Applied Mathematics and Computation 168 (2005) 431446

www.elsevier.com/locate/amc

On numerical improvement of the second

kind of GaussChebyshev quadrature rules
Mehdi Dehghan
a

a,*

, M. Masjed-Jamei

a,b,c

, M.R. Eslahchi

Department of Applied Mathematics, Faculty of Mathematics and Computer Science,

Amirkabir University of Technology, No. 424, Hafez Ave. Tehran, Iran
b
Department of Mathematics, K.N. Toosi University of Technology, Tehran, Iran
c
Center for Research and Studies, Sanjesh Organization, Ministry of Science,
Research and Technology, Tehran, Iran

Abstract
One of the integration methods is the Second Kind of GaussChebyshev quadrature
rule, denoted by:

 


Z 1
n
p
p X
kp
kp
f cos
f x 1
x2 dx
sin2
n 1 k1
n1
n1

1
p
f 2n g;
1 < g < 1:
2n1
2
2n!
According to Gauss quadrature rules, the precision degree of above formula is the highest, i.e. 2n
1. Hence, it is not possible to increase the precision degree of Second Kind
of GaussChebyshev integration formulas anymore. But, on the other hand, we claim
that we can improve the above formula numerically. To do this, we consider the integral
bounds as two unknown variables. This causes to numerically be extended the monomial space f(x) = xj from j = 0, 1, . . . , 2n
1 to j = 0, 1, . . . , 2n + 1. This means that we

Corresponding author.
E-mail addresses: mdehghan@aut.ac.ir (M. Dehghan), mmjamei@aut.ac.ir (M. Masjed-Jamei),
eslahchi@aut.ac.ir (M.R. Eslahchi).
0096-3003/$ - see front matter
2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.amc.2004.09.047

432

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

have two monomials more than Second Kind GaussChebyshev integration method. In
other words, we give an approximate formula as:
Z b
n
X
p
f x 1
x2 dx
wi f xi ;
a

i1

in which a,b and w1,w2, . . . , wn and x1,x2, . . . , xn are all unknowns and the formula is almost exact for the monomial basis f(x) = xj,j = 0, 1, . . . , 2n + 1. Some important examples are nally given to show the excellent superiority of the proposed nodes and
coecients with respect to the Second Kind GaussChebyshev nodes and coecients.
Let us add that in this part we give also some wonderful 2-point, 3-point and 4-point
formulas that are respectively comparable with 103-point, 261-point and 108-point formulas of Second Kind GaussChebyshev quadrature rules in average.

2004 Elsevier Inc. All rights reserved.
Keywords: Second Kind GaussChebyshev formula; Numerical integration methods; Degree of
accuracy; The method of undetermined coecient; The method of solving nonlinear systems

1. Introduction
It is known that the general form of Gauss quadrature rules is indicated as:
Z b
n
m
X
X
f x dwx
wj f xj
vk f zk Rn;m f ;
1
a

j1

k1

where the weights wj nj1 ; vk mk1 and nodes xj nj1 are unknowns and the nodes
zk mk1 are predetermined, w is also a positive measure on [a, b] (see [1012]).
The residue Rn,m[f] is determined (see for instance [13]) by:
f 2nm g
Rn;m f 
2n m!

"
#2
m
n
Y
Y
x
zk
x
xj dwx;
k1

a < g < b:

j1

2
p
By selecting dwx 1
x2 dx; a
1; b 1 and m = 0 we reach the Second
Kind GaussChebyshev formula. In other words we have:
Z 1
n
X
p
f x 1
x2 dx
wj f xj Rn f ;
3

1

j1

where,
f 2n g
Rn f
2n!

1

1

"

#2
m
Y
p
x
xi
1
x2 dx:
i1

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

433

Calculating wj nj1 ; xk nk1 and Rn f  in (3) yields (see [1]):


 


Z 1
n
p
p X
kp
kp
2
2
f x 1
x dx
sin
f cos
n 1 k1
n1
n1

1

p
f 2n g;
2n!

