RGMIA Research Report Collection, Vol. 1, No.

1, 1998
http://sci.vut.edu.au/rgmia/reports.html

AN OSTROWSKI TYPE INEQUALITY FOR DOUBLE

INTEGRALS AND APPLICATIONS FOR CUBATURE
FORMULAE

N.S. BARNETT AND S.S. DRAGOMIR

Abstract. An inequality of the Ostrowski type for double integrals and appli-
cations in Numerical Analysis in connection with cubature formulae are given.

1 Introduction

In 1938, A. Ostrowski proved the following integral inequality [5, p. 468]

Theorem 1.1. Let f : [a, b] R be a differentiable mapping on (a, b) whose
derivative f 0 : (a, b) R is bounded on (a, b), i.e., kf 0 k := sup |f 0 (t)| < .
t(a,b)
Then we have the inequality
 Z b  "
#
 1  1 x a+b
 
(1.1) f (x) f (t) dt + 2
2 (b a) kf 0 k
 ba a  4 (b a)

for all x [a, b] .

The constant 41 is the best possible.
For some generalizations of this classic fact see the book [5, p. 468-484] by
Mitrinovic, Pecaric and Fink.
Some applications of the above results in Numerical Integration and for
special means have been given in the recent paper [3] by S.S. Dragomir and S.
Wang. For other results of Ostrowskis type see the papers [1], [2] and [4].
In 1975, G.N. Milovanovic generalized Theorem 1.1 where f is a function of
several variables [5, p. 468]
Theorem 1.2. Let f : Rm R be a differentiable   function defined on D =
 f 
{(x1 , ..., xm ) |ai xi bi (i = 1, ..., m)} and let  x i
 Mi (Mi > 0, i = 1, ..., m)
in D. Furthermore, let function x 7 p (x) be integrable and p (x) > 0 for every
x D. Then for every x D, we have the inequality:

 R  P
m R
 p (y) f (y) dy  Mi p (y) |xi yi | dy
  i=1
(1.2) f (x) D R  D
R .
 p (y) dy  p (y) dy

D D

Date. November, 1998

1991 Mathematics Subject Classification. Primary 26D15; Secondary 41A55.
Key words and phrases. Ostrowskis Inequality, Cubature Formulae.

13
14 Barnett and Dragomir

In the present paper we point out an Ostrowski type inequality for double
integrals and apply it in Numerical Integration obtaining a general cubature
formula.

2 The Results

The following inequality of Ostrowskis type for mappings of two variables holds:
00 2f
Theorem 2.1. Let f : [a, b][c, d] R continuous on [a, b][c, d] , fx,y = xy
exists on (a, b) (c, d) and is bounded, i.e.,
 2 
00  f (x, y) 
fs,t := sup  
 xy  <
(x,y)(a,b)(c,d)

then we have the inequality:

Zb Zd Zd Zb
(2.1) | f (s, t) dsdt [(b a) f (x, t) dt + (d c) f (s, y) ds
a c c a

(d c) (b a) f (x, y)] |

"  2 # "  2 #
1 a+b 1 c+d 00

2
(b a) + x
2
(d c) + y fs,t
4 2 4 2

for all (x, y) [a, b] [c, d] .

