See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/1852891

Series Representation of the Modified Bessel Functions

Article · May 2001

Source: arXiv

CITATIONS READS
5 686

1 author:

Krzysztof Maslanka
Polish Academy of Sciences
25 PUBLICATIONS   84 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Krzysztof Maslanka on 19 July 2016.

The user has requested enhancement of the downloaded file.

 Series Representation of the Modified Bessel
Functions∗
arXiv:math-ph/0104018v1 12 Apr 2001

Krzysztof Maślanka
Astronomical Observatory of the Jagiellonian University
Orla 171, 30-244 Kraków, Poland

April 7, 1998

Abstract
Some power series representations of the modified Bessel functions
(McDonald functions Kα ) are derived using the little known formal-
ism of fractional derivatives. The resulting summation formulae are
believed to be new.

1 Fractional derivatives
There are several non-trivial examples in mathematics when some quantity,
originally defined as integer, can radically extend its original range and as-
sume fractional or even continuous values. The most common example is
the gamma-function of Euler which is a natural, unique generalization of the
ordinary factorial:
n Z∞
tn e−t dt
Y
n! ≡ i = Γ (n + 1) = (n > −1)
i=1 0

The same thing may be performed with the order of derivatives which can
also be made fractional. Although useful, fractional derivatives do not how-
ever create any essentially new calculus being rather some sort of particular,
relatively simple, integral transforms.
∗
Cracow Observatory preprint, no. 7/98

1
 Following Oldham and Spanier (see e.g. [1]; cf. also [2], [3] and [4]) we
s
s ∂
define the fractional derivative ∂x−a ≡ ∂x by an integral representation
known as the Riemann-Liouville integral. Given a real number s < 0, define
Zx
s 1
∂x−a f (x) = (x − t)−s−1 f (t) dt (1)
Γ (−s) a

where a < x is a fixed number, referred to as the boundary point. Since

s < 0, the integral is convergent, provided that f behaves well. For s ≥ 0 we
define
s
∂x−a ≡ ∂xn ∂x−a
s−n

where n is a positive integer chosen large enough so that s −n < 0 in order to

assure convergence of the integral in definition (1). It is not difficult to show
that, as expected, such a definition does not depend on n. One can further
prove that the familiar Leibniz rule for product differentiation has the form
!
s
∞
X s s−j j
∂x−a (f g) = ∂x−a f ∂x−a g (2)
j=0 j

where the generalized binomial is

(−1)j Γ (j − s)
!
s Γ (1 + s)
≡ =
j j!Γ (1 + s − j) j! Γ (−s)
It can also be shown that the following rule for the arbitrary order derivative
of the power function holds
Γ (p + 1) s=n p!
s
∂x−a (x − a)p = (x − a)p−s → (x − a)p−n (3)
Γ (p + 1 − s) (p − n)!
for any s and p > −1. However, the analogous formulae in the case of
the exponential and logarithm functions may, on the first sight, be quite
surprising
∂s s γ (−s, βx) s=n n
exp (βx) = β exp (βx) → β exp (βx)
∂xs Γ (−s)
∂s x−s 1 s=n (−1)n−1 (n − 1)!
 
ln x = ln x − ψ (−s) − C + →
∂xs Γ (1 − s) s xn
(Here n denotes a positive integer. For brevity we have set the boundary
point a = 0.)

2
2 Some definite integrals as fractional deriva-
tives
Let us now consider the following definite integrals (cf. [5], formulae 3.471.4
and 3.471.8)
Zx ! 1 ! !
µ−1 β β 2 −µ β β
−2µ
t (x − t) exp − dt = √ exp − Γ (µ) Kµ− 1 (4)
t πx 2x 2 2x
0

Zx ! !µ− 1 !

2 2 µ−1
 β 1 2 2
µ− 23 β
t −2µ
x −t exp − dt = √ x Γ (µ) Kµ− 1
t π β 2 x
0
(5)
which are valid for x > 0, Re β > 0, Re µ > 0. On using (1) it is evident that
both (4) and (5) may simply be interpreted as fractional derivatives of the
appropriate functions. Indeed
1
∂s β s+ 2
" !# ! !
β β β
s
x2s exp − = √ exp − Ks+ 1 (6)
∂x x πx 2x 2 2x
!s+ 1
∂s
" !# !
1 β 2 β 2
3 β
s
xs− 2 exp − √ =√ x 4
− 2s
Ks+ 1 (7)
∂x x π 2 2 x
It is thus possible even to define the functions Kν as fractional derivatives
with zero boundary point.

