Arxiv:1711.10112V2 [Math.NT]

arXiv:1711.10112v2 [math.NT] 30 Nov 2017 a nw ob ntl eeae!Snete,mn uhr h authors many then, Since generated! finitely be to known was a upre npr yNtoa cec onaingatD grant Foundation Science National by part in supported was enfrPr,Jh ogt n eai acetWo [PPVW Wood Matchett Melanie and Voight, John Park, Jennifer n eeoe ute n[a1] neetnl,telatte the Interestingly, [Wat15]. in further developed and uino ak.As,atog hr xssetniecomp extensive exists [BHK (e.g., there questions although some Also, to answers ranks. of bution o onens:teei n yRbnadSlebr o fa a discu for Granville, Silverberg by and another Rubin and 5.2], by and one 5.1 is Remarks there [RS00, boundedness: for 427 t jr onn n #550033. and Poonen) Bjorn (to #402472 h okoeepcainhsflpflpe tlatoc;se[ see once; least at flip-flopped has expectation folklore the n rsnsterm bu h oe htsgetthat suggest that curves model elliptic the many about theorems presents and [Mes82 ( [FGH07, 194], and summary. p. a 10.5], [Tat74, for Conjecture Tate 3.1] [Ulm02, 257], tion 1], p. Section [Cas66, [Bru92, 98], p. [Hon60, usin1.1. Question R question: Eric following and Lenstra, Hendrik Kane, a the M. of Daniel articles Bhargava, on Manjul primarily based article, survey present odl rvdthat proved Mordell ainlpit on points rational oeigrnsdrcl scalnigbcueteeaefe are there because challenging is directly ranks Modeling usin11wsipiil se yPicr n10 [Poi01 1901 in Poincaré by asked implicitly was 1.1 Question hssre ril a ensbitdt h rceig of Proceedings the to submitted been has article survey This phrases. and words Key 2010 Date h rsn uvydsrbsapoaiitcmdlfrtea the for model probabilistic a describes survey present The Let ERSISFRTEAIHEI FELPI CURVES ELLIPTIC OF ARITHMETIC THE FOR HEURISTICS E oebr3,2017. 30, November : ahmtc ujc Classification. Subject Mathematics Abstract. oe.I atclr h oe ugssta l u finitel but all that suggests model rank the correspond have particular, on based In ev curves theoretical model. elliptic the autho about discuss the We predictions of make Wood. articles and of Voight, series Rains, a Park, in developed model a curves, ea litccreover curve elliptic an be ≤ Is hsi nitouto oapoaiitcmdlfrtearit the for model probabilistic a to introduction an is This 21 E rk hc ol ml httern suiomybounded. uniformly is rank the that imply would which , h ru a on law group The . E E E n ec that hence and , ( ( litccre ak emrgop hfrvc–aegrou Shafarevich–Tate group, Selmer rank, curve, Elliptic Q Q ) ) one as bounded sfiieygnrtd[o2] let [Mor22]; generated finitely is Q se[i0]frbscdfiiin) Let definitions). basic for [Sil09] (see 1. JR POONEN BJORN rmr 10;Scnay1G0 42,1H2 14K15. 14H52, 14G25, 11G40, Secondary 11G05; Primary + Introduction E E rk 6) tsesta a oedt ol eneeded be would data more far that seems it 16]), aisoe l litccre over curves elliptic all over varies gives E ( 1 Q ) E sbudd usi o h rtheuristic first the not is Ours bounded. is ( Q ) h tutr fa bla ru,and group, abelian an of structure the S1096adSmn onaingrants Foundation Simons and MS-1601946 h 08IM h rtn fti article this of writing The ICM. 2018 the to ihEi an P1] with [PR12], Rains Eric with uthor losget on of bound a suggests also r ihBagv,Kn,Lenstra, Kane, Bhargava, with r o5,p 9,edo otoe( footnote of end 495, p. Poi50, n hoespoe bu the about proved theorems ing ayelpi uvsover curves elliptic many y rk dnefrtemdl n we and model, the for idence rk E .0] rse[PW6 Sec- [PPVW16, see or 5.20)], hoesaottedistri- the about theorems w ttoa aata suggests that data utational sdi [Wat in ssed ,[e8,I.. n II.1.2], and II.1.1 [Mes86, ], ( E .13,ee before even 173], p. , ihei felpi curves, elliptic of rithmetic Q is[KP1] n with and [BKLPR15], ains v u ot use,and guesses, forth put ave ( ) 6 scnendwt the with concerned is 16] Q iyo udai twists quadratic of mily ≤ ) mtco elliptic of hmetic eoeisrn.The rank. its denote 21 ,aeinvariety. abelian p, E o l u finitely but all for ( + Q Q 4 eto 11] Section 14, ) ? etestof set the be Q 21 . E ( 3 Q )], ) to suggest answers to others. Therefore, instead of modeling ranks in isolation, we model ranks, Selmer groups, and Shafarevich–Tate groups simultaneously, so that we can calibrate and corroborate the model using a diverse collection of known results.

