Journal of Applied Mathematics and Computational Mechanics 2014, 13(3), 123-128

POWER SERIES SOLUTION OF FIRST ORDER MATRIX

DIFFERENTIAL EQUATIONS

Stanisław Kukla, Izabela Zamorska

Institute of Mathematics, Czestochowa University of Technology
Częstochowa, Poland
stanislaw.kukla@im.pcz.pl, izabela.zamorska@im.pcz.pl

Abstract. In the paper, the solution of second order differential equations with various
coefficients is presented. The concerning equations are written as first order matrix differ-
ential equations and solved with the use of the power series method. Examples of applica-
tion of the proposed method to the equations occurring in the technical problems are
presented.

Keywords: power series, differential equations

Introduction

The goal of many researchers’ work is finding new and improving existing
various methods for solving ordinary differential equations. In various types of
differential equations (linear and non-linear, with constant or variable coefficients),
occurring, for example, in technical specification number of issues, it is possible to
designate the exact solution analysis, and in the others, however, it is necessary to
approximate methods. One of the classical methods, known since the seventeenth
century, in solving differential equations is the method of power series, involving
the appointment of a solution in the form of an infinite Taylor series (series coeffi-
cients are obtained from recursive equations). This method was applied e.g. by
Eisenberger in [1] to solve the second order differential equation describing rods
vibration problems and by Quaisi in [2] for the non-linear free vibration of beams
with restrained ends. In turn, Zhou in [3] proposed a method of differential trans-
formation - an improved method of power series, differing the way of determining
the coefficients of the series. A power series method with domain partition in
an implemented matrix formulation is another method, alternative to other tech-
niques of power series [4].
In this paper, the method of solving second order ordinary differential equation
will be presented by transforming this equation in the system of differential equa-
tions of the first order, then presenting it in matrix notation and solving with the
use of power series method. As examples there will be presented solutions of two
classic equations: the Airy equation and equation occurring in the description of
the rod’s vibration.
124 S. Kukla, I. Zamorska

1. Formulation and solution of the problem

Let us consider the first order matrix differential equation

Y′ ( x ) = B ( x ) Y ( x ) + F ( x ) (1)
T T
where Y ( x ) =  y1 ( x ) y2 ( x ) L ym ( x ) , F ( x ) =  f1 ( x ) f 2 ( x ) L f m ( x ) 
and B ( x ) = bi j ( x )  . The equation is completed by initial condition
1≤ i , j ≤ m

Y ( 0 ) = Y0 (2)

We are looking on solution of the equation (1) in the form of power series
∞
xn T
∞
xn
Y ( x ) = ∑ Yn where Yn = [ yn1 yn 2 L ynm ] . Assuming that B ( x ) = ∑ B n
n=0 n! n=0 n!
∞
xn
and F ( x ) = ∑ Fn , substituting into equation (1) we obtain:
n =0 n!
′
 ∞ xn   ∞ x n  ∞ xn  ∞ xn
 ∑ Yn  =  ∑ B n  ∑ Yn  + ∑ Fn (3)
 n=0 n !   n= 0 n!  n = 0 n!  n=0 n!

After transformations in equation (3) we get the following recursive relation

Y1 = F0 + B0 Y0
n
n
Yn +1 = Fn + ∑   B k Yn − k n = 1,2,... (4)
k =0  k 

The coefficients Yn may be written as:

Yn = Φn + Ψ n Y0 n = 0,1,2,... (5)
n
n
where Φ0 = 0, Ψ0 = E (the unit matrix) and Φ n +1 = Fn + ∑   B n − k Φ k ,
k =0  k 
n
n
Ψ n +1 = ∑   B n − k Φ k .
k =0  k 

Finally, a solution of an inhomogeneous matrix differential equation (1) can be

expressed as a sum
∞
xn  ∞ xn 
Y ( x ) = ∑ Φn +  ∑ Ψn  Y0 = Φ n ( x ) + Ψ n ( x ) Y0 (6)
n =0 n !  n =0 n! 
 Power series solution of first order matrix differential equations 125

2. Examples of method’s application

Let the ordinary differential equation of second order with variable coefficients

d 2 y ( x) dy ( x )
2
+ a2 ( x ) + a1 ( x ) y ( x ) = f ( x ) (7)
dx dx

with the conditions:

y ( 0 ) = y0 , y ( l ) = yl (8)

Introducing functions y1 ( x ) = y ( x ) , y2 ( x ) = y1′ ( x ) , equation (7) can be written as

a system
y1′ ( x ) = y2 ( x )
y2′ ( x ) = f ( x ) − a1 y1 ( x ) − a2 y2 ( x ) (9)

T
or in the matrix form as equation (1) where Y ( x ) =  y1 ( x ) y2 ( x )  ,
T  0 1 
F ( x ) = 0 f ( x )  and B ( x ) =  .
 − a1 ( x ) − a2 ( x ) 
Functions occurring in equation (7) are C ∞ class and may be written as
∞
xn ∞
xn ∞
xn
yi ( x ) = ∑ yin , ai ( x ) = ∑ ain (i = 1, 2), f ( x ) = ∑ f n (10)
n =0 n! n =0 n! n =0 n!

