Next Article in Journal
An Iterative Algorithm for Solving Generalized Variational Inequalities and Fixed Points Problems
Previous Article in Journal
Comparison of the Orthogonal Polynomial Solutions for Fractional Integral Equations
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Arithmetic Means for a Class of Functions and the Modified Bessel Functions of the First Kind

1
School of Mathematics and Informatics, Henan Polytechnic University, Jiaozuo 454010, Henan, China
2
College of Mathematics, Inner Mongolia University for Nationalities, Tongliao 028043, Inner Mongolia, China
3
School of Mathematical Sciences, Tianjin Polytechnic University, Tianjin 300387, China
*
Author to whom correspondence should be addressed.
Mathematics 2019, 7(1), 60; https://doi.org/10.3390/math7010060
Submission received: 21 November 2018 / Revised: 27 December 2018 / Accepted: 2 January 2019 / Published: 8 January 2019

Abstract

:
In the paper, by virtue of the residue theorem in the theory of complex functions, the authors establish several identities between arithmetic means for a class of functions and the modified Bessel functions of the first kind, present several identities between arithmetic means for a class of functions and infinite series, and find several series expressions for the modified Bessel functions of the first kind.

1. Preliminaries

Recall from ([1], Chapter 6) and ([2], Chapter 5) that the classical Euler gamma function can be defined by
Γ ( w ) = 0 t w 1 e t d t , ( w ) > 0
or by
Γ ( w ) = lim n n ! n w k = 0 n ( w + k ) , w C { 0 , 1 , 2 , }
and that the logarithmic derivative
ψ ( w ) = [ ln Γ ( w ) ] = Γ ( w ) Γ ( w )
is called the psi or digamma function. One can also find these notions in the papers [3,4], the handbooks [1,2], and closely related references therein.
In most calculus texts, the quantity 1 b a a b f ( x ) d x is called the arithmetic mean for f ( x ) on the interval [ a , b ] . This is justified by noting that the approximating Riemann sums are the arithmetic means of values of f ( x ) at points distributed across [ a , b ] . See ([5], p. 368).
Recall from ([1], Chapter 9) and ([2], Chapter 10) that the modified Bessel functions of the first kind I ν ( w ) can be defined by
I ν ( w ) = k = 0 1 k ! Γ ( ν + k + 1 ) w 2 2 k + ν , | arg w | < π , ν C .
One can find some inequalities for I ν ( w ) in the papers [6,7,8,9,10] and closely related references therein.

2. Motivations

In ([11], Lemma 2.1), by virtue of two approaches, the identity
2 π 0 π / 2 α cos 2 ϕ β sin 2 ϕ d ϕ = α β I 0 ln α β
for α , β > 0 was derived. By virtue of the identity (1) and a double inequality for I 0 ( w ) , the double inequality
L ( α , β ) < 2 π 0 π / 2 α cos 2 ϕ β sin 2 ϕ d ϕ < I ( α , β )
for α , β > 0 and α β was discovered in ([11], Theorem 1.1), where
I ( α , β ) = 1 e β β α α 1 / ( β α ) and L ( α , β ) = β α ln β ln α
for α , β > 0 and α β are respectively called the exponential and logarithmic means [5,12,13,14]. Subsequently, there have been several papers, such as [10,15,16,17,18,19,20], dedicated to estimating or bounding the so-called Toader–Qi mean
T Q ( α , β ) = 2 π 0 π / 2 α cos 2 ϕ β sin 2 ϕ d ϕ = 1 2 π 0 2 π α cos 2 ϕ β sin 2 ϕ d ϕ , α , β > 0 .
In this paper, we consider the following arithmetic means for a class of functions,
Q ( α , β ; k , ) = 1 2 π 0 2 π α cos k ϕ β sin ϕ d ϕ , α , β > 0 , k , N .
It is clear that Q ( α , β ; 2 , 2 ) is just the above Toader–Qi mean T Q ( α , β ) .

3. A Definition and a Lemma

In this paper, one of our main tools is the residue theorem in the theory of complex functions.
Definition 1 
(([21], p. 112) and ([2], p. 19)). Let f ( w ) have an isolated singularity at w = w 0 and let
f ( w ) = n = c n ( w w 0 ) n
be its Laurent expansion about w = w 0 . Then the residue of f ( w ) at w = w 0 is defined to be the coefficient c 1 and is denoted by Res [ f ( w ) , w 0 ] = c 1 .
Lemma 1 
(Residue theorem ([21], p. 112) and ([2], p. 19)). Let f be analytic in the domain D except for the isolated singularities w 1 , w 2 , , w n . If γ is a simple closed contour in D and does not pass through any of the points w k for 1 k n , then
1 2 π i γ f ( w ) d w = k = 1 n Res [ f ( w ) , w k ] .

