ) 2 + ρ n − u 1 ( 1 − ρ n − u 1 ) 2 ) (57) \le\left|\frac{q_{z'}}{(1-q_{z'})^2}\right|+\left|\frac{q_z}{(1-q_z)^2}\right|
+\sum_{n=1}^{\infty}\left(\frac{\rho^{n+v_1}}{(1-\rho^{n+v_1})^2}+\frac{\rho^{n-v_1}}{(1-\rho^{n-v_1})^2}\right)
+\sum_{n=1}^{\infty}\left(\frac{\rho^{n+u_1}}{(1-\rho^{n+u_1})^2}+\frac{\rho^{n-u_1}}{(1-\rho^{n-u_1})^2}\right) \tag*{(57)} ≤ ( 1 − q z ′ ) 2 q z ′ + ( 1 − q z ) 2 q z + n = 1 ∑ ∞ ( ( 1 − ρ n + v 1 ) 2 ρ n + v 1 + ( 1 − ρ n − v 1 ) 2 ρ n − v 1 ) + n = 1 ∑ ∞ ( ( 1 − ρ n + u 1 ) 2 ρ n + u 1 + ( 1 − ρ n − u 1 ) 2 ρ n − u 1 ) (57)
where z ′ = v 1 τ + v 2 z'=v_1\tau+v_2 z ′ = v 1 τ + v 2 and z = u 1 τ + u 2 z=u_1\tau+u_2 z = u 1 τ + u 2
≤ ∣ q z ′ ( 1 − q z ′ ) 2 ∣ + ∣ q z ( 1 − q z ) 2 ∣ + 2 ∑ n = 1 ∞ ( ρ n ( 1 − ρ n ) 2 + ρ n − 1 / 2 ( 1 − ρ n − 1 / 2 ) 2 ) . \le\left|\frac{q_{z'}}{(1-q_{z'})^2}\right|+\left|\frac{q_z}{(1-q_z)^2}\right|+2\sum_{n=1}^{\infty}\left(\frac{\rho^n}{(1-\rho^n)^2}+\frac{\rho^{n-1/2}}{(1-\rho^{n-1/2})^2}\right). ≤ ( 1 − q z ′ ) 2 q z ′ + ( 1 − q z ) 2 q z + 2 n = 1 ∑ ∞ ( ( 1 − ρ n ) 2 ρ n + ( 1 − ρ n − 1/2 ) 2 ρ n − 1/2 ) .
Observe that if u 1 = 0 u_1=0 u 1 = 0 , then
∣ q z ( 1 − q z ) 2 ∣ = 1 4 sin 2 ( π u 2 ) ≤ 1 4 sin 2 ( π / N ) ≤ N 2 32 due to sin x ≥ 2 2 π x ( 0 ≤ x ≤ π 4 ) and N > 4 , \begin{aligned}
\left|\frac{q_z}{(1-q_z)^2}\right|=\frac{1}{4\sin^2(\pi u_2)} \\
&\le\frac{1}{4\sin^2(\pi/N)} \\
&\le\frac{N^2}{32}\quad\text{due to }\sin x\ge\frac{2\sqrt{2}}{\pi}x\ \left(0\le x\le\frac{\pi}{4}\right)\text{ and }N>4,
\end{aligned} ( 1 − q z ) 2 q z = 4 sin 2 ( π u 2 ) 1 ≤ 4 sin 2 ( π / N ) 1 ≤ 32 N 2 due to sin x ≥ π 2 2 x ( 0 ≤ x ≤ 4 π ) and N > 4 ,
and if u 1 ≠ 0 u_1\ne0 u 1 = 0 , then
∣ q z ( 1 − q z ) 2 ∣ ≤ ∣ q z ∣ ( 1 − ∣ q z ∣ ) 2 = 1 4 sinh 2 ( π y ) where y = Im ( z ) < 1 ( 2 π y ) 2 since sinh x > x ( x > 0 ) ≤ N 2 3 π 2 because y = u 1 Im ( τ ) ≥ 3 2 N < N 2 25 . \begin{aligned}
\left|\frac{q_z}{(1-q_z)^2}\right|\le\frac{|q_z|}{(1-|q_z|)^2} \\
&=\frac{1}{4\sinh^2(\pi y)}\quad\text{where }y=\operatorname{Im}(z) \\
&<\frac{1}{(2\pi y)^2}\quad\text{since }\sinh x>x\ (x>0) \\
&\le\frac{N^2}{3\pi^2}\quad\text{because }y=u_1\operatorname{Im}(\tau)\ge\frac{\sqrt{3}}{2N} \\
&<\frac{N^2}{25}.
\end{aligned} ( 1 − q z ) 2 q z ≤ ( 1 − ∣ q z ∣ ) 2 ∣ q z ∣ = 4 sinh 2 ( π y ) 1 where y = Im ( z ) < ( 2 π y ) 2 1 since sinh x > x ( x > 0 ) ≤ 3 π 2 N 2 because y = u 1 Im ( τ ) ≥ 2 N 3 < 25 N 2 .
Hence it follows that
∣ q z ( 1 − q z ) 2 ∣ < N 2 25 , similarly ∣ q z ′ ( 1 − q z ′ ) 2 ∣ < N 2 25 . (58) \left|\frac{q_z}{(1-q_z)^2}\right|<\frac{N^2}{25},\qquad\text{similarly}\qquad\left|\frac{q_{z'}}{(1-q_{z'})^2}\right|<\frac{N^2}{25}. \tag*{(58)} ( 1 − q z ) 2 q z < 25 N 2 , similarly ( 1 − q z ′ ) 2 q z ′ < 25 N 2 . (58)
Furthermore, we find that
∑ n = 1 ∞ ( ρ n ( 1 − ρ n ) 2 + ρ n − 1 / 2 ( 1 − ρ n − 1 / 2 ) 2 ) ≤ ∑ n = 1 ∞ ( ρ n ( 1 − ρ ) 2 + ρ n − 1 / 2 ( 1 − ρ 1 / 2 ) 2 ) ≤ ρ ( 1 − ρ ) 3 + ρ 1 / 2 ( 1 − ρ ) ( 1 − ρ 1 / 2 ) 2 < 1 10 because Im ( τ ) ≥ 3 / 2 < N 2 100 since N ≥ 5. (59) \begin{aligned}
\sum_{n=1}^{\infty}\left(\frac{\rho^n}{(1-\rho^n)^2}+\frac{\rho^{n-1/2}}{(1-\rho^{n-1/2})^2}\right)\leq\sum_{n=1}^{\infty}\left(\frac{\rho^n}{(1-\rho)^2}+\frac{\rho^{n-1/2}}{(1-\rho^{1/2})^2}\right) \\
&\leq\frac{\rho}{(1-\rho)^3}+\frac{\rho^{1/2}}{(1-\rho)(1-\rho^{1/2})^2} \\
&<\frac{1}{10}\quad\text{because }\operatorname{Im}(\tau)\geq\sqrt{3}/2 \\
&<\frac{N^2}{100}\quad\text{since }N\geq5. \tag*{(59)}
\end{aligned} n = 1 ∑ ∞ ( ( 1 − ρ n ) 2 ρ n + ( 1 − ρ n − 1/2 ) 2 ρ n − 1/2 ) ≤ n = 1 ∑ ∞ ( ( 1 − ρ ) 2 ρ n + ( 1 − ρ 1/2 ) 2 ρ n − 1/2 ) ≤ ( 1 − ρ ) 3 ρ + ( 1 − ρ ) ( 1 − ρ 1/2 ) 2 ρ 1/2 < 10 1 because Im ( τ ) ≥ 3 /2 < 100 N 2 since N ≥ 5. (59)
Therefore, it follows from (57) –(59) that
∣ ℘ ( t ( w 1 τ + w 2 ) ; [ τ , 1 ] ) − ℘ ( w 1 τ + w 2 ; [ τ , 1 ] ) ∣ < ( 2 π ) 2 N 2 10 . \left|\wp(t(w_1\tau+w_2);[\tau,1])-\wp(w_1\tau+w_2;[\tau,1])\right|<\frac{(2\pi)^2N^2}{10}. ∣ ℘ ( t ( w 1 τ + w 2 ) ; [ τ , 1 ]) − ℘ ( w 1 τ + w 2 ; [ τ , 1 ]) ∣ < 10 ( 2 π ) 2 N 2 .
(ii) Put
r : = ∣ q τ K ∣ 1 / N = e − π ∣ d K ∣ / N and z : = τ K + s N . r:=|q_{\tau_K}|^{1/N}=e^{-\pi\sqrt{|d_K|}/N}\quad\text{and}\quad z:=\frac{\tau_K+s}{N}. r := ∣ q τ K ∣ 1/ N = e − π ∣ d K ∣ / N and z := N τ K + s .
The assumption ∣ d K ∣ ≥ 15 |d_K|\geq15 ∣ d K ∣ ≥ 15 implies
r ≤ e − π 15 / N < { 3 / 8 if N = 12 , 1 / 4 if N = 8 , 1 / 10 if N = 5. (60) r\leq e^{-\pi\sqrt{15}/N}<\begin{cases}
3/8 & \text{if }N=12,\\
1/4 & \text{if }N=8,\\
1/10 & \text{if }N=5.
\end{cases} \tag*{(60)} r ≤ e − π 15 / N < ⎩ ⎨ ⎧ 3/8 1/4 1/10 if N = 12 , if N = 8 , if N = 5. (60)
Since 0 < 1 / N < t / N < 1 0<1/N<t/N<1 0 < 1/ N < t / N < 1 , we see that neither z z z nor t z tz t z belongs to [ τ K , 1 ] [\tau_K,1] [ τ K , 1 ] . We then derive by (55) that
1 ( 2 π ) 2 ∣ ℘ ( t z ; [ τ K , 1 ] ) − ℘ ( z ; [ τ K , 1 ] ) ∣ ≥ ∣ q z ( 1 − q z ) 2 ∣ − ∣ q t z ( 1 − q t z ) 2 ∣ − T ≥ r ( 1 + r ) 2 − r t ( 1 − r t ) 2 − T , (61) \begin{aligned}
\frac{1}{(2\pi)^2}\left|\wp(tz;[\tau_K,1])-\wp(z;[\tau_K,1])\right|\geq\left|\frac{q_z}{(1-q_z)^2}\right|-\left|\frac{q_{tz}}{(1-q_{tz})^2}\right|-T \\
&\geq\frac{r}{(1+r)^2}-\frac{r^t}{(1-r^t)^2}-T, \tag*{(61)}
\end{aligned} ( 2 π ) 2 1 ∣ ℘ ( t z ; [ τ K , 1 ]) − ℘ ( z ; [ τ K , 1 ]) ∣ ≥ ( 1 − q z ) 2 q z − ( 1 − q t z ) 2 q t z − T ≥ ( 1 + r ) 2 r − ( 1 − r t ) 2 r t − T , (61)
where
T = ∑ n = 1 ∞ ( r N n + 1 ( 1 − r N n + 1 ) 2 + r N n − 1 ( 1 − r N n − 1 ) 2 + r N n + t ( 1 − r N n + t ) 2 + r N n − t ( 1 − r N n − t ) 2 ) . T=\sum_{n=1}^{\infty}\left(\frac{r^{Nn+1}}{(1-r^{Nn+1})^2}+\frac{r^{Nn-1}}{(1-r^{Nn-1})^2}+\frac{r^{Nn+t}}{(1-r^{Nn+t})^2}+\frac{r^{Nn-t}}{(1-r^{Nn-t})^2}\right). T = n = 1 ∑ ∞ ( ( 1 − r N n + 1 ) 2 r N n + 1 + ( 1 − r N n − 1 ) 2 r N n − 1 + ( 1 − r N n + t ) 2 r N n + t + ( 1 − r N n − t ) 2 r N n − t ) .
Meanwhile we also find that
T ≤ 1 ( 1 − r N − t ) 2 { ∑ n = 1 ∞ r N n + 1 + ∑ n = 1 ∞ r N n − 1 + ∑ n = 1 ∞ r N n + t + ∑ n = 1 ∞ r N n − t } = r N + 1 + r N − 1 + r N + t + r N − t ( 1 − r N − t ) 2 ( 1 − r N ) = r N − t ( 1 + r t − 1 ) ( 1 + r t + 1 ) ( 1 − r N − t ) 2 ( 1 − r N ) < 2 r N − t ( 1 − r N − t ) 2 ( 1 − r N ) since ( 1 + r t − 1 ) ( 1 + r t + 1 ) < 2 (62) \begin{aligned}
T\leq\frac{1}{(1-r^{N-t})^2}\left\{\sum_{n=1}^{\infty}r^{Nn+1}+\sum_{n=1}^{\infty}r^{Nn-1}+\sum_{n=1}^{\infty}r^{Nn+t}+\sum_{n=1}^{\infty}r^{Nn-t}\right\} \\
&=\frac{r^{N+1}+r^{N-1}+r^{N+t}+r^{N-t}}{(1-r^{N-t})^2(1-r^N)} \\
&=\frac{r^{N-t}(1+r^{t-1})(1+r^{t+1})}{(1-r^{N-t})^2(1-r^N)} \\
&<\frac{2r^{N-t}}{(1-r^{N-t})^2(1-r^N)}\quad\text{since }(1+r^{t-1})(1+r^{t+1})<2 \tag*{(62)}
\end{aligned} T ≤ ( 1 − r N − t ) 2 1 { n = 1 ∑ ∞ r N n + 1 + n = 1 ∑ ∞ r N n − 1 + n = 1 ∑ ∞ r N n + t + n = 1 ∑ ∞ r N n − t } = ( 1 − r N − t ) 2 ( 1 − r N ) r N + 1 + r N − 1 + r N + t + r N − t = ( 1 − r N − t ) 2 ( 1 − r N ) r N − t ( 1 + r t − 1 ) ( 1 + r t + 1 ) < ( 1 − r N − t ) 2 ( 1 − r N ) 2 r N − t since ( 1 + r t − 1 ) ( 1 + r t + 1 ) < 2 (62)
owing to (67) and the fact t ≥ 2 t\geq2 t ≥ 2 . Thus we achieve by (61) and (62) that
1 ( 2 π ) 2 ∣ ℘ ( t z ; [ τ K , 1 ] ) − ℘ ( z ; [ τ K , 1 ] ) ∣ > r ( 1 ( 1 + r ) 2 − r t − 1 ( 1 − r t ) 2 − 2 r N − t − 1 ( 1 − r N − t ) 2 ( 1 − r N ) ) > r ⋅ 1 2 \begin{aligned}
\frac{1}{(2\pi)^2}\left|\wp(tz;[\tau_K,1])-\wp(z;[\tau_K,1])\right| \\
&>r\left(\frac{1}{(1+r)^2}-\frac{r^{t-1}}{(1-r^t)^2}-\frac{2r^{N-t-1}}{(1-r^{N-t})^2(1-r^N)}\right) \\
&>r\cdot\frac{1}{2}
\end{aligned} ( 2 π ) 2 1 ∣ ℘ ( t z ; [ τ K , 1 ]) − ℘ ( z ; [ τ K , 1 ]) ∣ > r ( ( 1 + r ) 2 1 − ( 1 − r t ) 2 r t − 1 − ( 1 − r N − t ) 2 ( 1 − r N ) 2 r N − t − 1 ) > r ⋅ 2 1
by (60) and the fact ( N , t ) ∈ { ( 12 , 5 ) , ( 8 , 3 ) , ( 5 , 2 ) } (N,t)\in\{(12,5),(8,3),(5,2)\} ( N , t ) ∈ {( 12 , 5 ) , ( 8 , 3 ) , ( 5 , 2 )} , and so
∣ ℘ ( t ( τ K + s ) N ; [ τ K , 1 ] ) − ℘ ( τ K + s N ; [ τ K , 1 ] ) ∣ > ( 2 π ) 2 2 ∣ q τ K ∣ 1 / N . \left|\wp\left(\frac{t(\tau_K+s)}{N};[\tau_K,1]\right)-\wp\left(\frac{\tau_K+s}{N};[\tau_K,1]\right)\right|>\frac{(2\pi)^2}{2}|q_{\tau_K}|^{1/N}. ℘ ( N t ( τ K + s ) ; [ τ K , 1 ] ) − ℘ ( N τ K + s ; [ τ K , 1 ] ) > 2 ( 2 π ) 2 ∣ q τ K ∣ 1/ N .
Theorem 7.3. Let K K K be an imaginary quadratic field, and let E E E and h E h_E h E be as in (1) –(3) . If m \mathfrak m m is a nonzero proper ideal of O K \mathcal O_K O K that is the exact conductor of K m / K K_{\mathfrak m}/K K m / K and is one of three cases p 2 2 , s p l i t p 3 , n o n i n e r t \mathfrak p_2^2,\mathrm{split}\mathfrak p_3,\mathrm{noninert} p 2 2 , split p 3 , noninert , p 2 3 , s p l i t \mathfrak p_2^3,\mathrm{split} p 2 3 , split , p 5 , n o n i n e r t \mathfrak p_5,\mathrm{noninert} p 5 , noninert , then we have
K m = K ( h E ( P ) ) K_{\mathfrak m}=K(h_E(P)) K m = K ( h E ( P )) for every primitive m \mathfrak m m -division point P ∈ E ( C ) P\in E(\mathbb C) P ∈ E ( C ) .
Proof. If h K = 1 h_K=1 h K = 1 , then the assertion follows from (4) . So we assume that
h K > 1 , and so ∣ d K ∣ ≥ 15 and O K × = { ± 1 } (63) h_K>1,\quad\text{and so }|d_K|\ge15\text{ and }\mathcal O_K^\times=\{\pm1\} \tag*{(63)} h K > 1 , and so ∣ d K ∣ ≥ 15 and O K × = { ± 1 } (63)
by (6) and (8) . Set
( N , t ) = { ( 12 , 5 ) if m = p 2 2 , s p l i t p 3 , n o n i n e r t , ( 8 , 3 ) if m = p 2 3 , s p l i t , ( 5 , 2 ) if m = p 5 , n o n i n e r t . (N,t)=
\begin{cases}
(12,5)&\text{if }\mathfrak m=\mathfrak p_2^2,\mathrm{split}\mathfrak p_3,\mathrm{noninert},\\
(8,3)&\text{if }\mathfrak m=\mathfrak p_2^3,\mathrm{split},\\
(5,2)&\text{if }\mathfrak m=\mathfrak p_5,\mathrm{noninert}.
\end{cases} ( N , t ) = ⎩ ⎨ ⎧ ( 12 , 5 ) ( 8 , 3 ) ( 5 , 2 ) if m = p 2 2 , split p 3 , noninert , if m = p 2 3 , split , if m = p 5 , noninert .
Since N ( m ) = N N(\mathfrak m)=N N ( m ) = N and m ∩ Z = N Z \mathfrak m\cap\mathbb Z=N\mathbb Z m ∩ Z = N Z , it follows from [4 Theorem 6.9 and Proposition 6.11] that
m ‾ = [ τ K + s , N ] for some s ∈ Z , \overline{\mathfrak m}=[\tau_K+s,N]\quad\text{for some }s\in\mathbb Z, m = [ τ K + s , N ] for some s ∈ Z ,
and so
m − 1 = 1 N m ‾ = [ z , 1 ] with z = τ K + s N . \mathfrak m^{-1}=\frac{1}{N}\overline{\mathfrak m}=[z,1]\quad\text{with }z=\frac{\tau_K+s}{N}. m − 1 = N 1 m = [ z , 1 ] with z = N τ K + s .
We then find that
Ann O K ( ϕ E ( z + O K ) ) = { a ∈ O K ∣ a z ∈ O K } = { a ∈ O K ∣ a m − 1 ⊆ O K } = m , \operatorname{Ann}_{\mathcal O_K}\bigl(\phi_E(z+\mathcal O_K)\bigr)
=\{a\in\mathcal O_K\mid az\in\mathcal O_K\}
=\{a\in\mathcal O_K\mid a\mathfrak m^{-1}\subseteq\mathcal O_K\}
=\mathfrak m, Ann O K ( ϕ E ( z + O K ) ) = { a ∈ O K ∣ a z ∈ O K } = { a ∈ O K ∣ a m − 1 ⊆ O K } = m ,
which claims that ϕ E ( z + O K ) \phi_E(z+\mathcal O_K) ϕ E ( z + O K ) is a primitive m \mathfrak m m -division point. Since the Weber values at primitive m \mathfrak m m -division points are Galois conjugates over H K H_K H K , it suffices to prove the assertion for P = ϕ E ( z + O K ) P=\phi_E(z+\mathcal O_K) P = ϕ E ( z + O K ) .
Let P = ϕ E ( z + O K ) P=\phi_E(z+\mathcal O_K) P = ϕ E ( z + O K ) and L = K ( h E ( P ) ) L=K(h_E(P)) L = K ( h E ( P )) . Since K m / K K_{\mathfrak m}/K K m / K is an abelian extension, so is L / K L/K L / K . Suppose on the contrary that L L L is properly contained in K m K_{\mathfrak m} K m . Since ( t , N ) = 1 (t,N)=1 ( t , N ) = 1 , we get by Lemma 4.5 that
h E ( t P ) = h E ( P ) σ m ( [ t O K ] m ) ∈ L , h_E(tP)=h_E(P)^{\sigma_{\mathfrak m}([t\mathcal O_K]_{\mathfrak m})}\in L, h E ( tP ) = h E ( P ) σ m ([ t O K ] m ) ∈ L ,
and so
δ : = h E ( t P ) − h E ( P ) ∈ L . (64) \delta:= h_E(tP)-h_E(P) \in L. \tag*{(64)} δ := h E ( tP ) − h E ( P ) ∈ L . (64)
Here, we note by (23) , (40) , (63) and Lemma 7.2 (ii) that δ ≠ 0 \delta\ne0 δ = 0 . Take an element
ε ∈ Gal ( K m / L ) ∖ { id K m } , \varepsilon\in\operatorname{Gal}(K_m/L) \setminus\{\operatorname{id}_{K_m}\}, ε ∈ Gal ( K m / L ) ∖ { id K m } ,
and hence
δ ε = δ (65) \delta^\varepsilon=\delta \tag*{(65)} δ ε = δ (65)
by (64) . Furthermore, since H K L = K m H_KL=K_m H K L = K m by (4) , we see that
ε ∣ H K ≠ id ∣ H K . (66) \left.\varepsilon\right|_{H_K}\ne\left.\operatorname{id}\right|_{H_K}. \tag*{(66)} ε ∣ H K = id ∣ H K . (66)
Let C = σ m − 1 ( ε ) C=\sigma_m^{-1}(\varepsilon) C = σ m − 1 ( ε ) , and choose an ideal c \mathfrak c c of O K \mathcal O_K O K in C C C . Then we find that
h E ( P ) ε = f m ( C P ) σ m ( C ) by Remark 4.6 (i) = f m ( C P C ) by Proposition 4.4, (67) \begin{aligned}
h_E(P)^\varepsilon&=f_m(C_P)^{\sigma_m(C)} &&\text{by Remark 4.6 (i)}\\
&=f_m(C_PC) &&\text{by Proposition 4.4,} \tag*{(67)}
\end{aligned} h E ( P ) ε = f m ( C P ) σ m ( C ) = f m ( C P C ) by Remark 4.6 (i) by Proposition 4.4, (67)
and similarly
h E ( t P ) ε = ( h E ( P ) σ m ( [ t O K ] m ) ) σ m ( C ) by Lemma 4.5 = f m ( C P ) σ m ( [ t O K ] m C ) by Remark 4.6 (i) = f m ( [ t O K ] m C P C ) by Proposition 4.4. (68) \begin{aligned}
h_E(tP)^\varepsilon
&=\left(h_E(P)^{\sigma_m([t\mathcal O_K]_m)}\right)^{\sigma_m(C)}
&&\text{by Lemma 4.5}\\
&=f_m(C_P)^{\sigma_m([t\mathcal O_K]_mC)}
&&\text{by Remark 4.6 (i)}\\
&=f_m([t\mathcal O_K]_mC_PC)
&&\text{by Proposition 4.4.} \tag*{(68)}
\end{aligned} h E ( tP ) ε = ( h E ( P ) σ m ([ t O K ] m ) ) σ m ( C ) = f m ( C P ) σ m ([ t O K ] m C ) = f m ([ t O K ] m C P C ) by Lemma 4.5 by Remark 4.6 (i) by Proposition 4.4. (68)
Observe that C P C = [ z m c ] m C_PC=[z_m\mathfrak c]_m C P C = [ z m c ] m , and
m ( z m c ) − 1 = z m − 1 c − 1 = ν [ ω Q , 1 ] \mathfrak m(z_m\mathfrak c)^{-1}=z_m^{-1}\mathfrak c^{-1}=\nu[\omega_Q,1] m ( z m c ) − 1 = z m − 1 c − 1 = ν [ ω Q , 1 ]
for some ν ∈ K × \nu\in K^\times ν ∈ K × and Q = a x 2 + b x y + c y 2 ∈ Q r e d ( d K ) Q=ax^2+bxy+cy^2\in\mathcal Q_{\mathrm{red}}(d_K) Q = a x 2 + b x y + c y 2 ∈ Q red ( d K ) by (38). Here, Q Q Q is nonprincipal by (66) . Since z m c z_m\mathfrak c z m c is an ideal of O K \mathcal O_K O K and N ∈ m N\in\mathfrak m N ∈ m , there is a vector w = [ w 1 w 2 ] ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf w=[w_1\ \ w_2]\in\frac{1}{N}M_{1,2}(\mathbb Z)\setminus M_{1,2}(\mathbb Z) w = [ w 1 w 2 ] ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) such that
1 = w 1 ( ν ω Q ) + w 2 ν . 1=w_1(\nu\omega_Q)+w_2\nu. 1 = w 1 ( ν ω Q ) + w 2 ν .
Thus Definition 4.3 yields
f m ( C P C ) = f w ( ω Q ) . f_m(C_PC)=f_{\mathbf w}(\omega_Q). f m ( C P C ) = f w ( ω Q ) .
In a similar way, one derives
f m ( [ t O K ] m C P C ) = f t w ( ω Q ) , f_m([t\mathcal O_K]_mC_PC)=f_{t\mathbf w}(\omega_Q), f m ([ t O K ] m C P C ) = f t w ( ω Q ) ,
and hence
δ ε = f t w ( ω Q ) − f w ( ω Q ) (69) \delta^\varepsilon=f_{t\mathbf w}(\omega_Q)-f_{\mathbf w}(\omega_Q) \tag*{(69)} δ ε = f t w ( ω Q ) − f w ( ω Q ) (69)
by (67) and (68) .
Put
z Q : = w 1 ω Q + w 2 , ρ = ∣ q τ K ∣ = e − π ∣ d K ∣ , ρ Q : = ∣ q ω Q ∣ = ρ 1 / a . z_Q:=w_1\omega_Q+w_2,\qquad\rho=|q_{\tau_K}|=e^{-\pi\sqrt{|d_K|}},\qquad\rho_Q:=|q_{\omega_Q}|=\rho^{1/a}. z Q := w 1 ω Q + w 2 , ρ = ∣ q τ K ∣ = e − π ∣ d K ∣ , ρ Q := ∣ q ω Q ∣ = ρ 1/ a .
