Introduction and statement of results
A classical theme in arithmetic is to understand how the algebraic structure associated to an arithmetic object varies after reduction modulo primes. Let E / Q E/\mathbb{Q} E / Q be an elliptic curve with conductor N ( E ) N(E) N ( E ) . For every prime p ∤ N ( E ) p \nmid N(E) p ∤ N ( E ) , the reduction E ‾ / F p \overline{E}/\mathbb{F}_p E / F p is again an elliptic curve, and the set E ‾ ( F p ) \overline{E}(\mathbb{F}_p) E ( F p ) of its rational points forms a finite abelian group. It is therefore natural to ask how the group structure of E ‾ ( F p ) \overline{E}(\mathbb{F}_p) E ( F p ) varies as p p p ranges over primes of good reduction. In particular, one may count how often E ‾ ( F p ) \overline{E}(\mathbb{F}_p) E ( F p ) is cyclic using the function
π E ( x ) : = # { p ≤ x : ( p , N ( E ) ) = 1 , E ‾ ( F p ) is cyclic } . (1) \pi_E(x) := \#\{p \le x : (p, N(E)) = 1,\ \overline{E}(\mathbb{F}_p)\text{ is cyclic}\}. \tag*{(1)} π E ( x ) := # { p ≤ x : ( p , N ( E )) = 1 , E ( F p ) is cyclic } . (1)
The cyclicity problem has its origins in the work of Borosh, Moreno and Porta [4 ] , Lang, Trotter [24 ] and Serre [33 ] (see § 2.1 for a detailed description).
For m ∈ N m \in\mathbb{N} m ∈ N , let E [ m ] E[m] E [ m ] denote the group of m m m -torsion points of E E E , and let Q ( E [ m ] ) \mathbb{Q}(E[m]) Q ( E [ m ]) be the m m m th division field of E E E , obtained by adjoining to Q \mathbb{Q} Q the coordinates of all the points in E [ m ] E[m] E [ m ] . Under the generalized Riemann hypothesis (GRH), Serre showed that π E ( x ) = δ E li ( x ) + o ( x / log x ) \pi_E(x) = \delta_E \operatorname{li}(x) + o(x/\log x) π E ( x ) = δ E li ( x ) + o ( x / log x ) , where
li ( x ) = ∫ 2 x d t log t , \operatorname{li}(x) = \int_2^x \frac{dt}{\log t}, li ( x ) = ∫ 2 x log t d t ,
and δ E \delta_E δ E is given by the density
δ E : = ∑ m ≥ 1 μ ( m ) [ Q ( E [ m ] ) : Q ] . (2) \delta_E := \sum_{m \ge1} \frac{\mu(m)}{[\mathbb{Q}(E[m]) : \mathbb{Q}]}. \tag*{(2)} δ E := m ≥ 1 ∑ [ Q ( E [ m ]) : Q ] μ ( m ) . (2)
where μ \mu μ is the Möbius function. Significant progress has been made on the cyclicity problem over the years - we discuss this in § 2.1 . Recently, Akbal and Guloğlu [1 ] imposed an additional arithmetic progression condition on the primes in (1.1). Letting f ∈ N f \in\mathbb{N} f ∈ N and a ∈ ( Z / f Z ) ∗ a \in(\mathbb{Z}/f\mathbb{Z})^* a ∈ ( Z / f Z ) ∗ , they obtained asymptotic estimates for
π E ( x ; f , a ) : = # { p ≤ x : ( p , N ( E ) ) = 1 , p ≡ a ( m o d f ) , E ‾ ( F p ) is cyclic } . (3) \pi_E(x; f, a) := \#\{p \le x : (p, N(E)) = 1,\ p \equiv a \pmod f,\ \overline{E}(\mathbb{F}_p)\text{ is cyclic}\}. \tag*{(3)} π E ( x ; f , a ) := # { p ≤ x : ( p , N ( E )) = 1 , p ≡ a ( mod f ) , E ( F p ) is cyclic } . (3)
We give a detailed survey of the surrounding literature in § 2 .
Function fields provide a particularly attractive setting for the study of such problems. The polynomial ring F q [ T ] \mathbb{F}_q[T] F q [ T ] exhibits a close arithmetic analogy with the integers, while the distribution of prime polynomials is often more explicit and allows sharper counting estimates. It is therefore natural to ask for function-field analogues of classical questions concerning reductions of elliptic curves. This naturally leads to Drinfeld modules, whose arithmetic provides a framework for formulating and studying such questions over function fields.
Notation and setup
Let F q \mathbb{F}_q F q be a finite field of characteristic p p p , A = F q [ T ] A = \mathbb{F}_q[T] A = F q [ T ] and K = F q ( T ) K = \mathbb{F}_q(T) K = F q ( T ) . Let τ : x ↦ x q \tau: x \mapsto x^q τ : x ↦ x q be the Frobenius endomorphism of the additive group scheme G a \mathbb{G}_a G a over Spec K \operatorname{Spec} K Spec K . Let i : A → K i : A \to K i : A → K be the canonical inclusion. Let End K ( G a ) \operatorname{End}_K(\mathbb{G}_a) End K ( G a ) be the K K K -endomorphism ring G a \mathbb{G}_a G a over Spec K \operatorname{Spec} K Spec K . Define the subring K { τ } K\{\tau\} K { τ } of End K ( G a ) \operatorname{End}_K(\mathbb{G}_a) End K ( G a ) by K { τ } : = { ∑ j = 0 n a j τ j : n ≥ 0 , a j ∈ K } K\{\tau\} := \left\{\sum_{j=0}^{n} a_j\tau^j : n \ge0, a_j \in K\right\} K { τ } := { ∑ j = 0 n a j τ j : n ≥ 0 , a j ∈ K } , where a j ∈ K a_j \in K a j ∈ K is identified with the endomorphism defined by multiplication by a j a_j a j . For all a ∈ K a \in K a ∈ K , one has τ a = a q τ \tau a = a^q\tau τ a = a q τ , and hence K { τ } K\{\tau\} K { τ } is a noncommutative ring, usually called the ring of twisted polynomials. We have two homomorphisms: ϵ : K ⟶ K { τ } \epsilon: K \longrightarrow K\{\tau\} ϵ : K ⟶ K { τ } , ϵ ( a ) = a \epsilon(a) = a ϵ ( a ) = a , and D : K { τ } ⟶ K D : K\{\tau\} \longrightarrow K D : K { τ } ⟶ K with D ( ∑ j = 0 n a j τ j ) = a 0 D\left(\sum_{j=0}^{n} a_j\tau^j\right) = a_0 D ( ∑ j = 0 n a j τ j ) = a 0 .
A Drinfeld A A A -module ϕ \phi ϕ over K K K is defined as a ring homomorphism
ϕ : A → K { τ } , a ↦ ϕ a , \phi: A \to K\{\tau\},\quad a \mapsto\phi_a, ϕ : A → K { τ } , a ↦ ϕ a ,
such that i = D ∘ ϕ i = D \circ\phi i = D ∘ ϕ and ϕ ≠ ϵ ∘ i \phi\ne\epsilon\circ i ϕ = ϵ ∘ i . Let deg τ ϕ a \deg_\tau\phi_a deg τ ϕ a denote the degree of ϕ a \phi_a ϕ a in the variable τ \tau τ . The rank r r r of a Drinfeld module ϕ \phi ϕ is the unique positive integer [12 Proposition 2.1] such that deg τ ϕ a = r ⋅ deg a \deg_\tau\phi_a = r \cdot\deg a deg τ ϕ a = r ⋅ deg a for all a ∈ A a \in A a ∈ A with a ≠ 0 a \ne0 a = 0 . We write ϕ ≃ K ψ \phi\simeq_K \psi ϕ ≃ K ψ if there exists some u ∈ K ∗ u \in K^* u ∈ K ∗ such that ψ a = u − 1 ϕ a u \psi_a = u^{-1}\phi_a u ψ a = u − 1 ϕ a u for every a ∈ A a \in A a ∈ A . The canonical example of a Drinfeld A A A -module over K K K of rank 1 is the Carlitz module ψ C a r l \psi^{\mathrm{Carl}} ψ Carl , defined by ψ T C a r l = T + τ \psi^{\mathrm{Carl}}_T = T + \tau ψ T Carl = T + τ .
Throughout the paper, by a prime p ∈ A \mathfrak{p} \in A p ∈ A we mean a monic irreducible polynomial in A A A . A Drinfeld module has good reduction at all but finitely many primes [17 Definition 4.10.1] . We denote by P ϕ \mathcal{P}_\phi P ϕ the set of primes of good reduction for ϕ \phi ϕ . For each p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_\phi p ∈ P ϕ , let F p = A / p A \mathbb{F}_{\mathfrak{p}} = A/\mathfrak{p}A F p = A / p A be the residue field at p \mathfrak{p} p . The reduction of ϕ \phi ϕ modulo p \mathfrak{p} p defines an algebra homomorphism A → F p { τ } A \to\mathbb{F}_{\mathfrak{p}}\{\tau\} A → F p { τ } , and hence an A A A -module structure on F p \mathbb{F}_{\mathfrak{p}} F p , which we denote by ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) . Then, by the theory of torsion points for Drinfeld modules and that of finitely generated modules over a PID, there exist uniquely determined monic polynomials d 1 , p , … , d r , p ∈ A \mathfrak{d}_{1,\mathfrak{p}},\ldots,\mathfrak{d}_{r,\mathfrak{p}} \in A d 1 , p , … , d r , p ∈ A satisfying d i − 1 , p A ∣ d i , p A \mathfrak{d}_{i-1,\mathfrak{p}}A \mid\mathfrak{d}_{i,\mathfrak{p}}A d i − 1 , p A ∣ d i , p A for 2 ≤ i ≤ r 2 \le i \le r 2 ≤ i ≤ r such that
ϕ ( F p ) ≃ A d 1 , p A × A d 2 , p A × ⋯ × A d r , p A . (4) \phi(\mathbb{F}_{\mathfrak{p}}) \simeq\frac{A}{\mathfrak{d}_{1,\mathfrak{p}}A} \times\frac{A}{\mathfrak{d}_{2,\mathfrak{p}}A} \times\cdots\times\frac{A}{\mathfrak{d}_{r,\mathfrak{p}}A}. \tag*{(4)} ϕ ( F p ) ≃ d 1 , p A A × d 2 , p A A × ⋯ × d r , p A A . (4)
Let K ∞ = F q ( ( 1 / T ) ) K_\infty= \mathbb{F}_q((1/T)) K ∞ = F q (( 1/ T )) denote the completion of K K K at the prime at infinity in K K K . A sign function is a homomorphism sgn : K ∞ ∗ → F q ∗ \operatorname{sgn} : K_\infty^* \to\mathbb{F}_q^* sgn : K ∞ ∗ → F q ∗ , whose restriction to F q ∗ \mathbb{F}_q^* F q ∗ is the identity (see [17 Definition 7.2.1(1)] ). A Drinfeld A A A -module ψ \psi ψ over K K K of rank 1 is called sign-normalized if, for some sign function sgn \operatorname{sgn} sgn , the leading coefficient of ψ a \psi_a ψ a equals sgn ( a ) \operatorname{sgn}(a) sgn ( a ) for every nonzero a ∈ A a \in A a ∈ A (see [17 Definition 7.2.14] ). In our setting, this is equivalent to saying that ψ T = T + c τ \psi_T = T + c\tau ψ T = T + c τ for some c ∈ F q ∗ c \in\mathbb{F}_q^* c ∈ F q ∗ .
Summary of results
We follow the notation set up in § 1.1 . From (4) , one can see that the analogue of Serre’s cyclicity problem for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , is concerned with the quantity
π ϕ ( x ) : = # { p ∈ P ϕ : deg p = x , ϕ ( F p ) has at most r − 1 cyclic components } , \pi_\phi(x) := \#\left\{\mathfrak{p} \in\mathcal{P}_\phi: \deg\mathfrak{p} = x,\ \phi(\mathbb{F}_{\mathfrak{p}})\text{ has at most }r-1\text{ cyclic components}\right\}, π ϕ ( x ) := # { p ∈ P ϕ : deg p = x , ϕ ( F p ) has at most r − 1 cyclic components } ,
which for r = 2 r = 2 r = 2 is analogous to (1) . We discuss existing results for this problem in § 2.3 . It is natural to study the density of primes p \mathfrak{p} p lying in a given arithmetic progression, for which ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) has at most r − 1 r-1 r − 1 cyclic components. More precisely, for a monic polynomial t ∈ A t \in A t ∈ A and a ∈ ( A / ( t ) ) ∗ a \in(A/(t))^* a ∈ ( A / ( t ) ) ∗ , consider the quantity
π ϕ ( x ; t , a ) : = # { p ∈ P ϕ : deg p = x , p ≡ a ( m o d t ) , ϕ ( F p ) has at most r − 1 cyclic components } . (5) \pi_\phi(x;t,a) := \#\left\{\mathfrak{p} \in\mathcal{P}_\phi: \deg\mathfrak{p} = x,\ \mathfrak{p} \equiv a \pmod{t},\ \phi(\mathbb{F}_{\mathfrak{p}})\text{ has at most }r-1\text{ cyclic components}\right\}. \tag*{(5)} π ϕ ( x ; t , a ) := # { p ∈ P ϕ : deg p = x , p ≡ a ( mod t ) , ϕ ( F p ) has at most r − 1 cyclic components } . (5)
We may let deg t ≤ x \deg t \le x deg t ≤ x since π ϕ ( x ; t , a ) ≤ 1 \pi_\phi(x;t,a) \le1 π ϕ ( x ; t , a ) ≤ 1 if deg t > x \deg t > x deg t > x .
Cyclicity results: In this paper, we obtain an asymptotic formula for π ϕ ( x ; t , a ) \pi_\phi(x;t,a) π ϕ ( x ; t , a ) which holds uniformly for t t t in the range deg t ≤ x \deg t \le x deg t ≤ x . The exponents appearing in our error terms depend on deg t \deg t deg t , the rank r r r as well as the rank ρ \rho ρ of End K ‾ ( ϕ ) \operatorname{End}_{\overline{K}}(\phi) End K ( ϕ ) as an A A A -module. This is Theorem 1.1 below. We use inclusion-exclusion to express the counting function in terms of Frobenius conditions in the division fields of ϕ \phi ϕ . In particular, we encode the arithmetic progression via a Frobenius condition in the division fields of the Carlitz module, so that we work with Chebotarev conditions in composita of the form L m , t = K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) L_{m,t}=K(\phi[m])K(\psi^{\mathrm{Carl}}[t]) L m , t = K ( ϕ [ m ]) K ( ψ Carl [ t ]) . We then apply the Chebotarev density theorem to obtain the main term and estimate remainder terms to obtain the error term.
The density δ ϕ , t , a ( x ) \delta_{\phi,t,a}(x) δ ϕ , t , a ( x ) in Theorem 1.1 may depend on the degree x x x of the prime polynomials being counted. This dependence arises from the constant fields of the extensions L m , t / K L_{m,t}/K L m , t / K (see Corollary 3.9 and the discussion preceding it). However, it is possible to choose an integer B = B ( ϕ , K ) B=B(\phi,K) B = B ( ϕ , K ) , such that for given t , a t,a t , a , on each arithmetic progression x ≡ j ( m o d B ) x\equiv j\pmod{B} x ≡ j ( mod B ) , the density is constant, although its value may vary between residue classes. This is expressed more precisely in Remark 1.
In Theorem 1.2, we refine this result under the condition End K ( ϕ ) = A \operatorname{End}_{K}(\phi)=A End K ( ϕ ) = A . This is analogous to the non-CM condition for an elliptic curve E / Q E/\mathbb{Q} E / Q , namely End Q ( E ) = Z \operatorname{End}_{\mathbb{Q}}(E)=\mathbb{Z} End Q ( E ) = Z . In this setting, the adelic openness theorem of Pink and Rütsche [31 Theorem 0.1] plays the role of Serre’s open image theorem for non-CM elliptic curves. In particular, the Galois groups of the division fields K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) are isomorphic to GL r ( A / m A ) \operatorname{GL}_{r}(A/mA) GL r ( A / m A ) whenever m m m is coprime to a fixed nonzero polynomial depending on ϕ \phi ϕ (see Lemma 5.1). This stronger information allows us to obtain an asymptotic formula with improved error term in suitable ranges of the modulus in Theorem 1.2.
Positivity results: We next restrict to the case in which the field extension L m , t / K L_{m,t}/K L m , t / K is geometric (equivalently [ L m , t ∩ F ‾ q : F q ] = 1 [L_{m,t}\cap\overline{\mathbb{F}}_{q}:\mathbb{F}_{q}]=1 [ L m , t ∩ F q : F q ] = 1 ), for every monic m ∈ A m\in A m ∈ A . The density δ ϕ , t , a \delta_{\phi,t,a} δ ϕ , t , a is then independent of x x x , and we study conditions under which δ ϕ , t , a \delta_{\phi,t,a} δ ϕ , t , a is positive. It is at this point that an important distinction between the elliptic curve and Drinfeld module settings arises due to the dual role played by cyclotomic fields in the positivity problem for elliptic curves. First, the condition p ≡ a ( m o d f ) p\equiv a\pmod{f} p ≡ a ( mod f ) is encoded by the Frobenius action on Q ( ζ f ) \mathbb{Q}(\zeta_f) Q ( ζ f ) . Second, one has the containment Q ( ζ f ) ⊆ Q ( E [ f ] ) \mathbb{Q}(\zeta_f)\subseteq\mathbb{Q}(E[f]) Q ( ζ f ) ⊆ Q ( E [ f ]) for every f f f . These inclusions provide the auxiliary subfields and suitable disjointness results in the positivity arguments of [9 ] , [1 ] .
In the Drinfeld setting, these two roles need not be fulfilled by the same family of fields. The canonical rank 1 Drinfeld modules whose division fields give an explicit class field theory are precisely those that are sign-normalized. The arithmetic progression condition is thus encoded by the division field K ( ψ C a r l [ t ] ) K(\psi^{\mathrm{Carl}}[t]) K ( ψ Carl [ t ]) of the Carlitz module ψ C a r l \psi^{\mathrm{Carl}} ψ Carl (one may replace ψ C a r l \psi^{\mathrm{Carl}} ψ Carl by any chosen rank 1 sign-normalized Drinfeld module). However, K ( ψ C a r l [ t ] ) K(\psi^{\mathrm{Carl}}[t]) K ( ψ Carl [ t ]) need not be contained in K ( ϕ [ t ] ) K(\phi[t]) K ( ϕ [ t ]) . The analogue of the Weil pairing instead gives K ( ϕ ′ [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^{\prime}[m])\subseteq K(\phi[m]) K ( ϕ ′ [ m ]) ⊆ K ( ϕ [ m ]) , where ϕ ′ \phi^{\prime} ϕ ′ is the rank 1 Drinfeld module associated to ϕ \phi ϕ (see (1.16)), which need not be sign-normalized and need not satisfy ϕ ′ ≅ K ψ C a r l \phi^{\prime}\cong_{K}\psi^{\mathrm{Carl}} ϕ ′ ≅ K ψ Carl . In particular, when ϕ ′ ̸ ≅ K ψ C a r l \phi^{\prime}\not\cong_{K}\psi^{\mathrm{Carl}} ϕ ′ ≅ K ψ Carl , the division fields encoding the arithmetic progression cannot be identified with the ones naturally contained in the division fields of ϕ \phi ϕ , and thus the two roles of the cyclotomic fields in the elliptic case are now played by distinct entities. Indeed, one may still obtain positivity of δ ϕ , t , a \delta_{\phi,t,a} δ ϕ , t , a in such a scenario, as demonstrated by Example 1.7.
As a consequence, methods of [9 ] , [1 ] , [22 ] for positivity of density – which construct auxiliary fields using containment and disjointness relations between division fields – do not transfer to our setting. In this paper, we give a Galois-theoretic argument inspired by Milner and Shotton [27 ] , which characterizes the positivity problem in terms of the existence of a compatible automorphism in a finite Galois extension. This is Theorem 1.4 below.
As noted in [27 ] , the argument of [9 ] requires a correction, since pairwise linear disjointness of the auxiliary fields does not, in general, imply their joint linear disjointness. Our proof of Theorem 1.4 explicitly accounts for entanglements among the relevant division fields and gives a necessary and sufficient criterion for positivity of the density. When there is no arithmetic progression condition, it yields Corollary 1.5. This addresses the positivity problem considered in [22 ] and gives an alternative criterion for positivity, as compared to [22 Theorem 3] .
Finally, if ϕ ′ ≅ K ψ C a r l \phi^{\prime}\cong_{K}\psi^{\mathrm{Carl}} ϕ ′ ≅ K ψ Carl , the positivity criterion simplifies to yield Corollary 1.6.
Statement of results
In order to state our results, we need the following notation. Let π ( x ) \pi(x) π ( x ) denote the number of prime polynomials in A A A of degree x x x , μ q \mu_q μ q denote the Möbius function on $A, and φ ( m ) = ∣ ( A / m A ) ∗ ∣ \varphi(\mathfrak{m})=|(A/\mathfrak{m}A)^*| φ ( m ) = ∣ ( A / m A ) ∗ ∣ . Let t ∈ A \mathfrak{t}\in A t ∈ A be monic and let a ∈ ( A / ( t ) ) ∗ \mathfrak{a}\in(A/(\mathfrak{t}))^* a ∈ ( A / ( t ) ) ∗ . For a monic m ∈ A \mathfrak{m}\in A m ∈ A , put
L m , t : = K ( φ [ m ] ) K ( ψ C a r l [ t ] ) , r m , t : = [ L m , t ∩ F ‾ q : F q ] . (6) L_{\mathfrak{m},\mathfrak{t}}:=K(\varphi[\mathfrak{m}])K(\psi^{\mathrm{Carl}}[\mathfrak{t}]),\qquad r_{\mathfrak{m},\mathfrak{t}}:=[L_{\mathfrak{m},\mathfrak{t}}\cap\overline{\mathbb{F}}_q:\mathbb{F}_q]. \tag*{(6)} L m , t := K ( φ [ m ]) K ( ψ Carl [ t ]) , r m , t := [ L m , t ∩ F q : F q ] . (6)
For Galois extensions L / K L/K L / K , define
γ a , t ( L ) : = { 1 , ψ a C a r l fixes L ∩ K ( ψ C a r l [ t ] ) , 0 , otherwise . (7) \gamma_{\mathfrak{a},\mathfrak{t}}(L):=
\begin{cases}
1,&\psi^{\mathrm{Carl}}_{\mathfrak{a}}\text{ fixes }L\cap K(\psi^{\mathrm{Carl}}[\mathfrak{t}]),\\
0,&\text{otherwise}.
\end{cases} \tag*{(7)} γ a , t ( L ) := { 1 , 0 , ψ a Carl fixes L ∩ K ( ψ Carl [ t ]) , otherwise . (7)
The condition γ a , t ( K ( φ [ m ] ) ) = 1 \gamma_{\mathfrak{a},\mathfrak{t}}(K(\varphi[\mathfrak{m}]))=1 γ a , t ( K ( φ [ m ])) = 1 detects precisely when there is a unique automorphism ν m , t ∈ Gal ( L m , t / K ) \nu_{\mathfrak{m},\mathfrak{t}}\in\operatorname{Gal}(L_{\mathfrak{m},\mathfrak{t}}/K) ν m , t ∈ Gal ( L m , t / K ) , which fixes K ( φ [ m ] ) K(\varphi[\mathfrak{m}]) K ( φ [ m ]) and which acts as ψ a C a r l \psi^{\mathrm{Carl}}_{\mathfrak{a}} ψ a Carl on λ t \lambda_{\mathfrak{t}} λ t , where λ t \lambda_{\mathfrak{t}} λ t generates the A / t A A/\mathfrak{t}A A / t A -module ψ C a r l [ t ] \psi^{\mathrm{Carl}}[\mathfrak{t}] ψ Carl [ t ] . Moreover, there exists an integer 0 ≤ ϵ m , t < r m , t 0\leq\epsilon_{\mathfrak{m},\mathfrak{t}}<r_{\mathfrak{m},\mathfrak{t}} 0 ≤ ϵ m , t < r m , t such that the restriction of ν m , t \nu_{\mathfrak{m},\mathfrak{t}} ν m , t to the constant field L m , t ∩ F ‾ q L_{\mathfrak{m},\mathfrak{t}}\cap\overline{\mathbb{F}}_q L m , t ∩ F q is the ϵ m , t \epsilon_{\mathfrak{m},\mathfrak{t}} ϵ m , t -th power of the Frobenius τ \tau τ , that is,
ν m , t ∣ L m , t ∩ F ‾ q = τ ϵ m , t . (8) \left.\nu_{\mathfrak{m},\mathfrak{t}}\right|_{L_{\mathfrak{m},\mathfrak{t}}\cap\overline{\mathbb{F}}_q}=\tau^{\epsilon_{\mathfrak{m},\mathfrak{t}}}. \tag*{(8)} ν m , t ∣ L m , t ∩ F q = τ ϵ m , t . (8)
Whenever γ a , t ( K ( φ [ m ] ) ) = 0 \gamma_{\mathfrak{a},\mathfrak{t}}(K(\varphi[\mathfrak{m}]))=0 γ a , t ( K ( φ [ m ])) = 0 , we set ϵ m , t = 0 \epsilon_{\mathfrak{m},\mathfrak{t}}=0 ϵ m , t = 0 . Also, for d = p 1 e 1 ⋯ p k e k \mathfrak{d}=\mathfrak{p}_1^{e_1}\cdots\mathfrak{p}_k^{e_k} d = p 1 e 1 ⋯ p k e k , we define d 0 ( d ) : = p 1 ⌈ e 1 / r ⌉ ⋯ p k ⌈ e k / r ⌉ \mathfrak{d}_0(\mathfrak{d}):=\mathfrak{p}_1^{\lceil e_1/r\rceil}\cdots\mathfrak{p}_k^{\lceil e_k/r\rceil} d 0 ( d ) := p 1 ⌈ e 1 / r ⌉ ⋯ p k ⌈ e k / r ⌉ . Then, for a monic t ∈ A \mathfrak{t}\in A t ∈ A , set
J φ ( t ) : = ∑ d ∣ t ′ φ ( d ) q − deg d 0 , α φ ( t ) : = deg t − log q J φ ( t ) . (9) J_\varphi(\mathfrak{t}):=\sum_{\mathfrak{d}\mid\mathfrak{t}}'\varphi(\mathfrak{d})q^{-\deg\mathfrak{d}_0},\qquad\alpha_\varphi(\mathfrak{t}):=\deg\mathfrak{t}-\log_q J_\varphi(\mathfrak{t}). \tag*{(9)} J φ ( t ) := d ∣ t ∑ ′ φ ( d ) q − d e g d 0 , α φ ( t ) := deg t − log q J φ ( t ) . (9)
Furthermore, put θ r , ρ : = 3 r 2 − ( ρ + 2 ) r − ρ / ( 2 r 2 ) \theta_{r,\rho}:=3r^2-(\rho+2)r-\rho/(2r^2) θ r , ρ := 3 r 2 − ( ρ + 2 ) r − ρ / ( 2 r 2 ) , and define
β r : = r − 4 2 r , and Ξ r , ρ ( x , t ) : = θ r , ρ x + ρ ( r + 1 ) r 2 log q ( x + deg t log x ) . (10) \beta_r:=\frac{r-4}{2r},\qquad\text{and}\qquad\Xi_{r,\rho}(x,\mathfrak{t}):=\theta_{r,\rho}x+\frac{\rho(r+1)}{r^2}\log_q\left(\frac{x+\deg\mathfrak{t}}{\log x}\right). \tag*{(10)} β r := 2 r r − 4 , and Ξ r , ρ ( x , t ) := θ r , ρ x + r 2 ρ ( r + 1 ) log q ( log x x + deg t ) . (10)
Let End K ‾ ( φ ) \operatorname{End}_{\overline{K}}(\varphi) End K ( φ ) denote the endomorphism ring of the Drinfeld module φ \varphi φ over the algebraic closure K ‾ \overline{K} K of K K K . Let ρ \rho ρ denote the rank of End K ‾ ( φ ) \operatorname{End}_{\overline{K}}(\varphi) End K ( φ ) as an A A A -module.
For functions f f f and g g g , with g ≥ 0 g\geq0 g ≥ 0 , the notation f ≪ φ , K g f\ll_{\varphi,K}g f ≪ φ , K g or f = O φ , K ( g ) f=O_{\varphi,K}(g) f = O φ , K ( g ) means that ∣ f ∣ ≤ C g |f|\leq Cg ∣ f ∣ ≤ C g for some constant C > 0 C>0 C > 0 depending only on φ \varphi φ and K K K . Throughout the paper, let ∑ ′ \sum' ∑ ′ denote a sum over monics in A A A .
Our first main result is as follows.
Theorem 1.1. Let t ∈ A \mathfrak{t}\in A t ∈ A be monic and let a ∈ ( A / ( t ) ) ∗ \mathfrak{a}\in(A/(\mathfrak{t}))^* a ∈ ( A / ( t ) ) ∗ be arbitrary. Let φ \varphi φ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r\geq2 r ≥ 2 , and let ρ : = rank A End K ‾ ( φ ) \rho:=\operatorname{rank}_A\operatorname{End}_{\overline{K}}(\varphi) ρ := rank A End K ( φ ) . For monics m ∈ A \mathfrak{m}\in A m ∈ A , define
r m , t ( x ) : = { r m , t , if x ≡ ϵ m , t ( m o d r m , t ) , 0 , otherwise , (11) r_{\mathfrak{m},\mathfrak{t}}(x):=
\begin{cases}
r_{\mathfrak{m},\mathfrak{t}},&\text{if }x\equiv\epsilon_{\mathfrak{m},\mathfrak{t}}\pmod{r_{\mathfrak{m},\mathfrak{t}}},\\
0,&\text{otherwise},
\end{cases} \tag*{(11)} r m , t ( x ) := { r m , t , 0 , if x ≡ ϵ m , t ( mod r m , t ) , otherwise , (11) where r m , t r_{\mathfrak{m},\mathfrak{t}} r m , t , ϵ m , t \epsilon_{\mathfrak{m},\mathfrak{t}} ϵ m , t are as defined in (1.6), (1.8) respectively. Let γ a , t \gamma_{\mathfrak{a},\mathfrak{t}} γ a , t be as defined in (1.7) and define the density
δ φ , t , a ( x ) : = ∑ m ∈ A ′ μ q ( m ) r m , t ( x ) γ a , t ( K ( φ [ m ] ) ) [ K ( φ [ m ] ) K ( ψ C a r l [ t ] ) : K ] . (12) \delta_{\varphi,\mathfrak{t},\mathfrak{a}}(x):=\sum_{\mathfrak{m}\in A}'\frac{\mu_q(\mathfrak{m})r_{\mathfrak{m},\mathfrak{t}}(x)\gamma_{\mathfrak{a},\mathfrak{t}}(K(\varphi[\mathfrak{m}]))}{[K(\varphi[\mathfrak{m}])K(\psi^{\mathrm{Carl}}[\mathfrak{t}]):K]}. \tag*{(12)} δ φ , t , a ( x ) := m ∈ A ∑ ′ [ K ( φ [ m ]) K ( ψ Carl [ t ]) : K ] μ q ( m ) r m , t ( x ) γ a , t ( K ( φ [ m ])) . (12) where μ q \mu_q μ q denotes the Möbius function on A A A . Then, for integers x ≥ max { 2 , deg t } x\geq\max\{2,\deg\mathfrak{t}\} x ≥ max { 2 , deg t } ,
π φ ( x ; t , a ) = δ φ , t , a ( x ) π ( x ) + E φ , K ( x , t ) , \pi_\varphi(x;\mathfrak{t},\mathfrak{a})=\delta_{\varphi,\mathfrak{t},\mathfrak{a}}(x)\pi(x)+E_{\varphi,K}(x,\mathfrak{t}), π φ ( x ; t , a ) = δ φ , t , a ( x ) π ( x ) + E φ , K ( x , t ) , where
E φ , K ( x , t ) = { O φ , K ( q x 2 + x r ( 1 + deg t x ) ) , if α φ ( t ) ≤ β r x , O K ( x r q r − 1 r x ) + O φ , K ( q r 2 + 4 r − 2 2 r ( r + 1 ) x − α φ ( t ) r + 1 ( 1 + deg t x ) r r + 1 ) , if β r x < α φ ( t ) ≤ Ξ r , ρ ( x , t ) , O K ( x r q r − 1 r x ) + O φ , K ( q r 2 + ρ 2 r 2 x ( log x x + deg t ) ρ r 2 ( 1 + deg t x ) ) , if α φ ( t ) ≥ Ξ r , ρ ( x , t ) . E_{\varphi,K}(x,\mathfrak{t})=
\begin{cases}
O_{\varphi,K}\left(q^{\frac{x}{2}+\frac{x}{r}}\left(1+\frac{\deg\mathfrak{t}}{x}\right)\right),&\text{if }\alpha_\varphi(\mathfrak{t})\leq\beta_r x,\\
O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\varphi,K}\left(q^{\frac{r^2+4r-2}{2r(r+1)}x-\frac{\alpha_\varphi(\mathfrak{t})}{r+1}}\left(1+\frac{\deg\mathfrak{t}}{x}\right)^{\frac{r}{r+1}}\right),&\text{if }\beta_r x<\alpha_\varphi(\mathfrak{t})\leq\Xi_{r,\rho}(x,\mathfrak{t}),\\
O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\varphi,K}\left(q^{\frac{r^2+\rho}{2r^2}x}\left(\frac{\log x}{x+\deg\mathfrak{t}}\right)^{\frac{\rho}{r^2}}\left(1+\frac{\deg\mathfrak{t}}{x}\right)\right),&\text{if }\alpha_\varphi(\mathfrak{t})\geq\Xi_{r,\rho}(x,\mathfrak{t}).
\end{cases} E φ , K ( x , t ) = ⎩ ⎨ ⎧ O φ , K ( q 2 x + r x ( 1 + x d e g t ) ) , O K ( r x q r r − 1 x ) + O φ , K ( q 2 r ( r + 1 ) r 2 + 4 r − 2 x − r + 1 α φ ( t ) ( 1 + x d e g t ) r + 1 r ) , O K ( r x q r r − 1 x ) + O φ , K ( q 2 r 2 r 2 + ρ x ( x + d e g t l o g x ) r 2 ρ ( 1 + x d e g t ) ) , if α φ ( t ) ≤ β r x , if β r x < α φ ( t ) ≤ Ξ r , ρ ( x , t ) , if α φ ( t ) ≥ Ξ r , ρ ( x , t ) . where α φ ( t ) \alpha_\varphi(\mathfrak{t}) α φ ( t ) is as defined in (1.9), and β r \beta_r β r and Ξ r , ρ ( x , t ) \Xi_{r,\rho}(x,\mathfrak{t}) Ξ r , ρ ( x , t ) are as defined in (1.10).
When End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A , the next result improves the error estimate of Theorem 1.1 for large α ϕ ( t ) \alpha_\phi(t) α ϕ ( t ) . In particular, it extends the intermediate-range bound to all α ϕ ( t ) > β r x \alpha_\phi(t)>\beta_r x α ϕ ( t ) > β r x , allowing the error term to decrease further as α ϕ ( t ) \alpha_\phi(t) α ϕ ( t ) increases. For instance, when r = 2 r=2 r = 2 and t t t is irreducible of degree 3 x / 4 3x/4 3 x /4 , the exponent of q q q in the error bound decreases from 5 x / 8 5x/8 5 x /8 to 7 x / 12 7x/12 7 x /12 .
Theorem 1.2. Let t ∈ A t\in A t ∈ A be monic and let a ∈ ( A / ( t ) ) ∗ a\in(A/(t))^* a ∈ ( A / ( t ) ) ∗ be arbitrary. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r\ge2 r ≥ 2 such that End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A . Let δ ϕ , t , a ( x ) \delta_{\phi,t,a}(x) δ ϕ , t , a ( x ) be as defined in (12) . Then, for integers x ≥ max { 2 , deg t } x\ge\max\{2,\deg t\} x ≥ max { 2 , deg t } ,
π ϕ ( x ; t , a ) = δ ϕ , t , a ( x ) π ( x ) + E ϕ , K ( x , t ) , \pi_\phi(x;t,a)=\delta_{\phi,t,a}(x)\pi(x)+E_{\phi,K}(x,t), π ϕ ( x ; t , a ) = δ ϕ , t , a ( x ) π ( x ) + E ϕ , K ( x , t ) , where
E ϕ , K ( x , t ) = { O ϕ , K ( q x 2 + x r ( 1 + deg t x ) ) , if α ϕ ( t ) ≤ β r x , O K ( x r q r − 1 r x ) + O ϕ , K ( q r 2 + 4 r − 2 2 r ( r + 1 ) x − α ϕ ( t ) r + 1 ( 1 + deg t x ) r r + 1 ) , if α ϕ ( t ) > β r x , E_{\phi,K}(x,t)=
\begin{cases}
O_{\phi,K}\left(q^{\frac{x}{2}+\frac{x}{r}}\left(1+\frac{\deg t}{x}\right)\right), & \text{if }\alpha_\phi(t)\le\beta_r x,\\
O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{r^2+4r-2}{2r(r+1)}x-\frac{\alpha_\phi(t)}{r+1}}\left(1+\frac{\deg t}{x}\right)^{\frac{r}{r+1}}\right), & \text{if }\alpha_\phi(t)>\beta_r x,
\end{cases} E ϕ , K ( x , t ) = ⎩ ⎨ ⎧ O ϕ , K ( q 2 x + r x ( 1 + x d e g t ) ) , O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r ( r + 1 ) r 2 + 4 r − 2 x − r + 1 α ϕ ( t ) ( 1 + x d e g t ) r + 1 r ) , if α ϕ ( t ) ≤ β r x , if α ϕ ( t ) > β r x , where α ϕ ( t ) \alpha_\phi(t) α ϕ ( t ) is as defined in (9) and β r \beta_r β r is as defined in (10) .
Although our emphasis is on uniformity in deg t \deg t deg t rather than optimal error terms, we remark here that Theorem 1.2 recovers [22 Theorem 1] for K = F q ( T ) K=\mathbb{F}_q(T) K = F q ( T ) upon taking t = 1 t=1 t = 1 . Indeed, then a = 1 a=1 a = 1 , α ϕ ( 1 ) = 0 = ϵ m , 1 \alpha_\phi(1)=0=\epsilon_{m,1} α ϕ ( 1 ) = 0 = ϵ m , 1 and γ 1 , 1 ( K ( ϕ [ m ] ) ) = 1 \gamma_{1,1}(K(\phi[m]))=1 γ 1 , 1 ( K ( ϕ [ m ])) = 1 , so δ ϕ , 1 , 1 ( x ) = ∑ m ∈ A ′ μ q ( m ) r m ( x ) / [ K ( ϕ [ m ] ) : K ] \delta_{\phi,1,1}(x)=\sum_{m\in A}^{\prime}\mu_q(m)r_m(x)/[K(\phi[m]):K] δ ϕ , 1 , 1 ( x ) = ∑ m ∈ A ′ μ q ( m ) r m ( x ) / [ K ( ϕ [ m ]) : K ] , where r m ( x ) = r m , 1 1 r m , 1 ∣ x r_m(x)=r_{m,1}\mathbf{1}_{r_{m,1}\mid x} r m ( x ) = r m , 1 1 r m , 1 ∣ x and r m , 1 = [ K ( ϕ [ m ] ) ∩ F q : F q ] r_{m,1}=[K(\phi[m])\cap\mathbb{F}_q:\mathbb{F}_q] r m , 1 = [ K ( ϕ [ m ]) ∩ F q : F q ] . Moreover, E ϕ , K ( x , 1 ) ≪ ϕ , K q Δ r x E_{\phi,K}(x,1)\ll_{\phi,K}q^{\Delta_r x} E ϕ , K ( x , 1 ) ≪ ϕ , K q Δ r x , with Δ r \Delta_r Δ r matching the error exponent in [22 ] (see (18) ).
We next study when the density δ ϕ , t , a \delta_{\phi,t,a} δ ϕ , t , a is positive. Our next result is a corollary to Theorem 1.1 .
Corollary 1.3. Let t ∈ A t\in A t ∈ A be monic and let a ∈ ( A / ( t ) ) ∗ a\in(A/(t))^* a ∈ ( A / ( t ) ) ∗ be arbitrary. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r\ge2 r ≥ 2 such that r m , t = 1 r_{m,t}=1 r m , t = 1 for every monic m ∈ A m\in A m ∈ A . If K ( ϕ [ p ] ) = K K(\phi[\mathfrak p])=K K ( ϕ [ p ]) = K for some prime p ∈ A \mathfrak p\in A p ∈ A with deg p = 1 \deg\mathfrak p=1 deg p = 1 , then δ ϕ , t , a = 0 \delta_{\phi,t,a}=0 δ ϕ , t , a = 0 .
Note that when we put t = 1 t=1 t = 1 , Corollary 1.3 yields one direction of Theorem 3 of Kuo-Liu [22 ] .
We now show that assuming End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A , the positivity of δ ϕ , t , a \delta_{\phi,t,a} δ ϕ , t , a is equivalent to a condition on division fields. Under the assumption End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A , by a result of Pink and Rütsche [31 ] , there exists a nonzero monic polynomial M ( ϕ ) ∈ A M(\phi)\in A M ( ϕ ) ∈ A (see Lemma 5.1 ) such that the division fields K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , for m ∈ A m\in A m ∈ A coprime to M ( ϕ ) M(\phi) M ( ϕ ) , have full Galois image. Define N ( ϕ ) : = lcm ( M ( ϕ ) , ∏ deg p = 1 p ) N(\phi):=\operatorname{lcm}(M(\phi),\prod_{\deg\mathfrak p=1}\mathfrak p) N ( ϕ ) := lcm ( M ( ϕ ) , ∏ d e g p = 1 p ) . Let B \mathcal{B} B be the set of all primes in A A A dividing N ( ϕ ) N(\phi) N ( ϕ ) and G \mathcal{G} G be the set of all primes in A A A coprime to N ( ϕ ) N(\phi) N ( ϕ ) . Also define, for a finite set S ⊆ G \mathcal{S}\subseteq\mathcal{G} S ⊆ G ,
D B : = ∏ p ∈ B K ( ϕ [ p ] ) , L ϕ , t , S : = D B K ( ψ C a r l [ t ] ) . D_{\mathcal{B}}:=\prod_{\mathfrak p\in\mathcal{B}}K(\phi[\mathfrak p]),\qquad L_{\phi,t,\mathcal{S}}:=D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[t]). D B := p ∈ B ∏ K ( ϕ [ p ]) , L ϕ , t , S := D B K ( ψ Carl [ t ]) .
where the product denotes compositum of fields. We call a finite set S ⊆ G S \subseteq G S ⊆ G an entanglement set for ( ϕ , t ) (\phi,t) ( ϕ , t ) if, for every square-free monic m \mathfrak{m} m supported on G ∖ S G \setminus S G ∖ S , K ( ϕ [ m ] ) ∩ D B ∪ S K ( ψ C a r l [ t ] ) = K K(\phi[\mathfrak{m}]) \cap D_{\mathbb{B}\cup S}K(\psi^{\mathrm{Carl}}[t]) = K K ( ϕ [ m ]) ∩ D B ∪ S K ( ψ Carl [ t ]) = K . By Lemma 6.2, such a set S S S exists. We prove the following positivity criterion.
