5

I. Ramanujan's parameterizations

Ramanujan gave just two quadratic parametrizations to,

$$a^4+b^4+c^4+d^4+e^4 = f^4$$

where $f$ is always integrally divisible by $5$, one of which is,

$$(2x^2+12xy-6y^2)^4+(2x^2-12xy-6y^2)^4+(4x^2-12y^2)^4 \\ +(3^4+4^4)(x^2+3y^2)^4 = 5^4(x^2+3y^2)^4$$

In fact, using a more general identity, we can find infinitely many quadratic parameterizations such as,

$$(50x^2 + 300x y - 150y^2)^4 + (50x^2 - 300x y - 150y^2)^4 + (100x^2 - 300y^2)^4 \\ + (4^4 + 15^4)(x^2 + 3y^2)^4 = 103^4(x^2 + 3y^2)^4$$

as discussed in this post.


II. Pythagorean triples

Letting $(x,y)=(1,0)$ in Ramanujan's formula above, we get the equality,

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

One can notice the Pythagorean triple $(a,b,c)=(3,4,5)$. More generally, it turns out the equation,

$$(p + q)^4 + (-p + q)^4 + (2q)^4 + a^4 + b^4 = c^4$$

where $a^2+b^2=c^2$ is true if the simple condition is satisfied,

$$ab = p^2+3q^2$$

Note: Of course, one can swap $(p,q)\to (q,p)$ depending on the choice of notation.


III. Question

How do we show there are infinitely many primitive Pythagorean triples $(a,b,c)$ such that,

$$ab = p^2+3q^2$$

and $(p,q)$ are integers?


P.S. I did a computer search. Examples of the special triple $(a,b,b+1)$ are,

$$2^4+2^4+4^4+3^4+4^4=5^4,\qquad 3^2+4^2=5^2$$ $$2^4 + 32^4 + 34^4 + 13^4 + 84^4 = 85^4,\qquad 13^2 + 84^2 = 85^2$$ $$6^4 + 96^4 + 102^4 + 27^4 + 364^4 = 365^4,\qquad 27^2 + 364^2 = 365^2$$

and so on, though it is unknown if this quadratic-quartic subset has infinitely many members.


Update: In his answer, Tomita transformed the problem into a family of elliptic curves with an infinite number of rational points. As an bonus, the particular curve he chose with parameter $r=1$ shows an additional structure to the equation, namely,

$$(p + q)^4 + (p - q)^4 + (2p)^4 + a^4 + b^4 = (2p+a)^4$$

where $a^2 + b^2 = c^2 = (2p+a)^2$. Example:

$$4^4 + 2^4 + \color{red}{2}^4 + \color{blue}{3}^4 + 4^4 = (\color{red}{2}+\color{blue}{3})^4 = 5^4$$

$$1506^4 + 2154^4 + \color{red}{648}^4 + \color{blue}{2077}^4 + 1764^4 = (\color{red}{648} + \color{blue}{2077})^4 = 2725^4$$

$$22162738^4 + 22105616^4 + \color{red}{57122}^4 + \color{blue}{128068563}^4 + 3825484^4 \\= (\color{red}{57122} + \color{blue}{128068563})^4 = 128125685^4$$

and so on, for infinitely many primitive Pythagorean triples $(a,b,c)$.

  • 1
    I already checked $6$th powers. There doesn't seem to be any analogous relationships between $3$rd and $6$th powers. – Tito Piezas III Mar 29 '25 at 04:38
  • 1
    It's well-known which integers are represented by $p^2+3q^2$ (although I don't happen to know the answer), and it's well-known what $a,b$ look like for primitive Pythagorean triples, so it shouldn't be hard to work out an answer. – Gerry Myerson Mar 29 '25 at 06:02
  • 1
    The integers represented as $p^2+3q^2$ are tabulated at https://oeis.org/A003136 which says they are the numbers which, if divisible by a prime $\ell\equiv2\bmod3$, the exact power of $\ell$ dividing them is even. – Gerry Myerson Mar 29 '25 at 06:12
  • 1
    $a=16$, $b=63$, $p=24$, $q=12$ gives $36^4+12^4+24^4+16^4+63^4=65^4$. – Gerry Myerson Mar 29 '25 at 06:34
  • 1
    @GerryMyerson I did a computer search with $f$ below a bound, found $f=65$ and many others. Yes, essentially , one needs to solve $ab=2mn(m^2-n^2)=p^2+3q^2$ where GCD(a,b)=1. It is almost certain there are infinitely many primitive Pythagorean triples $(a,b,c)$ that qualify, but a proof would be nice. – Tito Piezas III Mar 29 '25 at 07:11

2 Answers2

3

$$ 3p^2+q^2 = 2mn(m^2-n^2) $$
Let $x=\dfrac{m}{n},\ y=\dfrac{q}{n^2},$ and $r=\dfrac{p}{n^2}.$

We have $$y^2 = 2x^3-2x-3r^2.$$ I searched the rational points $(x,y)$ where $p+q<10000$ and $r<100.$

Obviously, this equation can be transformed to Weierstrass form: $$Y^2=X^3-4X-12r^2$$ For the case of $r=1, Y^2=X^3-4X-12r^2$ has rank $1$ and generator is $P(X,Y)=(4,6).$ We get the solutions using group structure $kP$ with $k=1,2,3,4$ as follows.

[r][p+q, p-q, 2p,a,b,c]

[1][4, 2, 2, 3, 4, 5] [1][1506, 2154, 648, 2077, 1764, 2725] [1][22162738, 22105616, 57122, 128068563, 3825484, 128125685] [1][112480990768740, 22759941088740, 89721049680000, 72967048295281, 145407108099600, 162688097975281]

[p+q, p-q, 2p,a,b,c]

[24, 12, 36, 63, 16, 65] [92, 74, 166, 133, 156, 205] [106, 58, 164, 133, 156, 205] [124, 34, 158, 133, 156, 205] [142, 4, 146, 133, 156, 205] [210, 150, 60, 675, 52, 677] [310, 110, 200, 325, 228, 397] [346, 314, 32, 1443, 76, 1445] [368, 274, 94, 1443, 76, 1445] [374, 256, 118, 1443, 76, 1445] [382, 206, 176, 1443, 76, 1445] [470, 340, 130, 589, 300, 661] [534, 438, 972, 637, 1116, 1285] [666, 282, 948, 637, 1116, 1285] [610, 50, 560, 925, 372, 997] [650, 490, 160, 925, 372, 997] [702, 678, 24, 3843, 124, 3845] [726, 648, 78, 3843, 124, 3845] [798, 84, 882, 637, 1116, 1285] [812, 718, 94, 1333, 444, 1405] [834, 18, 852, 637, 1116, 1285] [844, 662, 182, 1333, 444, 1405] [910, 890, 20, 5475, 148, 5477] [942, 258, 684, 637, 1116, 1285] [948, 282, 666, 637, 1116, 1285] [980, 790, 190, 5475, 148, 5477] [1046, 862, 184, 1813, 516, 1885] [1110, 240, 1350, 819, 1900, 2069] [1114, 626, 488, 1813, 516, 1885] [1364, 428, 1792, 1273, 2064, 2425] [1370, 1240, 130, 2451, 700, 2549] [1386, 798, 2184, 2077, 1764, 2725] [1740, 570, 1170, 1525, 1548, 2173] [1772, 364, 1408, 1273, 2064, 2425] [1770, 780, 990, 1525, 1548, 2173] [1942, 1838, 104, 4453, 804, 4525] [1974, 126, 1848, 2077, 1764, 2725] [2084, 2012, 4096, 3913, 3216, 5065] [2166, 1464, 702, 2077, 1764, 2725] [2184, 798, 1386, 2077, 1764, 2725] [2202, 1266, 936, 2077, 1764, 2725] [2448, 2286, 162, 5427, 1036, 5525] [2946, 498, 2448, 3397, 2196, 4045] [3072, 2154, 918, 3397, 2196, 4045] [3138, 1842, 1296, 3397, 2196, 4045] [3356, 356, 3712, 3913, 3216, 5065] [3382, 3218, 164, 9373, 1164, 9445] [3894, 3882, 7776, 11557, 3924, 12205] [3988, 2804, 1184, 3913, 3216, 5065] [4036, 1412, 2624, 3913, 3216, 5065] [4044, 114, 4158, 5917, 2844, 6565] [4326, 492, 3834, 5917, 2844, 6565] [4604, 4498, 106, 47523, 436, 47525] [4800, 4680, 120, 50175, 448, 50177] [4962, 276, 5238, 7957, 3276, 8605] [5038, 4862, 176, 16093, 1524, 16165] [5422, 5186, 236, 17653, 1596, 17725] [5478, 4626, 852, 7957, 3276, 8605] [5676, 4218, 1458, 7957, 3276, 8605] [5930, 4820, 10750, 14259, 6100, 15509] [6576, 6534, 42, 77283, 556, 77285] [7218, 6114, 1104, 11557, 3924, 12205] [7446, 5664, 1782, 11557, 3924, 12205] [7670, 7540, 130, 28525, 2028, 28597] [7734, 4566, 3168, 11557, 3924, 12205] [7746, 3282, 4464, 11557, 3924, 12205] [7878, 4758, 12636, 22477, 5436, 23125] [8170, 7850, 320, 30589, 2100, 30661] [8316, 6468, 14784, 9457, 17424, 19825] [9042, 3498, 5544, 14317, 4356, 14965]