2n1

1 < g < 1:

As regards, the integration formula (5) has precision degree 2n
1. Note that
2n
1 is the highest precision in all integration formulas. This subject can be
proved by the following theorem:
Theorem 1. In the general n-point integration formula:
Z b
n
X
wxf x dx
wj f xj Rn f ;
a

j1

the highest precision degree is 2n
1, which w(x) is the weight function and Rn[f]
is the error of integration formula.
Proof. see [9].
Thus, for second kind of Chebyshev integration formula we can write:
Z 1
n
X
p
f x 1
x2 dx
wj f xj ; f x xi ; i 0; 1; . . . ; 2n
1:
7

1

j1

By applying above relation one can reach a nonlinear system with 2n equations
n
n
and 2n unknowns, so that the unknowns are xj j1 and wj j1 respectively.
Hence, solving this nonlinear system results the integration formula (5). Now
if we suppose that the bounds of integration formula (7) i.e. [
1,1], to be considered as two additional variables, say a and b then by extending the basis
space {1, x, x2, . . . , x2n
1} to {1, x, x2, . . . , x2n+1} we have:
Z a
n
X
p
f x 1
x2 dx
wj f xj ; f x xi ; i 0; 1; . . . ; 2n 1: 8
b

j1

Certainly, the nonlinear system (8) does not have any analytical solution, because as we said in the Theorem 1, the highest precision degree in general integration formulas is 2n
1.
On the other hand, we add that this nonlinear system can be solved
numerically, which causes to increase the monomial space two degrees. This
would be our main work in this article. In the next sections, some examples are
given to show the numerical superiority of the presented rules with respect to
the numerical results of Second Kind GaussChebyshev quadrature rules. h

434

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

2. Some conditions under which the algorithm of solving the nonlinear system (8)
is feasible
First, it should be noted that
the relation
is an ill-condition nonlinear

p(8)
system, because the statements 1
b2 and 1
a2 or arcsin(a) and arcsin(b)
exist in its left hand side always. For example, the left hand side (8) for
f(x) = x,x2 are respectively:
q q
3
3
Z b p
1
b2
1
a2

;
x 1
x2 dx

3
3
a
q
q
Z b p
2 3
b 1
b2 1
b
1


b

arcsinb

x2 1
x2 dx

8
8
4
a
q
p
3
a 1
a2
a 1
a2 1



arcsina:

8
4
8
Hence the algorithm of solving the nonlinear system (8) must have two
conditions:

1 < a < 1;

1 < b < 1;

and

in all iterations, otherwise the algorithm is stopped.

Furthermore, the algorithm should be designed such that the conditions:
i a < xi < b;

10

ii wi > 0;

11

hold in nal solution.

3. Determining [xj ]j1 ; [wj ]j1 , a and b in (8) numerically

n

As was said, the xj j1 ; wj j1 ; a and b in (8) are computable by Maple8 software numerically, but it should also be mentioned that to solve (8) we must apply an algorithm that satises the conditions of Section 2. By noting to this
matter, the values xj nj1 ; wj nj1 , a and b will be derived as follows:
2-point formula:
Z 0:002598931
p
f x 1
x2 dx 0:0002703255f 0:0001456453

0:0002808400

0:0002704076f
0:0001665397;

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

3-point formula:
Z 0:0008405871

435

p
f x 1
x2 dx 0:0004546046f 0:0006557356

0:0007960979

 0:0007276189f 0:0000221674
0:0004544615f
0:0006114110;

4-point formula:
Z 0:0001528194

p
f x 1
x2 dx 0:0000530105f 0:0001314802

0:0001587731

0:0001025828f 0:0000508364
0:0001026976f
0:0000564133
0:0000533017f
0:0001372376;

