Preprint
Article

This version is not peer-reviewed.

Complete and Partial Bell Polynomials at Bernoulli Numbers and Other Sequences, with Applications to Multiple Zeta Values

Feng Qi  *

Submitted:

15 November 2026

Posted:

17 November 2026

You are already at the latest version

Abstract
Comparing the Maclaurin series of a function with that of its logarithm or exponential yields closed-form values of the complete Bell polynomials. Combining this principle with the Faà di Bruno formula and Newton's forward differences, the author evaluates the complete and partial Bell polynomials at sequences built from the Bernoulli numbers, eta and zeta values, factorials, and Stirling, harmonic, and Catalan numbers, and at the Maclaurin coefficients of logarithms of several elementary functions. The main new results concern products of sine and hyperbolic sine functions, which involve the Bernoulli numbers B4k, and exponentials of six elementary functions. The radius of convergence of the series of \( \bigl(\frac{\tan z}{z}\bigr)^\lambda \) is determined for every real λ. As applications, the author derives a convolution identity for B4k, closed forms and monotonicity of the multiple zeta values \( \zeta(4,4,\dotsc,4) \) and \( \zeta^*(4,4,\dotsc,4) \) , and Ramanujan-type evaluations of series involving \( \frac1{\sinh(\pi m)} \).
Keywords: 
;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  

MSC:  Primary 05A19, 11B73, 11M32; Secondary 05A15, 11B68, 11B83, 26A24, 26A48, 30B10, 30E20, 33B10, 33C45, 41A58

1. Introduction

In 1992, Hoffman [20] and Zagier [48] introduced the multiple harmonic series (also known as the multiple zeta functions or the multiple zeta values) by
ζ * ( z 1 , z 2 , … , z k ) = ∑ n 1 ≥ n 2 ≥ … ≥ n k n 1 , n 2 , … , n k ∈ N 1 n 1 z 1 n 2 z 2 … n k z k
and
ζ ( z 1 , z 2 , … , z k ) = ∑ n 1 > n 2 > … > n k n 1 , n 2 , … , n k ∈ N 1 n 1 z 1 n 2 z 2 … n k z k
for k ∈ N = { 1 , 2 , … } and ℜ ( z 1 + z 2 + … + z j ) > j with j = 1 , 2 , … , k . These are natural generalizations of the classical Riemann zeta function
ζ ( z ) = ∑ n = 1 ∞ 1 n z , ℜ ( z ) > 1 .
For more information about ζ ( z ) , please refer to [45, Section 3.5]. The multiple zeta functions ζ * ( z 1 , z 2 , … , z k ) and ζ ( z 1 , z 2 , … , z k ) play important roles in quantum physics and knot theory; see the monographs [5,49].
In [11], the partial Bell polynomials, also known as the Bell polynomials of the second kind, denoted by B k , j ( a 1 , a 2 , … , a k − j + 1 ) , are defined by
B k , j ( a 1 , a 2 , … , a k − j + 1 ) k ! = ∑ ∑ i = 1 k − j + 1 i ℓ i = k ∑ i = 1 k − j + 1 ℓ i = j ℓ 1 , ℓ 2 , … , ℓ k − j + 1 ∈ N 0 ∏ i = 1 k − j + 1 1 ℓ i ! a i i ! ℓ i
for j , k ∈ N 0 satisfying k ≥ j , with the special cases B 0 , 0 ( a 1 ) = 1 and
B k , 0 ( a 1 , a 2 , … , a k + 1 ) = 0 , k ∈ N .
In [7, p. 412, Definition 11.1] and [11, p. 134], the complete Bell polynomials, denoted by Y k ( a 1 , a 2 , … , a k ) , are defined by Y 0 = 1 and
Y k ( a 1 , a 2 , … , a k ) = ∑ j = 1 k B k , j ( a 1 , a 2 , … , a k − j + 1 ) , k ∈ N .
For more information on the partial and complete Bell polynomials, please refer to [11, Section 3.3, pp. 133–137] and the survey article [40]. From the relation (1), it follows that
Y k ( a 1 , a 2 , … , a k ) k ! = ∑ ∑ i = 1 k i ℓ i = k ℓ 1 , ℓ 2 , … , ℓ k ∈ N 0 ∏ i = 1 k 1 ℓ i ! a i i ! ℓ i , k ∈ N .
The complete Bell polynomials Y k ( a 1 , a 2 , … , a k ) have an exponential generating function
exp ∑ k = 1 ∞ a k z k k ! = ∑ k = 0 ∞ Y k ( a 1 , a 2 , … , a k ) z k k ! ,
which can be found in [7, Chapter 11] and [11, p. 134].
In [9, Proposition 3], Chen, Chung, and Eie proved that
ζ ( 2 n , … , 2 n ︸ k ) = ( − 1 ) k ( n + 1 ) ( 2 π ) 2 n k ∑ ∑ i = 1 k i ℓ i = k ℓ 1 , ℓ 2 , … , ℓ k ∈ N 0 ∏ i = 1 k 1 ℓ i ! 1 2 i B 2 i n ( 2 i n ) ! ℓ i
and
ζ * ( 2 n , … , 2 n ︸ k ) = ( − 1 ) n k ( 2 π ) 2 n k ∑ ∑ i = 1 k i ℓ i = k ℓ 1 , ℓ 2 , … , ℓ k ∈ N 0 ∏ i = 1 k ( − 1 ) ℓ i ℓ i ! 1 2 i B 2 i n ( 2 i n ) ! ℓ i
for n , k ∈ N , where B i denotes the Bernoulli numbers generated [45, p. 3, Eq. (1.1)] by
z e z − 1 = ∑ i = 0 ∞ B i z i i ! = 1 − 1 2 z + ∑ i = 1 ∞ B 2 i z 2 i ( 2 i ) ! , | z | < 2 π .
For recent properties of the Bernoulli numbers B 2 i , see [2] and the references therein. Sum relations for multiple zeta (star) values with even arguments are also studied in [8,13]. In [30, Section 1.8.2], by (2), the author reformulated the formulas (4) and (5) as
ζ ( 2 n , 2 n , … , 2 n ︸ k ) = ( − 1 ) k ( n + 1 ) ( 2 π ) 2 n k k ! Y k 0 ! 2 B 2 n ( 2 n ) ! , 1 ! 2 B 4 n ( 4 n ) ! , … , ( k − 1 ) ! 2 B 2 k n ( 2 k n ) !
and
ζ * ( 2 n , 2 n , … , 2 n ︸ k ) = ( − 1 ) n k ( 2 π ) 2 n k k ! Y k − 0 ! 2 B 2 n ( 2 n ) ! , − 1 ! 2 B 4 n ( 4 n ) ! , … , − ( k − 1 ) ! 2 B 2 k n ( 2 k n ) !
for k , n ∈ N . These intrinsic observations and reformulations reveal a natural connection between the multiple zeta functions ζ * ( z 1 , … , z k ) and ζ ( z 1 , … , z k ) and the complete Bell polynomials Y k ( a 1 , a 2 , … , a k ) . In fact, this connection was earlier established in the paper [19].
Corresponding to n = 1 in (6) and (7), the complete Bell polynomials
Y k λ 0 ! B 2 2 ! , λ 1 ! B 4 4 ! , … , λ ( k − 1 ) ! B 2 k ( 2 k ) !
and
Y k λ 2 2 − 1 0 ! B 2 2 ! , λ 2 4 − 1 1 ! B 4 4 ! , … , λ 2 2 k − 1 ( k − 1 ) ! B 2 k ( 2 k ) !
for k ∈ N 0 and λ ∈ R were evaluated in the papers [12,19,27,28,46].
Corresponding to n = 2 in (6) and (7), the complete Bell polynomials
Y k λ 0 ! B 4 4 ! , λ 1 ! B 8 8 ! , … , λ ( k − 1 ) ! B 4 k ( 4 k ) !
for k ∈ N 0 and λ ∈ R were evaluated in the paper [30].
Series expansions of real powers of the sinc, inverse sine, inverse cosine, and inverse tangent functions, and related values of partial Bell polynomials, were studied in [16,17,21,31,34,38,41] and, especially, in the survey article [32].
In this paper, we further develop explicit evaluations of the complete and partial Bell polynomials Y k and B n , k at several sequences defined in terms of the Bernoulli numbers and other special numbers. As an application, we determine ζ ( 4 , 4 , … , 4 ) and ζ * ( 4 , 4 , … , 4 ) and study their monotonicity with respect to the number of arguments.
We organize this paper as follows. In Section 2, we fix notation and collect the lemmas used later. In Section 3, we evaluate Y k and B n , k at sequences related to the tangent and cosine functions, falling factorials, eta and zeta values, power-exponential functions, Gegenbauer polynomials, the inverse hyperbolic sine, and the logarithmic and arctangent functions. In Section 4, we treat products of sine and hyperbolic sine functions, which lead to the Bernoulli numbers B 4 k . In Section 5, we treat exponentials of six elementary functions and real powers of the generating function of the Catalan numbers. In Section 6, we derive a convolution identity for B 4 k , closed forms of ζ ( 4 , 4 , … , 4 ) and ζ * ( 4 , 4 , … , 4 ) , and monotonicity properties of these two sequences. Section 7 concludes the paper.

2. Preliminaries

In this section, we fix notation, collect the lemmas on Bell polynomials used in the proofs, and record some series expansions of elementary functions. Lemma 9 reduces partial Bell polynomials to complete ones, and Lemmas 12 and 13 prepare the results on products of sine and hyperbolic sine functions in Section 4. For further identities for partial Bell polynomials and Stirling numbers, see [22,42,43,47,50].

2.1. Notation and Special Numbers

In this subsection, we fix the notation used throughout the paper and recall the special numbers and polynomials in which our evaluations are expressed: the falling and rising factorials, the Stirling numbers of both kinds, the Lah numbers, and the generalized Bernoulli and Euler polynomials. Collecting these definitions in one place ensures that every symbol is defined before it is first used.
Throughout this paper, N = { 1 , 2 , … } and N 0 = { 0 , 1 , 2 , … } . For a function G analytic at the origin, [ x k ] G ( x ) denotes the coefficient of x k in the Maclaurin series of G ( x ) ; the same notation is used for other variables and for polynomials.
For z ∈ C and n ∈ N 0 , the falling factorial 〈 z 〉 n and the rising factorial ( z ) n are defined [39, p. 167] by
〈 z 〉 n = ∏ k = 0 n − 1 ( z − k ) and ( z ) n = ∏ k = 0 n − 1 ( z + k ) ,
with 〈 z 〉 0 = ( z ) 0 = 1 , and z n = 〈 z 〉 n n ! .
The Stirling numbers of the first kind s ( n , k ) are defined by
〈 z 〉 n = ∑ k = 0 n s ( n , k ) z k
and they are generated [25, Theorem 3.14] by
ln ( 1 + z ) z j = ∑ k = 0 ∞ s ( k + j , j ) k + j j z k k ! , | z | < 1 , j ∈ N 0 .
The Stirling numbers of the second kind S ( n , j ) are generated by
e z − 1 z j = ∑ n = 0 ∞ S ( n + j , j ) n + j j z n n ! , j ∈ N 0 , z ∈ C .
The unsigned Lah numbers are [11, p. 135]
L ( k , j ) = k − 1 j − 1 k ! j ! , k , j ∈ N , k ≥ j ,
with L ( 0 , 0 ) = 1 and L ( k , 0 ) = 0 for k ∈ N .
For σ ∈ C , the generalized Bernoulli polynomials B k ( σ ) ( x ) and the generalized Euler polynomials E k ( σ ) ( x ) are generated [45, Chapter 1] by
z e z − 1 σ e x z = ∑ k = 0 ∞ B k ( σ ) ( x ) z k k ! , | z | < 2 π
and
2 e z + 1 σ e x z = ∑ k = 0 ∞ E k ( σ ) ( x ) z k k ! , | z | < π .
In particular, B k ( 1 ) ( 0 ) = B k , and B k ( x ) = B k ( 1 ) ( x ) are the Bernoulli polynomials.
Finally, ψ ( z ) = Γ ′ ( z ) Γ ( z ) denotes the digamma function, so that
ψ ( k + 1 ) − ψ ( 1 ) = ∑ j = 1 k 1 j , k ∈ N .

