LETTER Related content

- The third exactly solvable hypergeometric

Exact solution of the Schrödinger equation for the quantum-mechanical potential
Artur Ishkhanyan
inverse square root potential - The generalized quantum isotonic
oscillator
J Sesma
To cite this article: A. M. Ishkhanyan 2015 EPL 112 10006
- The Schrodinger equation for the
interaction potential x2+x2/(1+gx2) and the
first Heun confluent equation
G Marcilhacy and R Pons
View the article online for updates and enhancements.

Recent citations
- A Conditionally Integrable Bi-confluent
Heun Potential Involving Inverse Square
Root and Centrifugal Barrier Terms
Tigran A. Ishkhanyan et al

- A new exactly integrable hypergeometric

potential for the Schrödinger equation
T. A. Ishkhanyan et al

- Exact solution of the Schrödinger equation

for a short-range exponential potential with
inverse square root singularity
A. M. Ishkhanyan

This content was downloaded from IP address 181.112.138.45 on 18/09/2018 at 23:01

 October 2015
EPL, 112 (2015) 10006 www.epljournal.org
doi: 10.1209/0295-5075/112/10006

Exact solution of the Schrödinger equation for the inverse

√
square root potential V0/ x

A. M. Ishkhanyan

Institute for Physical Research, NAS of Armenia - 0203 Ashtarak, Armenia

received 28 August 2015; accepted in final form 5 October 2015

published online 23 October 2015
PACS 03.65.Ge – Solutions of wave equations: bound states
PACS 02.30.Ik – Integrable systems
PACS 02.30.Gp – Special functions

Abstract – We √ present the exact solution of the stationary Schrödinger equation for the poten-
tial V = V0 / x. Each of the two fundamental solutions that compose the general solution of
the problem is given by a combination with non-constant coefficients of two confluent hyperge-
ometric functions of a shifted argument. Alternatively, the solution is written through the first
derivative of a tri-confluent Heun function. Apart from the quasi-polynomial solutions provided
by the energy specification En = E1 n−2/3 , we discuss the bound-state wave functions vanishing
both at infinity and in the origin. The exact spectrum equation involves two Hermite functions
of non-integer order which are not polynomials. An accurate approximation for the spectrum
providing a relative error less than 10−3 is En = E1 (n − 1/(2π))−2/3 . Each of the wave func-
tions of bound states in general involves a combination with non-constant coefficients of two
confluent hypergeometric and two non-integer order Hermite functions of a scaled and shifted
coordinate.

c EPLA, 2015
Copyright 

Introduction. – The exact solutions of the
Schrödinger equation are rare. The most known ones
are the potentials solvable in terms of the ordinary
and confluent hypergeometric functions. The list of the
known ordinary hypergeometric potentials is rather
large; however, there exist only two independent such
potentials; the Eckart [1] and the Pöschl-Teller [2] po-
tentials. Many of the widely used potentials such as the
Rosen-Morse [3], Woods-Saxon [4], Manning-Rosen [5],
Hulthén [6], Scarf [7] ones, are in fact particular speci-
fications of the mentioned two independent potentials.
In the confluent hypergeometric case there are three
independent potentials: the harmonic oscillator (plus
inverse square) potential discussed by Schrödinger [8], the
Kratzer potential (Coulomb plus inverse square) [8,9] and
the Morse potential [10].
In the present paper we present one more potential solv-
able in terms of the confluent hypergeometric √ functions
—the inverse square root potential V = V0 / x. The ex-
act solution of the Schrödinger equation for this potential
has a remarkable structure. Each of the two fundamental
solutions that compose the general solution of the prob-
lem presents a combination of two confluent hypergeomet-
ric functions. The coefficients of the combinations are not
constants. It is convenient to write one of the conflu-
ent hypergeometric functions involved in the fundamen-
tal solutions as a Hermite function of non-integer order
which in general is not a polynomial. Then the standard
polynomial reduction is achieved by the energy specifica-
tion En = E1 n−2/3 . Furthermore, we discuss the bound-
state wave functions vanishing both at infinity and in the
origin. The exact energy spectrum is derived through an
equation involving two Hermite functions. We show that
a highly accurate approximation for the energy levels is
given as En = E1 (n − 1/(2π))−2/3 . This approximation
provides an absolute error less than 10−3 for all quantum
numbers n ∈ N .
We note that the general solution of the problem is de-
rived via the reduction of the Schrödinger equation to an
equation obeyed by a function involving the first deriva-
tive of a tri-confluent Heun function [11–13]. The resultant
two-term confluent hypergeometric fundamental solution
for the inverse square root potential is in fact a specifica-
tion of the derivative of a tri-confluent Heun function for
given parameters. For this reason, the derivation is itself
of methodological interest because it introduces a differ-
ent feature in the search for solvable potentials. For this
reason, the derivation is separated out to an appendix.

10006-p1
A. M. Ishkhanyan

General solution and quasi-polynomial reduc- ψ (x)

tions. – It is straightforwardly checked by direct substi- n=1
tution that the general solution of the one-dimensional n=2
0.4
stationary Schrödinger equation for a particle of mass m n=3
and energy E
0.2
2
d ψ 2m
+ 2 (E − V (x)) ψ = 0, (1)
dx2  5 10 15 20
x

for the inverse square root potential

– 0.2
V0
V =√ (2)
x
Fig. 1: The first three bounded quasi-polynomial solutions for
with arbitrary V0 is written as the energy specification (8), V0 = −1,  = m = 1.

du
ψ(x) = e−δx/2 , (3) A peculiarity of this set of quasi-polynomial functions is
dy
that the solutions do not vanish at the origin (fig. 1).
where However, depending on the particular physical problem at
   hand (for instance, if one considers eq. (1) as the s-wave ra-
√ a 1 2
− 2a y
u=e c1 · Ha (y) + c2 ·1 F1 − ; ; y . (4) dial equation for the three-dimensional
√ Schrödinger equa-
2 2 tion with the potential V = V0 / r), it is useful to have a
set of bound wave functions that vanish at the origin [15].
Here c1,2 are arbitrary constants, Ha is the Hermite func-
tion [14] (for a non-negative integer a it becomes the Her- Bound states. – Consider the case when one demands
mite polynomial; however, in general a is arbitrary), 1 F1 is the wave function to vanish both in the origin and at the
the Kummer confluent hypergeometric function [14], the infinity: ψ(0) = 0 and ψ(+∞) = 0 (V0 < 0). The first
auxiliary dimensionless argument y defines a scaling of the condition imposes the following relation between the coef-
coordinate followed by deformation and shift: ficients c1,2 :
√ √    
y = sgn(V0 ) δx + 2a, (5) a 3 a 1
2a1 F1 1 − ; ; 2a +1 F1 − ; ; 2a
2 2 2 2
and the involved parameters δ and a are given as c1 = √ √  √  c2 , (13)
2a Ha−1 2a − Ha 2a

δ = −8m E/2, (6) while the second one, after matching the leading asymp-
2 2
m V0 totes of the involved Hermite and confluent hypergeomet-
a= 3/2
. (7)
 (−2mE) ric functions, results in a transcendental equation for the
parameter a, which is conveniently written through two
A standard set of bounded quasi-polynomial solutions for Hermite functions as
an attractive potential with V0 < 0 is achieved by putting √  √   √ 
a = n, n ∈ N . Then, the Hermite function in eq. (4) be- 2a H a−1 − 2a + H a − 2a = 0. (14)
comes the Hermite polynomial and one should put c2 = 0
to ensure vanishing of the solution at infinity. The energy This is the equation for the energy spectrum. It pos-
eigenvalues for this case are sesses a countable infinite set of roots which, together with
eq. (7), determine the energy eigenvalues.
√ To solve this
 1/3
V0 −mV0 equation, we divide it by Ha (− 2a) and use the known
En = 2
n−2/3 , n = 1, 2, 3, . . . , (8) recurrence relations and series expansions for the in-
2 
volved Hermite functions [14] to get the following key
and the corresponding solutions are written as approximation:
√ √ √ √ √ √
ψn = e− 2n y−δx/2 (Hn (y)− 2n Hn−1 (y)), y = 2n− δx. 2a Ha−1 (− 2a)
F ≡ √ +1≈
(9) Ha (− 2a)
To get an explicit representation, here are the first three  
sin π a + 1/2 − a e−2a
quasi-polynomials: √  √   √ ,
a 1+log(1−b1/ 2a) cos πa+1+log(1 − b2 / 2a)
√ √
ψ1 = e− 2y−δx/2 (1 − 2y), (10) (15)
ψ2 = e−2y−δx/2 (1 + 2y − 2y 2 ), (11)
√ √ √ where the parameters b1,2 are constants of the order of
ψ3 = e− 6y−δx/2 (3 − 3 6y − 6y 2 + 2 6y 3 ). (12) unity. The accuracy of this result is demonstrated in fig. 2,

10006-p2
 √
Exact solution of the Schrödinger equation for the inverse square root potential V0 / x

F ψ (x)
2 0.6 n=1

n=2
1 0.4 n=3

0.2
a
9.5 10.0 10.5 11.0 11.5

x
1 5 10 15 20 25

– 0.2
2

Fig. 4: Normalized wave functions of the ground state n = 1

Fig. 2: (Colour on-line) Approximation (10) —solid curve; and the next two states for the energy spectrum (16), V0 = −1,
exact numerical values —filled circles.  = m = 1.

E
x are not quasi-polynomials in contrast to the ordinary
10 20 30 40 50 60
spectrum (8). The wave functions of the ground state
0.2 n = 1 and the next two states are shown in fig. 4.
E3
0.4 E2 Discussion. – Thus, we have presented the general so-
Error lution of the one-dimensional stationary Schrödinger equa-
E1 0.0000 n
0.6 tion for the inverse square root potential. The solution is
0.0005 4 6 8 10
0.0010 written in terms of the confluent hypergeometric functions
V0 0.0015
0.8 0.0020
x and the Hermite functions of non-integer order. We have
0.0025
shown that the standard quasi-polynomial solutions pro-
1.0
vide the spectrum En = E1 n−2/3 .
Discussing the case of the bound-state wave functions
Fig. 3: (Colour on-line) Potential (2) and energy spectrum (17)
for V0 = −1,  = m = 1. The inset shows the relative error of vanishing both at infinity and in the origin, we have shown
the (n − 1/(2π))−2/3 approximation for the spectrum. that the exact spectrum equation for this case involves
two Hermite functions of non-integer order. We have
where the filled circles indicate the exact numerical values derived an accurate approximation for the energy spec-
and the solid curve presents the derived approximation. trum providing an absolute error less than 10−3 as En =
Having established this approximation, we arrive at a E1 (n − 1/(2π))−2/3 . Each of the bound-state wave func-
remarkably simple, yet highly accurate, approximation for tions involves a combination with non-constant coefficients
the exact eigenvalue equation (14) written in terms of the of four confluent hypergeometric functions (or, equiva-
familiar elementary functions: lently, non-integer order Hermite functions) of a scaled and
  shifted coordinate argument. A distinct feature of these
sin π a + 1/2 − a e−2a = 0. (16) wave functions is that they are not quasi-polynomials.
Compared with the exact equation, everywhere for We have shown that for the inverse square root potential
a > 1/2 this approximation provides a relative error of the the solution of the Schrödinger equation is equivalently
order or less than 10−3 . written through the derivative of the solution of a tri-
The smallest root of eq. (16) is close to the unity (a1 ≈ confluent Heun function. The employment of the equations
0.86); consequently, the exponential term is always small. obeyed by the derivatives of the Heun functions for con-
Hence, even if this term is neglected, one may expect a struction of exactly solvable potentials seems to be promis-
sufficiently good approximation for the energy spectrum. ing. This is because, on the one hand, these equations are
Thus, the roots of eq. (16) are approximately given as more complicated than the Heun equations (in that they
an = n − 1/(2π). Equation (7) then gives the spectrum have extra singularities) and, hence, they suggest a wider
 1/3  −2/3 set of effects, and, on the other hand, their solutions are
V0 −mV0 1 written through the solutions of the Heun equations and
En = n − , n = 1, 2, 3, . . .
2 2 2π thus these functions do not cause additional complications
(17) compared to the Heun functions.
This is indeed a rather accurate approximation. The rela- It is worth mentioning that the inverse square root po-
tive error is less than 10−3 for all n > 2 and it is less than tential is a member of one of the five exactly solvable
10−5 for n ≥ 7 (fig. 3). six-parametric bi-confluent Heun potentials presented by
We note that since the roots an are not integers the Lamieux and Bose [16], see also [17]. (Note that our poten-
wave functions of the bound states for spectrum (17) tial is actually three-parametric; apart from the parameter

10006-p3
A. M. Ishkhanyan

V0 there are two more available parameters —the space where z0 = q/α. It is seen that compared to the start-
origin and the energy origin that we have omitted.) Our ing tri-confluent equation (A.1) this equation possesses an
result thus shows that there exists a particular specifi- additional singularity located at z0 .
cation of the parameters of the bi-confluent Heun equa- We now try to reduce the Schrödinger equation to
tion for which the bi-confluent Heun function is expressed this equation by transforming both dependent and in-
through the tri-confluent Heun functions. There are other dependent variables ψ = ϕ (z) u(z), z = z(x). The
cases supporting this observation, e.g., several quantum approach suggested in [17] for construction of exactly
two-state models solvable in terms of the Heun functions integrable energy-independent potentials is based on the
suggest similar conclusions [18–20]. This applies not only assertion that if a potential is proportional to an energy-
to the bi-confluent Heun function but to all other Heun independent parameter, then the logarithmic z-derivative
functions including the general Heun function [20]. ρ (z)/ρ(z) of the function ρ(z) = dz/dx cannot have poles
The inverse square root potential appears in many other than the finite singularities of the target equation to
studies as a part of a more general potential (see, which the Schrödinger equation is reduced. It then follows
e.g., [15,21,22]). However, these are either conditionally that if the target equation is an equation with rational co-
solvable potentials when a parameter of the potential is efficients, then the equation for the coordinate transforma-
fixed or energy dependent potentials. We would like to tion should necessarily be of the form ρ(z) = Πi (z − zi )Ai ,
emphasize that the inverse square root potential we have where zi are the mentioned finite singularities of the target
presented is an independent exactly solvable one without equation (Manning form [25]). In addition, the exponents
restrictions so that the solution is not affected by addi- Ai should all be integers or half-integers.
tional influences coming from extra terms involved in the Since the only finite singularity of eq. (A.3) is z = z0 , the
potential. only permissible coordinate transformation for this equa-
A last remark concerns the Darboux transformation [23] tion is given as z  (x) = ρ = (z − z0 )m1 /σ with integer or
which is widely applied to construct new solvable poten- half-integer m1 . Following the lines of [17], we match the
tials by employing the known ones [24]. Though the solu- logarithmic derivative of ρ(z) with the corresponding term
tion that we have presented involves a special function and of the invariant of eq. (A.3):
its derivative, the potential is not derived by a Darboux    2
transformation. The latter assumes that the involved spe- 1 ρz 1 ρz 3/4
+ = . (A.4)
cial function is a solution of a (different) Schrödinger equa- 2 ρ z 4 ρ (z − z0 )2
tion, while in our case this is not the case. This assertion
is further supported by the observation that because of As a result we get m1 = −1. Consequently, the admissible
a logarithmic derivative involved in the Darboux trans- coordinate transformation is
formation it usually results in a conditionally integrable
2(x − x0 )
potential with a fixed parameter [24], while we present an z2 = . (A.5)
σ
exactly solvable one.
Now, we look for a solution of the Schrödinger equation of
∗∗∗ the form [17,18]
2 3
This research has been conducted within the scope of ψ = e−(γz+δz /2+εz /3)/2 w(z), (A.6)
the International Associated Laboratory IRMAS (CNRS-
France & SCS-Armenia). The work has been supported where w(z) is the solution of eq. (A.3). Substituting this ψ
by the Armenian State Committee of Science (SCS Grant and z(x) from (A.5) into the Schrödinger equation (1) and
No. 13RB-052). eliminating w (z) using eq. (A.3), we arrive at an equation
Appendix: derivation of the solution. – The which is proportional to w(z). This is because the pre-
derivatives of the Heun functions generally obey more factor in the wave function (A.6) is chosen so that the

complicated equations involving additional apparent sin- term proportional to the first derivative w (z) is cancelled.
gularities. For the tri-confluent Heun equation [11,12] The equation reads
√ √
d2 u  
2 du 2γ γ2 2 2 (2α − γδ + ε)  
+ γ +δz +εz + (α z − q)u = 0, (A.1) + − + 2 δ 2 + 2γ ε
dz 2 dz x3/2 x x1/2

if α is not zero, the equation obeyed by the function √ 1/2 2 16m(E −V) √
+ 4 2δεx + 4ε x+ 2
w ( 2x) = 0, (A.7)
2 3 du 
w = eγz+δz /2+εz /3 (A.2)
dz
where for simplicity we have put x0 = 0, σ = 1 and,
is written as without loss of the generality, q = 0. This specification
 
d2 w 2 1 dw of the parameters can always be applied because the tri-
− γ + δz + εz + + α (z − z0 )w = 0,
dz 2 z − z0 dz confluent Heun equation preserves its form when scaling
(A.3) and shifting the origin.

