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

net/publication/232252904

The fractional Schrödinger equation and the infinite potential well-numerical

results using the Riesz derivative

Article · October 2012

DOI: 10.1007/978-3-642-33335-4_10001 · Source: arXiv

CITATIONS READS

10 243

1 author:

Richard Herrmann
GigaHedron
39 PUBLICATIONS   812 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

numerical solution of fractional Schroedinger equation View project

nuclear physics View project

All content following this page was uploaded by Richard Herrmann on 05 February 2015.

The user has requested enhancement of the downloaded file.

 THE FRACTIONAL SCHRÖDINGER EQUATION
AND THE INFINITE POTENTIAL WELL -
NUMERICAL RESULTS USING THE RIESZ DERIVATIVE

1
RICHARD HERRMANN

Abstract. Based on the Riesz definition of the fractional derivative the fractional
Schrödinger equation with an infinite well potential is investigated. First it is shown
analytically, that the solutions of the free fractional Schrödinger equation are not
eigenfunctions, but good approximations for large k and for α ≈ 2. The first lowest
eigenfunctions are then calculated numerically and an approximate analytic formula
for the level spectrum is derived.
Key Words: Fractals and nonlinear dynamics, Quantum mechanics, numerical
analysis of boundary-value problems, Schrödinger equation
PACS: 05.45.Df, 03.65. -w, 02.60.Lj

1. Introduction
Wave equations play a significant role in the description of the dynamic development
of particles and fields; e.g. the Maxwell-equations describe the behavior of the electro-
magnetic field in terms of coupled partial differential equations. In quantum mechanics
a particle may be described by the non-relativistic Schrödinger wave equation, where
the kinetic term is given by the Laplace-operator.
Fractional calculus introduces the concept of non-locality to arbitrary hitherto local
operators [14], [15], [10]. This is a new property, which only recently attracted attention
on a broader basis. The interest in a non-local dynamic description of e.g. quantum
systems has been steadily increasing, because it is expected, that quantum phenomena
may be treated more elegantly from a generalized point of view.
Within this context, it is helpful to investigate fundamental properties of a fractional
wave equation and to study general features of its solution.
As an example, we will present in the following the main results of a solution of the
fractional Schrödinger equation with infinite potential well.
1
2 R. HERRMANN

2. The problem - analytic part

Let the one dimensional fractional stationary Schrödinger equation in scaled canon-
ical form be defined as
− ∆α/2 Ψ(x) = (E − V (x))Ψ(x) (2.1)
with the fractional Laplace-operator ∆α/2 .
The definition of a fractional order deriv-
ative is not unique, several definitions e.g. the Riemann[16], Caputo[3], Liouville [11],
Riesz [17], Feller[6], Weyl[18] coexist and are equally well suited for an extension of the
standard derivative.
In order to preserve Hermiticity for the fractional extension of the Laplace-operator[13],
we will explicitly apply the Riesz fractional derivative
∫
sin(πα/2) ∞ f (x + ξ) − 2f (x) + f (x − ξ)
∆α/2 ≡ ∞ RZ D α
f (x) = Γ(1 + α) dξ (2.2)
π 0 ξ α+1
0<α<2
where the left superscript in ∞ α
RZ D emphasizes the fact, that the integral domain is the
full space R.
Since the eigenfunctions of the Riesz derivative operator are given as
∞
RZ D cos(kx) = −|k|α cos(kx)
α
(2.3)
∞
RZ D sin(kx) = −|k| sin(kx)
α α
(2.4)
the eigenfunctions of the potential free (V (x) = 0) fractional Schrödinger equation
using the Riesz fractional derivative follow as
+
RZ Ψ (k, x) = cos(kx) (2.5)
−
RZ Ψ (k, x) = sin(kx) (2.6)
where the ±-sign indicates the parity and the corresponding continuous energy spec-
trum follows as:
Ekfree = |k|α k∈R (2.7)
∫ +∞ ∫ +∞
Since the integrals −∞ dx cos(kx) and −∞ dx sin(kx) respectively are divergent, the
eigenfunctions are not normalizable on the full domain R.
In the case α = 2, which corresponds to the classical quantum mechanics, of course
we may apply the classical box-normalization, which means we make a statement on
the behavior of the wave-function isolated in a box of size e.g. 2q of the form:
∫ +q
cos(kx)dx < ∞ (2.8)
−q
∫ +q
sin(kx)dx < ∞ (2.9)
−q
According to (2.3) and (2.4) a special feature of the Riesz derivative is that the type
of the eigenfunctions of the free fractional Schrödinger equation does not change for
 THE FRACTIONAL SCHRÖDINGER EQUATION . . . 3

arbitrary α. Hence the question arises, whether box-normalization is a legal procedure

in the case of non-local differential equations. At least, the physical meaning of box-
normalization may be different.
To avoid these interpretative difficulties we will switch on a potential V (x), such
that the eigenfunctions vanish at infinity. An ideal candidate for such a potential is
the infinite potential well, centered at the origin with finite size 2q, which is explicitly
given by:
{
0 |x| ≤ q
V (x) = (2.10)
∞ |x| > q

A reasonable ansatz for the corresponding eigenfunction follows as:

{
∞ Ψ(x) |x| ≤ q
⊔
Ψ(x) = (2.11)
0 |x| > q

Where the superscript ⊔ Ψ(x) emphasizes the fact, that the wave-function per defini-
tionem is now confined inside the infinite potential well and vanishes outside. The
superscript ∞ Ψ(x) indicates, that this function may at first be defined on the whole
domain, but is used only inside the bounded domain of the infinite potential well.
The normalization condition for ⊔ Ψ(x) is now given by:
∫ +∞ ∫ +q
dx⊔ Ψ(x) = dx∞ Ψ(x) = 1 (2.12)
−∞ −q

As a consequence, we may interpret ⊔ Ψ(x) physically as a normalizable wave-function

and its absolute value ⊔ Ψ(x)⊔ Ψ(x)∗ as a probability measure.
In the classical, local case (α = 2) we obtain immediately


cos k 2q x |x| ≤ q and k = 1, 3, 5, ...
π
⊔ ±
Ψ (x) = sin k 2q x |x| ≤ q and k = 2, 4, 6, ...
π (2.13)


0 |x| > q

and the continuous energy spectrum changes to a discrete one, since k ∈ N is an integer
now.
π
Eklocal = ( )2 k 2 k = 1, 2, 3, ... (2.14)
2q

In order to solve the fractional Schrödinger equation of the infinite potential well for
arbitrary α, we have to apply the Riesz derivative operator to ⊔ Ψ(x).
4 R. HERRMANN

For the positive semi-axis x ≥ 0 we obtain for x ≤ q:

∞ α ⊔ sin(πα/2)
RZ D ( Ψ(x)) = Γ(1 + α) ×
∫ ∞⊔ π
Ψ(x − ξ) − 2 ⊔ Ψ(x) + ⊔ Ψ(x + ξ)
dξ
0 ξ α+1
sin(πα/2) {
= Γ(1 + α) ×
π
∫ q−x ∞
Ψ(x − ξ) − 2 ∞ Ψ(x) + ∞ Ψ(x + ξ)
dξ + (2.15)
0 ξ α+1
∫ q+x ∞
Ψ(x − ξ) − 2 ∞ Ψ(x)
dξ +
q−x ξ α+1
∫ ∞
−2 ∞ Ψ(x) }
dξ
q+x ξ α+1
⊔ α ∞
= RZ D ( Ψ(x)) 0≤x≤q
In addition, for x > q due to the non-local character of the Riesz derivative operator
we obtain a non-vanishing finite term
∫ x+q ∞
∞ α ⊔ sin(πα/2) Ψ(x − ξ)
RZ D ( Ψ(x)) = Γ(1 + α) × dξ x>q (2.16)
π x−q ξ α+1
which is negligible only in the special case of the infinite potential well discussed here.
The corresponding equations for x ≤ 0 just interchange the roles of ∞ Ψ(x − ξ) and
∞ Ψ(x + ξ). Since parity is conserved for the infinite potential well, we may restrict to

(2.15) without loss of generality.

We can write (2.15) in short-hand notation:
∞ α ⊔
RZ D ( Ψ(x)) = ⊔ α ∞
RZ D ( Ψ(x)) (2.17)
This may be interpreted as a modification of the Riesz-operator, which now only covers
the inside region of the potential well. Obviously both operators differ significantly.
∞
RZ D
α
̸= ⊔
RZ D
α
(2.18)
This is a general feature of all fractional derivative definitions, which span over the full
R e.g. Liouville’s, Weyl’s and Feller’s definition.
Only for the Riemann R Dα and the Caputo C Dα definition of a fractional derivative
and only for the very special case of an infinite potential well centered at the origin the
operator is not altered:
∞ ⊔
R Dα = RD
α
(2.19)
∞ ⊔
C Dα = CD
α
(2.20)
because the corresponding derivative definitions only cover the inside region of the
potential well 0 ≤ x ≤ q.
 THE FRACTIONAL SCHRÖDINGER EQUATION . . . 5

As a direct consequence, eigenfunctions of the free fractional Schrödinger equation

based on the Riemann and Caputo derivative definition automatically determine the
eigenfunctions of the same Schrödinger equation with infinite potential well fulfilling
the additional constraint, that the functions should vanish at the boundaries of the
infinite well [9].
These eigenfunctions ∞ ±
R,C Ψ (x, α) are known analytically, normalizable on R for α <
2 and are given in terms of the Mittag-Leffler functions Eα (z) and Eα,β (z) as:

∞
Ψ+ (x, α) = x 2 −1 Eα, α2 (−xα )
α
R (2.21)
∞ − α−1 α
R Ψ (x, α) = x Eα,α (−x ) (2.22)
∞ + α
C Ψ (x, α) = Eα (−x ) (2.23)
∞
Ψ− (x, α) = x 2 Eα,1+ α2 (−xα )
α
C 0≤α≤2 (2.24)

which reduce to the trigonometric functions for α = 2 and the corresponding eigenval-
ues are determined from the zeros of these functions [9], [10].

In contrast to this simple classical behavior we now consider the case of the fractional
Schrödinger equation for the infinite potential well based on the Riesz definition of a
fractional derivative. The following questions arise:
• are plain waves still a solution for the infinite potential well?
• if not, are they at least a good approximation?
• what do the exact solutions look like?
To answer these questions, we rearrange terms in the integral-operators In of the mod-
ified Riesz-operator ⊔ α
RZ D :

( ) ( sin(πα/2) )
⊔ α ∞
RZ D ( Ψ) (x) = Γ(1 + α) (I1 + I2 + I3 )∞ Ψ (x) (2.25)
π

where In are given explicitely:

∫
Ψ(x + ξ) − 2Ψ(x) + Ψ(x − ξ)
q−x
I1 Ψ(x) = dξ (2.26)
0 ξ α+1
∫ q+x
Ψ(x − ξ)
I2 Ψ(x) = dξ (2.27)
q−x ξ α+1
∫ ∞ ∫ ∞
−2Ψ(x) 1
I3 Ψ(x) = α+1
dξ = −2Ψ(x) α+1
dξ (2.28)
q−x ξ q−x ξ
6 R. HERRMANN

+1.0 + +1.0

g1
+0.5 +0.5

0 0

-0.5 -0.5

g5+
-1.0
-1.0

+1.0 +1.0

g+ g 7+
+0.5
3 +0.5

0 0

-0.5
-0.5

-1.0
-1.0

-1.0 0 x +1.0 -1.0 0 x +1.0

Figure 1. The lowest solutions of −⊔ α

RZ D cos(kπ/2x) = Egk (α, x) for
q = 1 in the range (dashed line) 2 ≥ α ≥ 0.25 (dotted line) in ∆α = 0.25
steps. For the classical, local case α = 2 the solution is indeed an
eigenfunction, but for decreasing α deviations from cos(kπ/2x) become
more and more pronounced. On the other hand, for large k the error
becomes smaller and cos(kπ/2x) becomes a good first guess for the exact
eigenfunction.

Using the test functions Ψ+ (x) = cos(kπ/2x) and Ψ− (x) = sin(kπ/2x) we obtain
with the help of [1], [5] and a cup of tea:
{
cos(kπ/2x) (q − x)−α (
I1 =
sin(kπ/2x) 2α
k2 π2 q − x 2 ( α 3 α q − x 2)
( ) 1 F2 1 − ; , 2 − ; −k 2 π 2 ( ) +
α−2 q 2 2 2 4q
{
q−x ) cos(kπ/2x)
8 sin2 (kπ ) × (2.29)
4q sin(kπ/2x)
{ {
cos(kπ/2x) kπ α Re(Υ(x))
I2 = ( ) × (2.30)
sin(kπ/2x) 2q −Im(Υ(x))
{ {
cos(kπ/2x) −2(q − x)−α cos(kπ/2x)
I3 = × (2.31)
sin(kπ/2x) α sin(kπ/2x)
 THE FRACTIONAL SCHRÖDINGER EQUATION . . . 7

+1.0 + +1.0

Ψ1
α=0.25 +0.5

0
+0.5

-0.5
+
Ψ5
+0.1 -1.0

+1.0 +1.0
+
Ψ3 Ψ7+
+0.5 +0.5

0 0

-0.5
-0.5

-1.0
-1.0

-1.0 -0.5 0 +0.5 x +1.0 -1.0 -0.5 0 +0.5 x +1.0

Figure 2. The lowest numerically determined eigenfunctions Ψ+ k (α, x)

for q = 1 with positive parity for (dashed line) 2 ≥ α ≥ 0.25 (dotted
k (α → 2, x) = cos(kπ/2x)
line) in ∆α = 0.25 steps. Only in the limit Ψ+
holds.

with 1 F2 (a; b, c; x) is the hypergeometric function, Re(Υ(x)) is the real part and Im(Υ(x))
is the imaginary part of the complex function Υ(x) given by
( q−x q+x )
Υ(x) = e−i 2 (a+kx/q) Γ(−α, −ikπ
π
) − Γ(−α, −ikπ ) (2.32)
2q 2q
where Γ(a, z) denotes the incomplete Γ-function on the complex plane.
The kinetic part of the fractional Schrödinger equation may therefore be calculated
analytically and may be written with the testfunctions ∞ Ψ+ (x) = cos(kπ/2x) and
∞ Ψ− (x) = sin(kπ/2x):

−⊔ α
RZ D cos(kπ/2x) = Egk (x) k = 1, 3, 5, ... (2.33)
⊔
−RZ D sin(kπ/2x) = Egk (x) k = 2, 4, 6, ...
α
(2.34)
We define a pseudo-normalization condition
gk (x = 0) = 1 k = 1, 3, 5, ... (2.35)
gk (x = 1/k) = 1 k = 2, 4, 6, ... (2.36)
which allows to compare gk (x) with the trigonometric functions.
8 R. HERRMANN

+1.0 − +1.0

g
2 g −6
+0.5 +0.5

0
0

-0.5
-0.5

-1.0
-1.0

+1.0 +1.0

+0.5 +0.5

0 0

-0.5
−
-0.5
g− g
4 8
-1.0
-1.0

-1.0 0 x +1.0 -1.0 0 x +1.0

Figure 3. The lowest solutions of −⊔ α

RZ D sin(kπ/2x) = Egk (α, x) for
q = 1 in the range (dashed line) 2 ≥ α ≥ 0.25 (dotted line) in ∆α = 0.25
steps. For the classical, local case α = 2 the solution is indeed an
eigenfunction, but for decreasing α deviations from sin(kπ/2x) become
more and more pronounced. On the other hand, for large k the error
becomes smaller and sin(kπ/2x) becomes a good first guess for the exact
eigenfunction.

In figures 1 and 3 the results are sketched. The bad news is, that neither gk (x) =
cos(kπ/2x) nor gk (x) = sin(kπ/2x) holds. Consequently the solutions of the free frac-
tional Schrödinger equation based on the Riesz derivative definition are no eigenfunc-
tions of the same fractional Schrödinger equation with infinite potential well. This
answers the first question.
The good news is, that the deviations become more and more negligible for k ≫ 1.
This answers the second question: For large k, in the vicinity of α ≈ 2 and for x ≈ 0
the trigonometric functions seem to be a good first guess.
As a consequence of the pseudo-normalization condition (2.35), we can give an ana-
lytic expression for the approximate energy spectrum. With

−⊔ α ∼
RZ D cos(kπ/2x)|x=0 = Ek (α) k = 1, 3, 5, ... (2.37)
 THE FRACTIONAL SCHRÖDINGER EQUATION . . . 9

+1.0 +1.0
+ +
Ψ2 Ψ6
+0.5 +0.5
α=0.25
0 0

-0.5 -0.5

-1.0 -1.0

+1.0 +1.0

α=0.25
+0.5 +0.5

0 0

-0.5
-0.5
+
Ψ4 -1.0
Ψ8+
-1.0

-1.0 -0.5 0 +0.5 x +1.0 -1.0 -0.5 0 +0.5 x +1.0

Figure 4. The lowest numerically determined eigenfunctions Ψ− k (α, x)

with negative parity for q = 1 in the range (dashed line) 2 ≥ α ≥ 0.25
(dotted line) in ∆α = 0.25 steps. Only in the limit Ψ+ k (α → 2, x) =
sin(kπ/2x) holds.

we obtain (since I2 vanishes for x = 0):

1 sin( π2 α) ( k 2 π 2 ( α 3 α 1 2 2) π )
Ek∼ (α) = q −α Γ(α) 1 F2 1− ; , 2− ; − k π −4 cos(k ) (2.38)
2 π 2−α 2 2 2 16 2

which according to the above mentioned criteria will turn out to be a good approxima-
tion of the exact energy eigenvalues for large k in the vicinity of α ≈ 2.

3. The solution - numerical part

Since we have shown by explicit analytic evaluation of the fractional derivative, that
the trigonometric functions are no eigenfunctions of the infinite potential well, but only
good approximations, we will calculate the exact solutions numerically.
10 R. HERRMANN

comparison of
100 k=8
exact energies
energy

with approximations

10

1 k=1
0.5 1.0 1.5 α 2.0

Figure 5. A comparison of the numerically determined energy values

(triangles) with approximate formulas Ek∼ (α) from (2.38) (solid line)
and Ekfree (α) from (2.7) (dashed line) normalized to q = 1.

For that purpose, we expand the exact solution Ψ±

k (x) in a Taylor series

∑
N
Ψ+
k (x) = lim a2n x2n (3.1)
N →∞
n=0
∑N
Ψ−
k (x) = lim a2n+1 x2n+1 (3.2)
N →∞
n=0

and insert it into the fractional Schrödinger equation for the infinite well potential:

α ± ±
−⊔
RZ D Ψk (x) = Ek Ψk (x) (3.3)

The integrals on the left may be evaluated analytically and lead to transcendental
functions in x, which then are expanded up to order N in a Taylor-series too. This
 THE FRACTIONAL SCHRÖDINGER EQUATION . . . 11

leads to
∑
N ∑
N
b2n x2n = Ek a2n x2n (3.4)
n=0 n=0
∑
N ∑N
b2n+1 x2n+1 = Ek a2n+1 x2n+1 (3.5)
n=0 n=0
A term by term comparison results in a system of non-linear equations of the type
{Ek an = bn (an )} on the set of variables {an , Ek }, which is solved numerically. For
practical calculations we set N = 20, which yields an accuracy of the calculated energy
levels of about 0.25% for the ground state.
Results are presented in figures 2 and 4. For α = 2 we obtain the classical trigono-
metric functions. For decreasing α the eigenfunctions show a increasing tendency to
shift towards the walls. This is exactly the behavior, which is not modeled by the
trigonometric test functions presented in the last section. In figure 5 we compare the
determined energy levels with the presented energy formulas. Especially Ek∼ (α) from
(2.38) is a useful approximation.

