Demonstratio Mathematica 2022; 55: 710–736

Research Article

Feng Qi*

Taylor’s series expansions for real powers of

two functions containing squares of inverse
cosine function, closed-form formula for
specific partial Bell polynomials, and series
representations for real powers of Pi#
https://doi.org/10.1515/dema-2022-0157
received April 16, 2022; accepted August 12, 2022

Abstract: In this article, by virtue of expansions of two finite products of finitely many square sums, with the
aid of series expansions of composite functions of (hyperbolic) sine and cosine functions with inverse sine
and cosine functions, and in the light of properties of partial Bell polynomials, the author establishes
Taylor’s series expansions of real powers of two functions containing squares of inverse (hyperbolic) cosine
functions in terms of the Stirling numbers of the first kind, presents a closed-form formula of specific partial
Bell polynomials at a sequence of derivatives of a function containing the square of inverse cosine function,
derives several combinatorial identities involving the Stirling numbers of the first kind, demonstrates
several series representations of the circular constant Pi and its real powers, recovers Maclaurin’s series
expansions of positive integer powers of inverse (hyperbolic) sine functions in terms of the Stirling numbers
of the first kind, and also deduces other useful, meaningful, and significant conclusions and an application
to the Riemann zeta function.

Keywords: Taylor’s series expansion, Maclaurin’s series expansion, real power, inverse cosine function,
inverse hyperbolic cosine function, inverse sine function, inverse hyperbolic sine function, Stirling number
of the first kind, combinatorial identity, composite, series representation, circular constant, partial Bell
polynomial, closed-form formula, Riemann zeta function

MSC 2020: Primary 41A58, Secondary 05A19, 11B73, 11B83, 11C08, 11S05, 12D05, 26A24, 26C05, 33B10

1 Simple preliminaries
In this article, we use the notations
 = {1, 2, …} , 0 = {0, 1, 2, …} , − = { −1, −2, …} ,  = {0, ±1, ±2, …} .

The rising factorial, or say, the Pochhammer symbol, of β ∈  is defined [1, p. 7497] by


#
Dedicated to my 80-year-old father, Shu-Gong Qi, and to my 18-month-old grandson, Magnus Xi-Zhe Qi.


* Corresponding author: Feng Qi, Institute of Mathematics, Henan Polytechnic University, Jiaozuo 454010, China; Independent
Researcher, Dallas, TX 75252-8024, USA, e-mail: [email protected], [email protected], [email protected]

Open Access. © 2022 Feng Qi, published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0
International License.
 Expansions for real powers of two functions and Bell polynomials  711

n−1
β (β + 1)⋯(β + n − 1) , n ∈ ;
(β ) n = ∏ (β + k ) = ⎧⎨1, n = 0.
(1.1)
k=0 ⎩
For α, β ∈  with α + β ∈  ⧹{0, − 1, −2, …} , extended Pochhammer symbol (β )α is defined [2] by
Γ(α + β )
(β )α = , (1.2)
Γ(β )

where the classical Euler gamma function Γ(z ) can be defined [3, Chapter 3] by
n ! nz
Γ(z ) = lim n , z ∈  ⧹{0, −1, −2, …} .
k = 0 (z + k)
n →∞ ∏

The falling factorial for β ∈  and n ∈  is defined by

n−1
β (β − 1)⋯(β − n + 1) , n ∈ ;
⟨β⟩n = ∏ (β − k ) = ⎧⎨1, n = 0.
k=0 ⎩
The extended binomial coefficient ( ) is defined [4] by
z
w

⎧ Γ(z + 1)
, z ∉ − , w , z − w ∉ − ;
⎪ Γ(w + 1)Γ(z − w + 1)
⎪ 0, z ∉ − , w ∈ − or z − w ∈ − ;
⎪
⎪ ⟨z ⟩w
z , z ∈ − , w ∈ 0 ;
⎛ ⎞ = w! (1.3)
⎝w⎠ ⎨
⎪ ⟨z ⟩z − w , z , w ∈ − , z − w ∈ 0 ;
⎪ (z − w ) !
⎪
⎪ 0, z , w ∈ − , z − w ∈ − ;
⎩∞, z ∈ − , w ∉  .

The Stirling numbers of the first kind s(n , k ) for n ≥ k ≥ 0 can be generated [3, p. 20, (1.30)] by
∞
[ln(1 + x )]k xn
= ∑ s (n , k ) , ∣x∣ < 1 (1.4)
k! n=k
n!

and satisfy diagonal recursive relations

n ℓ
s (n + k , k ) ⟨k ⟩ℓ ℓ s (n + m , m )
= ∑ (−1)ℓ ∑ (−1)m⎛ ⎞ n + m
( )
n+k
k
ℓ= 0
ℓ ! m=0 ⎝m⎠ ( ) m

and
k−1
n ℓ−1
s(n , k ) = ( −1)n − k ∑ ( −1)ℓ⎛ ⎞⎛ ⎞s(n − ℓ , k − ℓ) .
ℓ= 0 ⎝ ℓ ⎠⎝ k − n − 1⎠

See [5, p. 23, Theorem 1.1], [6, Remark 2.1], and [7, p. 156, Theorem 4].
The partial Bell polynomials, or say, the Bell polynomials of the second kind, can be denoted and
defined by
n−k+1
n! x ℓi
Bn, k (x1, x2 , … , xn − k + 1) = ∑ ∏ ⎛ i⎞ .
1≤i≤n−k+1 ∏ni =−1k + 1ℓi! i=1 ⎝i! ⎠
ℓi ∈ {0} ∪ 
n−k+1
∑ i ℓi = n
i=1
n−k+1
∑ ℓi = k
i=1

in [8, p. 412, Definition 11.2] and [9, p. 134, Theorem A].

712  Feng Qi

According to [10, p. 20], the inverse cosine function arccos z can be defined by
π
[
⎧ + i ln (1 − z 2 )1 / 2 + iz , ] z ∈  ⧹( −∞ , −1) ∪ (1, ∞) ;
⎪2
arccos z = π ∓ i ln (x 2 − 1)1 / 2 − x ,
[ ] x ∈ ( −∞ , −1];
⎨
⎪ 2 [ 1/2
⎩∓ i ln (x − 1) + x , ] x ∈ [1, ∞) ,

where the upper/lower signs corresponding to the upper/lower sides of the rays ( −∞ , −1] and [1, ∞), and
the notation ln z denotes the principal value of the logarithmic function Lnz in z ∈  ⧹{0} .

2 Motivations
Let f (z ) and h(z ) be infinitely differentiable functions such that the function f (z ) has the formal series
expansion f (z ) = ∑∞ k
k = 0 ckz and the composite function h( f (z )) is defined on a non-empty open interval.
A natural problem is to find the series expansion of the composite function h( f (z )). This problem can be
regarded as how to compute derivatives of the composite function h( f (z )). There have been a long history
and a number of literature in textbooks, handbooks, monographs, and research articles on this problem.
See [2,3,11–13], for example.
In the above general theory, the cases h(z ) = z r for r ∈  ⧹{1} and f (z ) being concrete elementary
functions are of special interest and attract some mathematicians. We recall some results as follows.
Equation (1.4) can be rearranged as Maclaurin’s series expansions of the power function
k ∞
⎡ ln(1 + x ) ⎤ = s (n + k , k ) x n
∑
⎣ x ⎦ n=0 ( )
n+k
k
n!

for ∣x∣ < 1 and k ≥ 0. The Stirling numbers of the second kind S (n , k ) for n ≥ k ≥ 0 can be generated [14,
pp. 131–132] by
∞
(e x − 1)k xn
= ∑ S (n , k ) . (2.1)
k! n=k
n!

Equation (2.1) can be rearranged as Maclaurin’s series expansions of the power function
x k ∞
⎛e − 1⎞ = S (n + k , k ) x n
∑ , k ≥ 0.
⎝ x ⎠ n=0 ( )
n+k
k
n!

In [15, p. 377, (3.5)] and [16, pp. 109–110, Lemma 1], it was obtained that
∞
1 (μ + ν + n + 1)n x 2n + μ + ν
Iμ(x )Iν (x ) = ∑ ⎛ ⎞ ,
Γ(μ + 1)Γ(ν + 1) n = 0 n ! (μ + 1)n(ν + 1)n ⎝ 2 ⎠

where the first kind of modified Bessel function Iν (z ) can be represented [11, p. 375, 9.6.10] by
∞
1 z 2n + ν
Iν (z ) = ∑ ⎛ ⎞ , z ∈ .
n=0
n ! Γ(ν + n + 1) ⎝ 2 ⎠

In [17, p. 310], the power series expansion

∞ 2k + 2ν
1 ⎛ 2k + 2ν ⎞⎛ z ⎞
[Iν (z )]2 = ∑ 2
k=0
[Γ(ν + k + 1)] ⎝ k ⎠⎝ 2 ⎠

was listed. As for the series expansion of the function [Iν (z )]r for ν ∈  ⧹{ −1, −2, …} and r , z ∈  , recursive
formulas were investigated in [18,17,19–21]. One of the reasons why one investigated the series expansions
 Expansions for real powers of two functions and Bell polynomials  713

of the functions [Iν (z )]r is that the products of the (modified) Bessel functions of the first kind appear
frequently in problems of statistical mechanics and plasma physics, see [22–24].
In the articles [1,12,25–29], Maclaurin’s series expansions of the powers
m m m m
⎛ arcsin z ⎞ , (arcsin z ) , ⎛ arcsinhz ⎞ , (arcsinhz ) , (arctan z )m , (arctanhz )m ,
⎝ z ⎠ 1 − z2 ⎝ z ⎠ 1 + z2
sinm z , cosm z , tanm z , cotm z , secm z , cscmz

for m ≥ 2 and their history were reviewed, surveyed, established, discussed, and applied. Now we recite the
following two series expansions.

arcsin x k
Theorem 2.1. ([27, Theorem 1]) For k ∈  and ∣x∣ < 1, the function ( x ) , whose value at x = 0 is defined to
be 1, has Maclaurin’s series expansion
k ∞
⎛ arcsin x ⎞ = 1 + Q(k , 2m) (2x )2m
∑ (−1)m , (2.2)
⎝ x ⎠ m=1 ( k + 2m
k )
(2m) !

where
m
k + ℓ − 1⎞ k + m − 2 ⎞ℓ
Q (k , m ) = ∑⎛ s(k + m − 1, k + ℓ − 1)⎛ (2.3)
ℓ= 0 ⎝
k−1 ⎠ ⎝ 2 ⎠
for k ∈  and m ≥ 2 .

arcsinhx k
Theorem 2.2. ([27, Corollary 2]) For k ∈  and ∣x∣ < ∞, the function ( x ) , whose value at x = 0 is
defined to be 1, has Maclaurin’s series expansion
k ∞
⎛ arcsinhx ⎞ = 1 + Q(k , 2m) (2x )2m
∑ , (2.4)
⎝ x ⎠ m=1 (
k + 2m
)
(2m) !
k

where Q(k , 2m) is given by (2.3).

In the articles [1,27], the series expansion (2.2) has been applied to derive closed-form formulas for
specific partial Bell polynomials and to establish series representations of the generalized logsine function.
These results are needed and considered in [30–33], respectively.
In the community of mathematics, the circular constant Pi has attracted a number of mathematicians
spending long time and utilizing many methods to calculate it. The setting-up of the international Pi Day is
2
the best demonstration of the importance of the circular constant Pi. Taking x = 2
in (2.2) produces the
series representation
k ∞
⎛ π ⎞ = 1 + k ! ∑ ( −1)m2m Q(k , 2m) .
⎜ ⎟ (2.5)
⎝2 2 ⎠ m=1
(k + 2m) !

In this article, by virtue of expansions of two finite products of finitely many square sums
k k
∏(ℓ2 + α2) and ∏[(2ℓ − 1)2 + α2]
ℓ= 1 ℓ= 1

for k ∈  in Lemmas 3.1 and 3.2, with the aid of Taylor’s series expansions around x = 1 of the functions
cosh(α arccos x ) and cos(α arccos x ) in Lemma 3.3, and in the light of properties of partial Bell polynomials
selected in Lemma 3.4, we will
(1) establish Taylor’s series expansions around x = 1 of the functions
2 α 2 k
⎡ (arccos x ) ⎤ and ⎡ (arccoshx ) ⎤
⎢
⎣ 2(1 − x ) ⎥ ⎦ ⎢
⎣ 2(1 − x ) ⎥ ⎦
for α ∈  and k ∈  in terms of Q(k , m) in Theorems 4.1 and 5.2;
714  Feng Qi

(2) present a closed-form formula of the specific partial Bell polynomials

1 2 3 32 80 (2m − 2k + 2) !!
Bm, k ⎛− ,
⎜ ,− , ,− ,…, Q(2, 2m − 2k + 2)⎞ ⎟

⎝ 12 45 70 525 693 (2m − 2k + 4) ! ⎠

for m ≥ k ∈  in Theorem 5.1;
(3) derive some interesting combinatorial identities involving the Stirling numbers of the first kind s(n , k )
in Lemmas 3.1 and 3.2, in Corollaries 4.7 and 5.1, and in the proof of Theorem 5.1;
(4) demonstrate several series representations of the circular constant π and its powers π α for α ∈  in
Corollaries 4.2 and 5.2;
(5) recover Maclaurin’s series expansions (2.2) and (2.4) in Theorem 2.1 and 2.2; and
(6) deduce some other useful, meaningful, and significant conclusions and an application to the Riemann
zeta function ζ (x ) in Corollaries 4.1, 4.3, 4.4, 4.5, and 4.6, in Remarks 3.3 and 4.1, in Section 7, and
elsewhere in this article.

