Professional Documents
Culture Documents
2017
ABSTRACT
In this paper, the formulation of a class of second derivative blended block multistep methods for step
numbers k=3, 4,5 is considered through the Enright and the multistep collocation approaches. With these
approaches, we hope to improve the stability regions of the Adams Moultons Methods and thereby making
them suitable for the solution of stiff ordinary differential equations. These new methods proposed in this
paper turn out to be A-stable. Numerical examples obtained demonstrate the accuracy and efficiency of the
new blended block multistep methods.
Keywords: A-stable, Adams Moultons methods, Blended Block methods and Stiff ODEs,
1. INTRODUCTION
Most real life problems when modelled mathematically result in ordinary differential equations.
Some of the equations do not have analytic solutions as such the need for good numerical methods to
approximate their solutions. In this paper we are concern with the numerical solution of the stiff
initial value problem (1) using the second derivative linear multistep.
is continuous and differentiable. The second derivative k-step method takes the following form
k k k
åj= 0
a j yn + j = hå
j= 0
b j f n+ j + h2 å
j= 0
g j g n+ j (2)
Several methods have been developed to overcome this barrier theorem. Researchers like Gear
(1965),Lambert(1983), Butchers(1966), the second derivative methods of Enright (1974) ,Genin
(1974),Gamal and Iman (1998), Sahi et-al(2012) ,Mehdizadeh et-al (2012), Ehigie and
Okunuga(2014) and the third derivative method of Ezzeddine and Hojjati (2012), either relax the
condition to obtain A stable methods or incorporate off-step points to improve the stability
of the methods.
In this work, we consider the second derivative blended block multistep methods for step numbers
k=3, 4, 5, 6. With this approach, we hope to improve the stability regions of the conventional Adams
Moulton method and thereby making them suitable for the solution of (1)
24
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
numerical methods for solving stiff ODEs. Most conventional numerical methods are not A-stable
for k≥2 but blending two existing methods has the chances of producing one.
Blended linear multistep methods is a process where two existing linear multistep methods are
combined by multiplying the Jacobian of the blender with a constant multiplier to produce a single
method without altering the order of the Blended method for improved performance.
Enright (1972) blended the Adams Moultons method of order k+1 and the Backward Differentiation
Formula of order k to obtain the Enright Formula of order k+1. Enrightk 1 AMF k 1 khJ BDF k
Where J is the Jacobian and is a constant which determines the angle of the region of absolute
1 2
1
s 1 s s 1...s i 2ds
k!
k
k vi and vi
j 0 0
Skeel and Kong (1977) extended the Enright Blended Formulas and propose a sub class of the
Blended formulas as suitable for a general purpose ODE solver. The method possesses enhanced
stability properties with good region of absolute stability. Thus, the method proposed in this
paper is one that combines these desirable qualities for the direct solution of stiff system of
ODEs.
Enright approach and the same processes used by Enright was adopted to obtain the values of
the parameters k below
25
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
The Constant k
g k* Ok
To obtain the values of the constant we use the equation
1 2
1
s 1 s s 1...s i 2ds
k! 0
k
k vi v
where i (3.1)
j 0
1
1 7
v3
3! 0
s ( s 1)(s 1) 2 ds
360
1
1 17
v4 s ( s 1)(s 2)(s 1) 2 ds
4! 0 1440
1
1 41
v5
5! 0
s ( s 1)(s 2)(s 3)(s 1) 2 ds
5040
3 v0 v1 v2 v3
19
180
4 v0 v1 v2 v3 v4
3
32
5 v0 v1 v2 v3 v4 v5
863
10080
Adams Moulton Method k=3 is blended with the Backward Differentiation Formula (BDF) k=3.
Using the Enright’s approach
Adams Moulton k=3
1 5 19 9
yn + 3 - yn + 2 + hf n + hf n+ 1 - hf n+ 2 - hf n+ 3 = 0
24 24 24 24
BDF, k = 3
11 3 1
yn+ 3 - 3 yn+ 2 + yn+ 1 - yn - hf n+ 3 = 0
6 2 3
Blending the two gives
26
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
1 5 19 9 11 3 1
k Jh yn 3 3 yn 2 yn 1 yn hfn 3
yn 3 yn 2 hfn hfn 1 hfn 2 hfn 3 Ok
24 24 24 24 6 2 3
19
substituting γ* = from (3.3) into the above method
3 180
and simplifying yields
7 1 19
yn + 3 = yn + 2 + hf n - hf n+ 1 + hf n+ 2
1080 20 40
307 102 2 ''
- hf n+ 3 - h y n+ 3
5040 3960
Adams Moulton Method k=4 is blended with the Backward Differentiation Formula k=4.Using the
Enright’s approach.