Remark. Note that the above 4-point formula is in fact the numerical solution
of the following nonlinear system with 10 variables and 10 unknowns.
p
p
b 1
b2 1
a 1
a2 1
w1 w2 w3 w4

arcsina
arcsinb

2
2
2
2
1
b2 3=2 1
a2 3=2

3
3
p
b1
b2 3=2 b 1
b2 1
2
2
2
2
w1 x1 w2 x2 w3 x3 w4 x4

arcsinb
4 3=2
8 8
p
2
2
a1
a
a 1
a
1

arcsina
8
8
8
w1 x1 w2 x2 w3 x3 w4 x4

3=2

3=2

b2 1
b2
21
b2


5
15
3=2
3=2
a2 1
a2
21
a2

5
15

w1 x31 w2 x32 w3 x33 w4 x34

w1 x41

w2 x42

w3 x43

w4 x44

p
3=2
3=2
b3 1
b2
b1
b2
b 1
b2

6
8
16
3=2
1
a3 1
a2
arcsinb

16
6
p
3=2
a1
a2
a 1
a2
1
arcsina

16
8
16

436

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

3=2

3=2

3=2

3=2

b4 1
b2
4b2 1
b2


7
35
3=2
3=2
81
b2
a4 1
a2

105
7
3=2
2
2 3=2
4a 1
a
81
a2

35
105

w1 x51 w2 x52 w3 x53 w4 x54

b5 1
b2
5b3 1
b2


8
48
p
3=2
5b1
b2
5b 1
b2

64
128
3=2
5 arcsinb a5 1
a2

128
8
3=2
3=2
5a3 1
a2
5a1
a2

48
64
p
5a 1
a2 5 arcsina




128
128

w1 x61 w2 x62 w3 x63 w4 x64

b6 1
b2 3=2 2b4 1
b2 3=2

9
21
3=2
2
2 3=2
8b 1
b
161
a2




105
315
3=2
3=2
a6 1
a2
2a4 1
a2

9
21
8a2 1
a2 3=2 161
a2 3=2

105
315

w1 x71 w2 x72 w3 x73 w4 x74

b7 1
b2 3=2 7b5 1
b2 3=2

10
80
p
3
2
2 3=2
7b 1
b
7b1
b




128
96
p
7b 1
b2
7
arcsinb

256
256
a7 1
a2 3=2 7a5 1
a2 3=2

10
80
3=2
3
2
7a 1
a
7a1
a2 3=2

96
128
p
7a 1
a2
7
arcsina




256
256

w1 x81 w2 x82 w3 x83 w4 x84

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

3=2

437
3=2

b8 1
b2
8b6 1
b2


11
99
3=2
3=2
16b4 1
b2
64b2 1
b2




231
1155
1281
b2 3=2 a8 1
a2 3=2

3465
11
3=2
3=2
8a6 1
a2
16a4 1
a2

99
231
3=2
3=2
64a2 1
a2
1281
a2

1155
3465

w1 x91 w2 x92 w3 x93 w4 x94

5-point formula:
Z 0:0003109491

p
f x 1
x2 dx 0:0000427149f 0:0002970513

0:0003123990

0:0001652997f 0:0001914224
0:0002062081f
0:0000001641
0:0001655011f
0:0001918255
0:0000436244f
0:0002979199;
6-point formula:
Z 0:0002978505

p
f x 1
x2 dx 0:0000394402f 0:0002830130

0:0003492817

0:0001372437f 0:0001948552
0:0001831945f 0:0000310572
0:0001754422f
0:0001519539
0:0000007511f
0:0004249527
0:0001110605f
0:0002968243;
7-point formula:
Z 0:0005249564

p
f x 1
x2 dx 0:0000360803f 0:0004960943

0:0005249939

0:0001574751f 0:0004054700
0:0002155713f 0:0002244794
0:0002316639f
0:0000000012
0:0002155914f
0:0002244923
0:0001575097f
0:0004055060
0:0000360586f
0:0004961499;

438

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

8-point formula:
Z 0:0007656375

p
f x 1
x2 dx 0:0000653250f 0:0006749968

0:0006893659

0:0001648725f 0:0005857087
0:0002435306f 0:0004069214
0:0002917386f 0:0001478129
0:0002915273f
0:0001474111
0:0002420930f
0:0004059499
0:0001528959f
0:0005810051
0:0000030205f
0:0006893108:
Now let us dene the error:

Z

 b p X
n


xj 1
x2

wi xji  j 0; 1; . . . ; 2n 1;
ejn 

 a
i1

n 2; 3; . . . ; 8:
12

such that:
En Max ej
n ;

0 6 j 6 2n 1;

n 2; 3; . . . ; 8:

13

For example if n = 2, then we have:


 p
p

 b 1
b2 1
a 1
a2 1


0

arcsina
w1
w2 
e1 
arcsinb



2
2
2
2
1:3  10
11 ;
1
e1

2
e1

3
e1




 1
b2 3=2 1
a2 3=2



w1 x1
w2 x2  1:0  10
10 ;




3
3


3=2
 b1
b2 3=2 bp
1
b2 1
a1
a2


arcsinb



8
4
8
8

p

a 1
a2 1




arcsina
w1 x21
w2 x22  0;

8
8

 b2 1
b2 3=2 21
b2 3=2 a2 1
a2 3=2 21
a2 3=2





5
15
5
15




w1 x31
w2 x32  1:0  10
10 ;


M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

439

Table 1
Errors of improvement GaussChebychev quadrature rule
n

En

2
3
4
5
6
7
8

1.0 10
10
3.7 10
10
1.0 10
10
1.2 10
10
3.0 10
10
9.6 10
11
4.8 10
9

4
e1

5
e1

 b3 1
b2 3=2 b1
b2 3=2 bp
1
b2 1


arcsinb




16
16
6
8
p
a3 1
a2 3=2 a1
a2 3=2 a 1
a2
1

arcsina

16
16
6 
8



w1 x41
w2 x42  0;


 b4 1
b2 3=2 4b2 1
b2 3=2 81
b2 3=2 a4 1
a2 3=2





7
35
105
7


4a2 1
a2 3=2 81
a2 3=2


w1 x51
w2 x52  1:0  10
11 ;


35
105

Hence

n
o
0 1
5
E2 max e1 ; e1 ; . . . ; e1 1:0  10
10 :

14

For n = 2, 3, . . . , 8 the values En are given in Table 1.

4. Determining conditions for optimum use of obtained data

Since the nonlinear system (8) has been designed based on the monomial
space xj let us assume that f(x) has a convergent Taylor series:
1
X
f j 0
;
15
aj x j ; aj
f x
j!
j0
then we can write:
f x

2n1
X
j0

aj x j

f 2n2 w 2n2
x
;
2n 2!

1 < w < 1:

16

440

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

Now it is not dicult to verify, using xi and wi from Section 3 and relation (14)
that:
Z
!
!
n1
2X
n1
n
 b 2X

X
p

j 
j
2
aj x
aj
wi xi 
1
x dx



 a
j0
j0
i1


!
Z b p
 2X
2X
n1
n
n1
X


aj
xj 1
x2 dx

wi xji  6
jaj jej

n

 j0
a
i1
j0
6 En

2X
n1

jaj j:

17

j0

P2n1
Therefore if we take Q2n1 x j0 aj xj , in fact we have proved the following
inequality

Z

 b
n
2X
n1
X
p


Q2n1 x 1
x2 dx

wi q2n1 xi  6 En
jaj j;
18


 a
i1
j0
where xi,wi and En were obtained in Section 3.
P2n1
The inequality (17) or (18) will be very useful when j0 jaj j 6 1. In this
way we can see that the approximation
Z b
n
X
p
Q2n1 x 1
x2 dx
wi Q2n1 xi ;
19
a

i1

acts much better than for only monomial functions xj,j = 0, 1, 2, . . . , 2n + 1.

But as we said, the complete Second Kind GaussChebyshev quadrature
formula is in the following form:
Z 1
n
X
p
p
f 2n g; g 2
1; 1:
f x 1
x2 dx
wi f xi 2n1
2n!
2

1
i1
Table 2
Table of examples
Example

Function

Interval [a, b]

Taken from reference

x
f1 x p
x2
4

[3,3.5]

[5]

0; p4

3x

2
3

f2(x) = e sin(2x)
f3 x x22
4

[0,0.35]

[5]
[5]

4
5
6
7
8

f4(x) = cos2(x)
f5(x) = x2ln(x)
f6(x) = ecos(x)
2
f7 x e
x
1
f8 x 1e
x

0; p4
[1,1.5]
0; p4
[0,1]
[0,1]

[5]
[5]
[3]
[2]
[7]

f9 x 1 x2 2

[0,1]

[6]

[0,1]

[6]

10

f10 x

sin2 x
x

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

441

The
second term of the right hand side of this formula can be corresponded
R b f 2n2
P2n1 to
w 2n2
x
dx
and
the
additional
error
in
(17),
which
is
at
most
E
n
j0 jaj j.
a 2n2!
This is the cost of extending monomial xj, from j = 0, . . . , 2n
1, to
j = 0, 1,. . .,2n + 1. So if f(x) has a rapidly convergent Taylor series, our new
quadrature rules work better than corresponding Second Kind GaussChebyshev rules, as the examples (see Table 2) of the next section.

5. Numerical results
First we remind that transferring the integral interval [c, d] to [a, b], (a and b
are the obtained values from Section 3) is possible using the following:
Z d
Z b
/t dt
f x dx; where f x
c
a


d
c
d
c
bc
ad
/
x

:
20
b
a
b
a
b
a
Now consider the examples
in Table
2. It is worth pointing out that in the

p
above examples the weight 1
x2 is cancelled. This is because of the fact that
i x
i 1; 2; . . . ; 10. The functions, apwe convert fi(x) i = 1, 2, . . . , 10 by pg
1
x2
plied in these examples, are almost well behaved on their corresponding intervals, so the value
X1 f j 0
j0 j!
does not have much aection on the integration error, as following tables,
Tables 312 approve our claim. Note that in this tables, SGCQ(n) is n-point
Second Kind GaussChebyshev Quadrature value and ISGCQ(n) is improved
n-point Second Kind GaussChebyshev Quadrature value. Furthermore in

Table 3
R 3:5 x
Exact value of 3 p
dx 0:6362133458
x2
4
n

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

0.6361961339
0.6362119191
0.6362022381
0.6361748218
0.6361285104
0.6361097577

0.0000172119
0.0000014267
0.0000111077
0.0000385240
0.0000848354
0.0001035881

0.5765409181
0.6029466842
0.6150082747
0.6215194250
0.6254318024
0.6279656466

0.0714744407
0.0302845859
0.0167712628
0.0106583927
0.0106583927
0.0082476992

x
Table for f1 x p
.
x2
4

442

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

Table 4
Rp
Exact value of 04 e3x sin2x dx 2:5886286324
n

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

2.5913635373
2.5881743291
2.5854196440
2.5734843377
2.5785719326
2.5470062193

0.0027349049
0.0004543033
0.0032089884
0.0151442947
0.0100566998
0.0416224131

2.1718057877
2.3652133483
2.4481779302
2.4920133498
2.5180409280
2.5461820187

0.4168228447
0.2234152841
0.1404507022
0.0966152826
0.0705877044
0.0538514721

Table for f2(x) = e3xsin(2x).

Table 5
R 0:35
Exact value of 0

2
x2
4

dx
0:1768200201

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

0.1768189979

0.1768193834

0.1768181329

0.1768108286

0.1768138995

0.1767946345

0.0000010222
0.0000006367
0.0000018872
0.0000091915
0.0000061206
0.0000253856

0.1602505769

0.1675815269

0.1709309152

0.1727391261

0.1738256667

0.1745293754

0.0165694432
0.0092384932
0.0058891049
0.0040808940
0.0029943534
0.0022906447

Table for f3 x x22
4.

Table 6
Rp
Exact value of 04 cos2 x dx 0:6426990818
n

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

0.6423179858
0.6427390417
0.6428072923
0.6432153484
0.6430099714
0.6441259432

0.0003810960
0.0000399599
0.0001082105
0.0005162666
0.0003108896
0.0014268614

0.5887974725
0.6123828658
0.6233004634
0.6292308242
0.6328057432
0.6351254557

0.0539016093
0.0303162160
0.0193986184
0.0134682576
0.0098933386
0.0075736261

Table for f4(x) = cos2(x).

Table 7
R 1:5
Exact value of 1 x2 ln x dx 0:1922593577
n

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

0.1922701316
0.1922308382
0.1921955146
0.1919133470
0.1922001088
0.1912983040

0.0000107739
0.0000285195
0.0000638431
0.0003460107
0.0000592489
0.0009610537

0.1703034583
0.1802141237
0.1846293619
0.1869888182
0.1883991693
0.1893097728

0.0219558994
0.0120452340
0.0076299958
0.0052705395
0.0038601884
0.0029495849

Table for f5(x) = x2ln(x).

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

443

Table 8
Rp
Exact value of 04 ecosx dx 1:9397348506
n

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

1.9393769976
1.9397903055
1.9398887486
1.9404676959
1.9401814171
1.9417581054

0.0003578530
0.0000554549
0.0001538980
0.0007328453
0.0004465665
0.0020232548

1.7676020306
1.8433265289
1.8781645858
1.8970293104
1.9083826967
1.9157426023

0.1721328200
0.0964083217
0.0615702648
0.0427055402
0.0313521539
0.0239922483

Table for f6(x) = ecos(x).

Table 9
R1 2
Exact value of 0 e
x dx 0:7468241328
n

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

0.7465976903
0.7468602490
0.7469484070
0.7474305153
0.7471184910
0.7485038769

0.0002264425
0.0000361162
0.0001242742
0.0006063825
0.0002943582
0.0016797441

0.6843446355
0.7116329411
0.7243045723
0.7311880299
0.7353379796
0.7380309825

0.0624794973
0.0351911917
0.0225195605
0.0156361029
0.0114861532
0.0087931503

Table for f7 x e
x .

Table 10
R1
Exact value of 0

1
1ex

dx 0:3798854936

ISGCQ(n)

Error

SGCQ(n)

Error

2
3
4
5
6
7

0.3799077883
0.3798859899
0.3798661953
0.3798380689
0.3796782529
0.3797629548

0.0000222947
0.0000004963
0.0000192983
0.0000474247
0.0002072407
0.0001225388

0.3440073640
0.3598915004
0.3671434245
0.3710569135
0.3734080081
0.3749305272

0.0358781296
0.0199939932
0.0127420691
0.0088285801
0.0064774855
0.0049549664

1
Table for f8 x 1e
x.

Table 11
R1
3
Exact value of 0 1 x2 2 dx 1:5679519622
n

IGCQ(n)

Error

GCQ(n)

Error

2
3
4
5
6
7

1.5667494321
1.5677113568
1.5672340248
1.5645551131
1.5656154078
1.5586107956

0.0012025301
0.0002406054
0.0007179374
0.0033968491
0.0023365544
0.0093411666

1.3822624559
1.4665322297
1.5037937459
1.5236613629
1.5355245823
1.5431795238

0.1856895063
0.1014197325
0.0641582163
0.0442905993
0.0324273799
0.0247724384

Table for f9 x 1 x2 2 .