Theorem 1.4. Let t ∈ A t \in A t ∈ A be monic and let a ∈ ( A / ( t ) ) ∗ a \in(A/(t))^{*} a ∈ ( A / ( t ) ) ∗ be arbitrary. Let λ t \lambda_t λ t be a generator of ψ C a r l [ t ] \psi^{\mathrm{Carl}}[t] ψ Carl [ t ] , so that K ( ψ C a r l [ t ] ) = K ( λ t ) K(\psi^{\mathrm{Carl}}[t]) = K(\lambda_t) K ( ψ Carl [ t ]) = K ( λ t ) . Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 such that r m , t = 1 r_{\mathfrak{m},t} = 1 r m , t = 1 for every monic m ∈ A \mathfrak{m} \in A m ∈ A , and End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi) = A End K ( ϕ ) = A . Let S S S be an entanglement set for ( ϕ , t ) (\phi,t) ( ϕ , t ) , and define
Ω ϕ , t , a , S : = { σ ∈ Gal ( L ϕ , t , S / K ) : σ ∣ K ( ψ C a r l [ t ] ) ( λ t ) = ψ a C a r l ( λ t ) , σ ∣ K ( ϕ [ p ] ) ≠ 1 for every p ∈ B ∪ S } . (14) \Omega_{\phi,t,a,S}:=\left\{\sigma\in\operatorname{Gal}(L_{\phi,t,S}/K):\left.\sigma\right|_{K(\psi^{\mathrm{Carl}}[t])}(\lambda_t)=\psi_a^{\mathrm{Carl}}(\lambda_t),\left.\sigma\right|_{K(\phi[\mathfrak{p}])}\ne1\text{ for every }\mathfrak{p}\in\mathbb{B}\cup S\right\}. \tag*{(14)} Ω ϕ , t , a , S := { σ ∈ Gal ( L ϕ , t , S / K ) : σ ∣ K ( ψ Carl [ t ]) ( λ t ) = ψ a Carl ( λ t ) , σ ∣ K ( ϕ [ p ]) = 1 for every p ∈ B ∪ S } . (14) Then
δ ϕ , t , a = ∣ Ω ϕ , t , a , S ∣ [ L ϕ , t , S : K ] ∏ p ∉ B ∪ S ( 1 − 1 [ K ( ϕ [ p ] ) : K ] ) . (15) \delta_{\phi,t,a}=\frac{|\Omega_{\phi,t,a,S}|}{[L_{\phi,t,S}:K]}\prod_{\mathfrak{p}\notin\mathbb{B}\cup S}\left(1-\frac{1}{[K(\phi[\mathfrak{p}]):K]}\right). \tag*{(15)} δ ϕ , t , a = [ L ϕ , t , S : K ] ∣ Ω ϕ , t , a , S ∣ p ∈ / B ∪ S ∏ ( 1 − [ K ( ϕ [ p ]) : K ] 1 ) . (15) In particular, δ ϕ , t , a > 0 \delta_{\phi,t,a}>0 δ ϕ , t , a > 0 if and only if Ω ϕ , t , a , S \Omega_{\phi,t,a,S} Ω ϕ , t , a , S is nonempty.
In the case t = 1 t=1 t = 1 , let us denote the density (1.13) by δ ϕ \delta_{\phi} δ ϕ . We also have K ( ψ C a r l [ 1 ] ) = K K(\psi^{\mathrm{Carl}}[1])=K K ( ψ Carl [ 1 ]) = K , and L ϕ , t , S = L ϕ , S = ∏ p ∈ B ∪ S K ( ϕ [ p ] ) L_{\phi,t,S}=L_{\phi,S}=\prod_{\mathfrak{p}\in\mathbb{B}\cup S}K(\phi[\mathfrak{p}]) L ϕ , t , S = L ϕ , S = ∏ p ∈ B ∪ S K ( ϕ [ p ]) , so that the set Ω ϕ , t , a , S \Omega_{\phi,t,a,S} Ω ϕ , t , a , S reduces to
Ω ϕ , S = { σ ∈ Gal ( L ϕ , S / K ) : σ ∣ K ( ϕ [ p ] ) ≠ 1 for every p ∈ B ∪ S } . \Omega_{\phi,S}=\left\{\sigma\in\operatorname{Gal}(L_{\phi,S}/K):\left.\sigma\right|_{K(\phi[\mathfrak{p}])}\ne1\text{ for every }\mathfrak{p}\in\mathbb{B}\cup S\right\}. Ω ϕ , S = { σ ∈ Gal ( L ϕ , S / K ) : σ ∣ K ( ϕ [ p ]) = 1 for every p ∈ B ∪ S } .
We thus obtain the following.
Corollary 1.5. If ϕ \phi ϕ is a Drinfeld A A A -module over K K K of rank r ≥ 2 r\ge2 r ≥ 2 with End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A and all division fields of ϕ \phi ϕ are geometric, then δ ϕ > 0 ⟺ Ω ϕ , S ≠ ∅ \delta_{\phi}>0\Longleftrightarrow\Omega_{\phi,S}\ne\varnothing δ ϕ > 0 ⟺ Ω ϕ , S = ∅ .
For a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , given by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T=T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 Drinfeld A A A -module over K K K associated to ϕ \phi ϕ via the Weil pairing of Tate modules (see [36 ] , [30 Section 3.7] ) is given by
ϕ T 1 : = T + ( − 1 ) r − 1 a r τ . (16) \phi_T^1:=T+(-1)^{r-1}a_r\tau. \tag*{(16)} ϕ T 1 := T + ( − 1 ) r − 1 a r τ . (16)
As a corollary to Theorem 1.4, we have the following result.
Corollary 1.6. Let q > 2 q>2 q > 2 . Let t ∈ A t\in A t ∈ A be monic and let a ∈ ( A / ( t ) ) ∗ a\in(A/(t))^{*} a ∈ ( A / ( t ) ) ∗ be arbitrary. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r\ge2 r ≥ 2 such that End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A and r m , t = 1 r_{\mathfrak{m},t}=1 r m , t = 1 for every monic m ∈ A \mathfrak{m}\in A m ∈ A . Assume further that gcd ( t , N ( ϕ ) ) = 1 \gcd(t,N(\phi))=1 g cd( t , N ( ϕ )) = 1 and ϕ 1 ≃ K ψ C a r l \phi^1\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl . Then the density δ ϕ , t , a > 0 \delta_{\phi,t,a}>0 δ ϕ , t , a > 0 .
The following example shows that ϕ 1 ≃ K ψ C a r l \phi^1\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl is not a necessary condition for positivity of δ ϕ , t , a \delta_{\phi,t,a} δ ϕ , t , a .
Example 1.7. Let q = 5 q=5 q = 5 , and let ϕ \phi ϕ be the rank-2 Drinfeld A A A -module over K K K defined by
ϕ T = T − T τ − T 10 τ 2 . \phi_T=T-T\tau-T^{10}\tau^2. ϕ T = T − T τ − T 10 τ 2 . Then, the associated rank 1 Drinfeld module ϕ T 1 = T + T 10 τ \phi_T^1=T+T^{10}\tau ϕ T 1 = T + T 10 τ is not sign-normalized and ϕ 1 ̸ ≃ K ψ C a r l \phi^1\not\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl . It satisfies End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A and r m , T = 1 r_{\mathfrak{m},T}=1 r m , T = 1 for every monic m ∈ A \mathfrak{m}\in A m ∈ A . Moreover, for every a ∈ ( A / ( T ) ) ∗ a\in(A/(T))^{*} a ∈ ( A / ( T ) ) ∗ , δ ϕ , T , a > 0 \delta_{\phi,T,a}>0 δ ϕ , T , a > 0 .
The paper is organized as follows. Section 2 gives an account of the surrounding literature on the cyclicity problem for elliptic curves and Drinfeld modules. In Section 3, we recall the relation between the number of cyclic components of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) and complete splitting in the division fields of ϕ \phi ϕ . We also translate the arithmetic progression condition into a Frobenius condition in a Carlitz division field, recall the Chebotarev density theorem for function fields, and establish the uniform estimates required in the counting argument. Section 4 is devoted to the proof of Theorem 1.1. In Section 5, we consider the case when End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A and obtain improved error estimates in Theorem 1.2. Finally in Section 6, we study positivity of the density and give proofs of Corollary 1.3, Theorem 1.4, Corollary 1.6 and Example 1.7.
Surrounding literature
Cyclicity of elliptic curves
In 1975, Borosh, Moreno and Porta [4 ] conjectured that, for ’many’ elliptic curves E / Q E/\mathbb{Q} E / Q , the group E ( F p ) E(\mathbb{F}_p) E ( F p ) is cyclic for infinitely many primes p p p . Subsequently, in 1977, Lang and Trotter [24 ] formulated an elliptic curve analogue of Artin’s primitive root conjecture. The cyclicity conjecture is a subproblem of this, and concerns finding an asymptotic density estimate for π E ( x ) \pi_E(x) π E ( x ) (see (1) ). In 1978, assuming GRH for the Dedekind zeta functions of the division fields of E / Q E/\mathbb{Q} E / Q , Serre [33 ] established the asymptotic formula
π E ( x ) = δ E li ( x ) + o ( x / log x ) , (17) \pi_E(x)=\delta_E\operatorname{li}(x)+o(x/\log x), \tag*{(17)} π E ( x ) = δ E li ( x ) + o ( x / log x ) , (17)
where δ E \delta_E δ E is as in (1.2). In 1983, Murty [28 ] refined Serre’s argument, obtaining an improved error term under GRH. He also proved the asymptotic formula unconditionally for elliptic curves with complex multiplication (CM). Subsequently, in 1987, Murty [29 ] proved unconditionally the infinitude of cyclic reductions for certain non-CM elliptic curves, using the infinitude of supersingular primes. In 1990, Gupta and Murty [18 ] established unconditionally that for an arbitrary elliptic curve E / Q E/\mathbb{Q} E / Q satisfying Q ( E [ 2 ] ) ≠ Q \mathbb{Q}(E[2])\ne\mathbb{Q} Q ( E [ 2 ]) = Q , we have π E ( x ) ≫ E x / ( log x ) 2 \pi_E(x)\gg_E x/(\log x)^2 π E ( x ) ≫ E x / ( log x ) 2 .
Subsequent improvements by relaxing the assumption of GRH for non-CM elliptic curve and obtaining better error terms for CM elliptic curves were obtained by Cojocaru in [7 ] and [8 ] respectively. Cojocaru and Murty [9 ] improved the error term in (2.1) to O N ( E ) ( x 5 / 6 ( log x ) 2 / 3 ) O_{N(E)}(x^{5/6}(\log x)^{2/3}) O N ( E ) ( x 5/6 ( log x ) 2/3 ) under GRH for a non-CM elliptic curve E / Q E/\mathbb{Q} E / Q ; and to O ( ( x 3 / 4 ( log ( N ( E ) x ) ) 1 / 2 ) O((x^3/4(\log(N(E)x))^{1/2}) O (( x 3 /4 ( log ( N ( E ) x ) ) 1/2 ) if E E E has complex multiplication by the full ring of integers of an imaginary quadratic field. Further assuming Artin’s Holomorphy Conjecture and a Pair Correlation Conjecture for the associated Artin L L L -functions, they further improved the exponent 5 / 6 5/6 5/6 to 7 / 10 7/10 7/10 , up to a logarithmic factor. Their work also gives bounds for the least prime of cyclic reduction in terms of N ( E ) N(E) N ( E ) . Later results include average cyclicity estimates by Banks and Shparlinski [3 ] in 2009 and a Siegel–Walfisz-type theorem for CM elliptic curves by Akbary and V. K. Murty [2 ] in 2010.
The positivity of the density δ E \delta_E δ E was first studied by Cojocaru and Murty in [9 ] , following an observation of Serre. In 2017, Brau [5 ] obtained nonvanishing criteria for δ E \delta_E δ E using character sums for entanglement correction factors. More generally, in 2023, Campagna and Stevenhagen [6 ] studied the density of cyclic reductions for elliptic curves over number fields.
Cyclicity of elliptic curves in arithmetic progressions
Let f ≥ 1 f\ge1 f ≥ 1 and a ∈ ( Z / f Z ) ∗ a\in(\mathbb{Z}/f\mathbb{Z})^* a ∈ ( Z / f Z ) ∗ . In 2017, Brau [5 ] studied cyclicity density constants in arithmetic progressions using character sums, obtaining explicit formulas and nonvanishing results. Subsequently, Akbal and Güloğlu [1 ] investigated the counting function π E ( x ; f , a ) \pi_E(x;f,a) π E ( x ; f , a ) defined in (1.3).
Under GRH for the relevant Dedekind zeta functions, Akbal and Güloğlu [1 ] established
π E ( x ; f , a ) = δ E ( f , a ) li ( x ) + E E , f , a ( x ) , \pi_E(x;f,a)=\delta_E(f,a)\operatorname{li}(x)+E_{E,f,a}(x), π E ( x ; f , a ) = δ E ( f , a ) li ( x ) + E E , f , a ( x ) ,
with the density δ E ( f , a ) \delta_E(f,a) δ E ( f , a ) and bounds for errors E E , f , a ( x ) E_{E,f,a}(x) E E , f , a ( x ) explicitly given, for both non-CM elliptic curves and CM elliptic curves with multiplication by the full ring of integers of an imaginary quadratic field. They also obtained sufficient conditions for the positivity of δ E ( f , a ) \delta_E(f,a) δ E ( f , a ) .
In 2024, Wong [38 ] obtained an unconditional asymptotic formula for CM elliptic curves in arithmetic progressions. Under GRH, he also refined the estimates of Akbal and Güloğlu, obtaining logarithmic savings for suitable ranges of f f f . More recently, Lee, Mayle and Wang [26 ] and Lee [25 ] investigated average aspects of cyclicity in arithmetic progressions, while Jones and Lee [21 ] studied the vanishing of the associated densities.
Cyclicity of Drinfeld modules
In 2009, Kuo and Liu [22 ] studied the analogue of Serre’s cyclicity problem for Drinfeld A A A -modules over finite extensions of K K K . Let F / K F/K F / K be a finite extension, and let ϕ \phi ϕ be a Drinfeld A A A -module over F F F of rank r ≥ 2 r\ge2 r ≥ 2 . Let F F \mathbb{F}_F F F denote the constant field of F F F , and for monic m ∈ A \mathfrak{m}\in A m ∈ A , define r m : = [ F ( ϕ [ m ] ) ∩ F ‾ q : F ] r_{\mathfrak{m}}:=[F(\phi[\mathfrak{m}])\cap\overline{\mathbb{F}}_q:F] r m := [ F ( ϕ [ m ]) ∩ F q : F ] . Kuo and Liu defined the density δ ϕ , F ( x ) : = ∑ m ∈ A μ q ( m ) r m ( x ) / [ F ( ϕ [ m ] ) : F ] \delta_{\phi,F}(x) := \sum_{\mathfrak m \in A} \mu_q(\mathfrak m)r_{\mathfrak m}(x)/[F(\phi[\mathfrak m]):F] δ ϕ , F ( x ) := ∑ m ∈ A μ q ( m ) r m ( x ) / [ F ( ϕ [ m ]) : F ] , where r m ( x ) = r m r_{\mathfrak m}(x)=r_{\mathfrak m} r m ( x ) = r m whenever r m ∣ x r_{\mathfrak m}\mid x r m ∣ x , and zero otherwise. Let π F ( x ) \pi_F(x) π F ( x ) denote the number of primes of F F F of degree x x x , and let π ϕ , F ( x ) \pi_{\phi,F}(x) π ϕ , F ( x ) denote the number of primes of good reduction of degree x x x for which ϕ ( F p ) \phi(\mathbb F_{\mathfrak p}) ϕ ( F p ) has at most r − 1 r-1 r − 1 cyclic components. Assuming End F ( ϕ ) = A \operatorname{End}_F(\phi)=A End F ( ϕ ) = A , they established π ϕ , F ( x ) = δ ϕ , F ( x ) π F ( x ) + O ϕ , F ( q F Δ r x ) \pi_{\phi,F}(x)=\delta_{\phi,F}(x)\pi_F(x)+O_{\phi,F}(q_F^{\Delta_r x}) π ϕ , F ( x ) = δ ϕ , F ( x ) π F ( x ) + O ϕ , F ( q F Δ r x ) , where q F = ∣ F F ∣ q_F=|F_F| q F = ∣ F F ∣ and
Δ r : = { r 2 + 4 r − 2 2 r 2 + 2 r , r = 2 , 3 , r + 2 2 r , r ≥ 4. (18) \Delta_r := \begin{cases}
\frac{r^2+4r-2}{2r^2+2r}, & r=2,3,\\
\frac{r+2}{2r}, & r\ge4.
\end{cases} \tag*{(18)} Δ r := { 2 r 2 + 2 r r 2 + 4 r − 2 , 2 r r + 2 , r = 2 , 3 , r ≥ 4. (18)
They also investigated the positivity of δ ϕ , F ( x ) \delta_{\phi,F}(x) δ ϕ , F ( x ) , under the hypotheses that r m = 1 r_{\mathfrak m}=1 r m = 1 for every m ∈ A \mathfrak m\in A m ∈ A .
In 2013, Cojocaru and Shulman [10 ] established an unconditional average effective Chebotarev density theorem for the division fields of rank-2 Drinfeld modules ϕ \phi ϕ over K = F q ( T ) K=\mathbb F_q(T) K = F q ( T ) , with q q q odd, such that End K ( ϕ ) \operatorname{End}_K(\phi) End K ( ϕ ) is the full ring of integers of an imaginary quadratic extension of K K K . As an application, they obtained an asymptotic formula for cyclic reductions, with an error term of O ϕ ( q 3 x / 4 log x ) O_{\phi}(q^{3x/4}\log x) O ϕ ( q 3 x /4 log x ) . Subsequently, in 2015, Cojocaru and Shulman [11 ] studied the distribution of the first elementary divisor d 1 , p d_{1,\mathfrak p} d 1 , p (see (4) ) of the reductions of generic Drinfeld modules of arbitrary rank. For rank 2, they further studied the average growth of the second elementary divisor d 2 , p d_{2,\mathfrak p} d 2 , p .
In 2017, Virdol [37 ] generalized the result in [22 ] to Drinfeld modules with more general endomorphism rings, obtained better error terms and established an analogue of the Titchmarsh divisor problem for Drinfeld modules. In 2021, Kuo and Tweedle [23 ] studied the full cyclicity problem for Drinfeld modules of arbitrary rank. They established the existence of a density for the primes p \mathfrak p p for which ϕ ( F p ) \phi(\mathbb F_{\mathfrak p}) ϕ ( F p ) is cyclic. They further proved that this density is positive if and only if ϕ ( F p ) \phi(\mathbb F_{\mathfrak p}) ϕ ( F p ) is cyclic for infinitely many primes p \mathfrak p p .
Notation and Preliminaries
Basic notation
We have the following notation
π ( x ) \pi(x) π ( x ) : the number of primes in A A A of degree x x x ;
F q \mathbb F_q F q : the field of characteristic p p p having q q q elements;
A = F q [ T ] A=\mathbb F_q[T] A = F q [ T ] : the polynomial ring over F q \mathbb F_q F q in the variable T T T ;
μ q \mu_q μ q : the Möbius function on F q [ T ] \mathbb F_q[T] F q [ T ] ;
K = F q ( T ) K=\mathbb F_q(T) K = F q ( T ) : the field of fractions of A A A ;
∞ \infty ∞ : the prime at infinity in K K K ;
K ∞ = F q ( ( 1 / T ) ) K_\infty=\mathbb F_q((1/T)) K ∞ = F q (( 1/ T )) : the completion of K K K at ∞ \infty ∞ ;
C ∞ C_\infty C ∞ : the completion of K ∞ K_\infty K ∞ at ∞ \infty ∞ .
Characteristic polynomials of Frobenius
We will require some properties of the characteristic polynomial of the Frobenius map on the Tate module of ϕ \phi ϕ . For a prime l ∈ A \mathfrak l\in A l ∈ A , consider the l \mathfrak l l -adic completion of A A A , denoted by A l : = lim ← A / ( l n ) A_{\mathfrak l}:=\lim_{\leftarrow} A/(\mathfrak l^n) A l := lim ← A / ( l n ) . Let K l : = frac ( A l ) K_{\mathfrak l}:=\operatorname{frac}(A_{\mathfrak l}) K l := frac ( A l ) be the fraction field of A l A_{\mathfrak l} A l . For any Drinfeld module ϕ \phi ϕ of rank r r r , the l \mathfrak l l -adic Tate module T l ( ϕ ) T_{\mathfrak l}(\phi) T l ( ϕ ) is defined as T l ( ϕ ) : = lim ← ϕ [ l n ] T_{\mathfrak l}(\phi):=\lim_{\leftarrow}\phi[\mathfrak l^n] T l ( ϕ ) := lim ← ϕ [ l n ] , where the inverse limit is taken with respect to the morphisms ϕ [ l n + 1 ] → ϕ [ l n ] \phi[\mathfrak l^{n+1}]\to\phi[\mathfrak l^n] ϕ [ l n + 1 ] → ϕ [ l n ] , given by α ↦ ϕ l ( α ) \alpha\mapsto\phi_{\mathfrak l}(\alpha) α ↦ ϕ l ( α ) . Then T l ( ϕ ) ≃ A l r T_{\mathfrak l}(\phi)\simeq A_{\mathfrak l}^r T l ( ϕ ) ≃ A l r .
Let p ∈ P ϕ \mathfrak p\in\mathcal P_\phi p ∈ P ϕ be coprime to l \mathfrak l l and let ϕ ˉ \bar\phi ϕ ˉ denote the reduction mod p \mathfrak p p of our Drinfeld module ϕ \phi ϕ . Since ( p , l ) = 1 (\mathfrak p,\mathfrak l)=1 ( p , l ) = 1 , we have that the natural reduction mod p \mathfrak p p induces the isomorphism of free A l A_{\mathfrak l} A l -modules A l r ≃ T l ( ϕ ) ≃ T l ( ϕ ˉ ) A_{\mathfrak l}^r\simeq T_{\mathfrak l}(\phi)\simeq T_{\mathfrak l}(\bar\phi) A l r ≃ T l ( ϕ ) ≃ T l ( ϕ ˉ ) . Therefore, the Frobenius map η p \eta_{\mathfrak p} η p at p \mathfrak p p given by x ↦ x ∣ A / p ∣ x\mapsto x^{|A/\mathfrak p|} x ↦ x ∣ A / p ∣ , induces an endomorphism on T l ( ϕ ˉ ) T_{\mathfrak l}(\bar\phi) T l ( ϕ ˉ ) and hence on T l ( ϕ ) T_{\mathfrak l}(\phi) T l ( ϕ ) via the above isomorphism. The resulting induced action of η p \eta_{\mathfrak p} η p on the K l K_{\mathfrak l} K l -vector space T l ( ϕ ) ⊗ A l K l T_{\mathfrak l}(\phi)\otimes_{A_{\mathfrak l}}K_{\mathfrak l} T l ( ϕ ) ⊗ A l K l is considered to be the action of the Frobenius at p \mathfrak p p on T l ( ϕ ) ⊗ K l T_{\mathfrak l}(\phi)\otimes K_{\mathfrak l} T l ( ϕ ) ⊗ K l .
Let P ϕ , p l ( X ) P_{\phi,\mathfrak p}^{\mathfrak l}(X) P ϕ , p l ( X ) denote the characteristic polynomial of η p \eta_{\mathfrak p} η p , defined as
P ϕ , p l ( X ) : = det ( X Id − η p ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + a p , 0 . P_{\phi,\mathfrak p}^{\mathfrak l}(X):=\det(X\operatorname{Id}-\eta_{\mathfrak p})=X^r+a_{\mathfrak p,r-1}X^{r-1}+\cdots+a_{\mathfrak p,1}X+a_{\mathfrak p,0}. P ϕ , p l ( X ) := det ( X Id − η p ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + a p , 0 .
We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0}=u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i=0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A=d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T=T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (19) P_{\phi,\mathfrak{p}}(X)=X^r+a_{\mathfrak{p},r-1}X^{r-1}+\cdots+a_{\mathfrak{p},1}X+u_{\mathfrak{p}}\mathfrak{p},\qquad P_{\phi^1,\mathfrak{p}}(X)=X+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(19)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (19) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A\setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}}:=a_{\mathfrak{p},1}+\cdots+a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1)=1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1)=1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA}\cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m\in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m]\cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m]\cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m])=\prod_{\mathfrak{l}\mid m}K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0}=u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i=0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A=d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T=T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (20) P_{\phi,\mathfrak{p}}(X)=X^r+a_{\mathfrak{p},r-1}X^{r-1}+\cdots+a_{\mathfrak{p},1}X+u_{\mathfrak{p}}\mathfrak{p},\qquad P_{\phi^1,\mathfrak{p}}(X)=X+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(20)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (20) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A\setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}}:=a_{\mathfrak{p},1}+\cdots+a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1)=1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1)=1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA}\cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m\in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m]\cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m]\cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m])=\prod_{\mathfrak{l}\mid m}K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0}=u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i=0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A=d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T=T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (21) P_{\phi,\mathfrak{p}}(X)=X^r+a_{\mathfrak{p},r-1}X^{r-1}+\cdots+a_{\mathfrak{p},1}X+u_{\mathfrak{p}}\mathfrak{p},\qquad P_{\phi^1,\mathfrak{p}}(X)=X+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(21)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (21) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A\setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}}:=a_{\mathfrak{p},1}+\cdots+a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1)=1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1)=1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA}\cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m\in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m]\cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m]\cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m])=\prod_{\mathfrak{l}\mid m}K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0}=u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i=0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A=d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T=T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (22) P_{\phi,\mathfrak{p}}(X)=X^r+a_{\mathfrak{p},r-1}X^{r-1}+\cdots+a_{\mathfrak{p},1}X+u_{\mathfrak{p}}\mathfrak{p},\qquad P_{\phi^1,\mathfrak{p}}(X)=X+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(22)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (22) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A\setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}}:=a_{\mathfrak{p},1}+\cdots+a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1)=1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1)=1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA}\cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m\in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m]\cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m]\cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m])=\prod_{\mathfrak{l}\mid m}K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0}=u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i=0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A=d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T=T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (23) P_{\phi,\mathfrak{p}}(X)=X^r+a_{\mathfrak{p},r-1}X^{r-1}+\cdots+a_{\mathfrak{p},1}X+u_{\mathfrak{p}}\mathfrak{p},\qquad P_{\phi^1,\mathfrak{p}}(X)=X+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(23)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (23) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A\setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}}:=a_{\mathfrak{p},1}+\cdots+a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1)=1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1)=1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA}\cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m\in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m]\cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m]\cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m])=\prod_{\mathfrak{l}\mid m}K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0}=u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i=0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A=d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T=T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (24) P_{\phi,\mathfrak{p}}(X)=X^r+a_{\mathfrak{p},r-1}X^{r-1}+\cdots+a_{\mathfrak{p},1}X+u_{\mathfrak{p}}\mathfrak{p},\qquad P_{\phi^1,\mathfrak{p}}(X)=X+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(24)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (24) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A\setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}}:=a_{\mathfrak{p},1}+\cdots+a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1)=1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1)=1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in m a t h o r d K ( ϕ [ l ] ) mathord{K(\phi[\mathfrak{l}])} ma t h or d K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA}\cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m \in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m]\cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m]\cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m])=\prod_{\mathfrak{l}\mid m}K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0}=u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i = 0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A = d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T = T+a_1\tau+\cdots+a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (25) P_{\phi,\mathfrak{p}}(X) = X^r + a_{\mathfrak{p},r-1}X^{r-1} + \cdots+ a_{\mathfrak{p},1}X + u_{\mathfrak{p}}\mathfrak{p}, \qquad P_{\phi^1,\mathfrak{p}}(X) = X + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(25)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (25) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A \setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1 + b_{\mathfrak{p}} + u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1 + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}} := a_{\mathfrak{p},1} + \cdots+ a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1) = 1 + b_{\mathfrak{p}} + u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1) = 1 + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA} \cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m \in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m] \cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m] \cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m]) = \prod_{\mathfrak{l}\mid m} K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0} = u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i=0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A = d_{1,\mathfrak{p}}d_{2,\mathfrak{p}} \cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T = T + a_1\tau+ \cdots+ a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (26) P_{\phi,\mathfrak{p}}(X) = X^r + a_{\mathfrak{p},r-1}X^{r-1} + \cdots+ a_{\mathfrak{p},1}X + u_{\mathfrak{p}}\mathfrak{p}, \qquad P_{\phi^1,\mathfrak{p}}(X) = X + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(26)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (26) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A \setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1 + b_{\mathfrak{p}} + u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1 + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}} := a_{\mathfrak{p},1} + \cdots+ a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1) = 1 + b_{\mathfrak{p}} + u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1) = 1 + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in m a t h o r d K ( ϕ [ l ] ) mathord{K(\phi[\mathfrak{l}])} ma t h or d K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA} \cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m \in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m] \cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m] \cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m]) = \prod_{\mathfrak{l}\mid m} K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime @page 9/34 starts_mid=0 ends_mid=1 We state some of the properties of this polynomial, which will be beneficial in obtaining our estimation. We will recall the following properties as stated in [10 Proposition 8] .
Lemma 3.1. (see [16, Corollary 3.4, p.193; Theorem 5.1, p.199]) Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 1 r \ge1 r ≥ 1 . Let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and l \mathfrak{l} l be primes in K K K such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then,
(i) P ϕ , p l ( X ) ∈ A [ X ] P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) \in A[X] P ϕ , p l ( X ) ∈ A [ X ] ; in particular, P ϕ , p l ( X ) P^{\mathfrak{l}}_{\phi,\mathfrak{p}}(X) P ϕ , p l ( X ) is independent of l \mathfrak{l} l , and as such, we will drop the superscript l \mathfrak{l} l from notation and simply write P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) .
(ii) a p , 0 = u p p a_{\mathfrak{p},0} = u_{\mathfrak{p}}\mathfrak{p} a p , 0 = u p p for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ .
(iii) deg a p , i ≤ ( r − i ) deg p r \deg a_{\mathfrak{p},i} \le\frac{(r-i)\deg\mathfrak{p}}{r} deg a p , i ≤ r ( r − i ) d e g p , for i = 0 , … , r − 1 i = 0,\ldots,r-1 i = 0 , … , r − 1 .
(iv) P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A P_{\phi,\mathfrak{p}}(1)A = d_{1,\mathfrak{p}}d_{2,\mathfrak{p}}\cdots d_{r,\mathfrak{p}}A P ϕ , p ( 1 ) A = d 1 , p d 2 , p ⋯ d r , p A , where d i , p d_{i,\mathfrak{p}} d i , p are as defined in (4) .
For a positive integer k k k , the cyclotomic field Q ( ζ k ) \mathbb{Q}(\zeta_k) Q ( ζ k ) is a subset of the k k k -division field Q ( E [ k ] ) \mathbb{Q}(E[k]) Q ( E [ k ]) . Recall that for a Drinfeld A A A -module ϕ \phi ϕ over K K K of rank r r r , defined by ϕ T = T + a 1 τ + ⋯ + a r τ r \phi_T = T + a_1\tau+ \cdots+ a_r\tau^r ϕ T = T + a 1 τ + ⋯ + a r τ r , the rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , associated via Weil pairing, is given by (16) . Then, we have that K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[m]) \subseteq K(\phi[m]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) , for every m ∈ A m \in A m ∈ A (see [30 Theorem 3.7.1] ). With this in place, we state the following result of [10 ] .
Proposition 3.2. [10 Proposition 10] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 . There exists a rank 1 1 1 Drinfeld A A A -module ϕ 1 \phi^1 ϕ 1 over K K K , uniquely determined up to K ‾ \overline{K} K -isomorphism, such that:
(i) P ϕ ⊆ P ϕ 1 \mathcal{P}_{\phi} \subseteq\mathcal{P}_{\phi^1} P ϕ ⊆ P ϕ 1 ;
(ii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , the characteristic polynomials of ϕ \phi ϕ and ϕ 1 \phi^1 ϕ 1 at p \mathfrak{p} p satisfy the relation:
P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (27) P_{\phi,\mathfrak{p}}(X) = X^r + a_{\mathfrak{p},r-1}X^{r-1} + \cdots+ a_{\mathfrak{p},1}X + u_{\mathfrak{p}}\mathfrak{p}, \qquad P_{\phi^1,\mathfrak{p}}(X) = X + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p}, \tag*{(27)} P ϕ , p ( X ) = X r + a p , r − 1 X r − 1 + ⋯ + a p , 1 X + u p p , P ϕ 1 , p ( X ) = X + ( − 1 ) r − 1 u p p , (27) where, we recall, u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ ;
(iii) for any p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ and any nonzero m ∈ A m \in A m ∈ A coprime to p \mathfrak{p} p , we have that, if p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then
(a) p \mathfrak{p} p also splits completely in K ( ϕ 1 [ m ] ) K(\phi^1[m]) K ( ϕ 1 [ m ]) ;
(b) m r ∣ P ϕ , p ( 1 ) m^r \mid P_{\phi,\mathfrak{p}}(1) m r ∣ P ϕ , p ( 1 ) ;
(c) m ∣ P ϕ 1 , p ( 1 ) m \mid P_{\phi^1,\mathfrak{p}}(1) m ∣ P ϕ 1 , p ( 1 ) .
Corollary 3.3. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 and let m ∈ A ∖ { 0 } m \in A \setminus\{0\} m ∈ A ∖ { 0 } . If p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ is coprime to m m m , and p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , then there exist u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p , satisfying m r ∣ 1 + b p + u p p m^r \mid1 + b_{\mathfrak{p}} + u_{\mathfrak{p}}\mathfrak{p} m r ∣ 1 + b p + u p p and m ∣ 1 + ( − 1 ) r − 1 u p p m \mid1 + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ 1 + ( − 1 ) r − 1 u p p .
Proof. Consider the expressions (3.1) for the characteristic polynomials P ϕ , p ( X ) P_{\phi,\mathfrak{p}}(X) P ϕ , p ( X ) and P ϕ 1 , p ( X ) P_{\phi^1,\mathfrak{p}}(X) P ϕ 1 , p ( X ) . We define b p : = a p , 1 + ⋯ + a p , r − 1 b_{\mathfrak{p}} := a_{\mathfrak{p},1} + \cdots+ a_{\mathfrak{p},r-1} b p := a p , 1 + ⋯ + a p , r − 1 . Then, by Lemma 3.1(iii), deg b p ≤ ( r − 1 ) deg p r \deg b_{\mathfrak{p}} \le\frac{(r-1)\deg\mathfrak{p}}{r} deg b p ≤ r ( r − 1 ) d e g p . By Proposition 3.2(iii), if p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) , we have m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p m^r \mid P_{\phi,\mathfrak{p}}(1) = 1 + b_{\mathfrak{p}} + u_{\mathfrak{p}}\mathfrak{p} m r ∣ P ϕ , p ( 1 ) = 1 + b p + u p p and m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p m \mid P_{\phi^1,\mathfrak{p}}(1) = 1 + (-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} m ∣ P ϕ 1 , p ( 1 ) = 1 + ( − 1 ) r − 1 u p p , as required. □
Artin symbols and the Chebotarev Density Theorem
The following lemma of [22 ] relates the A A A -module structure of ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) to the splitting of primes in certain division fields.
Lemma 3.4. [22 Proposition 11] Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r \ge2 r ≥ 2 , let p ∈ P ϕ \mathfrak{p} \in\mathcal{P}_{\phi} p ∈ P ϕ , and let l ∈ A \mathfrak{l} \in A l ∈ A be a prime such that l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p . Then, the following are equivalent.
(1) ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r .
(2) The prime p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) .
Moreover, the A A A -module ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains at least r r r cyclic components if and only if the equivalent conditions above hold for some prime l ≠ p \mathfrak{l} \ne\mathfrak{p} l = p .
By the Chinese Remainder Theorem, we have an A A A -module isomorphism A m A ≅ ⨁ l ∣ m A l A \frac{A}{mA} \cong\bigoplus_{\mathfrak{l}\mid m}\frac{A}{\mathfrak{l}A} m A A ≅ ⨁ l ∣ m l A A , where m m m is a square-free monic in A A A . Since for every nonzero m ∈ A m \in A m ∈ A , ϕ [ m ] ≅ ( A / m A ) r \phi[m] \cong(A/mA)^r ϕ [ m ] ≅ ( A / m A ) r , we obtain ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] \phi[m] \cong\bigoplus_{\mathfrak{l}\mid m}\phi[\mathfrak{l}] ϕ [ m ] ≅ ⨁ l ∣ m ϕ [ l ] . This implies that we have the compositum K ( ϕ [ m ] ) = ∏ l ∣ m K ( ϕ [ l ] ) K(\phi[m]) = \prod_{\mathfrak{l}\mid m}K(\phi[\mathfrak{l}]) K ( ϕ [ m ]) = ∏ l ∣ m K ( ϕ [ l ]) . Therefore, a prime p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) if and only if it splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) for every prime l ∣ m \mathfrak{l}\mid\mathfrak{m} l ∣ m . Thus, if ( m , p ) = 1 (\mathfrak{m},\mathfrak{p})=1 ( m , p ) = 1 , then by Lemma 3.4 and the Chinese Remainder Theorem, p \mathfrak{p} p splits completely in K ( ϕ [ m ] ) K(\phi[\mathfrak{m}]) K ( ϕ [ m ]) if and only if ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / m A ) r (A/\mathfrak{m}A)^r ( A / m A ) r .
Using this, one obtains the following expression for π ϕ ( x ; t , a ) \pi_\phi(x;\mathfrak{t},\mathfrak{a}) π ϕ ( x ; t , a ) .
Lemma 3.5. Let t ∈ A \mathfrak{t}\in A t ∈ A be a monic polynomial and a ∈ ( A / ( t ) ) ∗ \mathfrak{a}\in(A/(\mathfrak{t}))^* a ∈ ( A / ( t ) ) ∗ be arbitrary. For x ≥ 2 x\ge2 x ≥ 2 , we have that
π ϕ ( x ; t , a ) = ∑ deg m ≤ x / r ′ μ q ( m ) π ϕ , m ( x ; t , a ) , (28) \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\sum_{\deg\mathfrak{m}\le x/r}'\mu_q(\mathfrak{m})\pi_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}), \tag*{(28)} π ϕ ( x ; t , a ) = d e g m ≤ x / r ∑ ′ μ q ( m ) π ϕ , m ( x ; t , a ) , (28) where
π ϕ , m ( x ; t , a ) : = # { p ∈ P ϕ : deg p = x , p ≡ a ( m o d t ) , p splits completely in K ( ϕ [ m ] ) } . (29) \pi_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}):=\#\{\mathfrak{p}\in\mathcal{P}_\phi:\deg\mathfrak{p}=x,\ \mathfrak{p}\equiv\mathfrak{a}\pmod{\mathfrak{t}},\ \mathfrak{p}\text{ splits completely in }K(\phi[\mathfrak{m}])\}. \tag*{(29)} π ϕ , m ( x ; t , a ) := # { p ∈ P ϕ : deg p = x , p ≡ a ( mod t ) , p splits completely in K ( ϕ [ m ])} . (29) Proof. By (1.5) and Lemma 3.4,
π ϕ ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( m o d t ) , p does not split completely in K ( ϕ [ l ] ) for all primes l ≠ p } . \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\#\{\mathfrak{p}\in\mathcal{P}_\phi:\deg\mathfrak{p}=x,\ \mathfrak{p}\equiv\mathfrak{a}\pmod{\mathfrak{t}},\ \mathfrak{p}\text{ does not split completely in }K(\phi[\mathfrak{l}])\text{ for all primes }\mathfrak{l}\ne\mathfrak{p}\}. π ϕ ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( mod t ) , p does not split completely in K ( ϕ [ l ]) for all primes l = p } . If p \mathfrak{p} p splits completely in K ( ϕ [ l ] ) K(\phi[\mathfrak{l}]) K ( ϕ [ l ]) , then by Lemma 3.4, ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / l A ) r (A/\mathfrak{l}A)^r ( A / l A ) r . Since # ϕ ( F p ) = # F p = q x \#\phi(\mathbb{F}_{\mathfrak{p}})=\#\mathbb{F}_{\mathfrak{p}}=q^x # ϕ ( F p ) = # F p = q x and # ( A / l A ) r = q r deg l \#(A/\mathfrak{l}A)^r=q^{r\deg\mathfrak{l}} # ( A / l A ) r = q r d e g l , it then follows that q r deg l ≤ q x q^{r\deg\mathfrak{l}}\le q^x q r d e g l ≤ q x , and hence deg l ≤ x / r \deg\mathfrak{l}\le x/r deg l ≤ x / r . In particular, since r ≥ 2 r\ge2 r ≥ 2 and deg p = x \deg\mathfrak{p}=x deg p = x , we must have l ≠ p \mathfrak{l}\ne\mathfrak{p} l = p . Hence,
π ϕ ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( m o d t ) , p does not split completely in K ( ϕ [ l ] ) for all primes l with deg l ≤ x / r } . \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\#\{\mathfrak{p}\in\mathcal{P}_\phi:\deg\mathfrak{p}=x,\ \mathfrak{p}\equiv\mathfrak{a}\pmod{\mathfrak{t}},\ \mathfrak{p}\text{ does not split completely in }K(\phi[\mathfrak{l}])\text{ for all primes }\mathfrak{l}\text{ with }\deg\mathfrak{l}\le x/r\}. π ϕ ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( mod t ) , p does not split completely in K ( ϕ [ l ]) for all primes l with deg l ≤ x / r } . For monics m ∈ A \mathfrak{m}\in A m ∈ A , define the set
S ϕ , m ( x ; t , a ) : = { p ∈ P ϕ : deg p = x , p ≡ a ( m o d t ) , p splits completely in K ( ϕ [ m ] ) } . S_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}):=\{\mathfrak{p}\in\mathcal{P}_\phi:\deg\mathfrak{p}=x,\ \mathfrak{p}\equiv\mathfrak{a}\pmod{\mathfrak{t}},\ \mathfrak{p}\text{ splits completely in }K(\phi[\mathfrak{m}])\}. S ϕ , m ( x ; t , a ) := { p ∈ P ϕ : deg p = x , p ≡ a ( mod t ) , p splits completely in K ( ϕ [ m ])} . Also define X : = { p ∈ P ϕ : deg p = x , p ≡ a ( m o d t ) } = S ϕ , 1 ( x ; t , a ) X:=\{\mathfrak{p}\in\mathcal{P}_\phi:\deg\mathfrak{p}=x,\ \mathfrak{p}\equiv\mathfrak{a}\pmod{\mathfrak{t}}\}=S_{\phi,1}(x;\mathfrak{t},\mathfrak{a}) X := { p ∈ P ϕ : deg p = x , p ≡ a ( mod t )} = S ϕ , 1 ( x ; t , a ) . As a result, we have
π ϕ ( x ; t , a ) = # ⋂ primes l deg l ≤ x / r ( X ∖ S ϕ , l ( x ; t , a ) ) = # ( X ∖ ⋃ primes l deg l ≤ x / r S ϕ , l ( x ; t , a ) ) . \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\#\bigcap_{\substack{\text{primes }\mathfrak{l}\\\deg\mathfrak{l}\le x/r}}\left(X\setminus S_{\phi,\mathfrak{l}}(x;\mathfrak{t},\mathfrak{a})\right)=\#\left(X\setminus\bigcup_{\substack{\text{primes }\mathfrak{l}\\\deg\mathfrak{l}\le x/r}}S_{\phi,\mathfrak{l}}(x;\mathfrak{t},\mathfrak{a})\right). π ϕ ( x ; t , a ) = # primes l d e g l ≤ x / r ⋂ ( X ∖ S ϕ , l ( x ; t , a ) ) = # X ∖ primes l d e g l ≤ x / r ⋃ S ϕ , l ( x ; t , a ) . The set of primes l \mathfrak{l} l with deg l ≤ x / r \deg\mathfrak{l}\le x/r deg l ≤ x / r is finite. Fix a total ordering of the prime polynomials of A A A . Hence, by the inclusion-exclusion principle,
π ϕ ( x ; t , a ) = # X − ∑ k ≥ 1 ( − 1 ) k + 1 ∑ l 1 < l 2 < ⋯ < l k deg l i ≤ x / r # ( ⋂ i = 1 k S ϕ , l i ( x ; t , a ) ) . \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\#X-\sum_{k\ge1}(-1)^{k+1}\sum_{\substack{\mathfrak{l}_1<\mathfrak{l}_2<\cdots<\mathfrak{l}_k\\\deg\mathfrak{l}_i\le x/r}}\#\left(\bigcap_{i=1}^{k}S_{\phi,\mathfrak{l}_i}(x;\mathfrak{t},\mathfrak{a})\right). π ϕ ( x ; t , a ) = # X − k ≥ 1 ∑ ( − 1 ) k + 1 l 1 < l 2 < ⋯ < l k d e g l i ≤ x / r ∑ # ( i = 1 ⋂ k S ϕ , l i ( x ; t , a ) ) . By the discussion preceding this lemma, this simplifies to
π ϕ ( x ; t , a ) = # X − ∑ k = 1 ∞ ( − 1 ) k + 1 ∑ l 1 < l 2 < ⋯ < l k deg l i ≤ x / r # S ϕ , l 1 ⋯ l k ( x ; t , a ) . (30) \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\#X-\sum_{k=1}^{\infty}(-1)^{k+1}\sum_{\substack{\mathfrak{l}_1<\mathfrak{l}_2<\cdots<\mathfrak{l}_k\\\deg\mathfrak{l}_i\le x/r}}\#S_{\phi,\mathfrak{l}_1\cdots\mathfrak{l}_k}(x;\mathfrak{t},\mathfrak{a}). \tag*{(30)} π ϕ ( x ; t , a ) = # X − k = 1 ∑ ∞ ( − 1 ) k + 1 l 1 < l 2 < ⋯ < l k d e g l i ≤ x / r ∑ # S ϕ , l 1 ⋯ l k ( x ; t , a ) . (30) Let m = l 1 ⋯ l k \mathfrak{m}=\mathfrak{l}_1\cdots\mathfrak{l}_k m = l 1 ⋯ l k be a product occurring in (3.4). If S ϕ , m ( x ; t , a ) S_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}) S ϕ , m ( x ; t , a ) is nonempty and p ∈ S ϕ , m ( x ; t , a ) \mathfrak{p}\in S_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}) p ∈ S ϕ , m ( x ; t , a ) , then deg l i ≤ x / r < x = deg p \deg\mathfrak{l}_i\le x/r<x=\deg\mathfrak{p} deg l i ≤ x / r < x = deg p for every i i i . Hence ( m , p ) = 1 (\mathfrak{m},\mathfrak{p})=1 ( m , p ) = 1 . Then, by the discussion preceding the lemma, ϕ ( F p ) \phi(\mathbb{F}_{\mathfrak{p}}) ϕ ( F p ) contains an A A A -submodule isomorphic to ( A / m A ) r (A/\mathfrak{m}A)^r ( A / m A ) r . Thus q r deg m = # ( A / m A ) r ≤ # ϕ ( F p ) = q x q^{r\deg\mathfrak{m}}=\#(A/\mathfrak{m}A)^r\le\#\phi(\mathbb{F}_{\mathfrak{p}})=q^x q r d e g m = # ( A / m A ) r ≤ # ϕ ( F p ) = q x , and hence deg m ≤ x / r \deg\mathfrak{m}\le x/r deg m ≤ x / r . Consequently, (3.4) is equivalent to
π ϕ ( x ; t , a ) = # X − ∑ k ≥ 1 ( − 1 ) k + 1 ∑ m square-free m has k distinct prime divisors deg m ≤ x / r # S ϕ , m ( x ; t , a ) , \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\#X-\sum_{k\ge1}(-1)^{k+1}\sum_{\substack{\mathfrak{m}\text{ square-free}\\\mathfrak{m}\text{ has }k\text{ distinct prime divisors}\\\deg\mathfrak{m}\le x/r}}\#S_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}), π ϕ ( x ; t , a ) = # X − k ≥ 1 ∑ ( − 1 ) k + 1 m square-free m has k distinct prime divisors d e g m ≤ x / r ∑ # S ϕ , m ( x ; t , a ) , or π ϕ ( x ; t , a ) = ∑ deg m ≤ x / r ′ μ q ( m ) # S ϕ , m ( x ; t , a ) \pi_\phi(x;\mathfrak{t},\mathfrak{a})=\sum_{\deg\mathfrak{m}\le x/r}'\mu_q(\mathfrak{m})\#S_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}) π ϕ ( x ; t , a ) = ∑ d e g m ≤ x / r ′ μ q ( m ) # S ϕ , m ( x ; t , a ) . Since # S ϕ , m ( x ; t , a ) = π ϕ , m ( x ; t , a ) \#S_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a})=\pi_{\phi,\mathfrak{m}}(x;\mathfrak{t},\mathfrak{a}) # S ϕ , m ( x ; t , a ) = π ϕ , m ( x ; t , a ) , this completes the proof. □
We now recall the definition of the Artin symbol [cf. [32 ] , pp. 121-122]. Let L / K L/K L / K be a Galois extension and O L \mathcal{O}_L O L denote the ring of integers of L L L . For a prime p ∈ K \mathfrak{p} \in K p ∈ K unramified in L L L and a prime P \mathfrak{P} P in L L L with P ∣ p \mathfrak{P} \mid\mathfrak{p} P ∣ p , there exists a unique element σ P ∈ Gal ( L / K ) \sigma_{\mathfrak{P}} \in\operatorname{Gal}(L/K) σ P ∈ Gal ( L / K ) such that for every α ∈ O L \alpha\in\mathcal{O}_L α ∈ O L ,
σ P ( α ) = α q deg p ( m o d P ) . \sigma_{\mathfrak{P}}(\alpha) = \alpha^{q^{\deg\mathfrak{p}}} \pmod{\mathfrak{P}}. σ P ( α ) = α q d e g p ( mod P ) .
This map is called the Frobenius automorphism at P \mathfrak{P} P . As P \mathfrak{P} P runs over all the primes dividing p \mathfrak{p} p in L L L , the corresponding Frobenius automorphisms form a conjugacy class in Gal ( L / K ) \operatorname{Gal}(L/K) Gal ( L / K ) , called the Artin symbol of p \mathfrak{p} p in L / K L/K L / K and denoted by ( p , L / K ) (\mathfrak{p}, L/K) ( p , L / K ) . When Gal ( L / K ) \operatorname{Gal}(L/K) Gal ( L / K ) is abelian, this conjugacy class has only one element, which we identify with that element.
Let 1 1 1 denote the conjugacy class of the identity element in the Galois group Gal ( L / K ) \operatorname{Gal}(L/K) Gal ( L / K ) . A prime p ∈ K \mathfrak{p} \in K p ∈ K splits completely in L / K L/K L / K if and only if p \mathfrak{p} p is unramified in L / K L/K L / K and the Artin symbol ( p , L / K ) = 1 (\mathfrak{p}, L/K) = 1 ( p , L / K ) = 1 . Therefore, the definition (29) of π ϕ , m ( x ; t , a ) \pi_{\phi,m}(x; t, a) π ϕ , m ( x ; t , a ) can be rewritten as
π ϕ , m ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( m o d t ) , ( p , K ( ϕ [ m ] ) / K ) = 1 } . (31) \pi_{\phi,m}(x; t, a) = \#\{\mathfrak{p} \in\mathcal{P}_{\phi} : \deg\mathfrak{p} = x,\ \mathfrak{p} \equiv a \pmod{t},\ (\mathfrak{p}, K(\phi[m])/K) = 1\}. \tag*{(31)} π ϕ , m ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( mod t ) , ( p , K ( ϕ [ m ]) / K ) = 1 } . (31)
We now translate the condition p ≡ a ( m o d t ) \mathfrak{p} \equiv a \pmod{t} p ≡ a ( mod t ) in the language of Artin symbols, using the Carlitz module ψ C a r l \psi^{\mathrm{Carl}} ψ Carl . Let λ t \lambda_t λ t generate the A / t A A/tA A / t A -module ψ C a r l [ t ] \psi^{\mathrm{Carl}}[t] ψ Carl [ t ] . From Theorem 12.10 of [32 ] , we see that for primes p \mathfrak{p} p coprime to t t t , the Artin symbol ( p , K ( ψ C a r l [ t ] ) / K ) (\mathfrak{p}, K(\psi^{\mathrm{Carl}}[t])/K) ( p , K ( ψ Carl [ t ]) / K ) is the automorphism which takes λ t \lambda_t λ t to ψ p C a r l ( λ t ) \psi_{\mathfrak{p}}^{\mathrm{Carl}}(\lambda_t) ψ p Carl ( λ t ) . For primes counted by π ϕ , m ( x ; t , a ) \pi_{\phi,m}(x; t, a) π ϕ , m ( x ; t , a ) , the condition p ≡ a ( m o d t ) \mathfrak{p} \equiv a \pmod{t} p ≡ a ( mod t ) , with a ∈ ( A / t A ) ∗ a \in(A/tA)^* a ∈ ( A / t A ) ∗ , implies p ∤ t \mathfrak{p} \nmid t p ∤ t .
Lemma 3.6. Let t ∈ A t \in A t ∈ A be a monic polynomial and a ∈ A a \in A a ∈ A be such that a ( m o d t ) a \pmod{t} a ( mod t ) is invertible in A / ( t ) A/(t) A / ( t ) . Let p ∈ A \mathfrak{p} \in A p ∈ A be a prime so that p ∤ t \mathfrak{p} \nmid t p ∤ t . Then the prime p ≡ a ( m o d t ) \mathfrak{p} \equiv a \pmod{t} p ≡ a ( mod t ) if and only if Artin symbol ( p , K ( ψ C a r l [ t ] ) / K ) (\mathfrak{p}, K(\psi^{\mathrm{Carl}}[t])/K) ( p , K ( ψ Carl [ t ]) / K ) acts as ψ a C a r l \psi_a^{\mathrm{Carl}} ψ a Carl on λ t \lambda_t λ t .
Proof. Since a ( m o d t ) ∈ ( A / ( t ) ) ∗ a \pmod{t} \in(A/(t))^* a ( mod t ) ∈ ( A / ( t ) ) ∗ , we know that ( p , K ( ψ C a r l [ t ] ) / K ) (\mathfrak{p}, K(\psi^{\mathrm{Carl}}[t])/K) ( p , K ( ψ Carl [ t ]) / K ) acts as ψ p C a r l \psi_{\mathfrak{p}}^{\mathrm{Carl}} ψ p Carl on λ t \lambda_t λ t . Now, by Theorem 2.3 of [19 ] , there exists an isomorphism
Φ : ( A / t A ) ∗ ⟶ Gal ( K ( ψ C a r l [ t ] ) / K ) \Phi: (A/tA)^* \longrightarrow\operatorname{Gal}(K(\psi^{\mathrm{Carl}}[t])/K) Φ : ( A / t A ) ∗ ⟶ Gal ( K ( ψ Carl [ t ]) / K )
m ⟼ ξ m , m \longmapsto\xi_m, m ⟼ ξ m ,
where ξ m ( λ t ) = ψ m C a r l ( λ t ) \xi_m(\lambda_t) = \psi_m^{\mathrm{Carl}}(\lambda_t) ξ m ( λ t ) = ψ m Carl ( λ t ) . Hence p ≡ a ( m o d t ) \mathfrak{p} \equiv a \pmod{t} p ≡ a ( mod t ) is equivalent to ψ p C a r l ( λ t ) = ψ a C a r l ( λ t ) \psi_{\mathfrak{p}}^{\mathrm{Carl}}(\lambda_t) = \psi_a^{\mathrm{Carl}}(\lambda_t) ψ p Carl ( λ t ) = ψ a Carl ( λ t ) .
As a result, we can modify and write (31) as
π ϕ , m ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , ( p , K ( ψ C a r l [ t ] ) / K ) ( λ t ) = ψ a C a r l ( λ t ) and ( p , K ( ϕ [ m ] ) / K ) = 1 } . \pi_{\phi,m}(x; t, a) = \#\{\mathfrak{p} \in\mathcal{P}_{\phi} : \deg\mathfrak{p} = x,\ (\mathfrak{p}, K(\psi^{\mathrm{Carl}}[t])/K)(\lambda_t) = \psi_a^{\mathrm{Carl}}(\lambda_t)\ \text{and}\ (\mathfrak{p}, K(\phi[m])/K) = 1\}. π ϕ , m ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , ( p , K ( ψ Carl [ t ]) / K ) ( λ t ) = ψ a Carl ( λ t ) and ( p , K ( ϕ [ m ]) / K ) = 1 } .
From Proposition 9.11 of [32 ] , we see that the Artin symbol of the compositum field
σ p = ( p , K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) / K ) \sigma_{\mathfrak{p}} = (\mathfrak{p}, K(\phi[m])K(\psi^{\mathrm{Carl}}[t])/K) σ p = ( p , K ( ϕ [ m ]) K ( ψ Carl [ t ]) / K )
captures the behaviour of the individual Artin maps when restricted to subfields as follows
σ p ∣ K ( ϕ [ m ] ) = ( p , K ( ϕ [ m ] ) / K ) , and σ p ∣ K ( ψ C a r l [ t ] ) = ( p , K ( ψ C a r l [ t ] ) / K ) . \sigma_{\mathfrak{p}}|_{K(\phi[m])} = (\mathfrak{p}, K(\phi[m])/K),\quad\text{and}\quad\sigma_{\mathfrak{p}}|_{K(\psi^{\mathrm{Carl}}[t])} = (\mathfrak{p}, K(\psi^{\mathrm{Carl}}[t])/K). σ p ∣ K ( ϕ [ m ]) = ( p , K ( ϕ [ m ]) / K ) , and σ p ∣ K ( ψ Carl [ t ]) = ( p , K ( ψ Carl [ t ]) / K ) .
Let ν m , t \nu_{m,t} ν m , t be the unique automorphism on K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) K(\phi[m])K(\psi^{\mathrm{Carl}}[t]) K ( ϕ [ m ]) K ( ψ Carl [ t ]) which fixes K ( ϕ [ m ] ) K(\phi[m]) K ( ϕ [ m ]) and acts as ψ a C a r l \psi_a^{\mathrm{Carl}} ψ a Carl on K ( ψ C a r l [ t ] ) K(\psi^{\mathrm{Carl}}[t]) K ( ψ Carl [ t ]) , provided it exists. Let C m , t C_{m,t} C m , t denote the conjugacy class of ν m , t \nu_{m,t} ν m , t in G = Gal ( L m , t / K ) G = \operatorname{Gal}(L_{m,t}/K) G = Gal ( L m , t / K ) . This implies that π ϕ , m ( x ; t , a ) \pi_{\phi,m}(x; t, a) π ϕ , m ( x ; t , a ) can be rewritten as
π ϕ , m ( x ; t , a ) = { # { p ∈ P ϕ : deg p = x , σ p = C m , t } if ν m , t exists , 0 otherwise . \pi_{\phi,m}(x; t, a) =
\begin{cases}
\#\{\mathfrak{p} \in\mathcal{P}_{\phi} : \deg\mathfrak{p} = x,\ \sigma_{\mathfrak{p}} = C_{m,t}\} & \text{if } \nu_{m,t}\text{ exists},\\
0 & \text{otherwise}.
\end{cases} π ϕ , m ( x ; t , a ) = { # { p ∈ P ϕ : deg p = x , σ p = C m , t } 0 if ν m , t exists , otherwise .
Such a ν m , t \nu_{m,t} ν m , t exists only if the intersection field K ( ϕ [ m ] ) ∩ K ( ψ C a r l [ t ] ) K(\phi[m]) \cap K(\psi^{\mathrm{Carl}}[t]) K ( ϕ [ m ]) ∩ K ( ψ Carl [ t ]) is fixed by ψ a C a r l \psi_a^{\mathrm{Carl}} ψ a Carl . Thus, by the definition of γ a , t \gamma_{a,t} γ a , t in (1.7), we may write
π ϕ , m ( x ; t , a ) = { # { p ∈ P ϕ : deg p = x , σ p = C m , t } if γ a , t ( K ( ϕ [ m ] ) ) = 1 , 0 otherwise . \pi_{\phi,m}(x; t, a) =
\begin{cases}
\#\{\mathfrak{p} \in\mathcal{P}_{\phi} : \deg\mathfrak{p} = x,\ \sigma_{\mathfrak{p}} = C_{m,t}\} & \text{if } \gamma_{a,t}(K(\phi[m])) = 1,\\
0 & \text{otherwise}.
\end{cases} π ϕ , m ( x ; t , a ) = { # { p ∈ P ϕ : deg p = x , σ p = C m , t } 0 if γ a , t ( K ( ϕ [ m ])) = 1 , otherwise .
Lemma 3.7. Let C m , t \mathcal{C}_{\mathfrak{m},t} C m , t denote the conjugacy class of ν m , t \nu_{\mathfrak{m},t} ν m , t in G = Gal ( L m , t / K ) G=\operatorname{Gal}(L_{\mathfrak{m},t}/K) G = Gal ( L m , t / K ) . Then C m , t = { ν m , t } \mathcal{C}_{\mathfrak{m},t}=\{\nu_{\mathfrak{m},t}\} C m , t = { ν m , t } .
Proof. Let ν m , t ′ \nu'_{\mathfrak{m},t} ν m , t ′ be a conjugate of ν m , t \nu_{\mathfrak{m},t} ν m , t in G G G . Since ν m , t \nu_{\mathfrak{m},t} ν m , t acts as identity on K ( ϕ [ m ] ) K(\phi[\mathfrak{m}]) K ( ϕ [ m ]) , so will ν m , t ′ \nu'_{\mathfrak{m},t} ν m , t ′ . Moreover, as K ( ψ Carl [ t ] ) / K K(\psi^{\operatorname{Carl}}[t])/K K ( ψ Carl [ t ]) / K is an abelian extension, ν m , t \nu_{\mathfrak{m},t} ν m , t restricted to K ( ψ Carl [ t ] ) K(\psi^{\operatorname{Carl}}[t]) K ( ψ Carl [ t ]) has only one conjugate automorphism. Thus, ν m , t ′ ( λ t ) = ν m , t ( λ t ) = ψ a Carl ( λ t ) \nu'_{\mathfrak{m},t}(\lambda_t)=\nu_{\mathfrak{m},t}(\lambda_t)=\psi_{\mathfrak{a}}^{\operatorname{Carl}}(\lambda_t) ν m , t ′ ( λ t ) = ν m , t ( λ t ) = ψ a Carl ( λ t ) . Hence, ν m , t ′ = ν m , t \nu'_{\mathfrak{m},t}=\nu_{\mathfrak{m},t} ν m , t ′ = ν m , t on L m , t L_{\mathfrak{m},t} L m , t .
This motivates invoking the Chebotarev density theorem over function fields. For a Galois extension L / K L/K L / K , let F ‾ L \overline{\mathbb{F}}_L F L denote the constant field L ∩ F ‾ q L\cap\overline{\mathbb{F}}_q L ∩ F q and r L = [ F ‾ L : F q ] r_L=[\overline{\mathbb{F}}_L:\mathbb{F}_q] r L = [ F L : F q ] . Also, recall that τ : c ↦ c q \tau:c\mapsto c^q τ : c ↦ c q denotes the Frobenius automorphism. For x ∈ N x\in\mathbb{N} x ∈ N , we define
π C ( x , L / K ) : = # { p ∈ A : deg p = x , p unramified in L / K and ( p , L / K ) ⊆ C } . \pi_{\mathcal{C}}(x,L/K):=\#\{\mathfrak{p}\in A:\deg\mathfrak{p}=x,\mathfrak{p}\text{ unramified in }L/K\text{ and }(\mathfrak{p},L/K)\subseteq\mathcal{C}\}. π C ( x , L / K ) := # { p ∈ A : deg p = x , p unramified in L / K and ( p , L / K ) ⊆ C } .
Theorem 3.8. [14 ] , Prop. 6.4.8, [22 ] , Theorem 7. Recall that K = F q ( T ) K=\mathbb{F}_q(T) K = F q ( T ) . Let L / K L/K L / K be a finite Galois extension with Galois group G G G and C ⊂ G \mathcal{C}\subset G C ⊂ G be a conjugacy class whose restriction to F ‾ L \overline{\mathbb{F}}_L F L is equal to τ n \tau^n τ n . Then, for x ≢ n ( m o d r L ) x\not\equiv n\pmod{r_L} x ≡ n ( mod r L ) , we have π C ( x , L / K ) = 0 \pi_{\mathcal{C}}(x,L/K)=0 π C ( x , L / K ) = 0 and for x ≡ n ( m o d r L ) x\equiv n\pmod{r_L} x ≡ n ( mod r L ) , we have
∣ π C ( x , L / K ) − r L ∣ C ∣ ∣ G ∣ q x x ∣ ≤ 2 ∣ C ∣ x ∣ G ∣ ( ( ∣ G ∣ + g L r L ) q x / 2 + ∣ G ∣ ( 2 g K + 1 ) q x / 4 + g L r L + ∣ G ∣ ) . \left|\pi_{\mathcal{C}}(x,L/K)-r_L\frac{|\mathcal{C}|}{|G|}\frac{q^x}{x}\right|\leq\frac{2|\mathcal{C}|}{x|G|}\left((|G|+g_Lr_L)q^{x/2}+|G|(2g_K+1)q^{x/4}+g_Lr_L+|G|\right). π C ( x , L / K ) − r L ∣ G ∣ ∣ C ∣ x q x ≤ x ∣ G ∣ 2∣ C ∣ ( ( ∣ G ∣ + g L r L ) q x /2 + ∣ G ∣ ( 2 g K + 1 ) q x /4 + g L r L + ∣ G ∣ ) . where g L g_L g L and g K g_K g K are the genus of L L L and K K K , respectively.
Let π ( x ) \pi(x) π ( x ) denote the number of primes in K K K of degree x x x . Substituting L = K L=K L = K in Theorem 3.8 yields the prime number theorem for polynomials:
π ( x ) = q x / x + O ( q x / 2 / x ) . (32) \pi(x)=q^x/x+O(q^{x/2}/x). \tag*{(32)} π ( x ) = q x / x + O ( q x /2 / x ) . (32)
We will apply the Chebotarev Density Theorem on the compositum fields L m , t L_{\mathfrak{m},t} L m , t , for monics m ∈ A \mathfrak{m}\in A m ∈ A (see (6) ). Let C m , t \mathcal{C}_{\mathfrak{m},t} C m , t be as defined in Lemma 3.7 and ϵ m , t \epsilon_{\mathfrak{m},t} ϵ m , t be a non-negative integer such that the restriction of ν m , t \nu_{\mathfrak{m},t} ν m , t to the field F ‾ L m , t \overline{\mathbb{F}}_{L_{\mathfrak{m},t}} F L m , t is equal to the map τ − ϵ m , t \tau^{-\epsilon_{\mathfrak{m},t}} τ − ϵ m , t . Using Lemma 3.7 , (7) , (32) , and the fact that there are only finitely many primes p \mathfrak{p} p of bad reduction, we have the following corollary of Theorem 3.8 . Henceforth, we let
r m , t = [ L m , t ∩ F ‾ q : F q ] , n m , t = [ L m , t : K ] , and g m , t = the genus of L m , t . r_{\mathfrak{m},t}=[L_{\mathfrak{m},t}\cap\overline{\mathbb{F}}_q:\mathbb{F}_q],\qquad n_{\mathfrak{m},t}=[L_{\mathfrak{m},t}:K],\qquad\text{and}\qquad g_{\mathfrak{m},t}=\text{the genus of }L_{\mathfrak{m},t}. r m , t = [ L m , t ∩ F q : F q ] , n m , t = [ L m , t : K ] , and g m , t = the genus of L m , t .
Corollary 3.9. For x ∈ N x\in\mathbb{N} x ∈ N , if γ a , t ( K ( ϕ [ m ] ) ) = 1 \gamma_{a,t}(K(\phi[\mathfrak{m}]))=1 γ a , t ( K ( ϕ [ m ])) = 1 and x ≡ ϵ m , t ( m o d r m , t ) x\equiv\epsilon_{\mathfrak{m},t}\pmod{r_{\mathfrak{m},t}} x ≡ ϵ m , t ( mod r m , t ) , then we have
π ϕ , m ( x ; t , a ) = r m , t n m , t π ( x ) + O ( ( 1 + r m , t g m , t n m , t ) q x / 2 x ) . \pi_{\phi,\mathfrak{m}}(x;t,a)=\frac{r_{\mathfrak{m},t}}{n_{\mathfrak{m},t}}\pi(x)+O\left(\left(1+\frac{r_{\mathfrak{m},t}g_{\mathfrak{m},t}}{n_{\mathfrak{m},t}}\right)\frac{q^{x/2}}{x}\right). π ϕ , m ( x ; t , a ) = n m , t r m , t π ( x ) + O ( ( 1 + n m , t r m , t g m , t ) x q x /2 ) . Otherwise π ϕ , m ( x ; t , a ) = 0 \pi_{\phi,\mathfrak{m}}(x;t,a)=0 π ϕ , m ( x ; t , a ) = 0 .
Some useful bounds
We now give some useful bounds below.
Theorem 3.10. [11 ] , Theorem 3.14. Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K of rank r ≥ 2 r\geq2 r ≥ 2 and define ρ : = rank A End K ( ϕ ) \rho:=\operatorname{rank}_A\operatorname{End}_K(\phi) ρ := rank A End K ( ϕ ) . Then, for any m ∈ A ∖ F q \mathfrak{m}\in A\setminus\mathbb{F}_q m ∈ A ∖ F q , we have
[ K ( ϕ [ m ] ) : K ] ≫ ϕ , K q r 2 deg m / ρ log ρ + log deg m + log log q . [K(\phi[\mathfrak{m}]):K]\gg_{\phi,K}\frac{q^{r^2\deg\mathfrak{m}/\rho}}{\log\rho+\log\deg\mathfrak{m}+\log\log q}. [ K ( ϕ [ m ]) : K ] ≫ ϕ , K log ρ + log deg m + log log q q r 2 d e g m / ρ .
Let K ∞ = F q ( ( 1 / T ) ) K_\infty=\mathbb{F}_q((1/T)) K ∞ = F q (( 1/ T )) be the completion of K K K at the prime at infinity. Then, K ∞ K_\infty K ∞ is the quotient field of the power series ring
F q [ [ 1 T ] ] = { ∑ i = 0 ∞ b i T − i : b i ∈ F q } . \mathbb{F}_q\left[\left[\frac{1}{T}\right]\right]=\left\{\sum_{i=0}^{\infty}b_iT^{-i}:b_i\in\mathbb{F}_q\right\}. F q [ [ T 1 ] ] = { i = 0 ∑ ∞ b i T − i : b i ∈ F q } .
Lemma 3.11. If H / K ∞ H/K_\infty H / K ∞ is a finite field extension, then H ∩ F ‾ q H\cap\overline{\mathbb{F}}_q H ∩ F q is a finite extension of F q \mathbb{F}_q F q . In particular, we have [ H ∩ F ‾ q : F q ] ≤ [ H : K ∞ ] [H\cap\overline{\mathbb{F}}_q:\mathbb{F}_q]\leq[H:K_\infty] [ H ∩ F q : F q ] ≤ [ H : K ∞ ] .
Proof. We proceed by contradiction. Let H / K ∞ H/K_\infty H / K ∞ be a finite field extension of degree d d d . Suppose H ∩ F ‾ q H \cap\overline{\mathbb{F}}_q H ∩ F q is an extension of F q \mathbb{F}_q F q of degree ≥ d + 1 \ge d+1 ≥ d + 1 . Then, there exists a set { α i } i = 1 d + 1 ⊆ H ∩ F ‾ q \{\alpha_i\}_{i=1}^{d+1} \subseteq H \cap\overline{\mathbb{F}}_q { α i } i = 1 d + 1 ⊆ H ∩ F q , which is linearly independent over F q \mathbb{F}_q F q . But since H / K ∞ H/K_\infty H / K ∞ is a finite extension, there exist c 1 , c 2 , ⋯ , c d + 1 ∈ K ∞ c_1,c_2,\cdots,c_{d+1} \in K_\infty c 1 , c 2 , ⋯ , c d + 1 ∈ K ∞ , not all zero, such that c 1 α 1 + c 2 α 2 + ⋯ + c d + 1 α d + 1 = 0 c_1\alpha_1+c_2\alpha_2+\cdots+c_{d+1}\alpha_{d+1}=0 c 1 α 1 + c 2 α 2 + ⋯ + c d + 1 α d + 1 = 0 . Clearing denominators, we obtain f i = ∑ j = 0 ∞ b i j T − j ∈ F q [ [ 1 T ] ] f_i=\sum_{j=0}^{\infty}b_{ij}T^{-j}\in\mathbb{F}_q\left[\left[\frac{1}{T}\right]\right] f i = ∑ j = 0 ∞ b ij T − j ∈ F q [ [ T 1 ] ] , for i = 1 , 2 , ⋯ , d + 1 i=1,2,\cdots,d+1 i = 1 , 2 , ⋯ , d + 1 , not all zero, such that
f 1 α 1 + f 2 α 2 + ⋯ + f d + 1 α d + 1 = 0. (33) f_1\alpha_1+f_2\alpha_2+\cdots+f_{d+1}\alpha_{d+1}=0. \tag*{(33)} f 1 α 1 + f 2 α 2 + ⋯ + f d + 1 α d + 1 = 0. (33)
Collecting coefficients of the same degree, we see that
∑ j = 0 ∞ ( b 1 j α 1 + b 2 j α 2 + ⋯ + b d + 1 , j α d + 1 ) T − j = 0. \sum_{j=0}^{\infty}(b_{1j}\alpha_1+b_{2j}\alpha_2+\cdots+b_{d+1,j}\alpha_{d+1})T^{-j}=0. j = 0 ∑ ∞ ( b 1 j α 1 + b 2 j α 2 + ⋯ + b d + 1 , j α d + 1 ) T − j = 0.
Hence, for every j ≥ 0 j\ge0 j ≥ 0 , we have that b 1 j α 1 + b 2 j α 2 + ⋯ + b d + 1 , j α d + 1 = 0 b_{1j}\alpha_1+b_{2j}\alpha_2+\cdots+b_{d+1,j}\alpha_{d+1}=0 b 1 j α 1 + b 2 j α 2 + ⋯ + b d + 1 , j α d + 1 = 0 , where each b i j ∈ F q b_{ij}\in\mathbb{F}_q b ij ∈ F q . By linear independence of α i \alpha_i α i 's over F q \mathbb{F}_q F q , we must then have that b i j = 0 b_{ij}=0 b ij = 0 , for all i , j i,j i , j . Thus, each f i = 0 f_i=0 f i = 0 , which is a contradiction since f i f_i f i 's were not all zero in (33) .
Hence, if H / K ∞ H/K_\infty H / K ∞ is a finite field extension, then H ∩ F ‾ q H\cap\overline{\mathbb{F}}_q H ∩ F q is a finite extension of F q \mathbb{F}_q F q .
Let C ∞ C_\infty C ∞ denote the completion of K ‾ ∞ \overline{K}_\infty K ∞ at the prime at infinity. For any Drinfeld A A A -module ϕ \phi ϕ over K K K , let ϕ t o r s \phi_{\mathrm{tors}} ϕ tors be defined as
ϕ t o r s : = { α ∈ C ∞ : ϕ a ( α ) = 0 for some non-zero a ∈ A } . (34) \phi_{\mathrm{tors}}:=\{\alpha\in C_\infty:\phi_a(\alpha)=0\text{ for some non-zero }a\in A\}. \tag*{(34)} ϕ tors := { α ∈ C ∞ : ϕ a ( α ) = 0 for some non-zero a ∈ A } . (34)
Note that ϕ t o r s \phi_{\mathrm{tors}} ϕ tors contains the torsion points ϕ [ m ] \phi[m] ϕ [ m ] , for every m ∈ A m\in A m ∈ A .
Lemma 3.12. Let L m , t L_{m,t} L m , t be as in (1.6). There exists a positive integer E ( ϕ , K ) E(\phi,K) E ( ϕ , K ) (depending only on ϕ \phi ϕ and K K K ) such that for every m , t ∈ A m,t\in A m , t ∈ A , we have r m , t = [ L m , t ∩ F ‾ q : F q ] < E ( ϕ , K ) r_{m,t}=[L_{m,t}\cap\overline{\mathbb{F}}_q:\mathbb{F}_q]<E(\phi,K) r m , t = [ L m , t ∩ F q : F q ] < E ( ϕ , K ) .
Proof. Let ϕ t o r s \phi_{\mathrm{tors}} ϕ tors and ψ t o r s C a r l \psi_{\mathrm{tors}}^{\mathrm{Carl}} ψ tors Carl be as defined in (34) . We have ϕ : A → K { τ } → K ∞ { τ } \phi:A\to K\{\tau\}\to K_\infty\{\tau\} ϕ : A → K { τ } → K ∞ { τ } . Then viewing ϕ \phi ϕ as a Drinfeld A A A -module over K ∞ K_\infty K ∞ , [17 Theorem 4.6.9] allows us to take M = K ∞ M=K_\infty M = K ∞ in [17 Remark 4.3.6.2] . Hence, there exist finite field extensions K ϕ / K ∞ K_\phi/K_\infty K ϕ / K ∞ and K C / K ∞ K_C/K_\infty K C / K ∞ such that ϕ t o r s ⊆ K ϕ \phi_{\mathrm{tors}}\subseteq K_\phi ϕ tors ⊆ K ϕ and ψ t o r s C a r l ⊆ K C \psi_{\mathrm{tors}}^{\mathrm{Carl}}\subseteq K_C ψ tors Carl ⊆ K C . Hence, we have the inclusion of compositums
⋃ m , t ∈ A K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) ⊆ K ϕ K C . (35) \bigcup_{m,t\in A}K(\phi[m])K(\psi^{\mathrm{Carl}}[t])\subseteq K_\phi K_C. \tag*{(35)} m , t ∈ A ⋃ K ( ϕ [ m ]) K ( ψ Carl [ t ]) ⊆ K ϕ K C . (35)
Since both K ϕ / K ∞ K_\phi/K_\infty K ϕ / K ∞ and K C / K ∞ K_C/K_\infty K C / K ∞ are finite extensions, we have that K ϕ K C / K ∞ K_\phi K_C/K_\infty K ϕ K C / K ∞ is also a finite extension and hence by Lemma 3.11 , K ϕ K C ∩ F ‾ q K_\phi K_C\cap\overline{\mathbb{F}}_q K ϕ K C ∩ F q is a finite extension of F q \mathbb{F}_q F q . Thus, by (35) , there exists a positive integer E ( ϕ , K ) E(\phi,K) E ( ϕ , K ) such that for every m , t ∈ A m,t\in A m , t ∈ A , [ L m , t ∩ F ‾ q : F q ] < E ( ϕ , K ) [L_{m,t}\cap\overline{\mathbb{F}}_q:\mathbb{F}_q]<E(\phi,K) [ L m , t ∩ F q : F q ] < E ( ϕ , K ) .
We observe that the genus g L g_L g L of a field L / K L/K L / K , as denoted in Theorem 3.8 , is essentially defined over the full constant field F L \mathbb{F}_L F L of L L L . So, we have that g L g_L g L is the genus of the field L / F L L/\mathbb{F}_L L / F L . To specify the role of the base constant field, we introduce the notation g L / F g_{L/F} g L / F for any constant field F ⊆ F L F\subseteq\mathbb{F}_L F ⊆ F L to denote the genus of the field L / F L/F L / F . Hence, we have that g L = g L / F L g_L=g_{L/\mathbb{F}_L} g L = g L / F L .
Lemma 3.13. For every m , t ∈ A m,t\in A m , t ∈ A , we have that
g m , t ≤ [ L m , t : K ( ϕ [ m ] ) ] ⋅ g K ( ϕ [ m ] ) + [ L m , t : K ( ψ C a r l [ t ] ) ] ⋅ g K ( ψ C a r l [ t ] ) + ( [ L m , t : K ( ϕ [ m ] ) ] − 1 ) ( [ L m , t : K ( ψ C a r l [ t ] ) ] − 1 ) . (36) g_{m,t}\le[L_{m,t}:K(\phi[m])] \cdot g_{K(\phi[m])}+[L_{m,t}:K(\psi^{\mathrm{Carl}}[t])] \cdot g_{K(\psi^{\mathrm{Carl}}[t])}
+([L_{m,t}:K(\phi[m])]-1)([L_{m,t}:K(\psi^{\mathrm{Carl}}[t])]-1). \tag*{(36)} g m , t ≤ [ L m , t : K ( ϕ [ m ])] ⋅ g K ( ϕ [ m ]) + [ L m , t : K ( ψ Carl [ t ])] ⋅ g K ( ψ Carl [ t ]) + ([ L m , t : K ( ϕ [ m ])] − 1 ) ([ L m , t : K ( ψ Carl [ t ])] − 1 ) . (36)
Proof. Let F = F L m , t F=\mathbb{F}_{L_{m,t}} F = F L m , t . We define the compositum fields L 1 : = K ( ϕ [ m ] ) F L_1:=K(\phi[m])F L 1 := K ( ϕ [ m ]) F and L 2 : = K ( ψ C a r l [ t ] ) F L_2:=K(\psi^{\mathrm{Carl}}[t])F L 2 := K ( ψ Carl [ t ]) F . Then, we note that L m , t = L 1 L 2 L_{m,t}=L_1L_2 L m , t = L 1 L 2 and F L 1 = F L 2 = F \mathbb{F}_{L_1}=\mathbb{F}_{L_2}=F F L 1 = F L 2 = F . Hence, Castelnuovo’s inequality [34 Theorem 3.11.3] implies that
g m , t ≤ [ L m , t : L 1 ] ⋅ g L 1 + [ L m , t : L 2 ] ⋅ g L 2 + ( [ L m , t : L 1 ] − 1 ) ( [ L m , t : L 2 ] − 1 ) . (37) g_{m,t}\le[L_{m,t}:L_1]\cdot g_{L_1}+[L_{m,t}:L_2]\cdot g_{L_2}+([L_{m,t}:L_1]-1)([L_{m,t}:L_2]-1). \tag*{(37)} g m , t ≤ [ L m , t : L 1 ] ⋅ g L 1 + [ L m , t : L 2 ] ⋅ g L 2 + ([ L m , t : L 1 ] − 1 ) ([ L m , t : L 2 ] − 1 ) . (37)
By [34 Theorem 3.6.3(b)] , we have
g L 1 = g L 1 / F = g K ( ϕ [ m ] ) and g L 2 = g L 2 / F = g K ( ψ C a r l [ t ] ) . (38) g_{L_1}=g_{L_1/F}=g_{K(\phi[m])}
\quad\text{and}\quad
g_{L_2}=g_{L_2/F}=g_{K(\psi^{\mathrm{Carl}}[t])}. \tag*{(38)} g L 1 = g L 1 / F = g K ( ϕ [ m ]) and g L 2 = g L 2 / F = g K ( ψ Carl [ t ]) . (38)
We also have that
[ L m , t : L 1 ] ≤ [ L m , t : K ( ϕ [ m ] ) ] and [ L m , t : L 2 ] ≤ [ L m , t : K ( ψ C a r l [ t ] ) ] . (39) [L_{m,t}:L_1]\le[L_{m,t}:K(\phi[m])] \quad\text{and}\quad[L_{m,t}:L_2]\le[L_{m,t}:K(\psi^{\mathrm{Carl}}[t])]. \tag*{(39)} [ L m , t : L 1 ] ≤ [ L m , t : K ( ϕ [ m ])] and [ L m , t : L 2 ] ≤ [ L m , t : K ( ψ Carl [ t ])] . (39)
Hence, using (3.12) and (39) in (3.11), we obtain (3.10). □
We will need the following inequality as stated in [11 Proposition 3.9] .
Proposition 3.14. [15 Corollary 7] Let ψ \psi ψ be a Drinfeld A A A -module over K K K . Let g K ( ψ [ m ] ) g_{K(\psi[m])} g K ( ψ [ m ]) denote the genus of the field K ( ψ [ m ] ) K(\psi[m]) K ( ψ [ m ]) . There exists a constant G ( ψ , K ) ∈ N G(\psi,K)\in\mathbb{N} G ( ψ , K ) ∈ N , depending on K K K and ψ \psi ψ , such that for any m ∈ A ∖ F q m\in A\setminus\mathbb{F}_{q} m ∈ A ∖ F q , we have g K ( ψ [ m ] ) ≤ G ( ψ , K ) ⋅ [ K ( ψ [ m ] ) : K ] ⋅ deg m g_{K(\psi[m])}\le G(\psi,K)\cdot[K(\psi[m]):K]\cdot\deg m g K ( ψ [ m ]) ≤ G ( ψ , K ) ⋅ [ K ( ψ [ m ]) : K ] ⋅ deg m .
This allows us to deduce
Proposition 3.15. There exist constants D ( ψ C a r l , K ) , D ( ϕ , K ) ∈ N D(\psi^{\mathrm{Carl}},K),D(\phi,K)\in\mathbb{N} D ( ψ Carl , K ) , D ( ϕ , K ) ∈ N such that for every m ∈ A ∖ F q m\in A\setminus\mathbb{F}_{q} m ∈ A ∖ F q ,
g m , t n m , t ≪ D ( ϕ , K ) deg m + D ( ψ C a r l , K ) deg t . \frac{g_{m,t}}{n_{m,t}}\ll D(\phi,K)\deg m+D(\psi^{\mathrm{Carl}},K)\deg t. n m , t g m , t ≪ D ( ϕ , K ) deg m + D ( ψ Carl , K ) deg t .
Proof. By Proposition 3.14, there exist constants D ( ψ C a r l , K ) ∈ N D(\psi^{\mathrm{Carl}},K)\in\mathbb{N} D ( ψ Carl , K ) ∈ N and D ( ϕ , K ) ∈ N D(\phi,K)\in\mathbb{N} D ( ϕ , K ) ∈ N satisfying
g K ( ψ C a r l [ t ] ) ≤ D ( ψ C a r l , K ) ⋅ [ K ( ψ C a r l [ t ] ) : K ] ⋅ deg t , g_{K(\psi^{\mathrm{Carl}}[t])}\le D(\psi^{\mathrm{Carl}},K)\cdot[K(\psi^{\mathrm{Carl}}[t]):K]\cdot\deg t, g K ( ψ Carl [ t ]) ≤ D ( ψ Carl , K ) ⋅ [ K ( ψ Carl [ t ]) : K ] ⋅ deg t ,
g K ( ϕ [ m ] ) ≤ D ( ϕ , K ) ⋅ [ K ( ϕ [ m ] ) : K ] ⋅ deg m . (40) g_{K(\phi[m])}\le D(\phi,K)\cdot[K(\phi[m]):K]\cdot\deg m. \tag*{(40)} g K ( ϕ [ m ]) ≤ D ( ϕ , K ) ⋅ [ K ( ϕ [ m ]) : K ] ⋅ deg m . (40)
By Lemma 3.13, we have the inequality
g m , t < [ L m , t : K ( ϕ [ m ] ) ] ⋅ g K ( ϕ [ m ] ) + [ L m , t : K ( ψ C a r l [ t ] ) ] ⋅ g K ( ψ C a r l [ t ] ) + [ L m , t : K ( ϕ [ m ] ) ] [ L m , t : K ( ψ C a r l [ t ] ) ] = [ L m , t : K ] ( g K ( ϕ [ m ] ) [ K ( ϕ [ m ] ) : K ] + g K ( ψ C a r l [ t ] ) [ K ( ψ C a r l [ t ] ) : K ] + [ L m , t : K ] [ K ( ϕ [ m ] ) : K ] [ K ( ψ C a r l [ t ] ) : K ] ) . \begin{aligned}
g_{m,t}&<[L_{m,t}:K(\phi[m])]\cdot g_{K(\phi[m])}+[L_{m,t}:K(\psi^{\mathrm{Carl}}[t])]\cdot g_{K(\psi^{\mathrm{Carl}}[t])}\\
&\quad+[L_{m,t}:K(\phi[m])][L_{m,t}:K(\psi^{\mathrm{Carl}}[t])]\\
&=[L_{m,t}:K]\left(\frac{g_{K(\phi[m])}}{[K(\phi[m]):K]}+\frac{g_{K(\psi^{\mathrm{Carl}}[t])}}{[K(\psi^{\mathrm{Carl}}[t]):K]}+\frac{[L_{m,t}:K]}{[K(\phi[m]):K][K(\psi^{\mathrm{Carl}}[t]):K]}\right).
