MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Is there a solution to: $a+b^m=b+c^n=c+a^l$ for l,m,n >1 and a, b, c distinct odd primes?

I've had a play around with specific possible solutions and there are lots of possibilities that may be systematically eliminated but I cannot see any obvious way to progress beyond specific cases. Is there any area of research that might be able to shed light on this, or is it a known result?

share|cite|improve this question
If you keep a, b and c fixed, log a/log b is well approximated by m/l. Making that a linear form in logarithms, that should force l and m to be fairly big. Which is less plausible than solutions at random for small values. This is an exponential Diophantine equation, but not hopeless. – Charles Matthews Jun 1 '10 at 14:24
I added tag "diophantine-equations". A nice sourse for exponential DEs is the book of Shorey and Tijdeman. I've not seen your equation, but as Charles says it is hardly hopeless. But most probably routine. – Wadim Zudilin Jun 1 '10 at 14:51

Here is somewhat interesting feature of such primes.

Without loss of generality, there are two cases to consider:

1) If $a < b < c$ then $a^{\ell} < c^n < b^m$ and thus $$0 < c^n - a^{\ell} = c-b < c-a$$

2) If $b < a < c$ then $a^{\ell} < b^m < c^n$ and thus $$0 < c^n - b^m = a-b < c-b$$

That is, two primes out of $a,b,c$ must be such that the difference between their non-trivial powers is strictly smaller than their own difference.

share|cite|improve this answer
Thanks. That is an interesting way of demonstrating how rare such numbers must be. – Francis Davey Jul 17 '10 at 6:41

for above question,Is there a solution to: $a+b^m=b+c^n=c+a^l$ for l,m,n >1 and a, b, c distinct positive integers?

share|cite|improve this answer
Yes, that would be a more unconstrained question, which I also don't know the answer to. – Francis Davey Jun 23 '10 at 7:19

assume that $a>b>c$,and $a,b,c$ are distinct odd primes,we write above equation to form:

$c^n=(b^{m/2}+i(a-b)^{1/2})(b^{m/2}-i(a-b)^{1/2})$ and


first equation: similar to $z[i]$,if $gcd(b^{m/2}+i(a-b)^{1/2},b^{m/2}-i(a-b)^{1/2})=d$,since

$NORM(2b^{m/2})=4b^m $and $NORM(b^{m/2}+/-(a-b)^{1/2})=b^m+(a-b)=c^n$ ,then $d=1$

so $b^{m/2}+i(a-b)^{1/2}=(a_1+ib_1)^n$ and $b^{m/2}-i(a-b)^{1/2}=(a_1-ib_1)^n$ ,so

$c=a_1^2+b_1^2$ and also $b^{m/2}+i(a-b)^{1/2}=r^n(cos(nx)+isin(nx))$ that

$r^2=a_1^2+b_1^2$,and$tan(x)=b_1/a_1$,then $b^{m/2}=c^{n/2}cos(nx)$ and $(a-b)^{1/2}=c^{n/2}sin(nx)$

with similar method from second equation ,we have:

$c^{n/2}=a^{l/2}cos(ly)$ and $(b-c)^{1/2}=a^{l/2}sin(ly)$

if we assume that $cos(nx)=cos(ly)$ so $c^{2n}=b^ma^l$ ,this is contradiction so above

equation has not any solution.

share|cite|improve this answer
Thanks. I'll have to go away and think about that. I'm not sure what you mean by "first equation: similar to z[i]". – Francis Davey Aug 20 '10 at 7:36
First question - are you assuming that m, n are even? If they are not then $b^m/2$ won't be an integer. Second, where does the assumption that l, m and n are greater than 1 enter? – Francis Davey Aug 20 '10 at 11:02
m,n,l are greater than 1,and $b_1$ and $a_1$ may be irrational or rational numbers,$z[i]$ is about integers,properties of this method is similar to $z[i]$,not the same – Hashem sazegar Aug 20 '10 at 11:27

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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