6

Let $a^4+b^4+c^4=2d^2$ such that $a,b,c,d$ are all nonzero Integers & $a+b+c \ne0$.

Already I know, Jacobi - Maiden Identity and it is Generalized by Ramanujan :

$$a^4 + b^4 + c^4 = 2(ab+ac+bc)^2$$ Where $a+b+c=0$

I Find four Solutions by Trial & Error $$(a,b,c,d)=(1,3,10,71),(1,9,22,347),(1,12,41,1193),(2,55,111,8971)$$

knowing one solution I will generate Infinitely many solutions by using Elliptic curves :

Consider $a^4+b^4+c^4=2d^2$ ---(1)

Wkt $1^4+3^4+10^4=2×71^2$

Let $a=x+3$, $b=x+10$, $c=1$ and $d=y$

$(x+3)^4+(x+10)^4+1^4=2(x^4+26x^3+327x^2+2054x+71^2)$

$x^4+26x^3+327x^2+2054x+5041=y^2$

We need to find Rational solutions for above Equation.

Since it has non trivial rational point so Above Quartic Equation is birationally equivalent to an Elliptic curve $E$ below

$E$ : $V^2=U^3-194643U-71422$

$$x=\frac{V-39U-17199}{6U+2646}$$

$$y=\frac{18x^2+234x-U+981}{18}$$

Rank of an Elliptic curve $E$ is 2

Generators:$(-297,5616)$,$(-9,1296)$

Since Rank of $E$ is 2 so (1) has Infinitely many Integer Solutions with $a+b+c \ne0$

I find Some Solutions by using Group structure of an Elliptic curve $E$ :

$$(a,b,c,d) = (3,7,14,143), (9,14,89,5603), (13,105,196,28261),(10,173,243,46811), (923,1596,4865,16843381)$$

Is there any parametrization that gives Infinitely many solutions ?

Is there any Complete Solution ?

  • 2
    @Angelo $2(ab)^2+2(ac)^2+2(bc)^2-a^4-b^4-c^4=(a+b+c)(-a+b+c)(a-b+c)(a+b-c)$ if we take $a+b+c=0$ or $a+b-c=0$ that gives same Jacobi Maiden Identity because in Lhs terms has an Even power So it works – Guruprasad Aug 15 '24 at 12:06
  • Guruprasad, since you wrote that $a,b,c,d$ are all positive, Jacobi-Maiden identity just solves the case $c=a+b$, not the case $a+b+c=0$ which is impossible if $a,b,c$ are positive or is equivalent to $a=b=c=0 $ if $a,b,c$ are non-negative. – Angelo Aug 15 '24 at 15:01
  • @Angelo, Thank You for finding a mistake in Statement. Now i edited the statement of the problem – Guruprasad Aug 15 '24 at 15:19
  • Guruprasad, why did you write that $a+b+c$ is non Zero at the beginning of your post and then you wrote that Jacobi - Maiden Identity solves the case $a+b+c=0$ ? You should be more coherent with yourself. – Angelo Aug 15 '24 at 16:23
  • @Angelo, There are known parameterization solutions for the case $a+b+c= 0$ where a,b,c, are Integers, so i restrict a problem with that condition to get an other parametrization Solutions. – Guruprasad Aug 15 '24 at 16:59
  • 1
    There are 301 co-prime solutions without a+b=c, with $a\lt b\lt c$ and $c<300$ including (3,7,14,143) and (13,15,16,269). If half the sum is around $c^4$, the chance it is square is around $1/2c^2$, and there are about $c^2/2$ possible pairs $a,b$. So there may well be a regular supply, even without the known $a+b=c$ ones. – Empy2 Aug 21 '24 at 13:52
  • @Empy2 thank you, is it possible to find any parameterization that gives infinitely many primitive solutions ? – Guruprasad Aug 21 '24 at 13:59
  • 2
    $(6n^2-2n-1)^4+(6n^2+2n-1)^4+(12n^2+2)^4=2*(108n^4+32n^2+3)^2$. There may be another with $(1,12,41),(19,31,96),(49,62,175)$ and $(91,105,278)$. – Empy2 Aug 21 '24 at 14:25
  • @Empy2 , how to find these kind of parameterization , Is there any reference to study these parameterization – Guruprasad Aug 21 '24 at 14:35
  • 1
    I ran a program to find all the solutions with $a\lt b\lt c\le 300$. I removed those with common factors. Then I looked at all sets of four solutions. Where each of a, b and c fit separate quadratics, the program told me. There were two in the list. I worked out the first one, and you can work out the other. – Empy2 Aug 21 '24 at 14:47
  • Mathematically I know how to generate more solutions but I don't know how to find a parametrization. – Guruprasad Dec 31 '24 at 09:25
  • 1
    @Empy2 I found generalizations to the first identity you gave above (as well as for the second). Kindly see answer below. – Tito Piezas III Jan 01 '25 at 08:08