4. Several Identities and Series for Arithmetic Means

We now state and prove the first identity for the arithmetic mean Q ( α , β ; 1 , 1 ) and the modified Bessel function of the first kind I 0 ( w ) as follows.
Theorem 1.
For α , β > 0 , we have
1 2 π 0 2 π α cos ϕ β sin ϕ d ϕ = I 0 ( ln α ) 2 + ( ln β ) 2 .
Proof. 
Let i = 1 and w = e i ϕ . Then
cos ϕ = w + w 1 2 and sin ϕ = w w 1 2 i .
Accordingly, by direct calculation and Lemma 1, we have
1 2 π 0 2 π α cos ϕ β sin ϕ d ϕ = 1 2 π i | w | = 1 α ( w + w 1 ) / 2 β ( w w 1 ) / ( 2 i ) w d w = Res α ( w + w 1 ) / 2 β ( w w 1 ) / ( 2 i ) w , 0 .
Straightforward computation yields
α ( w + w 1 ) / 2 β ( w w 1 ) / ( 2 i ) = exp w + w 1 2 ln α + w w 1 2 i ln β = exp ln α i ln β 2 w + ln α + i ln β 2 1 w = p = 0 1 p ! ln α i ln β 2 w + ln α + i ln β 2 1 w p = p = 0 1 p ! q = 0 p p q ln α i ln β 2 w q ln α + i ln β 2 1 w p q = p = 0 1 p ! q = 0 p p q ln α i ln β 2 q ln α + i ln β 2 p q w 2 q p = m = 0 1 ( 2 m ) ! q = 0 2 m 2 m q ln α i ln β 2 q ln α + i ln β 2 2 m q w 2 ( q m ) + m = 0 1 ( 2 m + 1 ) ! q = 0 2 m + 1 2 m + 1 q ln α i ln β 2 q ln α + i ln β 2 2 m q + 1 w 2 ( q m ) 1 = m = 0 1 ( 2 m ) ! 2 m m ln α i ln β 2 m ln α + i ln β 2 2 m m w 2 ( m m ) + m = 0 1 ( 2 m ) ! q = 0 m 1 + q = m + 1 2 m 2 m q ln α i ln β 2 q ln α + i ln β 2 2 m q w 2 ( q m ) + m = 0 1 ( 2 m + 1 ) ! q = 0 2 m + 1 2 m + 1 q ln α i ln β 2 q ln α + i ln β 2 2 m q + 1 w 2 ( q m ) 1 = m = 0 1 ( 2 m ) ! 2 m m ( ln α ) 2 + ( ln β ) 2 4 m + m = 0 1 ( 2 m ) ! q = 0 m 1 + q = m + 1 2 m 2 m q ln α i ln β 2 q ln α + i ln β 2 2 m q w 2 ( q m ) + m = 0 1 ( 2 m + 1 ) ! q = 0 2 m + 1 2 m + 1 q ln α i ln β 2 q ln α + i ln β 2 2 m q + 1 w 2 ( q m ) 1 ,
where any empty sum is taken to be zero. Therefore, by Definition 1, we obtain
Res α ( w + w 1 ) / 2 β ( w w 1 ) / ( 2 i ) w , 0 = m = 0 1 ( 2 m ) ! 2 m m ( ln α ) 2 + ( ln β ) 2 4 m = m = 0 1 ( m ! ) 2 ( ln α ) 2 + ( ln β ) 2 2 2 m = I 0 ( ln α ) 2 + ( ln β ) 2 .
Substituting this equation into (4) results in the identity (2). The proof of Theorem 1 is thus complete. □
We now state and prove the second and third identities between the arithmetic means Q ( α , β ; 1 , 2 ) and Q ( α , β ; 2 , 1 ) and two infinite series as follows.
Theorem 2.
For α , β > 0 , we have
1 2 π 0 2 π α cos ϕ β sin 2 ϕ d ϕ = m = 0 ( ln β ) 2 m ( 2 m ) ! n = 0 m 2 m 2 n ( 2 n 1 ) / 2 2 m n ln α ln β 2 n m = 0 ( ln β ) 2 m + 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( 2 n 1 ) / 2 2 m n + 1 ln α ln β 2 n
and
1 2 π 0 2 π α cos 2 ϕ β sin ϕ d ϕ = m = 0 ( ln β ) 2 m ( 2 m ) ! n = 0 m 1 2 2 ( m + n ) 2 m 2 n ln α ln β 2 n q = 0 m + n ( 1 ) q 2 n q 2 ( m n ) m + n q + 1 4 m = 0 ( 1 ) m ( ln β ) 2 m + 1 ( 2 m + 1 ) ! n = 0 m ( 1 ) n 2 m + 1 2 n + 1 [ 2 ( 2 n + 1 ) 1 ] / 2 m + n + 1 ln α ln β 2 n + 1 .
Proof. 
As did in the proof of Theorem 1, we can obtain
1 2 π 0 2 π α cos ϕ β sin 2 ϕ d ϕ = Res α ( w + w 1 ) / 2 β [ ( w w 1 ) / ( 2 i ) ] 2 w , 0 .
Direct computation yields
