Home Powered by MathJax

Classic vs Symmetric Addition Formulae

by

Gregg Kelly

In this section we show how the "classic" addition formulae can be obtained from the symmetric addition formulae for general genus 1 curves of degree 3 and 4 calculated on the previous pages.

In terms of the uniformising elliptic functions $f(z)$ and $g(z)$ of the curve $F(x,y)=0$, the symmetric level 2 addition formulae we have calculated in earlier sections have the form, when $n=3$ \begin{equation} \label{eq:symmetric0} f(z_3) \space = \space X(f(z_1),g(z_1),f(z_2),g(z_2))\qquad\qquad g(z_3) \space = \space Y(f(z_1),g(z_1),f(z_2),g(z_2)) \end{equation} where $z_1 + z_2 + z_3 = s$ is the "centre" point and $X$ and $Y$ are rational functions of their arguments and the coefficients of $F$.

It is convenient to write this addition formula as \begin{equation} \label{eq:symmetric} f(s - z_1 - z_2) \space = \space X(f(z_1),g(z_1),f(z_2),g(z_2))\qquad\qquad g(s - z_1 - z_2) \space = \space Y(f(z_1),g(z_1),f(z_2),g(z_2)) \end{equation}

From \eqref{eq:symmetric} we would like to derive a "classic" addition formulae of the form \begin{equation} \label{eq:classic} f(z_1 + z_2) \space = \space U(f(z_1),g(z_1),f(z_2),g(z_2))\qquad\qquad g(z_1 + z_2) \space = \space V(f(z_1),g(z_1),f(z_2),g(z_2)) \end{equation} where and $U$ and $V$ are rational functions.

If $s=0$ and $f$ and $g$ have even and odd symmetry respectively, as is the case for the Weiertrass curve with $f(z)=\wp(z)$ and $g(z)=\wp'(z)$, then it is trivially simple, using \eqref{eq:symmetric} we have \begin{equation} f(z_1 + z_2) \space = \space X(f(z_1),g(z_1),f(z_2),g(z_2))\qquad\qquad g(z_1 + z_2) \space = \space -Y(f(z_1),g(z_1),f(z_2),g(z_2)) \end{equation} and the classic and symmetric addition formulae are almost identical.

What is not immediately obvious is that we can carry out a similar process even when $s \ne 0$ and $f$ and $g$ have no special symmetries.

Substitute $z_1 \leftarrow s-z_1-z_2$ and $z_2 \leftarrow 0$ in \eqref{eq:symmetric} to get \begin{equation} \label{eq:cfs} \begin{aligned} f(z_1 + z_2) \space = \space X(f(s - z_1 - z_2),g(s - z_1 - z_2),f(0),g(0)) \\\\ g(z_1 + z_2) \space = \space Y(f(s - z_1 - z_2),g(s - z_1 - z_2),f(0),g(0)) \end{aligned} \end{equation} then apply \eqref{eq:symmetric} once more to the RHS of \eqref{eq:cfs} to get \begin{equation} \label{eq:classicfromsymmetric} \begin{aligned} f(z_1+z_2)\space &= \space X(X(f(z_1), g(z_1), f(z_2), g(z_2)), \thinspace Y(f(z_1), g(z_1), f(z_2), g(z_2)), \thinspace f(0), g(0)) \\\\ g(z_1+z_2)\space &= \space Y(X(f(z_1), g(z_1), f(z_2), g(z_2)), \thinspace Y(f(z_1), g(z_1), f(z_2), g(z_2)), \thinspace f(0), g(0)) \end{aligned} \end{equation} That is to say using the symmetric level 2 addition formula we are able to express $f(z_1+z_2)$ and $g(z_1+z_2)$ as rational functions of $f(z_1),g(z_1),f(z_2),g(z_2),f(0),g(0)$ and the coefficients of $F$.