3 Important lemmas
For attaining our aims mentioned just now, we need the following four important lemmas.

Lemma 3.1. For k ∈  and α ∈  , we have

k k 2k + 1
∏(ℓ2 + α2) = (−1)k ∑ (−1) j⎢ ∑
⎡ ⎛ ℓ ⎞s(2k + 1, ℓ)k ℓ− 2j − 1⎤α2j . (3.1)
⎜

2j + 1 ⎠
⎟
⎥
ℓ= 1 j=0 ⎣ ℓ= 2j + 1 ⎝ ⎦
In particular, when α = 0 in (3.1),
2k
∑ (ℓ + 1)s(2k + 1, ℓ + 1)k ℓ = (−1)k (k ! )2 . (3.2)
ℓ= 0

For 0 ≤ j ≤ k − 1, we have
2k − 1
∑ ⎜
⎛ ℓ ⎞s(2k − 1, ℓ)(k − 1)ℓ− 2j = −s(2k − 1, 2j ) .
⎟ (3.3)
2j
ℓ= 2j + 1 ⎝ ⎠

Proof. In [14, p. 165, (12.1)], there exists the formula

n
z
n !⎛ ⎞ = ∑ s (n , k )z k , z ∈ , n ≥ 0. (3.4)
n
⎝ ⎠ k=0

It is not difficult to verify that

k k−1
(iα)k Γ(iα + k )
∏[(ℓ − 1)2 + α2] = ∏(ℓ2 + α2) = (−1)k = ( −1)k iα
(iα + 1)−k Γ(iα − k + 1)
ℓ= 1 ℓ= 0
(3.5)
iα + k − 1 ⎞
= ( −1)k iα(2k − 1) ! ⎛
⎝ 2k − 1 ⎠
for α ∈  and k ∈  , where i = −1 is the imaginary unit and α ∈  . Substituting formula (3.4) into (3.5)
results in
k−1 2k − 1
∏(ℓ2 + α2) = (−1)k iα ∑ s(2k − 1, ℓ)(iα + k − 1)ℓ
ℓ= 0 ℓ= 0
2k − 1 ℓ
ℓ⎞
= ( −1)k iα ∑ s(2k − 1, ℓ) ∑ ⎛ ⎜(iα) j (k − 1)ℓ− j
⎟

ℓ= 0
j
j=0⎝ ⎠
 Expansions for real powers of two functions and Bell polynomials  715

2k − 1 2k − 1
⎡ ℓ⎞ ⎤
= ( −1)k ∑⎢∑⎛ s(2k − 1, ℓ)(k − 1)ℓ− j ⎥(iα) j + 1
⎜ ⎟

j
j = 0 ⎣ ℓ= j ⎝ ⎠ ⎦
2k − 1 2k − 1
⎡ ℓ⎞ ⎤ ( j + 1)π
= ( −1)k ∑⎢∑⎛ s(2k − 1, ℓ)(k − 1)ℓ− j ⎥α j + 1 cos
⎜ ⎟

j
j = 0 ⎣ ℓ= j ⎝ ⎠
2
⎦
2k − 1 2k − 1
⎡ ℓ⎞ ⎤ ( j + 1)π
+ i ( −1)k ∑⎢∑⎛ s(2k − 1, ℓ)(k − 1)ℓ− j ⎥α j + 1 sin
⎜ ⎟

j
j = 0 ⎣ ℓ= j ⎝ ⎠
2
⎦
for α ∈  and k ∈ , where we used the identity
kπ kπ
i k = cos + i sin , k ≥ 0. (3.6)
2 2
As a result, equating the real and imaginary parts, we obtain
k−1 2k − 1 2k − 1
( j + 1)π ⎡ ⎛ ℓ ⎞s(2k − 1, ℓ)(k − 1)ℓ− j ⎤α j + 1
∏(ℓ2 + α2) = (−1)k ∑ cos ⎢s(2k − 1, j ) + ∑ ⎜ ⎟
⎥ (3.7)
ℓ= 0 j=0
2 j
ℓ= j + 1 ⎝ ⎠
⎣ ⎦
and
2k − 1 2k − 1
( j + 1)π ⎡ ⎛ ℓ ⎞s(2k − 1, ℓ)(k − 1)ℓ− j ⎤α j + 1 = 0.
∑ sin ⎢s(2k − 1, j ) + ∑ ⎜ ⎟
⎥ (3.8)
j=0
2 j
ℓ= j + 1 ⎝ ⎠
⎣ ⎦
(2j + 1)π
Since cos( jπ ) = ( −1) j and cos 2 = 0 for j ∈  , equality (3.7) can be simplified as (3.1).
Taking α → 0 in (3.1) reduces to (3.2).
(2j + 1)π
Since sin( jπ ) = 0 and sin 2
= ( −1) j for j ∈  , equality (3.8) becomes
k−1 2k − 1
⎡
∑ (−1) j⎢s(2k − 1, 2j ) + ∑ ⎛ ℓ ⎞s(2k − 1, ℓ)(k − 1)ℓ− 2j ⎤α2j + 1 = 0.
2j
⎜ ⎟
⎥
j=0 ⎣ ℓ= 2j + 1 ⎝ ⎠ ⎦
Further regarding α as a variable leads to
2k − 1
⎡
( −1) j ⎢s(2k − 1, 2j ) + ∑ ⎛ ℓ ⎞s(2k − 1, ℓ)(k − 1)ℓ− 2j ⎤ = 0,
2j
⎜ ⎟
⎥
⎣ ℓ= 2j + 1 ⎝ ⎠ ⎦
which is equivalent to the combinatorial identity (3.3). The proof of Lemma 3.1 is complete. □

Lemma 3.2. For k ∈  and α ∈  , we have

k 2k 2k
⎡ s(2k , ℓ) ⎛ ℓ ⎞ ⎤
∏[(2ℓ − 1)2 + α2] = (−1)k 22k ∑ (−1) j⎢ ∑ ℓ
⎜(2k − 1)ℓ− 2j ⎥α2j .
⎟ (3.9)
j=0
2 2
⎝ ⎠j
ℓ= 1 ⎣ ℓ= 2j ⎦
In particular, when α = 0 in (3.9),
2k
1 ℓ (2k − 1) !! ⎤2
∑ s(2k , ℓ)⎛k − ⎞ = ( −1)k ⎡ . (3.10)
ℓ= 0 ⎝ 2⎠ ⎣ 2k ⎦

For 0 ≤ j < k ∈ , we have

2k ℓ
∑ ⎛ ℓ ⎞s(2k , ℓ)⎛k − 1 ⎞ = 0.
⎜ ⎟ (3.11)
ℓ= 2j + 1 ⎝
2j + 1 ⎠ ⎝ 2⎠

Proof. The identities in (3.5) can be rearranged as

k
(iα)k + 1 Γ(iα + k + 1) (2k + 1) ! ⎛ iα + k ⎞
∏(ℓ2 + α2) = (−1)k+1 = ( −1)k + 1i = ( −1)k + 1i
ℓ= 1
α2 (iα + 1)−(k + 1) αΓ(iα − k ) α ⎝ 2k + 1 ⎠
716  Feng Qi

for α ∈  ⧹{0} and k ∈  . By this, we acquire

2k
k ∏ℓ=1(ℓ2 + α2)
∏[(2ℓ − 1)2 + α2] = k
ℓ= 1 ∏ℓ=1[α2 + (2ℓ)2]
2k
∏ℓ=1(ℓ2 + α2)
= k
4k ∏[(α / 2)2 + ℓ2] (3.12)
ℓ= 1
1 Γ(iα + 2k + 1) 1 αΓ(iα / 2 − k )
= ( −1)2k + 1i
4k αΓ(iα − 2k ) ( −1)k + 1i 2Γ(iα / 2 + k + 1)
( −1)k Γ(iα / 2 − k ) Γ(iα + 2k + 1)
=
22k + 1 Γ(iα − 2k ) Γ(iα / 2 + k + 1)

for α ∈  ⧹{0} and k ∈  . Making use of the Gauss multiplication formula

n−1
nnz − 1 / 2 k
Γ(nz ) = ∏ Γ⎛z + ⎞
(2π )(n − 1) / 2 k = 0 ⎝ n⎠

in [11, p. 256, 6.1.20] leads to

Γ(iα − 2k ) 1 2iα − 2k − 1 / 2 ⎛ iα iα 1 2iα − 2k − 1 ⎛ 1 + iα
= 1 / 2
Γ − k ⎞Γ⎛ − k + ⎞ = Γ − k⎞
Γ(iα / 2 − k ) Γ(iα / 2 − k ) (2π ) ⎝2 ⎠ ⎝2 2⎠ π ⎝ 2 ⎠
and
Γ(iα + 2k + 1) (iα + 2k )
= Γ(2(iα / 2 + k ))
Γ(iα / 2 + k + 1) Γ(iα / 2 + k + 1)
(iα + 2k ) 22(iα / 2 + k ) − 1 / 2 1
= Γ(iα / 2 + k )Γ⎛iα / 2 + k + ⎞
Γ(iα / 2 + k + 1) (2π )1 / 2 ⎝ 2⎠
2iα + 2k ⎛ 1 + iα
= Γ + k⎞
π ⎝ 2 ⎠
for α ∈  ⧹{0} and k ∈ . Substituting these two equalities into (3.12), utilizing formula (3.4), interchanging
the order of double sums, and employing identity (3.6) yield
2iα + 2k
k
( −1)k
∏[(2ℓ − 1)2 + α2] = 2k+1 π (
1 + iα
Γ 2 ) +k
2iα − 2k − 1
Γ( − k)
2 1 + iα
ℓ= 1
π 2

= ( −1)k 22k
Γ( 1 + iα
2
+ k)

Γ( − k)
1 + iα
2
iα 1
⎛ 2
+k− 2 ⎞
= ( −1)k 22k(2k ) ! ⎜ ⎟
2k
⎝ ⎠
2k
iα 1 ℓ
= ( −1)k 22k ∑ s(2k , ℓ)⎛ +k− ⎞
ℓ= 0 ⎝2 2⎠
2k ℓ
s(2k , ℓ) ℓ⎞
= ( −1)k 22k ∑ ∑⎛ (iα) j (2k − 1)ℓ− j
⎜ ⎟

ℓ= 0
2ℓ j
j=0⎝ ⎠
2k 2k
⎡ s(2k , ℓ) ⎛ ℓ ⎞ ⎤
= ( −1)k 22k ∑ ⎢∑ ℓ
(2k − 1)ℓ− j ⎥(iα) j
⎜ ⎟

j = 0 ⎣ ℓ= j
2 ⎝ j⎠ ⎦
 Expansions for real powers of two functions and Bell polynomials  717

2k 2k
⎡ s(2k , ℓ) ⎛ ℓ ⎞ ⎤ jπ
= ( −1)k 22k ∑ ⎢∑ ℓ
⎜(2k − 1)ℓ− j ⎥⎛ cos ⎞α j
⎟

j = 0 ⎣ ℓ= j
2 j
⎝ ⎠ ⎝ 2⎠
⎦
2k 2k
⎡ s(2k , ℓ) ⎛ ℓ ⎞ ⎤ jπ
+ i ( −1)k 22k ∑ ⎢∑ ℓ
⎜ (2k − 1)ℓ− j ⎥⎛ sin ⎞α j
⎟

2 ⎝ j⎠ 2⎠
j = 0 ⎣ ℓ= j ⎦⎝
(2j + 1)π
for α ∈  and k ∈  . From the facts that cos( jπ ) = ( −1) j and cos 2
= 0 for j ∈  , equating the above real
part and simplifying produce identity (3.9).
(2j + 1)π
From the facts that sin( jπ ) = 0 and sin 2
= ( −1) j for j ∈  , the last imaginary part becomes
k−1 2k
⎡ s(2k , ℓ) ⎛ ℓ ⎞ ⎤
∑ (−1) j⎢ ∑ ℓ
⎜ (2k − 1)ℓ− 2j − 1⎥α2j + 1 = 0.
⎟

j=0
2 ⎝ 2j + 1 ⎠
⎣ ℓ= 2j + 1 ⎦
Regarding α as a variable means identity (3.11). The proof of Lemma 3.2 is complete. □

Remark 3.1. Formula (3.4) can be reformulated as

n
(z ) n = ∑ (−1)n −k s(n, k )z k . (3.13)
k=0

Comparing this equation with definition (1.1), we can regard (3.1) and (3.9) in Lemmas 3.1 and 3.2 as
generalizations of formula (3.13).

Remark 3.2. The identity (3.10) is a special case of identity (3.4) or (3.13).
In the monograph [14], we do not find the combinatorial identities (3.2), (3.3), and (3.11).

