Elkies found a solution
giving a counter examples to Euler’s conjecture that there are no solutions in positive integers to , the generalizations of Fermat’s equations to multiple variables .
In the case , a counterexample
was found by computer search, but was much harder, and only after Elkies found the above solution, Frye found the following solution by a computer search.
Finding integer solutions to this equation is equivalent to finding rational solutions to
by the map
We now concentrate on finding rational solutions to this equation. The main idea now is to relate to a slightly different surface, use it parametrizations to get parametrization for the surface . The parametrization reduce the problem to finding solutions on some elliptic curves.
The Surface can seen as the intersection of the surface with . So we need to find solutions to where is a square.
The Surface :
The surface can be parametrized a pencil of conics (that is for each fixed ), the solution is a conic -degree two in two variables)
We have .
Also changes to .
There are many ways to arrive at this parametrization : Bremner wrote as
and factored the expressions by arithmetic in Gaussian integers. (Zagier, Elkies also found them using different methods)
A rational solution on the surface corresponds to a rational , and from the above we can assume
for relatively prime odd integers and
So we have
Example:
and hence give
The solution gives .
If you use this one point and draw lines on the plane to find the other intersection points on the conic, we can find the parametrization for given by
which gives
The Surface :
Now imposing that is a square in the previous parametrization, we get the following parametrization for the surface
,
Example:
helps to get the parametrization
and hence
With
we get
Coordinate changes gives two Weierstrass equations
These elliptic curves have only finite many solutions and correspond to the trivial solutions
It can be checked (using quadratic reciprocity and properties of ternary quadratic forms) that the second and third equations for parametrization will have infinitely many solutions if and only if the squarefree parts of
, and
have all prime factors .
The smallest which satisfy these constraints are
, and .
gave us trivial solutions. doesn’t give solutions (simple local obstructions)
gives
The solution allows us to get the parametrization
Looking at , we see that only the plus sign is valid.
Now the change of variables,
gives the elliptic curve
Computer search gave the following solution:
which corresponds to
that is
More solutions: Using the solutions and on the elliptic curve we can generate more solutions by group law on the curve. For instance gives the following solution
for
Reference: https://www.jstor.org/stable/2008781