Research Article | | Peer-Reviewed

Characterization of a Large Family of Convergent Series That Leads to a Rapid Acceleration of Slowly Convergent Logarithmic Series

Received: 3 September 2024     Accepted: 19 September 2024     Published: 10 October 2024
Views:       Downloads:
Abstract

Logarithmic series are known to have a very slow rate of convergence. For example, it takes more than the first 20,000 terms of the sum of the reciprocals of squares of the natural numbers to attain 5 decimal places of accuracy. In this paper, I will devise an acceleration scheme that will yield the same level of accuracy with just the first 400 terms of that power series. To accomplish this, I establish a relationship between all monotonically decreasing sequence of positive terms whose sum converges, a positive number ρ and a differentiable function φ. Then, I use ρ and φ to define the Tφ, ρ transformations on the partial sums of any convergent series. Furthermore, I prove that these Tφ, ρ transformations yield a rapid rate of convergence for many slowly convergent logarithmic series. Finaly, I provide several examples on how to compute φ if one is given the convergent series of decreasing, positive terms.

Published in International Journal of Theoretical and Applied Mathematics (Volume 10, Issue 3)
DOI 10.11648/j.ijtam.20241003.11
Page(s) 33-37
Creative Commons

This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.

Copyright

Copyright © The Author(s), 2024. Published by Science Publishing Group

Keywords

Series, Accelerators, Logarithmic, Convergence

1. Introduction
This paper is devoted to accelerating the convergence of an infinite series n=1an=S<, that satisfies limnS-an+1S-an=1. We call such series logarithmic series. Note that if an+1an and if limnan+1an=1, then the convergent series n=1an, is a logarithmic series .
Throughout this paper, fx=ax shall denote a continuous, positive valued, and a monotonically decreasing function on [1,), that satisfies:
1. limxf(x+1)f(x)=1 and
2. 1fxdx<.
If μ is a positive real number greater than 1, then we extend the traditional definition of the partial sum of an infinite series to define the μth partial sum, S(μ), of n=1an as follows:
Sμ=a1+a2+aμ+μ-μaμ+1, where μ is the greatest integer less than μ.
If ann=1 and bnn=1 are sequences converging to A and B respectively, such that limnan-Abn-B=0, then we say that ann=1 converges more rapidly than bnn=1.
In , S. Mukherjee et. al cite a well-known theorem by J.P. Delahaye and Germain-Bonne which states that: it is impossible to construct a series accelerator which can accelerate the convergence of all convergent series.
In this paper, we show that, for a large group of convergent, logarithmic series, n=1an, there exists a transformation, T, on the partial sums of the series and an increasing function φ on [1,), such that
3. φ(n)>n,
4. TSφ(n) S,
5. limnTSφ-SSφn-S=0.
In (see pages 11 and 12), Brezinski shows that if:
Tφ,ρSn=Sφn+Dφ(n), then limnDφ(n)S-Sφ(n)=0 is equivalent to limnTφ,ρSn-SSφn-S=0.
Dφ(n) is then called a perfect estimation of the error of S(φn).
Thus, we shall prove that if ann=1 is a sequence of decreasing positive terms such that:
limnan+1an=1 and 0<limnSφn-S(φ(n-1))an=ρ<1,
then Dφ(n)=Sn-S(φn)1-1ρ will be a perfect estimation of the error of S(φn).
This result leads naturally to a class of series accelerators:
Tφ,ρSn=Sφn+Dφ(n)=Sφn+Sn-S(φn)1-1ρ
The Tφ,ρ transformations are extensions of the T+m accelerators by Clark and Gray (see ) since ρ works for all real numbers belonging to the interval 0,1.
Furthermore, we shall also establish an interesting relationship between ρ, a convergent series n=1an, and an equation:
φ'xfφx=ρf(x)
It is worthy to note that if φn=n2, then the transformation:
Tφ,ρ(Sn)=Sn-1ρS(φn)1-1ρ
