Introduction
Let E E E be an elliptic curve over Q \mathbb{Q} Q without complex multiplication. Its Galois image on torsion is open in GL 2 ( Z ^ ) \operatorname{GL}_{2}(\widehat{\mathbb{Z}}) GL 2 ( Z ) by Serre [39 ] , and Mazur’s Program B [34 p. 109] asks which images occur. Write N n s ( m ) N_{\mathrm{ns}}(m) N ns ( m ) for the normalizer of a non-split Cartan subgroup of GL 2 ( Z / m ) \operatorname{GL}_{2}(\mathbb{Z}/m) GL 2 ( Z / m ) and X n s + ( m ) X_{\mathrm{ns}}^{+}(m) X ns + ( m ) for its modular curve. For a prime ℓ \ell ℓ , Rouse, Sutherland and Zureick-Brown [36, Theorem 1.6] show that the ℓ \ell ℓ -adic image H H H of E E E falls in one of four cases: X H X_H X H is P 1 \mathbb{P}^{1} P 1 or an elliptic curve of rank one; X H X_H X H has an exceptional rational point and H H H is in an explicit list; H ≤ N n s ( m ) H \leq N_{\mathrm{ns}}(m) H ≤ N ns ( m ) for m = 3 3 , 5 2 , 7 2 , 11 2 m = 3^3, 5^2, 7^2, 11^2 m = 3 3 , 5 2 , 7 2 , 1 1 2 or a prime m = ℓ ≥ 19 m = \ell\geq19 m = ℓ ≥ 19 ; or H H H lies in one of two groups of level 49. They conjecture that in the last two cases X H X_H X H has no exceptional rational points, that is, none other than cusps and CM points [36 ] .
For m = 25 m = 25 m = 25 the curve X : = X n s + ( 25 ) X := X_{\mathrm{ns}}^+(25) X := X ns + ( 25 ) has genus 14, rational CM points and no rational cusps, and every isogeny factor of its Jacobian has analytic rank equal to its dimension [36 pp. 3–4 and Table 2] . Rouse, Sutherland and Zureick-Brown computed a canonical model of X X X and, searching the points of X n s + ( 5 ) ≅ P 1 X_{\mathrm{ns}}^+(5) \cong\mathbb{P}^1 X ns + ( 5 ) ≅ P 1 of height at most 1000 with a test on traces of Frobenius, found eight rational points, all CM [36 §9.2] ; the CM point of discriminant − 8 -8 − 8 , with j j j -invariant 8000, is a ninth. They could not determine X ( Q ) X(\mathbb{Q}) X ( Q ) [36 ] . The neighboring cases are settled: m = 27 m = 27 m = 27 by Balakrishnan, Betts, Hast, Jha and Müller [4 Theorem 1.3] , m = 49 m = 49 m = 49 by Furio and Lombardo [18 Theorem 1.4] , and the two groups of level 49 by Arango-Piñeros and Zureick-Brown [1 Theorem 2] . The case m = 121 m = 121 m = 121 is treated in [23 ] .
Let D : = { − 3 , − 7 , − 8 , − 12 , − 27 , − 28 , − 43 , − 67 , − 163 } D := \{-3, -7, -8, -12, -27, -28, -43, -67, -163\} D := { − 3 , − 7 , − 8 , − 12 , − 27 , − 28 , − 43 , − 67 , − 163 } . For D ∈ D D \in D D ∈ D let y D ∈ X ( Q ) y_D \in X(\mathbb{Q}) y D ∈ X ( Q ) be the CM point of discriminant D D D (Lemma 2.2).
Theorem 1.1. The curve X n s + ( 25 ) X_{\mathrm{ns}}^+(25) X ns + ( 25 ) has exactly nine rational points, the CM points y D y_D y D , D ∈ D D \in D D ∈ D .
The proof does not use the classification of imaginary quadratic orders of class number one, so, as each such order of conductor prime to 5 in whose field 5 is inert gives a rational point of X X X (proof of Lemma 2.2), Theorem 1.1 recovers that these orders are exactly those with discriminant in D D D . That rational points on non-split Cartan curves bear on the class number one problem in this way was noted by Serre [40 Appendix] .
Corollary 1.2. Let E / Q E/\mathbb{Q} E / Q be an elliptic curve without complex multiplication. Then the image of Gal ( Q ‾ / Q ) \operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q}) Gal ( Q / Q ) acting on E [ 25 ] E[25] E [ 25 ] is not contained in a conjugate of N n s ( 25 ) N_{\mathrm{ns}}(25) N ns ( 25 ) . So the case H ≤ N n s ( 5 2 ) H \leq N_{\mathrm{ns}}(5^2) H ≤ N ns ( 5 2 ) of [36 Theorem 1.6 (iii)] does not occur, and the 5-adic image of E E E is, up to conjugacy, one of the 25 subgroups of GL 2 ( Z 5 ) \operatorname{GL}_2(\mathbb{Z}_5) GL 2 ( Z 5 ) counted in [36 Table 3] , each of level at most 25 and index at most 120.
Proof. Such an E E E would give a rational point of X X X with j j j -invariant j ( E ) j(E) j ( E ) , which is not a CM j j j -invariant, against Theorem 1.1 . The groups in cases (iii) and (iv) of [36 Theorem 1.6] other than N n s ( 5 2 ) N_{\mathrm{ns}}(5^2) N ns ( 5 2 ) are proper subgroups of level prime to 5, so none contains H H H , which there is, up to conjugacy, the inverse image in GL 2 ( Z ^ ) \operatorname{GL}_2(\widehat{\mathbb{Z}}) GL 2 ( Z ) of the 5-adic image of E E E and so maps onto GL 2 ( Z / m ) \operatorname{GL}_2(\mathbb{Z}/m) GL 2 ( Z / m ) for every m m m prime to 5. By the first part the 5-adic image therefore falls under case (i) or (ii); at ℓ = 5 \ell= 5 ℓ = 5 these are the 23 groups H H H of 5-power level with X H ( Q ) X_H(\mathbb{Q}) X H ( Q ) infinite and the two exceptional groups of [36 Table 1] counted in [36 Table 3] , where the count assumes only, at ℓ = 5 \ell= 5 ℓ = 5 , that N n s ( 5 2 ) N_{\mathrm{ns}}(5^2) N ns ( 5 2 ) does not occur.
Karabiyik has announced a determination of X ( Q ) X(\mathbb{Q}) X ( Q ) by quadratic Chabauty in the framework of p p p -adic adelic metrics of Besser, Müller and Srinivasan [24], [6, §7].
The obstacle. The Jacobian J J J of X X X has rank 14, its genus (Proposition 10.1 ), so the method of Chabauty and Coleman does not apply, while quadratic Chabauty [5 ] does. For modular curves of rank equal to genus it works as follows [2 §1.5] , [3 Algorithm 3.12] . At an auxiliary prime q q q of good reduction, a correspondence Z Z Z of trace zero gives a locally analytic function on X ( Q q ) X(\mathbb{Q}_q) X ( Q q ) , the local height at q q q minus a bilinear form in the abelian logarithm, which on X ( Q ) X(\mathbb{Q}) X ( Q ) takes only the values allowed by the local heights at the bad primes. It is expanded as a power series on every residue disc modulo q q q , and X ( Q ) X(\mathbb{Q}) X ( Q ) lies among the points where it takes one of these values. The bilinear form is fixed by the heights at known rational points, each of which gives a linear equation, and the Hecke action reduces the number of points needed [2 §1.7] , [4 §2.5] ; generators of J ( Q ) J(\mathbb{Q}) J ( Q ) are not needed. In this way the rational points were determined at levels 13 and 17 [2 , 3 ] and 27 [4 ] . Hashimoto, Lido, Lombardo, Mascot and Parent work from the moduli interpretation instead of an equation of the curve, at prime level [20 ] . At the bad prime the local term has to be controlled: at level 27 it vanishes because every local point reduces to the one Galois-stable component of a stable model [4 §3.6] , and at level 17 the correspondences were chosen, through a known stable model, so that it vanishes [3 §5.5] .
Our last step is a first-order argument modulo q 2 q^2 q 2 at a known rational point: it expands the height identity only at a point y D y_D y D , so it can only show that the residue disc of y D y_D y D contains no second rational point. So every other disc has to be excluded first, by the descent below. At 5 the Jacobian has potentially good reduction (Lemma 7.1), and we choose correspondences for which the local term vanishes without a stable model. Leonhardt and Rendell computed a basis of H d R 1 H^1_{\mathrm{dR}} H dR 1 of X X X from modular forms and the Hodge filtration for a correspondence on it, the first steps of quadratic Chabauty for this curve [32 Theorem A] . We do not use them: the local heights at the auxiliary prime q q q that enter are, modulo q 2 q^2 q 2 , logarithms of values of functions (Lemma 8.1), so no splitting of the Hodge filtration is computed. Working modulo q 2 q^2 q 2 with integral data rather than with q q q -adic power series is also the point of view of geometric quadratic Chabauty [17 ] .
Residue discs from modular units. The ten cusps of X X X are defined over K : = Q ( ζ 25 ) + K := \mathbb{Q}(\zeta_{25})^+ K := Q ( ζ 25 ) + , and G : = Gal ( K / Q ) G := \operatorname{Gal}(K/\mathbb{Q}) G := Gal ( K / Q ) permutes them simply transitively. Products of Siegel functions along the orbits of N n s ( 25 ) ∩ SL 2 ( Z / 25 ) N_{\mathrm{ns}}(25) \cap\operatorname{SL}_2(\mathbb{Z}/25) N ns ( 25 ) ∩ SL 2 ( Z /25 ) give functions on X X X over K K K with divisors supported on the cusps. Those whose divisor is divisible by 5, modulo fifth powers and constants, form an F 5 [ G ] \mathbb{F}_5[G] F 5 [ G ] -module V V V of dimension 8 (Section 4). For rational points x x x and y y y the quotient u ( x ) / u ( y ) u(x)/u(y) u ( x ) / u ( y ) of the values of such a unit has valuation divisible by 5 at every prime of K K K not above 5, because X X X has a smooth model over Z [ 1 / 5 ] \mathbb{Z}[1/5] Z [ 1/5 ] . As 5 ∤ h ( K ) 5 \nmid h(K) 5 ∤ h ( K ) , it is a 5-unit up to a fifth power, and the 5-units modulo fifth powers form a free F 5 [ G ] \mathbb{F}_5[G] F 5 [ G ] -module of rank one on η : = ( 1 − ζ 25 ) ( 1 − ζ 25 − 1 ) \eta:= (1-\zeta_{25})(1-\zeta_{25}^{-1}) η := ( 1 − ζ 25 ) ( 1 − ζ 25 − 1 ) . So x x x has a label in Hom ( V , F 5 ) \operatorname{Hom}(V,\mathbb{F}_5) Hom ( V , F 5 ) , one of 5 8 5^8 5 8 , and at a prime q ≡ 1 ( m o d 25 ) q \equiv1 \pmod{25} q ≡ 1 ( mod 25 ) the label is read off the reduction of x x x by fifth-power residue symbols. The general statement is the following.
Theorem 1.3. Let X / Q X/\mathbb{Q} X / Q be a smooth projective geometrically connected curve, m ≥ 2 m \ge2 m ≥ 2 an integer, F / Q F/\mathbb{Q} F / Q a finite Galois extension with group G G G , and S S S a finite G G G -stable set of primes of F F F . Let Λ \Lambda Λ be a Z [ G ] \mathbb{Z}[G] Z [ G ] -module that is finitely generated over Z \mathbb{Z} Z , and e ↦ u e e \mapsto u_e e ↦ u e a homomorphism Λ → F ( X ) × \Lambda\to F(X)^\times Λ → F ( X ) × , such that:
(U1) the divisors of the u e u_e u e are supported on a finite set Σ ⊂ X ( F ) \Sigma\subset X(F) Σ ⊂ X ( F ) with Σ ∩ X ( Q ) = ∅ \Sigma\cap X(\mathbb{Q}) = \varnothing Σ ∩ X ( Q ) = ∅ , and σ ( u e ) ∈ F × u σ e \sigma(u_e) \in F^\times u_{\sigma e} σ ( u e ) ∈ F × u σ e for σ ∈ G \sigma\in G σ ∈ G , acting on coefficients;
(U2) for each prime q ∉ S \mathfrak{q} \notin S q ∈ / S of F F F , the curve X F X_F X F has a regular proper model over O F , q \mathcal{O}_{F,\mathfrak{q}} O F , q with irreducible special fiber.
Let Λ m : = { e ∈ Λ : div u e ∈ m Div ( X F ) } \Lambda_m := \{e \in\Lambda: \operatorname{div} u_e \in m\operatorname{Div}(X_F)\} Λ m := { e ∈ Λ : div u e ∈ m Div ( X F )} and Λ 0 : = { e ∈ Λ : u e ∈ F × } \Lambda_0 := \{e \in\Lambda: u_e \in F^\times\} Λ 0 := { e ∈ Λ : u e ∈ F × } , and put
V : = Λ m / ( m Λ + Λ 0 ) V := \Lambda_m/(m\Lambda+\Lambda_0) V := Λ m / ( m Λ + Λ 0 ) and
F ( S , m ) : = { a ∈ F × / F × m : m ∣ v q ( a ) for every prime q ∉ S } . F(S,m) := \{a \in F^\times/F^{\times m} : m \mid v_{\mathfrak{q}}(a)\text{ for every prime }\mathfrak{q} \notin S\}. F ( S , m ) := { a ∈ F × / F × m : m ∣ v q ( a ) for every prime q ∈ / S } . Fix y ∈ X ( Q ) y \in X(\mathbb{Q}) y ∈ X ( Q ) .
(a) For x ∈ X ( Q ) x \in X(\mathbb{Q}) x ∈ X ( Q ) , the map e ↦ u e ( x ) / u e ( y ) e \mapsto u_e(x)/u_e(y) e ↦ u e ( x ) / u e ( y ) induces a G G G -equivariant homomorphism
Φ x : V → F ( S , m ) . \Phi_x: V \to F(S,m). Φ x : V → F ( S , m ) .
(b) Let q ∉ S q \notin S q ∈ / S be a prime of F F F , unramified and of degree one over a prime q q q at which X X X has good reduction, let χ q : F ( S , m ) → F q × / F q × m \chi_q:F(S,m)\to\mathbb{F}_q^{\times}/\mathbb{F}_q^{\times m} χ q : F ( S , m ) → F q × / F q × m be the m m m -th power residue symbol at q \mathfrak{q} q , and let x ˉ \bar{x} x ˉ denote reduction modulo q \mathfrak{q} q . Then χ q ( Φ x ( e ) ) = u ~ e ( x ˉ ) / u ~ e ( y ˉ ) \chi_q(\Phi_x(e))=\tilde{u}_e(\bar{x})/\tilde{u}_e(\bar{y}) χ q ( Φ x ( e )) = u ~ e ( x ˉ ) / u ~ e ( y ˉ ) for e ∈ Λ m e\in\Lambda_m e ∈ Λ m , where u ~ e ( x ˉ ) ∈ F q × / F q × m \tilde{u}_e(\bar{x})\in\mathbb{F}_q^{\times}/\mathbb{F}_q^{\times m} u ~ e ( x ˉ ) ∈ F q × / F q × m is the leading coefficient of u e u_e u e at x ˉ \bar{x} x ˉ (the value of u ˉ e \bar{u}_e u ˉ e at x ˉ \bar{x} x ˉ when no point of Σ \Sigma Σ reduces to x ˉ \bar{x} x ˉ ), defined in § 3 .
So if Q Q Q is a finite set of such primes, Φ x \Phi_x Φ x lies in the set of Φ ∈ Hom G ( V , F ( S , m ) ) \Phi\in\operatorname{Hom}_G(V,F(S,m)) Φ ∈ Hom G ( V , F ( S , m )) such that for each q ∈ Q q\in Q q ∈ Q some z ˉ ∈ X ( F q ) \bar{z}\in X(\mathbb{F}_q) z ˉ ∈ X ( F q ) has χ q ( Φ ( e ) ) = u ~ e ( z ˉ ) / u ~ e ( y ˉ ) \chi_q(\Phi(e))=\tilde{u}_e(\bar{z})/\tilde{u}_e(\bar{y}) χ q ( Φ ( e )) = u ~ e ( z ˉ ) / u ~ e ( y ˉ ) for all e e e ; and x ˉ \bar{x} x ˉ is one such z ˉ \bar{z} z ˉ . On a modular curve X H X_H X H of level M M M with no rational cusp, the modular units give (U1), with F F F the field of the cusps, and the smooth model of Deligne and Rapoport over Z [ 1 / M ] \mathbb{Z}[1/M] Z [ 1/ M ] gives (U2) with S S S the primes above M M M . To use the theorem one needs F ( S , m ) F(S,m) F ( S , m ) explicitly, and Remark 3.1 gives a condition under which it is. At level 25, with m = 5 m=5 m = 5 , F = K F=K F = K and S S S the prime above 5, it gives:
Theorem 1.4. For every x ∈ X ( Q ) x\in X(\mathbb{Q}) x ∈ X ( Q ) there is D ∈ D D\in\mathcal{D} D ∈ D such that x x x and y D y_D y D have the same reduction modulo 1151.
The proof uses no rank, generator or Shafarevich–Tate group of J J J and no integrality of j j j . Evaluating functions whose divisors are m m m -th multiples, modulo m m m -th powers, is the classical descent by functions [38 §2.1, Lemma 2.1] , [8 §2] . For e ∈ Λ 5 e\in\Lambda_5 e ∈ Λ 5 put T e : = [ 1 5 div u e ] ∈ J ( K ) [ 5 ] T_e:=\left[\frac{1}{5}\operatorname{div}u_e\right]\in J(K)[5] T e := [ 5 1 div u e ] ∈ J ( K ) [ 5 ] . By Weil reciprocity, a rational divisor A A A of degree 0 away from the cusps gives e ↦ u e ( A ) e\mapsto u_e(A) e ↦ u e ( A ) modulo fifth powers, depending only on the class of A A A modulo 5 J ( Q ) 5J(\mathbb{Q}) 5 J ( Q ) ; this is the descent map attached to the classes T e T_e T e . So the label Φ x \Phi_x Φ x is the value at [ x − y ] [x-y] [ x − y ] of a homomorphism J ( Q ) / 5 → Hom G ( V , K ( S , 5 ) ) J(\mathbb{Q})/5\to\operatorname{Hom}_G(V,K(S,5)) J ( Q ) /5 → Hom G ( V , K ( S , 5 )) , computed without knowing J ( Q ) J(\mathbb{Q}) J ( Q ) . The sieve at the primes q q q has the shape of the Mordell–Weil sieve [9 ] , with Hom G ( V , K ( S , 5 ) ) \operatorname{Hom}_G(V,K(S,5)) Hom G ( V , K ( S , 5 )) in place of J ( Q ) / 5 J(\mathbb{Q})/5 J ( Q ) /5 . Here the functions are the modular units of the curve, their classes form a module over the Galois group of the field of the cusps, and the local conditions are fifth-power residue symbols at split primes. The proof at level 49 of Arango-Piñeros and Zureick-Brown is also a descent of Kummer type, on the superelliptic equations F ( x , y ) = k w 7 F(x,y)=kw^7 F ( x , y ) = k w 7 of Furio and Lombardo, where F F F is the binary cubic whose roots are the non-rational cusps of X n s + ( 7 ) X_{\mathrm{ns}}^+(7) X ns + ( 7 ) or of X s p + ( 7 ) X_{\mathrm{sp}}^+(7) X sp + ( 7 ) [1 pp. 2–3, 6] .
The term at 5. To finish we use quadratic Chabauty at q = 1151 q=1151 q = 1151 , with correspondences that fix the elliptic curve and change only the level structure. As q ≡ 1 ( m o d 25 ) q\equiv1\pmod{25} q ≡ 1 ( mod 25 ) , the labels can be read at q q q (§ 5 ), and as q ≡ 11 ( m o d 12 ) q\equiv11\pmod{12} q ≡ 11 ( mod 12 ) , the fixed points of the correspondences are supersingular modulo q q q ; so are eight of the nine y D y_D y D , as their D D D are not squares modulo q q q , and all computations except those at y − 163 y_{-163} y − 163 stay over F q 2 \mathbb{F}_{q^2} F q 2 (§ 11 ). For v ∈ F 5 × v\in\mathbb{F}_5^\times v ∈ F 5 × , S v S_v S v sends a Cartan subring ( Z / 25 ) [ x ] (\mathbb{Z}/25)[x] ( Z /25 ) [ x ] , with tr x = 0 \operatorname{tr}x=0 tr x = 0 and x 2 = 2 x^2=2 x 2 = 2 , to the six ( Z / 25 ) [ x ′ ] (\mathbb{Z}/25)[x'] ( Z /25 ) [ x ′ ] with x ′ = x + 5 w x'=x+5w x ′ = x + 5 w , where w w w anticommutes with x x x modulo 5 and w 2 = v w^2=v w 2 = v . On the Cartan subrings congruent to a given one modulo 5, S v S_v S v is a Cayley graph on F 25 \mathbb{F}_{25} F 25 , whose eigenvalues other than its degree 6 are the sums Υ ( m ) : = ∑ N γ = m Tr γ \Upsilon(m):=\sum_{N\gamma=m}\operatorname{Tr}\gamma Υ ( m ) := ∑ N γ = m Tr γ over γ ∈ F 25 \gamma\in\mathbb{F}_{25} γ ∈ F 25 , which are minus the Kloosterman sums ∑ x ∈ F 5 × ζ 5 x + m / x \sum_{x\in\mathbb{F}_5^\times}\zeta_5^{x+m/x} ∑ x ∈ F 5 × ζ 5 x + m / x . Put Z 1 : = S 2 − S 3 Z_1:=S_2-S_3 Z 1 := S 2 − S 3 and Z 2 : = S 1 − S 4 Z_2:=S_1-S_4 Z 2 := S 1 − S 4 . For the quartic character χ \chi χ of F 5 × \mathbb{F}_5^\times F 5 × with χ ( 2 ) = i \chi(2)=i χ ( 2 ) = i , Z 2 − i Z 1 = ∑ ν χ ˉ ( ν ) S ν Z_2-iZ_1=\sum_\nu\bar{\chi}(\nu)S_\nu Z 2 − i Z 1 = ∑ ν χ ˉ ( ν ) S ν , and by Hasse–Davenport its non-zero eigenvalues are squares of Gauss sums up to fourth roots of unity; so Z 1 Z_1 Z 1 and Z 2 Z_2 Z 2 have their eigenvalues in 5 Q \sqrt{5}\mathbb{Q} 5 Q (Proposition 6.4). So the local term at 5 vanishes, with no stable model, by the following criterion.
Proposition 1.5. Let p p p be a prime and X / Q p X/\mathbb{Q}_p X / Q p a smooth projective geometrically connected curve of genus at least 2 whose Jacobian J J J has potentially good reduction. Let C = ∑ i n i C i C = \sum_i n_i C_i C = ∑ i n i C i be a symmetric correspondence over Q p \mathbb{Q}_p Q p , where the C i ⊂ X × X C_i \subset X \times X C i ⊂ X × X are distinct irreducible curves other than the diagonal and the n i n_i n i are non-zero integers, with non-zero endomorphism Z Z Z of J J J ; let Fix C \operatorname{Fix} C Fix C and C ( y ) C(y) C ( y ) be its pull-backs by the diagonal X → X × X X \to X \times X X → X × X and, for y ∈ X y \in X y ∈ X , by x ↦ ( y , x ) x \mapsto(y,x) x ↦ ( y , x ) . Let ℓ ≠ p \ell\ne p ℓ = p be a prime and H : = H e ˊ t 1 ( X Q ‾ p , Q ‾ ℓ ) H := H^1_{\mathrm{\acute{e}t}}(X_{\overline{\mathbb{Q}}_p}, \overline{\mathbb{Q}}_\ell) H := H e ˊ t 1 ( X Q p , Q ℓ ) . Suppose there is α ∈ Q ‾ ℓ \alpha\in\overline{\mathbb{Q}}_\ell α ∈ Q ℓ with α ∉ Q \alpha\notin\mathbb{Q} α ∈ / Q such that Z Z Z acts on each isotypic component of H H H under the inertia group of Q p \mathbb{Q}_p Q p by a scalar in Q α \mathbb{Q}\alpha Q α . Then for all y , z ∈ X ( Q p ) y,z \in X(\mathbb{Q}_p) y , z ∈ X ( Q p ) such that none of ( y , y ) (y,y) ( y , y ) , ( z , z ) (z,z) ( z , z ) , ( y , z ) (y,z) ( y , z ) lies on a C i C_i C i ,
i p ( z − y , Fix C − C ( y ) − C ( z ) ) = 0 , i_p(z-y,\operatorname{Fix} C-C(y)-C(z))=0, i p ( z − y , Fix C − C ( y ) − C ( z )) = 0 , where i p i_p i p is the Néron local symbol.
The proof (§ 7 ) rests on the formula of Betts and Dogra for this local term [7 Corollary 12.1.3, Remark 12.1.4] . It expresses the term as a pairing with a measure on the stable reduction graph. Potentially good reduction makes the graph a tree; then the mass of the measure on each inertia orbit of vertices is half the trace of an endomorphism of an abelian variety, so rational, and also lies in Q α \mathbb{Q}\alpha Q α , so it is 0; the masses are constant on orbits, so the measure is 0. Betts and Dogra treat X n s + ( p ) X_{\mathrm{ns}}^+(p) X ns + ( p ) for primes p > 3 p > 3 p > 3 with p ≡ 3 ( m o d 4 ) p \equiv3 \pmod4 p ≡ 3 ( mod 4 ) through the components of its special fiber [7 §12.2.1] . The hypothesis of Proposition 1.5 holds for Z 1 Z_1 Z 1 and Z 2 Z_2 Z 2 (§ 7 ), which act on the ℓ \ell ℓ -adic representation σ ( π ) \sigma(\pi) σ ( π ) of each automorphic representation π \pi π occurring in the cohomology of X X X through the restriction π 5 ∣ G L 2 ( Z 5 ) \pi_5|_{\mathrm{GL}_2(\mathbb{Z}_5)} π 5 ∣ GL 2 ( Z 5 ) of its local component at 5. It does not hold for Hecke operators such as T 2 T_2 T 2 , whose eigenvalues are not determined by π 5 ∣ G L 2 ( Z 5 ) \pi_5|_{\mathrm{GL}_2(\mathbb{Z}_5)} π 5 ∣ GL 2 ( Z 5 ) .
The height pairing from the CM points. With all local terms away from q q q zero, the height identity for a pair of known rational points becomes a linear equation on the height pairing, with a right side in q Z q q\mathbb{Z}_q q Z q . The pairing is determined by a linear functional on the Hecke algebra modulo q q q , and at q = 1151 q=1151 q = 1151 the equations from the nine CM points (the control equations), for Z 1 Z_1 Z 1 and Z 2 Z_2 Z 2 , have rank 14 on the 14-dimensional space of such functionals; one correspondence would give at most eight (Remark 9.2 ). So they show that the pairing is integral on J ( Q ) ⊗ Z q J(\mathbb{Q}) \otimes\mathbb{Z}_q J ( Q ) ⊗ Z q and determine it modulo q 2 q^2 q 2 . On the disc of y D y_D y D , a second rational point would satisfy a congruence for each of Z 1 Z_1 Z 1 and Z 2 Z_2 Z 2 , with coefficients that we compute, and at each of the nine discs the two congruences have no common solution (Table 1 ). The general form is Proposition 9.1.
Table 1. The numbers of Proposition 9.1 (b) at q = 1151 q = 1151 q = 1151 for the tangent vectors δ D \delta_D δ D , as residues in [ 0 , 1150 ] [0, 1150] [ 0 , 1150 ] , with s Z = − L Z / Q Z s_Z = -L_Z/Q_Z s Z = − L Z / Q Z .
Outline
Discs (§ 3 , § 4 and § 5 ). Every x ∈ X ( Q ) x \in X(\mathbb{Q}) x ∈ X ( Q ) has the reduction modulo 1151 of some y D y_D y D .
Rank (§ 10 ). rank J ( Q ) = 14 \operatorname{rank} J(\mathbb{Q})=14 rank J ( Q ) = 14 .
Correspondences (§ 6 ). Z 1 Z_1 Z 1 and Z 2 Z_2 Z 2 have eigenvalues in 5 Q \sqrt{5}\mathbb{Q} 5 Q and fixed points over j = 0 j=0 j = 0 and 1728.
The term at 5 (§ 7 ). Their local term at 5 vanishes.
Heights (§ 8 and § 9 ). The control equations fix the height pairing, and Proposition 9.1 gives a criterion for a known point to be alone in its disc.
The prime 1151 (§ 11 ). The hypotheses hold at q = 1151 q=1151 q = 1151 , and the criterion holds at each of the nine points.
Computations. The computations on which the proofs rest use PARI/GP and Python and are exact, except for the L L L -values of Proposition 10.1, which are computed in FLINT’s ball arithmetic. The programs, which run in about three hours on one thread, are deposited with their outputs at [22 ] .
The curve
Let G : = GL 2 ( Z / 25 ) \mathcal{G}:=\operatorname{GL}_{2}(\mathbb{Z}/25) G := GL 2 ( Z /25 ) , let ξ : = ( 0 2 1 0 ) \xi:=\begin{pmatrix}0&2\\1&0\end{pmatrix} ξ := ( 0 1 2 0 ) , so that ξ 2 = 2 \xi^{2}=2 ξ 2 = 2 , and let C : = ( Z / 25 ) [ ξ ] ⊂ M 2 ( Z / 25 ) \mathcal{C}:=(\mathbb{Z}/25)[\xi]\subset M_{2}(\mathbb{Z}/25) C := ( Z /25 ) [ ξ ] ⊂ M 2 ( Z /25 ) . As 2 2 2 is not a square modulo 5 5 5 , C \mathcal{C} C is the unramified quadratic extension of Z / 25 \mathbb{Z}/25 Z /25 , and C \mathcal{C} C has order 600 600 600 . Let N ⊂ G N\subset\mathcal{G} N ⊂ G be the stabilizer of the subring C \mathcal{C} C under conjugation. It is the normalizer N n s ( 25 ) N_{\mathrm{ns}}(25) N ns ( 25 ) of the non-split Cartan subgroup C × \mathcal{C}^{\times} C × , it has order 1200 1200 1200 , and N = C × ⊔ C × s N=\mathcal{C}^{\times}\sqcup\mathcal{C}^{\times}s N = C × ⊔ C × s with s : = diag ( 1 , − 1 ) s:=\operatorname{diag}(1,-1) s := diag ( 1 , − 1 ) , which induces the non-trivial automorphism of C \mathcal{C} C . The elements of C × s \mathcal{C}^{\times}s C × s are the matrices ( a − 2 b b − a ) \begin{pmatrix}a&-2b\\b&-a\end{pmatrix} ( a b − 2 b − a ) , all of trace 0 0 0 , and det ( N ) = ( Z / 25 ) × \det(N)=(\mathbb{Z}/25)^{\times} det ( N ) = ( Z /25 ) × . Let Ω \Omega Ω be the set of conjugates of C \mathcal{C} C , which we call Cartan subrings; ∣ Ω ∣ = 250 |\Omega|=250 ∣Ω∣ = 250 . Each contains exactly two elements x x x with tr x = 0 \operatorname{tr}x=0 tr x = 0 and x 2 = 2 x^{2}=2 x 2 = 2 , namely ± x \pm x ± x , and every such x ∈ M 2 ( Z / 25 ) x\in M_{2}(\mathbb{Z}/25) x ∈ M 2 ( Z /25 ) generates a Cartan subring, since t 2 − 2 t^{2}-2 t 2 − 2 is irreducible modulo 5 5 5 .
Let Y Y Y be the compactified moduli scheme over Z [ 1 / 5 ] \mathbb{Z}[1/5] Z [ 1/5 ] of pairs ( E , ϕ ) (E,\phi) ( E , ϕ ) with ϕ : ( Z / 25 ) 2 → E [ 25 ] \phi:(\mathbb{Z}/25)^{2}\to E[25] ϕ : ( Z /25 ) 2 → E [ 25 ] an isomorphism, with the right action r g ( E , ϕ ) : = ( E , ϕ ∘ g ) r_g(E,\phi):=(E,\phi\circ g) r g ( E , ϕ ) := ( E , ϕ ∘ g ) of G \mathcal{G} G . The quotient X : = Y / N X:=Y/N X := Y / N , through ( E , ϕ ) ↦ ( E , ϕ C ϕ − 1 ) (E,\phi)\mapsto(E,\phi\mathcal{C}\phi^{-1}) ( E , ϕ ) ↦ ( E , ϕ C ϕ − 1 ) , is the curve of pairs ( E , R ) (E,R) ( E , R ) with R ⊂ End ( E [ 25 ] ) R\subset\operatorname{End}(E[25]) R ⊂ End ( E [ 25 ]) a Cartan subring, meaning that ϕ − 1 R ϕ ∈ Ω \phi^{-1}R\phi\in\Omega ϕ − 1 R ϕ ∈ Ω for one, hence every, ϕ \phi ϕ . It is X n s + ( 25 ) X_{\mathrm{ns}}^{+}(25) X ns + ( 25 ) , it is geometrically connected because det ( N ) = ( Z / 25 ) × \det(N)=(\mathbb{Z}/25)^{\times} det ( N ) = ( Z /25 ) × , it has degree 250 250 250 over the j j j -line, and its genus is 14 14 14 [36 Table 2] . As ξ \xi ξ is conjugate to its transpose, N N N is conjugate to its transpose, so the usual conventions for level structures give the same curve. By Deligne and Rapoport [16 VI, Proposition 6.7] , X X X has a smooth proper model X \mathcal{X} X over Z [ 1 / 5 ] \mathbb{Z}[1/5] Z [ 1/5 ] , the coarse moduli scheme of the same problem. For Q ( X ) \mathbb{Q}(X) Q ( X ) we use Shimura’s model [42 Theorem 6.6, Proposition 6.9] : the field F F F of modular functions for Γ ( 25 ) \Gamma(25) Γ ( 25 ) whose q 1 / 25 q^{1/25} q 1/25 -expansions have coefficients in Q ( ζ 25 ) \mathbb{Q}(\zeta_{25}) Q ( ζ 25 ) is Galois over Q ( j ) \mathbb{Q}(j) Q ( j ) with group G / { ± 1 } \mathcal{G}/\{\pm1\} G / { ± 1 } , an element of SL 2 ( Z / 25 ) \operatorname{SL}_{2}(\mathbb{Z}/25) SL 2 ( Z /25 ) acting by composition with a lift to SL 2 ( Z ) \operatorname{SL}_{2}(\mathbb{Z}) SL 2 ( Z ) and diag ( 1 , d ) \operatorname{diag}(1,d) diag ( 1 , d ) by ζ 25 ↦ ζ 25 d \zeta_{25}\mapsto\zeta_{25}^{d} ζ 25 ↦ ζ 25 d on coefficients, and Q ( X ) = F N \mathbb{Q}(X)=F^{N} Q ( X ) = F N .
Lemma 2.1. The curve X X X has ten cusps, each of width 25 25 25 . They are defined over K : = Q ( ζ 25 ) + K:=\mathbb{Q}(\zeta_{25})^{+} K := Q ( ζ 25 ) + , and G : = Gal ( K / Q ) G:=\operatorname{Gal}(K/\mathbb{Q}) G := Gal ( K / Q ) acts on them simply transitively. For d ∈ ( Z / 25 ) × d\in(\mathbb{Z}/25)^{\times} d ∈ ( Z /25 ) × choose g d ∈ N g_d\in N g d ∈ N with det g d = d \det g_d=d det g d = d and put B d : = g d − 1 diag ( 1 , d ) ∈ SL 2 ( Z / 25 ) B_d:=g_d^{-1}\operatorname{diag}(1,d)\in\operatorname{SL}_{2}(\mathbb{Z}/25) B d := g d − 1 diag ( 1 , d ) ∈ SL 2 ( Z /25 ) . Then σ d ( ∞ ) = B d ∞ \sigma_d(\infty)=B_d\infty σ d ( ∞ ) = B d ∞ , where σ d ∈ G \sigma_d\in G σ d ∈ G is induced by ζ 25 ↦ ζ 25 d \zeta_{25}\mapsto\zeta_{25}^{d} ζ 25 ↦ ζ 25 d .
Proof. The width of the cusp γ ∞ \gamma\infty γ ∞ is the least h > 0 h>0 h > 0 with ± ( 1 h 0 1 ) ∈ γ − 1 N γ \pm\begin{pmatrix}1&h\\0&1\end{pmatrix}\in\gamma^{-1}N\gamma ± ( 1 0 h 1 ) ∈ γ − 1 N γ . A conjugate of such an element is not in C × s \mathcal{C}^{\times}s C × s , whose elements have trace 0 0 0 . In C × \mathcal{C}^{\times} C × the elements a + b ξ a+b\xi a + b ξ with characteristic polynomial ( t ∓ 1 ) 2 (t\mp1)^{2} ( t ∓ 1 ) 2 have a = ± 1 a=\pm1 a = ± 1 and 5 ∣ b 5\mid b 5 ∣ b , so a + b ξ ∓ 1 = b ξ a+b\xi\mp1=b\xi a + b ξ ∓ 1 = b ξ is 5 5 5 times a matrix that is 0 0 0 or invertible modulo 5 5 5 , whereas, for 0 < h < 25 0<h<25 0 < h < 25 , ( 0 h 0 0 ) \begin{pmatrix}0&h\\0&0\end{pmatrix} ( 0 0 h 0 ) is non-zero modulo 5 5 5 if 5 ∤ h 5\nmid h 5 ∤ h and is 5 5 5 times a non-zero nilpotent if 5 ∣ h 5\mid h 5 ∣ h . So every width is 25 25 25 , and there are 250 / 25 = 10 250/25=10 250/25 = 10 cusps.
Let f ∈ K ( X ) = F N + f\in K(X)=F^{N^{+}} f ∈ K ( X ) = F N + , where N + : = { g ∈ N : det g = ± 1 } N^{+}:=\{g\in N:\det g=\pm1\} N + := { g ∈ N : det g = ± 1 } . The element g d g_d g d fixes Q ( X ) \mathbb{Q}(X) Q ( X ) and acts on Q ( ζ 25 ) \mathbb{Q}(\zeta_{25}) Q ( ζ 25 ) as σ d \sigma_d σ d , so it acts on K ( X ) K(X) K ( X ) as σ d \sigma_d σ d ; and diag ( 1 , d ) = g d B d \operatorname{diag}(1,d)=g_dB_d diag ( 1 , d ) = g d B d . If f f f and σ d f \sigma_d f σ d f are regular at the relevant cusps, ( σ d f ) ( ∞ ) (\sigma_d f)(\infty) ( σ d f ) ( ∞ ) is the constant term of f f f with σ d \sigma_d σ d applied to its coefficients, that is, the value at ∞ \infty ∞ of f diag ( 1 , d ) = ( σ d f ) ∘ B d f^{\operatorname{diag}(1,d)}=(\sigma_d f)\circ B_d f diag ( 1 , d ) = ( σ d f ) ∘ B d , which is ( σ d f ) ( B d ∞ ) (\sigma_d f)(B_d\infty) ( σ d f ) ( B d ∞ ) . As σ d ( f ( ∞ ) ) = ( σ d f ) ( σ d ∞ ) \sigma_d(f(\infty))=(\sigma_d f)(\sigma_d\infty) σ d ( f ( ∞ )) = ( σ d f ) ( σ d ∞ ) for all such f f f , σ d ∞ = B d ∞ \sigma_d\infty=B_d\infty σ d ∞ = B d ∞ . Now B d ∞ = ∞ B_d\infty=\infty B d ∞ = ∞ on X X X if and only if diag ( 1 , d ) ∈ N ⋅ { ± ( 1 h 0 1 ) } \operatorname{diag}(1,d)\in N\cdot\{\pm\begin{pmatrix}1&h\\0&1\end{pmatrix}\} diag ( 1 , d ) ∈ N ⋅ { ± ( 1 0 h 1 ) } , so if and only if N N N contains an upper triangular matrix with diagonal ± ( 1 , d ) \pm(1,d) ± ( 1 , d ) . The upper triangular elements of N N N are the diag ( a , ± a ) \operatorname{diag}(a,\pm a) diag ( a , ± a ) , so this happens exactly for d = ± 1 d=\pm1 d = ± 1 . Hence ∞ \infty ∞ is defined over K K K with stabilizer { 1 } \{1\} { 1 } in G G G , and its orbit has ten elements. ∎
For d ∈ ( Z / 25 ) × / { ± 1 } d \in(\mathbb{Z}/25)^{\times}/\{\pm1\} d ∈ ( Z /25 ) × / { ± 1 } we write c d : = σ d ( ∞ ) c_d := \sigma_d(\infty) c d := σ d ( ∞ ) for the cusps.
Lemma 2.2. Let D ∈ D D \in\mathcal{D} D ∈ D and let E D / Q E_D/\mathbb{Q} E D / Q have complex multiplication by the order O D \mathcal{O}_D O D of discriminant D D D . The image of O D \mathcal{O}_D O D in End ( E D [ 25 ] ) \operatorname{End}(E_D[25]) End ( E D [ 25 ]) is a Cartan subring stable under Gal ( Q ‾ / Q ) \operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q}) Gal ( Q / Q ) , so it defines a point y D ∈ X ( Q ) y_D \in X(\mathbb{Q}) y D ∈ X ( Q ) with j ( y D ) = j D j(y_D) = j_D j ( y D ) = j D . For D D D in the order of D \mathcal{D} D , j D = 0 j_D = 0 j D = 0 , − 3375 -3375 − 3375 , 8000 8000 8000 , 54000 54000 54000 , − 12288000 -12288000 − 12288000 , 16581375 16581375 16581375 , − 884736000 -884736000 − 884736000 , − 147197952000 -147197952000 − 147197952000 and − 262537412640768000 -262537412640768000 − 262537412640768000 . The map X → X ( 1 ) X \to X(1) X → X ( 1 ) is unramified at each y D y_D y D , so j − j D j-j_D j − j D is a local parameter at y D y_D y D . For D = − 3 D=-3 D = − 3 this holds modulo every prime q ≥ 7 q \ge7 q ≥ 7 , and for D ≠ − 3 D \ne-3 D = − 3 modulo every prime q ≥ 7 q \ge7 q ≥ 7 with j D ≢ 0 , 1728 ( m o d q ) j_D \not\equiv0,1728 \pmod q j D ≡ 0 , 1728 ( mod q ) .
Proof. As 5 5 5 is inert in Q ( D ) \mathbb{Q}(\sqrt{D}) Q ( D ) and prime to the conductor of O D \mathcal{O}_D O D , O D / 25 \mathcal{O}_D/25 O D /25 is the unramified quadratic extension of Z / 25 \mathbb{Z}/25 Z /25 . Since E D [ 25 ] E_D[25] E D [ 25 ] is free of rank one over O D / 25 \mathcal{O}_D/25 O D /25 , the image of O D \mathcal{O}_D O D is a Cartan subring. Galois conjugation carries O D ⊂ End ( E D , Q ‾ ) \mathcal{O}_D \subset\operatorname{End}(E_D,\overline{\mathbb{Q}}) O D ⊂ End ( E D , Q ) to itself, so the subring is Galois-stable. For j D ≠ 0 j_D \ne0 j D = 0 , Aut ( E D ) = { ± 1 } \operatorname{Aut}(E_D)=\{\pm1\} Aut ( E D ) = { ± 1 } and X → X ( 1 ) X \to X(1) X → X ( 1 ) is unramified. For D = − 3 D=-3 D = − 3 the automorphism of order 6 6 6 lies in O − 3 \mathcal{O}_{-3} O − 3 , so it normalizes the Cartan subring and the ramification index is again 1 1 1 . Modulo a prime q ≥ 7 q \ge7 q ≥ 7 the reduction of E − 3 E_{-3} E − 3 has automorphism group μ 6 \mu_6 μ 6 , and that of E D E_D E D , D ≠ − 3 D \ne-3 D = − 3 , has automorphism group { ± 1 } \{\pm1\} { ± 1 } when j D ≢ 0 , 1728 j_D \not\equiv0,1728 j D ≡ 0 , 1728 . □
The descent
We prove Theorem 1.3 . For q ∉ S q \notin S q ∈ / S let X q \mathcal{X}_q X q be the model of (U2), and write its special fiber as r X q , s r\mathcal{X}_{q,s} r X q , s with X q , s \mathcal{X}_{q,s} X q , s integral. A point x ∈ X ( Q ) x \in X(\mathbb{Q}) x ∈ X ( Q ) extends to a section x ~ \widetilde{x} x of X q \mathcal{X}_q X q , as X q \mathcal{X}_q X q is proper, and r x ~ ∗ X q , s r\widetilde{x}^{*}\mathcal{X}_{q,s} r x ∗ X q , s is the pull-back of the divisor of a uniformizer, of degree 1 1 1 . So r = 1 r=1 r = 1 : the special fiber is reduced and deg x ~ ∗ X q , s = 1 \deg\widetilde{x}^{*}\mathcal{X}_{q,s}=1 deg x ∗ X q , s = 1 .
Proof of Theorem 1.3 (a). As x , y ∉ Σ x,y \notin\Sigma x , y ∈ / Σ , the values u e ( x ) , u e ( y ) u_e(x),u_e(y) u e ( x ) , u e ( y ) are defined and non-zero. For e ∈ m Λ e \in m\Lambda e ∈ m Λ the quotient u e ( x ) / u e ( y ) u_e(x)/u_e(y) u e ( x ) / u e ( y ) is an m m m -th power, and for e ∈ Λ 0 e \in\Lambda_0 e ∈ Λ 0 it is 1 1 1 , so Φ x \Phi_x Φ x is well defined on V V V . For σ ∈ G \sigma\in G σ ∈ G , as x x x and y y y are rational, σ ( u e ( x ) / u e ( y ) ) = ( σ u e ) ( x ) / ( σ u e ) ( y ) = u σ e ( x ) / u σ e ( y ) \sigma(u_e(x)/u_e(y))=(\sigma u_e)(x)/(\sigma u_e)(y)=u_{\sigma e}(x)/u_{\sigma e}(y) σ ( u e ( x ) / u e ( y )) = ( σ u e ) ( x ) / ( σ u e ) ( y ) = u σ e ( x ) / u σ e ( y ) , the constant in (U1) cancelling. Let e ∈ Λ m e \in\Lambda_m e ∈ Λ m , with div u e = m D e \operatorname{div}u_e=mD_e div u e = m D e , and let q ∉ S q \notin S q ∈ / S . On the regular scheme X q \mathcal{X}_q X q the divisor of u e u_e u e is m D ‾ e + n X q , s m\overline{D}_e+n\mathcal{X}_{q,s} m D e + n X q , s , with D ‾ e \overline{D}_e D e the closure of D e D_e D e and n ∈ Z n \in\mathbb{Z} n ∈ Z , since X q , s \mathcal{X}_{q,s} X q , s is irreducible. Neither x ~ \widetilde{x} x nor y ~ \widetilde{y} y lies in its support, so
v q ( u e ( x ) u e ( y ) ) = deg x ~ ∗ ( div u e ) − deg y ~ ∗ ( div u e ) = m ( deg x ~ ∗ D ‾ e − deg y ~ ∗ D ‾ e ) ∈ m Z . v_q\left(\frac{u_e(x)}{u_e(y)}\right)=\deg\widetilde{x}^{*}(\operatorname{div}u_e)-\deg\widetilde{y}^{*}(\operatorname{div}u_e)=m\left(\deg\widetilde{x}^{*}\overline{D}_e-\deg\widetilde{y}^{*}\overline{D}_e\right)\in m\mathbb{Z}. v q ( u e ( y ) u e ( x ) ) = deg x ∗ ( div u e ) − deg y ∗ ( div u e ) = m ( deg x ∗ D e − deg y ∗ D e ) ∈ m Z .
For (b), let X \mathcal{X} X be a smooth proper model of X X X over Z q \mathbb{Z}_q Z q , unique if the genus is positive, on which reductions are taken; let x ‾ ∈ X ( F q ) \overline{x}\in\mathcal{X}(\mathbb{F}_q) x ∈ X ( F q ) , and τ \tau τ a parameter at x ‾ \overline{x} x over Z q \mathbb{Z}_q Z q , so that O ^ X , x ‾ = Z q [ [ τ ] ] \widehat{\mathcal{O}}_{\mathcal{X},\overline{x}}=\mathbb{Z}_q[[\tau]] O X , x = Z q [[ τ ]] . As q q q has degree one, u e ∈ Q q ( X ) u_e\in\mathbb{Q}_q(X) u e ∈ Q q ( X ) . The prime ideals of height one of Z q [ [ τ ] ] \mathbb{Z}_q[[\tau]] Z q [[ τ ]] are ( q ) (q) ( q ) and those generated by irreducible distinguished polynomials, and Z q [ [ τ ] ] \mathbb{Z}_q[[\tau]] Z q [[ τ ]] is factorial. For e ∈ Λ m e\in\Lambda_m e ∈ Λ m the multiplicity of each horizontal prime in div u e \operatorname{div}u_e div u e lies in m Z m\mathbb{Z} m Z , so
u e = q c ∏ P P n P ⋅ w , n P ∈ m Z , w ∈ Z q [ [ τ ] ] × , u_e=q^c\prod_P P^{n_P}\cdot w,\qquad n_P\in m\mathbb{Z},\quad w\in\mathbb{Z}_q[[\tau]]^\times, u e = q c P ∏ P n P ⋅ w , n P ∈ m Z , w ∈ Z q [[ τ ] ] × ,
with c c c the order of u e u_e u e along the special fiber, which does not depend on x ‾ \overline{x} x . We put u ~ e ( x ‾ ) : = w ( 0 ) m o d q \widetilde{u}_e(\overline{x}):=w(0)\bmod q u e ( x ) := w ( 0 ) mod q in F q × / F q × m \mathbb{F}_q^\times/\mathbb{F}_q^{\times m} F q × / F q × m . Another parameter changes each P P P by a unit factor, which enters to a power in m Z m\mathbb{Z} m Z , so u ~ e ( x ‾ ) \widetilde{u}_e(\overline{x}) u e ( x ) is well defined. When no point of Σ \Sigma Σ reduces to x ‾ \overline{x} x , no P P P occurs, and u ~ e ( x ‾ ) \widetilde{u}_e(\overline{x}) u e ( x ) is the value at x ‾ \overline{x} x of the reduction u ‾ e \overline{u}_e u e of q − c u e q^{-c}u_e q − c u e . No condition on the reductions of the points of Σ \Sigma Σ is needed.
Proof of Theorem 1.3 (b). For a ∈ F ( S , m ) a\in F(S,m) a ∈ F ( S , m ) , X q ( a ) \mathcal{X}_q(a) X q ( a ) is the class of a q − v q ( a ) aq^{-v_q(a)} a q − v q ( a ) modulo q q q ; as m ∣ v q ( a ) m\mid v_q(a) m ∣ v q ( a ) , any other uniformizer in place of q q q gives the same class. Let x x x reduce to x ‾ \overline{x} x and t : = τ ( x ) ∈ q Z q t:=\tau(x)\in q\mathbb{Z}_q t := τ ( x ) ∈ q Z q . Then u e ( x ) = q c ∏ P P ( t ) n P w ( t ) u_e(x)=q^c\prod_P P(t)^{n_P}w(t) u e ( x ) = q c ∏ P P ( t ) n P w ( t ) with ∏ P P ( t ) n P ∈ Q q × m \prod_P P(t)^{n_P}\in\mathbb{Q}_q^{\times m} ∏ P P ( t ) n P ∈ Q q × m and w ( t ) ≡ w ( 0 ) ( m o d q ) w(t) \equiv w(0) \pmod q w ( t ) ≡ w ( 0 ) ( mod q ) . The same holds at y y y with the same c c c , so u e ( x ) / u e ( y ) u_e(x)/u_e(y) u e ( x ) / u e ( y ) is w x ( 0 ) ‾ / w y ( 0 ) ‾ \overline{w_x(0)}/\overline{w_y(0)} w x ( 0 ) / w y ( 0 ) times a unit congruent to 1 1 1 modulo q q q times an m m m -th power, and its symbol is u ~ e ( x ‾ ) / u ~ e ( y ‾ ) \widetilde{u}_e(\overline{x})/\widetilde{u}_e(\overline{y}) u e ( x ) / u e ( y ) . □
Remark 3.1. There is an exact sequence 0 → O F , S × / m → F ( S , m ) → Cl S ( F ) [ m ] → 0 0 \to\mathcal{O}_{F,S}^{\times}/m \to F(S,m) \to\operatorname{Cl}_S(F)[m] \to0 0 → O F , S × / m → F ( S , m ) → Cl S ( F ) [ m ] → 0 , where the second map sends a a a to the class of the ideal a \mathfrak{a} a with ( a ) = a m (\mathfrak{a}) = \mathfrak{a}^m ( a ) = a m away from S S S . So F ( S , m ) F(S,m) F ( S , m ) is explicit when # Cl S ( F ) \#\operatorname{Cl}_S(F) # Cl S ( F ) is prime to m m m and a subgroup of O F , S × \mathcal{O}_{F,S}^{\times} O F , S × of index prime to m m m is known.
