Journal of Mathematics and Statistics 4 (3): 130-137, 2008

ISSN 1549-3644
2008 Science Publications
Corresponding Author: P.O. Babatola, Department of Mathematical Sciences, Federal University of Technology, Akure,
Ondo State, Nigeria
130

K-Step Rational Runge-Kutta Method for Solution of Stiff
System of Ordinary Differential Equations

P.O. Babatola, R.A. Ademiluyi and E.A. Areo
Department of Mathematical Sciences,
Federal University of Technology, Akure, Ondo State, Nigeria

Abstract: This study described the development, analysis and implementation of K-step implicit
rational Runge-Kutta schemes for solution of stiff system of ordinary differential equations. Its
development adopted taylor and binomial series expansion techniques to generate its parameters. The
analysis of its basic properties adopted dalhquist a-stability model test equation and the results showed
that the scheme was a-stable, consistent and convergent. Numerical results showed that the method
was accurate and effective.

Key words: K-step, rational runge-kutta, a-stability, stiff ODEs

INTRODUCTION

A differential equation of the form:


o o
y ) y(x y), f(x, y = = (1)

whose Jacobian possesses eigen values:


n j j j
1(1) j , iV U = + = (2)

where, 1 , satisfying the following conditions:

U
i
< < 0, j = 1(1)
n

(x) U min ) x ( U Max
j j
>>

or r(x) = 1
) x ( U min
) x ( U Max
j
j
>> (3)


o o
1
y ) y(x (x), E E(x) - (y y = + = (4)

where, E(x) is continuously differentiable, is a
complex constant with Re() <<0, with the exact
solution:


x
o
y(x) = E(x) +y e

(5)

consisting of two components E(x) which is slowly
varying in the interval of integration (x
o
, b), and the
second
x
o
y e

component decaying rapidly in the

= (6)

with | |
T
y(o) 1, 1 , 0 10 = whose solution is
obtained as:


Sin100x Cos100x
-0.00005 x
Cos100x - Sin100x
y(x) e
+ (
(

= (7)

Whose transitory phase is the entire interval of
integration 0 10 with 50 complete oscillation per
unit interval is an ODEs possessing these types of
properties are called stiff oscillating ODEs.
Most of the conventional Runge-Kutta schemes
cannot effectively solve them because they have small
region of absolute stability.
This perhaps motivated
[8]
to introduce a
rationalized Runge-Kutta scheme of the form confirm
existing phases in samples according to
Emmanualson
[4]
.


R
n i i
i 1
n 1 R
n i i
i 1
y WK
y
1 y VH
=
+
=
+
=
+

(8)
Where:
J. Math. & Stat., 4 (3): 130-137, 2008

131
K
i
=
|
|
.
|

\
|
+ +

=
S
1 i
j ij n i n
k a y , h c x hf


H
i
=
|
|
.
|

\
|
+ +

=
S
1 i
j ij n i n
k b z , h d x hg (9)

With

( ) ) y (x f Z z , x g
n n,
2
n n n
=

Subject to the constraints:

=
=
R
1 j
y 1
a c

=
=
R
1 j
y 1
b d (10)

Since method possesses adequate stability property
for solution of stiff ODEs, the papers consider the
extension of the scheme to a general step process so
that is can serve as a general purpose predictor for
multistep schemes:

=
+
=
+
+
+
+
=
R
1 i
1 1 1 m n
R
1 i
i i 1 m n
m n
H V y 1
K W y
y (11)

Where:


|
|
|
.
|

\
|
+ + =

=
+ +
R
1 j
1 ij 1 - m n 1 1 - m n 1
K b y , h C x hf K


|
|
|
.
|

\
|
+ + =

=
+ +
R
1 j
1 ij 1 - m n 1 1 - m n 1
H b Z , h d x hg H (12)

with

( ) ( )
n km1 n k-1 n m-1 n m-1
g x , Z f x , y
+ + + +
(13)

In the spirit of Ademiluyi and Babatola
[1]
the
scheme is classified into:
Explicit if the constraints (10) is such that a
ij
= 0
for j i
Semi-implicit if a
ij
= 0 for j > i
Implicit if a
ij
0 for at least one j > i

MATERIALS AND METHODS

Derivation of the Method: In this research, the
parameters Vi, Wi, Ci, di, a
ij
, b
ij
are to be determined
from the system of non-linear equations generated by
adopting the following steps:

Obtained the Taylor series expansion of Kis and
His about point (x
n
, y
n
) for i = 1(1)R
Insert the series expansion into (10)
Compare the final expansion with the Taylor series
expansion of y
n+1
about (x
n
,y
n
) in the power series
of h

The number of parameters normally exceeds the
numbers of equations, but in the spirit of
[9]
, Gill
[7]
and
Blum
[3]
, these parameters are chosen as to ensure that
(the resultant computation method has:

Adequate order of accuracy of the scheme is
achieved
Minimum bound of local truncation error
Large maximize interval of absolute stability
Minimum computer storage facilities

One-step one-stage schemes: By setting M = 1 and
R = 1, in Eq. 11 the general one-step one-stage scheme
is of the form:


1 1 n
1 1 n
1 n
H V y 1
K W y
y
+
+
=
+
(14)

Where:

( )
1 11 n 1 n 1
K a y , h c x hf K + + =

( )
1 11 n 1 n 1
H b z , h d x hg H + + = (15)

( ) ) y , f(x Z z x g
n n
2
n n n
= (16)

and


n
n
y
1
Z = (17)
with the constraints:
c
1
= a
11

J. Math. & Stat., 4 (3): 130-137, 2008

132
d
1
= b
11
(18)

The binomial expansion theorem of order one on
the right hand side of (10) yields:

s) order term (higher H V y k W y y
1 1
2
n 1 1 n 1 n
+ + =
+
(19)

While The Taylor series expansion of y
n+1
about y
n

gives:

5 ) 4 (
n
4
) 3 (
n
3 2
n n 1 n
h 0 y
24
h
y
6
h
y
2
h
y h y y + + + + + =
+
(20)

Adopting differential notations:


( )
( )
( )( ) ( )
n n
n x n y n
(3) 2
n xx n xy n yy y x n y
2
n y n
(4) 2 3
n xxx n xxy n xyy n yyy
2
y xx n xy n yy
2
x n y x y n y
3 2 2
n y n n y n
y f
y f f f Df
y f 2f f f f f f f f
D f f Df
y f 3f f 3f f f f
f f 2f f f f
f f f 3f 3f f fy
= D f + f D f + 3Df Df + fy Df
=
= + =
= + + + +
= +
= + + +
+ + +
+ + + +
(21)

substitute (21) into (20), we have:


( )
( )
5
n
2
y y n n
2
y n
3
4
n y n
2
3
n
2
n n 1 n
h 0 Df f Df Df 3 f D f f D
! 4
h
Df f f D
! 3
h
Df
! 2
h
hf y y
+ + + + +
+ + + + =
+
(22)

Similarly the Taylor series expansion of K
1
about
(x
n
y
n
) is:

( )
( )
)
1
2 n 1 x 11 1 y
4
1
2 2 2
1 xx 1 11 1 xy 11 1 yy
f c hf a k f
K h 0(h
c h f 2c ha k f a k f
| |
+ + +
|
= +
|
+ +
\ .
(23)

Collecting coefficients of equal powers of h,
Eq. 23 can be rewritten in the form:

K
1
= hA
1
+ h
2
B
1
+ h
3
D
1
+ 0h
4
(24)

where,

A
1
= f
n
,. B
1
= c
1
(f
x
+f
n
f
y
) = C
1
Df
n
2 2 2
1 1 1 y 1 xx n xy n yy n y n
D = c B f + C (f +2f f +f f ) = Df f +D f (25)

In a similar manner, expansion of H
1
about
yy
2
n n zz xxy
n
xx
xxz
f y f 2 g , f
y
f 2
g = +

= (x
n
, z
n
) yields:

H
1
= hN
1
+ h
2
M
1
+
h
3
R
1
+ 0h
4
(26)

Where:


( )
( )
1 n n n
1 1 x n z 1 n
2 2
1
2 1 1 1 z 1 xx n xz n zz
2 2
1
2 1 z n n
N =g(x , z ) = g
M = d (g + g g ) = d Dg
R = d Mg d g 2g g g g
d g Dg D g
+ + +
= +
(27)


n x xx
n x xx 2 2 2
n n n
n x
z y xz xy
n n
2
xzz x n xyy,
2 2 4
zzz n y n yy n yyy
f f f
g , g , g
y y y
2f 2f
g f , g f
y y
g 2f 2y f
g 4y f 6y f y f

= = =

= + = +
`

= + +
)
(28)

Substitute (28) into (26), we obtained:


2
n 1 n
1 1 n 2 2
n n n
f -d 2f
N , M Df
y y y
| |
= = +
|
\ .