2. The arithmetic of elliptic curves 2.1. Counting elliptic curves by height. Every elliptic curve E over Q is isomorphic to the projective closure of a unique curve y2 = x3 + Ax + B in which A and B are integers with 4A3 + 27B2 =6 0 (the smoothness condition) such that there is no prime p such that p4|A and p6|B. Let E be the set of elliptic curves of this form, so E contains one curve in each isomorphism class. Define the height of E ∈ E by ht E := max(|4A3|, |27B2|). (This definition is specific to the ground field Q, but it has analogues over other number fields.) Define E≤H := {E ∈ E : ht E ≤ H}. Ignoring constant factors, we have about H1/3 integers A with |4A3| ≤ H, and H1/2 integers B with |27B2| ≤ H. A positive fraction of such pairs (A, B) satisfy the smoothness 1/3 1/2 condition and divisibility conditions, so one should expect #E≤H to be about H H = H5/6. In fact, an elementary sieve argument [Bru92, Lemma 4.3] proves the following: Proposition 2.1. We have 4/3 −3/2 −1 5/6 #E≤H = (2 3 ζ(10) + o(1)) H as H →∞. Define the density of a subset S ⊆ E as

#(S ∩ E≤ ) lim H , H→∞ #E≤H if the limit exists. For example, it is a theorem that 100% of elliptic curves E over Q have no nontrivial rational torsion points; this statement is to be interpreted as saying that the density of the set S := {E ∈ E : E(Q)tors = 0} is 1 (even though there do exist E with E(Q)tors =06 ). 2.2. Elliptic curves over local fields. Our model will be inspired by theorems and conjectures about the arithmetic of elliptic curves over Q. But before studying elliptic curves over Q, we should thoroughly understand elliptic curves over local fields. Let Qv be the completion of Q at a place v. There is a natural injective homomorphism 2 inv: H (Qv, Gm) → Q/Z that is an isomorphism if v is nonarchimedean. Let E be an elliptic curve over Qv. Fix n ≥ 1. Taking Galois cohomology in the exact sequence 0 −→ E[n] −→ E −→n E −→ 0 1 yields a homomorphism E(Qv)/nE(Qv) → H (Qv, E[n]). Let Wv be its image. If v is a nonarchimedean place not dividing n and E has good reduction, then Wv equals the 1 subgroup of unramified classes in H (Qv, E[n]) [PR12, Proposition 4.13]. The theory of the Heisenberg group [Mum91, pp. 44–46] yields an exact sequence

1 −→ Gm −→ H −→ E[n] −→ 1, 2 which induces a map of sets

1 2 inv .qv : H (Qv, E[n]) −→ H (Qv, Gm) ֒→ Q/Z

It turns out that qv is a quadratic form in the sense that qv(x + y) − qv(x) − qv(y) is bi- additive [Zar74b, §2]. Moreover, qv|Wv = 0 [O’N02, Proposition 2.3]. In fact, using Tate 1 local duality one can show that Wv is a maximal isotropic subgroup of H (Qv, E[n]) with respect to qv [PR12, Proposition 4.11].

2.3. Selmer groups and Shafarevich–Tate groups. Now let E be an elliptic curve over A ′ Q. Let = v(Qv, Zv) be the adèle ring of Q; here v ranges over nontrivial places of A 1 A Q, Write E( ) for v E(Qv)/nE(Qv), and write H ( , E[n]) for the restricted product ′ (H1(Q , E[nQ]), W ). We have a commutative diagram v v vQ Q E(Q)/nE(Q) / H1(Q, E[n])

β E(A)/nE(A) α / H1(A, E[n]).

−1 1 The n-Selmer group is defined by Seln E := β (im α) ⊆ H (Q, E[n]). (This is equivalent to 1 1 A the classical definition; we have only replaced v H (Qv, E[n]) with a subgroup H ( , E[n]) into which α and β map.) The reason for defining Seln E is that it is a computable finite upper bound for (the image of) E(Q)/nE(Q).Q On the other hand, the Shafarevich–Tate group is defined by

1 1 X = X(E) := ker H (Q, E) → H (Qv, E) . v ! Y It is a torsion abelian group with an alternating pairing [ , ]: X × X → Q/Z defined by Cassels. Conjecturally, X is finite; in this case, [ , ] is nondegenerate and #X is a square [Cas62]. The definitions easily yield an exact sequence E(Q) (1) 0 −→ −→ Sel E −→ X[n] −→ 0, nE(Q) n so X[n] is measuring the difference between Seln E and the group E(Q)/nE(Q) it is trying to approximate. Each group in (1) decomposes according to the factorization of n into powers of distinct primes, so let us restrict to the case in which n = pe for some prime p and nonnegative integer e. Taking the direct limit over e yields an exact sequence

Qp ∞ 0 −→ E(Q) ⊗ −→ Selp∞ E −→ X[p ] −→ 0 Zp X ∞ X e of Zp-modules in which Selp∞ E := lim Selpe E and [p ] := e≥0 [p ]. Moreover, one −→ e can show that if E(Q)[p]=0 (as holds for 100% of curves), then Selpe E → (Selp∞ E)[p ] is an isomorphism (cf. [BKLPR15, Proposition 5.9(b)]), so no informationS about the individual pe-Selmer groups has been lost in passing to the limit. 3 2.4. The Selmer group as an intersection of maximal isotropic direct summands. 1 If ξ =(ξv) ∈ H (A, E[n]), then for all but finitely many v we have ξv ∈ Wv and hence qv(ξv)= 1 A 0, so we may define Q(ξ) := v qv(ξv). This defines a quadratic form Q: H ( , E[n]) → Q/Z. P Theorem 2.2. (a) Each of im α and im β is a maximal isotropic subgroup of H1(A, E[n]) with respect to Q [PR12, Theorem 4.14(a)]. (b) If n is prime or GQ → GL2(Z/nZ) is surjective then β is injective. (See [PR12, Propo- sition 3.3(e)] and [BKLPR15, Proposition 6.1].)