α ( w + w 1 ) / 2 β [ ( w w 1 ) / ( 2 i ) ] 2 = exp w + w 1 2 ln α + w w 1 2 i 2 ln β = exp ln α 2 w + 1 w ln β 4 w 1 w 2 = m = 0 1 m ! ln α 2 w + 1 w ln β 4 w 1 w 2 m = m = 0 1 m ! n = 0 m m n ln α 2 w + 1 w n ( 1 ) m n ln β 4 w 1 w 2 m n = m = 0 1 m ! n = 0 m ( 1 ) m n m n ( ln α ) n ( ln β ) m n 2 2 m n 1 + w 2 n w 2 1 2 ( m n ) w 2 m n = m = 0 1 ( 2 m ) ! n = 0 2 m ( 1 ) 2 m n 2 m n ( ln α ) n ( ln β ) 2 m n 2 4 m n 1 + w 2 n w 2 1 2 ( 2 m n ) w 4 m n + m = 0 1 ( 2 m + 1 ) ! n = 0 2 m + 1 ( 1 ) 2 m + 1 n 2 m + 1 n ( ln α ) n ( ln β ) 2 m + 1 n 2 2 ( 2 m + 1 ) n 1 + w 2 n w 2 1 2 ( 2 m + 1 n ) w 2 ( 2 m + 1 ) n = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n 2 2 ( 2 m n ) 1 + w 2 2 n w 2 1 2 ( 2 m 2 n ) w 2 ( 2 m n ) + m = 0 1 ( 2 m ) ! n = 0 m 1 2 m 2 n + 1 ( ln α ) 2 n + 1 ( ln β ) 2 m ( 2 n + 1 ) 2 4 m ( 2 n + 1 ) 1 + w 2 2 n + 1 w 2 1 2 [ 2 m ( 2 n + 1 ) ] w 4 m ( 2 n + 1 ) + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m + 1 2 n 2 2 ( 2 m + 1 ) 2 n 1 + w 2 2 n w 2 1 2 ( 2 m + 1 2 n ) w 2 ( 2 m + 1 ) 2 n + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( ln β ) 2 m + 1 ( 2 n + 1 ) 2 2 ( 2 m + 1 ) ( 2 n + 1 ) 1 + w 2 2 n + 1 w 2 1 2 [ 2 m + 1 ( 2 n + 1 ) ] w 2 ( 2 m + 1 ) ( 2 n + 1 ) .
In the second and fourth sums above, all terms are not constant and dependent of the variable w. The first and third sums above can be rewritten as
m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n 2 2 ( 2 m n ) 1 + w 2 2 n w 2 1 2 ( 2 m 2 n ) w 2 ( 2 m n ) = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) ( 2 w ) 2 ( 2 m n ) p = 0 2 n 2 n p w 2 p p = 0 4 ( m n ) 4 ( m n ) p ( 1 ) p w 2 p = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) ( 2 w ) 2 ( 2 m n ) p = 0 2 ( 2 m n ) q = 0 p 2 n q 4 ( m n ) p q ( 1 ) p q w 2 p = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) 2 2 ( 2 m n ) q = 0 2 m n 2 n q 4 ( m n ) 2 m n q ( 1 ) 2 m n q + m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) ( 2 w ) 2 ( 2 m n ) × p = 0 2 m n 1 + p = 2 m n + 1 2 ( 2 m n ) q = 0 p 2 n q 4 ( m n ) p q ( 1 ) p q w 2 p
and
m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m + 1 2 n 2 2 ( 2 m + 1 ) 2 n 1 + w 2 2 n w 2 1 2 ( 2 m + 1 2 n ) w 2 ( 2 m + 1 ) 2 n = m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n + 1 ( 2 w ) 2 ( 2 m n + 1 ) × p = 0 2 n 2 n p w 2 p p = 0 2 ( 2 m 2 n + 1 ) 2 ( 2 m 2 n + 1 ) p w 2 p ( 1 ) p = m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n + 1 ( 2 w ) 2 ( 2 m n + 1 ) × p = 0 2 ( 2 m n + 1 ) q = 0 p 2 n q 2 ( 2 m 2 n + 1 ) p q ( 1 ) p q w 2 p = m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n + 1 2 2 ( 2 m n + 1 ) × q = 0 2 m n + 1 2 n q 2 ( 2 m 2 n + 1 ) 2 m n + 1 q ( 1 ) 2 m n + 1 q + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n + 1 ( 2 w ) 2 ( 2 m n + 1 ) × p = 0 2 m n + p = 2 m n + 2 2 ( 2 m n + 1 ) q = 0 p 2 n q 2 ( 2 m 2 n + 1 ) p q ( 1 ) p q w 2 p .
Consequently, the constant term in the Laurent expansion of the function α ( w + w 1 ) / 2 β [ ( w w 1 ) / ( 2 i ) ] 2 at w = 0 is equal to
m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) 2 2 ( 2 m n ) q = 0 2 m n 2 n q 4 ( m n ) 2 m n q ( 1 ) 2 m n q + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n + 1 2 2 ( 2 m n + 1 ) × q = 0 2 m n + 1 2 n q 2 ( 2 m 2 n + 1 ) 2 m n + 1 q ( 1 ) 2 m n + 1 q = m = 0 1 ( 2 m ) ! n = 0 m ( 1 ) n 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) 2 2 ( 2 m n ) q = 0 2 m n ( 1 ) q 2 n q 4 ( m n ) 2 m n q + m = 0 1 ( 2 m + 1 ) ! n = 0 m ( 1 ) n 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n + 1 2 2 ( 2 m n + 1 ) q = 0 2 m n + 1 ( 1 ) q 2 n q 2 ( 2 m 2 n + 1 ) 2 m n + 1 q = m = 0 1 ( 2 m ) ! n = 0 m ( 1 ) n 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) 2 2 ( 2 m n ) ( 4 ) 2 m n ( 2 n 1 ) / 2 2 m n + m = 0 1 ( 2 m + 1 ) ! n = 0 m ( 1 ) n 2 m + 1 2 n ( ln α ) 2 n ( ln β ) 2 m 2 n + 1 2 2 ( 2 m n + 1 ) ( 4 ) 2 m n + 1 ( 2 n 1 ) / 2 2 m n + 1 = m = 0 ( ln β ) 2 m ( 2 m ) ! n = 0 m 2 m 2 n ( 2 n 1 ) / 2 2 m n ln α ln β 2 n m = 0 ( ln β ) 2 m + 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( 2 n 1 ) / 2 2 m n + 1 ln α ln β 2 n ,