Modular units at level 25
With the modular units of X X X , the hypotheses of Theorem 1.3 hold for m = 5 m=5 m = 5 , F = K F=K F = K and S S S the prime above 5 5 5 , and we determine V V V and K ( S , 5 ) K(S,5) K ( S , 5 ) , so that the theorem attaches to each rational point of X X X a label in Hom ( V , F 5 ) \operatorname{Hom}(V,\mathbb{F}_5) Hom ( V , F 5 ) .
For a = ( a 1 , a 2 ) ∈ Q 2 ∖ Z 2 a=(a_1,a_2)\in\mathbb{Q}^2\setminus\mathbb{Z}^2 a = ( a 1 , a 2 ) ∈ Q 2 ∖ Z 2 let g a g_a g a be the Siegel function, as in [33 §3.1] . On row vectors v = ( v 1 , v 2 ) ∈ ( Z / 25 ) 2 v=(v_1,v_2)\in(\mathbb{Z}/25)^2 v = ( v 1 , v 2 ) ∈ ( Z /25 ) 2 put Q ( v ) : = v 2 2 − 2 v 1 2 Q(v):=v_2^2-2v_1^2 Q ( v ) := v 2 2 − 2 v 1 2 . Then Q ( v g ) = det ( g ) Q ( v ) Q(vg)=\det(g)Q(v) Q ( v g ) = det ( g ) Q ( v ) for g ∈ C × g\in C^\times g ∈ C × and Q ( v s ) = Q ( v ) Q(vs)=Q(v) Q ( v s ) = Q ( v ) , so Q ( v g ) = ± det ( g ) Q ( v ) Q(vg)=\pm\det(g)Q(v) Q ( v g ) = ± det ( g ) Q ( v ) for g ∈ N g\in N g ∈ N . For n n n in I : = { 1 , 2 , 3 , 4 , 6 , 7 , 8 , 9 , 11 , 12 } I:=\{1,2,3,4,6,7,8,9,11,12\} I := { 1 , 2 , 3 , 4 , 6 , 7 , 8 , 9 , 11 , 12 } , a set of representatives of ( Z / 25 ) × / { ± 1 } (\mathbb{Z}/25)^\times/\{\pm1\} ( Z /25 ) × / { ± 1 } , let O n O_n O n be the set of the 30 classes ± v \pm v ± v with Q ( v ) = ± n Q(v)=\pm n Q ( v ) = ± n , and
U n : = ∏ ± v ∈ O n g v / 25 , A ( n ) : = 25 ∑ ± v ∈ O n 1 2 B 2 ( { v 1 / 25 } ) , U_n:=\prod_{\pm v\in O_n}g_{v/25},\qquad A(n):=25\sum_{\pm v\in O_n}\frac{1}{2}B_2(\{v_1/25\}), U n := ± v ∈ O n ∏ g v /25 , A ( n ) := 25 ± v ∈ O n ∑ 2 1 B 2 ({ v 1 /25 }) ,
with B 2 ( t ) = t 2 − t + 1 6 B_2(t)=t^2-t+\frac{1}{6} B 2 ( t ) = t 2 − t + 6 1 . By [33 Proposition 6 (i), (ii)] the choice of representatives changes U n U_n U n only by a constant, and by [33 Proposition 7] U n U_n U n has order A ( n ) A(n) A ( n ) at ∞ \infty ∞ in the parameter q 1 / 25 q^{1/25} q 1/25 . A computation in PARI/GP gives
( A ( n ) ) n ∈ I = ( 1 2 , 11 2 , − 9 2 , 1 2 , 1 2 , 11 2 , 1 2 , 1 2 , 1 2 , − 19 2 ) . \left(A(n)\right)_{n\in I}=\left(\frac{1}{2},\frac{11}{2},-\frac{9}{2},\frac{1}{2},\frac{1}{2},\frac{11}{2},\frac{1}{2},\frac{1}{2},\frac{1}{2},-\frac{19}{2}\right). ( A ( n ) ) n ∈ I = ( 2 1 , 2 11 , − 2 9 , 2 1 , 2 1 , 2 11 , 2 1 , 2 1 , 2 1 , − 2 19 ) .
all in 1 2 + Z \frac{1}{2}+\mathbb{Z} 2 1 + Z , of sum 0. Let Λ : = { e ∈ Z I : 2 ∣ ∑ n e n } \Lambda:=\{e\in\mathbb{Z}^{I}:2\mid\sum_n e_n\} Λ := { e ∈ Z I : 2 ∣ ∑ n e n } , with G G G acting by ( σ d e ) n : = e n / d (\sigma_d e)_n:=e_{n/d} ( σ d e ) n := e n / d .
Carlucci [11 ] writes t = t 1 + t 2 ϵ t=t_1+t_2\sqrt{\epsilon} t = t 1 + t 2 ϵ and [ t ] : = ( t 1 , t 2 ) / p k [t]:=(t_1,t_2)/p^k [ t ] := ( t 1 , t 2 ) / p k , for a squarefree ϵ ≡ 3 ( m o d 4 ) \epsilon\equiv3\pmod4 ϵ ≡ 3 ( mod 4 ) that is not a square modulo p p p . For ϵ ≡ 1 / 2 ( m o d 25 ) \epsilon\equiv1/2\pmod{25} ϵ ≡ 1/2 ( mod 25 ) , for instance ϵ = − 37 \epsilon=-37 ϵ = − 37 , his Cartan group at level 25 is C × C^\times C × , and t = v 1 + v 2 ϵ t=v_1+v_2\sqrt{\epsilon} t = v 1 + v 2 ϵ has [ t ] ≡ ± v / 25 ( m o d Z 2 ) [t]\equiv\pm v/25\pmod{\mathbb{Z}^2} [ t ] ≡ ± v /25 ( mod Z 2 ) and ∣ t ∣ = − Q ( v ) / 2 |t|=-Q(v)/2 ∣ t ∣ = − Q ( v ) /2 . So the functions U n U_n U n are, up to constants, his G n / 2 + G_{n/2}^+ G n /2 + , and Lemma 4.1 is the case p k = 25 p^k=25 p k = 25 of his results: for p ≠ 2 , 3 p\ne2,3 p = 2 , 3 , modulo constants, the modular units of X n s + ( p k ) X_{\mathrm{ns}}^+(p^k) X ns + ( p k ) are exactly the products ∏ h ( G h + ) e h \prod_h(G_h^+)^{e_h} ∏ h ( G h + ) e h with 12 / gcd ( 12 , p + 1 ) 12/\gcd(12,p+1) 12/ g cd( 12 , p + 1 ) dividing ∑ h e h \sum_h e_h ∑ h e h [11 Theorem 6.5] , a condition that at p = 5 p=5 p = 5 defines Λ \Lambda Λ . A constant multiple of each is defined over Q ( cos ( 2 π / p k ) ) \mathbb{Q}(\cos(2\pi/p^k)) Q ( cos ( 2 π / p k )) , which is K K K at p k = 25 p^k=25 p k = 25 , and the Galois action permutes the G h + G_h^+ G h + up to constants [11 Remark 6.6] . He describes the group of their divisors by a Stickelberger element built from the same sums of B 2 B_2 B 2 as the numbers A ( n ) A(n) A ( n ) [11 Main Theorem 7.1] . We include a proof in our notation, which gives the orders at the cusps c d c_d c d in the labeling used below.
Lemma 4.1. For each e ∈ Λ e\in\Lambda e ∈ Λ there is u e ∈ K ( X ) × u_e\in K(X)^\times u e ∈ K ( X ) × , unique up to K × K^\times K × , with
ord c d u e = ∑ n ∈ I e n A ( n / d ) ( d ∈ I ) , \operatorname{ord}_{c_d}u_e=\sum_{n\in I}e_nA(n/d)\qquad(d\in I), ord c d u e = n ∈ I ∑ e n A ( n / d ) ( d ∈ I ) , and u e u_e u e is a constant multiple of ∏ n ∈ I U n e n \prod_{n\in I}U_n^{e_n} ∏ n ∈ I U n e n . The choices can be made so that e ↦ u e e\mapsto u_e e ↦ u e is a homomorphism, and then σ ( u e ) ∈ K × u σ e \sigma(u_e)\in K^\times u_{\sigma e} σ ( u e ) ∈ K × u σ e for σ ∈ G \sigma\in G σ ∈ G .
Proof. Let f : = ∏ n U n e n f:=\prod_n U_n^{e_n} f := ∏ n U n e n . We first show that S ( n ) : = ∑ ± v ∈ O n v T v ≡ 0 ( m o d 25 ) S(n):=\sum_{\pm v\in O_n}v^{\mathsf{T}}v\equiv0\pmod{25} S ( n ) := ∑ ± v ∈ O n v T v ≡ 0 ( mod 25 ) , where v T v v^{\mathsf{T}}v v T v is the matrix ( v i v j ) (v_iv_j) ( v i v j ) . For g ∈ C × g\in C^\times g ∈ C × of determinant 1 1 1 , v ↦ v g v\mapsto vg v ↦ v g permutes O n O_n O n , so g T S ( n ) g = S ( n ) g^{\mathsf{T}}S(n)g=S(n) g T S ( n ) g = S ( n ) . With Δ : = diag ( − 2 , 1 ) \Delta:=\operatorname{diag}(-2,1) Δ := diag ( − 2 , 1 ) one has g T Δ − 1 g = det ( g ) Δ − 1 g^{\mathsf{T}}\Delta^{-1}g=\det(g)\Delta^{-1} g T Δ − 1 g = det ( g ) Δ − 1 for g ∈ C g\in C g ∈ C , and an element of order 3 3 3 in the group of determinant-one elements of C × C^\times C × , a cyclic group of order 30 30 30 , has a one-dimensional space of invariant symmetric matrices over F 5 \mathbb{F}_5 F 5 . As its order is prime to 5 5 5 , its invariant symmetric matrices over Z / 25 \mathbb{Z}/25 Z /25 are the multiples of Δ − 1 \Delta^{-1} Δ − 1 . So S ( n ) = c Δ − 1 S(n)=c\Delta^{-1} S ( n ) = c Δ − 1 . The matrix s ′ : = ( 1 − 2 1 − 1 ) = ( 1 + ξ ) s ∈ N s' := \begin{pmatrix}1&-2\\1&-1\end{pmatrix}=(1+\xi)s\in N s ′ := ( 1 1 − 2 − 1 ) = ( 1 + ξ ) s ∈ N has determinant 1 1 1 , permutes O n O_n O n , and satisfies s ′ T Δ − 1 s ′ = − Δ − 1 {s'}^{T}\Delta^{-1}s'=-\Delta^{-1} s ′ T Δ − 1 s ′ = − Δ − 1 . So S ( n ) = − S ( n ) S(n)=-S(n) S ( n ) = − S ( n ) and S ( n ) ≡ 0 S(n)\equiv0 S ( n ) ≡ 0 . Also f f f has 30 ∑ n e n ≡ 0 ( m o d 12 ) 30\sum_n e_n\equiv0\pmod{12} 30 ∑ n e n ≡ 0 ( mod 12 ) factors, counted with sign. By the theorem of Kubert and Lang [28 Chapter 3, Theorem 5.2] , [33 Theorem 8] , f f f is a modular unit for Γ ( 25 ) \Gamma(25) Γ ( 25 ) .
Let Γ X ⊂ SL 2 ( Z ) \Gamma_X\subset\operatorname{SL}_2(\mathbb{Z}) Γ X ⊂ SL 2 ( Z ) be the preimage of N ∩ SL 2 ( Z / 25 ) N\cap\operatorname{SL}_2(\mathbb{Z}/25) N ∩ SL 2 ( Z /25 ) . For γ ∈ Γ X \gamma\in\Gamma_X γ ∈ Γ X , v ↦ v γ v\mapsto v\gamma v ↦ v γ permutes the classes in each O n O_n O n up to sign, so f ∘ γ = c e ( γ ) f f\circ\gamma=c_e(\gamma)f f ∘ γ = c e ( γ ) f with c e ( γ ) ∈ C × c_e(\gamma)\in\mathbb{C}^{\times} c e ( γ ) ∈ C × [33 Proposition 6] . As f f f is invariant under Γ ( 25 ) \Gamma(25) Γ ( 25 ) , c e c_e c e is a character of ( N ∩ SL 2 ( Z / 25 ) ) / { ± 1 } (N\cap\operatorname{SL}_2(\mathbb{Z}/25))/\{\pm1\} ( N ∩ SL 2 ( Z /25 )) / { ± 1 } , a dihedral group of order 30 30 30 , so it is trivial on the rotations, a subgroup of odd order 15 15 15 . The reflection s ′ s' s ′ fixes 1 + i 1+i 1 + i in the upper half plane, and f ( 1 + i ) ≠ 0 f(1+i)\ne0 f ( 1 + i ) = 0 , so c e ( s ′ ) = 1 c_e(s')=1 c e ( s ′ ) = 1 and c e = 1 c_e=1 c e = 1 . Thus f f f is a function on X C X_{\mathbb{C}} X C with divisor supported on the cusps. Its divisor is K K K -rational by Lemma 2.1, and the line bundle it defines is trivial over C \mathbb{C} C , so it has a non-zero section over K K K : the divisor is that of some u e ∈ K ( X ) × u_e\in K(X)^{\times} u e ∈ K ( X ) × , and f ∈ C × u e f\in\mathbb{C}^{\times}u_e f ∈ C × u e .
For the orders, ord c d f = ord B d ∞ f = ord ∞ f ∘ B d \operatorname{ord}_{c_d}f=\operatorname{ord}_{B_d\infty}f=\operatorname{ord}_{\infty}f\circ B_d ord c d f = ord B d ∞ f = ord ∞ f ∘ B d , and by [33 Proposition 6 (iii)] f ∘ B d f\circ B_d f ∘ B d is a constant times ∏ n ∏ ± v ∈ O n q v B d / 25 e n \prod_n\prod_{\pm v\in O_n}q_{vB_d/25}^{e_n} ∏ n ∏ ± v ∈ O n q v B d /25 e n . The first coordinate of v B d = v g d − 1 diag ( 1 , d ) vB_d=vg_d^{-1}\operatorname{diag}(1,d) v B d = v g d − 1 diag ( 1 , d ) is that of v g d − 1 vg_d^{-1} v g d − 1 , and v ↦ v g d − 1 v\mapsto vg_d^{-1} v ↦ v g d − 1 carries O n O_n O n onto O n / d O_{n/d} O n / d . So ord c d U n = A ( n / d ) \operatorname{ord}_{c_d}U_n=A(n/d) ord c d U n = A ( n / d ) by [33 Proposition 7] . Choosing u e u_e u e on a basis of Λ \Lambda Λ gives a homomorphism. Finally div σ d ( u e ) = σ d ( div u e ) \operatorname{div}\sigma_d(u_e)=\sigma_d(\operatorname{div}u_e) div σ d ( u e ) = σ d ( div u e ) has order ∑ d ′ e n A ( n / d ′ ) \sum_{d'}e_nA(n/d') ∑ d ′ e n A ( n / d ′ ) at c d d ′ c_{dd'} c d d ′ , which is the order of u σ d e u_{\sigma_de} u σ d e there; two functions over K K K with the same divisor differ by a factor in K × K^{\times} K × . □ \square □
A computation in PARI/GP shows that the divisor map Λ → Z I \Lambda\to\mathbb{Z}^{I} Λ → Z I , e ↦ ( ord c d u e ) d e\mapsto(\operatorname{ord}_{c_d}u_e)_d e ↦ ( ord c d u e ) d , has elementary divisors 1 , 5 , 5 , 5 , 5 , 5 , 5 , 355 , 355 1,5,5,5,5,5,5,355,355 1 , 5 , 5 , 5 , 5 , 5 , 5 , 355 , 355 and a kernel of rank 1 1 1 , spanned by ( 1 , … , 1 ) (1,\ldots,1) ( 1 , … , 1 ) (the product of all the G h G_h G h is constant, as Carlucci notes after [11 Theorem 6.5] ). So in the notation of Theorem 1.3, with m = 5 m=5 m = 5 , Λ 5 \Lambda_5 Λ 5 has index 5 5 5 in Λ \Lambda Λ , Λ 0 = Z ( 1 , … , 1 ) \Lambda_0=\mathbb{Z}(1,\ldots,1) Λ 0 = Z ( 1 , … , 1 ) , and V : = Λ 5 / ( 5 Λ + Λ 0 ) V:=\Lambda_5/(5\Lambda+\Lambda_0) V := Λ 5 / ( 5Λ + Λ 0 ) has dimension 8 8 8 over F 5 \mathbb{F}_5 F 5 . At 71 71 71 , the only other prime dividing these elementary divisors, V V V would have dimension 2 2 2 .
The group K ( S , 5 ) K(S,5) K ( S , 5 ) . Let p 5 \mathfrak{p}_5 p 5 be the prime of K K K above 5 5 5 and S : = { p 5 } S:=\{\mathfrak{p}_5\} S := { p 5 } . Let ζ : = ζ 25 \zeta:=\zeta_{25} ζ := ζ 25 , η : = ( 1 − ζ ) ( 1 − ζ − 1 ) \eta:=(1-\zeta)(1-\zeta^{-1}) η := ( 1 − ζ ) ( 1 − ζ − 1 ) , and for a prime to 5 5 5 let ϑ a : = ζ ( 1 − a ) / 2 ( 1 − ζ a ) / ( 1 − ζ ) \vartheta_a:=\zeta^{(1-a)/2}(1-\zeta^a)/(1-\zeta) ϑ a := ζ ( 1 − a ) /2 ( 1 − ζ a ) / ( 1 − ζ ) , a unit of K K K with ϑ a 2 = σ a ( η ) / η \vartheta_a^2=\sigma_a(\eta)/\eta ϑ a 2 = σ a ( η ) / η .
Proposition 4.2. The class number of K K K is prime to 5 5 5 , and K ( S , 5 ) = O K , S × / 5 K(S,5)=O_{K,S}^{\times}/5 K ( S , 5 ) = O K , S × /5 , a free F 5 [ G ] \mathbb{F}_5[G] F 5 [ G ] -module of rank one on η \eta η .
Proof. The extension K / Q ( 5 ) K/\mathbb{Q}(\sqrt{5}) K / Q ( 5 ) is cyclic of degree 5 5 5 , with group Γ \Gamma Γ . It is totally ramified at the prime ( 5 ) (\sqrt{5}) ( 5 ) , as 5 5 5 is totally ramified in Q ( ζ 25 ) \mathbb{Q}(\zeta_{25}) Q ( ζ 25 ) , unramified at the other finite primes, and both fields are totally real. By the ambiguous class number formula [31 Theorem 1 (1)] , with one ramified place and h ( Q ( 5 ) ) = 1 h(\mathbb{Q}(\sqrt{5}))=1 h ( Q ( 5 )) = 1 , the group Cl ( K ) Γ \operatorname{Cl}(K)^{\Gamma} Cl ( K ) Γ of ambiguous classes has order 1 / ( E : E ∩ N K × ) 1/(E:E\cap NK^{\times}) 1/ ( E : E ∩ N K × ) , where E E E is the unit group of Q ( 5 ) \mathbb{Q}(\sqrt{5}) Q ( 5 ) ; so it is trivial. A group of order 5 5 5 acting on a non-trivial finite 5 5 5 -group has a non-trivial fixed point, so Cl ( K ) [ 5 ∞ ] = 0 \operatorname{Cl}(K)[5^{\infty}]=0 Cl ( K ) [ 5 ∞ ] = 0 . By Remark 3.1, K ( S , 5 ) = O K , S × / 5 K(S,5)=O_{K,S}^{\times}/5 K ( S , 5 ) = O K , S × /5 .
The prime p 5 \mathfrak{p}_5 p 5 is ( η ) (\eta) ( η ) , so O K , S × = O K × × η Z O_{K,S}^{\times}=O_K^{\times}\times\eta^{\mathbb{Z}} O K , S × = O K × × η Z . By [43 Lemma 8.1 (a)] the cyclotomic units of K K K are generated by − 1 -1 − 1 and the nine ϑ a \vartheta_a ϑ a , 1 < a < 25 / 2 1<a<25/2 1 < a < 25/2 , 5 ∤ a 5\nmid a 5 ∤ a , and by [43 Theorem 8.2] they have index h ( K ) h(K) h ( K ) in O K × O_K^{\times} O K × . As ϑ a 2 = σ a ( η ) / η \vartheta_a^2=\sigma_a(\eta)/\eta ϑ a 2 = σ a ( η ) / η , the group generated by − 1 -1 − 1 and the conjugates of η \eta η generated by − 1 -1 − 1 , η \eta η and the ϑ a \vartheta_a ϑ a has index 2 9 2^9 2 9 in the group generated by − 1 -1 − 1 , η \eta η and the ϑ a \vartheta_a ϑ a . So it has index 2 9 h ( K ) 2^9h(K) 2 9 h ( K ) in O K , S × O_{K,S}^{\times} O K , S × , prime to 5 5 5 . As O K , S × O_{K,S}^{\times} O K , S × has rank 10 = ∣ G ∣ 10=|G| 10 = ∣ G ∣ , the map Z [ G ] → O K , S × / { ± 1 } \mathbb{Z}[G]\to O_{K,S}^{\times}/\{\pm1\} Z [ G ] → O K , S × / { ± 1 } , x ↦ η x x\mapsto\eta^x x ↦ η x , is injective, and as − 1 -1 − 1 is a fifth power it induces F 5 [ G ] ≅ O K , S × / 5 \mathbb{F}_5[G]\cong O_{K,S}^{\times}/5 F 5 [ G ] ≅ O K , S × /5 . □ \square □
By Lemmas 2.1 and 4.1, (U1) holds with F = K F = K F = K , Σ \Sigma Σ the cusps and Λ \Lambda Λ , u e u_e u e as above; (U2) holds for q ≠ p 5 \mathfrak{q} \ne\mathfrak{p}_5 q = p 5 with the model X ⊗ O K , q X \otimes\mathcal{O}_{K,\mathfrak{q}} X ⊗ O K , q , whose fibers are smooth and connected. For x ∈ X ( Q ) x \in X(\mathbb{Q}) x ∈ X ( Q ) , Theorem 1.3 and Proposition 4.2 give a G G G -map Φ x : V → F 5 [ G ] η \Phi_x : V \to\mathbb{F}_5[G]_\eta Φ x : V → F 5 [ G ] η . Writing Φ x ( e ) = ∑ d ∈ I ϕ x ( σ d − 1 e ) σ d η \Phi_x(e) = \sum_{d \in I} \phi_x(\sigma_d^{-1}e)\sigma_d\eta Φ x ( e ) = ∑ d ∈ I ϕ x ( σ d − 1 e ) σ d η identifies Hom G ( V , F 5 [ G ] η ) \operatorname{Hom}_G(V,\mathbb{F}_5[G]_\eta) Hom G ( V , F 5 [ G ] η ) with Hom ( V , F 5 ) \operatorname{Hom}(V,\mathbb{F}_5) Hom ( V , F 5 ) ; we call ϕ x ∈ Hom ( V , F 5 ) \phi_x \in\operatorname{Hom}(V,\mathbb{F}_5) ϕ x ∈ Hom ( V , F 5 ) the label of x x x . We fix a base point y ∈ X ( Q ) y \in X(\mathbb{Q}) y ∈ X ( Q ) ; any rational point would do, and in the computations it is the point v − 7 v_{-7} v − 7 of § 5 .
The labels and the residue discs
We read the label of a rational point off its reductions modulo primes q ≡ 1 ( m o d 25 ) q \equiv1 \pmod{25} q ≡ 1 ( mod 25 ) and prove Theorem 1.4 .