Since Q ∈ Q r e d ( d K ) Q\in\mathcal Q_{\mathrm{red}}(d_K) Q ∈ Q red ( d K ) is a nonprincipal reduced form, we attain by (36) that
a ≥ 2 and 0 < ρ < ρ Q ≤ e − π 3 < 1 200 . (70) a\ge2\quad\text{and}\quad0<\rho<\rho_Q\le e^{-\pi\sqrt{3}}<\frac{1}{200}. \tag*{(70)} a ≥ 2 and 0 < ρ < ρ Q ≤ e − π 3 < 200 1 . (70)
We find that
∣ δ ∣ = ∣ g 2 ( τ K ) g 3 ( τ K ) Δ ( τ K ) ∣ ∣ ℘ ( t z ; [ τ K , 1 ] ) − ℘ ( z ; [ τ K , 1 ] ) ∣ |\delta|=\left|\frac{g_2(\tau_K)g_3(\tau_K)}{\Delta(\tau_K)}\right|\left|\wp(tz;[\tau_K,1])-\wp(z;[\tau_K,1])\right| ∣ δ ∣ = Δ ( τ K ) g 2 ( τ K ) g 3 ( τ K ) ∣ ℘ ( t z ; [ τ K , 1 ]) − ℘ ( z ; [ τ K , 1 ]) ∣
by the definitions (71), (3) and (26)
> ( 2 π ) 2 2 ∣ g 2 ( τ K ) g 3 ( τ K ) Δ ( τ K ) ∣ ρ 1 / N > \frac{(2\pi)^2}{2}\left|\frac{g_2(\tau_K)g_3(\tau_K)}{\Delta(\tau_K)}\right|\rho^{1/N} > 2 ( 2 π ) 2 Δ ( τ K ) g 2 ( τ K ) g 3 ( τ K ) ρ 1/ N
by Lemma 7.2 (ii)
= ( 2 π ) 2 2 ∣ j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 2 30 3 24 Δ ( τ K ) ∣ 1 / 6 ρ 1 / N = \frac{(2\pi)^2}{2}\left|\frac{j(\tau_K)^2(j(\tau_K)-1728)^3}{2^{30}3^{24}\Delta(\tau_K)}\right|^{1/6}\rho^{1/N} = 2 ( 2 π ) 2 2 30 3 24 Δ ( τ K ) j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 1/6 ρ 1/ N
by the definitions (20) and (24). {#eq:78}
and
∣ δ ξ ∣ = ∣ g 2 ( ω Q ) g 3 ( ω Q ) Δ ( ω Q ) ∣ ∣ ℘ ( t z Q ; [ ω Q , 1 ] ) − ℘ ( z Q ; [ ω Q , 1 ] ) ∣ |\delta^\xi|=\left|\frac{g_2(\omega_Q)g_3(\omega_Q)}{\Delta(\omega_Q)}\right|\left|\wp(tz_Q;[\omega_Q,1])-\wp(z_Q;[\omega_Q,1])\right| ∣ δ ξ ∣ = Δ ( ω Q ) g 2 ( ω Q ) g 3 ( ω Q ) ∣ ℘ ( t z Q ; [ ω Q , 1 ]) − ℘ ( z Q ; [ ω Q , 1 ]) ∣
by (76), the definitions (3) and (26)
< ( 2 π ) 2 N 2 10 ∣ g 2 ( ω Q ) g 3 ( ω Q ) Δ ( ω Q ) ∣ < \frac{(2\pi)^2N^2}{10}\left|\frac{g_2(\omega_Q)g_3(\omega_Q)}{\Delta(\omega_Q)}\right| < 10 ( 2 π ) 2 N 2 Δ ( ω Q ) g 2 ( ω Q ) g 3 ( ω Q )
by Lemma 7.2 (i)
= ( 2 π ) 2 N 2 10 ∣ j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 2 30 3 24 Δ ( ω Q ) ∣ 1 / 6 = \frac{(2\pi)^2N^2}{10}\left|\frac{j(\omega_Q)^2(j(\omega_Q)-1728)^3}{2^{30}3^{24}\Delta(\omega_Q)}\right|^{1/6} = 10 ( 2 π ) 2 N 2 2 30 3 24 Δ ( ω Q ) j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 1/6
by the definitions (20) and (24). {#eq:79}
Furthermore, we obtain that
∣ Δ ( τ K ) Δ ( ω Q ) ∣ ≤ ρ ρ Q ∏ n = 1 ∞ ( 1 + ρ n 1 − ρ Q n ) 24 \left|\frac{\Delta(\tau_K)}{\Delta(\omega_Q)}\right|\leq\frac{\rho}{\rho_Q}\prod_{n=1}^{\infty}\left(\frac{1+\rho^n}{1-\rho_Q^n}\right)^{24} Δ ( ω Q ) Δ ( τ K ) ≤ ρ Q ρ n = 1 ∏ ∞ ( 1 − ρ Q n 1 + ρ n ) 24
by (23) of Lemma 3.1
≤ ρ ρ Q ∏ n = 1 ∞ ( 1 ( 1 − ρ n ) ( 1 − ρ Q n ) ) 24 \leq\frac{\rho}{\rho_Q}\prod_{n=1}^{\infty}\left(\frac{1}{(1-\rho^n)(1-\rho_Q^n)}\right)^{24} ≤ ρ Q ρ n = 1 ∏ ∞ ( ( 1 − ρ n ) ( 1 − ρ Q n ) 1 ) 24
since 1 + x ≤ 1 1 − x 1+x\leq\frac{1}{1-x} 1 + x ≤ 1 − x 1 (0 ≤ x < 1 0\leq x<1 0 ≤ x < 1 )
≤ ρ ρ Q ( 1 − ρ 1 − 2 ρ 1 − ρ Q 1 − 2 ρ Q ) 24 (71) \leq\frac{\rho}{\rho_Q}\left(\frac{1-\rho}{1-2\rho}\frac{1-\rho_Q}{1-2\rho_Q}\right)^{24} \tag*{(71)} ≤ ρ Q ρ ( 1 − 2 ρ 1 − ρ 1 − 2 ρ Q 1 − ρ Q ) 24 (71)
because ∏ n = 1 ∞ ( 1 − x n ) ≥ 1 − ∑ n = 1 ∞ x n \prod_{n=1}^{\infty}(1-x_n)\geq1-\sum_{n=1}^{\infty}x_n ∏ n = 1 ∞ ( 1 − x n ) ≥ 1 − ∑ n = 1 ∞ x n (x n ≥ 0 x_n\geq0 x n ≥ 0 with ∑ n = 1 ∞ x n < 1 \sum_{n=1}^{\infty}x_n<1 ∑ n = 1 ∞ x n < 1 )
< ρ 1 / 2 ( 199 198 ) 48 <\rho^{1/2}\left(\frac{199}{198}\right)^{48} < ρ 1/2 ( 198 199 ) 48
by (77)
< 2 ρ 1 / 2 . <2\rho^{1/2}. < 2 ρ 1/2 .
Now, we achieve by (72), (78), (79) and (80) that
1 = ∣ δ ξ δ ∣ 6 < 2 ( N 2 5 ) 6 ρ ( 1 / 2 ) − ( 6 / N ) ∣ j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 ∣ . (72) 1=\left|\frac{\delta^\xi}{\delta}\right|^6<2\left(\frac{N^2}{5}\right)^6\rho^{(1/2)-(6/N)}\left|\frac{j(\omega_Q)^2(j(\omega_Q)-1728)^3}{j(\tau_K)^2(j(\tau_K)-1728)^3}\right|. \tag*{(72)} 1 = δ δ ξ 6 < 2 ( 5 N 2 ) 6 ρ ( 1/2 ) − ( 6/ N ) j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 . (72)
Case 1. Suppose that ∣ d K ∣ ≥ 20 |d_K|\geq20 ∣ d K ∣ ≥ 20 . We then find that
1 < 2 ( N 2 5 ) 6 ρ ( 1 / 2 ) − ( 6 / N ) ∣ j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 ∣ 1<2\left(\frac{N^2}{5}\right)^6\rho^{(1/2)-(6/N)}\left|\frac{j(\omega_Q)^2(j(\omega_Q)-1728)^3}{j(\tau_K)^2(j(\tau_K)-1728)^3}\right| 1 < 2 ( 5 N 2 ) 6 ρ ( 1/2 ) − ( 6/ N ) j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3
by (81)
< 2 ( N 2 5 ) 6 ρ ( 1 / 2 ) − ( 6 / N ) ⋅ 877383 ρ 5 / 2 <2\left(\frac{N^2}{5}\right)^6\rho^{(1/2)-(6/N)}\cdot877383\,\rho^{5/2} < 2 ( 5 N 2 ) 6 ρ ( 1/2 ) − ( 6/ N ) ⋅ 877383 ρ 5/2
by Lemma 7.1
= 1754766 ( N 2 5 ) 6 ρ 3 − ( 6 / N ) =1754766\left(\frac{N^2}{5}\right)^6\rho^{3-(6/N)} = 1754766 ( 5 N 2 ) 6 ρ 3 − ( 6/ N )
≤ 1754766 ( N 2 5 ) 6 ( e − π 20 ) 3 − ( 6 / N ) due to ∣ d K ∣ ≥ 20 < 1 for every N = 12 , 8 , 5 , \begin{aligned}
\le1754766\left(\frac{N^2}{5}\right)^6\left(e^{-\pi\sqrt{20}}\right)^{3-\left(6/N\right)} \quad\text{due to } \lvert d_K\rvert\ge20 \\
&<1 \quad\text{for every } N=12,8,5,
\end{aligned} ≤ 1754766 ( 5 N 2 ) 6 ( e − π 20 ) 3 − ( 6/ N ) due to ∣ d K ∣ ≥ 20 < 1 for every N = 12 , 8 , 5 ,
which yields a contradiction.
Case 2. Suppose that d K = − 15 d_K=-15 d K = − 15 . In this case, we have Q r e d ( d K ) = { x 2 + x y + 4 y 2 , 2 x 2 + x y + 2 y 2 } Q_{\mathrm{red}}(d_K)=\{x^2+xy+4y^2,2x^2+xy+2y^2\} Q red ( d K ) = { x 2 + x y + 4 y 2 , 2 x 2 + x y + 2 y 2 } , in which Q = 2 x 2 + x y + 2 y 2 Q=2x^2+xy+2y^2 Q = 2 x 2 + x y + 2 y 2 is the unique nonprincipal form. Moreover, we have
j ( τ K ) = − 191025 − 85995 5 2 and j ( ω Q ) = − 191025 + 85995 5 2 j(\tau_K)=\frac{-191025-85995\sqrt{5}}{2}
\quad\text{and}\quad
j(\omega_Q)=\frac{-191025+85995\sqrt{5}}{2} j ( τ K ) = 2 − 191025 − 85995 5 and j ( ω Q ) = 2 − 191025 + 85995 5
(see [22 Example 6.2.2 in Chapter II] ), and estimate that
∣ j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 ∣ < 2 ⋅ 10 − 12 . (73) \left\lvert\frac{j(\omega_Q)^2(j(\omega_Q)-1728)^3}{j(\tau_K)^2(j(\tau_K)-1728)^3}\right\rvert<2\cdot10^{-12}. \tag*{(73)} j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 < 2 ⋅ 1 0 − 12 . (73)
So, we establish that
1 < 2 ( N 2 5 ) 6 ρ ( 1 / 2 ) − ( 6 / N ) ∣ j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 ∣ by (81) < 2 ( N 2 5 ) 6 ( e − π 15 ) ( 1 / 2 ) − ( 6 / N ) ( 2 ⋅ 10 − 12 ) by (82) < 1 for every N = 12 , 8 , 5 , \begin{aligned}
1<2\left(\frac{N^2}{5}\right)^6\rho^{(1/2)-(6/N)}
\left\lvert\frac{j(\omega_Q)^2(j(\omega_Q)-1728)^3}{j(\tau_K)^2(j(\tau_K)-1728)^3}\right\rvert
\quad\text{by (81)} \\
&<2\left(\frac{N^2}{5}\right)^6\left(e^{-\pi\sqrt{15}}\right)^{(1/2)-(6/N)}(2\cdot10^{-12})
\quad\text{by (82)} \\
&<1 \quad\text{for every } N=12,8,5,
\end{aligned} 1 < 2 ( 5 N 2 ) 6 ρ ( 1/2 ) − ( 6/ N ) j ( τ K ) 2 ( j ( τ K ) − 1728 ) 3 j ( ω Q ) 2 ( j ( ω Q ) − 1728 ) 3 by (81) < 2 ( 5 N 2 ) 6 ( e − π 15 ) ( 1/2 ) − ( 6/ N ) ( 2 ⋅ 1 0 − 12 ) by (82) < 1 for every N = 12 , 8 , 5 ,
which yields a contradiction too.
Therefore we conclude that L = K m L=K_m L = K m as desired.
Combining the preceding results with Proposition 2.1 yields Theorem A.
Theorem 7.4. Let K K K be an imaginary quadratic field, and let E E E and h E h_E h E be as in (1)–(3). If m \mathfrak m m is a nonzero proper ideal of O K \mathcal O_K O K that is the exact conductor of K m / K K_m/K K m / K , then
K m = K ( h E ( P ) ) K_m=K(h_E(P)) K m = K ( h E ( P )) for every primitive m \mathfrak m m -division point P ∈ E ( C ) P\in E(\mathbb C) P ∈ E ( C ) .
Proof. The case h K = 1 h_K=1 h K = 1 follows from (4). For h K > 1 h_K>1 h K > 1 , Sugawara’s criteria (S1)–(S3) and [8 , 11 ] treat the cases excluded in Proposition 2.1. And, the remaining seven families are covered by Theorems 6.2 and 7.3.
Generation of intermediate fields using relative norms
Starting from a Weber value that generates K m K_m K m over K K K , we construct an explicit integer translate whose relative norms generate every intermediate field.
