If $x^4+y^4+z^4=mt^2$ has one solution $xyzt\neq0$ then, using an elliptic curve, one can generally find infinitely many primitive solutions. We focus on $m=1,2,3$.
I. m = 1
Fauquembergue found that given the Pythagorean triple $a^2+b^2 = c^2$, then,
$$(ab)^4+(ac)^4+(bc)^4 = (a^4+a^2b^2+b^4)^2$$
II. m = 2
Proth (and others) found that if $a+b+c = 0$, then,
$$a^4+b^4+c^4 = 2(ab+ac+bc)^2$$
Ramanujan extended this to,
$$a^4(b-c)^4 + b^4(a-c)^4 + c^4(a-b)^4 = 2(ab+ac+bc)^\color{red}4$$
$$(a^2b+b^2c+c^2a)^4 + (ab^2+bc^2+ca^2)^4 + (3abc)^4 = 2(ab+ac+bc)^\color{red}6$$
and all $(ab+ac+bc)^\color{red}{2k}$.
Note: Infinitely many polynomial identities where $a+b+c\neq0$ are here and here.
III. m = 3
S. Realis found,
$$(5n^4 + 4n^3 + 9n^2 + 10n + 5)^4 + (5n^4 + 10n^3 + 9n^2 + 4n + 5)^4 + (5n^4 + 16n^3 + 27n^2 + 16n + 5)^4=3t^2$$
where $t$ is a palindromic octic polynomial. This has the general form,
$$(n^4 + an^3 + bn^2 + cn + 1)^4 + (n^4 + cn^3 + bn^2 + an + 1)^4 + (n^4 + dn^3 + en^2 + dn + 1)^4=3t^2$$
for five rational numbers $(a,b,c,d,e) = (4/5,\,9/5,\,10/5,\,16/5,\,27/5)$ though how Realis derived this I have no idea.
IV. Computer search
The fact that Realis found a quartic polynomial piqued my interest. So I did a search for $x^4+y^4+z^4=3t^2$ with $x\leq y\leq z \leq B\,$ below a bound $B$ and $\text{GCD}(x,y,z)=1$,
$$x,\,y,\,z\\ 1,\, 1,\, 1\\ 11,\, \color{red}{11},\, \color{green}{23}\\ \color{red}{11},\, 13,\, \color{red}{47}\\ \;\color{green}{23},\, \color{green}{85},\, 157\\ 37,\, \color{red}{47},\, \color{green}{85}\\ 41,\, \color{blue}{89},\, \color{blue}{115}\\ 55,\, 103,\, 149 \\ \color{blue}{89},\, 113,\, 259 \\ 91,\, \color{blue}{115},\, 293$$
Those repeating numbers seem a bit suspicious. We have the "cyclic",
$$\color{blue}{11}^4+\color{blue}{23}^4+11^4=3t_1^2\\ \color{blue}{23}^4+\color{blue}{85}^4+157^4=3t_2^2\\ \color{blue}{85}^4+\color{blue}{47}^4+37^4=3t_3^2\\ \color{blue}{47}^4+\color{blue}{11}^4+13^4=3t_4^2$$
which is reminiscent of the cubic,
$$ (\color{blue}{-9})^3 + \color{blue}{16}^3 + 33^3 = {34}^3\\ \color{blue}{16}^3 + \color{blue}{51}^3 + 213^3 = 214^3\\ \color{blue}{51}^3 + \color{blue}{82}^3 + 477^3 = {478}^3\\ \color{blue}{82}^3 + \color{blue}{57}^3 + 495^3 = 496^3\\ \color{blue}{57}^3 + \color{blue}{22}^3 + 255^3 = {256}^3\\ \color{blue}{22}^3 + (-9)^3 + 57^3 = 58^3 $$
though there is an infinite family responsible for the cubic version discussed in this post.
V. Question
If we brute-search $x^4+y^4+z^4=3t^2$ with the same conditions above but for a higher bound, will there be more repeating numbers?
(1,1),(23,11),(2375,6227)– Dmitry Ezhov Apr 04 '25 at 14:58