Journal of Natural Sciences and Mathematics

Qassim University, Vol. 5, No. 2, PP 89-102 (January 2012/Safar 1433H.)

89

Nearly Pentadiagonal Linear Systems

A. A. Karawia * and S. A. El-Shehawy**
* Qassim University, Deanship of Educational Services, Computer Science Unit
Buraidah 51452, Kingdom of Saudi Arabia
Home Address: Mathematics Department, faculty of Science, Mansoura University, Mansoura 35516,
Egypt.
E-mail : [email protected]
** Qassim University, College of Science, Department of Mathematics
P.O. Box 6644 Buraidah 51452, Kingdom of Saudi Arabia
Home Address: Mathematics Department, faculty of Science, Menoufia University, Shebin El-Kom
32511, Egypt.
E-mail : [email protected]

(Received 22/11/2011, Accepted 17/03/2012)

Abstract. In this paper we present efficient computational and symbolic algorithms for solving nearly
pentadiagonal linear systems. The implementation of the algorithms using Computer Algebra Systems
(CAS) such as MAPLE, MATHEMATICA, and MATLAB is straightforward. Two examples are given in
order to illustrate the algorithms.

Keywords: Pentadiagonal matrices; Nearly pentadiagonal matrices; Linear systems; Determinants;
Computer algebra systems (CAS).

AMS Subject Classification: 15A15; 15A23; 68W30; 11Y05; 33F10; F.2.1; G.1.0.

A. A. Karawia and S. A. El-Shehawy

90
1. Introduction
Many authors studied pentadisgonal matrices, pentadiagonal linear systems and
some linear systems related to it such periodic and cyclic pentadiagonal linear
systems. In [10], authors developed an efficient algorithm to find the inverse of a
general pentadiagonal matrix using the general Doolittle factorization. Karawia and
Sogabe presented new algorithms for solving periodic pentadiagonal linear systems
[7, 8]. Authors in [5] described a reliable symbolic computational algorithm for
inverting cyclic petadiagonal matrices using parallel computing along with
recursion. Many problems in mathematics and applied science require the solution
of linear systems having nearly pentadiagonal coefficient matrices. This kind of
linear system arises in many fields of numerical computation and differential
Equations [2, 6, 9,].
The main goal of the current paper is to develop an efficient algorithm for
solving a general nearly pentadiagonal linear systems of the form:

A X Y = (1)
where
1 1 1
2 2 2 2
3 3 3 3 3
4 4 4 4 4
2 2 2 2 2
1 1 1 1
0 0 0 0
0 0 0 0
0 0
0 0
0
0
0
0 0
n n n n n
n n n n
n n n
d a a s
b d a a
b b d a a
b b d a a
A
b b d a a
b b d a
t b b d









=








% L
% L
%
% L M
%
% L M
M O O O O O O O M
M O O O O O O O
%
% M L L
%
M L L L
%
L L L
, (2)

1 2 1 2
( , , , ) , ( , , , )
T T
n n
X x x x Y y y y = = K K and 5 n .

A general n n nearly pentadiagonal matrix A of the form (2) can be stored in
5 4 n memory locations by using the vectors
1 2 2
( , , , )
n
a a a

= % % % % K a ,
1 2 1
( , , , , )
n
a a a s

= K a ,
2 3
( , , , , )
n
t b b b = K b ,
3 4
( , , , )
n
b b b =
% % % %
K b and
1 2
( , , , )
n
d d d = K d . When considering the system (1) it is advantageous to
introduce three additional vectors
1 2
( , , , )
n
c c c = K c ,
1 2
( , , , )
n
e e e = K e and
Nearly Pentadiagonal Linear Systems

91
1 2
( , , , )
n
f f f = K f . These vectors are related to the vectors % a , a , d , b and
%
b .
The current paper is organized as follows. In Section 2, the main results are
given. Illustrative examples are presented in Section 3. In Section 4, a conclusion is
given.