Let q ≡ 1 ( m o d 25 ) q \equiv1 \pmod{25} q ≡ 1 ( mod 25 ) be prime. Then q q q splits completely in K K K , and F q \mathbb{F}_q F q contains the 25 25 25 -th roots of unity, which is why we use these primes: for q ≡ − 1 ( m o d 25 ) q \equiv-1 \pmod{25} q ≡ − 1 ( mod 25 ) the group F q × / F q × 5 \mathbb{F}_q^\times/\mathbb{F}_q^{\times5} F q × / F q × 5 is trivial. Fix a prime of Q ( ζ ) \mathbb{Q}(\zeta) Q ( ζ ) above q q q , let ζ ˉ ∈ F q \bar{\zeta} \in\mathbb{F}_q ζ ˉ ∈ F q be the reduction of ζ \zeta ζ and q \mathfrak{q} q the prime of K K K below it. Identify F q × / F q × 5 \mathbb{F}_q^\times/\mathbb{F}_q^{\times5} F q × / F q × 5 with F 5 \mathbb{F}_5 F 5 by dlog, where dlog ( t ) = k \operatorname{dlog}(t) = k dlog ( t ) = k if t ( q − 1 ) / 5 = ζ ˉ 5 k t^{(q-1)/5} = \bar{\zeta}^{5k} t ( q − 1 ) /5 = ζ ˉ 5 k , and put
h d : = dlog ( ( 1 − ζ ˉ d ) ( 1 − ζ ˉ − d ) ) , ( Θ q ϕ ) ( e ) : = ∑ d ∈ I h d ϕ ( σ d − 1 e ) . h_d := \operatorname{dlog}\left((1-\bar{\zeta}^{d})(1-\bar{\zeta}^{-d})\right), \qquad(\Theta_q\phi)(e) := \sum_{d\in I} h_d\phi(\sigma_d^{-1}e). h d := dlog ( ( 1 − ζ ˉ d ) ( 1 − ζ ˉ − d ) ) , ( Θ q ϕ ) ( e ) := d ∈ I ∑ h d ϕ ( σ d − 1 e ) .
for ϕ ∈ Hom ( V , F 5 ) \phi\in\operatorname{Hom}(V,\mathbb{F}_5) ϕ ∈ Hom ( V , F 5 ) . For x ˉ ∈ X ( F q ) \bar{x} \in X(\mathbb{F}_q) x ˉ ∈ X ( F q ) let r x ˉ ∈ Hom ( V , F 5 ) r_{\bar{x}} \in\operatorname{Hom}(V,\mathbb{F}_5) r x ˉ ∈ Hom ( V , F 5 ) be e ↦ dlog ( u ~ e ( x ˉ ) / u ~ e ( y ˉ ) ) e \mapsto\operatorname{dlog}(\widetilde{u}_e(\bar{x})/\widetilde{u}_e(\bar{y})) e ↦ dlog ( u e ( x ˉ ) / u e ( y ˉ )) . By Theorem 1.3 (b), every x ∈ X ( Q ) x \in X(\mathbb{Q}) x ∈ X ( Q ) satisfies
Θ q ϕ x = r x ˉ . (1) \Theta_q\phi_x = r_{\bar{x}}. \tag*{(1)} Θ q ϕ x = r x ˉ . (1)
The map r x ˉ r_{\bar{x}} r x ˉ is the Tate pairing over F q \mathbb{F}_q F q of x ˉ − y ˉ \bar{x}-\bar{y} x ˉ − y ˉ with the classes 1 5 div u ˉ e \frac{1}{5}\operatorname{div}\bar{u}_e 5 1 div u ˉ e , so with 5 8 5^8 5 8 labels and about q q q points, heuristically a label that is not that of a rational point survives a prime with probability about q / 5 8 q/5^8 q / 5 8 .
The model. We compute on the canonical model. For v ∈ ( Z / 25 ) 2 v \in(\mathbb{Z}/25)^2 v ∈ ( Z /25 ) 2 not divisible by 5 5 5 , Hecke’s Eisenstein series e v e_v e v of weight one, defined as in [23 §5] with 121 121 121 replaced by 25 25 25 , has its expansion in 1 50 Z [ ζ 25 ] [ [ q 1 / 25 ] ] \frac{1}{50}\mathbb{Z}[\zeta_{25}][[q^{1/25}]] 50 1 Z [ ζ 25 ] [[ q 1/25 ]] and satisfies e v g = e v g e_v^g=e_{vg} e v g = e v g for the semilinear action of G G G . So, for v ′ v' v ′ also not divisible by 5 5 5 and s ∈ Z / 25 s \in\mathbb{Z}/25 s ∈ Z /25 , ∑ g ∈ N / { ± 1 } ζ 25 s det g e v g e v ′ g \sum_{g\in N/\{\pm1\}}\zeta_{25}^{s\det g}e_{vg}e_{v'g} ∑ g ∈ N / { ± 1 } ζ 25 s d e t g e v g e v ′ g is invariant under N N N , a form of weight 2 on X X X over Q \mathbb{Q} Q . A computation shows that forty of these span the forms of weight 2 on X X X , a space of dimension 23. As G G G permutes the cusps transitively, the combinations with constant term 0 at ∞ \infty ∞ are cusp forms, and a basis of those with expansions in Z [ ζ 25 ] [ [ q 1 / 25 ] ] \mathbb{Z}[\zeta_{25}][[q^{1/25}]] Z [ ζ 25 ] [[ q 1/25 ]] gives fourteen cusp forms of weight 2 on X X X , defined over Q \mathbb{Q} Q and integral away from 5. They are independent modulo each of the eighteen primes q ≡ 1 ( m o d 25 ) q \equiv1 \pmod{25} q ≡ 1 ( mod 25 ) with q ≤ 2351 q \le2351 q ≤ 2351 , so their reductions are a basis of H 0 ( X F q , Ω 1 ) H^0(X_{\mathbb{F}_q},\Omega^1) H 0 ( X F q , Ω 1 ) . Integrality and identities of forms of weight 2 are decided by finitely many coefficients: a form of weight 2 is a section of Ω 1 ( cusps ) \Omega^1(\text{cusps}) Ω 1 ( cusps ) , of degree 2 g − 2 + 10 = 36 2g-2+10=36 2 g − 2 + 10 = 36 , on X X X and on its reduction at each prime ℓ ≠ 5 \ell\ne5 ℓ = 5 , so a form whose first 37 coefficients at ∞ \infty ∞ vanish modulo ℓ \ell ℓ vanishes modulo ℓ \ell ℓ , and likewise modulo ℓ 2 \ell^2 ℓ 2 . These are checked on 160 coefficients, and on 60 in § 11 .
The products of the fourteen forms span a space of dimension 39 = 3 g − 3 =3g-3 = 3 g − 3 in weight 4, so X F q X_{\mathbb{F}_q} X F q is not hyperelliptic, the forms embed it in P 3 \mathbb{P}^3 P 3 , and the image lies on 66 = 105 − 39 =105-39 = 105 − 39 quadrics. By the theorem of Noether, Enriques and Petri [37 p. 157] , the forms map the symmetric algebra onto the canonical ring of X X X , with kernel generated in degrees 2 and 3. So the cubics through X X X form a space of dimension 560 − 5 ⋅ 13 = 495 -5\cdot13=495 − 5 ⋅ 13 = 495 , and as the products of the quadrics with the coordinates have rank 495 modulo 1151, they span it. Hence the quadrics generate the ideal of X X X and cut out X X X . The values of the forms at a CM point of discriminant D D D in the upper half plane (at D = − 3 D=-3 D = − 3 , where every form vanishes to order at least 2, their second derivatives) are proportional, to at least 90 digits, to an integer vector v D v_D v D that satisfies the quadrics, so v D ∈ X ( Q ) v_D \in X(\mathbb{Q}) v D ∈ X ( Q ) .
An exhaustive search on the quadrics over F q \mathbb{F}_q F q finds q + 1 − ∑ f a q ( f ) q+1-\sum_f a_q(f) q + 1 − ∑ f a q ( f ) points, the sum over the fourteen newforms f f f of Proposition 10.1 ; by [12 Theorem 1.4] this is # X ( F q ) \#X(\mathbb{F}_q) # X ( F q ) , so the search finds every point. Each u e u_e u e , for e e e in a basis of Λ 5 \Lambda_5 Λ 5 , is a quotient of two sections of 2 K X 2K_X 2 K X , found modulo q q q from the first coefficients of its expansion. This determines it, by Riemann–Roch: the pairs of sections with quotient u e u_e u e form a space of dimension 39 minus the degree of the poles of u e u_e u e , as that degree is at most 25, and the pairs that the coefficients allow form a space of the same dimension. Then u ~ e ( x ˉ ) \widetilde{u}_e(\bar{x}) u e ( x ˉ ) is computed at every x ˉ ∈ X ( F q ) \bar{x} \in X(\mathbb{F}_q) x ˉ ∈ X ( F q ) from the reductions of two such sections. Away from the cusps it is the quotient of their values, for a pair whose second section does not vanish at x ˉ \bar{x} x ˉ , and at the reduction of a cusp the quotient of their leading coefficients in a parameter there.
Proof of Theorem 1.4. The two smallest primes q ≡ 1 ( m o d 25 ) q \equiv1 \pmod{25} q ≡ 1 ( mod 25 ) are 101 and 151. A computation shows that Θ 101 \Theta_{101} Θ 101 and Θ 151 \Theta_{151} Θ 151 are invertible; that the 114 points of X ( F 101 ) X(\mathbb{F}_{101}) X ( F 101 ) have 114 distinct values of Θ 101 − 1 r x ˉ \Theta_{101}^{-1}r_{\bar{x}} Θ 101 − 1 r x ˉ ; that the 139 points of X ( F 151 ) X(\mathbb{F}_{151}) X ( F 151 ) have 136 distinct values of Θ 151 − 1 r x ˉ \Theta_{151}^{-1}r_{\bar{x}} Θ 151 − 1 r x ˉ ; and that exactly nine values are common to both lists, the labels ϕ v D \phi_{v_D} ϕ v D . So by (1) every x ∈ X ( Q ) x \in X(\mathbb{Q}) x ∈ X ( Q ) has ϕ x = ϕ v D \phi_x=\phi_{v_D} ϕ x = ϕ v D for some D ∈ D D \in\mathcal{D} D ∈ D , and at each of the eighteen primes a computation shows that the only x ˉ ∈ X ( F q ) \bar{x} \in X(\mathbb{F}_q) x ˉ ∈ X ( F q ) with r x ˉ = Θ q ϕ v D r_{\bar{x}}=\Theta_q\phi_{v_D} r x ˉ = Θ q ϕ v D is v ˉ D \bar{v}_D v ˉ D , so x ˉ = v ˉ D \bar{x}=\bar{v}_D x ˉ = v ˉ D . The nine points y D y_D y D are rational and distinct modulo 1151, as their j j j -invariants are (§ 11 ), and each of the nine labels has one point in its fiber at 1151; so each of the nine labels is that of exactly one y D y_D y D , and x ˉ \bar{x} x ˉ is the reduction of that y D y_D y D at each of the eighteen primes. □
The proof gives more: x x x and the same y D y_D y D have the same reduction modulo each of the eighteen primes q ≡ 1 ( m o d 25 ) q \equiv1 \pmod{25} q ≡ 1 ( mod 25 ) with q ≤ 2351 q \le2351 q ≤ 2351 .
Correspondences that change only the Cartan subring
For ν ∈ F 5 × \nu\in\mathbb{F}_5^\times ν ∈ F 5 × let R ν ⊂ Ω × Ω R_\nu\subset\Omega\times\Omega R ν ⊂ Ω × Ω be the set of pairs ( ( Z / 25 ) [ x ] , ( Z / 25 ) [ x ′ ] ) \bigl((\mathbb{Z}/25)[x],(\mathbb{Z}/25)[x']\bigr) ( ( Z /25 ) [ x ] , ( Z /25 ) [ x ′ ] ) with tr x = tr x ′ = 0 \operatorname{tr} x=\operatorname{tr} x'=0 tr x = tr x ′ = 0 , x 2 = x ′ 2 = 2 x^2=x'^2=2 x 2 = x ′2 = 2 and x ′ = x + 5 w x'=x+5w x ′ = x + 5 w , where w x ≡ − x w wx\equiv-xw w x ≡ − x w and w 2 ≡ ν ( m o d 5 ) w^2\equiv\nu\pmod{5} w 2 ≡ ν ( mod 5 ) . Here w w w is determined modulo 5. The relation is invariant under conjugation by G \mathcal{G} G and symmetric, as x = x ′ − 5 w x=x'-5w x = x ′ − 5 w and − w -w − w anticommutes with x ′ x' x ′ modulo 5. Let S ν S_\nu S ν be the correspondence ( E , R ) ↦ ∑ ( E , R ′ ) (E,R)\mapsto\sum_{(E,R')} ( E , R ) ↦ ∑ ( E , R ′ ) on X X X , the sum over the R ′ R' R ′ with ( ϕ − 1 R ϕ , ϕ − 1 R ′ ϕ ) ∈ R ν (\phi^{-1}R\phi,\phi^{-1}R'\phi)\in R_\nu ( ϕ − 1 R ϕ , ϕ − 1 R ′ ϕ ) ∈ R ν for some ϕ \phi ϕ . It is symmetric, defined over Q \mathbb{Q} Q and, by the same moduli description, over Z [ 1 / 5 ] \mathbb{Z}[1/5] Z [ 1/5 ] , and it commutes with the Hecke operators T n T_n T n , 5 ∤ n 5\nmid n 5 ∤ n , because an isogeny of degree prime to 5 carries E [ 25 ] E[25] E [ 25 ] , its Cartan subrings and R ν R_\nu R ν isomorphically. Put
Z 1 : = S 2 − S 3 , Z 2 : = S 1 − S 4 . Z_1:=S_2-S_3,\qquad Z_2:=S_1-S_4. Z 1 := S 2 − S 3 , Z 2 := S 1 − S 4 .
They are the real and imaginary parts of ∑ ν χ ‾ ( ν ) S ν \sum_{\nu}\overline{\chi}(\nu)S_\nu ∑ ν χ ( ν ) S ν for a quartic character χ \chi χ of F 5 × \mathbb{F}_5^\times F 5 × , whose non-zero eigenvalues on the Cartan subrings are squares of Gauss sums up to fourth roots of unity; that is what puts their eigenvalues in 5 Q \sqrt{5}\mathbb{Q} 5 Q , as Proposition 1.5 needs. Hypothesis (C) of Proposition 9.1 needs both here (Remark 9.2 ).
The fibers modulo 5. Fix x 1 x_1 x 1 with tr x 1 = 0 \operatorname{tr}x_1=0 tr x 1 = 0 and x 1 2 = 2 x_1^2=2 x 1 2 = 2 , and let x ˉ 1 \bar{x}_1 x ˉ 1 be its reduction, F 25 : = F 5 [ x ˉ 1 ] \mathbb{F}_{25}:=\mathbb{F}_5[\bar{x}_1] F 25 := F 5 [ x ˉ 1 ] , and A : = { u ∈ M 2 ( F 5 ) : u x ˉ 1 = − x ˉ 1 u } \mathcal{A}:=\{u\in M_2(\mathbb{F}_5):u\bar{x}_1=-\bar{x}_1u\} A := { u ∈ M 2 ( F 5 ) : u x ˉ 1 = − x ˉ 1 u } .
Lemma 6.1. (a) A = F 25 u 0 \mathcal{A}=\mathbb{F}_{25}u_0 A = F 25 u 0 for any non-zero u 0 ∈ A u_0\in\mathcal{A} u 0 ∈ A , and u 0 u_0 u 0 may be chosen with u 0 2 = 1 u_0^2=1 u 0 2 = 1 . Then ( γ u 0 ) 2 = N γ (\gamma u_0)^2=N\gamma ( γ u 0 ) 2 = N γ , and for a = α u 0 a=\alpha u_0 a = α u 0 , u = γ u 0 u=\gamma u_0 u = γ u 0 in A \mathcal{A} A the product a u = α γ ‾ au=\alpha\overline{\gamma} a u = α γ lies in F 25 \mathbb{F}_{25} F 25 , with tr ( a u ) = Tr F 25 / F 5 ( α γ ‾ ) \operatorname{tr}(au)=\operatorname{Tr}_{\mathbb{F}_{25}/\mathbb{F}_5}(\alpha\overline{\gamma}) tr ( a u ) = Tr F 25 / F 5 ( α γ ) .
(b) The Cartan subrings congruent to F 25 \mathbb{F}_{25} F 25 modulo 5 5 5 are the ( Z / 25 ) [ x 1 + 5 u ] (\mathbb{Z}/25)[x_1+5u] ( Z /25 ) [ x 1 + 5 u ] with u ∈ A u\in\mathcal{A} u ∈ A (lifted), and they are 25 25 25 distinct ones. In this parametrization R ν R_\nu R ν is the relation u ′ − u ∈ A ν : = { u ∈ A : u 2 = ν } u'-u\in\mathcal{A}_\nu:=\{u\in\mathcal{A}:u^2=\nu\} u ′ − u ∈ A ν := { u ∈ A : u 2 = ν } , a set of 6 6 6 elements.
(c) For a ∈ A a\in\mathcal{A} a ∈ A the function ψ a ( u ) : = ζ 5 tr ( a u ) \psi_a(u):=\zeta_5^{\operatorname{tr}(au)} ψ a ( u ) := ζ 5 tr ( a u ) on A \mathcal{A} A is an eigenfunction of the adjacency operator of R ν R_\nu R ν on this fiber, with eigenvalue 6 6 6 if a = 0 a=0 a = 0 and Υ ( a 2 ν ) \Upsilon(a^2\nu) Υ ( a 2 ν ) otherwise, where Υ ( m ) : = ∑ γ ∈ F 25 , N γ = m ζ 5 Tr γ \Upsilon(m):=\sum_{\gamma\in\mathbb{F}_{25},\,N\gamma=m}\zeta_5^{\operatorname{Tr}\gamma} Υ ( m ) := ∑ γ ∈ F 25 , N γ = m ζ 5 Tr γ . The 25 25 25 functions ψ a \psi_a ψ a are a basis.
Proof. (a) Conjugation by x ‾ 1 \overline{x}_1 x 1 is an involution of M 2 ( F 5 ) M_2(\mathbb{F}_5) M 2 ( F 5 ) with fixed space F 25 \mathbb{F}_{25} F 25 and ( − 1 ) (-1) ( − 1 ) -eigenspace A \mathcal{A} A , both of dimension 2 2 2 . For u 0 ∈ A u_0\in\mathcal{A} u 0 ∈ A non-zero, F 25 u 0 ⊂ A \mathbb{F}_{25}u_0\subset\mathcal{A} F 25 u 0 ⊂ A has dimension 2 2 2 , and u 0 γ = γ ‾ u 0 u_0\gamma=\overline{\gamma}u_0 u 0 γ = γ u 0 for γ ∈ F 25 \gamma\in\mathbb{F}_{25} γ ∈ F 25 . Elements of A \mathcal{A} A have trace 0 0 0 , so u 0 2 = − det u 0 u_0^2=-\det u_0 u 0 2 = − det u 0 is a scalar, and it is non-zero, as one sees on x ‾ 1 = ξ \overline{x}_1=\xi x 1 = ξ , u 0 = s u_0=s u 0 = s . Then ( γ u 0 ) 2 = N ( γ ) u 0 2 (\gamma u_0)^2=N(\gamma)u_0^2 ( γ u 0 ) 2 = N ( γ ) u 0 2 takes every value in F 5 × \mathbb{F}_5^\times F 5 × , so we may take u 0 2 = 1 u_0^2=1 u 0 2 = 1 , and α u 0 γ u 0 = α γ ‾ \alpha u_0\gamma u_0=\alpha\overline{\gamma} α u 0 γ u 0 = α γ . The trace of multiplication by an element of F 25 \mathbb{F}_{25} F 25 on F 5 2 \mathbb{F}_5^2 F 5 2 , a one-dimensional F 25 \mathbb{F}_{25} F 25 -space, is its trace to F 5 \mathbb{F}_5 F 5 .
(b) Such a Cartan subring has a unique generator x ≡ x ‾ 1 ( m o d 5 ) x\equiv\overline{x}_1\pmod5 x ≡ x 1 ( mod 5 ) with tr x = 0 \operatorname{tr}x=0 tr x = 0 and x 2 = 2 x^2=2 x 2 = 2 . Writing x = x 1 + 5 u x=x_1+5u x = x 1 + 5 u , x 2 = 2 x^2=2 x 2 = 2 means x 1 u + u x 1 ≡ 0 ( m o d 5 ) x_1u+ux_1\equiv0\pmod5 x 1 u + u x 1 ≡ 0 ( mod 5 ) , and then tr u ≡ 0 \operatorname{tr}u\equiv0 tr u ≡ 0 . For x ′ = x + 5 w x'=x+5w x ′ = x + 5 w the parameter becomes u + w u+w u + w . By (a), A ν = { γ u 0 : N γ = ν } \mathcal{A}_\nu=\{\gamma u_0:N\gamma=\nu\} A ν = { γ u 0 : N γ = ν } , which has 24 / 4 = 6 24/4=6 24/4 = 6 elements.
(c) ∑ w ∈ A ν ψ a ( u + w ) = ψ a ( u ) ∑ w ∈ A ν ζ 5 tr ( a w ) \sum_{w\in\mathcal{A}_\nu}\psi_a(u+w)=\psi_a(u)\sum_{w\in\mathcal{A}_\nu}\zeta_5^{\operatorname{tr}(aw)} ∑ w ∈ A ν ψ a ( u + w ) = ψ a ( u ) ∑ w ∈ A ν ζ 5 tr ( a w ) . For a = α u 0 ≠ 0 a=\alpha u_0\ne0 a = α u 0 = 0 and w = γ u 0 w=\gamma u_0 w = γ u 0 , β : = α γ ‾ \beta:=\alpha\overline{\gamma} β := α γ runs over the β \beta β with N β = N ( α ) ν = a 2 ν N\beta=N(\alpha)\nu=a^2\nu N β = N ( α ) ν = a 2 ν , and tr ( a w ) = Tr β \operatorname{tr}(aw)=\operatorname{Tr}\beta tr ( a w ) = Tr β . The pairing ( a , u ) ↦ tr ( a u ) (a,u)\mapsto\operatorname{tr}(au) ( a , u ) ↦ tr ( a u ) is non-degenerate on A \mathcal{A} A , so the ψ a \psi_a ψ a are distinct characters. □
The fibers of Ω \Omega Ω over the ten Cartan subrings of M 2 ( F 5 ) M_2(\mathbb{F}_5) M 2 ( F 5 ) are preserved by every R ν R_\nu R ν , so on Q [ Ω ] \mathbb{Q}[\Omega] Q [ Ω ] the adjacency operators A ν A_\nu A ν of the R ν R_\nu R ν commute. Put A Z 1 : = A 2 − A 3 A_{Z_1}:=A_2-A_3 A Z 1 := A 2 − A 3 and A Z 2 : = A 1 − A 4 A_{Z_2}:=A_1-A_4 A Z 2 := A 1 − A 4 , and let χ \chi χ be the character of F 5 × \mathbb{F}_5^\times F 5 × with χ ( 2 ) = i \chi(2)=i χ ( 2 ) = i . Then A Z 2 − i A Z 1 = ∑ ν χ ‾ ( ν ) A ν A_{Z_2}-iA_{Z_1}=\sum_\nu\overline{\chi}(\nu)A_\nu A Z 2 − i A Z 1 = ∑ ν χ ( ν ) A ν is 0 0 0 on ψ 0 \psi_0 ψ 0 , and on ψ a \psi_a ψ a , a ≠ 0 a\ne0 a = 0 , it is by Lemma 6.1 (c) the scalar χ ( a 2 ) ∑ γ ≠ 0 χ ‾ ( N γ ) ζ 5 Tr γ \chi(a^2)\sum_{\gamma\ne0}\overline{\chi}(N\gamma)\zeta_5^{\operatorname{Tr}\gamma} χ ( a 2 ) ∑ γ = 0 χ ( N γ ) ζ 5 Tr γ . By Hasse–Davenport [14 ] this sum is − g ( χ ‾ ) 2 -g(\overline{\chi})^2 − g ( χ ) 2 , where g g g is the Gauss sum over F 5 \mathbb{F}_5 F 5 , and g ( χ ‾ ) 2 = g ( χ 2 ) ∑ x χ ‾ ( x ) χ ‾ ( 1 − x ) = 5 ( − 1 + 2 i ) g(\overline{\chi})^2=g(\chi^2)\sum_x\overline{\chi}(x)\overline{\chi}(1-x)=\sqrt{5}(-1+2i) g ( χ ) 2 = g ( χ 2 ) ∑ x χ ( x ) χ ( 1 − x ) = 5 ( − 1 + 2 i ) , with 5 : = ζ 5 + ζ 5 − 1 − ζ 5 2 − ζ 5 − 2 \sqrt{5}:=\zeta_5+\zeta_5^{-1}-\zeta_5^2-\zeta_5^{-2} 5 := ζ 5 + ζ 5 − 1 − ζ 5 2 − ζ 5 − 2 the quadratic Gauss sum. So A Z 2 − i A Z 1 A_{Z_2}-iA_{Z_1} A Z 2 − i A Z 1 acts on ψ a \psi_a ψ a by χ ( a 2 ) 5 ( 1 − 2 i ) \chi(a^2)\sqrt{5}(1-2i) χ ( a 2 ) 5 ( 1 − 2 i ) . As A Z 1 A_{Z_1} A Z 1 and A Z 2 A_{Z_2} A Z 2 are real and symmetric, their eigenvalues on ψ a \psi_a ψ a are non-zero and lie in { ± 5 , ± 2 5 } \{\pm\sqrt{5},\pm2\sqrt{5}\} { ± 5 , ± 2 5 } . So the minimal polynomial P Z ∈ Q [ t ] P_Z\in\mathbb{Q}[t] P Z ∈ Q [ t ] of A Z A_Z A Z has distinct roots, all in 5 Q \sqrt{5}\mathbb{Q} 5 Q .
From the Cartan subrings to cohomology. Section 7 needs two facts about H : = H e ˊ t 1 ( X Q ‾ , Q ‾ ℓ ) H:=H^1_{\mathrm{\acute{e}t}}(X_{\overline{\mathbb{Q}}},\overline{\mathbb{Q}}_\ell) H := H e ˊ t 1 ( X Q , Q ℓ ) : the eigenvalues of Z 1 Z_1 Z 1 and Z 2 Z_2 Z 2 on H H H lie in 5 Q \sqrt{5}\mathbb{Q} 5 Q , which makes their traces vanish, and the scalar χ ( π ) \chi(\pi) χ ( π ) by which they act on the part σ ( π ) \sigma(\pi) σ ( π ) of H H H attached to an automorphic representation π \pi π depends only on π ∣ F 5 G L 2 ( F 5 ) \pi|_{\mathbb{F}_5\mathrm{GL}_2(\mathbb{F}_5)} π ∣ F 5 GL 2 ( F 5 ) , which Lemma 7.2 and Corollary 7.3 use. Both come from the action on the Cartan subrings.
Let k k k be a field of characteristic 0 0 0 . For a G \mathcal{G} G -stable subset R ⊂ Ω × Ω R\subset\Omega\times\Omega R ⊂ Ω × Ω let A R A_R A R be its adjacency operator on k [ Ω ] k[\Omega] k [ Ω ] , C ′ ↦ ∑ ( C ′ , C ′ ′ ) ∈ R C ′ ′ C'\mapsto\sum_{(C',C'')\in R}C'' C ′ ↦ ∑ ( C ′ , C ′′ ) ∈ R C ′′ , and D R : = { g ∈ G : ( C , g C g − 1 ) ∈ R } D_R:=\{g\in\mathcal{G}:(C,gCg^{-1})\in R\} D R := { g ∈ G : ( C , g C g − 1 ) ∈ R } , a union of double cosets of N N N .
Lemma 6.2. For a k [ G ] k[\mathcal{G}] k [ G ] -module W W W , w ↦ ( c g C − 1 ↦ g w ) w\mapsto(c_gC^{-1}\mapsto gw) w ↦ ( c g C − 1 ↦ g w ) is an isomorphism W N → Hom G ( k [ Ω ] , W ) W^N\to\operatorname{Hom}_{\mathcal{G}}(k[\Omega],W) W N → Hom G ( k [ Ω ] , W ) , with inverse Ψ ↦ Ψ ( C ) \Psi\mapsto\Psi(C) Ψ ↦ Ψ ( C ) . Under it, Ψ ↦ Ψ ∘ A R \Psi\mapsto\Psi\circ A_R Ψ ↦ Ψ ∘ A R corresponds to w ↦ ∑ h N ⊂ D R h w w\mapsto\sum_{hN\subset D_R}hw w ↦ ∑ h N ⊂ D R h w .
Proof. The map is well defined on W N W^N W N , as N N N is the stabilizer of C C C , and G \mathcal{G} G -equivariance gives the inverse. Then ( Ψ ∘ A R ) ( C ) = ∑ ( C , C ′ ′ ) ∈ R Ψ ( C ′ ′ ) = ∑ h N ⊂ D R h w (\Psi\circ A_R)(C)=\sum_{(C,C'')\in R}\Psi(C'')=\sum_{hN\subset D_R}hw ( Ψ ∘ A R ) ( C ) = ∑ ( C , C ′′ ) ∈ R Ψ ( C ′′ ) = ∑ h N ⊂ D R h w . □
Let H ( ⋅ ) : = H e ˊ t 1 ( Q ‾ , Q ‾ ℓ ) H(\cdot):=H^1_{\mathrm{\acute{e}t}}(\overline{\mathbb{Q}},\overline{\mathbb{Q}}_\ell) H ( ⋅ ) := H e ˊ t 1 ( Q , Q ℓ ) for a prime ℓ \ell ℓ , let pr : Y → X \operatorname{pr}:Y\to X pr : Y → X be the quotient map, and let G \mathcal{G} G act on H ( Y ) H(Y) H ( Y ) on the left by g ⋅ v : = r g ∗ v g\cdot v:=r_g^*v g ⋅ v := r g ∗ v . Write S R S_R S R for the correspondence of R R R on X X X , so that S ν = S R ν S_\nu=S_{R_\nu} S ν = S R ν , and S R ∗ S_R^* S R ∗ for its action on H ( X ) H(X) H ( X ) .
Lemma 6.3. The map pr ∗ \operatorname{pr}^* pr ∗ is an isomorphism H ( X ) → H ( Y ) N H(X)\to H(Y)^N H ( X ) → H ( Y ) N , and pr ∗ S R ∗ v = ∑ h N ⊂ D R h ⋅ pr ∗ v \operatorname{pr}^*S_R^*v=\sum_{hN\subset D_R}h\cdot\operatorname{pr}^*v pr ∗ S R ∗ v = ∑ h N ⊂ D R h ⋅ pr ∗ v .
Proof. The group N / { ± 1 } N/\{\pm1\} N / { ± 1 } acts faithfully on Y Y Y with quotient X X X , as ( E , ϕ ) ≅ ( E , − ϕ ) (E,\phi)\cong(E,-\phi) ( E , ϕ ) ≅ ( E , − ϕ ) , so pr ∗ \operatorname{pr}^* pr ∗ is an isomorphism onto the invariants and pr ∗ pr ∗ = ∑ n ∈ N / { ± 1 } r n ∗ \operatorname{pr}^*\operatorname{pr}_*=\sum_{n\in N/\{\pm1\}}r_n^* pr ∗ pr ∗ = ∑ n ∈ N / { ± 1 } r n ∗ . For g ∈ D R g\in D_R g ∈ D R , the map ( pr , pr ∘ r g ) : Y → X × X (\operatorname{pr},\operatorname{pr}\circ r_g):Y\to X\times X ( pr , pr ∘ r g ) : Y → X × X sends ( E , ϕ ) (E,\phi) ( E , ϕ ) to ( ( E , ϕ C ϕ − 1 ) , ( E , ϕ g C g − 1 ϕ − 1 ) ) ((E,\phi C\phi^{-1}),(E,\phi gCg^{-1}\phi^{-1})) (( E , ϕC ϕ − 1 ) , ( E , ϕ g C g − 1 ϕ − 1 )) . Its image S g S_g S g depends only on N g N NgN N g N , it has degree ∣ N ∩ g N g − 1 ∣ / 2 |N\cap gNg^{-1}|/2 ∣ N ∩ g N g − 1 ∣/2 onto S g S_g S g , since ( E , ϕ ) (E,\phi) ( E , ϕ ) and ( E , ϕ n ) (E,\phi n) ( E , ϕ n ) have the same image exactly when n ∈ N ∩ g N g − 1 n\in N\cap gNg^{-1} n ∈ N ∩ g N g − 1 , and S R = ∑ S g S_R=\sum S_g S R = ∑ S g over representatives g g g of the double cosets in D R D_R D R . So S g ∗ = 2 ∣ N ∩ g N g − 1 ∣ pr ∗ r g ∗ pr ∗ S_g^*=\frac{2}{|N\cap gNg^{-1}|}\operatorname{pr}_*r_g^*\operatorname{pr}^* S g ∗ = ∣ N ∩ g N g − 1 ∣ 2 pr ∗ r g ∗ pr ∗ , and as r n ∗ r g ∗ = ( n g ) ∗ r_n^*r_g^*=(ng)^* r n ∗ r g ∗ = ( n g ) ∗ ,
pr ∗ S g ∗ v = 1 ∣ N ∩ g N g − 1 ∣ ∑ n ∈ N ( n g ) ⋅ pr ∗ v = ∑ h N ⊂ N g N h ⋅ pr ∗ v . \operatorname{pr}^*S_g^*v=\frac{1}{|N\cap gNg^{-1}|}\sum_{n\in N}(ng)\cdot\operatorname{pr}^*v=\sum_{hN\subset NgN}h\cdot\operatorname{pr}^*v. pr ∗ S g ∗ v = ∣ N ∩ g N g − 1 ∣ 1 n ∈ N ∑ ( n g ) ⋅ pr ∗ v = h N ⊂ N g N ∑ h ⋅ pr ∗ v .
because n g N = n ′ g N ngN=n'gN n g N = n ′ g N if and only if n − 1 n ′ ∈ N ∩ g N g − 1 n^{-1}n'\in N\cap gNg^{-1} n − 1 n ′ ∈ N ∩ g N g − 1 .
By Lemmas 6.2 and 6.3, pr ∗ Z ∗ = ρ ( A Z ) pr ∗ \operatorname{pr}^*Z^*=\rho(A_Z)\operatorname{pr}^* pr ∗ Z ∗ = ρ ( A Z ) pr ∗ on H ( X ) H(X) H ( X ) for Z ∈ { Z 1 , Z 2 } Z\in\{Z_1,Z_2\} Z ∈ { Z 1 , Z 2 } , where ρ ( A ) \rho(A) ρ ( A ) is Ψ ↦ Ψ ∘ A \Psi\mapsto\Psi\circ A Ψ ↦ Ψ ∘ A on Hom G ( Q ‾ ℓ [ Ω ] , H ( Y ) ) = H ( Y ) N \operatorname{Hom}_{\mathcal{G}}(\overline{\mathbb{Q}}_\ell[\Omega],H(Y))=H(Y)^N Hom G ( Q ℓ [ Ω ] , H ( Y )) = H ( Y ) N . As A ↦ ρ ( A ) A\mapsto\rho(A) A ↦ ρ ( A ) reverses products and Q ‾ ℓ [ A Z ] \overline{\mathbb{Q}}_\ell[A_Z] Q ℓ [ A Z ] is commutative, P Z ( Z ∗ ) = 0 P_Z(Z^*)=0 P Z ( Z ∗ ) = 0 on H ( X ) H(X) H ( X ) .
Proposition 6.4. Let Z ∈ { Z 1 , Z 2 } Z\in\{Z_1,Z_2\} Z ∈ { Z 1 , Z 2 } and H : = H e ˊ t 1 ( X Q ‾ , Q ‾ ℓ ) H:=H^1_{\mathrm{\acute{e}t}}(X_{\overline{\mathbb{Q}}},\overline{\mathbb{Q}}_\ell) H := H e ˊ t 1 ( X Q , Q ℓ ) .
(a) Z Z Z acts semisimply and injectively on H H H with eigenvalues in 5 Q \sqrt{5}\mathbb{Q} 5 Q , and Tr ( Z ∣ H ) = 0 \operatorname{Tr}(Z\mid H)=0 Tr ( Z ∣ H ) = 0 .
(b) H = ⨁ π ∈ P σ ( π ) H=\bigoplus_{\pi\in P}\sigma(\pi) H = ⨁ π ∈ P σ ( π ) for a set P P P of fourteen cuspidal automorphic representations π \pi π of G L 2 ( A Q ) \mathrm{GL}_2(\mathbb{A}_{\mathbb{Q}}) GL 2 ( A Q ) of weight 2 2 2 whose central characters are trivial on Z 5 × \mathbb{Z}_5^\times Z 5 × , where σ ( π ) \sigma(\pi) σ ( π ) is the two-dimensional ℓ \ell ℓ -adic representation attached to π \pi π . The correspondence Z Z Z acts on σ ( π ) \sigma(\pi) σ ( π ) by a scalar z ( π ) ∈ 5 Q z(\pi)\in\sqrt{5}\mathbb{Q} z ( π ) ∈ 5 Q that depends only on the restriction of the local component π 5 \pi_5 π 5 to G L 2 ( Z 5 ) \mathrm{GL}_2(\mathbb{Z}_5) GL 2 ( Z 5 ) .
Proof. (a) P Z ( Z ) = 0 P_Z(Z)=0 P Z ( Z ) = 0 with P Z P_Z P Z as above. As A Z A_Z A Z is 0 0 0 on ψ 0 \psi_0 ψ 0 and has non-zero eigenvalues on the ψ a \psi_a ψ a , a ≠ 0 a\ne0 a = 0 , the kernel of A Z A_Z A Z on Q ‾ ℓ [ Ω ] \overline{\mathbb{Q}}_\ell[\Omega] Q ℓ [ Ω ] consists of the functions constant on the fibers over the Cartan subrings of M 2 ( F 5 ) M_2(\mathbb{F}_5) M 2 ( F 5 ) . If Z Z Z kills v ∈ H ( X ) v\in H(X) v ∈ H ( X ) , the Ψ \Psi Ψ corresponding to w : = pr ∗ v w:=\operatorname{pr}^*v w := pr ∗ v (Lemmas 6.2 and 6.3) vanishes on the image of A Z A_Z A Z , which contains C ′ − C ′ ′ C'-C'' C ′ − C ′′ whenever C ′ ≡ C ′ ′ ( m o d 5 ) C'\equiv C''\pmod5 C ′ ≡ C ′′ ( mod 5 ) (as A Z A_Z A Z is symmetric, its image is the orthogonal of its kernel). So w = Ψ ( C ) w=\Psi(C) w = Ψ ( C ) is invariant under the preimage N ′ N' N ′ of N n s ( 5 ) N_{\mathrm{ns}}(5) N ns ( 5 ) in G \mathcal{G} G , the stabilizer of C C C modulo 5 5 5 , and as in Lemma 6.3, H ( Y ) N ′ H(Y)^{N'} H ( Y ) N ′ is the pull-back of H ( X n s + ( 5 ) ) H(X^+_{\mathrm{ns}}(5)) H ( X ns + ( 5 )) , which is 0 0 0 as X n s + ( 5 ) X^+_{\mathrm{ns}}(5) X ns + ( 5 ) has genus 0 0 0 . Hence w = 0 w=0 w = 0 , so v = 0 v=0 v = 0 , and Z Z Z is injective on H H H . The trace is an integer, since Z Z Z acts on H 1 ( X ( C ) , Z ) H^1(X(\mathbb{C}),\mathbb{Z}) H 1 ( X ( C ) , Z ) , and it lies in 5 Q \sqrt{5}\mathbb{Q} 5 Q , so it is 0 0 0 .
(b) Let K ( 25 ) : = ker ( G L 2 ( Z ^ ) → G ) K(25):=\ker(\mathrm{GL}_2(\widehat{\mathbb{Z}})\to\mathcal{G}) K ( 25 ) := ker ( GL 2 ( Z ) → G ) . By [10 §2.3, p. 421] (see [30 ] for Q \mathbb{Q} Q ), there is an isomorphism of G × Gal ( Q ‾ / Q ) \mathcal{G}\times\operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q}) G × Gal ( Q / Q ) -modules
H ( Y ) ≅ ⨁ π V π θ ⊗ σ ( π ) , V π : = π f i n K ( 25 ) . H(Y)\cong\bigoplus_\pi V_\pi^\theta\otimes\sigma(\pi),\qquad V_\pi:=\pi_{\mathrm{fin}}^{K(25)}. H ( Y ) ≅ π ⨁ V π θ ⊗ σ ( π ) , V π := π fin K ( 25 ) .
over the cuspidal automorphic representations π \pi π of G L 2 \mathrm{GL}_2 GL 2 of weight 2 2 2 , where V θ V^\theta V θ denotes V V V with G \mathcal{G} G acting through an automorphism θ \theta θ of G \mathcal{G} G that depends only on the conventions for the two actions; nothing below depends on θ \theta θ . As a G \mathcal{G} G -module, V π V_\pi V π is 0 0 0 unless π \pi π is unramified outside 5 5 5 , and then it is π 5 ker ( G L 2 ( Z 5 ) → G ) \pi_5^{\ker(\mathrm{GL}_2(\mathbb{Z}_5)\to\mathcal{G})} π 5 k e r ( GL 2 ( Z 5 ) → G ) tensored with lines, so it depends only on π 5 ∣ G L 2 ( Z 5 ) \pi_5|_{\mathrm{GL}_2(\mathbb{Z}_5)} π 5 ∣ GL 2 ( Z 5 ) . Taking N N N -invariants, H ( X ) ≅ ⨁ π ( V π θ ) N ⊗ σ ( π ) H(X)\cong\bigoplus_\pi(V_\pi^\theta)^N\otimes\sigma(\pi) H ( X ) ≅ ⨁ π ( V π θ ) N ⊗ σ ( π ) . By Chen’s isogeny [12 Theorem 1.4] (proof of Lemma 7.1 , H H H is the sum of the irreducible representations σ ( f ) \sigma(f) σ ( f ) over the fourteen newforms f f f of level 625 with trivial character and Fricke eigenvalue + 1 +1 + 1 , and these are pairwise non-isomorphic by strong multiplicity one. Hence ( V π θ ) N (V_\pi^\theta)^N ( V π θ ) N has dimension at most 1 for every π \pi π ; let P P P be the fourteen π \pi π for which it is a line. The center of G G G lies in N N N , is preserved by θ \theta θ , and acts on V π V_\pi V π through the central character of π 5 \pi_5 π 5 , so that character is trivial on Z 5 × Z_5^\times Z 5 × for π ∈ P \pi\in P π ∈ P . By Lemmas 6.2 and 6.3, Z Z Z acts on ( V π θ ) N ⊗ σ ( π ) = Hom G ( Q ‾ [ Ω ] , V π θ ) ⊗ σ ( π ) (V_\pi^\theta)^N \otimes\sigma(\pi) = \operatorname{Hom}_G(\overline{\mathbb{Q}}[\Omega], V_\pi^\theta) \otimes\sigma(\pi) ( V π θ ) N ⊗ σ ( π ) = Hom G ( Q [ Ω ] , V π θ ) ⊗ σ ( π ) through ρ ( A Z ) \rho(A_Z) ρ ( A Z ) on the first factor, a line, so by a scalar z ( π ) z(\pi) z ( π ) that depends only on the G G G -module V π θ V_\pi^\theta V π θ . It lies in 5 Q \sqrt{5}\mathbb{Q} 5 Q by (a).
Fixed points. For μ ∈ F 5 × \mu\in\mathbb{F}_5^\times μ ∈ F 5 × , the six Cartan subrings S μ S_\mu S μ -related to that of y − 3 y_{-3} y − 3 form two orbits of size three under the automorphism of order 3 of E − 3 E_{-3} E − 3 (see the proof below), so S μ ( y − 3 ) = 3 Q μ + 3 Q μ ′ S_\mu(y_{-3}) = 3Q_\mu+ 3Q'_\mu S μ ( y − 3 ) = 3 Q μ + 3 Q μ ′ with Q μ ≠ Q μ ′ Q_\mu\ne Q'_\mu Q μ = Q μ ′ over j = 0 j = 0 j = 0 . For ν ∈ { 1 , 4 } \nu\in\{1,4\} ν ∈ { 1 , 4 } let Φ ν \Phi_\nu Φ ν be the sum of the ten points over j = 1728 j = 1728 j = 1728 described in (c) below.
Proposition 6.5. (a) Fix S ν S_\nu S ν has no point over j ≠ 0 , 1728 , ∞ j \ne0, 1728, \infty j = 0 , 1728 , ∞ .
(b) Over j = 0 j = 0 j = 0 , Fix S ν = 2 Q 2 ν + 2 Q 2 ν ′ \operatorname{Fix} S_\nu= 2Q_{2\nu} + 2Q'_{2\nu} Fix S ν = 2 Q 2 ν + 2 Q 2 ν ′ .
(c) Over j = 1728 j = 1728 j = 1728 , Fix S ν = 0 \operatorname{Fix} S_\nu= 0 Fix S ν = 0 for ν = 2 , 3 \nu= 2,3 ν = 2 , 3 , and Fix S ν = Φ ν \operatorname{Fix} S_\nu= \Phi_\nu Fix S ν = Φ ν , ten distinct points of multiplicity one, for ν = 1 , 4 \nu= 1,4 ν = 1 , 4 .
(d) Each cusp has multiplicity 2 in Fix S 1 \operatorname{Fix} S_1 Fix S 1 and Fix S 4 \operatorname{Fix} S_4 Fix S 4 , and 0 in Fix S 2 \operatorname{Fix} S_2 Fix S 2 and Fix S 3 \operatorname{Fix} S_3 Fix S 3 . Hence Fix Z 1 = 2 ( Q 4 + Q 4 ′ − Q 1 − Q 1 ′ ) \operatorname{Fix} Z_1 = 2(Q_4 + Q'_4 - Q_1 - Q'_1) Fix Z 1 = 2 ( Q 4 + Q 4 ′ − Q 1 − Q 1 ′ ) and Fix Z 2 = 2 ( Q 2 + Q 2 ′ − Q 3 − Q 3 ′ ) + Φ 1 − Φ 4 \operatorname{Fix} Z_2 = 2(Q_2 + Q'_2 - Q_3 - Q'_3) + \Phi_1 - \Phi_4 Fix Z 2 = 2 ( Q 2 + Q 2 ′ − Q 3 − Q 3 ′ ) + Φ 1 − Φ 4 .
The same holds over F ‾ q \overline{\mathbb{F}}_q F q for every prime q ≥ 7 q \ge7 q ≥ 7 , for the reductions of these points.
Proof. A point ( E , R ) (E,R) ( E , R ) is fixed when some R ′ = α R α − 1 R' = \alpha R\alpha^{-1} R ′ = α R α − 1 with α ∈ Aut ( E ) \alpha\in\operatorname{Aut}(E) α ∈ Aut ( E ) is S ν S_\nu S ν -related to R R R ; as ( R , R ) ∉ R ν (R,R) \notin R_\nu ( R , R ) ∈ / R ν , α ≠ ± 1 \alpha\ne\pm1 α = ± 1 . This gives (a). As Aut ( E ) / { ± 1 } \operatorname{Aut}(E)/\{\pm1\} Aut ( E ) / { ± 1 } has order 2 or 3 and α R α − 1 ≠ R \alpha R\alpha^{-1} \ne R α R α − 1 = R , only ± 1 \pm1 ± 1 fixes R R R , so a parameter t t t of X X X at ( E , R ) (E,R) ( E , R ) is a parameter of the stack. Each such α \alpha α , up to sign, gives a branch of S ν S_\nu S ν through ( ( E , R ) , ( E , R ) ) ((E,R),(E,R)) (( E , R ) , ( E , R )) which, in t t t , is the graph of t ↦ ϵ t t \mapsto\epsilon t t ↦ ϵ t with ϵ ≠ 1 \epsilon\ne1 ϵ = 1 the root of unity by which α \alpha α acts on the tangent space of the stack; it meets the diagonal with multiplicity one.
(b) Let ω ∈ Aut ( E − 3 ) \omega\in\operatorname{Aut}(E_{-3}) ω ∈ Aut ( E − 3 ) have order 3, and let R 0 = ( Z / 25 ) [ ω ] R_0 = (\mathbb{Z}/25)[\omega] R 0 = ( Z /25 ) [ ω ] , the Cartan subring of y − 3 y_{-3} y − 3 , generated by x 0 x_0 x 0 . If ω R ω − 1 \omega R\omega^{-1} ω R ω − 1 is S ν S_\nu S ν -related to R R R , then ω \omega ω normalizes R R R modulo 5, and an element of order 3 of the normalizer of a non-split Cartan subring of M 2 ( F 5 ) M_2(\mathbb{F}_5) M 2 ( F 5 ) lies in it; so R ≡ R 0 m o d 5 R \equiv R_0 \bmod5 R ≡ R 0 mod 5 , and R R R is generated by x = x 0 + 5 w ~ x = x_0 + 5\widetilde{w} x = x 0 + 5 w with w ~ \widetilde{w} w anticommuting with x 0 x_0 x 0 modulo 5. Then ω w ~ ω − 1 = ω ω ‾ − 1 w ~ = ω 2 w ~ \omega\widetilde{w}\omega^{-1} = \omega\overline{\omega}^{-1}\widetilde{w} = \omega^2\widetilde{w} ω w ω − 1 = ω ω − 1 w = ω 2 w modulo 5, so ω x ω − 1 = x + 5 ( ω 2 − 1 ) w ~ \omega x\omega^{-1} = x + 5(\omega^2 - 1)\widetilde{w} ω x ω − 1 = x + 5 ( ω 2 − 1 ) w , and ( ( ω 2 − 1 ) w ~ ) 2 = N ( ω 2 − 1 ) w ~ 2 = 3 w ~ 2 ((\omega^2 - 1)\widetilde{w})^2 = N(\omega^2 - 1)\widetilde{w}^2 = 3\widetilde{w}^2 (( ω 2 − 1 ) w ) 2 = N ( ω 2 − 1 ) w 2 = 3 w 2 . So the condition is w ~ 2 = ν / 3 = 2 ν \widetilde{w}^2 = \nu/3 = 2\nu w 2 = ν /3 = 2 ν , which describes S 2 ν ( R 0 ) S_{2\nu}(R_0) S 2 ν ( R 0 ) . The map w ~ ↦ ω 2 w ~ \widetilde{w} \mapsto\omega^2\widetilde{w} w ↦ ω 2 w has no non-zero fixed point, so these six subrings give the two points Q 2 ν , Q 2 ν ′ Q_{2\nu}, Q'_{2\nu} Q 2 ν , Q 2 ν ′ , each with the two branches of ω \omega ω and ω 2 \omega^2 ω 2 .
(c) Let ι ∈ Aut ( E 1728 ) \iota\in\operatorname{Aut}(E_{1728}) ι ∈ Aut ( E 1728 ) have order 4, and write a generator as x = a 0 ι + w x = a_0\iota+ w x = a 0 ι + w with w w w anticommuting with ι \iota ι . Then ι x ι − 1 = a 0 ι − w \iota x\iota^{-1} = a_0\iota- w ι x ι − 1 = a 0 ι − w . It is congruent to x x x modulo 5 only if w = 0 w = 0 w = 0 , which would give x 2 = a 0 2 = 2 x^2 = a_0^2 = 2 x 2 = a 0 2 = 2 , impossible; so the generator of ι R ι − 1 \iota R\iota^{-1} ι R ι − 1 congruent to x x x is − ι x ι − 1 -\iota x\iota^{-1} − ι x ι − 1 , which needs a 0 = 5 a 0 ′ a_0 = 5a'_0 a 0 = 5 a 0 ′ , and then − ι x ι − 1 = x + 5 ( − 2 a 0 ′ ) -\iota x\iota^{-1} = x + 5(-2a'_0) − ι x ι − 1 = x + 5 ( − 2 a 0 ′ ) with ( − 2 a 0 ′ ) 2 = − 4 a 0 ′ 2 = a 0 ′ 2 (-2a'_0)^2 = -4a_0'^2 = a_0'^2 ( − 2 a 0 ′ ) 2 = − 4 a 0 ′2 = a 0 ′2 . So ν = a 0 ′ 2 \nu= a_0'^2 ν = a 0 ′2 with a 0 ′ ≠ 0 a'_0 \ne0 a 0 ′ = 0 , which excludes ν = 2 , 3 \nu= 2,3 ν = 2 , 3 . For ν ∈ { 1 , 4 } \nu\in\{1,4\} ν ∈ { 1 , 4 } the subrings ( Z / 25 ) [ 5 a 0 ′ ι + w ] (\mathbb{Z}/25)[5a'_0\iota+ w] ( Z /25 ) [ 5 a 0 ′ ι + w ] with a 0 ′ 2 = ν a_0'^2 = \nu a 0 ′2 = ν and w 2 = 2 w^2 = 2 w 2 = 2 are twenty, and ι \iota ι pairs them into ten points, each with the single branch of ι \iota ι .