By definition, β(Seln E) = (im α) ∩ (im β). Thus, under either hypothesis in (b), Seln E is isomorphic to an intersection of maximal isotropic subgroups of H1(A, E[n]). Moreover, im α is a direct summand of H1(A, E[n]) [BKLPR15, Corollary 6.8]. It is conjectured that im β is a direct summand as well, at least for asymptotically 100% of elliptic curves over Q [BKLPR15, Conjecture 6.9], and it could hold for all of them. 2.5. The Birch and Swinnerton-Dyer conjecture. See [Wil06] for an introduction to the Birch and Swinnerton-Dyer conjecture more detailed than what we present here, and see [SW13, Section 8] for some more recent advances towards it. Let E ∈ E . To E one can associate its L-function L(E,s), a holomorphic function initially defined when Re s is sufficiently large, but known to extend to a holomorphic function on C (this is proved using the modularity of E). Just as the Dirichlet analytic class number formula expresses the residue at s =1 of the Dedekind zeta function of a number field k in terms of the arithmetic of k, the Birch and Swinnerton-Dyer conjecture expresses the leading term in the Taylor expansion of L(E,s) around s = 1 in terms of the arithmetic of E. We will state it only in the case that rk E(Q)=0 since that is all that we will need. In addition to the quantities previously associated to E, we need • the real period Ω, defined as the integral over E(R) of a certain 1-form; and • the Tamagawa number cp for each finite prime p, a p-adic volume analogous to the real period. Also define #X(E), if rk E(Q)=0; X0(E) := (0, if rk E(Q) > 0. Conjecture 2.3 (The rank 0 part of the Birch and Swinnerton-Dyer conjecture). If E ∈ E , then

X0 Ω cp (2) L(E, 1) = p . Q 2 #E( Q)tors Remark 2.4. In the case where the rank r := rk E(Q) is greater than 0, Conjecture 2.3 states only that L(E, 1)=0, whereas the full Birch and Swinnerton-Dyer conjecture predicts that ords=1 L(E,s)= r and predicts the leading coeﬃcient in the Taylor expansion of L(E,s) at s =1. Let H = ht E. Following Lang [Lan83] (see also [GS95], [dW98], [Hin07], [Wat08], and [HP16]), we estimate the typical size of X0 by estimating all the other quantities in (2) as 4 H →∞; see [PPVW16, Section 6] for details. The upshot is that if we average over E and o(1) −1 ignore factors that are H , then (2) simpliﬁes to 1 ∼ X0 Ω and we obtain X0 ∼ Ω ∼ H1/12. More precisely: o(1) • p cp = H [dW98, Theorem 3], [Hin07, Lemma 3.5], [Wat08, pp. 114–115], [PPVW16, Lemma 6.2.1]; Q • #E(Q)tors ≤ 16 [Maz77]; • Ω= H−1/12+o(1) [Hin07, Lemma 3.7], [PPVW16, Corollary 6.1.3]; and • the Riemann hypothesis for L(E,s) implies that L(E, 1) ≤ Ho(1) [IS00, p. 713]. In fact, it is reasonable to expect Average L(E, 1) ≍ 1. (The symbol ≍ means that the E∈E≤H left side is bounded above and below by positive constants times the right side.) Thus we expect 1/12+o(1) (3) Average X0(E)= H E∈E≤H as H →∞.

3. Modeling elliptic curves over Q

3.1. Modeling the p-Selmer group. According to Theorem 2.2, Selp E is isomorphic to an intersection of maximal isotropic subspaces in an infinite-dimensional quadratic space over Fp. So one might ask whether one could make sense of choosing maximal isotropic subspaces in an infinite-dimensional quadratic space at random, so that one could intersect two of them to obtain a space whose distribution is conjectured to be that of Selp E. This can be done by equipping an infinite-dimensional quadratic space with a locally compact topology [PR12, Section 2], but the resulting distribution can be obtained more simply by working within a 2n-dimensional quadratic space and taking the limit as n → ∞. Now every nondegenerate 2n-dimensional quadratic space with a maximal isotropic subspace is 2n isomorphic to the quadratic space Vn =(Fp , Q), where Q is the quadratic form

Q(x1,...,xn,y1,...,yn) := x1y1 + ··· + xnyn. E Therefore we conjecture that the distribution of dimFp Selp E as E varies over equals the limit as n → ∞ of the distribution of the dimension of the intersection of two maximal isotropic subspaces in Vn chosen uniformly at random from the ﬁnitely many possibilities. The limit exists and can be computed explicitly; this yields the formula on the right in the following: Conjecture 3.1 ([PR12, Conjecture 1.1]). For each s ≥ 0, the density of {E ∈ E : dimFp Selp E = s} equals s p (4) (1 + p−j)−1 . pj − 1 j≥0 j=1 Y Y 2 3 Remark 3.2. Let Ed be the elliptic curve dy = x − x over Q. Heath-Brown proved that the density of integers d such that dimF2 Sel2 Ed − 2= s equals s 2 (1+2−j)−1 , 2j − 1 j≥0 j=1 Y 5 Y matching (4) for p =2 [HB93,HB94]. (The −2 is there to remove the “causal” contribution to dim Sel2 Ed coming from Ed(Q)[2].) As we have explained, this result is a natural consequence of the theory of Section 2.4, but in fact Heath-Brown’s result came ﬁrst and the theory was reverse engineered from it [PR12]! Heath-Brown’s result was extended by Swinnerton-Dyer [SD08] and Kane [Kan13] to the family of quadratic twists of any E ∈ E with E[2] ⊆ E(Q) and no cyclic 4-isogeny.

e e 2n 3.2. Modeling the p -Selmer group. If p is replaced by p , then we should replace Fp e 2n by Vn := ((Z/p Z) , Q). But now there are diﬀerent types of maximal isotropic subgroups up to isomorphism. For example, if e = 2, then (Z/p2Z)n ×{0}n and (pZ/p2Z)2n are both maximal isotropic subgroups; of these, only the ﬁrst is a direct summand of Vn. In what follows, we will use only direct summands, for reasons to be explained at the end of this section.

