If I have an Elliptic curve $E: y^2=x^3-15x+22$ over $\mathbb{Q}$ with CM from the imaginary quadratic field $\mathbb{Q}(\sqrt{-3})$ then how do I compute the $24$-torsion subgroup $E[24]$ over $\overline{\mathbb{Q}}$?
I know how to compute the $2,3$ and $4$-torsion subgroups using the following algorithm: if $P=(x,y)$ is a point on $E$ then if $2P=\mathcal{O}$ we have $y=0$. And, if $3P=\mathcal{O}$ then $x(2P)=x$ further if instead $4P=\mathcal{O}$ then $y(2P)=0$.
So, what should I do next? Compute $6$-torsion points by setting $y(3P)=0$? And, then in a similar way the $12$ and $24$- torsion points?
Also, is there any software that allow us to compute $n$-torsion groups for large $n$?
Computing in this case means really computing. To see which is the complexity of the needed calculus, i am constrained to use some software helping me to get the answer quickly. When the answer stays here explicitly we can still argue if "easy / handy computations" may suffice.
In sage i am intializing the curve, asking for the $24$-division polynomial $f$ associated to it. In case of a nice information so far - we may still ask for the field where its roots live in, then make a base change and compute the generators (instead of computing all $24^2$ points). Well, instead of plotting the polynomial $f$ - which will soon have degree $289$, let us print the factors of $f$, or at least some information on them when they make the line explode to the right.
We get so far:
And you really do not want to see (here) the last polynomial of degree $96$. One can compute of course closer information, so for the last question regarding a possible soft aimed to help, well sage is a free choice.
Here is as supplementary information what we can immediately compute, and may help for the asked purpose.
We can ask for the torsion points over $\Bbb Q$.
We have some polynomials of degree two listed as factors, discriminant is sometimes $3$ times a square, so it may be interesting to switch to the field $F=\Bbb Q(\sqrt3)=\Bbb Q(a)$, $a=\sqrt 3$. Then the given curve, seen over $F$ has...
And the prints give the information:
Here, instead of getting $E[24]$, let us compute starting from $F$ the pieces $E[3]\subset E(\overline{\Bbb Q})$ and $E[2^k]\subset E(\overline{\Bbb Q})$, $k=1,2,3$ as far as possible.
and for the $8$-division let us better take each factor one by one...
Results were manually rearranged. Now let us exhibit the $3$-torsion. This is "hard enough" (to type). We ask for the corresponding division polynomial, which is of degree $3$, use $b$ for a "fixed symbolic root", so we expect to get a torsion point $T=(b,?)$, but the question mark involves taking the square root from $b^3 -15b+22$, denote it by $c$, so we have to extend $L=\Bbb Q(b)$ to $K=\Bbb Q(c)$. Over $K$ we have then at least a subgroup with $3^2$ elements with torsion order dividing $3$. Therin $3^2-1=8$ elements have order exactly $3$. Let us plot them in a final step... Here is my dialog with the sage interpreter:
In the last loop with lots of prints we see the two expected torsion points $(b,\pm c)$.