where we used the formula
r = 0 j ( 1 ) r q r 2 j q j r = ( 4 ) j ( q 1 ) / 2 j
in ([22], p. 6, Equation (1.40)) and ([23], p. 63, Equation (3.42)). In a word, it follows that
Res α ( w + w 1 ) / 2 β [ ( w w 1 ) / ( 2 i ) ] 2 w , 0 = m = 0 ( ln β ) 2 m ( 2 m ) ! n = 0 m 2 m 2 n ( 2 n 1 ) / 2 2 m n ln α ln β 2 n m = 0 ( ln β ) 2 m + 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( 2 n 1 ) / 2 2 m n + 1 ln α ln β 2 n .
The Equation (5) is thus proved.
As did in the above or in the proof of Theorem 1, we can also obtain
1 2 π 0 2 π α cos 2 ϕ β sin ϕ d ϕ = Res α [ ( w + w 1 ) / 2 ] 2 β ( w w 1 ) / ( 2 i ) w , 0 .
Straightforward calculation shows
α [ ( w + w 1 ) / 2 ] 2 β ( w w 1 ) / ( 2 i ) = exp w + w 1 2 2 ln α + w w 1 2 i ln β = exp ln α 4 w + 1 w 2 i ln β 2 w 1 w = m = 0 1 m ! ln α 4 w + 1 w 2 i ln β 2 w 1 w m = m = 0 1 m ! n = 0 m m n ln α 4 w + 1 w 2 n ( 1 ) m n i ln β 2 w 1 w m n = m = 0 1 m ! n = 0 m ( 1 ) m n m n ( ln α ) n ( i ln β ) m n 2 m + n 1 + w 2 2 n w 2 1 m n w m + n = m = 0 1 ( 2 m ) ! n = 0 2 m ( 1 ) 2 m n 2 m n ( ln α ) n ( i ln β ) 2 m n 2 2 m + n 1 + w 2 2 n w 2 1 2 m n w 2 m + n + m = 0 1 ( 2 m + 1 ) ! n = 0 2 m + 1 ( 1 ) 2 m + 1 n 2 m + 1 n ( ln α ) n ( i ln β ) 2 m + 1 n 2 2 m + 1 + n 1 + w 2 2 n w 2 1 2 m + 1 n w 2 m + 1 + n = m = 0 1 ( 2 m ) ! n = 0 m ( 1 ) 2 m 2 n 2 m 2 n ( ln α ) 2 n ( i ln β ) 2 m 2 n 2 2 m + 2 n 1 + w 2 4 n w 2 1 2 m 2 n w 2 m + 2 n + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n ( ln α ) 2 n ( i ln β ) 2 m + 1 2 n 2 2 m + 1 + 2 n 1 + w 2 4 n w 2 1 2 m + 1 2 n w 2 m + 1 + 2 n + m = 0 1 ( 2 m ) ! n = 0 m 1 2 m 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 m ( 2 n + 1 ) 2 2 m + 2 n + 1 1 + w 2 2 ( 2 n + 1 ) w 2 1 2 m ( 2 n + 1 ) w 2 m + 2 n + 1 + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 ( m n ) 2 2 ( m + n + 1 ) 1 + w 2 2 ( 2 n + 1 ) w 2 1 2 ( m n ) w 2 ( m + n + 1 ) .
In the second and third sums above, all terms are not constant and dependent of the variable w. The first and fourth sums above can be formulated as
m = 0 1 ( 2 m ) ! n = 0 m ( 1 ) 2 m 2 n 2 m 2 n ( ln α ) 2 n ( i ln β ) 2 m 2 n 2 2 m + 2 n 1 + w 2 4 n w 2 1 2 m 2 n w 2 m + 2 n = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( i ln β ) 2 ( m n ) ( 2 w ) 2 ( m + n ) p = 0 4 n 4 n p w 2 p p = 0 2 ( m n ) 2 ( m n ) p ( 1 ) p w 2 p = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( i ln β ) 2 ( m n ) ( 2 w ) 2 ( m + n ) p = 0 2 ( m + n ) q = 0 p 4 n q 2 ( m n ) p q ( 1 ) p q w 2 p = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( i ln β ) 2 ( m n ) 2 2 ( m + n ) q = 0 m + n 2 n q 2 ( m n ) m + n q ( 1 ) m + n q + m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( i ln β ) 2 ( m n ) ( 2 w ) 2 ( m + n ) × p = 0 m + n 1 + p = m + n + 1 2 ( m + n ) q = 0 p 4 n q 2 ( m n ) p q ( 1 ) p q w 2 p
and
m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 ( m n ) 2 2 ( m + n + 1 ) 1 + w 2 2 ( 2 n + 1 ) w 2 1 2 ( m n ) w 2 ( m + n + 1 ) = m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 ( m n ) ( 2 w ) 2 ( m + n + 1 ) × p = 0 2 ( 2 n + 1 ) 2 ( 2 n + 1 ) p w 2 p p = 0 2 ( m n ) 2 ( m n ) p w 2 p ( 1 ) 2 ( m n ) p = m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 ( m n ) ( 2 w ) 2 ( m + n + 1 ) × p = 0 2 ( m + n + 1 ) q = 0 p 2 ( 2 n + 1 ) q 2 ( m n ) p q ( 1 ) p q w 2 p = m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 ( m n ) 2 2 ( m + n + 1 ) × q = 0 m + n + 1 2 ( 2 n + 1 ) q 2 ( m n ) m + n + 1 q ( 1 ) m + n + 1 q + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 ( m n ) ( 2 w ) 2 ( m + n + 1 ) × p = 0 m + n + p = m + n + 2 2 ( m + n + 1 ) q = 0 p 2 ( 2 n + 1 ) q 2 ( m n ) p q ( 1 ) p q w 2 p .
