106 CHAPTER 3 Interpolation and Polynomial Approximation

3.1 Interpolation and the Lagrange Polynomial

One of the most useful and well-known classes of functions mapping the set of real numbers
into itself is the algebraic polynomials, the set of functions of the form

Pn (x) = an x n + an1 x n1 + + a1 x + a0 ,

where n is a nonnegative integer and a0 , . . . , an are real constants. One reason for their
importance is that they uniformly approximate continuous functions. By this we mean that
given any function, dened and continuous on a closed and bounded interval, there exists
a polynomial that is as close to the given function as desired. This result is expressed
precisely in the Weierstrass Approximation Theorem. (See Figure 3.1.)

Figure 3.1
y
y  f(x) 
y  P(x)
y  f(x)
y  f(x) 

a b x

Theorem 3.1 (Weierstrass Approximation Theorem)

Suppose that f is dened and continuous on [a, b]. For each  > 0, there exists a polynomial
P(x), with the property that

|f (x) P(x)| < , for all x in [a, b].

The proof of this theorem can be found in most elementary texts on real analysis (see,
for example, [Bart], pp. 165172).
Karl Weierstrass (18151897) is Another important reason for considering the class of polynomials in the approximation
often referred to as the father of of functions is that the derivative and indenite integral of a polynomial are easy to determine
modern analysis because of his and are also polynomials. For these reasons, polynomials are often used for approximating
insistence on rigor in the continuous functions.
demonstration of mathematical The Taylor polynomials were introduced in Section 1.1, where they were described
results. He was instrumental in as one of the fundamental building blocks of numerical analysis. Given this prominence,
developing tests for convergence you might expect that polynomial interpolation would make heavy use of these functions.
of series, and determining ways However this is not the case. The Taylor polynomials agree as closely as possible with
to rigorously dene irrational
a given function at a specic point, but they concentrate their accuracy near that point.
numbers. He was the rst to
A good interpolation polynomial needs to provide a relatively accurate approximation
demonstrate that a function could
be everywhere continuous but
over an entire interval, and Taylor polynomials do not generally do this. For example,
nowhere differentiable, a result suppose we calculate the rst six Taylor polynomials about x0 = 0 for f (x) = ex .
that shocked some of his Since the derivatives of f (x) are all ex , which evaluated at x0 = 0 gives 1, the Taylor
contemporaries. polynomials are

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 3.1 Interpolation and the Lagrange Polynomial 107

x2 x2 x3
Very little of Weierstrasss work P0 (x) = 1, P1 (x) = 1 + x, P2 (x) = 1 + x + , P3 (x) = 1 + x + + ,
was published during his lifetime, 2 2 6
but his lectures, particularly on x2 x3 x4 x2 x3 x4 x5
the theory of functions, had P4 (x) = 1 + x + + + , and P5 (x) = 1 + x + + + + .
2 6 24 2 6 24 120
signicant inuence on an entire
generation of students. The graphs of the polynomials are shown in Figure 3.2. (Notice that even for the
higher-degree polynomials, the error becomes progressively worse as we move away from
zero.)

Figure 3.2
y

20
y  P5(x)
y  ex
y  P4(x)
15
y  P3(x)

10
y  P2(x)

5
y  P1(x)

y  P0(x)
1 1 2 3 x

Although better approximations are obtained for f (x) = ex if higher-degree Taylor

polynomials are used, this is not true for all functions. Consider, as an extreme example,
using Taylor polynomials of various degrees for f (x) = 1/x expanded about x0 = 1 to
approximate f (3) = 1/3. Since

f (x) = x 1 , f  (x) = x 2 , f  (x) = (1)2 2 x 3 ,

and, in general,

f (k) (x) = (1)k k!x k1 ,

the Taylor polynomials are


n
f (k) (1) 
n
Pn (x) = (x 1)k = (1)k (x 1)k .
k!
k=0 k=0

To approximate f (3) = 1/3 by Pn (3) for increasing values of n, we obtain the values in
Table 3.1rather a dramatic failure! When we approximate f (3) = 1/3 by Pn (3) for larger
values of n, the approximations become increasingly inaccurate.