2 2
1 n n
1 y n 2
n n n
2
2 n n
1
2 n x
n n
d 2f f
R f Df
y y y
2f f
D f f
y y
( | || |
= + +
( | |
( \ .\ .
| ( | |
+ + ( |
|
( \ . \
(29)

Using (25) and (26) in (19), to get:

( )
( ) (
( ) ( )
( )
2 3 4
n 1 n 1 1 1 1
2 2 3 4
n 1 1 1 1
2 2 2
n 1 1 n 1 1 1 1 n 1 1
2 4
1 1 n 1 1
y y W hA h B h D 0h
y V hN h M h R 0h
= y WA y VN h WB -y VM h
WD y VR 0h
+
= + + + +
(
+ + +

+
+ +
(30)

Comparing the coefficients of the powers of h in
Eq. 22 and 30, we obtained:


n 1 1
2
n 1 1
f N V y A W = (31)

Since
J. Math. & Stat., 4 (3): 130-137, 2008

133
A
1
= f
n
N
1
=
2
n
n
y
f
, Eq. 31 yields:
W
1
+ V
1
= 1 (32)

Similarly from coefficients of h
2
in Eq. 22 and 30,
we have:

W
1
B
1

n 1 1
2
n
f D M V y = (33)

Also from (22) and (31) we obtained:

( )
2
Df
d V
y
f 2
Df d V C W
n
1 1
n
2
n
n 1 1 1 1
=
|
|
.
|

\
|
+ + (34)

W
1
C
1
+ V
1
d
1
= (35)

Putting Eq. 35 and 32 together, we have a system
of non linear simultaneous equations:

W
1
+ V
1
= 1
W
1
C
1
+ V
1
d
1
= (36)

with the constraints

a
11
= c
1

b
11
= d
1
(37)

Solving Eq. 37 and 38, we obtained:

V
1
= W
1
= c
1
= a
11
= , d
1
= b
11
=

Substituting these values in Eq. 14 we obtain a
family of one-step, one stage schemes of the form:


1
n
1
2
1
n
1 n
H
2
y
1
K y
y
+
+
=
+
(38)

Where:

K
1
= hf (x
n
+ h, y
n
+ K
1
)

H
1
= hg(x
n
+ h, z
n
+ H
1
) (39)

with

V
1
= , W
1
= , d
1
= c
1
= , a
11
= b
11
= Eq.
20 becomes:

1 n
1
4
1
n
1 n
H y
4
3
1
K y
y
+
+
=
+
(40)

Where:

K
1
= hf (x
n
+ h, y
n
+ K
1
)

H
1
= hg(x
n
+ h, z
n
+ H
1
) (41)

with

W
1
= 1/3 , V
1
= 2/3, a
11
= C
1
= 1/3, b
11
= d
1
= 7/12


1 n
1
3
1
n
1 n
H y
3
2
1
K y
y
+
+
=
+
(42)

Where:

K
1
= hf (x
n
+ 1/3 h, y
n
+ 1/3 K
1
)
H
1
= hg(x
n
+ 7/12 h, x
n
+ 7/12 H
1
) (43)

The basic properties of the method: The basic
properties required of a good computational method for
stiff ODEs includes consistency, convergence and
stability and A-stability.

Consistency: A scheme is said to be consistent, if the
difference equation of the computation formula exactly
approximate the differential equation it intends to
solve
[2]
.
To prove that Eq. 11 is consistent.
Recall that:

=
+
=
+
+
+
+
=
R
1 i
i i 1 k n
R
1 i
i i 1 - k n
k n
H V y 1
K W y
y
(44)

Subtract y
n+k-1
on both sides of Eq. 44:


1 - k n
R
1 i
i i 1 - k n
R
1 i
i i 1 - k n
1 - k n k n
y
H V y 1
K W y
y y
+
=
+
=
+
+ +

+ +
+
=

(45)

=
+
+
=
+

R
1 i
i i 1 - k n
i i
2
1 - k n
R
1 i
i
H V y 1
H V y Ki W
(46)
J. Math. & Stat., 4 (3): 130-137, 2008

134
with


|
|
|
.
|

\
|
+ + =

=
+
R
1 j
j ij n i 1 - k n i
k a y , h C x hf K


|
|
|
.
|

\
|
+ + =

=
+
R
1 j
j ij n i 1 - k n i
H b Z , h d x hg H (47)

Substituting (48) into (47), dividing through by h
and taking limit as h tends to zero, obtain:


)
R
n k n k 1
h o
i n k-1 n k-1
i 1
2
n k-1 1 n k 1 n k-1
limy y
Wf(x , y )
h
y Vg(x , y
+ +

+ +
=
+ + +

(48)

but


( )
n+k-1 n+k-1 n k-1, n k 1 2
n k 1
1
g(x , Z ) = f x y
y
+ +
+
(49)

then

Lt
( )
R
n k n k 1
i i n k 1, n k 1
i 1
y y
(W V)f x y
h
+ =
+ +
=

= +

, but
1 V W
R
1 L
i i
=
|
|
.
|

\
|
+

=


( )
1 - k n 1 - k n 1 k n
y , x f ) x ( y
+ + +
= (50)

Hence the method is consistent.

