A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A...
Transcript of A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A...
![Page 1: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/1.jpg)
A heuristic for abelian points on elliptic curves
Barry Mazur, Harvard UniversityKarl Rubin, UC Irvine
MIT, August 2018
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 2: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/2.jpg)
Growth of ranks in cyclic extensions
Fix an elliptic curve E over a number field K.
QuestionAs F runs through abelian extensions of K, how often isrank(E(F)) > rank(E(K))?
By considering the action of Gal(F/K) on E(F)⊗Q, the representationtheory of Q[Gal(F/K)] shows that it is enough to consider the casewhere F/K is cyclic.
General philosophy: it’s hard to find P ∈ E(Kab) with K(P)/K cyclic oflarge degree.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 3: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/3.jpg)
Growth of ranks: analytic approach
QuestionAs F runs through cyclic extensions of K, how often isrank(E(F)) > rank(E(K))?
Using BSD and the factorization
L(E/F, s) =∏
χ:Gal(F/K)→C×L(E, χ, s)
this is equivalent to:
QuestionAs χ runs through characters of Gal(K/K), how often is L(E, χ, 1) = 0?
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 4: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/4.jpg)
Modular symbols
Fix E/Q once and for all, and suppress it from the notation.
DefinitionFor r ∈ Q, define the (plus) modular symbol [r] = [r]E by
[r] :=12
(2πiΩ
∫ r
i∞fE(z)dz +
2πiΩ
∫ −r
i∞fE(z)dz
)∈ Q
where fE is the modular form attached to E, and Ω is the real period.
TheoremFor every primitive even Dirichlet character χ of conductor m,∑
a∈(Z/mZ)×χ(a)[a/m] =
τ(χ)L(E, χ, 1)
Ω.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 5: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/5.jpg)
Modular symbols
Fix E/Q once and for all, and suppress it from the notation.
DefinitionFor r ∈ Q, define the (plus) modular symbol [r] = [r]E by
[r] :=12
(2πiΩ
∫ r
i∞fE(z)dz +
2πiΩ
∫ −r
i∞fE(z)dz
)∈ Q
where fE is the modular form attached to E, and Ω is the real period.
TheoremFor every primitive even Dirichlet character χ of conductor m,∑
a∈(Z/mZ)×χ(a)[a/m] =
τ(χ)L(E, χ, 1)
Ω.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 6: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/6.jpg)
theta elements
In particular
L(E, χ, 1) = 0 ⇐⇒∑
a∈(Z/mZ)×χ(a)[a/m] = 0.
We want to use statistical properties of modular symbols to predicthow often this happens.
If m ≥ 1, and F/Q is cyclic of conductor m, let
σa,m ∈ Gal(Q(µm)/Q) the automorphism ζm 7→ ζam,
θm :=∑
a∈(Z/mZ)× [a/m]σa,m ∈ Q[Gal(Q(µm)/Q)],θF := θm|F ∈ Q[Gal(F/Q)].
If χ is an even character of Gal(F/Q), then
L(E, χ, 1) = 0 ⇐⇒ χ(θF) = 0.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 7: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/7.jpg)
theta elements
In particular
L(E, χ, 1) = 0 ⇐⇒∑
a∈(Z/mZ)×χ(a)[a/m] = 0.
We want to use statistical properties of modular symbols to predicthow often this happens.
If m ≥ 1, and F/Q is cyclic of conductor m, let
σa,m ∈ Gal(Q(µm)/Q) the automorphism ζm 7→ ζam,
θm :=∑
a∈(Z/mZ)× [a/m]σa,m ∈ Q[Gal(Q(µm)/Q)],θF := θm|F ∈ Q[Gal(F/Q)].
If χ is an even character of Gal(F/Q), then
L(E, χ, 1) = 0 ⇐⇒ χ(θF) = 0.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 8: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/8.jpg)
theta elements
In particular
L(E, χ, 1) = 0 ⇐⇒∑
a∈(Z/mZ)×χ(a)[a/m] = 0.
We want to use statistical properties of modular symbols to predicthow often this happens.
If m ≥ 1, and F/Q is cyclic of conductor m, let
σa,m ∈ Gal(Q(µm)/Q) the automorphism ζm 7→ ζam,
θm :=∑
a∈(Z/mZ)× [a/m]σa,m ∈ Q[Gal(Q(µm)/Q)],θF := θm|F ∈ Q[Gal(F/Q)].
