Georgian Math. J.

2016; 23 (4):579–586

Research Article

Mansour Mahmoud* and Ravi P. Agarwal

Bounds for Bateman’s G-function and its

applications
DOI: 10.1515/gmj-2016-0037
Received May 3, 2015; accepted November 20, 2015

Abstract: In this paper, we present an asymptotic formula for Bateman’s G-function G(x) and deduce the
double inequality
1 1 1
< G(x) − < , x > 0.
2x2 + 3/2 x 2x2
k−1
k=1 k+h , h ≠ −1, −2, −3, . . . .
We apply this result to find estimates for the error term of the alternating series ∑∞ (−1)

Also, we study the monotonicity of some functions involving the function G(x). Finally, we propose a sharp
double inequality for the function G(x) as a conjecture.

Keywords: Digamma function, Bateman’s G-function, asymptotic formula, Laplace transform, sharp
inequality, monotonicity, alternating series

MSC 2010: 33B15, 26D15, 41A60, 41A80

1 Introduction
Let ψ(x) denote the digamma function, which is defined by
d
ψ(x) = log Γ(x), x ≠ 0, −1, −2, . . . ,
dx
where Γ(x) is an ordinary Gamma function (see [1]). The English mathematician H. Bateman (1882–1946)
used the ψ-function to introduce
x+1 x
G(x) = ψ( ) − ψ( ), x ≠ 0, −1, −2, . . .
2 2
which Erdélyi et al. [4] called Bateman’s G-function.
This function G(x) satisfies the following functional equations [4]:
2
G(1 + x) = − G(x), G(1 − x) = 2π csc(πx) − G(x),
x
1 m−1
G(mx) = ∑ (−1)r G(x + r/m), m = 1, 3, 5, . . . .
m r=0

An integral representation of the function G(x) is

∞
e−xt
G(x) = 2 ∫ dt, x > 0, (1.1)
1 + e−t
0

*Corresponding author: Mansour Mahmoud: Department of Mathematics, Faculty of Science, Mansoura University,
Mansoura 35516, Egypt, e-mail: mansour@mans.edu.eg
Ravi P. Agarwal: Department of Mathematics, Texas A & M University – Kingsville, 700 University Blvd, Kingsville,
TX 78363-8202, USA, e-mail: agarwal@tamuk.edu

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35
580 | M. Mahmoud and R. P. Agarwal, Bounds for Bateman’s G-function and its applications

and thus G(x) is completely monotonic on (0, ∞), i.e.

(−1)k G(k) (x) ≥ 0, x > 0, k = 0, 1, 2, . . . .

The function G(x) is also related to the hypergeometric function by the relation
1
G(x) = 2 F 1 (1, 1; 1 + x; 1/2)
x
(see [3]) which follows from the Erdélyi relation
2
G(x) = 2 F 1 (1, x; 1 + x; −1)
x
(see [4]).
Bateman’s G-function is useful in summing certain numerical and algebraic series [15]. For example,
∞
(−1)k 1 c
∑ = G( ), c ≠ 0, −b, −2b, . . . , (1.2)
k=0
bk + c 2b b

and its nth partial sum is given by

n
(−1)k 1 c c
∑ = [G( ) + (−1)n G( + n + 1)], c ≠ 0, −b, −2b, . . . . (1.3)
k=0
bk + c 2b b b

