This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Algebraic Independence of Special Points on Shimura Varieties

Yu Fu and Roy Zhao
Abstract.

Given a correspondence VV between a connected Shimura variety SS and a commutative connected algebraic group GG, and nn\in\mathbb{N}, we prove that the VV-images of any nn special points on SS outside a proper Zariski closed subset are algebraically independent. Our result unifies previous unlikely intersection results on multiplicative independence and linear independence. We prove multiplicative independence of differences of singular moduli, generalizing previous results by Pila–Tsimerman, and Aslanyan–Eterović–Fowler. We also give an application to abelian varieties by proving that the special points of SS whose VV-images lie in a finite-rank subgroup of GG are contained in a finite union of proper special subvarieties of SS, only dependent on the rank of the subgroup. In this way, our result is a generalization of the works of Pila–Tsimerman and Buium–Poonen.

2020 Mathematics Subject Classification:
Primary: 11G18; Secondary: 03C64, 11G10, 14G35

1. Introduction

1.1. Statement of the results

Let SS be a connected Shimura variety, GG be a commutative connected algebraic group. Let VS×GV\subset S\times G be a proper irreducible closed subvariety such that the projection on each factor is dominant, and VV is finite over SS. Given a point (s,g)VS×G(s,g)\in V\subset S\times G, we call gg a VV-image of ss. Moreover, we say that a series of points g1,,gnGg_{1},\dots,g_{n}\in G are algebraically independent if they do not all lie in a proper algebraic subgroup.

The special subvarieties in GG are torsion cosets of connected algebraic subgroups and the special subvarieties in a Shimura variety are described in Section 2.A. We prove the independence of the VV-images of special points of SS. Previous work in the literature proved the linear independence of special points when VV is a correspondence from a modular curve or a Shimura curve to an elliptic curve, i.e. the linear dependences among Heegner points, which play an important role in the study of the Birch and Swinnerton-Dyer Conjecture. See [BP09], [Bal18], [PT22] for results in this direction. There have also been results proving multiplicative independence of special points when VV is the graph of the jj-function from Y(1)𝔾mY(1)\to\mathbb{G}_{m} (see [PT17], [AEF23]). The aim of this article is to provide a unified framework to generalize the previous works to the study of algebraic independence of special points in general commutative groups.

To state our main theorem, we give the definition of exemplary components following [PT22].

Definition 1.1.

With the notation as above, let πS\pi_{S} and πG\pi_{G} be the projections of S×GS\times G onto the first and second factors respectively.

  • A distinguished component is an irreducible component of WV(S×G)W\subset V\cap(S^{\prime}\times G^{\prime}), where S×GS^{\prime}\times G^{\prime} is a special subvariety of S×GS\times G, such that πS(W)=S\pi_{S}(W)=S^{\prime}.

  • Let WW be a distinguished component and let BπG(W)B\supset\pi_{G}(W) be the smallest special subvariety of GG containing πG(W)\pi_{G}(W). We say that WW is exemplary if there are no larger distinguished components WWW^{\prime}\supset W such that πG(W)B\pi_{G}(W^{\prime})\subset B.

  • We say a distinguished component WW in GG is dependent if GGG^{\prime}\subset G is strict. We note that the unique non-dependent exemplary component is VV itself.

Pila and Tsimerman [PT22, Theorem 1.1] were able to deal with the case where S=Y(1)nS=Y(1)^{n} is the nn-copy of a modular (or Shimura) curve and G=EnG=E^{n} is the self-nn-product of an elliptic curve for n1n\geq 1. We prove the following theorem, which is a generalization of their main theorem to arbitrary (connected) Shimura varieties and commutative groups. The theorem describes all algebraic dependence among VV-images of nn special points for any n1n\geq 1.

Theorem 1.2.

Let SS be a connected Shimura variety, GG be a commutative connected algebraic group. Suppose VS×GV\subset S\times G is an irreducible subvariety that maps finitely to SS. Then, there are only finitely many exemplary components in VV.

Our proof of Theorem 1.2 follows a similar path to [PT22], via point-counting on definable sets in an o-minimal structures and using an Ax–Schanuel theorem in a suitable form. The point-counting result follows from the work of Habegger and Pila in [HP16] and the Ax–Schanuel theorem follows from Blázquez-Sanz, Casale, Freitag, and Nagloo in [BSCFN23].

Results in this direction have various kinds of applications. As a first consequence of Theorem 1.2, we prove the following corollary and a weaker version, Corollary 5.2, which might be easier to apply in practice.

Corollary 1.3.

Let S,G,VS,G,V be as above. Suppose moreover that the projections of VV to SS and GG are both dominant. Fix nn\in\mathbb{N}. There exists a proper Zariski closed subset SSnS^{\prime}\subset S^{n} such that for any nn points s1,,snSs_{1},\dots,s_{n}\in S satisfying (s1,,sn)S(s_{1},\dots,s_{n})\not\in S^{\prime}, any VV-images g1,,gng_{1},\dots,g_{n} of s1,,sns_{1},\dots,s_{n} are algebraically independent.

Proof.

For n1n\geq 1, let 𝒔=(s1,,sn)\boldsymbol{s}=(s_{1},\cdots,s_{n}) be a special point in SnS^{n} and let 𝒈=(g1,,gn)\boldsymbol{g}=(g_{1},\cdots,g_{n}) be a point in GnG^{n} such that (𝒔,𝒈)Vn(\boldsymbol{s},\boldsymbol{g})\in V^{n}. It follows from the definition that if (𝒔,𝒈)W(\boldsymbol{s},\boldsymbol{g})\in W for some dependent exemplary component then g1,,gng_{1},\cdots,g_{n} are algebraically dependent in GG. Conversely, if g1,,gng_{1},\cdots,g_{n} are algebraically dependent in GG, then (𝒔,𝒈)(\boldsymbol{s},\boldsymbol{g}) is a distinguished point and hence contained in some dependent exemplary component. ∎

Let DD be a positive integer. For a set s1,,sns_{1},\cdots,s_{n} of special points in SS, we introduce the notion of DD-independent in Definition 5.3. The discriminant Δ(s)\Delta(s) of a special point is defined in Definition 3.1. We prove several results towards linear independence in abelian varieties of dimension g3g\leq 3, utilizing the classification of special subvarieties in the moduli space of abelian varieties 𝒜g\mathcal{A}_{g}.

Corollary 1.4.

Suppose that S𝒜gS\subset\mathcal{A}_{g} for g3g\leq 3. For n1n\geq 1 there exists a positive integer D=D(n,S,G,V)D=D(n,S,G,V) such that if s1,,sns_{1},\dots,s_{n} are DD-independent, then any VV-image x1,,xnx_{1},\dots,x_{n} of s1,,sns_{1},\dots,s_{n} are algebraically independent in GG.

Taking VV to be the graph of the function f:Y(1)n×Y(1)nnf\colon Y(1)^{n}\times Y(1)^{n}\to\mathbb{C}^{n} given by

f(z1,,zn,w1,,wn)=(z1w1,,znwn),f(z_{1},\dots,z_{n},w_{1},\dots,w_{n})=(z_{1}-w_{1},\dots,z_{n}-w_{n}),

we get the following generalization of [PT17, Thm. 1.2] and [AEF23, Thm. 1.1].

Corollary 1.5.

For n1n\geq 1, there exists a positive integer D=D(n)D=D(n) such that if x1,,xn,x_{1},\dots,x_{n}, y1,,ynY(1)y_{1},\dots,y_{n}\in Y(1) are singular moduli that are DD-independent, then there does not exist (a1,,an)n\{𝟎}(a_{1},\dots,a_{n})\in\mathbb{Z}^{n}\backslash\{\boldsymbol{0}\} such that

i=1n(xiyi)aiμ,\prod_{i=1}^{n}(x_{i}-y_{i})^{a_{i}}\in\mu_{\infty},

where μ\mu_{\infty} is the set of roots of unity.