Remark 3.3. The combinatorial identities (3.2) and (3.10) can be rearranged as
Q(2, 2k ) = ( −1)k (k ! )2 , k∈ (3.14)

and

(2k − 1) !! ⎤2
Q(1, 2k ) = ( −1)k ⎡ , k ∈ . (3.15)
⎣ 2k ⎦
The combinatorial identity (3.14) is also derived in [27, Remark 6].
Due to the trivial result s(n , n) = 1 for n ∈ 0, the combinatorial identities (3.3) and (3.11) can be rear-
ranged as
s(2j + 1, 2j ) = −j (2j + 1) , Q(2j + 1, 2m − 1) = 0

for m ≥ 2 and j ∈ 0, and

2m
2j + ℓ ⎞ 1 ℓ
∑⎛ ⎜ s(2j + 2m , 2j + ℓ)⎛ j + m − ⎞ = 0
⎟

ℓ= 1 ⎝
2j + 1 ⎠ ⎝ 2⎠

for j ∈ 0 and m ∈  . On the other hand, we have

Q(2j , 2m) = s(2j + 2m − 1, 2j − 1) + 2j ( j + m − 1)s(2j + 2m − 1, 2j )

2m − 1
1 ⎛ 2j + ℓ ⎞
+ 2j ( j + m − 1) ∑ ⎜ s(2j + 2m − 1, 2j + ℓ)( j + m − 1)ℓ
⎟

ℓ= 1 ℓ + 1 ⎝ 2j ⎠

for j , m ∈ . Can one give a simple form for the quantity

2m − 1
1 ⎛ 2j + ℓ ⎞
∑ ⎜ s(2j + 2m − 1, 2j + ℓ)( j + m − 1)ℓ
⎟

ℓ= 1
ℓ + 1 ⎝ 2j ⎠

for j , m ∈ ? Can one discover more simple forms, similar to Q(1, 2k ) and Q(2, 2k ) in (3.14) and (3.15), of
Q(k , m) for some k ∈  and m ≥ 2 ?
718  Feng Qi

Lemma 3.3. For α ∈  and ∣x∣ < 1, we have

∞ k
x 2k
cosh(α arcsin x ) = ∑ ⎛⎜∏[4(ℓ − 1)2 + α2]⎞⎟ , (3.16)
k = 0 ⎝ ℓ= 1
(2k ) !
⎠
∞ k
⎛ ⎞ x 2k + 1
sinh(α arcsin x ) = α ∑ ⎜∏[(2ℓ − 1)2 + α2]⎟ , (3.17)
k = 0 ⎝ ℓ= 1
(2k + 1) !
⎠
∞ k
( −1)k ⎛ (x − 1)k
cosh(α arccos x ) = ∑ ⎜∏[(ℓ − 1)2 + α2]⎞⎟ , (3.18)
(2k − 1) !! ℓ= 1 k!
k=0 ⎝ ⎠
∞ k
απ ⎛ ⎞ x 2k
cosh(α arccos x ) = ⎛cosh ⎞ ∑ ⎜∏[4(ℓ − 1)2 + α2]⎟
⎝ 2 ⎠ k = 0 ℓ= 1 (2k ) !
⎝ ⎠
(3.19)
∞ k
απ ⎛ ⎞ x 2k + 1
− α⎛ sinh ⎞ ∑ ⎜∏[(2ℓ − 1)2 + α2]⎟ ,
⎝ 2 ⎠ k = 0 ℓ= 1 (2k + 1) !
⎝ ⎠
∞ k
απ ⎛ ⎞ x 2k
sinh(α arccos x ) = ⎛sinh ⎞ ∑ ⎜∏[4(ℓ − 1)2 + α2]⎟
⎝ 2 ⎠ k = 0 ℓ= 1 (2k ) !
⎝ ⎠
(3.20)
∞ k
απ ⎛ ⎞ x 2k + 1
− α⎛ cosh ⎞ ∑ ⎜∏[(2ℓ − 1)2 + α2]⎟ ,
⎝ 2 ⎠ k = 0 ℓ= 1 (2k + 1) !
⎝ ⎠
∞ k
x 2k
cos(α arcsin x ) = ∑ ⎛⎜∏[4(ℓ − 1)2 − α2]⎞⎟ , (3.21)
k = 0 ⎝ ℓ= 1
(2k ) !
⎠
∞ k
⎛ ⎞ x 2k + 1
sin(α arcsin x ) = α ∑ ⎜∏[(2ℓ − 1)2 − α2]⎟ , (3.22)
k = 0 ⎝ ℓ= 1
(2k + 1) !
⎠
∞ k−1
( −1)k ⎛ (x − 1)k
cos(α arccos x ) = ∑ ⎜∏[(ℓ − 1)2 − α2]⎞⎟ (3.23)
(2k − 1) !! ℓ= 1 k!
k=0 ⎝ ⎠
∞ k
απ ⎛ ⎞ x 2k
= ⎛cos ⎞ ∑ ⎜∏[4(ℓ − 1)2 − α2]⎟
⎝ 2 ⎠ k = 0 ℓ= 1 (2k ) !
⎝ ⎠
(3.24)
∞ k
απ ⎛ ⎞ x 2k + 1
+ α⎛ sin ⎞ ∑ ⎜∏[(2ℓ − 1)2 − α2]⎟ ,
⎝ 2 ⎠ k = 0 ℓ= 1 (2k + 1) !
⎝ ⎠
∞ k
απ ⎛ ⎞ x 2k
sin(α arccos x ) = ⎛sin ⎞ ∑ ⎜∏[4(ℓ − 1)2 − α2]⎟
⎝ 2 ⎠ k = 0 ℓ= 1 (2k ) !
⎝ ⎠
(3.25)
∞ k
απ ⎛ ⎞ x 2k + 1
− α⎛ cos ⎞ ∑ ⎜∏[(2ℓ − 1)2 − α2]⎟ ,
⎝ 2 ⎠ k = 0 ℓ= 1 (2k + 1) !
⎝ ⎠
where ( −1) !! = 1 and any empty product is understood to be 1.

Proof. Let fα (x ) = cosh(α arcsin x ). Then consecutive differentiations and simplifications give
α α
f α′ (x ) = sinh(α arcsin x ) = fα2 (x ) − 1 ,
1− x2 1 − x2
(1 − x 2 )[ f α′(x )]2 − α2fα2 (x ) + α2 = 0,
(1 − x 2 )f α″(x ) − xf α′ (x ) − α2fα (x ) = 0,
(1 − x 2 )f α(3) (x ) − 3xf α″(x ) − (1 + α2 )f α′ (x ) = 0.
 Expansions for real powers of two functions and Bell polynomials  719

Accordingly, the differential equation

(1 − x 2 )f α(k + 2) (x ) − (2k + 1)xf α(k + 1) (x ) − (k 2 + α2 )f α(k ) (x ) = 0 (3.26)

is valid for k = 0, 1, respectively. Differentiating on both sides of (3.26) results in

(1 − x 2 )f α(k + 3) (x ) − (2k + 3)xf α(k + 2) (x ) − [(k + 1)2 + α2] f α(k + 1)(x ) = 0.

By induction, the equation (3.26) is valid for all k ≥ 0. Taking x → 0 in (3.26) gives

f α(k + 2) (0) − (k 2 + α2 )f α(k ) (0) = 0, k ≥ 0. (3.27)

It is clear that fα (0) = 1 and f α′ (0) = 0. Substituting these two initial values into the recursive relation (3.27)
and consecutively recursing reveal f α(2k − 1) (0) = 0 and
k
f α(2k ) (0) = ∏[4(ℓ − 1)2 + α2], k ∈ .
ℓ= 1

Consequently, the series expansion (3.16) follows.

Let fα (x ) = sinh(α arcsin x ). Then consecutive differentiations and simplifications give
α α
f α′ (x ) = cosh(α arcsin x ) = fα2 (x ) + 1 ,
1− x2 1− x2
(1 − x 2 )[ f α′(x )]2 − α2fα2 (x ) − α2 = 0,
(1 − x 2 )f α″(x ) − xf α′ (x ) − α2fα (x ) = 0,
(1 − x 2 )f α(3) (x ) − 3xf α″(x ) − (1 + α2 )f α′ (x ) = 0.

By the same argument as above, the derivative f α(k ) (0) for n ≥ 0 satisfy the recursive relation (3.27).
Furthermore, from the facts that fα (0) = 0 and f α′ (0) = α , we conclude f α(2k ) (0) = 0 and
k
f α(2k + 1) (0) = α∏[(2ℓ − 1)2 + α2], k ≥ 0.
ℓ= 1

Consequently, the series expansion (3.17) follows.

Let fα (x ) = cosh(α arccos x ). Then successively differentiating yields
α α
f α′ (x ) = − sinh(α arccos x ) = − fα2 (x ) − 1 ,
1− x2 1 − x2
(1 − x 2 )[ f α′ (x )]2 − α2[ fα2 (x ) − 1] = 0,
(1 − x 2 )f α″(x ) − xf α′ (x ) − α2fα (x ) = 0,
(1 − x 2 )f α‴ (x ) − 3xf α″(x ) − (1 + α2 )f α′ (x ) = 0.

As argued above, we conclude that the derivatives f α(k ) (x ) for k ≥ 0 satisfy equation (3.26). Letting x → 1 in
(3.26) gives

(2k + 1)f α(k + 1) (1) + (k 2 + α2 )f α(k ) (1) = 0. (3.28)

Setting x → 0 in (3.26) leads to (3.27). It is easy to see that

απ απ
fα (1) = 1, f α′ (1) = −α2 , fα (0) = cosh , f α′ (0) = −α sinh .
2 2
Substituting these four initial values into (3.28) and inductively recursing reveal
k
∏ℓ=1[(ℓ − 1)2 + α2]
f α(k ) (1) = ( −1)k ,
(2k − 1) !!
k
απ
f α(2k ) (0) = ⎛cosh ⎞ ∏[4(ℓ − 1)2 + α2],
⎝ 2 ⎠ ℓ= 1
720  Feng Qi

and
k
απ
f α(2k + 1) (0) = −α⎛ sinh ⎞∏[(2ℓ − 1)2 + α2]
⎝ 2 ⎠ ℓ= 1

for k ≥ 0. Consequently, the series expansions (3.18) and (3.19) follow.

Let fα (x ) = sinh(α arccos x ). Then
α α
f α′ (x ) = − cosh(α arccos x ) = − fα2 (x ) + 1 ,
1− x2 1− x2
(1 − x 2 )[ f α′(x )]2 − α2[ fα2 (x ) + 1] = 0,
(1 − x 2 )f α″(x ) − xf α′ (x ) − α2fα (x ) = 0,
απ
and, inductively, the derivatives f α(k ) (x ) for k ≥ 0 satisfy equations (3.26) and (3.27). Since fα (0) = sinh 2
απ
and f α′ (0) = −α cosh 2 , we obtain
k
⎛ ⎞ απ
f α(2k ) (0) = ⎜∏[4(ℓ − 1)2 + α2]⎟ sinh
2
⎝ ℓ= 1 ⎠
and
k
⎛ ⎞ απ
f α(2k + 1) (0) = −α⎜∏[(2ℓ − 1)2 + α2]⎟ cosh
2
⎝ ℓ= 1 ⎠
for k ≥ 0. Consequently, the series expansion (3.20) follows.
Let fα (x ) = cos(α arcsin x ). Then
α α
f α′ (x ) = − sin(α arcsin x ) = − 1 − fα2 (x )
1− x2 1 − x2
(1 − x 2 )[ f α′(x )]2 − α2[1 − fα2 (x )] = 0,
(1 − x 2 )f α″(x ) − xf α′ (x ) + α2fα (x ) = 0,
(1 − x 2 )f α(3) (x ) − 3xf α″(x ) + (α2 − 1)f α′ (x ) = 0,
(1 − x 2 )f α(4) (x ) − 5xf α(3) (x ) + (α2 − 4)f α″(x ) = 0,

and, inductively,

(1 − x 2 )f α(k + 2) (x ) − (2k + 1)xf α(k + 1) (x ) + (α2 − k 2 )f α(k ) (x ) = 0, k ≥ 0. (3.29)

Letting x → 0 in (3.29) results in

f α(k + 2) (0) + (α2 − k 2 )f α(k ) (0) = 0, k ≥ 0. (3.30)

Recusing the relation (3.30) and considering fα (0) = 1 and f α′ (0) = 0 arrive at
k
f (2k ) (0) = ∏[4(ℓ − 1)2 − α2] and f (2k + 1) (0) = 0
ℓ= 1

for k ≥ 0. Consequently, the series expansion (3.21) is valid.

Let fα (x ) = sin(α arcsin x ). Then
α α
f α′ (x ) = cos(α arcsin x ) = 1 − fα2 (x )
1− x2 1 − x2
(1 − x 2 )[ f α′(x )]2 − α2[1 − fα2 (x )] = 0,
(1 − x 2 )f α″(x ) − xf α′ (x ) + α2fα (x ) = 0,
(1 − x 2 )f α(3) (x ) − 3xf α″(x ) + (α2 − 1)f α′ (x ) = 0,
(1 − x 2 )f α(4) (x ) − 5xf α(3) (x ) + (α2 − 4)f α″(x ) = 0,
 Expansions for real powers of two functions and Bell polynomials  721

and, inductively,

(1 − x 2 )f α(k + 2) (x ) − (2k + 1)xf α(k + 1) (x ) + (α2 − k 2 )f α(k ) (x ) = 0, k ≥ 0. (3.31)

Letting x → 0 in (3.31) results in

f α(k + 2) (0) + (α2 − k 2 )f α(k ) (0) = 0, k ≥ 0. (3.32)

Considering fα (0) = 0 and f α′ (0) = α in (3.32) and recusing the relation (3.32) arrive at
k
f (2k ) (0) = 0 and f (2k + 1) (0) = α∏[(2ℓ − 1)2 − α2]
ℓ= 1

for k ≥ 0. Consequently, the series expansion (3.22) is valid.

