Read · 2 min
Stub explanation
We know that is irrational. But "how irrational" is it?
One fruitful way to "measure irrationality" is to look at the distances between and various rational numbers. More precisely, we can ask
How far is away from a given rational number ? The rationals are dense in the reals, so some fractions will be arbitrarily close to the real number : approximation gets easier as the denominator increases. But can we get a lower bound on the distance in terms of the denominator ?
(It will be more natural to try to lower bound instead, and then divide everything by at the end.)
Yes! Let be any rational with numerator and denominator . Since is irrational, , or more naturally, , or even more naturally, .
But are integers, so and are unequal integers, which means they differ by at least : Let , so . By difference of squares, this factors as
We want a lower bound on (basically), so it would suffice to have an upper bound on . From the triangle inequality, (with equality when have the same sign—which is always the case when is close enough to the positive number ). So So we wanted a lower bound in terms of , but right now we only have a lower bound in terms of and . That's not good enough if is a lot larger than , but let's rephrase our original question:
In the case that is small, how small can it be? In other words, our job is easier when is not small, because we don't have to prove anything then!
So here's one way to finish:
If and are close (say within of each other), then . Otherwise, if and are not within of each other, then certainly and must be more than apart, so .
Always one of the two cases holds, so we conclude that for any integers with , we have the lower bound for some positive constant (we're using rough asymptotic approximations here, to bound a complicated function in below by a simple power function in , which captures the most important information).
Finally, dividing by , gives . By the way, equality "basically holds" whenever is small enough (intuitively, whenever our lower bound above is "close to holding"), including when , a Pell-type equation discussed here.