(d) Near a cusp, X X X is described by the Tate curve with parameter t t t , t 25 t^{25} t 25 the Tate parameter, and u : = ( 1 1 0 1 ) u := \begin{pmatrix}1 & 1 \\ 0 & 1\end{pmatrix} u := ( 1 0 1 1 ) acts on level structures as t ↦ ζ 25 t t \mapsto\zeta_{25}t t ↦ ζ 25 t . So the branches of S ν S_\nu S ν through (cusp, cusp) come from the m ≠ 0 m o d 25 m \ne0 \bmod25 m = 0 mod 25 with ( R , u m R u − m ) ∈ R ν (R,u^mRu^{-m}) \in R_\nu ( R , u m R u − m ) ∈ R ν . Let x = ( a b c − a ) x = \begin{pmatrix}a & b \\ c & -a\end{pmatrix} x = ( a c b − a ) generate R R R ; then 5 ∤ c 5 \nmid c 5 ∤ c , as x ‾ \overline{x} x has no eigenvector. With n : = u − 1 n := u - 1 n := u − 1 , u m x u − m = x + m ( n x − x n ) − m 2 c n u^m x u^{-m} = x + m(nx - xn) - m^2 cn u m x u − m = x + m ( n x − x n ) − m 2 c n , and n x − x n nx - xn n x − x n has diagonal ( c , − c ) (c,-c) ( c , − c ) , so u m x u − m ≡ x m o d 5 u^m x u^{-m} \equiv x \bmod5 u m x u − m ≡ x mod 5 forces m = 5 m ′ m = 5m' m = 5 m ′ ; congruence to − x -x − x is excluded by the lower left entry. Then u m x u − m = x + 5 m ′ ( n x − x n ) u^{m}xu^{-m}=x+5m'(nx-xn) u m x u − m = x + 5 m ′ ( n x − x n ) , where n x − x n nx-xn n x − x n anticommutes with x x x and has square c 2 c^{2} c 2 . So ν = ( m ′ c ) 2 \nu=(m'c)^{2} ν = ( m ′ c ) 2 : there is no solution for ν = 2 , 3 \nu=2,3 ν = 2 , 3 and two values of m ′ m' m ′ for ν = 1 , 4 \nu=1,4 ν = 1 , 4 , with branches t ↦ ζ 5 ± m ′ t t\mapsto\zeta_{5}^{\pm m'}t t ↦ ζ 5 ± m ′ t .
The same arguments apply over F ‾ q \overline{\mathbb{F}}_{q} F q , q ≥ 7 q\geq7 q ≥ 7 , where the automorphism groups at j = 0 j=0 j = 0 , 1728 1728 1728 and elsewhere are the same and the Tate curve describes the cusps. □
The term at 5
We prove Proposition 1.5 1.5 1.5 and show that its hypotheses hold for X X X over Q 5 \mathbb{Q}_{5} Q 5 and the correspondences Z 1 Z_{1} Z 1 and Z 2 Z_{2} Z 2 (Corollary 7.3 7.3 7.3 ).
Proof of Proposition 1.5. Let d d d be the degree of C ( y ) C(y) C ( y ) , and write Z Z Z also for the action of the endomorphism on H H H . By the Lefschetz fixed point formula, deg Fix C = 2 d − Tr ( Z ∣ H ) \deg\operatorname{Fix}C=2d-\operatorname{Tr}(Z\mid H) deg Fix C = 2 d − Tr ( Z ∣ H ) . This trace is an integer, since the characteristic polynomial of an endomorphism of J J J has integer coefficients and is its characteristic polynomial on the Tate module [35 Theorem 10.9, Proposition 10.20] . It also lies in Q α \mathbb{Q}_{\alpha} Q α , and Q ∩ Q α = 0 \mathbb{Q}\cap\mathbb{Q}_{\alpha}=0 Q ∩ Q α = 0 as α ∉ Q \alpha\notin\mathbb{Q} α ∈ / Q . So Tr ( Z ∣ H ) = 0 \operatorname{Tr}(Z\mid H)=0 Tr ( Z ∣ H ) = 0 , and D : = Fix C − C ( y ) − C ( z ) D:=\operatorname{Fix}C-C(y)-C(z) D := Fix C − C ( y ) − C ( z ) has degree 0. As C C C is symmetric, transposition permutes the C i C_i C i , so by the hypotheses on y y y and z z z neither lies on Fix C i \operatorname{Fix}C_i Fix C i , C i ( y ) C_i(y) C i ( y ) or C i ( z ) C_i(z) C i ( z ) for any i i i , and neither lies in the support of D D D .
The reduction graph. Let K 0 K_{0} K 0 be the completion of the maximal unramified extension of Q p \mathbb{Q}_{p} Q p , with inertia group I = Gal ( K ‾ 0 / K 0 ) I=\operatorname{Gal}(\overline{K}_{0}/K_{0}) I = Gal ( K 0 / K 0 ) . By [15 Corollary (2.7)] there is a finite Galois extension L L L of K 0 K_{0} K 0 over which X X X has a stable model X L \mathcal{X}_{L} X L , and we take L L L so large that J J J has good reduction over L L L . By a theorem of Raynaud [15 Theorem (2.5)] , the identity component of the special fiber of the Néron model of J L J_{L} J L is Pic 0 \operatorname{Pic}^{0} Pic 0 of the special fiber of the minimal regular model of X L X_{L} X L , which is that of X L \mathcal{X}_{L} X L with each node replaced by a chain of rational curves [15 proof of Lemma (1.12)] . As for any nodal curve, this Pic 0 \operatorname{Pic}^{0} Pic 0 is an extension of ∏ v Jac ( C v ) \prod_{v}\operatorname{Jac}(C_{v}) ∏ v Jac ( C v ) by a torus of dimension b 1 ( Γ ) b_{1}(\Gamma) b 1 ( Γ ) . Here Γ \Gamma Γ is the dual graph of the special fiber, v v v runs over its vertices and C v C_{v} C v is the normalization of the component v v v . Good reduction leaves no torus, so Γ \Gamma Γ is a tree and J J J reduces to J ‾ = ∏ v Jac ( C v ) \overline{J}=\prod_{v}\operatorname{Jac}(C_{v}) J = ∏ v Jac ( C v ) . This decomposes H = ⨁ v H ( v ) H=\bigoplus_{v}H(v) H = ⨁ v H ( v ) with H ( v ) : = H e ˊ t 1 ( C v , Q ‾ ℓ ) H(v):=H^{1}_{\mathrm{\acute et}}(C_{v},\overline{\mathbb{Q}}_{\ell}) H ( v ) := H e ˊ t 1 ( C v , Q ℓ ) , and we let e v e_{v} e v be the projection onto H ( v ) H(v) H ( v ) . The group I I I acts on H H H through its finite quotient Gal ( L / K 0 ) \operatorname{Gal}(L/K_{0}) Gal ( L / K 0 ) . For σ \sigma σ in this quotient, the conjugate of X L \mathcal{X}_{L} X L by σ \sigma σ is a stable model of the same curve, so by [15 Lemma (1.12)] the action of σ \sigma σ on X L X_{L} X L extends to X L \mathcal{X}_{L} X L . Hence I I I acts on Γ \Gamma Γ , and σ H ( v ) = H ( σ v ) \sigma H(v)=H(\sigma v) σ H ( v ) = H ( σ v ) .
The formula of Betts and Dogra. The endomorphism Z Z Z is fixed by the Rosati involution, as C C C is symmetric, and has trace 0. By [7 Remark 12.1.4] , the function z ↦ i p ( z − y , Fix C − C ( y ) − C ( z ) ) z\mapsto i_{p}(z-y,\operatorname{Fix}C-C(y)-C(z)) z ↦ i p ( z − y , Fix C − C ( y ) − C ( z )) is a non-zero constant multiple of the coordinate j C jC j C of the unipotent Kummer map with base point y y y . By [7 Theorem 1.1.2, Corollary 12.1.3] (in the form of [3 Theorem 3.2] ), in the case of a tree [7 §12.2.1] ,
j C ( z ) = ⟨ red ( z ) − red ( y ) , μ ⟩ , μ = 1 2 ∑ v Tr ( e v Z e v ∣ H ( v ) ) v , jC(z)=\langle\operatorname{red}(z)-\operatorname{red}(y),\mu\rangle,\qquad\mu=\frac{1}{2}\sum_{v}\operatorname{Tr}(e_{v}Ze_{v}\mid H(v))v, j C ( z ) = ⟨ red ( z ) − red ( y ) , μ ⟩ , μ = 2 1 v ∑ Tr ( e v Z e v ∣ H ( v )) v ,
where red : X ( Q p ) → Γ \operatorname{red}:X(\mathbb{Q}_{p})\to\Gamma red : X ( Q p ) → Γ is the reduction map and ⟨ , ⟩ \langle\, ,\,\rangle ⟨ , ⟩ is the height pairing on the metrized graph Γ \Gamma Γ .
Each mass vanishes. Write μ = ∑ v μ v v \mu=\sum_{v}\mu_{v}v μ = ∑ v μ v v . Let O O O be an orbit of I I I on the vertices of Γ \Gamma Γ , and H O : = ⨁ v ∈ O H ( v ) H_{O}:=\bigoplus_{v\in O}H(v) H O := ⨁ v ∈ O H ( v ) . It is stable under I I I . As I I I acts on H H H semisimply, H O H_{O} H O is the sum of its intersections with the isotypic components of H H H under I I I , and Z Z Z acts on each of these by a scalar in Q α \mathbb{Q}_{\alpha} Q α . Hence Z Z Z preserves H O H_{O} H O , and
2 ∑ v ∈ O μ v = ∑ v ∈ O Tr ( e v Z e v ∣ H ( v ) ) = Tr ( Z ∣ H O ) ∈ Q α . 2\sum_{v\in O}\mu_{v}=\sum_{v\in O}\operatorname{Tr}(e_{v}Ze_{v}\mid H(v))=\operatorname{Tr}(Z\mid H_{O})\in\mathbb{Q}_{\alpha}. 2 v ∈ O ∑ μ v = v ∈ O ∑ Tr ( e v Z e v ∣ H ( v )) = Tr ( Z ∣ H O ) ∈ Q α .
On the other hand, Tr ( Z ∣ H O 1 ) \operatorname{Tr}(Z \mid H^{1}_{O}) Tr ( Z ∣ H O 1 ) is the trace on H 1 H^{1} H 1 of an endomorphism of the abelian variety ∏ v ∈ O Jac ( C v ) \prod_{v\in O}\operatorname{Jac}(C_{v}) ∏ v ∈ O Jac ( C v ) , namely the reduction of Z Z Z on J ‾ \overline{J} J restricted to this factor and projected back onto it, so it is an integer [35 Theorem 10.9, Proposition 10.20] , and therefore 0. As Z Z Z commutes with I I I (C C C is defined over Q p \mathbb{Q}_{p} Q p ) and σ e v σ − 1 = e σ v \sigma e_{v}\sigma^{-1}=e_{\sigma v} σ e v σ − 1 = e σ v , the masses μ v \mu_{v} μ v are constant on O O O , so each is 0. Hence μ = 0 \mu=0 μ = 0 , so j C ( z ) = 0 j_{C}(z)=0 j C ( z ) = 0 and i p ( z − y , D ) = 0 i_{p}(z-y,D)=0 i p ( z − y , D ) = 0 . □
Lemma 7.1. The Jacobian J J J has potentially good reduction at 5.
Proof. By Chen [12 Theorem 1.4] , J J J is isogenous over Q \mathbb{Q} Q to the new quotient of the Jacobian of X 0 + ( 625 ) X_{0}^{+}(625) X 0 + ( 625 ) , as X 0 + ( 25 ) X_{0}^{+}(25) X 0 + ( 25 ) has genus 0. So its Tate module is a sum of the ℓ \ell ℓ -adic representations of the newforms of weight 2 and level 625 with trivial character and Fricke eigenvalue + 1 +1 + 1 . The local component at 5 of such a newform is not a twist St ⊗ χ \operatorname{St}\otimes\chi St ⊗ χ of the Steinberg representation: its central character χ 2 \chi^{2} χ 2 would be trivial, and then its conductor would be 5 or 5 2 5^{2} 5 2 . So these representations have no monodromy at 5 [10 Théorème (A)] , inertia at 5 acts on them through a finite quotient, and J J J has potentially good reduction [41 §2, Theorem 2(i)] . □
Lemma 7.2. Let π , π ′ ∈ P \pi,\pi'\in P π , π ′ ∈ P (Proposition 6.4) and let I I I be the inertia group at 5. If σ ( π ) ∣ I \sigma(\pi)|_{I} σ ( π ) ∣ I and σ ( π ′ ) ∣ I \sigma(\pi')|_{I} σ ( π ′ ) ∣ I have a common irreducible constituent, then π 5 \pi_{5} π 5 and π 5 ′ \pi'_{5} π 5 ′ have isomorphic restrictions to GL 2 ( Z 5 ) \operatorname{GL}_{2}(\mathbb{Z}_{5}) GL 2 ( Z 5 ) .
Proof. Let W 5 W_{5} W 5 be the Weil group of Q 5 \mathbb{Q}_{5} Q 5 , and σ ( π 5 ) \sigma(\pi_{5}) σ ( π 5 ) the representation of the Weil–Deligne group of Q 5 \mathbb{Q}_{5} Q 5 attached to π 5 \pi_{5} π 5 by the correspondence normalized as in [10 (0.5)] . By Carayol’s theorem [10 Théorème (A)] it is the Weil–Deligne representation of σ ( π ) \sigma(\pi) σ ( π ) restricted to a decomposition group at 5. As σ ( π ) ⊂ H \sigma(\pi)\subset H σ ( π ) ⊂ H , inertia acts on it through a finite quotient (Lemma 7.1 ), so σ ( π 5 ) \sigma(\pi_{5}) σ ( π 5 ) has monodromy 0 and σ ( π 5 ) ∣ I = σ ( π ) ∣ I \sigma(\pi_{5})|_{I}=\sigma(\pi)|_{I} σ ( π 5 ) ∣ I = σ ( π ) ∣ I is semisimple. The same holds for π ′ \pi' π ′ , so Hom I ( σ ( π 5 ) , σ ( π 5 ′ ) ) ≠ 0 \operatorname{Hom}_{I}(\sigma(\pi_{5}),\sigma(\pi'_{5}))\ne0 Hom I ( σ ( π 5 ) , σ ( π 5 ′ )) = 0 . A Frobenius element Φ ∈ W 5 \Phi\in W_{5} Φ ∈ W 5 acts on this space by ϕ ↦ σ ( π 5 ′ ) ( Φ ) ∘ ϕ ∘ σ ( π 5 ) ( Φ ) − 1 \phi\mapsto\sigma(\pi'_{5})(\Phi)\circ\phi\circ\sigma(\pi_{5})(\Phi)^{-1} ϕ ↦ σ ( π 5 ′ ) ( Φ ) ∘ ϕ ∘ σ ( π 5 ) ( Φ ) − 1 , and an eigenvector ϕ \phi ϕ , with eigenvalue c c c , is a non-zero W 5 W_{5} W 5 -map σ ( π 5 ) ⊗ μ → σ ( π 5 ′ ) \sigma(\pi_{5})\otimes\mu\to\sigma(\pi'_{5}) σ ( π 5 ) ⊗ μ → σ ( π 5 ′ ) , where μ \mu μ is the unramified character with μ ( Φ ) = c \mu(\Phi)=c μ ( Φ ) = c .
If σ ( π 5 ) \sigma(\pi_{5}) σ ( π 5 ) or σ ( π 5 ′ ) \sigma(\pi'_{5}) σ ( π 5 ′ ) is irreducible, ϕ \phi ϕ is an isomorphism, and σ ( π 5 ′ ) ≅ σ ( π 5 ) ⊗ μ = σ ( π 5 ⊗ μ − 1 ) \sigma(\pi'_{5})\cong\sigma(\pi_{5})\otimes\mu=\sigma(\pi_{5}\otimes\mu^{-1}) σ ( π 5 ′ ) ≅ σ ( π 5 ) ⊗ μ = σ ( π 5 ⊗ μ − 1 ) by [10 (0.5) (b)] , where π 5 ⊗ μ − 1 \pi_{5}\otimes\mu^{-1} π 5 ⊗ μ − 1 means π 5 ⊗ ( μ − 1 ∘ det ) \pi_{5}\otimes(\mu^{-1}\circ\det) π 5 ⊗ ( μ − 1 ∘ det ) . The correspondence [29 ] is injective by the local converse theorem of Jacquet and Langlands [21 Corollary 2.19] , so π 5 ′ ≅ π 5 ⊗ ( μ − 1 ∘ det ) \pi'_{5}\cong\pi_{5}\otimes(\mu^{-1}\circ\det) π 5 ′ ≅ π 5 ⊗ ( μ − 1 ∘ det ) , and μ ∘ det \mu\circ\det μ ∘ det is trivial on GL 2 ( Z 5 ) \operatorname{GL}_{2}(\mathbb{Z}_{5}) GL 2 ( Z 5 ) .
Otherwise both are reducible. As the monodromy is 0, neither π 5 \pi_{5} π 5 nor π 5 ′ \pi'_{5} π 5 ′ is a twist of the Steinberg representation, and as σ ( π 5 ) \sigma(\pi_{5}) σ ( π 5 ) is reducible, π 5 \pi_{5} π 5 is not supercuspidal. So π 5 = Ind ( ξ 1 , ξ 2 ) \pi_{5}=\operatorname{Ind}(\xi_{1},\xi_{2}) π 5 = Ind ( ξ 1 , ξ 2 ) and π 5 ′ = Ind ( ξ 1 ′ , ξ 2 ′ ) \pi'_{5}=\operatorname{Ind}(\xi'_{1},\xi'_{2}) π 5 ′ = Ind ( ξ 1 ′ , ξ 2 ′ ) are irreducible principal series, and σ ( π 5 ) = ξ 1 − 1 ⊕ ξ 2 − 1 λ \sigma(\pi_{5})=\xi_{1}^{-1}\oplus\xi_{2}^{-1}\lambda σ ( π 5 ) = ξ 1 − 1 ⊕ ξ 2 − 1 λ with λ \lambda λ unramified [10 (0.5)] , and likewise for π 5 ′ \pi'_{5} π 5 ′ . Through local class field theory, a common constituent after renumbering, ξ 1 ′ = ξ 1 \xi'_{1}=\xi_{1} ξ 1 ′ = ξ 1 on Z 5 × \mathbb{Z}_{5}^{\times} Z 5 × . The central characters ξ 1 ξ 2 \xi_{1}\xi_{2} ξ 1 ξ 2 and ξ 1 ′ ξ 2 ′ \xi'_{1}\xi'_{2} ξ 1 ′ ξ 2 ′ are trivial on Z 5 × \mathbb{Z}_{5}^{\times} Z 5 × (Proposition 6.4), so also ξ 2 ′ = ξ 2 \xi'_{2}=\xi_{2} ξ 2 ′ = ξ 2 on Z 5 × \mathbb{Z}_{5}^{\times} Z 5 × . By the Iwasawa decomposition GL 2 ( Q 5 ) = B ( Q 5 ) GL 2 ( Z 5 ) \operatorname{GL}_{2}(\mathbb{Q}_{5})=B(\mathbb{Q}_{5})\operatorname{GL}_{2}(\mathbb{Z}_{5}) GL 2 ( Q 5 ) = B ( Q 5 ) GL 2 ( Z 5 ) , with B B B the upper triangular subgroup, the restriction of Ind ( ξ 1 , ξ 2 ) \operatorname{Ind}(\xi_{1},\xi_{2}) Ind ( ξ 1 , ξ 2 ) to GL 2 ( Z 5 ) \operatorname{GL}_{2}(\mathbb{Z}_{5}) GL 2 ( Z 5 ) is induced from the character ( a b 0 d ) ↦ ξ 1 ( a ) ξ 2 ( d ) \begin{pmatrix}a&b\\0&d\end{pmatrix}\mapsto\xi_{1}(a)\xi_{2}(d) ( a 0 b d ) ↦ ξ 1 ( a ) ξ 2 ( d ) of B ( Z 5 ) B(\mathbb{Z}_{5}) B ( Z 5 ) , so it depends only on ξ 1 \xi_{1} ξ 1 and ξ 2 \xi_{2} ξ 2 on Z 5 × \mathbb{Z}_{5}^{\times} Z 5 × . □
Corollary 7.3. Let Z = S v − S 4 v ∈ { Z 1 , Z 2 } Z=S_{v}-S_{4v}\in\{Z_{1},Z_{2}\} Z = S v − S 4 v ∈ { Z 1 , Z 2 } , and let y , z ∈ X ( Q 5 ) y,z\in X(\mathbb{Q}_{5}) y , z ∈ X ( Q 5 ) be such that none of ( y , y ) (y,y) ( y , y ) , ( z , z ) (z,z) ( z , z ) , ( y , z ) (y,z) ( y , z ) lies on S v S_{v} S v or S 4 v S_{4v} S 4 v . Then i 5 ( z − y , Fix Z − Z ( y ) − Z ( z ) ) = 0 i_{5}(z-y,\operatorname{Fix}Z-Z(y)-Z(z))=0 i 5 ( z − y , Fix Z − Z ( y ) − Z ( z )) = 0 .
Proof. The correspondence Z Z Z is symmetric, and its components lie on S v S_{v} S v or S 4 v S_{4v} S 4 v , none of them the diagonal, as R v R_{v} R v is symmetric and contains no pair ( R , R ) (R,R) ( R , R ) . It is non-zero, being injective on H H H by Proposition 6.4 (a), and J J J has potentially good reduction at 5 by Lemma 7.1 . By Proposition 6.4 (b), H = ⨁ π ∈ P σ ( π ) H=\bigoplus_{\pi\in P}\sigma(\pi) H = ⨁ π ∈ P σ ( π ) , with Z Z Z acting on σ ( π ) \sigma(\pi) σ ( π ) by z ( π ) ∈ 5 Q z(\pi)\in\sqrt{5}\mathbb{Q} z ( π ) ∈ 5 Q , and each σ ( π ) ∣ I \sigma(\pi)|_{I} σ ( π ) ∣ I is semisimple. The isotypic component of an irreducible representation τ \tau τ of I I I is the sum of the τ \tau τ -isotypic parts of the σ ( π ) \sigma(\pi) σ ( π ) that contain τ \tau τ , and by Lemma 7.2 and Proposition 6.4 (b) all these π \pi π have the same z ( π ) z(\pi) z ( π ) . So Proposition 1.5 applies with α = 5 \alpha= \sqrt{5} α = 5 .
Heights at the auxiliary prime
The argument of this section and the next works on any modular curve, and we state it in that generality. Let H ⊂ GL 2 ( Z / N X ) \mathcal{H} \subset\operatorname{GL}_{2}(\mathbb{Z}/N_X) H ⊂ GL 2 ( Z / N X ) contain the scalars, with det H = ( Z / N X ) × \det\mathcal{H} = (\mathbb{Z}/N_X)^{\times} det H = ( Z / N X ) × , let X = X H X = X_{\mathcal{H}} X = X H have genus g ≥ 2 g \ge2 g ≥ 2 and Jacobian J J J , and let T ⊂ End J \mathbb{T} \subset\operatorname{End} J T ⊂ End J be generated by the Hecke operators T n T_n T n , n n n prime to N X N_X N X . Let q ≥ 5 q \ge5 q ≥ 5 be a prime not dividing N X N_X N X , let X \mathcal{X} X be the smooth model of X X X over Z q \mathbb{Z}_q Z q [16 VI, Proposition 6.7] , and put
L : = H 0 ( X , Ω 1 ) \mathcal{L} := H^0(\mathcal{X}, \Omega^1) L := H 0 ( X , Ω 1 ) , Lie : = Hom ( L , Z q ) \operatorname{Lie} := \operatorname{Hom}(\mathcal{L}, \mathbb{Z}_q) Lie := Hom ( L , Z q ) , T q : = T ⊗ Z q \mathbb{T}_q := \mathbb{T} \otimes\mathbb{Z}_q T q := T ⊗ Z q , M : = J ( Q ) ⊗ Z q M := J(\mathbb{Q}) \otimes\mathbb{Z}_q M := J ( Q ) ⊗ Z q , N q : = # J ‾ ( F q ) N_q := \#\overline{J}(\mathbb{F}_q) N q := # J ( F q ) .
We assume
(B) q ∤ N q q \nmid N_q q ∤ N q ;
(M) Lie \operatorname{Lie} Lie is free of rank one over T q \mathbb{T}_q T q .
These are the hypotheses (B) and (M) of [23 Theorem 1.2] , with its prime p p p replaced by q q q , and we use [23 §§2–3] in that form. None of the results we use there needs its other hypothesis (Φ \Phi Φ ), which gives the integrality of the height pairing [23 Proposition 2.9] ; here integrality comes from the control equations instead (Proposition 9.1 (a)). By (M), L \mathcal{L} L has a T q \mathbb{T}_q T q -stable complement W W W in H d R 1 ( X / Z q ) H^1_{\mathrm{dR}}(\mathcal{X}/\mathbb{Z}_q) H dR 1 ( X / Z q ) , and W W W is isotropic [23 Lemma 2.1] . For a divisor b b b of degree 0 over a finite extension L L L of Q q \mathbb{Q}_q Q q , let ω b \omega_b ω b be the form of the third kind over L L L with residue divisor b b b and class in W ⊗ L W \otimes L W ⊗ L [23 §2] , and let h = ∑ v h v h = \sum_v h_v h = ∑ v h v be the height of Coleman and Gross [13 ] for W W W , with
h q ( a , b ) : = ∫ a ω b , h v ( a , b ) : = − i v ( a , b ) log q v ( v ≠ q ) h_q(a,b) := \int_a \omega_b,\qquad h_v(a,b) := -i_v(a,b)\log_q v\quad(v \ne q) h q ( a , b ) := ∫ a ω b , h v ( a , b ) := − i v ( a , b ) log q v ( v = q )
for divisors a , b a,b a , b of degree 0 with disjoint supports, and log q q = 0 \log_q q = 0 log q q = 0 . We never need W W W explicitly: modulo q 2 q^2 q 2 , the values of h q h_q h q that enter are given by Lemma 8.1 through values of functions. The local term h q h_q h q is symmetric, as W W W is isotropic, and T \mathbb{T} T acts on M M M by self-adjoint operators for h h h [23 Lemma 2.2] , as in [4 Lemma 2.13] . By (B), M M M is torsion-free and log M ⊂ q Lie \log M \subset q\operatorname{Lie} log M ⊂ q Lie , so
κ : = q − 1 log : M ⟶ Lie \kappa:= q^{-1}\log\colon M \longrightarrow\operatorname{Lie} κ := q − 1 log : M ⟶ Lie
is defined. A symmetric integer combination Z Z Z of double cosets H γ H \mathcal{H}\gamma\mathcal{H} H γ H commutes with T \mathbb{T} T , so by (M) it acts on M M M and on Lie \operatorname{Lie} Lie through an element of T q \mathbb{T}_q T q , and it is self-adjoint for h h h [23 §3] .