Let fα (x ) = cos(α arccos x ). Then the derivatives f α(k ) (x ) and f α(k ) (0) satisfy
α α
f α′ (x ) = sin(α arccos x ) = 1 − fα2 (x ) , (1 − x 2 )[ f α′(x )]2 − α2[1 − fα2 (x )] = 0,
1− x2 1− x2
and, inductively, the differential equation (3.29). Setting x → 1 in (3.29) acquires

(2k + 1)f α(k + 1) (1) = (α2 − k 2 )f α(k ) (1) , k ≥ 0. (3.33)

απ απ
Letting x → 0 in (3.29) leads to (3.30). Since fα (1) = 1, fα (0) = cos 2
, and f α′ (0) = α sin 2
, from (3.33) and
(3.30), we obtain
k−1 k
α2 − ℓ2 απ
f α(k ) (1) = ∏ , f α(2k ) (0) = ⎛cos ⎞∏[4(ℓ − 1)2 − α2]
ℓ= 0
2ℓ + 1 ⎝ 2 ⎠ ℓ= 1

and
k
απ
f α(2k + 1) (0) = α⎛ sin ⎞∏[(2ℓ − 1)2 − α2]
⎝ 2 ⎠ ℓ= 1

for k ≥ 0. As a result, we acquire the series expansions (3.23) and (3.24).

Let fα (x ) = sin(α arccos x ). Then
α α
f α′ (x ) = − cos(α arccos x ) = − 1 − fα2 (x ) ,
1 − x2 1 − x2
(1 − x 2 )[ f α′(x )]2 − α2[1 − fα2 (x )] = 0,
(1 − x 2 )f α″(x ) − xf α′ (x ) + α2fα (x ) = 0,
(1 − x 2 )f α(3) (x ) − 3xf α″(x ) + (α2 − 1)f α′ (x ) = 0,

and, inductively, the differential equation (3.29) and the recursive relation (3.30) are valid. Employing the
recursive relation (3.30) and using
απ απ
fα (0) = sin and f α′ (0) = −α cos
2 2
result in
k
απ
f α(2k ) (0) = ⎛sin ⎞∏[4(ℓ − 1)2 − α2]
⎝ 2 ⎠ ℓ= 1

and
k
απ
f α(2k + 1) (0) = −α⎛ cos ⎞∏[(2ℓ − 1)2 − α2]
⎝ 2 ⎠ ℓ= 1

for k ≥ 0. Accordingly, the series expansion (3.25) follows. The proof of Lemma 3.3 is complete. □
722  Feng Qi

Remark 3.4. On [34, pp. 1016–1017], four relations

sin(n arcsin z ) 1+n 1−n 3 2 sin(n arcsin z ) n n 3
= 2 F 1⎛ , ; ; z ⎞, = 2 F 1⎛1 + , 1 − ; ; z 2 ⎞ ,
nz ⎝ 2 2 2 ⎠ nz 1 − z 2 ⎝ 2 2 2 ⎠
n n 1 2 cos(n arcsin z ) 1+n 1−n 1 2
cos(n arcsin z ) = 2 F 1⎛ , − ; ; z ⎞ , = 2 F 1⎛ , ; ;z ⎞
⎝2 2 2 ⎠ 1 − z2 ⎝ 2 2 2 ⎠

for n ∈  were collected, where the Gauss hypergeometric function 2 F 1(α, β ; γ; z ) can be defined [3, Section
5.9] by
∞
(α)k (β )k z k
2 F 1 (α , β; γ; z ) = ∑ , ∣z∣ < 1
k=0
(γ )k k !

for complex numbers α, β ∈  ⧹{0} and γ ∈  ⧹{0, − 1, −2, …} , where (α)k , (β )k , and (γ )k are defined by (1.1)
or (1.2).
In particular, Maclaurin’s series expansion

∞ 2k − 1
⎛ 2 ⎞ t 2(k + 1)
cos(arcsin t ) = 1 + ∑ ( −1)k 22k ⎜ 2k +
k=0
1⎟ k + 1
⎝ ⎠
was derived in [27, Remark 18], where extended binomial coefficient ( ) is defined by (1.3).
z
w

Remark 3.5. In the article [35], among other things, three authors established series expansions at x = 0 or
x = 1 of the functions
exp(α arccos x ) arccos x sin(α arccos x )
exp(α arccos x ) , , , , exp(αarccoshx ) ,
1− x2 1− x2 1 − x2
sin(αarccoshx ) sinh(αarccoshx )
, , cos(α arccos x ), cosh(α arccos x ), cos(αarccoshx ),
x2 − 1 x2 − 1
cos(α arccos x ) cosh(α arccos x )
cosh(αarccoshx ) , sin(α arccos x ) , sinh(α arccos x ) , , ,
1− x2 1 − x2
cos(αarccoshx ) sinh(α arccos x ) cosh(αarccoshx )
, ,
1− x2 1− x2 1 − x2

for α ∈  ⧹{0} in terms of the Gauss hypergeometric function 2 F 1(α, β ; γ; z ).

By the way, we point out that the series expansions (2.1), (4.1), (4.3), (4.4), (4.7), and (4.9) in the article
[35] should be wrong.

Lemma 3.4. Let n ≥ k ≥ 0 and α, β ∈  .

1. The Faà di Bruno formula can be described in terms of partial Bell polynomials Bn, k by
n
dn
f ∘ h (t ) = ∑ f (k ) (h(t ))Bn,k (h′(t ), h″(t ), … , h(n −k+1)(t )), n ∈ 0 . (3.34)
dt n k=0

2. Partial Bell polynomials Bn, k satisfy the identities

Bn, k (αβx1, αβ 2 x2 , … , αβ n − k + 1xn − k + 1) = αk β n Bn, k (x1, x2 , … , xn − k + 1) , (3.35)

(n − k ) ! n ⎛ k ⎞ 2k − n
Bn, k (α, 1, 0, … ,0) = ⎛ ⎞ α , (3.36)
2n − k ⎝ k ⎠⎝ n − k ⎠

2n − k − 1 ⎞
Bn, k (( −1) !!, 1!!, 3!! , … ,[2(n − k ) − 1] !! ) = [2(n − k ) − 1] !! ⎛ , ⎜ ⎟ (3.37)
⎝ 2(n − k ) ⎠
 Expansions for real powers of two functions and Bell polynomials  723

and
∞ k ∞
1 ⎛ tm ⎞ tn
⎜ ∑ xm ⎟ = ∑ Bn,k (x1, x2, … , xn −k+1) . (3.38)
k ! ⎝m=1 m ! ⎠ n=k
n!

These formulas in Lemma 3.4 can be found in [8, p. 412], [9, pp. 134–135 and 139], [36, Theorem 4.1],
[37, p. 169, (3.6)], and [38, Theorem 1.2], respectively. These identities in Lemma 3.4 can also be found in the
survey and review article [12].

(arccosx )2 k (arccoshx )2 k
4 Taylor’s series expansions of ⎡ ⎤
2(1 − x ) ⎦
and ⎡ ⎤
2(1 − x ) ⎦
⎣ ⎣
In this section, by virtue of some conclusions in Lemmas 3.1, 3.2, and 3.3, we establish Taylor’s series
(arccos x )2 k (arccoshx )2 k
expansions around x = 1 of the functions ⎡ ⎤
2(1 − x ) ⎦
and ⎡ ⎤
2(1 − x ) ⎦
in terms of Q(k , m) defined by
⎣ ⎣
formula (2.3).

Theorem 4.1. For k ∈  and ∣x∣ < 1, we have

2 k ∞
⎡ (arccos x ) ⎤ = 1 + (2k ) ! ∑ Q(2k , 2n) [2(x − 1)]n (4.1)
⎢
⎣ 2(1 − x ) ⎥ ⎦ n=1
(2k + 2n) !

and
2 k ∞
⎡ (arccoshx ) ⎤ = 1 + (2k ) ! ∑ Q(2k , 2n) [2(x − 1)]n , (4.2)
⎢
⎣ 2(x − 1) ⎥ ⎦ n=1
(2k + 2n) !

where Q(2k , 2n) is defined by (2.3).

Proof. Replacing α by iα in (3.1) leads to

k k 2k + 1
∏(ℓ2 − α2) = (−1)k ∑ ⎢ ∑
⎡ ⎛ ℓ ⎞s(2k + 1, ℓ)k ℓ− 2j − 1⎤α2j . (4.3)
⎜

2j + 1 ⎠
⎟
⎥
ℓ= 1 j = 0 ⎣ ℓ= 2j + 1 ⎝ ⎦
From Taylor’s series expansion (3.23) and identity (4.3), it follows that
∞ ∞ k−1
(α arccos x )2k ( −1)k ⎡ (x − 1)k
∑ (−1)k = 1 + (x − 1)α2 − α2 ∑ ⎢∏(ℓ2 − α2)⎤⎥
(2k ) ! (2k − 1) !! ℓ= 1 k!
k=0 k=2 ⎣ ⎦
∞ k−1 2k − 1
1 ⎡ ℓ ⎞ ⎤ (x − 1)k
= 1 + (x − 1)α2 + α2 ∑ ⎢ ∑ α2j ∑ ⎛ s(2k − 1, ℓ)(k − 1)ℓ− 2j − 1⎥
⎜ ⎟

k=2
(2k − 1) !! j = 0 ℓ= 2j + 1 ⎝ 2j + 1 ⎠ k!
⎣ ⎦
∞ 2k − 1 k
1 ⎡ ℓ− 1⎤ (x − 1)
= 1 + (x − 1)α2 + α2 ∑ ∑ ℓ s ( 2k − 1, ℓ )( k − 1)
k=2
(2k − 1) !! ⎢
⎣ ℓ= 1
⎥ k!
⎦
∞ k 2k − 1
1 (x − 1)k ⎛ ℓ ⎞s(2k − 1, ℓ)(k − 1)ℓ− 2j + 1
+ α2 ∑ ∑ α2j−2 ∑ ⎜ ⎟

k=2
(2k − 1) !! k ! j=2 ℓ= 2j − 1 ⎝
2j − 1 ⎠
∞ 2k − 1
1 ⎡ (x − 1)k
= 1 + (x − 1)α2 + α2 ∑ ⎢ ∑ ℓs(2k − 1, ℓ)(k − 1)ℓ−1⎤⎥
k=2
(2k − 1) !! ⎣ ℓ= 1 ⎦ k!
∞ ∞ 2m − 1
1 (x − 1)m ℓ
+ ∑ ⎡⎢ ∑ ∑ ⎛ ⎞s(2m − 1, ℓ)(m − 1)ℓ− 2k + 1⎤α2k .
⎥
k=2 ⎣m=k
(2m − 1) !! m ! ℓ= 2k − 1 ⎝
2k − 1 ⎠ ⎦
724  Feng Qi

This means that

∞ 2k − 1
(arccos x )2 1 ⎡ (x − 1)k
− =x−1+ ∑ ⎢ ∑ ℓs(2k − 1, ℓ)(k − 1)ℓ−1⎤⎥
2! k=2
(2k − 1) !! ⎣ ℓ= 1 ⎦ k!
∞ 2k
[2(x − 1)]k + 1
=x−1+ ∑ ⎡⎢ ∑ (ℓ + 1)s(2k + 1, ℓ + 1)k ℓ⎤⎥ ,
k = 1 ⎣ ℓ= 0 ⎦ (2k + 2) !
where we used identity (3.2) in Lemma 3.1 or identity (3.14), and that
∞ 2m − 1
(arccos x )2k ℓ (x − 1)m
( −1)k = ∑⎡ ⎢ ∑ ⎛ ⎞s(2m − 1, ℓ)(m − 1)ℓ− 2k + 1⎤
⎥
(2k ) ! m = k ⎣ ℓ= 2k − 1 ⎝
2k − 1 ⎠ ⎦ (2m − 1) !!m!
∞
[2(x − 1)]m + k
= ∑ Q(2k , 2m)
m=0
(2k + 2m) !

for k ≥ 2 . Consequently, the series expansions

∞
(arccos x )2 m! (1 − x ) m + 1
= ∑ , ∣x∣ < 1 (4.4)
2! m=0
(2m + 1) !! m + 1

and
∞
(arccos x )2k [2(1 − x )]m + k
= ∑ (−1)mQ(2k , 2m) (4.5)
(2k ) ! m=0
(2k + 2m) !

for k ≥ 2 and ∣x∣ < 1 are valid.

