The elliptic curve functionality is restricted on general fields, for instance:
sage: R.<t> = QQ[]
sage: F = R.fraction_field()
sage: E = EllipticCurve(F, [1, t + t^3])
sage: E.point( (-t,0) )
(-t : 0 : 1)
sage: P = E.point( (-t,0) )
sage: P.is_finite_order()
---------------------------------------------------------------------------
NotImplementedError Traceback (most recent call last)
<ipython-input-39-27a368cc0c53> in <module>()
----> 1 P.is_finite_order()
::::: many lines
NotImplementedError: Computation of order of a point not implemented over general fields.
I took above a simpler curve with an obvious torsion point.
Even the method is_finite_order
failed to be executed.
Although in the above case we can at least see that 2P is 0, if we ask for it:
sage: 2*P
(0 : 1 : 0)
sage: 2*P == E(0)
True
So if we get problems with the method is_finite_order
, asking for the rank and/or for generators...
We can still use sage (and the own devices) to study the posted curve. After a quick pessimistic search of a polynomial x=x(t) (divided by a power of t maybe), such that x3+t4x+(t3+t) should at least have an even order of t, i've got an optimistic point of view, that there should be no point on the curve
y2=x3+t4x+(t3+t) .
Why? Since if there would be one, say (x,y)=( Px(t)/Qx(t) , Py(t)/Qy(t) ), then using specialization (not everywhere defined) morphisms Q(t)→Q, t→t0 we would get some rational points on the elliptic curves defined over Q,
y2=x3+t40x+(t30+t0) .
Unless there is a division by zero in the one or the other denominator, when specializing t to t0. But even so, the denominators would have been already very complicated, since...
# for t0 in [ -29..29 ]:
for t0 in [ -25, -24, -20, -8, -3, -2, 2, 8, 12, 18, 22, 24, 29 ]:
try:
E = EllipticCurve( QQ, ( t0^4, t0^3 + t0 ) )
ETP = E.torsion_points()
if len( ETP ) > 1:
print "... t0 = %3s | rank = ? | torsion points = %s" % ( t0, ETP )
continue
r = E.rank()
if r > 0:
print "... t0 = %3s | rank = %s | torsion points = %s" % ( t0, r, ETP )
continue
print "*** t0 = %3s | rank = %s | torsion points = %s" % ( t0, r, ETP )
except Exception:
pass
And we get:
*** t0 = -25 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = -24 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = -20 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = -8 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = -3 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = -2 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = 2 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = 8 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = 12 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = 18 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = 22 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = 24 | rank = 0 | torsion points = [(0 : 1 : 0)]
*** t0 = 29 | rank = 0 | torsion points = [(0 : 1 : 0)]
So an "unexpected" Q(t)-rational point on E is comming - in view of our first (naive) sieve - with denominators containing the factor
Q=(t+25)(t+24)(t+20)(t+8)(t+3)(t+2)(t−2)(t−8)(t−12)(t−18)(t−22)(t−24)(t−29) .
(In fact Q2 and Q3 in x and y.) This is not much, from an algorithmic point of view, but a good start.
For a family with one generating point A(x(t),y(t)), we can still use the specialization to "guess" A.