Gregory coefficients

Gregory coefficients $G n$ , also known as reciprocal logarithmic numbers, Bernoulli numbers of the second kind and the Cauchy numbers of the first kind,^[1]^[2]^[3]^[4]^[5]^[6]^[7]^[8]^[9]^[10]^[11]^[12]^[13] are the rational numbers

$n$	$1$	$2$	$3$	$4$	$5$	$6$	$7$	$8$	$9$	$10$	$11$	$...$	OEIS sequences
$G n$	$+ 1 / 2$	$- 1 / 12$	$+ 1 / 24$	$- 19 / 720$	$+ 3 / 160$	$- 863 / 60480$	$+ 275 / 24192$	$- 33953 / 3628800$	$+ 8183 / 1036800$	$- 3250433 / 479001600$	$+ 4671 / 788480$	$...$	A002206 (numerators), A002207 (denominators)

that occur in the Maclaurin series expansion of the reciprocal logarithm

{\begin{aligned}{\frac {z}{\ln(1+z)}}&=1+{\frac {1}{2}}z-{\frac {1}{12}}z^{2}+{\frac {1}{24}}z^{3}-{\frac {19}{720}}z^{4}+{\frac {3}{160}}z^{5}-{\frac {863}{60480}}z^{6}+\cdots \\&=1+\sum _{n=1}^{\infty }G_{n}z^{n}\,,\qquad |z|<1\,.\end{aligned}}

Gregory coefficients are alternating $G n = (-1) n -1 | G n |$ and decreasing in absolute value. These numbers are named after James Gregory who introduced them in 1670 in the numerical integration context. They were subsequently rediscovered by many famous mathematicians and often appear in works of modern authors who do not recognize them.^[1]^[5]^[14]^[15]^[16]^[17]

Computation and representations

The simplest way to compute Gregory coefficients is to use the recurrence formula

{\frac {G_{1}}{n}}-{\frac {G_{2}}{n-1}}+{\frac {G_{3}}{n-2}}-\cdots +(-1)^{n-1}{\frac {G_{n}}{1}}\,=\,{\frac {1}{n+1}}\,,\qquad n=2,3,4,\ldots

with $G 1 = 1 / 2$ .^[14]^[18] Gregory coefficients may be also computed explicitly via the following differential

G_{n}={\frac {1}{n!}}\left[{\frac {d^{n}}{dz^{n}}}{\frac {z}{\ln(1+z)}}\right]_{z=0},

the following integrals

{\begin{array}{l}\displaystyle G_{n}={\frac {1}{n!}}\int \limits _{0}^{1}x(x-1)(x-2)\cdots (x-n+1)\,dx\\[6mm]\displaystyle G_{n}=(-1)^{n-1}\!\int \limits _{0}^{\infty }{\frac {dx}{(1+x)^{n}\left(\ln ^{2}x+\pi ^{2}\right)}},\end{array}}

or the finite summation formula

G_{n}={\frac {1}{n!}}\sum \limits _{l=1}^{n}{\frac {s(n,l)}{l+1}},

where $s (n, l)$ are the signed Stirling numbers of the first kind.

Bounds and asymptotic behavior

The Gregory coefficients satisfy the bounds

{\frac {1}{6n(n-1)}}<{\big |}G_{n}{\big |}<{\frac {1}{6n}},\qquad n>2,

given by Johan Steffensen.^[15] These bounds were later improved by various authors. The best known bounds for them were given by Blagouchine.^[17] In particular,

{\frac {\,1\,}{\,n\ln ^{2}\!n\,}}\,-\,{\frac {\,2\,}{\,n\ln ^{3}\!n\,}}\leqslant \,{\big |}G_{n}{\big |}\,\leqslant \,{\frac {\,1\,}{\,n\ln ^{2}\!n\,}}-{\frac {\,2\gamma \,}{\,n\ln ^{3}\!n\,}}\,,\qquad \quad n\geqslant 5\,.

Asymptotically, at large index $n$ , these numbers behave as^[2]^[17]

{\big |}G_{n}{\big |}\sim {\frac {1}{n\ln ^{2}n}},\qquad n\to \infty .

