You are on page 1of 21

# UNIT 11 INTERPOLATION AT EQUALLY .

SPACED POINTS
Structure

11.1 Introduction
Objectives

11.2 Differences
11.2.1 Forward Differences
11.2.2 Backward Differences
1 1.2.3 Central Differences

11.3 Diffcrcnce Formulas
11.3.1 Newton's Forward Dilference Formula
11.3.2 Newton's Backward Difference Formula
11.3.3 Stirling's Central Difference Formula

11.4 Summary

1 1 . INTRODUCTION
Suppose that y is a function of x. The cxact functional relation y = f(x) bctwccn x and y
may or may not be known. But, the valucs of y at (n + 1) equally spaccd valucs of x are
supposed to bc known, i.e., (xi, y;); i = 0. .... n are known whcre xi- xi-, = h (fixcd).
i = 1.2. .... n. Suppose lhat we arc rcquircd lo dctcrminc an approximale value of ~ ( x )
or its dcrivativc f'(x) for some valucs of x in thc intcrval of intcrcst. Thc mctliods for
solving such problclns are bascd on thc conccpl of finite di~fcrcnccs.Wc havc
introduccd thc conccpt of forward, backward and ccntral dilTcrcnccs and discussed thcir
intcrrclalionship in Scc. 1I .2.

We havc alrcady introduced two important forms of thc interpolating polynomi31 in Units
9 and 10. Thcsc forms simplify whcn thc nodcs arc cquidishnt. For the case of equidistant
nodcs, wc have derivcd thc Ncwton's forward, backward diffcrcncc forms ant1 Stirling's
central diffcrcnce form of intcrpolating polynomial, each suitable for usc undcr a spccific
situation. Wc have dcrivcd thcse mcthods in Scc. 11.3, and also givcn the corresponding
error tcrm.

Objectives
After reading this unit, you should be able to

write a forward diffcrcnce in terms of function valucs fro~na table of forward
differences and locate a diffcrcnce of givcn order 31 a givcn poinl;

write a backward difference in terms of funclion valucs Crom a table of backward
differences and idcnlify differences of various orders at any givcn point from the
table;

expand a ccnual diflercncc in terms of funclion valucs and form a mble of c c n ~ a l
diffcrcnccs;
establish rclations between A, V ,6 arid divided diffcrcnce;
obtain the intcrpolating polynomial of f(x) for a givcn data by applying any one of
the inlerpolaling formulas;

compute f(x) approximately when x lies near the beginning of the table and estimate Inter. Interpolation at Equally
Spaccd Point
the error;
compute f(x) approximately when x lies near the end of the table and estimate the
error;
estimate the value of f(x) when x lies near the middle of the table and estimate the
error.

11.2 DIFFERENCES

Suppose that we are given a table of values (x,, yJ, i = 0 , 1.2, ...,N where yi = f(x9 = fi.
Let the nodal points be equidistant That is

For simplicity we introduce a linear change of variables

-
S = s(x) = -
h
X Xo
, sothat x = x(s) = xo + sh

and introduce the notation

The linear change of variables in Eqn. (2) transforms polynomials of degree n in x into
polynomials of degree n is s. We have already introduced the divided-difference table to
calculate a polynomial of degree < n which interpolates f(x) at xo, xl, ..., x,. For equally
spaced nodes, we shall deal with three types of differences, namely, forward, backward
and central and discuss their representation in the form of a table. We shall also derive the
relationship of these differences with divided differences and their interrelationship.

11.2.1 Forward Differences
We denote the forward diffcrenccs of f(x) of ith order at x = xo + sh by Aif, and define it
as follows:

\

Where A denotes forward difference operator.

When s = k, that is, x = xkrwe have

for i = 1 Afk = fk+ I - fk

for i = 2 A2fk.= Afk+ I- Afk

=fk+z-fk+l - [fk+l - fk]

T
= f k + ~ - ~ ~ +
~ +fk l

Similarly A3fk = fk+3- 3fk+2 + 3fk+, - fk.

We recall the binomial theorem

whcre s is a real and non-negative integer.

..b] and i times differentiable in ]a. . For i = 0.. This relation will be utilized to derive the Newton's forward-difference formula which .. then AnPn(xo)= a. 1. interpolates f(x) at xk + ih. Assuming that relation (5) holds for i = n 2 0.AnPn(xo) = hnn!a.i.q.. Hence (5) is proved. (6) and (3. n.hnn!a. i = 0.we get = hnn!a.= 0. We now give a result which immediately follows from this theorem in the following corollary.n ! hn arid An+'pn(xo)= 0... we have for i = n + 1 This shows that relation (5) holds for i = n + 1 also.. that is. both sides of relation (5) are same by convention.. This completes the proof The shift operator E is defined as Ef.b[ and 4 E la&[.and xo is an arbitrary point.We give below in Lemma 1 the relationship between the forward and divided differences.. all higher differences are zero. Since An ''pn(xo)= AnPn(xl). Taking i = n and f(x) = Pn(x) in Eqns. = fi + 1 In general Ef(x) = f(x + h). Corollary: If Pn(x) is a polynomial of degree n with leading coefficient an. Lemma I: For all i 2 0 Prook We prove the result by induction. Proof: Taking k = 0 in relation (5) we have Let us recall that where f(x) is a real-valued function defined o i [a.

fi = Efi . we get We now give in Table 1. Using (10)..) j=O ~j (-1l8-J w i n g use of relation (8) in Eqn.fi = (E .. (9).. x..We have Esfi = fi +.. Table 1:Forward.In Now. we get A'= (E. respectively.. Afi = fi +.f.Difference Table Note that the forward difference Akfolie on a straight line sloping downward to the right. ~~f~= fi+. we have for . XI. fl ..2.. be fo.. ~~~f~ = fi+ and E-IRfi r fi. Let the values of f(x) at n + 1 equally spaced points xo. 11.2 Backward Differences Let f be a real-valued function of x. the forward diffcrences of various orders using 5 val. Operating s times..I)'= k(..Thcy are defined as follows: where v denotes backward difference operator. For example. The backward differences of f(x) of ith order at xk = xo + kh are denoted by vlfk.l)fi Hence the shift and forward difference operations are related by A=E-l '. or E = 1 +A.s.

.

2fk + fk-1 Similarly. Also note that the coefficients of 6'fk are the same as those of the - binomial expansion of (1 x)'.2. s = 1. Notice that the even order differences at a tabular value xk are expressed in terms of tabular values o f f and odd order differences at a tabular value xk are expressed in terms of non . 11..'. / f and Z4fk= fk+ 2.2.Example 1: Evaluate the differences Intcrpolatlon at Equally Spaced Point Solution: (a) V3 [a2x2 + alx + a. is defined as Operating with 6. a r k + l. . The second order central difference is given by S2fk = =[fk + 112 ... Let us now discuss about the central differences.-. + fk. -: For example. denoted by 6fk..3.1/21 ' = f k + l-fk .3 Central Differences The first order central difference off at xk.fk + 1/11 = &[fir + 1/21 -8[fk .1. Also note that Afk = v fk + 1 = fk+ 1 fk. we obtain the higher order central differences as with Ffk = fk when s = 0. Since Wc have the operation relation The central differences at a non-tabular point xk+ In can be calculated in a similar way. - Try to show that ~~f~= v4f4.tabular value off.fk + fk-1 = f k + .z = fk + 1:.2.] = 0 Note that the backward differences vkf4lie on a straight line sloping upward to the right. fk ..4fk+ 1 + 6fk .4fk. .

.Interpolation S2fk+1n= f k + ~-2fk+ln n + fk-I. 63fk+lR= f k + 2 '3fk+1 + 3fk-fk-1 - S4fk+ln= f k + ~ n 4fk+3n + 6fk+ln-4fk-ln + fk-3n Relation (15) can be obtained easily by using the relation (14) We have The following formulas can also be established: \ We now give below the central difference table with 5 nodes. Table 3 :Central Difference Table Note that the differences S2"'fo lie on a horizon&l line shown by the dotted lines.

6 and p in terms of the operator E as follows A =E-1 Also E" =p 6 + - 2 E-'R= 6 P. V . (c) Express 62f2as a forward difference. We now define the mean operator p as follows 1 = 7 [E'" + E . gand p We have expressed A.7 Example 2: (a) Express A3f1as a backward difference. .I I ~ ] ~ ~ . (b) Express A3f1as a central difference. Spaced Point Note Lhat Lhe differences 62mf2lie on a horizontal line. Hence p = 1 [Ell2 + E-'I2] Relation Between the Operators A+ v.

H. .' ) ~ 4 El) Express v4f5in terms of function values. (E .l )+ (1-E-I)] = T ( ~ .E-112) = 7(E . the result. we have derived Newton's form of interpolating polynomial (using divided ." ' Sf2] = E .E-1) .~ .Intcrpolatiun Solution: (a) a3f1= (Evl3f1= E3v3f1= v ~ =E v3f4 ~ ~ ~(A=EV) (e) 2 S ~ ~ ~ = [ E .S. .' n )(E'R-E-'~) = -(E-E-') p6 = 2 2 R.E .' ) ~ -.2) 11.3 DIFFERENCE FORMULAS We shall now derive different difference formulas using the results obtained in the preceeding section (Section 11.1)p. E2) Show that (E + 1)6 = 2(E .' ) .' A ~ ~ ~ = A ~ E .H.E4.' ~ ~ =( SA=' E~-~' ~ A ) Example 3: Prove that (a) p2= 1 a2 + . 1 + 1 ~ . - 11. 4 Solution: (a) 1 [El" We have p = - 2 + E-"'1 = I + -S2 4 (b) L..' ) ~+ 4 -.3.S. 1 1 1 -(A 2 + V) = T [ ( ~ .1 Newton's Forward-Differcnce Formula ' In Unit 10. (c) We have 1 p = Z (Ell2 + E-112) (Ell2 . (E + E .1+p2l2=1+ (E . Hence.

We have also established in Sec.... (24).(25) of degree < n.k Anfk. is Substituting (22).=x0 + sh-[xo + (k+j)h]=(s-k-j)h Hence (23) can be rewritter! as (S . the following relationship between Interpolation at Equally Spaced Point divided differences and forward differences axk. we have X-xk+.. we obtain Setting k = 0.xk +. Setting k = 0 in (25) we get the formula The form (23).. Hence obtain the value of f(x) at x = 1... d*.k) (S . (25) or (26) is called h e Newton's forward-difference formula.k ) ( Sn! . . .I = n!h" a r k . xk + 1.we have the form Using the transformation (2).+ (S . 11. The Newton's form of interpolating polynomial interpolating at xk.I) ~ 2 9 . ?+ I The error term is now given by Example 4: Find the Newton's forward-difference interpolating polynomial which agrees with the table of values given below. (22) Substituting the divided differences in terms of the forward differences in the Newton's form.k . differences).1. and simplifying we get Newton's forward-difference form..2. .1) = fk + - (S k)Afk + 2! ~ .-.5. .~k + .

5)3 + z(1. and . we can use the formula (26). and the interpolation polynomial is not needed... as. Taking xo = 1 and the subsequent values of x as XI. The step length in the data is h = 1. f(l.5) + 7 = 3. Table 5 :Forward Differences Since the third order differences are constant.375 Note: If we want only the value of f(1.5)..Solution: We form a table of forward differences of f(x).5) s (l.375 + 3 + 7 = 13.x2. becomes which on simplification gives f(x) = x3 + 2x + 7 :. the Newton's forward-differencesinterpolation pdlynomial. . the higher order differences vanish and we can infer that f(x) is a polynomial of degree 3 and the Newton's forward-differences in~rpolationpolynomial exactly represents f(x) and is not an approximation to f(x). In this case.

125.4453 = 47. Estimate the population for the year 19f 5.1. Population: y 46 66 81 93 101 (in thousands) .8672 = 48 r . .5) (37) = 31 + 21.5625. we form a cumulative frequency table. 5 ) 24(-1.1.5) (-2. find the number of students who obtained less than 45 marks. - Example 5: From the following table. Table 6 :Frequency Table We have xo = 40. x = 45 and h = 10 + (0. In other words. Solution: We form a table'of the number of students f(x) whose marks are less han x. Inierpo~ntionnl Equally Spared Point E3) ' The population of a town in I@ decinnial census was given below.5) ( 4 .. The rlumber of students who obtained less than 45 darks is approximately 4% .1.

find the value of y (0.48 E10) The area A of a circle of diameter d is given in the following table.10 0. Find the value of tan (0. 10 and Rs. E12) The following table gives the'amount of a themical dissolved in water at dirrcrcnt temperatures. y(2) = 1 and y(3) = 10. ' El 1) In an examination.1511 0.891 I I Compute the amount dissolved at 8'.1003 0.97 21. Hence obtain the increase in population during the period 1915 and 1918.72 19. the number of candidates who secured marks in certain limits werc as follows: Marks 0-9 20 . x 0 .3. 14.30 tan x 0. Calculate the population in the year 1915 and 1918.12).2027 0.81 17. Tcmperature lo0 15' 20° 25" 30" 35" I Amount dissolved 19. x . Year x 1911 1921 1931 1941 1951 1961 Population y 12 15 20 . of men y 9 30 35 42 E9) The following table shows the monthly premiums to be paid to a company at different ages.27 15. 15 from the following table.51 22.10 10-20 20-30 30-40 No.E4) From the following table.2553 0.1 S x 5 0. Find the area of the circle when the diameter is 82 units.23): E5) Find the cubic polynomial which approximate y(x) given that y(0) = 1. 0.47 23. 27 39 52 (in thousands) E8) Find the number of men getting wages between Rs.52 24.15 0. Find the premium to be paid at the age of 26 years. y(1) = 0.20 0. Wages in Rs.3093 E7) The following table gives the population of a town in ten consecutive censuses.25 0.39 40-5P 60-79 80-991 I No. Age 20 24 28 32 36 Premium in Rs. of candidates 41 62 65 50 171 Find the number of candidates whose marks are 25 or less.96 22. .65 25. E6) The following table gives the values of tan x for 0.

l.] + (x-x.2 Newton's Backward-Difference Formula Reordering the interpolating nodes as x.)f[x.x) .q = x.(x ..l + (x-x. then x .] (27) We may also write Set x = x.xo and applying the Newton's divided difference form. x.i)h x-x.x. + (x . + sh ..x..... we get P.(x)=f[x.[x.x. at Equally 111terpolatlo11 E13) Find a polynomial which fits ihe following data : Spaced Poirlt 11..-l)f[xn-2*~n-1r~nI + .-j=(s+n-n+j)h=(s+j)h and Equation (28) becomes We have seen already that Hence. equation (29) can be written as ..(n--i)h]= (s + n . ..)(x-x. ....3. + sh.-l..) f[x0.

If we name the abscissas as xg. With h = 1. XS.. ..then x. the Newton's backward differences polynomial for the given data is given by (x-6) (x-5) (x-6) (x-5) (x-4) = 235 + (X-6) (93) +7 (30) + 6 (6) =235 + 93(x-6) + 15(x-6) + (x-4)(~-5)(x-6) which on simplification gives P(x) = x3 + 2x + 7. = xs = 6. Table 7 :Backward Difference Table Tables 5 and 7 are the same except that we consider the differences of Table 7 as . fn = fS = 235. I . Solution: We form the table of backward differences of f(x). In this case error is given by The backwatd-difference form is suitable for approximating the value of the function at x that lies towards the end of the table.Intcrpolatlon Equation (27). Example 6 Find the Newton's backward differences interpolating polynomial for the data of Example 4.. XI. backward differences. which is the same as the Newton's forward differences interpolation polynomial in Example 4. (28) or (29) is called the Newton's backward-difference fonn.

Interpolation at Equally Spaccd Poinr .

. The Stirling's interpolation is used for calculation when x lies 1 1 between xo .h and xo + .rh. This formula is used whenever interpolation is required of x near the middle of the table of values.3. then the same formula is used deleting the last term. For the central difference formulas.. Stirling's formula for interpolation is given by where s = (x .xo)/h and if n = 2p + 1 is odd.3 Stirling's Central Difference Form A number of central difference formulas are available which can be used according to a situation to maximum advantage. is chosen near the point being approximated and points below xo are labelled as xl. the origin xo. find the value of f(1.and those directly above as . Example 8: Using Stirling's formula.xe2. IT n = 2p is even.11. It may be noted from the Table 3. Even order differences at xo are those which lie along the horizontal line through xo. . that the odd order differences at x-]n are those which lie along the horizontal line between xo and x. Using this convention. But we shall consider only one such method known as Stirling's method.. (as in Table 3).. x. x2. .Sirnilafly..32) from the following table of values. Solution: Table 9: Central Difference . the odd order differences at xln arc those which lie along the horizontal line between xo and xl.

2.3 = o.. 3! 2 [s3f-.+-. In the following exercises.22) from the following table.] = T(0.1889 + 0.15) from the table.2059) = 0.7382. we get = 1.t+sfl. use the Stirling's interpolation formula.73816 = 1. + s3f112]+ s (s4!-1 ) 5df0. we have s2 s(s2-12) 1 2 2 2 f ( x ) = f 0 + \$ [sf~l. E21) Find f(l.. s = - (x-xo) = h 1.32-1. E24) Find the value of y when x = 30 from the table. Now.. - E23) Evaluate f(4. (32).1974 Substituting in the aboveequation.2]+z S2f.325) from the following. E22) Find the value of f(1. . 1 1 7 + ~f.1 From Eqn.725) from the following table.. E25) Find the approximate value of y (2. . I at Equally Spaced Point :. 0..

4 SUMMARY In this unit. (X .x0)/h 3. we have derived interpolation formulas for data wilh equally spaced values of the argument.xo)/h. Thus.Em)p = 2EYr8. The interpolation polynomial which fits a given data can be determined by using any of the formulas derived in this section which will be unique whatever be the interpolation formula that is used.I vkfn where s = (x . Newton's backward difference formula: Pn(x) = Pn(xn + sh) = f: (-'I)~ k -0 [-. If n = 2p is even. However. Table 10 . The interpolation formulas derived in this unit are listed below: 1 Newton's forward difference formula: where s . 2. We have seen how to find the value of f(x) for a given value of x by applying an appropriate interpolation formula derived in this section. (12) v4f. the formulas derived in this unit can only be applied to data with equally spaced arguments whereas the formulas derived in Units 9 and 10 can be applied for data with equally spaced or unequally spaced arguments. Stirling's central difference formula: if n = 2p + 1 is odd. the same formula is used deleting the last term.4 f + 6f3 . 11.Interpolallon 11.5 SOLUTIONSIANSWERS El) From Eqn. E 3) The forward differences table is given below.= f5 .vz)= Em 2pS = 2 ~ *pS LHS = E ' (Em RHS = ~E"(E* . The application of the formulas derived in this section is easier when compared to the application of the formulas derived in Units 9 ahd 10.4f2 + fl E2) ~ + E. the formulas derived in Units 9 and 10 are of a more general nature than those of Unit 11.

4241 E17) Population in 1954 is 43.+2 1~ 2 E16) 0.0. E19) Hint: The numbcr of candidates f(x) whose marks are less than or equal to x is as . E7) 12. Hence the increase in population is approximately 5.7081 E15) ~ ~ .2662. follows: (i) Take 79 as origin and determine f(70) We gct f(70) = 199.54 thousands.l thousands E8) 15 EY) 16.48 thousands. 'Taking xo = 1911. E20) 2x2 .33 thousands and [he population in 1958 is 48.64 thousands.85 thousands.25 Rs. 13. 1.79 (Hint: Take all differences into consideration) E13) 2x2-7x + 9 E14) 1. wc gct at Equally ~nterpo~aAon Spaced Point S = 1051-1911 10 = o.7x +9 . h = 10.4 = 54. E10) 5281 E l l ) \$8 E12) 18. x = 1915.81. (ii) Take 99 as origin and obtain f(89) = 232.8528 or y(1915) = 54.

1 Table 11 .725-1.7 = o.25 0. S = 1.:.