Consequently, the constant term in the Laurent expansion of the function α [ ( w + w 1 ) / 2 ] 2 β ( w w 1 ) / ( 2 i ) at w = 0 is equal to
m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( i ln β ) 2 ( m n ) 2 2 ( m + n ) q = 0 m + n 2 n q 2 ( m n ) m + n q ( 1 ) m + n q + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( i ln β ) 2 ( m n ) 2 2 ( m + n + 1 ) × q = 0 m + n + 1 2 ( 2 n + 1 ) q 2 ( m n ) m + n + 1 q ( 1 ) m + n + 1 q = m = 0 1 ( 2 m ) ! n = 0 m 2 m 2 n ( ln α ) 2 n ( ln β ) 2 ( m n ) 2 2 ( m + n ) q = 0 m + n 2 n q 2 ( m n ) m + n q ( 1 ) q + m = 0 1 ( 2 m + 1 ) ! n = 0 m 2 m + 1 2 n + 1 ( ln α ) 2 n + 1 ( ln β ) 2 ( m n ) 2 2 ( m + n + 1 ) × q = 0 m + n + 1 2 ( 2 n + 1 ) q 2 ( m n ) m + n + 1 q ( 1 ) q 1 = m = 0 ( ln β ) 2 m ( 2 m ) ! n = 0 m 2 m 2 n ln α ln β 2 n 1 2 2 ( m + n ) q = 0 m + n 2 n q 2 ( m n ) m + n q ( 1 ) q + 1 4 m = 0 ( 1 ) m ( ln β ) 2 m + 1 ( 2 m + 1 ) ! n = 0 m ( 1 ) n 2 m + 1 2 n + 1 ln α ln β 2 n + 1 [ 2 ( 2 n + 1 ) 1 ] / 2 m + n + 1 ,
where we used the formula (7). In conclusion, it follows that
Res α ( w + w 1 ) / 2 β [ ( w w 1 ) / ( 2 i ) ] 2 w , 0 = m = 0 ( ln β ) 2 m ( 2 m ) ! n = 0 m 2 m 2 n ln α ln β 2 n 1 2 2 ( m + n ) q = 0 m + n 2 n q 2 ( m n ) m + n q ( 1 ) q + 1 4 m = 0 ( 1 ) m ( ln β ) 2 m + 1 ( 2 m + 1 ) ! n = 0 m ( 1 ) n 2 m + 1 2 n + 1 ln α ln β 2 n + 1 [ 2 ( 2 n + 1 ) 1 ] / 2 m + n + 1 .
The Equation (6) is thus proved. The proof of Theorem 2 is complete. □
We now state and prove an identity between the arithmetic mean Q ( α , β ; k , k ) and an infinite series as follows.
Theorem 3.
For α , β > 0 and k N , we have
1 2 π 0 2 π α cos k ϕ β sin k ϕ d ϕ = j = 0 ( 1 ) k j ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( i ) k q 2 j q ( k q 1 ) / 2 k j ln α ln β q .
Proof. 
Employing the transforms in (3) and Lemma 1 reveals
1 2 π 0 2 π α cos k ϕ β sin k ϕ d ϕ = 1 2 π i | w | = 1 α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] k w d w = Res α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] k w , 0 .
Easy computation gives
α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] k = exp w + w 1 2 k ln α + w w 1 2 i k ln β = exp 1 + w 2 k ln α + 1 w 2 k i k ln β ( 2 w ) k = j = 0 1 j ! 1 + w 2 k ln α + 1 w 2 k i k ln β ( 2 w ) k j = j = 0 1 j ! 1 ( 2 w ) k j q = 0 j j q 1 + w 2 k ln α q 1 w 2 k i k ln β j q = j = 0 1 j ! 1 ( 2 w ) k j q = 0 j j q ( ln α ) q ( ln β ) j q i ( j q ) k 1 + w 2 q 1 w 2 j q k = j = 0 q = 0 2 j 2 j q ( ln α ) q ( ln β ) 2 j q i ( 2 j q ) k ( 2 j ) ! 1 + w 2 q 1 w 2 2 j q ( 4 w 2 ) j k + j = 0 q = 0 2 j + 1 2 j + 1 q ( ln α ) q ( ln β ) 2 j + 1 q i ( 2 j + 1 q ) k ( 2 j + 1 ) ! 1 + w 2 q 1 w 2 2 j + 1 q ( 2 w ) 2 j + 1 k .
It is easy to see that, in the above sum j = 0 q = 0 2 j + 1 , there is no any constant term independent of w. On the other hand, because
( 1 + w ) q ( 1 w ) 2 j q = r = 0 q q r w r s = 0 2 j q 2 j q s ( 1 ) s w s = p = 0 2 j r = 0 p q r 2 j q p r ( 1 ) p r w p = r = 0 j q r 2 j q j r ( 1 ) j r w j + p = 0 j 1 + p = j + 1 2 j r = 0 p q r 2 j q p r ( 1 ) p r w p ,
where any empty sum is taken to be zero and r s = 0 for s > r 0 and r , s N , we have
j = 0 q = 0 2 j 2 j q ( ln α ) q ( ln β ) 2 j q i ( 2 j q ) k ( 2 j ) ! 1 + w 2 q 1 w 2 2 j q ( 4 w 2 ) j k = j = 0 q = 0 2 j 2 j q ( ln α ) q ( ln β ) 2 j q i ( 2 j q ) k ( 2 j ) ! 1 + w 2 k q 1 w 2 k ( 2 j q ) ( 4 w 2 ) k j = j = 0 q = 0 2 j 2 j q ( ln α ) q ( ln β ) 2 j q i ( 2 j q ) k ( 2 j ) ! 4 k j r = 0 k j ( 1 ) k j r k q r k ( 2 j q ) k j r + j = 0 q = 0 2 j 2 j q ( ln α ) q ( ln β ) 2 j q i ( 2 j q ) k ( 2 j ) ! 1 ( 4 w 2 ) k j p = 0 k j 1 + p = k j + 1 2 k j r = 0 p k q r k ( 2 j q ) p r ( 1 ) p r w 2 p .
It is not difficult to see that, in the sum p = 0 k j 1 + p = k j + 1 2 k j , there is no any constant term independent of w. In a word, the constant term independent of w in the Laurent expansion of the function
α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] k
at z = 0 is equal to
j = 0 q = 0 2 j 2 j q ( ln α ) q ( ln β ) 2 j q i ( 2 j q ) k ( 2 j ) ! 4 k j r = 0 k j ( 1 ) k j r k q r k ( 2 j q ) k j r .
Consequently, by Definition 1, we deduce
Res α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] k w , 0 = j = 0 1 4 k j ( ln β ) 2 j ( 2 j ) ! q = 0 2 j 2 j q ln α ln β q ( i ) k q r = 0 k j ( 1 ) r k q r k ( 2 j q ) k j r .
Substituting the identity (7) into (10) and simplifying leads to
Res α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] k w , 0 = j = 0 ( 1 ) k j ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( i ) k q 2 j q ( k q 1 ) / 2 k j ln α ln β q .
Substituting this equation into (9) leads to (8). The proof of Theorem 3 is thus complete. □
We now state and prove an identity between the arithmetic mean Q ( α , β ; k , ) and an infinite series as follows.
Theorem 4.
For α , β > 0 and k , N , we have
1 2 π 0 2 π α cos k ϕ β sin ϕ d ϕ = m = 0 ( 1 ) [ k n + ( m n ) ] / 2 ( 1 ) ( m n ) + ( 1 ) k n 2 1 m ! × n = 0 m m n ln α 2 k n ln β ( 2 i ) m n r = 0 [ k n + ( m n ) ] / 2 ( 1 ) r k n r ( m n ) [ k n + ( m n ) ] / 2 r .
Proof. 
Utilizing the transforms in (3) and Lemma 1 results in
1 2 π 0 2 π α cos k ϕ β sin ϕ d ϕ = 1 2 π i | w | = 1 α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] w d w = Res α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] w , 0 .
Direct computation gives
α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] = exp w + w 1 2 k ln α + w w 1 2 i ln β = exp w + 1 w k ln α 2 k + w 1 w ln β ( 2 i ) = m = 0 1 m ! w + 1 w k ln α 2 k + w 1 w ln β ( 2 i ) m = m = 0 1 m ! n = 0 m m n w + 1 w k ln α 2 k n w 1 w ln β ( 2 i ) m n = m = 0 1 m ! n = 0 m m n ln α 2 k n ln β ( 2 i ) m n w + 1 w k n w 1 w ( m n ) = m = 0 1 m ! n = 0 m m n ln α 2 k n ln β ( 2 i ) m n p = 0 k n k n p w p 1 w k n p q = 0 ( m n ) ( m n ) q w q 1 w ( m n ) q = m = 0 1 m ! n = 0 m m n ln α 2 k n ln β ( 2 i ) m n p = 0 k n k n p w 2 p k n q = 0 ( m n ) ( m n ) q ( 1 ) ( m n ) q w 2 q ( m n ) = m = 0 1 m ! n = 0 m m n ln α 2 k n ln β ( 2 i ) m n ( 1 ) ( m n ) w k n + ( m n ) p = 0 k n k n p w 2 p q = 0 ( m n ) ( m n ) q ( 1 ) q w 2 q = m = 0 1 m ! n = 0 m m n ln α 2 k n ln β ( 2 i ) m n ( 1 ) ( m n ) w k n + ( m n ) j = 0 k n + ( m n ) r = 0 j ( 1 ) j r k n r ( m n ) j r w 2 j .
Therefore, by Definition 1, we deduce
Res α [ ( w + w 1 ) / 2 ] k β [ ( w w 1 ) / ( 2 i ) ] w , 0 = m = 0 1 m ! n = 0 m m n ln α 2 k n ln β ( 2 i ) m n × ( 1 ) ( m n ) 1 + ( 1 ) k n + ( m n ) 2 r = 0 [ k n + ( m n ) ] / 2 ( 1 ) [ k n + ( m n ) ] / 2 r k n r ( m n ) [ k n + ( m n ) ] / 2 r .
Substituting this result into (12) leads to the identity (11). The proof of Theorem 4 is thus complete. □

