# Looking for all sequences such that $a_i^2+a_j^2=a_k^2+a_l^2$ whenever $i^2+j^2=k^2+l^2$

I'm working in a difficult functional equation, and I have reduced the problem to the following question ($\mathbb{N}$ denotes the set of non negative integers $0,1,2,3,4\cdots$)

Question:

Can we find all sequences of non negative integers ($\forall i\in \mathbb{N} \,\, a_i\in \mathbb{N}$) such that \begin{align} \forall i,j,k,l\in\mathbb{N} && i^2+j^2=k^2+l^2 &\Rightarrow a_i^2+a_j^2=a_k^2+a_l^2\end{align}

My try: I know the Euler's complete solution to the Diophantine equation :

$$x^2+y^2=z^2+t^2$$ $(ab+cd,ac-bd,ab-cd,ac+bd)$ but this is not very useful.

I also tried another way which is if we suppose that the sequence is injective ($a_i\neq a_j$ if $i\neq j$, maybe later we can prove that either the sequence is constant or this will be true), and under this assumption we can deduce that if $a_i+a_j$ is odd then: $$d(|a_i^2-a_j^2|)\geq |i^2-j^2| \\ r_2(a_i^2+a_j^2)\geq r_2(i^2+j^2)$$

but the problem is we don't know much about the function $d(n)$ the number of divisors of $n$ and $r_2(n)$ the number of ways to write $n$ as the sum of two squares.

Expected solution: The only sequences I found are $a_i=c$ or $a_i=ki$ But is it the only sequences that fits our needs?, can we find a sequence with two values $c$ and $c'$ that fits the assertion?

Any comments, suggestions, any consequences of such assertion are welcome,Thank you for you help.

• Tiny note: you don't need your multiplier in the linear solution to be square - $a_i=ki$ works for any $k$. Also, it would help to clarify whether you're allowing zero values for $i/j/k/l$; you mention 'non-negative' for the $a_*$ but that doesn't say anything about the indices. Other than that, interesting question! – Steven Stadnicki Feb 24 '15 at 0:33
• The question is not entirely clear. Can write more formally? What is the relationship between the numbers in the system ? – individ Feb 24 '15 at 4:16
• @StevenStadnicki, I add the some improvement to the question, and I corrected the the linear solution, for the $indexes$ 0 is allowed, e.g : $a_0^2+a_5^2=a_3^2+a_4^2$, the problem when I write equations like this is: a sigle $a_i$ may verify infinity many equations, but for two elements $a_i$ and $a_j$ the number of such equation that contains $a_i$ and $a_j$ is finite (which allowed_me to conclude the two equations above). – Elaqqad Feb 24 '15 at 11:37
• Can it is necessary to solve the following system of equations? \left\{\begin{aligned}& i^2+j^2=k^2+l^2\\&(a+ib)^2+(a+jb)^2=(a+kb)^2+(a+lb)^2\end{aligned}\right. – individ Feb 24 '15 at 17:16
• This system of equations is equivalent to $i+j=k+l$ so ${i,j}={k,l}$ – Elaqqad Feb 24 '15 at 20:11