6

I. Quadruples

In this post, we saw how infinitely many Pythagorean triples such as,

$$3^2+4^2-5^2=0\quad$$

can lead to 4th power equalities,

$$2^4+2^4+4^4+3^4+4^4=5^4\quad$$

It turns out that infinitely many Pythagorean quadruples such as

$$\color{blue}{-65^2 + 52^2 + 15^2 + 36^2} = 0\qquad$$

can now lead to 6th power equalities,

$$\color{blue}{65^6 + 52^6 + 15^6} = \color{blue}{36^6} + (52 + 15)^6 + (52 - 15)^6$$

which is the second smallest $(6,3,3)$ known. Or another example,

$$\color{red}{-10947^2 + 10591^2 + 902^2 + 2618^2} = 0\qquad$$

leading to,

$$\color{red}{10947^6 + 10591^6 + 902^6} = \color{red}{2618^6} + (10591 + 902)^6 + (10591 - 902)^6$$

and so on, though the numbers get large fast since an elliptic curve is involved. As,

$$x_1^6+x_2^6+x_3^6 = y_1^6+y_2^6+y_3^6\qquad$$

they also obey,

$$x_1y_1(x_1^2-y_1^2)+x_2y_2(x_2^2-y_2^2)+x_3y_3(x_3^2-y_3^2)=0$$

hence are actually three right triangles whose areas (signed) sum to zero.


II. Elliptic curve

Consider the system,

\begin{align} a^2+b^2+c^2 &= d^2+(b+c)^2+(b-c)^2\\[5pt] a^6+b^6+c^6 &= d^6+(b+c)^6+(b-c)^6 \end{align}

This was first discussed in a 1974 paper by Brudno and Kaplansky. They used some complicated substitutions and didn't seem to notice that the first equation reduced to the simple quadruple $-a^2+b^2+c^2+d^2=0$. Using that fact, we can more quickly find a solution to the quadratic-sextic system as,

\begin{align} c &= \frac{a\sqrt{a^2-b^2}}{\sqrt{a^2+9b^2}}\\[6pt] d &= \frac{3b\sqrt{a^2-b^2}}{\sqrt{a^2+9b^2}} \end{align}

or to solve,

$$(a^2-b^2)(a^2+9b^2)=y^2$$

If it has a rational point, then a quartic to be made a square is birationally equivalent to an elliptic curve, hence there are infinitely many integer $(a,b)$ though avoiding the trivial $(a^2-b^2)(a^2-9b^2)(4a^2-9b^2)b^2=0$.


III. Question

Is is true that solutions to, $$(a^2-b^2)(a^2+9b^2)=y^2$$ come in unsigned pairs which yield the same $x_1^6+x_2^6+x_3^6 = y_1^6+y_2^6+y_3^6$ and one of the pair has $a+b=z_1^2$ while the other has $a-b=z_2^2$ ?

Example:

1st pair: $(a_1,b_1)=(13,3)$ where $13+3=4^2$ and $(a_1,b_1)=(5,4)$ where $5-4=1^2$.

2nd pair: $(a_2,b_2)=(267,22)$ where $267+22=17^2$ and $(a_2,b_2)=(123, 119)$ where $123-119=2^2$.

P.S. What are the 3rd, 4th, 5th, and do they obey the same?

  • 1
    The first report: $(x1,x2,x3,x4,x5,x6)=(16171009, 8612760, 7260991, 11601720, 15873751, 1351769), (a,b,c,d)=(2929, 1560, 7260991/5521, 11601720/5521).$ Neither $a+b$ nor $a-b$ is a square number. – Tomita Apr 01 '25 at 05:27
  • 1
    @Tomita It is a square number. $a+b = 2929+1560 =67^2$. – Tito Piezas III Apr 01 '25 at 05:39
  • @Tomita I used it to generate its partner $(a,b) = (11042, 4958)$ and $a-b=11042- 4958=78^2$. – Tito Piezas III Apr 01 '25 at 05:55
  • @Tito, I was wrong. – Tomita Apr 01 '25 at 06:23
  • @Tomita That's ok. Just a mistype, I guess. If you can find the first few smallest $(a_n,,b_n)$ so we can check their $a_n\pm b_n$ that may be enough for an answer. (No need for the $x_k$ as they get large fast.) – Tito Piezas III Apr 01 '25 at 06:27
  • 2
    Another paper on equal sums of three sixth powers is Kuwata, Masato. “Equal sums of sixth powers and quadratic line complexes.” The Rocky Mountain Journal of Mathematics 37, no. 2 (2007): 497–517. http://www.jstor.org/stable/44239196. It cites work of Bremner that might also be useful. – Gerry Myerson Apr 01 '25 at 09:39
  • 1
    @GerryMyerson Thanks. That paper mentions the family of the smallest solution, namely $23^6+ 10^6+ 15^6 = 3^6+ 19^6+ 22^6$ where relationships between the terms involve the quadratic form $x^2+xy\color{red}{-}y^2$ whereas $a^4+ b^4+ c^4 = d^4+ e^4+ f^4$ involve the similar $x^2+xy+y^2$ and the Eisenstein integers. I've been trying to research for something similar for $8$th powers but no luck so far. – Tito Piezas III Apr 01 '25 at 09:51
  • 1
    @TitoPiezasIII, does your method apply to non Pythagorian quadruples like $17^2=8^2 + 9^2 +12^2$? – user25406 Apr 01 '25 at 13:15
  • 2
    @user25406 A Pythagorean quadruple is defined as $a^2+b^2+c^2=d^2$ so your example is one of them. However, the method I described only works for special quadruples that is generated by the elliptic curve, but there are infinitely many of them. – Tito Piezas III Apr 01 '25 at 13:22
  • @Tomita Here a Diophantine question on $x^4+y^4+z^4=3t^2$. – Tito Piezas III Apr 04 '25 at 07:24

