• 検索結果がありません。

5-)++74)6--561)6-5.,-,1+)6-,1++)51.60-%;-)45.24.-5540)415418)56)8)

N/A
N/A
Protected

Academic year: 2022

シェア "5-)++74)6--561)6-5.,-,1+)6-,1++)51.60-%;-)45.24.-5540)415418)56)8)"

Copied!
3
0
0

読み込み中.... (全文を見る)

全文

(1)

Bulletin of Mathematical Analysis and Applications ISSN: 1821-1291, URL: http://www.bmathaa.org Volume 2 Issue 4(2010), Pages 137-139.

ON SOME ACCURATE ESTIMATES OF 𝜋

(DEDICATED IN OCCASION OF THE 70-YEARS OF PROFESSOR HARI M. SRIVASTAVA)

CRISTINEL MORTICI

Abstract. The aim of this paper is to establish some inequalities related to an accurate approximation formula of 𝜋. Being practically difficult, the computations arising in this problem were made using computer softwares such as Maple.

1. Introduction

Maybe the best known example of infinite product for estimating the constant 𝜋is the Wallis product [4]

𝜋 2 =2

1 ⋅2 3 ⋅4

3⋅ 4 5⋅ 6

5⋅6 7⋅ ⋅ ⋅=

𝑛=1

4𝑛2

4𝑛2−1, (1.1)

which is related to Euler’s gamma function Γ,since

𝑛

𝑘=1

4𝑘2

4𝑘2−1 = 16𝑛(Γ (𝑛+ 1))4

(2𝑛+ 1) (Γ (2𝑛+ 1))2. (1.2) Although it has a nice form, (1.1) is very slow, so it is not suitable for approximating the constant𝜋.

A possible starting point for accelerating (1.1) is the work of Fields [1] who shown that

Γ (𝑧+𝑎)

Γ (𝑧+𝑏) ∼ (𝑧+𝑎−𝜌)𝑎−𝑏

𝑁−1

𝑘=0

𝐵2𝑘(2𝜌)(𝜌) (𝑏−𝑎)2𝑘(𝑧+𝑎−𝜌)−2𝑘

(2𝑘)! (1.3)

+ (𝑧+𝑎−𝜌)𝑎−𝑏𝑂(

(𝑧+𝑎−𝜌)−2𝑁) , 2𝜌 = 1 +𝑎−𝑏, ∣arg (𝑧+𝑎)∣ ≤𝜋−𝜀, 𝜀 >0.

where the symbols𝐵2𝑘(2𝜌)stand for the generalized Bernoulli polynomials [2, 5].

2000Mathematics Subject Classification. 33B15, 26D15.

Key words and phrases. Wallis formula inequalities; asymptotic series; approximations.

c

⃝2010 Universiteti i Prishtin¨es, Prishtin¨e, Kosov¨e.

Submitted October 21, 2010. Published November 22, 2010.

137

(2)

138 C. MORTICI

With𝑧=𝑛, 𝑎=−𝑥, 𝑏= 1,and𝜌=−𝑥/2 in (1.3), we get ( 𝑥

𝑛 )

∼ (−1)𝑛(𝑛−𝑥/2)−(𝑥+1) Γ (−𝑥)

𝑘=0

𝐵2𝑘(2𝜌)(𝜌) (𝑥+ 1)2𝑘 (2𝑘)! (𝑛−𝑥/2)2𝑘

, 𝜌=−𝑥 2 , ( 𝑥

𝑛 )

∼(−1)𝑛(𝑛−𝑥/2)−(𝑥+1) Γ (−𝑥)

[

1 + (𝑥)3

24 (𝑛−𝑥/2)2+ (𝑥)5(5𝑥−2)

5760 (𝑛−𝑥/2)4 (1.4) +(𝑥)7(

35𝑥2−42𝑥+ 16) 2903040 (𝑛−𝑥/2)6 +⋅ ⋅ ⋅

] . See also [3, p. 142], where the following identity is stated

( 𝑥 𝑛

)

= (−1)𝑛 Γ (−𝑥)

Γ (𝑛−𝑥) Γ (𝑛+ 1).

Further, with𝑥=−1/2 in (1.4),we obtain the following formula 𝜋 = 4 (Γ (𝑛+ 1))416𝑛

(4𝑛+ 1) (Γ (2𝑛+ 1))2 [

1− 1

4 (4𝑛+ 1)2 + 21

32 (4𝑛+ 1)4 (1.5)

− 671

128 (4𝑛+ 1)6 + 180323

2048 (4𝑛+ 1)8 − 20898423 8192 (4𝑛+ 1)10 + 7426362705

65536 (4𝑛+ 1)12 − 1874409465055 262144 (4𝑛+ 1)14 +𝑂

( 1 𝑛16

)]2

.

The idea of expressing𝜋using the asymptotic expansion of the ratio Γ(𝑛+1/2)Γ(𝑛+1) was introduced by Tricomi and Erd´elyi in [3, p. 142, Rel. 23]. Here we make use of the asymptotic expansion for Γ(𝑛+1/2)Γ(𝑛+1) given in [1] to improve the results of Tricomi and Erd´elyi [3].

2. The results

By truncation of series (1.5), increasingly accurate approximations for𝜋can be derived. As example, if𝑛= 10, use of the first five terms in (1.5) gives𝜋with an error of 1.1639×10−12,while use of the first six terms in (1.5) gives𝜋with an error of 3.0431×10−14.