3
substituing the value γ 4 from (3.3) into the above method and simplifying yields
32
17 1 41 47 3133 3
yn 4 yn3 hf n hf n 1 hf n 2 hf n 3 hf n 4 h 2 y '' n 4
5760 45 480 90 5760 32
Adams Moulton Method k=5 is blended with the Backward Differentiation Formula k=5.
27
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
863
substituing the value γ5 from (3.3) into the above method and simplifying yields
10080
41 529 373 1271
yn 5 yn 4 hfn hfn 1 hfn 2 hfn 3
25200 40320 7560 10080
2837 317731 863 2 ''
hfn 4 hfn 5 h y n5 (3.11)
5040 604800 10080
The blended block linear multistep method for implementation purposes is constructed using the Onumanyi
et al (1994, 1999) approach where the general continuous k –step LMM is expressed as:
t 1 m 1
y ( x ) j ( x ) yn j h j ( x ) f n j (3.12)
j 0 j 0
and extending it to the second derivative gives the general form of the continuous k –step 2nd
t 1 m 1 m 1
y ( x ) j ( x ) yn j h j ( x ) f n j h 2 ( x) y
j
''
n j (3.13)
j 0 j 0 j 0
where h is the step length, m the number of distinct collocation points and t, the number of
interpolation points. But for the purpose of this work the general form of our continuous k –step 2nd
28
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
m 1
y ( x) k 1 ( x) y n k 1 h j ( x) f n j h 2 k ( x) y '' n k (3.14) wher
j 0 Ok
t m 1
k 1 ( x )
i 0
j , i 1 xi , j 0,1,..., t 1
t m 1
j ( x) i 0
j ,i 1 xi , j 0,1,..., m 1 (3.15)
t m 1
k ( x)
i 0
j , i 1 xi , j 0,1,..., m 1 being the continuous coefficients of the method.
To construct the three-step BBLMM, the general form of the Three Step Blended Linear Multistep
Where a 2 ( x), b 0 ( x), b1 ( x), b 2 ( x), b 3 ( x), l 3 ( x) are the continuous coefficients of the method.
In order to determine the continuous coefficients in (4), we use the approach in Onumanyi et.al. (1994) to
obtain the continuous coefficients as given below.
a 2 (t + xn ) = 1,
æ 13t 2 29t 3 t 4 t 5 ö
43h ÷
hb 0 (t + xn ) = ç
çt - + - + - ÷
÷
ç
è 12h 54h 2 8h 3 90h 4 ÷
135 ø
æ9t 2 7t 3 t 4 t 5 ö
7h ÷
hb1 (t + xn ) = ç
ç - + - - ÷
÷
ç
è 4h 4h 2 2h3 20 h 4 5 ÷
ø
æ- 9t 2
5t 3 7t 4 t 5 ö
h÷
hb 2 (t + xn ) = ç
ç + - + - ÷
÷
ç
è 4h 2h 2 8h 3
10 h 4 5÷
ø
æ13t 2 139t 3 t 4 11t 5 ö
11h ÷
hb 3 (t + xn ) = ç
ç - + + - ÷
÷
ç
è 12 h 108h 2
2h 3
180 h 4 ÷
135 ø
(3.36)
æ- t 2
11t 3 t 4 t 5 ö
2h ÷
hl 3 (t + xn ) = ç
ç + - + + ÷
÷
ç
è 2 18h 4h 2
30 h 3 ÷
45 ø
Substituting (3.36) into (3.32) yields the continuous form of the3-step BLMM (3.37).
29
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
æ13t 2 139t 3 t4 ö
11t 5 11h ÷
+ çç - + + - ÷
÷ f
çè 12h 108h 2
2h 3 4
180h 135 ø ÷ n+ 3
æ- t 2 11t 3 t 4 t5 ö
2h ÷
+ çç + - + + ÷
÷ g (3.37)
çè 2 18h 4h 2 30h3 45 ø ÷ n+ 3
43 7 1 11 2 2 ''
yn = yn + 2 - hf n - hf n+ 1 - hf n+ 2 - hf n+ 3 + h y n+ 3
135 5 5 135 45
23 9 29 83 11 2 ''
yn + 1 = yn + 2 + hf n - hf n+ 1 - hf n+ 2 + hf n+ 3 - h y n+ 3 (3.38)
1080 20 40 540 180
7 1 19 307 19 2 ''
yn + 3 = yn + 2 + hf n - hf n+ 1 + hf n+ 2 + hf n+ 3 - h y n+ 3
1080 20 40 540 180
To construct the four- step BLMM, the general form of the Four- Step Blended Linear Multistep
The continuous coefficient are obtained using the same procedure in (3.27) and substituting same
into (3.39) gives the continuous form of the four step blended linear multistep method.