If χ is an even character of Gal(F/Q), then
L(E, χ, 1) = 0 ⇐⇒ χ(θF) = 0.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 9: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/9.jpg)
theta coefficients
We haveθF =
∑γ∈Gal(F/Q)
cF,γγ
wherecF,γ =
∑σa,m|F=γ
[a/m].
How likely is it that χ(θF) = 0?
ExampleSuppose [F : Q] = p is prime, and χ : Gal(F/Q)→ C× is nontrivial. Theonly nontrivial Q-linear relation among the p-th roots of unity is thattheir sum is zero, so
χ(θF) = 0 ⇐⇒ cF,γ = cF,γ′ ∀γ, γ′ ∈ Gal(F/Q).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 10: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/10.jpg)
theta coefficients
We haveθF =
∑γ∈Gal(F/Q)
cF,γγ
wherecF,γ =
∑σa,m|F=γ
[a/m].
How likely is it that χ(θF) = 0?
ExampleSuppose [F : Q] = p is prime, and χ : Gal(F/Q)→ C× is nontrivial. Theonly nontrivial Q-linear relation among the p-th roots of unity is thattheir sum is zero, so
χ(θF) = 0 ⇐⇒ cF,γ = cF,γ′ ∀γ, γ′ ∈ Gal(F/Q).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 11: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/11.jpg)
Modular symbols
Let N be the conductor of E. For every r ∈ Q, modular symbols satisfy:
There is a δ ∈ Z independent of r such that δ[r] ∈ Z
[r] = [r + 1] since fE(z) = fE(z + 1)
[r] = [−r] by definition
Atkin-Lehner relation: if w is the global root number of E, andaa′N ≡ 1 (mod m), then [a′/m] = w[a/m]
Hecke relation: if a prime ` - N and a` is the `-th Fourier coefficient
of fE, then a`[r] = [`r] +∑`−1
i=0 [(r + i)/`]
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 12: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/12.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 10007, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 13: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/13.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 10007, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 14: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/14.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 100003, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 15: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/15.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 100003, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 16: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/16.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 1000003, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 17: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/17.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 1000003, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 18: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/18.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 10000019, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 19: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/19.jpg)
Distribution of modular symbols
Histogram of [a/m] : E = 11A1,m = 10000019, a ∈ (Z/mZ)×
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 20: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/20.jpg)
Distribution of modular symbols
This looks like a normal distribution.
How does the variance depend on m?
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 21: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/21.jpg)
Distribution of modular symbols
This looks like a normal distribution.
How does the variance depend on m?
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 22: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/22.jpg)
Distribution of modular symbols
Plot of variance vs. m, for E = 11A1:
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 23: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/23.jpg)
Distribution of modular symbols
Plot of variance vs. m, for E = 45A1:
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 24: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/24.jpg)
Distribution of modular symbols
For m ≥ 1 let Sm denote the data Sm = [a/m] : a ∈ (Z/mZ)×.
ConjectureThere is an explicit constant VE such that
1 as m→∞, the distribution of the 1√log(m)
Sm converge to a normal
distribution with mean zero and variance VE.2 for every divisor κ of N, lim m→∞
(m,N)=κVariance(Sm)− VE log(m) exists
and is finite.
Theorem (Petridis-Risager)The conjecture above holds if N is squarefree and we average over m.
The variance VE is essentially L(Sym2(E), 1), and Petridis & Risagercompute the limit in 2 in terms of L(Sym2(E), 1) and L′(Sym2(E), 1).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 25: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/25.jpg)
Distribution of modular symbols
For m ≥ 1 let Sm denote the data Sm = [a/m] : a ∈ (Z/mZ)×.
ConjectureThere is an explicit constant VE such that
1 as m→∞, the distribution of the 1√log(m)
Sm converge to a normal
distribution with mean zero and variance VE.2 for every divisor κ of N, lim m→∞
(m,N)=κVariance(Sm)− VE log(m) exists
and is finite.
Theorem (Petridis-Risager)The conjecture above holds if N is squarefree and we average over m.
The variance VE is essentially L(Sym2(E), 1), and Petridis & Risagercompute the limit in 2 in terms of L(Sym2(E), 1) and L′(Sym2(E), 1).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 26: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/26.jpg)
Distribution of theta coefficients
What does this tell us about the distribution of the theta coefficients?
