May I have a hint for this gcd problem?

172 Views Asked by At

I am trying to prove the following:

$$(2^a - 1, 2^b - 1) = 2^{(a,b)} - 1 \ \ \ \forall a,b \in \mathbb{N} $$ Where: $$ (a,b) := \gcd(a,b) $$

So far I have tried dividing $2^a-1$ by $2^b-1$ to see if there was anyway to get the remainder, and user the remainder to help inductively show what the $gcd$ tends toward, but all I could do through algebraic manipulation was a pathetic:

$$ \frac{2^a-1}{2^b-1} = \frac{2^a}{2^b-1} - \frac{1}{2^b-1}$$

Which wasn't in the form that I wanted. I needed a remainder.

I also, tried find a recurrence relation, which ends up being:

$$ a_1 = 1, a_n = 2a_{n-1} + 1$$

But, after that, I couldn't really think of anything I could do with it. (Don't even know why I bothered doing this in the first place...)

Then I started thinking in terms of induction:

Maybe let $a$ be an arbitrary natural number, and show that $$(2^a-1, 2^1-1) = (2^a-1,1) = 1 = 2^1 - 1 = 2^{(a,1)} - 1$$

But after the base case I was stumped:

Assume it holds true for all $k$ then if it holds for $k+1$ our claim is true:

$$ (2^a - 1, 2^{k+1} - 1) = (2^a - 1, 2 \cdot 2^k - 1) = ???$$

I don't even think you can use induction quite this way, I've just trying to come up with something. Anyone have any hints? Thank you.

1

There are 1 best solutions below

4
On BEST ANSWER

Somewhat in line with the division idea, you can use the property that $(a,b)=(a-b,b)$ (this is the basis for Euclid's algorithm). In your case, assuming $a>b$, \begin{align*} (2^a-1, 2^b-1) &= ((2^a-1) - 2^{a-b}(2^b-1), 2^b-1)\\ &= (2^{a-b} - 1, 2^b - 1). \end{align*} Notice that the exponents have a property analogous to the GCD one. Just to be clear, if we define $f(a,b) = (2^a-1,2^b-1)$, we can write $f(a,b)=f(a-b,b)$.

This means that if we follow all of the steps to compute the GCD of $a$ and $b$ by Euclid's algorithm, we can carry the steps over to get that $f(a,b)=2^{(a,b)}-1$.