Initial and boundary equations (8) are as follows:

 y ( 0)   y (l ) 
Y0 =  1  , Yl =  1  (11)
 y2 ( 0 )   y2 ( l ) 

where y1 ( 0 ) = y0 , y1 ( l ) = yl , y2 ( 0 ) and y2 ( l ) can be, respectively, determined

from the relations

1 ∞
ln  ∞
ln
y2 ( 0 ) = y11 =  yl − y0 + ∑ y1n  , y2 ( l ) = ∑ y1n +1 . (12)
l n =2 n!  n=0 n!

Airy equation
Schrödinger equation, one of the fundamental equations of non-relativistic
quantum mechanics, under certain assumptions can be reduced to the Airy equation
126 S. Kukla, I. Zamorska

y′′ ( x ) + xy ( x ) = 0 (13)

which has a well-known analytical solution in the form:

x  2 3  2 3 
y ( x) = C1 J 1  x  + C2 J 1  x  (14)
− 3 
 3  
3 3
3

where J(·) is a Bessel function of the first kind and constant values C1, C2 depend
on the selected initial conditions.
Using the matrix notation we have
∞
xn
F ( x ) = 0 , B ( x ) = ∑ Bn = B0 + B1 x (15)
n =0 n!

0 1   0 0
where B0 =   , B1 =  −1 0  , B 2 = B3 = K = 0 . Recursive relations for series
0 0  
coefficients Yn+1 are

Y1 = B 0 Y0 , Yn+1 = B0 Yn + nB1Yn −1 n = 1,2,... (16)

Solution of Airy equation (13) in a matrix form may be written as

 ∞ xn 
Y ( x ) =  ∑ Ψn  Y0 (17)
 n=0 n! 

where the first few values of Ψn are:

Ψ0 = E
Ψ1 = B 0
Ψ2 = B02 + B1
Ψ3 = B30 + B0 B1 + 2B1B0
Ψ4 = B04 + B02 B1 + 2B0 B1B0 + 3B1B02 + 3B12
Ψ5 = B50 + B30 B1 + 2B02 B1B 0 + 3B0 B1B02 + 3B0B12 + 4B1B30 + 4B1B0 B1 + 8B12 B0

By using designations:
1 0  1 0  0 1  0 0 0 0 
M0 =   , M1 =   , M2 =   , M3 =   , M4 =  
0 1  0 0 0 0  1 0  0 1 
(B0 = M2 , B1 = –M3), it can be seen that sets of matrices {Ο, M0 , M1 , M2 , M3 , M4 }
with sets multiplication create a semigroup. The algebraic properties of pair
({Ο, M0 , M1 , M2 , M3 , M4 }, ×) describes the multiplication table (Table 1).
 Power series solution of first order matrix differential equations 127

Table 1
Multiplication table

× M0 M1 M2 M3 M4
M0 M0 M1 M2 M3 M4
M1 M1 M1 M2 Ο Ο
M2 M2 Ο Ο M1 M2
M3 M3 M3 M4 Ο Ο
M4 M4 Ο Ο M3 M4

d  dy 
Equation  p( x)  + k p( x) y = 0 (18)
dx  dx 

This type of equation occurs frequently in problems of mechanics, for example

in the vibration’s description of rods or strings. The functions occurring in the
equation have a physical interpretation: y(x) is a function of deflection, p(x) is
a cross-section area and k is a parameter characterized vibration frequency of
the mechanical system under consideration.
Suppose that p ( x ) = e −α x where α is a constant, x ∈ 0,1 . The general solution
of homogeneous equation (18) is [5]

x  x  x 
α
y ( x ) = e 2 C1 cos  4k − α 2  + C2 sin  2 4k − α
2
 (19)
  2   

Matrices B and F existing in the equation (1) are as follows:

0 1 
   0 1
B ( x) =  − p′ ( x )  =  = B0 , B n = 0 n = 1,2,...
−k  −k α 
 p ( x ) 

F ( x) = 0 (20)

Series coefficients of the solution of (18) are:

Y0 , Y1 = B 0 Y0 , Y2 = B 02 Y0 , ..., Yn = B 0n Y0 (21)

and the solution may be written in the form:

 ∞ xn 
Y ( x ) =  ∑ B0n  Y0 (22)
 n=0 n! 
128 S. Kukla, I. Zamorska

Conclusions

The method of solving second order linear differential equations, presented

in the work, may be used for solving n-th order linear equations. Any of these
equations can be represented by a system of n first order linear differential equa-
tions with more than one dependent variable and, as the result, by one first order
differential matrix equation. Although this method shouldn’t be applied indiscrimi-
nately, it is suitable to solve, particularly, equations of arbitrary order with variable
coefficients which typically arise in vibration or heat transfer problems.

References
[1] Eisenberger M., Exact longitudinal vibration frequencies of a variable cross-section rods, Applied
Acoustics 1991, 34, 123-130.
[2] Quaisi M.I., A power series solution for the non-linear vibration of beams, Journal of Sound
and Vibration 1997, 199(4), 587-594.
[3] Zhou K., Differential Transformation and Its Application to Electrical Circuits, Wuhan People
Republic of China, Huazhong University Press 1986 (in Chinese).
[4] Inaudi J.A., Matusevich A.E., Domain-partition power series in vibration analysis of variable
cross-section rods, Journal of Sound and Vibration 2010, 329, 4534-4549.
[5] Zamorska I., Longitudinal vibrations of a non-uniform rods coupled by double spring-mass
systems, Scientific Research of the Institute of Mathematics and Computer Science 2007, 1(6),
277-284.