Proof. We have the equality:
Zx Zy
00
(2.2) (s a) (t c) fs,t (s, t) dtds
a c

Zx Zy
= (s a) [fs0 (s, y) (y c) fs0 (s, t) dt]ds
a c

Zx Zy Zx
= (y c) (s a) fs0 (s, y) ds (s a) fs0 (s, t) ds dt
a c a

Zx Zy Zx
= (y c) (x a) f (x, y) f (s, y) ds (x a) f (x, t) f (s, t) ds dt
a c a

Zx
= (y c) (x a) f (x, y) (y c) f (s, y) ds
a

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

Ostrowski Inequality for Double Integrals 15

Zy Zx Zy
(x a) f (x, t) dt + f (s, t) dsdt.
c a c

Also, by similar computations we have

Zx Zd
00
(2.3) (s a) (t d) fs,t (s, t) dsdt
a y

Zx Zd
= (s a) (d y) fs0 (s, y) fs0 (s, t) dt ds
a y

Zx Zd Zx
= (d y) (s a) fs0 (s, y) ds (s a) fs0 (s, t) ds dt
a y a

Zx Zd Zx
= (d y) (x a) f (x, y) f (s, y) ds (x a) f (x, t) f (s, t) ds dt
a y a

Zx
= (x a) (d y) f (x, y) (d y) f (s, y) ds
a

Zd Zx Zy
(x a) f (x, t) dt + f (s, t) dsdt.
y a c

Now,

Zb Zd
00
(2.4) (s b) (t d) fs,t (s, t) dsdt
x y

Zb Zd
= (s b) (d y) fs0 (s, y) fs0 (s, t) dt ds
x y

Zb Zd Zb
= (d y) (s b) fs0 (s, y) ds (s b) fs0 (s, t) ds dt
x y x

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

16 Barnett and Dragomir

Zb Zd Zb
= (d y) (b x) f (x, y) f (s, y) ds (b x) f (x, t) f (s, t) ds dt
x y x

Zb
= (d y) (b x) f (x, y) (d y) f (s, y) ds
x

Zd Zb Zd
(b x) f (x, t) dt + f (s, t) dsdt
y x y

and finally

Zb Zy
00
(2.5) (s b) (t c) fs,t (s, t) dsdt
x c

Zb Zy
= (s b) (y c)fs0 (s, y) fs0 (s, t) dt ds
x c

Zb Zy Zb
= (y c) (s b) fs0 (s, y) ds (s b) fs0 (s, t) ds dt
x c x

Zb Zy Zb
= (y c) (b x) f (x, y) f (s, y) ds (b x) f (x, t) f (s, t) ds dt
x c x

Zb
= (y c) (b x) f (x, y) (y c) f (s, y) ds
x

Zy Zb Zy
(b x) f (x, t) dt + f (s, t) dsdt.
c x c

If we add the equalities (2.2) (2.5) we get in the right membership:

[(y c) (x a) + (x a) (d y) + (d y) (b x) + (y c) (b x)] f (x, y)

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

Ostrowski Inequality for Double Integrals 17

Zx Zb Zy
(d c) f (s, y) ds (d c) f (s, y) ds (b a) f (x, t) dt
a x c

Zd Zx Zy Zx Zd
(b a) f (x, t) dt + f (s, t) dsdt + f (s, t) dsdt
y a c a y

Zb Zd Zb Zy
+ f (s, t) dsdt + f (s, t) dsdt
x y x c

Zb
= (d c) (b a) f (x, y) (d c) f (s, y) ds
a

Zb Zb Zd
(b a) f (x, t) dt + f (s, t) dsdt.
c a c
2
For the first membership, let us define the kernels: p : [a, b] R,
2
q : [c, d] R given by:

sa if s [a, x]
p (x, s) :=

sb if s (x, b]
and

tc if t [c, y]
q (y, t) := .

td if t (y, d]
Now, using these notations, we deduce that the left membership can be
represented as :
Zb Zd
00
p (x, s) q (y, t) fs,t (s, t) dsdt.
a c

Consequently, we get the identity

Zb Zd
00
(2.6) p (x, s) q (y, t) fs,t (s, t) dsdt
a c

Zb Zd Zb Zd
= (d c) (b a) f (x, y) (d c) f (x, y) ds (b a) f (x, t) dt + f (s, t) dsdt
a c a c

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

18 Barnett and Dragomir

for all (x, y) [a, b] [c, d] .

Now, using the identity (2.6) we get

Zb Zd Zd Zb
(2.7) | f (s, t) dsdt [(b a) f (x, t) dt + (d c) f (x, y) ds
a c c a

(d c) (b a) f (x, y)] |

Zb Zd Zb Zd
 00  00
|p (x, s)| |q (y, t)| fs,t (s, t) dsdt fs,t

|p (x, s)| |q (y, t)| dsdt.
a c a c

Now, observe that

Zb Zx Zb
|p (x, s)| ds = (s a) ds + (b s) ds
a a x

2 2  2
(x a) + (b x) 1 2 a+b
= = (b a) + x
2 4 2

and, similarly,

Zd  2
1 2 c+d
|q (y, t)| dt = (d c) + y .
4 2
c

Finally, using (2.7) , we get the desired inequality (2.1) .

