Chapter 07.

04
Romberg Rule of Integration
After reading this chapter, you should be able to:
1. derive the Romberg rule of integration, and
2. use the Romberg rule of integration to solve problems.
What is integration?
Integration is the process of measuring the area under a function plotted on a graph. Why
would we want to integrate a function? Among the most common examples are finding the
velocity of a body from an acceleration function, and displacement of a body from a velocity
function. hroughout many engineering fields, there are !what sometimes seems li"e#
countless applications for integral calculus. $ou can read about some of these applications in
%hapters &'.&&A(&'.&&).
*ometimes, the evaluation of expressions involving these integrals can become daunting, if
not indeterminate. +or this reason, a wide variety of numerical methods has been developed
to simplify the integral.
,ere, we will discuss the -omberg rule of approximating integrals of the form
( )

·
b
a
dx x f I
!1#
where
# !x f
is called the integrand
· a
lower limit of integration
· b upper limit of integration
&'.&..1
&'.&..2 %hapter &'.&.
Error in Multiple-Segment Trapezoidal Rule
he true error obtained when using the multiple segment trape/oidal rule with
n
segments
to approximate an integral
( )

b
a
dx x f
is given by
( )
( )
n
f
n
a b
E
n
i
i
t

·
′ ′

− ·
1
2
0
12
ξ
!2#
where for each i , i
ξ
is a point somewhere in the domain ( ) [ ] ih a h i a + − + , 1 , and
the term
( )
n
f
n
i
i ∑
·
′ ′
1
ξ
can be viewed as an approximate average value of
( ) x f ′ ′
in
[ ] b a,
.
his leads us to say that the true error t
E
in 12uation !2# is approximately proportional to
2
1
n
E
t
α ≈ !0#
Figure 1 Integration of a function.
-omberg rule of Integration &'.&..0
for the estimate of
( )

b
a
dx x f
using the
n
(segment trape/oidal rule.
able 1 shows the results obtained for
dt t
t

,
_

¸
¸

1
]
1

¸


0&
3
3 . 4
21&& 1.&&&&
1.&&&&
ln 2&&&
using the multiple(segment trape/oidal rule.
Table 1 5alues obtained using multiple segment trape/oidal rule for

,
_

¸
¸

1
]
1

¸


·
0&
3
3 . 4
21&& 1.&&&&
1.&&&&
ln 2&&& dt t
t
x
.

n
Approximate
5alue
t
E 6
t
∈ 6
a

1 11373 3&' − '.247 (((
2 11277 2&8 − 1.38. 8.0.0
0 11180 . . 41 − &.3278 1.&14
. 11110 8 . 81 − &..788 &.084.
8 11&4. & . 00 − &.2431 &.1774
7 11&3. 4 . 22 − &.2&'& &.&4&32
' 11&'3 3 . 17 − &.1821 &.&8.32
3 11&'. 4 . 12 − &.1178 &.&087&
he true error for the 1(segment trape/oidal rule is 3&' − , while for the 2(segment rule, the
true error is 2&8 − . he true error of 2&8 − is approximately a 2uarter of 3&' − . he
true error gets approximately 2uartered as the number of segments is doubled from 1 to 2.
he same trend is observed when the number of segments is doubled from 2 to . !the true
error for 2(segments is 2&8 − and for four segments is 8 . 81 − #. his follows 12uation !0#.
his information, although interesting, can also be used to get a better approximation of the
integral. hat is the basis of -ichardson9s extrapolation formula for integration by the
trape/oidal rule.
Richardson’s Extrapolation Formula or Trapezoidal Rule
he true error,
t
E
, in the
n
(segment trape/oidal rule is estimated as
2
1
n
E
t
α ≈
2
n
C
E
t
≈ !.#
where C is an approximate constant of proportionality.
*ince
n t
I T E − ·
!8#
where
&'.&... %hapter &'.&.
T : true value
n
I
: approximate value using
n
(segments
hen from 12uations !.# and !8#,
n
I T
n
C
− ≈
2
!7#
If the number of segments is doubled from
n
to n 2 in the trape/oidal rule,
( )
n
I T
n
C
2
2
2
− ≈
!'#
12uations !7# and !'# can be solved simultaneously to get
0
2
2
n n
n
I I
I T