We prove the following

Theorem 2.1. For every integer𝑛≥1, we have 4 (Γ (𝑛+ 1))416𝑛

(4𝑛+ 1) (Γ (2𝑛+ 1))2𝑎(𝑛)< 𝜋 < 4 (Γ (𝑛+ 1))416𝑛

(4𝑛+ 1) (Γ (2𝑛+ 1))2𝑏(𝑛), where

𝑎(𝑛) = (

1− 1

4 (4𝑛+ 1)2+ 21

32 (4𝑛+ 1)4 − 671

128 (4𝑛+ 1)6+ 180323 2048 (4𝑛+ 1)8

)2

and 𝑏(𝑛) =

(

1− 1

4 (4𝑛+ 1)2+ 21

32 (4𝑛+ 1)4 − 671

128 (4𝑛+ 1)6+ 180323

2048 (4𝑛+ 1)8 − 20898423 8192 (4𝑛+ 1)10

)2 .

(3)

ON SOME ACCURATE ESTIMATES OF𝜋 139

Proof. The sequences

𝑥𝑛 = 4 (Γ (𝑛+ 1))416𝑛

(4𝑛+ 1) (Γ (2𝑛+ 1))2𝑎(𝑛) , 𝑦𝑛= 4 (Γ (𝑛+ 1))416𝑛 (4𝑛+ 1) (Γ (2𝑛+ 1))2𝑏(𝑛) converge to𝜋and it suffices to show that𝑥𝑛 is strictly increasing and𝑦𝑛 is strictly decreasing. In this sense, we have

𝑥𝑛+1

𝑥𝑛 −1 = 4 (4𝑛+ 1) (𝑛+ 1)2 (4𝑛+ 5) (2𝑛+ 1)2

𝑎(𝑛+ 1) 𝑎(𝑛) −1

= − 𝑃(𝑛)

(4𝑛+ 5)17(2𝑛+ 1)2(134217728𝑛8+⋅ ⋅ ⋅+ 172467)2, where the polynomial

𝑃(𝑛) = 60235603222675842001797120𝑛24+⋅ ⋅ ⋅+ 22691018044772336786409 has all coefficients positive. In consequence,𝑥𝑛 is strictly increasing, convergent to 𝜋,so 𝑥𝑛< 𝜋.

Then 𝑦𝑛+1

𝑦𝑛

−1 = 4 (4𝑛+ 1) (𝑛+ 1)2 (4𝑛+ 5) (2𝑛+ 1)2

𝑏(𝑛+ 1) 𝑏(𝑛) −1

= 𝑄(𝑛)

(2𝑛+ 1)2(4𝑛+ 5)12(8589934592𝑛10+⋅ ⋅ ⋅ −20208555)2, where the polynomial

𝑄(𝑛+ 1) = 210420037966350927549442377646080𝑛30

+⋅ ⋅ ⋅+ 1909672653415578833630022434217112437351 has all coefficients positive. In consequence,𝑦𝑛 is strictly decreasing, convergent to

𝜋,so 𝑦𝑛> 𝜋. □

Remark. The computations in this paper were made using Maple software.

Acknowledgments. The authors would like to thank the anonymous referee for his/her comments that helped us improve this article.

References

[1] J. L. Fields,A note on the asymptotic expansion of a ratio of gamma functions, Proc. Edin- burgh Math. Soc.152 (1966) 43-45.

[2] N.E. Nørlund,Vorksungen ¨uber Differenzenrechnung, Springer (Berlin), 1924.

[3] F. G. Tricomi, A. Erd´elyi,The asymptotic expansion of a ratio of gamma functions, Pacific J. Math.,11 (1951) 133-142.

[4] J. Wallis,Computation of 𝜋by successive interpolations, (1655) in: A Source Book in Math- ematics, 1200-1800 (D. J. Struik, Ed.), Harvard University Press, Cambridge, MA, (1969), 244-253.

[5] N.M. Temme,Special Functions, Wiley (NewYork), 1996.

Cristinel Mortici, Valahia University of Tˆargovis¸te, Department of Mathematics, Bd. Unirii 18, 130082 Tˆargovis¸te, Romania

E-mail address:[email protected]; [email protected] URL:http://www.cristinelmortici.ro

参照

関連したドキュメント

Now we introduce the concept of n-Lipschitz mapping and prove that the n-Lipschitz map- ping satisfying the n-distance one preserving property is an n-isometry under some

In [14]-[15] it is proved the well-posedness of boundary value problems for a one-dimensional wave equation in a rectangular domain in case when boundary conditions are given on

In this paper, we define the notion of generalized (αη) EB -contraction and establish some new fixed point theorems in the context of extended b-metric spaces.. We also furnish

Subsequently, Abbas and Jungck [2], Abbas and Rhoades

It can be shown that cubic graphs with arbitrarily large girth exist (see Theorem 3.2) and so there is a well-defined integer µ 0 (g), the smallest number of vertices for which a

These numerical methods blend collocation, convolution, and approximations based on sinc basis functions with iterative schemes like the steepest descent and Newton’s method,

We derive a number-theoretic inequality involving sums of a certain class of Möbius functions and obtain a sufficient condition for the Riemann hypothesis depending on an

[1] Yasar Bolat, Omer akin; Oscillatory behaviour of higher order neutral type non-linear forced differential equation with oscillating coefficients, J.. Marini, Gab Villari; On