Conjecture 3.3. If we intersect two random maximal isotropic direct summands of Vn := ((Z/peZ)2n, Q) and take the limit as n → ∞ of the resulting distribution, we obtain the distribution of Selpe E as E varies over E . For m ≥ 1, let σ(m) denote the sum of the positive divisors of m. One can prove that the limit as n →∞ of the average size of the random intersection equals σ(pe), and there is an analogous result for positive integers m not of the form pe [BKLPR15, Proposition 5.20]. This suggests the following: Conjecture 3.4 ([PR12, Conjecture 1(b)], [BKLPR15, Section 5.7], [BS13a, Conjecture 4]). For each positive integer m,

Average # Selm E = σ(m). E∈E

(The average is interpreted as the limit as H →∞ of the average over E≤H .) One could similarly compute the higher moments of the conjectural distribution; see [PR12, Proposition 2.22(a)] and [BKLPR15, Section 5.5]. There are several reasons why insisting upon direct summands in Conjecture 3.3 seems right: • Conjecturally, both of the maximal isotropic subgroups arising in the arithmetic of the elliptic curve are direct summands: see the last paragraph of Section 2.4. • Requiring direct summands is essentially the only way to make the model for Selpe E consistent with the model for Selp E, given that Selp E ≃ (Selpe E)[p] for 100% of curves [BKLPR15, Remark 6.12]. • It leads to Conjecture 3.4, which has been proved for m ≤ 5 [BS15a, BS15b, BS13a, BS13b].

3.3. Modeling the p∞-Selmer group and the Shafarevich–Tate group. Choosing a maximal isotropic direct summand of ((Z/peZ)2n, Q) compatibly for all e is equivalent to 2n choosing a maximal isotropic direct summand of the quadratic Zp-module Vn := (Zp , Q). This observation will lead us to a process that models Selpe E for all e simultaneously, or equivalently, that models Selp∞ E directly. To simplify notation, for any Zp-module X, let X′ denote X ⊗ Qp ; if X is a Z -submodule of V , then X′ is a Z -submodule of V ′. Zp p n p n 6 Now choose maximal isotropic direct summands Z and W of Vn with respect to the measure arising from taking the inverse limit over e of the uniform measure on the set of maximal isotropic direct summands of (Z/peZ)2n [BKLPR15, Sections 2 and 4]; then we conjecture ′ ′ that the limiting distribution of Z ∩ W as n → ∞ equals the distribution of Selp∞ E as E varies over E . Again, the point is that this limiting distribution is compatible with the the previously conjectured distribution for Selpe E for each nonnegative integer e, and the conjecture for Selpe E was based on theorems about Selmer groups of elliptic curves (see Section 2.4). Even better, using the same ingredients, we can model rk E(Q) and X[p∞] at the same time: Conjecture 3.5 ([BKLPR15, Conjecture 1.3]). If we choose maximal isotropic direct sum- 2n mands Z and W of (Zp , Q) at random as above, and we deﬁne R :=(Z ∩ W )′, S := Z′ ∩ W ′, T := S/R, then the limit as n →∞ of the distribution of the exact sequence 0 −→ R −→ S −→ T −→ 0 equals the distribution of the sequence

Qp ∞ 0 −→ E(Q) ⊗ −→ Selp∞ E −→ X[p ] −→ 0 Zp as E varies over E . There are several pieces of indirect evidence for the rank and X predictions of Conjec- ture 3.5: • Each of R and E(Q) ⊗ Qp is isomorphic to (Q /Z )r for some nonnegative integer r, Zp p p called the Zp-corank of the module. • The Zp-corank of R is 0 or 1, with probability 1/2 each [BKLPR15, Proposition 5.6]. Likewise, a variant of a conjecture of Goldfeld (see [Gol79, Conjecture B] and [KS99a, KS99b]) predicts that rk E(Q) (which equals the Z -corank of E(Q)⊗ Qp ) is 0, 1, ≥ 2 p Zp with densities 1/2, 1/2, 0, respectively. • The group T is finite and carries a nondegenerate alternating pairing with values ∞ in Qp/Zp, just as X[p ] conjecturally does (the p-part of the Cassels pairing). In particular, #T is a square. • Smith has proved a result analogous to Conjecture 3.5 for the family of quadratic twists of any E ∈ E with E[2] ⊆ E(Q) and no cyclic 4-isogeny [Smi17]. Further evidence is that there are in fact three distributions that have been conjectured to be the distribution of X[p∞] as E varies over rank r elliptic curves, and these three distributions coincide [BKLPR15, Theorems 1.6(c) and 1.10(b)]. This is so even in the cases with r ≥ 2, which conjecturally occur with density 0. For a fixed nonnegative integer r, the three distributions are as follows: 1. A distribution defined by Delaunay [Del01,Del07,DJ14], who adapted the Cohen–Lenstra heuristics for class groups [CL84]. 2. The limit as n → ∞ of the distribution of T := (Z′ ∩ W ′)/(Z ∩ W )′ when (Z, W ) is 2n sampled from the set of pairs of maximal isotropic direct summands of (Zp , Q) satisfying 7 rkZp (Z ∩ W ) = r. (This set of pairs is the set of Zp-points of a scheme of finite type, so it carries a natural measure [BKLPR15, Sections 2 and 4].) 3. The limit as n → ∞ through integers of the same parity as r of the distribution of T (coker A)tors when A is sampled from the space of matrices in Mn(Zp) satisfying A = −A

and rkZp (ker A) = r; here ker A and coker A are deﬁned by viewing A as a Zp-linear n n homomorphism Zp → Zp . The last of these is inspired by the theorem of Friedman and Washington [FW89] that for each odd prime p, the limit as n → ∞ of the distribution coker A for A ∈ Mn(Zp) chosen at random with respect to Haar measure equals the distribution conjectured by Cohen and Lenstra to be the distribution of the p-primary part of the class group of a varying imaginary quadratic ﬁeld.