2.2. Lemmas on Bell Polynomials

In this subsection, we collect the tools on Bell polynomials on which all evaluations rest. Lemmas 1 and 2 turn the Maclaurin series of a function and of its logarithm, or of a function and of its exponential, into values of Y k . Lemmas 3 to 7 provide the Faà di Bruno formula, the homogeneity of B n , k , and closed forms of B n , k at special sequences, which we use to compute higher derivatives. Lemmas 8 and 9 recover B n , k from the integer powers of a function and from the values of Y n at integer multiples of a sequence, respectively.
Lemma 1. 
If
f ( x ) = ∑ k = 0 ∞ b k x k k !
and
ln f ( x ) = ∑ k = 1 ∞ a k x k k ! ,
then
Y k ( a 1 , a 2 , … , a k ) = b k , k ∈ N .
Proof. 
The series expansions (12) and (13) can be combined as
f ( x ) = exp ∑ k = 1 ∞ a k x k k ! = ∑ k = 0 ∞ b k x k k ! .
Comparing this equation with (3) immediately yields (14). Lemma 1 is thus proved. □
Lemma 2. 
If
f ( x ) = ∑ k = 1 ∞ b k x k k !
and
e f ( x ) = ∑ k = 0 ∞ a k x k k ! ,
then
Y k ( b 1 , b 2 , … , b k ) = a k , k ∈ N .
Proof. 
It is easy to see that the series expansions (15) and (16) can be combined as
exp ∑ k = 1 ∞ b k x k k ! = ∑ k = 0 ∞ a k x k k ! = e f ( x ) .
Comparing this equation with (3) immediately yields (17). Lemma 2 is thus proved. □
Lemma 3 
([7, Theorem 11.4, 10], and [11, p. 139, Theorem C]). In terms of the partial Bell polynomials B n , k ( z 1 , z 2 , … , z n − k + 1 ) , the Faà di Bruno formula states that
[ f ∘ h ( z ) ] ( n ) = ∑ k = 0 n f ( k ) ( h ( z ) ) B n , k h ′ ( z ) , h ′ ′ ( z ) , … , h ( n − k + 1 ) ( z ) , n ∈ N 0 ,
where f and h are n times differentiable, f at the point h ( z ) and h at the point z.
Lemma 4 
(Homogeneity [7] and [11]). For n , k ∈ N 0 with n ≥ k and a , b ∈ C , the partial Bell polynomials B n , k ( z 1 , z 2 , … , z n − k + 1 ) satisfy
B n , k a b x 1 , a b 2 x 2 , … , a b n − k + 1 x n − k + 1 = a k b n B n , k ( x 1 , x 2 , … , x n − k + 1 ) .
Lemma 5 
([11, p. 135, Theorem B]). For n , k ∈ N with n ≥ k , the partial Bell polynomials B n , k ( z 1 , z 2 , … , z n − k + 1 ) satisfy
B n , k ( 1 ! , 2 ! , … , ( n − k + 1 ) ! ) = n − 1 k − 1 n ! k ! .
Remark 1. 
The right-hand side of (20) is the unsigned Lah number L ( n , k ) defined in (10).
Lemma 6 
([36, Theorem 5.1] and [37, Theorem 5]). For n , k ∈ N 0 with n ≥ k , the partial Bell polynomials B n , k satisfy
B n , k ( z 1 , z 2 , 0 , … , 0 ) = ( n − k ) ! 2 n − k n k k n − k z 1 2 k − n z 2 n − k .
Lemma 7 
([29, Theorem 1.1]]). For n , k ∈ N 0 with n ≥ k , we have
B n , k 1 ! 2 , 2 ! 3 , … , ( n − k + 1 ) ! n − k + 2 = ( − 1 ) n − k k ! ∑ m = 0 k ( − 1 ) m k m s ( n + m , m ) n + m m ,
where s ( n , k ) denotes the Stirling numbers of the first kind defined in SubSection 2.1.
Lemma 8 
([38, Theorem 2.1]). Let f ( z ) be analytic at z = 0 with f ( 0 ) = 1 , and, for q ∈ N 0 , let
[ f ( z ) ] q = ∑ m = 0 ∞ C q , m z m m ! ,
so that C 0 , 0 = 1 and C 0 , m = 0 for m ∈ N . Then the partial Bell polynomials B n , k ( x 1 , x 2 , … , x n − k + 1 ) for n , k ∈ N 0 with n ≥ k satisfy
B n , k f ′ ( 0 ) , f ′ ′ ( 0 ) , … , f ( n − k + 1 ) ( 0 ) = ( − 1 ) k k ! ∑ q = 0 k ( − 1 ) q k q C q , n .
Lemma 9. 
Let n ∈ N and let a 1 , a 2 , … , a n be complex numbers. Then
Y n ( λ a 1 , λ a 2 , … , λ a n ) = ∑ k = 1 n λ k B n , k a 1 , a 2 , … , a n − k + 1 , λ ∈ C ,
and, for k = 1 , 2 , … , n ,
B n , k a 1 , a 2 , … , a n − k + 1 = ∑ j = k n s ( j , k ) j ! ∑ q = 0 j ( − 1 ) j − q j q Y n ( q a 1 , q a 2 , … , q a n ) .
Proof. 
Since, by (19),
B n , k ( λ x 1 , λ x 2 , … , λ x n − k + 1 ) = λ k B n , k ( x 1 , x 2 , … , x n − k + 1 )
and (1), we obtain (24). Thus p ( λ ) = Y n ( λ a 1 , … , λ a n ) is a polynomial in λ of degree at most n with p ( 0 ) = 0 . By Newton’s forward difference formula [1], we obtain
p ( λ ) = ∑ j = 0 n Δ j p ( 0 ) λ j and Δ j p ( 0 ) = ∑ q = 0 j ( − 1 ) j − q j q p ( q ) .
Employing the identity
λ j = 1 j ! ∑ k = 0 j s ( j , k ) λ k ,
see [11], and comparing the coefficients of λ k with those in (24) give (25). □

2.3. Series Expansions of Some Elementary Functions

In this subsection, we record series expansions of elementary functions that are used repeatedly in Section 3 to Section 6. Lemma 10 expresses real powers of sin z z and cos z through generalized Bernoulli and Euler polynomials, and Lemma 11 gives real powers of arctan x x . Lemmas 12 and 13 treat the product sin x sinh x , whose Maclaurin series contains only the powers x 4 m + 2 ; this lacunary structure is the source of the Bernoulli numbers B 4 k in Section 4.
Lemma 10 
([28, Remarks 3.1 and 3.2]). For r , λ ∈ R , we have
sin z z r = ∑ k = 0 ∞ ( − 1 ) k B 2 k ( − r ) − r 2 ( 2 z ) 2 k ( 2 k ) ! , | z | < π ,
and
( cos z ) λ = ∑ k = 0 ∞ ( − 1 ) k E 2 k ( − λ ) − λ 2 ( 2 z ) 2 k ( 2 k ) ! , | z | < π 2 ,
where the branches equal to 1 at z = 0 are taken. When r ∈ N 0 or λ ∈ N 0 , the corresponding series expansion holds for all z ∈ C . Moreover,
E 2 k + 1 ( 2 ϵ ) ( ϵ ) = 0 , ϵ ∈ C , k ∈ N 0 .
Remark 2. 
The domains of the series expansions (26) and (27) are sharp. The function sin z z is analytic and nonzero for | z | < π and has simple zeros at z = ± π , so sin z z r behaves like a nonzero constant multiple of ( π ∓ z ) r near z = ± π . Hence the radius of convergence of (26) equals π for r ∈ R ∖ N 0 and is infinite for r ∈ N 0 . Similarly, cos z is nonzero for | z | < π 2 and has simple zeros at z = ± π 2 , so the radius of convergence of (27) equals π 2 for λ ∈ R ∖ N 0 and is infinite for λ ∈ N 0 .
Lemma 11 
([15]). For | x | < 1 , we have
arctan x x n = n ! ∑ k = 0 ∞ ( − 1 ) k G ( n , k ) x 2 k , n ∈ N
and
arctan x x r = 1 + ∑ k = 1 ∞ ( − 1 ) k ∑ i = 1 2 k ( − r ) i ∑ j = 1 i ( − 1 ) j ( i − j ) ! G ( j , k ) x 2 k , r ∈ R ,
where
G ( n , k ) = ∑ ℓ = 0 2 k 2 ℓ 2 k + n − 1 2 k − ℓ s ( ℓ + n , n ) ( ℓ + n ) !
for n ∈ N and k ∈ N 0 .
Lemma 12. 
For x ∈ C , we have
sin x sinh x = ∑ m = 0 ∞ ( − 1 ) m 2 2 m + 1 ( 4 m + 2 ) ! x 4 m + 2 .
Proof. 
Since sin x = − i sinh ( i x ) , the product-to-sum formula gives
sin x sinh x = cosh ( 1 + i ) x − cosh ( 1 − i ) x 2 i = 1 2 i ∑ n = 0 ∞ ( 1 + i ) 2 n − ( 1 − i ) 2 n ( 2 n ) ! x 2 n .
Because ( 1 ± i ) 2 n = ( ± 2 i ) n , the numerator vanishes for even n, while for n = 2 m + 1 it equals 2 ( 2 i ) 2 m + 1 = ( − 1 ) m 2 2 m + 2 i . Dividing by 2 i gives (32). □
Lemma 13. 
Let
t m = ( − 1 ) m 2 · m ! ( 4 m + 2 ) ! , m ∈ N
be the Maclaurin coefficients of
T ( x ) = 1 + ∑ m = 1 ∞ t m x m m ! ,
so that T ( x ) = sin y sinh y y 2 with y = x 4 1 / 4 by Lemma 12. For k , n ∈ N with k ≥ n , the partial Bell polynomials B k , n satisfy
B k , n t 1 , t 2 , … , t k − n + 1 = k ! n ! 4 k ∑ r = 1 n ( − 1 ) n − r n r ( 4 i ) r ( 4 k + 2 r ) ! ∑ p , q = 0 r ( − 1 ) p + q r p r q ( r − 2 q ) + i ( r − 2 p ) 4 k + 2 r .
Proof. 
From the generating function of the partial Bell polynomials, we have
[ T ( x ) − 1 ] n n ! = ∑ k = n ∞ B k , n t 1 , t 2 , … , t k − n + 1 x k k ! .
Expanding [ T ( x ) − 1 ] n by the binomial theorem gives
B k , n t 1 , t 2 , … , t k − n + 1 = k ! n ! ∑ r = 0 n ( − 1 ) n − r n r [ x k ] [ T ( x ) ] r .
Since T ( x ) = sin y sinh y y 2 and y 4 k = 4 − k x k , it follows that
[ x k ] [ T ( x ) ] r = 4 − k [ y 4 k + 2 r ] ( sin y sinh y ) r .
In terms of exponential functions, we obtain
( sin y sinh y ) r = e i y − e − i y 2 i r e y − e − y 2 r = 1 ( 4 i ) r ∑ p , q = 0 r ( − 1 ) p + q r p r q e [ ( r − 2 q ) + i ( r − 2 p ) ] y ,
and the coefficient of y 4 k + 2 r is
1 ( 4 i ) r ( 4 k + 2 r ) ! ∑ p , q = 0 r ( − 1 ) p + q r p r q ( r − 2 q ) + i ( r − 2 p ) 4 k + 2 r .
For r = 0 and k ≥ 1 we have [ x k ] { 1 } = 0 . Substituting these expressions yields (33). □

