1

Chapter 3. Polynomial Interpolation/Function Approximation

3.1. Lagrange Interpolation Formula
For a set of data points: x i , y i , i 1, 2, , n and n 1 , the elementary
Lagrange interpolation formula is
x xj
,
x xj
j 1, j i i
n

l in ( x )

i 1, 2, 3, ..., n

(3.1.1)

l in (x) is a polynomial with degree no greater than n 1 . Its value at any data
point x k within the data set is either 1 or 0
l in ( x k )

xk x j
1,
x xj
j 1, j i i

for i k

(3.1.2)

l in ( x k )

xk x j
0,
x xj
j 1, j i i

for i k

(3.1.3)

which is equivalent to
1, i k

0, i k

l in ( x k ) ik

(3.1.4)

The Lagrange interpolation polynomial of degree n 1 is

n

p n 1 ( x ) y i l in ( x )

(3.1.5)

i 1

and its values at the data points are

n

i 1

i 1

p n1 ( x k ) y i l in ( x k ) y i ik y k ,

k 1, 2, , n

which means p n1 ( x ) passes through all the data points exactly.

(3.1.6)

2
For the simplest case where n 2 , there are only two data points and p1 ( x) is a
linear function which passes through the two data points. Thus, p1 ( x) is just a straight
line with its two end points being the two data points. From (3.1.6), we have
p1 ( x ) y1

x x2
x x1
y y1
y2
y1 2
( x x1 )
x1 x 2
x 2 x1
x 2 x1

(3.1.7)

For n 3 , p 2 ( x ) is the quadratic polynomial that passes through three data

points.
p 2 ( x ) y1

( x x 2 )( x x 3 )
( x x1 )( x x 3 )
( x x1 )( x x 2 )
y2
y3
( x1 x 2 )( x1 x 3 )
( x 2 x1 )( x 2 x 3 )
( x 3 x1 )( x 3 x 2 )

(3.1.8)

An advantageous property of the Lagrange interpolation polynomial is that the

data points need not be arranged in any particular order, as long as they are mutually
distinct. Thus the order of the data points is not important. For an application of the
Lagrange interpolation polynomial, say we know y1 and y 2 or y1 , y 2 and y 3 , then
we can estimate the function y ( x) anywhere in [ x1 , x 2 ] linearly and [ x1 , x3 ]
quadratically. This is what we do in finite element analysis.

3.2 Newton Interpolating Polynomial

Suppose there is a known polynomial p n1 ( x ) that interpolates the data set:
( xi , y i , i 1, 2, , n ) . When one more data point ( x n 1 , y n 1 ) , which is distinct
from all the old data points, is added to the data set, we can construct a new polynomial
that interpolates the new data set. Keep also in mind that the new data point need not be
at either end of the old data set. Consider the following polynomial of degree n
n

p n ( x ) p n1 ( x ) c n ( x x i )
i 1

(3.2.1)

where c n is an unknown constant. In the case of n 1 , we specify p 0 ( x ) as

p 0 ( x ) y1 , where data point 1 need not be at the beginning of the data set.
At the points of the old data set, the values of p n (x) are the same as those of
p n1 ( x ) . This is because the second term in Equation 3.2.1 is zero there. Since we
assume p n1 ( x ) interpolates the old data set, p n (x) does so too.
At the new data point, we want p n ( x n 1 ) y n 1 . This can be accomplished by
setting the coefficient c n to be

3
p (x
) p n1 ( x n1 ) y n1 p n1 ( x n1 )
c n n n1

( x n1 x i )

(3.2.2)

( x n1 x i )

i 1

i 1

which definitely exists since x n 1 is distinct from x i ( i 1, 2, , n) . Now p n (x) is a

polynomial that interpolates the new data set.
For any given data set: ( xi , y i , i 1, 2, , n ) , we can obtain the interpolating
polynomial by a recursive process that starts from p 0 ( x ) and uses the above
construction to get p1 ( x) , p 2 ( x ) , , p n 1 ( x ) . We will demonstrate this process
through the following example.
Example 3.2.1. Construct an interpolating polynomial for the following data set using the
formula in Equations 3.2.1 and 3.2.2
i
x
y