can be used to accelerate the convergence of the series n=11nln(n)2. I am not aware of any other series accelerator capable of accelerating the convergence of this series.
2. Section 1
We shall begin this section with an interesting result which characterizes all decreasing sequences ann=1, whose sum, n=1an, converges.
Theorem 1
Let f be a continuous, positive valued, and decreasing function on [1,). Then, for each ρ0,1, n=1f(n) converges if and only if there exists a differentiable function φ on [1,), such that:
6. φ(x)>x, and
7. φ'xfφx=ρf(x)
Proof:
Suppose that n=1f(n) converges. Then, 1ftdt converges . Furthermore, there exists F(x), a negative-valued antiderivative of f(x) that satisfies
limxFx=0 . Let φx=F-1(ρFx).
Differentiating φ(x) with respect to x, we get φ'xfφx=ρf(x).
Since F(x) is increasing on [1,), ρF(x)>F(x).
Thus, φx=F-1(ρFx) > F-1(Fx=x.
Next, suppose that φ(x)>x, and that φ'xfφx=ρf(x).
Then,  φ'xfφxf(x)=ρ<1. Therefore, by (see page 44), n=1f(n) converges.
Before proving our next Theorem, we shall give a couple of definitions. These definitions will mainly serve the purpose of simplifying the statement of Theorem 2.
Definition 1
If φ is a function satisfying the following conditions:
8. φ(x)>x and φ'(x)1, and if
9. limxφ'(ξx)φ'(ηx)=1, where ξx and ηxx,x+1,
then we say that φ is logarithmic on [1,).
Definition 2
Let φ(x) be a positive valued function and let ρ be an arbitrarily chosen positive number lying in the interval 0,1. We define the Tφ,ρ transformations as follows:
Tφ,ρ(Sn)=Sφn+Sn-S(φn)1-1ρ
Theorem 2
Suppose that n=1f(n) is a convergent, logarithmic series of a monotonically decreasing sequence of positive terms and that φ is logarithmic on [1,). Then, Tφ,ρ(Sn)=Sn-1ρS(φn)1-1ρ converges more rapidly than S(φn).
Proof:
First, we shall show that limnSφn-S(φ(n-1))f(n)=ρ.
To this end, note that
Sφn-S(φ(n-1))f(n)φn-φ(n-1))f(φn-1)f(n)
φ'(ξn)f(φ(n-1))f(n)
φ'n-1f(φn-1)f(n-1)f(n-1)f(n)φ'(ξn)φ'(n-1)
ρf(n-1)f(n)φ'(ξn)φ'(n-1), whereξnn-1,n.
Thus, lim̅Sφn-S(φ(n-1))f(n)ρ.
Similarly,
Sφn-S(φ(n-1))f(n)φn-φ(n-1))f(φ(n_)f(n)
φ'ηnf(φn)f(n)
φ'nf(φn)f(n)φ'(ηn)φ'(n)
ρφ'(ηn)φ'(n), whereηnn-1,n
Therefore, lim̲Sφn-S(φ(n-1))f(n)ρ. 
Hence, limnSφn-S(φ(n-1))f(n)=ρ .
Now, Sφn-S(n)) and Sφn-S both converge to 0 monotonically.
Therefore, by (see page 413) and above, we have that:
limnSφn-S(n)Sφn-S =1-limnf(n)Sφn-S(φ(n-1))=ρ-1ρ
Hence, limnTφ,ρSn-SS(φn)-S
=limnTφ,ρSn-Sφn+S(φn)-SS(φn)-S
=1-limnDφ(n)S(φn)-S
=1-11-1ρlimnSφn-S(n)Sφn-S 
=1-11-1ρ1-1ρ=0
Thus,
Dφ(n)=Sn-S(φn)1-1ρ
is a perfect estimation of the error of S(φn), whenever φ is logarithmic on [1,), and
Tφ,ρSn=Sφn+Dφ(n)
Corollary 2.1
Suppose that n=1f(n) is a convergent logarithmic series of a monotonically decreasing sequence of positive terms. If limxf(φ(x+1))f(φ(x))=1, then, Tφ,ρ(Sn)=Sn-1ρS(φn)1-1ρ converges more rapidly than S(φn).
Proof:
Note that φ'(x+1)φ'(x)=f(x+1)f(x).f(φ(x+1))f(φ(x)), and limxf(x+1)f(x)=1.
Next, we provide some examples showing how to determine φ(x) for a few convergent logarithmic series.
3. Examples
Example 1. Let fx=1x2. Then, Fx=-1x and therefore F-1x=-1x.
Hence, φx=xρ. If ρ=1m, where m is a positive integer, then the Tφ,ρ series accelerators reduce to the T+m transformations (see , page 268, definition 4.1).
Example 2. Let fx=1xln(x)2. Then, Fx=-1lnx and F-1x=e-1x. Thus,
φx=x1ρ. Note that φ is logarithmic on [1,).
Example 3. Let fx=1xlnxln(lnx))2. Then, Fx=-1ln(ln(x) and F-1x=ee-1x. Hence, φx=elnx1ρ and φ is logarithmic on [1,) since φ'x=lnx1-ρρρxφ(x).
Now, k=11k2=1.644.93 (5 decimal places) .
In the following table, Sn=k=1n1k2 (5 decimal places). We observe that it will take more than the first 20,000 terms of the series to attain 5 decimal places of accuracy. However, if one applies the Tφ,ρ series accelerators, one only needs at most the first 400 terms to achieve the same level of accuracy.
Table 1. Partial sums of k=11k2.