4. Products of Sine and Hyperbolic Sine Functions and the Bernoulli Numbers B 4 k

In this section, we evaluate Y k and B n , k for three functions built from the sine function: 1 + x sin x in Theorem 14; the ratio Φ ( x ) Φ ( x / 4 ) of products of the sine and hyperbolic sine functions in Theorems 15 and 16 and Corollary 2, which lead to the Bernoulli numbers B 4 k ; and sin ( a x ) a sin x in Theorem 17. Throughout, Lemmas 12 and 13 are used.
Theorem 14. 
Let r = 0.9320 … be the modulus of the zeros ± i y 0 of the function 1 + z sin z nearest to the origin, where y 0 > 0 is the unique positive solution of y sinh y = 1 . For λ ∈ R and | x | < r , we have
( 1 + x sin x ) λ = ∑ m = 0 ∞ ∑ j = 0 m 〈 λ 〉 j j ! ( − 1 ) m − j 4 m − j ( 2 m − 2 j ) ! B 2 m − 2 j ( − j ) − j 2 x 2 m
and
ln ( 1 + x sin x ) = ∑ m = 1 ∞ ( − 1 ) m − 1 ∑ n = 1 m 4 m − n n 1 ( 2 m − 2 n ) ! B 2 m − 2 n ( − n ) − n 2 x 2 m .
For m ∈ N and λ ∈ R , the complete Bell polynomials Y k ( a 1 , a 2 , … , a k ) satisfy
Y 2 m 0 , 2 λ , 0 , − 16 λ , 0 , 366 λ , … , a 2 m = ( 2 m ) ! ∑ j = 0 m 〈 λ 〉 j j ! ( − 1 ) m − j 4 m − j ( 2 m − 2 j ) ! B 2 m − 2 j ( − j ) − j 2 ,
where the arguments with odd indices are zero, the argument with index 2 i is
a 2 i = λ ( − 1 ) i − 1 ( 2 i ) ! ∑ n = 1 i 4 i − n n 1 ( 2 i − 2 n ) ! B 2 i − 2 n ( − n ) − n 2 ,
and Y 2 m + 1 of the same sequence equals 0.
Let
w i = − i cos i π 2 , i ∈ N .
For m ∈ N and 1 ≤ k ≤ 2 m , the partial Bell polynomials B n , k satisfy
B 2 m , k ( w 1 , w 2 , … , w 2 m − k + 1 ) = ( − 1 ) k ( 2 m ) ! k ! ∑ q = 0 k ( − 1 ) q k q ∑ j = 0 min { q , m } q j ( − 1 ) m − j 4 m − j ( 2 m − 2 j ) ! B 2 m − 2 j ( − j ) − j 2
and B n , k ( w 1 , w 2 , … , w n − k + 1 ) = 0 for odd n.
Proof. 
For j ∈ N 0 , the formula (26) with r = j gives
sin x x j = ∑ i = 0 ∞ ( − 1 ) i B 2 i ( − j ) − j 2 ( 2 x ) 2 i ( 2 i ) ! ,
and hence
( x sin x ) j = x 2 j sin x x j = ∑ i = 0 ∞ ( − 1 ) i 4 i B 2 i ( − j ) − j 2 x 2 j + 2 i ( 2 i ) ! .
Only even powers of x occur in (92). Putting m = j + i , we see that the coefficient of x 2 m in ( x sin x ) j is
c j , m = ( − 1 ) m − j 4 m − j ( 2 m − 2 j ) ! B 2 m − 2 j ( − j ) − j 2
for 0 ≤ j ≤ m and is zero for j > m and for odd powers.
The function 1 + x sin x is entire. Since | sin x | ≤ sinh | x | , we have | x sin x | ≤ | x | sinh | x | < r sinh r = 1 for | x | < r , while ( i y 0 ) sin ( i y 0 ) = − y 0 sinh y 0 = − 1 . Hence 1 + x sin x has no zeros in the disc | x | < r and vanishes at x = ± i y 0 , which justifies the description of r in the theorem. Consequently, ( 1 + x sin x ) λ and ln ( 1 + x sin x ) are analytic there, and their Maclaurin series converge for | x | < r . It therefore suffices to compute the coefficients for | x | so small that | x sin x | < 1 . There, the binomial series (49) with u = x sin x and the expansions (92) yield, after collecting the coefficients of x 2 m , the series expansion (88). For λ = n ∈ N 0 the series terminates, since 〈 n 〉 j = 0 for j > n , and this is the finite expansion
( 1 + x sin x ) n = ∑ j = 0 n n j ( x sin x ) j .
Similarly, from
ln ( 1 + u ) = ∑ n = 1 ∞ ( − 1 ) n − 1 u n n
with u = x sin x , we obtain
ln ( 1 + x sin x ) = ∑ m = 1 ∞ ∑ n = 1 m ( − 1 ) n − 1 n c n , m x 2 m ,
which is (89), because ( − 1 ) n − 1 ( − 1 ) m − n = ( − 1 ) m − 1 .
Since both series are even in x, the arguments a i = i ! [ x i ] ( λ ln ( 1 + x sin x ) ) with odd i are zero, and the coefficients a 2 i are those displayed in the theorem. Applying Lemma 1 to (88) and (89), multiplied by λ , gives (90) and the vanishing of the odd orders.
For the partial Bell polynomials, let
f ( x ) = 1 + x sin x = 1 + ∑ i = 0 ∞ ( − 1 ) i x 2 i + 2 ( 2 i + 1 ) ! .
Then f ( 0 ) = 1 , f ( 2 i + 2 ) ( 0 ) = ( − 1 ) i ( 2 i + 2 ) , and the odd-order derivatives vanish, that is, f ( i ) ( 0 ) = w i for i ∈ N . For q ∈ N 0 , the finite expansion above with n = q shows that
[ f ( x ) ] q = ∑ m = 0 ∞ C q , 2 m x 2 m ( 2 m ) ! , C q , 2 m = ( 2 m ) ! ∑ j = 0 min { q , m } q j c j , m ,
and C q , 2 m + 1 = 0 . Substituting these data into Lemma 8 gives (91) and the vanishing for odd n. The proof of Theorem 14 is complete. □
Remark 13. 
For λ = 1 , only the terms with j ≤ 1 in (90) contribute, and we obtain
Y 2 m ( 0 , 2 , 0 , − 16 , … ) = ( 2 m ) ! [ x 2 m ] ( 1 + x sin x ) = 2 m ( − 1 ) m − 1 .
For example, Y 2 ( 0 , 2 λ ) = 2 λ and
Y 4 ( 0 , 2 λ , 0 , − 16 λ ) = 12 λ 2 − 16 λ .
Theorem 15. 
Let
Φ ( x ) = ∑ m = 0 ∞ ( − 1 ) m x m ( 4 m + 2 ) ! .
For k ∈ N and λ ∈ R , the complete Bell polynomials Y k ( a 1 , a 2 , … , a k ) satisfy
Y k 3 λ B 4 4 ! , 15 λ B 8 8 ! , … , λ 2 2 k − 1 ( k − 1 ) ! B 4 k ( 4 k ) ! = k ! [ x k ] Φ ( x ) Φ ( x / 4 ) 2 λ ,
For n , k ∈ N with n ≥ k , the partial Bell polynomials B n , k at the first n − k + 1 terms of the sequence in (93) with λ = 1 satisfy
B n , k a 1 , … , a n − k + 1 = n ! ∑ j = k n s ( j , k ) j ! ∑ q = 0 j ( − 1 ) j − q j q [ x n ] Φ ( x ) Φ ( x / 4 ) 2 q .
Proof. 
Using the series expansion
ln sin y y = − ∑ k = 1 ∞ | B 2 k | 2 k ( 2 y ) 2 k ( 2 k ) ! , | y | < π
in [14, p. 55, Entry 1.518] and the relation sinh y = sin ( i y ) i in [14, p. 28], we obtain
ln sinh y y = ∑ k = 1 ∞ ( − 1 ) k + 1 | B 2 k | 2 k ( 2 y ) 2 k ( 2 k ) ! , | y | < π .
Adding (94) and (95), the terms with odd k cancel and, since | B 4 m | = − B 4 m , we have
ln sin y sinh y y 2 = ∑ m = 1 ∞ B 4 m 2 m ( 2 y ) 4 m ( 4 m ) ! .
Put a = x 4 2 and b = x 4 2 . Then a 2 = 2 b 2 , ( 2 a ) 4 m = 4 m x m , and ( 2 b ) 4 m = x m . Consequently,
ln sin a sinh a 2 sin b sinh b = ln sin a sinh a a 2 − ln sin b sinh b b 2 = 1 2 ∑ m = 1 ∞ 2 2 m − 1 B 4 m ( 4 m ) ! x m m .
Multiplying by 2 λ gives
ln sin a sinh a 2 sin b sinh b 2 λ = ∑ k = 1 ∞ λ 2 2 k − 1 ( k − 1 ) ! B 4 k ( 4 k ) ! x k k ! .
On the other hand, Lemma 12 with y = a and y = b gives
sin a sinh a = x Φ ( x ) and sin b sinh b = x 2 Φ x 4 .
Therefore
sin a sinh a 2 sin b sinh b = Φ ( x ) Φ ( x / 4 ) .
Applying Lemma 1 yields (93).
For the partial Bell polynomials, we use (93) with λ = q , which gives
Y n ( q a 1 , … , q a n ) = n ! [ x n ] Φ ( x ) Φ ( x / 4 ) 2 q ,
and substitute it into (25) in Lemma 9. The coefficients [ x n ] [ Φ ( x ) / Φ ( x / 4 ) ] 2 q are given explicitly in Corollary 2 below. The proof of Theorem 15 is complete. □
Theorem 16. 
Let Φ ( x ) be as in Theorem 15, let g ( x ) = Φ ( x ) Φ ( x / 4 ) , and let
c i = 1 2 2 2 i − 1 ( i − 1 ) ! B 4 i ( 4 i ) ! , i ∈ N .
For μ ∈ R and | x | < 4 π 4 , we have
[ g ( x ) ] μ = 1 2 sin x 4 2 sin x 4 2 sinh x 4 2 sinh x 4 2 μ = ∑ k = 0 ∞ Y k μ c 1 , μ c 2 , … , μ c k x k k ! .
For k ∈ N and μ ∈ R , the complete Bell polynomials Y k satisfy
Y k μ c 1 , μ c 2 , … , μ c k = k ! [ x k ] [ g ( x ) ] μ = k ! ∑ ∑ i = 1 k i ℓ i = k ℓ 1 , … , ℓ k ∈ N 0 ∏ i = 1 k 1 ℓ i ! μ c i i ! ℓ i .
Let w i = g ( i ) ( 0 ) = Y i ( c 1 , c 2 , … , c i ) for i ∈ N . For n , k ∈ N with n ≥ k , the partial Bell polynomials B n , k satisfy
B n , k ( w 1 , w 2 , … , w n − k + 1 ) = ( − 1 ) k k ! ∑ q = 0 k ( − 1 ) q k q Y n q c 1 , q c 2 , … , q c n = ∑ j = k n S ( j , k ) B n , j ( c 1 , c 2 , … , c n − j + 1 ) ,
where S ( j , k ) denotes the Stirling numbers of the second kind generated by (9).
Proof. 
In the proof of Theorem 15, we showed that
ln g ( x ) = 1 2 ∑ k = 1 ∞ 2 2 k − 1 B 4 k ( 4 k ) ! x k k = ∑ k = 1 ∞ c k x k k ! ,
and that g ( x ) = 1 2 sin a sinh a sin b sinh b with a = x 4 2 and b = x 4 2 . The zeros of Φ ( x ) are x = 4 π 4 n 4 for n ∈ N , and those of Φ x 4 are x = 16 π 4 n 4 . Hence g is analytic and nonzero in the disc | x | < 4 π 4 , and g μ has a convergent Maclaurin series there. Since
ln g μ = μ ln g = ∑ k = 1 ∞ μ c k x k k ! ,
the generating function (3) yields (96). Lemma 1 then gives (97), whose last expression is the definition of Y k as a sum over partitions.