1
0
0

2
5
2

3
7
-1

4
8
-2

5
10
20

Step1: for i 1
p 0 ( x ) c 0 y1 0

Step 2: adding point #2

p1 ( x ) p0 ( x ) c1 ( x x1 ) c1 x

Applying p1 ( x 2 ) y 2 , we get c1 0.4 . So

p1 ( x ) p0 ( x ) c1 ( x x1 ) 0.4 x

Step 3: adding point #3

p 2 ( x ) p1 ( x ) c 2 ( x x1 )( x x 2 ) 0.4 x c 2 x ( x 5)

Applying p 2 ( x3 ) y 3 , we get c 2 0.2714 . So

p 2 ( x ) 0.4 x 0.2714 x ( x 5)

Step 4: adding point #4

p 3 ( x ) p 2 ( x ) c 3 ( x x1 )( x x 2 )( x x 3 ) p 2 ( x ) c 3 x ( x 5)( x 7 )

4
Applying p 3 ( x 4 ) y 4 , we get c3 0.0548 . So
p 3 ( x ) p 2 ( x ) 0.0548 x ( x 5)( x 7)

Step 5: adding point #5

p4 ( x ) p 3 ( x ) c 4 ( x x1 )( x x 2 )( x x 3 )( x x 4 ) p 3 ( x ) c 4 x( x 5)( x 7)( x 8)

Applying p 4 ( x5 ) y 5 , we get c 4 0.0712 . So

p 4 ( x) p3 ( x ) 0.0712 x( x 5)( x 7)( x 8)

which is the final answer.

If we expand the recursive form, the r.h.s of Equation (3.2.1), we obtain the more
familiar form of a polynomial
p n 1 ( x) c 0 c1 ( x x1 ) c 2 ( x x1 )( x x 2 ) c n ( x x1 )( x x 2 ) ( x x n )

(3.2.3)

which is called the Newtons interpolation polynomial. Its constants can be determined
from the data set: ( xi , y i , i 1, 2, , n )
p 0 ( x 1 ) y1 c 0
p1 ( x 2 ) y 2 c 0 c1 ( x 2 x1 )
(3.2.4)
p 2 ( x 3 ) y 3 c 0 c1 ( x 3 x1 ) c 2 ( x 3 x1 )( x 3 x 2 )

which gives

c 0 y1
y c
y y1
c1 2 0 2
x 2 x1 x 2 x 1
y c c ( x x1 )
c2 3 0 1 3
( x 3 x1 )( x 3 x 2 )

(3.2.5)

We should note that forcing the polynomial through data with no regard for rates
of change in the data (i.e., derivatives) results in a C 0 continuous interpolating
polynomial. Alternatively, each data condition p ( xi ) y i is called a C 0 constraint.
Lets use the following notation for these constants

5
c 0 [ x 1 ] y y1

c1 [ x 1 , x 2 ] y

[ x 2 ] y [ x1 ] y
x 2 x1

c i 1 [ x1 , x 2 x i ] y ,

i 1, 2,

(3.2.6)

which has the following property

[ x1 x 2 x i ] y

[ x 2 , x 3 x i ] y [ x1 , x 2 x i 1 ] y
x i x1

(3.2.7)

so that it is called the divided difference. For the proof of this formula, please refer to
Gautschi (1997), but for example:
x1 , x 2 y y 2 y1 , x1 , x 2 , x 3 y x 2 , x 3 y x1 , x 2 y , etc. Using this recursion
x 2 x1
x 3 x1
property, a table of divided differences can be generated as follows
x1
x2
x3
x4

y1
y 2 [ x1 x 2 ] y
y 3 [ x 2 x 3 ] y [ x1 x 2 x 3 ] y
y 4 [ x 3 x 4 ] y [ x 2 x 3 x 4 ] y [ x1 x 2 x 3 x 4 ] y