2. Main Results
In this section we are going to formulate a new computational and symbolic
algorithms for solving general nearly pentadiagonal linear systems of the form (1).
To do this, we begin by considering the LU decomposition [1] of the matrix A in
the form:
3
1
4
2
2
4
1
3
3
1 1 1
2
2 2
3
3 3 3
4
4
2
1
1
1 0 0 0 0 0 0 0
0 0 0 0
1 0 0 0 0 0 0
0 0 0 0 0
1 0 0 0 0
0 0 0 0
0 1 0 0 0
0 0 0
0
0 1 0 0
0 1 0
0 0 1

=









%
%
%
%
L
% L
L
L
L M
% L M
L M
M O O O O O O O M
M O O O O O O O
M L L
M L L L
L L L
n
n
n
n
n
n
b
c
b
c
b
n c
b
n c
t
n c
c e a s
f
c e e
f
c e a
f
c e
A
f
f
f f
4 4
2 2 2
1 1
0
0
0 0 0
0 0 0
0 0 0 0 0
















% L M
M O O O O O O O M
M O O O O O O O
% M L L
M L L L
L L L
n n n
n n
n
a
c e a
c e
c
(3)

From (3) we obtain

2
1
2 2 1
1 2 2
1 2 1
if 1
if 2
, if 3(1) 1
if ,

=

= =

%
% %
%
i
i
i b
i i i i n c
n n n n
d i
d f e i
c
d f e a a e i n
d f a f e i n
(4)
where

A. A. Karawia and S. A. El-Shehawy

92

3
1
1
2 2 1
2 2
3 3
1
if 1
if 2
if
if 3
if 4(1) 1,
i n
b
n n c
i i i
a i
a f a i
e a f a i n
a a f e i
a f a i n

= =

%
%
%
%
(5)
and


3
2
1
2
3
3 2
1
3 1 2 1
( ) / if 1
if 2
( ) / if 3(1) 1
( ) / if .
n
i
i
n
t
n n n c
b
c
i
b
i i c
t
n n n n c
b e c i
i
f
b c i n
b a f e c i n

=

=

%
%
%
(6)

It is not difficult to prove that the LU decomposition (3) exists only if
0
i
c , 1(1) 1 i n = . Moreover, a general nearly pentadiagonal linear system (1)
possesses a unique solution if, in addition, 0
n
c . On the other hand, the
determinant of the matrix A is given by:
1
det( ) ,
n
i
i
A c
=
=

(7)

and this shows the importance of the vector c [4].
We may now formulate the following results.

2.1 Algorithm To solve the general nearly pentadiagonal linear system (1), we may
proceed as follows:

step 1: Set
1 1
c d = .

step 2: If
1
0 c = , then OUTPUT(the method is fails); STOP.

step 3: Set
2
1
1 1 1 2
, , , ,
b
n c
a s b t e a f = = = = and
2 2 2 1
c d f e = .

Nearly Pentadiagonal Linear Systems

93
step 4: If
2
0 c = , then OUTPUT(the method is fails); STOP.

step 5: Set
2 2 2 1 2 2
,
n n
e a f a e a f a = = % % , and
2 n
a e = % .

step 6: For 3, 4, , 1 i n = K Compute

2
2 1
( ) /
i
i
b
i i i i c
f b e c

=
%
,

2
1 2

=
%
%
i
i
b
i i i i i c
c d f e a ,
If 0
i
c = , then OUTPUT(the method is fails); STOP,
If 3 i = , then
2
1
i
i
b
i i i i n c
e a f a a

=
%
% else
1 i i i i
e a f a

= % .

step 7: Compute

1
3
1 3 2
( ) /
n
b
n n n c
f b e c

=
%
,
3
3 1 2 1
( ) /
n
t
n n n n n c
f b a f e c

= % ,
1 2 1 n n n n n
c d f a f e

= % ,

step 8: Set
1 1 2 2 2 1
, z y z y f z = = .

step 9: For 3, 4, , 1 i n = K Compute

2
1 2
i
i
b
i i i i i c
z y f z z

=
%
.

step 10: Set
3
1 1 2 3
n
t
n n n n n n c
z y f z f z z

= .

step 11: Compute the solution vector x using
1 1
1
1
,
n n n n
n n
z z e x
n n c c
x x

= = ,
For 2, 3, , 2 i n n = K Compute

1 2 i i i i i
i
z e x a x
i c
x
+ +

=
%
,
A. A. Karawia and S. A. El-Shehawy

94
Set
1 1 2 1 3 4
1
1
z e x a x s x
c
x

=
%
.

