13

Let all variables be positive integers. Consider the forms,

$$a^2+b^2=(b+1)^2\tag1$$ $$a^3+b^3+c^3=(c+1)^3\tag2$$

The smallest solutions are the well-known,

$$3^2+4^2=5^2$$ $$3^3+4^3+5^3=6^3$$

All solutions of Eq.1 are covered by only one polynomial parameterization. On the other hand, not all solutions of Eq.2 are covered by its three known parameterizations.


I. Family 1

$$(3 n^2)^3 + (6 n^2 - 3n + 1)^3 + (9 n^3 - 6n^2 + 3n - 1)^3 = (9 n^3 - 6n^2 + 3n)^3$$

$$3^3 + 4^3 + 5^3 = 6^3\\ 12^3 + 19^3 + 53^3 = 54^3\\ 27^3 + 46^3 + 197^3 = 198^3$$

Note: $F(n) = (6, 54, 198)/6 = 1, 9, 33,\dots$ is A005920, or the "tricapped prism numbers".


II. Family 2

$$(3 n^2)^3 + (6 n^2 + 3n + 1)^3 + (9 n^3 + 6n^2 + 3n)^3 = (9 n^3 + 6n^2 + 3n + 1)^3$$

$$3^3 + 10^3 + 18^3 = 19^3\\ 12^3 + 31^3 + 102^3 = 103^3\\ 27^3 + 64^3 + 306^3 = 307^3$$

Note: $F(n) = (18, 102, 306)/3 = 6, 34, 102,\dots$ is A067389.


III. Family 3

$$(n)^3 + (3n^2 + 2n + 1)^3 + (3n^3 + 3n^2 + 2n)^3 = (3n^3 + 3n^2 + 2n + 1)^3$$

$$1^3 + 6^3 + 8^3 = 9^3\\ 2^3 + 17^3 + 40^3 = 41^3\\ 3^3 + 34^3 + 114^3 = 115^3$$

Note: $F(n) = 3n^3 + 3n^2 + 2n = 8, 40, 114,\dots$ is A143943, or "the Wiener index of a chain of $n$ squares".


Update: (Jan 22, four months later)

Courtesy of an insight by Adam Bailey in this post, given $x^3+y^3+z^3=(z+1)^3$ we now realize that $(x,y)$ of Family 2 and 3 are just lattice points on the same ellipse. Do the substitution, $z = (k + 1)(x - 1) + k y\,$ where $k = \dfrac{3n^2+1}{3n}$ and, after removing a trivial factor, we get the ellipse,

$$3n^2(x^2 - xy + y^2) - (1 + 6n + 12n^2 + 18n^3 + 9n^4)x - (1 + 3n^2 + 9n^4)y + (1 + 3n + 9n^2 + 9n^3 + 9n^4) = 0$$

where two lattice points are $(x_1, y_1)$ and $(x_2, y_2)$ from Family 2 and 3, respectively. For example, let $n=3$ so,

$$27(x^2 - x y + y^2) - 1342x - 757y + 1063 = 0$$

which for this $n$ has only 2 lattice points,

enter image description here

thus we can use these coordinates,

$$27^3 + 64^3 + 306^3 = 307^3\;\\ 3^3 + 34^3 + 114^3 = 115^3$$

and so on for infinitely other $n$.


IV. Unknown Families?

$$14^3 + 23 ^3 + 70^3 = 71^3\\ 21^3 + 46^3 + 188^3 = 189^3\\ 16^3 + 51^3 + 213^3 = 214^3\\ \;9^3 \,+\, 58^3 + 255^3 = 256^3\\ 15^3 + 64^3 + 297^3 = 298^3$$

These and the families above are the primitive solutions with $c<307$.