Convergence: Since the proposed scheme is one step,
the numerical scheme (11) for solving ODEs (1) is said
to be convergent, and it is consistent, by Lambert
(1973) when it is applied to initial value problem (1)
generated a corresponding approximation y
n
which tend
to the exact solution y(x
n
) as n approaches infinity, that
is:

y
n
y(x
n
) as n

Let e
n+k
and T
n+k
denote the discretization and
Truncation errors generated by (1) respectively.
Adopting Binomial expansion and ignoring higher
term, Eq. 1:


R R
2
n+k n k-1 i i n k-1 i i
i 1 i 1
y = y WK y VH
+ (higher order term)
+ +
= =
+

(51)
If y(x
n+k
) is the approximate theoretical solution, it
seen to satisfy the difference equation:


R R
2
n k n k-1 i i (n k-1) i i
i 1 i 1
n+k
y(x ) y(x ) WK y VH
+ (higher order term) + T
+ + +
= =
= +

(52)

Subtract Eq. 51 from 52:

( )
( ) ) ( )
n
n k n k n k 1 n k 1
(x )
2 n n 2 n n
1 n n 1 n n n k
y y(x ) y y(x )
h x y h (x , y ;h)
h (x , y(x );h x , y ;h T
+ + + +
+
=
(
+

+ +
(53)

where,
2
(x
n
(y(x
n
);h) are assumed to be continuous
functions in the domain:

a x b, y < , 0 h h
o
defined as:

( )
( )
( )
y
2
n
R
1 n k 1 n i i 2 n k ; n k 1
i 1
R
i i n k 1 n k 1 2
i 1
(x k 1)
h x y(x ); h WK h x Z(x )h)
h
VH x y
y

+ + +
=
+ +
=
+
=

= =

(54)

This Eq. 54 modifies into:


( )
| ) {
n k k k 1 2 n k 1 n k-1
2 n k-1 n k-1 1 n k 1
1 n k 1 n k 1 n k
e e h x , y(x ;h
(x , y ;h) h y(x ;h)
(x , y ;h) T
+ = + + +
+ + + =
+ + +
+ (

+
+
(55)

By taking the absolute value on both sides of
Eq. 61 we have inequality:

T e hL e Kh e e
1 k n 1 k n 1 k n k n
+ + +
+ + + +
(56)

where, L and K are the Lipschitz constant for
1
(x,y;h)
and
2
(x,y;h) respectively and:


n k
a x b
T T
+

=
(57)

By setting N = L+ K
Inequality (57) becomes:


T ) hN 1 ( e e
1 k n k n
+ +
+ +
(58)

By adopting this theorem on convergence of
sequence of real numbers quoted without proof from
[7]
,
that is:
J. Math. & Stat., 4 (3): 130-137, 2008

135
If {e
j
, j = o(1)
n+k
} be set of real number. If there
exist finite constants R and S such that:


1 - k n i i
(1) 1 i , S e R e 1
+
= +
(59)


1 R , e R s
1 R
1 R
e then
o
i
i
i
+
|
|
.
|

\
|

(60)

Thus (59) becomes:


( )
( )
o
k n
k n
1
k n
e hN 1
hN
hN 1
e
+
+

+
+ +
+

(61)

since


) a x ( N hN ) k n (
k n
k n
e e hN 1
+
+
+
= = + (62)

and

x
n+k
b, then x
n+k
a b a

Consequently:

e
N(xn+k a)
e
N(b-a)
(63)


o
) a b ( N
) a b ( N
k n
e e T
hN
1 e
e

+
+
|
|
.
|

\
|

(64)