(3.2.8)

This table can be viewed as part of an n (n 1) matrix for a data set that has n
points. The first column is the x values of the data set and the second column is the y or
function values of the data set. For the rest of the ( n 1) ( n 1) lower triangle, the rule
for the construction of its elements, say, element(i, j), is as follows
(1) It takes the form of a division ( j 3, i j 1 )
element ( i , j )

element ( i , j 1) element ( i 1, j 1)
element ( i , 1) element ( i j 2, 1)

(3.2.9)

where
Element(i, j-1) is the element in the matrix immediately left of element(i, j);
Element(i-1, j-1) is the element above and immediately left;
Element(i, 1) is the x on the same row;
Element(i-j+2, 1). To find it, going from element(i, j) diagonally upward and leftward,
when reaching the second column, it is the element to the left.
(2) The denominator is easier to see in this form:

6
[ xk xl xq ] y

element on left element above left

xq x k

Example 3.2.2. Using the table of divided differences, construct the Newton interpolating
polynomial for the data set in Example 3.2.1.
0

10

20

20
0 .4
50
1 2
1 . 5
75
2 1
1
87
20 2
11
10 8

1 .5 0 .4
0.271
70
1 1 .5
0.167
85
11 1
4
10 7

0.167 0.271
0.055
80
4 0.167
0.767
10 5

0.767 0.055
0.071
10 0

where the number of decimal digits is reduced so that the table fits the page here. The
bracketed terms <> are the coefficients of the Newton polynomial. Then
p 4 ( x) 0 0.4 x 0.2714 x( x 5) 0.0548 x( x 5)( x 7) 0.0712 x( x 5)( x 7)( x 8)
0 x (0.4 ( x 5)( 0.2714 ( x 7)(0.0548 ( x 8)0.0712)))

The second line above is called a nested form that can save computation operations and
produces a more accurate solution numerically. Note in passing how similar this table is
to Romberg quadrature (or Richardsons extrapolation), only in here the formula for
calculating the rightmost terms is a simple ratio of differences.
A MATLAB code newton_example.m is written to incorporate the divided
difference scheme and calculate the Newton interpolation polynomial using the nested
form. The algorithm is as follows
(1) Input and plot data set points;
(2) Declare c(n, n + 1) as the matrix for the table; Note: c = element in Equation
(3.2.9);
(3) c(i, 1) x; c(i, 2) y;
(4) Calculate the rest of the table due to Equation (3.2.9);
(5) Calculate the polynomial using c(i, i +1) as the coefficients of the polynomial.
Figure 3.2.1 shows the resulting polynomial for the above example.

7
Newton Interpolating Polynomial
20

data points
Newton Polynomial

15

10
5
0
-5
-10
-2

10

12

Figure 3.2.1 Newton interpolation polynomial.

Through the above example, we can see the advantages of the divided difference
table over the algebraic approach we have in Example 3.2.1. First, it has less
computational operations. We do not need to write the polynomial and then use the C0
condition to calculate the constants. Second, it is much easier to incorporate it in a
computer code.
It is important to realize that both the Lagrange and Newton polynomials are C0
continuous and each would generate the same result. One effect of C0 continuity can be
seen in the large dip that occurs between the first two data points in the figure. Should it
really be there?

3.3 Hermite Interpolation Polynomial

The Hermite interpolation accounts for the derivatives of a given function. For
example, in the case of a beam finite element, suppose we need to obtain cubic
polynomials that satisfy the following cases:
(1) Consider: y = ax3 + bx2 + cx + d in [0, 1].
(2) Apply conditions
@x=0
@x=1
Case 1:
y = 1, y = 0 y = y= 0
Case 2:
y = 0, y= 1 y = y = 0
Case 3:
y = 0, y= 0 y = 1, y= 0
Case 4:
y = 0, y = 0 y = 0, y = 1
(3) Solve each case for a, b, c, d.
We recall the standard Newton form for a cubic polynomial
y ( x) c 0 c1 ( x x1 ) c 2 ( x x1 )( x x 2 ) c3 ( x x1 )( x x 2 )( x x3 )