V. Similar equalities

By reverse-engineering those identities,

$$(an^2+bn+c)^3+(dn^2+en+f)^3+(pn^3+qn^2+r)^3=(pn^3+qn^2+rn+s)^3$$

collecting powers of $n$, equating everything to zero, it seems those three are the only quadratic-cubic identities. Turns out there is also the form $a^3+b^3+c^3 = (c\color{blue}{+3})^3$,

$$(m^2)^3 + (2 m^2 + 3m + 3)^3 + (m^3 + 2m^2 + 3m)^3 = (m^3 + 2m^2 + 3m \color{blue}{+3})^3$$

$$(m^2)^3 + (2 m^2 - 3m + 3)^3 + (m^3 - 2m^2 + 3m \color{blue}{-3})^3 = (m^3 - 2m^2 + 3m)^3$$

which, for $m=3n$, can be factored out into $a^3+b^3+c^3 = (c+1)^3$.


VI. Question

Just like $a^3\pm b^3 \pm c^3 = 1$ has infinitely many polynomial parameterizations (where degrees are getting higher and higher), does $a^3+b^3+c^3 = (c+1)^3$ also have other families with degree higher than the cubic polynomials in this post?

  • 1
    As I suspected, there may be higher degree versions, but one term may have the wrong sign. For example, $$-( m^3 - m^2 + 3m)^3 + (m^3 + 2m^2 + 3)^3 + (m^4 + 2m^2)^3 = (m^4 + 2m^2 + 3)^3$$ If $m=3n$, this becomes $-a^3+b^3+c^3 = (c+1)^3.$ So close! – Tito Piezas III Sep 27 '23 at 14:19
  • 1
    This is a little above my current understanding, but does this link help? https://people.math.harvard.edu/~elkies/4cubes.html

    I coded a simple VB6 program and found $171$ solutions with $c<26755$

    Did you miss (22,27,255,256)? Do you want me to post my results and in what order?

    – Old Peter Sep 29 '23 at 18:50
  • @OldPeter Thanks! Yes, I missed that (though it should be 57, not 27). Kindly post it as a partial answer and maybe someone can find some patterns. For example, we now have two pairs of solutions with the same $c+1$, $$22^3 + 57^3 + 255^3 = 256^3\ 9^3 + 58^3 + 255^3 = 256^3\ 51^3 + 82^3 + 477^3 = 478^3\ 64^3 + 75^3 + 477^3 = 478^3$$ I've seen such behavior before and usually involves a parameterization with two terms with only even powers like $$(9t^3 - 1)^3 + (9t^4 - 3t)^3 + 1^3 = (9t^4)^3$$ hence the two terms are immune to sign changes of $t$. Pls arrange the data by increasing $c+1$. – Tito Piezas III Sep 30 '23 at 03:10
  • @TitoPiezasIII, I have found a simple way to generate solutions in different families. Following your advice to Old Peter, I will post it as a partial answer ( or a long comment). – user25406 Sep 30 '23 at 10:58
  • @TitoPiezasIII The following $c$ values give solution pairs $255, 477, 1331, 8898, 13088, 14526, 18755, 20154, 22968$. I’ll attempt to post the list later. – Old Peter Sep 30 '23 at 15:25
  • @Tito, I don't think this is the answer you want. The problem can be reduced to an elliptic curve $Y^2 = X^3+1728b^3-432$ with $(X,Y)=(12a, 72c+36).$
    Curious solutions.  
    $2287^3+  39222^3+ 4485150^3 = 4485151^3$  
    $1944^3+  39223^3+ 4485150^3 = 4485151^3$
    
    – Tomita Oct 02 '23 at 04:45
  • @Tomita Yes, the problem can be reduced to an elliptic curve, but with an integer condition. Using this elliptic curve and the 3 known families, we can generate an infinite more, but these new families will be rational and not necessarily integral. P.S. The "curious solutions" you found are doubly curious, since their second terms differ only by $1$. – Tito Piezas III Oct 02 '23 at 10:11
  • @Tomita You may like this question on elliptic curves for $a^4+b^4 = c^4+d^4$. – Tito Piezas III Oct 03 '23 at 05:40
  • @Tomita Peter and Adam have found additional families. There may be infinitely many. – Tito Piezas III Nov 09 '23 at 02:41
  • I'm trying to force six points of two point families making one conic family on the update of 2024-01-22. Betting triples can coincide, the two integer points by the pascal analogy should pick out a rational point with tangent parallel to the line through the two. This leads to searching for squares of the form $7\cdot(81n^8+162n^7+216n^6+189n^5+126n^4+63n^3+24n^2+6n+1).$ Thus far $n=-1$ works giving $3^3+4^3+(-6)^3=(-5)^3,$ and $n=-\frac13$ giving $(\frac13)^3+(\frac23)^3+(-\frac23)^3=(\frac13)^3.$ Sadly negative rational... – Jan-Magnus Økland Jan 23 '24 at 20:17
  • Interestingly, those two values of $n$ give the same ellipse, could there be a symmetry or duality here? – Jan-Magnus Økland Jan 23 '24 at 20:35

5 Answers5

4

I’ve produced $4405$ solutions to $a^3+b^3+c^3=(c+1)^3$.

Of these $293$ are of Family 1, $(3n^2,6n^2−3n+1,9n^3−6n^2+3n−1)$

Of these $292$ are of Family 2, $(3n^2,6n^2+3n+1,9n^3+6n^2+3n)$

Of these $216$ are of Family 3, $(n,3n^2+2n+1,3n^3+3n^2+2n)$

That leaves $3604$ to be classified. I’ve moved the $128$ cases, where there are pairs of solutions with equal $c$, to a new question, $a^3+b^3+c^3=(c+1)^3=A^3+B^3+c^3$ solutions. where I found

I’ve found $8$ cases of a new Family 4,

$$a=9n^3$$

$$b=27n^4+18n^3+9n^2+3n+1$$

$$c=81n^6+81n^5+54n^4+27n^3+9n^2+3n$$

And $8$ cases of a new Family 5, $(A,B,c)$

$$A=9n^3+9n^2+3n+1$$

$$B=27n^4+18n^3+9n^2+3n$$

$$c=81n^6+81n^5+54n^4+27n^3+9n^2+3n$$

So, $a^3+b^3+c^3 = (c+1)^3$ DOES also have other families with degree higher than the cubic polynomials in this post.

However, this just scratches the surface, there is plenty of life left in this question.

Old Peter
  • 1,357
  • 1
    +1 Beautiful! I especially love the simplicity of the first term of family 4, and how families 4 and 5 have the same third term $c$. – Tito Piezas III Nov 07 '23 at 01:28
  • 1
    @TitoPiezasIII, that is most kind of you to say so, it inspires me to continue. – Old Peter Nov 07 '23 at 11:54
  • 1
    There seems to be a connection between Family 2 and 4. Note that Family 2 is, $$\color{blue}{(3n^2)}^3 + (6n^2 + 3n + 1)^3 - \color{blue}{(9n^3 + 6n^2 + 3n + 1)}^3 = -z^3$$ while Family 4 is, $$\big(3n\color{blue}{(3n^2)}\big)^3 + \big((3n)\color{blue}{(9n^3 + 6n^2 + 3n + 1)} + 1\big)^3 + c^3 = (c + 1)^3$$ though it could be just coincidence. – Tito Piezas III Nov 07 '23 at 12:12
  • Of the the 4405 solutions to $a^3+b^3+c^3=(c+1)^3$, your data says that about 600 have the $a$-variable of form either $3n^2$ or $9n^3$. It might be informative to check other solutions which have the $a$-variable as powers or near-powers. – Tito Piezas III Nov 07 '23 at 12:19
  • What is $-z^3$ please? – Old Peter Nov 07 '23 at 14:12
3

Here is a further parametric solution, call it Family 6:

$\qquad a=27n^4+54n^3+45n^2+19n+3$

$\qquad b=54n^4+81n^3+63n^2+23n+4$

$\qquad c=243n^6+567n^5+675n^4+468n^3+201n^2+49n+5$

$\qquad c+1 = 243n^6+567n^5+675n^4+468n^3+201n^2+49n+6$

Alternatively,

$\qquad a= (3n + 1)(9n^3 + 15n^2 + 10n + 3)$

$\qquad b= (3n + 1)^2(6n^2 + 5n + 3) + 1$

$\qquad c= (3n + 1)^2(9n^2 + 9n + 5)(3n^2 + 2n + 1) $

$\qquad c+1 = (3n + 1)^2(9n^2 + 9n + 5)(3n^2 + 2n + 1) + 1$

It intersects with Families 1 and 3 when $n=0$ and $n=-1$ respectively.