444

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

Table 12
R1
Exact value of 0

sin2 x
x

dx 0:4596976941

ISGCQ(n)

Error

GSCQ(n)

Error

2
3
4
5
6
7

0.4231670749
0.4237269039
0.4238754908
0.4243757353
0.4248862011
0.4255435875

0.0005239334
0.0000358956
0.0001844825
0.0006847270
0.0011951928
0.0018525792

0.3919363811
0.4056634231
0.4121073762
0.4156316745
0.4177636760
0.4191499751

0.0317546272
0.0180275852
0.0115836321
0.0080593338
0.0059273323
0.0045410332

Table for f10 x sinxx.

Tables 1315 EX is abbreviation of examples and Error = jApproximation

Value
Exact Valuej.

6. Comparison of the solutions

In this section, we intend to compare the improved 2-point, 3-point and 4point formulas, given in Section 3 with n-point Second Kind GaussChebyshev
formulas. Tables 1315 show the eciency of improved 2-point, 3-point and 4point formulas against with n-point Second Kind GaussChebyshev for examples 1 to 10. Of course, this matter is also valid for all other values of n from
n = 5 to n = 8.
In this tables one can observe that the improved 2-point, 3-point and 4-point
formulas act in average with 103-point, 261-point and 108-point formulas of
Second Kind GaussChebyshev quadrature rules for the examples 1 to 10
respectively, which can be very considerable.

Table 13
Comparison ISGCQ(2) and SGCQ(n)
Ex

Exact Value

ISGCQ(2)

Error

SGCQ(n)

Error

1
2
3
4
5
6
7
8
9
10

0.6362133458
2.5886286324

0.1768200201
0.6426990818
0.1922593577
1.9397348506
0.7468241328
0.3798854936
1.5679519622
0.4236910083

0.6361961339
2.5913635373

0.1768189979
0.6423179858
0.1922701316
1.9393769976
0.7465976903
0.3799077883
1.5667494321
0.4231670749

0.0000172119
0.0027349049
0.0000010222
0.0003810960
0.0000107739
0.0003578530
0.0002264425
0.0000222947
0.0012025301
0.0005239334

173
34
375
34
130
64
48
118
35
22

0.6361959588
2.5858452285

0.1768189861
0.06423035823
0.1922484270
1.9393720010
0.7465898453
0.3798631630
1.5667367539
0.4231406863

0.0000173870
0.0027834039
0.0000010340
0.0003954995
0.0000109307
0.0003628496
0.0002342875
0.0000223306
0.0012152083
0.0005503220

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

445

Table 14
Comparison ISGCQ(3) and SGCQ(n)
EX

Exact Value

ISGCQ(3)

Error

SGCQ(n)

Error

1
2
3
4
5
6
7
8
9
10

0.6362133458
2.5886286324

0.1768200201
0.6426990818
0.1922593577
1.9397348506
0.7468241328
0.3798854936
1.5679519622
0.4236910083

0.6362119191
2.5881743291

0.1768193834
0.6427390417
0.1922308382
1.9397903055
0.7468602490
0.3798859899
1.5677113568
0.4237269039

0.0000014267
0.0004543033
0.0000006367
0.0000399599
0.0000285195
0.0000554549
0.0000361162
0.0000004963
0.0002406054
0.0000358956

606
85
478
109
80
165
123
797
79
89

0.6362119175
2.5881678403

0.1768193826
0.6426590426
0.1922307651
1.9396792186
0.7467875487
0.3798849958
1.5677059487
0.4236550606

0.0000014283
0.0004607921
0.0000006375
0.0000400392
0.0000285926
0.0000556320
0.0000365841
0.0000004978
0.0002460135
0.0000359477

Table 15
Comparison ISGCQ(4) and SGCQ(n)
EX

Exact Value

ISGCQ(4)

Error

SGCQ(n)

Error

1
2
3
4
5
6
7
8
9
10

0.6362133458
2.5886286324

0.1768200201
0.6426990818
0.1922593577
1.9397348506
0.7468241328
0.3798854936
1.5679519622
0.4236910083