(3.3.1)

This clearly is not what the Newton interpolation is meant for, but we can employ it as
follows: approximate y by using the divided difference between two points, then letting
one approach the other in the limit. For example, if y(xi) is used, consider adding another
point xm. Letting these two points converge to one point, we obtain
[ xi x m ] y

y m yi
y ( xi ), as x m xi
x m xi

From this we discover that the divided difference is an approximation of a derivative.

Then in the divided difference table, we will put two entries for xi in the data set and do
not calculate xi , xi y in its original form (which will overflow numerically), but rather
simply put the y (xi) value there.
For the case mentioned above, the table would be
x1
x1
x2
x2

y1
y1
y2
y2

y1
[ x1 x 2 ] y [ x1 x1 x 2 ] y
y 2
[ x1 x 2 x 2 ] y [ x1 x1 x 2 x 2 ] y

The tables for the four cases are best determined by hand. Then substitution of the
diagonal values for the ci s and for the x i s into Equation 3.3.1 yield the polynomials.
The results are:

9
Case 1 :
0

y ( x ) 1 0 x 1x 2 2 x 2 ( x 1) 2 x 3 3 x 2 1
Case 2 :
0 0
0

y ( x ) 0 1x 1x 2 1x 2 ( x 1) x 3 2 x 2 x
Case 3 :
0

y ( x ) 0 0 x 1x 2 2 x 2 ( x 1) 2 x 3 3 x 2
Case 4 :
0 0
0

y ( x) 0 0 x 0 x 2 1x 2 ( x 1) x 3 x 2

It is not hard to modify the code newton_example.m and incorporate these

changes. The resulting code is called hermite_example.m. It can compute the above
tables. The polynomials are plotted in Figure 3.3.1.
For cases involved with higher order derivatives, the principle is the same (see
Gautschi, 1997). One thing worth noting here is that when y(n)(xi) is used, all lower
derivatives and y(xi) itself must be included in the constraints. For example, you can not
have y(xi) as a constraint but not y(xi), nor y(2)(xi) but not y(xi) and y(xi).

10

0.8

0.4

0.6

0.8
0.4
0.2
0

case 1

-0.2
0.5
x

-0.4

case 2
0

0.5
x

0.5
x

0.8

0.6

0.8

0.4

0.6

0.2
0

0.4
0.2

data points
Hermite Polynomial

0.6

case 3

-0.2

0
0

0.2

0.5
x

-0.4

case 4
0

Figure 3.3.1 Hermite interpolation.

Example 3.3.1. Constructing displacements in a beam element from Hermite polynomials

Consider the beam of length L shown in Figure 3.3.2. The Hermite polynomials are:
x
x
N 1 ( x ) 2( ) 3 3( ) 2 1
L
L

x3
x2
N 2 ( x) 2 2
x
L
L
x
x
N 3 ( x) 2( ) 3 3( ) 2
L
L
3
2
x
x
N 3 ( x) 2
L
L

(3.3.2)
(3.3.3)
(3.3.4)
(3.3.5)

These polynomials interpolation functions may be thought of as the fundamental modes

of deflection, two of which are shown in Figure 3.3.2. The deflection w( x ) of any
statically loaded beam can be written in terms of these modes as
w( x) N 1W1 N 2 1 N 3W2 N 4 2

(3.3.6)
where the subscripts associate quantities with positions (or nodes) 1 and 2 on the beam
and Wi , i , i 1, 2, are the deflection and slope, respectively, at each node.

11

x/L
Figure 3.3.2 Static bending modes N1 and N3 in a beam.

Can you physically explain why N1 + N3 = 1 for all x in Figure 3.3.2? And why N2 +
N4 1?

References
Gautschi, Walter, Numerical Analysis, Birkhauser, Boston, 1997.