I found this solution by starting from the general parametric integer solution of $x^3+y^3+z^3=t^3$ due to Choudhry (1) and searching for a condition on the parameters that would ensure the necessary difference of $1$. Choudhry’s solution (with a change of letters) is:

$\qquad sx=r(-p^3-q^3+r^3)$

$\qquad sy=-p^4+2p^3q-3p^2q^2+2pq^3-q^4+(p+q)r^3$

$\qquad sz=p^4-2p^3q+3p^2q^2-2pq^3+q^4+(2p-q)r^3$

$\qquad st=r(p^3+(p-q)^3+r^3)$

To limit the search to manageable proportions, I set $r=1$ and looked for solutions with $x,y < 0$, $z,t > 0$ and $-y+1=z$. The point of the latter is that $x^3+y^3+(-y+1)^3=t^3$ can be rearranged as $(-x)^3+t^3+(-y)^3=(-y+1)^3$ with all terms positive, while $sy+sz=3pr^3=s(y-y+1)=s$, so (with $r=1$) the required common factor $s$ must equal simply $3p$. Aiming for $t=z+1$ looks more complicated.

For $3p$ to be a common factor it suffices that $3|p$ and $3p|(q^3-1)$. This can be seen by re-writing Choudhry’s solution (with $r=1$) as below:

$\qquad sx=-p(p^2)-(q^3-1)$

$\qquad sy= -p^2(p^2-2pq+3q^2)+3pq^3-(p+q)(q^3-1)$

$\qquad sz=p^2(p^2-2pq+3q^2)-(2p+q)(q^3-1)$

$\qquad st=2p(p^2)+3pq(-p+q)-(q^3-1)$

To satisfy $3|q^3-1$ it suffices to put $q=3n+1$ implying $q^3-1=27n^3+27n^2+9n$. Putting $p=(27n^3+27n^2+9n)/3n=9n^2+9n+3$ we then have both $3|p$ and $3p|q^3-1$.

It is then straightforward to substitute for $p$ and $q$ in the above rewritten solution and make the identification $(a,b,c,c+1)=(-x,t,-y,z)$ yielding the parametric form for Family 6 above.

This method can perhaps be adapted to find further families of solutions.

Reference:

  1. Choudhry A (1998) On Equal Sums of Cubes Rocky Mountain Journal of Mathematics Vol 28(4) p 1256
Adam Bailey
  • 4,197
  • An excellent post!

    I confirm that for $$n= 1$ to $6$ this matches my results.

    – Old Peter Nov 08 '23 at 19:41
  • +1 Great! With 6 families already, there could be infinitely many. And if there is, the polynomials don’t grow as fast as the ones for $x^3+y^3+z^3 =1$, so if there is a recursion, it would be simpler. – Tito Piezas III Nov 09 '23 at 00:19
  • @Adam There are infinitely many polynomial solutions of form cubic/quartic, distinct from your quartic/quartic. See my third answer to Peter's post. – Tito Piezas III Nov 18 '23 at 04:06
2

Partial answer

First, a prime must be chosen. We pick $7$. Then $7$ is decomposed the following way:
$7=1+6=2+5=3+4$
This decomposition is unique to this prime. In this case we need to consider the $3$ cases and find possible soltions. I chose to start with $3+4$ because it's a solution in the first family.
We are looking for a solution of the form:

$3^3 + 4^3 = a^3 - b^3= (a-b)(a^2 +ab + b^2)$

Since $a-b=1$ or $a=b+1$, the equation becomes: $3^3 + 4^3 =a^2 + ab +b^2$ which after substitution of $a$ by $b+1$ becomes: $3^4 + 4^3 =91=3b^2 + 3b +1$, a simple quadratic equation.

