27
$\begingroup$

We have the known BBP(Bailey–Borwein–Plouffe)-type formulas,

$$\ln3 = \sum_{n=0}^\infty\frac{1}{2^{2n}}\left(\frac{1}{2n+1}\right)$$

$$\ln5 = \frac{1}{2^2}\sum_{n=0}^\infty\frac{1}{2^{4n}}\left(\frac{2^2}{4n+1}+\frac{2^2}{4n+2}+\frac{1}{4n+3}\right)$$

However, I noticed that if we define the function,

$$R\big(a,b\big) = \sum_{n=0}^\infty\frac{1}{(2^a)^n}\left(\sum_{j=1}^{a-1}\frac{2^{a-1-j}}{an+j}+\sum_{k=1}^{a/b-1}(-1)^{k+1}\frac{2^{a-1-bk}}{an+bk}\right)\tag1$$

then it seems BBP-type formulas for Fermat numbers $2^{2^m}+1$ with $m>0$ have a common form,

$$\ln 5 = \frac1{2^{2}}R\big(2^2,2^1\big)$$

$$\ln 17 = \frac1{2^{13}}R\big(2^4,2^2\big)$$

$$\ln 257 = \frac1{2^{252}}R\big(2^8,2^3\big)$$

$$\color{brown}{\ln 65537 \overset{?}= \frac1{2^{65531}}R\big(2^{16},2^4\big)}$$

Q: Is the formula for $p=65537$ true?

I've used Mathematica to verify the $p=5,\,17,\,257$ to hundreds of decimal digits, and also $p=65537$ using its initial terms, but how do we rigorously prove that $(1)$ is true for Fermat numbers $>3$?

P.S. In this paper, the authors were not(?) able to find $p=65537$ and is missing in the list.

$\endgroup$
34
$\begingroup$

Yes, this actually holds for all Fermat numbers! Let's start with the identity $$\log 2=\sum_{k=1}^{\infty}\frac{1}{k2^k}$$ and try to work out an expression for $$\log (2^{2^s}+1)=2^{s}\log 2+\log\left(1-\frac{1}{2^{2^{s+1}}}\right)-\log\left(1-\frac{1}{2^{2^s}}\right)$$ $$=2^s\left(\sum_{k=1}^{\infty}\frac{1}{k2^k}\right)-\left(\sum_{k=1}^{\infty}\frac{1}{k2^{k2^{s+1}}}\right)+\left(\sum_{k=1}^{\infty}\frac{1}{k2^{k2^{s}}}\right)$$ $$=\sum_{n=0}^{\infty}\frac{1}{2^{n2^{2^s}}}\left(\sum_{j=1}^{2^{2^s}}\frac{2^{s-j}}{n2^{2^s}+j}-\sum_{h=1}^{2^{2^s-s-1}}\frac{2^{-h2^{s+1}}}{n2^{2^{s}-s-1}+h}+\sum_{l=1}^{2^{2^s-s}}\frac{2^{-l2^s}}{n2^{2^s-s}+l}\right)$$ $$=\sum_{n=0}^{\infty}\frac{1}{2^{n2^{2^s}}}\left(\sum_{j=1}^{2^{2^s}}\frac{2^{s-j}}{n2^{2^s}+j}-\sum_{h=1}^{2^{2^s-s-1}}\frac{2^{-(2h)2^{s+1}}}{n2^{2^{s}-s}+2h}+\sum_{l=1}^{2^{2^s-s}}\frac{2^{-l2^s}}{n2^{2^s-s}+l}\right)$$ $$=\sum_{n=0}^{\infty}\frac{1}{2^{n2^{2^s}}}\left(\sum_{j=1}^{2^{2^s}}\frac{2^{s-j}}{n2^{2^s}+j}+\sum_{l=1}^{2^{2^s-s}}\frac{(-1)^{l+1}2^{-l2^s}}{n2^{2^s-s}+l}\right)$$ here we see that the terms at $j=2^{2^s}$ and $l=2^{2^s-s}$ cancel out, so after factoring out $\frac{1}{2^{2^{2^s}-s-1}}$ we are left with the defining expression of $R$ with $a=2^{2^s}, b=2^s$: $$\log(2^{2^s}+1)=\frac{1}{2^{2^{2^s}-s-1}}R(2^{2^s},2^s).$$

$\endgroup$
  • 1
    $\begingroup$ Beautiful! In his compendium, Bailey (the first B in BBP), listed $p=3,5,7,17,257$ but apparently forgot $p=65537$ since he included larger primes in his list. $\endgroup$ – Tito Piezas III Jul 5 '17 at 13:52
  • $\begingroup$ Can you kindly look at this post as the counterpart formula for Mersenne numbers? Maybe a small tweak to your proof can settle that as well. $\endgroup$ – Tito Piezas III Jul 7 '17 at 16:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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