1 Answers1

4

$$ (a^2-b^2)(a^2+9b^2)= v^2 $$
Let $x=\dfrac{a}{b}, y= \dfrac{v}{b^2}$. We have

$$y^2 = x^4+8x^2-9 \tag{1}$$


I. First elliptic curve

This equation can be transformed to Weierstrass form with $(X,Y)=(x^2,xy)$ $$E_1: Y^2=X^3+8X^2-9X$$ $E_1$ has has rank $1$ and generator is $P(X,Y)=(-4,10).$
We get the solutions using the group structure $kP$ with $k=1\cdots10$ as follows.

Since the solutions are huge, we only show $(a,b,a-b,a+b).$

(2929, 1560, 37^2, 67^2)  
(112982885, 19106164, 9689^2, 11493^2)
(126901625119681, 125074345690320, 1351769^2, 15873751^2)
(23501593203523539455525, 6983371409262733246316, 128523234453^2, 174599440471^2)

Remark: The system of equations $(a^2-b^2=u^2,a^2+9b^2=v^2)$ is reduced to $y^2=10x^4-16x^2+10$ and it is birationally equivalent to $E_3: Y^2=X^3-X^2-30X+72.$ The previous $E_1: Y^2=X^3+8X^2-9X$ is also birationally equivalent to $E_3.$ This is why $E_1$ gives the solutions such that $a^2-b^2$ is square.


II. Second elliptic curve

Since the equation $(1)$ has a solution $(x,y)=(1,0)$, it is birationally equivalent to another elliptic curve

$$E_2:Y^2=X^3 -X^2 +15X + 225$$

yielding $(a,b,a-b,a+b)$

(3, 2, 1, 5)  
(13, 3, 10, 4^2)
(267, 22, 245, 17^2)
(617307, 564262, 53045, 1087^2)
(126690733, 111355677, 3916^2, 238046410)

where $(a_0,b_0) = (3,2)$ is trivial but continues to non-trivial solutions $(a_1,b_1) = (13, 3)$ and $(a_2,b_2) = (267, 22)$ mentioned in the post by Piezas.

Tomita
  • 3,209
  • 1
    Yes, $a+b$ and $a-b$ are squares, – Tomita Apr 01 '25 at 07:55
  • 1
    If $P_1=a^2-b^2$ is already a square, that implies $P_2=a^2+9b^2$ is a square too. I've come across cases where products of two quadratics $P_1(x)P_2(x)=y^2$ is a square only when together. – Tito Piezas III Apr 01 '25 at 08:00
  • 1
    To $c$ and $d$ are rational number, we need $a^2-b^2$ and $a^2+9b^2$ are square number. – Tomita Apr 01 '25 at 08:11
  • 1
    I may be mistaken. How do we get $c,d$? – Tomita Apr 01 '25 at 08:26
  • 1
    For $(c,d)$ to be rational, we only need $a^2-b^2 = nz_1^2$ and $a^2+9b^2=nz_2^2$. For example, my first solution is $(a,b)=(13,3)$, then $a^2-b^2 = 10\times4^2$ and $a^2+9b^2 = 10\times5^2$ and $(c,d)$ are rational. So I'm surprised your $(a-b,,a+b)$ are already squares. – Tito Piezas III Apr 01 '25 at 09:23
  • 1
    We can transform $(a^2-b^2)(a^2+9b^2)=v^2$ to another elliptic curve $Y^2=X^3 -X^2 +15X + 225,$ but this curve has no solution $x_k$. For some examples, $(a,b,a+b,a-b)=(3, 2, 5, 1),(13, 3, 16, 10),(267, 22, 289, 245),(617307, 564262, 1181569, 53045).$ – Tomita Apr 01 '25 at 09:49
  • 1
    I made some improvements to your answer. The second elliptic curve has smaller $a_k$ for $k=0,1,2,3$. Can you include $k=4$? – Tito Piezas III Apr 01 '25 at 10:35
  • 1
    System equation $(a^2-b^2=u^2,a^2+9b^2=v^2)$ is reduced to $v^2=10x^4-16x^2+10$ and it is transformed to $E_3: Y^2=X^3-X^2-30X+72.$ The previous $E_1: Y^2=X^3+8X^2-9X$ is birationally equivalent to $E_3.$ – Tomita Apr 01 '25 at 12:42
  • 1
    Ah, so the first elliptic curve was really designed to make $P_1=a^2-b^2$ and $P_2=a^2+9b^2$ as squares separately, while the second elliptic curve is together as $P_1P_2 = y^2$. Maybe you can add your comment to your answer. And also how $E_2$ was derived. – Tito Piezas III Apr 01 '25 at 13:00