Method for calculating the value of pi
Borwein's algorithm was devised by Jonathan and Peter Borwein to calculate the value of 1 / π {\displaystyle 1/\pi } . This and other algorithms can be found in the book Pi and the AGM – A Study in Analytic Number Theory and Computational Complexity .[ 1]
Ramanujan–Sato series[ edit ] These two are examples of a Ramanujan–Sato series . The related Chudnovsky algorithm uses a discriminant with class number 1.
Class number 2 (1989)[ edit ] Start by setting[ 2]
A = 212175710912 61 + 1657145277365 B = 13773980892672 61 + 107578229802750 C = ( 5280 ( 236674 + 30303 61 ) ) 3 {\displaystyle {\begin{aligned}A&=212175710912{\sqrt {61}}+1657145277365\\B&=13773980892672{\sqrt {61}}+107578229802750\\C&=\left(5280\left(236674+30303{\sqrt {61}}\right)\right)^{3}\end{aligned}}} Then
1 π = 12 ∑ n = 0 ∞ ( − 1 ) n ( 6 n ) ! ( A + n B ) ( n ! ) 3 ( 3 n ) ! C n + 1 2 {\displaystyle {\frac {1}{\pi }}=12\sum _{n=0}^{\infty }{\frac {(-1)^{n}(6n)!\,(A+nB)}{(n!)^{3}(3n)!\,C^{n+{\frac {1}{2}}}}}} Each additional term of the partial sum yields approximately 25 digits.
Class number 4 (1993)[ edit ] Start by setting[ 3]
A = 63365028312971999585426220 + 28337702140800842046825600 5 + 384 5 ( 10891728551171178200467436212395209160385656017 + 4870929086578810225077338534541688721351255040 5 ) 1 2 B = 7849910453496627210289749000 + 3510586678260932028965606400 5 + 2515968 3110 ( 6260208323789001636993322654444020882161 + 2799650273060444296577206890718825190235 5 ) 1 2 C = − 214772995063512240 − 96049403338648032 5 − 1296 5 ( 10985234579463550323713318473 + 4912746253692362754607395912 5 ) 1 2 {\displaystyle {\begin{aligned}A={}&63365028312971999585426220\\&{}+28337702140800842046825600{\sqrt {5}}\\&{}+384{\sqrt {5}}{\big (}10891728551171178200467436212395209160385656017\\&{}+\left.4870929086578810225077338534541688721351255040{\sqrt {5}}\right)^{\frac {1}{2}}\\B={}&7849910453496627210289749000\\&{}+3510586678260932028965606400{\sqrt {5}}\\&{}+2515968{\sqrt {3110}}{\big (}6260208323789001636993322654444020882161\\&{}+\left.2799650273060444296577206890718825190235{\sqrt {5}}\right)^{\frac {1}{2}}\\C={}&-214772995063512240\\&{}-96049403338648032{\sqrt {5}}\\&{}-1296{\sqrt {5}}{\big (}10985234579463550323713318473\\&{}+\left.4912746253692362754607395912{\sqrt {5}}\right)^{\frac {1}{2}}\end{aligned}}} Then
− C 3 π = ∑ n = 0 ∞ ( 6 n ) ! ( 3 n ) ! ( n ! ) 3 A + n B C 3 n {\displaystyle {\frac {\sqrt {-C^{3}}}{\pi }}=\sum _{n=0}^{\infty }{{\frac {(6n)!}{(3n)!(n!)^{3}}}{\frac {A+nB}{C^{3n}}}}} Each additional term of the series yields approximately 50 digits.
Iterative algorithms [ edit ] Quadratic convergence (1984)[ edit ] Start by setting[ 4]
a 0 = 2 b 0 = 0 p 0 = 2 + 2 {\displaystyle {\begin{aligned}a_{0}&={\sqrt {2}}\\b_{0}&=0\\p_{0}&=2+{\sqrt {2}}\end{aligned}}} Then iterate
a n + 1 = a n + 1 a n 2 b n + 1 = ( 1 + b n ) a n a n + b n p n + 1 = ( 1 + a n + 1 ) p n b n + 1 1 + b n + 1 {\displaystyle {\begin{aligned}a_{n+1}&={\frac {{\sqrt {a_{n}}}+{\frac {1}{\sqrt {a_{n}}}}}{2}}\\b_{n+1}&={\frac {(1+b_{n}){\sqrt {a_{n}}}}{a_{n}+b_{n}}}\\p_{n+1}&={\frac {(1+a_{n+1})\,p_{n}b_{n+1}}{1+b_{n+1}}}\end{aligned}}} Then p k converges quadratically to π ; that is, each iteration approximately doubles the number of correct digits. The algorithm is not self-correcting; each iteration must be performed with the desired number of correct digits for π 's final result.