$3b^2 + 3b +1= 91$ becomes $3b^2 + 3b +90 =0$ or $b^2 +b - 30=0$, basically a triangular number in disguise.
The solutions are $b=5$ and $b=-6$. Using this values we can write the solution involving both $a$ andd $b$ as:

$3^3 + 4^3 + 5^3 = 6^3$ which is the first equation of the first family.

Next we consider the decomposition $7=2+5$. The equation in $b$ is given by:
$b^2 + b -44 = 0$. The determinant $\Delta$ is not a square in this case so there will not be an integer solution for $(a,b)$.

Last, the decomposition $1+6$ is considered. Following the same steps above, we end up with the quadratic: $b^2 + b - 216=0$ whose solutions are $8$ and $-9$. The equation becomes: $1^3 +6^3 + 8^3 =9^3$ which is the first solution of the third family.

We can consider any prime and do a decompisition and find the corresponding cubic forms. For example, if we consider the prime $37$, we will end up with the following decompositions:
$37=3+34$ and $14+23$.
These decompositions will give us the third and first solution of the third family and fourth family.
If we consider the prime $67$, we will end up, among many solutions, with the $3$ following solutions of the forth family namely:

$21 + 46=67$
$16 + 51=67$
$9 + 58 =67$

This method doesn't distinguish between the families because it uses a quadratic equation to find the solutions. Here again, the triangular numbers keep showing up when they are the least expected.

In fact the method can be generalized to composites to solve $a^3 + b^3 +c^3= d^3$ with $d=c+2$, or $d=c+4$...We will give only one example.

The equation $a^3-b^3=(a-b)(a^2 +ab + b^2)$ becomes:

$2[(b+2)^2 +(b(b+2) +b^2]=6b^2 +12b + 8$

If we take the first equation of the first family, we get:

$3^3 + 5^3 = 152 = 6b^2 + 12b +8$ which becomes:

$6b^2 + 12b -144 =0$

a quadratic with the solutions $b=-6,4$. This equation corresponds to the decomposition of $8$ into $3+5$. Other decompositions of 8 may or may not admit integer solutions.

So again, we see that the method can handle other cases than $d=c+1$ and can use primes or composites as a starting point. The choice of $8$ in the previous example was used just to show that the method works. In general, the starting point is always an integer and its decomposition into a series of two numbers $N=x_{1} +x_{2}$ then using a quadratic derived from the condition $d = c+i$, with $i=1,2,3...$

user25406
  • 1,048
  • 1
  • 9
  • 16
  • 1
    A necessary (but not sufficient) condition for this to work is that the chosen prime is of the form $6n+1$. This is a consequence of the fact that if $a^3+b^3+c^3=d^3$ then $6 | (a+b+c-d)$. – Adam Bailey Sep 30 '23 at 14:22
  • @AdamBailey, in fact the method can be generalized to composites and to having c+2=d, c+3=d, ... For example, we can see that $6^3 - 4^3 =152 =8*19$. I will add this to the answer when I have a bit more time. – user25406 Sep 30 '23 at 18:06
  • 1
    @user25406. I like this method and intent to study it further when I can find the time. – Old Peter Sep 30 '23 at 19:06
  • 1
    @user25406: Yes, it has also occurred to me to express the equation as $a^3+b^3 = (c+1)^3-c^3$ so both sides factor. But I find it hard to generate polynomial solutions from this approach. – Tito Piezas III Oct 01 '23 at 10:00
  • @TitoPiezasIII, what is the motivation and the advantage of generating polynomial solutions. Can you provide references on how to generate polynomial solutions? – user25406 Oct 01 '23 at 11:04
  • 1
    @user25406 As an example of the usefulness of polynomial solutions, if you have two simultaneous quadratics to be made squares, you need a polynomial solution for one, then the problem reduces to solving an elliptic curve as in this post. As an example on how to generate them using Pell equations, like polynomial solutions to $x^3+y^3+z^3=1$, see this post. And I also suggest Dickson's book "History of the Theory of Numbers" for a more enlightening exposition on such matters. – Tito Piezas III Oct 01 '23 at 12:03
  • @user25406: What about {25, 38, 87, 90}? There is no prime you can produces this solution from. – azerbajdzan Oct 22 '23 at 21:10
  • It's not necessary to have a prime, sometimes you have a composite with few factors. I noticed that when we are dealing with composite numbers with more than two factors. Also, it's not always the smallest prime or composite that needs to be decomposed. I couldn't find a general rule. The example of $8$ with a decomposition into $3+5$ above shows that a prime is not always needed. – user25406 Oct 22 '23 at 21:18
  • The example you gave with $ {25,38, 87, 90} $ is a case when $d=c+3$ since $90=87+3$ so the quadratic equation involved is different from the case when $d=c+1$. I will try to work that example tomorrow and post the result as a comment if possible. – user25406 Oct 22 '23 at 21:23
  • Why I do not get notification on a new comment? Is there any problem with the site at the moment? It worked for me all the time... – azerbajdzan Oct 22 '23 at 21:52
  • I don't have any idea why you didn't get notified. – user25406 Oct 23 '23 at 10:45
  • @AdamBailey Peter found two new families. A very welcome development. – Tito Piezas III Nov 08 '23 at 14:40
1

Numerical solutions of $A^3+B^3+C^3=(C+1)^3$ arranged by $C$. There are nine pairs curiously with the same $C$, namely (255, 477, 1331, 8898, 13088, 14526, 18755, 20154, 22968), a sample of which,

\begin{align} 51^3+82^3+477^3 &= 478^3\\ 4^9+75^3+477^3 &= 478^3\\[4pt] 9^9 + 502^3 + 13088^3 &= 13089^3\\ 16^3 + 801^3 + 13088^3 &= 13089^3\\[4pt] 7^9 + 840^3 + 14526^3 &= 14527^3\\ 294^3 + 847^3 + 14526^3 &= 14527^3\\[4pt] 10^9 + 381^3 + 18755^3 &= 18756^3\\ 1018^3 + 69^3 + 18755^3 &= 18756^3 \end{align}

(3,4,5,6) (1,6,8,9) (3,10,18,19) (2,17,40,41) (12,19,53,54) (14,23,70,71) (12,31,102,103) (3,34,114,115) (21,46,188,189) (27,46,197,198) (16,51,213,214) (4,57,248,249) (22,57,255,256) (9,58,255,256) (15,64,297,298) (27,64,306,307) (58,75,453,454) (5,86,460,461) (64,75,477,478) (51,82,477,478) (20,89,487,488) (48,85,491,492) (57,82,495,496) (61,90,564,565) (66,97,632,633) (58,105,671,672) (48,109,684,685) (6,121,768,769) (91,120,909,910) (23,140,958,959) (75,136,989,990) (7,162,1190,1191) (75,166,1290,1291) (44,173,1324,1325) (136,141,1331,1332) (37,174,1331,1332) (76,171,1346,1347) (8,209,1744,1745) (108,199,1745,1746) (96,205,1779,1780) (149,212,2068,2069) (187,192,2130,2131) (108,235,2178,2179) (148,225,2208,2209) (66,247,2262,2263) (9,262,2448,2449) (118,279,2790,2791) (147,274,2813,2814) (10,321,3320,3321) (52,321,3327,3328) (147,316,3402,3403) (22,327,3414,3415) (248,275,3466,3467) (18,349,3764,3765) (48,355,3866,3867) (131,356,3973,3974) (67,372,4154,4155) (28,375,4193,4194) (192,361,4247,4248) (11,386,4378,4379) (270,343,4473,4474) (241,366,4583,4584) (146,395,4645,4646) (192,409,5016,5017) (232,399,5033,5034) (145,426,5175,5176) (335,368,5398,5399) (131,446,5506,5507) (12,457,5640,5641) (243,460,6101,6102) (165,478,6156,6157) (49,522,6888,6889) (243,514,7074,7075) (13,534,7124,7125) (305,506,7255,7256) (24,547,7386,7387) (225,538,7463,7464) (426,451,7506,7507) (376,495,7625,7626) (205,558,7796,7797) (246,571,8186,8187) (456,475,8205,8206) (97,594,8376,8377) (300,571,8429,8430) (448,519,8750,8751) (421,540,8795,8796) (14,617,8848,8849) (115,618,8898,8899) (72,619,8898,8899) (261,616,9156,9157) (314,605,9172,9173) (434,557,9211,9212) (90,643,9426,9427) (139,648,9570,9571) (300,631,9630,9631) (361,636,10071,10072) (114,673,10104,10105) (282,667,10314,10315) (100,693,10548,10549) (485,614,10732,10733) (15,706,10830,10831) (549,598,11244,11245) (363,694,11285,11286) (542,611,11362,11363) (513,634,11399,11400) (564,595,11402,11403) (493,654,11540,11541) (216,775,12590,12591) (363,760,12738,12739) (587,674,13018,13019) (502,729,13088,13089) (16,801,13088,13089) (190,801,13175,13176) (289,798,13320,13321) (264,823,13854,13855) (319,822,13998,13999) (333,820,14003,14004) (288,835,14213,14214) (627,712,14229,14230) (343,840,14526,14527) (294,847,14526,14527) (432,829,14723,14724) (318,865,15048,15049) (17,902,15640,15641) (621,802,15867,15868) (415,888,16038,16039) (432,901,16452,16453) (390,913,16536,16537) (467,902,16690,16691) (82,951,16