Considering Eq. 64 and adopting first mean value
theorem:

|
( ))
2 n k 1 n k 1 2 n k 1
n k
n k 1 1 n k 1 n k 1
1 n k 1 n k-1
2 n k 1 n k 1 2 n k 1
n k 1 2 n k 1 n k 1
n k 1
(x h, y(x h)) (x
T h
y(x ) h (x 0h, y(x 0h)
(x , y x
(x h, y(x h)) (x h,
h y(x ) (x h, (x h, y
(x
+ + +
+
+ + +
+ +
+ + +
+ + +
+
+ + (
=
(
+ + + +
(

(

+ +
= + + +
+
|
1 n k 1 n k 1
h)
(x h, y(x )) , 0 1
+ +
(
(
(
(

+
(65)

By taking the absolute value of (55) on (66) both
sides Eq. 46 into consideration, we have:


( )
2
n k 1 n k 1
2
n k 1 n k 1
T hL y (x 0h y(x ) jh
hk y(x h y(x ) Mh
+
+ +
= + +
+ + +
(66)

where, M and j are the partial derivative of
1

and
1
with respect of x respectively.
By setting Q = j + m and


a x b
y sup (y (x))

= (67)

Therefore, Eq. 54 yields:

T = h
2
(NY + Q) (68)

By substituting (57) into (52), we have:


o
) a b ( N ) a b ( N 2
k n
e e Q) (NY e h e

+
+ + (69)

Assuming no error in the input data, that is e
o
= 0.
Then the limit as h tends to zero, we obtain in
Eq. 70 yields:

0 e lim
k n
o h
n
=
+

which implies that:

) x ( y y lim
n n
o h
n
=

Thus establishing the convergence of scheme (11).

Stability Properties: To analyse the stability property
of this schemes, apply scheme (1) to Dalhquist
[4]

stability scalar test initial value problem:


o o
y ) y(x y, y = = (70)

to obtained a difference equation:


1 - k n k n
y ) z ( y
+ +
= (71)

with the stability function:


e ZB) - (I ZV 1
e ZA) - (I ZW 1
) z (
1 - T
-1 T
+
+
= (72)

Where:

W
T
= (W
1
, W
2
W
r
)
V
T
= (V
1
, V
2
V
r
)

To illustrate, this we consider the one-step, one-
stage scheme:


1 1 n
1 1 n
1 n
H V y 1
k W y
y
+
+
=
+
(73)
J. Math. & Stat., 4 (3): 130-137, 2008

136
Where:

K
1
= hf(x
n
+c
1
h, y
n
+ a
11
K
1
)

H
1
= hg(x
n
+ d
1
h, z
n +
b
11
H
1
) (74)

Applying (73) to the stability test Eq. 70 we obtain
the recurrent relation:


n 1 n
y ) z ( y =
+
(75)


with stability function:


1
1 11
1
i 11
1 WZ(1 a Z)
(z)
1 VZ(1 b Z)

+
=
+
(76)

For example, to analyze scheme (74) will introduce
a convergent and stable approximation to the solution
of stability function if:

1
Z 1
Z 1
) z (
4
3
4
1
p

+
= (77)

That is (- <x < o), the scheme is A-stable because
the interval of absolute stability is (-, o).

RESULTS AND DISCUSSION

In order to access performance of the schemes the
following sample problems were solved, with the
schemes adopting Fehlberg
[6]
approach.

Problem 1: Consider the stiff system of ODEs of the
form:

Y=AY (78)
Where A =
(
(
(

12 0 . 2 0
0 0 . 5 0
0 99 . 4 0 . 1
(79)

With initial condition y(o) = (2, 1, 2).

In the theoretical interval 0x1. Its general
solution is:

|
|
|
|
.
|

\
|
+ =
=
+ =

x 12 x 5
3
5x -
2
x 5 -x
1
e e ) x ( y
e ) x ( y
e e ) x ( y


Its numerical solution is found in Table 1.

Problem 2: The second sample problem considered is
the stiff system of initial values problems of ODEs
below:

1
2
3
4
y
0.5 0 0 0 0
y 0 1.0 0 0 0
y withinitial condition
0 0 9.0 0 0
y
0 0 0 0 10.0
y
(
| |
(
|

(
|
=
(
|

(
|
|
(
\ .



y (0) = [1 1 1 1]

The results are shown in Table 2.