5. Two New Series Expressions for I 0

Combining the Equation (8) in Theorem 3 with the identities (2) and (1) in sequence, we can find two series expressions for I 0 as follows.
Theorem 5.
For x R , we have
I 0 ( x ) = j = 0 ( 1 ) j 1 / 2 j x 2 j ( 2 j ) ! .
Proof. 
When taking k = 1 , the right hand side of the Equation (8) in Theorem 3 becomes
j = 0 ( 1 ) j ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( i ) q 2 j q ( q 1 ) / 2 j ln α ln β q = j = 0 ( 1 ) j ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( 1 ) q e q π i / 2 2 j q ( q 1 ) / 2 j ln α ln β q = j = 0 ( 1 ) j ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( 1 ) q cos q π 2 + i sin q π 2 2 j q ( q 1 ) / 2 j ln α ln β q = j = 0 ( 1 ) j ( ln β ) 2 j ( 2 j ) ! q = 0 j ( 1 ) q 2 j 2 q ( 2 q 1 ) / 2 j ln α ln β 2 q + i j = 0 ( 1 ) j ( ln β ) 2 j ( 2 j ) ! q = 0 j 1 ( 1 ) q + 1 2 j 2 q + 1 q j ln α ln β 2 q + 1 = q = 0 ( 1 ) q j = q ( 1 ) j ( ln β ) 2 j ( 2 j ) ! 2 j 2 q ( 2 q 1 ) / 2 j ln α ln β 2 q i q = 0 ( 1 ) q j = q + 1 ( 1 ) j ( 2 j ) ! 2 j 2 q + 1 q j ( ln β ) 2 j ln α ln β 2 q + 1 = q = 0 ( 1 ) q j = q ( 1 ) j ( 2 j ) ! 2 j 2 q ( 2 q 1 ) / 2 j ( ln β ) 2 j ln α ln β 2 q ,
where any empty sum is taken to be zero and r s = 0 for s > r 0 and r , s N . This means that
I 0 ( ln α ) 2 + ( ln β ) 2 = q = 0 ( 1 ) q j = q ( 1 ) j ( 2 j ) ! 2 j 2 q ( 2 q 1 ) / 2 j ( ln β ) 2 j ln α ln β 2 q .
Further letting α 1 and β = e x for x R reduces to (13). The proof of Theorem 5 is complete. □
Theorem 6.
For x R , we have
I 0 ( x ) = 1 cosh x j = 0 1 / 2 2 j ( 2 x ) 2 j ( 2 j ) ! .
Proof. 
When taking k = 2 , the right hand side of the Equation (8) in Theorem 3 becomes
j = 0 ( 1 ) 2 j ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( i ) 2 q 2 j q ( 2 q 1 ) / 2 2 j ln α ln β q = j = 0 ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( 1 ) q 2 j q ( 2 q 1 ) / 2 2 j ln α ln β q .
This implies that
α β I 0 ln α β = j = 0 ( ln β ) 2 j ( 2 j ) ! q = 0 2 j ( 1 ) q 2 j q ( 2 q 1 ) / 2 2 j ln α ln β q .
Further taking α 1 reveals
β I 0 ln 1 β = j = 0 ( ln β ) 2 j ( 2 j ) ! 1 / 2 2 j .
Again letting β = e ± 2 x results in
e ± x I 0 ( x ) = j = 0 ( ± 2 x ) 2 j ( 2 j ) ! 1 / 2 2 j
which can be rewritten as (14). The proof of Theorem 6 is complete. □

Author Contributions

The authors contributed equally to this work. All authors read and approved the final manuscript.

Funding

This research received no external funding.