For a nonzero proper ideal m \mathfrak m m of O K \mathcal O_K O K , let N m > 1 N_m>1 N m > 1 be the integer satisfying
m ∩ Z = N m Z . \mathfrak m\cap\mathbb Z=N_m\mathbb Z. m ∩ Z = N m Z .
We further define two integers A m A_m A m and B m B_m B m by
A m = 2 9 3 6 N m 6 and B m = ⌈ N m 6 e π ∣ d K ∣ ⌉ . A_m=2^9 3^6 N_m^6
\quad\text{and}\quad
B_m=\left\lceil N_m^6e^{\pi\sqrt{\lvert d_K\rvert}}\right\rceil. A m = 2 9 3 6 N m 6 and B m = ⌈ N m 6 e π ∣ d K ∣ ⌉ .
Lemma 8.1. Let N > 1 N>1 N > 1 be an integer and v ∈ 1 N M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{v}\in\frac{1}{N}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z}) v ∈ N 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) . Then the functions
2 7 3 5 N 2 f v ( 1 ) , 2 8 3 4 N 4 f v ( 2 ) , 2 9 3 6 N 6 f v ( 3 ) 2^{7}3^{5}N^{2}f_{\mathbf{v}}^{(1)},\qquad2^{8}3^{4}N^{4}f_{\mathbf{v}}^{(2)},\qquad2^{9}3^{6}N^{6}f_{\mathbf{v}}^{(3)} 2 7 3 5 N 2 f v ( 1 ) , 2 8 3 4 N 4 f v ( 2 ) , 2 9 3 6 N 6 f v ( 3 ) are all integral over Z [ j ] \mathbb{Z}[j] Z [ j ] .
Proof. It is well known that
the Fourier coefficients of 2 7 3 5 N 2 f v ( 1 ) ∘ γ with respect to q τ 1 / N \text{the Fourier coefficients of }2^{7}3^{5}N^{2}f_{\mathbf{v}}^{(1)}\circ\gamma\text{ with respect to }q_{\tau}^{1/N} the Fourier coefficients of 2 7 3 5 N 2 f v ( 1 ) ∘ γ with respect to q τ 1/ N are algebraic integers for every γ ∈ SL 2 ( Z ) \gamma\in\operatorname{SL}_{2}(\mathbb{Z}) γ ∈ SL 2 ( Z ) . (83)
(see [10 Lemma 5.6] and Proposition 3.3 (iii)). By the definitions (24) and (26), we find that
2 8 3 4 N 4 f v ( 2 ) = ( 2 7 3 5 N 2 f v ( 1 ) ) 2 j − 1728 and 2 9 3 6 N 6 f v ( 3 ) = ( 2 7 3 5 N 2 f v ( 1 ) ) 3 j ( j − 1728 ) . (74) 2^{8}3^{4}N^{4}f_{\mathbf{v}}^{(2)}=\frac{\left(2^{7}3^{5}N^{2}f_{\mathbf{v}}^{(1)}\right)^{2}}{j-1728}
\quad\text{and}\quad
2^{9}3^{6}N^{6}f_{\mathbf{v}}^{(3)}=\frac{\left(2^{7}3^{5}N^{2}f_{\mathbf{v}}^{(1)}\right)^{3}}{j(j-1728)}. \tag*{(74)} 2 8 3 4 N 4 f v ( 2 ) = j − 1728 ( 2 7 3 5 N 2 f v ( 1 ) ) 2 and 2 9 3 6 N 6 f v ( 3 ) = j ( j − 1728 ) ( 2 7 3 5 N 2 f v ( 1 ) ) 3 . (74) Observe from (25) that
q τ j , q τ ( j − 1728 ) ∈ 1 + q τ Z [ [ q τ ] ] , q_{\tau}j,\ q_{\tau}(j-1728)\in1+q_{\tau}\mathbb{Z}[[q_{\tau}]], q τ j , q τ ( j − 1728 ) ∈ 1 + q τ Z [[ q τ ]] , and so
1 j − 1728 ∈ q τ Z [ [ q τ ] ] and 1 j ( j − 1728 ) ∈ q τ 2 Z [ [ q τ ] ] . (75) \frac{1}{j-1728}\in q_{\tau}\mathbb{Z}[[q_{\tau}]]
\quad\text{and}\quad
\frac{1}{j(j-1728)}\in q_{\tau}^{2}\mathbb{Z}[[q_{\tau}]]. \tag*{(75)} j − 1728 1 ∈ q τ Z [[ q τ ]] and j ( j − 1728 ) 1 ∈ q τ 2 Z [[ q τ ]] . (75) Thus we get by the fact (83), the relations (84) and (85) that
the Fourier coefficients of 2 8 3 4 N 4 f v ( 2 ) ∘ γ and 2 9 3 6 N 6 f v ( 3 ) ∘ γ with respect to q τ 1 / N \text{the Fourier coefficients of }2^{8}3^{4}N^{4}f_{\mathbf{v}}^{(2)}\circ\gamma\text{ and }2^{9}3^{6}N^{6}f_{\mathbf{v}}^{(3)}\circ\gamma\text{ with respect to }q_{\tau}^{1/N} the Fourier coefficients of 2 8 3 4 N 4 f v ( 2 ) ∘ γ and 2 9 3 6 N 6 f v ( 3 ) ∘ γ with respect to q τ 1/ N are algebraic integers for every γ ∈ SL 2 ( Z ) \gamma\in\operatorname{SL}_{2}(\mathbb{Z}) γ ∈ SL 2 ( Z ) . (86)
Therefore, the fact from [12 Lemma 2.1 in Chapter 2] together with (83), (86) and the holomorphy of f v ( k ) f_{\mathbf{v}}^{(k)} f v ( k ) (k = 1 , 2 , 3 k=1,2,3 k = 1 , 2 , 3 ) on H \mathbb{H} H lead us to conclude that the functions 2 7 3 5 N 2 f v ( 1 ) 2^{7}3^{5}N^{2}f_{\mathbf{v}}^{(1)} 2 7 3 5 N 2 f v ( 1 ) , 2 8 3 4 N 4 f v ( 2 ) 2^{8}3^{4}N^{4}f_{\mathbf{v}}^{(2)} 2 8 3 4 N 4 f v ( 2 ) and 2 9 3 6 N 6 f v ( 3 ) 2^{9}3^{6}N^{6}f_{\mathbf{v}}^{(3)} 2 9 3 6 N 6 f v ( 3 ) are integral over Z [ j ] \mathbb{Z}[j] Z [ j ] .
For every positive integer N N N , we get by the theory of complex multiplication that
K N O K = K ( f ( τ K ) ∣ f ∈ F N is finite at τ K ) K_{N\mathcal{O}_{K}}=K(f(\tau_{K})\mid f\in\mathcal{F}_{N}\text{ is finite at }\tau_{K}) K N O K = K ( f ( τ K ) ∣ f ∈ F N is finite at τ K ) (see [13 Corollary to Theorem 2 in Chapter 10] ).
Lemma 8.2. If P P P is a primitive m \mathbf{m} m -division point of E ( C ) E(\mathbb{C}) E ( C ) , then A m h E ( P ) A_{\mathbf{m}}h_{E}(P) A m h E ( P ) is an algebraic integer.
Proof. Since O K = [ τ K , 1 ] \mathcal{O}_{K}=[\tau_{K},1] O K = [ τ K , 1 ] by (7) and N m ∈ m = Ann O K ( P ) N_{\mathbf{m}}\in\mathbf{m}=\operatorname{Ann}_{\mathcal{O}_{K}}(P) N m ∈ m = Ann O K ( P ) , we have
ϕ E − 1 ( P ) = ( v 1 τ K + v 2 ) + O K for some v = [ v 1 v 2 ] ∈ 1 N m M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) . \phi_{E}^{-1}(P)=(v_{1}\tau_{K}+v_{2})+\mathcal{O}_{K}\quad\text{for some }\mathbf{v}=\begin{bmatrix}v_{1}&v_{2}\end{bmatrix}\in\frac{1}{N_{\mathbf{m}}}M_{1,2}(\mathbb{Z})\setminus M_{1,2}(\mathbb{Z}). ϕ E − 1 ( P ) = ( v 1 τ K + v 2 ) + O K for some v = [ v 1 v 2 ] ∈ N m 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) . And, we see from the definitions (3) and (26) that
A m h E ( P ) = { 2 2 3 N m 4 ( 2 7 3 5 N m 2 f v ( 1 ) ( τ K ) ) , if j ( E ) ≠ 1728 , 0 , 2 1 3 2 N m 2 ( 2 8 3 4 N m 4 f v ( 2 ) ( τ K ) ) , if j ( E ) = 1728 , 2 0 3 0 N m 0 ( 2 9 3 6 N m 6 f v ( 3 ) ( τ K ) ) , if j ( E ) = 0. A_{\mathbf{m}}h_{E}(P)=
\begin{cases}
2^{2}3N_{\mathbf{m}}^{4}\left(2^{7}3^{5}N_{\mathbf{m}}^{2}f_{\mathbf{v}}^{(1)}(\tau_{K})\right), & \text{if }j(E)\ne1728,0,\\
2^{1}3^{2}N_{\mathbf{m}}^{2}\left(2^{8}3^{4}N_{\mathbf{m}}^{4}f_{\mathbf{v}}^{(2)}(\tau_{K})\right), & \text{if }j(E)=1728,\\
2^{0}3^{0}N_{\mathbf{m}}^{0}\left(2^{9}3^{6}N_{\mathbf{m}}^{6}f_{\mathbf{v}}^{(3)}(\tau_{K})\right), & \text{if }j(E)=0.
\end{cases} A m h E ( P ) = ⎩ ⎨ ⎧ 2 2 3 N m 4 ( 2 7 3 5 N m 2 f v ( 1 ) ( τ K ) ) , 2 1 3 2 N m 2 ( 2 8 3 4 N m 4 f v ( 2 ) ( τ K ) ) , 2 0 3 0 N m 0 ( 2 9 3 6 N m 6 f v ( 3 ) ( τ K ) ) , if j ( E ) = 1728 , 0 , if j ( E ) = 1728 , if j ( E ) = 0. Hence the value A m h E ( P ) A_{\mathbf{m}}h_{E}(P) A m h E ( P ) is an algebraic integer by Lemma 8.1 and the well-known fact that the singular modulus j ( τ K ) j(\tau_{K}) j ( τ K ) is an algebraic integer (see [13 Theorem 4 in Chapter 5] ).
Lemma 8.3. If P P P is a primitive m m m -division point of E ( C ) E(\mathbb{C}) E ( C ) , then
∣ ι ( h E ( P ) ) ∣ ≤ B m | \iota(h_E(P))| \leq\mathcal{B}_m ∣ ι ( h E ( P )) ∣ ≤ B m for every embedding ι : K m ↪ C \iota: K_m \hookrightarrow\mathbb{C} ι : K m ↪ C .
Proof. Put k = ∣ O K × ∣ / 2 k = |\mathcal{O}_K^\times|/2 k = ∣ O K × ∣/2 , and so
k = { 1 if K ≠ Q ( − 1 ) , Q ( − 3 ) , 2 if K = Q ( − 1 ) , 3 if K = Q ( − 3 ) . k =
\begin{cases}
1 & \text{if } K \neq\mathbb{Q}(\sqrt{-1}), \mathbb{Q}(\sqrt{-3}),\\
2 & \text{if } K = \mathbb{Q}(\sqrt{-1}),\\
3 & \text{if } K = \mathbb{Q}(\sqrt{-3}).
\end{cases} k = ⎩ ⎨ ⎧ 1 2 3 if K = Q ( − 1 ) , Q ( − 3 ) , if K = Q ( − 1 ) , if K = Q ( − 3 ) .
by (8). Let ι : K m → C \iota: K_m \to\mathbb{C} ι : K m → C be an embedding. Since K K K is imaginary quadratic, ι ∣ K \iota|_K ι ∣ K is either the identity or the complex conjugation. In the second case, the map
ι ′ : K m → C , x ↦ ι ( x ) ‾ \iota': K_m \to\mathbb{C}, \qquad x \mapsto\overline{\iota(x)} ι ′ : K m → C , x ↦ ι ( x )
fixes K K K and
∣ ι ( h E ( P ) ) ∣ = ∣ ι ( h E ( P ) ) ‾ ∣ = ∣ ι ′ ( h E ( P ) ) ∣ . |\iota(h_E(P))| = |\overline{\iota(h_E(P))}| = |\iota'(h_E(P))|. ∣ ι ( h E ( P )) ∣ = ∣ ι ( h E ( P )) ∣ = ∣ ι ′ ( h E ( P )) ∣.
Thus we may assume that ι \iota ι fixes K K K . By Lemma 4.7, it suffices to show that
∣ f w ( k ) ( ω Q ) ∣ ≤ B m |f_{\mathbf{w}}^{(k)}(\omega_Q)| \leq\mathcal{B}_m ∣ f w ( k ) ( ω Q ) ∣ ≤ B m
for all w = [ w 1 w 2 ] ∈ 1 N m M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{w} = \begin{bmatrix} w_1 & w_2 \end{bmatrix} \in\frac{1}{N_m}M_{1,2}(\mathbb{Z}) \setminus M_{1,2}(\mathbb{Z}) w = [ w 1 w 2 ] ∈ N m 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) with 0 ≤ w 1 ≤ 1 / 2 0 \leq w_1 \leq1/2 0 ≤ w 1 ≤ 1/2 and for all Q ∈ Q r e d ( d K ) Q \in\mathcal{Q}_{\mathrm{red}}(d_K) Q ∈ Q red ( d K ) .