The new algorithm 2.1 will be referred to as KNPENTA algorithm.
KNPENTA algorithm for solving the nearly pentadiagonal system (1) is generally
preferable because the conditions 0, 1(1)
i
c i n = are sufficient for its validity.
The advantage of the vector c is now clear.

The following symbolic algorithm is developed in order to remove the
cases where the numeric algorithm KNPENTA fails.
2.2 Algorithm To solve the general nearly pentadiagonal linear system (1), we may
proceed as follows:
step 1: Set
1 1
c d = .

step 2: Set
1
c x = and
1
d x = ( x is just a symbolic name) whenever
1
0 c = .

step 3: Set
n
a s = ,
1
b t = ,
1 1
e a = ,
2
1
2
b
c
f = , and
2 2 2 1
c d f e = .

step 4: Set
2
c x = whenever
2
0 c = .

step 5: Set
2 2 2 1 2 2
,
n n
e a f a e a f a = = % % , and
2 n
a e = % .

step 6: For 3, 4, , 1 i n = K Compute

2
2 1
( ) /
i
i
b
i i i i c
f b e c

=
%
,

1 2 2
i
i
b
i i i i i c
c d f e a

=
%
% ,
Set
i
c x = whenever 0
i
c = ,
If 3 i = , then
2
1
i
i
b
i i i i n c
e a f a a

=
%
% else
1 i i i i
e a f a

= % .

step 7: Compute

1
3
1 3 2
( ) /
n
b
n n n c
f b e c

=
%
,
3
3 1 2 1
( ) /
n
t
n n n n n c
f b a f e c

= % ,
Nearly Pentadiagonal Linear Systems

95
1 2 1 n n n n n
c d f a f e

= % ,
Set
n
c x = whenever 0
n
c = .

step 8: Set
1 1 2 2 2 1
, z y z y f z = = .

step 9: For 3, 4, , 1 i n = K Compute

2
1 2
i
i
b
i i i i i c
z y f z z

=
%
.

step 10: Set
3
1 1 2 3
n
t
n n n n n n c
z y f z f z z

= .

step 11: Compute the solution vector x using
1 1
1
1
,
n n n n
n n
z z e x
n n c c
x x

= = ,
For 2, 3, , 2 i n n = K Compute

1 2 i i i i i
i
z e x a x
i c
x
+ +

=
%
,
Set
1 1 2 1 3 4
1
1
z e x a x s x
c
x

=
%
.

step 12: Substitute 0 x = in all expressions of the solution vector. n i x
i
,..., 2 , 1 , =

The symbolic algorithm 2.2 will be referred to as KSNPENTA algorithm.

In [3], Claerbout showed that the two-dimensional Laplacian operator, which
appears in 3-D finite-difference migration, has the form of pentadiagonal matrix. If
we choose 4, 1(1)
i
d i n = = , 0 s t = = and 1,
i i i i
a b a b i = = = =
%
% , we
can obtain it.

3. Illustrative Examples
In this section we are going to give illustrative examples.
Example 3.1 Solve the nearly pentadiagonal linear system of size 10 given by

A. A. Karawia and S. A. El-Shehawy

96
1
2
3
4
5
6
7
8
9
10
3 1 3 5 0 0 0 0 0 0
2 2 1 2 0 0 0 0 0 0
3 4 5 5 1 0 0 0 0 0
0 3 2 1 1 3 0 0 0 0
0 0 6 1 2 5 1 0 0 0
0 0 0 3 3 2 7 5 0 0
0 0 0 0 8 1 12 3 4 0
0 0 0 0 0 2 5 3 1 20
0 0 0 0 0 0 3 11 21 3
0 0 0 0 0 0 2 4 9 31
x
x
x
x
x
x
x
x
x
x

30
13
35
27
69
18
38
280
328
247







=








(8)

by using the KNPENTA algorithm and KSNPENTA algorithm.