Tomita
  • 3,209
  • I don't get it. The first entry in your table corresponds to $p=18$, $q=6$. What is $r$ in $y^2=2x^3-2x-3r^2$ (or $Y^2=X^3-4X-12r^2$)? And then, what's the corresponding rational point? – Gerry Myerson Mar 30 '25 at 02:41
  • 1
    @GerryMyerson I think given Tomita's family of elliptic curves $y^2 = 2x^3-2x-3r^2$, he let the computer run through various rational $r$, searching for rational $(x,y)$. Once found, then one can recover $(m,n,p,q)$ from the relations $(x, y, r) = (m/n, q/n^2, p/n^2)$. For the first entry, I believe he used $(x,y,r)=(8, 6, 18)$. This yields $(m,p,q)=(8n, 18n^2, 6n^2)$ where one can set $n=1$ to get your $(p,q)=(18,6)$. – Tito Piezas III Mar 30 '25 at 03:23
  • @Tomita Since $y^2 = 2x^3-2x-3r^2$ is a family of elliptic curves where presumably infinitely many rational $r$ will do, I think a subset of $r$ can be given in polynomial form, hence also $(x,y)$. I will check. – Tito Piezas III Mar 30 '25 at 03:29
  • A solution $(p+q,p-q,2p,a,b,c)=(24, 12, 36, 63, 16, 65)$ was got using $(x,y,r)=(8, 6, 18)$ and $(p,q,m,n)=(18, 6, 8, 1).$ – Tomita Mar 30 '25 at 03:45
  • @Tomita Since this family of elliptic curves provides infinitely many rational points, then that answers the question. By the way, did you notice the additional structure $$(p + q)^4 + (p - q)^4 + (2p)^4 + a^4 + b^4 = \color{blue}{(2p+a)}^4$$ provided by the elliptic curve you chose? – Tito Piezas III Mar 31 '25 at 06:58
  • @Tito, I didn't notice that at all. It's interesting! – Tomita Mar 31 '25 at 10:55
  • @Tito, $(a,b,c)=(m^2-n^2,2mn,m^2+n^2).$ Since $c-a=2n^2$ and $x=m/n, $then $c-a=2,72,648$ with $n=1,6,18.$ – Tomita Mar 31 '25 at 11:19
  • @Tomita Ah, one mystery solved. – Tito Piezas III Mar 31 '25 at 11:38
  • @Tito, If $p=n^2,$ then $a+2p=m^2-n^2+2n^2=m^2+n^2=c.$ – Tomita Mar 31 '25 at 12:40
  • @Tomita Here's a new one with 6th powers. – Tito Piezas III Apr 01 '25 at 04:01
1

Seiji Tomita has given an Identity on his web site. The link is shown below:

http://www.maroon.dti.ne.jp/fermat/dioph57e.html

Click the fourth power section & select the posting # 86.

