. 3 : CS Ses i k woserl aulyss is te, ban of Mabe tha "sles as povlons Ust nypentncdin is i TAPTED fe By ia bcd oe CEFR | inite y t “541 INTRODUCTION The calculus of finite differences Plays an important role in Numerical methods. It deals with the Variations in a function when the independent aa variable changes by finite jumps which may be equal or unequal. In contrast, the infinitesimal calculus deals with the relationships that exist between the values assumed by the function, whenever the independent variable changes continuously in a given interval. In this chapter, we shall study the variations ina function due to the changes in the independent variable by equal intervals. - 52. FINITE DIFFERENCES Lt y=/(x) bea discrete furiction, Ifxya,+h,x,+2h,.. Successive values 4Aythen the corre in «sy: mhare the of x, where two consecutive values differ by a quantity ~~~ = ‘sponding values Of yare y,,Y.,Y---,y,. The value of the t Variable xj led the argument and the corresponding functions Variable x is usualy called the argu po al Value is known as the entry. The arguments and entries can be n in a tabular form as follows: ~ y=/GFm) Pye ae =S@,+h) =fe,+2h) ‘Scanned with CamSeannerF 5.2. Narerkal Metros Es ‘ lues of f (x) oF f(x) et determine the val ate, 8 sine the following thee types of differences ane (i) Forward differences id (ii) Backward differences and {iii Central differences | : 5.3 FORWARD DIFFERENCES we subtract from each value of y (except) the preceding vale BEY, ~ Ye Ba“) _ Fespectively, known as theft of y. These results which may be denoted Ay,, dy,,....4) ic. Ay y Yas where A iy a symbol representing an operation of forward diffe called first forward differences. Thus, the first forward diff given by : LAMENT O 12am Now, the second forward differences are defined as the dif the first differences, that is, ay, = Alay.) = AO, =») = Mar AY, = Sar Wear t Yn scalled second forward difference operator. the third forward differences are : Ay, = MS) = ary, — Aty, = ALA = AG: ~y,)- AW =O, -y)- 2 Simitarly, 7 ye ary, os "nfoction atten the Forward diferences ar *°™" rhe = fas LO) = firs rm h , d= f(x + Aye SO) fl and so 20) = £04 3h) - ye apy n Se S00. where his the interval of differencin&- ‘Scanned with CamSeannesFinite Differences — 5.3 “| 3rd sth difference | sitfereve y=xth | yo Sa) \ ay,) fed 7 I Ps Aye cd 4% neat 24| ese Ay, oy, ay, Ix,=,4 3a yet) 7 dy, ay; 74 4h] YES) Ay, i Ay y= 4+ SAL RS (x) The first ferm in the table y, is called the leading term and the differences Ay, A?y;, Wy, ..- are called leading differences. It can be seen that the a differences A'y, with a subscript ‘i’ lie along the diagonai sloping downwards; that is, forward with respect to the direction of x. The above difference table is known as Forward difference table or Diaynul difference table. Properties of A The operator ‘A’ satisfies the following propeniies (i) AUG) £ g(x) =4f (x) 4 Age). ic. Ais linear Gi) Aa f(x) = aA f(x), @ being a constant (iii) Of (x) = ANA (x) = AATF (x). where m and m are Positive integers (i) AZ (x). ¢ (ef). AE OO Observation We can express any highct order forward ditterence ot + MM terms of the entries Jy Fe: From A y,=¥,-3¥, 43%, -¥, and SO on, we can sce that the coefficients of the entries vn # "nomial coefficients. Therefore, in general. =" Yn HCY es ieee RY ‘Scanned with CamSeannes6:4 Numerical Methods Observation? We can express any value Of y in terms We know that yy ~ Yo : WOM AI (LADY, Now at AM RAY CF AYy, similarly, y, = (144)? y, and so on. In general, Hy UAB Yo "GAY NONI, bpp Of be w AY, Oa BACKWARD DIFFERENCES The differences Y,~Vy> Ya~ Vir +009 Va Ya when dented ty iy .++4 Vy, respectively, are called the first backward differences, the backward difference operator called nabla operator. me Vy, =, ~Yor Va = Va Wyre Wy YH, Now the second backward differences are defined as the di the first backward difrerences, i.c. pe Vy, = VV) = VO;-Y) = V,- VY, = Or Y- W.-W) =, - 2%, * Yo Vy, = Vy,-Vy, = ¥y-2y, + 1, and so on. Ingencral, - Ve VY Ves In function notation, these are written as VF) = f(x) -Se- A) a W(x th) =f(e+hy-f@) ’ + . Vf (e+ Qh= f(x + 2h) - Et MASH Vfix+ 3h) = f(x +3h)- 3p VE and so on, where h is the interval of differencing. 4d These bac bub following mai a inl . ‘? lies along tothe directi kward differences are arranged in a !2 nner. In this table, the difference V'y, with? the diagonal sloping upwards; that is, back™ 'on of increasing argument x. ile _ tl ‘Scanned with CamSeannerFinite Differences 5.5 3rd ath Sth a eo difference | difference| difference . Jou peath | 5 = 3 any, a, =35t 3h) yy s=x¢ 4h oy, a SH oe La Properties of V we () @)+ eG] = WG) Va(x), i.e V isa linear operator. (®°Viaf()] = a Vf(x), @being a constant, Gil) V°V" (2) = Vor" (2), m and.n being positive integers. _ &) VV) ge) # IYO). 2). Observation We can express an differences Vy,, V2y,, etc. i By definition, ¥ value of y in terms of y, and the backward ; YnYur= VY, of Yni= Yy~Vy, =(1- Vy, 66) i a Yna= Vy4~ Vy, = (l= Vy, = (1 ~— V/y; [using Eqn (5:6)) iil, Y,3 =(1- Vy, and so on. Ingeneral, Joe = (1 Vy, 6D Leyte, Wy, +4, Vy, — <9 CIP Vy, ) 5 nde sik z i°5 CENTRAL DIFFERENCES #8 times, it is More convenient to employ another’ system of differences f at Central differences. In this system the symbol 5 is used instead id is Y dite wn as central difference operator. The subscript of Sy for eten one is the average of the subscripts of the two members of the. Fe THF in WF ins W-M A SYon ‘Scanned with CamSeannerSamet 4 eihods a6 vor hygher order differences, we have , . Py. 6 Sy = Sy. By FH OHS TH FI gy My 8). and s0 on ty [he central differences are tabulated below, 6. = ragueser | y=F0) [st 2nd saa cnt, _|siterence diference | iter | dite! %e Y% 6 ‘ te a On eat Ms By, ey oy, ” 8, By, On ay, % By, 8B, Xn BY, ve By, Bra s ‘We can see from the table that central differences on the same iri linc have the same suffix. Also, all odd differences have a fraction! and the even differences have integer suffix. Note ; 1. “From all the three tables, we can see that only the notation not the differences, For examples, : Ww~W= A= A= Bin 2. If we write y=/(x)asy=f, or y=y, then the entries core © N04, xh, x+2h are Ys, Vege Yeaaare! _ respective! MEY, 44-Yy MY, = Ay, ., Ay, and $0 0n- Sinai, W.= y=, | Ms Yevun Jey ANd $0 On, y 5.6 DIFFERENCES OF POLYNOMIAL ” ces are cont a ial of nth degree, its nth differet ifferences are zero, ato “+a_x+ta, be the Po the int” re constants and a, * 0. L*t ++) G, Then, Yeas Ofe + hp + afer hp tt eel differencing. c ote Cie ‘Scanned with CamSeanner“AY, Sead 3 ‘ aoa alte + hir—w]eai(es hy et) FFG (FHA Bagh beta ce tt Rett . KI" are constants involving # but not x. Thus, the first her polynomial of the ere b,c’, « ference of a polynomial of the nth degree is anot! degree (1 - 1). Consider now > Aty,= ACA) = ay, 4, © Sa, nhl(x + hy =x!) + Bet byt — aay te + RQ AX) = a m(n—1 iP xt BY tt cl et RY The second difference, A? ,, is thus a polynomial of degree (1 - 2). | Proceeding in this manner we obtain the result i Ary, = a.m 1) (n—2)... 21x I nu! if = a constant independent of x. Therefore, the nth difference is constant. Hence, A” yO Ave | and soon “Note~the converse of above is true, That is, ifthe nth differences of a i — function tabulated at equally spaced intervals are zonstant, the function is & apolynomial of degree n. 5 5.7 FACTORIAL NOTATION OR FACTORIAL POLYNOMIAL, 2 | Consider the continued product, 4 (x-fi) (x= 2h)... = (= DA] : 4 q {Containing r factors of which x is the first one and the successive factors * are decreased hy a constant difference h. This is known as factorial | polynomial and is denoted x. 2 9 = x(x — A) (x 2h) 2. [x -(r- DA] Hence, x= x, x @= x(x - A), x =x (x= hx - 3h) and so on. » Differences of x 4x9 = hee Ay oe ai F+A)x he... 2==2)A]=3(2=h).. 5x h)... fe Date he -x- nn =x yh ; Le-(r- A ‘Scanned with CamSeannes: 2TESO 2302) 15= 04 77 Again, from Observation 2 of Section 5.3, we have 215 = GplSt= 665+ 17459 + 1865) + 60)°7 “ siyosill frample§2 Evaluate (i) A cosx, (ii) A log /«:). (iil) sil A tan, and (vy Arent Solute? Let hbe the interval of differencing. i) cose: ES c0s(x + h) — cose * 2 sin» +) sia! 2. 2 ‘Scanned with CamSeannesce Ente Ditorerene 5:14 Linicss otherwise stated, the interval of differencing mss oe taker cate Uni Note ‘Tas! = . rae 58 Von! the second difference rn w= V2e + 4254 20x + sath nterval of differencing a Soliton sytheticdivision method : Dividing the given polynomial successi ely 2, x-4 and x-6, we have bye? S() =< Tae + 428 30" +9 In factorial notation LO) = x0 — 20+ 14a +9 F00) = 4D + 1431 (using Av = rh x) AY (2)= 42) 3x 2202" = 48x (x-2)-16 = 48 - 96x- 16 5. 10 ERROR PROPAGATION IN DIFFERENCE TABLE lery te ‘ys. ..¥ he the true values of a function and suppose in the value Shows i a0 error c. Then the effect of this error in the differences is 9 TiN the table piven the next page. 7 ‘Scanned with CamSeannesservation 5.16 . Numerical Methods , 3 ie y ay Ay x, ‘ wy x, n a Ay . x ys ay, Ay, x, % ay, 4, b- x, YW A-A | ayyte A-- 7 [ayes a yer Ay, 2] a ~~ | ee . yO. [Ate ‘ & Po x, ; ay, ay, A ay, oy, oa ee This table shows that 4 {| the effect of an error increases with the order of differen Gi) the errors in any column are given by the binomial coeff 3 (I~ ey. Thus, in the third cotamn the errors are ee 2nd in the fourth column the errors are 6 Ae, 66, Ae &1 (iil) the algebraic sum of the errors in any difference columnisa® (iv) “the maximum error in each column exists opposite 19 containing the error i, Ve : le. These facts enable us to detect errors ina difference tad : i - The sum of the entries in any’column of die Citference between the lat entry and the first entry in the pre Consider the sum at ihe entries in the third column: OW ty, 4 At by.t yt Ay, Ay, ai) = (Ay Ay.) 4( y= Ay.) + (ay —ady 20 OED = by, ay, —° ” ‘Scanned with CamSeannerNumerical Methods Solution Since the values of y belongs toa polynomial of degree fou, be differences must be constant, Then from the table, Hes = AY, SM =F ONE i Hence the missing entries corresponding to x= 12, 14 cre and 5.11 OTHER DIFFERENCE OPERATORS 4 6 Now westll So far we ave studied i operato: fal 7 < ied the s 4, lay a vital role! other operators like i methods, Z0£, # Dete,-which also! Shitt Operator E “fhis the inte a 3 - athe al i i i ment x then defined as of differencing in the argu Ep@)=sorh. ‘Scanned with CamSeannesFinite erences 5.19 called sranslation operator due to the reason that it results the ve of the function. E2f(x) = LES (3)) = ES (x + Y= f(x + 2h) similarly, Ef) = $+ 3h), E*f (x)= f(x 4h) and En f(xy= fi (xt mh). ify isthe function f(x) then Ey, = y,,,.--and Ey, =y, ,,, the inverse operator E's defined as Iris extol E“f()= fx-h) Ao? Similarly, E-f(x)= f(x—nh). te # In general, t 1 E*f(x)= f(x+ mh) for any realn. (5.11) Notet If yay -- y, are the consecutive values of y, then EV = WE My SEL, Yay” ‘Averaging Operator ys _ The averaging operator jis defined by : HS (x)= Af (x + W12) + f(x h12)) BY = ALY cam * Yacnal 6.12) {| Diterenta! Operator D ie, * diferential operator D is defined as D f(x) = fy ) ‘Beneral, a Df (x)= a F(x) (5.13) M Operator 4 an a OPerator:1 is defined as 13f(x) = f(x) * All the above operators are linear and obey index laws. ton between A and E H AS =f(x4h)-f@) = Ef(s) -Sx) = (E- DF) | eg A=E-lorE=1+4 ‘Scanned with CamSeanner5.20 Numerical Methods Relation between & and v Of (x) =S)-S (A) =f®-E"f()=(1- gE =v E1-EtorE1=1-V Ey Teea-y CEMA Relation between £ and & Bf) =feet hI2-SO- hi2) 7 =E*f()-E"f@) =(E%-E*S@) 6 =E*-E% Again, 6= £%(1-E")=E"V and 5 = E-*(E-1)=E"A 16 =EYV=E“V : Relation between & and yw - Use) = Lf + Ad) + fe — H2)) =%[E*S@tE*S@)] ~ “ =%(E*+ E*) f(x) : HO =A(E*+E*) “Relation “of D- with other: Operators Weknow tat D/G)= Lf) =f"aete. By Taylor's series : “a h 7 Peay) + Loh) =f) TF) + E prey + BIO ic or Ef(s)=f0)+ WD fe) + * py) +E vp 2p? 4p? =[Ptap ey PP 4s ~ 2 3! a4 ‘Scanned with CamSeannes