+ ≈ !3#
Example 1
he vertical distance in meters covered by a roc"et from 3 · t to 0& · t seconds is given by

,
_

¸
¸

1
]
1

¸


·
0&
3
3 . 4
21&& 1.&&&&
1.&&&&
ln 2&&& dt t
t
x
a# ;se -omberg9s rule to find the distance covered. ;se the 2(segment and .(segment
trape/oidal rule results given in able 1.
b# +ind the true error for part !a#.
c# +ind the absolute relative true error for part !a#.
Solution
a# m 11277
2
· I
m 11110
.
· I
;sing -ichardson9s extrapolation formula for the trape/oidal rule, the true value is given by
0
2
2
n n
n
I I
I T

+ ≈
and choosing
2 · n
,
0
2 .
.
I I
I T

+ ≈

0
11277 11110
11110

+ ·

m 11&72 ·
b# he exact value of the above integral is

,
_

¸
¸

1
]
1

¸


·
0&
3
3 . 4
21&& 1.&&&&
1.&&&&
ln 2&&& dt t
t
x

m 11&71 ·
so the true error
5alue e Approximat 5alue rue − ·
t
E
11&72 11&71− ·

m 1 − ·
c# he absolute relative true error, t

, would then be
-omberg rule of Integration &'.&..8
1&&
5alue rue
1rror rue
× · ∈
t

1&&
11&71
11&72 11&71
×

·
6 &&4&. . & ·
able 2 shows the -ichardson9s extrapolation results using 1, 2, ., and 3 segments. -esults
are compared with those of the trape/oidal rule.
Table ! 5alues obtained using -ichardson9s extrapolation formula for the trape/oidal rule
for

,
_

¸
¸

1
]
1

¸


·
0&
3
3 . 4
21&& 1.&&&&
1.&&&&
ln 2&&& dt t
t
x
.
n
rape/oidal -ule
6
t

for rape/oidal
-ule
-ichardson9s
1xtrapolation
6
t

for -ichardson9s
1xtrapolation
1
2
.
3
11373
11277
11110
11&'.
'.247
1.38.
&..788
&.1178
((
11&78
11&72
11&71
((
&.&0717
&.&&4&.1
&.&&&&
Romberg "ntegration
-omberg integration is the same as -ichardson9s extrapolation formula as given by
!3# 12uation
. ,owever, -omberg used a recursive algorithm for the extrapolation as
follows.
he estimate of the true error in the trape/oidal rule is given by
( )
( )
n
f
n
a b
E
n
i
i
t

·
′ ′

− ·
1
2
0
12
ξ
*ince the segment width, h , is given by
n
a b
h

·
12uation !2# can be written as
( )
( )
n
f
a b h
E
n
i
i
t

·
′ ′

− ·
1
2
12
ξ
!4#
he estimate of true error is given by
2
Ch E
t
≈ !1&#
It can be shown that the exact true error could be written as
...
7
0
.
2
2
1
+ + + · h A h A h A E
t
!11#
and for small h ,
( )
. 2
1
h ! h A E
t
+ · !12#
&'.&..7 %hapter &'.&.
*ince we used
2
Ch E
t
≈ in the formula !12uation !12##, the result obtained from
!1&# 12uation
has an error of ( )
.
h ! and can be written as
( )
0
2
2 2
n n
n R n
I I
I I

+ ·

1 .
1 2
2
2


+ ·

n n
n
I I
I !10#
where the variable T is replaced by
( )
R n
I
2
as the value obtained using -ichardson9s
extrapolation formula. <ote also that the sign ≈ is replaced by the sign :.
,ence the estimate of the true value now is
( )
.
2
Ch I T
R n
+ ≈
=etermine another integral value with further halving the step si/e !doubling the number of
segments#,
( )
0
2 .
. .
n n
n R n
I I
I I

+ · !1.#
then
( )
.
.
2

,
_

¸
¸
+ ≈
h
C I T
R n
+rom 12uation !10# and !1.#,
( )
( ) ( )
18
2 .
.
R n R n
R n
I I
I T

+ ≈
( )
( ) ( )
1 .
1 0
2 .
.


+ ·

R n R n
R n
I I
I !18#
he above e2uation now has the error of ( )
7
h ! . he above procedure can be further
improved by using the new values of the estimate of the true value that has the error of
( )
7
h ! to give an estimate of ( )
3
h ! .
>ased on this procedure, a general expression for -omberg integration can be written as
2 ,
1 .
1
, 1 1 , 1
1 , 1 ,



+ ·

− + −
+ −
"
I I
I I
"
# " # "
# " # "

!17#
he index " represents the order of extrapolation. +or example, 1 · " represents the
values obtained from the regular trape/oidal rule, 2 · " represents the values obtained using
the true error estimate as ( )
2
h ! , etc. he index
#
represents the more and less accurate
estimate of the integral. he value of an integral with a
1 + #
index is more accurate than the
value of the integral with a
#
index.
+or 2 · " ,
1 · #
,
1 .
1 2
1 , 1 2 , 1
2 , 1 1 , 2


+ ·

I I
I I
-omberg rule of Integration &'.&..'

0
1 , 1 2 , 1
2 , 1
I I
I

+ ·
+or 0 · " ,
1 · #
,
1 .
1 0
1 , 2 2 , 2
2 , 2 1 , 0


+ ·

I I
I I
18
1 , 2 2 , 2
2 , 2
I I
I

+ · !1'#
Example !
he vertical distance in meters covered by a roc"et from 3 · t to 0& · t seconds is given by

,
_

¸
¸

1
]
1

¸


·
0&
3
3 . 4
21&& 1.&&&&
1.&&&&
ln 2&&& dt t
t
x
;se -omberg9s rule to find the distance covered. ;se the 1, 2, ., and 3(segment trape/oidal
rule results as given in able 1.
Solution
+rom able 1, the needed values from the original the trape/oidal rule are
11373
1 , 1
· I
11277
2 , 1
· I
11110
0 , 1
· I
11&'.
. , 1
· I
where the above four values correspond to using 1, 2, . and 3 segment trape/oidal rule,
respectively. o get the first order extrapolation values,
0
1 , 1 2 , 1
2 , 1 1 , 2
I I
I I

+ ·

0
11373 11277
11277

+ ·
11&78 ·
*imilarly
0
2 , 1 0 , 1
0 , 1 2 , 2
I I
I I

+ ·

0
11277 11110
11110

+ ·
11&72 ·
0
0 , 1 . , 1
. , 1 0 , 2
I I
I I

+ ·

0
11110 11&'.
11&'.

+ ·
11&71 ·
+or the second order extrapolation values,
18
1 , 2 2 , 2
2 , 2 1 , 0
I I
I I

+ ·
&'.&..3 %hapter &'.&.

18
11&78 11&72
11&72

+ ·
11&72 ·
*imilarly
18
2 , 2 0 , 2
0 , 2 2 , 0
I I
I I

+ ·

18
11&72 11&71
11&71

+ ·
11&71 ·
+or the third order extrapolation values,
70
1 , 0 2 , 0
2 , 0 1 , .
I I
I I

+ ·

70
11&72 11&71
11&71

+ ·

m 11&71 ·
able 0 shows these increasingly correct values in a tree graph.
Table # Improved estimates of the value of an integral using -omberg integration.
11373
11277
11110
11&'.
11&78
73
11&72
73
11&71
373
11&72
373
11&71
373
11&71
373
1(segment
2(segment
.(segment
3(segment
+irst ?rder *econd ?rder hird ?rder
I<1)-AI?<
opic -omberg -ule
*ummary extboo" notes of -omberg -ule of integration.
@aAor )eneral 1ngineering
Authors Autar Baw
=ate Canuary 27, 2&1.
Web *ite httpDEEnumericalmethods.eng.usf.edu