A residue disc is a fiber of the reduction map X ( C q ) → X ( F ‾ q ) X(\mathbb{C}_q) \to\mathcal{X}(\overline{\mathbb{F}}_q) X ( C q ) → X ( F q ) . For a residue point x ˉ \bar{x} x ˉ rational over F q f \mathbb{F}_{q^f} F q f , let O O O be the ring of Witt vectors of F q f \mathbb{F}_{q^f} F q f ; an integral parameter at x ˉ \bar{x} x ˉ is a function τ \tau τ with O ^ X O , x ˉ = O [ [ τ ] ] \widehat{\mathcal{O}}_{\mathcal{X}_O,\bar{x}} = O[[\tau]] O X O , x ˉ = O [[ τ ]] . For a Q q \mathbb{Q}_q Q q -rational divisor D D D of degree 0, (B) gives a function f ˉ D \bar{f}_D f ˉ D on X F q \mathcal{X}_{\mathbb{F}_q} X F q with divisor N q D ˉ N_q\bar{D} N q D ˉ , and it lifts to a function f D f_D f D over Q q \mathbb{Q}_q Q q whose zeros and poles lie in the residue discs of the points of D D D and for which D 1 : = N q D − div f D D_1 := N_qD - \operatorname{div} f_D D 1 := N q D − div f D has degree 0 on every residue disc [23 proof of Lemma 2.8] . We fix such an f D f_D f D . For a divisor A = ∑ n P P A = \sum n_P P A = ∑ n P P at whose points a function f f f is a unit, log f ( A ) : = ∑ n P log f ( P ) \log f(A) := \sum n_P \log f(P) log f ( A ) := ∑ n P log f ( P ) .
Lemma 8.1. Assume (B), and let D D D be a Q q \mathbb{Q}_q Q q -rational divisor of degree 0.
(a) On a residue disc that contains no point of D D D , with integral parameter τ \tau τ , ω D ∈ O [ [ τ ] ] d τ \omega_D \in O[[\tau]]\,d\tau ω D ∈ O [[ τ ]] d τ and ω D ≡ N q − 1 d f D / f D \omega_D \equiv N_q^{-1}\,df_D/f_D ω D ≡ N q − 1 d f D / f D modulo q q q .
(b) If D D D is rational over Q \mathbb{Q} Q , then κ ( [ D ] , ω ) = ( q N q ) − 1 ∫ D 1 ω \kappa([D],\omega) = (qN_q)^{-1}\int_{D_1}\omega κ ([ D ] , ω ) = ( q N q ) − 1 ∫ D 1 ω for ω ∈ L \omega\in\mathcal{L} ω ∈ L , and ∫ D 1 ω \int_{D_1}\omega ∫ D 1 ω is a sum of integrals over residue discs, each in q O qO q O .
(c) Let y , y ′ ∈ X ( Q q ) y,y' \in X(\mathbb{Q}_q) y , y ′ ∈ X ( Q q ) lie in residue discs that contain no point of D D D , and put A : = y ′ − y A:=y'-y A := y ′ − y . Then N q h q ( A , D ) ∈ q Z q N_qh_q(A,D)\in q\mathbb{Z}_q N q h q ( A , D ) ∈ q Z q and
N q h q ( A , D ) ≡ log f D ( A ) + N q − 1 log f A ( D 1 ) ( m o d q 2 ) . N_qh_q(A,D)\equiv\log f_D(A)+N_q^{-1}\log f_A(D_1)\pmod{q^2}. N q h q ( A , D ) ≡ log f D ( A ) + N q − 1 log f A ( D 1 ) ( mod q 2 ) .
Proof. (a) By the proof of [23 Lemma 2.8] , N q ω D = ω D 1 + d f D / f D N_q\omega_D=\omega_{D_1}+df_D/f_D N q ω D = ω D 1 + d f D / f D . The divisor D 1 D_1 D 1 lies in the residue discs of the points of D D D , so ω D 1 \omega_{D_1} ω D 1 is q q q times an integral form on the disc [23 Proposition 2.6 (a)] , and f D f_D f D is a unit there.
(b) As [ div f D ] = 0 [\operatorname{div}f_D]=0 [ div f D ] = 0 , N q [ D ] = [ D 1 ] N_q[D]=[D_1] N q [ D ] = [ D 1 ] , and ⟨ log [ D 1 ] , ω ⟩ = ∫ D 1 ω \langle\log[D_1],\omega\rangle=\int_{D_1}\omega ⟨ log [ D 1 ] , ω ⟩ = ∫ D 1 ω . As D 1 D_1 D 1 has degree 0 on every disc, this is the sum over the discs of the integrals of ω \omega ω over the part of D 1 D_1 D 1 in the disc, and each lies in q O q\mathcal{O} q O by [23 Lemma 2.3] , as ω \omega ω is integral.
(c) By (a), N q h q ( A , D ) = ∫ A ω D 1 + log f D ( A ) N_qh_q(A,D)=\int_A\omega_{D_1}+\log f_D(A) N q h q ( A , D ) = ∫ A ω D 1 + log f D ( A ) , and log f D ( A ) \log f_D(A) log f D ( A ) is the logarithm of a unit of Z q \mathbb{Z}_q Z q , so it lies in q Z q q\mathbb{Z}_q q Z q . As h q h_q h q is symmetric, ∫ A ω D 1 = h q ( A , D 1 ) = ∫ D 1 ω A \int_A\omega_{D_1}=h_q(A,D_1)=\int_{D_1}\omega_A ∫ A ω D 1 = h q ( A , D 1 ) = ∫ D 1 ω A . On the discs of the points of D D D , which contain no point of A A A , (a) gives ω A ≡ N q − 1 d f A / f A \omega_A\equiv N_q^{-1}df_A/f_A ω A ≡ N q − 1 d f A / f A modulo q q q , with both forms integral. As D 1 D_1 D 1 has degree 0 on each of these discs, [23 Lemma 2.3] gives ∫ D 1 ω A ∈ q Z q \int_{D_1}\omega_A\in q\mathbb{Z}_q ∫ D 1 ω A ∈ q Z q and ∫ D 1 ω A ≡ N q − 1 ∫ D 1 d f A / f A = N q − 1 log f A ( D 1 ) \int_{D_1}\omega_A\equiv N_q^{-1}\int_{D_1}df_A/f_A=N_q^{-1}\log f_A(D_1) ∫ D 1 ω A ≡ N q − 1 ∫ D 1 d f A / f A = N q − 1 log f A ( D 1 ) modulo q 2 q^2 q 2 .
Control equations and the disc of a known point
We keep the setting of Section 8 and assume (B) and (M). Let y 1 , … , y n ∈ X ( Q ) y_1,\ldots,y_n\in X(\mathbb{Q}) y 1 , … , y n ∈ X ( Q ) , and let Z \mathcal{Z} Z be a finite set of correspondences Z Z Z on X X X , each a symmetric integer combination of double cosets H γ H H\gamma H H γ H with γ ∉ H \gamma\notin H γ ∈ / H , with Tr ( Z ∣ L ) = 0 \operatorname{Tr}(Z\mid\mathcal{L})=0 Tr ( Z ∣ L ) = 0 , and such that:
(F) for all i , j i,j i , j , the closure of each term of Z Z Z in X × X \mathcal{X}\times\mathcal{X} X × X does not contain ( y ˉ i , y ˉ j ) (\bar y_i,\bar y_j) ( y ˉ i , y ˉ j ) ;
(V) at each prime v ∣ N X v\mid N_X v ∣ N X , i v ( z − y , Fix Z − Z ( y ) − Z ( z ) ) = 0 i_v(z-y,\operatorname{Fix}Z-Z(y)-Z(z))=0 i v ( z − y , Fix Z − Z ( y ) − Z ( z )) = 0 for all y , z ∈ X ( Q ) y,z\in X(\mathbb{Q}) y , z ∈ X ( Q ) whose reductions modulo q q q lie in { y ˉ 1 , … , y ˉ n } \{\bar y_1,\ldots,\bar y_n\} { y ˉ 1 , … , y ˉ n } .
For Z ∈ Z Z\in\mathcal{Z} Z ∈ Z and i < j i<j i < j , B i j Z : = Fix Z − Z ( y i ) − Z ( y j ) B^Z_{ij}:=\operatorname{Fix}Z-Z(y_i)-Z(y_j) B ij Z := Fix Z − Z ( y i ) − Z ( y j ) is a rational divisor of degree 0. Fix a generator e e e of Lie over T q \mathbb{T}_q T q , and define a i j , b i j Z ∈ T q a_{ij},b^Z_{ij}\in\mathbb{T}_q a ij , b ij Z ∈ T q by
κ [ y j − y i ] = a i j e , κ [ B i j Z ] = b i j Z e . \kappa[y_j-y_i]=a_{ij}e,\qquad\kappa[B^Z_{ij}]=b^Z_{ij}e. κ [ y j − y i ] = a ij e , κ [ B ij Z ] = b ij Z e .
For each Z Z Z and i < j i<j i < j , the height identity for y j − y i y_j-y_i y j − y i and B i j Z B^Z_{ij} B ij Z is a linear equation on the height pairing, with right side q − 1 h q ( y j − y i , B i j Z ) q^{-1}h_q(y_j-y_i,B^Z_{ij}) q − 1 h q ( y j − y i , B ij Z ) (proof of (a) below); we call these the control equations. As B i j Z B^Z_{ij} B ij Z is symmetric in i i i and j j j , the pairs with i > j i>j i > j give nothing new.
Proposition 9.1. In this setting, assume also:
(R) rank J ( Q ) ≤ g J(\mathbb{Q})\le g J ( Q ) ≤ g ;
(C) the products a i j b i j Z a_{ij}b^Z_{ij} a ij b ij Z , for Z ∈ Z Z\in\mathcal{Z} Z ∈ Z and i < j i<j i < j , span T q / q \mathbb{T}_q/q T q / q over F q \mathbb{F}_q F q .
Then:
(a) The map κ : M → Lie \kappa:M\to\operatorname{Lie} κ : M → Lie is an isomorphism, and h ( M , M ) ⊂ q Z q h(M,M)\subset q\mathbb{Z}_q h ( M , M ) ⊂ q Z q . There is a unique linear map λ ˉ : T q / q → F q \bar\lambda:\mathbb{T}_q/q\to\mathbb{F}_q λ ˉ : T q / q → F q with λ ˉ ( a i j b i j Z ) = q − 1 h q ( y j − y i , B i j Z ) ( m o d q ) \bar\lambda(a_{ij}b^Z_{ij})=q^{-1}h_q(y_j-y_i,B^Z_{ij})\pmod q λ ˉ ( a ij b ij Z ) = q − 1 h q ( y j − y i , B ij Z ) ( mod q ) for all Z Z Z and i < j i<j i < j , and the form b ~ \tilde b b ~ on Lie / q \operatorname{Lie}/q Lie / q with b ~ ( κ a , κ a ′ ) ≡ q − 1 h ( a , a ′ ) ( m o d q ) \tilde b(\kappa a,\kappa a')\equiv q^{-1}h(a,a')\pmod q b ~ ( κa , κ a ′ ) ≡ q − 1 h ( a , a ′ ) ( mod q ) is b ~ ( t e , t ′ e ) = λ ˉ ( t t ′ ) \tilde b(te,t'e)=\bar\lambda(tt') b ~ ( t e , t ′ e ) = λ ˉ ( t t ′ ) .
(b) Let y : = y i y:=y_i y := y i , let δ ≠ 0 \delta\ne0 δ = 0 be a tangent vector to X F q \mathcal{X}_{\mathbb{F}_q} X F q at y ˉ \bar y y ˉ , and let ev δ ∈ Lie / q = Hom ( L / q , F q ) \operatorname{ev}_\delta\in\operatorname{Lie}/q=\operatorname{Hom}(\mathcal{L}/q,\mathbb{F}_q) ev δ ∈ Lie / q = Hom ( L / q , F q ) be ω ↦ ω ( δ ) \omega\mapsto\omega(\delta) ω ↦ ω ( δ ) . For Z ∈ Z Z\in\mathcal{Z} Z ∈ Z let D Z : = Fix Z − 2 Z ( y ) D_Z:=\operatorname{Fix}Z-2Z(y) D Z := Fix Z − 2 Z ( y ) and ω ˉ D Z : = N q − 1 d f D Z / f D Z \bar\omega_{D_Z}:=N_q^{-1}df_{D_Z}/f_{D_Z} ω ˉ D Z := N q − 1 d f D Z / f D Z , and put
L Z : = ω ˉ D Z ( δ ) − b ~ ( ev δ , κ [ D Z ] ) , Q Z : = b ~ ( ev δ , Z ev δ ) . L_Z:=\bar\omega_{D_Z}(\delta)-\tilde b(\operatorname{ev}_\delta,\kappa[D_Z]),\qquad Q_Z:=\tilde b(\operatorname{ev}_\delta,Z\operatorname{ev}_\delta). L Z := ω ˉ D Z ( δ ) − b ~ ( ev δ , κ [ D Z ]) , Q Z := b ~ ( ev δ , Z ev δ ) .
If there is no c ∈ F q c\in\mathbb{F}_q c ∈ F q with L Z = c Q Z L_Z=cQ_Z L Z = c Q Z for every Z ∈ Z Z\in\mathcal{Z} Z ∈ Z , for instance if L Z Q Z ′ ≠ L Z ′ Q Z L_ZQ_{Z'}\ne L_{Z'}Q_Z L Z Q Z ′ = L Z ′ Q Z for some Z , Z ′ ∈ Z Z,Z'\in\mathcal{Z} Z , Z ′ ∈ Z , then y y y is the only rational point in its residue disc. Proof. (a) By (B), M M M is torsion-free, and κ \kappa κ is T q \mathbb{T}_q T q -linear. The ideal of T q / q \mathbb{T}_q/q T q / q generated by the a i j a_{ij} a ij contains every a i j b i j Z a_{ij}b_{ij}^{Z} a ij b ij Z , so by (C) it is the unit ideal. Hence κ ( M ) + q Lie = Lie \kappa(M)+q\operatorname{Lie}=\operatorname{Lie} κ ( M ) + q Lie = Lie , and κ ( M ) = Lie \kappa(M)=\operatorname{Lie} κ ( M ) = Lie by Nakayama’s lemma. As M M M is free of rank at most g g g by (R) and Lie \operatorname{Lie} Lie is free of rank g g g , κ \kappa κ is an isomorphism. So β ( x , x ′ ) : = q − 1 h ( κ − 1 x , κ − 1 x ′ ) \beta(x,x'):=q^{-1}h(\kappa^{-1}x,\kappa^{-1}x') β ( x , x ′ ) := q − 1 h ( κ − 1 x , κ − 1 x ′ ) is a symmetric Q q \mathbb{Q}_q Q q -valued form on Lie \operatorname{Lie} Lie with β ( t x , x ′ ) = β ( x , t x ′ ) \beta(tx,x')=\beta(x,tx') β ( t x , x ′ ) = β ( x , t x ′ ) for t ∈ T q t\in\mathbb{T}_q t ∈ T q , and β ( t e , t ′ e ) = λ ( t t ′ ) \beta(te,t'e)=\lambda(tt') β ( t e , t ′ e ) = λ ( t t ′ ) for λ ( t ) : = β ( e , t e ) \lambda(t):=\beta(e,te) λ ( t ) := β ( e , t e ) .
Fix Z Z Z and i < j i<j i < j , and put A : = y j − y i A:=y_j-y_i A := y j − y i and B : = B i j Z B:=B_{ij}^{Z} B := B ij Z . By (F), no term of Z Z Z passes through ( y i , y i ) (y_i,y_i) ( y i , y i ) , ( y i , y j ) (y_i,y_j) ( y i , y j ) , ( y j , y i ) (y_j,y_i) ( y j , y i ) or ( y j , y j ) (y_j,y_j) ( y j , y j ) , and no residue disc meets the supports of both A A A and B B B . In h ( [ A ] , [ B ] ) = ∑ v h v ( A , B ) h([A],[B])=\sum_v h_v(A,B) h ([ A ] , [ B ]) = ∑ v h v ( A , B ) , the terms at v ∤ q N X v\nmid qN_X v ∤ q N X vanish by [23 Lemma 3.1] , as X X X has good reduction at v v v [16 VI, Proposition 6.7] , and those at v ∣ N X v\mid N_X v ∣ N X vanish by (V). So
λ ( a i j b i j Z ) = β ( κ [ A ] , κ [ B ] ) = q − 1 h q ( A , B ) , \lambda(a_{ij}b_{ij}^{Z})=\beta(\kappa[A],\kappa[B])=q^{-1}h_q(A,B), λ ( a ij b ij Z ) = β ( κ [ A ] , κ [ B ]) = q − 1 h q ( A , B ) ,
which lies in Z q \mathbb{Z}_q Z q by Lemma 8.1 (c). By (C) and Nakayama’s lemma the a i j b i j Z a_{ij}b_{ij}^{Z} a ij b ij Z span T q \mathbb{T}_q T q over Z q \mathbb{Z}_q Z q , so λ ( T q ) ⊂ Z q \lambda(\mathbb{T}_q)\subset\mathbb{Z}_q λ ( T q ) ⊂ Z q . Thus β \beta β is Z q \mathbb{Z}_q Z q -valued, h ( M , M ) ⊂ q Z q h(M,M)\subset q\mathbb{Z}_q h ( M , M ) ⊂ q Z q , and the reduction λ ˉ \bar{\lambda} λ ˉ of λ \lambda λ is as stated; (C) makes it unique.
(b) Let z ∈ X ( Q ) z\in X(\mathbb{Q}) z ∈ X ( Q ) , z ≠ y z\ne y z = y , with z ˉ = y ˉ \bar{z}=\bar{y} z ˉ = y ˉ . Let w w w be a parameter of X X X at y ˉ \bar{y} y ˉ over Z q \mathbb{Z}_q Z q with w ( y ) = 0 w(y)=0 w ( y ) = 0 and d w ˉ ( δ ) = 1 d\bar{w}(\delta)=1 d w ˉ ( δ ) = 1 , and put u : = w ( z ) u:=w(z) u := w ( z ) and k : = v q ( u ) ≥ 1 k:=v_q(u)\ge1 k := v q ( u ) ≥ 1 . Fix Z ∈ Z Z\in\mathcal{Z} Z ∈ Z , and let P : = [ z − y ] P:=[z-y] P := [ z − y ] and D : = Fix Z − Z ( y ) − Z ( z ) D:=\operatorname{Fix}Z-Z(y)-Z(z) D := Fix Z − Z ( y ) − Z ( z ) . By (F), no point of D Z D_Z D Z or of Z ( z ) − Z ( y ) Z(z)-Z(y) Z ( z ) − Z ( y ) lies in the residue disc of y ˉ \bar{y} y ˉ , and no term of Z Z Z passes through ( y , y ) (y,y) ( y , y ) , ( y , z ) (y,z) ( y , z ) , ( z , y ) (z,y) ( z , y ) or ( z , z ) (z,z) ( z , z ) . As in (a), all local terms of h ( P , [ D ] ) h(P,[D]) h ( P , [ D ]) other than h q h_q h q vanish, and [ D ] = [ D Z ] − Z P [D]=[D_Z]-ZP [ D ] = [ D Z ] − Z P , so
h q ( z − y , D ) = h ( P , [ D Z ] ) − h ( P , Z P ) . (2) h_q(z-y,D)=h(P,[D_Z])-h(P,ZP). \tag*{(2)} h q ( z − y , D ) = h ( P , [ D Z ]) − h ( P , Z P ) . (2)
Below, the product of u u u or of u 2 / q u^2/q u 2 / q with an element of F q \mathbb{F}_q F q is the product with any lift; modulo q k + 1 q^{k+1} q k + 1 it does not depend on the lift, as v q ( u 2 / q ) = 2 k − 1 ≥ k v_q(u^2/q)=2k-1\ge k v q ( u 2 / q ) = 2 k − 1 ≥ k .
The divisor Z ( z ) − Z ( y ) Z(z)-Z(y) Z ( z ) − Z ( y ) is rational and misses the disc of y ˉ \bar{y} y ˉ , and it has degree 0 on every residue disc, as z ˉ = y ˉ \bar{z}=\bar{y} z ˉ = y ˉ and each term of Z Z Z extends to a finite flat correspondence on X X X . Hence ω Z ( z ) − Z ( y ) \omega_{Z(z)-Z(y)} ω Z ( z ) − Z ( y ) is q q q times an integral form on the disc of y ˉ \bar{y} y ˉ [23 Proposition 2.6 (a)] . By Lemma 8.1 (a), ω D Z \omega_{D_Z} ω D Z is integral there, with reduction ω ˉ D Z \bar{\omega}_{D_Z} ω ˉ D Z . Integrating from y y y to z z z term by term [23 (3.1)] ,
h q ( z − y , D ) ≡ u ω ˉ D Z ( δ ) ( m o d q k + 1 ) . h_q(z-y,D)\equiv u\bar{\omega}_{D_Z}(\delta)\pmod{q^{k+1}}. h q ( z − y , D ) ≡ u ω ˉ D Z ( δ ) ( mod q k + 1 ) .
For ω ∈ L \omega\in\mathcal{L} ω ∈ L , ⟨ log P , ω ⟩ = ∫ y z ω ≡ u ( ω / d w ) ( y ) ( m o d q k + 1 ) \langle\log P,\omega\rangle=\int_y^z\omega\equiv u(\omega/dw)(y)\pmod{q^{k+1}} ⟨ log P , ω ⟩ = ∫ y z ω ≡ u ( ω / d w ) ( y ) ( mod q k + 1 ) , in the same way. So κ P ≡ ( u / q ) ev ~ \kappa P\equiv(u/q)\widetilde{\operatorname{ev}} κ P ≡ ( u / q ) ev modulo q k Lie q^k\operatorname{Lie} q k Lie , where ev ~ ∈ Lie \widetilde{\operatorname{ev}}\in\operatorname{Lie} ev ∈ Lie is ω ↦ ( ω / d w ) ( y ) \omega\mapsto(\omega/dw)(y) ω ↦ ( ω / d w ) ( y ) , a lift of ev δ \operatorname{ev}_{\delta} ev δ . As β \beta β is Z q \mathbb{Z}_q Z q -valued by (a) and 2 k ≥ k + 1 2k\ge k+1 2 k ≥ k + 1 , modulo q k + 1 q^{k+1} q k + 1
h ( P , [ D Z ] ) = q β ( κ P , κ [ D Z ] ) ≡ u b ~ ( ev δ , κ [ D Z ] ) , h ( P , Z P ) = q β ( κ P , Z κ P ) ≡ u 2 q Q Z . h(P,[D_Z])=q\beta(\kappa P,\kappa[D_Z])\equiv u\widetilde{b}(\operatorname{ev}_{\delta},\kappa[D_Z]),\qquad h(P,ZP)=q\beta(\kappa P,Z\kappa P)\equiv\frac{u^2}{q}Q_Z. h ( P , [ D Z ]) = q β ( κ P , κ [ D Z ]) ≡ u b ( ev δ , κ [ D Z ]) , h ( P , Z P ) = q β ( κ P , Z κ P ) ≡ q u 2 Q Z .
So (9.1) becomes
u L Z + u 2 q Q Z ≡ 0 ( m o d q k + 1 ) . uL_Z+\frac{u^2}{q}Q_Z\equiv0\pmod{q^{k+1}}. u L Z + q u 2 Q Z ≡ 0 ( mod q k + 1 ) .
If k ≥ 2 k\ge2 k ≥ 2 , this says L Z = 0 L_Z=0 L Z = 0 . If k = 1 k=1 k = 1 , then s : = u / q ( m o d q ) s:=u/q\pmod q s := u / q ( mod q ) is non-zero and L Z + s Q Z = 0 L_Z+sQ_Z=0 L Z + s Q Z = 0 . So L Z = c Q Z L_Z=cQ_Z L Z = c Q Z for every Z ∈ Z Z\in\mathcal{Z} Z ∈ Z , with c = 0 c=0 c = 0 or c = − s c=-s c = − s , contrary to the hypothesis. □ \square □
The numbers in (b) depend on δ \delta δ only through a scalar: replacing δ \delta δ by μ δ \mu\delta μ δ , μ ∈ F q × \mu\in\mathbb{F}_q^\times μ ∈ F q × , multiplies L Z L_Z L Z by μ \mu μ and Q Z Q_Z Q Z by μ 2 \mu^2 μ 2 . So the hypothesis of (b) does not depend on δ \delta δ , and, when Q Z ≠ 0 Q_Z\ne0 Q Z = 0 , we can use s Z : = L Z / Q Z s_Z:=L_Z/Q_Z s Z := L Z / Q Z of w ( z ) / q ( m o d q ) w(z)/q\pmod q w ( z ) / q ( mod q ) that Z Z Z allows becomes s Z / μ s_Z/\mu s Z / μ . Hence the ratio s Z / s Z ′ s_Z/s_{Z'} s Z / s Z ′ does not depend on δ \delta δ , and the condition s Z ≠ s Z ′ s_Z\ne s_{Z'} s Z = s Z ′ is that this ratio is not 1 1 1 .