Solution

(a) The implementation of the KNPENTA algorithm gives:

1
3 c = (Step 1).


2
1 1 2 3
5, 2, 1,
n
a b e f

= = = = and
4
2 3
c = (Step 3).


16
2 3
3,
n
e e = = and
16
2 3
a = % (Step 5).


9 253 105 368
4012 204567
3 4 5 6 7 8 9 4 35 184 2271 101 186433
[ , , , , , , ] [ , 1, , , , , ] f f f f f f f

= ,

1313
35 552 757 14341 4112262
3 4 5 6 7 8 9 4 35 92 1514 303 186433
[ , , , , , , ] [ , 1, , , , , ]

= c c c c c c c and

1393 1371 3532041
934 26873
3 4 5 6 7 8 9 35 184 2271 101 186433
[ , , , , , , ] [12, 2, , , , , ] e e e e e e e

= (Step 6).


91736 1203361
1 186433 4112262
, ,
n
f f

= = and
701215
19866 n
c = (Step 7).


1 2
[ , ] [30, 33] z z = (Step 8).


317 8609 11475 238883 138311 129996
3 4 5 6 7 8 9 4 35 184 4542 303 14341
32 [ , , , , , , ] [ , , , , , , ] = z z z z z z z (Step 9).

Nearly Pentadiagonal Linear Systems

97

3506075
10 9933
z = (Step 10).


1 2 3 4 5 6 7 8 9 10
[ , , , , , , , , , ] [1, 2, 3, 4, 5, 6, 7, 8, 9,10] = x x x x x x x x x x (Step 11).

Also the determinant of A is det( ) 145151505 A = by using (7).

(b) The implementation of the KSNPENTA algorithm gives:


_ : ( , , , , , ) [1, 2, 3, 4, 5, 6, 7, 8, 9,10]
T
nearly penta X b b d a a y = =
%
% .
Example 3.2 Solve the nearly pentadiagonal linear system of size given by

1
2
3
4
5
6
7
0 3 1 5 0 0 0 13
9 1 4 2 0 0 0 9
7 6 2 3 7 0 0 41
0 5 2 3 1 11 0 52
0 0 2 4 1 6 1 54
0 0 0 3 4 5 1 27
0 0 0 4 13 6 1 44
w
w
w
w
w
w
w





=

(9)


by using the KNPENTA algorithm and KSNPENTA algorithm.

Solution

(a) The implementation of the KNPENTA algorithm gives:
The method is break down since
1 1
0 c d = = .

(b) The implementation of the KSNPENTA algorithm gives:


A. A. Karawia and S. A. El-Shehawy

98

-73002
3998p+36501
1018p-36501
3998p+36501
5(646p-36501)
3998p+36501
3(8320p+36501)
: ( , , , , , )
3998p+36501
2(6385p+36501)
3998p+36501
12(2133p+12167)
3998p+36501
2(21917p-109503)
3998p+36501
nearly_penta W b b d a a y

= =

%
%
0
2
1
5
3 .
2
4
6
p =














=



















Also the determinant of coefficients matrix equals -73002 and for more details about
how to call this procedure, see Appendix 1.

4. Conclusion
The computational algorithm KNPENTA described here is very effective,
provided that optimal LU factorization is used, because the conditions
0, 1(1)
i
c i n = are sufficient for its validity. The symbolic algorithm
KSNPENTA is developed in order to remove the cases where the computational
algorithm KNPENTA fails. The symbolic algorithm KSNPENTA is competitive
with the computational algorithm KNPENTA for solving a nearly pentadiagonal
linear system which appears in many applications.