Qiu and Vuorinen [23] deduced that the function

1
h1 (x) = G(2x) −
2x
is strictly decreasing and convex from (0, ∞) onto (0, ∞). Also, they proved that the function x2 h1 (x) is
strictly increasing from (0, ∞) onto (0, 1/8), and in particular they obtained
1 4(1.5 − log 4) 1 1 1
+ < G(x) < + 2 , x> . (1.4)
x x2 x 2x 2
Motivated by this result, Mortici [9] showed that the function f a (x) : (0, ∞) → (0, ∞),
a
f a (x) = ψ(x + a) − ψ(x) − , a ∈ (0, 1),
x
(k)
is strictly completely monotonic, that is, (−1)k f a (x) > 0 for x > 0 and k = 0, 1, 2, . . . . This result improves
the result of Qiu and Vuorinen. As a direct consequence, Mortici established the inequality
1
0 < ψ(x + a) − ψ(x) ≤ ψ(a) + γ + − a, x ≥ 1, a ∈ (0, 1),
a
where γ is the Euler constant. For more information about the digamma function and related inequalities, we
refer to [5, 9–14, 17–21] and the closely related references therein.
In this paper, we deduce an asymptotic expansion of the function G(x) and prove that the function
G(x) − xar for x > 0 is strictly completely monotonic if a < 0 (a ≤ 1) with r = 2, 3, 4, . . . (r = 1). We also present
an improvement of the lower bound of inequity (1.4). Our results generalize the main theorem of [29] and a
result of [7] on the error term of the alternating series. Finally, we conjecture a sharp double inequality for
Bateman’s G-function.

2 Main results
In the following lemma, we develop an asymptotic formula for the function G(x).

Lemma 2.1. The asymptotic formula for Bateman’s G-function is given by

1 ∞ (22k − 1)B2k
G(x) ∼ +∑ , x → ∞. (2.1)
x k=1 kx2k

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35
 M. Mahmoud and R. P. Agarwal, Bounds for Bateman’s G-function and its applications | 581

Proof. We note that the Laplace transform

∞
2
∫ e−xt dt, x > 0,
1 + e−t
0

converges for all sufficiently large x, and the function q(t) = 2

1+e−t
is infinitely differentiable in a neighborhood
of the origin. Now since
∞
2(22m − 1)B2m 2m−1
q(t) = 1 + tanh(t/2) = 1 + ∑ t , |t| < π,
m=1
(2m)!

where B r are the Bernoulli numbers, by Watson’s lemma (see [2, 16]) it follows that
∞
2 ∞
q(k) (0)
∫ e−xt dt ∼ ∑ k+1 , x → ∞,
1+e −t
k=0
x
0

where
(22k − 1)B2k
q(0) (0) = 1, q(2k) (0) = 0, q(2k−1) (0) = , k = 1, 2, 3, . . . ,
k
and hence
∞
2 1 ∞ (22k − 1)B2k
∫ e−xt dt ∼ + ∑ , x → ∞.
1+e −t x k=1 kx2k
0

Motivated by the results of Mortici [9], in the following lemma, we will prove the monotonicity of the function
M a,r (x) = G(x) − xar for x > 0, r ∈ ℕ and some different values of a.

Lemma 2.2. The functions

a
M a,1 (x) = G(x) − , a ≤ 1, x > 0,
x
and
a
M a,r (x) = G(x) − , a < 0, x > 0, r = 2, 3, 4, . . . ,
xr
are strictly completely monotonic.

Proof. Using the formula

∞
1 1
= ∫ t r−1 e−xt dt, r ∈ ℕ,
x r (r − 1)!
0

and relation (1.1), we obtain for n = 0, 1, 2, 3, . . . that

∞ ∞
(n) 2 a t n e−xt
(−1) n
M a,r (x) = ∫( − t r−1 n −xt
)t e dt = ∫ ϕ r (t) dt,
1 + e−t (r − 1)! e2t − 1
0 0

where
a
ϕ r (t) = 2(e2t − e t ) − t r−1 (e2t − 1).
(r − 1)!
Clearly,
∞
2k+1 t k
ϕ r (t) = ∑ P r (t, k),
k=1
k!

where
1 k a
P r (t, k) = 1 − ( ) − t r−1 .
2 2(r − 1)!
Now we consider the following two cases.