3 Polynomials Vk
In order to evaluate the left hand side of (6) and (7) let us introduce

(α) ∂k
Vk (βxα ) ≡ xk exp (βxα ) exp (−βxα ) (8)
∂xk
It is easy to prove that, contrary to their appearance, Vk are just polynomials
k
(α) (α)
Akj z j
X
Vk (z) =
j=0

3
with coefficients given by

(α) k
j
X (−1)i Γ (k − αi)
Akj = (−1)
i=0 i! (j − i)! Γ (−αi)

In particular, for α = −1 we have

(−1)
j
(−1)i Γ (k + i) (−1)k+j k! (k − 1)!
= (−1)k
X
Akj =
i=0 i! (j − i)! Γ (i) (k − j)! j! (j − 1)!

(The last equality may be rigorously proved using elementary methods pre-
sented in e.g. [7]; Mathematica 3.0 effectively simplifies such sums.) We shall
also need fractional derivatives of the following expressions

xν exp (−βxα )

with α, ν real. Using Leibniz rule (2) and the property (3) we have

∂s ν α ν−s Γ (ν + 1) α
X∞
(−1)k Γ (k − s) (α)
s
[x exp (−βx )] = x exp (−βx ) Vk
∂x Γ (−s) k=0 k! Γ (k − s + ν + 1)
(9)

4 Series expansion of the Modified Bessel func-

tions
Inserting (9) into (6) or (7) we can get a general expansion of the modified
Bessel functions (called also McDonald functions; cf. e.g. [6]) holding for
any positive α
√ 1
 
π Γ (2s) X ∞
(−1)k Γ k + 2 − s (−1)
Ks (z) = exp (−z)  Vk (2z) (10)
(2z)s
  
Γ 21 − s k=0 k! Γ k + 12 + s

Substituting polynomials (8) into (10) we further get, after some manipula-
tions
 
1
 
−s k
! j
∞
2 k − 1 (−2z) 
Ks (z) = 2s−1 Γ (s) z −s exp (−z) 1 + k
X X

k=1
1
+s j=1 j−1 j!
2 k

4
 where
Γ (k + a)
(a)k ≡
Γ (a)
denotes the Pochhammer symbol. The same method applied to (7) gives
 
k 1
s−1
√ Γ (2s) −s ∞
X (−1) Γ k + 2 − s (−1/2)
Ks (z) = 2 π 1  z exp (−z) V (z)
k! Γ k + 1 + s k

Γ 2 −s k=0 2
(11)
which is yet another power series convergent to the McDonald function. The
familiar ratio of Pochhammer symbols in (10) and (11) allows us to call these
expansions ’hypergeometric-like’ since they formally resemble the well-known
hypergeometric Kummer function 1 F1 .

References
[1] K. B. Oldham, J. Spanier, The Fractional Calculus, Academic Press, New
York, 1974

[2] V. S. Vladimiroff, Obobshchennye funkcii v matematicheskoj fizikie,

Nauka, Moskva, 1979 (in Russian)

[3] R. Camporesi, Harmonic Analysis and Propagators on Homogeneous

Spaces, Physics Reports 196 no. 1&2, 1990

[4] K. Maślanka, Efekty kwantowe w wielowymiarowych modelach kosmolog-

icznych, Ph. D. thesis, 1986, (unpublished, in Polish)

[5] I. S. Ryzhik, I. M. Gradstein, Tablicy intiegralov, sum, riadow i proizve-

denij, Nauka, Moskva, 1971, (in Russian)

[6] M. Abramowitz, I. A. Stegun, eds., Handbook of Mathematical Functions,

U.S. National Bureau of Standards, Dover, New York, 1965

[7] R. L. Graham, D. E. Knuth, O. Patashnik, Concrete Mathematics. A

Foundation for Computer Science, Addison-Wesley, 1994

View publication stats