a/b + b/c + c/a = n
Dean Hickerson
dean at math.ucdavis.edu
Sat Jul 12 20:37:56 CEST 2003
Hans Havermann asked:
> Can it be shown that, as a minimum, all solutions require (a*b*c) to be
> a cube?
Yes: Without loss of generality, gcd(a,b,c)=1. Let d=gcd(b,c),
e=gcd(c,a), and f=gcd(a,b). Then d, e, and f are relatively prime in
pairs. Also, a is divisible by e and f, so it's divisible by ef; let
a = e f A. Similarly, let b = f d B and c = d e C.
We have d = gcd(b,c) = gcd(f d B, d e C) = d gcd(f B, e C), so B is
relatively prime to both e and C. Similarly, C is r.p. to both f and A,
and A is r.p. to both d and B. In other words, A, B, and C are r.p. in
pairs, and gcd(A,d) = gcd(B,e) = gcd(C,f) = 1.
Also,
2 2 2 2 2 2
A B d f + B C e d + C A f e
a/b + b/c + c/a = ---------------------------------
A B C d e f
So A divides B C^2 e d^2, but is relatively prime to B, C, and d, so
A divides e. Also, e divides A B^2 d f^2, but is r.p. to d, f, and B,
so e divides A. Hence A=e. Similarly, B=f and C=d. Hence:
a = e^2 f, b = f^2 d, c = d^2 e,
so a b c = (d e f)^3. Also,
3 3 3
d + e + f
a/b + b/c + c/a = ------------.
d e f
Now we can look for integer values of a/b + b/c + c/a more efficiently:
Look for triples (d,e,f) with 1 <= d <= e <= f, relatively prime in pairs,
such that d^3 + e^3 + f^3 is divisible by d e f. For each such triple, we
get two triples (a,b,c): (e^2 f, f^2 d, d^2 e) and (e f^2, d f^2, e d^2).
Incidentally, the first few values of n are given by sequence A072716:
%S A072716 3,5,6,9,10,13,14,17,18,19,21
%N A072716 Integers expressible as (x^3 + y^3 + z^3)/xyz with positive
integers x, y, and z. (Alternatively, the integers expressible
as x/y + y/z + z/x with positive integers x, y, and z.)
...
%A A072716 Tadaaki Ohno (t-ohno(AT)hyper.ocn.ne.jp), Aug 07 2002
Apparently the author doesn't know if 22 is in the sequence.
Dean Hickerson
dean at math.ucdavis.edu
More information about the SeqFan
mailing list