Applied Mathematics Letters 20 (2007) 38–42

www.elsevier.com/locate/aml

Evaluation of integrals involving orthogonal polynomials: Laguerre

polynomial and Bessel function example
A.D. Alhaidari ∗
Shura Council, Riyadh 11212, Saudi Arabia
Physics Department, King Fahd University of Petroleum & Minerals, Dhahran 31261, Saudi Arabia

Received 8 December 2005; received in revised form 16 February 2006; accepted 22 February 2006

Abstract

Using the theory of orthogonal polynomials, their associated recursion relations and differential formulas we develop a new
method for evaluating integrals that include orthogonal polynomials. The method is illustrated by obtaining the following integral
result that involves the Bessel function and associated Laguerre polynomial:
 ∞  
1 ν+ 1
x ν e−x/2 Jν (μx)L 2ν (x)dx = νΓ ν + 1 1
(sin θ)ν+ 2 Cn 2 (cos θ),
n 2 √
0 2 πμ

where μ and ν are real parameters such that μ ≥ 0 and ν > − 12 , cos θ = μ2 −1/4 , and Cnλ (x) is a Gegenbauer (ultraspherical)
2
μ +1/4
polynomial.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: Definite integrals; Bessel function; Associated Laguerre polynomial; Gegenbauer polynomial; Recursion relation; Function spectral
decomposition

The use of spherical Bessel functions1 in the theoretical physics literature is overwhelming. One reason is that
these functions are eigen-solutions of the three-dimensional Laplacian in spherical coordinates [1]. Another is the
fact that they make up the radial component of the wavefunctions for free particles in three dimensions [1,2]. These
free-particle wavefunctions are the reference in scattering calculations for spherically symmetric interactions. Con-
sequently, the projection of the Bessel function onto various bases used in numerical schemes is important. These
weighted projections (typically, integrals) are necessary for computing the expectation values of selected observables
to be compared with measurements. All integral formulas having the general form shown√ in the abstract, which one
could find in various tables of integration, are with the integration variable x replaced by x in the argument of the
Bessel function [3]. This is useful for many applications where the Gaussian bases (a.k.a. “oscillator bases”) are widely
used in the calculation [4]. The elements of such a basis could be written as x ν e−x /2 L λn (x 2 ), or a linear combination
2

thereof. Nonetheless, an equally useful basis, which is sometimes referred to as the “Laguerre basis”, has elements of

∗ Corresponding address: Shura Council, Riyadh 11212, Saudi Arabia.

E-mail address: haidari@mailaps.org. 
1 If J (x) is the Bessel function, then the spherical Bessel function is defined as j (x) = π J
ν ν 2x ν+ 1 (x).
2

0893-9659/$ - see front matter 

c 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aml.2006.02.020
 A.D. Alhaidari / Applied Mathematics Letters 20 (2007) 38–42 39

the form x ν e−x/2 L λn (x). In typical scattering problems with x being the radial coordinate and ν =  + 1, where  is
the angular momentum quantum number, the former basis is orthogonal whereas the latter is not; it is “trithogonal”
(i.e., the basis overlap matrix – the identity representation – is tridiagonal). In the J -matrix method of quantum scat-
tering both bases are used in the calculation of the phase shift and resonance structure [5]. They are chosen because
they produce a tridiagonal matrix representation of the reference Hamiltonian. Consequently, an exact solution of the
reference problem is obtained in terms of orthogonal polynomials. These are solutions of the three-term recursion re-
lation resulting from the tridiagonal structure of the matrix representation of the reference wave equation. The method
developed here is an integral version of the differential method used in the J -matrix to obtain a spectral decomposi-
tion of the reference wavefunction in the chosen bases (i.e., the expansion coefficients of the wavefunction in these
bases). In this Letter, we use the theory of orthogonal polynomials, their associated recursion relations and differential
formulas to develop a method for evaluating a new integral that involves the Bessel function Jν (x) and the associated
Laguerre polynomials L 2ν n (x). In all subsequent developments
  we restrict our treatment to real spaces and fields.
ν
For Jν (x) defined on the positive real line, the limx→0 2x Jν (x) = 1/Γ (ν +1) and we could, in principle, expand
it as an infinite series in terms of a complete set of basis functions which are compatible with the range of Jν (x) and
its limiting values. Therefore, we assume that we can make a separable expansion of Jν (μx) in the space spanned by
the square integrable functions {φn (x)}∞ ∞
n=0 with real expansion coefficients {cn (μ)}n=0 . That is, we set out to find real
∞
functions {cn (μ)}n=0 such that

∞ 
∞
Jν (μx) = cn (μ)φn (x) = cn (μ)x ν e−αx L λn (x), (1)
n=0 n=0

where μ ≥ 0. The real basis parameters α and λ will be fixed as we go but meanwhile the only constraints are α > 0
and λ > −1. This expansion could also be interpreted as a spectral decomposition of the Bessel function on the chosen
Laguerre basis. Now, by using the orthogonality property of the associated Laguerre polynomials [6] we can write
 ∞
Γ (n + 1)
cn (μ) = x λ−ν e(α−1)x Jν (μx)L λn (x)dx. (2)
Γ (n + λ + 1) 0
Therefore, we must impose the stronger constraint that 1 > α > 0. The differential equation for the Bessel function
could be written as the eigenvalue equation DJν (x) = −Jν (x), where D is the second-order linear differential operator
d2
− νx 2 [6]. Using the differential equation of the associated Laguerre polynomials and their differential
2
dx 2
+ 1x dx
d

formula, x dx L n = n L λn − (n + λ)L λn−1 , we can write

d λ

   
2ν − λ 2n + 2ν + 1 2ν − λ 1 − 2α
Dφn = n −α + α φn −
2
+ (n + λ)φn−1 . (3)
x2 x x2 x
∞
We introduce a conjugate space spanned by the real L 2 functions {φ̃n (x)}n=0 such that the basis overlap matrix (i.e., the
inner product φ̃n |φm with integration measure dx) is tridiagonal. That is, φ̃n |φm = 0 if |n − m| |≥ 2. Therefore, if
we write φ̃n (x) = x ρ e−β x L λn (x), then the orthogonality of the associated Laguerre polynomials dictates that
β =1−α and ρ = −ν + λ + δ, (4)
where δ = 0 or 1. Now, the matrix representation of the eigenvalue equation DJν (μx) = −μ2 Jν (μx) can be written as

∞
φ̃n |(D + μ2 )|Jν = cm φ̃n |(D + μ2 )|φm = 0. (5)
m=0

Additionally, we require that the matrix representation of the differential operator D be at most tridiagonal. That is,
φ̃n |D|φm = 0 if |n − m| ≥ 2. Using the action of D on the basis given by Eq. (3) and the parameter relation (4) this
requirement translates into λ = 2ν and δ = 1. Hence, we could write




1

Inm ≡ φ̃n |(D + μ )|φm = (α + μ )φ̃n |φm − α(2m + 2ν + 1) φ̃n

φm
2 2 2
x




1

+ (2α − 1)(m + 2ν) φ̃n

φm−1 . (6)
x
40 A.D. Alhaidari / Applied Mathematics Letters 20 (2007) 38–42

Again, using the orthogonality relation of the associated Laguerre polynomials we obtain
Γ (n + 2ν + 1)
Inm = [(2n + 2ν + 1)(α 2 + μ2 − α)δnm − n(α 2 + μ2 )δn,m+1
Γ (n + 1)
− (n + 2ν + 1)(α 2 + μ2 + 1 − 2α)δn,m−1 ]. (7)
Thus, Eq. (5) results in a three-term recursion relation for the expansion coefficients which is written as In,n−1 cn−1 +
∞
In,n cn + In,n+1 cn+1 = 0. In terms of the polynomials {Pnα,ν }n=0 , which are defined as
Γ (n + 2ν + 1)
Pnα,ν (μ) ≡ cn (μ), (8)
Γ (n + 1)
this recursion relation reads as follows:
α,ν α,ν
(2n + 2ν + 1)(α 2 + μ2 − α)Pnα,ν = (n + 1)(α 2 + μ2 + 1 − 2α)Pn+1 + (n + 2ν)(α 2 + μ2 )Pn−1 . (9)
The two-parameter polynomial, Pnα,ν (μ),
which satisfies this recursion relation is, to the best of our knowledge, new.
However, for α = 12 Eq. (9) reduces to the three-term recursion relation of the Gegenbauer (ultraspherical) polynomial
ν+ 12 μ2 −1/4
Cn (y) [6], where y = y(μ) = μ2 +1/4
≡ cos θ and 0 < θ ≤ π. Consequently, we can write
1
,ν ν+ 12
Pnν (μ) ≡ Pn2 (μ) = f ν (μ) Cn (y), (10)
where f ν (μ) is an arbitrary real function of μ which is independent of the index n. Combining Eq. (2) with Eq. (8)
we get
 ∞
ν ν+ 21
Pn (μ) = x ν e−x/2 Jν (μx)L 2ν
n (x)dx = f ν (μ)Cn (y). (11)
0
Therefore, what remains is only to evaluate f ν (μ). To do that, we differentiate the above integral with respect to μ
using the chain rule on Jν (μx) (in the sequence μ → μx → x) and then integrate by parts since the integrand
vanishes at the end points. The result is as follows:
 ∞ 
dPnν ν −x/2 x 2ν d 2ν
μ =− x e Jν (μx) (ν + 1)L n − L n + x L n dx
2ν
dμ 0 2 dx
 ∞  
1
=− x ν e−x/2 Jν (μx) L 2ν 2ν 2ν
n + (n + 1)L n+1 − (n + 2ν)L n−1 dx. (12)
2 0
In going from the first to the second line of this equation, we have used the differential formula and recursion relation
n (x). Thus, Eq. (12) states that
for L 2ν
dPnν
2μ = −Pnν − (n + 1)Pn+1
ν ν
+ (n + 2ν)Pn−1 . (13)
dμ
On the other hand, differentiating the right side of Eq. (11) with respect to μ and using μ dμ
d
= (1 − y 2 ) dy
d
in the
differential relation of the Gegenbauer polynomial,
d λ 1 n(n + 1) λ 1 (n + 2λ − 1)(n + 2λ) λ
(1 − y 2 ) Cn (y) = − Cn+1 (y) + Cn−1 (y), (14)
dy 2 n+λ 2 n+λ
we obtain
 