3 Answers3

4

It can be shown there are infinitely many polynomial identities to $x^4+y^4+z^4=2t^2$ such that $\color{red}{x+y+z\neq0}$ and $\text{gcd}(x,y,z)=1.$


I. Family 1

$$2^4(an^2+b)^4 + (an^2 - 2mn - b)^4 + (an^2 + 2mn - b)^4 = 2(3a^2n^4 + 32m^2n^2+3b^2)^2$$

where $b=6m^2/\color{red}a\,$ for any $(m,a)$. For example, let $m=1$, thus $b=6/a$, but we choose $a$ such that $b$ is also an integer.

Let $a=6$, $b=1$, so,

$$2^4(6n^2+1)^4+(6n^2−2n−1)^4+(6n^2+2n−1)^4=2(108n^4+32n^2+3)^2$$

Let $a=3$, $b=2$, so,

$$2^4(3n^2+2)^4+(3n^2−2n−2)^4+(3n^2+2n−2)^4=2(27n^4+32n^2+12)^2$$

where the first is Empy2's identity in the comments, and so on for any $(m,a)$.


II. Family 2

$$(an^2-b)^4 + ((an^2 + b)m + cn)^4 + ((an^2 + b)(m+1) + dn)^4 = 2t^2$$

where,

\begin{align} b &= (2m^2 + 2m + 1)(m^2 + m + 1)(m + 1)m/\color{red}a\\[4pt] c &= 2m^4 + 2m^3 - 2m - 1\\[4pt] d &= 2m^4 + 6m^3 + 6m^2 + 4m + 1 \end{align}

for any $(m,a)$. As usual, we choose $a$ such that $b$ is also an integer. For example, let $m=1$, then $b=30/a$, and we have many integer choices for $a$.

Let $a=6$, $b=5$, so,

$$(6n^2-5)^4+(6n^2+5+n)^4+(2(6n^2+5)+19n)^4 =\\ 2(108n^4 + 306n^3 + 449n^2 + 255n + 75)^2$$

Let $a=3$, $b=10$, so,

$$\;(3n^2-10)^4+(3n^2+10+n)^4+(2(3n^2+10)+19n)^4 =\\ 2(27n^4 + 153n^3 + 449n^2 + 510n + 300)^2$$

and so on for any $(m,a)$.

Update: Empy2 suggested a more symmetrical version,

$$((a n^2 + b)(x + y))^4 + ((a n^2 - b)x + c n)^4 + ((a n^2 - b)y - d n)^4 = 2z^2$$

where,

\begin{align} b &= (x^2 + y^2)(x^2 + x y + y^2)x y/\color{red}a\\[4pt] c &= 2x^3y - (x + y)^3(x - y)\\[4pt] d &= 2xy^3 + (x + y)^3(x - y) \end{align}


III. Derivation

To find such identities, what you do is assume three quadratic polynomials $P_k(x)$ and one quartic $Q_1(x)$, then equate,

$$P_1(x)^4+P_2(x)^4+P_3(x)^4 = 2Q_1(x)^2$$

and collect powers of $x$. Set all coefficients equal to zero, and you have a system of $m$ equations in $n$ unknowns, which is near-impossible to solve if you don't have Mathematica, Maple, or other computer algebra system.

But if one has an existing example like Empy2's identity above where there are obvious patterns in the coefficients, then that helps in reducing the number of unknowns and/or equations.