We prove also that on low dimensional abelian varieties, the intersection of any finite-rank subgroup of GG with the set of VV-images of special points on SS is contained in a finite union of proper special subvarieties. This generalizes Pila and Tsimerman’s result [PT22, Corollary 1.4] to higher dimensional abelian varieties, and Buium and Poonen’s [BP09, Theorem 1.1] in a uniform way, depending only on the rank of the subgroup.

Theorem 1.6.

Suppose GG is defined over a field KK of characteristic 0 and suppose the Shimura datum for SS has GSG_{S} simple. Let ΓG(K)\Gamma\subset G(K) be a subgroup of rank rr and let Γ\Gamma^{\prime} be its division group. Then, there exist N(r,S,G,V)N(r,S,G,V) proper special subvarieties of SS such that the set of special points of SS with a VV-image in Γ\Gamma^{\prime} is contained in a union of these proper special subvarieties.

A similar argument to [PT22, Proposition 3.2] proves that Theorem 1.2 is a consequence of the Zilber–Pink Conjecture. Also, Theorem 1.6 may be thought as a weak version of the André–Oort–Mordell–Lang conjecture in the case of subgroups of finite rank formulated by Baldi in [Bal18, Conjecture 5.2]. The original formulation of the André–Oort–Mordell–Lang conjecture is due to Pink, which can be found in [Pin05].

1.2. Organization of the paper

The notions of Shimura varieties, special and weakly special subvarieties are explained in Section 2. We also prove a mixed version of the Ax–Schanuel theorem for the uniformization of the product S×GS\times G in the form needed for our main theorem in this section. In Section 3, we collect and prove some arithmetic estimates. The main theorem is proved in Section 4 and its applications and corollaries are proved in Section 5.

2. Shimura varieties and the mixed Ax–Schanuel

Let SS be an algebraic variety and q:XSq\colon X\to S be a universal covering of SS. The Ax–Schanuel theorem gives information about the bi-algebraic varieties, algebraic varieties VXV\subset X such that q(V)q(V) is also algebraic, for the transcendental covering map qq. The rough statement of the Ax–Schanuel theorem is that varieties of X×SX\times S should have a proper intersection with the graph of qq unless the projection of VV to SS is contained within a bi-algebraic variety. These bi-algebraic varieties are precisely the weakly special subvarieties. We first give precise definitions of bi-algebraic varieties and results in the case when SS is a commutative algebraic group or a Shimura variety, and then prove the result when SS is a product of the two.

2.1. Shimura Varieties

Let S:=ShK(GS,X)S:=\operatorname{Sh}_{K}(G_{S},X) be the (connected) Shimura variety associated with a connected Shimura datum (GS,X)(G_{S},X), where GG is a semisimple group of adjoint type and KK is a compact open subgroup of G(𝔸f).G(\mathbb{A}_{f}). More information about Shimura varieties can be found in Milne ([Mil05]).

For any Shimura subvariety ZZ of SS and any aG(𝔸f)a\in G(\mathbb{A}_{f}), we refer to any irreducible component of the Hecke correspondence TK,a(Z)T_{K,a}(Z) as a special subvariety of SS. A special point is a special subvariety of dimension zero.

For any xXx\in X, let 𝐌:=MT(x)\mathbf{M}:=\operatorname{MT}(x) be the Mumford-Tate group of xx, which is defined as the smallest \mathbb{Q}-subgroup HH of GG such that xx factors through HH_{\mathbb{R}}. Let X𝐌X_{\mathbf{M}} be the 𝐌()\mathbf{M}(\mathbb{R})-conjugacy class of xx. Then the image of X𝐌X_{\mathbf{M}} in S=Γg\XS=\Gamma_{g}\backslash X is a special subvariety of SS. It is not hard to see that every special subvariety of SS arises this way. The action of 𝐌()\mathbf{M}(\mathbb{R}) on X𝐌X_{\mathbf{M}} factors through the adjoint group 𝐌ad()\mathbf{M}^{\operatorname{ad}}(\mathbb{R}), which is a direct product of its \mathbb{Q}-simple factors. Therefore, we can write X𝐌X_{\mathbf{M}} as a product

X𝐌=X1×X2,X_{\mathbf{M}}=X_{1}\times X_{2},

corresponding to the action of Mad()M^{\operatorname{ad}}(\mathbb{R}). A weakly special subvariety is the image of the fiber {x1}×X2\{x_{1}\}\times X_{2} or X1×{x2}X_{1}\times\{x_{2}\} for any x1X1x_{1}\in X_{1} or x2X2x_{2}\in X_{2}. Therefore, a weakly special subvariety of SS is a special subvariety if and only if it contains a special point. The weakly special subvarieties of SS are precisely those subvarieties that are totally geodesic in SS ([Moo98, Section 4]), and they are also precisely the bi-algebraic subvarieties under the uniformization map q:XSq\colon X\to S. More detailed information on weakly special subvarieties of Shimura varieties can be found in [UY11].

The Ax–Schanuel theorem for Shimura varieties is a statement about the functional transcendence of the uniformization map qq.

Theorem 2.1 ([MPT19, Theorem 1.1]).

Let VX×SV\subset X\times S be an algebraic subvariety and let DX×SD\subset X\times S be the graph of q:XSq\colon X\to S. Let UVDU\subset V\cap D be an irreducible component such that

dimU>dimVdimS.\dim U>\dim V-\dim S.

Then, the projection of UU to SS is contained in a proper weakly special subvariety of SS.

2.2. Commutative Algebraic Groups

Let GG be a connected commutative algebraic group of dimension gg defined over \mathbb{C} and let exp:gG\exp\colon\mathbb{C}^{g}\to G be the exponential map from LieGg\operatorname{Lie}G\cong\mathbb{C}^{g} to GG.

Definition 2.2.

For each connected subgroup BGB\subset G and closed point pG()p\in G(\mathbb{C}), we say that the algebraic subvariety B+pGB+p\subset G is a weakly special subvariety of GG. We say that it is a special subvariety of GG if pGtorsp\in G_{\mathrm{tors}} is a torsion point.

The special points of GG are special subvarieties of dimension 0, and hence they are precisely the torsion points of GG. As before, a weakly special subvariety of GG is special precisely when it contains a special point.

The Ax–Schanuel theorem for commutative groups is a statement about covering map given by the Lie\operatorname{Lie}-exponential exp\exp.

Theorem 2.3 ([Ax72, Theorem 3]).

Let Vg×GV\subset\mathbb{C}^{g}\times G be an algebraic subvariety and let Dg×GD\subset\mathbb{C}^{g}\times G be the graph of exp:gG\exp\colon\mathbb{C}^{g}\to G. Let UVDU\subset V\cap D be an irreducible component such that

dimU>dimVdimG.\dim U>\dim V-\dim G.

Then the projection of UU to GG is contained in a proper weakly special subvariety of GG.

2.3. Mixed Ax–Schanuel

We will prove a version of the Ax–Schanuel theorem for the product uniformization map q×exp:X×gS×Gq\times\exp\colon X\times\mathbb{C}^{g}\to S\times G. This was proven by Pila and Tsimerman in [PT22] for the case when SS is a product of modular curves and GG is a product of elliptic curves. In that case, it was shown that the Ax–Schanuel theorem for the product follows from an Ax–Schanuel theorem for mixed Shimura varieties by Gao in [Gao20]. Using the same method, one can prove that the Ax–Schanuel theorem holds for the product S×GS\times G whenever SS is a Shimura variety of abelian type and GG is an abelian variety. However, to prove it for general Shimura varieties and general commutative groups, we will need stronger machinery from Blázquez-Sanz, Casale, Freitag, and Nagloo ([BSCFN23]). The translation into the language of (weakly) special subvarieties is given by Chiu ([Chi22]).

Definition 2.4.

Let SS be a Shimura variety and GG a connected commutative algebraic group. We say that an algebraic subvariety VS×GV\subset S\times G is a weakly special subvariety of S×GS\times G if there exist weakly special subvarieties SHSS_{H}\subset S and BGB\subset G such that V=SH×BV=S_{H}\times B. We say that VV is a special subvariety if SH,BS_{H},B are both special.

Theorem 2.5.