n

S(n)

S(2n)

T2n,.5= 2S2n-S(n)

S(4n)

T4n, .25= 4Sn-S(n)3

10

1.54977

1.59616

1.64256

1.62024

1.64374

100

1.63498

1.63995

1.64491

1.64244

1.64492

200

1.63995

1.64244

1.64493

1.64369

1.64493

1000

1.64393

1.64443

1.64493

1.64468

1.64493

10000

1.64483

1.64488

1.64493

1.64491

1.64493

100000

1.64492

1.64493

1.64493

1.64493

1.64493

4. Conclusion
In conclusion, we have shown if n=1an is a convergent logarithmic series and if the φ function associated with this series is logarithmic, then the convergence of n=1an can be accelerated by the Tφ,n series accelerators.
Consequently, any convergent logarithmic series that can be written as a linear combination of logarithmically convergent series such that the associated φ functions converge logarithmically, can be accelerated by the Tφ,ρ series accelerators.
Acknowledgments
I wish to express my heartfelt thanks to Dr. William Ford for the useful insights he provided me in the field of series acceleration.
Author Contributions
Joseph Gaskin is the sole author. The author read and approved the final manuscript.
Conflicts of Interest
The author declares no conflict of interest.
References
[1] Brezinski, C., and Zaglia, R., Extrapolation Methods, Theory and Practice, Studies in Computational Mathematics 2, Elsevier, 2013.
[2] Bromwich, T. J., An Introduction to the Theory of Infinite Series, Alpha Editions,
[3] Belghaba K., On the Transformation T+m due to Gray and Clark, Journal of Mathematics and Statistics, Vol. 3,
[4] H. L. Gray, and W. D. Clark, “On a Class of Nonlinear Transformation and their Applications to the Evaluation of Infinite Series,” Journal of Research of the National Bureau of Standards-B. Mathematical Sciences, Vol. 73B, No. 3, July-September 1969.
[5] J. P. Delahaye and B. Germain-Bonne, SIAM Journal on Numerical Analysis, Vol. 19, No. 4, 1982, pp 840-844.
[6] C. Brezinski, Convergence acceleration during the 20th century, Journal of Computational and Applied Mathematics, Volume 122, Issues 1-2, October 2000, pp 1-2.
[7] d’AsperemontA, Scieur D, Taylor A2021, Acceleration Methods, Foundations and Trends (R) in Optimization, 5(1-2), pp1-245
[8] Briggs, Cochran, Gillet, Shultz, Calculus Early Transcendentals, 3rd Edition, ISBN13: 978-0-13-476684-3, 2019, pp 674.
[9] Howard Anton, I. C. Bivens, Stephen Davis, Calculus, 12th Edition, 2021, ISBN-13:978-1119778127, ISBN-10:1119778123, pp 290.
[10] Andrew H. Van Tuyl, Acceleration of Convergence of a family of logarithmically convergent sequences, Mathematics of Computation, Vol 63, Number 207, July 1994, pp 229.
[11] S. Mukherjee, B-S Hua, N. Umetani, D. Meister, Neural sequence transformation, Computer Graphics Forum, The Eurographics Association and John Wiley & Sons Ltd, 2021, pp 132.
[12] Khan SA, International Journal of Applied Mathematics, ISBN: 314-8060, Volume 33, 2020.
[13] Robert G. Bartle, Donald R. Sherbert, Introduction to Real Analysis, 4th Edition, ISBN-13: 978-0471433316, 2011, pp 400-418.
[14] James Stewart, Daniel K. Clegg, Saleem Watson, Calculus: Early Transcendentals, ISBN-13: 978-1337624183, 9th Edition, April 2020, Chapter 7.8, pp 534.
[15] Britannica, The Editors Encyclopedia, “Infinite series”, Encyclopedia Britannica, August 2024.
Cite This Article
  • APA Style