By Lemma 3.2, we obtain the expansion formula
1 ( 2 π i ) 2 ℘ ( z ; [ τ , 1 ] ) = 1 12 + q z ( 1 − q z ) 2 + ∑ n = 1 ∞ ( q τ n q z ( 1 − q τ n q z ) 2 + q τ n q z − 1 ( 1 − q τ n q z − 1 ) 2 ) − 2 ∑ n = 1 ∞ n q τ n 1 − q τ n ( z ∈ C , τ ∈ H ) . (76) \frac{1}{(2\pi i)^2}\wp(z;[\tau,1]) =
\frac{1}{12}+\frac{q_z}{(1-q_z)^2}
+\sum_{n=1}^{\infty}\left(\frac{q_\tau^nq_z}{(1-q_\tau^nq_z)^2}+\frac{q_\tau^nq_z^{-1}}{(1-q_\tau^nq_z^{-1})^2}\right)-2\sum_{n=1}^{\infty}\frac{nq_\tau^n}{1-q_\tau^n}
\qquad(z\in\mathbb{C},\ \tau\in\mathbb{H}). \tag*{(76)} ( 2 π i ) 2 1 ℘ ( z ; [ τ , 1 ]) = 12 1 + ( 1 − q z ) 2 q z + n = 1 ∑ ∞ ( ( 1 − q τ n q z ) 2 q τ n q z + ( 1 − q τ n q z − 1 ) 2 q τ n q z − 1 ) − 2 n = 1 ∑ ∞ 1 − q τ n n q τ n ( z ∈ C , τ ∈ H ) . (76)
Let w = [ w 1 w 2 ] ∈ 1 N m M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) \mathbf{w} = \begin{bmatrix} w_1 & w_2 \end{bmatrix} \in\frac{1}{N_m}M_{1,2}(\mathbb{Z}) \setminus M_{1,2}(\mathbb{Z}) w = [ w 1 w 2 ] ∈ N m 1 M 1 , 2 ( Z ) ∖ M 1 , 2 ( Z ) with 0 ≤ w 1 ≤ 1 / 2 0 \leq w_1 \leq1/2 0 ≤ w 1 ≤ 1/2 , and let Q = a x 2 + b x y + c y 2 ∈ Q r e d ( d K ) Q=ax^2+bxy+cy^2 \in\mathcal{Q}_{\mathrm{red}}(d_K) Q = a x 2 + b x y + c y 2 ∈ Q red ( d K ) . Set
z Q : = w 1 ω Q + w 2 and y Q : = Im ( ω Q ) = ∣ d K ∣ 2 a . z_Q := w_1\omega_Q+w_2
\qquad\text{and}\qquad
y_Q := \operatorname{Im}(\omega_Q)=\frac{\sqrt{|d_K|}}{2a}. z Q := w 1 ω Q + w 2 and y Q := Im ( ω Q ) = 2 a ∣ d K ∣ .
Then we know by (37) that
y Q ≥ 3 2 . (77) y_Q \geq\frac{\sqrt{3}}{2}. \tag*{(77)} y Q ≥ 2 3 . (77)
If w 1 = 0 w_1=0 w 1 = 0 , then we attain that w 2 ∉ Z w_2\notin\mathbb{Z} w 2 ∈ / Z and
∣ 1 − q z Q ∣ = 2 ∣ sin ( π w 2 ) ∣ ≥ 2 sin π N m ≥ 4 N m since sin ( π x ) ≥ 2 x ( 0 ≤ x ≤ 1 / 2 ) , |1-q_{z_Q}|=2|\sin(\pi w_2)|\geq2\sin\frac{\pi}{N_m}\geq\frac{4}{N_m}
\qquad\text{since }\sin(\pi x)\geq2x\ (0\leq x\leq1/2), ∣1 − q z Q ∣ = 2∣ sin ( π w 2 ) ∣ ≥ 2 sin N m π ≥ N m 4 since sin ( π x ) ≥ 2 x ( 0 ≤ x ≤ 1/2 ) ,
and so
∣ q z Q ( 1 − q z Q ) 2 ∣ ≤ N m 2 16 . (78) \left|\frac{q_{z_Q}}{(1-q_{z_Q})^2}\right|\leq\frac{N_m^2}{16}. \tag*{(78)} ( 1 − q z Q ) 2 q z Q ≤ 16 N m 2 . (78)
If 0 < w 1 ≤ 1 / 2 0<w_1\leq1/2 0 < w 1 ≤ 1/2 , then
∣ q Z Q ( 1 − q Z Q ) 2 ∣ ≤ ∣ q Z Q ∣ ( 1 − ∣ q Z Q ∣ ) 2 = 1 4 sinh 2 ( π y Q w 1 ) < 1 4 ( π y Q w 1 ) 2 because sinh x > x ( x > 0 ) < N m 2 16 by the fact w 1 ≥ 1 N m and (88). (79) \left|\frac{q_{Z_Q}}{(1-q_{Z_Q})^2}\right|\leq\frac{|q_{Z_Q}|}{(1-|q_{Z_Q}|)^2}
=\frac{1}{4\sinh^2(\pi y_Qw_1)}
<\frac{1}{4(\pi y_Qw_1)^2}\quad\text{because }\sinh x>x\ (x>0)
<\frac{N_m^2}{16}\quad\text{by the fact }w_1\geq\frac{1}{N_m}\text{ and (88).} \tag*{(79)} ( 1 − q Z Q ) 2 q Z Q ≤ ( 1 − ∣ q Z Q ∣ ) 2 ∣ q Z Q ∣ = 4 sinh 2 ( π y Q w 1 ) 1 < 4 ( π y Q w 1 ) 2 1 because sinh x > x ( x > 0 ) < 16 N m 2 by the fact w 1 ≥ N m 1 and (88). (79)
And if we let ρ : = ∣ q ω Q ∣ \rho:=|q_{\omega_Q}| ρ := ∣ q ω Q ∣ , then we get by (88) that
ρ = e − 2 π y Q ≤ e − π 3 < 1 200 . (80) \rho=e^{-2\pi y_Q}\leq e^{-\pi\sqrt{3}}<\frac{1}{200}. \tag*{(80)} ρ = e − 2 π y Q ≤ e − π 3 < 200 1 . (80)
Since 0 ≤ w 1 ≤ 1 / 2 0\leq w_1\leq1/2 0 ≤ w 1 ≤ 1/2 , we establish that
∑ n = 1 ∞ ∣ q ω Q n q Z Q ( 1 − q ω Q n q Z Q ) 2 ∣ ≤ ∑ n = 1 ∞ ρ n + w 1 ( 1 − ρ n + w 1 ) 2 ≤ ∑ n = 1 ∞ ρ n ( 1 − ρ ) 2 = ρ ( 1 − ρ ) 3 , (81) \sum_{n=1}^{\infty}\left|\frac{q_{\omega_Q}^nq_{Z_Q}}{(1-q_{\omega_Q}^nq_{Z_Q})^2}\right|
\leq\sum_{n=1}^{\infty}\frac{\rho^{n+w_1}}{(1-\rho^{n+w_1})^2}
\leq\sum_{n=1}^{\infty}\frac{\rho^n}{(1-\rho)^2}
=\frac{\rho}{(1-\rho)^3}, \tag*{(81)} n = 1 ∑ ∞ ( 1 − q ω Q n q Z Q ) 2 q ω Q n q Z Q ≤ n = 1 ∑ ∞ ( 1 − ρ n + w 1 ) 2 ρ n + w 1 ≤ n = 1 ∑ ∞ ( 1 − ρ ) 2 ρ n = ( 1 − ρ ) 3 ρ , (81)
∑ n = 1 ∞ ∣ q ω Q n q Z Q − 1 ( 1 − q ω Q n q Z Q − 1 ) 2 ∣ ≤ ∑ n = 1 ∞ ρ n − w 1 ( 1 − ρ n − w 1 ) 2 ≤ ∑ n = 1 ∞ ρ n − 1 / 2 ( 1 − ρ 1 / 2 ) 2 = ρ 1 / 2 ( 1 − ρ ) ( 1 − ρ 1 / 2 ) 2 , (82) \sum_{n=1}^{\infty}\left|\frac{q_{\omega_Q}^nq_{Z_Q}^{-1}}{(1-q_{\omega_Q}^nq_{Z_Q}^{-1})^2}\right|
\leq\sum_{n=1}^{\infty}\frac{\rho^{n-w_1}}{(1-\rho^{n-w_1})^2}
\leq\sum_{n=1}^{\infty}\frac{\rho^{n-1/2}}{(1-\rho^{1/2})^2}
=\frac{\rho^{1/2}}{(1-\rho)(1-\rho^{1/2})^2}, \tag*{(82)} n = 1 ∑ ∞ ( 1 − q ω Q n q Z Q − 1 ) 2 q ω Q n q Z Q − 1 ≤ n = 1 ∑ ∞ ( 1 − ρ n − w 1 ) 2 ρ n − w 1 ≤ n = 1 ∑ ∞ ( 1 − ρ 1/2 ) 2 ρ n − 1/2 = ( 1 − ρ ) ( 1 − ρ 1/2 ) 2 ρ 1/2 , (82)
2 ∑ n = 1 ∞ ∣ n q ω Q n 1 − q ω Q n ∣ ≤ 2 ∑ n = 1 ∞ n ρ n 1 − ρ n ≤ 2 ∑ n = 1 ∞ n ρ n 1 − ρ = 2 ρ ( 1 − ρ ) 3 . (83) 2\sum_{n=1}^{\infty}\left|\frac{nq_{\omega_Q}^n}{1-q_{\omega_Q}^n}\right|
\leq2\sum_{n=1}^{\infty}\frac{n\rho^n}{1-\rho^n}
\leq2\sum_{n=1}^{\infty}\frac{n\rho^n}{1-\rho}
=\frac{2\rho}{(1-\rho)^3}. \tag*{(83)} 2 n = 1 ∑ ∞ 1 − q ω Q n n q ω Q n ≤ 2 n = 1 ∑ ∞ 1 − ρ n n ρ n ≤ 2 n = 1 ∑ ∞ 1 − ρ n ρ n = ( 1 − ρ ) 3 2 ρ . (83)
We then derive that
1 ( 2 π ) 2 ∣ g ( z Q ; [ ω Q , 1 ] ) ∣ < 1 12 + N m 2 16 + ρ ( 1 − ρ ) 3 + ρ 1 / 2 ( 1 − ρ ) ( 1 − ρ 1 / 2 ) 2 + 2 ρ ( 1 − ρ ) 3 \frac{1}{(2\pi)^2}|\mathfrak{g}(z_Q;[\omega_Q,1])|
<\frac{1}{12}+\frac{N_m^2}{16}+\frac{\rho}{(1-\rho)^3}+\frac{\rho^{1/2}}{(1-\rho)(1-\rho^{1/2})^2}+\frac{2\rho}{(1-\rho)^3} ( 2 π ) 2 1 ∣ g ( z Q ; [ ω Q , 1 ]) ∣ < 12 1 + 16 N m 2 + ( 1 − ρ ) 3 ρ + ( 1 − ρ ) ( 1 − ρ 1/2 ) 2 ρ 1/2 + ( 1 − ρ ) 3 2 ρ
by (87) and (89)–(94)
= 1 12 + N m 2 16 + ( 3 ρ ( 1 − ρ ) 3 + ρ 1 / 2 ( 1 − ρ ) ( 1 − ρ 1 / 2 ) 2 ) < 1 12 + N m 2 16 + 1 10 because ρ < 1 200 by (91) and \begin{aligned}
=\frac{1}{12}+\frac{N_m^2}{16}+\left(\frac{3\rho}{(1-\rho)^3}+\frac{\rho^{1/2}}{(1-\rho)(1-\rho^{1/2})^2}\right) \\
&<\frac{1}{12}+\frac{N_m^2}{16}+\frac{1}{10}\quad\text{because }\rho<\frac{1}{200}\text{ by (91) and}
\end{aligned} = 12 1 + 16 N m 2 + ( ( 1 − ρ ) 3 3 ρ + ( 1 − ρ ) ( 1 − ρ 1/2 ) 2 ρ 1/2 ) < 12 1 + 16 N m 2 + 10 1 because ρ < 200 1 by (91) and
3 x ( 1 − x ) 3 + x 1 / 2 ( 1 − x ) ( 1 − x 1 / 2 ) 2 is increasing for 0 < x < 1 < N m 2 8 since N m ≥ 2 , \begin{aligned}
\frac{3x}{(1-x)^3}+\frac{x^{1/2}}{(1-x)(1-x^{1/2})^2}\text{ is increasing for }0<x<1 \\
&<\frac{N_m^2}{8}\quad\text{since }N_m\geq2,
\end{aligned} ( 1 − x ) 3 3 x + ( 1 − x ) ( 1 − x 1/2 ) 2 x 1/2 is increasing for 0 < x < 1 < 8 N m 2 since N m ≥ 2 ,
from which it follows that
∣ g ( z Q ; [ ω Q , 1 ] ) ∣ < π 2 2 N m 2 . (84) |\mathfrak{g}(z_Q;[\omega_Q,1])|<\frac{\pi^2}{2}N_m^2. \tag*{(84)} ∣ g ( z Q ; [ ω Q , 1 ]) ∣ < 2 π 2 N m 2 . (84)
Using (21) in Lemma 3.1, we deduce that
12 ( 2 π ) 4 ∣ g 2 ( ω Q ) ∣ ≤ 1 + 240 ∑ n = 1 ∞ n 3 ρ n 1 − ρ n < 1 + 240 ∑ n = 1 ∞ 4 n ρ n 1 − ρ due to n 3 < 4 n ( n ≥ 1 ) = 1 + 240 4 ρ ( 1 − ρ ) ( 1 − 4 ρ ) since 4 ρ < 1 by x r e f ‘ ( 91 ) ‘ < 6 because ρ < 1 200 by x r e f ‘ ( 91 ) ‘ , \begin{aligned}
\frac{12}{(2\pi)^4}|g_2(\omega_Q)|\leq1+240\sum_{n=1}^{\infty}\frac{n^3\rho^n}{1-\rho^n} \\
&< 1 + 240 \sum_{n=1}^{\infty} \frac{4^n \rho^n}{1-\rho}
\quad\text{due to } n^3 < 4^n \ (n \ge1) \\
&= 1 + 240 \frac{4\rho}{(1-\rho)(1-4\rho)}
\quad\text{since } 4\rho< 1 \text{ by } {xref}`(91)` \\
&< 6 \quad\text{because } \rho< \frac{1}{200} \text{ by } {xref}`(91)`,
\end{aligned} ( 2 π ) 4 12 ∣ g 2 ( ω Q ) ∣ ≤ 1 + 240 n = 1 ∑ ∞ 1 − ρ n n 3 ρ n < 1 + 240 n = 1 ∑ ∞ 1 − ρ 4 n ρ n due to n 3 < 4 n ( n ≥ 1 ) = 1 + 240 ( 1 − ρ ) ( 1 − 4 ρ ) 4 ρ since 4 ρ < 1 by x r e f ‘ ( 91 ) ‘ < 6 because ρ < 200 1 by x r e f ‘ ( 91 ) ‘ ,
and similarly we achieve that
6 3 ( 2 π ) 6 ∣ g 3 ( ω Q ) ∣ ≤ 1 + 504 ∑ n = 1 ∞ n 5 ρ n 1 − ρ n by x r e f ‘ ( 22 ) ‘ < 1 + 504 ∑ n = 1 ∞ 7 n ρ n 1 − ρ owing to the fact n 5 < 7 n ( n ≥ 1 ) = 1 + 504 7 ρ ( 1 − ρ ) ( 1 − 7 ρ ) since 7 ρ < 1 by x r e f ‘ ( 91 ) ‘ < 20 because ρ < 1 200 by x r e f ‘ ( 91 ) ‘ . \begin{aligned}
\frac{6^3}{(2\pi)^6}|g_3(\omega_Q)| \le1 + 504 \sum_{n=1}^{\infty} \frac{n^5 \rho^n}{1-\rho^n}
\quad\text{by } {xref}`(22)` \\
&< 1 + 504 \sum_{n=1}^{\infty} \frac{7^n \rho^n}{1-\rho}
\quad\text{owing to the fact } n^5 < 7^n \ (n \ge1) \\
&= 1 + 504 \frac{7\rho}{(1-\rho)(1-7\rho)}
\quad\text{since } 7\rho< 1 \text{ by } {xref}`(91)` \\
&< 20 \quad\text{because } \rho< \frac{1}{200} \text{ by } {xref}`(91)`.