Table 3.1 n 0 1 2 3 4 5 6 7
Pn (3) 1 1 3 5 11 21 43 85

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
108 CHAPTER 3 Interpolation and Polynomial Approximation

For the Taylor polynomials all the information used in the approximation is concentrated
at the single number x0 , so these polynomials will generally give inaccurate approximations
as we move away from x0 . This limits Taylor polynomial approximation to the situation in
which approximations are needed only at numbers close to x0 . For ordinary computational
purposes it is more efcient to use methods that include information at various points. We
consider this in the remainder of the chapter. The primary use of Taylor polynomials in
numerical analysis is not for approximation purposes, but for the derivation of numerical
techniques and error estimation.

Lagrange Interpolating Polynomials

The problem of determining a polynomial of degree one that passes through the distinct
points (x0 , y0 ) and (x1 , y1 ) is the same as approximating a function f for which f (x0 ) = y0
and f (x1 ) = y1 by means of a rst-degree polynomial interpolating, or agreeing with, the
values of f at the given points. Using this polynomial for approximation within the interval
given by the endpoints is called polynomial interpolation.
Dene the functions
x x1 x x0
L0 (x) = and L1 (x) = .
x0 x 1 x1 x 0

The linear Lagrange interpolating polynomial through (x0 , y0 ) and (x1 , y1 ) is

x x1 x x0
P(x) = L0 (x)f (x0 ) + L1 (x)f (x1 ) = f (x0 ) + f (x1 ).
x0 x 1 x1 x 0

Note that

L0 (x0 ) = 1, L0 (x1 ) = 0, L1 (x0 ) = 0, and L1 (x1 ) = 1,

which implies that

P(x0 ) = 1 f (x0 ) + 0 f (x1 ) = f (x0 ) = y0

and

P(x1 ) = 0 f (x0 ) + 1 f (x1 ) = f (x1 ) = y1 .

So P is the unique polynomial of degree at most one that passes through (x0 , y0 ) and
(x1 , y1 ).

Example 1 Determine the linear Lagrange interpolating polynomial that passes through the points (2, 4)
and (5, 1).
Solution In this case we have

x5 1 x2 1
L0 (x) = = (x 5) and L1 (x) = = (x 2),
25 3 52 3
so
1 1 4 20 1 2
P(x) = (x 5) 4 + (x 2) 1 = x + + x = x + 6.
3 3 3 3 3 3
The graph of y = P(x) is shown in Figure 3.3.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 3.1 Interpolation and the Lagrange Polynomial 109

Figure 3.3
y

(2,4)
4
3
2
y  P(x) = x  6
(5,1)
1

1 2 3 4 5 x

To generalize the concept of linear interpolation, consider the construction of a poly-

nomial of degree at most n that passes through the n + 1 points

(x0 , f (x0 )), (x1 , f (x1 )), . . . , (xn , f (xn )).

(See Figure 3.4.)

Figure 3.4
y

y  f (x)

y  P(x)

x0 x1 x2 xn x

In this case we rst construct, for each k = 0, 1, . . . , n, a function Ln,k (x) with the
property that Ln,k (xi ) = 0 when i  = k and Ln,k (xk ) = 1. To satisfy Ln,k (xi ) = 0 for each
i  = k requires that the numerator of Ln,k (x) contain the term

(x x0 )(x x1 ) (x xk1 )(x xk+1 ) (x xn ).

To satisfy Ln,k (xk ) = 1, the denominator of Ln,k (x) must be this same term but evaluated at
x = xk . Thus

(x x0 ) (x xk1 )(x xk+1 ) (x xn )

Ln,k (x) = .
(xk x0 ) (xk xk1 )(xk xk+1 ) (xk xn )

A sketch of the graph of a typical Ln,k (when n is even) is shown in Figure 3.5.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
110 CHAPTER 3 Interpolation and Polynomial Approximation

Figure 3.5
L n,k(x)

x0 x1 ... x k1 xk x k1 ... x n1 xn x

The interpolating polynomial is easily described once the form of Ln,k is known. This
polynomial, called the nth Lagrange interpolating polynomial, is dened in the following
theorem.

Theorem 3.2 If x0 , x1 , . . . , xn are n + 1 distinct numbers and f is a function whose values are given at
The interpolation formula named
these numbers, then a unique polynomial P(x) of degree at most n exists with
for Joseph Louis Lagrange f (xk ) = P(xk ), for each k = 0, 1, . . . , n.
(17361813) was likely known
by Isaac Newton around 1675, This polynomial is given by
but it appears to rst have been
published in 1779 by Edward 
n

Waring (17361798). Lagrange P(x) = f (x0 )Ln,0 (x) + + f (xn )Ln,n (x) = f (xk )Ln,k (x), (3.1)
wrote extensively on the subject k=0
of interpolation and his work had where, for each k = 0, 1, . . . , n,
signicant inuence on later
mathematicians. He published (x x0 )(x x1 ) (x xk1 )(x xk+1 ) (x xn )
Ln,k (x) = (3.2)
this result in 1795. (xk x0 )(xk x1 ) (xk xk1 )(xk xk+1 ) (xk xn )
 n
The symbol is used to write (x xi )
= .
products compactly and parallels
 i=0
(x k xi )
the symbol , which is used for i =k
writing sums.
We will write Ln,k (x) simply as Lk (x) when there is no confusion as to its degree.

Example 2 (a) Use the numbers (called nodes) x0 = 2, x1 = 2.75, and x2 = 4 to nd the second
Lagrange interpolating polynomial for f (x) = 1/x.
(b) Use this polynomial to approximate f (3) = 1/3.
Solution (a) We rst determine the coefcient polynomials L0 (x), L1 (x), and L2 (x). In
nested form they are
(x 2.75)(x 4) 2
L0 (x) = = (x 2.75)(x 4),
(2 2.5)(2 4) 3
(x 2)(x 4) 16
L1 (x) = = (x 2)(x 4),
(2.75 2)(2.75 4) 15
and
(x 2)(x 2.75) 2
L2 (x) = = (x 2)(x 2.75).
(4 2)(4 2.5) 5

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 3.1 Interpolation and the Lagrange Polynomial 111

Also, f (x0 ) = f (2) = 1/2, f (x1 ) = f (2.75) = 4/11, and f (x2 ) = f (4) = 1/4, so


2
P(x) = f (xk )Lk (x)
k=0

1 64 1
= (x 2.75)(x 4) (x 2)(x 4) + (x 2)(x 2.75)
3 165 10
1 2 35 49
= x x+ .
22 88 44
(b) An approximation to f (3) = 1/3 (see Figure 3.6) is
9 105 49 29
f (3) P(3) = + = 0.32955.
22 88 44 88
Recall that in the opening section of this chapter (see Table 3.1) we found that no Taylor
polynomial expanded about x0 = 1 could be used to reasonably approximate f (x) = 1/x
at x = 3.

Figure 3.6
y

2 y  f (x)

1
y  P(x)

1 2 3 4 5 x

The interpolating polynomial P of degree less than or equal to 3 is dened in Maple

with
P := x interp([2, 11/4, 4], [1/2, 4/11, 1/4], x)
   
11 1 4 1
x interp 2, , 4 , , , ,x
4 2 11 4
To see the polynomial, enter
P(x)
1 2 35 49
x x+
22 88 44
Evaluating P(3) as an approximation to f (3) = 1/3, is found with
evalf(P(3))

0.3295454545

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
112 CHAPTER 3 Interpolation and Polynomial Approximation

The interpolating polynomial can also be dened in Maple using the CurveFitting package
and the call PolynomialInterpolation.
The next step is to calculate a remainder term or bound for the error involved in
approximating a function by an interpolating polynomial.