    Gaskin, J. (2024). Characterization of a Large Family of Convergent Series That Leads to a Rapid Acceleration of Slowly Convergent Logarithmic Series. International Journal of Theoretical and Applied Mathematics, 10(3), 33-37. https://doi.org/10.11648/j.ijtam.20241003.11

    Copy | Download

    ACS Style

    Gaskin, J. Characterization of a Large Family of Convergent Series That Leads to a Rapid Acceleration of Slowly Convergent Logarithmic Series. Int. J. Theor. Appl. Math. 2024, 10(3), 33-37. doi: 10.11648/j.ijtam.20241003.11

    Copy | Download

    AMA Style

    Gaskin J. Characterization of a Large Family of Convergent Series That Leads to a Rapid Acceleration of Slowly Convergent Logarithmic Series. Int J Theor Appl Math. 2024;10(3):33-37. doi: 10.11648/j.ijtam.20241003.11

    Copy | Download

  • @article{10.11648/j.ijtam.20241003.11,
      author = {Joseph Gaskin},
      title = {Characterization of a Large Family of Convergent Series That Leads to a Rapid Acceleration of Slowly Convergent Logarithmic Series
    },
      journal = {International Journal of Theoretical and Applied Mathematics},
      volume = {10},
      number = {3},
      pages = {33-37},
      doi = {10.11648/j.ijtam.20241003.11},
      url = {https://doi.org/10.11648/j.ijtam.20241003.11},
      eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ijtam.20241003.11},
      abstract = {Logarithmic series are known to have a very slow rate of convergence. For example, it takes more than the first 20,000 terms of the sum of the reciprocals of squares of the natural numbers to attain 5 decimal places of accuracy. In this paper, I will devise an acceleration scheme that will yield the same level of accuracy with just the first 400 terms of that power series. To accomplish this, I establish a relationship between all monotonically decreasing sequence of positive terms whose sum converges, a positive number ρ and a differentiable function φ. Then, I use ρ and φ to define the Tφ, ρ transformations on the partial sums of any convergent series. Furthermore, I prove that these Tφ, ρ transformations yield a rapid rate of convergence for many slowly convergent logarithmic series. Finaly, I provide several examples on how to compute φ if one is given the convergent series of decreasing, positive terms.
    },
     year = {2024}
    }
    

    Copy | Download

  • TY  - JOUR
    T1  - Characterization of a Large Family of Convergent Series That Leads to a Rapid Acceleration of Slowly Convergent Logarithmic Series
    
    AU  - Joseph Gaskin
    Y1  - 2024/10/10
    PY  - 2024
    N1  - https://doi.org/10.11648/j.ijtam.20241003.11
    DO  - 10.11648/j.ijtam.20241003.11
    T2  - International Journal of Theoretical and Applied Mathematics
    JF  - International Journal of Theoretical and Applied Mathematics
    JO  - International Journal of Theoretical and Applied Mathematics
    SP  - 33
    EP  - 37
    PB  - Science Publishing Group
    SN  - 2575-5080
    UR  - https://doi.org/10.11648/j.ijtam.20241003.11
    AB  - Logarithmic series are known to have a very slow rate of convergence. For example, it takes more than the first 20,000 terms of the sum of the reciprocals of squares of the natural numbers to attain 5 decimal places of accuracy. In this paper, I will devise an acceleration scheme that will yield the same level of accuracy with just the first 400 terms of that power series. To accomplish this, I establish a relationship between all monotonically decreasing sequence of positive terms whose sum converges, a positive number ρ and a differentiable function φ. Then, I use ρ and φ to define the Tφ, ρ transformations on the partial sums of any convergent series. Furthermore, I prove that these Tφ, ρ transformations yield a rapid rate of convergence for many slowly convergent logarithmic series. Finaly, I provide several examples on how to compute φ if one is given the convergent series of decreasing, positive terms.
    
    VL  - 10
    IS  - 3
    ER  - 

    Copy | Download

Author Information
  • Department of Mathematics and Physics, Southern University and A&M College, Baton Rouge, USA

    Research Fields: Real Analysis, Sequences and series, Number Theory, Differential equations, Point Set Topology