Leonhard Euler · 1770 (1828 English ed.) · Internet Archive
Elements of Algebra
All 627 leaves · leaf 467
Transcription
430 ELEMENTS PART II.
*Demonstration*. We have seen that $p$ is odd, and $q$ even, because $p^2 + q^2$ cannot be a square, except when $q = 2rs$, and $p = r^2 - s^2$; and $p^2 + 3q^2$ cannot be a square, except when $q = 2tu$, and $p = t^2 - 3u^2$, or $p = 3u^2 - t^2$.
Now, as in both cases $q$ must be a double product, let us suppose for both, $q = 2abcd$; and, for the first formula, let us make $r = ab$, and $s = cd$; for the second, let $t = ac$, and $u = bd$. We shall have for the former $p = a^2b^2 - c^2d^2$, and for the latter $p = a^2c^2 - 3b^2d^2$, or $p = 3b^2d^2 - a^2c^2$, and these two values must be equal; so that we have either $a^2b^2 - c^2d^2 = a^2c^2 - 3b^2d^2$, or $a^2b^2 - c^2d^2 = 3b^2d^2 - a^2c^2$; and it will be perceived that the numbers $a, b, c$, and $d$, are each less than $p$ and $q$. We must, however, consider each case separately: the first gives $a^2b^2 + 3b^2d^2 = c^2d^2 + a^2c^2$, or $b^2(a^2 + 3d^2) = c^2(a^2 + d^2)$, whence $\frac{b^2}{c^2} = \frac{a^2+d^2}{a^2+3d^2}$ a fraction that must be a square.
Now, the numerator and denominator can here have no other common divisor than 2, because their difference is $2d^2$. If, therefore, 2 were a common divisor, both $\frac{a^2+d^2}{2}$, and $\frac{a^2+3d^2}{2}$, must be a square; but the numbers $a$ and $d$ are in this case both odd, so that their squares have the form $8n+1$, and the formula $\frac{a^2+3d^2}{2}$ is contained in the expression $4n+2$, and cannot be a square; wherefore 2 cannot be a common divisor; the numerator $a^2+d^2$, and the denominator $a^2+3d^2$ are therefore prime to each other, and each of them must of itself be a square.
But these formulae are similar to the former, and if the last were squares, similar formulae, though composed of the smallest numbers, would have also been squares; so that we conclude, reciprocally, from our not having found squares in small numbers, that there are none in great.
This conclusion, however, is not admissible, unless the second case, $a^2b^2 - c^2d^2 = 3b^2d^2 - a^2c^2$, furnishes a similar one. Now, this equation gives $a^2b^2 + a^2c^2 = 3b^2d^2 + c^2d^2$, or $a^2(b^2+c^2) = d^2(3b^2+c^2)$; and, consequently, $\frac{a^2}{d^2} = \frac{b^2+c^2}{3b^2+c^2} = \frac{c^2+b^2}{c^2+3b^2}$; so that as this fraction ought to be a square, the foregoing conclusion is fully confirmed; for, if in great numbers there were cases in which $p^2+q^2$, and $p^2+3q^2$, were squares, such cases must have also existed with regard to smaller numbers; but this is not the fact.
This text was produced by machine and has not been corrected by hand. Where it matters, read the scan. Confidence is the model's own, not an external check.