Theorem 3.3 Suppose x0 , x1 , . . . , xn are distinct numbers in the interval [a, b] and f C n+1 [a, b]. Then,
for each x in [a, b], a number (x) (generally unknown) between x0 , x1 , . . . , xn , and hence
in (a, b), exists with

f (n+1) ((x))
f (x) = P(x) + (x x0 )(x x1 ) (x xn ), (3.3)
(n + 1)!
There are other ways that the where P(x) is the interpolating polynomial given in Eq. (3.1).
error term for the Lagrange
polynomial can be expressed, but
this is the most useful form and Proof Note rst that if x = xk , for any k = 0, 1, . . . , n, then f (xk ) = P(xk ), and choosing
the one that most closely agrees
(xk ) arbitrarily in (a, b) yields Eq. (3.3).
with the standard Taylor
If x  = xk , for all k = 0, 1, . . . , n, dene the function g for t in [a, b] by
polynomial error form.
(t x0 )(t x1 ) (t xn )
g(t) = f (t) P(t) [f (x) P(x)]
(x x0 )(x x1 ) (x xn )
n
(t xi )
= f (t) P(t) [f (x) P(x)] .
i=0
(x xi )

Since f C n+1 [a, b], and P C [a, b], it follows that g C n+1 [a, b]. For t = xk , we have
n
(xk xi )
g(xk ) = f (xk ) P(xk ) [f (x) P(x)] = 0 [f (x) P(x)] 0 = 0.
i=0
(x xi )

Moreover,

n
(x xi )
g(x) = f (x) P(x) [f (x) P(x)] = f (x) P(x) [f (x) P(x)] = 0.
i=0
(x xi )

Thus g C n+1 [a, b], and g is zero at the n + 2 distinct numbers x, x0 , x1 , . . . , xn . By

Generalized Rolles Theorem 1.10, there exists a number in (a, b) for which g(n+1) ( ) = 0.
So

n 
(n+1) (n+1) (n+1) d n+1  (t xi )
0=g ( ) = f ( ) P ( ) [f (x) P(x)] n+1 . (3.4)
dt i=0
(x xi )
t=

However P(x) is a polynomial

n of degree at most n, so the (n+1)st derivative, P(n+1) (x),
is identically zero. Also, i=0 [(t xi )/(x xi )] is a polynomial of degree (n + 1), so

n  
(t xi ) 1
= n t n+1 + (lower-degree terms in t),
i=0
(x x i ) i=0 (x x i )

and

d n+1  (t xi )
n
(n + 1)!
= n .
dt n+1
i=0
(x xi ) i=0 (x xi )

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 3.1 Interpolation and the Lagrange Polynomial 113

Equation (3.4) now becomes

(n + 1)!
0 = f (n+1) ( ) 0 [f (x) P(x)] n ,
i=0 (x xi )

and, upon solving for f (x), we have

f (n+1) ( ) 
n
f (x) = P(x) + (x xi ).
(n + 1)! i=0

The error formula in Theorem 3.3 is an important theoretical result because Lagrange
polynomials are used extensively for deriving numerical differentiation and integration
methods. Error bounds for these techniques are obtained from the Lagrange error formula.
Note that the error form for the Lagrange polynomial is quite similar to that for the Tay-
lor polynomial. The nth Taylor polynomial about x0 concentrates all the known information
at x0 and has an error term of the form
f (n+1) ((x))
(x x0 )n+1 .
(n + 1)!
The Lagrange polynomial of degree n uses information at the distinct numbers x0 , x1 , . . . ,
xn and, in place of (x x0 )n , its error formula uses a product of the n + 1 terms (x x0 ),
(x x1 ), . . . , (x xn ):
f (n+1) ((x))
(x x0 )(x x1 ) (x xn ).
(n + 1)!

