0% found this document useful (0 votes)
115 views7 pages

π π π = 4 · arctan (1) : How to approximate π? Use: tan 4 = 1 ⇒ 4 = arctan (1) ⇒ Know: Taylor expansion of arctan x

The document provides several methods for approximating the value of π: 1) Using the Taylor series expansion of arctan(x) and the identity π = 4arctan(1), which converges very slowly to π. 2) Approximating π as the sum of arctan(1/2) and arctan(1/3), which converges much faster than the first method. 3) Other identities involving arctan can also be used to approximate π to increasing levels of precision. The Chudnovsky algorithm in particular converges extremely quickly.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
115 views7 pages

π π π = 4 · arctan (1) : How to approximate π? Use: tan 4 = 1 ⇒ 4 = arctan (1) ⇒ Know: Taylor expansion of arctan x

The document provides several methods for approximating the value of π: 1) Using the Taylor series expansion of arctan(x) and the identity π = 4arctan(1), which converges very slowly to π. 2) Approximating π as the sum of arctan(1/2) and arctan(1/3), which converges much faster than the first method. 3) Other identities involving arctan can also be used to approximate π to increasing levels of precision. The Chudnovsky algorithm in particular converges extremely quickly.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

How to approximate π?

π π
Use: tan = 1 ⇒ = arctan(1) ⇒ π = 4 · arctan(1)
4 4
Know: Taylor expansion of arctan x:

x3 x5 x7 x9
T (x) = x − + − + − ···
3 5 7 9

which is convergent at for −1 < x ≤ 1.

Take x = 1:
1 1 1 1
arctan(1) = 1 − + − + − ···
3 5 7 9
1 1 1 1
 
π = 4 arctan(1) = 4 1 − + − + − · · ·
3 5 7 9
1 1 1 1
 
Expansion: π =4 1− + − + − ···
3 5 7 9
Approximation by truncated Taylor polynomials:
1 1 1 1
 
Up to the x9 -term: π ≈ 4 1 − + − + = 3.339682540.
 3 5 7 9
1 1 1 1

19
Up to the x -term: π ≈ 4 1 − + − + · · · − = 3.041839619
 3 5 7 19 
1 1 1 1
Up to the x100 -term: π ≈ 4 1 − + − + · · · − = 3.121594653
3 5 7 99
Up to the x1000 -term: π ≈ 3.139592656

Up to the x50000 -term: π ≈ 3.141552653


Exact value of π: π = 3.1415926535897932385....

• Observation:

The series will converge to the exact value of π

But it converges to π very slowly, unlike the convergence of e.


Question: why is this convergence process that slow?
Key: Taylor series is expanded at 0:

f 00 (0) f 000 (0) 3 f (n) (0) n


T (x) = f (0) + f 0 (0)x + + x + ··· x + ···
2! 3! n!

which is an approximation of f (x) around x = 0.

• So: if x is closer to 0, the convergence of T (x) to f (x) is faster. And if x is

further away from 0, the convergence gets slower.


• The convergence interval of Taylor series of arctan x is −1 < x ≤ 1.

x = 1 is the furthest point in the convergence range. This is why the

convergence of at x = 1 is quite slow.


Question: Any faster algorithm to approximate π?
1 1
Identity: π = 4(arctan + arctan )
2 3
x3 x5 x7 x9
Expansion: arctan x = x − + − + − ···
3 5 7 9
Approximations:
1 1
Degree arctan 2
arctan 3
π
3 3
(1/2) (1/3)
up to x3 -term 1
2
− 3
= 0.45833 1
3
− 3
= 0.32098 3.11728
3 5 3 5
1
2
− (1/2)
3
+ (1/2)
5
1
3
− (1/3)
3
+ (1/3)
5
5
up to x -term = 0.464583 = 0.321810 3.145576

Degree Approximation Error


x20 -term: π≈ 3.141592579... 10−7
x50 -term: π≈ 3.141592653589793266... 10−17
x100 -term: π≈ 3.14159265358979323846264338327949... 10−32
Exact value: π = 3.14159265358979323846264338327950...
• This series converges to π much faster than the previous one: π = 4 arctan 1.
Other identities of π:

(1) π = 4 arctan(1)
1 1
 
(2) π = 4 4 arctan + arctan
2 3
1 1
 
(3) π = 4 4 arctan − arctan (Marchin’s formula)
5 239
1 1 1 1
 
(4) π = 4 12 arctan + 32 arctan − 5 arctan + 12 arctan
49 57 239 110443

!−1
X (−1)n · (6n)! · (13591409 + 545140134n)
(5) π= 12 (Chudnovsky’s)
(3n)! · (n!)3 · 6403203n+3/2
n=0

Compare the algorithms of approximating π :

Algorithm (1) (2) (3) (4) (5)


Error of approx. by 10 terms 0.2 10−4 10−8 10−18 10−156
Error of approx. by 100 terms 0.02 10−32 10−72 10−172 10−1433
The latest approximation of e and π, and other mathematical constants:

http://www.numberworld.org/

Approximation of 2: Newton’s method
• For any positive number A, the sequence defined by recurrence relation

an A
an+1 = + , with a1 = 1
2 2an

will converge to A.
an 1 √
• When A = 2: the sequence an+1 = + converges to 2.
2 an

Approximations: a1 = 1
a1 1 1 1
a2 = + = + = 1.5
2 a1 2 1
a2 1 1.5 1
a3 = + = + = 1.41666...
2 a2 2 1.5
a3 1 1.41666 1
a4 = + = + = 1.414215686...
2 a3 2 1.41666
a5 = 1.4142135623746... (Error: 10−12 )

a6 = 1.4142135623730950488016896235... (Error: 10−25 )



Exact value: 2 = 1.4142135623730950488016887242...

You might also like