dPnν d fν ν+ 1 n(n + 1) ν (n + 2ν)(n + 2ν + 1) ν
2μ = 2μ Cn 2 − Pn+1 + Pn−1 . (15)
dμ dμ n + ν + 1/2 n + ν + 1/2
Equating the right hand side of Eq. (13) with that of Eq. (15) and using the recursion relation of the Gegenbauer
polynomial we conclude that
   
d 1 1
μ + f ν (μ) = − ν + y f ν (μ). (16)
dμ 2 2
 A.D. Alhaidari / Applied Mathematics Letters 20 (2007) 38–42 41

√  
Defining gν (y) = μ f ν (μ), this could be rewritten for gν as (1 − y 2 ) dy
d
gν = − ν + 12 ygν which is easily
μ
integrated giving the solution f ν (μ) = Aν √1μ (sin θ )ν+ 2 , where sin θ = μ2 +1/4
1
and Aν is a constant independent of μ.
Putting all of the above together, we obtain the following realization of the real expansion coefficients cn (μ):
Γ (n + 1) Aν 1 ν+ 1
cn (μ) = √ (sin θ )ν+ 2 Cn 2 (cos θ ). (17)
Γ (n + 2ν + 1) μ
To determine Aν we substitute this in the expansion (1) and take the simultaneous limits x → 0 and μ → 0. Using
Γ (n+2λ)
Cnλ (−1) = (−1)n Γ (n+1) Γ (2λ) , we obtain
 μx ν 1 1 ∞
= Aν (μx)ν 22ν+1 (−1)n L 2ν
n (0). (18)
2 Γ (ν + 1) Γ (2ν + 1) n=0
  
Using the generating function ∞ λ
n=0 L n (x)t = (1 − t)
n −1−λ exp xt , which is valid for −1 ≤ t < 1 [7], we get the
t −1
 
following value for the constant Aν = √1π 2ν Γ ν + 12 . Inserting this in the expression fν (μ) = Aν √1μ (sin θ )ν+ 2
1

and substituting in Eq. (11) we, finally, get the sought-after result
 ∞  
1 1 1 ν+ 1
x ν e−x/2 Jν (μx)L 2ν
n (x)dx = 2 ν
Γ ν + √ (sin θ )ν+ 2 Cn 2 (cos θ ). (19)
0 2 πμ
It is to be noted that for negative ν the limit as μ → 0 of Jν (μx) does not exist but μ−ν Jν (μx) does. Now, since
√
ν > − 12 then μJν (μx) is non-singular at μ = 0. Therefore, this result should be rewritten for μ = 0 by multiplying
√
both sides by μ.

It is also worth noting that one could verify our result by writing the sum ∞ ν ν
n=0 Pn (μ) t , where Pn (μ) is given
n

by its integral representation in Eq. (11). Using the generating function of the Laguerre polynomials stated below
 ∞ ν −σ x ν −ν− 1
Eq. (18) and the integral formula 0 x e Jν (μx)dx = √π (2μ) Γ ν + 2 (μ + σ )
1 1 2 2 2 [8], one obtains


∞    −ν− 1
1 1
Pnν (μ)t n ν
(sin θ )ν+ 2 1 + t 2 − 2t cos θ
1 2
=2 Γ ν+ √ . (20)
n=0
2 πμ

