6

$$S=\{a^3+b^3+c^3-3abc|a,b,c\in\Bbb Z\}$$

Can we decide $S$? that is, we want to find all integers of the form $a^3+b^3+c^3-3abc$.

obviously,

  1. if $m,n\in S$, then $mn\in S$, so we only need to consider primes;
  2. if $n\in S$, then $-n\in S$.

Let $f(a,b,c)=a^3+b^3+c^3-3abc$. $f(0,0,0)=0$, $f(1,0,0)=1$, $f(1,1,0)=2$, we get that $0,1,2\in S$

p.s. I find a solution for $a,b,c \geq0$: $n=2^rp_1^{r_1}\dotsb p_s^{r_s}$, $p_1< p_2<...$ are odd primes, then a suficient and necessary condition for $n$ can be expressed as $a^3+b^3+c^3-3abc$($a,b,c \geq0$) is $p_1>3$ or $p_1=3$ with $r_1\geq2$

Parcly Taxel
  • 103,344
ziang chen
  • 7,771
  • 1
    For your claim in part 1, just because $m$ and $n$ are in $S$ implies $mn$ is in $S$ does not imply the converse - that if $mn$ is in $S$ then $m$ and $n$ are also in $S$. And, in fact, 1 is not true. Consider $m=3$ and $n=9$. – Foo Barrigno Mar 24 '14 at 17:28
  • 1
    I think there is a nice factorization lurking around... – Ian Mateus Mar 24 '14 at 17:28
  • I've seen this problem a long time ago. If only I could remember the solution... – user2345215 Mar 24 '14 at 17:31
  • 2
    You might find it useful that $a^3+b^3+c^3-3abc=(a+b+c)(a^2+b^2+c^2-ab-ac-bc)=\frac{1}{2}(a+b+c)((a-b)^2+(b-c)^2+(c-a)^2)$. – user134824 Mar 24 '14 at 17:31
  • @FooBarrigno we only need to find all primes of the form $a^3+b^3+c^3-3abc$ – ziang chen Mar 24 '14 at 17:33
  • http://math.stackexchange.com/questions/722878/how-find-this-all-positive-integer-number-n-x3y3z3-3xyz – lab bhattacharjee Mar 24 '14 at 17:51
  • compare http://math.stackexchange.com/questions/329936/primes-represented-integrally-by-a-homogeneous-cubic-form AND http://math.stackexchange.com/questions/336191/numbers-represented-by-a-cubic-form – Will Jagy Mar 24 '14 at 18:45
  • Would you mind showing the obviousness of 1? It's true (from looking at the accepted answer) but I don't see what your obvious way of seeing it is.

    Here's an answer to my question, but it's certainly not obvious: https://math.stackexchange.com/questions/1860381/given-integers-m-n-find-integers-a-b-c-such-that-a3b3c3-3abc-m-n?rq=1

    – Geoffrey Sangston May 24 '17 at 19:47

2 Answers2

8

Note that $$(a\pm1)^3+a^3+a^3-3(a\pm1)a^2=3a^3\pm3a^2+3a\pm1-3a^3\mp3a^2=3a\pm1$$ and $$(a+1)^3+a^3+(a-1)^3-3(a+1)a(a-1)=3a^3+6a-3a^3+3a=9a$$ Suppose $n=a^3+b^3+c^3-3abc$ where $n\in\mathbb Z$.

So we can get all $n$ such that $3\nmid n$ or $9\mid n$.

Note that $(3p+k)^3\equiv k^3\equiv k\ \ (\text{mod 9})$ for $k\in\{-1,0,1\}$. We can write $a=3p+k$, $b=3q+l$ and $c=3r+m$, where $k,l,m\in\{-1,0,1\}$. Then $$a^3+b^3+c^3-3abc\equiv k+l+m-3(3p+k)(3q+l)(3r+m)\equiv k+l+m-3klm\ \ (\text{mod }9)$$ Additionally $$k+l+m-3klm\equiv k+l+m\ \ (\text{mod 3})$$ Suppose $3\mid n$. Then $3\mid k+l+m$.

