3
$\begingroup$

It is well-known that the product of two squares is a square, and the product of sums of two squares is a sum of two squares (Brahmagupta-Fibonacci identity): $$(a^2+b^2)(c^2+d^2)=(ac-bd)^2+(ad+bc)^2$$ and an analogous result holds for sums of four squares (Euler's four square identity), eight squares (Degen's eight square identity) and 16 squares (Pfister's 16 square identity).

According to the Wikipedia article, analogous results hold for all sums of $2^n$ squares. But what about sums of $2n$ squares? Is it known whether there is a formula for the product of sums of 6 squares as a sum of 6 squares, i.e. whether, given $a_i,b_j\in\mathbb{Z}$, $$\left(\sum_{i=1}^6a_i^2\right)\left(\sum_{j=1}^6b_j^2\right)=\sum_{k=1}^6c_k^2$$ where $c_k=f_k(a_1,\ldots,a_6,b_1,\ldots,b_6)$ for some polynomials $f_k$?

$\endgroup$
6
  • $\begingroup$ This is trivially true, since every non-negative integer is the sum of six squares. But there is no such formula as for powers of two. $\endgroup$ Aug 3, 2021 at 11:51
  • $\begingroup$ @DietrichBurde Thank you, so doesn't this mean it's trivial for every sum of $\ge 4$ squares? What is the significance of the 8 and 16 square identities? $\endgroup$
    – a1402
    Aug 3, 2021 at 11:54
  • 1
    $\begingroup$ There are no nice formulas, but do such formulas exist? $\endgroup$
    – a1402
    Aug 3, 2021 at 11:54
  • $\begingroup$ No, I don't think so. Wikipedia certainly would have mentioned those, right? K. Conrad's article is cited for $m=16$, there should be an argument given there, why we should have $m=2^n$. $\endgroup$ Aug 3, 2021 at 11:56
  • 1
    $\begingroup$ I found it in Conrad's article. It is called "A Converse to Pfister's Theorem." $\endgroup$ Aug 3, 2021 at 12:02

1 Answer 1

2
$\begingroup$

There is no such formula.

Pfister’s theorem tells us that in any field, when $n$ is a power of 2, the set of sums of $n$ squares is closed under multiplication. It follows that when $n$ is a power of 2, there is a formula for the $c_k$ in terms of $a_i,b_i$, where $(a_1^2+\ldots+a_n^2)(b_1^2+\ldots+b_n^2)=c_1^2+\ldots+c_n^2$ .

Another theorem of Pfister (converse of Pfister’s above theorem) tells us that for every positive integer $n$ that is not a power of 2, the nonzero sums of $n$ squares are not closed under multiplication in the field $\mathbb{R}(x_1,x_2,\ldots)$. This means that there are two sums of $n$ squares in this field such that their product is not a sum of $n$ squares in the field. Hence no formula exists when $n$ is not a power of 2.

Source: https://kconrad.math.uconn.edu/blurbs/linmultialg/pfister.pdf

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .