3

(Note: Positive integers throughout, except where specified.)

I. Parameterization

Following on from this excellent question from Tito Piezas III, "Sums of Cubes of form $a^3+b^3+c^3=(c+1)^3$". Piezas observed there were pairs that shared the same $c$, hence the same sum $c+1$, for example,

\begin{align} 9^3+58^3+255^3 &=256^3\\ 22^3+57^3+255^3 &=256^3 \end{align}

So I have produced 4405 solutions to $a^3+b^3+c^3=(c+1)^3$ up to $c=3.15\times10^8,$ then extracted $128$ solutions (given in next section) that share the same sum,

$$a^3+b^3+c^3=A^3+B^3+c^3=(c+1)^3$$

For the eight cases where $b-B=1$, I have found a parametric solution

$$a=9n^3$$ $$b=27n^4+18n^3+9n^2+3n+1$$ $$c=81n^6+81n^5+54n^4+27n^3+9n^2+3n$$ $$A=9n^3+9n^2+3n+1$$ $$B=27n^4+18n^3+9n^2+3n$$

Proof: Just expand and equate.

Non-positive $n$ give incorrect results if negative results are moved to change sign. For the majority of cases where $b-B\gt1,$ I have not yet found parametric solutions.

Results:

(n,a,b,c,A,B)

(1,9,58,255,22,57)
(2,72,619,8898,115,618)
(3,243,2764,83925,334,2763)
(4,576,8221,430428,733,8220)
(5,1125,19366,1556115,1366,19365)
(6,1944,39223,4485150,2287,39222)
(7,3087,71464,11030313,3550,71463)
(8,4608,120409,24123480,5209,120408)

For $n=1$ to $8$ these agree with the expected results.

My question:

Can you please find parametric solutions for one or more of the following groups of numerical solutions below to $a^3+b^3+c^3=A^3+B^3+c^3=(c+1)^3$ ?

P.S. Failing that, a link to a reliable sequence formula calculator, or a detailed worked example of converting difference tables to formula manually.


II. Raw Data

The 128 solutions below are sorted by $b-B = k$ for $k\leq 103$, including 28 with $k> 103.$

128 = 16 + 12 + 8 + 2 + 8 + 2 + 6 + 2 + 4 + 4 + 2 + 4 + 2 + 4 + 4 + 2 + 4 + 4 + 4 + 2 + 4 + 28

Data for $b-B=1$

(22,57,255)
(9,58,255)
(115,618,8898)
(72,619,8898)
(334,2763,83925)
(243,2764,83925)
(733,8220,430428)
(576,8221,430428)
(1366,19365,1556115)
(1125,19366,1556115)
(2287,39222,4485150)
(1944,39223,4485150)
(3550,71463,11030313)
(3087,71464,11030313)
(5209,120408,24123480)
(4608,120409,24123480)

Data for $b-B=7$

(51,82,477)
(64,75,477)
(294,847,14526)
(343,840,14526)
(1302,5971,267762)
(1435,5964,267762)
(2841,16672,1245927)
(3064,16665,1245927)
(6333,48004,6079299)
(6712,47997,6079299)
(10302,91321,15944352)
(10825,91314,15944352)

Data for $b-B=13$

(1956,6631,315726)
(2095,6618,315726)
(2892,11137,684480)
(3073,11124,684480)
(13683,87400,14946453)
(14194,87387,14946453)
(16941,116026,22853211)
(17530,116013,22853211)

Data for $b-B=18$

(69,1018,18755)
(381,1000,18755)

Data for $b-B=19$

(594,1003,20154)
(643,984,20154)
(1719,4258,165609)
(1828,4239,165609)
(11673,54700,7421991)
(12076,54681,7421991)
(18738,102595,19030386)
(19291,102576,19030386)

Data for $b-B=30$

(7739,92810,16328917)
(10739,92780,16328917)

Data for $b-B=31$

(822,1009,22968)
(865,978,22968)
(7401,21256,1826583)
(7648,21225,1826583)
(23403,98914,18079365)
(23944,98883,18079365)

Data for $b-B=33$

(37,174,1331)
(136,141,1331)

Data for $b-B=37$

(2709,4762,206451)
(2818,4725,206451)
(5796,13477,938532)
(5989,13440,938532)

Data for $b-B=43$

(19689,62740,9212247)
(20116,62697,9212247)
(35439,137650,29735697)
(36076,137607,29735697)

Data for $b-B=48$

(46,6441,298448)
(1810,6393,298448)

Data for $b-B=49$

(31359,106888,20429049)
(31918,106839,20429049)
(38523,140740,30794541)
(39166,140691,30794541)

Data for $b-B=57$

(1102,14421,1000067)
(3325,14364,1000067)

Data for $b-B=61$

(8124,14653,1107900)
(8317,14592,1107900)
(14685,32986,3608235)
(14986,32925,3608235)

Data for $b-B=67$

(5898,8629,531564)
(6037,8562,531564)
(26373,68104,10554939)
(26812,68037,10554939)

Data for $b-B=72$

(16,801,13088)
(502,729,13088)

Data for $b-B=73$

(4203,4780,247293)
(4294,4707,247293)
(44064,127975,26965902)
(44671,127902,26965902)

Raw data for $b-B=79$

(6558,8635,555546)
(6691,8556,555546)
(42855,116638,23561961)
(43432,116559,23561961)

Data for $b-B=91$

(19215,35266,4121265)
(19516,35175,4121265)
(31194,68563,10842210)
(31627,68472,10842210)

Data for $b-B=97$ (Second pair added by Piezas)

(11040,15253,1277268)
(11221,15156,1277268)
(60117, 159286, 37677027)
(60790, 159189, 37677027)

Data for $b-B=103$

(16545,25864,2697567)
(16792,25761,2697567)
(55326,136441,30051936)
(55945,136338,30051936)


More raw data for $k>103$

(39006,72487,12113622) (39439,72360,12113622)

(13380,14689,1361952) (13537,14556,1361952)

(21129,25888,2987895) (21352,25737,2987895)

(47781,78856,14135451)
(48220,78693,14135451)

(71169,133498,30219399)
(71758,133329,30219399)

(54477,80212,15030579)
(54904,80013,15030579)

(32835,35326,5147325)
(33076,35115,5147325)

(62517,80236,15925707)
(62920,79989,15925707)

(67881,78916,16373271)
(68260,78633,16373271)

(68346,72577,15292152)
(68689,72270,15292152)

(201,3166,102863)
(2097,2824,102863)

(3029,12038,768604)
(8021,10790,768604)

(865,20886,1742760)
(15273,17704,1742760)

(8781,72346,11244731)
(47997,64534,11244731)

Note: Despite checking and double-checking, mistakes happen (sorry). Please check before use.

Old Peter
  • 1,357
  • 1
    I think it is interesting that $b-B = 1,7,13,19,31 = 6k+1$ has relatively many solutions. Furthermore, all their $b$ is of the form $b \equiv 1,(\text{mod},3)$ which automatically implies $B \equiv 0,(\text{mod},3)$. If the case $b-B=1$ has a parameterization, it is tempting to speculate that the others might have as well. – Tito Piezas III Nov 08 '23 at 03:19
  • 1
    However, $b-B =1$ is easier because of $a=9, 72, 243, 576 = 9n^3$ and, even if only four solutions are known, the rest can be extrapolated. For $b-B = 7$, the $a$-variable unfortunately has no clear pattern and, if it does have a parameterization, the difference tables seem to imply that not all belong to the same family. – Tito Piezas III Nov 08 '23 at 03:50
  • 1
    I have copy-edited your post to smoothen its flow, plus some minor correction of typos. I hope it is ok. – Tito Piezas III Nov 09 '23 at 04:15
  • Edit away, by all means. I’m struggling with this mark-up; for this post it took me longer than the research. My guess is that a two parameter system is needed, perhaps even more than two. – Old Peter Nov 09 '23 at 14:16
  • Not sure about “More raw data (unsorted)” as it is sorted by $b-B,a$ before $A$. – Old Peter Nov 09 '23 at 14:28
  • Ok. By the way, there are now 10 polynomial solutions to $a^3+b^3+c^3 = (c+1)^3$. Yours actually was a triplet. Kindly see this new post. – Tito Piezas III Nov 09 '23 at 14:29
  • I have changed the "unsorted" wording as requested. Also, I've analyzed your data for $k=7$ and posted a partial answer. There are promising results, but the given data is insufficient. Do you think you can provide the next few pairs for $k=7$? – Tito Piezas III Nov 10 '23 at 08:15
  • Never mind a deeper search for $k = 7$. I managed to find their families by filling in some blanks. Give me a while to post them. – Tito Piezas III Nov 10 '23 at 13:36
  • Answer had been updated with four families for $k=7$. – Tito Piezas III Nov 10 '23 at 15:34
  • Doesn’t this intersect with https://math.stackexchange.com/questions/1933638/does-the-3-3-3-equation-x-13x-23x-33-y-13y-23y-33-have-a-complete-in, and in particular the @TitoPiezasIII answer at https://math.stackexchange.com/a/1935275/93271? – Kieren MacMillan Nov 12 '23 at 16:23
  • 1
    @KierenMacMillan Not exactly. The first post is about $x_1^3+x_2^3+x_3^3 = y_1^3+y_2^3+y_3^3,$ presumably in distinct integers. The second post allows rational solutions. Peter's post is about a special case of $x_1^3+x_2^3 = y_1^3+y_2^3$ which also satisfies $12(x_1^3+x_2^3)-3 = \square$ specifically in the integers, which makes it a tad more difficult. – Tito Piezas III Nov 12 '23 at 16:35

3 Answers3

2

(New answer, a few hours later.)

As I suspected (see old answer below), while the case $k=1$ involved just one pair of families that solved,

$$a^3+b^3+c^3 = A^3+B^3+c^3 = (c+1)^3$$

the case $k=7$ involved two pairs of families.


I. Pair 1

\begin{align}a_1 &= 3 (7 n - 2) (21 n^2 - 30n + 12) + 6 = \color{blue}{51,\, 1302,\, 6333},\dots\\[6pt] b_1 &= 3 (441 n^4 - 966 n^3 + 795 n^2 - 275 n + 30) + \color{red}7\\[6pt] c_1 &= 3(9261 n^6 - 30429 n^5 + 41706 n^4 - 29835 n^3 + 11451 n^2 - 2147n + 152)\\[6pt] A_1 &= 3 (7 n - 1) (21 n^2 - 30n + 13) - 8 = \color{blue}{64,\, 1435,\, 6712},\dots\\[6pt] B_1 &= 3 (441 n^4 - 966 n^3 + 795 n^2 - 275 n + 30)\\[6pt] \end{align}

Thus $b_1-B_1 = 7.\,$ (Note that powers of $7$ appear throughout, such as $441 = 21^2,$ and $9261 = 21^3.$) Example for $n=1$,

$$51^3 + (75 + 7)^3 + 477^3 = 478^3\\ 64^3 + (75 + 0)^3 + 477^3 = 478^3$$


II. Pair 2

\begin{align}a_2 &= 3(7 n + 1) (21 n^2 - 12n + 3) + 6 = \color{blue}{294,\, 2841,\, 10302},\dots\\[6pt] b_2 &= 3 (441 n^4 - 210 n^3 + 39 n^2 + 13n - 3) + \color{red}7\\[6pt] c_2 &= 3 (9261 n^6 - 6615 n^5 + 2016 n^4 + 351 n^3 - 213 n^2 + 31n + 11)\\[6pt] A_2 &= 3 (7 n + 2) (21 n^2 - 12n + 4) - 8 = \color{blue}{343,\, 3064,\, 10825},\dots\\[6pt] B_2 &= 3 (441 n^4 - 210 n^3 + 39 n^2 + 13n - 3)\\[6pt] \end{align}

Likewise, $b_2-B_2 = 7.\,$ Example for $n=1$,

$$294^3 + (840 + 7)^3 + 14526^3 = 14527^3\\ 343^3 + (840 + 0)^3 + 14526^3 = 14527^3$$

Peter's 12 blue numbers were basically all I had to work with, and I had to separate them properly to find these two pairs of families. Plus a lot of guesswork based on the structure of the case $k=1,$ and help from Mathematica.

I assume that, with enough data, parameterizations of $b-B = k\,$ for other $k$ may be found. (They have also been found for $k=13,19,31$. Will be posted later.)


(Old answer)

I. Case $b-B = 1$

The differences of the first 6 pairs,

$$(22 - 9,\; 115 - 72,\; 334 - 243,\; 733 - 576,\; 1366 - 1125,\; 2287 - 1944)$$

are $(13,\; 43,\; 91,\; 157,\; 241,\; 343).\,$ Their differences divided by 6 are,

$$(5,\; 8,\; 11,\; 14,\; 17)$$

with a constant difference of $\color{blue}{3\times1 = 3}$. They are in arithmetic progression and, testing for the rest of the pairs, it also holds true. Thus $k=1$ has a polynomial parameterization.


II. Case $b-B = 7$

The differences of the 6 known pairs,

$$(64 - 51,\; 343 - 294,\; 1435 - 1302,\; 3064 - 2841,\; 6712 - 6333,\; 10825 - 10302)$$

are $(13,\; 49,\; 133,\; 223,\; 379,\; 523).\,$ Their differences are also divisible by 6 but do not yield an arithmetic progression. However, if we take differences of every other number,

$$-(13-133,\; 133-379)/6 = (20,\,41)$$ $$-(49-223,\; 223-523)/6 = (29,\,50)$$

then we seem to have a constant difference of $\color{blue}{3\times7 = 21}$. Unfortunately, with limited data, this is hard to test. Maybe more data points are needed. (See new answer above.)

  • I’m not exactly clear on your method, (not unusual for me), but if you have a method to predict candidate $c$ values I can run code to look for solutions. – Old Peter Nov 10 '23 at 15:39
  • @OldPeter Actually, it's the $a$-values that are more "accessible" since for both $k=1$ and $k=7$ they are determined by just cubic polynomials, but $c$ is a sextic. You are in fact familiar with my method of getting differences between the terms of a sequence (my old answer). The complication is if there is more than one family, as is the case with $k=7$, and I was lucky you just had enough data for me to see a pattern. The case $k=13$ probably will be trickier. – Tito Piezas III Nov 10 '23 at 15:55
  • I have an idea and will try it tomorow. – Old Peter Nov 10 '23 at 19:29
2

Thanks to Peter's database and a lot of painstaking data-mining and trial-and-error using Mathematica, and guided by the form for cases $k=1$ and $k=7$, we can show there are infinitely many polynomial identities to,

$$a^3+b^3+c^3 = A^3+B^3+c^3 = (c+1)^3$$

Parameterizations can exist if $k=b-B = \color{blue}{9m^2\pm3m+1} = 1,7,13,31,43,\dots$ as well as for others such as $k=19$. Recall from this post that this polynomial is quite important to "real cubic fields".

We can solve the equation as a quadratic in $c$,

$$c = \frac{-3+\sqrt{12a^3+12b^3-3}}6 = \frac{-3+\sqrt{12A^3+12B^3-3}}6$$

Thus we need to find appropriate $(a,b,A,B)$. First, define some variables for the first pair,

$$p = 9m^2,\quad q = 9m^2+1$$

and for the second pair,

$$r = -3m,\quad s = -3m+1$$

Then the two pairs are related by,

$$k = \color{blue}{p+s = q+r} = 9m^2-3m+1 $$


I. First Pair

Using $p = 9m^2,\;q = 9m^2+1,$

\begin{align} a_1 &= 9k^2n^3 + 9k p n^2 + 3p^2 n + (54m^4 - 9m^3 + 9m^2 - 3m)\\[6pt] b_1 &= B_1+k\\[6pt] A_1 &= 9k^2n^3 + 9k q n^2 + 3q^2 n + (54m^4 - 9m^3 + 18m^2 +1)\\[6pt] B_1 &= 27k^2 n^4 + 18k(18m^2 + 1)n^3 + 9(162m^4 + 18m^2 + 1)n^2 + 3(135m^4 + 18m^3 + 27m^2 + 1)n + (54m^4 + 9m^3 + 9m^2 + 3m) \end{align}

so $b_1-B_1 = k$. Example, let $m=2$, so $k = 31$, and $n=0,1$. Then,

$$822^3 + (978 + 31)^3 + 22968^3 = 22969^3\\ 865^3 \,+ (978 + 0)^3 \,+ 22968^3 = 22969^3\\ 23403^3 + (98883 + 31)^3 + 18079365^3 = 18079366^3\\ 23944^3 + (98883 + 0)^3 \,+\, 18079365^3 = 18079366^3$$


II. Second Pair

Using $r = -3m,\; s = -3m+1,$

\begin{align} a_2 &= 9k^2n^3 + 9k r n^2 + 3r^2 n + (27m^4 - 18m^3 + 9m^2 - 3m)\\[6pt] b_2 &= B_2+k\\[6pt] A_2 &= 9k^2n^3 + 9k s n^2 + 3s^2 n + (27m^4 - 18m^3 + 9m^2 - 3m + 1)\\[6pt] B_2 &= 27k^2n^4 - 18k(6m - 1)n^3 + 9(18m^2 - 6m + 1)n^2 + 3(27 m^4 - 18m^3 + 9m^2 - 6m + 1)n - 9m^3 \end{align}

Likewise, $b_2-B_2 = k$. Example, let $m=2$, so $k = 31$ again, and $n=1,2$. Then,

$$7401^3 + (21225 + 31)^3 + 1826583^3 = 1826584^3\\ 7648^3 \,+ (21225 + 0)^3 \,+ 1826583^3 = 1826584^3\\ 63030^3 + (370050 + 31)^3 + 130303080^3 = 130303081^3\\ 64081^3 + (370050 + 0)^3 \,+\, 130303080^3 = 130303081^3$$

For $k=31$, Peter found the first six, but not the 7th and 8th as it had $c$ beyond his search range. These four identities "explain" why the number of his solutions come in multiples of 4, and why pairs of first terms are relatively close to each other. However, there is probably another superfamily that contains $k=19, 37,$ and others.

  • 1
    Congratulations, this answer certainly looks like a major breakthrough, but sadly it’s probably beyond my skillset to check. Have you found any smallish solutions? – Old Peter Nov 12 '23 at 14:35
  • 1
    @OldPeter This actually covers all of your solutions as long as $k = 9m^2\pm3m+1$. The four $a_1, A_1, a_2, A_2$ explains why the number of your solutions per $k$ come in multiples of $4$: they are the smallest members of the family. I'll revise the examples to illustrate the case $k=13$. Don't you have Mathematica? – Tito Piezas III Nov 12 '23 at 15:28
  • No, I don’t, sorry. Maths is just one of my many hobbies. – Old Peter Nov 12 '23 at 18:56
  • @OldPeter I assume you used a program to find these numerical solutions. May I know what? Sage? – Tito Piezas III Nov 12 '23 at 21:29
  • Coded in VB6 and run, single thread, on an old desktop. I can post any results wanted, just let me know. – Old Peter Nov 13 '23 at 10:52
  • 1
    @OldPeter Oh, I see. I found another family $k = 3m^2-3m+1 = 1,7,19,37,61,91.$ So what remains unexplained for primes $k\leq103$ are $k = 67,79,103$. If possible, can you find 4 more solutions for each of these last three primes? – Tito Piezas III Nov 13 '23 at 11:09
  • That’s good. As is the way with such programs, their performance drops dramatically as the control variable increases. I estimate that request would require months. If you have reasonable estimates for $c$, it’s practical. – Old Peter Nov 13 '23 at 14:22
  • 1
    @OldPeter I just finished my session with Mathematica. Since you have 4 solutions each for $k=67,79,103$, I used those as a "seed" to generate 4 x 3 = 12 identities, with each solution being the smallest value. It seems then if a given $k$ has one solution, then it has infinitely many, though I can't prove it in the general case. I'll post the identities tomorrow. – Tito Piezas III Nov 13 '23 at 16:08
2

This third answer neatly generalizes both my first and second answers. When I noticed similarities between the polynomial parameterizations I found for ALL primes $p = 6m+1$ with $p<100,$ I knew there had to be a common form. After some effort, I found it. Given,

$$a^3+b^3+C^3 = A^3+B^3+C^3 = (C+1)^3$$

then the two-parameter identity was quite simple as,

\begin{align} a &= \frac{p^3+q^3-1}{3p}\\[4pt] b &= \frac{p^3(q-1)+(q^2+q+1)^2}{3p^2}+\color{blue}p\\[4pt] A &= \frac{p^3+(q+1)^3+1}{3p}\\[4pt] B &= \frac{p^3(q-1)+(q^2+q+1)^2}{3p^2}\\[4pt] C &= \frac{p^6+9p^3q^2+2p^3(q-1)^3+(q^2+q+1)^3}{9p^3} \end{align}

Of course, $b-B = p.\,$ (The new variables $(p,q)$ were chosen for aesthetics.) For example, for $p = 1$, then there is just one choice such that all terms are integers, namely $q = 3n$, and we recover Peter's,

\begin{align} a &=9n^3\\ b &=27n^4+18n^3+9n^2+3n+1\\ A &=9n^3+9n^2+3n+1\\ B &=27n^4+18n^3+9n^2+3n\\ C &=81n^6+81n^5+54n^4+27n^3+9n^2+3n \end{align}

However for $p>1$ and $p = 6m+1 \neq (6e-1)(6f-1)$, then we can have more than one choice depending on the distinct prime factorization of $p$.


Examples:

For $p=7$, then $q = 3(7n+\alpha)$ for $\alpha$ = (3, 6) which explains the two families in my first answer.

For $p = 7\times13=91$, then $q = 3(91n+\alpha)$ for $\alpha$ = (3, 27, 55, 66) which yields four families.

For $p = 7\times13\times19=1729$, then $q = 3(1729n+\alpha)$ for $\alpha$ = (276, 300, 794, 846, 965, 1340, 1459, 1511) which yields eight families.

And so on. In general, if $p$ is prime, then there are two choices of $\alpha$. If $p$ is not prime, then there may be more. Thus, finding integer polynomial parameterizations to,

$$a^3+b^3+C^3 = A^3+B^3+C^3 = (C+1)^3$$

where $b-B = p$ for some desired $p=6m+1$ can be reduced to the easier divisibility problem,

$$B = \frac{p^3(q-1)+(q^2+q+1)^2}{3p^2}$$

where $q = 3(pn+\alpha)$ and finding appropriate integers $\alpha$ within the small range $\alpha<p$ such that $B$'s denominator vanishes, easily done by Mathematica in seconds.

  • This is excellent, thank you. – Old Peter Nov 18 '23 at 10:53
  • 1
    @OldPeter Thanks. It was so satisfying to see there is pattern to your data, and that it was not random. Well, at least most if it. However, it gave me an idea for a new post regarding solving equations of degree $p = 6m+1$ in radicals since I've seen these numbers $\alpha$ before. – Tito Piezas III Nov 18 '23 at 11:34