In the language of quadratic Chabauty, the left side u L Z + ( u 2 / q ) Q Z uL_Z+(u^2/q)Q_Z u L Z + ( u 2 / q ) Q Z of the last congruence in the proof, with u = w ( z ) u=w(z) u = w ( z ) and k = v q ( u ) k=v_q(u) k = v q ( u ) , is the expansion at y y y modulo q k + 1 q^{k+1} q k + 1 of the quadratic Chabauty function of Z Z Z with base point y y y : L Z ≠ 0 L_Z \ne0 L Z = 0 implies that y y y is a simple zero of it, and s Z ≠ s Z ′ s_Z \ne s_{Z'} s Z = s Z ′ that the functions of Z Z Z and Z ′ Z' Z ′ have no common zero z z z with v q ( w ( z ) ) = 1 v_q(w(z)) = 1 v q ( w ( z )) = 1 .
The rank
Proposition 10.1. rank J ( Q ) = 14 \operatorname{rank} J(\mathbb{Q}) = 14 rank J ( Q ) = 14 .
Proof. By Chen’s isogeny (proof of Lemma 7.1 ), J J J is isogenous over Q \mathbb{Q} Q to ∏ f A f \prod_f A_f ∏ f A f , over the Galois orbits of newforms f f f of weight 2 2 2 and level 625 625 625 with Fricke eigenvalue + 1 +1 + 1 , where A f A_f A f is the quotient of J 0 ( 625 ) J_0(625) J 0 ( 625 ) attached to f f f . A computation in PARI/GP gives four orbits, of dimensions 2 2 2 , 2 2 2 , 2 2 2 and 8 8 8 . The root number of such an f f f is − 1 -1 − 1 , so
L ′ ( f , 1 ) = 2 ∑ n ≥ 1 a n n E 1 ( 2 π n 25 ) , E 1 ( x ) : = ∫ 1 ∞ e − x t d t t . L'(f,1) = 2\sum_{n\geq1}\frac{a_n}{n}E_1\left(\frac{2\pi n}{25}\right), \qquad E_1(x) := \int_1^\infty e^{-xt}\frac{\mathrm{d}t}{t}. L ′ ( f , 1 ) = 2 n ≥ 1 ∑ n a n E 1 ( 25 2 π n ) , E 1 ( x ) := ∫ 1 ∞ e − x t t d t .
A computation in ball arithmetic with python-flint gives L ′ ( f , 1 ) > 0 L'(f,1) > 0 L ′ ( f , 1 ) > 0 for all fourteen f f f ; the smallest of these values is 0.782729318982775 ± 1.7 ⋅ 10 − 16 0.782729318982775 \pm1.7 \cdot10^{-16} 0.782729318982775 ± 1.7 ⋅ 1 0 − 16 . By S.-W. Zhang’s theorem [44 Theorem A] , which generalizes results of Gross and Zagier [19 ] and of Kolyvagin [27 ] to totally real fields, rank A f ( Q ) = [ O f : Z ] ⋅ ord s = 1 L ( f , s ) = dim A f \operatorname{rank} A_f(\mathbb{Q}) = [\mathcal{O}_f : \mathbb{Z}]\cdot\operatorname{ord}_{s=1} L(f,s) = \dim A_f rank A f ( Q ) = [ O f : Z ] ⋅ ord s = 1 L ( f , s ) = dim A f for these f f f , where O f \mathcal{O}_f O f is the ring generated by the coefficients of f f f . So rank J ( Q ) = 2 + 2 + 2 + 8 = 14 \operatorname{rank} J(\mathbb{Q}) = 2 + 2 + 2 + 8 = 14 rank J ( Q ) = 2 + 2 + 2 + 8 = 14 . □
The prime 1151
We apply Proposition 9.1 to X X X , with H = N \mathcal{H} = \mathbb{N} H = N , N X = 25 N_X = 25 N X = 25 , q : = 1151 q := 1151 q := 1151 , the nine points y D y_D y D , D ∈ D D \in\mathcal{D} D ∈ D , and Z : = { Z 1 , Z 2 } \mathcal{Z} := \{Z_1, Z_2\} Z := { Z 1 , Z 2 } .
The points y D y_D y D modulo q 2 q^2 q 2 . The computations below use v D v_D v D (§ 5 ) for y D y_D y D modulo q 2 q^2 q 2 ; we now show that v D ≡ y D ( m o d q 2 ) v_D \equiv y_D \pmod{q^2} v D ≡ y D ( mod q 2 ) . By the proof of Theorem 1.4 , each y ˉ D \bar{y}_D y ˉ D is one of the v ˉ D ′ \bar{v}_{D'} v ˉ D ′ . For D ≠ − 3 D \ne-3 D = − 3 , the moduli description of (C) below, on a curve E E E with j j j -invariant j D j_D j D over the Witt vectors of F q 2 \mathbb{F}_{q^2} F q 2 (of F q 5 \mathbb{F}_{q^5} F q 5 for D = − 163 D = -163 D = − 163 ) on which E [ 25 ] E[25] E [ 25 ] is rational, finds a point over j D j_D j D at which the values of the forms are proportional to v D v_D v D modulo q 2 q^2 q 2 . So j ( v D ) ≡ j D ( m o d q 2 ) j(v_D) \equiv j_D \pmod{q^2} j ( v D ) ≡ j D ( mod q 2 ) for D ≠ − 3 D \ne-3 D = − 3 ; as the j D j_D j D are distinct modulo q q q , y ˉ D = v ˉ D \bar{y}_D = \bar{v}_D y ˉ D = v ˉ D for every D D D , including − 3 -3 − 3 , and as j − j D j-j_D j − j D is a parameter at y ˉ D \bar{y}_D y ˉ D (Lemma 2.2), v D ≡ y D ( m o d q 2 ) v_D \equiv y_D \pmod{q^2} v D ≡ y D ( mod q 2 ) for D ≠ − 3 D \ne-3 D = − 3 .
At y − 3 y_{-3} y − 3 the fourteen values vanish, as the automorphisms of E − 3 E_{-3} E − 3 fix its Cartan subgroup, so we use a unit instead. Let e ∈ Λ e \in\Lambda e ∈ Λ have e n = − 2 e_n = -2 e n = − 2 for n ≡ ± 2 ( m o d 5 ) n \equiv\pm2 \pmod5 n ≡ ± 2 ( mod 5 ) and e n = 0 e_n = 0 e n = 0 otherwise. By Lemma 4.1 and the values of Λ ( n ) \Lambda(n) Λ ( n ) , the divisor of u e u_e u e is 5 ∑ d ≡ ± 1 c d − 5 ∑ d ≡ ± 2 c d 5\sum_{d\equiv\pm1} c_d - 5\sum_{d\equiv\pm2} c_d 5 ∑ d ≡ ± 1 c d − 5 ∑ d ≡ ± 2 c d . The map X → X n s + ( 5 ) X \to X^+_{\mathrm{ns}}(5) X → X ns + ( 5 ) , ( E , R ) ↦ ( E , R m o d 5 ) (E,R) \mapsto(E,R\bmod5) ( E , R ) ↦ ( E , R mod 5 ) , is defined over Q \mathbb{Q} Q , and σ d \sigma_d σ d fixes or swaps the two cusps of X n s + ( 5 ) X^+_{\mathrm{ns}}(5) X ns + ( 5 ) , which are conjugate over Q ( 5 ) \mathbb{Q}(\sqrt{5}) Q ( 5 ) , according as d ≡ ± 1 d \equiv\pm1 d ≡ ± 1 or ± 2 \pm2 ± 2 mod 5 5 5 . So this divisor is the pull-back of the difference of the two cusps (each cusp of X X X , of width 25 25 25 , has ramification index 5 5 5 over its image, of width 5 5 5 ), and as X n s + ( 5 ) X^+_{\mathrm{ns}}(5) X ns + ( 5 ) has genus 0 0 0 , u e u_e u e is constant on the fibers of X → X n s + ( 5 ) X \to X^+_{\mathrm{ns}}(5) X → X ns + ( 5 ) . Each S ν ( x ) S_\nu(x) S ν ( x ) lies in the fiber of x x x . A computation of u e u_e u e modulo q 2 q^2 q 2 , as a quotient of two forms of weight 4 4 4 , shows that u e ( v − 3 ) ≡ u e ( P ) ( m o d q 2 ) u_e(v_{-3}) \equiv u_e(P) \pmod{q^2} u e ( v − 3 ) ≡ u e ( P ) ( mod q 2 ) at every point P P P of S ν ( y − 3 ) S_\nu(y_{-3}) S ν ( y − 3 ) , ν ∈ F 5 × \nu\in\mathbb{F}_5^\times ν ∈ F 5 × , found by the moduli description on a curve E E E with j j j -invariant 0 0 0 over the Witt vectors of F q 2 \mathbb{F}_{q^2} F q 2 on which E [ 25 ] E[25] E [ 25 ] is rational, and that u e u_e u e is regular with non-zero differential at y ˉ − 3 \bar{y}_{-3} y ˉ − 3 modulo q q q . So u e − u e ( y − 3 ) u_e-u_e(y_{-3}) u e − u e ( y − 3 ) is a parameter at y ˉ − 3 \bar{y}_{-3} y ˉ − 3 , and v − 3 ≡ y − 3 m o d q 2 v_{-3}\equiv y_{-3}\bmod q^2 v − 3 ≡ y − 3 mod q 2 .
Why 1151. As q ≡ 1 m o d 25 q\equiv1\bmod25 q ≡ 1 mod 25 , the labels can be read at q q q , and Theorem 1.4 puts every rational point in the residue disc of some y D y_D y D at q q q . As q ≡ 11 m o d 12 q\equiv11\bmod12 q ≡ 11 mod 12 , the curves with j = 0 j=0 j = 0 and j = 1728 j=1728 j = 1728 are supersingular modulo q q q , and so are the y D y_D y D with D ≠ − 163 D\ne-163 D = − 163 , as these D D D are not squares modulo q q q ; while 4 q = 23 2 + 163 ⋅ 5 2 4q=23^2+163\cdot5^2 4 q = 2 3 2 + 163 ⋅ 5 2 , so y − 163 y_{-163} y − 163 is ordinary. A supersingular curve E ˉ \bar{E} E ˉ over F q \mathbb{F}_q F q has π q 2 = − q \pi_q^2=-q π q 2 = − q , which acts on E ˉ [ 25 ] \bar{E}[25] E ˉ [ 25 ] as − 1 -1 − 1 because q ≡ 1 m o d 25 q\equiv1\bmod25 q ≡ 1 mod 25 . So every point of X \mathcal{X} X over a supersingular j ∈ F q j\in\mathbb{F}_q j ∈ F q is defined over F q 2 \mathbb{F}_{q^2} F q 2 : the fixed points of the S ν S_\nu S ν , over j = 0 j=0 j = 0 and 1728 1728 1728 , and the points of S ν ( y ˉ D ) S_\nu(\bar{y}_D) S ν ( y ˉ D ) for D ≠ − 163 D\ne-163 D = − 163 . This keeps the computations at q q q over F q 2 \mathbb{F}_{q^2} F q 2 , except at y − 163 y_{-163} y − 163 : its Frobenius ± ( 23 + 5 − 163 ) / 2 \pm(23+5\sqrt{-163})/2 ± ( 23 + 5 − 163 ) /2 is a scalar modulo 5 5 5 , so the points of S ν ( y ˉ − 163 ) S_\nu(\bar{y}_{-163}) S ν ( y ˉ − 163 ) are defined over F q \mathbb{F}_q F q , and they are computed from E ˉ [ 25 ] \bar{E}[25] E ˉ [ 25 ] over F q 5 \mathbb{F}_{q^5} F q 5 . At 101 101 101 , the curve with j = 1728 j=1728 j = 1728 is ordinary and the points Φ ν \Phi_\nu Φ ν are not all defined over F 101 2 \mathbb{F}_{101^2} F 10 1 2 , so over F 101 2 \mathbb{F}_{101^2} F 10 1 2 only Z 1 Z_1 Z 1 is usable, and one correspondence gives control equations of rank at most 8 8 8 (Remark 9.2 ). At 151 151 151 , (B) fails: 151 151 151 divides # J ˉ ( F 151 ) \#\bar{J}(\mathbb{F}_{151}) # J ˉ ( F 151 ) . Below 1151 1151 1151 , the only other prime ≡ 1 m o d 25 \equiv1\bmod25 ≡ 1 mod 25 and ≡ 11 m o d 12 \equiv11\bmod12 ≡ 11 mod 12 is 251 251 251 , where three of the y D y_D y D are ordinary, none with Frobenius scalar modulo 5 5 5 , unlike y − 163 y_{-163} y − 163 at 1151 1151 1151 .
The hypotheses. The correspondences Z 1 Z_1 Z 1 and Z 2 Z_2 Z 2 are symmetric integer combinations of double cosets N g N NgN N g N with g ∉ N g\notin N g ∈ / N (proof of Lemma 6.3), and Tr ( Z ∣ L ) = 1 2 Tr ( Z ∣ H ) = 0 \operatorname{Tr}(Z\mid\mathcal{L})=\frac{1}{2}\operatorname{Tr}(Z\mid H)=0 Tr ( Z ∣ L ) = 2 1 Tr ( Z ∣ H ) = 0 by Proposition 6.4.
(B) A computation in PARI/GP gives
N q = # J ˉ ( F 1151 ) = 8164058494096884485556947262955286865234375 = 3 6 ⋅ 5 13 ⋅ 71 2 ⋅ 89 2 ⋅ 56149 ⋅ 60029 ⋅ 72211 ⋅ 943985489 , N_q=\#\bar{J}(\mathbb{F}_{1151})=8164058494096884485556947262955286865234375
=3^6\cdot5^{13}\cdot71^2\cdot89^2\cdot56149\cdot60029\cdot72211\cdot943985489, N q = # J ˉ ( F 1151 ) = 8164058494096884485556947262955286865234375 = 3 6 ⋅ 5 13 ⋅ 7 1 2 ⋅ 8 9 2 ⋅ 56149 ⋅ 60029 ⋅ 72211 ⋅ 943985489 ,
which is prime to q q q .
(M) The characteristic polynomial of T 7 T_7 T 7 on L / q \mathcal{L}/q L / q is square-free, with six linear and four quadratic factors over F q \mathbb{F}_q F q (a computation in PARI/GP). So Z q [ T 7 ] ⊂ End L \mathbb{Z}_q[T_7]\subset\operatorname{End}\mathcal{L} Z q [ T 7 ] ⊂ End L is étale of rank 14 14 14 over Z q \mathbb{Z}_q Z q , and L \mathcal{L} L and its dual Lie are free of rank one over it. As T q \mathbb{T}_q T q is commutative and contains T 7 T_7 T 7 , T q = Z q [ T 7 ] \mathbb{T}_q=\mathbb{Z}_q[T_7] T q = Z q [ T 7 ] , and (M) holds.
(R) This is Proposition 10.1 .
(F) The closure of S ν S_\nu S ν in X × X \mathcal{X}\times\mathcal{X} X × X is the correspondence S ˉ ν \bar{S}_\nu S ˉ ν given by the same moduli description over Z q \mathbb{Z}_q Z q , which lies over the diagonal of the j j j -line. In the order of D D D , the j D j_D j D are 0 0 0 , 78 78 78 , 1094 1094 1094 , 1054 1054 1054 , 76 76 76 , 69 69 69 , 868 868 868 , 1114 1114 1114 and 658 658 658 modulo q q q , distinct, and 1728 ≡ 577 1728\equiv577 1728 ≡ 577 . So a point ( y ˉ D , y ˉ D ′ ) (\bar{y}_D,\bar{y}_{D'}) ( y ˉ D , y ˉ D ′ ) of S ˉ ν \bar{S}_\nu S ˉ ν has D = D ′ D=D' D = D ′ and is a fixed point of S ˉ ν \bar{S}_\nu S ˉ ν , which by Proposition 6.5 lies over j = 0 j=0 j = 0 , 1728 1728 1728 or ∞ \infty ∞ . Only D = − 3 D=-3 D = − 3 remains, and the fixed points over j = 0 j=0 j = 0 are the reductions of Q μ Q_\mu Q μ and Q μ ′ Q'_\mu Q μ ′ , which differ from y ˉ − 3 \bar{y}_{-3} y ˉ − 3 : the automorphisms of E − 3 E_{-3} E − 3 fix the Cartan subring of y − 3 y_{-3} y − 3 and not those of Q μ Q_\mu Q μ , Q μ ′ Q'_\mu Q μ ′ .
(V) This is Corollary 7.3: by (F), for points y , z ∈ X ( Q ) y,z\in X(\mathbb{Q}) y , z ∈ X ( Q ) with reductions among the y ˉ D \bar{y}_D y ˉ D , none of ( y , y ) (y,y) ( y , y ) , ( z , z ) (z,z) ( z , z ) , ( y , z ) (y,z) ( y , z ) lies on an S ν S_\nu S ν .
(C) A computation in Python gives the a i j a_{ij} a ij and b i j Z b^Z_{ij} b ij Z modulo q q q and the right sides of the control equations, with the points of Fix S ν \operatorname{Fix}S_\nu Fix S ν and S ˉ ν ( y ˉ D ) \bar{S}_\nu(\bar{y}_D) S ˉ ν ( y ˉ D ) found as above. The matrices of the S ν S_\nu S ν on L / q 2 \mathcal{L}/q^2 L / q 2 come from the moduli description: over a supersingular j j j , the values of the fourteen forms at the 250 points ( E , R ) (E,R) ( E , R ) , for a fixed differential on E E E , are computed over the Witt vectors of F q 2 \mathbb{F}_{q^2} F q 2 modulo q 2 q^2 q 2 from the values of the weight-one Eisenstein series at the points of E [ 25 ] E[25] E [ 25 ] [26 ] , as in [20 ] , §§4.1, 6.1, since modulo q 2 q^2 q 2 the fourteen forms are combinations of the averaged products of Section 5 (an identity checked on 60 coefficients); and S ν S_\nu S ν is the matrix that sends the values at each point to the sum of the values at its six S ν S_\nu S ν -related points. These matrices have traces − 11 -11 − 11 , 4 4 4 , − 11 -11 − 11 , as Proposition 6.5 and the Lefschetz formula require, and 1 + S 1 + S 2 + S 3 + S 4 = 0 1+S_1+S_2+S_3+S_4=0 1 + S 1 + S 2 + S 3 + S 4 = 0 on L / q 2 L/q^2 L / q 2 , as 1 + ∑ ν S ν 1+\sum_\nu S_\nu 1 + ∑ ν S ν factors through X n s + ( 5 ) X^+_{\mathrm{ns}}(5) X ns + ( 5 ) , of genus 0.
For each divisor D D D that enters, the function f D f_D f D is computed on the canonical curve over Z / q 2 \mathbb{Z}/q^2 Z / q 2 by linear algebra in the sections of 4 K X 4K_X 4 K X , in the manner of Khuri-Makdisi [25 ] : an effective divisor of degree 28 is represented by the sections of 4 K X 4K_X 4 K X that vanish on it. Products of sections are read on their first 30 coefficients at ∞ \infty ∞ , which lie in ( Z / q 2 ) [ ζ 25 ] (\mathbb{Z}/q^2)[\zeta_{25}] ( Z / q 2 ) [ ζ 25 ] and determine a section of 8 K X 8K_X 8 K X : its expansions at the ten cusps are the conjugates of that at ∞ \infty ∞ , it vanishes at a cusp to order 8 less than its expansion there, and deg 8 K X = 208 < 10 ⋅ ( 30 − 8 ) \deg8K_X=208<10\cdot(30-8) deg 8 K X = 208 < 10 ⋅ ( 30 − 8 ) . A double-and-add chain along N q N_q N q keeps, for each n n n it reaches, an effective divisor of degree 28 linearly equivalent to n D + D 0 nD+D_0 n D + D 0 , and a product of quotients of sections of 4 K X 4K_X 4 K X whose divisor is the difference, starting from D 0 D_0 D 0 itself at n = 0 n=0 n = 0 . Here D 0 D_0 D 0 is a sum of 28 points over a supersingular j ∈ F q j\in\mathbb{F}_q j ∈ F q other than 0, 1728 and the j j j -invariants of the nine CM points (in the computation j = 62 j=62 j = 62 ); every divisor D D D that enters lies over these, so the residue discs of D 0 D_0 D 0 contain no point of D D D . At n = N q n=N_q n = N q the reduction of the divisor kept is linearly equivalent to D ‾ 0 \overline{D}_0 D 0 , as N q [ D ‾ ] = 0 N_q[\overline{D}]=0 N q [ D ] = 0 , and one more quotient replaces it by a divisor D ‾ 0 ′ ≡ D 0 ( m o d q ) \overline{D}'_0\equiv D_0\pmod q D 0 ′ ≡ D 0 ( mod q ) with one point in the residue disc of each point of D 0 D_0 D 0 . This gives f D f_D f D , with div f D = N q D + D 0 − D 0 ′ \operatorname{div} f_D=N_qD+D_0-D'_0 div f D = N q D + D 0 − D 0 ′ , so that D 1 = D 0 ′ − D 0 D_1=D'_0-D_0 D 1 = D 0 ′ − D 0 has degree 0 on every residue disc, and f ‾ D \overline{f}_D f D , with divisor N q D ‾ N_q\overline{D} N q D . For the two divisors of each control equation, the divisors D 0 D_0 D 0 lie in disjoint sets of residue discs. The proof of Lemma 8.1 (a)–(c) then applies with the residue discs of D 0 D_0 D 0 counted among those of the points of D D D , and gives κ \kappa κ and h q h_q h q modulo q 2 q^2 q 2 .
The 36 products a i j b i j Z 1 a_{ij}b_{ij}^{Z_1} a ij b ij Z 1 span a space of dimension 8, and together with the 36 for Z 2 Z_2 Z 2 they span T q / q \mathbb{T}_q/q T q / q . The right sides telescope like the products (Remark 9.2 ), so the 72 control equations reduce to 16 in 14 unknowns, and they are consistent: two conditions modulo q q q that a non-zero local term at 5 would be expected to break. So Proposition 9.1 (a) gives h ( M , M ) ⊂ q Z q h(M,M)\subset q\mathbb{Z}_q h ( M , M ) ⊂ q Z q and the form b ~ \widetilde{b} b .
The disc coordinate. For D ≠ − 3 D\ne-3 D = − 3 let k D : = j D ( j D − 1728 ) k_D:=j_D(j_D-1728) k D := j D ( j D − 1728 ) , a D : = − 3 k D a_D:=-3k_D a D := − 3 k D and b D : = − 2 k D ( j D − 1728 ) b_D:=-2k_D(j_D-1728) b D := − 2 k D ( j D − 1728 ) , so that y 2 = x 3 + a D x + b D y^2=x^3+a_Dx+b_D y 2 = x 3 + a D x + b D has j j j -invariant j D j_D j D . As X → X ( 1 ) X\to X(1) X → X ( 1 ) is unramified at y ‾ D \overline{y}_D y D (Lemma 2.2), the family y 2 = x 3 + a D x + b D + t y^2=x^3+a_Dx+b_D+t y 2 = x 3 + a D x + b D + t over F q [ [ t ] ] \mathbb{F}_q[[t]] F q [[ t ]] lifts to a unique formal curve through y ‾ D \overline{y}_D y D on X F q X_{\mathbb{F}_q} X F q , and we let δ D \delta_D δ D be its tangent vector at t = 0 t=0 t = 0 . Equivalently, d j ( δ D ) = ∂ j / ∂ b ( a D , b D ) dj(\delta_D)=\partial j/\partial b(a_D,b_D) d j ( δ D ) = ∂ j / ∂ b ( a D , b D ) for j ( a , b ) = 6912 a 3 / ( 4 a 3 + 27 b 2 ) j(a,b)=6912a^3/(4a^3+27b^2) j ( a , b ) = 6912 a 3 / ( 4 a 3 + 27 b 2 ) . At y ‾ − 3 \overline{y}_{-3} y − 3 , where a D = b D = 0 a_D=b_D=0 a D = b D = 0 , we take δ − 3 \delta_{-3} δ − 3 with e δ − 3 = 716 v ‾ − 3 e\delta_{-3}=716\overline{v}_{-3} e δ − 3 = 716 v − 3 in the basis dual to the fourteen forms, for the primitive vector v − 3 v_{-3} v − 3 of the deposite [22 ] ; at each y ‾ D \overline{y}_D y D , e v D e v_D e v D is a multiple of v ‾ D \overline{v}_D v D , as the forms give the canonical embedding.
The proof of Theorem 1.1 . Let x ∈ X ( Q ) x\in X(\mathbb{Q}) x ∈ X ( Q ) . By Theorem 1.4 , x x x has the reduction modulo 1151 of some y D y_D y D , D ∈ D D\in\mathcal{D} D ∈ D . The hypotheses of Proposition 9.1 hold at q = 1151 q=1151 q = 1151 , as checked above. In every row of Table 1 , L Z 1 L_{Z_1} L Z 1 and L Z 2 L_{Z_2} L Z 2 are non-zero and s Z 1 / s Z 2 ≠ 1 s_{Z_1}/s_{Z_2}\ne1 s Z 1 / s Z 2 = 1 , that is, L Z 1 Q Z 2 ≠ L Z 2 Q Z 1 L_{Z_1}Q_{Z_2}\ne L_{Z_2}Q_{Z_1} L Z 1 Q Z 2 = L Z 2 Q Z 1 . So Proposition 9.1 (b) gives x = y D x=y_D x = y D . The nine points y D y_D y D are distinct, as their j j j -invariants are. ∎
Declarations
Use of AI. This paper and its results were produced by an AI research system that the author designed, built and directs, running on Anthropic’s Claude Opus 5.5 in Claude Code and OpenAI’s GPT-6 Astra in Codex. The author chooses the problems and prioritizes the work. Separate AI agents checked each result, and a second program confirmed the computations the proofs use, with any exceptions listed in the public deposit. The system also searched the literature and checked the citations against reliable sources. The author read the paper, reran the computations from the deposit and takes full responsibility for all of the paper’s contents.
Data availability. The programs and data used in this paper, with instructions for rerunning them, are deposited at Zenodo [22 ] .
References
Toronto, Canada
Email address: manvirjaswal@icloud.com
URL: https://manvirjaswal.github.io