Also observe that putting $z_1=z_2=0$ in \eqref{eq:symmetric} gives us the coordinates of the centre point $s$ in terms of the coordinates of the zero point $f(0),g(0)$ \begin{equation} \label{eq:centre} f(s) \space = \space X(f(0),g(0),f(0),g(0))\qquad\qquad g(s) \space = \space Y(f(0),g(0),f(0),g(0)) \end{equation} then by putting $z_1=z$ and $z_2=s$ in \eqref{eq:symmetric} gives us a "negation" formula \begin{equation} \label{eq:negation} f(-z) \space = \space X(f(z),g(z),f(s),g(s))\qquad\qquad g(-z) \space = \space Y(f(z),g(z),f(s),g(s)) \end{equation} That is to say using the symmetric level 2 addition formula we are able to express $f(-z)$ and $g(-z)$ as rational functions of $f(z),g(z),f(0),g(0)$ and the coefficients of $F$.

If the symmetric addition formula \eqref{eq:symmetric0} has $n$ variables instead of 3, the same results hold, all we need to do is set the surplus $z$ variables to zero.

The above formulae have a simple geometric interpretation. It is most easily described when the interpolating curves associated with the level 1 symmetric addition formula are straight lines as is the case for the Weierstrass curve.

To make things clearer we use the parameterisation $f(z) = \wp(z + z_0)$ and $g(z) = \wp'(z + z_0)$ so that the zero point $O$ corresponding to $z=0$ is an arbitrarily chosen point $(\wp(z_0),\wp'(z_0))$.

Draw a straight line through two points $P_1$ and $P_2$ on the curve and it will intersect the curve at a third point $P_3$. If $P_1$ and $P_2$ correspond to $z_1$ and $z_2$ repectively then $P_3$ will correspond to $z_3 = s - z_1 - z_2$ where $s = 3z_0$.

If we draw a line through $P_3$ and $O$ it will intersect the curve at a fourth point say $P_4$ which corresponds to $z_1 + z_2$. This is the classic addition formula $P_4 = P_1 \oplus P_2$.

The centre point $S$ corresponding to $z = s$ is the point where a tangent drawn at the zero point $O$ intersects the curve. To find the negation of a point $P_1$ draw a line through $P_1$ and the centre point $S$, it will intersect the curve at point $P_2$ which corresponds to $z = -z_1$. This is the negation formula $P_2 = \ominus P_1$.

To illustrate the four variable case, consider the quartic curve parameterised by the elliptic functions $f$ and $f'$ satisfying the differential equation \begin{equation} f'(z)^2 \enspace = \enspace R\left(f(z)\right) \end{equation} with boundary condition $f(0) = x_0$ so that $z=0$ at some arbitrary point $(x_0,y_0)$ on the quartic. Then the classic addition formula arises from two applications of the symmetric addition formula.

In this diagram the $x,y$-coordinates of the points are given by $\left(x_i,y_i\right) = \left(f(z_i),f'(z_i)\right)$. If $R$ has four real roots and negative leading coefficient then it will look something like this:

  1. Run a parabola through the points $z_1,z_2,z_3$ to give a point $z_4$ such that $z_1 + z_2 + z_3 + z_4 \equiv s$ where $s$ is twice the sum of the position of the poles of $f$.
  2. Run a second parabola tangential to the quartic at $0$ and passing through $z_4$ to give a point $z_5$ such that $0 + 0 + z_4 + z_5 \equiv s$.
  3. From this it follows that $z_5 \equiv z_1 + z_2 + z_3$.
Note that under the boundary condition $f'(0) = 0$ we have $s \equiv 0$ and the second parabola degenerates into a pair of vertical lines.

References

[1] Harold M. Edwards A Normal Form For Elliptic Curves Bull. Amer. Math. Soc. 44 (2007), 393-422

[2] F.G. Frobenius and L. Stickelberger Zur Theorie der elliptischen Functionen J. reine angew. Math 83 (1877), 175–179