4
$\begingroup$

We are searching for rank $8$ elliptic curves with the torsion subgroup $\mathbb{Z}/6$ using newly discovered families similar to Kihara's as described in

A. Dujella, J.C. Peral, P. Tadić, Elliptic curves with torsion group $\mathbb{Z}/6\mathbb{Z}$, Glas. Mat. Ser. III 51 (2016), 321-333 doi:10.3336/gm.51.2.03, 1503.03667

and came across a curve

[0,1,0,-60313024735007362096072931339173916555726439220,3303762732437764940265112114690488828303891527290367159038723810320068]

Both Magma Calculator and mwrank (with $-b14$) return $7$ generators for this curve:

SetClassGroupBounds("GRH");
E:=EllipticCurve([0,1,0,-60313024735007362096072931339173916555726439220,3303762732437764940265112114690488828303891527290367159038723810320068]);
MordellWeilShaInformation(E);

Using model [ 0, 1, 0, -60313024735007362096072931339173916555726439220, 3303762732437764940265112114690488828303891527290367159038723810320068 ]
Torsion Subgroup = Z/6
The 2-Selmer group has rank 9
New point of infinite order (x = 5260000916960497219694209104164/9078169)
New point of infinite order (x = -146551684206472947976069)
New point of infinite order (x = -151681843950496144133344)
New point of infinite order (x = 211401387771733499670296)
New point of infinite order (x = 1521645572343712794396956)
New point of infinite order (x = 14998786693919437193768749863407/42159049)
New point of infinite order (x = 1185363853402839599279348827593704/4311629569)
After 2-descent:
    7 <= Rank(E) <= 8
    Sha(E)[2] <= (Z/2)^1
(Searched up to height 10000 on the 2-coverings.)

Both Magma and mwrank return $8$ for the upper bound on rank:

E:=EllipticCurve([0,1,0,-60313024735007362096072931339173916555726439220,3303762732437764940265112114690488828303891527290367159038723810320068]);
TwoPowerIsogenyDescentRankBound(E);

8 [ 3, 3, 3, 3, 3 ]
[ 7, 7, 7, 7, 7 ]
mwrank -v0 -p200 -s
[0,1,0,-60313024735007362096072931339173916555726439220,3303762732437764940265112114690488828303891527290367159038723810320068]

Version compiled on Oct 29 2018 at 22:35:09 by GCC 7.3.0
using NTL bigints and NTL real and complex multiprecision floating point
Enter curve: [0,1,0,-60313024735007362096072931339173916555726439220,3303762732437764940265112114690488828303891527290367159038723810320068]
Curve [0,1,0,-60313024735007362096072931339173916555726439220,3303762732437764940265112114690488828303891527290367159038723810320068] :  selmer-rank = 9
upper bound on rank = 8

Considering parity, there should be one more generator on the curve.

Is there a way to find it?

We would greatly appreciate any hint leading to the discovery of the extra generator.

A bounty of $100$ will be offered for obtaining it.

Also, if you can compute an extra generator, your name will be published at the bottom of the page here: https://web.math.pmf.unizg.hr/~duje/tors/z6.html.

So far, we are unsuccsessful applying Zev Klagsbrun's approach.

$\endgroup$
  • 2
    $\begingroup$ I'd appreciate the title not to contain the OP's opinion of the difficulty of the question (and other meta information). $\endgroup$ – YCor Mar 2 at 16:02
  • 2
    $\begingroup$ Did you try applying either method to the 3-isogenous curve E' ? It's possible that it will turn up one or more points whose image(s) in E aren't all in the rank-7 subgroup you've already found. $\endgroup$ – Noam D. Elkies Mar 2 at 16:28
  • 2
    $\begingroup$ No, I mean 3-isogenous. Since there's a 3-torsion point there's a 3-isogenous curve. Using the 2-isogenous curve will produce nothing new because mwrank already works with a 2-isogenous pair when there's a rational 2-torsion point. But the subgroup that pulls back to the 3-isogenous curve E' is easier to compute on E', and may include something new. $\endgroup$ – Noam D. Elkies Mar 2 at 23:18
  • 2
    $\begingroup$ Is it "the same" rank-7 subgroup, though? One easy test is to compare the regulators. If their ratio is a small power of 3 then it must be the same group. If it looks like a random real number then mapping back to E should give something independent of your known rank-7 group on E. $\endgroup$ – Noam D. Elkies Mar 3 at 1:23
  • 2
    $\begingroup$ Your computation seems correct. Here is the Magma code to compute isogenies using Vélu's formulas: F3 := DivisionPolynomial(E,3); F := Factorization(F3)[1][1]; E3, phi := IsogenyFromKernel(E,F); $\endgroup$ – François Brunault Mar 3 at 8:57
9
+100
$\begingroup$

In this case, it pays to work on the curve $E'$ that is 2-isogenous to $E$, which is given by the equation $$ y^2 = x^3 + 404100192598226941365253x^2+ 1470175712258164849983363482095324897635296971x. $$

It is relatively easy to find the 7 independent points

(110776963853866550724016 : -80505869468630089210131377497504980 : 1),
(32748343004768539696321 : 22729843039610504103217516288626615 : 1),
(33849219894746702769856 : -23485175376256902691040244680173080 : 1),
(83632124830832490757371607899/221801449 : 2583856339678270035941408483841167567865756495/3303288979957 : 1),
(708467443931328192488072059/9 : 18905656692867793914636996097237707997855/27 : 1),
(82669556642133426513025 : -58721781791249872164660769545564855 : 1),
(15653787556726119039025946369377024/458311398169 : 7352032091260403572244890413120572309377214021131920/310270858512236803 : 1)

on $E'$.

To find the final generator for $E'$, we need to apply 4-descent. We can find the final point by applying 4-descent to any number of 2-covers of $E'$. I got a hit on the second one I tried, which was given by $$ C: y^2 = 40629834885797531781124x^4 + 168889375516514838017136x^3 - 522073637939703266788148x^2 - 1468041555233000587306308x + 3129498870121696083297961. $$

One 4-covering of $C$ is given by the intersection of the quadrics $$ 26478x^2 + 149391xy + 147873xz + 592534xw - 114021y^2 - 58434yz + 336829yw + 118629z^2 + 510074zw + 438488w^2 $$ and $$ 148610x^2 + 19042xy + 1022361xz - 1631065xw - 112833y^2 + 39500yz + 44513yw - 182441z^2 + 822710zw + 972880w^2 $$ in $\mathbb{P}^3$. This has a rational point at (9608:-18440:7168:6485).

The corresponding rational point $Q$ on $E'$ is given by

(-1117913472469704108682566452343804675555656978809017451310789/5063218268474760928272238797769806961 : 1057252852298703272836151916049253048501046484070626117498677044566639368499019825133763515/11393049242085218620499121263510902732524161971013459959 : 1),

which has canonical height $106.35$. Using the other generators, we can find the somewhat smaller point $Q'$

(-101978171213582065261997068234175010796855613305889831/9689020674223383052192046420224 : 159945556170312399327355677945959801245147861170127803803586963373362965000839965/30159200458008252634403235853082432918163525632 : 1)

having canonical height $91.18$ that is independent of the first seven generators.

The discriminant of $E'$ is somewhat smaller than that of $E$, so arguably $E'$ is the model that you'd want to present. However, if you want to work on $E$, then we can map the generators of $E'$ to $E$. A "nice" choice of points that generator $E(\mathbb{Q})$ modulo torsion is:

(225814423074482435222996 : -34626227198383025293628840137707870 : 1),
(-151681843950496144133344 : 94669696550541536347147233612637650 : 1),
(-209328450874403020690564 : 82198277984719522705484038947202890 : 1),
(30089244387022819458738989/4 : -164963185085547559044466862956833045285/8 : 1),
(1521645572343712794396956 : -1853309164504557572407788096216777750 : 1),
(-108427296385410370841188 : 92566828570027620341222482486229910 : 1),
(50259473549510044315364816963047/529 : -356309090633909063646983926957844863796356762164/12167 : 1),
(-825240948184709245504424305468714422091334601593572177730440863632644222609685944791784/3476235906470724630424489414741262801265768781870915852371028441 : 13350668242169750290895245364589704907786881444602849078579635591912831036781435950810619013243973916848136198892088368285547068910/204957522013550460549477456089656119391219129674638303351158680324230445521351939753256298608739 : 1).

The final generator has canonical height $169.76$, which means that it would be more difficult to find by searching directly on 4-coverings of $E$.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.