3.4. Modeling the rank of an elliptic curve. In the previous section, we saw in the

third construction that conditioning on rkZp (ker A)= r yields the conjectural distribution of X[p∞] for rank r curves. The simplest possible explanation for this would be that sampling A T at random from Mn(Zp)alt := {A ∈ Mn(Zp): A = −A} without conditioning on rkZp (ker A)

caused rkZp (ker A) to be distributed like the rank of an elliptic curve.

What is the distribution of rkZp (ker A)? If n is even, then the locus in Mn(Zp)alt deﬁned by det A = 0 is the set of Zp-points of a hypersurface, which has Haar measure 0, so

rkZp (ker A)=0 with probability 1. If n is odd, however, then rkZp (ker A) cannot be 0,

because n − rkZp (ker A) is the rank of A, which is even for an alternating matrix. For n odd,

it turns out that rkZp (ker A)=1 with probability 1. If we imagine that n was chosen large

and with random parity, then the result is that rkZp (ker A) is 0 or 1, with probability 1/2 each. This result agrees with the variant of Goldfeld’s conjecture mentioned above. This model cannot, however, distinguish the relative frequencies of curves of various ranks ≥ 2, because in the model the event rkZp (ker A) ≥ 2 occurs with probability 0. Therefore we propose a refined model in which instead of sampling A from Mn(Zp)alt, we sample A from the set Mn(Z)alt,≤X of matrices in Mn(Z)alt with entries bounded by a number X depending on the height H of the elliptic curve being modeled, tending to ∞ as H →∞. This way, for elliptic curves of a given height H, the model predicts a potentially positive but diminishing probability of each rank ≥ 2 (the probability that an integer point in a box lies on a certain subvariety), and we can quantify the rate at which this probability tends to 0 as H →∞ in order to count curves of height up to H having each given rank. In fact, we let n grow with H as well. Here, more precisely, is the refined model. To model an elliptic curve E of height H, using functions η(H) and X(H) to be specified later, we do the following: 1. Choose n to be an integer of size about η(H) of random parity (e.g., we could choose n uniformly at random from {⌈η(H)⌉, ⌈η(H)⌉ +1}). 2. Choose AE ∈ Mn(Z)alt,≤X(H) uniformly at random, independently for each E. X′ ′ 3. Define random variables E := (coker A)tors and rkE := rkZ(ker A). X′ ′ Think of E as the “pseudo-Shafarevich–Tate group” of E and rkE as the “pseudo-rank” of E; their behavior is intended to model the actual X and rank. To complete the description of the model, we must specify the functions η(H) and X(H). We do this by asking “How large is X0 on average?”, both in the model and in reality. Recall 8 from (3) that we expect 1/12+o(1) (5) Average X0(E)= H . E∈E≤H Define X′ ′ X′ # E, if rkE =0; E,0 := ′ (0, if rkE > 0. Using that the determinant of an n × n matrix is given by a polynomial of degree n in the entries, we can prove that X′ η(H)(1+o(1)) (6) Average E,0 = X(H) , E∈E≤H assuming that η(H) does not grow too quickly with H. Comparing (5) and (6) suggests choosing η(H) and X(H) so that X(H)η(H) = H1/12+o(1). We assume this from now on. It turns out that we will not need to know any more about η(H) and X(H) than this. 3.5. Consequences of the model. To see what distribution of ranks is predicted by the refined model, we must calculate the distribution of ranks of alternating matrices whose entries are integers with bounded absolute value; the relevant result, whose proof is adapted from [EK95], is the following: Theorem 3.6 (cf. [PPVW16, Theorem 9.1.1]). If 1 ≤ r ≤ n and n − r is even, and A ∈ Mn(Z)alt,≤X is chosen uniformly at random, then n −(r−1)/2 Prob(rk(ker A) ≥ r) ≍n (X ) . (The subscript n on the symbol ≍ means that the implied constants depend on n.) Theorem 3.6 implies that for fixed r ≥ 1 and E ∈ E of height H, ′ η(H) −(r−1)/2+o(1) −(r−1)/24+o(1) (7) Prob(rkE ≥ r)=(X(H) ) = H . 5/6 20/24 Using this, and the fact #E≤H ≍ H = H (Proposition 2.1), we can now sum (7) over E ∈ E≤H to prove the following theorem about our model: Theorem 3.7 ([PPVW16, Theorem 7.3.3]). The following hold with probability 1: E ′ 20/24+o(1) #{E ∈ ≤H : rkE =0} = H E ′ 20/24+o(1) #{E ∈ ≤H : rkE =1} = H E ′ 19/24+o(1) #{E ∈ ≤H : rkE ≥ 2} = H E ′ 18/24+o(1) #{E ∈ ≤H : rkE ≥ 3} = H . . E ′ 1/24+o(1) #{E ∈ ≤H : rkE ≥ 20} = H E ′ o(1) #{E ∈ ≤H : rkE ≥ 21} ≤ H , E ′ #{E ∈ : rkE > 21} is finite. This suggests the conjecture that the same statements hold for the actual ranks of elliptic curves over Q. In particular, we conjecture that rk E(Q) is uniformly bounded, bounded by the maximum of the ranks of the conjecturally finitely many elliptic curves of rank > 21. 9 Remark 3.8. Elkies has found infinitely many elliptic curves over Q of rank ≥ 19, and one of rank ≥ 28; these have remained the records since 2006 [Elk06].

4. Generalizations 4.1. Elliptic curves over global ﬁelds. What about elliptic curves over other global ﬁelds K? Let EK be a set of representatives for the isomorphism classes of elliptic curves over K.

Let BK := lim supE∈EK rk E(K). For example, the conjecture suggested by Theorem 3.7 predicts that 20 ≤ BQ ≤ 21.

Theorem 4.1 ([TŠ67], [Ulm02]). If K is a global function ﬁeld, then BK = ∞.

Even for number fields, BK can be arbitrarily large (but maybe still always finite): Theorem 4.2 ([PPVW16, Theorem 12.4.2]). There exist number fields K of arbitrarily high degree such that BK ≥ [K : Q].

Examples of number fields K for which BK is large include fields in anticyclotomic towers and certain multiquadratic fields; see [PPVW16, Section 12.4]. A naive adaptation of our heuristic (see [PPVW16, Sections 12.2 and 12.3]) would suggest 20 ≤ BK ≤ 21 for every global field K, but Theorems 4.1 and 4.2 contradict this conclusion. Our rationalization of this is that the elliptic curves of high rank in Theorems 4.1 and 4.2 are special in that they are definable over a proper subfield of K, and these special curves exhibit arithmetic phenomena that our model does not take into account. To exclude these E ◦ E curves, let K be the set of E ∈ K such that E is not a base change of a curve from a ◦ ◦ proper subfield, and let B := lim sup E ◦ rk E(K). It is possible that B < ∞ for every K E∈ K K global field K. ◦ Remark 4.3. On the other hand, it is not true that BK ≤ 21 for all number fields, as we now explain. Shioda proved that y2 = x3+t360+1 has rank 68 over C(t) [Shi92]. In fact, it has rank 68 also over K(t) for a suitable number field K. For this K, specialization yields infinitely E ◦ ◦ many elliptic curves in K of rank ≥ 68. Thus BK ≥ 68. See [PPVW16, Remark 12.3.1] for details. 4.2. Abelian varieties. Question 4.4. For abelian varieties A over number fields K, is there a bound on rk A(K) depending only on dim A and [K : Q]? By restriction of scalars, we can reduce to the case K = Q at the expense of increasing the dimension. By “Zarhin’s trick” that A4 × (A∨)4 is principally polarized [Zar74a], we can reduce to the case that A is principally polarized, again at the expense of increasing the dimension. For fixed g ≥ 0, one can write down a family of projective varieties including all g-dimensional principally polarized abelian varieties over Q, probably with each isomorphism class represented infinitely many times. We can assume that each abelian variety A is defined by a system of homogeneous polynomials with integer coefficients, in which the number of variables, the number of polynomials, and their degrees are bounded in terms of g. Define the height of A to be the maximum of the absolute values of the coefficients. Then the number of g-dimensional principally polarized abelian varieties over Q of height ≤ H is bounded by ′ a polynomial in H. If there is a model involving a pseudo-rank rkA whose probability of 10 exceeding r gets divided by at least a fixed fractional power of H each time r is incremented by 1, as we had for elliptic curves, then the pseudo-ranks are bounded with probability 1. This might suggest a positive answer to Question 4.4, though the evidence is much flimsier than in the case of elliptic curves.

Acknowledgments I thank Nicolas Billerey, Serge Cantat, Andrew Granville, Eric Rains, Michael Stoll, and John Voight for comments.