Let V(X×g)×(S×G)V\subset(X\times\mathbb{C}^{g})\times(S\times G) be an algebraic subvariety and let D(X×g)×(S×G)D\subset(X\times\mathbb{C}^{g})\times(S\times G) be the graph of q×exp:(X×g)(S×G)q\times\exp\colon(X\times\mathbb{C}^{g})\to(S\times G). Let UVDU\subset V\cap D be an irreducible component such that

dimU>dimVdim(S×G).\dim U>\dim V-\dim(S\times G).

Then the projection of UU to S×GS\times G is contained in a proper weakly special subvariety of S×GS\times G.

Proof.

Let the Shimura datum of SS be (GS,X)(G_{S},X). We may take SS to be a connected Shimura variety and GG to be a derived group. Applying [BSCFN23, Thm. 3.6] to D(X×g)×(S×G)D\subset(X\times\mathbb{C}^{g})\times(S\times G) gives that the projection of UU in S×GS\times G is contained in a proper subvariety whose Galois group is a strict algebraic subgroup HGS()×𝔾a()gH\subset G_{S}(\mathbb{C})\times\mathbb{G}_{a}(\mathbb{C})^{g}. Since GG is a derived group, it has no abelian quotients and hence Goursat’s lemma says that the projection of HH to G()G(\mathbb{C}) or 𝔾a()g\mathbb{G}_{a}(\mathbb{C})^{g} is not onto. By [Chi22, Thm. 3.2] for the Shimura variety side and [Ax72, Thm. 3] for the commutative group side, that means the projection of UU to SS or GG is contained in a proper weakly special subvariety. ∎

3. Bounds on the Galois orbits

In this section we give some arithmetic estimates which will be used later in the proof of the theorems.

Fix a choice of fundamental domain S\mathcal{F}_{S} for the uniformization q:XSq:X\to S of the Shimura variety SS. Let FF be the number field over which SS admits a canonical model. The degree of FF is bounded in terms of the datum (GS,X)(G_{S},X) and KK. All the special points of SS are algebraic points defined over abelian extensions of FF.

For a special point sSs\in S, let xx be a preimage of ss, i.e. xx is a pre-special point. By definition, the Mumford–Tate group of a special point is an algebraic torus. Let K𝐌mK^{m}_{\mathbf{M}} be the maximal compact open subgroup of 𝐌(𝔸f)\mathbf{M}(\mathbb{A}_{f}) and K𝐌K_{\mathbf{M}} the compact open subgroup K𝐌(𝔸f)K\cap\mathbf{M}(\mathbb{A}_{f}) of 𝐌(𝔸f)\mathbf{M}(\mathbb{A}_{f}). Let EE be the splitting field of 𝐌\mathbf{M}. Since GSG_{S} is of adjoint type, EE is a (Galois) CM field. Let DED_{E} be the absolute value of the discriminant of EE.

Definition 3.1.

The discriminant of ss is

Δ(s)[K𝐌m:K𝐌]DE.\Delta(s)\coloneqq[K^{m}_{\mathbf{M}}:K_{\mathbf{M}}]D_{E}.

We want to estimate the heights and degrees of special and pre-special points in terms of Δ(s).\Delta(s). Fortunately, recent progress allows us to have the following proposition.

Proposition 3.2.

Let sSs\in S be a special point with discriminant Δ(s)\Delta(s) and let xx be a preimage of ss under the uniformization map qq. Let hh be the canonical height on SS, which is a logarithmic Weil height, and let HH be the multiplicative Weil height for a fixed realization of XX. Then we have

  • (a)

    h(s)|Δ(s)|o(1)h(s)\leq|\Delta(s)|^{o(1)};

  • (b)

    H(x)C1|Δ(s)|C2H(x)\leq C_{1}|\Delta(s)|^{C_{2}};

  • (c)

    [F(s):F]|Δ(s)|1/2+ε[F(s):F]\ll|\Delta(s)|^{1/2+\varepsilon} for any ε>0\varepsilon>0;

  • (d)

    [F(s):F]C3|Δ(s)|c[F(s):F]\gg C_{3}|\Delta(s)|^{c} for some fixed c>0c>0.

Where the constants C1,C2,C_{1},C_{2},\dots are depend on GS,X,F,SG_{S},X,F,\mathcal{F}_{S} and the realization.

Proof.
  • (a)

    Follows from [PST+21, Theorem 9.11];

  • (b)

    See [DO16, Theorem 1.1, 4.1];

  • (c)

    Follows from the Brauer-Siegel theorem for arithmetic tori in [Tsi11, Theorem 1.3];

  • (d)

    See [BSY22, Theorem 1].

Over ¯\overline{\mathbb{Q}}, any commutative connected group GG can be written as a product of a semi-abelian variety with affine space G×𝔾amG^{\prime}\times\mathbb{G}_{a}^{m} (see [NW14, Prop. 5.1.12]). We will need height bounds on the semi-abelian factor, so suppose GG is a semi-abelian variety over a number field LL with toric part TT and abelian quotient π:GA\pi:G\to A. Let gg be the dimension of AA. Let KK be any number field. Note that the Weil height on a semi-abelian variety might be negative. However, the canonical height h^L\widehat{h}_{L} on GG defined by Kühne [Kü20, Sec. 3], with respect to a TT-effective line bundle (M,ϱ)(M,\varrho) on a TT-equivariant compactification T¯\overline{T} and an ample line bundle NN on AA, remedies this issue. As for abelian varieties, the zero set of h^L\widehat{h}_{L} coincides with the torsion points of GG. Let

η:=η(G,L)=infh^L(P),\eta:=\eta(G,L)=\operatorname{inf}\widehat{h}_{L}(P),

where the infimum is taken over all non-torsion PP in G(K)G(K), and we write

ω:=ω(G,K)\omega:=\omega(G,K)

for the cardinality of the torsion group of G(K)G(K). Note that η\eta and ω\omega depend on the embedding of GG in a projective space. Suppressing this dependence in our notation, we can prove the following theorem using an argument similar to the proof of [Mas88, Theorem 𝔾m,A\mathbb{G}_{m},A].

Theorem 3.3.

Suppose P1,,PnP_{1},\ldots,P_{n} on G(K)G(K) have canonical heights at most qηq\geq\eta. Then the relation group of P1,,PnP_{1},\ldots,P_{n} generated by

𝒎:m1P1+mnPn=0G\boldsymbol{m}:m_{1}P_{1}+\cdots m_{n}P_{n}=0_{G}

satisfy

|mi|nn1ω(q/η)n1.|m_{i}|\leq n^{n-1}\omega(q/\eta)^{n-1}.
Proof.

We refer to [Kü20, Lem. 8] for the definition and properties of the canonical height on a semi-abelian variety. By definition,

h^L=h^G(M,ϱ)+h^πN,\widehat{h}_{L}=\widehat{h}_{G(M,\varrho)}+\widehat{h}_{\pi^{*}N},

where h^G(M,ϱ)\widehat{h}_{G(M,\varrho)} is linear and h^πN\widehat{h}_{\pi^{*}N} is quadratic. We define a convex distance function ff on n\mathbb{Z}^{n} by

f(𝒎)=h^G(M,ϱ)(𝒎)+h^πN(𝒎)1/2.f(\boldsymbol{m})=\widehat{h}_{G(M,\varrho)}(\boldsymbol{m})+\widehat{h}_{\pi^{*}N}(\boldsymbol{m})^{1/2}.

Let Γ=Γ(f)\Gamma=\Gamma(f) be the set consisting of all 𝒎\boldsymbol{m} such that m1P1++mnPnm_{1}P_{1}+\ldots+m_{n}P_{n} is a torsion point of G(K)G(K), and take Γ0\Gamma_{0} as the relation group of P1,,PnP_{1},\ldots,P_{n}. We can choose E=q,ε=η1/2E=q,\varepsilon=\eta^{1/2}, and the theorem follows immediately from [Mas88, Prop.].

General semi-abelian subvarieties of GG are generated by relation groups where the miEnd(G)m_{i}\in\operatorname{End}(G). We accommodate for End(G)\operatorname{End}(G)\neq\mathbb{Z} in the following way. Let {1,τ1,,τk}End(G)\{1,\tau_{1},\dots,\tau_{k}\}\in\operatorname{End}(G) be a set of generators. Then, we can work in G(k+1)nG^{(k+1)n} with Pi,τ1Pi,τkPiP_{i},\tau_{1}P_{i},\cdots\tau_{k}P_{i} for 1in1\leq i\leq n. Due to the dominance of a height associated to an ample divisor (see [Lan83, Chap. 4, Prop. 5.4] or [Poo01, Lemma 6]),

(3.0.1) h^L(τi(P))Cih^L(P)\displaystyle\widehat{h}_{L}(\tau_{i}(P))\leq C^{i}\widehat{h}_{L}(P)

where each CiC^{i} is an absolute constant depending on τi\tau_{i} and the embedding of GG. Let

a1+j=1kaj+1τj=max1jk+1{|aj|}\left\lVert a_{1}+\sum_{j=1}^{k}a_{j+1}\tau_{j}\right\rVert=\textup{max}_{1\leq j\leq k+1}\{|a_{j}|\}

denote the norm of 𝒂=a1+j=1kaj+1τj\boldsymbol{a}=a_{1}+\sum_{j=1}^{k}a_{j+1}\tau_{j}.

Corollary 3.4.
(3.0.2) miC¯(2ng)2ng1ω(q/η)(2ng1)/2\displaystyle\left\lVert m_{i}\right\rVert\leq\overline{C}(2ng)^{2ng-1}\omega(q/\eta)^{(2ng-1)/2}

where C¯\overline{C} is taken as the maximal of (Ci)(2ng1)/2(C^{i})^{(2ng-1)/2} and q:=max1in{h^L(Pi)}q:=\textup{max}_{1\leq i\leq n}\{\widehat{h}_{L}(P_{i})\}.

To estimate 3.0.2 in terms of degree of the algebraic points, we need some estimates for η,ω\eta,\omega, which is given for tori and abelian varieties in [Mas88]. Let D=[K:]D=[K:\mathbb{Q}] and =log(D+2)\mathcal{L}=\log(D+2) we have:

ηC1D(2g+1)2g\eta\geq C^{-1}D^{-(2g+1)}\mathcal{L}^{-2g}

and

ωCDgg.\omega\leq CD^{g}\mathcal{L}^{g}.

To bound the height of rational points on a suitable definable set in Section 4, we need the following upper bound of the norm of the generating set for the linear relations satisfied by nn fixed VV-images of special points in terms of their discriminants.

Proposition 3.5.

Let (s1,x1),,(sn,xn)S×G×𝔾am(s_{1},x_{1}),\cdots,(s_{n},x_{n})\in S\times G\times\mathbb{G}_{a}^{m} be points of VV with sis_{i} special and with discriminants Δ(si)\Delta(s_{i}). Define the complexity of s=(s1,,sn)s=(s_{1},\dots,s_{n}) by

𝚫(s)=𝚫(s1,,sn)=max|Δ(si)|.\boldsymbol{\Delta}(s)=\boldsymbol{\Delta}(s_{1},\cdots,s_{n})=\operatorname{max}|\Delta(s_{i})|.

Then there are constants C,C,cC,C^{\prime},c^{\prime} depending on S,G,V,nS,G,V,n such that for 𝚫C\boldsymbol{\Delta}\geq C, there is a generating set for the linear relations satisfied by the xiGx_{i}\in G such that

miC𝚫(s)c.||m_{i}||\leq C^{\prime}\boldsymbol{\Delta}(s)^{c^{\prime}}.
Proof.

Since VV is a correspondence in S×GS\times G, (s,x)V(s,x)\in V is an algebraic point, using standard properties of heights we have that

H(x)C1H(s)cH(x)\leq C_{1}^{\prime}H(s)^{c}

and since the projection of VV to SS is finite and both projections are dominant on each factor, we have

[L(x):L]C2[F(s):F].[L(x):L]\leq C_{2}^{\prime}[F(s):F].

Take C=max{C1,C2}.C^{\prime}=\operatorname{max}\{C_{1}^{\prime},C_{2}^{\prime}\}. Therefore the degree D=[L:]D=[L^{\prime}:\mathbb{Q}] of the field of definition of x1,,xix_{1},\cdots,x_{i} is bounded in terms of 𝚫\boldsymbol{\Delta} by Proposition 3.2(c).

By [Kü20, Lem. 8(a)], after fixing the embedding corresponding to the TT-effective and ample line bundles on GG, the differences of h^L(P)\widehat{h}_{L}(P) and h(P)h(P) are bounded globally on G(¯)G(\overline{\mathbb{Q}}) by an absolute constant δ\delta. Therefore for h:=max{h(x1),,h(xn)}h:=\textup{max}\{h(x_{1}),\cdots,h(x_{n})\} sufficiently large, we have

ηhδ\eta\leq h-\delta

and

hδh\geq\delta

which implies

q2h.q\leq 2h.

By Proposition 3.2(a) η\eta and qq are bounded in terms of 𝚫\boldsymbol{\Delta}, and the proposition follows from Corollary 3.4.

4. Exemplary Components

4.1. Proof of the main theorem over ¯\overline{\mathbb{Q}}

Now we can prove Theorem 1.2. We first prove the theorem for varieties VV defined over ¯\overline{\mathbb{Q}}, and then show the result for all VV.

First, we prove a criterion for inclusion of complex algebraic varieties defined over ¯\overline{\mathbb{Q}} and a finiteness result for pre-special subvarieties.

Lemma 4.1.

Let Vn×mV\subset\mathbb{C}^{n}\times\mathbb{C}^{m} and WnW\subset\mathbb{C}^{n} be irreducible algebraic varieties defined over ¯\overline{\mathbb{Q}}. There exists a finite set of points P1,,Pm+1nP_{1},\dots,P_{m+1}\in\mathbb{C}^{n} such that for any QmQ\in\mathbb{C}^{m}, we have W×{Q}VW\times\{Q\}\subset V if and only if Pi×QVP_{i}\times Q\in V for all ii.

Proof.

Let d=dimWd=\dim W. We can find a generic point P1×P2××Pm+1Wm+1P_{1}\times P_{2}\times\cdots\times P_{m+1}\in W^{m+1} such that tr.deg.(P1,,Pm+1)=d(m+1)\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(P_{1},\dots,P_{m+1})=d(m+1), and we claim that these PiP_{i} satisfy the above property. The forward direction of the if and only if is clear. Now suppose that Pi×QVP_{i}\times Q\in V for all ii. Viewing PiP_{i} as a set, let Pi~Pi\tilde{P_{i}}\subset P_{i} be a minimal transcendental basis for tr.deg.(Pi)\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(P_{i}). Write Q=(q1,,qn)nQ=(q_{1},\dots,q_{n})\in\mathbb{C}^{n} and let Qi{1,2,,n}Q_{i}\subset\{1,2,\dots,n\} be the set of indices jj so that qjq_{j} is algebraic over (Pi~)\mathbb{Q}(\tilde{P_{i}}). By construction, |Pi~|=d\lvert\tilde{P_{i}}\rvert=d and the Pi~\tilde{P_{i}} are algebraically independent. This implies that QiQj=Q_{i}\cap Q_{j}=\varnothing for iji\neq j and so i=1m+1|Qi|m\sum_{i=1}^{m+1}\lvert Q_{i}\rvert\leq m. Thus, there exists an ii such that |Qi|=0\lvert Q_{i}\rvert=0 and for this ii we have

tr.deg.(Pi,Q)=tr.deg.(Pi)+tr.deg.(Q).\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(P_{i},Q)=\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(P_{i})+\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(Q).