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35
582 | M. Mahmoud and R. P. Agarwal, Bounds for Bateman’s G-function and its applications

Case 1: If r = 1, then
1 k a
P1 (t, k) = 1 − ( ) − ,
2 2
which is positive if and only if
a 1 k
≤1−( ) , k = 1, 2, 3, . . . .
2 2
Thus ϕ1 (t) > 0 for all t ≥ 0 if and only if a ≤ 1.

Case 2: If r = 2, 3, 4, . . . , then P r (t, k) is increasing as a function of t if a < 0 with P r (0, k) = 1 − ( 12 )k > 0.

Thus, ϕ r (t) > 0 for all t ≥ 0, r = 2, 3, . . . if and only if a < 0.

As a consequence, we find that the function M a,1 (x) is strictly decreasing for a ≤ 1, i.e.,

M a,1 (x) > lim (M a,1 (x)), a ≤ 1,

x→∞

and by the asymptotic formula (2.1), we have

a
lim (M a,1 (x)) = lim (G(x) − ) = 0.
x→∞ x→∞ x
Hence we have the following result.

Corollary 2.3. The following inequality holds:

1
G(x) > , x > 0.
x
We are now ready to obtain the bounds for the function G(x) using the second term of our asymptotic expan-
sion (2.1).

Theorem 2.4. For x > 0,

1 1 1
< G(x) − < . (2.2)
2x2 + 3/2 x 2x2
Proof. Consider the function
et − 1 t
f(t) = − , t ≥ 0.
et + 1 2
Then
(e t − 1)2
f 󸀠 (t) = −
2(e t + 1)2
shows that the function f(t) is decreasing with f(0) = 0. Thus f(t) < 0, t ≥ 0, and hence the Laplace transform
of the function f(t) satisfies
Lf(t) < 0,
which gives
1 1
G(x) − < , x > 0.
x 2x2
Now, consider the function
t3
K(t) = f(t) + , t ≥ 0,
24
for which
β(t) α(t)
K 󸀠 (t) = ,
8(e t + 1)2
where
β(t) = (t − 2) + (2 + t)e t and α(t) = (t + 2) + (t − 2)e t .
Clearly, β(t) is an increasing function and β(0) = 0, hence β(t) ≥ 0 for t ≥ 0. Also, the function α(t) is convex
for t ≥ 0 and its minimum is at t = 0. Thus α(t) ≥ α(0) = 0 for t ≥ 0. Hence it follows that
et − 1 t t3
≥ − , t ≥ 0. (2.3)
e + 1 2 24
t

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35
 M. Mahmoud and R. P. Agarwal, Bounds for Bateman’s G-function and its applications | 583

Using
z3 z5 π
sin z ≤ z − + , 0≤z≤
6 120 2
(see [22] and also [8]), we get

sin( √23 t) t t3 3t5 π

− ≥ −( − + ), 0≤t≤ . (2.4)
√3 2 16 1280 √3

Using inequalities (2.3) and (2.4), we find that the function

√
e t − 1 sin( 23 t)
g(t) = − , t ≥ 0,
et + 1 √3

satisfies
t3 (80 − 9t2 ) π
g(t) > > 0, 0≤t≤ . (2.5)
3840 √3
Also,
et − 1 1 et − 3
g(t) > − > >0 if t > ln 3. (2.6)
e t + 1 2 2(e t + 1)
Using (2.5) and (2.6), we obtain g(t) > 0 for t ≥ 0. Thus the Laplace transform of the function g(t) must satisfy

Lg(t) > 0

and hence
1 1
G(x) − > , x > 0.
x 2x2 + 3/2
Remark 2.5. The lower bound of inequality (2.2) improves the lower bound of inequality (1.4) for

9 − 6 ln 4
x>√ ≈ 2.748.
8 ln 4 − 11

3 Applications
Let ∑∞k=1 (−1) a k be an alternating series such that {a k } is a positive decreasing sequence converging to zero.
k

Using Leibniz’ convergence test, this series converges and | ∑∞ k=n+1 (−1) a k | < a n+1 . Several authors studied
k

the problem of finding estimates for the error term | ∑k=n+1 (−1) a k | for a specific sequence a k , see [7, 24–29].
∞ k