937,16938) (216,997,18267,18268) (408,985,18471,18472) (18,1009,18504,18505) (547,954,18546,18547) (381,1000,18755,18756) (69,1018,18755,18756) (507,976,18797,18798) (522,979,18978,18979) (655,948,19433,19434) (619,978,19770,19771) (589,990,19787,19788) (643,984,20154,20155) (594,1003,20154,20155) (808,885,20171,20172) (305,1064,20272,20273) (618,1009,20520,20521) (673,990,20616,20617) (507,1054,20826,20827) (58,1095,20921,20922) (648,1015,20958,20959) (367,1086,21057,21058) (774,967,21353,21354) (45,1114,21467,21468) (19,1122,21698,21699) (496,1095,21870,21871) (823,990,22566,22567) (7,1152,22574,22575) (798,1015,22758,22759) (847,984,22806,22807) (865,978,22968,22969) (822,1009,22968,22969) (840,1003,23106,23107) (913,954,23304,23305) (888,979,23370,23371) (588,1135,23561,23562) (911,980,23785,23786) (535,1158,23846,23847) (411,1192,24242,24243) (20,1241,25240,25241) (370,1251,25874,25875) (588,1219,25914,25915)

If anybody could remind me how to produce LF I'll do it.

Old Peter
  • 1,357
  • OldPeter: I removed the latex format so it loads faster. I hope you don't mind. These are the 171 smallest primitive solutions? – Tito Piezas III Oct 01 '23 at 09:31
  • @TitoPiezasIII, thanks for your help, it’s a long time since I used this markup language. An offshoot question has occurred to me; are you ok if I post it? – Old Peter Oct 01 '23 at 13:50
  • Sure, feel free to post your question. Perhaps put a link to this post though. – Tito Piezas III Oct 01 '23 at 15:16
  • I think I found a possible explanation for when we get pairs with the same $c$. I am taking the equations $22^3 +57^3 + 255^3 = 256^3$ and $9^3 + 58^3 +255^3 = 256^3$. Using the quadratic equation from my answer below, we end up with the equation: $b^2 + b - 65280 = 0$ with solutions $b=255,-256$ confirming that the decomposition of $22 + 57 = 79$ and $9+58 = 67$ produce the same $c$ and $c+1$. Now $256^3 - 255^3 = 195841$ with $195841 = 37* 67*79$. Notice that both $67$ and $79$ are factor of $195841$. I think that is the reason why we get a pair with the same $c$. (cont) – user25406 Oct 01 '23 at 15:48
  • (Cont) - Though I have not checked other solutions (because I do everything by hand with a small calculator). I suspect a code can check quickly if other solutions will also show that common factors of $(c+1)^3 -c^3$ is the reason ( sorry I can't code). – user25406 Oct 01 '23 at 15:53
  • In fact, it's enough to take the difference of $(c+1)^3 - c^3$ and factor it. So the next pair is $478^3 -477^3 = 684019$ which factors into $71937139=37133139$. Again we see the common factor $139= 4^3 +75$ and $133 = 719 = 51 + 82 $. I still don't know what role the third factor $37$ plays if any. – user25406 Oct 01 '23 at 16:12
  • I think that if the difference between $(c+1)^3 -c^3$ has more than $2$ prime factors though it's hard to guess how many more, we may end up with pairs with the same $c$. For example, $13089^3 -13088^3=71943731231 – user25406 Nov 08 '23 at 18:59
0