\end{aligned} g m , t < [ L m , t : K ( ϕ [ m ])] ⋅ g K ( ϕ [ m ]) + [ L m , t : K ( ψ Carl [ t ])] ⋅ g K ( ψ Carl [ t ]) + [ L m , t : K ( ϕ [ m ])] [ L m , t : K ( ψ Carl [ t ])] = [ L m , t : K ] ( [ K ( ϕ [ m ]) : K ] g K ( ϕ [ m ]) + [ K ( ψ Carl [ t ]) : K ] g K ( ψ Carl [ t ]) + [ K ( ϕ [ m ]) : K ] [ K ( ψ Carl [ t ]) : K ] [ L m , t : K ] ) .
Hence, by (3.14), we obtain
g m , t n m , t < 1 [ K ( ϕ [ m ] ) : K ] g K ( ϕ [ m ] ) + 1 [ K ( ψ C a r l [ t ] ) : K ] g K ( ψ C a r l [ t ] ) + 1 [ K ( ϕ [ m ] ) ∩ K ( ψ C a r l [ t ] ) : K ] ≤ D ( ϕ , K ) deg m + D ( ψ C a r l , K ) deg t + 1 ≪ D ( ϕ , K ) deg m + D ( ψ C a r l , K ) deg t . \begin{aligned}
\frac{g_{m,t}}{n_{m,t}}&<\frac{1}{[K(\phi[m]):K]}g_{K(\phi[m])}+\frac{1}{[K(\psi^{\mathrm{Carl}}[t]):K]}g_{K(\psi^{\mathrm{Carl}}[t])}\\
&\quad+\frac{1}{[K(\phi[m])\cap K(\psi^{\mathrm{Carl}}[t]):K]}\\
&\le D(\phi,K)\deg m+D(\psi^{\mathrm{Carl}},K)\deg t+1\\
&\ll D(\phi,K)\deg m+D(\psi^{\mathrm{Carl}},K)\deg t.
\end{aligned} n m , t g m , t < [ K ( ϕ [ m ]) : K ] 1 g K ( ϕ [ m ]) + [ K ( ψ Carl [ t ]) : K ] 1 g K ( ψ Carl [ t ]) + [ K ( ϕ [ m ]) ∩ K ( ψ Carl [ t ]) : K ] 1 ≤ D ( ϕ , K ) deg m + D ( ψ Carl , K ) deg t + 1 ≪ D ( ϕ , K ) deg m + D ( ψ Carl , K ) deg t .
This proves the proposition. □
We note that the estimate in Proposition 3.15 also holds when m = 1 m=1 m = 1 . Indeed, if t = 1 t=1 t = 1 , then L 1 , 1 = K L_{1,1}=K L 1 , 1 = K , and hence g 1 , 1 = 0 g_{1,1}=0 g 1 , 1 = 0 . If deg t ≥ 1 \deg t\ge1 deg t ≥ 1 , then L 1 , t = K ( ψ C a r l [ t ] ) L_{1,t}=K(\psi^{\mathrm{Carl}}[t]) L 1 , t = K ( ψ Carl [ t ]) , and by Proposition 3.14,
g 1 , t n 1 , t = g K ( ψ C a r l [ t ] ) [ K ( ψ C a r l [ t ] ) : K ] ≪ D ( ψ C a r l , K ) deg t . \frac{g_{1,t}}{n_{1,t}}=\frac{g_{K(\psi^{\mathrm{Carl}}[t])}}{[K(\psi^{\mathrm{Carl}}[t]):K]}\ll D(\psi^{\mathrm{Carl}},K)\deg t. n 1 , t g 1 , t = [ K ( ψ Carl [ t ]) : K ] g K ( ψ Carl [ t ]) ≪ D ( ψ Carl , K ) deg t .
Thus, the estimate in Proposition 3.15 holds for every monic m ∈ A m\in A m ∈ A .
We apply Lemma 3.12 and Proposition 3.15 to Corollary 3.9. When either m ≠ 1 m\ne1 m = 1 or t ≠ 1 t\ne1 t = 1 , the initial 1 1 1 in the error term of Corollary 3.9 is absorbed by the bounds from Lemma 3.12 and Proposition 3.15. Then, for x ∈ N x\in\mathbb{N} x ∈ N , if γ a , t ( K ( ϕ [ m ] ) ) = 1 \gamma_{a,t}(K(\phi[m]))=1 γ a , t ( K ( ϕ [ m ])) = 1 (see (1.7)) and x ≡ ε m , t ( m o d r m , t ) x\equiv\varepsilon_{m,t}\pmod{r_{m,t}} x ≡ ε m , t ( mod r m , t ) , we have
π ϕ , m ( x ; t , a ) = r m , t n m , t π ( x ) + O ( E ( ϕ , K ) D ( ϕ , K ) deg m q x / 2 x ) + O ( E ( ϕ , K ) D ( ψ C a r l , K ) deg t q x / 2 x ) . (41) \pi_{\phi,m}(x;t,a)=\frac{r_{m,t}}{n_{m,t}}\pi(x)+O\left(E(\phi,K)D(\phi,K)\deg m\frac{q^{x/2}}{x}\right)+O\left(E(\phi,K)D(\psi^{\mathrm{Carl}},K)\deg t\frac{q^{x/2}}{x}\right). \tag*{(41)} π ϕ , m ( x ; t , a ) = n m , t r m , t π ( x ) + O ( E ( ϕ , K ) D ( ϕ , K ) deg m x q x /2 ) + O ( E ( ϕ , K ) D ( ψ Carl , K ) deg t x q x /2 ) . (41)
Otherwise π ϕ , m ( x ; t , a ) = 0 \pi_{\phi,m}(x;t,a)=0 π ϕ , m ( x ; t , a ) = 0 . When m = t = 1 m=t=1 m = t = 1 , we have L 1 , 1 = K L_{1,1}=K L 1 , 1 = K , and hence by Corollary 3.9, we have an additional O ϕ , K ( q x / 2 / x ) O_{\phi,K}(q^{x/2}/x) O ϕ , K ( q x /2 / x ) , which is absorbed into the first error term below. Hence, the Möbius sum in (3.2) yields
π ϕ ( x ; t , a ) = ∑ deg m ≤ x / r ′ μ q ( m ) γ a , t ( K ( ϕ [ m ] ) ) r m , t ( x ) n m , t π ( x ) + O ( q x / 2 x E ( ϕ , K ) D ( ϕ , K ) ∑ deg m ≤ x / r ′ deg m ) + O ( q x / 2 + x / r x E ( ϕ , K ) D ( ψ C a r l , K ) deg t ) , (42) \begin{aligned}
\pi_{\phi}(x;t,a)=\sum_{\deg m\le x/r}^{\prime}\frac{\mu_q(m)\gamma_{a,t}(K(\phi[m]))r_{m,t}(x)}{n_{m,t}}\pi(x)+O\left(\frac{q^{x/2}}{x}E(\phi,K)D(\phi,K)\sum_{\deg m\le x/r}^{\prime}\deg m\right) \\
&+ O\left(\frac{q^{x/2+x/r}}{x}E(\phi,K)D(\psi^{\mathrm{Carl}},K)\deg t\right), \tag*{(42)}
\end{aligned} π ϕ ( x ; t , a ) = d e g m ≤ x / r ∑ ′ n m , t μ q ( m ) γ a , t ( K ( ϕ [ m ])) r m , t ( x ) π ( x ) + O x q x /2 E ( ϕ , K ) D ( ϕ , K ) d e g m ≤ x / r ∑ ′ deg m + O ( x q x /2 + x / r E ( ϕ , K ) D ( ψ Carl , K ) deg t ) , (42)
where r m , t ( x ) r_{m,t}(x) r m , t ( x ) is equal to r m , t r_{m,t} r m , t whenever x ≡ ϵ m , t ( m o d r m , t ) x \equiv\epsilon_{m,t} \pmod{r_{m,t}} x ≡ ϵ m , t ( mod r m , t ) , and 0 otherwise.
Proof of Theorem 1.1
We wish to estimate the quantity (1.5). By (3.2), we have that
π ϕ ( x ; t , a ) = ∑ deg m ≤ x / r ′ μ q ( m ) π ϕ , m ( x ; t , a ) . (43) \pi_{\phi}(x;t,a)=\sum_{\substack{\deg m\le x/r}}' \mu_q(m)\pi_{\phi,m}(x;t,a). \tag*{(43)} π ϕ ( x ; t , a ) = d e g m ≤ x / r ∑ ′ μ q ( m ) π ϕ , m ( x ; t , a ) . (43)
One can observe that in the expression (3.16) for π ϕ ( x ; t , a ) \pi_{\phi}(x;t,a) π ϕ ( x ; t , a ) , when r = 2 r=2 r = 2 , the error terms are not O ( q x / x ) O(q^x/x) O ( q x / x ) , and are thus as large as π ( x ) \pi(x) π ( x ) (see (3.6)). Hence we will split the sum (4.1) into two parts and use (3.15) only for one of them. We will let y = y ( x ) y=y(x) y = y ( x ) be a parameter to be chosen later, satisfying y ( x ) → ∞ y(x)\to\infty y ( x ) → ∞ as x → ∞ x\to\infty x → ∞ . Then
π ϕ ( x ; t , a ) = ∑ m a i n + ∑ e r r o r , (44) \pi_{\phi}(x;t,a)=\sum_{\mathrm{main}}+\sum_{\mathrm{error}}, \tag*{(44)} π ϕ ( x ; t , a ) = main ∑ + error ∑ , (44)
where,
∑ m a i n : = ∑ deg m ≤ y μ q ( m ) π ϕ , m ( x ; t , a ) , ∑ e r r o r : = ∑ y < deg m ≤ x / r ′ μ q ( m ) π ϕ , m ( x ; t , a ) . (45) \sum_{\mathrm{main}}:=\sum_{\deg m\le y}\mu_q(m)\pi_{\phi,m}(x;t,a),\qquad
\sum_{\mathrm{error}}:=\sum_{y<\deg m\le x/r}'\mu_q(m)\pi_{\phi,m}(x;t,a). \tag*{(45)} main ∑ := d e g m ≤ y ∑ μ q ( m ) π ϕ , m ( x ; t , a ) , error ∑ := y < d e g m ≤ x / r ∑ ′ μ q ( m ) π ϕ , m ( x ; t , a ) . (45)
Main term
Using the asymptotic formula (3.15) for π ϕ , m ( x ; t , a ) \pi_{\phi,m}(x;t,a) π ϕ , m ( x ; t , a ) in (4.3), we obtain
∑ m a i n = ∑ deg m ≤ y μ q ( m ) γ a , t ( K ( ϕ [ m ] ) ) r m , t ( x ) n m , t π ( x ) + O ( q x / 2 x E ( ϕ , K ) D ( ϕ , K ) ∑ deg m ≤ y ′ deg m ) + O ( q x / 2 + y x E ( ϕ , K ) D ( ψ C a r l , K ) deg t ) , (46) \begin{aligned}
\sum_{\mathrm{main}}=\sum_{\deg m\le y}\frac{\mu_q(m)\gamma_{a,t}(K(\phi[m]))r_{m,t}(x)}{n_{m,t}}\pi(x)+O\left(\frac{q^{x/2}}{x}E(\phi,K)D(\phi,K)\sum_{\deg m\le y}'\deg m\right) \\
&+O\left(\frac{q^{x/2+y}}{x}E(\phi,K)D(\psi^{\mathrm{Carl}},K)\deg t\right), \tag*{(46)}
\end{aligned} main ∑ = d e g m ≤ y ∑ n m , t μ q ( m ) γ a , t ( K ( ϕ [ m ])) r m , t ( x ) π ( x ) + O x q x /2 E ( ϕ , K ) D ( ϕ , K ) d e g m ≤ y ∑ ′ deg m + O ( x q x /2 + y E ( ϕ , K ) D ( ψ Carl , K ) deg t ) , (46)
where r m , t ( x ) r_{m,t}(x) r m , t ( x ) is as defined in (1.11). Since ∑ deg m ≤ y deg m ≪ y q y \sum_{\deg m\le y}\deg m\ll yq^y ∑ d e g m ≤ y deg m ≪ y q y , the first error term above is ≪ ϕ , K q x / 2 + y y / x \ll_{\phi,K}q^{x/2+y}y/x ≪ ϕ , K q x /2 + y y / x . The first and second error terms thus are
≪ ϕ , K q x / 2 + y ( y x + deg t x ) . \ll_{\phi,K}q^{x/2+y}\left(\frac{y}{x}+\frac{\deg t}{x}\right). ≪ ϕ , K q x /2 + y ( x y + x deg t ) .
The first sum in (4.4) can be extended to run over all monics m ∈ A m\in A m ∈ A satisfying x ≡ ϵ m , t ( m o d r m , t ) x\equiv\epsilon_{m,t}\pmod{r_{m,t}} x ≡ ϵ m , t ( mod r m , t ) , up to an error E t a i l E_{\mathrm{tail}} E tail which is
≪ π ( x ) ∑ deg m > y ′ ∣ μ q ( m ) ∣ r m , t ( x ) γ a , t ( K ( ϕ [ m ] ) ) n m , t . (47) \ll\pi(x)\sum_{\deg m>y}'\frac{|\mu_q(m)|r_{m,t}(x)\gamma_{a,t}(K(\phi[m]))}{n_{m,t}}. \tag*{(47)} ≪ π ( x ) d e g m > y ∑ ′ n m , t ∣ μ q ( m ) ∣ r m , t ( x ) γ a , t ( K ( ϕ [ m ])) . (47)
Recalling that n m , t = [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] n_{m,t}=[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K] n m , t = [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] , we see that
∑ m a i n = δ ϕ , t , a ( x ) ⋅ π ( x ) + E t a i l + O ϕ , K ( q x / 2 + y ( y x + deg t x ) ) , (48) \sum_{\mathrm{main}}=\delta_{\phi,t,a}(x)\cdot\pi(x)+E_{\mathrm{tail}}+O_{\phi,K}\left(q^{x/2+y}\left(\frac{y}{x}+\frac{\deg t}{x}\right)\right), \tag*{(48)} main ∑ = δ ϕ , t , a ( x ) ⋅ π ( x ) + E tail + O ϕ , K ( q x /2 + y ( x y + x deg t ) ) , (48)
where δ ϕ , t , a ( x ) \delta_{\phi,t,a}(x) δ ϕ , t , a ( x ) is as defined in (1.12), and is shown to be an absolutely convergent sum below.
Using the bound n m , t ≥ [ K ( ϕ [ m ] ) : K ] n_{m,t}\ge[K(\phi[m]):K] n m , t ≥ [ K ( ϕ [ m ]) : K ] , and using Lemma 3.12 to bound r m , t ( x ) r_{m,t}(x) r m , t ( x ) , we see that
E t a i l ≪ E ( ϕ , K ) π ( x ) ∑ deg m > y ′ 1 [ K ( ϕ [ m ] ) : K ] . (49) E_{\mathrm{tail}}\ll E(\phi,K)\pi(x)\sum_{\deg m>y}'\frac{1}{[K(\phi[m]):K]}. \tag*{(49)} E tail ≪ E ( ϕ , K ) π ( x ) d e g m > y ∑ ′ [ K ( ϕ [ m ]) : K ] 1 . (49)
Invoking the lower bound for [ K ( ϕ [ m ] ) : K ] [K(\phi[m]):K] [ K ( ϕ [ m ]) : K ] given in Theorem 3.10, we see that (4.7) is
≪ ϕ , K π ( x ) ∑ deg m > y log ρ + log deg m + log log q q r 2 deg m / ρ ≪ K π ( x ) ∑ n > y q n ⋅ log ρ + log n + log log q q r 2 n / ρ . (50) \ll_{\phi,K}\pi(x)\sum_{\deg m>y}\frac{\log\rho+\log\deg m+\log\log q}{q^{r^2\deg m/\rho}}
\ll_K\pi(x)\sum_{n>y}q^n\cdot\frac{\log\rho+\log n+\log\log q}{q^{r^2n/\rho}}. \tag*{(50)} ≪ ϕ , K π ( x ) d e g m > y ∑ q r 2 d e g m / ρ log ρ + log deg m + log log q ≪ K π ( x ) n > y ∑ q n ⋅ q r 2 n / ρ log ρ + log n + log log q . (50)
where ρ = rank A End K ( ϕ ) \rho= \operatorname{rank}_{A}\operatorname{End}_{K}(\phi) ρ = rank A End K ( ϕ ) .
Note that y → ∞ y \to\infty y → ∞ as x → ∞ x \to\infty x → ∞ , q ≥ 2 q \ge2 q ≥ 2 , r ≥ 2 r \ge2 r ≥ 2 and ρ ≤ r \rho\le r ρ ≤ r (see [11 Theorem 3.2(ii)] ). For n n n sufficiently large, we have log ρ + log n + log log q ≪ ϕ , K log n \log\rho+ \log n + \log\log q \ll_{\phi,K} \log n log ρ + log n + log log q ≪ ϕ , K log n . Denote the exponent of 1 / q 1/q 1/ q in the summand of (50) by n a na na , where a = a ( r , ρ ) = ( r 2 / ρ ) − 1 a = a(r,\rho) = \left(r^{2}/\rho\right) - 1 a = a ( r , ρ ) = ( r 2 / ρ ) − 1 . It is easy to see that a ≥ 1 a \ge1 a ≥ 1 .
The sum over n n n in (50) is then
≪ ϕ , K ∑ n > y log n q n a / 2 ⋅ 1 q n a / 2 ≪ ϕ , K log y q y a / 2 ∑ n ≥ y 1 q n a / 2 , \ll_{\phi,K} \sum_{n>y} \frac{\log n}{q^{na/2}} \cdot\frac{1}{q^{na/2}} \ll_{\phi,K} \frac{\log y}{q^{ya/2}} \sum_{n\ge y} \frac{1}{q^{na/2}}, ≪ ϕ , K n > y ∑ q na /2 log n ⋅ q na /2 1 ≪ ϕ , K q y a /2 log y n ≥ y ∑ q na /2 1 ,
since log n q n a / 2 \frac{\log n}{q^{na/2}} q na /2 l o g n is a decreasing function of n n n as n → ∞ n \to\infty n → ∞ . The final sum above is a geometric progression summing to ≪ 1 q y a / 2 \ll\frac{1}{q^{ya/2}} ≪ q y a /2 1 .
We have thus obtained
E t a i l ≪ ϕ , K π ( x ) ⋅ log y q y a ≪ ϕ , K q x − ( r 2 ρ − 1 ) y log y x , (51) E_{\mathrm{tail}} \ll_{\phi,K} \pi(x) \cdot\frac{\log y}{q^{ya}} \ll_{\phi,K} \frac{q^{x-\left(\frac{r^{2}}{\rho}-1\right)y}\log y}{x}, \tag*{(51)} E tail ≪ ϕ , K π ( x ) ⋅ q y a log y ≪ ϕ , K x q x − ( ρ r 2 − 1 ) y log y , (51)
using the prime number theorem (32) .
Moreover, the above discussion also allows us to conclude that the density δ ϕ , t , a ( x ) \delta_{\phi,t,a}(x) δ ϕ , t , a ( x ) is finite. Indeed, from the bound n m , t ≥ [ K ( ϕ [ m ] ) : K ] n_{m,t} \ge[K(\phi[\mathfrak{m}]) : K] n m , t ≥ [ K ( ϕ [ m ]) : K ] , and Lemma 3.12 for r m , t r_{m,t} r m , t , we see that the δ ϕ , t , a ( x ) \delta_{\phi,t,a}(x) δ ϕ , t , a ( x ) is dominated in absolute value by the series
E ( ϕ , K ) ∑ m ∈ A ′ 1 [ K ( ϕ [ m ] ) : K ] . (52) E(\phi,K)\sum_{\mathfrak{m}\in A}' \frac{1}{[K(\phi[\mathfrak{m}]) : K]}. \tag*{(52)} E ( ϕ , K ) m ∈ A ∑ ′ [ K ( ϕ [ m ]) : K ] 1 . (52)
However the tail of (52) is simply the sum in (49) , which we showed to be ≪ ϕ , K log y q y a \ll_{\phi,K} \frac{\log y}{q^{ya}} ≪ ϕ , K q y a l o g y in the discussion above (51) , where a = r 2 ρ − 1 a = \frac{r^{2}}{\rho} - 1 a = ρ r 2 − 1 . As q ≥ 2 q \ge2 q ≥ 2 and a ≥ 1 a \ge1 a ≥ 1 , the tail converges to 0 0 0 as y → ∞ y \to\infty y → ∞ . By Cauchy’s criterion, the series (12) converges absolutely, as required.
Error term
In order to estimate the term ∑ e r r o r \sum_{\mathrm{error}} ∑ error in (44) , recall the definition of π ϕ , m ( x ; t , a ) \pi_{\phi,\mathfrak{m}}(x;t,a) π ϕ , m ( x ; t , a ) from (31) . Using Corollary 3.3, we see that each prime p \mathfrak{p} p being counted by π ϕ , m ( x ; t , a ) \pi_{\phi,\mathfrak{m}}(x;t,a) π ϕ , m ( x ; t , a ) satisfies
1 + b p + u p p ≡ 0 ( m o d m r ) and 1 + ( − 1 ) r − 1 u p p ≡ 0 ( m o d m ) , (53) 1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} \equiv0 \pmod{\mathfrak{m}^{r}} \quad\text{and} \quad1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} \equiv0 \pmod{\mathfrak{m}}, \tag*{(53)} 1 + b p + u p p ≡ 0 ( mod m r ) and 1 + ( − 1 ) r − 1 u p p ≡ 0 ( mod m ) , (53)
for some u p ∈ F q ∗ u_{\mathfrak{p}} \in\mathbb{F}_{q}^{*} u p ∈ F q ∗ and some b p ∈ A b_{\mathfrak{p}} \in A b p ∈ A with deg b p ≤ ( r − 1 ) x r \deg b_{\mathfrak{p}} \le\frac{(r-1)x}{r} deg b p ≤ r ( r − 1 ) x . Since 1 + b p + u p p 1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} 1 + b p + u p p and 1 + ( − 1 ) r − 1 u p p 1+(-1)^{r-1}u_{\mathfrak{p}}\mathfrak{p} 1 + ( − 1 ) r − 1 u p p are both divisible by m \mathfrak{m} m , we may eliminate p \mathfrak{p} p from the second congruence in (53) to obtain the equivalent system of congruences
1 + b p + u p p ≡ 0 ( m o d m r ) and 1 + b p + ( − 1 ) r ≡ 0 ( m o d m ) . 1+b_{\mathfrak{p}}+u_{\mathfrak{p}}\mathfrak{p} \equiv0 \pmod{\mathfrak{m}^{r}} \quad\text{and} \quad1+b_{\mathfrak{p}}+(-1)^{r} \equiv0 \pmod{\mathfrak{m}}. 1 + b p + u p p ≡ 0 ( mod m r ) and 1 + b p + ( − 1 ) r ≡ 0 ( mod m ) .
Summing over all possibilities u p = u u_{\mathfrak{p}}=u u p = u and b p = b b_{\mathfrak{p}}=b b p = b , where u ∈ F q ∗ u \in\mathbb{F}_{q}^{*} u ∈ F q ∗ and b ∈ A b \in A b ∈ A with deg b ≤ ( r − 1 ) x r \deg b \le\frac{(r-1)x}{r} deg b ≤ r ( r − 1 ) x , we see that
π ϕ , m ( x ; t , a ) ≤ ∑ u ∈ F q ∗ ∑ b ∈ A deg b ≤ ( r − 1 ) x r 1 + b + ( − 1 ) r ≡ 0 ( m o d m ) ∑ deg p = x p ≡ a ( m o d t ) 1 + b + u p ≡ 0 ( m o d m r ) 1. (54) \pi_{\phi,\mathfrak{m}}(x;t,a) \le\sum_{\substack{u\in\mathbb{F}_{q}^{*}}} \sum_{\substack{b\in A\\ \deg b\le\frac{(r-1)x}{r}\\ 1+b+(-1)^{r}\equiv0\pmod{\mathfrak{m}}}} \sum_{\substack{\deg\mathfrak{p}=x\\ \mathfrak{p}\equiv a\pmod t\\ 1+b+u\mathfrak{p}\equiv0\pmod{\mathfrak{m}^{r}}}} 1. \tag*{(54)} π ϕ , m ( x ; t , a ) ≤ u ∈ F q ∗ ∑ b ∈ A d e g b ≤ r ( r − 1 ) x 1 + b + ( − 1 ) r ≡ 0 ( mod m ) ∑ d e g p = x p ≡ a ( mod t ) 1 + b + u p ≡ 0 ( mod m r ) ∑ 1. (54)
We observe that if the congruences p ≡ a ( m o d t ) \mathfrak{p} \equiv a \pmod t p ≡ a ( mod t ) , p ≡ − ( 1 + b ) u − 1 ( m o d m r ) \mathfrak{p} \equiv-(1+b)u^{-1} \pmod{\mathfrak{m}^{r}} p ≡ − ( 1 + b ) u − 1 ( mod m r ) are incompatible modulo gcd ( t , m r ) \gcd(t,\mathfrak{m}^{r}) g cd( t , m r ) , then the innermost sum is empty. Otherwise, by the Chinese remainder theorem, they can be written as a single congruence p ≡ c ( m o d [ t , m r ] ) \mathfrak{p} \equiv c \pmod{[t,\mathfrak{m}^{r}]} p ≡ c ( mod [ t , m r ]) for some nonzero c ∈ A c \in A c ∈ A with deg c < deg [ t , m r ] \deg c < \deg[t,\mathfrak{m}^{r}] deg c < deg [ t , m r ] and where [ t , m r ] [t,\mathfrak{m}^{r}] [ t , m r ] is the least common multiple of the monics t t t and m r \mathfrak{m}^{r} m r in A A A .
Since the prime p ∈ A \mathfrak{p} \in A p ∈ A is monic, we may write p \mathfrak{p} p as c + [ t , m r ] ⋅ h 1 c+[t,\mathfrak{m}^{r}]\cdot h_{1} c + [ t , m r ] ⋅ h 1 for some h 1 ∈ A h_{1}\in A h 1 ∈ A where h 1 = 0 h_{1}=0 h 1 = 0 or h 1 h_{1} h 1 is monic with deg h 1 = x − deg [ t , m r ] \deg h_{1}=x-\deg[t,\mathfrak{m}^{r}] deg h 1 = x − deg [ t , m r ] . Hence, the innermost sum in (54) is at most 1 + q x − deg [ t , m r ] 1+q^{x-\deg[t,m^r]} 1 + q x − d e g [ t , m r ] , so that (4.12) is
≪ ( 1 + q x − deg [ t , m r ] ) ∑ u ∈ F q ∗ ∑ b ∈ A deg b ≤ ( r − 1 ) x r b ≡ − 1 − ( − 1 ) r ( m o d m ) 1. (55) \ll\left(1+q^{x-\deg[t,m^r]}\right)\sum_{u\in\mathbb{F}_q^*}\sum_{\substack{b\in A\\ \deg b\leq\frac{(r-1)x}{r}\\ b\equiv-1-(-1)^r\pmod m}}1. \tag*{(55)} ≪ ( 1 + q x − d e g [ t , m r ] ) u ∈ F q ∗ ∑ b ∈ A d e g b ≤ r ( r − 1 ) x b ≡ − 1 − ( − 1 ) r ( mod m ) ∑ 1. (55)
To estimate the innermost sum above, we may again write b b b as b = − 1 − ( − 1 ) r + m ⋅ h 2 b=-1-(-1)^r+m\cdot h_2 b = − 1 − ( − 1 ) r + m ⋅ h 2 for some h 2 ∈ A h_2\in A h 2 ∈ A . Since deg b ≤ ( r − 1 ) x / r \deg b\leq(r-1)x/r deg b ≤ ( r − 1 ) x / r , we must have deg h 2 ≤ ( r − 1 ) x / r − deg m \deg h_2\leq(r-1)x/r-\deg m deg h 2 ≤ ( r − 1 ) x / r − deg m . This means that the innermost sum in (4.13) is ≤ q ( r − 1 ) x r − deg m + 1 \leq q^{\frac{(r-1)x}{r}-\deg m+1} ≤ q r ( r − 1 ) x − d e g m + 1 . Putting these bounds back into (4.12), we obtain
π ϕ , m ( x ; t , a ) ≤ ( q r − 1 r x − deg m + 1 + q 2 r − 1 r x − deg m − deg [ t , m r ] + 1 ) ∑ u ∈ F q ∗ 1 \pi_{\phi,m}(x;t,a)\leq\left(q^{\frac{r-1}{r}x-\deg m+1}+q^{\frac{2r-1}{r}x-\deg m-\deg[t,m^r]+1}\right)\sum_{u\in\mathbb{F}_q^*}1 π ϕ , m ( x ; t , a ) ≤ ( q r r − 1 x − d e g m + 1 + q r 2 r − 1 x − d e g m − d e g [ t , m r ] + 1 ) u ∈ F q ∗ ∑ 1
≪ K q r − 1 r x − deg m ( 1 + q x − deg [ t , m r ] ) . \ll_K q^{\frac{r-1}{r}x-\deg m}\left(1+q^{x-\deg[t,m^r]}\right). ≪ K q r r − 1 x − d e g m ( 1 + q x − d e g [ t , m r ] ) .
We now come back to the expression (4.3) for ∑ e r r o r \sum_{\mathrm{error}} ∑ error to see that
∑ e r r o r ≪ K q r − 1 r x ∑ y < deg m ≤ x / r ′ q − deg m + q 2 r − 1 r x ∑ y < deg m ≤ x / r ′ q − deg m − deg [ t , m r ] = q r − 1 r x ∑ y < deg m ≤ x / r ′ q − deg m + q 2 r − 1 r x − deg t ∑ y < deg m ≤ x / r ′ q − ( r + 1 ) deg m + deg ( t , m r ) , (56) \begin{aligned}
\sum_{\mathrm{error}}\ll_K q^{\frac{r-1}{r}x}\sum_{y<\deg m\leq x/r}^{\prime}q^{-\deg m}+q^{\frac{2r-1}{r}x}\sum_{y<\deg m\leq x/r}^{\prime}q^{-\deg m-\deg[t,m^r]} \\
&=q^{\frac{r-1}{r}x}\sum_{y<\deg m\leq x/r}^{\prime}q^{-\deg m}+q^{\frac{2r-1}{r}x-\deg t}\sum_{y<\deg m\leq x/r}^{\prime}q^{-(r+1)\deg m+\deg(t,m^r)}, \tag*{(56)}
\end{aligned} error ∑ ≪ K q r r − 1 x y < d e g m ≤ x / r ∑ ′ q − d e g m + q r 2 r − 1 x y < d e g m ≤ x / r ∑ ′ q − d e g m − d e g [ t , m r ] = q r r − 1 x y < d e g m ≤ x / r ∑ ′ q − d e g m + q r 2 r − 1 x − d e g t y < d e g m ≤ x / r ∑ ′ q − ( r + 1 ) d e g m + d e g ( t , m r ) , (56)
where we have used deg [ t , m r ] = deg t + r deg m − deg ( t , m r ) \deg[t,m^r]=\deg t+r\deg m-\deg(t,m^r) deg [ t , m r ] = deg t + r deg m − deg ( t , m r ) , and ( t , m r ) ∈ A (t,m^r)\in A ( t , m r ) ∈ A denotes the monic polynomial which is the gcd of t , m r t,m^r t , m r . We rewrite the first term of (4.14) as
q r − 1 r x ∑ y < ℓ ≤ x / r q − ℓ ∑ m : deg m = ℓ ′ 1 = q r − 1 r x ∑ y < ℓ ≤ x / r 1 ≤ x r q r − 1 r x . (57) q^{\frac{r-1}{r}x}\sum_{y<\ell\leq x/r}q^{-\ell}\sum_{\substack{m:\deg m=\ell}}^{\prime}1=q^{\frac{r-1}{r}x}\sum_{y<\ell\leq x/r}1\leq\frac{x}{r}q^{\frac{r-1}{r}x}. \tag*{(57)} q r r − 1 x y < ℓ ≤ x / r ∑ q − ℓ m : d e g m = ℓ ∑ ′ 1 = q r r − 1 x y < ℓ ≤ x / r ∑ 1 ≤ r x q r r − 1 x . (57)
We use the identity ∑ d ∣ t , d ∣ m r φ ( d ) = q deg ( t , m r ) \sum_{d\mid t,d\mid m^r}\varphi(d)=q^{\deg(t,m^r)} ∑ d ∣ t , d ∣ m r φ ( d ) = q d e g ( t , m r ) , to rewrite the sum in the second term of (4.14) as
∑ y < deg m ≤ x / r ′ q − ( r + 1 ) deg m ∑ d ∣ t d ∣ m r ′ φ ( d ) . (58) \sum_{y<\deg m\leq x/r}^{\prime}q^{-(r+1)\deg m}\sum_{\substack{d\mid t\\d\mid m^r}}^{\prime}\varphi(d). \tag*{(58)} y < d e g m ≤ x / r ∑ ′ q − ( r + 1 ) d e g m d ∣ t d ∣ m r ∑ ′ φ ( d ) . (58)
Interchanging the order of summation, we get that (4.16) is equal to
∑ d ∣ t ′ φ ( d ) ∑ y < deg m ≤ x / r d ∣ m r ′ q − ( r + 1 ) deg m . (59) \sum_{d\mid t}^{\prime}\varphi(d)\sum_{\substack{y<\deg m\leq x/r\\d\mid m^r}}^{\prime}q^{-(r+1)\deg m}. \tag*{(59)} d ∣ t ∑ ′ φ ( d ) y < d e g m ≤ x / r d ∣ m r ∑ ′ q − ( r + 1 ) d e g m . (59)
Let d 0 : = d 0 ( d ) d_0:=d_0(d) d 0 := d 0 ( d ) be defined such that for d = p 1 α 1 p 2 α 2 ⋯ p k α k d=p_1^{\alpha_1}p_2^{\alpha_2}\cdots p_k^{\alpha_k} d = p 1 α 1 p 2 α 2 ⋯ p k α k , we have that
d 0 ( p 1 α 1 p 2 α 2 ⋯ p k α k ) = p 1 ⌈ α 1 / r ⌉ p 2 ⌈ α 2 / r ⌉ ⋯ p k ⌈ α k / r ⌉ . (60) d_0(p_1^{\alpha_1}p_2^{\alpha_2}\cdots p_k^{\alpha_k})=p_1^{\lceil\alpha_1/r\rceil}p_2^{\lceil\alpha_2/r\rceil}\cdots p_k^{\lceil\alpha_k/r\rceil}. \tag*{(60)} d 0 ( p 1 α 1 p 2 α 2 ⋯ p k α k ) = p 1 ⌈ α 1 / r ⌉ p 2 ⌈ α 2 / r ⌉ ⋯ p k ⌈ α k / r ⌉ . (60)
Then, d ∣ m r d\mid m^r d ∣ m r if and only if d 0 ∣ m d_0\mid m d 0 ∣ m . Writing m = d 0 h 3 m=d_0h_3 m = d 0 h 3 for some monic h 3 ∈ A h_3\in A h 3 ∈ A , we rewrite the innermost sum in (4.17) as
q − ( r + 1 ) deg d 0 ∑ h 3 : y < deg d 0 + deg h 3 ≤ x / r ′ q − ( r + 1 ) deg h 3 . (61) q^{-(r+1)\deg d_0}\sum_{\substack{h_3:\\y<\deg d_0+\deg h_3\leq x/r}}^{\prime}q^{-(r+1)\deg h_3}. \tag*{(61)} q − ( r + 1 ) d e g d 0 h 3 : y < d e g d 0 + d e g h 3 ≤ x / r ∑ ′ q − ( r + 1 ) d e g h 3 . (61)
Then (4.19) is bounded above by
q − ( r + 1 ) deg d 0 ∑ ℓ > y − deg d 0 q − ( r + 1 ) ℓ ∑ h 3 : deg h 3 = ℓ ′ 1 = q − ( r + 1 ) deg d 0 ∑ ℓ > y − deg d 0 q − r ℓ q^{-(r+1)\deg d_0}\sum_{\ell>y-\deg d_0}q^{-(r+1)\ell}\sum_{\substack{h_3:\deg h_3=\ell}}^{\prime}1=q^{-(r+1)\deg d_0}\sum_{\ell>y-\deg d_0}q^{-r\ell} q − ( r + 1 ) d e g d 0 ℓ > y − d e g d 0 ∑ q − ( r + 1 ) ℓ h 3 : d e g h 3 = ℓ ∑ ′ 1 = q − ( r + 1 ) d e g d 0 ℓ > y − d e g d 0 ∑ q − r ℓ
≪ K q − deg d 0 − r y , (62) \ll_K q^{-\deg d_0-ry}, \tag*{(62)} ≪ K q − d e g d 0 − r y , (62)
since the sum ∑ ℓ > y − deg d 0 q − r ℓ \sum_{\ell>y-\deg d_0}q^{-r\ell} ∑ ℓ > y − d e g d 0 q − r ℓ is the tail sum of a geometric progression.
Hence, by (58) , (59) , (61) and (62) , the sum in the second term of (56) is bounded above by
q − r y ∑ d ∣ t ′ φ ( d ) q − deg d 0 . (63) q^{-ry}\sum_{\substack{d\mid t}}' \varphi(d)q^{-\deg d_0}. \tag*{(63)} q − r y d ∣ t ∑ ′ φ ( d ) q − d e g d 0 . (63)
Recall that J ϕ ( t ) = ∑ d ∣ t ′ φ ( d ) q − deg d 0 J_\phi(t)=\sum_{\substack{d\mid t}}'\varphi(d)q^{-\deg d_0} J ϕ ( t ) = ∑ d ∣ t ′ φ ( d ) q − d e g d 0 , as defined in (9) . Thus, by (56) , (57) and (63) , we have obtained
∑ e r r o r ′ ≪ K x q q ( r − 1 r ) x + q ( 2 r − 1 r ) x − deg t − r y J ϕ ( t ) . (64) \sum_{\mathrm{error}}' \ll_{K} \frac{x}{q}q^{\left(\frac{r-1}{r}\right)x}+q^{\left(\frac{2r-1}{r}\right)x-\deg t-ry}J_\phi(t). \tag*{(64)} error ∑ ′ ≪ K q x q ( r r − 1 ) x + q ( r 2 r − 1 ) x − d e g t − r y J ϕ ( t ) . (64)
Remark 3. Analogous to Wong’s refinement [38 Section 4.2, p. 771] , one may estimate the innermost sum in (54) by discarding the congruence condition modulo t \mathfrak{t} t and applying the function field Brun–Titchmarsh theorem [20 Theorem 4.3] to the residue class modulo m r \mathfrak{m}^{r} m r . This gives, for y < x / r y<x/r y < x / r , the alternative estimate
∑ e r r o r ′ ( y ) ≪ q , r , K q ( r − 1 ) x / r ( 1 + q x − r y x − r y ) . \sum_{\mathrm{error}}'(y)\ll_{q,r,K}q^{(r-1)x/r}\left(1+\frac{q^{x-ry}}{x-ry}\right). error ∑ ′ ( y ) ≪ q , r , K q ( r − 1 ) x / r ( 1 + x − r y q x − r y ) .
Then, ∑ e r r o r ′ \sum_{\mathrm{error}}' ∑ error ′ is better precisely when 1 − J ϕ ( t ) ( x − r y ) q deg t < ( x r − 1 ) x − r y q x − r y 1-\frac{J_\phi(\mathfrak{t})(x-ry)}{q^{\deg\mathfrak{t}}}<\left(\frac{x}{r}-1\right)\frac{x-ry}{q^{x-ry}} 1 − q d e g t J ϕ ( t ) ( x − r y ) < ( r x − 1 ) q x − r y x − r y . Hence, if y y y is so close to x / r x/r x / r that q x − r y x − r y ≤ x r − 1 \frac{q^{x-ry}}{x-ry}\leq\frac{x}{r}-1 x − r y q x − r y ≤ r x − 1 , then ∑ e r r o r ′ \sum_{\mathrm{error}}' ∑ error ′ is better for every t \mathfrak{t} t . Otherwise, ∑ e r r o r ′ \sum_{\mathrm{error}}' ∑ error ′ is better exactly when
q deg t J ϕ ( t ) < x − r y 1 − ( x r − 1 ) x − r y q x − r y . \frac{q^{\deg\mathfrak{t}}}{J_\phi(\mathfrak{t})}<\frac{x-ry}{1-\left(\frac{x}{r}-1\right)\frac{x-ry}{q^{x-ry}}}. J ϕ ( t ) q d e g t < 1 − ( r x − 1 ) q x − r y x − r y x − r y .
Estimation of J ϕ ( t ) J_\phi(t) J ϕ ( t )
Observe that the summands φ ( d ) q − deg d 0 \varphi(d)q^{-\deg d_0} φ ( d ) q − d e g d 0 are multiplicative in d d d and hence J ϕ ( t ) J_\phi(t) J ϕ ( t ) is multiplicative in t t t .
Since d ∣ d 0 r d\mid d_0^r d ∣ d 0 r by definition, we have deg d 0 ≥ 1 r deg d \deg d_0\geq\frac{1}{r}\deg d deg d 0 ≥ r 1 deg d . Hence each summand of J ϕ ( t ) J_\phi(t) J ϕ ( t ) is bounded above by φ ( d ) q − deg d r ≤ q ( 1 − 1 r ) deg d \varphi(d)q^{-\frac{\deg d}{r}}\leq q^{(1-\frac{1}{r})\deg d} φ ( d ) q − r d e g d ≤ q ( 1 − r 1 ) d e g d . Thus one has J ϕ ( t ) ≤ τ ( t ) q ( r − 1 r ) deg t J_\phi(t)\leq\tau(t)q^{(\frac{r-1}{r})\deg t} J ϕ ( t ) ≤ τ ( t ) q ( r r − 1 ) d e g t , where τ ( t ) \tau(t) τ ( t ) is the number of monic polynomials in A A A dividing t t t .