Corollary 2.2. Under the above assumptions, we have the inequality:

 b d
Z Z Zd  
 a+b
(2.8) 
f (s, t) dsdt (b a) f , t dt
 2

a c c


Zb     
c+d a + b c + d 
+ (d c) f s, ds (d c) (b a) f , 
2 2 2 
a

1 2 00
(b a) (d c) fs,t .
2

16

Remark 2.1. The constants 41 from the first and the second bracket are optimal
in the sense that not both of them can be less than 14 .

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

Ostrowski Inequality for Double Integrals 19

Indeed, if we had assumed that there exists c1 , c2 0, 41 so that

Zb Zd Zd Zb
(2.9) | f (s, t) dsdt [(b a) f (x, t) dt + (d c) f (s, y) ds
a c c a

(d c) (b a) f (x, y)]|

"  2 # "  2 #
a+b c+d 00
c1 (b a) + x
2 2
c2 (d c) + y fs,t
2 2

for all f as in Theorem 2.1 and (x, y) [a, b] [c, d] , then we would have had
for f (s, t) = st and x = a, y = c that:
Zb Zd


b2 a2 d2 c2
f (s, t) dsdt = ,
4
a c

Zd Zb
d2 c2 b2 a2
f (x, t) dt = a , f (s, y) ds = c ,
2 2
c a

00
fs,t = 1

and by (2.9) , the inequality:

 
 b2 a2
d2 c2
d2 c2 b2 a2 
 
 (b a) a (d c) c + (d c) (b a) ac
 4 2 2 

   
2 1 2 1
(b a) c1 + (d c) c2 +
4 4
i.e.
2 2   
(b a) (d c) 2 2 1 1
(b a) (d c) c1 + c2 +
4 4 4
i.e.
  
1 1 1
(2.10) c1 + c2 + .
4 4 4


Now, as we have assumed that c1 , c2 0, 14 , we get
1 1 1 1
c1 + < , c2 + <
4 2 4 2



and then c1 + 14 c2 + 14 < 1
4 which contradicts the inequality (2.10) , and the
statement is proved.

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

20 Barnett and Dragomir

Remark 2.2. Now, if we assume that f (s, t) = h (s) h (t) , h : [a, b] R, h is

continuous and suppose that kh0 k < , then from (2.1) we get (for x = y)
 b
Z Zb Zb

 h (s) ds h (s) ds h (x) (b a) h (s) ds


a a a

Zb 

h (x) (b a) h (s) ds + (b a) h (x)
2 2

"  2 #2
1 2 a+b 2
(b a) + x khk
4 2

i.e.
b 2
Z   2
h (s) ds h (x) (b a) 1 (b a)2 + x a + b 2
khk
4 2
a

which is clearly equivalent to Ostrowskis inequality.

Consequently (2.1) can be also regarded as a generalization for double inte-
grals of the classical result due to Ostrowski.

3 Applications for Cubature Formulae

Let us consider the arbitrary division In : a = x1 < x1 < ... < xn1 <
xn = b and Jm : c = y0 < y1 < ... < ym1 < ym = b and i [xi , xi+1 ]
(i = 0, ..., n 1) , j [yj , yj+1 ] (j = 0, ..., m 1) be intermediate points.
Consider the sum
yZj+1
n1
X m1
X
(3.1) C (f, In , Jm , , ) := hi f ( i , t) dt
i=0 j=0 yj

x
Zi+1
n1
X m1
X n1
X m1
X



+ lj f s, j ds hi lj f i , j
i=0 j=0 xi i=0 j=0

for which we assume that the involved integrals can more easily be computed
than the original double integral
Zb Zd
D := f (s, t) dsdt,
a c

and

hi := xi+1 xi (i = 0, ..., n 1) , lj := yj+1 yj (j = 0, ..., m 1) .