Taking μ = 1 in (96) gives g ( i ) ( 0 ) = Y i ( c 1 , … , c i ) = w i . For q ∈ N 0 , the series (96) with μ = q is the expansion [ g ( x ) ] q = ∑ n = 0 ∞ C q , n x n n ! with C q , n = Y n ( q c 1 , … , q c n ) for n ≥ 1 and C 0 , 0 = 1 , C 0 , n = 0 for n ≥ 1 . Substituting these data into Lemma 8 gives the first expression in (98). By the homogeneity
Y n ( q c 1 , … , q c n ) = ∑ j = 1 n q j B n , j ( c 1 , … , c n − j + 1 )
and the identity
∑ q = 0 k ( − 1 ) q k q q j = ( − 1 ) k k ! S ( j , k ) ,
we obtain the second expression. The proof of Theorem 16 is complete. □
Remark 14. 
We have
c 1 = − 1 480 , c 2 = − 1 161280 , c 3 = − 691 20756736000 ,
and
Y 1 ( μ c 1 ) = − μ 480 , Y 2 ( μ c 1 , μ c 2 ) = μ ( 7 μ − 10 ) 1612800 , Y 3 ( μ c 1 , μ c 2 , μ c 3 ) = − μ 3003 μ 2 − 12870 μ + 11056 332107776000 .
Moreover,
w 1 = − 1 480 , w 2 = − 1 537600 , w 3 = − 1189 332107776000 ,
and, for example,
B 2 , 2 = w 1 2 = 1 230400 , B 3 , 2 = 3 w 1 w 2 = 1 86016000 , B 3 , 3 = w 1 3 = − 1 110592000 .
If μ ∈ N , the radius of convergence of (96) is 16 π 4 .
Corollary 2. 
Let Φ ( x ) , g ( x ) = Φ ( x ) Φ ( x / 4 ) , and c i be as in Theorems 15 and 16, and let
T ( x ) = 2 Φ ( x ) = 1 + ∑ m = 1 ∞ t m x m m !
with t m as in Lemma 13. For r ∈ N and j ∈ N 0 , let
ρ j , r = 1 4 j ( 4 i ) r ( 4 j + 2 r ) ! ∑ p , q = 0 r ( − 1 ) p + q r p r q ( r − 2 q ) + i ( r − 2 p ) 4 j + 2 r ,
and, for μ ∈ R , let d 0 ( μ ) = 1 and
d k ( μ ) = 1 k ! ∑ n = 1 k 〈 μ 〉 n B k , n t 1 , t 2 , … , t k − n + 1 = ∑ n = 1 k μ n ∑ r = 1 n ( − 1 ) n − r n r ρ k , r , k ∈ N ,
where μ n = 〈 μ 〉 n n ! . Then, for μ ∈ R and | x | < 4 π 4 ,
[ g ( x ) ] μ = 1 2 sin x 4 2 sin x 4 2 sinh x 4 2 sinh x 4 2 μ = ∑ k = 0 ∞ ∑ i = 0 k 4 i − k d i ( μ ) d k − i ( − μ ) x k ,
and, for k ∈ N ,
Y k μ c 1 , μ c 2 , … , μ c k = k ! ∑ i = 0 k 4 i − k d i ( μ ) d k − i ( − μ ) .
In particular, taking μ = 2 λ , the complete Bell polynomials in (93) satisfy
Y k 3 λ B 4 4 ! , 15 λ B 8 8 ! , … , λ 2 2 k − 1 ( k − 1 ) ! B 4 k ( 4 k ) ! = k ! ∑ i = 0 k 4 i − k d i ( 2 λ ) d k − i ( − 2 λ ) , λ ∈ R ,
and, for k = 0 , 1 , 2 , the coefficient [ x k ] [ g ( x ) ] 2 λ equals 1, − λ 240 , and λ ( 7 λ − 5 ) 806400 , respectively.
Proof. 
Lemma 12 with y = a = x 4 2 , for which 4 a 4 = x and 2 a 2 = x , and y = b = x 4 2 , for which 4 b 4 = x 4 and 2 b 2 = x 2 , yields
sin a sinh a = x Φ ( x ) and sin b sinh b = x 2 Φ x 4 .
Therefore the middle expression in (100) equals
[ g ( x ) ] μ = [ T ( x ) ] μ T x 4 − μ
for | x | < 4 π 4 , with the branches equal to 1 at x = 0 ; here T ( x ) ≠ 0 for | x | < 4 π 4 , because T ( x ) = sin y sinh y y 2 with y = x 4 1 / 4 by Lemma 12.
Applying the Faà di Bruno formula (18) to f ( u ) = ( 1 + u ) μ and h ( x ) = T ( x ) − 1 , for which h ( 0 ) = 0 and h ( m ) ( 0 ) = t m , we obtain
d k d x k T ( x ) μ | x = 0 = ∑ n = 1 k 〈 μ 〉 n B k , n t 1 , … , t k − n + 1 = k ! d k ( μ )
for k ∈ N , that is, [ T ( x ) ] μ = ∑ k = 0 ∞ d k ( μ ) x k . Substituting the explicit formula (33) in Lemma 13 into the first expression in (99), and using 〈 μ 〉 n n ! = μ n , gives the second expression in (99). Replacing x by x 4 and μ by − μ gives
T x 4 − μ = ∑ k = 0 ∞ 4 − k d k ( − μ ) x k .
The Cauchy product of these two series yields (100). Comparing (100) with (96) gives (101), and taking μ = 2 λ , for which μ c i is the ith argument in (93), gives (102). The values for k = 0 , 1 , 2 follow from d 1 ( μ ) = − μ 360 and d 2 ( μ ) = μ ( 7 μ − 6 ) 1814400 . The proof of Corollary 2 is complete. □
Remark 15. 
We have d 1 ( μ ) = − μ 360 and d 2 ( μ ) = μ ( 7 μ − 6 ) 1814400 . Hence the coefficient of x in (100) is
d 1 ( μ ) + 1 4 d 1 ( − μ ) = − μ 480 ,
in agreement with Y 1 ( μ c 1 ) = − μ 480 in (101).
Theorem 17. 
Let a , b ∈ R with a ≠ 0 . For | x | < π max { 1 , | a | } , we have the series expansion
sin ( a x ) / ( a x ) sin x / x b = ∑ k = 0 ∞ ( − 1 ) k A k ( a , b ) x 2 k ( 2 k ) ! ,
where
A k ( a , b ) = 4 k ∑ j = 0 k 2 k 2 j a 2 j B 2 j ( − b ) − b 2 B 2 k − 2 j ( b ) b 2 .
For k ∈ N , the complete Bell polynomials Y k ( a 1 , a 2 , … , a k ) satisfy
Y 2 k 0 , 2 2 | B 2 | ( 1 − a 2 ) 2 b , 0 , 2 4 | B 4 | ( 1 − a 4 ) 4 b , … , 2 2 k | B 2 k | ( 1 − a 2 k ) 2 k b = ( − 1 ) k A k ( a , b ) ,
where the arguments with odd indices are zero, and Y 2 k + 1 of the same sequence equals 0. In particular, A 0 ( a , b ) = 1 ,
A 1 ( a , b ) = b a 2 − 1 3 , A 2 ( a , b ) = b a 2 − 1 ( 5 b − 2 ) a 2 − ( 5 b + 2 ) 15 .
Let
w i = − cos i π 2 ∑ j = 0 ⌊ i / 2 ⌋ i 2 j 2 i − 2 j − 2 B i − 2 j 2 j + 1 a 2 j , i ∈ N ,
that is, w i = 0 for odd i,
w 2 = 1 − a 2 3 , and w 4 = ( a 2 − 1 ) ( 3 a 2 − 7 ) 15 .
For m ∈ N and 1 ≤ k ≤ 2 m , the partial Bell polynomials B n , k satisfy
B 2 m , k ( w 1 , w 2 , … , w 2 m − k + 1 ) = ( − 1 ) m + k k ! ∑ q = 0 k ( − 1 ) q k q A m ( a , q ) ,
and B n , k ( w 1 , w 2 , … , w n − k + 1 ) = 0 for odd n.
Proof. 
Write
sin ( a x ) / ( a x ) sin x / x b = sin ( a x ) a x b x sin x b .
The function sin ( a x ) a sin x is analytic and nonzero for | x | < π max { 1 , | a | } , since its poles are at x ∈ π Z ∖ { 0 } and its zeros are at x ∈ π | a | Z ∖ { 0 } , apart from the removable singularity at x = 0 . Hence its real power and its logarithm have Maclaurin series that converge in this disc.
By the formula (26) with r = b and z = a x , and with r = − b and z = x , we have
sin ( a x ) a x b = ∑ j = 0 ∞ ( − 1 ) j B 2 j ( − b ) − b 2 ( 2 a x ) 2 j ( 2 j ) !
and
x sin x b = ∑ i = 0 ∞ ( − 1 ) i B 2 i ( b ) b 2 ( 2 x ) 2 i ( 2 i ) ! .
Multiplying these two series, collecting the coefficient of x 2 k with k = i + j , and using
1 ( 2 j ) ! ( 2 k − 2 j ) ! = 2 k 2 j 1 ( 2 k ) ! ,
we obtain (103) and (104), because
( − 1 ) j ( − 1 ) k − j = ( − 1 ) k and ( 2 a ) 2 j 2 2 k − 2 j = 4 k a 2 j .
For the logarithm, the series (94) with y = a x and y = x yields
ln sin ( a x ) a sin x = ln sin ( a x ) a x − ln sin x x = ∑ i = 1 ∞ | B 2 i | 2 i 2 2 i 1 − a 2 i x 2 i ( 2 i ) ! .
Therefore
ln sin ( a x ) a sin x b = ∑ i = 1 ∞ 2 2 i | B 2 i | 1 − a 2 i 2 i b x 2 i ( 2 i ) ! .
Since this series is even in x, the arguments of odd index are zero. Applying Lemma 1 to this series and to (103), whose coefficient of x 2 k ( 2 k ) ! is ( − 1 ) k A k ( a , b ) , gives (105), and the odd orders vanish because both series are even.
For the partial Bell polynomials, let
f ( x ) = sin ( a x ) a sin x = sin ( a x ) a x sin x x .
Using the classical series
sin ( a x ) a x = ∑ j = 0 ∞ ( − 1 ) j ( a x ) 2 j ( 2 j + 1 ) ! and x sin x = ∑ m = 0 ∞ ( − 1 ) m + 1 2 2 m − 2 B 2 m ( 2 m ) ! x 2 m ,
the latter being the classical Maclaurin series of the cosecant function, and multiplying them, we obtain f ( 0 ) = 1 , f ( i ) ( 0 ) = w i for i ∈ N , with w i as in the theorem. For q ∈ N 0 , the series (103) with b = q gives
[ f ( x ) ] q = ∑ n = 0 ∞ C q , n x n n ! , C q , 2 m = ( − 1 ) m A m ( a , q )
and C q , 2 m + 1 = 0 . Substituting these data into Lemma 8 gives (106) and the vanishing for odd n.
Finally, B 0 ( σ ) ( · ) = 1 and B 2 ( σ ) σ 2 = − σ 12 , which follow from the series
z sin z σ = 1 + σ 6 z 2 + …
and give the value of A 1 ( a , b ) . The expression for A 2 ( a , b ) follows in the same way from the term of order x 4 . The proof of Theorem 17 is complete. □
Remark 16. 
If a = 1 , then the left-hand side of (103) is identically 1, so A k ( 1 , b ) = 0 for k ∈ N and all the arguments in (105) vanish. The limit a → 0 recovers the expansion of x sin x b .
Remark 17. 
In [18], we find the series expansions
∑ k = 0 ∞ a 2 k ( 2 k + 1 ) ! B 2 k + 1 ( x ) = 1 2 csch a 2 sinh a x − 1 2 , | a | < 2 π ,
and
∑ k = 0 ∞ ( − 1 ) k a 2 k ( 2 k + 1 ) ! B 2 k + 1 ( x ) = 1 2 csc a 2 sin a x − 1 2 , | a | < 2 π ,
where B n ( x ) denotes the Bernoulli polynomials.