Let X={Pi}¯ZarX=\overline{\{P_{i}\}}^{\mathrm{Zar}} and Y={Q}¯ZarY=\overline{\{Q\}}^{\mathrm{Zar}} be the ¯\overline{\mathbb{Q}}-Zariski closure of PiP_{i} and QQ. Then dimX=tr.deg.(Pi)=d\dim X=\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(P_{i})=d and dimY=tr.deg.(Q)\dim Y=\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(Q). Note that the Zariski closure of Pi×QP_{i}\times Q has dimension tr.deg.(Pi,Q)=dimX+dimY\operatorname{tr.deg.}_{\mathbb{Q}}\mathbb{Q}(P_{i},Q)=\dim X+\dim Y, showing that the ¯\overline{\mathbb{Q}}-Zariski closure of Pi×QP_{i}\times Q is X×YX\times Y. But X×YVX\times Y\subset V by definition and X=WX=W showing inclusion in the other direction. ∎

Lemma 4.2.

Suppose VS×GV\subset S\times G is an algebraic subvariety of the product of a Shimura variety with a connected commutative algebraic group. There exists a finite set Σ\Sigma of sub-Shimura datum (H,XH)(H,X_{H}) of (GS,X)(G_{S},X) and splittings (Had,XHad)=(H1,X1)×(H2,X2)(H^{\operatorname{ad}},X_{H}^{\operatorname{ad}})=(H_{1},X_{1})\times(H_{2},X_{2}) such that if WVW\subset V is exemplary, then there exist one such splitting in Σ\Sigma such that πS(W)\pi_{S}(W) is the image of X1×{x2}X_{1}\times\{x_{2}\}, for some x2X2x_{2}\in X_{2}, under the uniformization map of SS.

Moreover, if 𝒱W\mathcal{V}\to W is a definable family of VS×GV\subset S\times G, then the result still holds.

Proof.

The result for a fixed VS×GV\subset S\times G is proven by [PT22, Prop. 3.4] and [DR18, Prop. 6.10]. The version for family is given by the uniform Ax–Schanuel theorem, which is implied by the Ax–Schanuel theorem. This is proven in [ES23, Prop. 2.20]. ∎

We will need the following stronger form of the Pila–Wilkie point-counting theorem for families.

Theorem 4.3.

[HP16, Cor. 7.2] Let F×m×nF\subset\mathbb{R}^{\ell}\times\mathbb{R}^{m}\times\mathbb{R}^{n} be a definable family parametrized by the first factor \mathbb{R}^{\ell}. Let ε>0\varepsilon>0 and kk\in\mathbb{N} and let π1:m×nm\pi_{1}\colon\mathbb{R}^{m}\times\mathbb{R}^{n}\to\mathbb{R}^{m} and π2:m×nn\pi_{2}\colon\mathbb{R}^{m}\times\mathbb{R}^{n}\to\mathbb{R}^{n} denote the projections onto the first and second factors. There exists a constant c=c(F,k,ε)>0c=c(F,k,\varepsilon)>0 satisfying the following.

Let xx\in\mathbb{R}^{\ell} and let Fxm×nF_{x}\subset\mathbb{R}^{m}\times\mathbb{R}^{n} denote the fiber of FF over xx. If T1T\geq 1 and there exists a subset

Σ{(y,z)Fx:Hk(y)T}\Sigma\subset\{(y,z)\in F_{x}:H_{k}(y)\leq T\}

such that |π2(Σ)|>cTε\lvert\pi_{2}(\Sigma)\rvert>cT^{\varepsilon}, then there exists a continuous definable function β:[0,1]Fx\beta\colon[0,1]\to F_{x} satisfying the following four properties:

  1. (1)

    The composition π1β\pi_{1}\circ\beta is semialgebraic and its restriction to (0,1)(0,1) is real analytic;

  2. (2)

    The composition π2β\pi_{2}\circ\beta is non-constant;

  3. (3)

    π2(β(0))π2(Σ)\pi_{2}(\beta(0))\in\pi_{2}(\Sigma);

  4. (4)

    and the restriction of β\beta to (0,1)(0,1) is real analytic.

Theorem 4.4.

Let SS be a connected Shimura variety and let GG be a connected commutative algebraic group. Suppose VS×GV\subset S\times G is a proper irreducible subvariety that is finite over SS. Moreover, suppose that GG and VV are defined over ¯\overline{\mathbb{Q}}. Then, there are only finitely many exemplary components in VV.

Proof.

Let πS:VS\pi_{S}\colon V\to S and πG:VG\pi_{G}\colon V\to G be the projections onto the two factors. Let WVW\subset V be an exemplary component and let SπS(W)S^{\prime}\coloneqq\pi_{S}(W) be the special subvariety of SS which WW maps onto, and let GπG(W)G^{\prime}\supset\pi_{G}(W) be the smallest special subvariety of GG containing πG(W)\pi_{G}(W). SS^{\prime} is a special subvariety of SS and hence there is a sub-Shimura datum (H,XH)(H,X_{H}) of (GS,X)(G_{S},X), a decomposition (Had,XHad)=(H1,X1)×(H2,X2)(H^{\operatorname{ad}},X_{H}^{\operatorname{ad}})=(H_{1},X_{1})\times(H_{2},X_{2}), and a point y2X2y_{2}\in X_{2} so that SS^{\prime} is the image of X1×{y2}X_{1}\times\{y_{2}\}. By Lemma 4.2, there are only finitely many choices of XHX_{H} and X1X_{1}. Thus, to show finiteness of exemplary components, it suffices to show that for a fixed X1X_{1} and fixed dimension of GG^{\prime}, there are only finitely many points y2X2y_{2}\in X_{2} arising from exemplary subvarieties.

Let q~:X1×X2XS\tilde{q}\colon X_{1}\times X_{2}\to X\to S denote the uniformization map of SS restricted to X1×X2X_{1}\times X_{2} and let q:F1×F2Sq\colon F_{1}\times F_{2}\to S denote the restriction of q~\tilde{q} to a fundamental domain F1×F2F_{1}\times F_{2} of X1×X2X_{1}\times X_{2}. Let exp:gG\exp\colon\mathbb{C}^{g}\to G denote the uniformization map of the algebraic group and let e:FGGe\colon F_{G}\to G denote the restriction of that map to a fundamental domain. Using Lemma 4.1, let x1,,xrF1x_{1},\dots,x_{r}\in F_{1} denote a set of points to determine if q(X1×{x2})×exp(z)Vq(X_{1}\times\{x_{2}\})\times\exp(z)\subset V. Let ξ1,,ξg\xi_{1},\dots,\xi_{g} denote the coordinates in GG and let τ1,,τkEnd(G)\tau_{1},\dots,\tau_{k}\in\operatorname{End}(G) be a \mathbb{Z}-basis for End(G)\operatorname{End}(G). Take a generating set of all equations of the form

mijτjξi=0,mij\sum m_{ij}\tau_{j}\xi_{i}=0,m_{ij}\in\mathbb{Z}

that all the points of GG^{\prime} satisfy, and let G0G_{0} be the identity component of the algebraic subgroup of GG defined by these equations. We can extend each map τi:GG\tau_{i}\colon G\to G to an endomorphism of its covering space τi~:gg\tilde{\tau_{i}}\colon\mathbb{C}^{g}\to\mathbb{C}^{g} satisfying τi~(0)=0\tilde{\tau_{i}}(0)=0. Suppose that G0G_{0} is cut out by \ell such equations. Let

Y=\displaystyle Y= {(y,z1,,zg,m111,,mgk,b1,,b)F2×FG×gk×:\displaystyle\left\{(y,z_{1},\dots,z_{g},m_{111},\dots,m_{gk\ell},b_{1},\dots,b_{\ell})\in F_{2}\times F_{G}\times\mathbb{R}^{gk\ell}\times\mathbb{R}^{\ell}:\right.
1ir,(q(xi,y),e(𝒛))V and 1l,i,jmijlτj~(zi)=bl}\displaystyle\left.\forall 1\leq i\leq r,(q(x_{i},y),e(\boldsymbol{z}))\in V\text{ and }\forall 1\leq l\leq\ell,\sum_{i,j}m_{ijl}\tilde{\tau_{j}}(z_{i})=b_{l}\right\}

and set ZZ to be the projection of YY to F2×gk×F_{2}\times\mathbb{R}^{gk\ell}\times\mathbb{R}^{\ell}. Both YY and ZZ are definable sets. The set YY parametrizes points yF2y\in F_{2} so that a VV-image of X1×{y}X_{1}\times\{y\} lies within a proper special subvariety cut out by the 𝒎\boldsymbol{m} of GG (but not all choices of 𝒎\boldsymbol{m} and 𝒃\boldsymbol{b} correspond to algebraic subvarieties).