  • 1
    I almost forgot. The $P(n)$ identities above generally obey $x+y+z\neq 0$ for infinitely many $n$, as long as $n\neq 0$. – Tito Piezas III Jan 01 '25 at 08:18
  • For family -1 you defined $t$ but for family -2 you dont defined $t$. Thank you for your Answer. – Guruprasad Jan 02 '25 at 07:21
  • @Guruprasad It was a longer polynomial and would clutter up the post. But one can find it since the LHS is given. – Tito Piezas III Jan 02 '25 at 08:22
  • 1
    If you replace $(1,-m,m+1)$ in your Family 2, with $(x+y,x,y)$, you get $b=(x^2+y^2)(x^2+xy+y^2)xy/a, c=2x^3y+(x+y)^3(y-x),d=-2xy^3+(x+y)^3(y-x),[(x+y)(an^2+b)]^4+[x(an^2-b)+cn]^4+[y(an^2-b)+dn]^4=2z^2$ – Empy2 Jan 03 '25 at 11:09
  • @Empy2 Great! I updated the post with your input. I noticed something curious though. My variable $b$ is a deg-6 polynomial in $m$, but your $b$ is only a deg-5 in $x$, though both versions have $(c,d)$ as deg-4. If $x = y$ and after scaling of variables, it seems it simplifies into Family 1. – Tito Piezas III Jan 04 '25 at 05:56
  • Thanks for that. Both $x$ and $y$ contain $m$, so $b$ translates into a degree-six in $m$. – Empy2 Jan 04 '25 at 06:35
  • 1
    FWIW and to have it written here, $z$ in the RHS is $(x^2+xy+y^2)(an^2-b)^2+(x^4+y^4)(x-y)(an^2-b)+x^8+2x^7y+8x^6y^2+14x^5y^3+18x^4y^4+14x^3y^5+8x^2y^6+2xy^7+y^8$ – Empy2 Jan 04 '25 at 07:53
  • 1
    @Empy2 And the palindromic octic polynomial simplifies as $4x^4y^4\left(4\dfrac{u^5-1}{u-1}+3\right)$ where $u=\dfrac{x^2+y^2}{2xy}$. – Tito Piezas III Jan 04 '25 at 08:20
  • @Empy2 Here's a new post on the similar $x^4+y^4+z^4=3t^2$. Might be of interest. – Tito Piezas III Apr 04 '25 at 07:37
4

The answer by @TitoPiezasIII has a long list of comments, so I start another answer.

Here are some similar quadratic sets of answers, but not based on $x,y,x+y$. They are based on other simple solutions.

$$[14(n^2+6)]^4+[3(n^2-6)+34n]^4+[7(n^2-6)+14n]^4=2z^2\\ [3(n^2-20)]^4+[14(n^2+20)+119n]^4+[7(n^2+20)+49n]^4=2z^2\\ [1(n^2-30)]^4+[3(n^2+30)+31n]^4+[10(n^2+30)+109n]^4=2z^2$$

Edit : The form of these solutions, based on a solution to $x^4+y^4+z^4-2w^2$, is $$[x(n^2-b)]^4+[y(n^2+b)+cn]^4+[z(n^2+b)+dn]^4 = \\ 2[w(n^2+b)^2+p(n^2+b)n+qn^2]^2$$ This is a degree-8 polynomial in $n$. Equate coefficients. The nine equations, which connect $b,c,d,p,q$ can be reduced to a single, degree-six polynomial in the variable $c/d$. If this polynomial has a rational root, we get a solution to the quadratic.
For example, if $x=1,y=3,z=10,w=71$, the degree six polynomial has $18$-digit coefficients. It seems to have two rational roots, $31/109$ and $1199/3799$, and four complex ones. The rational roots can be used to produce the third solution above, and $$[1(n^2-35700)]^4+[3(n^2+35700)+1199n]^4+[10(n^2+35700)+3799n]^4=\\ 2[71(n^2+35700)^2+53963(n^2+35700)n+ 10256249n^2]^2$$

Empy2
  • 52,372
1

$a=55k, b=2k, c=k, d=2139k^2$ ($k\in \mathbb{N}$) can be one example of parametrization such that $a\neq b+c$ or anything else.

이상원
  • 410