We now obtain a nontrivial bound for J ϕ ( t ) J_\phi(t) J ϕ ( t ) . Let t = p α t=\mathfrak{p}^{\alpha} t = p α , a prime power. Then
J ϕ ( p α ) = 1 + ∑ i = 1 α ( q i deg p − q ( i − 1 ) deg p ) q − ⌈ i / r ⌉ deg p = ( 1 − 1 q deg p ) H ϕ ( p α ) + 1 q deg p , (65) J_\phi(\mathfrak{p}^{\alpha})=1+\sum_{i=1}^{\alpha}\left(q^{i\deg\mathfrak{p}}-q^{(i-1)\deg\mathfrak{p}}\right)q^{-\lceil i/r\rceil\deg\mathfrak{p}}=\left(1-\frac{1}{q^{\deg\mathfrak{p}}}\right)H_\phi(\mathfrak{p}^{\alpha})+\frac{1}{q^{\deg\mathfrak{p}}}, \tag*{(65)} J ϕ ( p α ) = 1 + i = 1 ∑ α ( q i d e g p − q ( i − 1 ) d e g p ) q − ⌈ i / r ⌉ d e g p = ( 1 − q d e g p 1 ) H ϕ ( p α ) + q d e g p 1 , (65)
where
H ϕ ( p α ) : = ∑ i = 0 α q ( i − ⌈ i / r ⌉ ) deg p . (66) H_\phi(\mathfrak{p}^{\alpha}):=\sum_{i=0}^{\alpha}q^{(i-\lceil i/r\rceil)\deg\mathfrak{p}}. \tag*{(66)} H ϕ ( p α ) := i = 0 ∑ α q ( i − ⌈ i / r ⌉) d e g p . (66)
This definition of H ϕ H_\phi H ϕ on prime powers yields the multiplicative function
H ϕ ( f ) : = ∑ g ∣ f ′ ∑ h ∈ A : g ∣ h r deg h < deg g 1 , H_\phi(f):=\sum_{g\mid f}'\sum_{\substack{h\in A:g\mid h^r\\\deg h<\deg g}}1, H ϕ ( f ) := g ∣ f ∑ ′ h ∈ A : g ∣ h r d e g h < d e g g ∑ 1 ,
where the inner sum is over all polynomials h ∈ A h\in A h ∈ A , including h = 0 h=0 h = 0 . Indeed, for g = p i g=\mathfrak{p}^{i} g = p i , the condition p i ∣ h r \mathfrak{p}^{i}\mid h^r p i ∣ h r is equivalent to p ⌈ i / r ⌉ ∣ h \mathfrak{p}^{\lceil i/r\rceil}\mid h p ⌈ i / r ⌉ ∣ h , and hence the inner sum has exactly q ( i − ⌈ i / r ⌉ ) deg p q^{(i-\lceil i/r\rceil)\deg\mathfrak{p}} q ( i − ⌈ i / r ⌉) d e g p elements. This function H ϕ ( f ) H_\phi(f) H ϕ ( f ) is the function field analogue of the function H ( f ) H(f) H ( f ) defined in [1 Theorem 3] .
From (66) , it is easy to see that H ϕ ( p α ) ≥ 2 H_\phi(\mathfrak{p}^{\alpha})\geq2 H ϕ ( p α ) ≥ 2 whenever α ≥ 1 \alpha\geq1 α ≥ 1 . Hence, for α ≥ 1 \alpha\geq1 α ≥ 1 , one has
1 q deg p ≤ H ϕ ( p α ) 2 q deg p , \frac{1}{q^{\deg\mathfrak{p}}}\leq\frac{H_\phi(\mathfrak{p}^{\alpha})}{2q^{\deg\mathfrak{p}}}, q d e g p 1 ≤ 2 q d e g p H ϕ ( p α ) ,
and substituting this inequality into (65) yields
( 1 − 1 q deg p ) H ϕ ( p α ) ≤ J ϕ ( p α ) ≤ ( 1 − 1 2 q deg p ) H ϕ ( p α ) . \left(1-\frac{1}{q^{\deg\mathfrak{p}}}\right)H_\phi(\mathfrak{p}^{\alpha})\leq J_\phi(\mathfrak{p}^{\alpha})\leq\left(1-\frac{1}{2q^{\deg\mathfrak{p}}}\right)H_\phi(\mathfrak{p}^{\alpha}). ( 1 − q d e g p 1 ) H ϕ ( p α ) ≤ J ϕ ( p α ) ≤ ( 1 − 2 q d e g p 1 ) H ϕ ( p α ) .
for prime powers p α p^\alpha p α with α ≥ 1 \alpha\ge1 α ≥ 1 . Moreover, J ϕ ( 1 ) = H ϕ ( 1 ) J_\phi(1)=H_\phi(1) J ϕ ( 1 ) = H ϕ ( 1 ) . Hence, by multiplicativity of H ϕ H_\phi H ϕ and J ϕ J_\phi J ϕ , for any monic t ∈ A t\in A t ∈ A ,
J ϕ ( t ) ≤ H ϕ ( t ) ∏ primes p ∣ t ( 1 − 1 2 q deg p ) . (67) J_\phi(t)\le H_\phi(t)\prod_{\substack{\text{primes }p\mid t}}\left(1-\frac{1}{2q^{\deg p}}\right). \tag*{(67)} J ϕ ( t ) ≤ H ϕ ( t ) primes p ∣ t ∏ ( 1 − 2 q d e g p 1 ) . (67)
Following [1 Remark 1, p.1284] , one can show analogously that
H ϕ ( ∏ i ≤ k p i α i ) ≤ 2 k σ ( ∏ i ≤ k p i ⌊ α i ( r − 1 ) / r ⌋ ) , and H ϕ ( t ) ≤ 9 4 2 ω ( t ) q ( r − 1 r ) deg t , (68) H_\phi\left(\prod_{i\le k}p_i^{\alpha_i}\right)\le2^k\sigma\left(\prod_{i\le k}p_i^{\left\lfloor\alpha_i(r-1)/r\right\rfloor}\right),\qquad\text{and}\qquad H_\phi(t)\le\frac{9}{4}2^{\omega(t)}q^{\left(\frac{r-1}{r}\right)\deg t}, \tag*{(68)} H ϕ ( i ≤ k ∏ p i α i ) ≤ 2 k σ ( i ≤ k ∏ p i ⌊ α i ( r − 1 ) / r ⌋ ) , and H ϕ ( t ) ≤ 4 9 2 ω ( t ) q ( r r − 1 ) d e g t , (68)
where σ ( f ) : = ∑ d ∣ f q deg d \sigma(f):=\sum_{d\mid f}q^{\deg d} σ ( f ) := ∑ d ∣ f q d e g d , and ω ( t ) \omega(t) ω ( t ) denotes the number of distinct prime divisors of t t t . Thus, by (4.25) and (4.26), we have
J ϕ ( t ) ≤ 9 4 2 ω ( t ) q ( r − 1 r ) deg t ∏ primes p ∣ t ( 1 − 1 2 q deg p ) . J_\phi(t)\le\frac{9}{4}2^{\omega(t)}q^{\left(\frac{r-1}{r}\right)\deg t}\prod_{\substack{\text{primes }p\mid t}}\left(1-\frac{1}{2q^{\deg p}}\right). J ϕ ( t ) ≤ 4 9 2 ω ( t ) q ( r r − 1 ) d e g t primes p ∣ t ∏ ( 1 − 2 q d e g p 1 ) .
Final expression for π ϕ ( x ; t , a ) \pi_\phi(x;t,a) π ϕ ( x ; t , a )
Combining (4.2), (4.6), (4.9), (4.22), we have obtained
π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + E 1 + E 2 + E 3 a + E 3 b , (69) \pi_\phi(x;t,a)=\pi(x)\cdot\delta_{\phi,t,a}(x)+E_1+E_2+E_{3a}+E_{3b}, \tag*{(69)} π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + E 1 + E 2 + E 3 a + E 3 b , (69)
where
E 1 = O ϕ , K ( q x 2 + y ( y x + deg t x ) ) , E 2 = O ϕ , K ( q x − ( ρ 2 ρ − 1 ) y log y x ) , E_1=O_{\phi,K}\left(q^{\frac{x}{2}+y}\left(\frac{y}{x}+\frac{\deg t}{x}\right)\right),\qquad E_2=O_{\phi,K}\left(\frac{q^{x-\left(\frac{\rho^2}{\rho}-1\right)y}\log y}{x}\right), E 1 = O ϕ , K ( q 2 x + y ( x y + x deg t ) ) , E 2 = O ϕ , K x q x − ( ρ ρ 2 − 1 ) y log y ,
E 3 a = O K ( x r q ( r − 1 r ) x ) , E 3 b = O K ( q ( 2 r − 1 r ) x − α ϕ ( t ) − r y ) , E_{3a}=O_K\left(\frac{x}{r}q^{\left(\frac{r-1}{r}\right)x}\right),\qquad E_{3b}=O_K\left(q^{\left(\frac{2r-1}{r}\right)x-\alpha_\phi(t)-ry}\right), E 3 a = O K ( r x q ( r r − 1 ) x ) , E 3 b = O K ( q ( r 2 r − 1 ) x − α ϕ ( t ) − r y ) ,
where, for monics t ∈ A t\in A t ∈ A , we define
α ϕ ( t ) : = deg t − log q J ϕ ( t ) , (70) \alpha_\phi(t):=\deg t-\log_q J_\phi(t), \tag*{(70)} α ϕ ( t ) := deg t − log q J ϕ ( t ) , (70)
with log q \log_q log q denoting logarithm to the base q q q and J ϕ ( t ) : A → R + J_\phi(t):A\to\mathbb{R}^{+} J ϕ ( t ) : A → R + as in (1.9). In order to optimise our error terms, we make some observations about the function α ϕ ( t ) \alpha_\phi(t) α ϕ ( t ) .
Lemma 4.1. Let α ϕ ( t ) \alpha_\phi(t) α ϕ ( t ) be as defined in (4.28). Then α ϕ ( t ) ≥ 0 \alpha_\phi(t)\ge0 α ϕ ( t ) ≥ 0 for every monic t ∈ A t\in A t ∈ A .
Proof. We first note that α ( 1 ) = 0 \alpha(1)=0 α ( 1 ) = 0 . By the multiplicativity of J ϕ ( t ) J_\phi(t) J ϕ ( t ) , it follows that when a , b ∈ A a,b\in A a , b ∈ A are coprime, then α ( a b ) = α ( a ) + α ( b ) \alpha(ab)=\alpha(a)+\alpha(b) α ( ab ) = α ( a ) + α ( b ) . Hence, it suffices to show that α ( p k ) ≥ 0 \alpha(p^k)\ge0 α ( p k ) ≥ 0 for every prime p p p and k ∈ N k\in\mathbb{N} k ∈ N . Then, by (1.9) and (4.18),
J ϕ ( p k ) = 1 + ∑ i = 1 k ( q i deg p − q ( i − 1 ) deg p ) q − ⌈ i / r ⌉ deg p . (71) J_\phi(p^k)=1+\sum_{i=1}^{k}\left(q^{i\deg p}-q^{(i-1)\deg p}\right)q^{-\lceil i/r\rceil\deg p}. \tag*{(71)} J ϕ ( p k ) = 1 + i = 1 ∑ k ( q i d e g p − q ( i − 1 ) d e g p ) q − ⌈ i / r ⌉ d e g p . (71)
Since ⌈ i / r ⌉ ≥ 1 \lceil i/r\rceil\ge1 ⌈ i / r ⌉ ≥ 1 for every 1 ≤ i ≤ k 1\le i\le k 1 ≤ i ≤ k , (4.29) is bounded above by
1 + q − deg p ∑ i = 1 k ( q i deg p − q ( i − 1 ) deg p ) = 1 + q − deg p ( q k deg p − 1 ) . (72) 1+q^{-\deg p}\sum_{i=1}^{k}\left(q^{i\deg p}-q^{(i-1)\deg p}\right)=1+q^{-\deg p}\left(q^{k\deg p}-1\right). \tag*{(72)} 1 + q − d e g p i = 1 ∑ k ( q i d e g p − q ( i − 1 ) d e g p ) = 1 + q − d e g p ( q k d e g p − 1 ) . (72)
We consider the difference q k deg p − J ϕ ( p k ) q^{k\deg p}-J_\phi(p^k) q k d e g p − J ϕ ( p k ) . By (4.29) and (4.30), we have that
q k deg p − J ϕ ( p k ) ≥ q k deg p − q ( k − 1 ) deg p + q − deg p − 1 = ( q ( k − 1 ) deg p − q − deg p ) ( q deg p − 1 ) ≥ 0. \begin{aligned}
q^{k\deg p}-J_\phi(p^k)\ge q^{k\deg p}-q^{(k-1)\deg p}+q^{-\deg p}-1 \\
&=\left(q^{(k-1)\deg p}-q^{-\deg p}\right)\left(q^{\deg p}-1\right)\ge0.
\end{aligned} q k d e g p − J ϕ ( p k ) ≥ q k d e g p − q ( k − 1 ) d e g p + q − d e g p − 1 = ( q ( k − 1 ) d e g p − q − d e g p ) ( q d e g p − 1 ) ≥ 0.
As a result, we obtain q k deg p ≥ J ϕ ( p k ) q^{k\deg p}\ge J_\phi(p^k) q k d e g p ≥ J ϕ ( p k ) and so α ( p k ) = log q ( q k deg p J ϕ ( p k ) ) ≥ 0 \alpha(p^k)=\log_q\left(\frac{q^{k\deg p}}{J_\phi(p^k)}\right)\ge0 α ( p k ) = log q ( J ϕ ( p k ) q k d e g p ) ≥ 0 . □
:::
Choice of parameter y y y
The displayed error bounds are compared up to bounded factors. We note that E 3 a E_{3a} E 3 a is independent of y ( x ) y(x) y ( x ) . Furthermore, as y ( x ) y(x) y ( x ) increases, E 1 E_1 E 1 increases whereas E 2 E_2 E 2 and E 3 b E_{3b} E 3 b decrease. So, an optimal choice of y ( x ) y(x) y ( x ) must come from comparing the cases E 1 = E 2 E_1=E_2 E 1 = E 2 and E 1 = E 3 b E_1=E_{3b} E 1 = E 3 b . We define
θ r , ρ : = 3 r 2 − ( ρ + 2 ) r − ρ 2 r 2 , Ξ r , ρ ( x , t ) : = θ r , ρ x + ρ ( r + 1 ) r 2 log q ( x + deg t log x ) . \theta_{r,\rho}:=\frac{3r^2-(\rho+2)r-\rho}{2r^2}, \qquad\Xi_{r,\rho}(x,t):=\theta_{r,\rho}x+\frac{\rho(r+1)}{r^2}\log_q\left(\frac{x+\deg t}{\log x}\right). θ r , ρ := 2 r 2 3 r 2 − ( ρ + 2 ) r − ρ , Ξ r , ρ ( x , t ) := θ r , ρ x + r 2 ρ ( r + 1 ) log q ( log x x + deg t ) .
Case 1: Let E 1 = E 2 E_1=E_2 E 1 = E 2 . Then, a suitable choice of y y y would be
y 1 ( x ) : = ρ r 2 ( x 2 − log q ( x + deg t ) + log q log x ) . y_1(x):=\frac{\rho}{r^2}\left(\frac{x}{2}-\log_q(x+\deg t)+\log_q\log x\right). y 1 ( x ) := r 2 ρ ( 2 x − log q ( x + deg t ) + log q log x ) .
One can check that for y = y 1 ( x ) y=y_1(x) y = y 1 ( x ) , when α ϕ ( t ) < Ξ r , ρ ( x , t ) \alpha_\phi(t)<\Xi_{r,\rho}(x,t) α ϕ ( t ) < Ξ r , ρ ( x , t ) , E 3 b E_{3b} E 3 b dominates over E 1 = E 2 E_1=E_2 E 1 = E 2 and hence
E 1 + E 2 + E 3 a + E 3 b = O K ( x r q r − 1 r x ) + O ϕ , K ( q 4 r − ρ − 2 2 r x − α ϕ ( t ) ( x + deg t log x ) ρ / r ) . E_1+E_2+E_{3a}+E_{3b}=O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{4r-\rho-2}{2r}x-\alpha_\phi(t)}\left(\frac{x+\deg t}{\log x}\right)^{\rho/r}\right). E 1 + E 2 + E 3 a + E 3 b = O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r 4 r − ρ − 2 x − α ϕ ( t ) ( log x x + deg t ) ρ / r ) .
and when α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) \alpha_\phi(t)\geq\Xi_{r,\rho}(x,t) α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) , then E 1 = E 2 E_1=E_2 E 1 = E 2 dominates over E 3 b E_{3b} E 3 b and hence
E 1 + E 2 + E 3 a + E 3 b = O K ( x r q r − 1 r x ) + O ϕ , K ( q r 2 + 2 ρ 2 r 2 x ( log x x + deg t ) ρ / r 2 ( 1 + deg t x ) ) . E_1+E_2+E_{3a}+E_{3b}=O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{r^2+2\rho}{2r^2}x}\left(\frac{\log x}{x+\deg t}\right)^{\rho/r^2}\left(1+\frac{\deg t}{x}\right)\right). E 1 + E 2 + E 3 a + E 3 b = O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r 2 r 2 + 2 ρ x ( x + deg t log x ) ρ / r 2 ( 1 + x deg t ) ) .
Case 2: Let E 1 = E 3 b E_1=E_{3b} E 1 = E 3 b . The following comparisons assume y 2 ( x ) ≤ x / r y_2(x)\leq x/r y 2 ( x ) ≤ x / r and y 2 ( x ) → ∞ y_2(x)\to\infty y 2 ( x ) → ∞ . Then, a suitable choice of y y y would be
y 2 ( x ) : = 1 r + 1 ( 3 r − 2 2 r x − α ϕ ( t ) − log q ( 1 + deg t x ) ) . y_2(x):=\frac{1}{r+1}\left(\frac{3r-2}{2r}x-\alpha_\phi(t)-\log_q\left(1+\frac{\deg t}{x}\right)\right). y 2 ( x ) := r + 1 1 ( 2 r 3 r − 2 x − α ϕ ( t ) − log q ( 1 + x deg t ) ) .
One can check that for y = y 2 ( x ) y=y_2(x) y = y 2 ( x ) , when α ϕ ( t ) < Ξ r , ρ ( x , t ) \alpha_\phi(t)<\Xi_{r,\rho}(x,t) α ϕ ( t ) < Ξ r , ρ ( x , t ) , then E 1 = E 3 b E_1=E_{3b} E 1 = E 3 b dominates over E 2 E_2 E 2 and hence
E 1 + E 2 + E 3 a + E 3 b = O K ( x r q r − 1 r x ) + O ϕ , K ( q r 2 + 4 r − 2 2 r ( 1 + r ) x − α ϕ ( t ) r + 1 ( 1 + deg t x ) r r + 1 ) . E_1+E_2+E_{3a}+E_{3b}=O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{r^2+4r-2}{2r(1+r)}x-\frac{\alpha_\phi(t)}{r+1}}\left(1+\frac{\deg t}{x}\right)^{\frac{r}{r+1}}\right). E 1 + E 2 + E 3 a + E 3 b = O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r ( 1 + r ) r 2 + 4 r − 2 x − r + 1 α ϕ ( t ) ( 1 + x deg t ) r + 1 r ) .
and when α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) \alpha_\phi(t)\geq\Xi_{r,\rho}(x,t) α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) , then E 2 E_2 E 2 dominates over E 1 = E 3 b E_1=E_{3b} E 1 = E 3 b and hence
E 1 + E 2 + E 3 a + E 3 b = O K ( x r q r − 1 r x ) + O ϕ , K ( q − 3 r 3 + ( 2 ρ + 2 ) r 2 + 5 ρ r − 2 ρ 2 ρ r ( 1 + r ) x + r 2 − ρ ρ ( r + 1 ) α ϕ ( t ) ( 1 + deg t x ) r 2 − ρ ρ ( r + 1 ) log y 2 ( x ) x ) . E_1+E_2+E_{3a}+E_{3b}=O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{-3r^3+(2\rho+2)r^2+5\rho r-2\rho}{2\rho r(1+r)}x+\frac{r^2-\rho}{\rho(r+1)}\alpha_\phi(t)}\left(1+\frac{\deg t}{x}\right)^{\frac{r^2-\rho}{\rho(r+1)}}\frac{\log y_2(x)}{x}\right). E 1 + E 2 + E 3 a + E 3 b = O K ( r x q r r − 1 x ) + O ϕ , K q 2 ρ r ( 1 + r ) − 3 r 3 + ( 2 ρ + 2 ) r 2 + 5 ρ r − 2 ρ x + ρ ( r + 1 ) r 2 − ρ α ϕ ( t ) ( 1 + x deg t ) ρ ( r + 1 ) r 2 − ρ x log y 2 ( x ) .
For sufficiently large x x x , we have 0 < y 1 ( x ) ≤ x / r 0<y_1(x)\leq x/r 0 < y 1 ( x ) ≤ x / r . In the middle range below, y 2 ( x ) ≤ x / r y_2(x)\leq x/r y 2 ( x ) ≤ x / r and y 2 ( x ) → ∞ y_2(x)\to\infty y 2 ( x ) → ∞ as x → ∞ x\to\infty x → ∞ . Comparing the exponents of q q q in the two cases, we choose
y ( x ) : = { x r when α ϕ ( t ) ≤ r − 4 2 r x , y 2 ( x ) when r − 4 2 r x < α ϕ ( t ) < Ξ r , ρ ( x , t ) , y 1 ( x ) when α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) . y(x):=
\begin{cases}
\frac{x}{r} & \text{when }\alpha_\phi(t)\leq\frac{r-4}{2r}x,\\
y_2(x) & \text{when }\frac{r-4}{2r}x<\alpha_\phi(t)<\Xi_{r,\rho}(x,t),\\
y_1(x) & \text{when }\alpha_\phi(t)\geq\Xi_{r,\rho}(x,t).
\end{cases} y ( x ) := ⎩ ⎨ ⎧ r x y 2 ( x ) y 1 ( x ) when α ϕ ( t ) ≤ 2 r r − 4 x , when 2 r r − 4 x < α ϕ ( t ) < Ξ r , ρ ( x , t ) , when α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) .
Thus, with this choice of y ( x ) y(x) y ( x ) , when α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) \alpha_\phi(t)\geq\Xi_{r,\rho}(x,t) α ϕ ( t ) ≥ Ξ r , ρ ( x , t ) , we obtain
π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O K ( x r q r − 1 r x ) + O ϕ , K ( q r 2 + 2 ρ 2 r 2 x ( log x x + deg t ) ρ / r 2 ( 1 + deg t x ) ) , \pi_\phi(x;t,\mathfrak a)=\pi(x)\cdot\delta_{\phi,t,\mathfrak a}(x)+O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{r^2+2\rho}{2r^2}x}\left(\frac{\log x}{x+\deg t}\right)^{\rho/r^2}\left(1+\frac{\deg t}{x}\right)\right), π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r 2 r 2 + 2 ρ x ( x + deg t log x ) ρ / r 2 ( 1 + x deg t ) ) ,
and when r − 4 2 r x < α ϕ ( t ) < Ξ r , ρ ( x , t ) \frac{r-4}{2r}x<\alpha_\phi(t)<\Xi_{r,\rho}(x,t) 2 r r − 4 x < α ϕ ( t ) < Ξ r , ρ ( x , t ) , we obtain
π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O K ( x r q r − 1 r x ) + O ϕ , K ( q r 2 + 4 r − 2 2 r ( 1 + r ) x − α ϕ ( t ) r + 1 ( 1 + deg t x ) r r + 1 ) , \pi_\phi(x;t,\mathfrak a)=\pi(x)\cdot\delta_{\phi,t,\mathfrak a}(x)+O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{r^2+4r-2}{2r(1+r)}x-\frac{\alpha_\phi(t)}{r+1}}\left(1+\frac{\deg t}{x}\right)^{\frac{r}{r+1}}\right), π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r ( 1 + r ) r 2 + 4 r − 2 x − r + 1 α ϕ ( t ) ( 1 + x deg t ) r + 1 r ) ,
Lastly, when α ϕ ( t ) ≤ r − 4 2 x \alpha_{\phi}(t) \le\frac{r-4}{2}x α ϕ ( t ) ≤ 2 r − 4 x , we choose y ( x ) = x / r y(x)=x/r y ( x ) = x / r which makes the sum ∑ e r r o r \sum_{\mathrm{error}} ∑ error in (4.3) vanish. Hence (4.27) reduces to π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + E 1 + E 2 \pi_{\phi}(x;t,\mathfrak{a})=\pi(x)\cdot\delta_{\phi,t,\mathfrak{a}}(x)+E_1+E_2 π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + E 1 + E 2 , which, on using ρ ≤ r \rho\le r ρ ≤ r , yields
π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O ϕ , K ( q x 2 + x r ( 1 + deg t x ) ) . \pi_{\phi}(x;t,\mathfrak{a})=\pi(x)\cdot\delta_{\phi,t,\mathfrak{a}}(x)+O_{\phi,K}\left(q^{\frac{x}{2}+\frac{x}{r}}\left(1+\frac{\deg t}{x}\right)\right). π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O ϕ , K ( q 2 x + r x ( 1 + x deg t ) ) .
This proves Theorem 1.1.
The Case when End K ( ϕ ) = A \operatorname{End}_{K}(\phi)=A End K ( ϕ ) = A
Preliminaries for Theorem 1.2
Let ϕ \phi ϕ be a Drinfeld A A A -module over K K K . Throughout this subsection, we assume that the endomorphism ring End K ( ϕ ) = A \operatorname{End}_{K}(\phi)=A End K ( ϕ ) = A . We say that Galois extensions L 1 / K L_1/K L 1 / K and L 2 / K L_2/K L 2 / K are linearly disjoint over K K K when Gal ( L 1 L 2 / K ) ≅ Gal ( L 1 / K ) × Gal ( L 2 / K ) \operatorname{Gal}(L_1L_2/K)\cong\operatorname{Gal}(L_1/K)\times\operatorname{Gal}(L_2/K) Gal ( L 1 L 2 / K ) ≅ Gal ( L 1 / K ) × Gal ( L 2 / K ) . The following lemma is a corollary to the work of Pink and Rütsche [31 Theorem 0.1] .
Lemma 5.1. [23 Corollary 9] There exists M ( ϕ ) ∈ A M(\phi)\in A M ( ϕ ) ∈ A such that
if gcd ( m , M ( ϕ ) ) = 1 \gcd(\mathfrak{m},M(\phi))=1 g cd( m , M ( ϕ )) = 1 , then Gal ( K ( ϕ [ m ] ) / K ) ≅ GL r ( A / m A ) \operatorname{Gal}(K(\phi[\mathfrak{m}])/K)\cong\operatorname{GL}_{r}(A/\mathfrak{m}A) Gal ( K ( ϕ [ m ]) / K ) ≅ GL r ( A / m A ) ;
if m = m 1 m 2 \mathfrak{m}=\mathfrak{m}_1\mathfrak{m}_2 m = m 1 m 2 , where m 1 \mathfrak{m}_1 m 1 is composed of primes that divide M ( ϕ ) M(\phi) M ( ϕ ) and m 2 \mathfrak{m}_2 m 2 is coprime to M ( ϕ ) M(\phi) M ( ϕ ) , then K ( ϕ [ m 1 ] ) K(\phi[\mathfrak{m}_1]) K ( ϕ [ m 1 ]) and K ( ϕ [ m 2 ] ) K(\phi[\mathfrak{m}_2]) K ( ϕ [ m 2 ]) are linearly disjoint over K K K ;
if gcd ( m 1 , M ( ϕ ) ) = gcd ( m 2 , M ( ϕ ) ) = gcd ( m 1 , m 2 ) = 1 \gcd(\mathfrak{m}_1,M(\phi))=\gcd(\mathfrak{m}_2,M(\phi))=\gcd(\mathfrak{m}_1,\mathfrak{m}_2)=1 g cd( m 1 , M ( ϕ )) = g cd( m 2 , M ( ϕ )) = g cd( m 1 , m 2 ) = 1 , then K ( ϕ [ m 1 ] ) K(\phi[\mathfrak{m}_1]) K ( ϕ [ m 1 ]) and K ( ϕ [ m 2 ] ) K(\phi[\mathfrak{m}_2]) K ( ϕ [ m 2 ]) are linearly disjoint over K K K .
For the proof of Theorem 1.4, we will need to append primes of degree 1 to M ( ϕ ) M(\phi) M ( ϕ ) , by defining
N ( ϕ ) : = lcm ( M ( ϕ ) , ∏ deg p = 1 p ) (73) N(\phi):=\operatorname{lcm}\left(M(\phi),\prod_{\deg\mathfrak{p}=1}\mathfrak{p}\right) \tag*{(73)} N ( ϕ ) := lcm M ( ϕ ) , d e g p = 1 ∏ p (73)
Then, note that all the points of Lemma 5.1 still hold if we replace M ( ϕ ) M(\phi) M ( ϕ ) by N ( ϕ ) N(\phi) N ( ϕ ) .
Lemma 5.2. Let N ( ϕ ) N(\phi) N ( ϕ ) be as defined in (5.1). Then, we have that
if gcd ( m , N ( ϕ ) ) = 1 \gcd(\mathfrak{m},N(\phi))=1 g cd( m , N ( ϕ )) = 1 , then Gal ( K ( ϕ [ m ] ) / K ) ≅ GL r ( A / m A ) \operatorname{Gal}(K(\phi[\mathfrak{m}])/K)\cong\operatorname{GL}_{r}(A/\mathfrak{m}A) Gal ( K ( ϕ [ m ]) / K ) ≅ GL r ( A / m A ) ;
if m = m 1 m 2 \mathfrak{m}=\mathfrak{m}_1\mathfrak{m}_2 m = m 1 m 2 , where m 1 \mathfrak{m}_1 m 1 is composed of primes that divide N ( ϕ ) N(\phi) N ( ϕ ) and m 2 \mathfrak{m}_2 m 2 is coprime to N ( ϕ ) N(\phi) N ( ϕ ) , then K ( ϕ [ m 1 ] ) K(\phi[\mathfrak{m}_1]) K ( ϕ [ m 1 ]) and K ( ϕ [ m 2 ] ) K(\phi[\mathfrak{m}_2]) K ( ϕ [ m 2 ]) are linearly disjoint over K K K ;
if gcd ( m 1 , N ( ϕ ) ) = gcd ( m 2 , N ( ϕ ) ) = gcd ( m 1 , m 2 ) = 1 \gcd(\mathfrak{m}_1,N(\phi))=\gcd(\mathfrak{m}_2,N(\phi))=\gcd(\mathfrak{m}_1,\mathfrak{m}_2)=1 g cd( m 1 , N ( ϕ )) = g cd( m 2 , N ( ϕ )) = g cd( m 1 , m 2 ) = 1 , then K ( ϕ [ m 1 ] ) K(\phi[\mathfrak{m}_1]) K ( ϕ [ m 1 ]) and K ( ϕ [ m 2 ] ) K(\phi[\mathfrak{m}_2]) K ( ϕ [ m 2 ]) are linearly disjoint over K K K .
Proof. We note that for any a ∈ A a\in A a ∈ A with gcd ( a , N ( ϕ ) ) = 1 \gcd(a,N(\phi))=1 g cd( a , N ( ϕ )) = 1 , we also have gcd ( a , M ( ϕ ) ) = 1 \gcd(a,M(\phi))=1 g cd( a , M ( ϕ )) = 1 . As a result, parts (1) and (3) follow from parts (1) and (3) of Lemma 5.1.
We now proceed to prove (2). Note that for coprime a , b ∈ A a,b\in A a , b ∈ A , we have that K ( ϕ [ a b ] ) = K ( ϕ [ a ] ) K ( ϕ [ b ] ) K(\phi[ab])=K(\phi[a])K(\phi[b]) K ( ϕ [ ab ]) = K ( ϕ [ a ]) K ( ϕ [ b ]) . Let m = m 1 m 2 \mathfrak{m}=\mathfrak{m}_1\mathfrak{m}_2 m = m 1 m 2 , where m 1 \mathfrak{m}_1 m 1 is made up of prime divisors of N ( ϕ ) N(\phi) N ( ϕ ) and gcd ( m 2 , N ( ϕ ) ) = 1 \gcd(\mathfrak{m}_2,N(\phi))=1 g cd( m 2 , N ( ϕ )) = 1 . As a result, gcd ( m 2 , M ( ϕ ) ) = 1 \gcd(\mathfrak{m}_2,M(\phi))=1 g cd( m 2 , M ( ϕ )) = 1 as well. Also, let m 1 = m 11 m 12 \mathfrak{m}_1=\mathfrak{m}_{11}\mathfrak{m}_{12} m 1 = m 11 m 12 , where m 11 \mathfrak{m}_{11} m 11 is made up of prime divisors of M ( ϕ ) M(\phi) M ( ϕ ) and gcd ( m 12 , M ( ϕ ) ) = 1 \gcd(\mathfrak{m}_{12},M(\phi))=1 g cd( m 12 , M ( ϕ )) = 1 . Note that ( m 11 , m 12 ) = 1 (\mathfrak{m}_{11},\mathfrak{m}_{12})=1 ( m 11 , m 12 ) = 1 . Then,
(i) by part (2) of Lemma 5.1, K ( ϕ [ m 11 ] ) K(\phi[\mathfrak{m}_{11}]) K ( ϕ [ m 11 ]) and K ( ϕ [ m 12 m 2 ] ) K(\phi[\mathfrak{m}_{12}\mathfrak{m}_2]) K ( ϕ [ m 12 m 2 ]) are linearly disjoint over K K K ;
(ii) by part (3) of Lemma 5.1, K ( ϕ [ m 12 ] ) K(\phi[\mathfrak{m}_{12}]) K ( ϕ [ m 12 ]) and K ( ϕ [ m 2 ] ) K(\phi[\mathfrak{m}_2]) K ( ϕ [ m 2 ]) are linearly disjoint over K K K ;
(iii) by part (2) of Lemma 5.1,
Gal ( K ( ϕ [ m 11 ] ) / K ) × Gal ( K ( ϕ [ m 12 ] ) / K ) ≅ Gal ( K ( ϕ [ m 1 ] ) / K ) . \operatorname{Gal}(K(\phi[\mathfrak{m}_{11}])/K)\times\operatorname{Gal}(K(\phi[\mathfrak{m}_{12}])/K)\cong\operatorname{Gal}(K(\phi[\mathfrak{m}_1])/K). Gal ( K ( ϕ [ m 11 ]) / K ) × Gal ( K ( ϕ [ m 12 ]) / K ) ≅ Gal ( K ( ϕ [ m 1 ]) / K ) .
Hence, we have that K ( ϕ [ m 1 ] ) K(\phi[\mathfrak{m}_1]) K ( ϕ [ m 1 ]) and K ( ϕ [ m 2 ] ) K(\phi[\mathfrak{m}_2]) K ( ϕ [ m 2 ]) are linearly disjoint over K K K .
:::
In the course of the proof, we come across intersections of the form K ( ϕ [ m ] ) ∩ K ( ψ , C a l [ t ] ) K(\phi[\mathfrak{m}])\cap K(\psi,\mathcal{C}\!al[\mathfrak{t}]) K ( ϕ [ m ]) ∩ K ( ψ , C a l [ t ]) . These can be explicitly characterized for m \mathfrak{m} m coprime to N ( ϕ ) N(\phi) N ( ϕ ) . We first compute the commutator subgroup of GL r ( A / m A ) \operatorname{GL}_{r}(A/\mathfrak{m}A) GL r ( A / m A ) , which will be useful later.
Lemma 5.3. Let m ∈ A \mathfrak{m}\in A m ∈ A . Then for r ≥ 3 r\ge3 r ≥ 3 , the commutator subgroup
[ GL r ( A / m A ) , GL r ( A / m A ) ] = SL r ( A / m A ) . [\operatorname{GL}_{r}(A/\mathfrak{m}A),\operatorname{GL}_{r}(A/\mathfrak{m}A)]=\operatorname{SL}_{r}(A/\mathfrak{m}A). [ GL r ( A / m A ) , GL r ( A / m A )] = SL r ( A / m A ) .
Proof. Since A A A is a Principal Ideal Domain, by the Chinese Remainder Theorem it suffices to show
H r : = [ GL r ( A / p k A ) , GL r ( A / p k A ) ] = SL r ( A / p k A ) , H_r := [\operatorname{GL}_r(A/\mathfrak{p}^k A),\operatorname{GL}_r(A/\mathfrak{p}^k A)] = \operatorname{SL}_r(A/\mathfrak{p}^k A), H r := [ GL r ( A / p k A ) , GL r ( A / p k A )] = SL r ( A / p k A ) ,
for all primes p ∈ A \mathfrak{p} \in A p ∈ A and k ∈ N k \in\mathbb{N} k ∈ N . It is clear that H r ⊆ SL r ( A / p k A ) H_r \subseteq\operatorname{SL}_r(A/\mathfrak{p}^k A) H r ⊆ SL r ( A / p k A ) since every element in the commutator H r H_r H r has determinant 1. We will now show that H r ⊇ SL r ( A / p k A ) H_r \supseteq\operatorname{SL}_r(A/\mathfrak{p}^k A) H r ⊇ SL r ( A / p k A ) .
For 1 ≤ i , j ≤ r 1 \leq i,j \leq r 1 ≤ i , j ≤ r with i ≠ j i \neq j i = j and λ ∈ A / p k A \lambda\in A/\mathfrak{p}^k A λ ∈ A / p k A , let E i j ( λ ) E_{ij}(\lambda) E ij ( λ ) denote the elementary r × r r \times r r × r matrix which has 1’s along the diagonal, λ \lambda λ at the ( i , j ) (i,j) ( i , j ) -th entry and 0’s elsewhere. Since R = A / p k A R = A/\mathfrak{p}^k A R = A / p k A is a local ring, these matrices generate SL r ( R ) \operatorname{SL}_r(R) SL r ( R ) (see [13 Lemma 2.1(iii)] ). By [13 Lemma 2.1(i)(b)] , for r > 2 r > 2 r > 2 , and 1 ≤ i , j , k ≤ r 1 \leq i,j,k \leq r 1 ≤ i , j , k ≤ r all distinct, we have the identity [ E i j ( α ) , E j k ( β ) ] = E i k ( α β ) [E_{ij}(\alpha), E_{jk}(\beta)] = E_{ik}(\alpha\beta) [ E ij ( α ) , E j k ( β )] = E ik ( α β ) . As a result, all elementary matrices E i k ( λ ) E_{ik}(\lambda) E ik ( λ ) belong to the commutator subgroup, and hence SL r ( A / p k A ) ⊆ H r \operatorname{SL}_r(A/\mathfrak{p}^k A) \subseteq H_r SL r ( A / p k A ) ⊆ H r , as needed.
Lemma 5.4
Lemma 5.4. Let m ∈ A \mathfrak{m} \in A m ∈ A be such that deg p ≥ 2 \deg\mathfrak{p} \geq2 deg p ≥ 2 for all primes p ∣ m \mathfrak{p} \mid\mathfrak{m} p ∣ m . Then, the commutator subgroup
[ GL 2 ( A / m A ) , GL 2 ( A / m A ) ] = SL 2 ( A / m A ) . [\operatorname{GL}_2(A/\mathfrak{m}A),\operatorname{GL}_2(A/\mathfrak{m}A)] = \operatorname{SL}_2(A/\mathfrak{m}A). [ GL 2 ( A / m A ) , GL 2 ( A / m A )] = SL 2 ( A / m A ) .
Proof. As in the beginning of the previous proof, we only need to show that SL 2 ( A / p k A ) ⊆ [ GL 2 ( A / p k A ) , GL 2 ( A / p k A ) ] \operatorname{SL}_2(A/\mathfrak{p}^k A) \subseteq[\operatorname{GL}_2(A/\mathfrak{p}^k A),\operatorname{GL}_2(A/\mathfrak{p}^k A)] SL 2 ( A / p k A ) ⊆ [ GL 2 ( A / p k A ) , GL 2 ( A / p k A )] . Consider the quotient map
ξ : A / p k A ⟶ A / p k A p A / p k A ≅ A p A . \xi: A/\mathfrak{p}^k A \longrightarrow\frac{A/\mathfrak{p}^k A}{\mathfrak{p}A/\mathfrak{p}^k A} \cong\frac{A}{\mathfrak{p}A}. ξ : A / p k A ⟶ p A / p k A A / p k A ≅ p A A .
Since deg ( p ) ≥ 2 \deg(\mathfrak{p}) \geq2 deg ( p ) ≥ 2 , the field A / p A A/\mathfrak{p}A A / p A has size q deg p ≥ 4 q^{\deg\mathfrak{p}} \geq4 q d e g p ≥ 4 . Thus, there exists a β ′ ∈ A / p A \beta' \in A/\mathfrak{p}A β ′ ∈ A / p A such that β ′ ≠ 0 \beta' \neq0 β ′ = 0 and β ′ − 1 ≠ 0 \beta' - 1 \neq0 β ′ − 1 = 0 . Let β ∈ ξ − 1 ( β ′ ) \beta\in\xi^{-1}(\beta') β ∈ ξ − 1 ( β ′ ) be a preimage of β ′ \beta' β ′ . Then β ∈ A / p k A \beta\in A/\mathfrak{p}^k A β ∈ A / p k A is such that β ∉ p A / p k A \beta\notin\mathfrak{p}A/\mathfrak{p}^k A β ∈ / p A / p k A and β − 1 ∉ p A / p k A \beta- 1 \notin\mathfrak{p}A/\mathfrak{p}^k A β − 1 ∈ / p A / p k A .
Since p A / p k A \mathfrak{p}A/\mathfrak{p}^k A p A / p k A is the unique maximal ideal of A / p k A A/\mathfrak{p}^k A A / p k A , we obtain that β \beta β and β − 1 \beta- 1 β − 1 are units in A / p k A A/\mathfrak{p}^k A A / p k A . Then for any λ ∈ A / p k A \lambda\in A/\mathfrak{p}^k A λ ∈ A / p k A , we have
E 12 ( λ ) = [ Diag ( β , 1 ) , E 12 ( ( β − 1 ) − 1 λ ) ] , (74) E_{12}(\lambda) = [\operatorname{Diag}(\beta,1),E_{12}((\beta-1)^{-1}\lambda)], \tag*{(74)} E 12 ( λ ) = [ Diag ( β , 1 ) , E 12 (( β − 1 ) − 1 λ )] , (74)
where Diag ( β , 1 ) \operatorname{Diag}(\beta,1) Diag ( β , 1 ) is the matrix
( β 0 0 1 ) . \begin{pmatrix}
\beta& 0\\
0 & 1
\end{pmatrix}. ( β 0 0 1 ) .
By taking transpose on (74) , we see that E 21 ( λ ) E_{21}(\lambda) E 21 ( λ ) is also a commutator. By [13 Lemma 2.1(iii)] , SL 2 ( A / p k A ) \operatorname{SL}_2(A/\mathfrak{p}^k A) SL 2 ( A / p k A ) is generated by the matrices E 12 ( λ ) E_{12}(\lambda) E 12 ( λ ) , E 21 ( λ ) E_{21}(\lambda) E 21 ( λ ) , λ ∈ A / p k A \lambda\in A/\mathfrak{p}^k A λ ∈ A / p k A . Hence, SL 2 ( A / p k A ) ⊆ [ GL 2 ( A / p k A ) , GL 2 ( A / p k A ) ] \operatorname{SL}_2(A/\mathfrak{p}^k A) \subseteq[\operatorname{GL}_2(A/\mathfrak{p}^k A),\operatorname{GL}_2(A/\mathfrak{p}^k A)] SL 2 ( A / p k A ) ⊆ [ GL 2 ( A / p k A ) , GL 2 ( A / p k A )] .