Example 3 In Example 2 we found the second Lagrange polynomial for f (x) = 1/x on [2, 4] using the
nodes x0 = 2, x1 = 2.75, and x2 = 4. Determine the error form for this polynomial, and
the maximum error when the polynomial is used to approximate f (x) for x [2, 4].
Solution Because f (x) = x 1 , we have

f  (x) = x 2 , f  (x) = 2x 3 , and f  (x) = 6x 4 .

As a consequence, the second Lagrange polynomial has the error form
f  ((x))
(x x0 )(x x1 )(x x2 ) = ((x))4 (x 2)(x 2.75)(x 4), for (x) in (2, 4).
3!
The maximum value of ((x))4 on the interval is 24 = 1/16. We now need to determine
the maximum value on this interval of the absolute value of the polynomial
35 2 49
g(x) = (x 2)(x 2.75)(x 4) = x 3 x + x 22.
4 2
Because

35 49 35 49 1
Dx x 3 x 2 + x 22 = 3x 2 x + = (3x 7)(2x 7),
4 2 2 2 2
the critical points occur at
 
7 7 25 7 7 9
x = , with g = , and x= , with g = .
3 3 108 2 2 16
Hence, the maximum error is
 
f  ((x)) 1  9  3
|(x x0 )(x x1 )(x x2 )| = 0.00586.
3! 16 6  16  512

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
114 CHAPTER 3 Interpolation and Polynomial Approximation

The next example illustrates how the error formula can be used to prepare a table of
data that will ensure a specied interpolation error within a specied bound.
Example 4 Suppose a table is to be prepared for the function f (x) = ex , for x in [0, 1]. Assume the
number of decimal places to be given per entry is d 8 and that the difference between
adjacent x-values, the step size, is h. What step size h will ensure that linear interpolation
gives an absolute error of at most 106 for all x in [0, 1]?
Solution Let x0 , x1 , . . . be the numbers at which f is evaluated, x be in [0,1], and suppose
j satises xj x xj+1 . Eq. (3.3) implies that the error in linear interpolation is
 (2) 
 f ( )  |f (2) ( )|

|f (x) P(x)| =  (x xj )(x xj+1 ) = |(x xj )||(x xj+1 )|.
2! 2

The step size is h, so xj = jh, xj+1 = (j + 1)h, and

|f (2) ( )|
|f (x) P(x)| |(x jh)(x (j + 1)h)|.
2!
Hence

max [0,1] e
|f (x) P(x)| max |(x jh)(x (j + 1)h)|
2 xj xxj+1

e
max |(x jh)(x (j + 1)h)|.
2 xj xxj+1

Consider the function g(x) = (x jh)(x (j + 1)h), for jh x (j + 1)h. Because


h
g (x) = (x (j + 1)h) + (x jh) = 2 x jh ,
2

the only critical point for g is at x = jh + h/2, with g(jh + h/2) = (h/2)2 = h2 /4.
Since g(jh) = 0 and g((j + 1)h) = 0, the maximum value of |g (x)| in [jh, (j + 1)h]
must occur at the critical point which implies that

e e h2 eh2
|f (x) P(x)| max |g(x)| = .
2 xj xxj+1 2 4 8

Consequently, to ensure that the the error in linear interpolation is bounded by 106 , it is
sufcient for h to be chosen so that

eh2
106 . This implies that h < 1.72 103 .
8
Because n = (1 0)/h must be an integer, a reasonable choice for the step size is
h = 0.001.

E X E R C I S E S E T 3.1
1. For the given functions f (x), let x0 = 0, x1 = 0.6, and x2 = 0.9. Construct interpolation polynomials
of degree at most one and at most two to approximate f (0.45), and nd the absolute error.
a. f (x) = cos x c. f (x) = ln(x + 1)

b. f (x) = 1 + x d. f (x) = tan x

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.