1

By the Lander-Parkin conjecture, it is assumed there are no non-trivial integer solutions to either, $$a^8+b^8+c^8 = d^8+e^8+f^8\tag1$$ $$a^8+b^8 = c^8+d^8\tag2$$

However, if we relax $(1)$ a bit, $$1118^8 + 1937^8 + 2502045^4 = 455^8 + 1846^8 + 3200691^4\tag{3a}$$ then remove the common factor $13^8$ (pointed out by Adam Bailey), $$86^8 + 149^8 + 14805^4 = 35^8 + 142^8 + 18939^4\tag{3b}$$

In fact, there are infinitely many other primitive solutions like it by solving,

$$(a+b)^8+(a−b)^8+(4ab)^4=(c+d)^8+(c−d)^8+(4cd)^4$$

Question: Likewise, if we relax $(2)$ to either,

$$a^8+b^4 = c^8+d^4\tag4$$ $$a^8+b^4 = c^4+d^4\tag5$$

then are there non-trivial solutions to $(4)$ and $(5)$?

P.S. In response to Bailey's answer below, I forgot to add that solutions must be primitive. Otherwise, $(5)$ is easily solved.

  • Do you mean to ask whether at least one of (4), (5) has a nontrivial solution? [I guess you aren't asking about both since one implies the other.] – coffeemath Nov 17 '22 at 11:19
  • @coffeemath: Yes, I was hoping for solutions to the more symmetric (4), but I'll settle for (5) for now. – Tito Piezas III Nov 17 '22 at 11:51

2 Answers2

3

Extending on On the eighth powers $A^8+B^8=C^8+D^8$: You provided a link to Wroblewski's database

Running a Python script$^1$ against that text file, I get:

#1: line 3 has 1 square: 292^4 + 193^4 = 257^4 + 16^8
#2: line 16 has 1 square: 4849^4 + 58^8 = 4303^4 + 4288^4
#3: line 54 has 1 square: 17332^4 + 23^8 = 17236^4 + 6673^4
#4: line 218 has 1 square: 213672^4 + 116309^4 = 203349^4 + 392^8
#5: line 253 has 1 square: 276598^4 + 223^8 = 253814^4 + 203329^4
#6: line 433 has 1 square: 747633^4 + 465236^4 = 682161^4 + 784^8
#7: line 441 has 1 square: 872^8 + 251873^4 = 676769^4 + 598772^4
#8: line 496 has 1 square: 923339^4 + 292^8 = 922933^4 + 190984^4
Done 1420 lines.

So 8 out of 1420 primitive 4-tuples are of kind $(5)$, and there are none of kind $(4)$.


In case it is useful, here are formulae that generate infinite families of such 4-tuples. See also Finding formula that solves $w^4+x^4=y^4+z^4$ over the integers:

$$\begin{align} f_1(a) &= a^{7} + a^{5} - 2 a^{3} + 3 a^{2} + a \\ f_2(a) &= a^{13} - a^{12} - a^{11} - 5 a^{10} - 6 a^{9} + 12 a^{8} + 4 a^{7} - 7 a^{6} + 3 a^{5} + 3 a^{4} - 4 a^{3} - 2 a^{2} + a - 1 \\ f_5(a) &= a^{13} + 27 a^{12} - 214 a^{11} - 186 a^{10} - 2481 a^{9} + 861 a^{8} - 2804 a^{7} - 972 a^{6} - 2481 a^{5} - 27 a^{4} - 214 a^{3} + 294 a^{2} + a + 3 \end{align}$$

Then the 4-tuples are $(^h\!f(a,b), {}^h\!f(b,-a), {}^h\!f(a,-b),{}^h\!f(b,a))$ where $^h\!f$ denotes the homogenized version of eiter function.

Feed in $(a,b)=1$ and notice that the resulting 4-tuple might have a common factor of 2 (found empirically).


$^1$For reference, here is the script so you can play around with it:

#!/usr/bin/env python3.8

from math import isqrt # Needs Python 3.8 or higher from urllib.request import urlopen import sys, re

URL = "https://www.math.uni.wroc.pl/~jwr/422/422-10m.txt"

Pattern for 4 decimal numbers separated by white-space.

line_pat = re.compile (r'\s(\d+)\s+(\d+)\s+(\d+)\s+(\d+)\s')

count = 0

def check4 (vals, lineno): """ Vals is a 4-tuple of values that satisfy a^4 + b^4 = c^4 + d^4. Work out which of them are squares, and neaty print the result. """ assert len(vals) == 4 assert vals[0]4 + vals[1]4 == vals[2]4 + vals[3]4

# Tuple of square-roots if perfect square, 0 otherwise.
qs = tuple ( isqrt(v) if v == isqrt(v)**2 else 0 for v in vals )

# How many squares do we have?
n_qs = sum (bool(q) for q in qs)

if n_qs != 0:
    global count
    count += 1
    print ("#%d: line %d has %d square" % (count, lineno, n_qs),
           end = ": " if n_qs == 1 else "s: ")
    for i in range(4):
        print ("%s^%d%s" % (qs[i] if qs[i] else vals[i],
                            8 if qs[i] else 4,
                            (" + ", " = ", " + ", "\n")[i]),
               end = "")

lineno = 0

with open ("*.txt") as data:

with urlopen (URL) as data: for bline in data: lineno += 1 # Lines are decoded as "bytes", thus decode them as UTF-8. line = bline.decode().strip()

    match = line_pat.match (line)
    if match:
        values = tuple ( int(match.group(i)) for i in range(1,5) )
        check4 (values, lineno)
    else:
        print ("unrecognized line:", line, file=sys.stderr)
        sys.exit (1)

print ("Done %d lines." % lineno)

emacs drives me nuts
  • 10,390
  • 2
  • 12
  • 31
  • Thanks! So Wroblewski's database has more than a thousand solutions. I was really hoping that at least one of them was of form $(4)$, namely $a^8+b^4 = c^8+d^4$. I'll have to settle for form $(5)$ for now. – Tito Piezas III Nov 17 '22 at 11:48
  • @Tito Piezas III: Maybe one can use the result to estimate how likely it is to find a $(4)$? The data set has only values up to $10^7$, maybe one can use that to extrapolate the number of $a^4+b^4=c^4+d^4$ given $\max(a,b,c,d)$? – emacs drives me nuts Nov 17 '22 at 11:59
  • That's a good question, regarding the density of solutions to $a^4+b^4 = c^4+d^4$ given $\max(a,b,c,d)$. However, I do find it odd there are no form $(5)$ beyond line 500 of the database. I guess it is sheer chance some entries are squares, and the higher you go, the smaller the chance. – Tito Piezas III Nov 17 '22 at 12:06
  • Some statistics: The 4-tuples are sorted according to $a=\max(a,b,c,d)$ (and not according to $a^4+b^4$). In 4/8 cases, the square is the smallest value; in 3/8 cases the 2nd-smallest is the square; and in 1/8 cases the largest number is the square. For the $n$-th 4-tuple, it's $\max\approx n^{2.25}$ but that might just be an artifact due to few samples (and because I only tried $\max\sim n^x$. – emacs drives me nuts Nov 17 '22 at 13:27
  • ...added formulae to my answer that generate infinite families of solutions. Maybe you have luck finding a 4-tuple you need. – emacs drives me nuts Nov 17 '22 at 15:43
  • Thanks. However I have a collection of identities which should be about 400 pages by now. The problem is the degree of $f(x) = y^2$ is much higher than $4$, so we are dealing with hyperelliptic curves which are notoriously difficult. – Tito Piezas III Nov 17 '22 at 15:54
  • If interested, can you run a Python script on Wroblewski's database https://www.math.uni.wroc.pl/~jwr/eslp/tables.htm for $x_1^3+x_2^3+x_3^3 = x_4^3$ and find square $x_i$? There are infinitely many primitive solutions to $a^3+b^6+c^6 = d^6$ (by using an elliptic curve) such as the two smallest $$118^3 + 15^6 + 18^6 = 19^6$$ $$25402^3+27^6+138^6 = 169^6$$ If you can find other small solutions in Wroblewski's database, the discussion is on this post. – Tito Piezas III Sep 19 '23 at 01:55
2

Solutions of (5) can easily be obtained from solutions of the 4-2-2 equation. Given any integers $A,B,C,D$ such that:

$$A^4+B^4=C^4+D^4$$

we have:

$$A^8+(AB)^4=(AC)^4+(AD)^4$$

Taking for example the smallest 4-2-2 solution:

$$59^4+158^4=133^4+134^4$$

we have:

$$59^8+9322^4=7847^4+7906^4$$

Unfortunately there doesn't appear to be any similar way to infer solutions of (4).

Adam Bailey
  • 4,197
  • I forgot to add the constraint that solutions $(a,b,c,d)$ must be *primitive* (no common factor). That's why Wroblewski's database of a thousand primitive solutions was searched for a form (4) or (5). Only the latter was found. Then again, a thousand solutions is a far cry from infinity, so there yet might be hope. – Tito Piezas III Nov 18 '22 at 01:05
  • @TitoPiezasIII Thanks, I probably should have queried first whether you were looking for solutions with no common factor. I am inclined to think that for an equation in unlike powers, implying that a common factor cannot be simply cancelled to yield a smaller solution of the same equation, a solution with a common factor is still of some interest, as indeed is your (3) which has $13$ as a common factor. – Adam Bailey Nov 18 '22 at 19:13
  • 1
    I didn't realize (3) had the common factor $13^8$ (and which can be removed and still maintain the powers). The general solution is, $$(a+b)^8+(a-b)^8 +(4ab)^4 = (c+d)^8+(c-d)^8 +(4cd)^4\tag6$$ where $(a,b,c,d)$ must satisfy, $$a^4+14a^2b^2+b^4 = c^4+14c^2d^2+d^4\tag7$$ The equation (7) has infinitely many primitive solutions, *then* one has to check if it yields primitive (6). For my example in the post, I forgot to do the latter, hehe, but has now been rectified as, $$86^8 + 149^8 + 14805^4 = 35^8 + 142^8 + 18939^4$$ – Tito Piezas III Nov 19 '22 at 01:52