By similar arguments as done above, from the series expansion (3.18), we can also recover the series
expansions (4.4) and (4.5).
The series expansions (4.4) and (4.5) can be reformulated as
∞
(arccos x )2 (m ! ) 2 (1 − x ) m
=1+ ∑
2(1 − x ) m=1
(2m + 1) !! (m + 1) m !

and
2 k ∞
( −1)m (1 − x ) m
⎡ (arccos x ) ⎤ = 1 + ∑ Q(2k , 2m)
⎢
⎣ 2(1 − x ) ⎥ ⎦ m=1 (2m − 1) !! ( 2m + 2k
2k ) m!

for k ≥ 2 . Making use of identity (3.2) or (3.14), we obtain

( −1)m (m ! ) 2
Q(2, 2m) = , m ∈ .
(2m − 1) !! ( 2m + 2
2 ) (2m + 1) !! (m + 1)

Consequently, the series expansions (4.4) and (4.5) can be unified as the series expansion (4.1).
By the relation
arccos x = −i arccoshx , (4.6)

from (4.1), we deduce (4.2). The proof of Theorem 4.1 is complete. □

Corollary 4.1. For k ∈  and ∣x∣ < 1, we have

2 k ∞
⎡ (π − arccos x ) ⎤ = 1 + (2k ) ! ∑ ( −1)m Q(2k , 2m) [2(x + 1)]m (4.7)
⎢
⎣ 2(1 + x ) ⎥
⎦ m=1
(2k + 2m) !
 Expansions for real powers of two functions and Bell polynomials  725

and
k ∞
(π + i arccoshx )2 ⎤ Q(2k , 2m)
( −1)k ⎡ = 1 + (2k ) ! ∑ ( −1)m [2(x + 1)]m , (4.8)
⎢
⎣ 2(1 + x ) ⎥
⎦ m=1
( 2k + 2m) !

where Q(2k , 2m) is defined by (2.3).

Proof. This follows from replacing x by −x in (4.1) and (4.2) and utilizing the relations arccos x +
arccos( −x ) = π and (4.6). □

Corollary 4.2. For k ∈ , we have

2 k ∞
⎛ π ⎞ = 1 + (2k ) ! ∑ ( −1)m2m Q(2k , 2m) .
⎜ ⎟ (4.9)
⎝8⎠ m=1
(2k + 2m) !

In particular, we have
∞
π2 2m 1
= ∑ . (4.10)
8 m=1
m2
( ) 2m
m

Proof. The series representation (4.9) of π 2k follows from letting x = 0 in either (4.1), (4.2), (4.7), or (4.8).
The series representation (4.10) follows from taking k = 1 in (4.9), or setting k = 2 in (2.5), and then
making use of identity (3.2) or (3.14). □

Remark 4.1. The series representation (4.10) recovers a conclusion in [39, Theorem 5.1].
As for series representations of π 2 , in [40, p. 453, (14)] and the article [41], among other things, we find
∞ ∞ ∞ ∞
π2 1 π2 1 π2 ( −1)m π2 1 1
= ∑ , = ∑ , = ∑ , = ∑ , (4.11)
6 m=0
( m + 1)2 8 m=0
( 2m + 1)2 12 m=0
(m + 1)2 18 m=1
m 2
( )
2m
m

1
which are different from (4.10). The last series representation in (4.11) is derived from letting k = 2 and x = 2
in (2.2).
Because
1/m
⎡ 2m 1/m 1/m
8 ⎤ 1 6 8
lim ⎢ 2 ⎥ = , lim ⎡ ⎤ = 1, lim ⎡ ⎤ = 1,
m →∞ ⎢ m
⎣ ( )
2m
m ⎦
⎥ 2 m →∞ ⎢
⎣ ( m + 1)2 ⎥
⎦ m →∞ ⎢
⎣ ( 2m + 1)2 ⎥
⎦
(4.12)
1/m
1/m ⎡ 18
( −1)m12 1 ⎤ 1
lim ⎡ ⎤ = 1, lim ⎢ 2 ⎥ = ,
m →∞ ⎢
⎣ (m + 1)2 ⎥
⎦ m →∞ ⎢ m
⎣ ( )
2m
m ⎦
⎥ 4

we regard that the last series representation in (4.11) in [40, p. 453, (14)] converges to π 2 quicker than (4.10)
and other three in (4.11). The first unsolved problem posed on December 13, 2010 by Herbert S. Wilf
(1931–2012) is about the convergent speed of rational approximations of the circular constant π . For
more details on this unsolved problem, see [39, Remark 7.7].
For k ∈  , let
1/ m 1/ m
Q(2k , 2m) ⎤ Q(2k , 2m) ⎤
L(k ) = lim ⎡(2k ) ! ( −1)m2m = 2 lim ⎡( −1)m .
⎢
m →∞⎣ (2k + 2m) ! ⎥
⎦ m →∞⎣⎢ (2k + 2m) ! ⎥
⎦
1
The first limit in (4.12) means that L(1) = 2 . What is the convergent speed of the hypergeometric term in (4.9)
for k ≥ 2 ? Equivalently speaking, what is the limit L(k ) for k ≥ 2 ? Is the limit L(k ) a decreasing sequence in
k ≥ 2 ? What is the limit limk →∞[L(k )]1 / k ?
726  Feng Qi

Corollary 4.3. For k , m ∈  , we have

k (m)
⎛⎡ (arccos x )2 ⎤ ⎞ (2k ) ! (2m) !!
⎜⎢ 2(1 − x ) ⎥ ⎟ = Q(2k , 2m) ,
(2k + 2m) !
⎝⎣ ⎦ ⎠
x=1
k (m)
⎛⎡ (arccoshx )2
⎤ ⎞ (2k ) ! (2m) !!
⎜⎢ 2(1 − x ) ⎥ ⎟ = ( −1)k Q(2k , 2m) ,
(2k + 2m) !
⎝⎣ ⎦ ⎠
x=1
k (m)
⎛⎡ (π − arccos x )2 ⎤ ⎞ (2k ) ! (2m) !!
⎜⎢ ⎟ = ( −1)m Q(2k , 2m) ,
2(1 + x ) ⎥ (2k + 2m) !
⎝⎣ ⎦ ⎠
x = (−1)+

and

k (m)
⎛⎡ (π + i arccoshx )2 ⎤ ⎞ (2k ) ! (2m) !!
⎜⎢ ⎟ = ( −1)k + m Q(2k , 2m) ,
2(1 + x ) ⎥ (2k + 2m) !
⎝⎣ ⎦ ⎠
x = (−1)+

where Q(2k , 2m) is defined by (2.3).

Proof. These derivatives follow from series expansions (4.1), (4.2), (4.7), and (4.8). □

Corollary 4.4. For k , n ∈ , we have

⎧ 0, n < k;
⎪ ( −1)k (2k ) !! , n = k;
[(arccos x )2k ](n) ∣ x = 1 = (4.13)
⎨ k (2k ) !
⎪ ( −1) (2n − 1) !! Q(2k , 2n − 2k ) , n>k
⎩
and

⎧ 0, n < k;
⎪ (2k ) !! , n = k;
[(arccoshx )2k ](n) ∣ x = 1 = (4.14)
⎨ (2k ) !
⎪ (2n − 1) !! Q(2k , 2n − 2k ) , n > k.
⎩

Proof. For k ∈  and ∣x∣ < 1, the series expansions (4.1) and (4.2) can be reformulated as
∞
(x − 1)k Q(2k , 2m) (x − 1)k + m
(arccos x )2k = ( −1)k (2k ) !! + ( −1)k (2k ) ! ∑ (4.15)
k! m=1
(2k + 2m − 1) !! (k + m) !

and
∞
(x − 1)k Q(2k , 2m) (x − 1)k + m
(arccoshx )2k = (2k ) !! + (2k ) ! ∑ . (4.16)
k! m=1
(2k + 2m − 1) !! (k + m) !

These forms of series expansions (4.15) and (4.16) imply the formulas in (4.13) and (4.14). □

Corollary 4.5. For k ∈  and ∣x∣ < 1, we have

k ∞
k⎞ ( −1)m Q(2k , 2m) ⎛ k + m ⎞⎤ j
(arccos x )2k = ∑ (−1) j⎡⎢2k⎛ ⎜ + (2k ) ! ∑
⎟ x ⎜ ⎟

j=0 ⎣ ⎝ j⎠ m=1
(k + m) ! (2k + 2m − 1) !! ⎝ j ⎠⎥⎦
(4.17)
∞ ∞
( −1) j + m Q(2k , 2j + 2m − 2k ) ⎛ j + m ⎞⎤ j
+ ( −1)k (2k ) ! ∑ ( −1) j ⎡
⎢ ∑ ( j + m) ! ⎥x⎜ ⎟

j=k+1 ⎣m=0 (2j − 1) !! ⎝ j ⎠⎦

 Expansions for real powers of two functions and Bell polynomials  727

and
k ∞
k⎛ k ⎞ ( −1)m Q(2k , 2m) ⎛ k + m ⎞⎤ j
(arccoshx )2k = ( −1)k ∑ ( −1) j ⎡
⎢2 j + (2k ) ! ∑ (k + m) ! (2k + 2m − 1) !!
⎜ ⎟
⎥x ⎜ ⎟

j=0 ⎣ ⎝ ⎠ m=1 ⎝ j ⎠⎦
(4.18)
∞ ∞
( −1) j + m Q(2k , 2j + 2m − 2k ) ⎛ j + m ⎞⎤ j
+ (2k ) ! ∑ ( −1) j ⎡
⎢ ∑ ( j + m) ! ⎥x , ⎜ ⎟

j=k+1 ⎣m=0 (2j − 1) !! ⎝ j ⎠⎦

where Q(2k , 2m) is given by (2.3).

Proof. For k ∈  and ∣x∣ < 1, by the binomial theorem, the series expansion (4.15) can be rewritten as
k ∞ k+m
k⎞ j ( −1)m − j Q(2k , 2m) ⎛ k + m ⎞ j
(arccos x )2k = 2k ∑ (−1) j⎛ ⎜ x + (2k ) ! ∑ ∑
⎟ x ⎜ ⎟

j=0
j
⎝ ⎠ m=1 j=0
( k + m ) ! (2k + 2m − 1) !! ⎝ j ⎠
k k ∞ ∞ ∞
k⎞ j ⎛ ⎞ ( −1)m − j Q(2k , 2m) ⎛ k + m ⎞ j
= 2k ∑ (−1) j⎛ ⎜ x + (2k ) ! ⎜ ∑ ∑ +
⎟ ∑ ∑ ⎟ (k + m) ! (2k + 2m − 1) !! x ⎜ ⎟

j=0 ⎝ j⎠ ⎝ j=0 m=1 j=k+1 m=j−k ⎠ ⎝ j ⎠

k ∞
k⎞ ( −1)m Q(2k , 2m) ⎛ k + m ⎞⎤ j
= ∑ (−1) j⎡⎢2k⎛ + (2k ) ! ∑
⎜ ⎟ x ⎜ ⎟

j=0
j
⎣ ⎝ ⎠ m=1
(k + m ) ! (2k + 2m − 1) !! ⎝ j ⎠⎥⎦
∞ ∞
⎡ ( −1)m Q(2k , 2m) ⎛ k + m ⎞⎤ j
+ (2k ) ! ∑ ( −1) j ⎢ ∑ x.
(k + m) ! (2k + 2m − 1) !! ⎝ j ⎠⎥
⎜ ⎟

j=k+1 ⎣m=j−k ⎦
The series expansion (4.17) is thus proved.
Similarly, from (4.16), we conclude (4.18). Theorem 4.5 is thus proved. □

Remark 4.2. What are closed-form expressions of coefficients in Maclaurin’s series expansion around the
point x = 0 of the real power (arccos x )α for α ∈  and ∣x∣ < 1? For answers to this question, please refer to
[39, Section 3].

Corollary 4.6. For m , n ∈ , we have

[(arccos x )2n − 1](m) ∣ x = 1 = { 0,

∞,
m < n;
m≥n
(4.19)

and

[(arccoshx )2n − 1](m) ∣ x = 1 = { 0,

∞,
m < n;
m ≥ n.
(4.20)

Consequently, the functions (arccos x )2n − 1 and (arccoshx )2n − 1 for n ∈  cannot be expanded into Taylor’s series
expansions at the point x = 1.

Proof. It is easy to verify that

(arccos x )2 ~ 2(1 − x ) , x → 1. (4.21)

By the Faà di Bruno formula (3.34) and in the light of identities (3.35) and (3.36), we obtain
(m) m
⎛⎜ 1 ⎞⎟ 1
= ∑ − (1 − x 2 )−1 / 2 − j Bm, j ( −2x , −2, 0, … ,0)
⎝ 1− x2 ⎠ j=0
2 j
m
(2j − 1) !! ( −2) j
= ∑ (−1) j j
Bm, j (x , 1, 0, … ,0)
j=0
2 (1 − x 2 ) j + 1 / 2
728  Feng Qi

m
(2j − 1) !! (m − j ) ! ⎛ m ⎞⎛ j ⎞ 2j − m
= ∑ ⎜ ⎟⎜x ⎟

j=0
( 1 − x 2 ) j + 1 / 2 2m − j ⎝ j ⎠⎝ m − j ⎠
m
(2j − 1) !! (m − j ) ! ⎛ m ⎞⎛ j ⎞ x 2j − m
= ∑ ⎜ ⎟⎜ ⎟

j=0
(1 + x ) j + 1 / 2 2m − j ⎝ j ⎠⎝ m − j ⎠ (1 − x ) j + 1 / 2

for m ∈ 0. This implies that

(m)
⎛⎜ 1 ⎞⎟ (2m − 1) !!
~ , x → 1, m ∈ 0 . (4.22)
2
⎝ 1−x ⎠ [2(1 − x )]m + 1 / 2

Utilizing the Faà di Bruno formula (3.34), employing identities (3.35) and (3.37), and making use of
(4.21) and (4.22), we acquire
m (m − j )
⎛ 1 1 1 ⎞
[(arccos x )2n − 1](m) = ∑ ⟨2n − 1⟩j (arccos x )2n −j−1Bm,j⎜− , ⎜⎛− ⎞ ⎛
⎟′ , … , ⎜−
⎞
⎟
2
1−x ⎝ 2
1−x ⎠ 1 − x 2 ⎟
j=0 ⎝ ⎝ ⎠ ⎠
m
( −1) !! 1 !! 3 !!
~ ∑ ⟨2n − 1⟩j [2(1 − x )]n −( j+1) / 2 (−1) jBm,j⎛ ⎜ , ,
1 ∕ 2 [2(1 − x )]3 ∕ 2 [2(1 − x )]5 ∕ 2
,…
j=0 ⎝ [2( 1 − x )]
[2(m − j ) − 1] !! ⎞
, ⎟

[2(1 − x )]m − j + 1 ∕ 2 ⎠
m
[2(1 − x )] j ∕ 2
= ∑ (−1) j⟨2n − 1⟩j [2(1 − x )]n −( j+ 1) ∕ 2 Bm, j (( −1) !!, 1!!, 3!! , … ,[2(m − j ) − 1] !! )
j=0
[2(1 − x )]m
m
2m − j − 1 ⎞
= [2(1 − x )]n − m − 1 ∕ 2 ∑ ( −1) j ⟨2n − 1⟩j [2(m − j ) − 1] !! ⎛ ⎜ ⎟

j=0 ⎝ 2(m − j ) ⎠

→ { 0,
∞,
n>m
n≤m

as x → 1 for m , n ∈  . The results in (4.19) are thus proved.

Substituting (4.6) into (4.19) leads to (4.20). Theorem 4.6 is therefore proved. □

Corollary 4.7. For k ∈ 0, we have

k
2k − j − 1 ⎞
∑ (−1) j⟨2k⟩j [2(k − j ) − 1] !! ⎛ ⎜ = ( −1)k (2k ) !! .
⎟ (4.23)
j=0 ⎝ 2( k − j ) ⎠

Proof. As done in the proof of Theorem 4.6, we arrive at

n (n − j )
⎛ 1 1 1 ⎞
[(arccos x )2k ](n) = ∑ ⟨2k⟩j (arccos x )2k−jBn,j⎜− , ⎛⎜− ⎞⎟′ , … , ⎛⎜− ⎞⎟
2
1−x ⎝ 2
1−x ⎠ 1 − x 2 ⎟
j=0 ⎝ ⎝ ⎠ ⎠
n
( −1) !! 1 !! [2(n − j ) − 1] !! ⎞
~ ∑ ⟨2k ⟩j [2(1 − x )]k − j ∕ 2 ( −1) j Bn, j⎛ ⎜ ,
1 ∕ 2 [2(1 − x )]3 ∕ 2
,…, ⎟

j=0 ⎝ [2( 1 − x )] [2(1 − x )]n − j + 1 ∕ 2 ⎠

n
[2(1 − x )] j ∕ 2
= ∑ (−1) j⟨2k⟩j [2(1 − x )]k−j ∕ 2 Bn, j (( −1) !!, 1!!, 3!! , … ,[2(n − j ) − 1] !! )
j=0
[2(1 − x )]n
n
2n − j − 1 ⎞
= [2(1 − x )]k − n ∑ ( −1) j ⟨2k ⟩j [2(n − j ) − 1] !! ⎛ ⎜ ⎟

j=0 ⎝ 2(n − j ) ⎠
⎧ 0, k>n
⎪ k
→ 2k − j − 1 ⎞
⎨ ∑ ( −1) j ⟨2k ⟩j [2(k − j ) − 1] !! ⎛ ⎜ , ⎟ k=n
⎪ j=0
⎩ ⎝ 2(k − j ) ⎠

as x → 1 for k , n ∈ . Comparing this result with (4.13) gives (4.23). □

 Expansions for real powers of two functions and Bell polynomials  729

Remark 4.3. Identity (4.23) is similar to

n
2n − k − 1 ⎞
∑ k ! [2(n − k ) − 1] !! ⎛ ⎜ = (2n − 1) !! ,
⎟ n ∈ 0 ,
k=0 ⎝ 2(n − k ) ⎠
which was established in [42, Theorem 4.2] and [43, p. 10, (3.12)], respectively.

(arccos x )2 α
5 Taylor’s series expansions at x = 1 of ⎡ ⎤
2(1 − x ) ⎦
⎣
In this section, via establishing a closed-form expression for the specific partial Bell polynomials at a
(arccos x )2 α
sequence of the derivatives at x = 1 of the function ⎡ ⎤,
2(1 − x ) ⎦
we present Taylor’s series expansion at
⎣
(arccos x )2 α
x = 1 of the function ⎡ ⎤
2(1 − x ) ⎦
for α ∈  .
⎣

Theorem 5.1. For m ≥ k ∈ , we have

(m − k + 1)
⎛ (arccos x )2⎤′ (arccos x )2⎤″ (arccos x )2⎤ ⎞
Bm, k ⎜⎡ ,⎡ ,…,⎡ ⎟
⎢ 2(1 − x ) ⎥⎦ x=1 ⎢ ⎣ 2(1 − x ) ⎥ ⎢
⎣ 2(1 − x ) ⎥
⎝⎣ ⎦ x=1 ⎦ x = 1⎠

1 2 3 32 80 (2m − 2k + 2) !! (5.1)
= 2kBm, k ⎛− , ⎜ ,− , ,− ,…, Q(2, 2m − 2k + 2)⎞ ⎟

⎝ 12 45 70 525 693 (2m − 2k + 4) ! ⎠

k
m k Q(2j , 2m)
= ( −2)k [2(m − k )] !! ⎛ ⎞ ∑ ( −1) j (2j ) ! ⎛ ⎞ ⎜ ⎟ ,
⎝ k ⎠ j=1 ⎝ j ⎠ (2j + 2m) !

where Q(2j , 2m) is defined by (2.3).

Proof. Let
(m)
(arccos x )2⎤
xm = ⎡ , m ∈ .
⎢
⎣ 2(1 − x ) ⎥⎦ x=1

Then, from (3.38) and (4.1), it follows that

∞ k
n + k⎞ dn x
Bn + k , k (x1, x2 , … , xn + 1) = ⎛ lim n ⎡ ∑ m +1 t m⎤⎥
⎝ k ⎠ t → 0 dt ⎢
⎣m=0 ( m + 1) ! ⎦
k
n + k⎞ dn ⎛ 1
∞
(arccos x ) ⎤ 2 (m) tm ⎞
=⎛ lim n ⎜ ∑ ⎡⎢
⎝ k ⎠ t → 0 dt t m=1⎣
2(1 − x ) ⎥⎦ m! ⎟
⎝ x=1 ⎠
k
∞ (m )
n + k⎞ dn ⎛ 1 (arccos x )2 ⎤ (x − 1)m ⎞
=⎛ lim n ⎜ ∑ ⎡⎢
⎝ k ⎠ x → 1 dx x − 1 m=1⎣
2(1 − x ) ⎥ ⎦ m! ⎟
⎝ x=1 ⎠
k
n + k⎞ dn⎛ 1 ⎡ (arccos x )2
=⎛ lim ⎜ − 1⎤⎟⎞
⎝ k ⎠ x → 1 dx n ⎝ x − 1 ⎢
⎣ 2(1 − x ) ⎥
⎦⎠
k j
n + k⎞ dn ⎛ 1 k ⎞⎡ (arccos x )2 ⎤ ⎞
=⎛ lim n ⎜ ∑ (−1)k−j⎛ ⎜ ⎟
⎟
⎝ k ⎠ x → 1 dx (x − 1)k j=0 ⎝ j ⎠⎢
⎣ 2(1 − x ) ⎥ ⎦⎠
⎝
730  Feng Qi

k ∞
n + k⎞ dn ⎡ ( −1)k 1 k ⎞⎛ Q(2j , 2m) ⎤
=⎛ lim n ⎢ + ∑ (−1)k−j⎛ ⎜ ⎜1 + (2j ) ! ∑
⎟ [2(x − 1)]m ⎞⎟⎥
⎝ k ⎠ x → 1 dx (x − 1)k (x − 1)k j=1
j
⎝ ⎠⎝ m=1
( 2j + 2m ) ! ⎠⎦
⎣
∞ k
n + k⎞ dn ⎛ ⎡ k Q(2j , 2m) ⎤ ⎞
= ( −1)k ⎛ lim n ⎜ ∑ 2m⎢ ∑ ( −1) j (2j ) ! ⎛ ⎞ m−k
⎥(x − 1) ⎟
⎜ ⎟

⎝ k ⎠ x → 1 dx m = 1 j
⎝ ⎠ ( 2j + 2m ) !
⎝ ⎣ j=1 ⎦ ⎠
for k ∈  . This implies that
k
k ⎞ Q(2j , 2m)
∑ (−1) j(2j ) ! ⎛ ⎜ ⎟ = 0, 1 ≤ m < k. (5.2)
j=1 ⎝ j ⎠ (2j + 2m) !

Accordingly, we derive
∞ k
n + k⎞ dn ⎡ k ⎞ Q(2j , 2m) ⎤
Bn + k , k (x1, x2 , … , xn + 1) = ( −1)k ⎛ lim n ∑ 2m⎢ ∑ (−1) j(2j ) ! ⎛ ⎜ ⎟
⎥(x − 1)
m−k
⎝ k ⎠ x → 1 dx m=k
j
⎝ ⎠ ( 2j + 2m ) !
⎣ j=1 ⎦
∞ k
n + k⎞ dn ⎛ ⎡ k Q(2j , 2m + 2k ) ⎤ ⎞
= ( −2)k ⎛ lim n ⎜ ∑ 2m⎢ ∑ ( −1) j (2j ) ! ⎛ ⎞ ⎥(x − 1)m⎟ ⎜ ⎟

⎝ k ⎠ x → 1 dx m = 0 ⎝ j ⎠ (2j + 2m + 2k ) ! ⎦
⎝ ⎣ j=1 ⎠
∞ k
n + k⎞ ⎡ k ⎞ Q(2j , 2m + 2k ) ⎤
= ( −2)k ⎛ lim ∑ 2m⎢ ∑ (−1) j(2j ) ! ⎛ ⎜ ⎟
⎥⟨m⟩n (x − 1)
m−n
⎝ k ⎠ x→1 m=n
j
⎝ ⎠ ( 2j + 2m + 2k ) !
⎣ j=1 ⎦
k
n + k⎞ k Q(2j , 2n + 2k )
= ( −2)k (2n) !! ⎛ ∑ (−1) j(2j ) ! ⎛ ⎞ ⎜ ⎟

⎝ k ⎠ j=1 ⎝ j ⎠ (2j + 2n + 2k ) !

for n ≥ k ∈  . Replacing n + k by m results in

k
m k Q(2j , 2m)
Bm, k (x1, x2 , … , xm − k + 1) = ( −2)k [2(m − k )] !! ⎛ ⎞ ∑ ( −1) j (2j ) ! ⎛ ⎞ ⎜ ⎟

⎝ k ⎠ j=1 ⎝ j ⎠ (2j + 2m) !

for m ≥ k ∈ . The required result is thus proved. □

Theorem 5.2. For α ∈  , we have

2 α ∞ n
⟨α⟩j j
j ⎞ Q(2ℓ , 2n) ⎤
⎡ (arccos x ) ⎤ = 1 + ⎡
∑ ⎢ ∑ (−1) j ∑ (−1)ℓ(2ℓ) ! ⎛ n (5.3)
⎢ ⎥[2(x − 1)] .
⎣ 2(1 − x ) ⎥ ⎦ n=1⎣ j=1
j! ℓ= 1 ⎝ ℓ ⎠ (2ℓ + 2n) ! ⎦

Proof. By virtue of the Faà di Bruno formula (3.34) and formula (5.1) in Theorem 5.1, we obtain

2 α (n) n α−j (n − j + 1)
⎛⎜⎡ (arccos x ) ⎤ ⎟⎞ (arccos x )2 ⎤ ⎛ (arccos x )2 ⎤′ (arccos x )2 ⎤ ⎞
⎢ ⎥
= ∑ ⟨α⟩j ⎡⎢ Bn, j⎜⎡ , …,⎡ ⎟
⎝⎣ 2(1 − x ) ⎦ ⎠ j=1 ⎣ 2(1 − x ) ⎥ ⎦ ⎢
⎝⎣
2( 1 − x ) ⎥
⎦ ⎢
⎣ 2( 1 − x ) ⎥
⎦ ⎠
for n ∈  . Taking the limit x → 1 and employing (5.1) in Theorem 5.1 lead to
α (n) n (n − j + 1)
(arccos x )2 ⎤ ⎞ ⎛ (arccos x )2 ⎤′ (arccos x )2 ⎤ ⎞
lim ⎛⎜⎡ ⎟ = ∑ ⟨α⟩j Bn, j⎜lim ⎡ , … ,lim ⎡ ⎟
x→1 ⎢ ⎥ x→1 ⎢ ⎥ ⎢ ⎥
⎝⎣ 2(1 − x ) ⎦ ⎠ j=1 ⎝ ⎣ 2( 1 − x ) ⎦ x → 1 ⎣ 2( 1 − x ) ⎦ ⎠
n j
n j Q(2ℓ , 2n)
= ∑ ⟨α⟩j ( −2) j [2(n − j )] !! ⎛ ⎞ ∑ ( −1)ℓ(2ℓ) ! ⎛ ⎞ ⎜ ⎟

j=1 ⎝ j ⎠ ℓ= 1 ⎝ ℓ ⎠ (2ℓ + 2n) !

for n ∈  . Consequently, the required result (5.3) is proved. □

 Expansions for real powers of two functions and Bell polynomials  731

Corollary 5.1. For k , n ∈ , we have

n j
⟨k ⟩j j ⎞ Q(2ℓ , 2n) Q(2k , 2n)
∑ (−1) j ∑ (−1)ℓ(2ℓ) ! ⎛ = (2k ) ! . (5.4)
j=1
j! ℓ= 1
ℓ
⎝ ⎠ ( 2ℓ + 2n ) ! ( 2k + 2n) !