Among the early important results of the calculus there were the following results of Leibniz and Gregory
on the alternating series [6]:
∞
(−1)k−1 ∞
(−1)k−1 π
∑ = ln 2 and ∑ = .
k=1
k k=1
2k − 1 4

By considering the function

π/4

I(n) = ∫ tann θdθ, n = 0, 1, 2, . . . ,

0

Kazarinoff [6] deduced the relations

󵄨󵄨 󵄨 󵄨 󵄨
󵄨π n
(−1)k 󵄨󵄨󵄨 󵄨󵄨󵄨 ∞ (−1)k−1 󵄨󵄨󵄨
I(2n) = 󵄨󵄨󵄨󵄨 − ∑ 󵄨󵄨 = 󵄨󵄨 ∑ 󵄨󵄨,
󵄨󵄨 4 k=1 2k − 1 󵄨󵄨󵄨 󵄨󵄨󵄨k=n+1 2k − 1 󵄨󵄨󵄨
󵄨 󵄨 󵄨 󵄨
1 󵄨󵄨 n
(−1)k+1 󵄨󵄨󵄨 1 󵄨󵄨󵄨 ∞ (−1)k−1 󵄨󵄨󵄨
I(2n + 1) = 󵄨󵄨󵄨󵄨ln 2 − ∑ 󵄨󵄨 = 󵄨󵄨 ∑ 󵄨󵄨.
2 󵄨󵄨 k=1
k 󵄨󵄨󵄨 2 󵄨󵄨󵄨k=n+1 k 󵄨󵄨󵄨

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35
584 | M. Mahmoud and R. P. Agarwal, Bounds for Bateman’s G-function and its applications

He also deduced the estimates

1 1
< I(2n) < (3.1)
2(2n + 1) 2(2n − 1)
and
1 1
< I(2n + 1) < . (3.2)
4(n + 1) 4n
Tóth and Bukor [29] refined (3.2) to
󵄨󵄨 ∞ 󵄨
1 󵄨 (−1)k−1 󵄨󵄨󵄨 1
≤ 󵄨󵄨󵄨󵄨 ∑ 󵄨󵄨 < , n ≥ 1, (3.3)
2n + C1 󵄨󵄨k=n+1 k 󵄨󵄨󵄨 2n + C2

where C1 = 21−ln 2 and C 2 = 1 are the best constants.

ln 2−1

Koumandos [7] refined (3.1) as follows:

󵄨󵄨 ∞ 󵄨
1 󵄨 (−1)k−1 󵄨󵄨󵄨 1
≤ 󵄨󵄨󵄨󵄨 ∑ 󵄨󵄨 < , n ≥ 1, (3.4)
4n + C3 󵄨󵄨k=n+1 2k − 1 󵄨󵄨󵄨 4n + C4

where the constants C3 = 4−π

4
− 4 and C4 = 0 are the best possible ones.
Our next result generalizes the bounds of inequalities (3.3) and (3.4).

Lemma 3.1. For h ≠ −1, −2, −3, . . . and h + n + 1 > 0, one has
󵄨󵄨 ∞ 󵄨
4(h + n)2 + 10(h + n) + 9 󵄨󵄨 (−1)k−1 󵄨󵄨󵄨 2(h + n) + 3
< 󵄨󵄨 ∑ 󵄨󵄨 < .
󵄨 󵄨
2(h + n + 1)[4(h + n) + 8(h + n) + 7] 󵄨󵄨k=n+1 k + h 󵄨󵄨 4(h + n + 1)2
2

