574 Chapter 20Numerical methods
The area of the trapezium betweenx
j
andx
j+ 1
is(f
j
1 + 1 f
j+ 1
)h 22 , and the total area is
(20.32)
wheref
j
1 = 1 f(a 1 + 1 jh).
For an integrand that is continuous and differentiable in the intervala 1 ≤ 1 x 1 ≤ 1 b, it
can be shown that the error in this formula is
(20.33)
wheref′′(c)is the second derivative of the function at some pointx 1 = 1 cin the interval.
Alternatively, when his small enough, the error can be approximated as (see the
Euler–MacLaurin formula (20.39) below)
(20.34)
wheref
0
′andf
n
′are the derivatives off(x)at the end points. It follows that halving
h(or doubling the number of subintervals n) reduces the error by a factor of four.
EXAMPLE 20.10Find estimates of the integral by means of the
trapezoidal rule (20.32) and the error formula (20.34) forn 1 = 1 3, 6and 12.
The required values of the integrandf(x) 1 = 1 xe
−x
2
are listed in the following table.
Table 20.7 Values off(x) 1 = 1 xe
−
x
2
x 0.0 0.1 0.2 0.3 0.4 0.5 0.6
f(x) 0.000000 0.099005 0.192158 0.274179 0.340858 0.389400 0.418606
x 0.7 0.8 0.9 1.0 1.1 1.2
f(x) 0.428838 0.421834 0.400372 0.367879 0.328017 0.284313
Then, for example, forn 1 = 13 we have
h 1 = 1 0.4, f
0
1 = 1 0, f
1
1 = 1 0.340858, f
2
1 = 1 0.421834, f
3
1 = 1 0.284313
and
T(h) 1 = 1 0.4[0 1 + 1 0.340858 1 + 1 0.421834 1 +× 1 0.284313] 1 = 1 0.361939
1
2
Ixedx
x
=
−
Z
0
12. 2
ε≈− ′− ′
1
12
2
0
hf f
n
Z
a
b
fxdx Th
ba
() −==−() hf c()
−
ε ′′
12
2
Z
a
b
nn
fxdx Th h f f f() ≈= +++++() f f
−
1
2
1
2
012 1