Recall that ϕ 1 \phi^1 ϕ 1 , as defined in (16) , denotes the rank 1 Drinfeld A A A -module over K K K associated to ϕ \phi ϕ via the Weil pairing of Tate modules.
Lemma 5.5
Lemma 5.5. Let End K ( ϕ ) = A \operatorname{End}_K(\phi) = A End K ( ϕ ) = A . Let N ( ϕ ) N(\phi) N ( ϕ ) be as defined in (73) . Let m ∈ A \mathfrak{m} \in A m ∈ A with gcd ( m , N ( ϕ ) ) = 1 \gcd(\mathfrak{m},N(\phi)) = 1 g cd( m , N ( ϕ )) = 1 . Then the maximal abelian Galois extension of K K K contained in K ( ϕ [ m ] ) K(\phi[\mathfrak{m}]) K ( ϕ [ m ]) is K ( ϕ 1 [ m ] ) K(\phi^1[\mathfrak{m}]) K ( ϕ 1 [ m ]) .
Proof. Let L L L be the maximal abelian Galois extension of K K K in K ( ϕ [ m ] ) K(\phi[\mathfrak{m}]) K ( ϕ [ m ]) . We have the group isomorphism
Gal ( L / K ) ≅ Gal ( K ( ϕ [ m ] ) / K ) [ Gal ( K ( ϕ [ m ] ) / K ) , Gal ( K ( ϕ [ m ] ) / K ) ] . \operatorname{Gal}(L/K) \cong\frac{\operatorname{Gal}(K(\phi[\mathfrak{m}])/K)}{[\operatorname{Gal}(K(\phi[\mathfrak{m}])/K),\operatorname{Gal}(K(\phi[\mathfrak{m}])/K)]}. Gal ( L / K ) ≅ [ Gal ( K ( ϕ [ m ]) / K ) , Gal ( K ( ϕ [ m ]) / K )] Gal ( K ( ϕ [ m ]) / K ) .
Since gcd ( m , N ( ϕ ) ) = 1 \gcd(\mathfrak{m},N(\phi)) = 1 g cd( m , N ( ϕ )) = 1 , we have that Gal ( K ( ϕ [ m ] ) / K ) ≅ GL r ( A / m A ) \operatorname{Gal}(K(\phi[\mathfrak{m}])/K) \cong\operatorname{GL}_r(A/\mathfrak{m}A) Gal ( K ( ϕ [ m ]) / K ) ≅ GL r ( A / m A ) by Lemma 5.2(1). By the construction of N ( ϕ ) N(\phi) N ( ϕ ) (see (73) ) and Lemmas 5.3, 5.4, the commutator subgroup of GL r ( A / m A ) \operatorname{GL}_r(A/\mathfrak{m}A) GL r ( A / m A ) is SL r ( A / m A ) \operatorname{SL}_r(A/\mathfrak{m}A) SL r ( A / m A ) . Thus, Gal ( L / K ) ≅ ( GL r ( A / m A ) / SL r ( A / m A ) ) ≅ ( A / m A ) ∗ \operatorname{Gal}(L/K) \cong(\operatorname{GL}_r(A/\mathfrak{m}A)/\operatorname{SL}_r(A/\mathfrak{m}A)) \cong(A/\mathfrak{m}A)^* Gal ( L / K ) ≅ ( GL r ( A / m A ) / SL r ( A / m A )) ≅ ( A / m A ) ∗ . By [35 Theorem 3.1.3] , K ( ϕ 1 [ m ] ) ⊆ K ( ϕ [ m ] ) K(\phi^1[\mathfrak{m}]) \subseteq K(\phi[\mathfrak{m}]) K ( ϕ 1 [ m ]) ⊆ K ( ϕ [ m ]) is an abelian extension of K K K . Hence we must have K ( ϕ 1 [ m ] ) ⊆ L K(\phi^1[\mathfrak{m}]) \subseteq L K ( ϕ 1 [ m ]) ⊆ L . Moreover since Gal ( K ( ϕ 1 [ m ] ) / K ) ≅ ( A / m A ) ∗ \operatorname{Gal}(K(\phi^1[\mathfrak{m}])/K) \cong(A/\mathfrak{m}A)^* Gal ( K ( ϕ 1 [ m ]) / K ) ≅ ( A / m A ) ∗ (again by [35 Theorem 3.1.3] ), we see that L = K ( ϕ 1 [ m ] ) L = K(\phi^1[\mathfrak{m}]) L = K ( ϕ 1 [ m ]) .
Lemma 5.6
Lemma 5.6. Let ϕ \phi ϕ be an arbitrary Drinfeld A A A -module over K K K . For any a , b ∈ A \mathfrak{a}, \mathfrak{b} \in A a , b ∈ A , we have that the compositum K ( ϕ [ a ] ) K ( ϕ [ b ] ) = K ( ϕ [ lcm ( a , b ) ] ) K(\phi[\mathfrak{a}])K(\phi[\mathfrak{b}]) = K(\phi[\operatorname{lcm}(\mathfrak{a},\mathfrak{b})]) K ( ϕ [ a ]) K ( ϕ [ b ]) = K ( ϕ [ lcm ( a , b )]) , where lcm ( a , b ) \operatorname{lcm}(\mathfrak{a},\mathfrak{b}) lcm ( a , b ) is the least common multiple of a , b \mathfrak{a},\mathfrak{b} a , b in A A A . Proof. Recall that for a monic a ∈ A a \in A a ∈ A , the torsion group ϕ [ a ] = { α ∈ K ‾ : ϕ a ( α ) = 0 } \phi[a] = \{\alpha\in\overline{K} : \phi_a(\alpha) = 0\} ϕ [ a ] = { α ∈ K : ϕ a ( α ) = 0 } . Hence, we have the set inclusions ϕ [ a ] ⊆ ϕ [ lcm ( a , b ) ] \phi[a] \subseteq\phi[\operatorname{lcm}(a,b)] ϕ [ a ] ⊆ ϕ [ lcm ( a , b )] and ϕ [ b ] ⊆ ϕ [ lcm ( a , b ) ] \phi[b] \subseteq\phi[\operatorname{lcm}(a,b)] ϕ [ b ] ⊆ ϕ [ lcm ( a , b )] . Thus,
K ( ϕ [ a ] ) K ( ϕ [ b ] ) ⊆ K ( ϕ [ lcm ( a , b ) ] ) . K(\phi[a])K(\phi[b]) \subseteq K(\phi[\operatorname{lcm}(a,b)]). K ( ϕ [ a ]) K ( ϕ [ b ]) ⊆ K ( ϕ [ lcm ( a , b )]) .
To prove the reverse inclusion, let a ′ = lcm ( a , b ) / a a' = \operatorname{lcm}(a,b)/a a ′ = lcm ( a , b ) / a and b ′ = lcm ( a , b ) / b b' = \operatorname{lcm}(a,b)/b b ′ = lcm ( a , b ) / b . Then, as gcd ( a ′ , b ′ ) = 1 \gcd(a',b') = 1 g cd( a ′ , b ′ ) = 1 , there exist e , f ∈ A e, f \in A e , f ∈ A such that e a ′ + f b ′ = 1 ea' + fb' = 1 e a ′ + f b ′ = 1 . So, we have that for every α ∈ K ‾ \alpha\in\overline{K} α ∈ K ,
ϕ e ( ϕ a ′ ( α ) ) + ϕ f ( ϕ b ′ ( α ) ) = ϕ e a ′ + f b ′ ( α ) = α . (75) \phi_e(\phi_{a'}(\alpha)) + \phi_f(\phi_{b'}(\alpha)) = \phi_{ea'+fb'}(\alpha) = \alpha. \tag*{(75)} ϕ e ( ϕ a ′ ( α )) + ϕ f ( ϕ b ′ ( α )) = ϕ e a ′ + f b ′ ( α ) = α . (75)
Now, let α ∈ ϕ [ lcm ( a , b ) ] \alpha\in\phi[\operatorname{lcm}(a,b)] α ∈ ϕ [ lcm ( a , b )] be arbitrary. Then, we have
ϕ a ( ϕ e ( ϕ a ′ ( α ) ) ) = ϕ e ( ϕ lcm ( a , b ) ( α ) ) = ϕ e ( 0 ) = 0 \phi_a(\phi_e(\phi_{a'}(\alpha))) = \phi_e(\phi_{\operatorname{lcm}(a,b)}(\alpha)) = \phi_e(0) = 0 ϕ a ( ϕ e ( ϕ a ′ ( α ))) = ϕ e ( ϕ lcm ( a , b ) ( α )) = ϕ e ( 0 ) = 0
ϕ b ( ϕ f ( ϕ b ′ ( α ) ) ) = ϕ f ( ϕ lcm ( a , b ) ( α ) ) = ϕ f ( 0 ) = 0 \phi_b(\phi_f(\phi_{b'}(\alpha))) = \phi_f(\phi_{\operatorname{lcm}(a,b)}(\alpha)) = \phi_f(0) = 0 ϕ b ( ϕ f ( ϕ b ′ ( α ))) = ϕ f ( ϕ lcm ( a , b ) ( α )) = ϕ f ( 0 ) = 0
Hence, we obtain ϕ e ( ϕ a ′ ( α ) ) ∈ ϕ [ a ] \phi_e(\phi_{a'}(\alpha)) \in\phi[a] ϕ e ( ϕ a ′ ( α )) ∈ ϕ [ a ] and ϕ f ( ϕ b ′ ( α ) ) ∈ ϕ [ b ] \phi_f(\phi_{b'}(\alpha)) \in\phi[b] ϕ f ( ϕ b ′ ( α )) ∈ ϕ [ b ] , and as a result, by (5.3), α ∈ ϕ [ a ] + ϕ [ b ] \alpha\in\phi[a] + \phi[b] α ∈ ϕ [ a ] + ϕ [ b ] . Thus, K ( ϕ [ lcm ( a , b ) ] ) ⊆ K ( ϕ [ a ] ) K ( ϕ [ b ] ) K(\phi[\operatorname{lcm}(a,b)]) \subseteq K(\phi[a])K(\phi[b]) K ( ϕ [ lcm ( a , b )]) ⊆ K ( ϕ [ a ]) K ( ϕ [ b ]) .
Lemma 5.7. For any nonzero monic a , b ∈ A a,b \in A a , b ∈ A , we have K ( ψ C a r l [ a ] ) ∩ K ( ψ C a r l [ b ] ) = K ( ψ C a r l [ gcd ( a , b ) ] ) K(\psi^{\mathrm{Carl}}[a]) \cap K(\psi^{\mathrm{Carl}}[b]) = K(\psi^{\mathrm{Carl}}[\gcd(a,b)]) K ( ψ Carl [ a ]) ∩ K ( ψ Carl [ b ]) = K ( ψ Carl [ g cd( a , b )]) .
Proof. Since gcd ( a , b ) \gcd(a,b) g cd( a , b ) divides both a a a and b b b , we have
K ( ψ C a r l [ gcd ( a , b ) ] ) ⊆ K ( ψ C a r l [ a ] ) ∩ K ( ψ C a r l [ b ] ) . (76) K(\psi^{\mathrm{Carl}}[\gcd(a,b)]) \subseteq K(\psi^{\mathrm{Carl}}[a]) \cap K(\psi^{\mathrm{Carl}}[b]). \tag*{(76)} K ( ψ Carl [ g cd( a , b )]) ⊆ K ( ψ Carl [ a ]) ∩ K ( ψ Carl [ b ]) . (76)
On the other hand, by Lemma 5.6, K ( ψ C a r l [ a ] ) K ( ψ C a r l [ b ] ) = K ( ψ C a r l [ lcm ( a , b ) ] ) K(\psi^{\mathrm{Carl}}[a])K(\psi^{\mathrm{Carl}}[b]) = K(\psi^{\mathrm{Carl}}[\operatorname{lcm}(a,b)]) K ( ψ Carl [ a ]) K ( ψ Carl [ b ]) = K ( ψ Carl [ lcm ( a , b )]) . Hence,
[ K ( ψ C a r l [ a ] ) ∩ K ( ψ C a r l [ b ] ) : K ] = [ K ( ψ C a r l [ a ] ) : K ] [ K ( ψ C a r l [ b ] ) : K ] [ K ( ψ C a r l [ lcm ( a , b ) ] ) : K ] = φ ( a ) φ ( b ) φ ( lcm ( a , b ) ) = φ ( gcd ( a , b ) ) . \begin{aligned}
[K(\psi^{\mathrm{Carl}}[a]) \cap K(\psi^{\mathrm{Carl}}[b]) : K]
&= \frac{[K(\psi^{\mathrm{Carl}}[a]) : K][K(\psi^{\mathrm{Carl}}[b]) : K]}{[K(\psi^{\mathrm{Carl}}[\operatorname{lcm}(a,b)]) : K]} \\
&= \frac{\varphi(a)\varphi(b)}{\varphi(\operatorname{lcm}(a,b))} = \varphi(\gcd(a,b)).
\end{aligned} [ K ( ψ Carl [ a ]) ∩ K ( ψ Carl [ b ]) : K ] = [ K ( ψ Carl [ lcm ( a , b )]) : K ] [ K ( ψ Carl [ a ]) : K ] [ K ( ψ Carl [ b ]) : K ] = φ ( lcm ( a , b )) φ ( a ) φ ( b ) = φ ( g cd( a , b )) .
Since [ K ( ψ C a r l [ gcd ( a , b ) ] ) : K ] = φ ( gcd ( a , b ) ) [K(\psi^{\mathrm{Carl}}[\gcd(a,b)]) : K] = \varphi(\gcd(a,b)) [ K ( ψ Carl [ g cd( a , b )]) : K ] = φ ( g cd( a , b )) , the result follows by (5.4).
Before proving the next lemma, recall that for the rank 1 Drinfeld A A A -modules ψ C a r l \psi^{\mathrm{Carl}} ψ Carl and ϕ 1 \phi^1 ϕ 1 , by [35 Example 2.2.5, p.38] , there exists an isomorphism of Drinfeld modules given by
ψ C a r l ⟶ ϕ 1 , x ↦ u ⋅ x (77) \psi^{\mathrm{Carl}} \longrightarrow\phi^1,\qquad x \mapsto u \cdot x \tag*{(77)} ψ Carl ⟶ ϕ 1 , x ↦ u ⋅ x (77)
for some u ∈ K ‾ u \in\overline{K} u ∈ K satisfying u q − 1 ∈ K u^{q-1} \in K u q − 1 ∈ K .
Lemma 5.8. Let End K ( ϕ ) = A \operatorname{End}_K(\phi) = A End K ( ϕ ) = A . Let N ( ϕ ) N(\phi) N ( ϕ ) be as defined in (5.1). Let m ∈ A m \in A m ∈ A with gcd ( m , N ( ϕ ) ) = 1 \gcd(m,N(\phi)) = 1 g cd( m , N ( ϕ )) = 1 and let u u u be as in (5.5) above. Then, we have
[ K ( ϕ [ m ] ) ∩ K ( ψ C a r l [ t ] ) : K ] = d m , t c m , t φ ( gcd ( m , t ) ) , (78) [K(\phi[m]) \cap K(\psi^{\mathrm{Carl}}[t]) : K] = \frac{d_{m,t}}{c_{m,t}}\varphi(\gcd(m,t)), \tag*{(78)} [ K ( ϕ [ m ]) ∩ K ( ψ Carl [ t ]) : K ] = c m , t d m , t φ ( g cd( m , t )) , (78) where c m , t c_{m,t} c m , t and d m , t d_{m,t} d m , t are certain positive divisors of q − 1 q-1 q − 1 , given by
c m , t = [ K ( ψ C a r l [ lcm ( m , t ) ] ) ( u ) : K ( ψ C a r l [ lcm ( m , t ) ] ) ] , c_{m,t} = [K(\psi^{\mathrm{Carl}}[\operatorname{lcm}(m,t)])(u) : K(\psi^{\mathrm{Carl}}[\operatorname{lcm}(m,t)])], c m , t = [ K ( ψ Carl [ lcm ( m , t )]) ( u ) : K ( ψ Carl [ lcm ( m , t )])] , and
d m , t = [ K ( u ψ C a r l [ m ] ) K ( ψ C a r l [ t ] ) ( u ) : K ( u ψ C a r l [ m ] ) K ( ψ C a r l [ t ] ) ] . (79) d_{m,t} = [K(u\psi^{\mathrm{Carl}}[m])K(\psi^{\mathrm{Carl}}[t])(u) : K(u\psi^{\mathrm{Carl}}[m])K(\psi^{\mathrm{Carl}}[t])]. \tag*{(79)} d m , t = [ K ( u ψ Carl [ m ]) K ( ψ Carl [ t ]) ( u ) : K ( u ψ Carl [ m ]) K ( ψ Carl [ t ])] . (79)
Proof. Consider the diagram By [35 Theorem 3.1.3] , H : = K ( ψ C a r l [ t ] ) H := K(\psi^{\mathrm{Carl}}[t]) H := K ( ψ Carl [ t ]) is an abelian extension of K K K (in fact Gal ( H / K ) ≅ ( A / t A ) ∗ \operatorname{Gal}(H/K) \cong(A/tA)^* Gal ( H / K ) ≅ ( A / t A ) ∗ ). The abelian Galois extension ( K ( ϕ [ m ] ) ∩ H ) / K (K(\phi[m]) \cap H)/K ( K ( ϕ [ m ]) ∩ H ) / K must hence be contained inside L L L by Lemma 5.5. We thus have
Figure 1.
[ K ( ϕ [ m ] ) ∩ H : K ] = [ L ∩ H : K ] = [ L : K ] [ H : K ] [ L H : K ] = φ ( m ) φ ( t ) [ L H : K ] . (80) [K(\phi[m]) \cap H : K] = [L \cap H : K] = \frac{[L : K][H : K]}{[LH : K]} = \frac{\varphi(m)\varphi(t)}{[LH : K]}. \tag*{(80)} [ K ( ϕ [ m ]) ∩ H : K ] = [ L ∩ H : K ] = [ L H : K ] [ L : K ] [ H : K ] = [ L H : K ] φ ( m ) φ ( t ) . (80)
where L H = K ( ϕ 1 [ m ] ) K ( ψ C a r l [ t ] ) LH = K(\phi^1[m])K(\psi^{\mathrm{Carl}}[t]) L H = K ( ϕ 1 [ m ]) K ( ψ Carl [ t ]) .
The isomorphism (5.5) induces an isomorphism of the torsion points ψ C a r l [ m ] ≃ ϕ 1 [ m ] \psi^{\mathrm{Carl}}[m] \simeq\phi^1[m] ψ Carl [ m ] ≃ ϕ 1 [ m ] , over K ‾ \overline{K} K given by x ↦ u x x \mapsto ux x ↦ ux . As a consequence we have K ( ϕ 1 [ m ] ) = K ( u ⋅ ψ C a r l [ m ] ) K(\phi^1[m]) = K(u \cdot\psi^{\mathrm{Carl}}[m]) K ( ϕ 1 [ m ]) = K ( u ⋅ ψ Carl [ m ]) . Hence we have L H = K ( u ψ C a r l [ m ] ) K ( ψ C a r l [ t ] ) LH = K(u\psi^{\mathrm{Carl}}[m])K(\psi^{\mathrm{Carl}}[t]) L H = K ( u ψ Carl [ m ]) K ( ψ Carl [ t ]) . We claim that
L H ( u ) = K ( ψ C a r l [ m ] ) K ( ψ C a r l [ t ] ) ( u ) . (81) LH(u) = K(\psi^{\mathrm{Carl}}[m])K(\psi^{\mathrm{Carl}}[t])(u). \tag*{(81)} L H ( u ) = K ( ψ Carl [ m ]) K ( ψ Carl [ t ]) ( u ) . (81)
In order to show this, first observe that one has K ( u ψ C a r l [ m ] ) ⊆ K ( ψ C a r l [ m ] ) ( u ) K(u\psi^{\mathrm{Carl}}[m]) \subseteq K(\psi^{\mathrm{Carl}}[m])(u) K ( u ψ Carl [ m ]) ⊆ K ( ψ Carl [ m ]) ( u ) , which gives the forward inclusion in (5.9). For the reverse inclusion, note that if λ ∈ ψ C a r l [ m ] \lambda\in\psi^{\mathrm{Carl}}[m] λ ∈ ψ Carl [ m ] , then writing λ \lambda λ as u − 1 ( u λ ) u^{-1}(u\lambda) u − 1 ( u λ ) , we see that λ ∈ K ( u ψ C a r l [ m ] ) ( u ) \lambda\in K(u\psi^{\mathrm{Carl}}[m])(u) λ ∈ K ( u ψ Carl [ m ]) ( u ) . This yields K ( ψ C a r l [ m ] ) ⊆ K ( u ψ C a r l [ m ] ) ( u ) K(\psi^{\mathrm{Carl}}[m]) \subseteq K(u\psi^{\mathrm{Carl}}[m])(u) K ( ψ Carl [ m ]) ⊆ K ( u ψ Carl [ m ]) ( u ) , as needed.
Applying Lemma 5.6 to (5.9), we see that L H ( u ) = K ( ψ C a r l [ lcm ( m , t ) ] ) ( u ) = K 1 ( u ) LH(u) = K(\psi^{\mathrm{Carl}}[\operatorname{lcm}(m,t)])(u) = K_1(u) L H ( u ) = K ( ψ Carl [ lcm ( m , t )]) ( u ) = K 1 ( u ) , denoting K ( ψ C a r l [ lcm ( m , t ) ] ) K(\psi^{\mathrm{Carl}}[\operatorname{lcm}(m,t)]) K ( ψ Carl [ lcm ( m , t )]) as K 1 K_1 K 1 . Hence
[ L H : K ] = [ L H ( u ) : K ] [ L H ( u ) : L H ] = [ K 1 ( u ) : K 1 ] [ K 1 : K ] [ L H ( u ) : L H ] = c m , t d m , t φ ( lcm ( m , t ) ) . (82) [LH : K] = \frac{[LH(u) : K]}{[LH(u) : LH]} = \frac{[K_1(u) : K_1][K_1 : K]}{[LH(u) : LH]} = \frac{c_{m,t}}{d_{m,t}}\varphi(\operatorname{lcm}(m,t)). \tag*{(82)} [ L H : K ] = [ L H ( u ) : L H ] [ L H ( u ) : K ] = [ L H ( u ) : L H ] [ K 1 ( u ) : K 1 ] [ K 1 : K ] = d m , t c m , t φ ( lcm ( m , t )) . (82)
where c m , t c_{m,t} c m , t and d m , t d_{m,t} d m , t are as defined in (5.7). Since u q − 1 ∈ K ∗ u^{q-1} \in K^* u q − 1 ∈ K ∗ , we see that c m , t c_{m,t} c m , t and d m , t d_{m,t} d m , t are divisors of q − 1 q-1 q − 1 . Substituting (5.10) into (5.8), by multiplicativity of φ \varphi φ , we obtain (5.6), as needed. □
We will also use the following bound.
Lemma 5.9. For every integer n ≥ 1 n \ge1 n ≥ 1 , we have
∑ m s q u a r e - f r e e deg m = n ′ 1 φ ( m ) ≪ 1. (83) \sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m=n}}' \frac{1}{\varphi(m)} \ll1. \tag*{(83)} m square - free d e g m = n ∑ ′ φ ( m ) 1 ≪ 1. (83)
Proof. We have
∑ m s q u a r e - f r e e deg m = n ′ 1 φ ( m ) = q − n ∑ m s q u a r e - f r e e deg m = n ′ q deg m φ ( m ) = q − n ∑ m s q u a r e - f r e e deg m = n ′ ∏ p ∣ m ( 1 − q − deg p ) − 1 = q − n ∑ m s q u a r e - f r e e deg m = n ′ ∏ p ∣ m ( 1 + 1 q deg p − 1 ) = q − n ∑ m s q u a r e - f r e e deg m = n ′ ∏ p ∣ m ( 1 + 1 φ ( p ) ) . (84) \begin{aligned}
\sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m=n}}' \frac{1}{\varphi(m)}
= q^{-n}\sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m=n}}' \frac{q^{\deg m}}{\varphi(m)}
= q^{-n}\sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m=n}}' \prod_{p\mid m}(1-q^{-\deg p})^{-1} \\
&= q^{-n}\sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m=n}}^{\prime}\prod_{\mathfrak p\mid m}\left(1+\frac{1}{q^{\deg\mathfrak p}-1}\right)=q^{-n}\sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m=n}}^{\prime}\prod_{\mathfrak p\mid m}\left(1+\frac{1}{\varphi(\mathfrak p)}\right). \tag*{(84)}
\end{aligned} m square - free d e g m = n ∑ ′ φ ( m ) 1 = q − n m square - free d e g m = n ∑ ′ φ ( m ) q d e g m = q − n m square - free d e g m = n ∑ ′ p ∣ m ∏ ( 1 − q − d e g p ) − 1 = q − n m square - free d e g m = n ∑ ′ p ∣ m ∏ ( 1 + q d e g p − 1 1 ) = q − n m square - free d e g m = n ∑ ′ p ∣ m ∏ ( 1 + φ ( p ) 1 ) . (84)
Writing the inner product in (5.12) as a sum then reveals that (5.12) is equal to
q − n ∑ m s q u a r e - f r e e deg m = n ′ ∑ d ∣ m 1 φ ( d ) ≤ q − n ∑ d s q u a r e - f r e e deg d ≤ n ′ q n − deg d φ ( d ) ≤ ∏ p ( 1 + 1 q deg p ( q deg p − 1 ) ) . (85) q^{-n}\sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m=n}}^{\prime}\sum_{\mathfrak d\mid m}\frac{1}{\varphi(\mathfrak d)}
\le q^{-n}\sum_{\substack{\mathfrak d\ \mathrm{square\text{-}free}\\ \deg\mathfrak d\le n}}^{\prime}\frac{q^{n-\deg\mathfrak d}}{\varphi(\mathfrak d)}
\le\prod_{\mathfrak p}\left(1+\frac{1}{q^{\deg\mathfrak p}(q^{\deg\mathfrak p}-1)}\right). \tag*{(85)} q − n m square - free d e g m = n ∑ ′ d ∣ m ∑ φ ( d ) 1 ≤ q − n d square - free d e g d ≤ n ∑ ′ φ ( d ) q n − d e g d ≤ p ∏ ( 1 + q d e g p ( q d e g p − 1 ) 1 ) . (85)
The last product converges, since
∑ p 1 q deg p ( q deg p − 1 ) ≪ ∑ p q − 2 deg p ≤ ∑ j ≥ 1 q − j < ∞ . \sum_{\mathfrak p}\frac{1}{q^{\deg\mathfrak p}(q^{\deg\mathfrak p}-1)}
\ll\sum_{\mathfrak p}q^{-2\deg\mathfrak p}\le\sum_{j\ge1}q^{-j}<\infty. p ∑ q d e g p ( q d e g p − 1 ) 1 ≪ p ∑ q − 2 d e g p ≤ j ≥ 1 ∑ q − j < ∞.
Hence, by (5.12) and (5.13), we obtain (5.11). □
Proof of Theorem 1.2
The proof follows through similar to most part of Section 4, by splitting (4.1) into a main sum and an error sum (see (4.3)). The estimation of the error sum follows exactly as in subsections 4.2 and 4.3. For the main sum, the error follows as in subsection 4.1 until (4.6). We refine the estimate for E t a i l E_{\mathrm{tail}} E tail (see (4.7)) in the case when End K ( ϕ ) = A \operatorname{End}_{K}(\phi)=A End K ( ϕ ) = A . The argument below is analogous to the refinement of Wong [38, pp. 771-772].
We have from (4.5) that
E t a i l ≪ ϕ , K π ( x ) ∑ m s q u a r e - f r e e deg m > y ′ 1 [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] . (86) E_{\mathrm{tail}}\ll_{\phi,K}\pi(x)\sum_{\substack{m\ \mathrm{square\text{-}free}\\ \deg m>y}}^{\prime}\frac{1}{[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]}. \tag*{(86)} E tail ≪ ϕ , K π ( x ) m square - free d e g m > y ∑ ′ [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] 1 . (86)
Let m = m 1 m 2 \mathfrak m=\mathfrak m_1\mathfrak m_2 m = m 1 m 2 and t = t 1 t 2 \mathfrak t=\mathfrak t_1\mathfrak t_2 t = t 1 t 2 , where m 1 , t 1 \mathfrak m_1,\mathfrak t_1 m 1 , t 1 are composed of prime divisors of N ( ϕ ) N(\phi) N ( ϕ ) and m 2 , t 2 \mathfrak m_2,\mathfrak t_2 m 2 , t 2 are coprime to N ( ϕ ) N(\phi) N ( ϕ ) . Then, we note that
[ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] ≥ [ K ( ϕ [ m 2 ] ) K ( ψ C a r l [ t ] ) : K ] = [ K ( ϕ [ m 2 ] ) : K ] [ K ( ψ C a r l [ t ] ) : K ] [ K ( ϕ [ m 2 ] ) ∩ K ( ψ C a r l [ t ] ) : K ] . (87) [K(\phi[\mathfrak m])K(\psi^{\mathrm{Carl}}[\mathfrak t]):K]\ge[K(\phi[\mathfrak m_2])K(\psi^{\mathrm{Carl}}[\mathfrak t]):K]=\frac{[K(\phi[\mathfrak m_2]):K][K(\psi^{\mathrm{Carl}}[\mathfrak t]):K]}{[K(\phi[\mathfrak m_2])\cap K(\psi^{\mathrm{Carl}}[\mathfrak t]):K]}. \tag*{(87)} [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] ≥ [ K ( ϕ [ m 2 ]) K ( ψ Carl [ t ]) : K ] = [ K ( ϕ [ m 2 ]) ∩ K ( ψ Carl [ t ]) : K ] [ K ( ϕ [ m 2 ]) : K ] [ K ( ψ Carl [ t ]) : K ] . (87)
By [22 Proposition 12] , we have the bound
[ K ( ϕ [ m 2 ] ) : K ] ≫ ϕ , K φ ( m 2 ) q ( r 2 − 1 ) deg m 2 . (88) [K(\phi[\mathfrak m_2]):K]\gg_{\phi,K}\varphi(\mathfrak m_2)q^{(r^2-1)\deg\mathfrak m_2}. \tag*{(88)} [ K ( ϕ [ m 2 ]) : K ] ≫ ϕ , K φ ( m 2 ) q ( r 2 − 1 ) d e g m 2 . (88)
By Lemma 5.8, we have that [ K ( ϕ [ m 2 ] ) ∩ K ( ψ C a r l [ t ] ) : K ] = d m 2 , t φ ( gcd ( m 2 , t ) ) / c m 2 , t [K(\phi[\mathfrak m_2])\cap K(\psi^{\mathrm{Carl}}[\mathfrak t]):K]=d_{\mathfrak m_2,\mathfrak t}\varphi(\gcd(\mathfrak m_2,\mathfrak t))/c_{\mathfrak m_2,\mathfrak t} [ K ( ϕ [ m 2 ]) ∩ K ( ψ Carl [ t ]) : K ] = d m 2 , t φ ( g cd( m 2 , t )) / c m 2 , t , where c m 2 , t c_{\mathfrak m_2,\mathfrak t} c m 2 , t and d m 2 , t d_{\mathfrak m_2,\mathfrak t} d m 2 , t are positive divisors of q − 1 q-1 q − 1 . In particular,
[ K ( ϕ [ m 2 ] ) ∩ K ( ψ C a r l [ t ] ) : K ] ≤ ( q − 1 ) φ ( gcd ( m 2 , t ) ) . (89) [K(\phi[\mathfrak m_2])\cap K(\psi^{\mathrm{Carl}}[\mathfrak t]):K]\le(q-1)\varphi(\gcd(\mathfrak m_2,\mathfrak t)). \tag*{(89)} [ K ( ϕ [ m 2 ]) ∩ K ( ψ Carl [ t ]) : K ] ≤ ( q − 1 ) φ ( g cd( m 2 , t )) . (89)
Hence, plugging (5.16) and (5.17) into (5.15), we have
[ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] ≫ ϕ , K φ ( m 2 ) q ( r 2 − 1 ) deg m 2 φ ( t ) φ ( gcd ( m 2 , t ) ) . [K(\phi[\mathfrak m])K(\psi^{\mathrm{Carl}}[\mathfrak t]):K]\gg_{\phi,K}\frac{\varphi(\mathfrak m_2)q^{(r^2-1)\deg\mathfrak m_2}\varphi(\mathfrak t)}{\varphi(\gcd(\mathfrak m_2,\mathfrak t))}. [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] ≫ ϕ , K φ ( g cd( m 2 , t )) φ ( m 2 ) q ( r 2 − 1 ) d e g m 2 φ ( t ) .
Thus, by (5.14), we have
E t a i l ≪ ϕ , K π ( x ) ∑ m 1 ∣ N ( ϕ ) ∑ m 2 s q u a r e - f r e e gcd ( m 2 , N ( ϕ ) ) = 1 deg m 2 > y − deg m 1 ′ φ ( gcd ( m 2 , t ) ) φ ( m 2 ) q ( r 2 − 1 ) deg m 2 φ ( t ) E_{\mathrm{tail}}\ll_{\phi,K}\pi(x)\sum_{\mathfrak m_1\mid N(\phi)}\sum_{\substack{\mathfrak m_2\ \mathrm{square\text{-}free}\\ \gcd(\mathfrak m_2,N(\phi))=1\\ \deg\mathfrak m_2>y-\deg\mathfrak m_1}}^{\prime}\frac{\varphi(\gcd(\mathfrak m_2,\mathfrak t))}{\varphi(\mathfrak m_2)q^{(r^2-1)\deg\mathfrak m_2}\varphi(\mathfrak t)} E tail ≪ ϕ , K π ( x ) m 1 ∣ N ( ϕ ) ∑ m 2 square - free g c d ( m 2 , N ( ϕ )) = 1 d e g m 2 > y − d e g m 1 ∑ ′ φ ( m 2 ) q ( r 2 − 1 ) d e g m 2 φ ( t ) φ ( g cd( m 2 , t ))
≪ ϕ , K π ( x ) 1 φ ( t ) ∑ m 1 ∣ N ( ϕ ) ∑ m 2 s q u a r e - f r e e gcd ( m 2 , N ( ϕ ) ) = 1 deg m 2 > y − deg m 1 ′ φ ( gcd ( m 2 , t ) ) φ ( m 2 ) q ( r 2 − 1 ) deg m 2 . (90) \ll_{\phi,K}\pi(x)\frac{1}{\varphi(\mathfrak t)}\sum_{\mathfrak m_1\mid N(\phi)}\sum_{\substack{\mathfrak m_2\ \mathrm{square\text{-}free}\\ \gcd(\mathfrak m_2,N(\phi))=1\\ \deg\mathfrak m_2>y-\deg\mathfrak m_1}}^{\prime}\frac{\varphi(\gcd(\mathfrak m_2,\mathfrak t))}{\varphi(\mathfrak m_2)q^{(r^2-1)\deg\mathfrak m_2}}. \tag*{(90)} ≪ ϕ , K π ( x ) φ ( t ) 1 m 1 ∣ N ( ϕ ) ∑ m 2 square - free g c d ( m 2 , N ( ϕ )) = 1 d e g m 2 > y − d e g m 1 ∑ ′ φ ( m 2 ) q ( r 2 − 1 ) d e g m 2 φ ( g cd( m 2 , t )) . (90)
Since gcd ( m 2 , N ( ϕ ) ) = 1 \gcd(m_2,N(\phi))=1 g cd( m 2 , N ( ϕ )) = 1 , we have gcd ( m 2 , t ) = gcd ( m 2 , t 2 ) \gcd(m_2,t)=\gcd(m_2,t_2) g cd( m 2 , t ) = g cd( m 2 , t 2 ) . Let g = gcd ( m 2 , t 2 ) g=\gcd(m_2,t_2) g = g cd( m 2 , t 2 ) , and write m 2 = g h m_2=gh m 2 = g h . For squarefree m 2 m_2 m 2 , gcd ( g , h ) = 1 \gcd(g,h)=1 g cd( g , h ) = 1 and hence ϕ ( m 2 ) = ϕ ( g ) ϕ ( h ) \phi(m_2)=\phi(g)\phi(h) ϕ ( m 2 ) = ϕ ( g ) ϕ ( h ) . Thus, the innermost sum in (90) is
≤ ∑ g ∣ t 2 ′ q − ( r 2 − 1 ) deg g ∑ h s q u a r e - f r e e deg h > y − deg m 1 − deg g ′ 1 ϕ ( h ) q ( r 2 − 1 ) deg h . (91) \leq\sum_{g\mid t_2}'q^{-(r^2-1)\deg g}\sum_{\substack{h\ \mathrm{square\text{-}free}\\ \deg h>y-\deg m_1-\deg g}}'\frac{1}{\phi(h)q^{(r^2-1)\deg h}}. \tag*{(91)} ≤ g ∣ t 2 ∑ ′ q − ( r 2 − 1 ) d e g g h square - free d e g h > y − d e g m 1 − d e g g ∑ ′ ϕ ( h ) q ( r 2 − 1 ) d e g h 1 . (91)
The inner sum in (91) is then bounded above by
∑ d > y − deg m 1 − deg g 1 q ( r 2 − 1 ) d ∑ h s q u a r e - f r e e deg h = d ′ 1 ϕ ( h ) . (92) \sum_{d>y-\deg m_1-\deg g}\frac{1}{q^{(r^2-1)d}}\sum_{\substack{h\ \mathrm{square\text{-}free}\\ \deg h=d}}'\frac{1}{\phi(h)}. \tag*{(92)} d > y − d e g m 1 − d e g g ∑ q ( r 2 − 1 ) d 1 h square - free d e g h = d ∑ ′ ϕ ( h ) 1 . (92)
By Lemma 5.9 and the properties of geometric sums, we have that (92) is
≪ K q − ( r 2 − 1 ) ( y − deg m 1 − deg g ) . (93) \ll_K q^{-(r^2-1)(y-\deg m_1-\deg g)}. \tag*{(93)} ≪ K q − ( r 2 − 1 ) ( y − d e g m 1 − d e g g ) . (93)
Hence, by (91) –(93) , (90) simplifies to
E t a i l ≪ ϕ , K π ( x ) 1 ϕ ( t ) ∑ m 1 ∣ N ( ϕ ) ′ ∑ g ∣ t 2 ′ q − ( r 2 − 1 ) deg g q − ( r 2 − 1 ) ( y − deg m 1 − deg g ) E_{\mathrm{tail}}\ll_{\phi,K}\pi(x)\frac{1}{\phi(t)}\sum_{m_1\mid N(\phi)}'\sum_{g\mid t_2}'q^{-(r^2-1)\deg g}q^{-(r^2-1)(y-\deg m_1-\deg g)} E tail ≪ ϕ , K π ( x ) ϕ ( t ) 1 m 1 ∣ N ( ϕ ) ∑ ′ g ∣ t 2 ∑ ′ q − ( r 2 − 1 ) d e g g q − ( r 2 − 1 ) ( y − d e g m 1 − d e g g )
≪ ϕ , K π ( x ) τ ( t 2 ) ϕ ( t ) q − ( r 2 − 1 ) y ∑ m 1 ∣ N ( ϕ ) ′ q ( r 2 − 1 ) deg m 1 ≪ ϕ , K π ( x ) τ ( t 2 ) ϕ ( t ) q − ( r 2 − 1 ) y . \ll_{\phi,K}\pi(x)\frac{\tau(t_2)}{\phi(t)}q^{-(r^2-1)y}\sum_{m_1\mid N(\phi)}'q^{(r^2-1)\deg m_1}\ll_{\phi,K}\pi(x)\frac{\tau(t_2)}{\phi(t)}q^{-(r^2-1)y}. ≪ ϕ , K π ( x ) ϕ ( t ) τ ( t 2 ) q − ( r 2 − 1 ) y m 1 ∣ N ( ϕ ) ∑ ′ q ( r 2 − 1 ) d e g m 1 ≪ ϕ , K π ( x ) ϕ ( t ) τ ( t 2 ) q − ( r 2 − 1 ) y .
Using the prime number theorem, we thus obtain
E t a i l ≪ ϕ , K q x − ( r 2 − 1 ) y x τ ( t 2 ) ϕ ( t ) . E_{\mathrm{tail}}\ll_{\phi,K}\frac{q^{x-(r^2-1)y}}{x}\frac{\tau(t_2)}{\phi(t)}. E tail ≪ ϕ , K x q x − ( r 2 − 1 ) y ϕ ( t ) τ ( t 2 ) .
We now determine the corresponding change in the optimization. We now have
E 1 = O ϕ , K ( q x 2 + y ( y x + deg t x ) ) , E 2 = O ϕ , K ( q x − ( r 2 − 1 ) y x τ ( t 2 ) ϕ ( t ) ) , (94) E_1=O_{\phi,K}\left(q^{\frac{x}{2}+y}\left(\frac{y}{x}+\frac{\deg t}{x}\right)\right),\qquad E_2=O_{\phi,K}\left(\frac{q^{x-(r^2-1)y}}{x}\frac{\tau(t_2)}{\phi(t)}\right), \tag*{(94)} E 1 = O ϕ , K ( q 2 x + y ( x y + x deg t ) ) , E 2 = O ϕ , K ( x q x − ( r 2 − 1 ) y ϕ ( t ) τ ( t 2 ) ) , (94)
E 3 a = O K ( x r q ( r − 1 r ) x ) , E 3 b = O K ( q ( 2 r − 1 r ) x − α ϕ ( t ) − r y ) . E_{3a}=O_K\left(\frac{x}{r}q^{\left(\frac{r-1}{r}\right)x}\right),\qquad E_{3b}=O_K\left(q^{\left(\frac{2r-1}{r}\right)x-\alpha\phi(t)-ry}\right). E 3 a = O K ( r x q ( r r − 1 ) x ) , E 3 b = O K ( q ( r 2 r − 1 ) x − α ϕ ( t ) − r y ) .
By Lemma 5.9 and the multiplicativity of ϕ \phi ϕ , we have ϕ ( t ) ≫ K q deg t / ( 1 + deg t ) \phi(t)\gg_K q^{\deg t}/(1+\deg t) ϕ ( t ) ≫ K q d e g t / ( 1 + deg t ) . Furthermore, since every prime divisor of t 2 t_2 t 2 has degree at least 2, we have τ ( t 2 ) ≤ q deg t / 2 \tau(t_2)\leq q^{\deg t/2} τ ( t 2 ) ≤ q d e g t /2 . Hence,
τ ( t 2 ) ϕ ( t ) ≪ K ( 1 + deg t ) q − deg t 2 . (95) \frac{\tau(t_2)}{\phi(t)}\ll_K(1+\deg t)q^{-\frac{\deg t}{2}}. \tag*{(95)} ϕ ( t ) τ ( t 2 ) ≪ K ( 1 + deg t ) q − 2 d e g t . (95)
The displayed error bounds are compared up to bounded factors. We note that E 3 a E_{3a} E 3 a is independent of y ( x ) y(x) y ( x ) . Furthermore, as y ( x ) y(x) y ( x ) increases, E 1 E_1 E 1 increases whereas E 2 E_2 E 2 and E 3 b E_{3b} E 3 b decrease. We consider the choice obtained by balancing E 1 E_1 E 1 and E 3 b E_{3b} E 3 b , namely
y 2 ( x ) : = max { 0 , 1 r + 1 ( 3 r − 2 2 r x − α ϕ ( t ) − log q ( 1 + deg t x ) ) } . y_2(x):=\max\left\{0,\frac{1}{r+1}\left(\frac{3r-2}{2r}x-\alpha\phi(t)-\log_q\left(1+\frac{\deg t}{x}\right)\right)\right\}. y 2 ( x ) := max { 0 , r + 1 1 ( 2 r 3 r − 2 x − α ϕ ( t ) − log q ( 1 + x deg t ) ) } .
One can check that when α ϕ ( t ) > r − 4 2 r x \alpha\phi(t)>\frac{r-4}{2r}x α ϕ ( t ) > 2 r r − 4 x , we have 0 ≤ y 2 ( x ) ≤ x / r 0\leq y_2(x)\leq x/r 0 ≤ y 2 ( x ) ≤ x / r . Moreover, by (95) , α ϕ ( t ) ≤ deg t ≤ x \alpha\phi(t)\leq\deg t\leq x α ϕ ( t ) ≤ deg t ≤ x , and r ≥ 2 r\geq2 r ≥ 2 , the bounds for E 2 E_2 E 2 and E 3 b E_{3b} E 3 b are both dominated by that for E 1 E_1 E 1 when y = y 2 ( x ) y=y_2(x) y = y 2 ( x ) . Hence,
E 1 + E 2 + E 3 a + E 3 b = O K ( x r q r − 1 r x ) + O ϕ , K ( q r 2 + 4 r − 2 2 r ( r + 1 ) x − α ϕ ( t ) r + 1 ( 1 + deg t x ) r r + 1 ) . E_1+E_2+E_{3a}+E_{3b}=O_K\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{r^2+4r-2}{2r(r+1)}x-\frac{\alpha\phi(t)}{r+1}}\left(1+\frac{\deg t}{x}\right)^{\frac{r}{r+1}}\right). E 1 + E 2 + E 3 a + E 3 b = O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r ( r + 1 ) r 2 + 4 r − 2 x − r + 1 α ϕ ( t ) ( 1 + x deg t ) r + 1 r ) .
Thus, we choose
y ( x ) : = { x r when α ϕ ( t ) ≤ r − 4 2 r x , y 2 ( x ) when α ϕ ( t ) > r − 4 2 r x . y(x) := \begin{cases}
\frac{x}{r} & \text{when }\alpha_{\phi}(t) \le\frac{r-4}{2r}x,\\
y_{2}(x) & \text{when }\alpha_{\phi}(t) > \frac{r-4}{2r}x.
\end{cases} y ( x ) := { r x y 2 ( x ) when α ϕ ( t ) ≤ 2 r r − 4 x , when α ϕ ( t ) > 2 r r − 4 x .
With this choice of y ( x ) y(x) y ( x ) , when α ϕ ( t ) > r − 4 2 r x \alpha_{\phi}(t) > \frac{r-4}{2r}x α ϕ ( t ) > 2 r r − 4 x , we obtain
π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O K ( x r q r − 1 r x ) + O ϕ , K ( q r 2 + 4 r − 2 2 r ( r + 1 ) x − α ϕ ( t ) r + 1 ( 1 + deg t x ) r r + 1 ) . \pi_{\phi}(x;t,\mathfrak{a})=\pi(x)\cdot\delta_{\phi,t,\mathfrak{a}}(x)+O_{K}\left(\frac{x}{r}q^{\frac{r-1}{r}x}\right)+O_{\phi,K}\left(q^{\frac{r^{2}+4r-2}{2r(r+1)}x^{-\frac{\alpha_{\phi}(t)}{r+1}}}\left(1+\frac{\deg t}{x}\right)^{\frac{r}{r+1}}\right). π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O K ( r x q r r − 1 x ) + O ϕ , K ( q 2 r ( r + 1 ) r 2 + 4 r − 2 x − r + 1 α ϕ ( t ) ( 1 + x deg t ) r + 1 r ) .
Lastly, when α ϕ ( t ) ≤ r − 4 2 r x \alpha_{\phi}(t) \le\frac{r-4}{2r}x α ϕ ( t ) ≤ 2 r r − 4 x , we choose y ( x ) = x / r y(x)=x/r y ( x ) = x / r , which makes the sum ∑ e r r o r \sum_{\mathrm{error}} ∑ error in (4.3) vanish. Hence (4.27) reduces to π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + E 1 + E 2 \pi_{\phi}(x;t,\mathfrak{a})=\pi(x)\cdot\delta_{\phi,t,\mathfrak{a}}(x)+E_{1}+E_{2} π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + E 1 + E 2 , where E 1 , E 2 E_{1},E_{2} E 1 , E 2 are as in (5.22). Using (5.23), this yields
π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O ϕ , K ( q x 2 + x r ( 1 + deg t x ) ) . \pi_{\phi}(x;t,\mathfrak{a})=\pi(x)\cdot\delta_{\phi,t,\mathfrak{a}}(x)+O_{\phi,K}\left(q^{\frac{x}{2}+\frac{x}{r}}\left(1+\frac{\deg t}{x}\right)\right). π ϕ ( x ; t , a ) = π ( x ) ⋅ δ ϕ , t , a ( x ) + O ϕ , K ( q 2 x + r x ( 1 + x deg t ) ) .
This proves Theorem 1.2 .
Positivity of density
Recall from Remark 2 that if r m , t = 1 r_{m,t}=1 r m , t = 1 for every monic m ∈ A m\in A m ∈ A , then the density δ ϕ , t , a ( x ) \delta_{\phi,t,\mathfrak{a}}(x) δ ϕ , t , a ( x ) is independent of x x x . In this case, we denote it by δ ϕ , t , a \delta_{\phi,t,\mathfrak{a}} δ ϕ , t , a (see (13) ). Then, since π ϕ ( x ; t , a ) ≥ 0 \pi_{\phi}(x;t,\mathfrak{a})\ge0 π ϕ ( x ; t , a ) ≥ 0 by definition (see (5) ), Theorem 1.1 implies δ ϕ , t , a ≥ 0 \delta_{\phi,t,\mathfrak{a}}\ge0 δ ϕ , t , a ≥ 0 .
Proof of Corollary 1.3
Define
ϕ ( K ) t o r s : = { α ∈ K : ϕ a ( α ) = 0 for some non-zero a ∈ A } . \phi(K)_{\mathrm{tors}}:=\{\alpha\in K:\phi_{a}(\alpha)=0\text{ for some non-zero }a\in A\}. ϕ ( K ) tors := { α ∈ K : ϕ a ( α ) = 0 for some non-zero a ∈ A } .
Let p ∈ A \mathfrak{p}\in A p ∈ A be a prime with deg p = 1 \deg\mathfrak{p}=1 deg p = 1 such that K ( ϕ [ p ] ) = K K(\phi[\mathfrak{p}])=K K ( ϕ [ p ]) = K . By definition, we have ϕ [ p ] ⊆ ϕ ( K ) t o r s \phi[\mathfrak{p}]\subseteq\phi(K)_{\mathrm{tors}} ϕ [ p ] ⊆ ϕ ( K ) tors . Since ϕ [ p ] ≃ ( A / p A ) r \phi[\mathfrak{p}]\simeq(A/\mathfrak{p}A)^{r} ϕ [ p ] ≃ ( A / p A ) r , we get that ϕ ( K ) t o r s \phi(K)_{\mathrm{tors}} ϕ ( K ) tors has a submodule isomorphic to ( A / p A ) r (A/\mathfrak{p}A)^{r} ( A / p A ) r . Since reduction embeds ϕ ( K ) t o r s \phi(K)_{\mathrm{tors}} ϕ ( K ) tors into ϕ ( F q ) \phi(\mathbb{F}_{q}) ϕ ( F q ) for all but finitely many primes q \mathfrak{q} q (see [22 Proof of Theorem 3’(i)] ), there are only finitely many prime reductions ϕ ( F q ) \phi(\mathbb{F}_{q}) ϕ ( F q ) that have at most ( r − 1 ) (r-1) ( r − 1 ) cyclic components. Recall by (5) that
π ϕ ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( t ) and ϕ ( F p ) has at most ( r − 1 ) cyclic components } . \pi_{\phi}(x;t,\mathfrak{a})=\#\{\mathfrak{p}\in\mathcal{P}_{\phi}:\deg\mathfrak{p}=x,\mathfrak{p}\equiv\mathfrak{a}(t)\text{ and }\phi(\mathbb{F}_{\mathfrak{p}})\text{ has at most }(r-1)\text{ cyclic components}\}. π ϕ ( x ; t , a ) = # { p ∈ P ϕ : deg p = x , p ≡ a ( t ) and ϕ ( F p ) has at most ( r − 1 ) cyclic components } .
Hence, by Theorem 1.1 , we have that the density δ ϕ , t , a = 0 \delta_{\phi,t,\mathfrak{a}}=0 δ ϕ , t , a = 0 .
We can also obtain the result by the following computation. For any prime p ∈ A \mathfrak{p}\in A p ∈ A , we can rewrite (1.13) as
δ ϕ , t , a = ∑ m gcd ( m , p ) = 1 ′ ( μ q ( m ) γ a , t ( K ( ϕ [ m ] ) ) [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] + μ q ( p m ) γ a , t ( K ( ϕ [ p m ] ) ) [ K ( ϕ [ p m ] ) K ( ψ C a r l [ t ] ) : K ] ) . (96) \delta_{\phi,t,\mathfrak{a}}=\sum_{\substack{m\\ \gcd(m,\mathfrak{p})=1}}^{\prime}\left(\frac{\mu_{q}(m)\gamma_{\mathfrak{a},t}(K(\phi[m]))}{[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]}+\frac{\mu_{q}(\mathfrak{p}m)\gamma_{\mathfrak{a},t}(K(\phi[\mathfrak{p}m]))}{[K(\phi[\mathfrak{p}m])K(\psi^{\mathrm{Carl}}[t]):K]}\right). \tag*{(96)} δ ϕ , t , a = m g c d ( m , p ) = 1 ∑ ′ ( [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] μ q ( m ) γ a , t ( K ( ϕ [ m ])) + [ K ( ϕ [ p m ]) K ( ψ Carl [ t ]) : K ] μ q ( p m ) γ a , t ( K ( ϕ [ p m ])) ) . (96)
Let p \mathfrak{p} p be such that deg p = 1 \deg\mathfrak{p}=1 deg p = 1 and K ( ϕ [ p ] ) = K K(\phi[\mathfrak{p}])=K K ( ϕ [ p ]) = K . Then, for m ∈ A m\in A m ∈ A with gcd ( m , p ) = 1 \gcd(m,\mathfrak{p})=1 g cd( m , p ) = 1 , Lemma 5.6 gives K ( ϕ [ p m ] ) = K ( ϕ [ p ] ) K ( ϕ [ m ] ) = K ( ϕ [ m ] ) K(\phi[\mathfrak{p}m])=K(\phi[\mathfrak{p}])K(\phi[m])=K(\phi[m]) K ( ϕ [ p m ]) = K ( ϕ [ p ]) K ( ϕ [ m ]) = K ( ϕ [ m ]) . Furthermore, for m ∈ A m\in A m ∈ A with gcd ( m , p ) = 1 \gcd(m,\mathfrak{p})=1 g cd( m , p ) = 1 , μ q ( p m ) = − μ q ( m ) \mu_{q}(\mathfrak{p}m)=-\mu_{q}(m) μ q ( p m ) = − μ q ( m ) . Hence, (6.1) is the same as
∑ m gcd ( m , p ) = 1 ′ ( μ q ( m ) γ a , t ( K ( ϕ [ m ] ) ) [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] − μ q ( m ) γ a , t ( K ( ϕ [ m ] ) ) [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] ) = 0. \sum_{\substack{m\\ \gcd(m,\mathfrak{p})=1}}^{\prime}\left(\frac{\mu_{q}(m)\gamma_{\mathfrak{a},t}(K(\phi[m]))}{[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]}-\frac{\mu_{q}(m)\gamma_{\mathfrak{a},t}(K(\phi[m]))}{[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]}\right)=0. m g c d ( m , p ) = 1 ∑ ′ ( [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] μ q ( m ) γ a , t ( K ( ϕ [ m ])) − [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] μ q ( m ) γ a , t ( K ( ϕ [ m ])) ) = 0.
This proves the corollary.
Preliminaries for Theorem 1.4
Throughout this subsection, we assume that End K ( ϕ ) = A \operatorname{End}_{K}(\phi)=A End K ( ϕ ) = A . Let N ( ϕ ) N(\phi) N ( ϕ ) be as defined in (73) , and define
B = { p ∈ A : p prime , p ∣ N ( ϕ ) } , G = { p ∈ A : p prime , gcd ( p , N ( ϕ ) ) = 1 } . \mathcal{B}=\{\mathfrak{p}\in A:\mathfrak{p}\text{ prime},\ \mathfrak{p}\mid N(\phi)\},\qquad\mathcal{G}=\{\mathfrak{p}\in A:\mathfrak{p}\text{ prime},\ \gcd(\mathfrak{p},N(\phi))=1\}. B = { p ∈ A : p prime , p ∣ N ( ϕ )} , G = { p ∈ A : p prime , g cd( p , N ( ϕ )) = 1 } .
For a set S \mathcal{S} S of primes, put D S : = ∏ p ∈ S K ( ϕ [ p ] ) D_{\mathcal{S}}:=\prod_{\mathfrak{p}\in\mathcal{S}}K(\phi[\mathfrak{p}]) D S := ∏ p ∈ S K ( ϕ [ p ]) , where the product denotes the compositum. Then, by Lemma 5.6, for a finite S \mathcal{S} S , D S = K ( ϕ [ s ] ) D_{\mathcal{S}}=K(\phi[\mathfrak{s}]) D S = K ( ϕ [ s ]) , s = ∏ p ∈ S p \mathfrak{s}=\prod_{\mathfrak{p}\in\mathcal{S}}\mathfrak{p} s = ∏ p ∈ S p . We first record a simple field-theoretic observation.
Lemma 6.1. Let L 1 , L 2 , L 3 L_1,L_2,L_3 L 1 , L 2 , L 3 be finite Galois extensions of K K K with L 1 ∩ L 2 = K L_1\cap L_2=K L 1 ∩ L 2 = K . Then L 1 ∩ L 2 L 3 = K L_1\cap L_2L_3=K L 1 ∩ L 2 L 3 = K if and only if L 1 L 2 ∩ L 3 = L 2 ∩ L 3 L_1L_2\cap L_3=L_2\cap L_3 L 1 L 2 ∩ L 3 = L 2 ∩ L 3 .
Proof. We write the degree [ L 1 L 2 L 3 : K ] [L_1L_2L_3:K] [ L 1 L 2 L 3 : K ] in two different ways as
[ L 1 L 2 : K ] [ L 3 : K ] [ L 1 L 2 ∩ L 3 : K ] = [ L 1 : K ] [ L 2 L 3 : K ] [ L 1 ∩ L 2 L 3 : K ] . (97) \frac{[L_1L_2:K][L_3:K]}{[L_1L_2\cap L_3:K]}=\frac{[L_1:K][L_2L_3:K]}{[L_1\cap L_2L_3:K]}. \tag*{(97)} [ L 1 L 2 ∩ L 3 : K ] [ L 1 L 2 : K ] [ L 3 : K ] = [ L 1 ∩ L 2 L 3 : K ] [ L 1 : K ] [ L 2 L 3 : K ] . (97)
Since L 1 ∩ L 2 = K L_1\cap L_2=K L 1 ∩ L 2 = K , the left hand side of (97) is equal to
[ L 1 : K ] [ L 2 : K ] [ L 3 : K ] [ L 1 L 2 ∩ L 3 : K ] . (98) \frac{[L_1:K][L_2:K][L_3:K]}{[L_1L_2\cap L_3:K]}. \tag*{(98)} [ L 1 L 2 ∩ L 3 : K ] [ L 1 : K ] [ L 2 : K ] [ L 3 : K ] . (98)
Simplifying the numerator on the right hand side of (97) gives
[ L 1 : K ] [ L 2 : K ] [ L 3 : K ] [ L 1 ∩ L 2 L 3 : K ] [ L 2 ∩ L 3 : K ] . (99) \frac{[L_1:K][L_2:K][L_3:K]}{[L_1\cap L_2L_3:K][L_2\cap L_3:K]}. \tag*{(99)} [ L 1 ∩ L 2 L 3 : K ] [ L 2 ∩ L 3 : K ] [ L 1 : K ] [ L 2 : K ] [ L 3 : K ] . (99)
Equating (98) and (99) yields
[ L 1 L 2 ∩ L 3 : K ] = [ L 1 ∩ L 2 L 3 : K ] [ L 2 ∩ L 3 : K ] . (100) [L_1L_2\cap L_3:K]=[L_1\cap L_2L_3:K][L_2\cap L_3:K]. \tag*{(100)} [ L 1 L 2 ∩ L 3 : K ] = [ L 1 ∩ L 2 L 3 : K ] [ L 2 ∩ L 3 : K ] . (100)
Hence, if L 1 ∩ L 2 L 3 = K L_1\cap L_2L_3=K L 1 ∩ L 2 L 3 = K , then by (100) , [ L 1 L 2 ∩ L 3 : K ] = [ L 2 ∩ L 3 : K ] [L_1L_2\cap L_3:K]=[L_2\cap L_3:K] [ L 1 L 2 ∩ L 3 : K ] = [ L 2 ∩ L 3 : K ] . But since L 2 ∩ L 3 ⊆ L 1 L 2 ∩ L 3 L_2\cap L_3\subseteq L_1L_2\cap L_3 L 2 ∩ L 3 ⊆ L 1 L 2 ∩ L 3 , we thus have L 1 L 2 ∩ L 3 = L 2 ∩ L 3 L_1L_2\cap L_3=L_2\cap L_3 L 1 L 2 ∩ L 3 = L 2 ∩ L 3 . Conversely, if L 1 L 2 ∩ L 3 = L 2 ∩ L 3 L_1L_2\cap L_3=L_2\cap L_3 L 1 L 2 ∩ L 3 = L 2 ∩ L 3 , then by (100) , [ L 1 ∩ L 2 L 3 : K ] = 1 [L_1\cap L_2L_3:K]=1 [ L 1 ∩ L 2 L 3 : K ] = 1 , and hence L 1 ∩ L 2 L 3 = K L_1\cap L_2L_3=K L 1 ∩ L 2 L 3 = K . □ \square □
The following lemma isolates all the entanglement which is relevant to the positivity of the density into finitely many prime-division fields.
Lemma 6.2. Let End K ( ϕ ) = A \operatorname{End}_{K}(\phi)=A End K ( ϕ ) = A . Then, there exists a finite set S : = S ϕ , t ⊆ G \mathcal{S}:=\mathcal{S}_{\phi,t}\subseteq\mathcal{G} S := S ϕ , t ⊆ G such that for every square-free monic m ∈ A \mathfrak{m}\in A m ∈ A supported on G ∖ S \mathcal{G}\setminus\mathcal{S} G ∖ S , one has K ( ϕ [ m ] ) ∩ D B ∪ S K ( ψ C a r l [ t ] ) = K K(\phi[\mathfrak{m}])\cap D_{\mathcal{B}\cup\mathcal{S}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}])=K K ( ϕ [ m ]) ∩ D B ∪ S K ( ψ Carl [ t ]) = K .
Proof. Let S ⊆ G \mathcal{S}\subseteq\mathcal{G} S ⊆ G be an arbitrary finite set. Then, by part (3) of Lemma 5.2, for every square-free monic m ∈ A \mathfrak{m}\in A m ∈ A supported on G ∖ S \mathcal{G}\setminus\mathcal{S} G ∖ S , we have that K ( ϕ [ m ] ) ∩ D S = K K(\phi[\mathfrak{m}])\cap D_{\mathcal{S}}=K K ( ϕ [ m ]) ∩ D S = K . Hence, by Lemma 6.1, it suffices to show that there exists a finite S ⊆ G \mathcal{S}\subseteq\mathcal{G} S ⊆ G satisfying
K ( ϕ [ m ] ) D S ∩ D B K ( ψ C a r l [ t ] ) = D S ∩ D B K ( ψ C a r l [ t ] ) . (101) K(\phi[\mathfrak{m}])D_{\mathcal{S}}\cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}])=D_{\mathcal{S}}\cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}]). \tag*{(101)} K ( ϕ [ m ]) D S ∩ D B K ( ψ Carl [ t ]) = D S ∩ D B K ( ψ Carl [ t ]) . (101)
Since K ( ϕ [ m ] ) D S ⊆ D G K(\phi[\mathfrak{m}])D_{\mathcal{S}}\subseteq D_{\mathcal{G}} K ( ϕ [ m ]) D S ⊆ D G , we consider the field F : = D G ∩ D B K ( ψ C a r l [ t ] ) F:=D_{\mathcal{G}}\cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}]) F := D G ∩ D B K ( ψ Carl [ t ]) . Since B \mathcal{B} B is finite, the extension D B K ( ψ C a r l [ t ] ) / K D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}])/K D B K ( ψ Carl [ t ]) / K is finite, and hence F / K F/K F / K is also finite. Moreover, F ⊆ D G F\subseteq D_{\mathcal{G}} F ⊆ D G . Therefore, there exists a finite set S ⊆ G \mathcal{S}\subseteq\mathcal{G} S ⊆ G such that
F = D S ∩ D B K ( ψ C a r l [ t ] ) . (102) F=D_{\mathcal{S}}\cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}]). \tag*{(102)} F = D S ∩ D B K ( ψ Carl [ t ]) . (102)
Therefore,
K ( ϕ [ m ] ) D S ∩ D B K ( ψ C a r l [ t ] ) ⊆ D G ∩ D B K ( ψ C a r l [ t ] ) = D S ∩ D B K ( ψ C a r l [ t ] ) , K(\phi[\mathfrak{m}])D_{\mathcal{S}}\cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}])\subseteq D_{\mathcal{G}}\cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}])=D_{\mathcal{S}}\cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}]), K ( ϕ [ m ]) D S ∩ D B K ( ψ Carl [ t ]) ⊆ D G ∩ D B K ( ψ Carl [ t ]) = D S ∩ D B K ( ψ Carl [ t ]) ,
where the last equality follows from (102) . The reverse inclusion is immediate, giving (101) and completing the proof. □ \square □
Let S \mathcal{S} S be an entanglement set for ( ϕ , t ) (\phi,t) ( ϕ , t ) as in Lemma 6.2, and put L ϕ , t , S : = D B ∪ S K ( ψ C a r l [ t ] ) L_{\phi,t,\mathcal{S}}:=D_{\mathcal{B}\cup\mathcal{S}}K(\psi^{\mathrm{Carl}}[\mathfrak{t}]) L ϕ , t , S := D B ∪ S K ( ψ Carl [ t ]) . Let λ t \lambda_t λ t generate the A / t A A/tA A / t A -module ψ C a r l [ t ] \psi^{\mathrm{Carl}}[\mathfrak{t}] ψ Carl [ t ] . Recall, by (14) , that
Ω ϕ , t , a , S : = { σ ∈ Gal ( L ϕ , t , S / K ) : σ ∣ K ( ψ C a r l [ t ] ) ( λ t ) = ψ a C a r l ( λ t ) , σ ∣ K ( ϕ [ p ] ) ≠ 1 for every p ∈ B ∪ S } . \Omega_{\phi,t,a,\mathcal{S}}:=\{\sigma\in\operatorname{Gal}(L_{\phi,t,\mathcal{S}}/K):\sigma|_{K(\psi^{\mathrm{Carl}}[\mathfrak{t}])}(\lambda_t)=\psi^{\mathrm{Carl}}_{a}(\lambda_t),\ \sigma|_{K(\phi[\mathfrak{p}])}\ne1\text{ for every }\mathfrak{p}\in\mathcal{B}\cup\mathcal{S}\}. Ω ϕ , t , a , S := { σ ∈ Gal ( L ϕ , t , S / K ) : σ ∣ K ( ψ Carl [ t ]) ( λ t ) = ψ a Carl ( λ t ) , σ ∣ K ( ϕ [ p ]) = 1 for every p ∈ B ∪ S } .
Lemma 6.3. With the notation above,
∑ m p ∣ m ⇒ p ∈ B ∪ S ′ μ q ( m ) γ a , t ( K ( ϕ [ m ] ) ) [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] = ∣ Ω ϕ , t , a , S ∣ [ L ϕ , t , S : K ] . \sum_{\substack{m\\ p\mid m\Rightarrow p\in\mathcal{B}\cup\mathcal{S}}}^{\prime}
\frac{\mu_q(m)\gamma_{a,t}(K(\phi[m]))}{[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]}
=
\frac{|\Omega_{\phi,t,a,\mathcal{S}}|}{[L_{\phi,t,\mathcal{S}}:K]}. m p ∣ m ⇒ p ∈ B ∪ S ∑ ′ [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] μ q ( m ) γ a , t ( K ( ϕ [ m ])) = [ L ϕ , t , S : K ] ∣ Ω ϕ , t , a , S ∣ .
Proof. Put G : = Gal ( L ϕ , t , S / K ) G:=\operatorname{Gal}(L_{\phi,t,\mathcal{S}}/K) G := Gal ( L ϕ , t , S / K ) , and define H p : = { σ ∈ G : σ ∣ K ( ϕ [ p ] ) = 1 } H_p:=\{\sigma\in G:\sigma|_{K(\phi[p])}=1\} H p := { σ ∈ G : σ ∣ K ( ϕ [ p ]) = 1 } , ( p ∈ B ∪ S ) (p\in\mathcal{B}\cup\mathcal{S}) ( p ∈ B ∪ S ) , and X a , t : = { σ ∈ G : σ ( λ t ) = ψ a C a r l ( λ t ) } X_{a,t}:=\{\sigma\in G:\sigma(\lambda_t)=\psi^{\mathrm{Carl}}_a(\lambda_t)\} X a , t := { σ ∈ G : σ ( λ t ) = ψ a Carl ( λ t )} . Then
Ω ϕ , t , a , S = X a , t ∖ ⋃ p ∈ B ∪ S H p . (103) \Omega_{\phi,t,a,\mathcal{S}}=X_{a,t}\setminus\bigcup_{p\in\mathcal{B}\cup\mathcal{S}}H_p. \tag*{(103)} Ω ϕ , t , a , S = X a , t ∖ p ∈ B ∪ S ⋃ H p . (103)
For square-free monic m m m supported on B ∪ S \mathcal{B}\cup\mathcal{S} B ∪ S , the identity K ( ϕ [ m ] ) = ∏ p ∣ m K ( ϕ [ p ] ) K(\phi[m])=\prod_{p\mid m}K(\phi[p]) K ( ϕ [ m ]) = ∏ p ∣ m K ( ϕ [ p ]) gives
X a , t ∩ ⋂ p ∣ m H p = { σ ∈ G : σ ∣ K ( ϕ [ m ] ) = 1 , σ ( λ t ) = ψ a C a r l ( λ t ) } . (104) X_{a,t}\cap\bigcap_{p\mid m}H_p=\{\sigma\in G:\sigma|_{K(\phi[m])}=1,\ \sigma(\lambda_t)=\psi^{\mathrm{Carl}}_a(\lambda_t)\}. \tag*{(104)} X a , t ∩ p ∣ m ⋂ H p = { σ ∈ G : σ ∣ K ( ϕ [ m ]) = 1 , σ ( λ t ) = ψ a Carl ( λ t )} . (104)
The prescribed automorphisms agree on K ( ϕ [ m ] ) ∩ K ( ψ C a r l [ t ] ) K(\phi[m])\cap K(\psi^{\mathrm{Carl}}[t]) K ( ϕ [ m ]) ∩ K ( ψ Carl [ t ]) if and only if γ a , t ( K ( ϕ [ m ] ) ) = 1 \gamma_{a,t}(K(\phi[m]))=1 γ a , t ( K ( ϕ [ m ])) = 1 . In this case, they extend to the compositum K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) K(\phi[m])K(\psi^{\mathrm{Carl}}[t]) K ( ϕ [ m ]) K ( ψ Carl [ t ]) , and then to L ϕ , t , S L_{\phi,t,\mathcal{S}} L ϕ , t , S . Thus the set in (104) is empty when γ a , t ( K ( ϕ [ m ] ) ) = 0 \gamma_{a,t}(K(\phi[m]))=0 γ a , t ( K ( ϕ [ m ])) = 0 , and otherwise is a coset of Gal ( L ϕ , t , S / K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) ) \operatorname{Gal}(L_{\phi,t,\mathcal{S}}/K(\phi[m])K(\psi^{\mathrm{Carl}}[t])) Gal ( L ϕ , t , S / K ( ϕ [ m ]) K ( ψ Carl [ t ])) . Consequently,
∣ X a , t ∩ ⋂ p ∣ m H p ∣ = γ a , t ( K ( ϕ [ m ] ) ) [ L ϕ , t , S : K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) ] . \left|X_{a,t}\cap\bigcap_{p\mid m}H_p\right|=\gamma_{a,t}(K(\phi[m]))[L_{\phi,t,\mathcal{S}}:K(\phi[m])K(\psi^{\mathrm{Carl}}[t])]. X a , t ∩ p ∣ m ⋂ H p = γ a , t ( K ( ϕ [ m ])) [ L ϕ , t , S : K ( ϕ [ m ]) K ( ψ Carl [ t ])] .
Dividing by ∣ G ∣ = [ L ϕ , t , S : K ] |G|=[L_{\phi,t,\mathcal{S}}:K] ∣ G ∣ = [ L ϕ , t , S : K ] yields
1 ∣ G ∣ ∣ X a , t ∩ ⋂ p ∣ m H p ∣ = γ a , t ( K ( ϕ [ m ] ) ) [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] . (105) \frac{1}{|G|}\left|X_{a,t}\cap\bigcap_{p\mid m}H_p\right|=\frac{\gamma_{a,t}(K(\phi[m]))}{[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]}. \tag*{(105)} ∣ G ∣ 1 X a , t ∩ p ∣ m ⋂ H p = [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] γ a , t ( K ( ϕ [ m ])) . (105)
Since μ q ( m ) = 0 \mu_q(m)=0 μ q ( m ) = 0 unless m m m is square-free, inclusion-exclusion, (103) , and (105) give
∑ m p ∣ m ⇒ p ∈ B ∪ S ′ μ q ( m ) γ a , t ( K ( ϕ [ m ] ) ) [ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] = 1 ∣ G ∣ ∑ m s q u a r e - f r e e p ∣ m ⇒ p ∈ B ∪ S ′ μ q ( m ) ∣ X a , t ∩ ⋂ p ∣ m H p ∣ = 1 ∣ G ∣ ∣ X a , t ∖ ⋃ p ∈ B ∪ S H p ∣ = ∣ Ω ϕ , t , a , S ∣ [ L ϕ , t , S : K ] . \begin{aligned}
\sum_{\substack{m\\ p\mid m\Rightarrow p\in\mathcal{B}\cup\mathcal{S}}}^{\prime}
\frac{\mu_q(m)\gamma_{a,t}(K(\phi[m]))}{[K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]}
=
\frac{1}{|G|}
\sum_{\substack{m\ \mathrm{square\text{-}free}\\ p\mid m\Rightarrow p\in\mathcal{B}\cup\mathcal{S}}}^{\prime}
\mu_q(m)\left|X_{a,t}\cap\bigcap_{p\mid m}H_p\right| \\
&=\frac{1}{|G|}\left|X_{a,t}\setminus\bigcup_{p\in\mathcal{B}\cup\mathcal{S}}H_p\right|
=\frac{|\Omega_{\phi,t,a,\mathcal{S}}|}{[L_{\phi,t,\mathcal{S}}:K]}.
\end{aligned} m p ∣ m ⇒ p ∈ B ∪ S ∑ ′ [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] μ q ( m ) γ a , t ( K ( ϕ [ m ])) = ∣ G ∣ 1 m square - free p ∣ m ⇒ p ∈ B ∪ S ∑ ′ μ q ( m ) X a , t ∩ p ∣ m ⋂ H p = ∣ G ∣ 1 X a , t ∖ p ∈ B ∪ S ⋃ H p = [ L ϕ , t , S : K ] ∣ Ω ϕ , t , a , S ∣ .
Proof of Theorem 1.4
Every square-free m ∈ A m\in A m ∈ A may be written uniquely as m = m 0 m 1 m=m_0m_1 m = m 0 m 1 , where m 0 m_0 m 0 is supported on B ∪ S \mathcal{B}\cup\mathcal{S} B ∪ S and m 1 m_1 m 1 is supported on G ∖ S \mathcal{G}\setminus\mathcal{S} G ∖ S . In particular, gcd ( m 0 , m 1 ) = 1 \gcd(m_0,m_1)=1 g cd( m 0 , m 1 ) = 1 and by Lemma 5.2 ,
K ( ϕ [ m ] ) = K ( ϕ [ m 0 ] ) K ( ϕ [ m 1 ] ) . (106) K(\phi[m])=K(\phi[m_0])K(\phi[m_1]). \tag*{(106)} K ( ϕ [ m ]) = K ( ϕ [ m 0 ]) K ( ϕ [ m 1 ]) . (106)
Since S \mathcal{S} S is an entanglement set for ( ϕ , t ) (\phi,t) ( ϕ , t ) , we have by Lemma 6.2 ,
K ( ϕ [ m 1 ] ) ∩ L ϕ , t , S = K . (107) K(\phi[m_1])\cap L_{\phi,t,\mathcal{S}}=K. \tag*{(107)} K ( ϕ [ m 1 ]) ∩ L ϕ , t , S = K . (107)
Since K ( ϕ [ m 0 ] ) K ( ψ C a r l [ t ] ) ⊆ L ϕ , t , S K(\phi[m_0])K(\psi^{\mathrm{Carl}}[t])\subseteq L_{\phi,t,\mathcal{S}} K ( ϕ [ m 0 ]) K ( ψ Carl [ t ]) ⊆ L ϕ , t , S , we obtain from (107) that
K ( ϕ [ m 1 ] ) ∩ K ( ϕ [ m 0 ] ) K ( ψ C a r l [ t ] ) = K . (108) K(\phi[m_1])\cap K(\phi[m_0])K(\psi^{\mathrm{Carl}}[t])=K. \tag*{(108)} K ( ϕ [ m 1 ]) ∩ K ( ϕ [ m 0 ]) K ( ψ Carl [ t ]) = K . (108)
Therefore,
[ K ( ϕ [ m ] ) K ( ψ C a r l [ t ] ) : K ] = [ K ( ϕ [ m 0 ] ) K ( ψ C a r l [ t ] ) : K ] [ K ( ϕ [ m 1 ] ) : K ] . (109) [K(\phi[m])K(\psi^{\mathrm{Carl}}[t]):K]=[K(\phi[m_0])K(\psi^{\mathrm{Carl}}[t]):K][K(\phi[m_1]):K]. \tag*{(109)} [ K ( ϕ [ m ]) K ( ψ Carl [ t ]) : K ] = [ K ( ϕ [ m 0 ]) K ( ψ Carl [ t ]) : K ] [ K ( ϕ [ m 1 ]) : K ] . (109)
We also claim that
K ( ϕ [ m ] ) ∩ K ( ψ C a r l [ t ] ) = K ( ϕ [ m 0 ] ) ∩ K ( ψ C a r l [ t ] ) . (110) K(\phi[m])\cap K(\psi^{\mathrm{Carl}}[t])=K(\phi[m_0])\cap K(\psi^{\mathrm{Carl}}[t]). \tag*{(110)} K ( ϕ [ m ]) ∩ K ( ψ Carl [ t ]) = K ( ϕ [ m 0 ]) ∩ K ( ψ Carl [ t ]) . (110)
Indeed, by (6.12), K ( ϕ [ m 1 ] ) ∩ K ( ϕ [ m 0 ] ) = K K(\phi[\mathfrak{m}_1]) \cap K(\phi[\mathfrak{m}_0]) = K K ( ϕ [ m 1 ]) ∩ K ( ϕ [ m 0 ]) = K , and by (6.13), K ( ϕ [ m 1 ] ) ∩ K ( ϕ [ m 0 ] ) K ( ψ C a r l [ t ] ) = K K(\phi[\mathfrak{m}_1]) \cap K(\phi[\mathfrak{m}_0])K(\psi^{\mathrm{Carl}}[\mathfrak{t}]) = K K ( ϕ [ m 1 ]) ∩ K ( ϕ [ m 0 ]) K ( ψ Carl [ t ]) = K . Thus, applying Lemma 6.1 with L 1 L_1 L 1 , L 2 L_2 L 2 , L 3 L_3 L 3 equal to K ( ϕ [ m 1 ] ) K(\phi[\mathfrak{m}_1]) K ( ϕ [ m 1 ]) , K ( ϕ [ m 0 ] ) K(\phi[\mathfrak{m}_0]) K ( ϕ [ m 0 ]) , K ( ψ C a r l [ t ] ) K(\psi^{\mathrm{Carl}}[\mathfrak{t}]) K ( ψ Carl [ t ]) respectively, and using (6.11), we obtain (6.15). Hence,
γ a , t ( K ( ϕ [ m ] ) ) = γ a , t ( K ( ϕ [ m 0 ] ) ) . (111) \gamma_{a,t}(K(\phi[\mathfrak{m}]))=\gamma_{a,t}(K(\phi[\mathfrak{m}_0])). \tag*{(111)} γ a , t ( K ( ϕ [ m ])) = γ a , t ( K ( ϕ [ m 0 ])) . (111)
The series defining δ ϕ , t , a \delta_{\phi,t,a} δ ϕ , t , a is absolutely convergent, as proved at the end of subsection 4.1. Therefore, using (6.14), (6.16), and the multiplicativity of μ q \mu_q μ q , we may write
δ ϕ , t , a = ∑ m 0 p ∣ m 0 ⇒ p ∈ B ∪ S ′ ∑ m 1 p ∣ m 1 ⇒ p ∈ G ∖ S ′ μ q ( m 0 ) μ q ( m 1 ) γ a , t ( K ( ϕ [ m 0 ] ) ) [ K ( ϕ [ m 0 ] ) K ( ψ C a r l [ t ] ) : K ] [ K ( ϕ [ m 1 ] ) : K ] = ( ∑ m 0 p ∣ m 0 ⇒ p ∈ B ∪ S ′ μ q ( m 0 ) γ a , t ( K ( ϕ [ m 0 ] ) ) [ K ( ϕ [ m 0 ] ) K ( ψ C a r l [ t ] ) : K ] ) ( ∑ m 1 p ∣ m 1 ⇒ p ∈ G ∖ S ′ μ q ( m 1 ) [ K ( ϕ [ m 1 ] ) : K ] ) . (112) \begin{aligned}
\delta_{\phi,t,a}
&=\sum_{\substack{\mathfrak{m}_0\\p\mid\mathfrak{m}_0\Rightarrow p\in\mathcal{B}\cup S}}^{\prime}
\sum_{\substack{\mathfrak{m}_1\\p\mid\mathfrak{m}_1\Rightarrow p\in\mathcal{G}\setminus S}}^{\prime}
\frac{\mu_q(\mathfrak{m}_0)\mu_q(\mathfrak{m}_1)\gamma_{a,t}(K(\phi[\mathfrak{m}_0]))}
{[K(\phi[\mathfrak{m}_0])K(\psi^{\mathrm{Carl}}[\mathfrak{t}]):K][K(\phi[\mathfrak{m}_1]):K]} \\
&=\left(\sum_{\substack{\mathfrak{m}_0\\p\mid\mathfrak{m}_0\Rightarrow p\in\mathcal{B}\cup S}}^{\prime}
\frac{\mu_q(\mathfrak{m}_0)\gamma_{a,t}(K(\phi[\mathfrak{m}_0]))}
{[K(\phi[\mathfrak{m}_0])K(\psi^{\mathrm{Carl}}[\mathfrak{t}]):K]}\right)
\left(\sum_{\substack{\mathfrak{m}_1\\p\mid\mathfrak{m}_1\Rightarrow p\in\mathcal{G}\setminus S}}^{\prime}
\frac{\mu_q(\mathfrak{m}_1)}{[K(\phi[\mathfrak{m}_1]):K]}\right). \tag*{(112)}
\end{aligned} δ ϕ , t , a = m 0 p ∣ m 0 ⇒ p ∈ B ∪ S ∑ ′ m 1 p ∣ m 1 ⇒ p ∈ G ∖ S ∑ ′ [ K ( ϕ [ m 0 ]) K ( ψ Carl [ t ]) : K ] [ K ( ϕ [ m 1 ]) : K ] μ q ( m 0 ) μ q ( m 1 ) γ a , t ( K ( ϕ [ m 0 ])) = m 0 p ∣ m 0 ⇒ p ∈ B ∪ S ∑ ′ [ K ( ϕ [ m 0 ]) K ( ψ Carl [ t ]) : K ] μ q ( m 0 ) γ a , t ( K ( ϕ [ m 0 ])) m 1 p ∣ m 1 ⇒ p ∈ G ∖ S ∑ ′ [ K ( ϕ [ m 1 ]) : K ] μ q ( m 1 ) . (112)
By Lemma 6.3, the first factor in (6.17) is equal to
∣ Ω ϕ , t , a , S ∣ [ L ϕ , t , S : K ] . (113) \frac{|\Omega_{\phi,t,a,S}|}{[L_{\phi,t,S}:K]}. \tag*{(113)} [ L ϕ , t , S : K ] ∣ Ω ϕ , t , a , S ∣ . (113)
We now consider the second factor. If m 1 \mathfrak{m}_1 m 1 is square-free and supported on G ∖ S \mathcal{G}\setminus S G ∖ S , then by Lemma 5.2 and the Chinese Remainder Theorem, [ K ( ϕ [ m 1 ] ) : K ] = ∏ p ∣ m 1 [ K ( ϕ [ p ] ) : K ] [K(\phi[\mathfrak{m}_1]):K]=\prod_{p\mid\mathfrak{m}_1}[K(\phi[p]):K] [ K ( ϕ [ m 1 ]) : K ] = ∏ p ∣ m 1 [ K ( ϕ [ p ]) : K ] . Hence, again using absolute convergence,
∑ m 1 p ∣ m 1 ⇒ p ∈ G ∖ S ′ μ q ( m 1 ) [ K ( ϕ [ m 1 ] ) : K ] = ∏ p ∉ B ∪ S ( 1 − 1 [ K ( ϕ [ p ] ) : K ] ) . (114) \sum_{\substack{\mathfrak{m}_1\\p\mid\mathfrak{m}_1\Rightarrow p\in\mathcal{G}\setminus S}}^{\prime}
\frac{\mu_q(\mathfrak{m}_1)}{[K(\phi[\mathfrak{m}_1]):K]}
=
\prod_{p\notin\mathcal{B}\cup S}\left(1-\frac{1}{[K(\phi[p]):K]}\right). \tag*{(114)} m 1 p ∣ m 1 ⇒ p ∈ G ∖ S ∑ ′ [ K ( ϕ [ m 1 ]) : K ] μ q ( m 1 ) = p ∈ / B ∪ S ∏ ( 1 − [ K ( ϕ [ p ]) : K ] 1 ) . (114)
Combining (6.17), (6.18), and (6.19), we obtain (1.15). It remains to prove the positivity criterion. By part (1) of Lemma 5.2, [ K ( ϕ [ p ] ) : K ] > 1 [K(\phi[p]):K]>1 [ K ( ϕ [ p ]) : K ] > 1 for every p ∉ B ∪ S p\notin\mathcal{B}\cup S p ∈ / B ∪ S . Moreover, the absolute convergence established earlier gives ∑ p ∉ B ∪ S 1 [ K ( ϕ [ p ] ) : K ] < ∞ \sum_{p\notin\mathcal{B}\cup S}\frac{1}{[K(\phi[p]):K]}<\infty ∑ p ∈ / B ∪ S [ K ( ϕ [ p ]) : K ] 1 < ∞ . Therefore,
∏ p ∉ B ∪ S ( 1 − 1 [ K ( ϕ [ p ] ) : K ] ) > 0. \prod_{p\notin\mathcal{B}\cup S}\left(1-\frac{1}{[K(\phi[p]):K]}\right)>0. p ∈ / B ∪ S ∏ ( 1 − [ K ( ϕ [ p ]) : K ] 1 ) > 0.
It then follows that δ ϕ , t , a > 0 \delta_{\phi,t,a}>0 δ ϕ , t , a > 0 if and only if Ω ϕ , t , a , S ≠ ∅ \Omega_{\phi,t,a,S}\ne\varnothing Ω ϕ , t , a , S = ∅ . This proves the theorem.
Proof of Corollary 1.6
We show that δ ϕ , t , a > 0 \delta_{\phi,t,a}>0 δ ϕ , t , a > 0 under the assumptions gcd ( t , N ( ϕ ) ) = 1 \gcd(t,N(\phi))=1 g cd( t , N ( ϕ )) = 1 , End K ‾ ( ϕ ) = A \operatorname{End}_{\overline{K}}(\phi)=A End K ( ϕ ) = A and ϕ 1 ≃ K ψ C a r l \phi^1\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl . By Theorem 1.4, it suffices to show that Ω ϕ , t , a , S ≠ ∅ \Omega_{\phi,t,a,S}\ne\varnothing Ω ϕ , t , a , S = ∅ . Let S S S be as in Lemma 6.2. Recall that
Ω ϕ , t , a , S : = { σ ∈ Gal ( L ϕ , t , S / K ) : σ ∣ K ( ψ C a r l [ t ] ) ( λ t ) = ψ a C a r l ( λ t ) , σ ∣ K ( ϕ [ p ] ) ≠ 1 for every p ∈ B ∪ S } . \Omega_{\phi,t,a,S}:=\left\{\sigma\in\operatorname{Gal}(L_{\phi,t,S}/K):\left.\sigma\right|_{K(\psi^{\mathrm{Carl}}[\mathfrak{t}])}(\lambda_t)=\psi^{\mathrm{Carl}}_a(\lambda_t),\ \left.\sigma\right|_{K(\phi[p])}\ne1\text{ for every }p\in\mathcal{B}\cup S\right\}. Ω ϕ , t , a , S := { σ ∈ Gal ( L ϕ , t , S / K ) : σ ∣ K ( ψ Carl [ t ]) ( λ t ) = ψ a Carl ( λ t ) , σ ∣ K ( ϕ [ p ]) = 1 for every p ∈ B ∪ S } .
We first construct an automorphism on D B D_{\mathcal{B}} D B . Put b : = ∏ p ∈ B p \mathfrak{b}:=\prod_{p\in\mathcal{B}}p b := ∏ p ∈ B p . Since ϕ 1 ≃ K ψ C a r l \phi^1\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl , we have
K ( ψ C a r l [ b ] ) = K ( ϕ 1 [ b ] ) ⊆ K ( ϕ [ b ] ) = D B . K(\psi^{\mathrm{Carl}}[\mathfrak{b}])=K(\phi^1[\mathfrak{b}])\subseteq K(\phi[\mathfrak{b}])=D_{\mathcal{B}}. K ( ψ Carl [ b ]) = K ( ϕ 1 [ b ]) ⊆ K ( ϕ [ b ]) = D B .
Since q > 2 q>2 q > 2 , there exists c ∈ F q ∗ c\in\mathbb{F}_q^* c ∈ F q ∗ with c ≠ 1 c\ne1 c = 1 . Recall that we have Gal ( K ( ψ C a r l [ b ] ) / K ) ≃ ( A / b A ) ∗ \operatorname{Gal}(K(\psi^{\mathrm{Carl}}[\mathfrak{b}])/K)\simeq(A/\mathfrak{b}A)^* Gal ( K ( ψ Carl [ b ]) / K ) ≃ ( A / b A ) ∗ . Let σ c ∈ Gal ( K ( ψ C a r l [ b ] ) / K ) \sigma_c\in\operatorname{Gal}(K(\psi^{\mathrm{Carl}}[\mathfrak{b}])/K) σ c ∈ Gal ( K ( ψ Carl [ b ]) / K ) be the automorphism corresponding to c ( m o d b ) c\pmod{\mathfrak{b}} c ( mod b ) . Since c − 1 c-1 c − 1 is nonzero in ( A / b A ) ∗ (A/\mathfrak{b}A)^* ( A / b A ) ∗ , we have that c ≢ 1 ( m o d p ) c\not\equiv1\pmod p c ≡ 1 ( mod p ) for every p ∈ B p\in\mathcal{B} p ∈ B . Thus, σ c \sigma_c σ c acts nontrivially on K ( ψ C a r l [ p ] ) K(\psi^{\mathrm{Carl}}[p]) K ( ψ Carl [ p ]) for every p ∈ B p\in\mathcal{B} p ∈ B . Since D B / K D_{\mathcal{B}}/K D B / K is Galois, we may extend σ c \sigma_c σ c to an automorphism σ 1 ∈ Gal ( D B / K ) \sigma_1\in\operatorname{Gal}(D_{\mathcal{B}}/K) σ 1 ∈ Gal ( D B / K ) such that for every p ∈ B p\in\mathcal{B} p ∈ B , σ 1 ∣ K ( ϕ [ p ] ) ≠ 1 \left.\sigma_1\right|_{K(\phi[p])}\ne1 σ 1 ∣ K ( ϕ [ p ]) = 1 .
We next consider D S K ( ψ C a r l [ t ] ) D_S K(\psi^{\mathrm{Carl}}[\mathfrak{t}]) D S K ( ψ Carl [ t ]) . First, let p ∈ S p\in S p ∈ S be such that p ∤ t p\nmid t p ∤ t . Then we choose any 1 ≠ τ p ∈ Gal ( K ( ϕ [ p ] ) / K ) 1\ne\tau_p\in\operatorname{Gal}(K(\phi[p])/K) 1 = τ p ∈ Gal ( K ( ϕ [ p ]) / K ) . Now, let p ∈ S p\in S p ∈ S be such that p ∣ t p\mid t p ∣ t . Then K ( ψ C a r l [ p ] ) ⊆ K ( ψ C a r l [ t ] ) K(\psi^{\mathrm{Carl}}[p])\subseteq K(\psi^{\mathrm{Carl}}[\mathfrak{t}]) K ( ψ Carl [ p ]) ⊆ K ( ψ Carl [ t ]) . Let λ t \lambda_t λ t generate the A / t A A/\mathfrak{t}A A / t A -module ψ C a r l [ t ] \psi^{\mathrm{Carl}}[\mathfrak{t}] ψ Carl [ t ] . Consider the automorphism σ a , t ∈ Gal ( K ( ψ C a r l [ t ] ) / K ) \sigma_{a,t}\in\operatorname{Gal}(K(\psi^{\mathrm{Carl}}[\mathfrak{t}])/K) σ a , t ∈ Gal ( K ( ψ Carl [ t ]) / K ) such that σ a , t ( λ t ) = ψ C a r l ( λ t ) \sigma_{a,t}(\lambda_t)=\psi^{\mathrm{Carl}}(\lambda_t) σ a , t ( λ t ) = ψ Carl ( λ t ) . If σ a , t ∣ K ( ψ C a r l [ p ] ) ≠ 1 \left.\sigma_{a,t}\right|_{K(\psi^{\mathrm{Carl}}[\mathfrak p])}\ne1 σ a , t ∣ K ( ψ Carl [ p ]) = 1 , then choose any extension τ p \tau_{\mathfrak p} τ p of σ a , t ∣ K ( ψ C a r l [ p ] ) \left.\sigma_{a,t}\right|_{K(\psi^{\mathrm{Carl}}[\mathfrak p])} σ a , t ∣ K ( ψ Carl [ p ]) to K ( ϕ [ p ] ) K(\phi[\mathfrak p]) K ( ϕ [ p ]) . If σ a , t ∣ K ( ψ C a r l [ p ] ) = 1 \left.\sigma_{a,t}\right|_{K(\psi^{\mathrm{Carl}}[\mathfrak p])}=1 σ a , t ∣ K ( ψ Carl [ p ]) = 1 , then [ K ( ϕ [ p ] ) : K ( ψ C a r l [ p ] ) ] = ∣ G L r ( A / p A ) ∣ / ∣ ( A / p A ) ∗ ∣ > 1 [K(\phi[\mathfrak p]):K(\psi^{\mathrm{Carl}}[\mathfrak p])]=|\mathrm{GL}_{r}(A/\mathfrak p A)|/|(A/\mathfrak p A)^{*}|>1 [ K ( ϕ [ p ]) : K ( ψ Carl [ p ])] = ∣ GL r ( A / p A ) ∣/∣ ( A / p A ) ∗ ∣ > 1 , and hence we may choose an element 1 ≠ τ p ∈ Gal ( K ( ϕ [ p ] ) / K ( ψ C a r l [ p ] ) ) 1\ne\tau_{\mathfrak p}\in\operatorname{Gal}(K(\phi[\mathfrak p])/K(\psi^{\mathrm{Carl}}[\mathfrak p])) 1 = τ p ∈ Gal ( K ( ϕ [ p ]) / K ( ψ Carl [ p ])) .
By Lemma 5.2 , the fields in the family { K ( ϕ [ p ] ) } p ∈ S \{K(\phi[\mathfrak p])\}_{\mathfrak p\in\mathcal S} { K ( ϕ [ p ]) } p ∈ S are jointly linearly disjoint over K K K . Hence the automorphisms τ p \tau_{\mathfrak p} τ p combine to an automorphism τ s ∈ Gal ( D S / K ) \tau_s\in\operatorname{Gal}(D_{\mathcal S}/K) τ s ∈ Gal ( D S / K ) . By construction,
τ s ∣ K ( ϕ [ p ] ) ≠ 1 , for every p ∈ S , τ s ∣ K ( ψ C a r l [ q ] ) = σ a , t ∣ K ( ψ C a r l [ q ] ) , for q ∈ S , q ∣ t . (115) \left.\tau_s\right|_{K(\phi[\mathfrak p])}\ne1,\quad\text{for every }\mathfrak p\in\mathcal S,\qquad
\left.\tau_s\right|_{K(\psi^{\mathrm{Carl}}[\mathfrak q])}=\left.\sigma_{a,t}\right|_{K(\psi^{\mathrm{Carl}}[\mathfrak q])},\quad\text{for }\mathfrak q\in\mathcal S,\ \mathfrak q\mid t. \tag*{(115)} τ s ∣ K ( ϕ [ p ]) = 1 , for every p ∈ S , τ s ∣ K ( ψ Carl [ q ]) = σ a , t ∣ K ( ψ Carl [ q ]) , for q ∈ S , q ∣ t . (115)
Now, let s : = ∏ p ∈ S p \mathfrak s:=\prod_{\mathfrak p\in\mathcal S}\mathfrak p s := ∏ p ∈ S p . Then, D S = K ( ϕ [ s ] ) D_{\mathcal S}=K(\phi[\mathfrak s]) D S = K ( ϕ [ s ]) . Since K ( ψ C a r l [ t ] ) / K K(\psi^{\mathrm{Carl}}[t])/K K ( ψ Carl [ t ]) / K is an abelian extension, we have by Lemma 5.6 and Lemma 5.7 ,
D S ∩ K ( ψ C a r l [ t ] ) = K ( ϕ 1 [ s ] ) ∩ K ( ψ C a r l [ t ] ) = K ( ψ C a r l [ s ] ) ∩ K ( ψ C a r l [ t ] ) = K ( ψ C a r l [ gcd ( s , t ) ] ) = ∏ q ∈ S , q ∣ t K ( ψ C a r l [ q ] ) . \begin{aligned}
D_{\mathcal S}\cap K(\psi^{\mathrm{Carl}}[t])=K(\phi^{1}[\mathfrak s])\cap K(\psi^{\mathrm{Carl}}[t])=K(\psi^{\mathrm{Carl}}[\mathfrak s])\cap K(\psi^{\mathrm{Carl}}[t]) \\
&=K(\psi^{\mathrm{Carl}}[\gcd(\mathfrak s,t)])=\prod_{\mathfrak q\in\mathcal S,\ \mathfrak q\mid t}K(\psi^{\mathrm{Carl}}[\mathfrak q]).
\end{aligned} D S ∩ K ( ψ Carl [ t ]) = K ( ϕ 1 [ s ]) ∩ K ( ψ Carl [ t ]) = K ( ψ Carl [ s ]) ∩ K ( ψ Carl [ t ]) = K ( ψ Carl [ g cd( s , t )]) = q ∈ S , q ∣ t ∏ K ( ψ Carl [ q ]) .
Then, by (6.20), τ s ∣ D S ∩ K ( ψ C a r l [ t ] ) = σ a , t ∣ D S ∩ K ( ψ C a r l [ t ] ) \left.\tau_s\right|_{D_{\mathcal S}\cap K(\psi^{\mathrm{Carl}}[t])}=\left.\sigma_{a,t}\right|_{D_{\mathcal S}\cap K(\psi^{\mathrm{Carl}}[t])} τ s ∣ D S ∩ K ( ψ Carl [ t ]) = σ a , t ∣ D S ∩ K ( ψ Carl [ t ]) . Therefore, τ s \tau_s τ s and σ a , t \sigma_{a,t} σ a , t combine to an automorphism σ 2 ∈ Gal ( D S K ( ψ C a r l [ t ] ) / K ) \sigma_2\in\operatorname{Gal}(D_{\mathcal S}K(\psi^{\mathrm{Carl}}[t])/K) σ 2 ∈ Gal ( D S K ( ψ Carl [ t ]) / K ) satisfying
σ 2 ∣ K ( ϕ [ p ] ) ≠ 1 for every p ∈ S , σ 2 ∣ K ( ψ C a r l [ t ] ) = σ a , t . \left.\sigma_2\right|_{K(\phi[\mathfrak p])}\ne1\quad\text{for every }\mathfrak p\in\mathcal S,\qquad
\left.\sigma_2\right|_{K(\psi^{\mathrm{Carl}}[t])}=\sigma_{a,t}. σ 2 ∣ K ( ϕ [ p ]) = 1 for every p ∈ S , σ 2 ∣ K ( ψ Carl [ t ]) = σ a , t .
Finally, since ϕ 1 ≃ K ψ C a r l \phi^{1}\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl , we have D S K ( ψ C a r l [ t ] ) ⊆ K ( ϕ [ lcm ( s , t ) ] ) D_{\mathcal S}K(\psi^{\mathrm{Carl}}[t])\subseteq K(\phi[\operatorname{lcm}(\mathfrak s,t)]) D S K ( ψ Carl [ t ]) ⊆ K ( ϕ [ lcm ( s , t )]) . Since gcd ( t , N ( ϕ ) ) = 1 \gcd(t,N(\phi))=1 g cd( t , N ( ϕ )) = 1 and S ⊆ G \mathcal S\subseteq\mathcal G S ⊆ G , we have gcd ( lcm ( s , t ) , N ( ϕ ) ) = 1 \gcd(\operatorname{lcm}(\mathfrak s,t),N(\phi))=1 g cd( lcm ( s , t ) , N ( ϕ )) = 1 . Hence, by Lemma 5.2 , D B ∩ D S K ( ψ C a r l [ t ] ) = K D_{\mathcal B}\cap D_{\mathcal S}K(\psi^{\mathrm{Carl}}[t])=K D B ∩ D S K ( ψ Carl [ t ]) = K . Therefore, σ 1 \sigma_1 σ 1 and σ 2 \sigma_2 σ 2 combine to an automorphism σ ∈ Gal ( D B ∪ S K ( ψ C a r l [ t ] ) / K ) \sigma\in\operatorname{Gal}(D_{\mathcal B\cup\mathcal S}K(\psi^{\mathrm{Carl}}[t])/K) σ ∈ Gal ( D B ∪ S K ( ψ Carl [ t ]) / K ) , and by construction σ ∈ Ω ϕ , t , a , S \sigma\in\Omega_{\phi,t,a,\mathcal S} σ ∈ Ω ϕ , t , a , S . Thus Ω ϕ , t , a , S ≠ ∅ \Omega_{\phi,t,a,\mathcal S}\ne\varnothing Ω ϕ , t , a , S = ∅ , and the result follows from Theorem 1.4 .
Proof of Example 1.7
Let ϕ 0 \phi_0 ϕ 0 be the rank-2 Drinfeld A A A -module over K K K defined by ϕ 0 T = T + τ − T 4 τ 2 \phi_{0T}=T+\tau-T^{4}\tau^{2} ϕ 0 T = T + τ − T 4 τ 2 . By [39 Theorem 1.2] , for every nonzero monic m ∈ A \mathfrak m\in A m ∈ A ,
Gal ( K ( ϕ 0 [ m ] ) / K ) ≃ G L 2 ( A / m A ) . (116) \operatorname{Gal}(K(\phi_0[\mathfrak m])/K)\simeq\mathrm{GL}_2(A/\mathfrak m A). \tag*{(116)} Gal ( K ( ϕ 0 [ m ]) / K ) ≃ GL 2 ( A / m A ) . (116)
The associated rank-1 Drinfeld module is ϕ 0 T 1 = T + T 4 τ \phi^1_{0T}=T+T^4\tau ϕ 0 T 1 = T + T 4 τ . Note that T − 1 ψ T C a r l T = T − 1 ( T + τ ) T = T + T 4 τ T^{-1}\psi^{\mathrm{Carl}}_T T=T^{-1}(T+\tau)T=T+T^4\tau T − 1 ψ T Carl T = T − 1 ( T + τ ) T = T + T 4 τ as well and hence ϕ 0 1 ≃ K ψ C a r l \phi^1_0\simeq_K\psi^{\mathrm{Carl}} ϕ 0 1 ≃ K ψ Carl , and
K ( ϕ 0 1 [ m ] ) = K ( ψ C a r l [ m ] ) (117) K(\phi^1_0[\mathfrak m])=K(\psi^{\mathrm{Carl}}[\mathfrak m]) \tag*{(117)} K ( ϕ 0 1 [ m ]) = K ( ψ Carl [ m ]) (117)
for every nonzero monic m ∈ A \mathfrak m\in A m ∈ A . Let u u u be a nonzero T T T -torsion point of ψ C a r l \psi^{\mathrm{Carl}} ψ Carl . Then u 4 = − T u^4=-T u 4 = − T and K ( u ) = K ( ψ C a r l [ T ] ) K(u)=K(\psi^{\mathrm{Carl}}[T]) K ( u ) = K ( ψ Carl [ T ]) . We consider the Drinfeld module ϕ : = u − 1 ϕ 0 u \phi:=u^{-1}\phi_0u ϕ := u − 1 ϕ 0 u given by ϕ T = T + u 4 τ − T 4 u 24 τ 2 \phi_T=T+u^4\tau-T^4u^{24}\tau^2 ϕ T = T + u 4 τ − T 4 u 24 τ 2 . Using u 4 = − T u^4=-T u 4 = − T , we have ϕ T = T − T τ − T 10 τ 2 \phi_T=T-T\tau-T^{10}\tau^2 ϕ T = T − T τ − T 10 τ 2 . We state some relevant properties of ϕ \phi ϕ .
Lemma 6.4. Let ϕ \phi ϕ be the Drinfeld module defined in Example 1.7. Then the following hold.
End K ( ϕ ) = A \operatorname{End}_{K}(\phi)=A End K ( ϕ ) = A .
The associated rank-1 Drinfeld module given by ϕ T 1 = T + T 10 τ \phi^1_T=T+T^{10}\tau ϕ T 1 = T + T 10 τ is such that ϕ 1 \phi^1 ϕ 1 is not sign-normalized, ϕ 1 ̸ ≃ K ψ C a r l \phi^1\not\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl , and K ( ϕ 1 [ T ] ) = K ( ψ C a r l [ T ] ) K(\phi^1[T])=K(\psi^{\mathrm{Carl}}[T]) K ( ϕ 1 [ T ]) = K ( ψ Carl [ T ]) .
For every nonzero monic m ∈ A \mathfrak m\in A m ∈ A , Gal ( K ( ϕ [ m ] ) / K ) ≃ G L 2 ( A / m A ) \operatorname{Gal}(K(\phi[\mathfrak m])/K)\simeq\mathrm{GL}_2(A/\mathfrak m A) Gal ( K ( ϕ [ m ]) / K ) ≃ GL 2 ( A / m A ) .
For every monic m ∈ A \mathfrak m\in A m ∈ A , r m , T = 1 r_{\mathfrak m,T}=1 r m , T = 1 .
Proof. Recall that ϕ = u − 1 ϕ 0 u \phi=u^{-1}\phi_0u ϕ = u − 1 ϕ 0 u , where u 4 = − T u^4=-T u 4 = − T . Since ϕ \phi ϕ and ϕ 0 \phi_0 ϕ 0 are isomorphic over K ‾ \overline K K , we have an isomorphism of A A A -algebras
End K ‾ ( ϕ ) ≅ A End K ‾ ( ϕ 0 ) . (118) \operatorname{End}_{\overline K}(\phi)\cong_A\operatorname{End}_{\overline K}(\phi_0). \tag*{(118)} End K ( ϕ ) ≅ A End K ( ϕ 0 ) . (118)
Now, it follows from [30 Theorem 7.5.17(5)] , that since the j j j -invariant (see [30 Definition 3.8.3] ) j ( ϕ 0 ) = − 1 6 / T 4 = − T − 4 j(\phi_0)=-1^6/T^4=-T^{-4} j ( ϕ 0 ) = − 1 6 / T 4 = − T − 4 is not integral over A A A , we have End K ‾ ( ϕ 0 ) = A \operatorname{End}_{\overline K}(\phi_0)=A End K ( ϕ 0 ) = A . Thus, by (6.23), End K ‾ ( ϕ ) = A \operatorname{End}_{\overline K}(\phi)=A End K ( ϕ ) = A . This proves part (1).
We next prove part (2). By (1.16), the associated rank-1 Drinfeld module is given by ϕ T 1 = T + T 10 τ \phi^1_T=T+T^{10}\tau ϕ T 1 = T + T 10 τ , and is not sign-normalized since T 10 ∉ F 5 ∗ T^{10}\notin\mathbb F_5^* T 10 ∈ / F 5 ∗ . If ϕ 1 ≃ K ψ C a r l \phi^1\simeq_K\psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl , then there exists some v ∈ K ∗ v\in K^* v ∈ K ∗ such that ϕ T 1 = v − 1 ψ T C a r l v = T + v 4 τ \phi_T^1 = v^{-1}\psi^{\mathrm{Carl}}_T v = T + v^4\tau ϕ T 1 = v − 1 ψ T Carl v = T + v 4 τ , and hence T 10 T^{10} T 10 must be a fourth power in K ∗ K^* K ∗ , which is not possible. Hence ϕ 1 ̸ ≃ K ψ C a r l \phi^1 \not\simeq_K \psi^{\mathrm{Carl}} ϕ 1 ≃ K ψ Carl .
Now, put η : = T − 2 u − 1 \eta:= T^{-2}u^{-1} η := T − 2 u − 1 . Since u 4 = − T u^4 = -T u 4 = − T , we have η 4 = − T − 9 \eta^4 = -T^{-9} η 4 = − T − 9 . Therefore, ϕ T 1 ( η ) = T η + T 10 η 5 = η ( T + T 10 η 4 ) = 0 \phi_T^1(\eta) = T\eta+ T^{10}\eta^5 = \eta(T + T^{10}\eta^4) = 0 ϕ T 1 ( η ) = T η + T 10 η 5 = η ( T + T 10 η 4 ) = 0 . Thus, η \eta η is a nonzero T T T -torsion point of ϕ 1 \phi^1 ϕ 1 . Since ϕ 1 [ T ] ≃ A / ( T ) \phi^1[T] \simeq A/(T) ϕ 1 [ T ] ≃ A / ( T ) is a free A / ( T ) A/(T) A / ( T ) -module of rank 1, K ( ϕ 1 [ T ] ) = K ( η ) K(\phi^1[T]) = K(\eta) K ( ϕ 1 [ T ]) = K ( η ) . On the other hand, K ( η ) = K ( u ) K(\eta) = K(u) K ( η ) = K ( u ) , and u u u is a nonzero T T T -torsion point of ψ C a r l \psi^{\mathrm{Carl}} ψ Carl . Hence K ( ϕ 1 [ T ] ) = K ( u ) = K ( ψ C a r l [ T ] ) K(\phi^1[T]) = K(u) = K(\psi^{\mathrm{Carl}}[T]) K ( ϕ 1 [ T ]) = K ( u ) = K ( ψ Carl [ T ]) . This proves part (2).
We now prove part (3). Since ϕ = u − 1 ϕ 0 u \phi= u^{-1}\phi_0u ϕ = u − 1 ϕ 0 u and u ∈ K ( ψ C a r l [ T ] ) u \in K(\psi^{\mathrm{Carl}}[T]) u ∈ K ( ψ Carl [ T ]) , we have
K ( ϕ [ m ] ) K ( ψ C a r l [ T ] ) = K ( ϕ 0 [ m ] ) K ( ψ C a r l [ T ] ) (119) K(\phi[\mathfrak m])K(\psi^{\mathrm{Carl}}[T]) = K(\phi_0[\mathfrak m])K(\psi^{\mathrm{Carl}}[T]) \tag*{(119)} K ( ϕ [ m ]) K ( ψ Carl [ T ]) = K ( ϕ 0 [ m ]) K ( ψ Carl [ T ]) (119)
for every nonzero monic m ∈ A \mathfrak m \in A m ∈ A . Suppose first that T ∣ m T \mid\mathfrak m T ∣ m . By part (2), K ( ψ C a r l [ T ] ) = K ( ϕ 1 [ T ] ) ⊆ K ( ϕ [ m ] ) K(\psi^{\mathrm{Carl}}[T]) = K(\phi^1[T]) \subseteq K(\phi[\mathfrak m]) K ( ψ Carl [ T ]) = K ( ϕ 1 [ T ]) ⊆ K ( ϕ [ m ]) , which implies that the left hand side of (119) is equal to K ( ϕ [ m ] ) K(\phi[\mathfrak m]) K ( ϕ [ m ]) . On the other hand, (6.22) gives K ( ψ C a r l [ T ] ) = K ( ϕ 0 1 [ T ] ) ⊆ K ( ϕ 0 [ m ] ) K(\psi^{\mathrm{Carl}}[T]) = K(\phi_0^1[T]) \subseteq K(\phi_0[\mathfrak m]) K ( ψ Carl [ T ]) = K ( ϕ 0 1 [ T ]) ⊆ K ( ϕ 0 [ m ]) , which implies that the right hand side of (119) is equal to K ( ϕ 0 [ m ] ) K(\phi_0[\mathfrak m]) K ( ϕ 0 [ m ]) . Hence, we obtain K ( ϕ [ m ] ) = K ( ϕ 0 [ m ] ) K(\phi[\mathfrak m]) = K(\phi_0[\mathfrak m]) K ( ϕ [ m ]) = K ( ϕ 0 [ m ]) , and by (6.21), Gal ( K ( ϕ 0 [ m ] ) / K ) ≃ GL 2 ( A / m A ) \operatorname{Gal}(K(\phi_0[\mathfrak m])/K) \simeq\operatorname{GL}_2(A/\mathfrak m A) Gal ( K ( ϕ 0 [ m ]) / K ) ≃ GL 2 ( A / m A ) . Now suppose that gcd ( m , T ) = 1 \gcd(\mathfrak m,T) = 1 g cd( m , T ) = 1 . Applying (6.21) and using CRT gives
K ( ϕ 0 [ m ] ) ∩ K ( ϕ 0 [ T ] ) = K . (120) K(\phi_0[\mathfrak m]) \cap K(\phi_0[T]) = K. \tag*{(120)} K ( ϕ 0 [ m ]) ∩ K ( ϕ 0 [ T ]) = K . (120)
Since K ( ψ C a r l [ T ] ) = K ( ϕ 0 1 [ T ] ) ⊆ K ( ϕ 0 [ T ] ) K(\psi^{\mathrm{Carl}}[T]) = K(\phi_0^1[T]) \subseteq K(\phi_0[T]) K ( ψ Carl [ T ]) = K ( ϕ 0 1 [ T ]) ⊆ K ( ϕ 0 [ T ]) , it follows from (120) that K ( ϕ 0 [ m ] ) ∩ K ( ψ C a r l [ T ] ) = K K(\phi_0[\mathfrak m]) \cap K(\psi^{\mathrm{Carl}}[T]) = K K ( ϕ 0 [ m ]) ∩ K ( ψ Carl [ T ]) = K . Therefore, by (119) ,
∣ GL 2 ( A / m A ) ∣ = [ K ( ϕ 0 [ m ] ) : K ] = [ K ( ϕ [ m ] ) K ( ψ C a r l [ T ] ) : K ( ψ C a r l [ T ] ) ] = [ K ( ϕ [ m ] ) : K ( ϕ [ m ] ) ∩ K ( ψ C a r l [ T ] ) ] ≤ [ K ( ϕ [ m ] ) : K ] ≤ ∣ GL 2 ( A / m A ) ∣ . \begin{aligned}
|\operatorname{GL}_2(A/\mathfrak m A)| &= [K(\phi_0[\mathfrak m]) : K] = [K(\phi[\mathfrak m])K(\psi^{\mathrm{Carl}}[T]) : K(\psi^{\mathrm{Carl}}[T])] \\
&= [K(\phi[\mathfrak m]) : K(\phi[\mathfrak m]) \cap K(\psi^{\mathrm{Carl}}[T])] \\
&\le[K(\phi[\mathfrak m]) : K] \le|\operatorname{GL}_2(A/\mathfrak m A)|.
\end{aligned} ∣ GL 2 ( A / m A ) ∣ = [ K ( ϕ 0 [ m ]) : K ] = [ K ( ϕ [ m ]) K ( ψ Carl [ T ]) : K ( ψ Carl [ T ])] = [ K ( ϕ [ m ]) : K ( ϕ [ m ]) ∩ K ( ψ Carl [ T ])] ≤ [ K ( ϕ [ m ]) : K ] ≤ ∣ GL 2 ( A / m A ) ∣.
Hence [ K ( ϕ [ m ] ) : K ] = ∣ GL 2 ( A / m A ) ∣ [K(\phi[\mathfrak m]) : K] = |\operatorname{GL}_2(A/\mathfrak m A)| [ K ( ϕ [ m ]) : K ] = ∣ GL 2 ( A / m A ) ∣ , which proves part (3).
It remains to prove part (4). The second assertion of [39 Theorem 1.2] implies that, for every nonzero monic m ∈ A \mathfrak m \in A m ∈ A ,
K ( ϕ 0 [ m ] ) ∩ F ‾ 5 = F 5 . (121) K(\phi_0[\mathfrak m]) \cap\overline{\mathbb{F}}_5 = \mathbb{F}_5. \tag*{(121)} K ( ϕ 0 [ m ]) ∩ F 5 = F 5 . (121)
Using (119) , the inclusion K ( ψ C a r l [ T ] ) = K ( ϕ 0 1 [ T ] ) ⊆ K ( ϕ 0 [ T ] ) K(\psi^{\mathrm{Carl}}[T]) = K(\phi_0^1[T]) \subseteq K(\phi_0[T]) K ( ψ Carl [ T ]) = K ( ϕ 0 1 [ T ]) ⊆ K ( ϕ 0 [ T ]) , and Lemma 5.6, we obtain
K ( ϕ [ m ] ) K ( ψ C a r l [ T ] ) = K ( ϕ 0 [ m ] ) K ( ψ C a r l [ T ] ) ⊆ K ( ϕ 0 [ lcm ( m , T ) ] ) . K(\phi[\mathfrak m])K(\psi^{\mathrm{Carl}}[T]) = K(\phi_0[\mathfrak m])K(\psi^{\mathrm{Carl}}[T]) \subseteq K(\phi_0[\operatorname{lcm}(\mathfrak m,T)]). K ( ϕ [ m ]) K ( ψ Carl [ T ]) = K ( ϕ 0 [ m ]) K ( ψ Carl [ T ]) ⊆ K ( ϕ 0 [ lcm ( m , T )]) .
Thus, by (121) , K ( ϕ [ m ] ) K ( ψ C a r l [ T ] ) ∩ F ‾ 5 = F 5 K(\phi[\mathfrak m])K(\psi^{\mathrm{Carl}}[T]) \cap\overline{\mathbb{F}}_5 = \mathbb{F}_5 K ( ϕ [ m ]) K ( ψ Carl [ T ]) ∩ F 5 = F 5 . Consequently, r m , T = 1 r_{\mathfrak m,T} = 1 r m , T = 1 for every monic m ∈ A \mathfrak m \in A m ∈ A . This proves part (4). □
Now, by part (3) of Lemma 6.4 and the Chinese Remainder Theorem, we may take M ( ϕ ) = 1 M(\phi) = 1 M ( ϕ ) = 1 (see Lemma 5.1). Hence,
N ( ϕ ) = ∏ deg p = 1 p , B = { p primes : deg p = 1 } , G = { p primes : deg p ≠ 1 } . N(\phi) = \prod_{\deg\mathfrak p=1} \mathfrak p,\qquad\mathcal{B} = \{\mathfrak p\text{ primes} : \deg\mathfrak p = 1\},\qquad\mathcal{G} = \{\mathfrak p\text{ primes} : \deg\mathfrak p \ne1\}. N ( ϕ ) = d e g p = 1 ∏ p , B = { p primes : deg p = 1 } , G = { p primes : deg p = 1 } .
Furthermore, part (2) of Lemma 6.4 gives
K ( ψ C a r l [ T ] ) = K ( ϕ 1 [ T ] ) ⊆ K ( ϕ [ T ] ) ⊆ D B . (122) K(\psi^{\mathrm{Carl}}[T]) = K(\phi^1[T]) \subseteq K(\phi[T]) \subseteq D_{\mathcal{B}}. \tag*{(122)} K ( ψ Carl [ T ]) = K ( ϕ 1 [ T ]) ⊆ K ( ϕ [ T ]) ⊆ D B . (122)
Also, by Lemma 5.2, for every square-free monic m \mathfrak m m supported on G \mathcal{G} G , K ( ϕ [ m ] ) ∩ D B = K K(\phi[\mathfrak m]) \cap D_{\mathcal{B}} = K K ( ϕ [ m ]) ∩ D B = K . Together with (122) , this gives K ( ϕ [ m ] ) ∩ D B K ( ψ C a r l [ T ] ) = K K(\phi[\mathfrak m]) \cap D_{\mathcal{B}}K(\psi^{\mathrm{Carl}}[T]) = K K ( ϕ [ m ]) ∩ D B K ( ψ Carl [ T ]) = K . Thus, we may take S = ∅ S = \varnothing S = ∅ as an entanglement set for ( ϕ , T ) (\phi,T) ( ϕ , T ) , and L ϕ , T , ∅ = D B L_{\phi,T,\varnothing} = D_{\mathcal{B}} L ϕ , T , ∅ = D B . Therefore, by Theorem 1.4, in order to prove Example 1.7, it remains only to show that Ω ϕ , T , a , ∅ ≠ ∅ \Omega_{\phi,T,a,\varnothing} \ne\varnothing Ω ϕ , T , a , ∅ = ∅ for every a ∈ ( A / ( T ) ) ∗ a \in(A/(T))^* a ∈ ( A / ( T ) ) ∗ .
Proposition 6.5. For every a ∈ ( A / ( T ) ) ∗ a \in(A/(T))^* a ∈ ( A / ( T ) ) ∗ , one has Ω ϕ , T , a , ∅ ≠ ∅ \Omega_{\phi,T,a,\varnothing} \ne\varnothing Ω ϕ , T , a , ∅ = ∅ . Proof. Fix a ∈ ( A / ( T ) ) ∗ \mathfrak{a} \in(A/(T))^{*} a ∈ ( A / ( T ) ) ∗ . Put b : = ∏ p ∈ B p \mathfrak{b} := \prod_{\mathfrak{p}\in\mathcal{B}} \mathfrak{p} b := ∏ p ∈ B p . Then, by Lemma 5.6, D B = K ( ϕ [ b ] ) D_{\mathfrak{B}} = K(\phi[\mathfrak{b}]) D B = K ( ϕ [ b ]) . Recall that L ϕ , T , ∅ = D B L_{\phi,T,\varnothing} = D_{\mathfrak{B}} L ϕ , T , ∅ = D B . Hence, part (3) of Lemma 6.4, together with the Chinese Remainder Theorem, gives
Gal ( L ϕ , T , ∅ / K ) ≃ ∏ p ∈ B Gal ( K ( ϕ [ p ] ) / K ) . (123) \operatorname{Gal}(L_{\phi,T,\varnothing}/K) \simeq\prod_{\mathfrak{p}\in\mathcal{B}} \operatorname{Gal}(K(\phi[\mathfrak{p}])/K). \tag*{(123)} Gal ( L ϕ , T , ∅ / K ) ≃ p ∈ B ∏ Gal ( K ( ϕ [ p ]) / K ) . (123)
For every p ∈ B \mathfrak{p} \in\mathcal{B} p ∈ B with p ≠ T \mathfrak{p} \ne T p = T , choose any 1 ≠ σ p ∈ Gal ( K ( ϕ [ p ] ) / K ) 1 \ne\sigma_{\mathfrak{p}} \in\operatorname{Gal}(K(\phi[\mathfrak{p}])/K) 1 = σ p ∈ Gal ( K ( ϕ [ p ]) / K ) . Now, since K ( ψ C a r l [ T ] ) = K ( ϕ 1 [ T ] ) ⊆ K ( ϕ [ T ] ) K(\psi^{\mathrm{Carl}}[T]) = K(\phi^{1}[T]) \subseteq K(\phi[T]) K ( ψ Carl [ T ]) = K ( ϕ 1 [ T ]) ⊆ K ( ϕ [ T ]) , the restriction map Gal ( K ( ϕ [ T ] ) / K ) ⟶ Gal ( K ( ψ C a r l [ T ] ) / K ) \operatorname{Gal}(K(\phi[T])/K) \longrightarrow\operatorname{Gal}(K(\psi^{\mathrm{Carl}}[T])/K) Gal ( K ( ϕ [ T ]) / K ) ⟶ Gal ( K ( ψ Carl [ T ]) / K ) is surjective. Hence, we may choose σ T ∈ Gal ( K ( ϕ [ T ] ) / K ) \sigma_T \in\operatorname{Gal}(K(\phi[T])/K) σ T ∈ Gal ( K ( ϕ [ T ]) / K ) such that σ T ( λ T ) = ψ a C a r l ( λ T ) \sigma_T(\lambda_T) = \psi_{\mathfrak{a}}^{\mathrm{Carl}}(\lambda_T) σ T ( λ T ) = ψ a Carl ( λ T ) . We may further require σ T ≠ 1 \sigma_T \ne1 σ T = 1 . Indeed, this is automatic if ψ a C a r l ( λ T ) ≠ λ T \psi_{\mathfrak{a}}^{\mathrm{Carl}}(\lambda_T) \ne\lambda_T ψ a Carl ( λ T ) = λ T , while if ψ a C a r l ( λ T ) = λ T \psi_{\mathfrak{a}}^{\mathrm{Carl}}(\lambda_T) = \lambda_T ψ a Carl ( λ T ) = λ T , we choose any element 1 ≠ σ T ∈ Gal ( K ( ϕ [ T ] ) / K ( ψ C a r l [ T ] ) ) 1 \ne\sigma_T \in\operatorname{Gal}(K(\phi[T])/K(\psi^{\mathrm{Carl}}[T])) 1 = σ T ∈ Gal ( K ( ϕ [ T ]) / K ( ψ Carl [ T ])) , which is nontrivial since
[ K ( ϕ [ T ] ) : K ] = ∣ GL 2 ( F 5 ) ∣ > 4 = [ K ( ψ C a r l [ T ] ) : K ] . [K(\phi[T]) : K] = |\operatorname{GL}_{2}(\mathbb{F}_{5})| > 4 = [K(\psi^{\mathrm{Carl}}[T]) : K]. [ K ( ϕ [ T ]) : K ] = ∣ GL 2 ( F 5 ) ∣ > 4 = [ K ( ψ Carl [ T ]) : K ] .
Thus, by construction, σ T ( λ T ) = ψ a C a r l ( λ T ) \sigma_T(\lambda_T) = \psi_{\mathfrak{a}}^{\mathrm{Carl}}(\lambda_T) σ T ( λ T ) = ψ a Carl ( λ T ) and σ p ≠ 1 \sigma_{\mathfrak{p}} \ne1 σ p = 1 for every p ∈ B \mathfrak{p} \in\mathcal{B} p ∈ B . By (6.28), we then obtain σ ∈ Gal ( L ϕ , T , ∅ / K ) \sigma\in\operatorname{Gal}(L_{\phi,T,\varnothing}/K) σ ∈ Gal ( L ϕ , T , ∅ / K ) such that σ ∣ K ( ϕ [ p ] ) = σ p \sigma|_{K(\phi[\mathfrak{p}])} = \sigma_{\mathfrak{p}} σ ∣ K ( ϕ [ p ]) = σ p for every p ∈ B \mathfrak{p} \in\mathcal{B} p ∈ B . Since
Ω ϕ , T , a , ∅ = { σ ∈ Gal ( K ( ϕ [ b ] ) / K ) : σ ∣ K ( ψ C a r l [ T ] ) ( λ T ) = ψ a C a r l ( λ T ) , σ ∣ K ( ϕ [ p ] ) ≠ 1 for every p ∈ B } , \Omega_{\phi,T,\mathfrak{a},\varnothing} = \{\sigma\in\operatorname{Gal}(K(\phi[\mathfrak{b}])/K) : \sigma|_{K(\psi^{\mathrm{Carl}}[T])}(\lambda_T) = \psi_{\mathfrak{a}}^{\mathrm{Carl}}(\lambda_T), \sigma|_{K(\phi[\mathfrak{p}])} \ne1 \text{ for every } \mathfrak{p} \in\mathcal{B}\}, Ω ϕ , T , a , ∅ = { σ ∈ Gal ( K ( ϕ [ b ]) / K ) : σ ∣ K ( ψ Carl [ T ]) ( λ T ) = ψ a Carl ( λ T ) , σ ∣ K ( ϕ [ p ]) = 1 for every p ∈ B } ,
it then follows that σ ∈ Ω ϕ , T , a , ∅ \sigma\in\Omega_{\phi,T,\mathfrak{a},\varnothing} σ ∈ Ω ϕ , T , a , ∅ . Hence Ω ϕ , T , a , ∅ ≠ ∅ \Omega_{\phi,T,\mathfrak{a},\varnothing} \ne\varnothing Ω ϕ , T , a , ∅ = ∅ .
Thus, by Proposition 6.5 and Theorem 1.4, δ ϕ , T , a > 0 \delta_{\phi,T,\mathfrak{a}} > 0 δ ϕ , T , a > 0 for every a ∈ ( A / ( T ) ) ∗ \mathfrak{a} \in(A/(T))^{*} a ∈ ( A / ( T ) ) ∗ . The remaining assertions of Example 1.7 follow from Lemma 6.4. This completes the proof.
Acknowledgment
Arnab Saha is supported by the ARG-MATRICS Grant (Grant No. ANRF/ARGM/2025/001294/MTR)
References
Department of Mathematics, Indian Institute of Technology Gandhinagar, Gandhinagar, Gujarat 382355, India
Email address: prasanna.jha@iitgn.ac.in
Department of Mathematics, Indian Institute of Technology Gandhinagar, Gandhinagar, Gujarat 382355, India
Email address: arnab.saha@iitgn.ac.in
Department of Mathematics, Indian Institute of Technology Gandhinagar, Gandhinagar, Gujarat 382355, India
Email address: akshaa.vatwani@iitgn.ac.in