Suppose for the sake of contradiction that there were infinitely many ¯\overline{\mathbb{Q}}-algebraic exemplary subvarieties WW^{\prime} with fibers over X1×X2X_{1}\times X_{2}, and with dimension of the smallest special subvariety of GG containing πG(W)\pi_{G}(W^{\prime}) equal to a fixed dimension d=dimGd=\dim G^{\prime}. Each one gives a ¯\overline{\mathbb{Q}}-point (y,𝒎,𝒃)Z(y^{\prime},\boldsymbol{m},\boldsymbol{b})\in Z, and the Gal(¯/K)\operatorname{Gal}(\overline{\mathbb{Q}}/K)-orbits of WW^{\prime} also lie in ZZ for KK the defining field of S,G,End(G)S,G,\operatorname{End}(G). Over ¯\overline{\mathbb{Q}}, any commutative connected group GG can be written as a product of a semi-abelian variety with affine space G×𝔾anG^{\prime}\times\mathbb{G}_{a}^{n} (see [NW14, Prop. 5.1.12]). The tuple 𝒎\boldsymbol{m} consists integers and for semi-abelian varieties, we may take the fundamental domain so that the real part of each ziz_{i} is in the interval [0,1][0,1]. Thus, each blb_{l} corresponding to an equation on the semi-abelian variety is an integer bounded by |mijl|\sum\lvert m_{ijl}\rvert, which by Proposition 3.5, is bounded by C|Δ(y)|cC\lvert\Delta(y^{\prime})\rvert^{c}. Special subvarieties of 𝔾an\mathbb{G}_{a}^{n} are given by the equations mijτj~(zi)=0\sum m_{ij}\tilde{\tau_{j}}(z_{i})=0, so bl=0b_{l}=0 for those equations. Thus, taking the Galois orbit of WW^{\prime} gives a point with different yy^{\prime} as well as different blb_{l}, but the same 𝒎\boldsymbol{m}. View ZZ is fibered over the 𝒎\boldsymbol{m} variable and let ΣZ𝒎\Sigma\subset Z_{\boldsymbol{m}} be the set of points arising from exemplary subvarieties WW^{\prime} as well as their Galois conjugates. By Proposition 3.2, there are at least C|Δ(y)|cC^{\prime}\lvert\Delta(y^{\prime})\rvert^{c^{\prime}} points of height less than C|Δ(y)|cC\lvert\Delta(y^{\prime})\rvert^{c}. Then Theorem 4.3 gives the existence of a set RZ𝒎R\subset Z_{\boldsymbol{m}} of positive dimension whose projection to F2F_{2} is connected semialgebraic and whose projection to \mathbb{R}^{\ell} is non-constant.

Let exp¯:g/G0~G/G0\overline{\exp}\colon\mathbb{C}^{g}/\tilde{G_{0}}\to G/G_{0} be the exponentiation map of G/G0G/G_{0}. Let FG0F_{G_{0}} denote the image of the fundamental domain FGF_{G} under the quotient map, which will serve as a fundamental domain for exp¯\overline{\exp}. First, take the preimage of RZ𝒎R\subset Z_{\boldsymbol{m}} under the projection map YZY\to Z and then project the preimage under the map

Y𝒎F2×FG0×F2×FG0.Y_{\boldsymbol{m}}\to F_{2}\times F_{G_{0}}\times\mathbb{R}^{\ell}\to F_{2}\times F_{G_{0}}.

Let the image of RR under these transformations be RF2×FG0R^{\prime}\subset F_{2}\times F_{G_{0}}. By applying the point counting theorem if necessary, we may take RR^{\prime} to be connected and semialgebraic. Since the projection of RR to \mathbb{R}^{\ell} was non-constant, the projection of RR^{\prime} to FG0F_{G_{0}} is also non-constant. Moreover, we may assume that G/G0G/G_{0} is a semi-abelian variety by expanding G0G_{0} if necessary because the projection to the affine space 𝔾a\mathbb{G}_{a} is constant (bl=0b_{l}=0). By applying the Ax–Schanuel theorem (Theorem 2.5) to the Zariski closure of (F1×R)×(q×exp)(F1×R)(F_{1}\times R^{\prime})\times(q\times\exp)(F_{1}\times R^{\prime}), we get that (q×exp)(F1×R)S×G/G0(q\times\exp)(F_{1}\times R^{\prime})\subset S\times G/G_{0} is contained in a proper weakly special subvariety. Since RR^{\prime} contains preimages of special points, its image must lie in a proper special subvariety S×GS×G/G0S^{\prime}\times G^{\prime}\subset S\times G/G_{0}.

Let VS×G/G0V^{\prime}\subset S\times G/G_{0} be the image of VV under the quotient map GG/G0G\to G/G_{0}. By construction, we have that (q×exp)(F1×R)V(q\times\exp)(F_{1}\times R^{\prime})\subset V^{\prime}, and so VV^{\prime} contains a proper special subvariety S×GS^{\prime}\times G^{\prime}. However, the map VSV\to S is finite, and hence VSV^{\prime}\to S is also finite. Hence, GG^{\prime} is a single point, meaning that RR was originally of the form R′′×{g}R^{\prime\prime}\times\{g\}, with gg corresponding to a torsion translation of G0G_{0}, which by abuse of notation we will also denote G0G_{0}. Let G0~FG\tilde{G_{0}}\subset F_{G} be the preimage of G0G_{0}. Applying Ax–Schanuel to the Zariski closure of

U(F1×R′′×G0~)×(q(F1×R′′)×G0)V,U\coloneqq(F_{1}\times R^{\prime\prime}\times\tilde{G_{0}})\times(q(F_{1}\times R^{\prime\prime})\times G_{0})\cap V,

we get that q(F1×R′′)×G0q(F_{1}\times R^{\prime\prime})\times G_{0} is contained in a proper special subvariety of VV, which must be of the form S×G0VS^{\prime}\times G_{0}\subset V. This shows that our original exemplary component WW was not exemplary because there is a larger Shimura variety SS^{\prime} properly containing πS(W)\pi_{S}(W) whose VV image lies within G0=GG_{0}=G^{\prime}. ∎

4.2. From ¯\overline{\mathbb{Q}} to \mathbb{C}

Armed with the result over number fields, through a continuity argument, we are able to prove Theorem 1.2 over any characteristic 0 field.

Proof of Theorem 1.2.

Let FF be the finitely generated subfield over \mathbb{Q} which S,G,End(G)S,G,\operatorname{End}(G), and VV are all defined. Then, there exists a quasi-projective geometrically irreducible variety ZZ over a number field KK such that FF is the function field of ZZ, and SS is defined over KK. By taking a Zariski open subset of ZZ, we may assume that GG extends to an abelian scheme 𝒢\mathcal{G} over ZZ and VV extends to a variety 𝒱\mathcal{V} that is flat over ZZ. Choose a generic closed point z0Z()z_{0}\in Z(\mathbb{C}) such that the field gotten by localizing at z0z_{0} is isomorphic to FF. Choose a simply connected Euclidean open neighborhood Uz0U\ni z_{0} of z0z_{0} so that the homology of 𝒢\mathcal{G} can be trivialized over UU, and so we have 𝒢UG×U\mathcal{G}_{U}\cong G\times U as analytic varieties.