Equating powers of t on both sides should reproduce (19). Alternatively, one may use the generating
 −λ ∞ λ
function 1 − 2yt + t 2 = n=0 Cn (y)t [9]. Additionally, using the more general formula for the integral
n
 ∞ τ −σ x
0 x e Jν (μx)dx, with τ + ν > −1 [8], one should in principle be able to obtain a more general result.
In conclusion, we summarize the method in the following steps:
(1) We write Jν (μx) as an infinite series of products of functions, {φn (x)}∞ ∞
n=0 and {cn (μ)}n=0 , that are compatible
with the range of definition of Jν (x) and its limit values: Eq. (1).
(2) Using the orthogonality property of the polynomials in the basis functions {φn } we write an integral expression
for the expansion coefficients {cn }: Eq. (2).
∞
(3) We construct the tridiagonal conjugate space spanned by {φ̃n (x)}n=0 and require that it also supports a tridiagonal
matrix representation for the Bessel differential operator D: Eq. (6).
(4) The resulting three-term recursion relation is solved in terms of orthogonal polynomials giving the expansion
coefficients {cn } modulo an arbitrary real function fν (μ): Eq. (10).
(5) The function f ν (μ) is determined (up to an overall constant factor Aν ) by solving a simple first-order linear
differential equation which is obtained by differentiating the resulting integral formula: Eqs. (11)–(15).
(6) The remaining constant Aν is determined by taking the limit of the expansion of Jν (μx) as (μ, x) → 0: Eq. (18).
It should also be noted that we needed knowledge of only two properties of the Bessel function to obtain the
analytic closed-form result. These were the differential equation of Jν (x) and its values at the limits. This is a highly
nontrivial observation that could have a major impact on the application of the method to a wider range of functions. It
also adds value to the method and could motivate its development into a powerful integration tool for such problems.
42 A.D. Alhaidari / Applied Mathematics Letters 20 (2007) 38–42

Acknowledgments

The author is grateful to H.A. Mavromatis (Physics, KFUPM) and R.S. Alassar (Math, KFUPM) for fruitful
consultations. Help in the calculation provided by A. Al-Hasan (Physics, KFU) is highly appreciated. I am also
grateful to the referee for the suggestion of including the elementary proof in Eq. (20).

References

[1] P.M. Morse, H. Feshbach, Methods of Theoretical Physics, vol. I, McGraw-Hill, New York, 1953;
R. Courant, D. Hilbert, Methods of Mathematical Physics, vol. I, Interscience, New York, 1966;
G. Arfken, Mathematical Methods for Physicists, 2nd ed., Academic, New York, 1970, pp. 478–533.
[2] A. Messiah, Quantum Mechanics, vol. I, North-Holland, Amsterdam, 1965;
E. Merzbacher, Quantum Mechanics, 2nd ed., Wiley, New York, 1970;
R.L. Liboff, Quantum Mechanics, 2nd ed., Addison-Wesley, Reading, 1992.
[3] See, for example I.S. Gradshteyn, I.M. Ryzhik, Tables of Integrals, Series, and Products, Academic, New York, 1980.
[4] R.G. Newton, Scattering Theory of Waves and Particles, 2nd ed., Springer, New York, 1966;
S. Gelman, Topics in Atomic Collision Theory, Academic, New York, 1969;
B.H. Bransden, Atomic Collision Theory, Benjamin, New York, 1970;
J.R. Taylor, Scattering Theory, Wiley, New York, 1972;
P.G. Burke, Potential Scattering in Atomic Physics, Plenum, New York, 1977.
[5] E.J. Heller, H.A. Yamani, Phys. Rev. A 9 (1974) 1201;
H.A. Yamani, L. Fishman, J. Math. Phys. 16 (1975) 410;
A.D. Alhaidari, E.J. Heller, H.A. Yamani, M.S. Abdelmonem (Eds.), The J -matrix method: Recent Developments and Selected Applications,
Springer, Amsterdam, 2006.
[6] A. Erdélyi (Ed.), Higher Transcendental Functions, vol. I, McGraw-Hil1, New York, 1953;
M. Abramowitz, I.A. Stegun (Eds.), Handbook of Mathematical Functions, Dover, New York, 1965;
T.S. Chihara, An Introduction to Orthogonal Polynomials, Gordon and Breach, New York, 1978;
G. Szegö, Orthogonal polynomials, 4th ed., Am. Math. Soc., Providence, RI, 1997.
[7] W. Magnus, F. Oberhettinger, R.P. Soni, Formulas and Theorems for the Special Functions of Mathematical Physics, 3rd ed., Springer, New
York, 1966, p. 242.
[8] Page 91 of Ref. [7].
[9] Page 222 of Ref. [7].