Cubic convergence (1991)[ edit ] Start by setting
a 0 = 1 3 s 0 = 3 − 1 2 {\displaystyle {\begin{aligned}a_{0}&={\frac {1}{3}}\\s_{0}&={\frac {{\sqrt {3}}-1}{2}}\end{aligned}}} Then iterate
r k + 1 = 3 1 + 2 ( 1 − s k 3 ) 1 3 s k + 1 = r k + 1 − 1 2 a k + 1 = r k + 1 2 a k − 3 k ( r k + 1 2 − 1 ) {\displaystyle {\begin{aligned}r_{k+1}&={\frac {3}{1+2\left(1-s_{k}^{3}\right)^{\frac {1}{3}}}}\\s_{k+1}&={\frac {r_{k+1}-1}{2}}\\a_{k+1}&=r_{k+1}^{2}a_{k}-3^{k}\left(r_{k+1}^{2}-1\right)\end{aligned}}} Then ak converges cubically to 1 / π ; that is, each iteration approximately triples the number of correct digits.
Quartic convergence (1985)[ edit ] Start by setting[ 5]
a 0 = 2 ( 2 − 1 ) 2 y 0 = 2 − 1 {\displaystyle {\begin{aligned}a_{0}&=2\left({\sqrt {2}}-1\right)^{2}\\y_{0}&={\sqrt {2}}-1\end{aligned}}} Then iterate
y k + 1 = 1 − ( 1 − y k 4 ) 1 4 1 + ( 1 − y k 4 ) 1 4 a k + 1 = a k ( 1 + y k + 1 ) 4 − 2 2 k + 3 y k + 1 ( 1 + y k + 1 + y k + 1 2 ) {\displaystyle {\begin{aligned}y_{k+1}&={\frac {1-\left(1-y_{k}^{4}\right)^{\frac {1}{4}}}{1+\left(1-y_{k}^{4}\right)^{\frac {1}{4}}}}\\a_{k+1}&=a_{k}\left(1+y_{k+1}\right)^{4}-2^{2k+3}y_{k+1}\left(1+y_{k+1}+y_{k+1}^{2}\right)\end{aligned}}} Then a k converges quartically against 1 / π ; that is, each iteration approximately quadruples the number of correct digits. The algorithm is not self-correcting; each iteration must be performed with the desired number of correct digits for π 's final result.
One iteration of this algorithm is equivalent to two iterations of the Gauss–Legendre algorithm . A proof of these algorithms can be found here:[ 6]
Quintic convergence [ edit ] Start by setting
a 0 = 1 2 s 0 = 5 ( 5 − 2 ) = 5 ϕ 3 {\displaystyle {\begin{aligned}a_{0}&={\frac {1}{2}}\\s_{0}&=5\left({\sqrt {5}}-2\right)={\frac {5}{\phi ^{3}}}\end{aligned}}} where ϕ = 1 + 5 2 {\displaystyle \phi ={\tfrac {1+{\sqrt {5}}}{2}}} is the golden ratio . Then iterate
x n + 1 = 5 s n − 1 y n + 1 = ( x n + 1 − 1 ) 2 + 7 z n + 1 = ( 1 2 x n + 1 ( y n + 1 + y n + 1 2 − 4 x n + 1 3 ) ) 1 5 a n + 1 = s n 2 a n − 5 n ( s n 2 − 5 2 + s n ( s n 2 − 2 s n + 5 ) ) s n + 1 = 25 ( z n + 1 + x n + 1 z n + 1 + 1 ) 2 s n {\displaystyle {\begin{aligned}x_{n+1}&={\frac {5}{s_{n}}}-1\\y_{n+1}&=\left(x_{n+1}-1\right)^{2}+7\\z_{n+1}&=\left({\frac {1}{2}}x_{n+1}\left(y_{n+1}+{\sqrt {y_{n+1}^{2}-4x_{n+1}^{3}}}\right)\right)^{\frac {1}{5}}\\a_{n+1}&=s_{n}^{2}a_{n}-5^{n}\left({\frac {s_{n}^{2}-5}{2}}+{\sqrt {s_{n}\left(s_{n}^{2}-2s_{n}+5\right)}}\right)\\s_{n+1}&={\frac {25}{\left(z_{n+1}+{\frac {x_{n+1}}{z_{n+1}}}+1\right)^{2}s_{n}}}\end{aligned}}} Then ak converges quintically to 1 / π (that is, each iteration approximately quintuples the number of correct digits), and the following condition holds:
0 < a n − 1 π < 16 ⋅ 5 n ⋅ e − 5 n π {\displaystyle 0<a_{n}-{\frac {1}{\pi }}<16\cdot 5^{n}\cdot e^{-5^{n}}\pi \,\!} Start by setting
a 0 = 1 3 r 0 = 3 − 1 2 s 0 = ( 1 − r 0 3 ) 1 3 {\displaystyle {\begin{aligned}a_{0}&={\frac {1}{3}}\\r_{0}&={\frac {{\sqrt {3}}-1}{2}}\\s_{0}&=\left(1-r_{0}^{3}\right)^{\frac {1}{3}}\end{aligned}}} Then iterate
t n + 1 = 1 + 2 r n u n + 1 = ( 9 r n ( 1 + r n + r n 2 ) ) 1 3 v n + 1 = t n + 1 2 + t n + 1 u n + 1 + u n + 1 2 w n + 1 = 27 ( 1 + s n + s n 2 ) v n + 1 a n + 1 = w n + 1 a n + 3 2 n − 1 ( 1 − w n + 1 ) s n + 1 = ( 1 − r n ) 3 ( t n + 1 + 2 u n + 1 ) v n + 1 r n + 1 = ( 1 − s n + 1 3 ) 1 3 {\displaystyle {\begin{aligned}t_{n+1}&=1+2r_{n}\\u_{n+1}&=\left(9r_{n}\left(1+r_{n}+r_{n}^{2}\right)\right)^{\frac {1}{3}}\\v_{n+1}&=t_{n+1}^{2}+t_{n+1}u_{n+1}+u_{n+1}^{2}\\w_{n+1}&={\frac {27\left(1+s_{n}+s_{n}^{2}\right)}{v_{n+1}}}\\a_{n+1}&=w_{n+1}a_{n}+3^{2n-1}\left(1-w_{n+1}\right)\\s_{n+1}&={\frac {\left(1-r_{n}\right)^{3}}{\left(t_{n+1}+2u_{n+1}\right)v_{n+1}}}\\r_{n+1}&=\left(1-s_{n+1}^{3}\right)^{\frac {1}{3}}\end{aligned}}} Then a k converges nonically to 1 / π ; that is, each iteration approximately multiplies the number of correct digits by nine.[ 7]
^ Jonathan M. Borwein, Peter B. Borwein, Pi and the AGM – A Study in Analytic Number Theory and Computational Complexity , Wiley, New York, 1987. Many of their results are available in: Jorg Arndt, Christoph Haenel, Pi Unleashed, Springer, Berlin, 2001, ISBN 3-540-66572-2 ^ Bailey, David H (2023-04-01). "Peter Borwein: A Visionary Mathematician". Notices of the American Mathematical Society . 70 (04): 610–613. doi :10.1090/noti2675 . ISSN 0002-9920 . ^ Borwein, J.M.; Borwein, P.B. (1993). "Class number three Ramanujan type series for 1/π" . Journal of Computational and Applied Mathematics . 46 (1–2): 281–290. doi :10.1016/0377-0427(93)90302-R . ^ Arndt, Jörg; Haenel, Christoph (1998). π Unleashed . Springer-Verlag. p. 236. ISBN 3-540-66572-2 . ^ Mak, Ronald (2003). The Java Programmers Guide to Numerical Computation . Pearson Educational. p. 353. ISBN 0-13-046041-9 . ^ Milla, Lorenz (2019), Easy Proof of Three Recursive π -Algorithms , arXiv :1907.04110 ^ Henrik Vestermark (4 November 2016). "Practical implementation of π Algorithms" (PDF) . Retrieved 29 November 2020 .