Wikipedia:Reference desk/Archives/Mathematics/2022 February 12

Welcome to the Wikipedia Mathematics Reference Desk Archives
The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

February 12

Proofs of rational bounds on π

In Proof that 22/7 exceeds pi a very pretty integral is given:

$$\int _{0}^{1}{\frac {x^{4}\left(1-x\right)^{4}}{1+x^{2}}}\,dx={\frac {22}{7}}-\pi ,$$

which is evaluated via rather tedious polynomial long division, but simplifies nicely. $22/7$ is one of the continued fraction convergents to $\pi$, but the analogous integrals for the next convergents ($333/106$, $355/113$, etc.) are far more complicated—and they must be, since there must be a factor of 53 and 113 in each antiderivative, respectively. My question is simply whether this is a grand coincidence. The natural generalization (replacing the exponent 4 with $4n$, n>1) never seems to give a "good" approximation in Diophantine terms. Cheers, Ovinus (talk) 21:56, 12 February 2022 (UTC)

Are you asking whether it is a coincidence that the largest prime factor of the denominators of the next convergents is so much higher? A few steps later, in $104348/33215,$ we have $33215=5\times 7\times 13\times 73,$ so there is not some rule that the largest prime factor must keep increasing. In the continued-fraction expansion of $e,$ we even have a convergent $87/32.$  --Lambiam 23:49, 12 February 2022 (UTC)

@Lambiam: Indeed there's no bound on the convergents' prime factors; I was just noting that for pi in particular, it would be impossible to achieve a value of, for example, $355/113-\pi$ with something like $\int _{0}^{1}P(x)/q(1+x^{2})\,dx$, with $P$ being a polynomial with integer coefficients and $q,\operatorname {deg} P<112$. My question was more about intuition for why the long division magically simplifies to such a low common denominator; in the original integral, the antiderivative's coefficients' GCD is $1/1230$ or something like that.

But as a tangent, your example of $104348/33215$ makes me wonder if there are low-degree rational functions that could be used to obtain those approximations too. I'm sure there are, but I don't see a systematic way to find them. Ovinus (talk) 01:24, 13 February 2022 (UTC)

You should probably read the Stephen Lucas paper linked to in the article. (Amazingly, you don't have to go through a paywall to get to it.) I think the answer is yes, it is (more or less) a grand coincidence. The Lucas paper does mention that some "experimenting" was needed to produce integrals for 355/113 - pi, which I interpret as meaning that the method used would not be easily generalized. --RDBury (talk) 03:42, 13 February 2022 (UTC)