30
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
æ 7t 2 95t 3 25t 4 ö
÷
ççt - + - ÷
çç 3÷
% 6h 144h 2
128h ÷ ÷
y (t + xn ) = yn+ 3 + çç ÷ fOk
t6 ÷
n
çç 7t 5 210h ÷
÷
ççè 240h 4 -
+ - ÷
÷
640 576h5 ø
æ8t 2 64t 3 31t 4 13t 5 t6 7h ÷ö
+ çç - + - + - ÷
÷ f n+ 1
çè 3h 27 h 2 36h3 90h 4 108h 5
5÷ ø
æ- 3t 2
11t 3 51t 4 3t 5 t6 99h ö
÷
+ çç + - + - - ÷ f n+ 2
çè h 3h 2
32h 3
10h 4
48h 5
160 ÷
÷
ø
æ8t 2 32t 3 7t 4 11t 5 t6 9h ö
÷
+ çç - + - + - ÷
÷ f
çè 3h 9h 2
4h 3
30h 4
36h 5 ÷ n+ 3
10 ø
æ 7t 2 691t 3 947t 4 131t 5 25t 6 ÷ ö
çç- + - + - ÷
ç 5÷
+ çç 6h 432h 2
1152h 3
720h 4
1728h ÷ ÷
çç ÷
÷
f n+ 4
çç-
9h ÷
÷
è ÷
÷
ø
10
æt 2 25t 3 35t 4 t5 t6 3h 2 ö
÷
+ çç - + - + - ÷
÷ g (3.41)
çè 2 36h 96h 2
12h 3
144h 5
32 ø÷ n+ 4
Evaluating the continuous scheme (3.41) at t = 0, h, 2h, 4h yields the Four-Step BBLMM
11 17 19 17 1
yn + 1 = yn + 3 - hf n - hf n+ 1 - hf n+ 2 - hf n+ 3 + hf n+ 4
90 45 15 45 90
201 7 99 9 149 3 2 ''
yn = yn + 3 - hf n - hf n+ 1 - hf n+ 2 - hf n+ 3 + hf n+ 4 + h y n+ 4
640 5 160 10 640 32
11 7 83
yn + 2 = yn + 3 -
hf n + hf n + 1 - hf n + 2
1920 135 160
19 1831 1
- hf n + 3 + hf n + 4 + h 2 y '' n + 4
30 17280 6
17 1 41
yn + 4 = yn + 3 - hf n + hf n + 1 - hf n + 2
5760 45 480
47 3133 3
+ hf n + 3 + hf n + 4 + h 2 y '' n + 4 (3.42)
90 5760 32
The continuous coefficient are obtained using the same procedure in (3.27) and substituting same
into (3.25) gives the continuous form of the five step blended linear multistep method.
31
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
Evaluating the continuous scheme (3.39) at t = 0, h, 2h, 3h, 5h yields the Five-Step
BBLMM
158 52 344
yn = yn + 4 - hf n - hf n + 1 - hf n + 2
252 35 945
176 2 548 16
- hf n + 3 - hf n + 4 - hf n + 5 + h 2 y '' n + 5
105 35 4725 315
33 1737 337
yn + 1 = yn + 4 + hf n - hf n + 1 - hf n + 2
2800 4480 280
1023 339 2201 863
- hf n + 3 - hf n + 4 + hf n + 5 - h 2 y '' n + 5
1120 560 22400 1120
1 53 382
yn + 2 = yn + 4 - hf n - hf n + 1 - hf n + 2
630 2520 945
773 263 221 1
- hf n + 3 - hf n + 4 + hf n + 5 + h 2 y '' n + 5
630 630 7560 126
19 89 677
yn + 3 = yn + 4 - hf n - hf n + 1 + hf n + 2
8400 4480 7560
1933 323 48467 271
- hf n + 3 - hf n + 4 + hf n + 5 - h 2 y '' n + 5
3360 560 604800 10080
41 529 373
yn + 5 = yn + 4 + hf n - hf n + 1 + hf n + 2 (3.46)
25200 40320 7560
1271 2837 317731 863
- hf n + 3 + hf n + 4 + hf n + 5 - h 2 y '' n + 5
10080 5040 604800 10080
Following the work of Ehigie and Okunuga (2014), we observed that the three step block method is zero
32
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
2
K=5
K=3
K=5
im(z)
0
K=4
-1
-2
-3
0 0.5 1 1.5 2 2.5 3 3.5 4
Re(z)
Numerical Experiment
Problem 1
Problem 2
Problem3
33
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
Fig2: Solution to Problem 1 using the Three Step Blended Block Linear Multistep Second derivative Methods
1
0.8
Ok y2
y3
y1
ODE 23
ODE 23
Blended
solver
solver
Block
y2 Blended Block
y3 Blended Block
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25 30 35 40
Figure 2: Solution of problem 1 Computed with Three - Step Blended Block Adams Moutons Methods
Fig 3: Solution to Problem 1 using the Four Step Blended Block Linear Multistep Second derivative Methods
1
0.6
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25 30 35 40
Figure 3: Solution of problem 1 Computed with Four - Step Blended Block Adams Moutons
Fig4: Solution to Problem 1 using the Five Step Blended Block Linear Multistep Second derivative Methods
Methods 1
0.6
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25 30 35 40
Figure 4: Solution of problem 1 Computed with Five - Step Blended Block Adams Moutons Methods
y1 Exact
0.9 y2 Exact
y1 Blended
y2 Blended
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25 30 35 40
Figure 5: Solution of problem 2 Computed with Three - Step Blended Block Adams Moultons Methods
34
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
y1 Exact
0.9
Ok y2
y1
y2
Exact
Blended
Blended
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25 30 35 40
Figure 6: Solution of problem 2 Computed with Four - Step Blended Block Adams Moultons Methods
y1 Exact
0.9 y2 Exact
y1 Blended
y2 Blended
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25 30 35 40
Figure 7: Solution of problem 2 Computed with Five - Step Blended Block Adams Moutons Methods
3
y1 Exact
y2 Exact
2.5 y1 Blended
y2 Blended
1.5
0.5
-0.5
-1
-1.5
0 5 10 15 20 25 30 35 40
Figure 8: Solution of problem 3 Computed with Three - Step Blended Block Adams Moutons Methods
35
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
3
y1 Exact
y2 Exact
2.5
Ok y1
y2
Blended
Blended
1.5
0.5
-0.5
-1
-1.5
0 5 10 15 20 25 30 35 40
Figure 9: Solution of problem 3 Computed with Four - Step Blended Block Adams Moutons Methods
3
y1 Exact
y2 Exact
2.5 y1 Blended
y2 Blended
1.5
0.5
-0.5
-1
-1.5
0 5 10 15 20 25 30 35 40
Figure 10: Solution of problem 3 Computed with Five - Step Blended Block Adams Moutons Methods
CONCLUSION
A class of blended block linear Multistep second derivative methods have been constructed through the
Enright and multistep collocation approaches. The Region of absolute stability of the conventional Adams
Moulton methods has been greatly enhanced. (A-stable). These methods are all convergent. Numerical results
reveal the efficiency of the methods in solving stiff systems ODEs
REFERENCES
Dahlquist, G. (1963). A Special Stability problem for Linear Multistep Methods. Bit 3, 27-43.
Gear, C.W. (1965). Hybrid Methods for Initial Value Problems in Ordinary Differential Equations. SIAM J.
Numer. Anal. Ser B, 2, 69 – 86.
Lambert, J.D. (1973). Computational Methods in Ordinary Differential Equations. John Willey. New York.
Butcher, J.C. (1966). On the Convergence of Numerical Solutions to Ordinary Differential Equations. Math.
Comp. 20, 1 -10.
Chollom J.P. (2005): A study of Block Hybrid Methods with link o two-step Runge Kutta Methods for first
order Ordinary Differential Equations. PhD Thesis (Unpublished) University of Jos, Jos Nigeria.
Chollom J.P., Olatunbusin I.O. and Omagwu S (2012): A Class of A-Stable Block Explicit Methods for the
Solution of Ordinary Differential Equations. Research Journal of Mathematics and Statistics 4(2): 52-
56.
Dalquist, G. (1963). A Special Stability problem for linear Multistep Methods BIT. 3 27-43
Ehigie J.O. & Okunuga S.A. (2014). L( )-Stable Second Derivative Block Multistep Formula for stiff initial
Value Problems. International Journal of applied Mathematics 44(3). 7 -13
Enright, W.H (1972). Numerical solution of stiff differential equations 46(1972), 321-331, Dept of Computer
Science, University of Toronto, Toronto, Canada.
36
Omagwu et al..…. Int. J. Inno. Maths, Statistics & Energy Policies 5(1):24-37, 2017
Fatunla, S.O. (1991). Block Method for Second Differential Equations. International Journals of Computer
Mathematics. 41, 55-663
Sahi, R.K.,Jator,S.N and Khan, N.A. (2012). A Simpson’s Ok type second derivative method for stiff systems,
International journal of pure and applied mathematics. 81(4), 619-633.
Mehdizadeh, K. M., Nasehi,O,N. and Hojjati, G. (2012). A class of second derivative multistep methods for
stiff systems, Acta universitatis Apulensis, 30 171-188.
Onumanyi, P., Awoyemi, D.O., Jator, S.N. and Sirisena, U.W. (1994).New Linear Multistep Methods with
continuous Coefficients for first order Initial Value Problems. Journal of the Nigerian Mathematical
Society .13, 37 – 51.
Henrici, P. (1962). Discrete Variable Method in Ordinary Differential Equations. John Willey. New York
37