References

[BHK+16] Jennifer S. Balakrishnan, Wei Ho, Nathan Kaplan, Simon Spicer, William Stein, and James Weigandt, Databases of elliptic curves ordered by height and distributions of Selmer groups and ranks, LMS J. Comput. Math. 19 (2016), no. suppl. A, 351–370. MR3540965 ↑1 [BKLPR15] Manjul Bhargava, Daniel M. Kane, Hendrik W. Lenstra Jr., Bjorn Poonen, and Eric Rains, Modeling the distribution of ranks, Selmer groups, and Shafarevich-Tate groups of elliptic curves, Camb. J. Math. 3 (2015), no. 3, 275–321. MR3393023 ↑1, 2.3, b, 2.4, 3.2, 3.4, 3.2, 3.3, 3.5, 3.3, 2 [BS15a] Manjul Bhargava and Arul Shankar, Binary quartic forms having bounded invariants, and the boundedness of the average rank of elliptic curves, Ann. of Math. (2) 181 (2015), no. 1, 191–242, DOI 10.4007/annals.2015.181.1.3. MR3272925 ↑3.2 [BS15b] Manjul Bhargava and Arul Shankar, Ternary cubic forms having bounded invariants, and the existence of a positive proportion of elliptic curves having rank 0, Ann. of Math. (2) 181 (2015), no. 2, 587–621, DOI 10.4007/annals.2015.181.2.4. MR3275847 ↑3.2 [BS13a] Manjul Bhargava and Arul Shankar, The average number of elements in the 4-Selmer groups of elliptic curves is 7, December 27, 2013. Preprint, arXiv:1312.7333v1 . ↑3.4, 3.2 [BS13b] Manjul Bhargava and Arul Shankar, The average size of the 5-Selmer group of elliptic curves is 6, and the average rank is less than 1, December 30, 2013. Preprint, arXiv:1312.7859v1 . ↑3.2 [Bru92] Armand Brumer, The average rank of elliptic curves. I, Invent. Math. 109 (1992), no. 3, 445–472, DOI 10.1007/BF01232033. MR1176198 (93g:11057) ↑1, 2.1 [Cas62] J. W. S. Cassels, Arithmetic on curves of genus 1. IV. Proof of the Hauptvermutung, J. Reine Angew. Math. 211 (1962), 95–112. MR0163915 (29 #1214) ↑2.3 [Cas66] J. W. S. Cassels, Diophantine equations with special reference to elliptic curves, J. London Math. Soc. 41 (1966), 193–291. MR0199150 (33 #7299) ↑1 [CL84] H. Cohen and H. W. Lenstra Jr., Heuristics on class groups of number fields, Number theory, Noordwijkerhout 1983 (Noordwijkerhout, 1983), Lecture Notes in Math., vol. 1068, Springer, Berlin, 1984, pp. 33–62, DOI 10.1007/BFb0099440. MR756082 (85j:11144) ↑1 [Del01] Christophe Delaunay, Heuristics on Tate-Shafarevitch groups of elliptic curves defined over Q, Experiment. Math. 10 (2001), no. 2, 191–196. MR1837670 (2003a:11065) ↑1 [Del07] Christophe Delaunay, Heuristics on class groups and on Tate-Shafarevich groups: the magic of the Cohen-Lenstra heuristics, Ranks of elliptic curves and random matrix theory, London Math. Soc. Lecture Note Ser., vol. 341, Cambridge Univ. Press, Cambridge, 2007, pp. 323–340. MR2322355 (2008i:11089) ↑1 [DJ14] Christophe Delaunay and Frédéric Jouhet, pℓ-torsion points in finite abelian groups and combi- natorial identities, Adv. Math. 258 (2014), 13–45, DOI 10.1016/j.aim.2014.02.033. MR3190422 ↑1 [Elk06] Noam D. Elkies, Z28 in E(Q), etc., May 3, 2006. Email to the [email protected] mailing list. ↑3.8 [EK95] Alex Eskin and Yonatan R. Katznelson, Singular symmetric matrices, Duke Math. J. 79 (1995), no. 2, 515–547, DOI 10.1215/S0012-7094-95-07913-7. MR1344769 (96h:11099) ↑3.5 11 [FGH07] David W. Farmer, S. M. Gonek, and C. P. Hughes, The maximum size of L-functions, J. Reine Angew. Math. 609 (2007), 215–236, DOI 10.1515/CRELLE.2007.064. MR2350784 (2009b:11140) ↑1 [FW89] Eduardo Friedman and Lawrence C. Washington, On the distribution of divisor class groups of curves over a finite field, Théorie des nombres (Quebec, PQ, 1987), de Gruyter, Berlin, 1989, pp. 227–239. MR1024565 (91e:11138) ↑3.3 [Gol79] Dorian Goldfeld, Conjectures on elliptic curves over quadratic fields, Number theory, Carbondale 1979 (Proc. Southern Illinois Conf., Southern Illinois Univ., Carbondale, Ill., 1979), Lecture Notes in Math., vol. 751, Springer, Berlin, 1979, pp. 108–118. MR564926 (81i:12014) ↑3.3 [GS95] Dorian Goldfeld and Lucien Szpiro, Bounds for the order of the Tate-Shafarevich group, Com- positio Math. 97 (1995), no. 1-2, 71–87. Special issue in honour of Frans Oort. MR1355118 (97a:11102) ↑2.5 [HB93] D. R. Heath-Brown, The size of Selmer groups for the congruent number problem, Invent. Math. 111 (1993), no. 1, 171–195, DOI 10.1007/BF01231285. MR1193603 (93j:11038) ↑3.2 [HB94] D. R. Heath-Brown, The size of Selmer groups for the congruent number problem. II, Invent. Math. 118 (1994), no. 2, 331–370, DOI 10.1007/BF01231536. With an appendix by P. Monsky. MR1292115 (95h:11064) ↑3.2 [Hin07] Marc Hindry, Why is it difficult to compute the Mordell-Weil group?, Diophantine geometry, CRM Series, vol. 4, Ed. Norm., Pisa, 2007, pp. 197–219. MR2349656 (2008i:11074) ↑2.5 [HP16] Marc Hindry and Amílcar Pacheco, An analogue of the Brauer–Siegel theorem for abelian varieties in positive characteristic, Mosc. Math. J. 16 (2016), no. 1, 45–93. MR3470576 ↑2.5 [Hon60] Taira Honda, Isogenies, rational points and section points of group varieties, Japan. J. Math. 30 (1960), 84–101. MR0155828 (27 #5762) ↑1 [IS00] H. Iwaniec and P. Sarnak, Perspectives on the analytic theory of L-functions, Geom. Funct. Anal. Special Volume (2000), 705–741. GAFA 2000 (Tel Aviv, 1999). MR1826269 (2002b:11117) ↑2.5 [Kan13] Daniel Kane, On the ranks of the 2-Selmer groups of twists of a given elliptic curve, Algebra Number Theory 7 (2013), no. 5, 1253–1279, DOI 10.2140/ant.2013.7.1253. MR3101079 ↑3.2 [KS99a] Nicholas M. Katz and Peter Sarnak, Random matrices, Frobenius eigenvalues, and monodromy, American Mathematical Society Colloquium Publications, vol. 45, Amer. Math. Soc., Provi- dence, RI, 1999. MR2000b:11070 ↑3.3 [KS99b] Nicholas M. Katz and Peter Sarnak, Zeroes of zeta functions and symmetry, Bull. Amer. Math. Soc. (N.S.) 36 (1999), no. 1, 1–26, DOI 10.1090/S0273-0979-99-00766-1. MR1640151 (2000f:11114) ↑3.3 [Lan83] William E. Lang, On Enriques surfaces in characteristic p.I, Math. Ann. 265 (1983), no. 1, 45–65. MR719350 (86c:14031) ↑2.5 [Maz77] B. Mazur, Modular curves and the Eisenstein ideal, Inst. Hautes Études Sci. Publ. Math. 47 (1977), 33–186 (1978). MR488287 (80c:14015) ↑2.5 [Mes82] Jean-François Mestre, Construction d’une courbe elliptique de rang ≥ 12, C. R. Acad. Sci. Paris Sér. I Math. 295 (1982), no. 12, 643–644 (French, with English summary). MR688896 (84b:14019) ↑1 [Mes86] Jean-François Mestre, Formules explicites et minorations de conducteurs de variétés algébriques, Compositio Math. 58 (1986), no. 2, 209–232 (French). MR844410 (87j:11059) ↑1 [Mor22] L. J. Mordell, On the rational solutions of the indeterminate equations of the third and fourth degrees, Proc. Cambridge Phil. Soc. 21 (1922), 179–192. ↑1 [Mum91] David Mumford, Tata lectures on theta. III, Progress in Mathematics, vol. 97, Birkhäuser Boston Inc., Boston, MA, 1991. With the collaboration of Madhav Nori and Peter Norman. MR1116553 (93d:14065) ↑2.2 [O’N02] Catherine O’Neil, The period-index obstruction for elliptic curves, J. Number Theory 95 (2002), no. 2, 329–339, DOI 10.1016/S0022-314X(01)92770-2. Erratum in J. Number Theory 109 (2004), no. 2, 390. MR1924106 (2003f:11079) ↑2.2 [PPVW16] Jennifer Park, Bjorn Poonen, John Voight, and Melanie Matchett Wood, A heuristic for boundedness of ranks of elliptic curves, February 3, 2016. Preprint, arXiv:1602.01431v1, to appear in J. Europ. Math. Soc. ↑1, 1, 2.5, 3.6, 3.7, 4.2, 4.1, 4.3 12 [Poi01] H. Poincaré, Sur les propriétés arithmétiques des courbes algébriques, J. Pures Appl. Math. (5) 7 (1901), 161–234. ↑1 [Poi50] Henri Poincaré, Œuvres d’Henri Poincaré, Volume 5 (Albert Châtelet, ed.), Gauthier-Villars, Paris, 1950. ↑1 [PR12] Bjorn Poonen and Eric Rains, Random maximal isotropic subspaces and Selmer groups, J. Amer. Math. Soc. 25 (2012), no. 1, 245–269, DOI 10.1090/S0894-0347-2011-00710-8. MR2833483 ↑1, 2.2, a, b, 3.1, 3.1, 3.2, 3.4, 3.2 [RS00] Karl Rubin and Alice Silverberg, Ranks of elliptic curves in families of quadratic twists, Exper- iment. Math. 9 (2000), no. 4, 583–590. MR1806293 (2001k:11105) ↑1 [Shi92] Tetsuji Shioda, Some remarks on elliptic curves over function fields, Astérisque 209 (1992), 12, 99–114. Journées Arithmétiques, 1991 (Geneva). MR1211006 (94d:11046) ↑4.3 [Sil09] Joseph H. Silverman, The arithmetic of elliptic curves, 2nd ed., Graduate Texts in Mathematics, vol. 106, Springer, Dordrecht, 2009. MR2514094 (2010i:11005) ↑1 [Smi17] Alexander Smith, 2∞-Selmer groups, 2∞-class groups, and Goldfeld’s conjecture, June 7, 2017. Preprint, arXiv:1702.02325v2. ↑3.3 [SW13] William Stein and Christian Wuthrich, Algorithms for the arithmetic of elliptic curves using Iwasawa theory, Math. Comp. 82 (2013), no. 283, 1757–1792, DOI 10.1090/S0025-5718-2012- 02649-4. MR3042584 ↑2.5 [SD08] Peter Swinnerton-Dyer, The effect of twisting on the 2-Selmer group, Math. Proc. Cam- bridge Philos. Soc. 145 (2008), no. 3, 513–526, DOI 10.1017/S0305004108001588. MR2464773 (2010d:11059) ↑3.2 [Tat74] John T. Tate, The arithmetic of elliptic curves, Invent. Math. 23 (1974), 179–206. MR0419359 (54 #7380) ↑1 [TŠ67] D. T. Tè˘ıt and I. R. Šafarevič, The rank of elliptic curves, Dokl. Akad. Nauk SSSR 175 (1967), 770–773 (Russian). MR0237508 (38 #5790) ↑4.1 [Ulm02] Douglas Ulmer, Elliptic curves with large rank over function fields, Ann. of Math. (2) 155 (2002), no. 1, 295–315, DOI 10.2307/3062158. MR1888802 (2003b:11059) ↑1, 4.1 [Wat08] Mark Watkins, Some heuristics about elliptic curves, Experiment. Math. 17 (2008), no. 1, 105– 125. MR2410120 (2009g:11076) ↑2.5 [Wat15] Mark Watkins, A discursus on 21 as a bound for ranks of elliptic curves over Q, and sundry related topics, August 20, 2015. Available at http://magma.maths.usyd.edu.au/~watkins/papers/DISCURSUS.pdf . ↑1 [Wat+14] Mark Watkins, Stephen Donnelly, Noam D. Elkies, Tom Fisher, Andrew Granville, and Nicholas F. Rogers, Ranks of quadratic twists of elliptic curves, Publ. math. de Besançon 2014/2 (2014), 63–98 (English, with English and French summaries). ↑1 [dW98] Benjamin M. M. de Weger, A + B = C and big X’s, Quart. J. Math. Oxford Ser. (2) 49 (1998), no. 193, 105–128, DOI 10.1093/qjmath/49.193.105 (English). MR1617347 (99j:11065) ↑2.5 [Wil06] Andrew Wiles, The Birch and Swinnerton-Dyer conjecture, The millennium prize problems, Clay Math. Inst., Cambridge, MA, 2006, pp. 31–41. MR2238272 ↑2.5 [Zar74a] Ju. G. Zarhin, A remark on endomorphisms of abelian varieties over function fields of finite characteristic, Izv. Akad. Nauk SSSR Ser. Mat. 38 (1974), 471–474 (Russian). MR0354689 (50 #7166) ↑4.2 [Zar74b] Ju. G. Zarhin, Noncommutative cohomology and Mumford groups, Mat. Zametki 15 (1974), 415–419 (Russian). MR0354612 (50 #7090) ↑2.2

Department of Mathematics, Massachusetts Institute of Technology, Cambridge, MA 02139-4307, USA E-mail address: [email protected] URL: http://math.mit.edu/~poonen/