More accurate description of $G n$ at large $n$ may be found in works of Van Veen,^[18] Davis,^[3] Coffey,^[19] Nemes^[6] and Blagouchine.^[17]

Series with Gregory coefficients

Series involving Gregory coefficients may be often calculated in a closed-form. Basic series with these numbers include

{\begin{aligned}\sum _{n=1}^{\infty }{\big |}G_{n}{\big |}=1\\[2mm]\sum _{n=1}^{\infty }G_{n}={\frac {1}{\ln 2}}-1\\[2mm]\sum _{n=1}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n}}=\gamma ,\end{aligned}}

where $γ = 0.5772156649...$ is Euler's constant. These results are very old, and their history may be traced back to the works of Gregorio Fontana and Lorenzo Mascheroni.^[17]^[20] More complicated series with the Gregory coefficients were calculated by various authors. Kowalenko,^[8] Alabdulmohsin ^[10] and some other authors calculated

{\begin{array}{l}\displaystyle \sum _{n=2}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n-1}}=-{\frac {1}{2}}+{\frac {\ln 2\pi }{2}}-{\frac {\gamma }{2}}\\[6mm]\displaystyle \displaystyle \sum _{n=1}^{\infty }\!{\frac {{\big |}G_{n}{\big |}}{n+1}}=1-\ln 2.\end{array}}

Alabdulmohsin^[10] also gives these identities

{\begin{aligned}&{\big |}G_{1}{\big |}+{\big |}G_{2}{\big |}-{\big |}G_{4}{\big |}-{\big |}G_{5}{\big |}+{\big |}G_{7}{\big |}+{\big |}G_{8}{\big |}-{\big |}G_{10}{\big |}-{\big |}G_{11}{\big |}+\cdots ={\frac {\sqrt {3}}{\pi }}\\[2mm]&{\big |}G_{2}{\big |}+{\big |}G_{3}{\big |}-{\big |}G_{5}{\big |}-{\big |}G_{6}{\big |}+{\big |}G_{8}{\big |}+{\big |}G_{9}{\big |}-{\big |}G_{11}{\big |}-{\big |}G_{12}{\big |}+\cdots ={\frac {2{\sqrt {3}}}{\pi }}-1\\[2mm]&{\big |}G_{1}{\big |}-{\big |}G_{3}{\big |}-{\big |}G_{4}{\big |}+{\big |}G_{6}{\big |}+{\big |}G_{7}{\big |}-{\big |}G_{9}{\big |}-{\big |}G_{10}{\big |}+{\big |}G_{12}{\big |}+\cdots =1-{\frac {\sqrt {3}}{\pi }}.\end{aligned}}

Candelperger, Coppo^[21]^[22] and Young^[7] showed that

\sum _{n=1}^{\infty }{\frac {{\big |}G_{n}{\big |}\cdot H_{n}}{n}}={\frac {\pi ^{2}}{6}}-1,

where $H n$ are the harmonic numbers. Blagouchine^[17]^[23]^[24]^[25] provides the following identities

{\begin{aligned}&\sum _{n=1}^{\infty }{\frac {G_{n}}{n}}=\operatorname {li} (2)-\gamma \\[2mm]&\sum _{n=3}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n-2}}=-{\frac {1}{8}}+{\frac {\ln 2\pi }{12}}-{\frac {\zeta '(2)}{\,2\pi ^{2}}}\\[2mm]&\sum _{n=4}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n-3}}=-{\frac {1}{16}}+{\frac {\ln 2\pi }{24}}-{\frac {\zeta '(2)}{4\pi ^{2}}}+{\frac {\zeta (3)}{8\pi ^{2}}}\\[2mm]&\sum _{n=1}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n+2}}={\frac {1}{2}}-2\ln 2+\ln 3\\[2mm]&\sum _{n=1}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n+3}}={\frac {1}{3}}-5\ln 2+3\ln 3\\[2mm]&\sum _{n=1}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n+k}}={\frac {1}{k}}+\sum _{m=1}^{k}(-1)^{m}{\binom {k}{m}}\ln(m+1)\,,\qquad k=1,2,3,\ldots \\[2mm]&\sum \limits _{n=1}^{\infty }{\frac {{\big |}G_{n}{\big |}}{n^{2}}}=\int \limits _{0}^{1}{\frac {-\operatorname {li} (1-x)+\gamma +\ln x}{x}}\,dx\\[2mm]&\sum \limits _{n=1}^{\infty }{\frac {G_{n}}{n^{2}}}=\int \limits _{0}^{1}{\frac {\operatorname {li} (1+x)-\gamma -\ln x}{x}}\,dx,\end{aligned}}