4. Conclusion
We have demonstrated, that the non-local character of the fractional operators used
in the fractional Schrödinger equation indeed needs special attention. Concepts like box
normalization, WKB-approximation or piecewise solution may work well in a classical
local approach, but cause errors when applied to non-local problems.
On the other hand, we have shown, that such local strategies may lead to useful
approximations e.g. in low-level (α = 2 − ϵ) fractional problems.
This may be understood from
lim ⊔ Dα =∞
RZ D
α
(4.6)
q→∞ RZ
1
as a consequence of the specific weight w(ξ) = ξα+1 in the integral definition of the
Riesz fractional derivative definition (2.2) with the property
lim w(ξ) = 0 (4.7)
ξ→∞
The infinite potential well serves as a helpful tool to demonstrate the consequences
of different approaches.
 12 R. HERRMANN

Acknowledgment
We thank A. Friedrich for useful discussions.

References
[1] Abramowitz, M. and Stegun, I. A. (1965). Handbook of mathematical functions Dover Publications,
New York
[2] Bayin, S. S. (2012). On the consistency of the solutions of the space fractional Schrödinger equation
arXiv:1203.4556v1 [math-ph], J. Math. Phys. 53 042105
[3] Caputo, M. (1967) Linear model of dissipation whose Q is almost frequency independent Part II
Geophys. J. R. Astr. Soc 13, 529–539
[4] Dong, J. and Xu, M. (2007). Some solutions to the space fractional Schrödinger equation using
momentum representation method J. Math. Phys. 48, 072105
[5] Erdelyi, A., Magnus, A., Oberhettinger, F. and Tricomy, F. G. (1953) Higher transcendental
functions Vol. II, Bateman Manuscript Project, California Institute of Technology, McGraw Hill,
New York
[6] Feller, W. (1952). On a generalization of Marcel Riesz’ potentials and the semi-groups generated
by them Comm. Sem. Mathem. Universite de Lund, 72–81
[7] Guo, X. and Xu, M. (2006). Some physical applications of fractional Schrödinger equation J. Math.
Phys. 47, 082104
[8] Hawkins, E. and Schwarz, J. M. (2012). Comment ”On the consistency of solutions of the space
fractional Schrödinger equation” arXiv:1210.1447 [math-ph]
[9] Herrmann, R. (2005) Properties of a fractional derivative Schrödinger type wave equation and a
new interpretation of the charmonium spectrum arXiv:math-ph/0510099
[10] Herrmann, R. (2011) Fractional calculus - an introduction for physicists, World Scientific Publish-
ing, Singapore
[11] Liouville, J. (1832). Sur le calcul des differentielles á indices quelconques J. École Polytechnique
13, 1–162
[12] Jeng, M., Xu, S.-L.-Y., Hawkins, E. and Schwarz J. M. (2008). On the non-locality of the fractional
Schrödinger equation arXiv:0810.1543v1 [math-ph], J. Math. Phys. 51, 062102 (2010)
[13] Laskin, N. (2002). Fractional Schrödinger equation Phys. Rev. E 66, 056108–0561014
[14] Oldham, K. B. and Spanier, J. (1976) The fractional calculus, Dover Publications, Mineola, New
York
[15] Podlubny, I. (1999) Fractional differential equations, Academic Press, New York
[16] Riemann, B. (1847) Versuch einer allgemeinen Auffassung der Integration und Differentiation
in: Weber, H. and Dedekind, R. (Eds.) (1892) Bernhard Riemann’s gesammelte mathematische
Werke und wissenschaftlicher Nachlass, Teubner, Leipzig, reprinted in Collected works of Bernhard
Riemann, Dover Publications (1953), 353–366
[17] Riesz, M. (1949). L’integrale de Riemann-Liouville et le probléme de Cauchy Acta Math. 81, 1–222
[18] Weyl, H. (1917). Bemerkungen zum Begriff des Differentialquotienten gebrochener Ordnung
Vierteljahresschr. Naturforsch. Ges. Zürich, 62, 296–302

1 gigaHedron
Berliner Ring 80
D-63303 Dreieich, Germany
e-mail: herrmann@gigahedron.com

View publication stats