Suppose for the sake of contradiction that there are infinitely many exemplary subvarieties of V𝒱z0V\cong\mathcal{V}_{z_{0}}. By Lemma 4.2, special subvarieties that arise as projections of the exemplary subvarieties of VV to SS come from finitely many splittings of special subvarieties of SS. So, there is a splitting of sub-Shimura datum (H1,X1)×(H2,X2)(H_{1},X_{1})\times(H_{2},X_{2}) with image S1×S2SS_{1}\times S_{2}\subset S and infinitely many special points piS2p_{i}\in S_{2} such that there are exemplary subvarieties WiW_{i} of VV mapping surjectively to S1×{pi}S_{1}\times\{p_{i}\}. By the André–Oort conjecture, proven by [PST+21], the Zariski closure of these S1×{pi}S_{1}\times\{p_{i}\} is a finite union R1,,RnSR_{1},\dots,R_{n}\subset S of special subvarieties of SS. For each ii, let GiGG_{i}\subset G be the smallest special subvariety containing πG(Wi)𝒢z0\pi_{G}(W_{i})\subset\mathcal{G}_{z_{0}}, and we extend GiG_{i} to a family 𝒢i\mathcal{G}_{i} over UU. For every zUz\in U, we have that the 𝒱z\mathcal{V}_{z} image of S1×{pi}S_{1}\times\{p_{i}\} lies within Gi,zG_{i,z}. However, for each zU(¯)z\in U(\overline{\mathbb{Q}}), Theorem 4.4 says that there are only finitely many exemplary components and hence the image of each RiR_{i} must also lie within a proper special subvariety of GG.

The 𝒱\mathcal{V}-images of each RiR_{i} gives a family of subvarieties \mathcal{R} of 𝒢\mathcal{G} such that for each zU(¯)z\in U(\overline{\mathbb{Q}}), the fiber z\mathcal{R}_{z} lies within a proper special subvariety of 𝒢z\mathcal{G}_{z}. We claim that this holds at z0z_{0} as well. By replacing \mathcal{R} with a gg self-sum, where g=dimGg=\dim G, we may assume the \mathcal{R} is a coset of an abelian subscheme of 𝒢\mathcal{G} and we still have that z\mathcal{R}_{z} lies within a proper special subvariety of GG. By quotienting by the identity component of \mathcal{R}, we may assume that \mathcal{R} is finite over UU and after finite base change that \mathcal{R} is a section of UU. Applying the Main Theorem of [Mas89] and the extension to semi-abelian varieties given in Section 55 of loc. cit. to ×τ1,,τkGk+1\mathcal{R}\times\tau_{1}\mathcal{R},\dots,\tau_{k}\mathcal{R}\subset G^{k+1}, we get that z\mathcal{R}_{z} must lie within a proper special subvariety of 𝒢z\mathcal{G}_{z}. ∎

5. Applications of the main theorem

In this section, we see how Theorem 1.2 can be used to get results on linear dependence of special points. We prove a weaker version of Corollary 1.3 which may be easier to apply. First we prove a lemma that is probably already known to experts but we could not find a reference for.

Lemma 5.1.

Let SS be a Shimura variety and let πij:SnS2\pi_{ij}\colon S^{n}\to S^{2} be the projection onto the iith and jjth coordinates. If SSnS^{\prime}\subset S^{n} is a proper Shimura subvariety, then there exist i,ji,j\in\mathbb{N} with 1i<jn1\leq i<j\leq n such that πij(S)\pi_{ij}(S^{\prime}) is a proper Shimura subvariety of S2S^{2}.

Proof.

We may assume that SS is a connected Shimura variety whose Shimura datum (GS,X)(G_{S},X) is such that GSG_{S} is a product of simple groups and that SS^{\prime} is also a connected Shimura variety whose Shimura datum is (G,X)(G^{\prime},X^{\prime}). By viewing SS^{\prime} as the orbit of a point under the action of a subgroup of GSnG_{S}^{n}, it suffices to show that if GG1×GnG^{\prime}\subset G_{1}\times\cdots G_{n} is a proper subgroup of a product of simple groups, then there exist i,ji,j such that πij(G)Gi×Gj\pi_{ij}(G^{\prime})\subset G_{i}\times G_{j} is a proper subgroup.

Suppose otherwise, then since π12\pi_{12} and π13\pi_{13} are surjective, we can find elements of GG^{\prime} of the form (g,1,g,)(g,1,g^{\prime},\dots) and (h,h,1,)G(h,h^{\prime},1,\dots)\in G^{\prime} for g,hg,h in G1,hG2,gG3G_{1},h^{\prime}\in G_{2},g^{\prime}\in G_{3}. Taking their commutator, we find that the projection of GG^{\prime} to G1×G2×G3G_{1}\times G_{2}\times G_{3} by taking the first three coordinates contains the element (ghg1h1,1,1)(ghg^{-1}h^{-1},1,1). Since g,hG1g,h\in G_{1} were arbitrary, we get that [G1,G1]×{1}G1×G2×G3[G_{1},G_{1}]\times\{1\}\subset G_{1}\times G_{2}\times G_{3}. Now since G1G_{1} is a simple group, we get that [G1,G1]=G1[G_{1},G_{1}]=G_{1}. By symmetry, we see that π123(G)=G1×G2×G3\pi_{123}(G^{\prime})=G_{1}\times G_{2}\times G_{3}. By inductively repeating this process, we see that G=G1×GnG^{\prime}=G_{1}\times\cdots G_{n}, contradicting the properness of GG^{\prime}. ∎

Corollary 5.2.

Let S,G,VS,G,V be as in Corollary 1.3. Fix nn\in\mathbb{N}. There exists a proper Zariski closed subset SS2S^{\prime}\subset S^{2} such that for any nn points s1,,snSs_{1},\dots,s_{n}\in S such that (si,sj)S(s_{i},s_{j})\not\in S^{\prime} for any pair iji\neq j, then any VV-images g1,,gng_{1},\dots,g_{n} for s1,,sns_{1},\dots,s_{n} are linearly independent.

In the case that S=𝒜2,𝒜3S=\mathcal{A}_{2},\mathcal{A}_{3} is the moduli space of abelian surfaces or threefolds, we can give a slightly more explicit condition for the independence of VV-images, by studying the subvarieties of S2S^{2}.

Definition 5.3.

We say that a set of CM abelian varieties s1,,sn𝒜gs_{1},\dots,s_{n}\in\mathcal{A}_{g} for g3g\leq 3 are DD-independent, for some integer DD, if each sis_{i} is a simple abelian variety, there exists no isogeny of degree D\leq D between some pair sisjs_{i}\to s_{j} for iji\neq j, and for each ii, we have Δ(si)>D\Delta(s_{i})>D.

We now give the proof of Corollary 1.4.

Proof of Corollary 1.4.

If x1,,xnx_{1},\cdots,x_{n} are linearly dependent in GG then by Corollary 5.2, there must exist some iji\neq j such that (si,sj)(s_{i},s_{j}) lies within a finite set of proper special subvarieties of S2S^{2}. So we need to describe the special subvarieties in 𝒜g×𝒜g\mathcal{A}_{g}\times\mathcal{A}_{g}.

Let W𝒜g×𝒜gW\subset\mathcal{A}_{g}\times\mathcal{A}_{g} be one of these finitely many proper special subvarieties. First, suppose that the projection to one of the factors is proper. This projection must be a proper special subvariety of 𝒜g\mathcal{A}_{g}. Since g3g\leq 3, by [MZ99], the Shimura datum associated with the projection of WW must be PEL-type Shimura datum and so the endomorphism ring of some sis_{i}, tensored up to \mathbb{Q}, must contain a \mathbb{Q}-algebra BB. Since we assumed our sis_{i} were DD-independent, they are simple and hence BB is a number field or division algebra. Since there are finitely many proper special subvarieties of 𝒜g2\mathcal{A}_{g}^{2}, there are only finitely many such number fields or division algebras that can appear. Setting DD larger than the discriminants of the number fields and division algebras that appear prevent the projection of sis_{i} from being contained in a WW of this form.