where $li(z)$ is the integral logarithm and ${\tbinom {k}{m}}$ is the binomial coefficient. It is also known that the zeta function, the gamma function, the polygamma functions, the Stieltjes constants and many other special functions and constants may be expressed in terms of infinite series containing these numbers.^[1]^[17]^[18]^[26]^[27]

Generalizations

Various generalizations are possible for the Gregory coefficients. Many of them may be obtained by modifying the parent generating equation. For example, Van Veen^[18] consider

\left({\frac {\ln(1+z)}{z}}\right)^{s}=s\sum _{n=0}^{\infty }{\frac {z^{n}}{n!}}K_{n}^{(s)}\,,\qquad |z|<1\,,

and hence

G_{n}=-{\frac {K_{n}^{(-1)}}{n!}}

Equivalent generalizations were later proposed by Kowalenko^[9] and Rubinstein.^[28] In a similar manner, Gregory coefficients are related to the generalized Bernoulli numbers

\left({\frac {t}{e^{t}-1}}\right)^{s}=\sum _{k=0}^{\infty }{\frac {t^{k}}{k!}}B_{k}^{(s)},\qquad |t|<2\pi \,,

see,^[18]^[26] so that

G_{n}=-{\frac {B_{n}^{(n-1)}}{(n-1)\,n!}}

Jordan^[1]^[16]^[29] defines polynomials $ψ n (s)$ such that

{\frac {z(1+z)^{s}}{\ln(1+z)}}=\sum _{n=0}^{\infty }z^{n}\psi _{n}(s)\,,\qquad |z|<1\,,

and call them Bernoulli polynomials of the second kind. From the above, it is clear that $G n = ψ n (0)$ . Carlitz^[16] generalized Jordan's polynomials $ψ n (s)$ by introducing polynomials $β$

\left({\frac {z}{\ln(1+z)}}\right)^{s}\!\!\cdot (1+z)^{x}=\sum _{n=0}^{\infty }{\frac {z^{n}}{n!}}\,\beta _{n}^{(s)}(x)\,,\qquad |z|<1\,,

and therefore

G_{n}={\frac {\beta _{n}^{(1)}(0)}{n!}}

Blagouchine^[17] introduced numbers $G n (k)$ such that

G_{n}(k)={\frac {1}{n!}}\sum \limits _{l=1}^{n}{\frac {s(n,l)}{l+k}},

and studied their asymptotics at large $n$ . Clearly, $G n = G n (1)$ . A different generalization of the same kind was also proposed by Komatsu^[29]

c_{n}^{(k)}=\sum \limits _{l=0}^{n}{\frac {s(n,l)}{(l+1)^{k}}},

so that $G n = c n (1) / n!$ Numbers $c n (k)$ are called by the author poly-Cauchy numbers.^[29] Coffey^[19] defines polynomials

P_{n+1}(y)={\frac {1}{n!}}\int \limits _{0}^{y}x(1-x)(2-x)\cdots (n-1-x)\,dx

and therefore $| G n | = P n +1 (1)$ .