Appendix 1. A Maple procedure for solving a nearly pentadiagonal linear system


> # A Maple Procedure.
> # To compute the solution of a general nearly pentadiagonal Linear system.
> restart:
> with(linalg,vector,vectdim):
> nearly-penta:=proc(bb::vector,b::vector, d::vector,a::vector,aa::vector,y::vector)
local i,j,k,n; global T,e,c,f,z,X;
n:=vectdim(d):e:=array(1..n):c:=array(1..n):
f:=array(1..n):z:=array(1..n):X:=array(1..n):
Nearly Pentadiagonal Linear Systems

99
#components of the vectors e, c, and f #
c[1]:=d[1]:if c[1]=0 then c[1]:=x; d[1]:=x;fi: e[1]:=a[1]:f[2]:=simplify(b[2]/c[1]):
c[2]:=simplify(d[2]-e[1]*f[2]):if c[2]=0 then c[2]:=x;fi:
e[2]:=simplify(a[2]-f[2]*aa[1]):e[n]:=simplify(aa[2]-f[2]*a[n]): aa[2]:=e[n]:
for i from 3 to n-1 do
f[i]:=simplify((b[i]-bb[i]*e[i-2]/c[i-2])/c[i-1]):
if i=3 then e[i]:=simplify(a[i]-f[i]*aa[i-1]-bb[i]*a[n]/c[i-2]); else
e[i]:=simplify(a[i]-f[i]*aa[i-1]);
fi:
c[i]:=simplify(d[i]-bb[i]*aa[i-2]/c[i-2]-e[i-1]*f[i]); if c[i]=0 then c[i]:=x; fi:
end do:
f[1]:=simplify((bb[n]-b[1]*e[n-3]/c[n-3])/c[n-2]):
f[n]:=simplify((b[n]-b[1]*aa[n-3]/c[n-3]-f[1]*e[n-2])/c[n-1]):
c[n]:=simplify(d[n]-f[1]*aa[n-2]-e[n-1]*f[n]):if c[n]=0 then c[n]:=x; fi:
# To compute the vector Z #
z[1]:=y[1]:z[2]:=y[2]-f[2]*z[1]:i:=i:
for i from 3 to n-1 do
z[i]:=simplify(y[i]-bb[i]*z[i-2]/c[i-2]-f[i]*z[i-1]):
end do:
z[n]:=simplify(y[n]-f[n]*z[n-1]-f[1]*z[n-2]-b[1]*z[n-3]/c[n-3]):
# To compute the Solution of the system X. #
X[n]:=z[n]/c[n]:i:=i:
X[n-1]:=simplify((z[n-1]-e[n-1]*X[n])/c[n-1]):
for i from n-2 by -1 to 2 do
X[i]:=simplify((z[i]-e[i]*X[i+1]-aa[i]*X[i+2])/c[i]):
end do:
X[1]:=simplify((z[1]-e[1]*X[2]-aa[1]*X[3]-a[n]*X[4])/c[i]):
# To compute the determinant T #
T:=subs(x=0,simplify(product(c[r],r=1..n)));
eval(X):
end:
> # Call no. 1 for the procedure nearly penta. #
> x:=x:
> aa:=aa:=vector([3,2,1,3,1,-5,-4,20]);
aa := [3, 2, 1, 3, 1, -5, -4, 20]
>a:=vector([-1,1,5,1,5,7,3,1,3,5]);
a:=[-1,1,5,1,5,7,3,1,3,5]
> d:=vector([3,2,5,1,2,2,12,3,21,31]);
d:=[3,2,5,1,2,2,12,3,21,31]
> b:=vector([-2,-2,-4,-2,1,-3,1,5,11,-9]);
b:=[-2,-2,-4,-2,1,-3,1,5,11,-9]
> bb:=vector([0,0,3,3,6,3,-8,2,3,4]);
bb:=[0,0,3,3,6,3,-8,2,3,4]
> y:=vector([30,13,35,27,69,18,38,280,328,247]);
A. A. Karawia and S. A. El-Shehawy

100
y:=[30,13,35,27,69,18,38,280,328,247]
> X:=nearly penta(bb,b,d,a,aa,y);
X := [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
T
> T;
-145151505
> # End of call no. 1. #

> # Call no. 2 for the procedure nearly penta. #
> a:=vector([3,4,3,1,6,1,5]);
a:=[3,4,3,1,6,1,5]
> d:=vector([0,1,2,-3,1,5,1]);
d:=[0,1,2,-3,1,5,1]
> b:=vector([-4,9,-6,2,4,-4,6]);
b:=[-4,9,-6,2,4,-4,6]
> aa:=vector([-1,2,7,11,1,0,0]);
aa:=[-1,2,7,11,1,0,0]
> bb:=vector([0,0,-7,5,2,3,13]);
bb:=[0,0,-7,5,2,3,13]
> y:=vector([13,9,41,52,54,27,44]);
y:=[13,9,41,52,54,27,44]
> W:=nearly_penta(d,a,a_,b,b_,y);


> T;
-73002
> p:=0:X:=map(eval,op(W));
X := [-2, 1, 5, 3, 2, 4, 6]
T

> # End of call no. 2. #

5. Acknowledgement
This work has been supported by the Deanship of Scientific Research at Qassim
University. The authors would like to thank the committee of the journal and the
referees for their constructive comments.

6. References
[1] Allen, M. B. and Isaacson,E. L. Numerical Analysis for Applied Science, Wiley-
Interscience, John Wiley & Sons, 1997.
Nearly Pentadiagonal Linear Systems

101
[2] Cheney, W. and Kincaid, D. Numerical Mathematics and Computing (Second
Edition), Wadsworth, Inc., 1985.
[3] Claerbout, J, F. Imaging the earths interior, Blackwell Scientific Publications,
1985.
[4] El-Mikkawy, M. E. A. "A Fast Algorithm for Evaluating nth Order tridiagonal
Determinants". J. Comput. Appl. Math., Vol. 166(2004), pp. 581-584.
[5] El-Mikkawy, M. E. A. and Rahmo, D. Symbolic algorithm for inverting cyclic
pentadiagonal matrices recursively-Derivation and implementation.
Computers & Mathematics with Applications, Vol. 59, Issue 4, (2010), pp.
1386-1396.
[6] LeVeque, R. J., Finite Difference Methods for Differential Equations, University
of Washington, 2006.
http://www.imsc.res.in/_rjoy/WWW/research/numerics/assets/am585winter06.pdf
[7] Karawia, A. A. A computational algorithm for solving periodic penta-diagonal
linear systems. Applied Mathematics and Computation, Vol. 174, Issue 1,
(2006), pp. 613-618.
[8] Sogabe, T. New algorithms for solving periodic tridiagonal and periodic
pentadiagonal linear systems. Applied Mathematics and Computation, Vol.
202, Issue 2, (2008), pp. 850-856.
[9] Stone, H. L. "Iterative Solution of Implicit Approximations of Multidimensional
Partial Differential equations". SIAM J. Numer. Anal., Vol. 5, No. 3(1968),
pp. 530-558.
[10] Zhao, X. and Huang, T. On the inverse of a general pentadiagonal matrix.
Applied Mathematics and Computation, Vol. 202, Issue 2, (2008), pp. 639-
646.

A. A. Karawia and S. A. El-Shehawy

102


. * . **
* ,.. ,.. :.- - .,.,.... .... :... - ,,... ....,
** ..,,. ,.. .., - ,... .,.s - ,,... ....,

: .... . . . .... _, . . (

. .-,. ..- _ .. ..,., .-.. ,,... _. _; _. _.| . ..;
,. ... ,... .,_.. . .... ..,., ..| .-.., .; ,.. _ _.. _,..
,... .<...,.. .... ; .... _,.,.. ..,., .