Now suppose that the projection of WW to each factor of 𝒜g\mathcal{A}_{g} is surjective. Choose a generic wWw\in W and let GSp2g()×Sp2g()G\subset\mathrm{Sp}_{2g}(\mathbb{C})\times\mathrm{Sp}_{2g}(\mathbb{C}) be the set of gg such that gwWgw\in W. This is a Lie group of Sp2g()×Sp2g()\mathrm{Sp}_{2g}(\mathbb{C})\times\mathrm{Sp}_{2g}(\mathbb{C}) and the projection onto each factor of Sp2g()\mathrm{Sp}_{2g}(\mathbb{C}) is surjective. By Goursat’s lemma, the group GG is given as the graph of an isomorphism between Sp2g()/N1Sp2g()/N2\mathrm{Sp}_{2g}(\mathbb{C})/N_{1}\to\mathrm{Sp}_{2g}(\mathbb{C})/N_{2}, with N1,N2N_{1},N_{2} normal subgroups. However, since GG is a Lie group and Sp2g\mathrm{Sp}_{2g} is a simple Lie group, we must have that N1,N2N_{1},N_{2} are trivial and so WW is a Hecke correspondence. Therefore, by choosing DD large enough, we may also exclude all Hecke correspondences that appear by giving a lower bound on the isogenies that appear.

We give a generalization of [PT22, Corollary 1.4]. Although there are infinitely many abelian varieties with fixed discriminant of their generic \mathbb{Z}-Hodge structure, the number of irreducible special subvarieties that contains them is finite.

Proof of Theorem 1.6.

By applying Corollary 5.2 to Vr+1V^{r+1} inside of Sr+1×Gr+1S^{r+1}\times G^{r+1}, we obtain that there exists finitely many proper special subvarieties SnS_{n} of S2S^{2} such that for any r+1r+1 points s1,,sr+1Ss_{1},\dots,s_{r+1}\in S with (si,sj)nSn(s_{i},s_{j})\not\in\bigcup_{n}S_{n} for all pairs i,ji,j, any VV-images of the sis_{i} are linearly independent. Since the Shimura datum is simple, the argument in the proof of Corollary 1.4 gives that either the projection of SnS_{n} on one factor is a proper special subvariety or SnS_{n} is a correspondence on S2S^{2}. Let SS^{\prime} be the union of all of the proper projections and let DD be the maximum degree of the projections of SnS_{n} to a factor of SS when SnS_{n} is a correspondence. Then, there exists an N=N(r+1,D)N=N(r+1,D) such that given s1,,sNS\Ss_{1},\dots,s_{N}\in S\backslash S^{\prime} special, we can find a subset s1,,sr+1s_{1}^{\prime},\dots,s_{r+1}^{\prime} such that no (si,sj)(s_{i}^{\prime},s_{j}^{\prime}) lies in a SnS_{n}, and thus their VV-images are linearly independent. Thus, outside of SSS^{\prime}\subset S, there are less than NN other special points of SS whose VV-images lie in Γ\Gamma^{\prime}. Both SS^{\prime} and NN depend on S,G,V,rS,G,V,r, and are independent of Γ\Gamma, proving the theorem. ∎

Acknowledgements

The authors were introduced to the problem at the 2023 Arizona Winter School under Jacob Tsimerman’s project group. We would like to heartily thank Jacob for suggesting this question to us and for the helpful conversations, and Niven Achenjang, Camille Amoyal, Pam (Miao) Gu and Dong Gyu Lim for the helpful discussions during the winter school. We would also like to thank Sebastian Eterović for pointing out the application to multiplicative independence of singular moduli. Finally, we thank the organizers of the Arizona Winter School for creating a wonderful research experience.

References

  • [AEF23] Vahagn Aslanyan, Sebastian Eterović, and Guy Fowler. Multiplicative relations among differences of singular moduli, 2023.
  • [Ax72] James Ax. Some topics in differential algebraic geometry. I. Analytic subgroups of algebraic groups. Amer. J. Math., 94:1195–1204, 1972.
  • [Bal18] Gregorio Baldi. On a conjecture of Buium and Poonen. Annales de l’Institut Fourier, 2018.
  • [BP09] Alexandru Buium and Bjorn Poonen. Independence of points on elliptic curves arising from special points on modular and Shimura curves, I: Global results. Duke Mathematical Journal, 147(1):181 – 191, 2009.
  • [BSCFN23] David Blázquez-Sanz, Guy Casale, James Freitag, and Joel Nagloo. A differential approach to Ax-Schanuel, i, 2023.
  • [BSY22] Gal Binyamini, Harry Schmidt, and Andrei Yafaev. Lower bounds for Galois orbits of special points on Shimura varieties: a point-counting approach. Mathematische Annalen, 385, 02 2022.
  • [Chi22] Kenneth Chung Tak Chiu. Ax-Schanuel with derivatives for mixed period mappings, 2022.
  • [DO16] Chris Daw and Martin Orr. Heights of pre-special points of Shimura varieties. Mathematische Annalen, 365, 08 2016.
  • [DR18] Christopher Daw and Jinbo Ren. Applications of the hyperbolic Ax-Schanuel conjecture. Compos. Math., 154(9):1843–1888, 2018.
  • [ES23] Sebastian Eterović and Thomas Scanlon. Likely intersections, 2023.
  • [Gao20] Ziyang Gao. Mixed Ax-Schanuel for the universal abelian varieties and some applications. Compos. Math., 156(11):2263–2297, 2020.
  • [HP16] Philipp Habegger and Jonathan Pila. O-minimality and certain atypical intersections. Ann. Sci. Éc. Norm. Supér. (4), 49(4):813–858, 2016.
  • [Kü20] Lars Kühne. The bounded height conjecture for semiabelian varieties. Compositio Mathematica, 156(7):1405–1456, 2020.
  • [Lan83] Serge Lang. Fundamentals of Diophantine geometry. Springer-Verlag, New York, 1983.
  • [Mas88] D. W. Masser. Linear relations on algebraic groups. In New advances in transcendence theory (Durham, 1986), pages 248–262. Cambridge Univ. Press, Cambridge, 1988.
  • [Mas89] D. W. Masser. Specializations of finitely generated subgroups of abelian varieties. Trans. Amer. Math. Soc., 311(1):413–424, 1989.
  • [Mil05] J. S. Milne. Introduction to Shimura varieties. In Harmonic analysis, the trace formula, and Shimura varieties, volume 4 of Clay Math. Proc., pages 265–378. Amer. Math. Soc., Providence, RI, 2005. Revised version in https://www.jmilne.org/math/xnotes/svi.pdf.
  • [Moo98] Ben Moonen. Linearity properties of Shimura varieties. i. J. Algebraic Geom., 7(3):539–567, 1998.
  • [MPT19] Ngaiming Mok, Jonathan Pila, and Jacob Tsimerman. Ax-Schanuel for Shimura varieties. Ann. of Math. (2), 189(3):945–978, 2019.
  • [MZ99] B.J.J. Moonen and Yu.G. Zarhin. Hodge classes on abelian varieties of low dimension. Mathematische Annalen, 315(4):711–733, 12 1999.
  • [NW14] Junjiro Noguchi and Jörg Winkelmann. Nevanlinna theory in several complex variables and Diophantine approximation, volume 350 of Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. Springer, Tokyo, 2014.
  • [Pin05] Richard Pink. A Combination of the Conjectures of Mordell-Lang and André-Oort, pages 251–282. Birkhäuser Boston, Boston, MA, 2005.
  • [Poo01] Bjorn Poonen. Néron-Tate projection of algebraic points. Internat. Math. Res. Notices, (9):435–440, 2001.
  • [PST+21] Jonathan Pila, Ananth Shankar, Jacob Tsimerman, Hélène Esnault, and Michael Groechenig. Canonical heights on Shimura varieties and the André-Oort conjecture. Preprint, 09 2021.
  • [PT17] Jonathan Pila and Jacob Tsimerman. Multiplicative relations among singular moduli. Ann. Sc. Norm. Super. Pisa Cl. Sci. (5), 17(4):1357–1382, 2017.
  • [PT22] Jonathan Pila and Jacob Tsimerman. Independence of CM points in elliptic curves. J. Eur. Math. Soc. (JEMS), 24(9):3161–3182, 2022.
  • [Tsi11] Jacob Tsimerman. Brauer-siegel for arithmetic tori and lower bounds for galois orbits of special points. Journal of the American Mathematical Society, 25, 03 2011.
  • [UY11] Emmanuel Ullmo and Andrei Yafaev. A characterization of special subvarieties. Mathematika, 57(2):263–273, 2011.