Proof. Let h ≠ −1, −2, −3, . . . and h + n + 1 > 0. Using (1.2) and (1.3), we have
∞
(−1)k−1 (−1)n
∑ = G(h + n + 1)
k=n+1
k+h 2

and hence
󵄨󵄨 ∞ 󵄨
󵄨󵄨 (−1)k−1 󵄨󵄨󵄨 1
󵄨󵄨 ∑ 󵄨󵄨 = G(h + n + 1).
󵄨󵄨 󵄨
󵄨k=n+1 k + h 󵄨󵄨 2
Using (2.2), we get
󵄨󵄨 ∞ 󵄨
1 1 󵄨󵄨 (−1)k−1 󵄨󵄨󵄨 1 1
+ < 󵄨󵄨 ∑ 󵄨󵄨 < + .
2 󵄨 󵄨
4(h + n + 1) + 3 2(h + n + 1) 󵄨󵄨k=n+1 k + h 󵄨󵄨 4(h + n + 1)2 2(h + n + 1)

Special cases. (1) In the case of h = 0, we get

󵄨󵄨 ∞ 󵄨
4n2 + 10n + 9 󵄨󵄨 (−1)k−1 󵄨󵄨󵄨 2n + 3
< 󵄨󵄨 ∑ 󵄨󵄨 < ,
2(n + 1)(4n2 + 8n + 7) 󵄨󵄨󵄨k=n+1 k 󵄨󵄨󵄨 4(n + 1)2

which improves inequality (3.3) for n > 1.

(2) In the case of h = −1/2, we get
󵄨󵄨 ∞ 󵄨
󵄨󵄨 (−1)k−1 󵄨󵄨󵄨 1
󵄨󵄨 ∑ 󵄨󵄨 = G(n + 1/2)
󵄨󵄨 󵄨
󵄨k=n+1 2k − 1 󵄨󵄨 4

and hence
󵄨󵄨 ∞ 󵄨
4n2 + 6n + 5 󵄨󵄨 (−1)k−1 󵄨󵄨󵄨 (n + 1)
< 󵄨󵄨 ∑ 󵄨󵄨 < ,
8(2n + 1)(n2 + n + 1) 󵄨󵄨󵄨k=n+1 2k − 1 󵄨󵄨󵄨 (2n + 1)2

which improves inequality (3.4) for n > 1.

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35
 M. Mahmoud and R. P. Agarwal, Bounds for Bateman’s G-function and its applications | 585

4 A conjecture
Motivated by the asymptotic expansion of the function G(x), we make the following conjecture.

Conjecture 4.1. For m = 1, 2, 3, . . . , we have

2m
(22k − 1)B2k 1 2m−1 (22k − 1)B2k
∑ < G(x) − < ∑ , x > 0,
k=1
kx2k x k=1
kx2k

where the bounds are sharp.