With this assumption, we can state the following cubature formula:

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

Ostrowski Inequality for Double Integrals 21

Theorem 3.1. Let f : [a, b] [c, d] R be as in Theorem 2.1 and In , Jm ,

and be as above. Then we have the cubature formula:
Zb Zd
(3.2) f (s, t) dsdt = C (f, In , Jm , , ) + R (f, In , Jm , , )
a c

where the remainder term R (f, In , Jm , , ) satisfies the estimation:

(3.3) |R (f, In , Jm , , )|

"  2 # "  2 #
00 n1
X m1
X 1 xi + xi+1 1 y j + y j+1
fs,t 2
h + i 2
l + j
i=0 j=0
4 i 2 4 j 2

1 n1
X m1 X
fs,t
00
h2i lj2 .
4
i=0 j=0

Proof. Apply Theorem 2.1 on the interval [xi , xi+1 ] [yj , yj+1 ]
(i = 0, ..., n 1; j = 0, ..., m 1) to get:

 
 xZi+1 yZj+1 yZj+1 x
Zi+1 




 f (s, t) dsdt hi f ( i , t) dt + lj f s, j ds hi lj f i , j 
 
 xi y j yj xi 

"  2 # "  2 #
1 2 xi + xi+1 1 2 yj + yj+1 00
h + i l + j fs,t
4 i 2 4 j 2

for all i = 0, ..., n 1; j = 0, ..., m 1.

Summing over i from 0 to n 1 and over j from 0 to m 1 and using the
generalized triangle inequality we deduce the first inequality in (3.3) .
For the second part we observe that
   
   
 i xi + xi+1  1 hi and  j yj + yj+1  1 lj
 2  2  2  2

for all i, j as above.

Remark 3.1. As
n1
X n1
X
h2i (h) hi = (b a) (h)
i=0 i=0

and
m1
X m1
X
lj2 (l) lj = (d c) (l)
j=0 j=0

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

22 Barnett and Dragomir

where

(h) = max {hi : i = 0, ..., n 1} ,

(l) = max {lj : j = 0, ..., m 1} ,

the right membership of (3.3) can be bounded by

1 fs,t
00


(b a) (d c) (h) (l)
4
which is of order 2 precision.
Now, define the sum

n1
X m1
X
yZj+1  
xi + xi+1
CM (f, In , Jm ) := hi f , t dt
i=0 j=0
2
yj

n1
X m1
X
x
Zi+1   n1
X m1
X  
yj + yj+1 xi + xi+1 yj + yj+1
+ lj f s, ds hi lj f , .
i=0 j=0
2 i=0 j=0
2 2
xi

Then we have the best cubature formula we can get from (3.2) .
Corollary 3.2. Under the above assumptions we have

Zb Zd
(3.4) f (s, t) dsdt = CM (f, In , Jm ) + R (f, In , Jm )
a c

when the remainder R (f, In , Jm ) satisfies the estimation:

00 n1 m1
fs,t X X
|R (f, In , Jm )|
h2i lj2 .
16 i=0 j=0

References

[1] S.S. DRAGOMIR and S. WANG , A new inequality of Ostrowskis type

in L1 norm and applictions to some special means and to some numerical
quadrature rules, Tamkang J. of Math., 28(1997), 239-244.

[2] S.S. DRAGOMIR and S. WANG, An inequality of Ostrowski-Gruss type

and its applications to the estimation of error bounds for some special
means and for some numerical quadrature rules, Computers Math. Applic.,
33(1997), 15-20.

[3] S.S. DRAGOMIR and S. WANG, Applications of Ostrowskis inequality to

the estimation of error bounds for some special means and some numerical
quadrature rules, Appl. Math. Lett., 11(1998), 105-109.

RGMIA Research Report Collection, Vol. 1, No. 1, 1998

Ostrowski Inequality for Double Integrals 23

[4] S.S. DRAGOMIR and S. WANG, A new inequality of Ostrowskis type in

Lp norm and applications to some special means and to some numerical
quadrature rules, submitted for publications

[5] D.S. MITRINOVIC, J.E. PECARIC and A.M. FINK, Inequalities Involving
Functions and Their Integrals and Derivatives, Kluwer Academic Publishers,
Dordrecht, 1994.

School of Communications and Informatics, Victoria University of Technology, PO

Box 14428, MCMC Melbourne, 8001 Victoria, Australia
E-mail address: {neil, sever}@matilda.vut.edu.au

RGMIA Research Report Collection, Vol. 1, No. 1, 1998