Acknowledgments

The authors are thankful to the anonymous referees for their careful corrections to and valuable comments on the original version of this paper.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Abramowitz, M.; Stegun, I.A. (Eds.) Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables; Applied Mathematics Series 55; National Bureau of Standards: Washington, DC, USA, 1972.
  2. Olver, F.W.; Lozier, D.W.; Boisvert, R.F.; Clark, C.W. (Eds.) NIST Handbook of Mathematical Functions; Cambridge University Press: New York, NY, USA, 2010. [Google Scholar]
  3. Qi, F.; Guo, B.-N. Integral representations and complete monotonicity of remainders of the Binet and Stirling formulas for the gamma function. Rev. R. Acad. Cienc. Exactas Fís. Nat. Ser. A Math. RACSAM 2017, 111, 425–434. [Google Scholar] [CrossRef]
  4. Qi, F. Limit formulas for ratios between derivatives of the gamma and digamma functions at their singularities. Filomat 2013, 27, 601–604. [Google Scholar] [CrossRef] [Green Version]
  5. Handbook of Means and Their Inequalities; Revised from the 1988 Original [P. S. Bullen, D. S. Mitrinović and P. M. Vasić, Means and Their Inequalities, Reidel, Dordrecht], Mathematics and Its Applications 560; Kluwer Academic Publishers Group: Dordrecht, The Netherlands, 2003.
  6. Qi, F. Properties of modified Bessel functions and completely monotonic degrees of differences between exponential and trigamma functions. Math. Inequal. Appl. 2015, 18, 493–518. [Google Scholar] [CrossRef]
  7. Qi, F.; Berg, C. Complete monotonicity of a difference between the exponential and trigamma functions and properties related to a modified Bessel function. Mediterr. J. Math. 2013, 10, 1685–1696. [Google Scholar] [CrossRef]
  8. Guo, B.-N.; Qi, F. Some inequalities and absolute monotonicity for modified Bessel functions of the first kind. Commun. Korean Math. Soc. 2016, 31, 355–363. [Google Scholar] [CrossRef]
  9. Qi, F.; Wang, S.-H. Complete monotonicity, completely monotonic degree, integral representations, and an inequality related to the exponential, trigamma, and modified Bessel functions. Glob. J. Math. Anal. 2014, 2, 91–97. [Google Scholar] [CrossRef]
  10. Yang, Z.-H.; Chu, Y.-M. On approximating the modified Bessel function of the first kind and Toader-Qi mean. J. Inequal. Appl. 2016, 40, 21. [Google Scholar] [CrossRef]
  11. Qi, F.; Shi, X.T.; Liu, F.F.; Yang, Z.H. A double inequality for an integral mean in terms of the exponential and logarithmic means. Period. Math. Hungar. 2017, 75, 180–189. [Google Scholar] [CrossRef]
  12. Guo, B.-N.; Qi, F. On inequalities for the exponential and logarithmic functions and means. Malays. J. Math. Sci. 2016, 10, 23–34. [Google Scholar]
  13. Qi, F.; Guo, B.-N. The reciprocal of the weighted geometric mean of many positive numbers is a Stieltjes function. Quaest. Math. 2018, 41, 653–664. [Google Scholar] [CrossRef]
  14. Qi, F.; Lim, D. Integral representations of bivariate complex geometric mean and their applications. J. Comput. Appl. Math. 2018, 330, 41–58. [Google Scholar] [CrossRef]
  15. Qi, F.; Guo, B.-N. Lévy–Khintchine representation of Toader–Qi mean. Math. Inequal. Appl. 2018, 21, 421–431. [Google Scholar] [CrossRef]
  16. Qian, W.-M.; Zhang, X.-H.; Chu, Y.-M. Sharp bounds for the Toader-Qi mean in terms of harmonic and geometric means. J. Math. Inequal. 2017, 11, 121–127. [Google Scholar] [CrossRef]
  17. Xu, H.Z.; Qian, W.M. Some sharp bounds for Toader-Qi mean and other bivariate means. J. Zhejiang Univ. Sci. Ed. 2017, 44, 526–530. (In Chinese) [Google Scholar]
  18. Yang, Z.-H. Some Sharp Inequalities for the Toader-Qi Mean. 2015. Available online: https://arxiv.org/abs/1507.05430 (accessed on 18 November 2018).
  19. Yang, Z.-H.; Chu, Y.-M. A sharp lower bound for Toader-Qi mean with applications. J. Funct. Spaces 2016, 2016, 4165601. [Google Scholar] [CrossRef]
  20. Yang, Z.-H.; Chu, Y.-M. Sharp bounds for Toader-Qi mean in terms of logarithmic and identric means. Math. Inequal. Appl. 2016, 19, 721–730. [Google Scholar] [CrossRef]
  21. Conway, J.B. Functions of One Complex Variable, 2nd ed.; Springer: New York, NY, USA; Heildeberg/Berlin, Germany, 1978. [Google Scholar]
  22. Gould, H.W. Combinatorial Identities: Table I: Intermediate Techniques for Summing Finite Series; Volume 4, Eight Tables Based on Seven Unpublished Manuscript Notebooks (1945–1990) of H.W. Gould, Includes Series Techniques and Certain Special Numbers; West Virginia University: Morgantown, WV, USA, 2010. [Google Scholar]
  23. Sprugnoli, R. Riordan Array Proofs of Identities in Gould’s Book; University of Florence: Florence, Italy, 2006. [Google Scholar]

Share and Cite

MDPI and ACS Style

Qi, F.; Yao, S.-W.; Guo, B.-N. Arithmetic Means for a Class of Functions and the Modified Bessel Functions of the First Kind. Mathematics 2019, 7, 60. https://doi.org/10.3390/math7010060

AMA Style

Qi F, Yao S-W, Guo B-N. Arithmetic Means for a Class of Functions and the Modified Bessel Functions of the First Kind. Mathematics. 2019; 7(1):60. https://doi.org/10.3390/math7010060

Chicago/Turabian Style

Qi, Feng, Shao-Wen Yao, and Bai-Ni Guo. 2019. "Arithmetic Means for a Class of Functions and the Modified Bessel Functions of the First Kind" Mathematics 7, no. 1: 60. https://doi.org/10.3390/math7010060

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop