I would like to find the solutions to $$i)\qquad a(a+b+c)=bc \\ii)\qquad a(a+b+c)=2bc \\iii)\qquad a(a+b+c)=3bc$$ for $0< a \le b \le c$ and of course: $\textrm{gcd}(a,b,c) = 1$ (since those are the interesting cases).

To be honest, Diophantine equations seem like a black art. Guess and check has always been the only sure method that I can understand.

At least the last case ($iii$) seems to be "easy" as the only solution is apparently $ a = b= c=1$. I don't see how to prove this though.

For $i$, I've found solutions such as: $\{\{1,2,3\}, \{2,3,10\}, \{3,4,21\}, \{3,5,12\}, \ldots\}$

And For $ii$, I've found solutions such as: $\{\{1,1,2\}, \{4,5,6\}, \{15,20,21\}, \{20,22,35\}, \ldots\}$

Notwithstanding my ability to exhaust over integers, I don't have any deeper understanding of what is going on here mathematically. I would really like to understand these better and would appreciate some help! Thanks in advance!

  • $\begingroup$ these are indefinite ternary quadratic forms (put everything one one side, set equal to zero); if there are any solutions in integers, there are infinitely many. The inequalities are not a natural part of the setup, and may reduce to finitely many solutions. $\endgroup$ – Will Jagy Dec 10 '14 at 23:34
  • $\begingroup$ I guess the inequalities are not really needed. I'll search on "indefinite ternary quadratic forms." That is a mouthful. $\endgroup$ – amcalde Dec 10 '14 at 23:35
  • $\begingroup$ The diophantine equation $a(a+b+c)=nbc$ is equivalent to $$a^2(1/n+1/n^2)=(b-a/n)(c-a/n).$$ $\endgroup$ – Ian Mateus Dec 11 '14 at 0:38
  • $\begingroup$ It's quite easy to give a parametric solution to $a(a+b+c)=nbc$, but the tricky part is the OP wants $0<a<b<c$ which restricts the solution space. I'll give a partial answer below. $\endgroup$ – Tito Piezas III Dec 11 '14 at 0:54

One can give a parameterization to,

$$a(a+b+c) = nbc\tag{1}$$

However, since you want $0<a<b<c$, this makes it trickier.

For $n=1$: A look at your solutions and one can see that $a,b$ are in arithmetic progression. Hence,

$$a,\;b,\;c = n,\;n+1,\;2n^2+n$$

For $n=2$: Most of your $b,c$ differ by one. Hence,

$$a,\;b,\;c = 2pq,\;p^2+pq-1,\;p^2+pq$$

where $p,q$ solve the Pell equation $p^2-3q^2=1$. Since $pq<p^2$, this guarantees that $a<b$.

For $n\ge3$: I checked this with Mathematica and, curiously, there doesn't seem to be any.

P.S. Without the inequalities, then $(1)$ has a two-parameter solution,

$$a,\;b,\;c = (-u + n v)u, \; (u + v)u, \; (-u + n v)v$$

Edit: This answers the comment below. First, we do the substitution $b,\;c = x+y,\;x-y\,$, on $(1)$ then solve for $a$,

$$a =-x\pm\sqrt{(n+1)x^2-ny^2}$$

Thus we have to solve, as W. Jagy pointed out, a ternary quadratic form,


and initial solution of which is $x,y,z =1,1,1$. Given an initial solution, we generally can find an infinite more. Two methods are described here in Form 2b.

| cite | improve this answer | |
  • $\begingroup$ I am most interested in your PS. Like I mentioned in my post, Diophantine equations seem opaque to me. How did you arrive at your two-parameter equation?? $\endgroup$ – amcalde Dec 11 '14 at 15:00
  • $\begingroup$ @amcalde: I've modified my answer... $\endgroup$ – Tito Piezas III Dec 11 '14 at 15:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.