References

1 2 3 4 Ch. Jordan. The Calculus of Finite Differences Chelsea Publishing Company, USA, 1947.
1 2 L. Comtet. Advanced combinatorics (2nd Edn.) D. Reidel Publishing Company, Boston, USA, 1974.
1 2 H.T. Davis. The approximation of logarithmic numbers. Amer. Math. Monthly, vol. 64, no. 8, pp. 11–18, 1957.
↑ P. C. Stamper. Table of Gregory coefficients. Math. Comp. vol. 20, p. 465, 1966.
1 2 D. Merlini, R. Sprugnoli, M. C. Verri. The Cauchy numbers. Discrete Math., vol. 306, pp. 1906–1920, 2006.
1 2 G. Nemes. An asymptotic expansion for the Bernoulli numbers of the second kind. J. Integer Seq, vol. 14, 11.4.8, 2011
1 2 P.T. Young. A 2-adic formula for Bernoulli numbers of the second kind and for the Nörlund numbers. J. Number Theory, vol. 128, pp. 2951-2962, 2008.
1 2 V. Kowalenko. Properties and Applications of the Reciprocal Logarithm Numbers. Acta Appl. Math., vol. 109, pp. 413-437, 2010.
1 2 V. Kowalenko. Generalizing the reciprocal logarithm numbers by adapting the partition method for a power series expansion. Acta Appl. Math., vol. 106, pp. 369–420, 2009.
1 2 3 I. M. Alabdulmohsin. Summability calculus, arXiv:1209.5739, 2012.
↑ I. Mezo. Gompertz Constant, Gregory Coefficients and a Series of the Logarithm Function. J. Ana. Num. Theor. 2, No. 2, pp. 33–36, 2014.
↑ F. Qi and X.-J. Zhang An integral representation, some inequalities, and complete monotonicity of Bernoulli numbers of the second kind. Bull. Korean Math. Soc., vol. 52, no. 3, pp. 987?98, 2015.
↑ Weisstein, Eric W. "Logarithmic Number." From MathWorld—A Wolfram Web Resource.
1 2 J. C. Kluyver. Euler's constant and natural numbers. Proc. K. Ned. Akad. Wet., vol. 27(1-2), 1924.
1 2 J.F. Steffensen. Interpolation (2nd Edn.). Chelsea Publishing Company, NewYork, USA, 1950.
1 2 3 L. Carlitz. A note on Bernoulli and Euler polynomials of the second kind. Scripta Math., vol. 25, pp. 323–330,1961.
1 2 3 4 5 6 7 8 Ia.V. Blagouchine. Two series expansions for the logarithm of the gamma function involving Stirling numbers and containing only rational coefficients for certain arguments related to $π$ ⁻¹. J.Math. Anal. Appl., 2015.
1 2 3 4 5 S.C. Van Veen. Asymptotic expansion of the generalized Bernoulli numbers Bn(n-1) for large values of n (n integer). Indag. Math. (Proc.), vol. 13, pp. 335–341, 1951.
1 2 M.W. Coffey. Series representations for the Stieltjes constants. Rocky Mountain J. Math., vol. 44, pp. 443–477, 2014.
↑ Ia.V. Blagouchine. A theorem for the closed-form evaluation of the first generalized Stieltjes constant at rational arguments and some related summations J. Number Theory, vol. 148, pp. 537–592 and vol. 151, pp. 276–277, 2015.
↑ B. Candelpergher and M.-A. Coppo. A new class of identities involving Cauchy numbers, harmonic numbers and zeta values. Ramanujan J., vol. 27, pp. 305–328, 2012.
↑ B. Candelpergher and M.-A. Coppo. A new class of identities involving Cauchy numbers, harmonic numbers and zeta values. Ramanujan J., vol. 27, pp. 305–328, 2012
↑ A269330
↑ A270857
↑ A270859
1 2 N. Nörlund. Vorlesungen über Differenzenrechnung. Springer, Berlin, 1924.
↑ Ia.V. Blagouchine. Expansions of generalized Euler's constants into the series of polynomials in $π$ ⁻² and into the formal enveloping series with rational coefficients only J. Number Theory, vol. 158, pp. 365-396, 2016.
↑ M. O. Rubinstein. Identities for the Riemann zeta function Ramanujan J., vol. 27, pp. 29–42, 2012.
1 2 3 Takao Komatsu. On poly-Cauchy numbers and polynomials, 2012.

This article is issued from Wikipedia - version of the 11/27/2016. The text is available under the Creative Commons Attribution/Share Alike but additional terms may apply for the media files.