Proof. This follows from letting α = k ∈  in (5.3) and equating coefficients of factors (x − 1)n in (4.1). □

Corollary 5.2. For α ∈  , we have

2 α ∞ n
⟨α⟩j j
j ⎞ Q(2ℓ , 2n)
⎛π ⎞ = 1 +
⎜ ⎟ ∑ (−1)n ∑ (−1) j ∑ (−1)ℓ(2ℓ) ! ⎛ . (5.5)
⎝9⎠ n=1 j=1
j! ℓ= 1 ⎝ ℓ ⎠ (2ℓ + 2n) !

1
Proof. This follows from setting x = 2
in (5.3). □

Remark 5.1. Formula (5.1) in Theorem 5.1 can be used to compute Taylor’s series expansions of functions
like f ( (arccos x )2
2(1 − x ) ) around the point x = 1, only if all the derivatives of f (x) at x = 1 are explicitly computable.

arcsin x k
6 Recovering Maclaurin’s series expansion at x = 0 of ( x )
In this section, by virtue of some conclusions in Lemmas 3.1, 3.2, and 3.3, we recover Maclaurin’s series
expansions (2.2) and (2.4) in Theorems 2.1 and 2.2.
It is easy to see that
∞
e t + e−t t 2k
cosh t = = ∑ .
2 k=0
(2k ) !

Then, making use of identity (3.1) in Lemma 3.1, the series expansion (3.16) can be reformulated as
∞ ∞ k−1
(arcsin x )2k 2k x2 ⎛ α 2 ⎞ x 2k
∑ α = 1 + α2 + α2 ∑ 4k − 1⎜∏⎡ℓ2 + ⎛ ⎞ ⎤⎟
(2k ) ! 2 ⎢ ⎝2⎠ ⎥
⎦⎠ (2k ) !
k=0 k=2 ⎝ ℓ= 1 ⎣
∞ k
x2 2 x 2k ( −1) j − 1
=1 + α + ∑ (−4)k−1 ∑
2 k=2
(2k ) ! j=1
4 j−1
2k − 1 2 ∞ 2k − 1
⎡ ⎛ ℓ ⎞ ℓ− 2j + 1⎤α 2j = 1 + x α 2 + α 2 x 2k ⎡ ℓ− 1⎤
⎢ ∑ 2j − 1 s(2k − 1, ℓ)(k − 1)
⎜ ⎟
⎥ 2
∑ (−4)k−1
( 2k ) ! ⎢ ∑ ℓs(2k − 1, ℓ)(k − 1) ⎥
⎣ ℓ= 2j − 1 ⎝ ⎠ ⎦ k=2 ⎣ ℓ= 1 ⎦
∞ k 2k − 1
x 2k ⎡ ⎛ ℓ ⎞s(2k − 1, ℓ)(k − 1)ℓ− 2j + 1⎤α2j
+ ∑ ∑ (−4)k−j⎢ ∑ ⎜ ⎟
⎥
(2k ) ! 2j − 1 ⎠
k=2 j=2 ⎣ ℓ= 2j − 1 ⎝ ⎦
∞
x2 2 x 2k
=1 + α + α2 ∑ ( −4)k − 1 ( −1)k − 1[(k − 1) ! ]2
2 k=2
(2k ) !
∞ ∞ 2k − 1
x 2k ⎡ ℓ ⎞ ⎤
+ ∑∑ ( −4)k − j ⎢ ∑ ⎛ ⎜ s(2k − 1, ℓ)(k − 1)ℓ− 2j + 1⎥α2j
⎟

( 2k ) ! 2 j − 1
j=2 k=j ⎣ ℓ= 2j − 1 ⎝ ⎠ ⎦
∞
x2 2 x 2k
= 1 + α + α2 ∑ [(2k − 2) !! ]2
2 k=2
(2k ) !
∞ ∞ 2m − 1
x 2m ℓ
+ ∑∑ ( −4)m − k ⎡
⎢ ∑⎛ ⎞s(2m − 1, ℓ)(m − 1)ℓ− 2k + 1⎤α2k ,
⎥
k=2 m=k
( 2m ) ! ⎝
⎣ℓ= 2k − 1 2k − 1⎠ ⎦
732  Feng Qi

where we used identity (3.2) or (3.14). Regarding α as a variable and equating coefficients of α2k we arrive at
∞ ∞
(arcsin x )2 x2 x 2k 1 (2k − 2) !! x 2k
= + ∑ [(2k − 2) !! ]2 = ∑ (6.1)
2 2 k=2
(2k ) ! 2 k = 1 (2k − 1) !! k

and
∞ 2m − 1
(arcsin x )2k ℓ x 2m
= ∑ ( −4)m − k ⎡
⎢ ∑ ⎛ ⎞s(2m − 1, ℓ)(m − 1)ℓ− 2k + 1⎤
⎥
(2k ) ! m=k ⎣ ℓ= 2k − 1 ⎝ 2k − 1 ⎠ ⎦ (2m) !
∞ (6.2)
x 2k ( −1)m (2x )2m
= ∑ Q(2k , 2m) , k ≥ 2.
(2k ) ! m=0 ( 2k + 2m
2k ) (2m) !

Making use of the series expansion (3.17) and identity (3.9) in Lemma 3.2, we obtain
∞ ∞ 2k 2k
(arcsin x )2k + 1 2k + 1 x 2k + 1 ⎡ s(2k , ℓ) ⎛ ℓ ⎞ ⎤
∑ α = α ∑ ( −1)k 22k ∑ (−1) j⎢ ∑ ℓ
⎜ (2k − 1)ℓ− 2j ⎥α2j
⎟

k=0
(2k + 1) ! k=0
(2k + 1) ! j=0
2 ⎝ 2j ⎠
⎣ ℓ= 2j ⎦
∞ ∞ 2k
⎡ (2x )2k + 1 s(2k , ℓ) ⎛ ℓ ⎞ ⎤
= ∑ (−1) j⎢ ∑ ( −1)k ∑ ℓ+ 1
⎜ (2k − 1)ℓ− 2j ⎥α2j + 1
⎟

j=0
(2k + 1) ! 2 ⎝ 2j ⎠
⎣ k =⌈j / 2⌉ ℓ= 2j ⎦
∞ ∞ 2m
(2x )2m + 1 s(2m , ℓ) ℓ
= ∑ (−1)k ⎡⎢ ∑ ( −1)m ∑ ℓ+ 1
⎛ ⎞(2m − 1)ℓ− 2k ⎤ 2k + 1
⎥α ,
k=0
(2m + 1) ! 2 ⎝ 2k ⎠
⎣ m =⌈k / 2⌉ ℓ= 2k ⎦
where we used identity (3.10) or (3.15) and ⌈x⌉ stands for the ceiling function which gives the smallest
integer not less than x . Regarding α as a variable and equating coefficients of α2k + 1 reduce to
∞ 2m
(arcsin x )2k + 1 s(2m , ℓ) ℓ (2x )2m + 1
= ( −1)k ∑ ( −1)m⎡ ⎢ ∑ ℓ+ 1
⎛ ⎞(2m − 1)ℓ− 2k ⎤
⎥ (2m + 1) !
(2k + 1) ! m =⌈k / 2⌉ ⎣ ℓ= 2k 2 ⎝ 2k ⎠ ⎦
∞ (6.3)
x 2k + 1 ( −1)m (2x )2m
= ∑ Q(2k + 1, 2m) .
(2k + 1) ! m=0 ( 2k + 2m + 1
2k + 1 ) (2m) !

Combining the series expansions (6.1), (6.2), and (6.3) leads to the series expansion (2.2).
By similar arguments as above, from the series expansions (3.21) and (3.22), we can recover series
expansion (2.2) once again.
Utilizing the relation arcsinht = −i arcsin(it ) or, equivalently, the relation arcsin t = −i arcsinh(it ), the
series expansions (2.2) and (2.4) can be derived from each other.

arcsin x k
7 A simple application of the series expansion of ( x ) to
Riemann zeta function
The special cases k = 1, 2, 3, 4 of Maclaurin’s series expansion (2.2) in Theorem 2.1 have been applied in
[30–33]. These applications have been reviewed and generalized in [1,27].
Now we quote a paragraph in [32, Section 5] as follows.
Wolfram alpha (https://www.wolframalpha.com) says that
1 π /2
(arcsin x )3 π3 9
∫ x
dx = ∫ u3 cot udu = 8
ln2 −
16
πζ (3) (7.1)
0 0

and
 Expansions for real powers of two functions and Bell polynomials  733

1 π /2
(arcsin x )4
∫ x
dx = ∫ u4 cot udu = 321 [2π 4 ln2 − 18π 2ζ (3) + 93ζ (5)], (7.2)
0 0

where the Riemann zeta function ζ (z ) can be defined [44, Fact 13.3] for z ∈  with R(z ) > 1 by
∞ ∞ ∞
1 1 1 1 1
ζ (z ) = ∑ z
= −z ∑ (2k − 1) z
= 1−z ∑
( −1)k − 1 z . (7.3)
k=1
k 1 − 2 k=1
1 − 2 k=1
k

It should be possible to describe such integrals as certain infinite sums with or without numbers
(2n − 1) !! 1 2n
w2n = = 2n ⎛ ⎞ . (7.4)
(2n) !! 2 ⎝n⎠

We plan to study those details in subsequent publication.

Substituting Maclaurin’s series expansion (2.2) for k = 3, 4 into the left hand sides of (7.1) and (7.2) and
simplifying produce
1
3 ∞
Q(3, 2m) 22m + 1
∫ (arcsin
x
x)
dx =
1
3
+ 3 ∑ ( −1)m
(3 + 2m) ! 2m + 3
(7.5)
m=1
0

and
1
4 ∞
Q(4, 2m) 22m + 3
∫ (arcsin
x
x)
dx =
1
4
+ 3 ∑ ( −1)m
(4 + 2m) ! 2m + 4
. (7.6)
m=1
0

Combining (7.1) and (7.2) with (7.5) and (7.6) and simplifying yield
∞
16 Q(3, 2m) 22m + 3
2π3 ln2 − 9πζ (3) = + 12 ∑ ( −1)m
3 m=1
(3 + 2m) ! 2m + 3

and
∞
Q(4, 2m) 22m + 4
2π 4 ln2 − 18π 2ζ (3) + 93ζ (5) = 8 + 48 ∑ ( −1)m ,
m=1
(4 + 2m) ! 2m + 4

where Q(3, 2m) and Q(4, 2m) are defined by the quantity in (2.3).

Remark 7.1. On the Riemann zeta function ζ (z ) defined in (7.3) and its recent applications, we recommend
the articles [19,41,45–48]. On the Wallis ratio defined in (7.4) and its properties, we recommend the articles
[49–51] and closely related references therein.

8 Conclusion
In this article, by virtue of Lemmas 3.1 and 3.2, with the aid of Taylor’s series expansion (3.23) or (3.18) in
Lemma 3.3, and in the light of properties recited in Lemma 3.4 of partial Bell polynomials, the author
establishes Taylor’s series expansions (4.1), (4.2), and (5.3) in Theorems 4.1 and 5.2, presents a closed-form
formula (5.1), derives several combinatorial identities (3.2), (3.3), (3.11), (4.23), (5.2), and (5.4), demonstrates
several series representations (4.9), (4.10), and (5.5) in Corollaries 4.2 and 5.2 of the circular constant π and
its real powers, recovers Maclaurin’s series expansions (2.2) and (2.4) in Section 6, and finally applies the
series expansion (2.2) in Theorem 2.1 to derive series representations for quantities containing π and its
powers, ζ (3), and ζ (5).
Those conclusions stated in Corollaries 4.1, 4.3, 4.4, 4.5, and 4.6 are useful, meaningful, and significant.
734  Feng Qi

Several Maclaurin’s series expansions of the functions (arccos x )m and (arccoshx )m for m ∈  have been
surveyed, reviewed, collected, and mentioned in [27, Section 7], but comparatively their forms or formula-
tions are not more beautiful, not more satisfactory, not simpler, not more concise, or not nicer than these
newly established ones in this article.
This article is a revised version of the preprint [52] and a companion of the papers [1,27,39,43,53].

Acknowledgment: The author is thankful to anonymous referees for their careful corrections to and helpful
comments on the original version of this article.

Funding information: None declared.

Author contributions: The author has accepted responsibility for the entire content of this manuscript and
approved its submission.

Conflict of interest: The author states no conflict of interest.

Ethical approval: The conducted research is not related to either human or animal use.

Data availability statement: Data sharing is not applicable to this article as no datasets were generated or
analyzed during the current study.

References
[1] B.-N. Guo, D. Lim, and F. Qi, Series expansions of powers of arcsine, closed forms for special values of Bell polynomials,
and series representations of generalized logsine functions, AIMS Math. 6 (2021), no. 7, 7494–7517, DOI: https://doi.org/
10.3934/math.2021438.
[2] E. R. Hansen, A Table of Series and Products, Prentice-Hall, Englewood Cliffs, NJ, USA, 1975.
[3] N. M. Temme, Special Functions: An Introduction to Classical Functions of Mathematical Physics, A Wiley-Interscience
Publication, John Wiley & Sons, Inc., New York, 1996, DOI: http://dx.doi.org/10.1002/9781118032572.
[4] C.-F. Wei, Integral representations and inequalities of extended central binomial coefficients, Math. Methods Appl. Sci. 45
(2022), no. 9, 5412–5422, DOI: https://doi.org/10.1002/mma.8115.
[5] F. Qi, Diagonal recurrence relations for the Stirling numbers of the first kind, Contrib. Discrete Math. 11 (2016), no. 1,
22–30, DOI: https://doi.org/10.11575/cdm.v11i1.62389.
[6] F. Qi, Diagonal recurrence relations, inequalities, and monotonicity related to the Stirling numbers of the second kind,
Math. Inequal. Appl. 19 (2016), no. 1, 313–323, DOI: https://doi.org/10.7153/mia-19-23.
[7] F. Qi and B.-N. Guo, A diagonal recurrence relation for the Stirling numbers of the first kind, Appl. Anal. Discrete Math. 12
(2018), no. 1, 153–165, DOI: https://doi.org/10.2298/AADM170405004Q.
[8] C. A. Charalambides, Enumerative Combinatorics, CRC Press Series on Discrete Mathematics and its Applications.
Chapman and Hall/CRC, Boca Raton, FL, 2002.
[9] L. Comtet, Advanced Combinatorics: The Art of Finite and Infinite Expansions, Revised and Enlarged Edition, D. Reidel
Publishing Co., 1974, DOI: https://doi.org/10.1007/978-94-010-2196-8.
[10] F. W. J. Olver, D. W. Lozier, R. F. Boisvert, and C. W. Clark (eds.), NIST Handbook of Mathematical Functions, Cambridge
University Press, New York, 2010, http://dlmf.nist.gov/.
[11] M. Abramowitz and I. A. Stegun (Eds), Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical
Tables, National Bureau of Standards, Applied Mathematics Series 55, 10th printing, Washington, 1972.
[12] F. Qi, D.-W. Niu, D. Lim, and Y.-H. Yao, Special values of the Bell polynomials of the second kind for some sequences and
functions, J. Math. Anal. Appl. 491 (2020), no. 2, Article 124382, 31 pages, DOI: https://doi.org/10.1016/
j.jmaa.2020.124382.
[13] F. Qi, X.-T. Shi, and F.-F. Liu, Expansions of the exponential and the logarithm of power series and applications, Arab. J.
Math. (Springer) 6 (2017), no. 2, 95–108, DOI: https://doi.org/10.1007/s40065-017-0166-4.
[14] J. Quaintance and H. W. Gould, Combinatorial Identities for Stirling Numbers, The unpublished notes of H. W. Gould. With a
foreword by George E. Andrews, World Scientific Publishing Co. Pte. Ltd, Singapore, 2016.
[15] V. R. Thiruvenkatachar and T. S. Nanjundiah, Inequalities concerning Bessel functions and orthogonal polynomials, Proc.
Ind. Acad. Sci. Sect. A 33 (1951), 373–384.
 Expansions for real powers of two functions and Bell polynomials  735

[16] Z.-H. Yang and S.-Z. Zheng, Monotonicity and convexity of the ratios of the first kind modified Bessel functions and
applications, Math. Inequal. Appl. 21 (2018), no. 1, 107–125, DOI: https://doi.org/10.7153/mia-2018-21-09.
[17] C. M. Bender, D. C. Brody, and B. K. Meister, On powers of Bessel functions, J. Math. Phys. 44 (2003), no. 1, 309–314,
DOI: https://doi.org/10.1063/1.1526940.
[18] Á. Baricz, Powers of modified Bessel functions of the first kind, Appl. Math. Lett. 23 (2010), no. 6, 722–724, DOI: https://
doi.org/10.1016/j.aml.2010.02.015.
[19] Y. Hong, B.-N. Guo, and F. Qi, Determinantal expressions and recursive relations for the Bessel zeta function and for a
sequence originating from a series expansion of the power of modified Bessel function of the first kind, CMES Comput.
Model. Eng. Sci. 129 (2021), no. 1, 409–423, DOI: https://doi.org/10.32604/cmes.2021.016431.
[20] F. T. Howard, Integers related to the Bessel function J1 (z ), Fibonacci. Quart. 23 (1985), no. 3, 249–257.
[21] V. H. Moll and C. Vignat, On polynomials connected to powers of Bessel functions, Int. J. Number Theory 10 (2014), no. 5,
1245–1257, DOI: https://doi.org/10.1142/S1793042114500249.
[22] M. Bakker and N. M. Temme, Sum rule for products of Bessel functions: comments on a paper by Newberger, J. Math. Phys.
25 (1984), no. 5, 1266–1267, DOI: https://doi.org/10.1063/1.526282.
[23] B. S. Newberger, Erratum: New sum rule for products of Bessel functions with application to plasma physics, J. Math. Phys.
24 (1983), no. 8, 2250–2250, DOI: https://doi.org/10.1063/1.525940.
[24] B. S. Newberger, New sum rule for products of Bessel functions with application to plasma physics, J. Math. Phys. 23
(1982), no. 7, 1278–1281, DOI: https://doi.org/10.1063/1.525510.
[25] J. M. Borwein and M. Chamberland, Integer powers of arcsin, Int. J. Math. Math. Sci. 2007 (2007), Art. ID 19381, 10 pages,
DOI: https://doi.org/10.1155/2007/19381.
[26] Yu. A. Brychkov, Power expansions of powers of trigonometric functions and series containing Bernoulli and Euler poly-
nomials, Integral Transforms Spec. Funct. 20 (2009), no. 11–12, 797–804, DOI: https://doi.org/10.1080/
10652460902867718.
[27] B.-N. Guo, D. Lim, and F. Qi, Maclaurin’s series expansions for positive integer powers of inverse (hyperbolic) sine and
tangent functions, closed-form formula of specific partial Bell polynomials, and series representation of generalized
logsine function, Appl. Anal. Discrete Math. 17 (2023), no. 1, in Press. DOI: https://doi.org/10.2298/AADM210401017G.
[28] F. Qi, Derivatives of tangent function and tangent numbers, Appl. Math. Comput. 268 (2015), 844–858, DOI: http://
dx.doi.org/10.1016/j.amc.2015.06.123.
[29] F. Qi, C.-P. Chen, and D. Lim, Several identities containing central binomial coefficients and derived from series expansions
of powers of the arcsine function, Results Nonlinear Anal. 4 (2021), no. 1, 57–64, DOI: https://doi.org/10.53006/
rna.867047.
[30] A. I. Davydychev and M. Y Kalmykov, New results for the ε -expansion of certain one-, two- and three-loop Feynman
diagrams, Nuclear Phys. B 605 (2001), no. 1–3, 266–318, DOI: https://doi.org/10.1016/S0550-3213(01)00095-5.
[31] M. Yu. Kalmykov and A. Sheplyakov, lsjk–a C++ library for arbitrary-precision numeric evaluation of the generalized log-
sine functions, Computer Phys. Commun. 172 (2005), no. 1, 45–59, DOI: https://doi.org/10.1016/j.cpc.2005.04.013.
[32] M. Kobayashi, Integral representations for local dilogarithm and trilogarithm functions, Open J. Math. Sci. 5 (2021), no. 1,
337–352, DOI: https://doi.org/10.30538/oms2021.0169.
[33] F. Oertel, Grothendieck’s inequality and completely correlation preserving functions–a summary of recent results and an
indication of related research problems, arXiv: https://arxiv.org/abs/2010.00746v2.
[34] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, Translated from the Russian, Translation edited
and with a preface by Daniel Zwillinger and Victor Moll, Eighth edition, Revised from the seventh edition, Elsevier/
Academic Press, Amsterdam, 2015, DOI: https://doi.org/10.1016/B978-0-12-384933-5.00013-8.
[35] M. I. Qureshi, J. Majid, and A. H. Bhat, Hypergeometric forms of some composite functions containing arccosine(x ) using
Maclaurin’s expansion, South East Asian J. Math. Math. Sci. 16 (2020), no. 3, 83–95.
[36] F. Qi and B.-N. Guo, Explicit formulas for special values of the Bell polynomials of the second kind and for the Euler
numbers and polynomials, Mediterr. J. Math. 14 (2017), no. 3, Art. 140, 14 pages, DOI: https://doi.org/10.1007/s00009-
017-0939-1.
[37] F. Qi, D.-W. Niu, D. Lim, and B.-N. Guo, Closed formulas and identities for the Bell polynomials and falling factorials,
Contrib. Discrete Math. 15 (2020), no. 1, 163–174, DOI: https://doi.org/10.11575/cdm.v15i1.68111.
[38] F. Qi, X.-T. Shi, F.-F. Liu, and D. V. Kruchinin, Several formulas for special values of the Bell polynomials of the second kind
and applications, J. Appl. Anal. Comput. 7 (2017), no. 3, 857–871, DOI: https://doi.org/10.11948/2017054.
[39] F. Qi, Explicit formulas for partial Bell polynomials, Maclaurin’s series expansions of real powers of inverse (hyperbolic)
cosine and sine, and series representations of powers of Pi, Research Square (2021), DOI: https://doi.org/10.21203/
rs.3.rs-959177/v3.
[40] D. H. Lehmer, Interesting series involving the central binomial coefficient, Amer. Math. Monthly 92 (1985), no. 7, 449–457,
DOI: http://dx.doi.org/10.2307/2322496.
[41] Q.-M. Luo, B.-N. Guo, and F. Qi, On evaluation of Riemann zeta function ζ (s), Adv. Stud. Contemp. Math. (Kyungshang) 7
(2003), no. 2, 135–144.
[42] S. Jin, B.-N. Guo, and F. Qi, Partial Bell polynomials, falling and rising factorials, Stirling numbers, and combinatorial
identities, CMES Comput. Model. Eng. Sci. 132 (2022), no. 3, 781–799, DOI: https://dx.doi.org/10.32604/
cmes.2022.019941.
736  Feng Qi

[43] F. Qi and M. D. Ward, Closed-form Formulas and Properties of Coefficients in Maclaurin’s Series Expansion of Wilf’s
Function Composited by Inverse Tangent, Square Root, and Exponential Functions, 2022, arXiv: https://arxiv.org/abs/
2110.08576v2.
[44] D. S. Bernstein, Scalar, Vector, and Matrix Mathematics: Theory, Facts, and Formulas, Revised and expanded edition,
Princeton University Press, Princeton, NJ, 2018.
[45] F. Qi, A double inequality for the ratio of two non-zero neighbouring Bernoulli numbers, J. Comput. Appl. Math. 351 (2019),
1–5, DOI: https://doi.org/10.1016/j.cam.2018.10.049.
[46] Y. Shuang, B.-N. Guo, and F. Qi, Logarithmic convexity and increasing property of the Bernoulli numbers and their ratios,
Rev. R. Acad. Cienc. Exactas Fís. Nat. Ser. A Mat. RACSAM 115 (2021), no. 3, Paper No. 135, 12 pages, DOI: https://doi.org/
10.1007/s13398-021-01071-x.
[47] Z.-H. Yang and J.-F. Tian, Sharp bounds for the ratio of two zeta functions, J. Comput. Appl. Math. 364 (2020), 112359, 14
pages, DOI: https://doi.org/10.1016/j.cam.2019.112359.
[48] L. Zhu, New bounds for the ratio of two adjacent even-indexed Bernoulli numbers, Rev. R. Acad. Cienc. Exactas Fís. Nat.
Ser. A Mat. RACSAM 114 (2020), no. 2, Paper No. 83, 13 pages, DOI: https://doi.org/10.1007/s13398-020-00814-6.
[49] C.-P. Chen and F. Qi, The best bounds in Wallis’ inequality, Proc. Amer. Math. Soc. 133 (2005), no. 2, 397–401, DOI: http://
dx.doi.org/10.1090/S0002-9939-04-07499-4.
[50] S. Guo, J.-G. Xu, and F. Qi, Some exact constants for the approximation of the quantity in the Wallis’ formula, J. Inequal.
Appl. 2013 (2013), Paper No. 67, 7 pages, DOI: https://doi.org/10.1186/1029-242X-2013-67.
[51] F. Qi and C. Mortici, Some best approximation formulas and inequalities for the Wallis ratio, Appl. Math. Comput. 253
(2015), 363–368, DOI: https://doi.org/10.1016/j.amc.2014.12.039.
[52] F. Qi, Taylor’s series expansions for real powers of functions containing squares of inverse (hyperbolic) cosine functions,
explicit formulas for special partial Bell polynomials, and series representations for powers of circular constant, arXiv:
https://arxiv.org/abs/2110.02749v2.
[53] F. Qi and P. Taylor, Several series expansions for real powers and several formulas for partial Bell polynomials of sinc and
sinhc functions in terms of central factorial and Stirling numbers of second kind, arXiv: https://arxiv.org/abs/2204.
05612v4.