The Identity (1), is shown below:

$(-290-500k+40k^2)^4 + (250-80k-290k^2)^4 + (-40-580k-250k^2)^4 + 325^4(1+k+k^2)^4 + 228^4(1+k+k^2)^4 = 397^4(1+k+k^2)^4$

In the above:

$a=228(1+k+k^2)$

$b=325(1+k+k^2)$

$c=397(1+k+k^2)$

$p+q=(250-80k-290k^2)$

$p-q=(-40-580k-250k^2)$

$2q=(-290-500k+40k^2)$

In the above, the condition $ab=p^2+3q^2$ is satisfied so:

$[p-q,\,p+q,\,2q,\,a,\,b]^4=[c]^4$

where $a^2+b^2=c^2$. For $k=0$:

$(290,250,40,325,228)^4=(397)^4$

and $325^2+228^2=397^2$.

Note: Since equation (1) is an identity, it means that that there are many numerical solutions for different values of the parameter 'k'.

Also note that the numerical solutions given by Tito Piezas & Gerry Myerson has the variable 'c' divisible by '5' while the solution in the answer is (not) divisible by '5'. Implying that this identity given in the answer belongs to a different parametric family group.

Note added:

Alternate way to arrive at (4-1-5) quartic equation:

$(q+p)^4+(q-p)^4+(2q)^4+a^4+b^4=c^4$ ------(1)

We take:

$a=(8k-3)$

$b=4(4k-1)(2k-1)$

$c=(32k^2-24k+5)$

the above satisfies, $(c^2=a^2+b^2)$

we have from (1):

$(c^4-a^4-b^4)=(q+p)^4+(q-p)^4+(2q)^4$

After substitution from above we get:

$(c^4-a^4-b^4)=2[(8k-3)(4)(4k-1)(2k-1)]^2 =2(ab)^2$

We also have:

$(q+p)^4+(q-p)^4+(2q)^4=2(p^2+3q^2)^2$

hence: $2(p^2+3q^2)^2=2(ab)^2$

Therefore: $p^2+3q^2=ab$ -- (the condition required)

For, k=2, we get:

$(a,b,c)=(13,84,85)$

from, $(p^2+3q^2)=(ab)=(13*84)=1092$ ----(2)

(p,q)=(3,19) satisfies equation (2)

therefore we get:

$(p,q,a,b,c)=(3,19,13,84,85)$

And substituting above in equation (1) we get:

$(22,16,38,13,84)^4=(85)^4$


David
  • 235
  • 1
    I made some edits to improve the formatting. While this family has the required form $$[p-q,,p+q,,2q,,a,,b]^4=[c]^4$$ where $a^2+b^2=c^2$ note that its Pythagorean triple $(a,b,c)$ is not primitive since it has the common factor $1+k+k^2$. This belongs to the "super-family" I discussed in this other post. – Tito Piezas III Mar 30 '25 at 03:54
  • @Tito Piezas III. The identity given by Tomita is okay, but the (RHS) grows very fast as 'k' increases. In, [(p+q),(p-q),2p,a,b]^4=c^4 we have, c=(2p+a) & b^2=4p(p+a). thus the equation needs to be solved for only two unknowns, (p,a). So, I would'nt be surprised, if there is a much simpler identity lurking around which supplies us much smaller numerical solutions. – David Apr 02 '25 at 18:35
  • Also see my note added to my answer on 03 April 2025. – David Apr 03 '25 at 08:49