If [F : Q] = d, then each theta coefficient cF,γ is a sum of ϕ(m)/dmodular symbols. We (think we) know how the modular symbols aredistributed, but are they independent? If so, then the
cF,γ√VE log(m)(ϕ(m)/d)
should satisfy a normal distribution with variance 1.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 27: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/27.jpg)
Distribution of normalized theta coefficients, d = 3
E = 11A1, primes m ≡ 1 (mod 3), L ⊂ Q(µm), [L : Q] = 3,
10000 < m < 20000:
The red curve is theexpected normaldistribution.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 28: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/28.jpg)
Distribution of normalized theta coefficients, d = 3
E = 11A1, primes m ≡ 1 (mod 3), L ⊂ Q(µm), [L : Q] = 3,
20000 < m < 40000:
The red curve is theexpected normaldistribution.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 29: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/29.jpg)
Distribution of normalized theta coefficients, d = 3
E = 11A1, primes m ≡ 1 (mod 3), L ⊂ Q(µm), [L : Q] = 3,
40000 < m < 80000:
The red curve is theexpected normaldistribution.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 30: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/30.jpg)
Distribution of normalized theta coefficients, d = 3
E = 11A1, primes m ≡ 1 (mod 3), L ⊂ Q(µm), [L : Q] = 3,
80000 < m < 160000:
The red curve is theexpected normaldistribution.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 31: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/31.jpg)
Distribution of normalized theta coefficients, d = 3
E = 11A1, primes m ≡ 1 (mod 3), L ⊂ Q(µm), [L : Q] = 3,
160000 < m < 320000:
The red curve is theexpected normaldistribution.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 32: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/32.jpg)
Distribution of normalized theta coefficients, d = 3
E = 11A1, primes m ≡ 1 (mod 3), L ⊂ Q(µm), [L : Q] = 3,
320000 < m < 640000:
The red curve is theexpected normaldistribution.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 33: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/33.jpg)
Distribution of normalized theta coefficients, d = 3
E = 11A1, primes m ≡ 1 (mod 3), L ⊂ Q(µm), [L : Q] = 3,
10000 < m < 640000:
The red curve is theexpected normaldistribution.
The data does seem to beconverging (so the scalingfactor looks correct), butnot to the expected limit.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 34: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/34.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 3
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 35: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/35.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 5
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 36: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/36.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 7
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 37: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/37.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 11
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 38: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/38.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 13
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 39: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/39.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 17
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 40: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/40.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 23
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 41: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/41.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 31
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 42: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/42.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 41
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 43: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/43.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 53
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 44: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/44.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 97
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 45: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/45.jpg)
Distribution of theta coefficients, small d
E = 11A1, m ≡ 1 (mod d), L ⊂ Q(µm), [L : Q] = d,
d = 293
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 46: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/46.jpg)
Oversimplified picture
Suppose F/Q is a cyclic extension of degree d and conductor m.
Very roughly, θF lies in a cube of side√
VE log(m)ϕ(m)/d in thed-dimensional lattice Z[Gal(F/Q)].
Suppose χ : Gal(F/Q)∼−→ µd is a faithful character. Then
L(E, χ, 1) = 0 ⇐⇒ θF ∈ ker(χ : Z[Gal(F/Q)]→ C).
That kernel is a sublattice of codimension ϕ(d), so we might expect the“probability” that L(E, χ, 1) = 0 should be about(
CE√log(m)ϕ(m)/d
)ϕ(d)for some constant CE.
This goes to zero very fast as d and m grow.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 47: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/47.jpg)
Oversimplified picture
This isn’t quite right:
The previous argument ignores the Atkin-Lehner relation, which“pairs up” the coefficients and forces θF into a sublattice ofZ[Gal(F/Q)] with rank approximately d/2. Taking this into accountchanges the expectation to(
CE dlog(m)ϕ(m)
)ϕ(d)/4
.
The distribution of the (normalized) θL is not uniform in a box, andwe don’t fully understand what the correct distribution is.Fortunately, for applications, it doesn’t seem to matter very muchwhat the distribution is, only that there is one and it’s boundedindependent of d.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 48: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/48.jpg)
Oversimplified picture
This isn’t quite right:
The previous argument ignores the Atkin-Lehner relation, which“pairs up” the coefficients and forces θF into a sublattice ofZ[Gal(F/Q)] with rank approximately d/2. Taking this into accountchanges the expectation to(
CE dlog(m)ϕ(m)
)ϕ(d)/4
.
The distribution of the (normalized) θL is not uniform in a box, andwe don’t fully understand what the correct distribution is.Fortunately, for applications, it doesn’t seem to matter very muchwhat the distribution is, only that there is one and it’s boundedindependent of d.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 49: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/49.jpg)
The heuristic
This all leads to the folowing heuristic estimate:
HeuristicThere is a constant CE, depending only on E, such that
Prob[L(E, χ, 1) = 0] ≤(
CE dlog(m)ϕ(m)
)ϕ(d)/4
where d > 2 is the order of χ and m its conductor.
The exponent ϕ(d)/4 comes from the assumption that the thetacoefficients are independent (except for the Atkin-Lehner relation).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 50: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/50.jpg)
Consequences of the heuristic
Heuristic
Prob[L(E, χ, 1) = 0] ≤( CE dlog(m)ϕ(m)
)ϕ(d)/4.
Random matrix theory prediction:
RMT prediction (David, Fearnley, Kisilevsky)For fixed prime d,
Prob[L(E, χ, 1) = 0] ≈ CE,d
(√log(m)
m
)ϕ(d)/4
.
These agree up to power of log.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 51: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/51.jpg)
Consequences of the heuristicHeuristic
Prob[L(E, χ, 1) = 0] ≤( CE dlog(m)ϕ(m)
)ϕ(d)/4.
PropositionSuppose t : Z>0 → R≥0 is a function, and t(d) log(d). Then
∑d : t(d)>1
∑χ order d
(CE d
log(m)ϕ(m)
)t(d)
converges.
Applying this with t(d) = ϕ(d)/4 shows
Heuristic ∑d : ϕ(d)>4
∑χ order d
Prob[L(E, χ, 1) = 0] converges.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 52: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/52.jpg)
Consequences of the heuristic
This leads to:
ConjectureSuppose L/Q is an abelian extension with only finitely many subfieldsof degree 2, 3, or 5 over Q.
Then for every elliptic curve E/Q, we expect that E(L) is finitelygenerated.
For example, these conditions hold when L is:
the Z-extension of Q,the maximal abelian `-extension of Q, for ` ≥ 7,the compositum of all of the above.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 53: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/53.jpg)
Consequences of the heuristic
This leads to:
ConjectureSuppose L/Q is an abelian extension with only finitely many subfieldsof degree 2, 3, or 5 over Q.
Then for every elliptic curve E/Q, we expect that E(L) is finitelygenerated.
For example, these conditions hold when L is:
the Z-extension of Q,the maximal abelian `-extension of Q, for ` ≥ 7,the compositum of all of the above.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 54: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/54.jpg)
Consequences of the heuristic
Alternatively:
ConjectureSuppose E is an elliptic curve over Q, and let M denote thecompositum of all abelian fields of degree at most 5.
Then E(Qab)/E(M) is finitely generated.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 55: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/55.jpg)
(In)dependence: star-like structure
Suppose d is an odd prime. Consider F/Q cyclic of degree d and Ewith global root number −1. Then the Atkin-Lehner relation tells us thatone of the d theta coefficients is zero, and the others come in (d − 1)/2pairs (c,−c).
If d = 5 or 7 we can plot the (d − 1)/2-tuples of (normalized) thetacoefficients. If they are indeed independent, we should get a cloud ofdata points concentrated near the origin without much other structure.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 56: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/56.jpg)
(In)dependence: star-like structure
Example: E = 37A1, d = 5
If we view a point (x, y) as an elementxγ+yγ2−yγ3−xγ4 ∈ R[Gal(F/Q)], thenthe asymptotes are the lines where oneof the 2 complex conjugate pairs ofcharacters vanishes.
In other words, this says that if oneof the values χ(θF) is large, theother has to be small.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 57: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/57.jpg)
(In)dependence: star-like structure
Example: E = 37A1, d = 5If we view a point (x, y) as an elementxγ+yγ2−yγ3−xγ4 ∈ R[Gal(F/Q)], thenthe asymptotes are the lines where oneof the 2 complex conjugate pairs ofcharacters vanishes.
In other words, this says that if oneof the values χ(θF) is large, theother has to be small.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 58: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/58.jpg)
(In)dependence: star-like structure
Example: E = 37A1, d = 7
[3-dimensional Mathematica graphic]
In this example the asymptote lines are the lines where all except oneof the (d − 1)/2 pairs of characters vanish.
In other words, if one of the values χ(θF) is large, all the others seemto be (relatively) small.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 59: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/59.jpg)
(In)dependence: star-like structure
Example: E = 37A1, d = 7
[3-dimensional Mathematica graphic]
In this example the asymptote lines are the lines where all except oneof the (d − 1)/2 pairs of characters vanish.
In other words, if one of the values χ(θF) is large, all the others seemto be (relatively) small.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 60: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/60.jpg)
(In)dependence: star-like structure
When d = 5 and the root number is +1, then there are 3 (potentially)independent theta coefficients. However, the Hecke relation on themodular symbols says that∑
γ
cF,γ =
(∏`|m
(a` − 2)
)[0]
√m
Since the cF,γ have size roughly√
m log(m), this says that the sum ofall the theta coefficients is essentially zero for large m.
Example: E = 11A1, d = 5
[3-dimensional Mathematica graphic]
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 61: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/61.jpg)
(In)dependence: star-like structure
Another way to try to measure this phenomenon: Consider the monicpolynomial
fF(x) = x(p−1)/2 + c1x(p−3)/2 + · · ·+ c(p−1)/2
whose roots are the (p− 1)/2 positive real numbers
|χ(θF)|2
mχ log(mχ)
for nontrivial χ (note that the |χ(θF)|2 are positive, real, conjugatecyclotomic integers).
The assertion that at most one |χ(θF)|2 is ‘large’ is similar to askingthat in the set c1/n
n : 1 ≤ n ≤ (p− 1)/2, only c1 can be ‘large’.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 62: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/62.jpg)
(In)dependence: star-like structure
Another way to try to measure this phenomenon: Consider the monicpolynomial
fF(x) = x(p−1)/2 + c1x(p−3)/2 + · · ·+ c(p−1)/2
whose roots are the (p− 1)/2 positive real numbers
|χ(θF)|2
mχ log(mχ)
for nontrivial χ (note that the |χ(θF)|2 are positive, real, conjugatecyclotomic integers).
The assertion that at most one |χ(θF)|2 is ‘large’ is similar to askingthat in the set c1/n
n : 1 ≤ n ≤ (p− 1)/2, only c1 can be ‘large’.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 63: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/63.jpg)
Constructing abelian points
Suppose we have a diagram of curves
C
cyclic, degree dφ
E Y
If P ∈ Y(Q), then the points in the fiber φ−1(P) are defined over a cyclicextension L/Q of degree (dividing) d. These project to points in E(L).
If Y = P1, and d is prime, this produces ∼ Xα cyclic extensions L/Q ofdegree d with rank E(L) > rank E(Q) and disc(L) < X, where α dependson deg(φ).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 64: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/64.jpg)
Constructing abelian points
Suppose we have a diagram of curves
C
cyclic, degree dφ
E Y
If P ∈ Y(Q), then the points in the fiber φ−1(P) are defined over a cyclicextension L/Q of degree (dividing) d. These project to points in E(L).
If Y = P1, and d is prime, this produces ∼ Xα cyclic extensions L/Q ofdegree d with rank E(L) > rank E(Q) and disc(L) < X, where α dependson deg(φ).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 65: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/65.jpg)
Constructing abelian points
Suppose we have a diagram of curves
C
cyclic, degree dφ
E Y
If P ∈ Y(Q), then the points in the fiber φ−1(P) are defined over a cyclicextension L/Q of degree (dividing) d. These project to points in E(L).
If Y = P1, and d is prime, this produces ∼ Xα cyclic extensions L/Q ofdegree d with rank E(L) > rank E(Q) and disc(L) < X, where α dependson deg(φ).
If d = 3, then for many elliptic curves E with a 3-isogeny, we canconstruct such a diagram with Y = P1.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 66: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/66.jpg)
Constructing abelian points
Suppose we have a diagram of curves
C
cyclic, degree dφ
E Y
If P ∈ Y(Q), then the points in the fiber φ−1(P) are defined over a cyclicextension L/Q of degree (dividing) d. These project to points in E(L).
If Y = P1, and d is prime, this produces ∼ Xα cyclic extensions L/Q ofdegree d with rank E(L) > rank E(Q) and disc(L) < X, where α dependson deg(φ).
If d = 5 we know one example of such a diagram (C =“Bring’scurve”, of genus 4) with E in the isogeny class 50A and with Y ofgenus zero but Y(Q) = ∅. For many imaginary quadratic fields Kthis gives many cyclic degree 5 extensions L/K withrank E(L) > rank E(K).
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 67: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/67.jpg)
Extensions and generalizations
Extension to other base fields: Suppose now that K is a number fieldand E is an elliptic curve over K. In this case there can be characters χof Gal(K/K) such that L(E, χ, 1) vanishes because of root numberconsiderations.
For all other χ we can ask whether
Prob[L(E, χ, 1) = 0](
CEdχlog(mχ)ϕ(mχ)
)ϕ(dχ)/4
(*)
where dχ is the order of χ and mχ is the norm of its conductor.
(The motivation for (*) depends on the distribution of theta coefficientsfor abelian extensions F/K. Maarten Derickx and Alex Best arecurrently working to compute these general theta coefficients.)
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 68: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/68.jpg)
Extensions and generalizations
Extension to other base fields: Suppose now that K is a number fieldand E is an elliptic curve over K. In this case there can be characters χof Gal(K/K) such that L(E, χ, 1) vanishes because of root numberconsiderations.
For all other χ we can ask whether
Prob[L(E, χ, 1) = 0](
CEdχlog(mχ)ϕ(mχ)
)ϕ(dχ)/4
(*)
where dχ is the order of χ and mχ is the norm of its conductor.
(The motivation for (*) depends on the distribution of theta coefficientsfor abelian extensions F/K. Maarten Derickx and Alex Best arecurrently working to compute these general theta coefficients.)
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 69: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/69.jpg)
Extensions and generalizations
ConjectureSuppose L/K is an abelian extension with only finitely many subfieldsof degree 2, 3, or 5.
Then for every elliptic curve E/K, if we exclude those characters thatvanish for root number considerations, then we expect L(E, χ, 1) = 0for only finitely many other characters χ of Gal(L/K).
ConjectureSuppose L/Q is an abelian extension with only finitely many subfieldsof degree 2, 3, or 5.
Then for every elliptic curve E/L, we expect that E(L) is finitelygenerated.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 70: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/70.jpg)
Extensions and generalizations
ConjectureSuppose L/K is an abelian extension with only finitely many subfieldsof degree 2, 3, or 5.
Then for every elliptic curve E/K, if we exclude those characters thatvanish for root number considerations, then we expect L(E, χ, 1) = 0for only finitely many other characters χ of Gal(L/K).
ConjectureSuppose L/Q is an abelian extension with only finitely many subfieldsof degree 2, 3, or 5.
Then for every elliptic curve E/L, we expect that E(L) is finitelygenerated.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 71: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/71.jpg)
Extensions and generalizations
Studying p-Selmer: Instead of asking how often L(E, χ, 1) = 0, we canask how often L(E, χ, 1)/ΩE is divisible by (some prime above) p. Bythe Birch & Swinnerton-Dyer conjecture, this should tell us about thep-Selmer group Selp(E/L).
It seems reasonable to expect that if the θ-coefficents cL,χ,g are not allthe same (mod p), then they are equidistributed (mod p).
For example, this leads to the following:
ConjectureLet S be a finite set of rational primes, not containing p. Let L be thecompositum of the cyclotomic Z`-extensions of Q for ` ∈ S. If E is anelliptic curve over Q whose mod p representation is irreducible, thendimFp Selp(E/L) is finite.
The heuristic does not predict finite p-Selmer rank when S is infinite.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018
![Page 72: A heuristic for abelian points on elliptic curveskrubin/lectures/simons.pdf · 2018. 8. 21. · A heuristic for abelian points on elliptic curves Barry Mazur, Harvard University Karl](https://reader035.fdocuments.in/reader035/viewer/2022071006/5fc3af51ce4ffa0de53a5058/html5/thumbnails/72.jpg)
Extensions and generalizations
Studying p-Selmer: Instead of asking how often L(E, χ, 1) = 0, we canask how often L(E, χ, 1)/ΩE is divisible by (some prime above) p. Bythe Birch & Swinnerton-Dyer conjecture, this should tell us about thep-Selmer group Selp(E/L).
It seems reasonable to expect that if the θ-coefficents cL,χ,g are not allthe same (mod p), then they are equidistributed (mod p).
For example, this leads to the following:
ConjectureLet S be a finite set of rational primes, not containing p. Let L be thecompositum of the cyclotomic Z`-extensions of Q for ` ∈ S. If E is anelliptic curve over Q whose mod p representation is irreducible, thendimFp Selp(E/L) is finite.
The heuristic does not predict finite p-Selmer rank when S is infinite.
Mazur & Rubin Heuristic for abelian points on elliptic curves MIT, August 2018