References
[1] G. E. Andrews, R. Askey and R. Roy, Special Functions, Encyclopedia Math. Appl. 71, Cambridge University Press, Cam-
bridge, 1999.
[2] E. T. Copson, Asymptotic Expansions, Cambridge Tracts in Math. Math. Phys. 55, Cambridge University Press, New York,
1965.
[3] W. Ehrhardt, The AMath and DAMath special functions: Reference manual and implementation notes, version 2.03, 2014,
www.wolfgang-ehrhardt.de/specialfunctions.pdf.
[4] A. Erdélyi, Higher Transcendental Functions. Vols. I–III. Based on Notes Left by Harry Bateman, McGraw-Hill Book Com-
pany, New York, 1953–1955.
[5] B.-N. Guo and F. Qi, Sharp inequalities for the psi function and harmonic numbers, Analysis (Berlin) 34 (2014), no. 2,
201–208.
[6] N. D. Kazarinoff, Analytic Inequalities, Holt, Rinehart and Winston, New York, 1961.
[7] S. Koumandos, Monotonicity of some functions involving the psi function and sharp estimates for the remainders of cer-
tain series, Int. J. Contemp. Math. Sci. 2 (2007), no. 13–16, 713–720.
[8] J.-C. Kuang, Applied Inequalities (in Chinese), 3rd ed., Shandong Science and Technology Press, Jinan City, 2004.
[9] C. Mortici, A sharp inequality involving the psi function, Acta Univ. Apulensis Math. Inform. 21 (2010), 41–45.
[10] C. Mortici, Estimating the digamma and trigamma functions by completely monotonicity arguments, Appl. Math. Comput.
217 (2010), no. 8, 4081–4085.
[11] C. Mortici, New approximations of the gamma function in terms of the digamma function, Appl. Math. Lett. 23 (2010),
no. 1, 97–100.
[12] C. Mortici, Accurate estimates of the gamma function involving the PSI function, Numer. Funct. Anal. Optim. 32 (2011),
no. 4, 469–476.
[13] C. Mortici, The quotient of gamma functions by the psi function, Comput. Appl. Math. 30 (2011), no. 3, 627–638.
[14] C. Mortici and C.-P. Chen, New sharp double inequalities for bounding the gamma and digamma function, An. Univ. Vest
Timiş. Ser. Mat.-Inform. 49 (2011), no. 2, 69–75.
[15] K. Oldham, J. Myland and J. Spanier, An Atlas of Functions. With Equator, the Atlas Function Calculator, 2nd ed., Springer,
New York, 2009.
[16] F. W. J. Olver, D. W. Lozier, R. F. Boisvert and C. W. Clark, NIST Handbook of Mathematical Functions, Cambridge University
Press, Cambridge, 2010.
[17] F. Qi, Limit formulas for ratios between derivatives of the gamma and digamma functions at their singularities, Filomat 27
(2013), no. 4, 601–604.
[18] F. Qi, P. Cerone and S. S. Dragomir, Complete monotonicity of a function involving the divided difference of PSI functions,
Bull. Aust. Math. Soc. 88 (2013), no. 2, 309–319.
[19] F. Qi, S.-X. Chen and W.-S. Cheung, Logarithmically completely monotonic functions concerning gamma and digamma
functions, Integral Transforms Spec. Funct. 18 (2007), no. 5–6, 435–443.
[20] F. Qi and B.-N. Guo, Complete monotonicities of functions involving the gamma and digamma functions, RGMIA Res. Rep.
Coll. 7 (2014), no. 1, 63–72.
[21] F. Qi and B.-N. Guo, Complete monotonicity of divided differences of the di- and tri-gamma functions with applications,
Georgian Math. J. 23 (2016), no. 2, 279–291.
[22] F. Qi, D.-W. Niu and B.-N. Guo, Refinements, generalizations, and applications of Jordan’s inequality and related problems,
J. Inequal. Appl. 2009 (2009), Article ID 271923.
[23] S.-L. Qiu and M. Vuorinen, Some properties of the gamma and psi functions, with applications, Math. Comp. 74 (2005),
no. 250, 723–742.
[24] A. Sîntămărian, A new proof for estimating the remainder of the alternating harmonic series, Creat. Math. Inform. 21
(2012), no. 2, 221–225.

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35
586 | M. Mahmoud and R. P. Agarwal, Bounds for Bateman’s G-function and its applications

[25] A. Sîntămărian, Sharp estimates regarding the remainder of the alternating harmonic series, Math. Inequal. Appl. 18
(2015), no. 1, 347–352.
[26] V. Timofte, On Leibniz series defined by convex functions, J. Math. Anal. Appl. 300 (2004), no. 1, 160–171.
[27] V. Timofte, Integral estimates for convergent positive series, J. Math. Anal. Appl. 303 (2005), no. 1, 90–102.
[28] L. Tóth, On a class of Leibniz series, Rev. Anal. Numér. Théor. Approx. 21 (1992), no. 2, 195–199.
[29] L. Tóth and J. Bukor, On the alternating series 1 − 12 + 13 − 14 + ⋅ ⋅ ⋅, J. Math. Anal. Appl. 282 (2003), no. 1, 21–25.

Bereitgestellt von | De Gruyter / TCS

Angemeldet
Heruntergeladen am | 23.01.17 09:35