5. Exponentials of Elementary Functions and the Catalan Numbers

In this section, we expand
e cos x − 1 , e x sin x , e arctan x , e x cos x , e x sec x , e tan x
in power series, and we expand real powers of the generating function of the Catalan numbers; see [33,44] for properties of these numbers.
Theorem 18. 
Let ε 2 m and τ 2 m + 1 denote the secant and tangent numbers, defined by
sec x = ∑ m = 0 ∞ ε 2 m x 2 m ( 2 m ) ! and tan x = ∑ m = 0 ∞ τ 2 m + 1 x 2 m + 1 ( 2 m + 1 ) ! .
The following series expansions are valid.
1.
For x ∈ C ,
e cos x − 1 = ∑ n = 0 ∞ ( − 1 ) n ∑ j = 0 n 1 ( 2 j ) ! ! ( 2 n − 2 j ) ! B 2 n − 2 j ( − 2 j ) ( − j ) x 2 n .
2.
For x ∈ C ,
e x sin x = ∑ n = 0 ∞ ∑ j = 0 n ( − 1 ) n − j 4 n − j j ! ( 2 n − 2 j ) ! B 2 n − 2 j ( − j ) − j 2 x 2 n .
3.
For | x | < 1 ,
e arctan x = 1 + ∑ n = 1 ∞ ∑ j + 2 k = n , j ≥ 1 ( − 1 ) k G ( j , k ) x n ,
where G ( j , k ) is defined by (31).
4.
For x ∈ C ,
e x cos x = ∑ n = 0 ∞ ∑ m = 0 ⌊ n / 2 ⌋ ( − 1 ) m 4 m ( n − 2 m ) ! ( 2 m ) ! E 2 m ( 2 m − n ) 2 m − n 2 x n .
5.
For | x | < π 2 ,
e x sec x = ∑ n = 0 ∞ ∑ m = 0 ⌊ n / 2 ⌋ ( − 1 ) m 4 m ( n − 2 m ) ! ( 2 m ) ! E 2 m ( n − 2 m ) n − 2 m 2 x n .
6.
For | x | < π 2 ,
e tan x = ∑ n = 0 ∞ [ ∑ m = 0 ⌊ n / 2 ⌋ ( − 1 ) m 4 m ( n − 2 m ) ! ( 2 m ) ! × ∑ i = 0 m 2 m 2 i B 2 i ( 2 m − n ) 2 m − n 2 E 2 m − 2 i ( n − 2 m ) n − 2 m 2 ] x n .
Consequently, denoting by c n the coefficient of x n in the corresponding series, for k ∈ N the complete Bell polynomials Y k ( b 1 , b 2 , … , b k ) satisfy
Y k ( b 1 , b 2 , … , b k ) = k ! c k ,
where
1.
b i = cos i π 2 , that is, ( b i ) = ( 0 , − 1 , 0 , 1 , 0 , − 1 , … ) , for e cos x − 1 ;
2.
b i = − i cos i π 2 , that is, ( 0 , 2 , 0 , − 4 , 0 , 6 , … ) , for e x sin x ;
3.
b i = sin i π 2 ( i − 1 ) ! , that is, ( 1 , 0 , − 2 , 0 , 24 , 0 , − 720 , … ) , for e arctan x ;
4.
b i = sin i π 2 i , that is, ( 1 , 0 , − 3 , 0 , 5 , 0 , − 7 , … ) , for e x cos x ;
5.
b i = 1 − ( − 1 ) i 2 i ε i − 1 , that is, ( 1 , 0 , 3 , 0 , 25 , 0 , 427 , … ) , for e x sec x ;
6.
b i = 1 − ( − 1 ) i 2 τ i , that is, ( 1 , 0 , 2 , 0 , 16 , 0 , 272 , … ) , for e tan x .
For n , k ∈ N with n ≥ k , the partial Bell polynomials B n , k ( b 1 , b 2 , … , b n − k + 1 ) satisfy
B n , k ( b 1 , b 2 , … , b n − k + 1 ) = n ! k ! [ x n ] { f ( x ) k } ,
where f is the function in the corresponding part. Explicitly,
1.
B n , k = cos n π 2 k ! ∑ j = 0 k ( − 1 ) k − j 2 j k j ∑ i = 0 j j i ( j − 2 i ) n ,
for f ( x ) = cos x − 1 ;
2.
B n , k = n ! k ! ( − 1 ) m 2 2 m ( 2 m ) ! B 2 m ( − k ) − k 2
if n = 2 k + 2 m with m ∈ N 0 , and B n , k = 0 otherwise, for f ( x ) = x sin x ;
3.
B n , k = n ! ( − 1 ) m G ( k , m )
if n = k + 2 m with m ∈ N 0 , and B n , k = 0 otherwise, for f ( x ) = arctan x ;
4.
B n , k = n ! k ! ( − 1 ) m 2 2 m ( 2 m ) ! E 2 m ( − k ) − k 2
if n = k + 2 m with m ∈ N 0 , and B n , k = 0 otherwise, for f ( x ) = x cos x ;
5.
B n , k = n ! k ! ( − 1 ) m 2 2 m ( 2 m ) ! E 2 m ( k ) k 2
if n = k + 2 m with m ∈ N 0 , and B n , k = 0 otherwise, for f ( x ) = x sec x ;
6.
B n , k = n ! k ! ( − 1 ) m 2 2 m ( 2 m ) ! ∑ i = 0 m 2 m 2 i B 2 i ( − k ) − k 2 E 2 m − 2 i ( k ) k 2
if n = k + 2 m with m ∈ N 0 , and B n , k = 0 otherwise, for f ( x ) = tan x .
Proof. 
Let f be one of the six functions
cos x − 1 , x sin x , arctan x , x cos x , x sec x , tan x .
Each is analytic with f ( 0 ) = 0 in the stated domain, so
e f ( x ) = ∑ j = 0 ∞ [ f ( x ) ] j j ! ,
and the Maclaurin series of [ f ( x ) ] j may be summed term by term. We compute [ f ( x ) ] j for j ∈ N 0 and then collect the coefficient of x n .
(a) Since cos x − 1 = − 2 sin 2 x 2 , we have
( cos x − 1 ) j = ( − 2 ) j x 2 2 j sin ( x / 2 ) x / 2 2 j .
By the formula (26) with r = 2 j and z = x 2 ,
( cos x − 1 ) j = − 1 2 j x 2 j ∑ m = 0 ∞ ( − 1 ) m B 2 m ( − 2 j ) ( − j ) x 2 m ( 2 m ) ! .
Dividing by j ! , summing over j, and putting n = j + m gives (a).
(b) Since
( x sin x ) j = x 2 j sin x x j ,
the formula (26) with r = j and z = x yields (92), and (b) follows by putting n = j + i .
(c) We have
( arctan x ) j = x j arctan x x j = j ! ∑ k = 0 ∞ ( − 1 ) k G ( j , k ) x j + 2 k
for j ∈ N by the formula (29). Dividing by j ! and putting n = j + 2 k gives (c).
(d) and (e) Since ( x cos x ) j = x j cos j x and ( x sec x ) j = x j ( cos x ) − j , the formula (27) with λ = j and with λ = − j gives
cos j x = ∑ m = 0 ∞ ( − 1 ) m E 2 m ( − j ) − j 2 ( 2 x ) 2 m ( 2 m ) !
and
( cos x ) − j = ∑ m = 0 ∞ ( − 1 ) m E 2 m ( j ) j 2 ( 2 x ) 2 m ( 2 m ) ! .
Dividing by j ! and putting n = j + 2 m gives (d) and (e).
(f) Since tan j x = x j tan x x j , the expansion of tan z z λ in the proof of Theorem 1 with λ = j and z = x reads
tan x x j = ∑ m = 0 ∞ ( − 1 ) m ∑ i = 0 m 2 m 2 i B 2 i ( − j ) − j 2 E 2 m − 2 i ( j ) j 2 ( 2 x ) 2 m ( 2 m ) ! .
Dividing by j ! and putting n = j + 2 m gives (f).
For the Bell polynomials, write
e f ( x ) = ∑ k = 0 ∞ c k x k and f ( x ) = ∑ i = 1 ∞ b i x i i ! ,
where b i = f ( i ) ( 0 ) . By Lemma 2, Y k ( b 1 , … , b k ) = k ! c k . The values of b i follow from the Maclaurin series
cos x − 1 = ∑ m = 1 ∞ ( − 1 ) m x 2 m ( 2 m ) ! , x sin x = ∑ m = 0 ∞ ( − 1 ) m x 2 m + 2 ( 2 m + 1 ) ! , arctan x = ∑ m = 0 ∞ ( − 1 ) m x 2 m + 1 2 m + 1 , x cos x = ∑ m = 0 ∞ ( − 1 ) m x 2 m + 1 ( 2 m ) ! , x sec x = ∑ m = 0 ∞ ε 2 m x 2 m + 1 ( 2 m ) ! , tan x = ∑ m = 0 ∞ τ 2 m + 1 x 2 m + 1 ( 2 m + 1 ) ! .
For the partial Bell polynomials, the definition of B n , k gives
B n , k ( b 1 , b 2 , … , b n − k + 1 ) = n ! k ! [ x n ] f ( x ) k .
For the last five functions, the coefficients of f ( x ) j with j = k were computed above, for example, ( x sin x ) k = x 2 k sin x x k , and we use them with n ! k ! in place of the division by j ! . For the first function, we write
( cos x − 1 ) k = ∑ j = 0 k ( − 1 ) k − j k j cos j x
and
cos j x = 2 − j ∑ i = 0 j j i cos ( ( j − 2 i ) x ) .
The proof of Theorem 18 is complete. □
Remark 18. 
The first terms of these expansions are
e cos x − 1 = 1 − x 2 2 + x 4 6 − 31 x 6 720 + … , e x sin x = 1 + x 2 + x 4 3 + x 6 120 + … , e arctan x = 1 + x + x 2 2 − x 3 6 − 7 x 4 24 + x 5 24 + … , e x cos x = 1 + x + x 2 2 − x 3 3 − 11 x 4 24 − x 5 5 + … , e x sec x = 1 + x + x 2 2 + 2 x 3 3 + 13 x 4 24 + 7 x 5 15 + … , e tan x = 1 + x + x 2 2 + x 3 2 + 3 x 4 8 + 37 x 5 120 + … .
Remark 19. 
The first two expansions in Theorem 18 agree with [18, (5.20.4) and (5.20.9), pp. 81–82], and the other four agree with [18, (5.24.6)–(5.24.9), p. 90]. Hansen gives the general terms for
e cos x − 1 , e x cos x , e x sec x , e tan x
as finite sums of powers and factorials, only the recurrence
( k + 2 ) c k + 2 − c k + 1 + k c k = 0
with c 0 = c 1 = 1 for the coefficients c k of e arctan x , and only the first terms of e x sin x . For e x cos x , Hansen’s general term agrees term by term with that in Theorem 18, by the identity (46). Theorem 18 expresses all six series through generalized Bernoulli and Euler polynomials or the numbers G ( k , m ) , states their domains of validity, and gives the corresponding values of Y k and B n , k , which are not considered in [18].
Theorem 19. 
Let F ( x ) = 1 + 1 − 4 x 2 and let C n = 1 n + 1 2 n n be the Catalan numbers. For λ ∈ R and | x | < 1 4 , we have
1 + 1 − 4 x 2 λ = 1 − λ ∑ n = 1 ∞ 1 n 2 n − λ − 1 n − 1 x n
and
ln 1 + 1 − 4 x 2 = − ∑ n = 1 ∞ ( 2 n − 1 ) ! ( n ! ) 2 x n ,
where a m = 〈 a 〉 m m ! for a ∈ R and m ∈ N 0 .
For k ∈ N and λ ∈ R , the complete Bell polynomials Y k ( a 1 , a 2 , … , a k ) satisfy
Y k − λ 1 ! 1 ! , − λ 3 ! 2 ! , − λ 5 ! 3 ! , … , − λ ( 2 k − 1 ) ! k ! = − λ ( k − 1 ) ! 2 k − λ − 1 k − 1 .
Let w i = − i ! C i − 1 for i ∈ N , that is, ( w i ) i ≥ 1 = ( − 1 , − 2 , − 12 , − 120 , … ) . For n , k ∈ N with n ≥ k , the partial Bell polynomials B n , k satisfy
B n , k ( w 1 , w 2 , … , w n − k + 1 ) = ( − 1 ) k + 1 ( n − 1 ) ! ( k − 1 ) ! ∑ q = 1 k ( − 1 ) q k − 1 q − 1 2 n − q − 1 n − 1 .
Proof. 
Let
y = y ( x ) = 1 − 1 − 4 x 2 .
Then F ( x ) = 1 − y , y ( 0 ) = 0 , and y satisfies y = x φ ( y ) with φ ( y ) = 1 1 − y , because y ( 1 − y ) = x . Since y is analytic for | x | < 1 4 , the Lagrange inversion formula [11] states that, for an analytic function H and n ∈ N ,
[ x n ] H ( y ( x ) ) = 1 n [ y n − 1 ] H ′ ( y ) φ ( y ) n .
Taking H ( y ) = ( 1 − y ) λ gives
[ x n ] F ( x ) λ = − λ n [ y n − 1 ] ( 1 − y ) λ − n − 1 = − λ n ( − 1 ) n − 1 λ − n − 1 n − 1 = − λ n 2 n − λ − 1 n − 1 ,
where we used
( − 1 ) m − a m = a + m − 1 m .
This proves (107).
Differentiating (107) with respect to λ at λ = 0 yields
ln F ( x ) = − ∑ n = 1 ∞ 1 n 2 n − 1 n − 1 x n = − ∑ n = 1 ∞ ( 2 n − 1 ) ! ( n ! ) 2 x n ,
which is (108). Hence
ln F ( x ) λ = − λ ∑ n = 1 ∞ ( 2 n − 1 ) ! n ! x n n ! .
Applying Lemma 1 to this series and to (107), whose coefficient of x k k ! is − λ ( k − 1 ) ! 2 k − λ − 1 k − 1 , gives (109).
Since
F ( x ) = 1 − y ( x ) = 1 − ∑ i = 1 ∞ C i − 1 x i ,
we have F ( 0 ) = 1 and F ( i ) ( 0 ) = − i ! C i − 1 = w i . For q ∈ N 0 , the series (107) with λ = q shows that
[ F ( x ) ] q = ∑ n = 0 ∞ C q , n x n n !
with
C q , n = − q ( n − 1 ) ! 2 n − q − 1 n − 1 , n ≥ 1 .
Substituting these data into Lemma 8 and using q k q = k k − 1 q − 1 yields (110). The proof of Theorem 19 is complete. □
Remark 20. 
The first terms of (107) are
1 − λ x + λ ( λ − 3 ) 2 x 2 − λ ( λ − 4 ) ( λ − 5 ) 6 x 3 + … .
The series (108) is equivalent to the expansion of − ln 1 + 1 − 4 x 2 in [18]. For λ = 1 , we have F ( x ) = 1 − x C ( x ) , where C ( x ) = 1 − 1 − 4 x 2 x is the generating function of the Catalan numbers; see [23,44].
Remark 21. 
Since
1 + 1 − 4 x 1 − 1 − 4 x = 4 x ,
we have
1 F ( x ) = 2 1 + 1 − 4 x = 1 − 1 − 4 x 2 x = C ( x ) .
Hence, replacing λ by − λ in (107) and using 1 n 2 n + λ − 1 n − 1 = 1 2 n + λ 2 n + λ n , we recover the classical expansion
[ C ( x ) ] λ = 1 + ∑ n = 1 ∞ λ 2 n + λ 2 n + λ n x n , | x | < 1 4 ,
which reduces to C ( x ) = ∑ n = 0 ∞ C n x n for λ = 1 . Correspondingly, replacing λ by − λ in (109) gives
Y k λ 1 ! 1 ! , λ 3 ! 2 ! , λ 5 ! 3 ! , … , λ ( 2 k − 1 ) ! k ! = λ ( k − 1 ) ! 2 k + λ − 1 k − 1 , k ∈ N ,
and, in particular,
Y k 1 ! 1 ! , 3 ! 2 ! , … , ( 2 k − 1 ) ! k ! = k ! C k .

