Applied Mathematics and Computation 256 (2015) 885904

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc

Integral relations for solutions of the conuent Heun equation

La Jaccoud El-Jaick, Bartolomeu D.B. Figueiredo
a

Centro Brasileiro de Pesquisas Fsicas (CBPF), Rua Dr. Xavier Sigaud, 150, 22290-180 Rio de Janeiro, RJ, Brazil

a r t i c l e

i n f o

a b s t r a c t
Firstly, we construct kernels for integral relations among solutions of the conuent Heun
equation (CHE). Additional kernels are systematically generated by applying substitutions
of variables. Secondly, we establish integral relations between known solutions of the CHE
that are power series and solutions that are series of special functions. Thirdly, by using one
of the integral relations as an integral transformation we obtain a new series solution of the
ordinary spheroidal wave equation (a particular CHE). From this solution we construct new
series solutions of the general CHE, and show that these are suitable for solving the radial
part of the two-center problem in quantum mechanics.
2015 Elsevier Inc. All rights reserved.

Keywords:
Ordinary differential equations
Integral relations
Conuent Heun equation
Spheroidal wave equation

1. Introductory remarks
Recently we have found that the transformations of variables which preserve the form of the general Heun equation correspond to transformations which preserve the form of the equation for kernels of integral relations among solutions of the
Heun equation [1]. In fact, by using the known transformations of the Heun equation [2,3] we have found prescriptions for
transforming kernels and, in this manner, we have generated several new kernels for the equation.
The above correspondence can be extended to the conuent equations of the Heun family, that is, to the (single) conuent, double-conuent, biconuent and triconuent Heun equations [4,5], as well as to the reduced forms of such equations
[6,7]. In the present study we investigate only the conuent Heun equation (CHE). Specically:
 we deal with the construction and transformations of integral kernels for the CHE;
 from some of these kernels, we establish integral relations between known solutions of the CHE;
 from another kernel, we obtain new solutions in series of conuent hypergeometric functions for the CHE; we show that
these solutions are suitable to solve the radial part of the two-centre problem in quantum mechanics [8].
We write the CHE as [9]
2

zz  z0

d U
2

dz

B1 B2 z


dU 
B3  2xgz  z0 x2 zz  z0 U 0;
dz

CHE

where z0 ; Bi ; g and x are constants. This CHE is called generalized spheroidal wave equation by Leaver [9], but sometimes the
last expression refers to a particular case of the CHE [5,10]. Excepting the Mathieu equation, the CHE is the most studied of
the conuent Heun equations and includes the (ordinary) spheroidal equation [5] as a particular case. Its recent occurrence
Corresponding author.
E-mail addresses: leajj@cbpf.br (L.J. El-Jaick), barto@cbpf.br (B.D.B. Figueiredo).
http://dx.doi.org/10.1016/j.amc.2015.01.094
0096-3003/ 2015 Elsevier Inc. All rights reserved.

886

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

in several classes of quantum two-state systems [11], certainly will require new solutions for Eq. (1). On the other side, the
reduced conuent Heun equation (RCHE) is written as
2

zz  z0

d U
2

dz

B1 B2 z

dU
B3 qz  z0 U 0;
dz

RCHE

where z0 ; Bi and q (q 0) are constants. The RCHE describes the angular part of the Schrdinger equation for an electron in
the eld of a point electric dipole [12,13]. It appears as well in the study of two-level systems [14], polymer dynamics [15]
and theory of gravitation [16]. The form (2) for the RCHE results from the CHE (1) by means of the limits

x ! 0; g ! 1 such that

2gx q;

Whittaker-Ince limit:

In both equations, z 0 and z z0 are regular singular points with exponents (0; 1 B1 =z0 ) and (0; 1  B2  B1 =z0 ), respectively, that is, from ascending power series solutions we nd
B

limUz  1;
z!0

limUz  z
z!0

1 z 1
0

limUz  1;

z!z0

1B2  z 1

limUz  z  z0

z!z0

CHE; RCHE:

In contrast, at the irregular singular point z 1 the behaviour of the solutions is different for each equation since

limUz  eixz zigB2 =2 for the CHE1;

z!1

limUz  e2i

p
qz

z!1

z1=4B2 =2 for the RCHE2;

as follow from the normal and the subnormal Thom solutions [17] for the CHE and RCHE, respectively.
By the concepts of Ref. [7], the s-rank of the singularity at z 1 is 2 for the CHE, and 3=2 for the RCHE. However, more
important is the fact that the solutions exhibit the above behaviour predicted by the normal or subnormal Thom solutions,
and fact that the WhittakerInce limit (3) may generate solutions to the RCHE. In effect, most of the known solutions for the
RCHE [1820] has been obtained from solutions of the CHE by means of the limit (3). On account of this, and for the sake of
brevity, we restrict ourselves to integral relations among solutions of the CHE. We also do not consider kernels for
double-conuent Heun equations which result by taking the Leaver limit z0 ! 0 of Eqs. (1) and (2).
Integral relations, in principle, enable us to transform known solutions into solutions with different properties. However,
apart from the Mathieu equation, only in rare cases this task has been accomplished successfully. One case is constituted by
the solutions in series of associated Legendre functions for the Lam equation, obtained by Erdlyi in 1948 as a transformation of solutions in FourierJacobi series [21]; however, as far we are aware, his solutions have not been extended for the
general Heun equation (of which Lam equation is a particular case). Another example is an expansion in series of irregular
conuent hypergeometric functions for the CHE, which was obtained by Leaver in 1985 as an integral transformation of a
power-series solution [9]: the transformation was constructed for g iB2 =2  1 and, then, the result was extended for
arbitrary g.
To establish integral relations for solutions the CHE it is necessary to get appropriate integral kernels. To this end, in
Section 2 we proceed as in case of the general Heun equation [1]. In other words, rstly we insert into the integral connecting
two solutions a weight function wz; t which allows to write the CHE and the equation for its kernels in terms of differential
operators functionally identical (respecting z and t). In this manner, by examining each variable substitution which leaves
invariant the form of the CHE (one variable, z) we nd prescriptions for the variables transformations which preserve the
form of the equation for the kernels (two variables, z and t). By using these substitutions, we may systematically convert
a given (initial) kernel into new kernels. As initial kernels we use the ones obtained as limits of kernels of the general Heun
equation [1], adapting them to the form (1) for the CHE.
In Section 3 we nd integral relations which transform the Jaff power-series solutions [22] into known expansions in
series of irregular conuent hypergeometric functions, including the aforementioned solution given by Leaver. Similarly,
the power-series solutions of Baber and Hass [23] are transformed into known expansions in series of regular conuent
hypergeometric functions. We also consider non-integral transformations (involving only substitutions of variables) which
do not modify the type of series: they transform, for example, a power-series solution into another power-series solution,
and an expansion in series of special functions into another expansion in series special functions. In contrast, by applying
an integral transformation on an asymptotic (Thom) solution, in Section 4 we obtain a new solution in series of irregular
conuent hypergeometric functions for the spheroidal equation. That solution is extended to any CHE (not just the spheroidal equation). Then, by substitutions of variables (non-integral transformations) we obtain a group of solutions for the CHE;
we show that these solutions afford bounded and convergent solutions to the radial part of the Schrdinger equation for an
electron in the eld of two Coulomb centres [8] (the two-centre problem). In Section 5 we present concluding remarks and
mention open issues, while in Appendix A we give some formulas concerning special functions, and in Appendix B we discuss
asymptotic solutions for the CHE.
2. Kernels for the conuent Heun equation
In this section we regard kernels for the CHE (1). In particular,

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

887

 In Section 2.1 we get the correspondences among substitutions of variables which preserve the form of the CHE and the
substitutions which preserve the equation of the kernels of the CHE.
 In Section 2.2 we construct a group of kernels with an arbitrary constant of separation k, given by products of two conuent hypergeometric functions and elementary functions.
 In Section 2.3 we nd another group of kernels with an arbitrary constant of separation k, given by products of conuent
hypergeometric functions and Gauss hypergeometric functions (and elementary functions).
 Taking suitable values for k we get kernels given products of elementary functions and one special function; thus, in Sections 2.4 and 2.5 we nd products of elementary and conuent hypergeometric functions and, in Section 2.6, products of
elementary and Gauss hypergeometric functions.
Later on, in Section 4, we will need kernels for the ordinary spheroidal wave equation [5]


 

 dXx
d 
l2
c2 1  x2 k 
Xx 0:
1  x2
dx
dx
1  x2

Such kernels are obtained from the ones of the CHE through the substitutions

Xx zl=2 z  1l=2 Uz;

x 1  2z;

which give
2

zz  1

d U
2

dz

l 1 2l 1z


dU 
ll 1  k 4c2 zz  1 U 0;
dz

8a

that is, the CHE (1) with

B2 2B1 2l 1;

z0 1;

B3 ll 1  k;

x2 4c2 :

g 0;

8b

Thus, the spheroidal Eq. (6) will be treated as a CHE with z0 1; g 0 and B2 2B1 , namely,
2

zz  1

d U
2

dz

B1  2B1 z


dU 
B3 x2 zz  1 U 0:
dz

2.1. Transformations of the CHE and its kernels

Dening the operator Lz by

Lz zz  z0 


@2
@ 
B1 B2 z x2 zz  z0  2xgz
@z
@z2

10

and interpreting this as an ordinary differential operator, the CHE (1) reads

Lz B3 2gxz0 Uz 0:

11

The adjoint operator Lz corresponding to Lz is [24]

Lz zz  z0


@2
@ 
2z0  B1 4  B2 z x2 zz  z0  2xgz 2  B2 :
@z
@z2

12

On the other side, if Uz is a known solution of equation CHE, we seek new solutions Uz having the form

Uz

t2

Kz; tUtdt

t1

t2

wz; tGz; tUtdt

t1
B

wz; t t

1 z 1
0

t2

1 z 1
0

B2 z 1 1

t  z0

Gz; tUtdt;

13

t1

t  z0

B2 z 1 1
0

where the kernel Kz; t or Gz; t is determined from a partial differential equation. The general theory is usually established
for the function Kz; t [24], but to study the transformations of kernels we will deal with Gz; t. If the integration endpoints
t 1 and t 2 are independent of z, by applying Lz to integral (13) we nd

Lz Uz

t2

Lz Kz; tUtdt

t1

t2



Ut Lz  Lt Kz; tdt

t1

t2

UtLt Kz; tdt;

14

t1

Lt being obtained from Lz by replacing z with t. Now we demand that



Lz  Lt Kz; t 0

()

Lz  Lt Gz; t 0:

15

888

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

Thence, by using the Lagrange identity

UtLt Kz; t  Kz; tLt Ut

@
Pz; t;
@t

where the bilinear concomitant Pz; t is given by



@Kz; t
dUt
 B2  2t B1 z0 UtKz; t
Pz; t tt  z0 Ut
 Kz; t
@t
dt
B

 z1
0

B2 z 1

t  z0

Ut


@Gz; t
dUt
;
 Gz; t
@t
dt

16

Eq. (14) reduces to

Lz Uz



Z t2
Z t2
@Pz; t
@Pz; t
dt B3 2gxz0
Kz; tLt Ut
Kz; tUtdt
dt;
@t
@t
t1
t1

t2
t1

where in the last step we have used Eq. (11). Using Eq. (13) as well, this yields

Lz B3 2gxz0 Uz Pz; t 2  Pz; t1 :

17

Therefore, Uz is also a solution of the CHE if: (i) the kernel satises Eq. (15), (ii) the integral (13) exists and (iii) the righthand side of Eq. (17) vanishes.
Now let us examine the transformations of the solutions Uz and kernels Gz; t. If Uz UB1 ; B2 ; B3 ; z0 ; x; g; z denotes
one solution of the CHE, the following transformations [18,25,26] T 1 ; T 2 ; T 3 and T 4 leave invariant the form of the CHE:
B1

1B 

T 1 Uz z z0 UC 1 ; C 2 ; C 3 ; z0 ; x; g; z;
T 3 Uz UB1 ; B2 ; B3 ; z0 ; x; g; z;
where

C 1 B1  2z0 ;
D2 2  B2 

B1

T 2 Uz z  z0 2 z0 UB1 ; D2 ; D3 ; z0 ; x; g; z;
T 4 Uz UB1  B2 z0 ; B2 ; B3 2gxz0 ; z0 ; x; g; z0  z;

18




2B1
B1
B1
B2
;
;
C 3 B3 1
z0
z0
z0

B1 B1
D3 B3
B2  1 :
z0 z0

C 2 2 B2

2B1
;
z0

19

By composition of these transformations, from an initial solution we may generate a group containing up to 16 solutions. To
get the corresponding transformations for the kernels, we notice that the operators Lz and Lt which appear in the CHE (11)
and Lz  Lt Gz; t 0 have the same functional form. Hence, if Gz; t GB1 ; B2 ; z0 ; x; g; z; t is a solution of the Eq. (15), we
nd that the transformations R1 ; R2 ; R3 and R4 , given by
B

R1 Gz; t zt

1 z 1
0

GC 1 ; C 2 ; z0 ; x; g; z; t;

R3 Gz; t GB1 ; B2 ; z0 ; x; g; z; t;

R2 Gz; t z  z0 t  z0 

1B2  z 1
0

GB1 ; D2 ; z0 ; x; g; z; t;

20

R4 Gz; t GB1  B2 z0 ; B2 ; z0 ; x; g; z0  z; z0  t

do not change the form of the kernel Eq. (15). These transformations may generate a group containing up to 16 kernels when
applied to an initial kernel.
For another version of the CHE we have obtained initial kernels as limits of kernels for the general Heun equation [1]. For
the version (1), in the following we reobtain these kernels by solving Eq. (15) and use the transformations (20) to generate
groups of kernels closed under such transformations.
2.2. First group of kernels: products of conuent hypergeometric functions
For a particular problem, kernels given by products of conuent hypergeometric functions have already appeared in the
i;j

literature [27]. In the rst place we show that Eq. Lz  Lt Gz; t 0 is satised by 16 of such products, denoted by G1
dened as (i; j 1; 2; 3; 4)
i;j

 j f;
G1 z; t eixzt ui n  u

and

21

 j f are the conuent hypergeometric functions (A.2), having the following arguments and parameters:
where u n and u

ui n :

n

 j f :
u

2ix
z  z0 t  z0 ;
z0

2ix
zt;
z0

a k;

c

a
B1
;
z0

B2
 ig  k;
2

c B2

B1
;
z0

22a
22b

where k is an arbitrary constant of separation. In the second place, by the transformation R3 we may get another set of keri;j

nels, G2 , given by

889

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904
i;j

i;j

i;j

G2 z; t R3 G1 z; t G1 z; tjg;x#g;x ;

23

The transformations R1 ; R2 and R4 are superuous in this case.

To obtain the kernels (21), rst we write

Gz; t eixzt f z; t;

24

in Eq. (15). This leads to

zz  z0

 @f
@2f 
B1 B2 2ixz0 z  2ixz2
@z2
@z

tt  z0


@2f 
B2
2 @f
z  tf 0:

B

2i
x
z
t

2i
x
t

2i
x

i
g
1
2
0
@t
2
@t2

25

Then, by the substitutions

n

2ixz  z0 t  z0
;
z0

2ixzt
;
z0

f XnYf

26

we nd the conuent hypergeometric equations





B1
dX
B2
X 0;


n


i
g

k
2
dn
z0
2
dn2



B1
dY



f
 kY 0;
df
z0
df2

d X

d Y

27

where k is the constant of separation. The solutions for the above equations are: Xn ui n with a B2 =2  ig  k and
 j f with a k and c B1 =z0 . Inserting these solutions into (24) and (26) we nd the kernels
c B2 B1 =z0 ; and Yf u
(21). Thence, the kernels given by regular conuent hypergeometric functions are
1;1

G1


 

B2
B1
2ix
B1 2ix
z; t eixzt U
 ig  k; B2 ; 
z  z0 t  z0 U k;  ;
zt ;
2
z0
z0
z0 z0
ixzt2iz xzt

1;2

z; t e

2;1

z; t e

G1
G1

1 z 1
0

2;2

ixzt2iz xzz0 tz0

0

 U 1 ig k 

G1

zt

28

 

B2
B1
2ix
B1
2ix
 ig  k; B2 ; 
z  z0 t  z0 U 1  k; 2 ; 
zt ;
2
z0
z0
z0
z0

29

z  z0 t  z0 

1B2  z 1
0

 

B2
B1 2ix
B1 2ix
; 2  B2  ;
z  z0 t  z0 U k;  ;
zt ;
2
z0 z0
z0 z0

B1

30

B1

1B 

z; t eixzt zt z0 z  z0 t  z0  2 z0

 

B2
B1 2ix
B1
2ix
 U 1 ig k  ; 2  B2  ;
z  z0 t  z0 U 1  k; 2 ; 
zt :
2
z0 z0
z0
z0

31

The remaining kernels are obtained by replacing one or both functions U by W. In this manner we obtain the 16 kernels. The
transformations R1 ; R2 and R4 are superuous because they simply rearrange these kernels. For instance,
1;1

R1 G1

ixzt2iz xzt
0

1 z 1

zt

U 2

B1
B1
2ix
 k1 ; 2 ; 
zt
z0
z0
z0



B1 B2
B1
2ix
1  ig  k1 ; B2 ; 
z  z0 t  z0 ;
U
z0
2
z0
z0
1;1

where we have transformed k into k1 . By putting k1 k 1 B1 =z0 , we see that R1 G1

1;2

G1

2.3. Second group: product of hypergeometric and conuent hypergeometric functions

Now we nd a group of kernels Gi;j given by products of the four conuent hypergeometric ui given in (A.2) with the six
Gauss hypergeometric functions F j written in (A.5), (A.6) and (A.7). These kernels take the form

Gi;j z; t eixzt z t  z0 k ui n  F j f;

i 1; . . . ; 4; j 1; . . . ; 6

32

where k is a constant of separation, whereas the arguments and parameters for the hypergeometric functions are

ui n : n 2ixz t  z0 ; a

B2
 ig  k;
2

c B2  2k;

33a

890

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

F j f : f

zt
;
z0 z t  z0

a k; b B2  1  k;

c

B1
:
z0

33b

We can show that the transformations Ri do not generate new kernels.

The above kernels are constructed by inserting

Gz; t eixzt f z; t eixzt gn; f

34

into Lz  Lt Gz; t 0, where n and f are dened in Eqs. (33a) and (33b) . Thus we nd

"

@g
B2
@2g
B1
@g
n n 2 B2  n 
 ig g f1  f 2   B2 f
0:
@n
@f
2
z0
@n
@f
@2g

The separation of variables gn; f XnYf leads to

2

d X
dn2

B2  n


k
dX
B2

 ig  X 0;
n
dn
2

f1  f

d Y
df2



B1
dY 
B2 f
 kY 0;
df
z0

35

where 
k is a constant of separation. Putting 
k kB2  1  k, we nd that Yf is given by hypergeometric functions
Yf F j f as in Eqs. (32) and (33b), while Xn obeys the equation
2

d X
dn2

B2  n



dX
B2
kB2  1  k
X 0:

 ig 
dn
n
2

The substitution Xn nk Xx gives the conuent hypergeometric equation

2

d X
dn

B2  2k  n



dX
B2

 ig  k X 0;
dn
2

36

whose solutions are Xn ui n. In this manner, by inserting the previous solutions for Xn and Yf into

Gz; t eixzt z t  z0 k Xn Yf

37

we obtain kernels having the form (32).

The kernels G1; j and G2; j in terms of regular conuent hypergeometric functions are



B2
G1;j z; t eixzt z t  z0 k F j f U
 ig  k; B2  2k; 2ixz t  z0 ;
2

38



B2
G2;j z; t eixzt z t  z0 1B2 k F j f U 1 ig k  ; 2 2k  B2 ; 2ixz t  z0 ;
2

39

whereas G3; j and G4; j in terms of irregular functions are obtained by substituting Wa; c; u for Ua; c; u, that is,

G3; j z; t G1; j z; tjU#W ;

G4; j z; t G2; j z; tjU#W :

40

The functions F j f are given by



B1
zt
;
F 1 f F k; B2  1  k;  ;
z0 z0 z t  z0

41

1Bz1 

0
zt
B1
B1
B1
zt
F f
F k 1 ; B2  k; 2 ;
;
z0 z t  z0
z0
z0
z0 z0 z t  z0

42



B1 z  z0 t  z0
;
F 3 f F k; B2  1  k; B2 ;
z0 z0 z0  z  t

43


1B2 Bz 1 

0
z  z0 t  z0
B1
B1
B1 z  z0 t  z0
F f
F k  ; k 1  B2  ; 2  B2  ;
;
z0 z t  z0
z0
z0
z0 z0 z0  z  t

44

F 5 f


k 

z0 z t  z0
B1
z0 z t  z0
;
F k; k 1 ; 2 2k  B2 ;
zt
zt
z0

45

F 6 f


B 1k 

z0 z t  z0 2
B1
z0 z t  z0
:
F B2  k; B2  1  k; B2  2k;
zt
zt
z0

46

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

891

By using the explicit form for the kernels and the fact that the separation constant is arbitrary, it is possible to show that the
transformations Ri simply rearrange the previous kernels. For instance, we get



B2
R3 G1;j z; t eixzt z t  z0 k3 U
ig  k3 ; B2  2k3 ; 2ixz t  z0 H j f;
2
where H j f is obtained by substituting k3 for k in F j f. Thence, putting k3 B2  k  1 and taking into account that
Fa; b; c; u Fb; a; c; u, we nd that H5 f F 6 f; H6 f F 5 f and H j f F j f if j 1; 2; 3; 4. For this reason, R3 G1;j is
equivalent to G2;j .
2.4. Third group: conuent hypergeometric functions
An initial set has the form
i

G1 z; t eixzt ui n;

i 1; 2; 3; 4

47

where the ui n denote the four solutions (A.2) for the conuent hypergeometric equation with the following argument and
parameters:

n

2ix
z  z0 t  z0 ;
z0

B2
 ig;
2

c B2

B1
:
z0

48

The set (47) is obtained by putting k 0 and Y constant in (27). Besides this, from (47) we form four sets by using the
rules R2 and R4 , namely,
i

G1 z; t;

G2 z; t R1 G1 z; t;

G3 z; t R4 G2 z; t;

G4 z; t R2 G3 z; t:

49

The four pairs in terms of regular conuent hypergeometric functions Ua; c; u read



B2
B1
2ix
1
G1 z; t eixzt U
 ig; B2 ; 
z  z0 t  z0 ;
2
z0
z0


B
B2
B1 2ix
ixzt2izxzt
1B2  z 1
2
0 U 1 ig 
0 z  z t  z 
G1 z; t e
;
2

B

;
z

z
t

z

0
0
2
0
0 ;
2
z0 z0

B1 B2
B1
2ix
; B2 ; 
z  z0 t  z0 ;
z0
2
z0
z0


B
B
B1 B2
B1 2ix
ixzt2izxzt
1 z 1
1B2  z 1
2
0 z  z t  z 
0 U ig 
0 zt
G2 z; t e
 ; 2  B2  ;
z  z0 t  z0 ;
0
0
z0
2
z0 z0
1

G2 z; t eixzt zt

1 z 1
0

50

U 1  ig

51

B1 B2
B1 2ixzt
;
 ; ;
z0
z0
2
z0

B
B
B1 B2
B1
2ixzt
ixzt2izxzt
1 z 1
1B2  z 1
2
0 z  z t  z 
0 U ig
0 zt
;
G3 z; t e
;2 ;
0
0
z0
z0
2
z0

52

B2
B1 2ixzt
1
;
G4 z; t eixzt U
 ig;  ;
z0
2
z0

B
B2
B1
2ixzt
ixzt2izxzt
1 z 1
2
0 U 1 ig 
0 zt
G4 z; t e
:
;2 ;
z0
2
z0

53

G3 z; t eixzt z  z0 t  z0 

1B2  z 1
0

U 1  ig 

We get the pairs in terms of irregular conuent hypergeometric functions by replacing Ua; c; u by Wa; c; u:



3
1
Gi z; t Gi z; t

U!W



4
2
Gi z; t Gi z; t

U!W

; i 1; 2; 3; 4:

54

Thus, by using also R3 , we nd that this group is constituted by 32 kernels. Notice that R1 ; R2 and R4 generate only four sets of
kernels instead of 16 sets because in some cases these transformations rearrange the kernels of a given set in a different
i

order: we can test this by computing, for example, R2 G1 or R2 G2 . Notice that kernels whose arguments of the hypergeometric functions are 2ixzt=z0 have been known since long [28].
For the spheroidal equation g 0; z0 1; B2 2B1 , sixteen of the previous kernels reduce to four kernels in terms of
elementary functions, namely,


G1 z; t eixzt2ixzt ;

G2 z; t eixzt2ixzt ztz  1t  11B1 :



55

892

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

For instance,
1

G4 z; t / G4 z; t / G1 z; t;

G3 z; t / G3 z; t / G2 z; t:

G1 z; t / G1 z; t / G1 z; t;
G2 z; t / G2 z; t / G2 z; t;

The kernel G2 z; t will be used in Section 4.1.

2.5. Fourth group: conuent hypergeometric functions again
To obtain new kernels given by conuent hypergeometric functions we take



i
G1 z; t Gi;1 z; t

k0

where the Gi;1 denote the kernels (32) with j 1. Since the above choice for k eliminates the Gauss hypergeometric function
[F0; b; c; f 1] we nd
i

G1 z; t eixzt ui n;

i 1; 2; 3; 4

56a

where ui n denote the solutions (A.2) for the conuent hypergeometric equation with

n 2ixz t  z0 ;

B2
 ig;
2

c B2

see Eq: 33a:

56b

Other choices for k also lead to kernels in terms of conuent hypergeometric functions. However, such kernels are obtained
from the initial set (56a) by using the transformations Ri . In this manner we nd four sets, namely,
i

G1 z; t;

G2 z; t R1 G1 z; t;

G3 z; t R2 G2 z; t;

G4 z; t R1 G3 z; t;

57

since R4 does not generate new kernels. The kernels given by regular conuent hypergeometric functions are



B2
1
G1 z; t eixzt U
 ig; B2 ; 2ixz t  z0 ;
2


B2
2
ixzt
z t  z0 1B2 U 1 ig  ; 2  B2 ; 2ixz t  z0 ;
G1 z; t e
2

B1 B2
2B1
; 2 B2
; 2ixz t  z0 ;
z0
2
z0


B
2B
B1 B2
2B1
1 z 1
1B2  z 1
2
ixzt
0 U ig 
zt 0 z t  z0 
 ; B2 
; 2ixz t  z0 ;
G2 z; t e
z0
2
z0
1

G2 z; t eixzt zt

1 z 1
0

U 1  ig


B2
; 4  B2 ; 2ixz t  z0 ;
2


B
B
B
1 z 1
1B2  z 1
2
B2 3
ixzt
0 z t  z 
zt 0 z  z0 t  z0 
U ig  1 2 ; B2  2; 2ixz t  z0 ;
G3 z; t e
0
2
1

G3 z; t eixzt zt

1 z 1
0

58

1B2  z 1

z  z0 t  z0 

U 2  ig 


B1 B2
2B1
 ; 2  B2 
; 2ixz t  z0 ;
z0
2
z0


B
2B
B1 B2
2B
1B2  z 1
B2 z 1 1
2
ixzt
0 z t  z 
0
z  z0 t  z0 
U ig ; B2 1 ; 2ixz t  z0 :
G4 z; t e
0
z0
2
z0
B

1B2  z 1

G4 z; t eixzt z  z0 t  z0 

59

60

U 1  ig 

61

Replacing Ua; c; u by Wa; c; u and using the transformation R3 , once more we get a group with 32 kernels. Some particular
cases of these kernels are already known [4]. Furthermore, if g 0 this group can be expressed in terms of Bessel functions
by means of (A.12).
2.6. Fifth group: hypergeometric functions
i

To get kernels given by hypergeometric functions we take G1 z; t G1;i z; tjkB2 =2ig , where G1;i are the kernels given
in (38). In fact, for this choice for k we obtain U0; c; n 1 and, thence,
B2

G1 z; t eixzt z t  z0 ig 2 F i f;
i

f zt=z0 z t  z0 ;

62

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

893

where the hypergeometric functions F i f are obtained by putting k B2 =2  ig in Eqs. (41)(46). Explicitly
B2

G1 z; t eixzt z t  z0 ig 2 F
1

G1 z; t eixzt z t  z0 



B2
B2
B1
zt
;
 ig; ig  1;  ;
2
2
z0 z0 z t  z0

ig1 z 1  22
0

B2

G1 x; t eixzt z t  z0 ig 2 F
3



B1 B2
B1 B2
B1
zt
;
F 1  ig ; ig ; 2 ;
z0
2
z0
2
z0 z0 z t  z0



B2
B2
B1 z  z0 t  z0
;
 ig; ig  1; B2 ;
2
2
z0 z0 z0  z  t

ig1 1

1 z 1

zt 

B2

1B 

B2

64

65

B1

z0
2
G1 x; t eixzt z t  z0 
z  z0 t  z0  2 z0


B1 B2
B1 B2
B1 z  z0 t  z0
;
 F ig   ; 1  ig   ; 2  B2  ;
z0
2
z0
2
z0 z0 z0  z  t

G1 x; t eixzt zt ig 2 F

63


B2
B1 B2
z0 z t  z0
;
 ig; 1  ig ; 2  2ig;
zt
2
z0
2



B2
B1 B2
B2
z0 z t  z0
6
:
G1 x; t eixzt z t  z0 2ig1 zt1ig 2 F ig ; ig  1 ; 2ig;
zt
z0
2
2

66

67

68

The transformations R1 ; R2 and R4 at most rearrange the preceding kernels. For example,
1

R1 G1 z; t G1 z; t;

R2 G1 z; t G1 z; t;

R4 G1 z; t G1 z; t:

However, we nd six additional kernels G2 z; t by using the transformations R3 as

i

G2 z; t R3 G1 z; t
i

69
i

So, G2 is obtained by replacing g; x by g; x in G1 .

3. Integral relations between known solutions
In this section we use some kernels to obtain integral relations among solutions of the CHE. We nd that:
 The Jaff solutions in power series are tranformed into Leavers expansions in series of irregular conuent hypergeometric
series.
 The BaberHass solutions in power series are transformed into solutions given by series of regular conuent hypergeometric functions.
Relations for solutions generated by transformations of the CHEs may be obtained by transforming also the kernel, since each
transformation of a solution corresponds to a transformation of a kernel.
3.1. Jaffs solutions in power series and Leavers solutions
By U J1 z and U L1 z we denote respectively the Jaff [22] and the Leaver [9] solutions for the CHE, namely,
B2

U J1 z eixz zig 2

U L1 z eixz

1
z  z n
X
0
a1n
;
z
n0

1
X
B1
B
B
a1n C n B2
W n ig 2 ;  1 ; 2ixz ;
z0
2
z0
n0

where the recurrence relations for the a1n are (a11 0)






B1 1
B1
B1
B2
an1 2n n B2 ig  ixz0 B3 B2
ixz0  ig 
a1n
n 1 n B2
z0
z0
z0
2



B2
B2 B1
n ig a1n1 0:
n  1 ig
2
2 z0

70a

70b

894

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

The convergence of solutions (70a) and (70b) is discussed in the Leaver paper [9], where it is used the minimal solution
for the coefcients a1n . In fact, three-term recurrence relations as the above ones admit two independent solutions, say, f n and
g n . If limn!1 f n =g n 0; f n is called minimal solution [29,30]. In addition, it is necessary to suppose that there is an arbitrary
parameter in the CHE. The series converges only for special values of that parameter, determined from a transcendental
(characteristic) equation which results from the recurrence relations [9].
By supposing that U J1 z converge for jzj P jz0 j and by using Eq. (13), we nd the relation

U L1 z C 1

1 z 1
0

t  z0 

B2 z 1 1
0

z0

G4 z; tU J1 tdt;



B1
> 0;
Re n B2
z0

Reixz < 0;

71

where C 1 is a constant and Gz; t G4 z; t is the kernel indicated in (54). In fact, by setting y t=z0 , we nd that the righthand side of (71) is equivalent to

eixz z

1 z 1
0

Z
1
X
a1n
n0


B
B2
B2
B1
nB 1 1
dy e2ixzy y  1 2 z0 ynig 2 W 1 ig  ; 2 ; 2ixzy :
2
z0

1
z0

Then, by using the integral [31]

1
1
1
eay y  1l1 yakl2 W
a  k; 2a 1; ay dy Clea W
a l  k; 2a 1; a ;
2
2

Re

l > 0; Re a > 0;

72

we obtain the relation (71).

For the bilinear concomitant (16) we nd

n
1
X
t  z0
an
t
n0




z
nz0
1
B1 B2
@W
;
ig 1
 2ix
1 t

Wt
z0
t  z0 t
z0
2
@t

P1 z; t z

B
1 z 1 ixz
0

2t
z0 1

B2

tig 2 t  z0

B2 z 1
0

where

W W 1 ig 

B2
B1
2ixzt
;
;2 ;
z0
2
z0

@ W 2ixz
B2
B
B
2ixzt
:

1 ig 
W 2 ig  2 ; 3 1 ; 
z0
z0
@t
2
2
z0

Since Wa; b; y ya when jyj ! 1 and Reixz < 0, the exponential factor assures that P 1 z; t vanishes when t=z0 ! 1. On
h
i
the other hand, the condition Re B2 Bz01 > 0 assures that P 1 z; t vanishes also for t z0 since t  z0 B2 B1 =z0 ! 0.
In this manner, we have extended the results of Leaver [9] who has considered only relations between solutions with
ig B2 =2  1. Notice also that the conditions given in (71) are necessary only to assure the integral relation between
the solutions. In fact the Leaver solutions can be derived directly from the differential equation without imposing those conditions [9].


For the present case the transformation T 1 is ineffective and, so, from U J1 ; U L1 we can obtain only 8 pairs of solutions by
composition of the transformations (18); to each pair corresponds a kernel generated by the transformations (20). For example, taking U J2 z T 2 U J1 z and U L2 z T 2 U L1 z, we nd
B2

U J2 z eixz z  z0 1B2  2 z
B2

U L2 z eixz z  z0 1B2  2

ig1 z 1 22
0

1
z  z n
X
0
a2n
;
z
n0

73a

1
X
B1
B
B
B
a2n C n 2  B2 
W n ig 1  2  1 ;  1 ; 2ixz ;
z0
2 z0
z0
n0

where the recurrence relations for a2n are (a21 0)



B1 2
a
n 1 n 2  B2 
z0 n1



B1
B1
B1
B2
B1 B2
B3 2  B2
1 
a2n
2n n 2 ig  ixz0  B2 
ixz0  ig 1 
z0
z0
z0
2
z0
2



B2
B2 B1 2
n ig  
a
n 1 ig 
0:
2
2 z0 n1

Using Eq. (13), we nd that

U L2 z C 2

z0

dt t

1 z 1
0

B2 z 1 1

t  z0 

U J2 tR2 G4 z; t;



B1
> 0;
Re n 2  B2 
z0

Reixz < 0;

73b

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

895

where C 2 is a constant, G4 z; t is the kernel indicated in (54), and the transformation R2 is given in (20); then,
ixzt2izxzt

R2 G4 e

z  z0 t  z0 

1B2  z 1
0

zt

1 z 1
0

W ig

B1 B2
B1
2ixzt
:
;2 ;
z0
z0
2
z0

We have supposed that the Jaff solutions converge for jzj P jz0 j, but we must be careful about the point z 1, since [9]
B2

lim U J1 z eixz zig 2

z!1

1
X

a1n

a1n1
1
a1n

with

n0

p
2ixz0 ig  xz0  3=4
p

n
n

74

where the ratio a1n1 =a1n holds for the minimal solution of the recurrence relations when n ! 1. Thus, the DAlambert test is
P 1
inconclusive as to the convergence of
an . For the radial part of the two-center problem we could use the Raabe test for
convergence, as in Eq. (106).
3.2. Solutions in power series and solutions in series of conuent hypergeometric functions
We nd another pair of solutions for the CHE which are again connected by the integral (13). By one side we have the
Baber-Hass expansion [18,23]

U baber
z eixz
1

1
X
a1n z  z0 n ;

75a

n0

where the coefcients satisfy the relations (a11 0)

B1
B2
z0 n B2
n 1a1n1 b1n a1n 2ix n ig  1 a1n1 0;
z0
2

75b

with b1n nn B2  1 2ixz0 B3 ixz0 B2 B1 =z0 . The minimal solutions for a1n yield solutions convergent for any nite
value of z. On the other side, if B2 =2  ig is not zero or negative integer we have the solution [18]

U 1 z eixz

B2
1
bn U
 ig; n B2 ; 2ixz ;
2
n0

1
X

76a

where the recurrence relations for bn are obtained from the previous ones by taking
1

bn

Cz0 n Cn B2 B1 =z0 1
an ;
Cn B2

C constant:

This yields
1

n B2 n 1bn1 b1n bn  2ixz0




n B2 Bz01  1 n ig B22  1
n B2  1

bn1 0:

76b

Now, if we insert U baber

t and the kernel G4 z; t given in (53) into Eq. (13), we nd the solution U 1 z, that is,
1

U1 K

t2

1 z 1
0

t  z0 

B2 z 1 1
0

t1

G4 z; tU baber
tdt;
1



B1
> 0;
Re n B2
z0



B1
Re 
> 0;
z0

77

where K is a constant. In effect, by taking t1 0 and t2 z0 , the above integral is proportional to

Z
1
X
eixz z0 n an1

n0

2
3


1Bz 1

nB2 1Bz1

0
0
t 4 t
t
B2
B1
t 5
d
1
U
 ig;  ; 2ixz
z0
z0
z0
z0
2
z0

Then, by using the relation [31]

Z
0

1
CkC1 2l  k
1
xk1 1  x2lk U
l  m; k; yx dx
U l  m; 1 2l; y ;
2
2
C1 2l

Rek > 0;

78

Re1 2l  k > 0;

we nd the solution U 1 z given in (76a) provided that Ren B2 B1 =z0  > 0 and ReB1 =z0  > 0. On the other side, from
dUa; b; n=dn a=bUa 1; b 1; n for n 2ixzt=z0 ; a B2 =2  ig and b B1 =z0 , we nd that the bilinear concomitant
(16) is given by

P1 z; t eixz t

z1
0

B2 z 1

t  z0

Ua; b; n

)

X
1
1
X
az
na1n t  z0 n1 2ix Ua; b; n 
Ua 1; b 1; n
a1n t  z0 n :
bz0
n1
n0

Therefore, P1 z; t 0 P1 z; t z0 0 due to the conditions ReB1 =z0 > 0 and ReB2 B1 =z0 > 0.

896

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904



Observe that from the pair U baber
; U 1 we can obtain 16 pairs of solutions by using the four transformations (18) and
1
composition of them: to each pair corresponds a kernel which is obtained by using the transformations (20).
4. New solutions for the conuent equation
In Section 4.1, by an integral transformation we nd a new solution in series of irregular conuent hypergeometric functions for the ordinary spheroidal equation. Then, in Section 4.2 we extend that solution to the general case (no restriction on
the parameters of the CHE). In this manner, we obtain an initial solution, U1 z, which allows to generate a group of solutions
Ui z for the CHE by by means of transformations (18). Finally, in Section 4.3, we show that the new solutions are suitable for
the radial part of the two-center problem of the quantum mechanics.
Initially we make some comments on the recurrence relations and the ratio test for convergence. As in the preceding seci

tion, the three-term recurrence relations for the series coefcients bn of Ui z have the form

ai0 bi1 bi0 bi0 0;

ain bin1 bin bin cin bin1 0 n P 1

79

where ain ; bin and cin depend on the parameters of the differential equation and on the summation index n. By omitting the
superscripts, these relations take the form

b0

6c
6 1
6
60
6
6
6
6
6
6
6
6
6
4

32

a0

b1

a1

c2 b2 a2
..

..

..

cN

bN

aN

cN1 bN1 aN1

..

..

2 3
0
76
7 6 7
76 b1 7 6 0 7
76
7
7
76 b2 7 6
76
7 607
7
76 . 7 6
6
76 . 7 6 .. 7
76 . 7 6 . 7
7;
76
7
7
76 bN 7 6
0
6
76
7 6 7
76
7 6 7
76 bN1 7 4 0 7
5
54
5
..
..
..
.
.
.
b0

80

where we have split the matrix into blocks. This system of homogeneous linear equations has nontrivial solutions for bn only
if the determinant of the above tridiagonal matrix vanishes: this demands some arbitrary parameter in the matrix elements
and, as a consequence, in the differential equation. The condition on the determinant can also be expressed by an (characteristic) equation given by the continued fraction [9]

b0

a0 c1 a1 c2 a2 c3
b1  b2  b3 

81

The solution of the characteristic equation and the computation of the series coefcients are important aspects concerning
applications of solutions of the CHE [32,33]. However, we do not consider these questions connected with numerical study of
solutions. We only mention that the problem is simplied if cinN1 0 for some N P 0; then, the series terminates at n N
leading to a nite-series solution with 0 6 n 6 N (see page 146 of [34]) which is called polynomial or quasi-polynomial solution. In this case, only the left upper block of the matrix is relevant.
P
On the other side, the convergence of a series like 1
n0 f n z is obtained by computing the limit



f z
Lz lim  n1 :
n!1 f z

82a

By the DAlembert ratio test the series converges in the region where Lz < 1 and diverges where L1 z > 1. If Lz 1, the
DAlembert test is inconclusive; however, by the Raabe test [35,36], if

Lz 1

A
1
O 2
n
n

82b

(where A is a constant) the series converges if A < 1 and diverges if A > 1; the test is inconclusive if A 1.
4.1. An integral transformation for the spheroidal equation
For the spheroidal equation in the form (9) we will nd a solution U1 z given by

U1 z eixz z1B1 z  11B1

1
X
1
bn W2 B1 ; 2 B1  n; 2ixz;

B1  2; 3; . . .

83a

n0
1

where the coefcients bn satisfy the relations (b1 0)

1

2ixn 1bn1 nn 1 2ix ix2 B1  B1 1 B1 B3 bn  nn B1 1bn1 0:

83b

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

897

U1 z is not valid if B1 2; 3; . . ., because in these cases the function Wa; c; y becomes a polynomial of xed degree and,
accordingly, (83a) is not a series expansion. This follows from the relation [37]

Wl; a 1; y 1l l! Lal y;

l 0; 1; 

84

where the Ll y denote Laguerre polynomials of degree l. Besides this, the above expansion in general does not hold at z 0
because in most cases Wa; c; y goes to innity at z 0 [5]. The convergence of W for z 0 will be discussed later on.
We get the expansion (83a) by applying an integral transformation to the asymptotic expansion W2 z given in Eq. (B.4).
1

First, for the spheroidal equation, by writing Wz W2 z and a2n bn , we nd

Wz eixz z  11B1

1
X
1
bn zn1 ;

Eq:B:4for the spheroidal equation

85

n0
1

where the coefcients bn satisfy (83b). In the second place, the solution U1 is obtained by inserting Ut Wt and
Gz; t G
2 z; t see Eq. (55) into the right-hand side of Eq. (13), and by integrating from t 1 to t 1, that is,
13

U1 z

55

t 1B1 t  11B1 G2 z; t Wtdt eixz zz  11B1

eixt2ixzt Wtdt

which gives
85

U1 z eixz zz  11B1

Z
1
X
1
bn

e2ixzt t  11B1 t n1 dt:

86

n0

Thence, we obtain (83a) by using [5]

eyt t  1a1 tca1 dt Caey Wa; c; y;

Re a > 0; Re y > 0:

The integrability conditions on the right-hand side require that

Re2 B1  > 0andReixz < 0:

87

On the other side, the bilinear concomitant (16) reads



@G z; t
dWt
;
Pz; t t B1 t  1B1 Wt 2
 G2 z; t
@t
dt
(
ixz2t1

zz  1

1B1

t  1

2B1

1
1
X
X
1
1
2ixz  1t B1  bn t n1
nbn tn1
n0

)
88

n0

Since the series converge at t 1, the conditions (87) assure that Pz; t 1 0. However, the concomitant is undetermined at t 1 because [for Re2 B1 > 0] Pz; t is given by the product of the vanishing factor t  12B1 by a divergent
series. Despite this, we can check directly [38] that U1 z is indeed a solution of the spheroidal Eq. (9).
Now we use the ratio test to get the convergence of U1 . Thus, when n ! 1, we nd that the minimal solution of (83b)
satises [38]
1

bn1
1

bn

1

B1
b
B1
) n1
1 :
1
n
n
bn

89

To get the ratio between successive W, we use the relation [5]

a 1  cWa; c  1; y c  1 yWa; c; y  yWa; c 1; y 0:

Hence, by taking

a 2 B1 ;

c 2 B1  n;

y 2ixz;

Wn y W2 B1 ; 2 B1  n; 2ixz

we obtain

n 1

Wn1
Wn1
 n  1  B1 2ixz 2ixz
0:
Wn
Wn

If z is bounded (that is, if 2ixz=n ! 0), then when n ! 1 this equation is satised by

Wn1
1
Wn1
1
 1  B1 2 ()
 1 B1 2 or
n
n
Wn
Wn



Wn1 2ixz
B1
Wn1
n
1
()

1

1  1 B1 :
n
2ixz
n
Wn
n
Wn

90

898

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

Only the rst ratio is consistent with the fact that, if jcj ! 1 while a and y remain xed and bounded, then [37]

"

Wa; c; y  ca 1a

#
p
ca3


2
1
2p c
1
3
;
ya2 eya2 1 O
jcj
Ca ey

c ! 1; a 0; 1; 2; . . . ; j argcj < p:

Thus, using (89) and (90), we nd

1

lim

n!1

bn1 Wn1
1
bn

Wn

1

2
1
for U1 :
O 2
n
n

91a

Therefore, by the Raabe test the series may converge for any nite value of z (the ratios (90) are valid if z is nite); however,
we must exclude the point z 0 because in general the function Wa; c; y goes to innity at y 0. On the other side, from
limy!1 Wa; c; y ya , we nd

b 
Re B1
1
 n1 89
O 2 ;
 1 1
 b 
n
n
n

1
X
1
z
bn ;

ixz

limU1 z e

B1

z!1

n0

Thus, according to the Raabe test again, the series

n ! 1:

91b

bn converges at z 1 if ReB1 < 1.

4.2. Solutions for the conuent Heun equation

Now the solution U1 z for the spheroidal equation, given in (83a), is extended for any CHE. In fact, we can construct a
i

group of solutions Ui z whose series coefcient bn satisfy the relations (79). To this end, in the right-hand side of (83a)
we perform the substitutions
B

1 z 1

z1B1 z  11B1 # z

1B2  z 1

z  z0

W2 B1 ; 2 B1  n; 2ixz # Wa; b  n; 2ixz;

where, in the substitutions of the rst line we have used 1 B1 =z0 and 1  B2  B1 =z0 because these are indicial exponents at
z 0 and z z0 , respectively. By using the properties of Wa; c; y we nd that a 2 ig  B2 =2 and b 2 B1 =z0 [38]. Thus,
U1 is given

U1 z eixz z

1 z 1
0

1B2  z 1

z  z0 

1
X
B2
B1
1
bn W 2 ig  ; 2  n; 2ixz ;
2
z0
n0

ig  B2 =2  2; 3; 

92a

where the coefcients bn satisfy the recurrence relations (79) with [38]


 


2B1
B1
B1
2  B2 
2  B2 B3 ;
b1n n n 1  B2 
2ixz0 ixz0  1 
z0
z0
z0



B
B
B
c1n  n ig  1  2 n 1  B2  1 :
z0
2
z0

a1n 2ixz0 n 1;

92b

By the transformations (18), U1 produces a group constituted by 16 solutions, Ui . Eight of these can be constructed as

U1 z;

U2 z T 1 U1 z; U3 z T 2 U2 z; U4 z T 1 U3 z;

93

U5 z T 4 U1 z; U6 z T 4 U2 z; U7 z T 4 U3 z; U8 z T 4 U4 z;
while the others result by the transformation T 3 which changes (g; x) by (g; x) in the above solutions. Thus,

U2 z eixz z  z0 

1B2  z 1
0

1
X
B2 B1
B1
2
bn W 1 ig   ;   n; 2ixz ;
2 z0
z0
n0

ig 

B2 B1

 1; 2; . . .
2 z0


94a

where, in the recurrence relations (79) for bn ,



B1
2  B2 B3 ;
b2n nn 3  B2 2ixz0  ixz0 2  B2 
z0



B
B
c2n  n ig 1  2 n 1  B2  1
2
z0

a2n 2ixz0 n 1;

94b

The third solution reads

U3 z eixz

1
X
B2
B1
3
bn W ig ;   n; 2ixz ;
2
z0
n0



B2
ig
0; 1; . . . ;
2

95a

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

899

with


 


2B1
B1
B1
1 ixz0 B3 ;
b3n n n 1 B2
2ixz0 B2
z0
z0
z0



B
B
B
c3n  n ig 2 1 n  1 B2 1 :
2 z0
z0

a3n 2ixz0 n 1;

95b

At last, we write

U4 z eixz z

1 z 1
0

1
X
B2 B1
B1
4
bn W 1 ig ; 2  n; 2ixz ; ig B2 =2 B1 =z0  1; 2; 
2 z0
z0
n0

96a

with



B1
B3 ;
b4n nn  1 B2 2ixz0  ixz0 B2
z0



B
B
c4n  n ig  1 2 n  1 B2 1 :
2
z0

a4n 2ixz0 n 1;

96b

The relation (91a) is valid also for the present case, whereas (91b) is replaced by [38]
ix z

lim U1 z e

z!1

ig 22

1
X
1
bn ;
n0

b 
1
B2
1



O
when n ! 1:
with  n1

Re
i
g


 b1 
n
n2
2

97

Then, U1 converges at z 1 if Reig  B2 =2 < 1. By using the transformations as in (93), we nd that the Ui converge for
nite values of z, excepting possibly the points z 0 (if i 1; 2; 3; 4) and z z0 (if i 5; 6; 7; 8). According to the Raabe test,
these Ui converge also at z 1 if



B2
Re ig  1 < 0 : U1 ; U5 ;
2


B2
Re ig  1 < 0 : U3 ; U7 ;
2



B1 B2
< 0 : U2 ; U8 ;
Re ig  
z0
2


B1 B2
< 0 : U4 ; U6 :
Re ig
z0
2

98

4.3. The radial part of the two-center problem

Now we consider the equations of the two-center problem of quantum mechanics, as the one describing the electron of
the ionized hydrogen molecule. Using Leavers conventions [9], the wave function w of the time-independent Schrdinger
equation for an electron in the eld of two Coulombian centers has the form

w eimu Rk Sl;

r1 r2
;
2a

r1  r2
;
2a

m 0; 1; 2; . . . ;

99

where r 1 and r 2 are the distances from the electron to the two nuclei, and 2a the intercenter distance. By the denitions
m

Sz Sk z 2 2  z 2 U  z;
m
2

m
2

Rz Rl z z  2 U z;

z l 1;
z k 1;

0 6 z 6 2;

100

z P 2;

Leaver obtained CHEs in the form (1) for U  , with the parameters g for g , B
3 for B3 )

x2 2a2 E;

z0 2;
B3

xg aN1  N2 ;

B1 2m 1;

B2 2m 1;

x 2aN1  N2 mm 1  Alm :

101

where Alm is a separation constant, and N 1 and N 2 are the charges on the two nuclei. Thus, there are two CHEs, one for the
angular coordinate l and one for the radial coordinate k. Each CHE is associated with a characteristic Eq. (81) which
determines the possible values of the constants Alm and E.
Now we consider Rz, the radial solution given in (100). For bound states (E < 0) we take

p
ix a 2jEj

.p
ig ig N1 N2
2jEj

102

in order to assure that the factor expixz remains nite when z ! 1. Then, if jEj is nite,





B1 B2
B1 B2
N1 N2
Re ig
 p < 0;
Re ig  
z0
2
z0
2
2jEj
and, consequently, four of the solutions listed in (98) converge at z 1. To get wavefunctions bounded also at z 2, we
select U2 if m 6 0 and U4 if m P 0. Thus, we nd

900

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

Rz ea

jmj

jmj

2jEj z

z 2 z  2 2

1
X
p
N1 N2
2
bn W 1  p ; 1  jmj  n; a 8jEj z
2jEj
n0

!
103a

where the coefcients bn satisfy the relations (79) with

h
h
pi
pi
b2n n n 1 2jmj  2a 8jEj jmj 1 jmj  a 8jEj 2aN1 N2  ajEj  Alm ;
#
N1 N2
2
p

cn n jmj n jmj 
:
103b
2jEj

a2n

p
8jEj an 1;
"

The expansion (103a) holds only if

p
N1 N2 = 2jEj l 1;

l 0; 2; . . .

104

a condition which assures that Wa; b; y is not a polynomial of degree l in y.

The condition (104) is also required by the Jaff expansions. In effect, by using the solutions U J1 (if m P 0) and U J2 (if
m 6 0) given in Eqs. (70a) and (73a), respectively, we nd

RJ z ea

2jEj z

1

jmj N1 N 2
p
2
2jEj

jmj

z  2 2

n
1
X
z2
a1n
;
z
n0

105a

where the recurrence relations for a1n have the form (79) with

a1n n 1n jmj 1;

"
#
"
#
p N1 N 2
p
N1 N2
b1n 2n n 1 jmj a 8jEj  p jmj 1 p  a 8jEj  1 2aN 1 N2  ajEj  Alm ;
2jEj
2jEj
"
#"
#
N1 N2
N1 N2

n  p :
c1n  n jmj  p
2jEj
2jEj

105b

p
Thus, cl1 0 if N 1 N 2 = 2jEj l 1 and, then, RJ z becomes a nite-series solution with 0 6 n 6 l, as stated after Eq.
(81). In this case, the constant Alm would be determined from the characteristic equation associated with the recurrence relations for a1n . However, if E and Alm are both determined from the radial solution, we cannot satisfy the characteristic equation
corresponding to the angular solutions (these are usually given by series where the summation begins at n 0 and, so,
present recurrence relations having the form (80)). Therefore, also for the Jaff solutions it is necessary that
p
N 1 N 2 = 2jEjl 1. The same is true respecting Hylleraas expansions in series of Laguerre polynomials [9,39].
The convergence of solution (103a) follows immediately from the Raabe test. As to the Jaff solution (105a), we have to
examine its behavior at z 1. By using (101) together with (102), the expressions (74) imply that
J

limR z e

z!1

a

2jEj z

1
p 1 X 1
2jEj
an ;
n0

N1 N 2

"
#

q
p
p N1 N2
a1n1
1 3
1
p

O
:

1


2
an
2jEj

2a
2jEj

n 4
n2
a1n
2jEj

106

Then, by a convenient choice of n, the constant A which appears in (82b) becomes less than 1 and so, by the Raabe test, the
solution converges at z 1.
5. Final remarks: conclusion and open problems
By inserting a suitable weight function wz; t into the integral relation (13) we have found the kernel Eq. (15) where the
differential operators Lz and Lt have the same functional dependence as the operator of the CHE (11), a fact which allows to
get transformations of the kernels by examining the known transformations of the solutions for the CHE. As mentioned, this
is an extension of a similar correspondence found in 2011 for the general Heun equation (HE) [1].
Actually, in 1942 Erdlyi used the appropriate weight function for the HE but he could not infer how to transform the
kernels because the transformations of the HE were fully established only in 2007 [3]. On the other side, transformations
of conuent Heun equations are known since 1978 [25,26] but have not been applied to transform kernels see, for example,
references [4,6,7,27]. In the present study we have considered transformations of kernels of the CHE, where the initial kernels (to be transformed) come from kernels of the HE by a process of conuence [1]; however, for the sake of completeness,
in Section 2 we have reobtained them by solving the kernel equation.
By separation of variables we have found two groups of kernels presenting an arbitrary constant of separation. One group,
with products of two conuent hypergeometric functions, includes some particular kernels already known in the literature
[27]; the other group, with products of conuent hypergeometric functions and Gauss hypergeometric functions, is new as
far as we known. For particular values of the constant of separation we have obtained three groups given by produtc of ele-

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

901

mentary functions with one special function: this is represented by conuent hypergeometric functions (two groups) and by
Gauss hypergeometric functions (one group).
In Section 3 we have found some integral transformations among known solutions of the conuent Heun equations. We
have used two singularities as endpoints of integration and have supposed that the solutions to be transformed are convergent at both endpoints (this assures that the bilinear concomitants vanish there). If the solutions are modied by the rules
(18), the kernels must be modied by the rules (20). This emphasizes that the correspondence between the transformations
of the Heun equations and of the respective kernels are connected parts of the transformation theory.
The applications of Section 3 simply relate known solutions without affording new ones. In contrast, in Section 4, by
means of an integral transformation we have obtained a new solution for the spheroidal wave equation, which in turn leads
to a group of new solutions for the CHE. We have seen that these solutions may be used to compute the radial part of the
wavefunctions for bound states of hydrogen moleculelike ions and, by this reason, can play the role of the expansions in series of Laguerre polynomials proposed by Hylleraas in 1931 [39] and the Jaff power-series solutions which have been used
from 1934 [22] up to now [40].
It is possible to realize further properties of the solutions by considering other problems, as the Lorentzian model of a
quantum two-state system given by Ishkhanyan and Gregoryan [11]. This is ruled by a CHE with z 1 it=2 and z0 1,
where t denotes the time. According to the authors, for certain values of a parameter R, the problem admits nite-series solutions which are bounded for any admissible value of t and assure that the system returns to the initial state after the interaction. By using the solutions of Section 4.2, we have veried that the previous statement is correct; it seems that no other
known solution of the CHE permits to prove the statement. In addition, since in this case jzj P 1=2, it would be interesting to
check if the solutions of Section 4.2 lead to innite-series solutions suitable for some range of the parameter R (the Hylleraas
and Jaff solutions do not converge for jzj < 1).
We have omitted details concerning the derivation of the new solutions of the CHE. In addition, the solutions must be
improved as follows: (i) by considering also expansions in series of regular conuent hypergeometric functions to get solution valid in the neighborhood of z 0 [38], (ii) by using the WhittakerInce limit as in Ref. [18] to obtain solutions for the
RCHE (2), (iii) by inserting a characteristic parameter m and letting that the series summation runs from minus to plus innite [38] (two-sided series) in order to obtain solutions for a CHE without free parameters.
As mentioned in the rst section, along with the RCHE we have disregarded two other equations which are associated
with the CHE by formal limits. These are the double-conuent Heun equation (DHE) and the reduced DHE (RDHE) which
appear when we allow that z0 ! 0 in the CHE and RCHE, respectively, that is,

DHE :



2
z2 ddzU2 B1 B2 z dU
B3  2gxz x2 z2 U 0; B1 0; x0
dz
2

RDHE : z2 ddzU2 B1 B2 z dU
B3 qzU 0;
dz

q 0; B1 0

where now z 0 and z 1 are irregular singularities. At z 1 the behaviour is again given by Eq. (5), that is,

lim Uz  eixz zigB2 =2 for the DHE 1;

z!1

limUz  e2i

z!1

p
qz

z1=4B2 =2 for the RDHE 2;

while at z 0 the normal Thom solutions affords [18]

limUz  1
z!0

or

limUz  eB1 =z z2B2

z!0

for DHE and RDCE:

We show elsewhere [41] that the application of the WhitakkerInce limit (3) and the Leaver limit (z0 ! 0) on kernels of
the CHE lead to new kernels for the RCHE, DHE and RDHE. This is in accordance with a previous conjecture [1] but, by integrating the kernel equations, we also nd kernels which are not connected with known kernels of the CHE: for the RCHE we
obtain a group of kernels expressed by products of Bessel and hypergeometric functions, while for the DHE and RDHE we
obtain kernels given in terms of elementary functions. Therefore, in addition to the use of limiting procedures, we can solve
the kernel equations to generate other kernels for the RCHE, DHE and RDHE.
Appendix A. Hypergeometric functions
The regular and irregular conuent hypergeometric functions are denoted by Ua; c; u and Wa; c; u, respectively. They
satisfy the conuent hypergeometric equation [37]
2

d uu
2

du

c  u

duu
 auu 0
du

which admits the solutions

A:1

902

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

u1 u Ua; c; u; u2 u eu u1c U1  a; 2  c; u; u3 u Wa; c; u; u4 u eu u1c W1  a; 2  c; u:

A:2
All of them are dened and distinct only if c is not an integer. Alternative forms for these solutions follow from the relations

Ua; c; u eu Uc  a; c; u;

Wa; c; u u1c W1 a  c; 2  c; u:

A:3

On the other side, solutions for the (Gauss) hypergeometric equation [37],
2

u1  u

d F
du

c  a b 1u

dF
 abF 0;
du

A:4

are given by hypergeometric functions Fa; b; c; u Fb; a; c; u. In fact, in the vicinity of the singular points 0, 1 and 1, the
formal solutions for the hypergeometric Eq. (A.4) are, respectively,

F 1 u F a; b; c; u;

F 2 u u1c F a 1  c; b 1  c; 2  c; u;

F 3 u F a; b; a b 1  c; 1  u;

F 4 u 1  u

1
F 5 u ua F a; a 1  c; a 1  b; ;
u

cab

A:5

F c  a; c  b; 1 c  a  b; 1  u;

A:6

1
F 6 u ub F b 1  c; b; b 1  a; :
u

A:7

Each of these may be written in four forms by using the relations

Fa; b; c; u 1  ucab Fc  a; c  b; c; u;

Fa; b; c; u 1  ua F a; c  b; c; u=u  1:

A:8

On the other side, the usual form for the Bessel equation is [5]
2

y2

d Zy
dy


dZy  2
y  a2 Zy 0:
dy

A:9

The solutions for this equation are denoted by Z j

a y according as [5,34]

Z 1
a y J a y;

Z 2
a y Y a y;

1
Z 3
a y Ha y;

2
Z 4
a y Ha y

A:10
1

where J a y and Y a y are the Bessel functions of the rst and second kind, respectively; Ha y and Ha y are the rst and
the second Hankel functions. There are formulas connecting these functions ([5]). For example,

Ya

i cosapJ  J
1 h 1
a
a
Ha  H2
:

a
2i
sinap

A:11

Bessel and conuent hypergeometric functions are connected by [37]

1
2

U a ; 2a 1; 2iy Ca 1 eiy

 y a

J a y;

2
p
i p iyap
e
2ya H1

a y;
2

1
2

W a ; 2a 1; 2iy

1
2

W a ; 2a 1; 2iy 

eiyap 2ya H2
a y:

A:12

Appendix B. Wilsons asymptotic expansions for the CHE

Such solutions were considered in 1928 by Wilson [10]. Actually they are given by 8 asymptotic Thom expansions [17]
which we denote by Wi z (i 1; 2; 3; 4) and T 3 Wi z. For the CHE in the form (1) we nd
B2

W1 z eixz zig 2

1
X
a1n zn

B:1

n0

where the coefcients a1n satisfy the three-term recurrence relations (a11 0)


B1
B2
B2
B3
a1n
2ixn 1a1n1  nn 1 2ig 2ixz0 ixz0 B2
ig 1 ig 
z0
2
2

B1 B2
B2
n ig  1 a1n1 0:
z0 n ig
z0
2
2

B:2

For the other solutions we take

W2 z T 2 W1 z;

W3 z T 4 W1 z;

W4 z T 4 W2 z T 1 W3 z

B:3

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

903

and Wi4 T 3 Wi (i 1; . . . 4). Thus, from the rst solution we get

W2 z eixz z  z0

1B2  z 1
0

ig1 z 1 22
0

1
X
a2n zn ;

where

B:4

n0


B1
B2
B2
B3
a2n
2ixn 1a2n1  nn 1 2ig 2ixz0 ixz0 2  B2 
ig 1 ig 
z0
2
2

B1 B2
B2 2
n ig 1 
a
z0 n ig  
0:
z0
2
2 n1

B:5

This W2 z is the only solution relevant for Section 5. For this reason we omit the other solutions.
By the DAlembert test the solutions W1 z and W2 z converge for jzj > jz0 j, whereas W3 z and W4 z converge for
jz  z0 j > jz0 j. However, by the Raabe test they converge also at jzj jz0 j and jz  z0 j jz0 j provided that

8 h
i
>
< Re B2 Bz 1 < 1 in W1 z;
0
h
i
jzj P jz0 j if
>
: Re B2 Bz 1 > 1 in W2 z;
0

8 h i
>
< Re Bz 1 > 1 in W3 z;
0
h i
jz  z0 j P jz0 j if
>
: Re Bz 1 < 1 in W4 z;
0

B:6

where the restrictions on parameters of the equation are necessary only to assure convergence at jzj jz0 j or jz  z0 j jz0 j.
The above regions of convergence suppose the minimal solutions for the series coefcients [29,30]. In the following we
consider only the series which appears in W 1 z, the convergence for the other solutions being obtained by using the transformations as indicated above. Thus, when n ! 1 in W 1 z we have

2ix

1
a1n1
B1
an1


2i
g

2i
x
z

z
n

2i
g


1
0
0
0
2
a1n
z0
a1n

whose minimal solution for a1n1 =a1n when n ! 1 satises





a1n1
1
B1
a1n1
1
1
B1
)
:

z
1


2

1



2
B
B
0
2
2
n
z0
n
a1n
z0
a1n
z0
Thence, when n ! 1,

 1 n1 






an1 z
 jz0 j
a1n1 zn1 z0
1
B1
1
B1


)
:

1


2
1

Re
B


2
B
2
2
 a1 zn 
n
jzj
n
a1n zn
z
z0
z0
n
So, by the DAlambert test the series converge absolutely for jzj > jz0 j. However, by the Raabe test, the series converge even
for jzj jz0 j provided that ReB2 B1 =z0  < 1.
References
[1] L.J. El-Jaick, B.D.B. Figueiredo, Transformations of Heuns equation and its integral relations, J. Phys. A: Math. Theor. 44 (2011) 075204.
arXiv:1002.4559v3.
[2] K. Heun, Zur theorie der riemanschen functionen zweiter ordnung mit vier verzweigungspunkten, Math. Ann. 33 (1899) 161179.
[3] R.S. Maier, The 192 solutions of the Heun equation, Math. Comput. 76 (2007) 811843.
[4] A. Ronveaux (Ed.), Heuns Differential Equations, Oxford University Press, 1995.
[5] F.W.J. Olver, D.W. Lozier, R.F. Boisvert, C.W. Clark (Eds.), NIST Handbook of Mathematical Functions, (National Institute of Standards and Tecnology)
Cambridge University Press, 2010.
[6] A.Ya. Kazakov, S.Yu. Slavyanov, Integral relations for Heun-class special functions, Theor. Math. Phys. 107 (1996) 733739.
[7] S.Yu. Slavyanov, W. Lay, Special Functions: A Unied Theory Based on Singularities, Oxford University Press, 2000.
[8] A.H. Wilson, The ionised hydrogen molecule, Proc. R. Soc. London A118 (1928) 635647.
[9] E.W. Leaver, Solutions to a generalized spheroidal wave equation: Teukolsky equations in general relativity, and the two-center problem in molecular
quantum mechanics, J. Math. Phys. 27 (1986) 12381265.
[10] A.H. Wilson, A generalised spheroidal wave equation, Proc. R. Soc. London A118 (1928) 617635.
[11] A.M. Ishkhanyan, A. Grigoryan, Fifteen classes of solutions of the quantum two-state problem in terms of the conuent Heun function, J. Phys. A Math.
Theor. 47 (2014) 465205.
[12] J.M. Lvy-Leblond, Electron capture by polar molecules, Phys. Rev. 153 (1967) 14.
[13] L.J. El-Jaick, B.D.B. Figueiredo, A limit of the conuent Heun equation and the Schrdinger equation for an inverted potential and for an electric dipole, J.
Math. Phys. 50 (2009) 123511.
[14] P.K. Jha, Y.V. Rostovtsev, Coherent excitation of a two-level atom driven by a far-off-resonant classical eld: analytical solutions, Phys. Rev. A 81 (2010)
033827.
[15] M. Renardy, On the eigenfunctions for Hookean and FENE dumbbell models, J. Rheol 57 (2013) 13111324.
[16] N. Barbosa-Cendejas, A. Herrera-Aguilar, K. Kanakoglou, U. Nucamendi, I. Quiros, Mass hierarchy, mass gap and corrections to Newtons law on thick
branes with Poincar symmetry, Gen. Relativ. Gravitation 46 (2014) 1631.
[17] F.M.J. Olver, Asymptotics and Special Functions, Academic Press, 1974.
[18] L.J. El-Jaick, B.D.B. Figueiredo, Solutions for conuent and double-conuent Heun equations, J. Math. Phys. 49 (2008) 083508. arXiv:0807.2219v2.
[19] B.D.B. Figueiredo, Inces limits for conuent and double-conuent Heun equations, J. Math. Phys. 46 (2005) 113503.
[20] L.J. El-Jaick, B.D.B. Figueiredo, Conuent and double-conuent heun equations: convergence of solutions in series of coulomb wavefunctions, arXiv:
1209.4673 [math-ph].
[21] A. Erdlyi, Expansions of Lam functions into series of Legendre functions, Proc. R. Soc. Edimb. 62 (1948) 247267.
[22] G. Jaff, Zur theorie des wasserstoffmoleklions, Z. Phys. 87 (1934) 535544.

904

L.J. El-Jaick, B.D.B. Figueiredo / Applied Mathematics and Computation 256 (2015) 885904

[23] W.G. Baber, H.R. Hass, The two centre problem in wave mechanics, Proc. Cambridge Philos. Soc. 25 (1935) 564581.
[24] E.L. Ince, Ordinary Differential Equations, Dover, 1926.
[25] A. Decarreau, M.C. Dumont-Lepage, P. Maroni, A. Robert, A. Ronveaux, Formes canoniques des quations conuentes de lquation de Heun, Ann. Soc.
Sci. Brux. T92 (III) (1978) 5378.
[26] A. Decarreau, P. Maroni, A. Robert, Sur les quations conuentes de lquation de Heun, Ann. Soc. Sci. Brux. T92 (III) (1978) 151189.
[27] T. Masuda, H. Suzuki, Integral equations of elds on a rotating black hole, J. Math. Phys. 38 (1997) 36693678.
[28] C.G. Lambe, D.R. Ward, Some differential equations and associated integral equations, Q. J. Math. 5 (1934) 8197. Oxford series.
[29] W. Gautschi, Computational aspects of three-term recurrence relations, SIAM Rev. 9 (1967) 2482.
[30] W. Gautschi, Minimal solutions of three-term recurrence relations and orthogonal polynomials, Math. Comput. 36 (154) (1981) 447554.
[31] I.S. Gradshteyn, I.M. Ryzhik, Table of Integrals, Series and Products, Academic Press, 2007.
[32] D.B. Hodge, Eigenvalues and eigenfunctions of the spheroidal wave equation, J. Math. Phys. 11 (1970) 23082312.
[33] P.E. Falloon, P.C. Abbott, J.B. Wang, Theory and computation of spheroidal wavefunctions, J. Phys. A Math. Gen. 36 (2003) 54775495.
[34] F.M. Arscott, Periodic Differential Equations, Pergamon Press, 1964.
[35] E.T. Whittaker, G.N. Watson, A Course of Modern Analysis, Cambridge University Press, 1954.
[36] K. Knopp, Innite Sequences and Series, Dover, 1956.
[37] A. Erdlyi (Ed.), Higher Transcendental Functions, vol. 1, McGraw-Hill, 1953.
[38] L.J. El-Jaick, B.D.B. Figueiredo, New solutions for the conuent Heun equation and quasiexact solvability, arXiv: 1311.7677v1.
[39] E.A. Hylleraas, ber die elektronenterme des wasserstoffmolekls, Z. Phys. 71 (1931) 739763.
[40] M.G. Santos, S.O. Kepler, Theoretical study of the line proles of the hydrogen perturbed by collisions with protons, Mon. Not. R. Astron. Soc. 423 (2012)
6879.
[41] L.J. El-Jaick, B.D.B. Figueiredo, Integral relations for solutions of conuent Heun equations, arXiv:1311.3703 [math-ph].