0.6362022381
2.5854196440

0.1768181329
0.6428072923
0.1921955146
1.9398887486
0.7469484070
0.3798661953
1.5672340248
0.4238754908

0.0000111077
0.0032089884
0.0000018872
0.0001082105
0.0000638431
0.0001538980
0.0001242742
0.0000192983
0.0007179374
0.0001844825

216
102
277
65
53
98
66
127
45
38

0.6362021665
2.5883074028

0.1768181287
0.6425878610
0.1921950200
1.9395784363
0.7466988217
0.3798661925
1.5672077734
0.4234995807

0.0000111793
0.0003212296
0.0000018914
0.0001112208
0.0000643377
0.0001564143
0.0001253111
0.0000193011
0.0007441888
0.0001914276

7. Conclusion
In this research, new numerical integration formulas which are more accurate than the corresponding Second Kind GaussChebyshev rules were presented. The main idea behind our approach was that the lower and upper
bounds were considered as two additional unknowns. This helped us to
develop new numerical integration rules for Second Kind GaussChebyshev
formulas. Numerical examples were also given to show the eciency of our approach. We nally mention that this technique can be applied for Gauss
Legendre, GaussLobatto and rst kind of GaussChebyshev quadrature rules
too [4,8,9,1417].

References
[1] M. Abramowitz, I.A. Stegun, Handbook of Mathematical Functions, Dover, 1972.

446

M. Dehghan et al. / Appl. Math. Comput. 168 (2005) 431446

[2] K. Atkinson, An Introduction to Numerical Analysis, Second, Ed., John Wiley, 1989.
[3] K. Atkinson, Elementary Numerical Analysis, Second Ed., John Wiley, 1993.
[4] E. Babolian, M. MasjedJamei, M.R. Eslahchi, On numerical improvement of GaussLegendre
quadrature rules, Applied Mathematics and Computation, in press.
[5] R.L. Burden, J. Douglas Faires, Numerical Analysis, Seventh Ed., Thomson Learning, 2001.
[6] SD. Conte, C. de Boor, Elementary Numerical AnalysisAn Algorithmic Approach,
McGraw-Hill, 1981.
[7] R. Davis, P. Rabinowitz, Methods of Numerical Integration, Second Ed., Academic Press,
New York, 1984.
[8] M.R. Eslahchi, M. MasjedJamei, E. Babolian, On numerical improvement of GaussLobatto
quadrature rules, Applied Mathematics and Computation, in press.
[9] M.R. Eslahchi, M. Dehghan, M. MasjedJamei, On numerical improvement of the rst kind
GaussChebyshev quadrature rules, Applied Mathematics and Computation, in press.
[10] W. Gautschi, Construction of GaussChristoel quadrature formulas, Math. Comp. 22 (1968)
251270.
[11] W. Gautschi, Orthogonal polynomials-construction theory and applications, J. Comp. Appl.
Math. 1213 (1985) 6167.
[12] G.H. Golub, J.H. Welsch, Calculation of Gauss quadrature rule, Math. Comp. 23 (1969) 221
230.
[13] J. Stoer, R. Bulirsch, Introduction to Numerical Analysis, Second Ed., Springer Verlag, New
York, 1993.
[14] M.R. Eslahchi, M. Dehghan, M. MasjedJamei, On numerical improvement of rst kind
ChebyshevNewtonCotes quadrature rules (closed type), Applied Mathematics and Computation, in press.
[15] M. MasjedJamei, M.R. Eslahchi, M. Dehghan, On numerical improvement of Gauss-Radau
quadrature rules, Applied Mathematics and Computation, in press.
[16] M. Dehghan, M. MasjedJamei, M.R. Eslahchi, On numerical improvement of closed Newton
Cotes quadrature rules, Applied Mathematics and Computation, in press.
[17] M. Dehghan, M. MasjedJamei, M.R. Eslahchi, On numerical improvement of open Newton
Cotes quadrature rules, Applied Mathematics and Computation, submitted.