If $k=0$, $l=0$ or $m=0$ the other two must be opposite, so $k+l+m-3klm=0$ and $9\mid n$. Else we can only have $k,l,m=-1$ or $k,l,m=1$. But in both of these cases we get $9\mid n$ too.

We can conclude the only possible values are such $n\in\mathbb Z$ that $3\nmid n$ or $9\mid n$.

user2345215
  • 16,422
  • 1
    Nice answer. Overkill for determining algorithmic decidability of the set $S$ but it definitely gives a much simpler algorithm for deciding! – user2566092 Mar 24 '14 at 18:45
  • @user2566092, compare http://math.stackexchange.com/questions/329936/primes-represented-integrally-by-a-homogeneous-cubic-form and http://math.stackexchange.com/questions/336191/numbers-represented-by-a-cubic-form – Will Jagy Mar 24 '14 at 19:09
  • @WillJagy Those questions are asking for formulaic characterizations, not determining decidability. However I have the feeling maybe the OP meant "determine" rather than "decide" in this question which is unfortunate for my posted answer, because strict (algorithmic) decidability in Diophantine equations is typically much easier to establish based on prior works, rather than coming up with a formula or a simple characterization for solutions to a particular equation. – user2566092 Mar 24 '14 at 19:41
  • @user2566092, right. It is a bit of guesswork when someone posts what turns out to be a problem for contest training (or an old contest) and you don't know what the level or intent might be. I really like the "formulaic" type, I thought you and user2345215 might enjoy looking at those, although already answered, here or on MO. – Will Jagy Mar 24 '14 at 19:45
  • 2
    $a^3+b^3+c^3-3abc=\frac12(a+b+c)(3a^2+3b^2+3c^2-(a+b+c)^2)$, if $3| (a^3+b^3+c^3-3abc)$, then $3| (a+b+c)$, so $9| (a^3+b^3+c^3-3abc)$ – ziang chen Mar 25 '14 at 09:48
2

Using the factorization $(a + b + c)(a^2 + b^2 + c^2 - ab - ac - bc)$, you can set equal to a given number $n$ and then use the linear substitution $a = m - b - c$ where $m$ divides $n$, and then you get a quadratic Diophantine equation in two variables from $a^2 + b^2 + c^2 - ab - ac - bc = n/m$. According to http://www.cs.nyu.edu/pipermail/fom/2006-December/011182.html , a quadratic is decidable over ${\mathbb Z}$. There are only finitely many possibilities for $m$ for a given $n$, so yes, the set $S$ of integers which have an integer solution to your equation is decidable in the algorithmic sense. See also https://mathoverflow.net/questions/51987/which-types-of-diophantine-equations-are-solvable, where the following quote is found:

"In a 1972 paper, C. L. Siegel [Nachr. Akad. Wiss. Göttingen Math.-Phys. Kl. II 1972, 21-46; MR0311578 (47 #140)] constructed an algorithm to determine whether an arbitrary quadratic equation had integer solutions."

user2566092
  • 26,142
  • 1
    Can down-voters explain why they down-voted? It really does look like Siegel showed quadratics are decidable over ${\mathbb Z}$ and the rest of the argument is trivial. – user2566092 Mar 24 '14 at 18:03
  • See my future answer. (I think this would be more appropriate as a comment) – user2345215 Mar 24 '14 at 18:05
  • @user2345215 I suppose you're implying that you're going to give some explicit characterization of the set $S$ that is more illuminating than the original definition. However the question was about "decidability" which has a strict algorithmic interpretation in the Diophantine equation literature, so my answer is valid and I don't understand the down-vote. – user2566092 Mar 24 '14 at 18:32
  • Well I assumed he just meant determine by decide... I can see your point. Too bad it's now locked. – user2345215 Mar 24 '14 at 18:41