Table 1: Numerical result of k-step implicit rational runge-kutta schemes for solving stiff systems of ordinary differential equations
Y1 Y2 Y3
------------------------------------------------------------------------------------------------------
X Control step size E1 E2 E3
0.1980099667D+01 0.9706425830D+00 0.8869204674D+00
0.3000000000D-01 0.3000000000D-01 0.8291942688D-09 0.3281419103D-07 0.8161313500D-05
0.1885147337D+01 0.8379203859D+00 0.4917945068D+00
0.1774236000D+00 0.1771470000D-01 0.9577894033D-01 0.3422855333D-08 0.5357828618D-06
0.1791235536D+01 0.7191953586D+00 0.2663621637D+00
0.3307246652D+00 0.1046033532D-01 0.11050933794D-10 0.35587255336D-09 0.3474808041D-07
0.1694213422D+01 6088845946D+00 0.1365392880D+00
0.4977858155D+00 0.6176733963D-02 0.1269873096D-11 0.3655098446D-10 0.2146555961D-08
0.1556933815D+01 0.4729421983D+00 0.4953161076D-01
0.7512863895D+00 0.3647299638D-01 0.1425978891D-08 0.3505060447D-07 0.1010194837D-05
0.1435390902D+01 0.3709037123D+00 0.1867601194D-01
0.9951298893D+00 0.2153693963D-01 0.1594313570D-09 0.3316564301D-08 0.4481540687D-07
J. Math. & Stat., 4 (3): 130-137, 2008

137
Table 2: Numerical result of k-step implicit rational runge-kutta schemes for solving stiff systems of ordinary differential equations
Y1 Y2 Y3 Y4
--------------------------------------------------------------------------------------------------------------------
X Control step size E1 E2 E3 E4
0.9950124792D+00 0.9900498337D+00 0.9139311928+00 0.9048374306D+00
0.3000000000D-01 0.3000000000D-01 0.2597677629D-10 0.4145971344D-09 0.2617874150D-05 0.3971726602D-05
0.9708623323D+00 0.9425736684D+00 0.5872698932D+00 0.5535451450D+00
0.1774236000D+00 0.1771470000D-01 0.3078315380D-11 0.4788947017D-10 0.2005591107D-06 0.2890213078D-06
0.9402798026D+00 0.8841261072D+00 0.3300866691D+00 0.2918382654D+00
0.3694667141D+00 0.1046033532D-01 0.3621547506D-12 0.5454525720D-11 0.1355160001D-07 0.1829417523D-07
0.9144602205D+00 0.8362374949D+00 0.1999708940D+00 0.1672231757D+00
0.5365278644D+00 0.6176733963D-02 0.4285460875D+13 0.6268319197D-12 0.9915873955D-09 0.1265158728D-08
0.8693495443D+00 0.7557686301D+00 0.8044517344D-01 0.6079796167D-01
0.8400599835D+00 0.3647299638D-01 0.4961209221D-10 0.6922001861D-09 0.5087490103D-06 0.5899525189D-06

CONCLUSION

From the above results (Table 1 and 2), it can be
seen that the proposed schemes are quite accurate,
convergent and stable.

REFERENCES

1. Ademiluyi, R.A. and P.O. Babatola, 2001. Semi-
Implicit RR-K formula for approximation of stiff
initial values problem in ODEs. J. Math. Sci.
Educ., 3: 1-25.
2. Ademiluyi, R.A., 2001. A new one-step method
for integration of ODEs. J. Nigerian Math. Soc.,
21: 41-57.
3. Blum, E.K., 1952. A modification of runge-kutta
fourth order method. Maths Comput., 16: 176-187.
4. Dalhquist, G., 1963. A special stability problem for
linear multistep methods. BIT3., 27-43.
5. Fatunla, S.O., 1987. Recent advances in ODFEs.
Proceeding of the 1st International Conference on
Numerical Analysis, London, pp: 27-31.



















6. Fehlberg, E., 1964. New higher order R-K,
formulae with step size control for systems of first
and second order differential equations. Angew
Math. Mech., 44: 83-88.
7. Gill, S., 1951. A process for step-by-step
integration of differential equations in an automatic
digital computing machine. Proc. Cambridge
Philors Soc., 47: 95-108.
http://serv1.ist.psu.edu:8080/showciting;jsessionid
=6385285C3878AEB94269786C54943BFF?cid=1
984208&sort=recent.
8. Hong Yuanfu, 1982. A class of A-stable or A ()
stable Explicit Schemes. Proceeding of BAILJI
Conference Computational and Asymptotic
Method for Boundary and Interior Layer, Trinity
College, Dublin, pp: 236-241.
9. King, R., 1966. Runge-Kutta methods with
constraint minimum error bound. Maths Comp.,
20: 386-391.
10. Lambert, J.D., 1973. Predictor-corrector algorithm
with identical region of stability. SIAM J.
Numercal Anal., 8: 337-344.