Re: [HM] 25 = 27 - 2

Julio Gonzalez Cabillon (jgc@adinet.com.uy)
Sat, 09 Jan 1999 18:09:05 -0200

On Tue, 5 Jan 1999, Jim Propp <propp@math.mit.edu> wrote:
| ...
| According to Dickson (volume II, pages 533-534), Euler, in his
| treatment of Fermat's assertion (Algebra, II, chapter 12,
| Articles 187-195; or, Opera Omnia, (1), I, 429-434), made axx+byy
| a cube by assuming that x sqrt(a) + y sqrt(-b) is of the form
| (p sqrt(a) + q sqrt(-b))^3 for some p, q. But Dickson goes on
| to say that in the same chapter of his Algebra (Articles 195-196)
| Euler himself demonstrates that the assumption is invalid in general!
| (Specifically, Euler shows that it fails for a=2, x=4, b=-5, y=1.)
|
| So, what did Euler believe he'd done? And what did his colleagues
| and successors think he'd done?
|
| Judging from what I've written above, you might suppose that it would
| have been common knowledge that Euler's proof was incomplete. However,
| Dickson tells us that in 1875 "T. Pepin criticized Euler's proofs,
| noting that there may exist sets of formulas for x and y other than
| the set deduced by Euler's assumption." (Dickson does not say whether
| Pepin acknowledged that Euler himself had given a counterexample.)
|
| So, how did Euler's proof come to be judged definitive when he himself
| apparently did not deem it so?
|
| Perhaps the passage in Euler's Algebra will help resolve the question.
| Have any of you read it?
| ...

Dear Jim,

Euler, in his *Article 188* ("Algebra", II, chapter 12; or, Opera Omnia,
(1), I, 429-430), wonders under which conditions $ axx + cyy $ can be a
cube.

A *crude* translation (and intepretation!) follows:

Article 188:
In order to make the formula axx + cyy a cube, we may set things in a
similar fashion [ to that of previous *Articles* ]. Let us assume that

x R(a) + y R(-c) = ( p R(a) + q R(-c) )^3
and
x R(a) - y R(-c) = ( p R(a) - q R(-c) )^3

[ R(x) denotes the square root of $x$ ]

So, the product becomes

axx + cyy = (app + cqq)^3,

and thus our formula (axx + cyy) is a cube. This depends, however,
on whether $x$ and $y$ can be determined as rationals -- and this may
succeed somehow in a 'lucky way'.

Therefore:

xR(a) + y R(-c) = ap^3 R(a) + 3appq R(-c) - 3cpqq R(a) - cq^3 R(-c)

and

xR(a) - y R(-c) = ap^3 R(a) - 3appq R(-c) - 3cpqq R(a) + cq^3 R(-c)

from which it evidently follows that

x = ap^3 - 3cpqq and y = 3appq - cq^3 .

We look for two squares, xx and yy, so that xx + yy is a cube. For
instance, let be a = 1 and c = 1. Therefore:

x = p^3 - 3pqq and y = 3ppq - q^3,

and then
xx + yy = (pp + qq)^3.

We have then p = 2 and q = 1, and so x = 2 and y = 11; whence
xx + yy = 125 = 5^3.

============

[Gauss's original follows:]

188

Wann demnach diese Formel axx + cyy zu einem Cubo gemacht
werden soll, so setze man auf eine aehnliche. Weise als vorher

xR(a) + yR(-c) = ( pR(a) + qR(-c) )^3
und
xR(a) - yR(-c) = ( pR(a) - qR(-c) )^3

dann daraus wird das Product axx + cyy = (app + cqq)^3, und
also unsere Formel ein Cubus: es kommt aber nur darauf an, ob
auch hier $x$ und $y$ auf eine rationale Art bestimmt werden
koenne? welches gluecklicher weise gelingt; dann wann die
angesetzte Cubi wuercklich genommen werden, so erhalten wir
diese zwei Gleichungen

xR(a)+yR(-c) = ap^3R(a)+3appq R(-c)-3cpqqR(a)-cq^3R(-c)
und
xR(a)-yR(-c) = ap^3R(a)-3appq R(-c)-3cpqqR(a)+cq^3R(-c)

woraus offenbahr folgt, dass

x = ap^3 - 3 cpqq und y = 3appq - cq^3 .

Man suche z. E. zwey Quadrate xx und yy, deren Summ xx + yy
einen Cubus ausmache: weil nun hier a = 1 und c = 1, so bekommen
wir
x = p^3 - 3pqq und y = 3ppq - q^2,

und alsdann wird xx + yy = (pp + qq)^3. Es sey nun p=2 und
q=1, so wird x=2 und y=11; hieraus xx + yy = 125 = 5^3.

============

195

[Here Euler looks for integral squares (x^2) such that 2xx - 5 must
be a cube.]

For these cases we may wonder in general when 2xx - 5yy becomes a
cube. But, according to *Article 188* [see above], this case takes
place when

x = 2p^3 + 15pqq and y = 6ppq + 5q^3 [a = 2, c = -5]

However, $y$ must be 1 (or -1) [see Art. 188], and therefore

6ppq + 5q^3 = q(6pp + 5qq) = 1 (or -1)

These equations have neither integral solutions nor rationals. Hence
this example is very peculiar, since a solution can still be found
for x=4, so that 2xx - 5 = 27 is the cube of 3. The motive must be
examined because of its great importance.

[Gauss's original follows:]

195

IV. Frage: Man suche solche Quadrate in gantzen Zahlen,
welche doppelt genommen wann davon 5 subtrahirt wird,
dass ein Cubus heraus komme; oder 2xx - 5 soll ein
Cubus seyn.

Man suche erstlich diejenigen Faelle da 2xx - 5yy ein
Cubus wird, welches nach dem 188ten Articul, wo a=2 und
c=-5, geschieht, wann x=2p^3 + 15pqq und y=6ppq + 5q^3.
Hier aber muss seyn y=1 (oder -1), und folglich

6ppq + 5q^3 = q(6pp + 5qq) = 1 (oder -1),

welches in gantzen Zahlen nicht geschehen kann, und auch
nicht einmahl in Bruechen; dahero dieser Fall sehr
merckwuerdig ist, da gleichwohl eine Aufloesung statt
findet, wann nemlich x=4, dann da wird 2xx - 5 = 27,
welches der Cubus ist von 3; und hievon ist es von der
groessten Wichtigkeit den Grund zu untersuchen.

===========

Any comments? ...

With best regards,
Julio Gonzalez Cabillon