We explore solving the equation for a given value of $c$.

Start by expanding the binomial power $(c+1)^3$ and rearranging:

$a^3+b^3=3c^2+3c+1$

Let $3c^2+3c+1=pq$ for any divisor $p$ of $3c^2+3c+1$. Then we have

$a+b=p$

$a^2-ab+b^2=(a+b)^2-3ab=q$

and thus

$ab=(p^2-3q)/3.$

Factors of $3c^2+3c+1$ are all $\equiv1\bmod3$, so $ab$ will properly be an integer. But $p$ must be chosen large enough to render this product positive yet small enough for the discriminant

$\Delta=(a+b)^2-4ab=(a-b)^2$

to be nonnegative. Thus

$(3c^2+3c+1)^{1/3}\le p\le[(3/4)(3c^2+3c+1)]^{1/2}.$

Solutions are then obtained for those cases where $\Delta$ is a square and thus $a-b$ may be rendered a whole number.

Let us drop in $c=70$ as an example. Then $3c^2+3c+1=14911=13×31×37$, and we have potenially eight choices for $p$ (and thus $q$). But $p=13$ is too small to render a positive value for $ab$ and $p=13×31$ is too large giving a negative $\Delta$. Therefore the only possible solutions are given by $p=31$ or $p=37$.

For $p=31$ we calculate:

$q=13×37=481$

$ab=(31^2-481)/3=160$

$\Delta=31^2-(4×160)=321.$

The discriminant, divisible only by $3$ and not by $9$, fails to be a square.

For $p=37$, however:

$q=13×31=403$

$ab=(37^2-403)/3=322$

$\Delta=37^2-(4×322)=81=9^2.$

So we accept $a+b=p=37$ and $a-b=9$, giving the quoted solution $23^3+14^3+70^3=71^3$.

Oscar Lanzi
  • 39,403
  • Your answer is the same as my partial answer above. We end up having to solve a quadratic equation in $c$. – user25406 Nov 08 '23 at 19:04
  • We have a slight difference in sequencing; you sort solutions accodring to $a+b$ and I sort them according to $c$. – Oscar Lanzi Nov 08 '23 at 20:50
  • not really. I used $a+b$ for the formula $a^3 -b^3$ not the $a,b$ of the expression $a^3 +b^3 +c^3 =(c+1)^3$. I know, it's a bit confusing but the formula on the web provided $a^3-b^3=...$ I didn't bother changing it to $(c+1)^3 -c^3$. – user25406 Nov 08 '23 at 20:56
  • Here another answer of mine with the "same sequencing" as yours on mathoverflow last October. https://mathoverflow.net/questions/413031/how-many-cubes-are-the-sum-of-three-positive-cubes/456151?noredirect=1#comment1181358_456151 – user25406 Nov 09 '23 at 11:10
  • I got lost when you wrote $ab=(p^2-3ab)/3$ shouldn't it be $ab=(p^2-q)/3$? since you got that $a^2-ab+b^2=(a+b)^2-3ab=p^2-3ab=q\ $ and where does the discriminant $\Delta=(a+b)^2-4ab=(a-b)^2$ comes from? I don't see something like $ax+(a+b)x+b=0$ – Dabed Nov 09 '23 at 13:48
  • 1
    Mistyped, sorry. – Oscar Lanzi Nov 09 '23 at 13:51
  • I mistyped too I wrote $ax+(a+b)x+b=0$ but I meant $ax^2+(a+b)x+b=0$ – Dabed Nov 09 '23 at 13:56