Proceedings of the 18th International Conference

on Computational and Mathematical Methods

in Science and Engineering, CMMSE 2018
July 9–14, 2018.

Noisy Pais-Uhlenbeck Oscillator

E. Urenda-Cázares1 , P. B. Espinoza1 , A. Gallegos1 , R. Jaimes-Reátegui1 ,

J. E. Macı́as-Dı́az2 and H. Vargas-Rodrı́guez1
1 Departamento de Ciencias Exactas y Tecnologı́a, Centro Universitario de los Lagos,
Universidad de Guadalajara, Mexico
2 Departamento de Matemáticas y Fı́sica, Universidad Autónoma de Aguascalientes
emails: ernesto.urenda@academicos.udg.mx, wolfgang@culagos.udg.mx,
gallegos@culagos.udg.mx, rjaimes@culagos.udg.mx, jemacias@correo.uaa.mx,
hvargas@culagos.udg.mx

Abstract

In this work, we consider the Pais-Uhlenbeck oscillator with additive and multiplica-
tive white noise. The equation is solved with the fourth order Runge-Kutta method and
the stochastic variable is generated via Box-Müller algorithm. We show the correspond-
ing numeric solutions and compare the results with and without noise. We note that
the noise (additive and multiplicative) under similar conditions is less influential if a
forced term is included in the oscillator.
Key words: Pais-Uhlenbeck oscillator, additive noise, multiplicative noise, Runge-
Kutta method, Box-Müller algorithm
MSC 2000: 34F05, 60H10, 93E03

1 Introduction
One of the most known models of high-order derivatives is the Pais-Uhlenbeck oscillator
(PUO) [1]. This model was originally proposed in order to solve some divergence problems
in quantum field theory. The PUO is usually written in the following form:

d4 x d2 x
+ c + ex = f (t) (1)
dt4 dt2
where c and e are real constants, note that we have included the forced term f (t). This
oscillator can be interpreted as two coupled harmonic oscillators with frequencies ω1 and

CMMSE
c ISBN: 978-84-697-7861-6
 NOISY PAIS-UHLENBECK OSCILLATOR

ω2 , where c = ω12 + ω22 and e = ω12 ω22 [2]. The central point of this work is to take account
possible stochastic perturbations in form of white noise in the PUO. This idea is widely
discussed by Gitterman in [3]. In particular we want to include additive noise which can
be interpreted as an external perturbation to the system and multiplicative noise that it is
viewed as internal noise.

2 Noise in the PUO

In order to include noise in the equation (1), it is convenient to write it in the following
form:

dXt = a(t, Xt )dt + b(t, Xt )dBt , (2)

where Bt is the stochastic variable and the vectors Xt , a(t, Xt ) and b(t, Xt ) are:

     
dx ẋ 0
 , b(t, Xt ) =  0 
 dẋ   ẍ   
 dẍ  , a(t, Xt ) = 
dXt =    ...  0  (3)
x 
...
dx −cẍ − ex + f (t) αxn

where the point means temporal derivative, α is the noise intensity and the power n is equal
to 0 and 1 that it corresponds to additive and multiplicative noise respectively. Of course,
we can use a general function of x for multiplicative noise but we use the simplest form.
We solve the equation (2) using the usual fourth order Runge-Kutta method and we generate
the normalized stochastic variable using the Box-Müller algorithm [4].

3 Numerical results
In order to obtain numerical results we set c = 5 and e = 4, with initial conditions x(0) = 1,
...
ẋ(0) = 0, ẍ(0) = 1 and x (0) = 0. Also, we want to study the homogeneus PUO with
f (t) = 0 and the forced PUO with f (t) = sin2 (3t). For the intensity of additive noise we
use α = 0.02, 0.1 that it corresponds to 1% and 5% respectively of the amplitude of the
solution (approximately). For multiplicative noise we use α = 0.04, 0.2 that it corresponds
to 1% and 5% respectively of the parameter e. The results for additive and multiplicative
noise are shown in figures 1 and 2 respectively.

CMMSE
c ISBN: 978-84-697-7861-6
E. URENDA-CÁZARES ET AL.

No noise
1.5 Noise 1%
Noise 5%
1

0.5

0
x

-0.5

-1

-1.5

-2

-2.5
0 5 10 15 20 25 30 35

No noise
Noise 1%
1.5
Noise 5%

0.5

0
x

-0.5

-1

-1.5

-2

-2.5
0 5 10 15 20 25 30 35

Figure 1: Solution of homogeneous(top) and forced(down) PUO with additive noise. Red
color corresponds to the solution without noise, black color corresponds to the solution with
noise intensity α = 0.02 (1%) and blue color corresponds to the solution with noise intensity
α = 0.1 (5%)

CMMSE
c ISBN: 978-84-697-7861-6
 NOISY PAIS-UHLENBECK OSCILLATOR

2 No noise
Noise 1%
Noise 5%
1.5

0.5

0
x

-0.5

-1

-1.5

-2

-2.5

0 5 10 15 20 25 30 35

No noise
Noise 1%
1.5
Noise 5%

0.5

0
x

-0.5

-1

-1.5

-2

0 5 10 15 20 25 30 35

Figure 2: Solution of homogeneous(top) and forced(down) PUO with multiplicative noise.

Red color corresponds to the solution without noise, black color corresponds to the solution
with noise intensity α = 0.04 (1%) and blue color corresponds to the solution with noise
intensity α = 0.2 (5%).

CMMSE
c ISBN: 978-84-697-7861-6
E. URENDA-CÁZARES ET AL.

In order to visualize the noise influence quantitatively, we obtain the corresponding average
deviation S of the noisy solutions respect the solution without noise.1 The results are shown
in the next table:

Noise percent 1% 5%
S for Additive forced PUO 0.01190 0.05466
S for Additive homogeneous PUO 0.02592 0.09115
S for Multiplicative forced PUO 0.04865 0.14491
S for Multiplicative homogeneous PUO 0.06053 0.19479

Obviously we can see more noise influence with more percentage. However, note that the
additive and multiplicative (separately) noises affect more the solution for the homogeneous
case than the forced case.
Remark 1. In this work, we will extend the methodology described in this abstract adding
simultaneously additive noise to (1) and multiplicative noise to frequencies ω1 and ω2 that
they are associated to parameters c and e as it was mentioned in the introduction.

References
[1] A. Pais and G.E. Uhlenbeck, On Field Theories with Non-Localized Action, Phys.
Rev. 79 (1950) 145–165.

[2] A. Mostafazadeh, A Hamiltonian formulation of the Pais–Uhlenbeck oscillator that

yields a stable and unitary quantum system, Phys. Lett. A 375 (2010) 93–98.

[3] M. Gitterman, The Noisy Oscillator: The First Hundred Years, From Einstein Until
Now, World Scientific, Singapore, 2005.

[4] G. E. P. Box and Mervin E. Müller, A Note on the Generation of Random Normal
Deviates, Ann. Math. Stat. 29 (1958), 610-611.

1
We calculate S as s
PN
(xi,nonoise − xi,noise )2
i=1
S= (4)
N
where N is the number of steps in Runge-Kutta method.

CMMSE
c ISBN: 978-84-697-7861-6