10006-p4

Exact solution of the Schrödinger equation for the inverse square root potential V0 / x
The last step is now straightforward. First, eq. (A.7)
shows that V (x) is necessarily of the form
V5 V4 V0
V (x) = + + 1/2 + V1 + V2 x1/2 + V3 x. (A.8)
x3/2 x x
Substituting this into eq. (A.7) and requiring the vanishing
of the coefficients at powers of x, we readily get that if
the parameters V0,1,2,3,4,5 of the potential are supposed
independent, then V5 = V4 = V3 = V2 = 0 and
 √
2 2m(−E + V1 ) 2 2mV0
δ=± , α=− , γ = ε = 0.
 2
(A.9)
These parameters determine the tri-confluent Heun func-
tion that gives, through eqs. (A.2), (A.6), the solution
of the Schrödinger equation for the inverse square root
potential (2).
Notably, the parameter ε for this solution is zero. It
is known that then the tri-confluent Heun equation is re-
duced to the Kummer confluent hypergeometric equation
via a transformation of the variables. This observation
accomplishes the development. Because it is the deriva-
tive w(z) = u (z) of the tri-confluent function that is
involved in the wave function (A.6) and since the par-
ticular tri-confluent Heun function under consideration
is a product of a confluent hypergeometric function and
a non-constant pre-factor, we get a solution that gener-
ally involves four confluent hypergeometric functions (each
fundamental solution of the corresponding confluent hy-
pergeometric equation leads to a combination with non-
constant coefficients of a pair of confluent hypergeometric
functions).
REFERENCES
[1] Eckart C., Phys. Rev., 35 (1930) 1303.
[2] Pöschl G. and Teller E., Z. Phys., 83 (1933) 143.
10006-p5
√
[3] Rosen N. and Morse P. M., Phys. Rev., 42 (1932) 210.
[4] Woods R. D. and Saxon D. S., Phys. Rev., 95 (1954)
577.
[5] Manning M. F. and Rosen N., Phys. Rev., 44 (1933)
953.
[6] Hulthén L., Ark. Mat. Astron. Fys., 28A (1942) 5;
Hulthén L., Ark. Mat. Astron. Fys., 29B (1942) 1.
[7] Scarf F., Phys. Rev., 112 (1958) 1137.
[8] Schrödinger E., Ann. Phys. (Berlin), 76 (1926) 361.
[9] Kratzer A., Z. Phys., 3 (1920) 289.
[10] Morse P. M., Phys. Rev., 34 (1929) 57.
[11] Ronveaux A. (Editor), Heun’s Differential Equations
(Oxford University, London) 1995.
[12] Slavyanov S. Yu. and Lay W., Special Functions
(Oxford University Press, Oxford) 2000.
[13] Heun functions, their generalizations and applications,
http://theheunproject.org/bibliography.html.
[14] Andrews G. E., Askey R. and Roy R., Special
Functions (Cambridge University Press, Cambridge)
1999.
[15] Znojil M., Phys. Rev. A, 61 (2000) 066101.
[16] Lamieux A. and Bose A. K., Ann. Inst. Henri Poincaré
A, 10 (1969) 259.
[17] Ishkhanyan A. and Krainov V., Discretization of
Natanzon potentials, arXiv:1508.06989 (2015).
[18] Shahverdyan T. A., Ishkhanyan T. A., Grigoryan
A. E. and Ishkhanyan A. M., J. Contemp. Phys., 50
(2015) 211.
[19] Ishkhanyan A. M. and Grigoryan A. E., J. Phys. A,
47 (2014) 465205.
[20] Ishkhanyan A. M., Shahverdyan T. A. and
Ishkhanyan T. A., Eur. Phys. J. D, 69 (2015) 10.
[21] Stillinger F. H., J. Math. Phys., 20 (1979) 1891.
[22] Batic D., Mills-Howell D. and Nowakowski M.,
J. Math. Phys., 56 (2015) 052106; Batic D., Williams
R. and Nowakowski M., J. Phys. A, 46 (2013) 245204.
[23] Darboux G., C. R. Acad. Sci. Paris, 94 (1882) 1456.
[24] Matveev V. B. and Salle M. A., Darboux Transforma-
tions and Solitons (Springer, Berlin) 1991.
[25] Manning M. F., Phys. Rev., 48 (1935) 161.