6. Applications

In this section, we apply the results of Section 4 to the Bernoulli numbers B 4 k and to multiple zeta values. SubSection 6.1 evaluates Y k at the sequence
a m = ( m − 1 ) ! 2 B 4 m ( 4 m ) ! , m ∈ N ,
and derives a convolution identity for B 4 k . SubSection 6.2 deduces closed forms of the multiple zeta values ζ ( 4 , 4 , … , 4 ) and ζ * ( 4 , 4 , … , 4 ) , and SubSection 6.3 studies their monotonicity with respect to the number of arguments.

6.1. Evaluation at the Bernoulli Numbers B 4 k and a Convolution Identity

In this subsection, we specialize the results of Section 4 to the sequence (111), which consists of the arguments in (6) for n = 2 . In Theorem 20, we evaluate Y k at this sequence in closed form. Combining this evaluation with the recurrence satisfied by the complete Bell polynomials, we derive in Theorem 21 a convolution identity for the Bernoulli numbers B 4 k .
Theorem 20. 
For k ∈ N , the complete Bell polynomials Y k satisfy
Y k ( a 1 , a 2 , … , a k ) = ( − 1 ) k k ! 2 1 − 2 k ( 4 k + 2 ) ! .
Proof. 
In the proof of Theorem 15, we have
ln sin y sinh y y 2 = ∑ m = 1 ∞ B 4 m 2 m ( 2 y ) 4 m ( 4 m ) ! .
Put z = ( 2 y ) 4 . Since B 4 m 2 m = ( m − 1 ) ! 2 B 4 m m ! , the right-hand side equals ∑ m = 1 ∞ a m z m m ! . By Lemma 12, the function itself is
f ( z ) = sin y sinh y y 2 = 2 Φ z 4 = ∑ m = 0 ∞ ( − 1 ) m 2 1 − 2 m ( 4 m + 2 ) ! z m .
Lemma 1 gives (112). □
Remark 22 
([30, Theorem 3.2]). For k ∈ N and ϵ ∈ R , we have
Y k ϵ 0 ! B 4 4 ! , ϵ 1 ! B 8 8 ! , … , ϵ ( k − 1 ) ! B 4 k ( 4 k ) ! = k ! ∑ j = 0 2 k ( − 1 ) j B 2 j ( − 2 ϵ ) ( − ϵ ) ( 2 j ) ! B 2 ( 2 k − j ) ( − 2 ϵ ) ( − ϵ ) [ 2 ( 2 k − j ) ] ! .
For ϵ = 1 2 , the left-hand side is (112), and B 2 j ( − 1 ) ( − 1 / 2 ) ( 2 j ) ! = 2 − 2 j ( 2 j + 1 ) ! . Hence (113) is consistent with (112), and the two together are equivalent to the elementary identity
∑ j = 0 2 k ( − 1 ) j 4 k + 2 2 j + 1 = ( − 1 ) k 2 2 k + 1 .
Theorem 21. 
For k ∈ N ,
∑ i = 1 k ( − 1 ) i 4 i B 4 i ( 4 i ) ! ( 4 k − 4 i + 2 ) ! = 2 k ( 4 k + 2 ) ! .
Proof. 
Differentiating
f ( z ) = exp ∑ m = 1 ∞ a m z m m !
gives
f ′ = f ∑ m = 0 ∞ a m + 1 z m m ! ,
and comparing coefficients yields the recurrence
Y k ( a 1 , … , a k ) = ∑ j = 0 k − 1 k − 1 j a j + 1 Y k − 1 − j ( a 1 , … , a k − 1 − j ) , Y 0 = 1 .
Inserting (111) and (112) and using k − 1 j j ! ( k − 1 − j ) ! = ( k − 1 ) ! , we obtain
( − 1 ) k k ! 2 1 − 2 k ( 4 k + 2 ) ! = ( k − 1 ) ! ∑ i = 1 k ( − 1 ) k − i 2 2 i − 2 k B 4 i ( 4 i ) ! ( 4 k − 4 i + 2 ) ! .
Multiplying by ( − 1 ) k 2 2 k ( k − 1 ) ! gives (114). □
Remark 23. 
Every evaluation Y k ( a 1 , a 2 , … , a k ) = k ! c k in Section 3, Section 4 and Section 5 yields, in the same way, a recurrence between the arguments a i and the coefficients c k . This gives convolution identities for the Bernoulli numbers, Stirling numbers, harmonic numbers, and Euler and Bernoulli polynomials that occur in those evaluations. For example, Theorems 4 and 5 yield the recurrences
η ( 1 − k ) = 1 2 − 1 2 ∑ j = 0 k − 2 k − 1 j η ( − j ) and ∑ j = 0 k − 1 k − 1 j ζ ( − j ) k − j = − 1 k + 1
for k ∈ N , the second of which is equivalent to a classical recurrence for the Bernoulli numbers, because ζ ( − j ) = − B j + 1 j + 1 for j ∈ N .

6.2. Multiple Zeta Values with All Arguments Equal to Four

In this subsection, we substitute the evaluation (112) into the formulas (6) and (7) with n = 2 . This gives, in Theorem 22, a closed form of ζ ( 4 , 4 , … , 4 ) and an explicit formula for ζ * ( 4 , 4 , … , 4 ) in terms of the partial Bell polynomials evaluated in Lemma 13. We then compare these results with known formulas and list some values.
Theorem 22. 
For k ∈ N , the multiple zeta values satisfy
ζ ( 4 , 4 , … , 4 ︸ k ) = 2 2 k + 1 π 4 k ( 4 k + 2 ) !
and
ζ * ( 4 , 4 , … , 4 ︸ k ) = 4 π 4 k k ! ∑ n = 1 k ( − 1 ) n n ! B k , n t 1 , t 2 , … , t k − n + 1 ,
where t m = ( − 1 ) m 2 · m ! ( 4 m + 2 ) ! , and the partial Bell polynomials B k , n ( t 1 , t 2 , … , t k − n + 1 ) are given explicitly in Lemma 13.
Proof. 
The formulas (6) and (7) with n = 2 read
ζ ( 4 , 4 , … , 4 ︸ k ) = ( − 1 ) k ( 2 π ) 4 k k ! Y k ( a 1 , a 2 , … , a k ) , ζ * ( 4 , 4 , … , 4 ︸ k ) = ( 2 π ) 4 k k ! Y k ( − a 1 , − a 2 , … , − a k ) .
Substituting (112) into the first one yields (115). For the second one, Lemma 2 applied to
exp − ∑ m = 1 ∞ a m z m m ! = 1 f ( z )
gives
Y k ( − a 1 , − a 2 , … , − a k ) = k ! [ z k ] 1 2 Φ ( z / 4 ) .
Hence
ζ * ( 4 , 4 , … , 4 ︸ k ) = 4 π 4 k [ u k ] 1 2 Φ ( u ) .
Now
2 Φ ( u ) = T ( u ) = 1 + ∑ m ≥ 1 t m u m m ! ,
and the generating function of B k , n gives
1 T ( u ) = ∑ n = 0 ∞ ( − 1 ) n [ T ( u ) − 1 ] n = ∑ n = 0 ∞ ( − 1 ) n n ! ∑ k = n ∞ B k , n ( t 1 , t 2 , … , t k − n + 1 ) u k k ! .
Comparing coefficients gives (116). □
Remark 24 
([30, Theorem 3.2]). For k ∈ N , the multiple zeta values also satisfy
ζ ( 4 , 4 , … , 4 ︸ k ) = ( − 1 ) 3 k ( 2 π ) 4 k ∑ j = 0 2 k ( − 1 ) j B 2 j ( − 1 ) − 1 2 ( 2 j ) ! B 2 ( 2 k − j ) ( − 1 ) − 1 2 [ 2 ( 2 k − j ) ] !
and
ζ * ( 4 , 4 , … , 4 ︸ k ) = ( 2 π ) 4 k ∑ j = 0 2 k ( − 1 ) j B 2 j 1 2 ( 2 j ) ! B 2 ( 2 k − j ) 1 2 [ 2 ( 2 k − j ) ] ! .
The formula (117) is equivalent to (115), while (118) gives another expression for ζ * ( 4 , 4 , … , 4 ) in terms of Bernoulli polynomials.
Remark 25. 
The value (115) is known, and it is recovered here from the Bernoulli-number form (6) through Lemma 12. For example,
ζ ( 4 ) = ζ * ( 4 ) = π 4 90 , ζ ( 4 , 4 ) = π 8 113400 , ζ * ( 4 , 4 ) = 13 π 8 113400 ,
and
ζ * ( 4 , 4 , 4 ) = 4009 π 12 3405402000 .

6.3. Monotonicity of ζ ( 4 , 4 , … , 4 ) and ζ * ( 4 , 4 , … , 4 )

In this subsection, we study how the multiple zeta values in Theorem 22 behave as the number of arguments grows. We show that ζ ( 4 , 4 , … , 4 ) is strictly logarithmically concave and eventually strictly decreasing, whereas ζ * ( 4 , 4 , … , 4 ) is strictly increasing and strictly concave with an explicit limit. We also express ζ * ( 4 , 4 , … , 4 ) as a rapidly convergent series, which leads to the Ramanujan-type evaluations in Corollary 3.
For k ∈ N , let
Z k = ζ ( 4 , 4 , … , 4 ︸ k ) and Z k * = ζ * ( 4 , 4 , … , 4 ︸ k ) ,
and let Z 0 = Z 0 * = 1 . With this convention, the formulas (115) and (116) also hold for k = 0 .
Theorem 23. 
The sequences ( Z k ) k ≥ 0 and ( Z k * ) k ≥ 0 have the following properties.
1.
For k ∈ N 0 , we have
Z k + 1 Z k = 4 π 4 ( 4 k + 3 ) ( 4 k + 4 ) ( 4 k + 5 ) ( 4 k + 6 ) ,
which is strictly decreasing in k. Consequently, the sequence ( Z k ) k ≥ 0 is strictly logarithmically concave, Z 1 > Z 0 , the sequence ( Z k ) k ≥ 1 is strictly decreasing, and Z k → 0 as k → ∞ .
2.
The sequence ( Z k * ) k ≥ 0 is strictly increasing and strictly concave. More precisely, the increments δ k = Z k * − Z k − 1 * for k ∈ N are positive and satisfy
1 16 < δ k + 1 δ k < ζ ( 4 ) − 1 = π 4 90 − 1 , k ∈ N .
3.
For k ∈ N 0 , we have
Z k * = 4 π ∑ m = 1 ∞ ( − 1 ) m − 1 m 4 k − 1 sinh ( π m ) .
4.
The limit of Z k * is
lim k → ∞ Z k * = ∏ n = 2 ∞ 1 − 1 n 4 − 1 = 4 π sinh π = 1.088 … ,
and, for k ∈ N ,
4 π sinh π − 8 π 16 k sinh ( 2 π ) < Z k * < 4 π sinh π − 8 π 16 k sinh ( 2 π ) + 12 π 81 k sinh ( 3 π ) .
In particular,
4 π sinh π − Z k * ∼ 8 π 16 k sinh ( 2 π )
as k → ∞ .
Proof. (a) By (115), we have
Z k + 1 Z k = 2 2 k + 3 π 4 k + 4 ( 4 k + 6 ) ! ( 4 k + 2 ) ! 2 2 k + 1 π 4 k = 4 π 4 ( 4 k + 3 ) ( 4 k + 4 ) ( 4 k + 5 ) ( 4 k + 6 ) ,
which is (119). Its denominator strictly increases with k, so the ratio strictly decreases; equivalently, Z k − 1 Z k + 1 < Z k 2 for k ∈ N . Moreover, Z 1 Z 0 = π 4 90 > 1 , while
Z k + 1 Z k ≤ Z 2 Z 1 = π 4 1260 < 1
for k ∈ N . Hence ( Z k ) k ≥ 1 is strictly decreasing, and Z k → 0 because the ratio (119) tends to 0.
(b) Let k ∈ N . Splitting the series defining Z k * according to whether n k = 1 or n k ≥ 2 , we obtain Z k * = Z k − 1 * + δ k with
δ k = ∑ n 1 ≥ n 2 ≥ … ≥ n k ≥ 2 1 n 1 4 n 2 4 … n k 4 > 0 ;
for k = 1 this reads ζ ( 4 ) = 1 + ∑ n = 2 ∞ 1 n 4 . Hence ( Z k * ) k ≥ 0 is strictly increasing. Summing first over n k + 1 , we have
δ k + 1 = ∑ n 1 ≥ n 2 ≥ … ≥ n k ≥ 2 1 n 1 4 n 2 4 … n k 4 ∑ m = 2 n k 1 m 4 .
The inner sum satisfies
1 16 ≤ ∑ m = 2 n k 1 m 4 < ∑ m = 2 ∞ 1 m 4 = ζ ( 4 ) − 1 ,
and the left inequality is strict whenever n k ≥ 3 . Since terms with n k ≥ 3 occur, we obtain (120). Because ζ ( 4 ) − 1 < 1 , the increments δ k strictly decrease, that is, Z k + 1 * − Z k * < Z k * − Z k − 1 * for k ∈ N , and ( Z k * ) k ≥ 0 is strictly concave.
(c) Let
P ( t ) = ∏ n = 1 ∞ 1 − t n 4 .
The product converges absolutely and locally uniformly, so P is an entire function of t whose zeros are the simple zeros t = m 4 for m ∈ N . For | t | < 1 , expanding every factor 1 − t n 4 − 1 into a geometric series and multiplying out, which is legitimate because all terms are nonnegative for 0 ≤ t < 1 , we obtain
1 P ( t ) = ∑ k = 0 ∞ Z k * t k , | t | < 1 ,
since every tuple n 1 ≥ n 2 ≥ … ≥ n k ≥ 1 corresponds to exactly one choice of exponents e n ∈ N 0 with ∑ n e n = k , namely the multiplicities of the values n in the tuple. By the product formulas
sin ( π x ) = π x ∏ n = 1 ∞ 1 − x 2 n 2 and sinh ( π x ) = π x ∏ n = 1 ∞ 1 + x 2 n 2 ,
we have
P x 4 = sin ( π x ) sinh ( π x ) π 2 x 2 , x ∈ C .
By Lemma 12, we have P ( t ) = T 4 π 4 t ; thus (123) is consistent with the formula (116).
Let m ∈ N . As x → m , we have
sin ( π x ) = ( − 1 ) m π ( x − m ) + O ( x − m ) 2
and
1 − x 4 m 4 = − 4 m ( x − m ) + O ( x − m ) 2 .
Hence, by (124),
∏ n = 1 n ≠ m ∞ 1 − m 4 n 4 = lim t → m 4 P ( t ) 1 − t / m 4 = ( − 1 ) m − 1 sinh ( π m ) 4 π m ,
so that
P ′ ( m 4 ) = − 1 m 4 ( − 1 ) m − 1 sinh ( π m ) 4 π m .
Therefore, for k ∈ N 0 , the residue of 1 t k + 1 P ( t ) at the simple pole t = m 4 equals
1 m 4 k + 4 P ′ ( m 4 ) = − 4 π ( − 1 ) m − 1 m 1 − 4 k sinh ( π m ) .
For N ∈ N , let ρ = N + 1 2 and let C N be the circle | t | = ρ 4 . For t ∈ C N , choose x with x 4 = t and | arg x | ≤ π 4 , and write x = u + i v . Then | x | = ρ and | u | ≥ ρ 2 , so
| sinh ( π x ) | ≥ sinh ( π | u | ) ≥ sinh π ρ 2 ,
because
| sinh ( π x ) | 2 = sinh 2 ( π u ) + sin 2 ( π v ) .
Moreover,
| sin ( π x ) | 2 = sin 2 ( π u ) + sinh 2 ( π v ) .
If | v | ≥ 1 4 , then
| sin ( π x ) | ≥ sinh π 4 > 4 5 .
If | v | < 1 4 , then
0 ≤ ρ − | u | = v 2 ρ + | u | < 1 16 ρ ≤ 1 24 ,
so the distance from | u | to the nearest integer exceeds 1 2 − 1 24 , and
| sin ( π x ) | ≥ | sin ( π u ) | ≥ cos π 24 > 4 5 .
By (124), we conclude that
| P ( t ) | ≥ 4 sinh ( π ρ / 2 ) 5 π 2 ρ 2 , t ∈ C N ,
and hence
1 2 π i ∮ C N d t t k + 1 P ( t ) ≤ 5 π 2 ρ 2 − 4 k 4 sinh ( π ρ / 2 ) → 0 , N → ∞ ,
for every k ∈ N 0 . By the residue theorem and (123), the integral over C N equals
Z k * − 4 π ∑ m = 1 N ( − 1 ) m − 1 m 4 k − 1 sinh ( π m ) ,
where Z k * is the residue at t = 0 . Letting N → ∞ yields (121).
(d) Let k ∈ N . The terms u m = 1 m 4 k − 1 sinh ( π m ) strictly decrease in m and tend to 0, so the alternating series in (121) satisfies
u 1 − u 2 < ∑ m = 1 ∞ ( − 1 ) m − 1 u m < u 1 − u 2 + u 3 .
Multiplying by 4 π and using
4 π u 1 = 4 π sinh π , 4 π u 2 = 8 π 16 k sinh ( 2 π ) , 4 π u 3 = 12 π 81 k sinh ( 3 π )
gives (122). Letting k → ∞ in (122) shows Z k * → 4 π sinh π , and (125) with m = 1 shows that this limit equals ∏ n = 2 ∞ 1 − 1 n 4 − 1 . The proof of Theorem 23 is complete. □
Remark 26. 
The proof of Theorem 23(b) works for any real s ≥ 2 : the sequence k ↦ ζ * ( { s } k ) , where { s } k denotes s repeated k times, is strictly increasing and strictly concave, its increments satisfy
1 2 s < δ k + 1 δ k < ζ ( s ) − 1 < 1 ,
and, by summing all increments,
lim k → ∞ ζ * ( { s } k ) = ∏ n = 2 ∞ 1 − n − s − 1 .
For s = 2 , this limit equals 2, and the residue argument in the proof of Theorem 23(c), applied to
∏ n = 1 ∞ 1 − t n 2 = sin π t π t ,
gives the classical formula
ζ * ( { 2 } k ) = 2 η ( 2 k ) = 2 1 − 2 1 − 2 k ζ ( 2 k )
for k ∈ N . Numerically, the ratios δ k + 1 δ k for s = 4 decrease to 1 16 , in accordance with the asymptotic relation in Theorem 23(d).
Combining (121) with (116), we obtain closed forms of a family of series involving the hyperbolic sine.
Corollary 3. 
For k ∈ N 0 , we have
∑ m = 1 ∞ ( − 1 ) m − 1 m 4 k − 1 sinh ( π m ) = 4 k − 1 π 4 k − 1 k ! ∑ n = 0 k ( − 1 ) n n ! B k , n t 1 , t 2 , … , t k − n + 1 ,
where t m = ( − 1 ) m 2 · m ! ( 4 m + 2 ) ! as in Lemma 13. In particular, the left-hand side of (126) is a rational multiple of π 4 k − 1 , and
∑ m = 1 ∞ ( − 1 ) m − 1 m sinh ( π m ) = 1 4 π , ∑ m = 1 ∞ ( − 1 ) m − 1 m 3 sinh ( π m ) = π 3 360 , ∑ m = 1 ∞ ( − 1 ) m − 1 m 7 sinh ( π m ) = 13 π 7 453600 , ∑ m = 1 ∞ ( − 1 ) m − 1 m 11 sinh ( π m ) = 4009 π 11 13621608000 .
Proof. 
By (121) and (116), the left-hand side of (126) equals Z k * 4 π ; this gives (126); for k = 0 , we use Z 0 * = 1 and B 0 , 0 = 1 . The partial Bell polynomials
B k , n ( t 1 , t 2 , … , t k − n + 1 )
are rational, so the right-hand side is a rational multiple of π 4 k − 1 . The special values follow from Z 0 * = 1 and the values of Z 1 * , Z 2 * , and Z 3 * listed in Remark 25. □
Remark 27. 
The second special value in Corollary 3 is a classical formula of Ramanujan; see [3]. Formula (126) expresses the whole family, with exponents 4 k − 1 , through partial Bell polynomials at the explicit sequence ( t m ) , whose values are given in closed form by (33).

7. Conclusions

In this paper, we evaluated the complete Bell polynomials Y k and the partial Bell polynomials B n , k at a wide range of sequences in closed form.
The method rests on Lemmas 1 and 2: if the Maclaurin series of a function and of its logarithm, or of a function and of its exponential, are both known, then the coefficients of one series are the arguments of Y k and the coefficients of the other are its values. The partial Bell polynomials B n , k then follow from the complete ones, either by Lemma 8, which uses the integer powers of the function, or by Lemma 9, which uses Newton’s forward differences and the Stirling numbers of the first kind. Where a derivative formula is needed, it comes from the Faà di Bruno formula in Lemma 3.
Applying this method, we obtained evaluations at sequences built from the Bernoulli numbers, the Dirichlet eta and Riemann zeta values at non-positive integers, factorials, Stirling numbers, harmonic numbers, and Catalan numbers, and at the Maclaurin coefficients of the logarithms of the tangent, cosine, power-exponential, Gegenbauer-type, inverse hyperbolic sine, logarithmic, and arctangent functions. The values are expressed in terms of generalized Bernoulli and Euler polynomials, Stirling and Lah numbers, and Gegenbauer polynomials. Along the way, we derived series expansions of real powers of
tan z z , 1 + x 1 − x , ( 1 + x ) 1 + x , ( 1 − x ) 1 / ( 1 − x ) , 1 − 2 x cos ϕ + x 2 , ln ( 1 + x ) x , 1 + x sin x , sin ( a x ) a sin x , 1 + 1 − 4 x 2 ,
and we determined in Corollary 1 the radius of convergence of the series of tan z z λ for every real λ .
The main new results are those in Section 4 and Section 5. Products of the sine and hyperbolic sine functions extend the evaluations of Y k at the Bernoulli numbers B 2 k , which correspond to n = 1 in (6) and (7), to arguments involving the Bernoulli numbers B 4 k , which correspond to n = 2 . Exponentials of six elementary functions and real powers of the generating function of the Catalan numbers give further closed forms of Y k and B n , k .
As applications, we obtained in Section 6 a convolution identity for the Bernoulli numbers B 4 k in Theorem 21, and closed forms of the multiple zeta values ζ ( 4 , 4 , … , 4 ) and ζ * ( 4 , 4 , … , 4 ) in Theorem 22. In Theorem 23, we proved that ζ ( 4 , 4 , … , 4 ) is strictly logarithmically concave and eventually strictly decreasing in the number of arguments, while ζ * ( 4 , 4 , … , 4 ) is strictly increasing and strictly concave and converges to 4 π sinh π at the geometric rate 16 − k ; the residue argument behind (121) also gives the Ramanujan-type evaluations in Corollary 3. Every evaluation Y k ( a 1 , a 2 , … , a k ) = k ! c k in Section 3, Section 4 and Section 5 yields, in the same way, a convolution identity between the arguments and the values.
In future work, we plan to study evaluations at the Bernoulli numbers B 2 i n for n ≥ 3 , which correspond to the values of ζ ( 2 n , … , 2 n ) and of ζ * ( 2 n , … , 2 n ) , including the analogues of Theorem 23 for ζ * ( 2 n , … , 2 n ) , and we plan to find closed forms of B n , k in the cases where only finite-difference sums, as in Theorems 1 and 15, are available at present.

Author Contributions

The author contributed to all parts of the manuscript and read and approved the final manuscript.

Funding

The author was partially supported by the Natural Science Foundation of Inner Mongolia Autonomous Region (Grant No. 2025QN01041) and by the Youth Project of Hulunbuir City for Basic Research and Applied Basic Research (Grant No. GH2024020).

Institutional Review Board Statement

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 new data were created or analyzed in this study.

Acknowledgments

Not applicable.

Conflicts of Interest

The author declares no competing interests.

Use of Artificial Intelligence

The author used the AI assistant Claude (Anthropic) in preparing this article: for language editing and proofreading; for restructuring the article, including the collection of notation in SubSection 2.1, the division of the results into Section 3 to Section 5, and the revision of the title, abstract, keywords, and subject classification; for drafting the introductions to the sections and subsections and other short connecting paragraphs; for assistance in deriving and checking some of the new formulas and their proofs, in particular the radius of convergence in Corollary 1 and the monotonicity results, the series representation, and the Ramanujan-type evaluations in Theorem 23 and Corollary 3; and for verifying the formulas in the lemmas, theorems, and corollaries by exact computer-algebra computations for small indices and by high-precision numerical computations, which led to corrections of several statements. The author checked all mathematical results, proofs, and computations, and takes full responsibility for the content of the article..

References

  1. Abramowitz, M.; Stegun, I. A. (Eds.) Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables; National Bureau of Standards Applied Mathematics Series 55, reprint of the 1972 ed.; Dover Publications: New York, 1992. [Google Scholar]
  2. Bao, L.-Y.; He, C.-Y.; Qi, F. Monotonic sequences and inequalities involving the ratio between two adjacent nonzero Bernoulli numbers. Math. Inequal. Appl. 2026, 29(no. 1), 1–14. [Google Scholar] [CrossRef]
  3. Berndt, B. C. Ramanujan’s Notebooks, Part II; Springer-Verlag: New York, 1989. [Google Scholar]
  4. Bernstein, D. S. Scalar, Vector, and Matrix Mathematics: Theory, Facts, and Formulas, revised and expanded ed.; Princeton University Press: Princeton, NJ, 2018. [Google Scholar]
  5. Burgos Gil, J. I.; Fresán, J. Multiple Zeta Values: from Numbers to Motives. Clay Mathematics Proceedings, Amer. Math. Soc., Providence, RI, to appear. Available online: https://javier.fresan.perso.math.cnrs.fr/mzv.pdf.
  6. Cao, J.; Qi, F.; Du, W.-S. Closed-form formulas for the nth derivative of the power-exponential function xx. Symmetry 2023, 15(no. 2, Art. 323), 13 pp. [Google Scholar] [CrossRef]
  7. Charalambides, C. A. Enumerative Combinatorics; Chapman & Hall/CRC: Boca Raton, FL, 2002. [Google Scholar]
  8. Chen, K.-W.; Chung, C.-L. Sum relations of multiple zeta star values with even arguments. Mediterr. J. Math. 2017, 14(no. 3, Art. 110), 13 pp. [Google Scholar] [CrossRef]
  9. Chen, K.-W.; Chung, C.-L.; Eie, M. Sum formulas of multiple zeta values with arguments multiples of a common positive integer. J. Number Theory 2017, 177, 479–496. [Google Scholar] [CrossRef]
  10. Chiţescu, I. Around the formula of Faà di Bruno; Éditions universitaires européennes: Mauritius, 2017. [Google Scholar]
  11. Comtet, L. Advanced Combinatorics, revised and enlarged ed.; D. Reidel Publishing Co.: Dordrecht, 1974. [Google Scholar] [CrossRef]
  12. Genčev, M. Extension of Hoffman’s combinatorial identity via specific zeta-like series. Results Math. 2024, 79(no. 1, Art. 2), 25 pp. [Google Scholar] [CrossRef]
  13. Genčev, M. On restricted sum formulas for multiple zeta values with even arguments. Arch. Math. 2016, 107(no. 1), 9–22. [Google Scholar] [CrossRef]
  14. Gradshteyn, I. S.; Ryzhik, I. M. Table of Integrals, Series, and Products, translated from the Russian, edited and with a preface by D. Zwillinger and V. Moll, 8th ed.; Elsevier/Academic Press: Amsterdam, 2015. [Google Scholar] [CrossRef]
  15. Grishaev, A. V.; Sazonov, V. F. New formulas for Maclaurin’s series expansions of powers of arctangent and other related topics. Integral Transform. Spec. Funct. 2025, 36(no. 1), 39–52. [Google Scholar] [CrossRef]
  16. Guo, B.-N.; Lim, D.; Qi, F. 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. Discret. Math. 2022, 16(no. 2), 427–466. [Google Scholar] [CrossRef]
  17. Guo, B.-N.; Lim, D.; Qi, F. Series expansions of powers of arcsine, closed forms for special values of Bell polynomials, and series representations of generalized logsine functions. AIMS Math. 2021, 6(no. 7), 7494–7517. [Google Scholar] [CrossRef]
  18. Hansen, E. R. A Table of Series and Products; Prentice-Hall: Englewood Cliffs, NJ, 1975. [Google Scholar]
  19. He, C.-Y.; Qi, F. Reformulations and generalizations of Hoffman’s and Genčev’s combinatorial identities. Results Math. 2024, 79(no. 4, Art. 131), 17 pp. [Google Scholar] [CrossRef]
  20. Hoffman, M. E. Multiple harmonic series. Pac. J. Math. 1992, 152(no. 2), 275–290. Available online: http://projecteuclid.org/euclid.pjm/1102636166. [CrossRef]
  21. Hong, Y.; Guo, B.-N.; Qi, F. 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. 2021, 129(no. 1), 409–423. [Google Scholar] [CrossRef]
  22. Jin, S.; Guo, B.-N.; Qi, F. Partial Bell polynomials, falling and rising factorials, Stirling numbers, and combinatorial identities. CMES Comput. Model. Eng. Sci. 2022, 132(no. 3), 781–799. [Google Scholar] [CrossRef]
  23. Koshy, T. Catalan Numbers with Applications; Oxford University Press: Oxford, 2009. [Google Scholar]
  24. Li, W.-H.; Lim, D.; Qi, F. Expanding the function ln(1+ex) into power series in terms of the Dirichlet eta function and the Stirling numbers of the second kind. Carpathian Math. Publ. 2024, 16(no. 1), 320–327. [Google Scholar] [CrossRef]
  25. Mansour, T.; Schork, M. Commutation Relations, Normal Ordering, and Stirling Numbers; CRC Press: Boca Raton, FL, 2016. [Google Scholar]
  26. Olver, F. W. J.; Lozier, D. W.; Boisvert, R. F.; Clark, C. W. (Eds.) NIST Handbook of Mathematical Functions; Cambridge University Press: New York, 2010; Available online: http://dlmf.nist.gov/.
  27. Qi, F. Alternative proofs of Xu’s forms for He–Qi’s combinatorial identities reformulating and generalizing Hoffman’s and Genčev’s combinatorial identities for complete Bell polynomials. Filomat. 10 February 2026, 40. Available online: https://www.researchgate.net/publication/400661711.
  28. Qi, F. Concise and elegant proofs of three formulas for complete Bell polynomials. arXiv 2026, arXiv:2605.22959. [Google Scholar] [CrossRef]
  29. Qi, F. Diagonal recurrence relations for the Stirling numbers of the first kind. Contrib. Discret. Math. 2016, 11(no. 1), 22–30. [Google Scholar] [CrossRef]
  30. Qi, F. Evaluations of complete Bell polynomials for quadruple-indexed Bernoulli numbers in terms of generalized Bernoulli polynomials; Qeios, 2026. [Google Scholar] [CrossRef]
  31. Qi, F. Power series expansions of real powers of inverse cosine and sine functions, closed-form formulas of partial Bell polynomials at specific arguments, and series representations of real powers of circular constant. Symmetry 2024, 16(no. 9, Art. 1145), 21 pp. [Google Scholar] [CrossRef]
  32. Qi, F. Series and connections among central factorial numbers, Stirling numbers, inverse of Vandermonde matrix, and normalized remainders of Maclaurin series expansions. Mathematics 2025, 13(no. 2, Art. 223), 52 pp. [Google Scholar] [CrossRef]
  33. Qi, F. Some properties of the Catalan numbers. Appl. Anal. Discret. Math. 2025, 19(no. 1), 176–184. [Google Scholar] [CrossRef]
  34. Qi, F. 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. Demonstr. Math. 2022, 55(no. 1), 710–736. [Google Scholar] [CrossRef]
  35. Qi, F. Uniform treatments of Bernoulli numbers, Stirling numbers, and their generating functions. arXiv 2025, arXiv:2504.16965. [Google Scholar] [CrossRef]
  36. Qi, F.; Guo, B.-N. Explicit formulas for special values of the Bell polynomials of the second kind and for the Euler numbers and polynomials. Mediterr. J. Math. 2017, 14(no. 3, Art. 140), 14 pp. [Google Scholar] [CrossRef]
  37. Qi, F.; He, C.-Y.; Lim, D. Explicit formulas of two Gauss hypergeometric functions and several combinatorial identities. Discret. Appl. Math. 2026, 393, 215–229. [Google Scholar] [CrossRef]
  38. Qi, F.; Milovanović, G. V.; Lim, D. Specific values of partial Bell polynomials and series expansions for real powers of functions and for composite functions. Filomat 2023, 37(no. 28), 9469–9485. [Google Scholar] [CrossRef]
  39. Qi, F.; Niu, D.-W.; Lim, D.; Guo, B.-N. Closed formulas and identities for the Bell polynomials and falling factorials. Contrib. Discret. Math. 2020, 15(no. 1), 163–174. [Google Scholar] [CrossRef]
  40. Qi, F.; Niu, D.-W.; Lim, D.; Yao, Y.-H. Special values of the Bell polynomials of the second kind for some sequences and functions. J. Math. Anal. Appl. 2020, 491(no. 2, Art. 124382), 31 pp. [Google Scholar] [CrossRef]
  41. Qi, F.; Taylor, P. Series expansions for powers of sinc function and closed-form expressions for specific partial Bell polynomials. Appl. Anal. Discret. Math. 2024, 18(no. 1), 92–115. [Google Scholar] [CrossRef]
  42. Quaintance, J.; Gould, H. W. Combinatorial Identities for Stirling Numbers; World Scientific Publishing: Singapore, 2016. [Google Scholar]
  43. Riordan, J. Combinatorial Identities, reprint of the 1968 original; Robert E. Krieger Publishing Co.: Huntington, NY, 1979. [Google Scholar]
  44. Stanley, R. P. Catalan Numbers; Cambridge University Press: New York, 2015. [Google Scholar] [CrossRef]
  45. Temme, N. M. Special Functions; John Wiley & Sons: New York, 1996. [Google Scholar] [CrossRef]
  46. Xu, A. Concise forms of He–Qi combinatorial identities in terms of generalized Bernoulli polynomials, Euler polynomials, and Catalan numbers. ResearchGate 2026. [Google Scholar] [CrossRef]
  47. Xu, A.-M.; Cen, G.-D. Closed formulas for computing higher-order derivatives of functions involving exponential functions. Appl. Math. Comput. 2015, 270, 136–141. [Google Scholar] [CrossRef]
  48. Zagier, D. Values of zeta functions and their applications. In First European Congress of Mathematics, vol. II (Paris, 1992), Progr. Math.; Birkhäuser, Basel, 1994; vol. 120, pp. 497–512. [Google Scholar]
  49. Zhao, J. Multiple Zeta Functions, Multiple Polylogarithms and Their Special Values. In Series on Number Theory and its Applications; World Scientific Publishing: Hackensack, NJ, 2016; vol. 12. [Google Scholar] [CrossRef]
  50. Zhang, Z.-Z.; Yang, J.-Z. Notes on some identities related to the partial Bell polynomials. Tamsui Oxf. J. Inf. Math. Sci. 2012, 28(no. 1), 39–48. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.