\end{aligned} ( 2 π ) 6 6 3 ∣ g 3 ( ω Q ) ∣ ≤ 1 + 504 n = 1 ∑ ∞ 1 − ρ n n 5 ρ n by x r e f ‘ ( 22 ) ‘ < 1 + 504 n = 1 ∑ ∞ 1 − ρ 7 n ρ n owing to the fact n 5 < 7 n ( n ≥ 1 ) = 1 + 504 ( 1 − ρ ) ( 1 − 7 ρ ) 7 ρ since 7 ρ < 1 by x r e f ‘ ( 91 ) ‘ < 20 because ρ < 200 1 by x r e f ‘ ( 91 ) ‘.
Furthermore, we obtain that
1 ( 2 π ) 12 ∣ Δ ( ω Q ) ∣ ≥ ρ ∏ n = 1 ∞ ( 1 − ρ n ) 24 ≥ ρ ( 1 − ∑ n = 1 ∞ ρ n ) 24 = ρ ( 1 − 2 ρ 1 − ρ ) 24 > ρ ( 198 199 ) 24 since ρ < 1 200 by x r e f ‘ ( 91 ) ‘ > 7 8 e − 2 π y Q ≥ 7 8 e − π ∣ d K ∣ because y Q = ∣ d K ∣ 2 a ≤ ∣ d K ∣ 2 . \begin{aligned}
\frac{1}{(2\pi)^{12}}|\Delta(\omega_Q)| \ge\rho\prod_{n=1}^{\infty}(1-\rho^n)^{24} \\
&\ge\rho\left(1-\sum_{n=1}^{\infty}\rho^n\right)^{24} \\
&= \rho\left(\frac{1-2\rho}{1-\rho}\right)^{24} \\
&> \rho\left(\frac{198}{199}\right)^{24}
\quad\text{since } \rho< \frac{1}{200} \text{ by } {xref}`(91)` \\
&> \frac{7}{8}e^{-2\pi y_Q} \\
&\ge\frac{7}{8}e^{-\pi\sqrt{|d_K|}}
\quad\text{because } y_Q=\frac{\sqrt{|d_K|}}{2a}\le\frac{\sqrt{|d_K|}}{2}.
\end{aligned} ( 2 π ) 12 1 ∣Δ ( ω Q ) ∣ ≥ ρ n = 1 ∏ ∞ ( 1 − ρ n ) 24 ≥ ρ ( 1 − n = 1 ∑ ∞ ρ n ) 24 = ρ ( 1 − ρ 1 − 2 ρ ) 24 > ρ ( 199 198 ) 24 since ρ < 200 1 by x r e f ‘ ( 91 ) ‘ > 8 7 e − 2 π y Q ≥ 8 7 e − π ∣ d K ∣ because y Q = 2 a ∣ d K ∣ ≤ 2 ∣ d K ∣ .
Thus it follows that
∣ g 2 ( ω Q ) ∣ < 6 ( 2 π ) 4 12 , ∣ g 3 ( ω Q ) ∣ < 20 ( 2 π ) 6 6 3 , ∣ Δ ( ω Q ) ∣ > 7 ( 2 π ) 12 8 e − π ∣ d K ∣ . (85) |g_2(\omega_Q)| < \frac{6(2\pi)^4}{12}, \quad|g_3(\omega_Q)| < \frac{20(2\pi)^6}{6^3}, \quad|\Delta(\omega_Q)| > \frac{7(2\pi)^{12}}{8}e^{-\pi\sqrt{|d_K|}}. \tag*{(85)} ∣ g 2 ( ω Q ) ∣ < 12 6 ( 2 π ) 4 , ∣ g 3 ( ω Q ) ∣ < 6 3 20 ( 2 π ) 6 , ∣Δ ( ω Q ) ∣ > 8 7 ( 2 π ) 12 e − π ∣ d K ∣ . (85)
Case 1. If K ≠ Q ( − 1 ) , Q ( − 3 ) K \ne\mathbb{Q}(\sqrt{-1}), \mathbb{Q}(\sqrt{-3}) K = Q ( − 1 ) , Q ( − 3 ) , then
∣ f W ( k ) ( ω Q ) ∣ = ∣ f W ( 1 ) ( ω Q ) ∣ = ∣ g 2 ( ω Q ) ∣ ∣ g 3 ( ω Q ) ∣ ∣ 1 Δ ( ω Q ) ∣ ∣ ℘ ( z Q ; [ ω Q , 1 ] ) ∣ < ( 6 ( 2 π ) 4 12 ) ( 20 ( 2 π ) 6 6 3 ) ( 8 7 ( 2 π ) 12 e π ∣ d K ∣ ) ( π 2 2 N m 2 ) by x r e f ‘ ( 95 ) ‘ and x r e f ‘ ( 96 ) ‘ = 5 756 N m 2 e π ∣ d K ∣ < B m since N m ≥ 2. \begin{aligned}
|f_{\mathcal{W}}^{(k)}(\omega_Q)|=|f_{\mathcal{W}}^{(1)}(\omega_Q)| \\
&=|g_2(\omega_Q)||g_3(\omega_Q)|\left|\frac{1}{\Delta(\omega_Q)}\right||\wp(z_Q;[\omega_Q,1])| \\
&<\left(\frac{6(2\pi)^4}{12}\right)\left(\frac{20(2\pi)^6}{6^3}\right)\left(\frac{8}{7(2\pi)^{12}}e^{\pi\sqrt{|d_K|}}\right)\left(\frac{\pi^2}{2}N_m^2\right)
\quad\text{by } {xref}`(95)` \text{ and } {xref}`(96)` \\
&=\frac{5}{756}N_m^2e^{\pi\sqrt{|d_K|}} \\
&< B_m \quad\text{since } N_m \ge2.
\end{aligned} ∣ f W ( k ) ( ω Q ) ∣ = ∣ f W ( 1 ) ( ω Q ) ∣ = ∣ g 2 ( ω Q ) ∣∣ g 3 ( ω Q ) ∣ Δ ( ω Q ) 1 ∣ ℘ ( z Q ; [ ω Q , 1 ]) ∣ < ( 12 6 ( 2 π ) 4 ) ( 6 3 20 ( 2 π ) 6 ) ( 7 ( 2 π ) 12 8 e π ∣ d K ∣ ) ( 2 π 2 N m 2 ) by x r e f ‘ ( 95 ) ‘ and x r e f ‘ ( 96 ) ‘ = 756 5 N m 2 e π ∣ d K ∣ < B m since N m ≥ 2.
Case 2. If K = Q ( − 1 ) K = \mathbb{Q}(\sqrt{-1}) K = Q ( − 1 ) , then Q r e d ( d K ) = { x 2 + y 2 } \mathcal{Q}_{\mathrm{red}}(d_K) = \{x^2 + y^2\} Q red ( d K ) = { x 2 + y 2 } and so ω Q = i \omega_Q = i ω Q = i . Note by Lemma 3.1 and the fact 0 < q i = e − 2 π < 1 0 < q_i = e^{-2\pi} < 1 0 < q i = e − 2 π < 1 that
12 ( 2 π ) 4 ∣ g 2 ( i ) ∣ > 1. (86) \frac{12}{(2\pi)^4} |g_2(i)| > 1. \tag*{(86)} ( 2 π ) 4 12 ∣ g 2 ( i ) ∣ > 1. (86)
Then we have that
∣ f w ( k ) ( ω Q ) ∣ = ∣ f w ( 2 ) ( i ) ∣ = ∣ ℘ ( w 1 i + w 2 ; [ i , 1 ] ) ∣ 2 ∣ g 2 ( i ) ∣ \begin{aligned}
|f_w^{(k)}(\omega_Q)| = |f_w^{(2)}(i)| \\
&= \frac{|\wp(w_1i + w_2; [i,1])|^2}{|g_2(i)|}
\end{aligned} ∣ f w ( k ) ( ω Q ) ∣ = ∣ f w ( 2 ) ( i ) ∣ = ∣ g 2 ( i ) ∣ ∣ ℘ ( w 1 i + w 2 ; [ i , 1 ]) ∣ 2
since g 3 ( i ) = 0 g_3(i) = 0 g 3 ( i ) = 0 (see [13 Theorem 3 in Chapter 3] )
< ( π 2 2 N m 2 ) 2 12 ( 2 π ) 4 < \left(\frac{\pi^2}{2}N_{\mathfrak{m}}^2\right)^2 \frac{12}{(2\pi)^4} < ( 2 π 2 N m 2 ) 2 ( 2 π ) 4 12
by (95) and (97)
< N m 4 < B m . \begin{aligned}
< N_{\mathfrak{m}}^4 \\
&< B_{\mathfrak{m}}.
\end{aligned} < N m 4 < B m .
Case 3. If K = Q ( − 3 ) K = \mathbb{Q}(\sqrt{-3}) K = Q ( − 3 ) , then Q r e d ( d K ) = { x 2 + x y + y 2 } \mathcal{Q}_{\mathrm{red}}(d_K) = \{x^2 + xy + y^2\} Q red ( d K ) = { x 2 + x y + y 2 } and so ω Q = ζ 3 \omega_Q = \zeta_3 ω Q = ζ 3 . We see from (96) that
∣ Δ ( ζ 3 ) ∣ > 7 ( 2 π ) 12 8 e − π 3 . (87) |\Delta(\zeta_3)| > \frac{7(2\pi)^{12}}{8}e^{-\pi\sqrt{3}}. \tag*{(87)} ∣Δ ( ζ 3 ) ∣ > 8 7 ( 2 π ) 12 e − π 3 . (87)
And, we find that
∣ f w ( k ) ( ω Q ) ∣ = ∣ f w ( 3 ) ( ζ 3 ) ∣ = ∣ ℘ ( w 1 ζ 3 + w 2 ; [ ζ 3 , 1 ] ) ∣ 3 27 ∣ Δ ( ζ 3 ) ∣ \begin{aligned}
|f_w^{(k)}(\omega_Q)| = |f_w^{(3)}(\zeta_3)| \\
&= \frac{|\wp(w_1\zeta_3 + w_2; [\zeta_3,1])|^3}{\sqrt{27|\Delta(\zeta_3)|}}
\end{aligned} ∣ f w ( k ) ( ω Q ) ∣ = ∣ f w ( 3 ) ( ζ 3 ) ∣ = 27∣Δ ( ζ 3 ) ∣ ∣ ℘ ( w 1 ζ 3 + w 2 ; [ ζ 3 , 1 ]) ∣ 3
since g 2 ( ζ 3 ) = 0 g_2(\zeta_3) = 0 g 2 ( ζ 3 ) = 0 (see [13 Theorem 3 in Chapter 3] )
and so Δ ( ζ 3 ) = − 27 g 3 ( ζ 3 ) 2 \Delta(\zeta_3) = -27g_3(\zeta_3)^2 Δ ( ζ 3 ) = − 27 g 3 ( ζ 3 ) 2
< ( π 2 2 N m 2 ) 3 8 e π 3 27 ⋅ 7 ( 2 π ) 12 < \left(\frac{\pi^2}{2}N_{\mathfrak{m}}^2\right)^3 \sqrt{\frac{8e^{\pi\sqrt{3}}}{27\cdot7(2\pi)^{12}}} < ( 2 π 2 N m 2 ) 3 27 ⋅ 7 ( 2 π ) 12 8 e π 3
by (95) and (98)
< N m 6 < B m . \begin{aligned}
< N_{\mathfrak{m}}^6 \\
&< B_{\mathfrak{m}}.
\end{aligned} < N m 6 < B m .
This completes the proof.
For a nonzero proper ideal m \mathfrak{m} m of O K \mathcal{O}_K O K , define an integer
C m : = 2 + 2 3 d m ( A m B m ) 2 d m where d m : = [ K m : K ] . (88) \mathfrak{C}_{\mathfrak{m}} := 2 + 2^{3d_{\mathfrak{m}}}(\mathcal{A}_{\mathfrak{m}}B_{\mathfrak{m}})^{2d_{\mathfrak{m}}}
\quad\text{where } d_{\mathfrak{m}} := [K_{\mathfrak{m}}:K]. \tag*{(88)} C m := 2 + 2 3 d m ( A m B m ) 2 d m where d m := [ K m : K ] . (88)
Theorem 8.4. Let P ∈ E ( C ) P \in E(\mathbb{C}) P ∈ E ( C ) be a primitive m \mathfrak{m} m -division point. If K m = K ( h E ( P ) ) K_{\mathfrak{m}} = K(h_E(P)) K m = K ( h E ( P )) (for example, when f K m / K = m \mathfrak{f}_{K_{\mathfrak{m}}/K} = \mathfrak{m} f K m / K = m by Theorem A), then for every intermediate field K ⊆ L ⊆ K m K \subseteq L \subseteq K_{\mathfrak{m}} K ⊆ L ⊆ K m
L = K ( N K m / L ( h E ( P ) − C m ) ) . (89) L = K\left(N_{K_{\mathfrak{m}}/L}(h_E(P) - \mathfrak{C}_{\mathfrak{m}})\right). \tag*{(89)} L = K ( N K m / L ( h E ( P ) − C m ) ) . (89)
Proof. Let u = h E ( P ) u = h_E(P) u = h E ( P ) . Since K m = K ( u ) K_{\mathfrak{m}} = K(u) K m = K ( u ) by assumption, the Galois conjugates u g u^g u g for g ∈ Gal ( K m / K ) g \in\operatorname{Gal}(K_{\mathfrak{m}}/K) g ∈ Gal ( K m / K ) are all distinct. Furthermore, we note by Lemmas 8.2 and 8.3 that
A m u is an algebraic integer (90) \mathcal{A}_{\mathfrak{m}}u \text{ is an algebraic integer} \tag*{(90)} A m u is an algebraic integer (90)
and
∣ ι ( u ) ∣ ≤ B m for every embedding ι : K m ↪ C . (91) |\iota(u)| \le B_{\mathfrak{m}} \quad\text{for every embedding } \iota: K_{\mathfrak{m}} \hookrightarrow\mathbb{C}. \tag*{(91)} ∣ ι ( u ) ∣ ≤ B m for every embedding ι : K m ↪ C . (91)
Case 1. If L = K m L = K_m L = K m , then
N K m / L ( h E ( P ) − C m ) = h E ( P ) − C m . N_{K_m/L}(h_E(P)-C_m)=h_E(P)-C_m. N K m / L ( h E ( P ) − C m ) = h E ( P ) − C m .
Since C m ∈ Z C_m \in\mathbb{Z} C m ∈ Z and K m = K ( h E ( P ) ) K_m=K(h_E(P)) K m = K ( h E ( P )) by hypothesis, (89) follows.
Case 2. If L = K L=K L = K , then the relative norm belongs to K K K , and so (89) is obvious.
Case 3. Now, consider the case where K ⊆ L ⊆ K m K \subseteq L \subseteq K_m K ⊆ L ⊆ K m . If we put
H = Gal ( K m / L ) , d = ∣ H ∣ = [ K m : L ] , ℓ = [ L : Q ] , H=\operatorname{Gal}(K_m/L),\quad d=|H|=[K_m:L],\quad\ell=[L:\mathbb{Q}], H = Gal ( K m / L ) , d = ∣ H ∣ = [ K m : L ] , ℓ = [ L : Q ] ,
then we observe that
d ≥ 2 , d ℓ = [ K m : Q ] = 2 d m , ℓ ≤ d m . (92) d\geq2,\quad d\ell=[K_m:\mathbb{Q}]=2d_m,\quad\ell\leq d_m. \tag*{(92)} d ≥ 2 , d ℓ = [ K m : Q ] = 2 d m , ℓ ≤ d m . (92)
For g ∈ Gal ( K m / K ) g\in\operatorname{Gal}(K_m/K) g ∈ Gal ( K m / K ) , let
F g , H ( x ) = ∏ σ ∈ H ( x − ( u g ) σ ) ∈ L [ x ] . (93) F_{g,H}(x)=\prod_{\sigma\in H}\left(x-(u^g)^\sigma\right)\in L[x]. \tag*{(93)} F g , H ( x ) = σ ∈ H ∏ ( x − ( u g ) σ ) ∈ L [ x ] . (93)
Furthermore, for g ∉ H g\notin H g ∈ / H , let
D g , H ( x ) = F g , H ( x ) − F id , H ( x ) ∈ L [ x ] . (94) D_{g,H}(x)=F_{g,H}(x)-F_{\operatorname{id},H}(x)\in L[x]. \tag*{(94)} D g , H ( x ) = F g , H ( x ) − F id , H ( x ) ∈ L [ x ] . (94)
Note that if g ∉ H g\notin H g ∈ / H and so g H ∩ H = ∅ gH\cap H=\varnothing g H ∩ H = ∅ , then the sets of roots of F g , H ( x ) F_{g,H}(x) F g , H ( x ) and F id , H ( x ) F_{\operatorname{id},H}(x) F id , H ( x ) are disjoint because the associated Galois conjugates of u u u are distinct. This implies that if g ∉ H g\notin H g ∈ / H , then the polynomial D g , H D_{g,H} D g , H is nonzero.
Now, let g ∉ H g\notin H g ∈ / H and write
F g , H ( x ) = a d x d + a d − 1 x d − 1 + ⋯ + a 0 , F id , H ( x ) = b d x d + b d − 1 x d − 1 + ⋯ + b 0 , D g , H ( x ) = c d x d + c d − 1 x d − 1 + ⋯ + c 0 . \begin{aligned}
F_{g,H}(x)&=a_dx^d+a_{d-1}x^{d-1}+\cdots+a_0,\\
F_{\operatorname{id},H}(x)&=b_dx^d+b_{d-1}x^{d-1}+\cdots+b_0,\\
D_{g,H}(x)&=c_dx^d+c_{d-1}x^{d-1}+\cdots+c_0.
\end{aligned} F g , H ( x ) F id , H ( x ) D g , H ( x ) = a d x d + a d − 1 x d − 1 + ⋯ + a 0 , = b d x d + b d − 1 x d − 1 + ⋯ + b 0 , = c d x d + c d − 1 x d − 1 + ⋯ + c 0 .
Then we achieve by the definitions (93) , (94) and (91) that for any embedding ι : L ↪ C \iota:L\hookrightarrow\mathbb{C} ι : L ↪ C (which can be extended to an embedding of K m K_m K m into C \mathbb{C} C ) and 0 ≤ j ≤ d 0\leq j\leq d 0 ≤ j ≤ d
∣ ι ( c j ) ∣ ≤ ∣ ι ( a j ) ∣ + ∣ ι ( b j ) ∣ ≤ ( d j ) B m d − j + ( d j ) B m d − j ≤ U , where U = 2 d + 1 B m d . (95) |\iota(c_j)|\leq|\iota(a_j)|+|\iota(b_j)|\leq\binom{d}{j}B_m^{d-j}+\binom{d}{j}B_m^{d-j}\leq U,\quad\text{where }U=2^{d+1}B_m^d. \tag*{(95)} ∣ ι ( c j ) ∣ ≤ ∣ ι ( a j ) ∣ + ∣ ι ( b j ) ∣ ≤ ( j d ) B m d − j + ( j d ) B m d − j ≤ U , where U = 2 d + 1 B m d . (95)
And, we know by (90) that the coefficients of A m d F g , H ( x ) A_m^dF_{g,H}(x) A m d F g , H ( x ) and A m d F id , H ( x ) A_m^dF_{\operatorname{id},H}(x) A m d F id , H ( x ) are all algebraic integers. Let δ \delta δ be the nonzero leading coefficient of D g , H ( x ) D_{g,H}(x) D g , H ( x ) . We then derive that
1 ≤ ∣ N L / Q ( A m d δ ) ∣ since A m d δ is a nonzero algebraic integer = A m d ℓ ∣ N L / Q ( δ ) ∣ = A m d ℓ ∣ δ ∣ ⋅ ∏ ι ∣ ι ( δ ) ∣ where ι runs over all nonidentity embeddings L ↪ C ≤ A m d ℓ ∣ δ ∣ U ℓ − 1 by (106), \begin{aligned}
1&\leq|N_{L/\mathbb{Q}}(A_m^d\delta)|\quad\text{since }A_m^d\delta\text{ is a nonzero algebraic integer}\\
&=A_m^{d\ell}|N_{L/\mathbb{Q}}(\delta)|\\
&=A_m^{d\ell}|\delta|\cdot\prod_{\iota}|\iota(\delta)|\quad\text{where }\iota\text{ runs over all nonidentity embeddings }L\hookrightarrow\mathbb{C}\\
&\leq A_m^{d\ell}|\delta|U^{\ell-1}\quad\text{by (106),}
\end{aligned} 1 ≤ ∣ N L / Q ( A m d δ ) ∣ since A m d δ is a nonzero algebraic integer = A m d ℓ ∣ N L / Q ( δ ) ∣ = A m d ℓ ∣ δ ∣ ⋅ ι ∏ ∣ ι ( δ ) ∣ where ι runs over all nonidentity embeddings L ↪ C ≤ A m d ℓ ∣ δ ∣ U ℓ − 1 by (106),
and hence
∣ δ ∣ ≥ A m − d ℓ U − ( ℓ − 1 ) . (96) |\delta| \ge\mathcal{A}_m^{-d\ell}U^{-(\ell-1)}. \tag*{(96)} ∣ δ ∣ ≥ A m − d ℓ U − ( ℓ − 1 ) . (96)
If D g , H ( x ) D_{g,H}(x) D g , H ( x ) is constant, then it has no zeros. Otherwise, every zero z z z of D g , H ( x ) D_{g,H}(x) D g , H ( x ) satisfies
∣ z ∣ ≤ 1 + U ∣ δ ∣ by Cauchy’s bound (see [14, p. 123]) and (106) ≤ 1 + A m d ℓ U ℓ by (107) = 1 + 2 2 d m + ℓ ( A m B m ) 2 d m because d ℓ = 2 d m by (103) and U = 2 d + 1 B m d ≤ 1 + 2 3 d m ( A m B m ) 2 d m since ℓ ≤ d m by (103) < C m by the definition (99). \begin{aligned}
|z| &\le1+\frac{U}{|\delta|} &&\text{by Cauchy’s bound (see [14, p. 123]) and (106)}\\
&\le1+\mathcal{A}_m^{d\ell}U^\ell&&\text{by (107)}\\
&=1+2^{2d_m+\ell}(\mathcal{A}_m\mathcal{B}_m)^{2d_m} &&\text{because }d\ell=2d_m\text{ by (103) and }U=2^{d+1}\mathcal{B}_m^d\\
&\le1+2^{3d_m}(\mathcal{A}_m\mathcal{B}_m)^{2d_m} &&\text{since }\ell\le d_m\text{ by (103)}\\
&<\mathcal{C}_m &&\text{by the definition (99).}
\end{aligned} ∣ z ∣ ≤ 1 + ∣ δ ∣ U ≤ 1 + A m d ℓ U ℓ = 1 + 2 2 d m + ℓ ( A m B m ) 2 d m ≤ 1 + 2 3 d m ( A m B m ) 2 d m < C m by Cauchy’s bound (see [14, p. 123]) and (106) by (107) because d ℓ = 2 d m by (103) and U = 2 d + 1 B m d since ℓ ≤ d m by (103) by the definition (99).
This observation yields that
D g , H ( C m ) ≠ 0 for every g ∈ Gal ( K m / K ) ∖ H . (97) D_{g,H}(\mathcal{C}_m)\ne0 \quad\text{for every }g\in\operatorname{Gal}(K_m/K)\setminus H. \tag*{(97)} D g , H ( C m ) = 0 for every g ∈ Gal ( K m / K ) ∖ H . (97)
Let
v = N K m / L ( u − C m ) . v=N_{K_m/L}(u-\mathcal{C}_m). v = N K m / L ( u − C m ) .
We find by the definition (93) that for g ∈ Gal ( K m / K ) g\in\operatorname{Gal}(K_m/K) g ∈ Gal ( K m / K )
v g = ( − 1 ) d F g , H ( C m ) . v^g=(-1)^dF_{g,H}(\mathcal{C}_m). v g = ( − 1 ) d F g , H ( C m ) .
Thus every element of H H H fixes v v v , whereas no element of Gal ( K m / K ) ∖ H \operatorname{Gal}(K_m/K)\setminus H Gal ( K m / K ) ∖ H fixes v v v by the definition (94) and (97) . Therefore we conclude by Galois theory that
L = K ( v ) = K ( N K m / L ( h E ( P ) − C m ) ) . L=K(v)=K\left(N_{K_m/L}(h_E(P)-\mathcal{C}_m)\right). L = K ( v ) = K ( N K m / L ( h E ( P ) − C m ) ) .
Hilbert’s twelfth problem for imaginary quadratic fields
We combine the relative norm construction with the description of ray class groups by extended form class groups. This gives explicit generators for all finite abelian extensions of an imaginary quadratic field in terms of special values of Fricke functions.
Lemma 9.1. If m = N O K \mathfrak{m}=N\mathcal{O}_K m = N O K for a positive integer N N N divisible by 4 4 4 , then we have
K m = K ( h E ( ϕ E ( 1 N + O K ) ) ) . K_m=K\left(h_E\left(\phi_E\left(\frac{1}{N}+\mathcal{O}_K\right)\right)\right). K m = K ( h E ( ϕ E ( N 1 + O K ) ) ) .
Proof. Note that the point ϕ E ( 1 N + O K ) ∈ E ( C ) \phi_E\left(\frac{1}{N}+\mathcal{O}_K\right)\in E(\mathbb{C}) ϕ E ( N 1 + O K ) ∈ E ( C ) is a primitive m \mathfrak{m} m -division point. If N = 4 N=4 N = 4 , the assertion follows from [8 Corollary 7.1] . If N > 4 N>4 N > 4 , then m \mathfrak{m} m satisfies (S1), so the assertion follows from [25 p. 190] .
For a positive integer N N N , let Q ( d K , N ) \mathcal{Q}(d_K,N) Q ( d K , N ) be the subset of Q ( d K ) \mathcal{Q}(d_K) Q ( d K ) defined by
Q ( d K , N ) = { a x 2 + b x y + c y 2 ∈ Q ( d K ) ∣ gcd ( a , N ) = 1 } . \mathcal{Q}(d_K,N)=\{ax^2+bxy+cy^2\in\mathcal{Q}(d_K)\mid\gcd(a,N)=1\}. Q ( d K , N ) = { a x 2 + b x y + c y 2 ∈ Q ( d K ) ∣ g cd( a , N ) = 1 } .
The congruence subgroup
Γ 1 ( N ) = { γ ∈ S L 2 ( Z ) ∣ γ ≡ [ 1 ∗ 0 1 ] ( m o d N M 2 ( Z ) ) } \Gamma_1(N)=\left\{\gamma\in\mathrm{SL}_2(\mathbb{Z})\mid\gamma\equiv\begin{bmatrix}1&*\\0&1\end{bmatrix}\pmod{N M_2(\mathbb{Z})}\right\} Γ 1 ( N ) = { γ ∈ SL 2 ( Z ) ∣ γ ≡ [ 1 0 ∗ 1 ] ( mod N M 2 ( Z )) }
acts from the right on the set Q ( d K , N ) \mathcal{Q}(d_K,N) Q ( d K , N ) by
Q γ = Q ( γ [ x y ] ) ( Q ∈ Q ( d K , N ) , γ ∈ Γ 1 ( N ) ) , Q^\gamma=Q\left(\gamma\begin{bmatrix}x\\y\end{bmatrix}\right)\qquad(Q\in\mathcal{Q}(d_K,N),\ \gamma\in\Gamma_1(N)), Q γ = Q ( γ [ x y ] ) ( Q ∈ Q ( d K , N ) , γ ∈ Γ 1 ( N )) ,
which yields an equivalence relation ∼ Γ 1 ( N ) \sim_{\Gamma_1(N)} ∼ Γ 1 ( N ) on Q ( d K , N ) \mathcal{Q}(d_K,N) Q ( d K , N ) . For Q ∈ Q ( d K , N ) Q\in\mathcal{Q}(d_K,N) Q ∈ Q ( d K , N ) , let [ Q ] Γ 1 ( N ) [Q]_{\Gamma_1(N)} [ Q ] Γ 1 ( N ) be its equivalence class under ∼ Γ 1 ( N ) \sim_{\Gamma_1(N)} ∼ Γ 1 ( N ) . Furthermore, let C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) denote the set of these equivalence classes, namely,
C Γ 1 ( N ) ( d K , N ) = Q ( d K , N ) / ∼ Γ 1 ( N ) = { [ Q ] Γ 1 ( N ) ∣ Q ∈ Q ( d K , N ) } . C_{\Gamma_1(N)}(d_K,N)=\mathcal{Q}(d_K,N)/\mathord{\sim_{\Gamma_1(N)}}=\{[Q]_{\Gamma_1(N)}\mid Q\in\mathcal{Q}(d_K,N)\}. C Γ 1 ( N ) ( d K , N ) = Q ( d K , N ) / ∼ Γ 1 ( N ) = {[ Q ] Γ 1 ( N ) ∣ Q ∈ Q ( d K , N )} .
For Q ∈ Q ( d K , N ) Q\in\mathcal{Q}(d_K,N) Q ∈ Q ( d K , N ) , write
Q = a Q x 2 + b Q x y + c Q y 2 , Q=a_Qx^2+b_Qxy+c_Qy^2, Q = a Q x 2 + b Q x y + c Q y 2 ,
and choose a Q ′ ∈ Z a'_Q\in\mathbb{Z} a Q ′ ∈ Z with a Q a Q ′ ≡ 1 ( m o d N ) a_Qa'_Q\equiv1\pmod N a Q a Q ′ ≡ 1 ( mod N ) . Recall that ω Q \omega_Q ω Q is the zero of Q ( x , 1 ) Q(x,1) Q ( x , 1 ) in H \mathbb{H} H . Put
b K = { 0 if d K ≡ 0 ( m o d 4 ) , 1 if d K ≡ 1 ( m o d 4 ) b_K=\begin{cases}0&\text{if }d_K\equiv0\pmod4,\\1&\text{if }d_K\equiv1\pmod4\end{cases} b K = { 0 1 if d K ≡ 0 ( mod 4 ) , if d K ≡ 1 ( mod 4 )
so that τ K = ( − b K + d K ) / 2 \tau_K=(-b_K+\sqrt{d_K})/2 τ K = ( − b K + d K ) /2 .
Proposition 9.2. Let m = N O K \mathfrak{m}=N\mathcal{O}_K m = N O K for a positive integer N N N . There is a binary operation on C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) which makes C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) a group isomorphic to the ray class group C ( m ) C(\mathfrak{m}) C ( m ) via the mapping
C Γ 1 ( N ) ( d K , N ) ⟶ C ( m ) , [ Q ] Γ 1 ( N ) ⟼ [ [ ω Q , 1 ] ] m . C_{\Gamma_1(N)}(d_K,N)\longrightarrow C(\mathfrak{m}),\qquad[Q]_{\Gamma_1(N)}\longmapsto[[\omega_Q,1]]_{\mathfrak{m}}. C Γ 1 ( N ) ( d K , N ) ⟶ C ( m ) , [ Q ] Γ 1 ( N ) ⟼ [[ ω Q , 1 ] ] m . Moreover, the mapping
σ N : C Γ 1 ( N ) ( d K , N ) ⟶ Gal ( K m / K ) , \sigma_N:C_{\Gamma_1(N)}(d_K,N)\longrightarrow\operatorname{Gal}(K_{\mathfrak{m}}/K), σ N : C Γ 1 ( N ) ( d K , N ) ⟶ Gal ( K m / K ) , [ Q ] Γ 1 ( N ) ⟼ ( f ( τ K ) ⟼ f ( [ 1 − a Q ′ ( b Q + b K 2 ) 0 a Q ′ ] ( − ω Q ‾ ) ) ∣ f ∈ F N is finite at τ K ) [Q]_{\Gamma_1(N)}\longmapsto\left(f(\tau_K)\longmapsto f\left(\begin{bmatrix}1&-a'_Q\left(\frac{b_Q+b_K}{2}\right)\\0&a'_Q\end{bmatrix}(-\overline{\omega_Q})\right)\mid f\in\mathcal{F}_N\text{ is finite at }\tau_K\right) [ Q ] Γ 1 ( N ) ⟼ ( f ( τ K ) ⟼ f ( [ 1 0 − a Q ′ ( 2 b Q + b K ) a Q ′ ] ( − ω Q ) ) ∣ f ∈ F N is finite at τ K ) is a well-defined isomorphism.
Proof. See [9 Theorems 9.4 and 12.3] .
Remark 9.3. We call the group C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) the extended form class group of discriminant d K d_K d K for Γ 1 ( N ) \Gamma_1(N) Γ 1 ( N ) . When N = 1 N=1 N = 1 , this group is none other than Gauss’s classical form class group (see [5 Theorems 3.9 and 7.7] ).
Theorem 9.4. Let K K K be an imaginary quadratic field of discriminant d K d_K d K , and let L L L be a finite extension of K K K . Then L / K L/K L / K is abelian if and only if there exist a positive integer N N N divisible by 4 4 4 and a subgroup H H H of the extended form class group C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) such that
L = K ( ∏ [ Q ] Γ 1 ( N ) ∈ H ( f [ 0 a Q ′ / N ] ( ∣ O K × ∣ / 2 ) ( − ω ‾ Q ) − C N O K ) ) . L=K\left(\prod_{[Q]_{\Gamma_1(N)}\in H}\left(f_{\left[\begin{smallmatrix}0&a'_Q/N\end{smallmatrix}\right]}^{(|\mathcal{O}_K^\times|/2)}(-\overline{\omega}_Q)-C_{N\mathcal{O}_K}\right)\right). L = K [ Q ] Γ 1 ( N ) ∈ H ∏ ( f [ 0 a Q ′ / N ] ( ∣ O K × ∣/2 ) ( − ω Q ) − C N O K ) .
Proof. For a positive integer N N N , let σ N : C Γ 1 ( N ) ( d K , N ) → ∼ Gal ( K N O K / K ) \sigma_N:C_{\Gamma_1(N)}(d_K,N)\xrightarrow{\sim}\operatorname{Gal}(K_{N\mathcal{O}_K}/K) σ N : C Γ 1 ( N ) ( d K , N ) ∼ Gal ( K N O K / K ) be the isomorphism stated in Proposition 9.2. Then we establish that
L / K is abelian L/K\text{ is abelian} L / K is abelian
⟺ L ⊂ K m for some nonzero ideal m of O K \Longleftrightarrow\quad L\subset K_{\mathfrak m}\text{ for some nonzero ideal }\mathfrak m\text{ of }\mathcal{O}_K ⟺ L ⊂ K m for some nonzero ideal m of O K
⟺ L ⊆ K N O K for some positive integer N divisible by 4 \Longleftrightarrow\quad L\subseteq K_{N\mathcal{O}_K}\text{ for some positive integer }N\text{ divisible by }4 ⟺ L ⊆ K N O K for some positive integer N divisible by 4
⟺ L is the subfield of K N O K which is the fixed field of a subgroup of Gal ( K N O K / K ) \Longleftrightarrow\quad L\text{ is the subfield of }K_{N\mathcal{O}_K}\text{ which is the fixed field of a subgroup of }\operatorname{Gal}(K_{N\mathcal{O}_K}/K) ⟺ L is the subfield of K N O K which is the fixed field of a subgroup of Gal ( K N O K / K )
for some positive integer N N N divisible by 4 4 4
⟺ L is the subfield of K N O K which is the fixed field of σ N ( H ) \Longleftrightarrow\quad L\text{ is the subfield of }K_{N\mathcal{O}_K}\text{ which is the fixed field of }\sigma_N(H) ⟺ L is the subfield of K N O K which is the fixed field of σ N ( H )
with H H H a subgroup of C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) for some positive integer N N N divisible by 4 4 4 , by Proposition 9.2
⟺ L = K ( ∏ [ Q ] Γ 1 ( N ) ∈ H ( f [ 0 1 / N ] ( ∣ O K × ∣ / 2 ) ( τ K ) − C N O K ) σ N ( [ Q ] Γ 1 ( N ) ) ) \Longleftrightarrow\quad L=K\left(\prod_{[Q]_{\Gamma_1(N)}\in H}\left(f_{\left[\begin{smallmatrix}0&1/N\end{smallmatrix}\right]}^{(|\mathcal{O}_K^\times|/2)}(\tau_K)-C_{N\mathcal{O}_K}\right)^{\sigma_N([Q]_{\Gamma_1(N)})}\right) ⟺ L = K [ Q ] Γ 1 ( N ) ∈ H ∏ ( f [ 0 1/ N ] ( ∣ O K × ∣/2 ) ( τ K ) − C N O K ) σ N ([ Q ] Γ 1 ( N ) )
with H H H a subgroup of C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) for some positive integer N N N divisible by 4 4 4 , by Theorem 8.4, Lemma 9.1 , the definitions (3) and (25)
⟺ L = K ( ∏ [ Q ] Γ 1 ( N ) ∈ H ( f [ 0 a Q ′ / N ] ( ∣ O K × ∣ / 2 ) ( − ω ‾ Q ) − C N O K ) ) \Longleftrightarrow\quad L=K\left(\prod_{[Q]_{\Gamma_1(N)}\in H}\left(f_{\left[\begin{smallmatrix}0&a'_Q/N\end{smallmatrix}\right]}^{(|\mathcal{O}_K^\times|/2)}(-\overline{\omega}_Q)-C_{N\mathcal{O}_K}\right)\right) ⟺ L = K [ Q ] Γ 1 ( N ) ∈ H ∏ ( f [ 0 a Q ′ / N ] ( ∣ O K × ∣/2 ) ( − ω Q ) − C N O K )
with H H H a subgroup of C Γ 1 ( N ) ( d K , N ) C_{\Gamma_1(N)}(d_K,N) C Γ 1 ( N ) ( d K , N ) for some positive integer N N N divisible by 4 4 4 , by Propositions 3.3 (ii), 9.2 and (74) .
Statements & Declarations
Funding
The first author was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. RS-2026-25606107). The second author was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. RS-2026-25588491). The third author was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. RS-2026-25588491) and Hankuk University of Foreign Studies Research Fund of 2026.
Competing interests
The authors have not disclosed any competing interests.
Data availability
Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
Department of Mathematics Dankook University Cheonan-si, Chungnam 31116 Republic of Korea E-mail address: hoyunjung@dankook.ac.kr
Department of Mathematical Sciences KAIST Daejeon 34141 Republic of Korea E-mail address: jkgoo@kaist.ac.kr
DEPARTMENT OF MATHEMATICS HANKUK UNIVERSITY OF FOREIGN STUDIES YONGIN-SI, GYEONGGI-DO 17035 REPUBLIC OF KOREA E-mail address: dhshin@hufs.ac.kr