6 Finite Differences 1. Introduction. 2, Finite differences. 3. Differences of a polynomial. - Factorial Notation. 5. Effect of an error ona difference table. 6. Other ‘tence operators. 7, Relations between the operators. 8, Application to summation of series. Bs 6.1. Introduction, The calculus of finite differences deals with © changes that tah variable), due to fini 6.2. Finite differences. Suppose that the function Y = fix) is tabulated for the equally spaced values x = 0, Xo + hy Xo + Why sey x9 + nh giving y = YO> V1, Y 25 «nee » Yn- To determine the values of f(x) or f(x) for some intermediate values of x, the following three types of differen- ces are found useful : (1) Forward differences. The differences 1 — yo, yo — 1, vases Yn — Yn-1 when denoted by Ayo, A eerere » Ayn_1 respectively are called the first forward differences where A is the forward difference operater. “Thus the first forward differences are AY, = Irs —Jp- Similarly the second forward differences are defined by Ayp = Ayr — Aye In general, APy, = AP-4y,,, — AP-ly, defines the pth forward differences. i at These differences are systematically set out as follows in wh is calleda 152FINITE DIFFERENCES 153 Forward Difference Table Value Ist rd tth Sth ofy aiff diff diff diff Yo Ay xp th yr A*v9 AY Vo 9 + 2h Ya ay, aty Ay a’y, s*y¥9 xp + 3h Ys A452 ty, Ays A*y2 xp+4h ye A’y3 Aye Xo+ 5h ys Ina difference table, x is called the argument and y the fine’ tion or the entry. yo, the first entry is called the leading term and Ayo. A*yo, ASyo ete. are called the leading differences. Obs. 1. Any higher onder forward difference car. be expressed in terms of the enteries. Wehave A°¥o = Ay: — Avo = 2-1) - 2-30) '2— 21 +Yo 91 —A'¥0 = Ws—By2 +93) — 2-2; +0)N ENGINEERING, NUMERICAL METHODS | Wb Sey, Nt \iflerences are exhibited in the following These differe rd Difference Table Velue Ist ond oi ut Sth a diff diff, diff. diff, . Yo Vy1 c ahh vu we P 4 ye ws I xo + 2h Je or 3 es ys ‘ v4 P V5 xo + Sh y3 Viy4 , We Vya V5 Xo + 4h v4 v'ys Vy5 Xo + 5h ys = dy Yo- yy = dya/2, igher o} rder central differences are defined as 52 - Oy = yy, dys — ae = yp, .. oy, 8), = ®y5p and so on. These differences are shown in the following Central Differ Similarly, h ence Table a —eteheg Tables _ sap Value Value 154 2nd ang th 5th ofx i diff diff diff. Hes aig AES a dy ip xoth ” oy, Ovyy Boyar 3/2 Xp + 2h Je d4yo r 52 ye Syste Xy + 3h Js dys Ying Yap Xo + 4h M4 Oyo Xo + 5hfINITE DIFFERENCES 155 We see from this table that the norizontal line have the same suffix ij ‘alues of the . only integral values of the sulfix, C ~ Tt is often required to fi same column of difforences, We Central difference Also the diffe sultix On the game fences of odd op< are known only for halt y And those of, en order ad the mean of adjacent values in the denote (hi Ms ean hy yy Tho l 2 5 HTS 5 OV & Oyg 2D) Wy yy 1p, + 84yy) ote Obs. The reader should note that it is only the nol and not the differences, e. g Hon which changes Yi—Jo = Vy = Ay, = Of all the formulae, those involving centr; al practice as the coefficients in such formul Example 6.1. Healuate (i) (iit) A 7 /cos ¥ HV) A® eas ERG, differences © most useful in eg ese much inore rapidly, Aba! vii) \ (0° log 2x) (P.TU., B. Tech., 2001) @ Atany = tan! th)—tan ty ancl [ez eget eae = tay (ir pele tan” ee ul) A (e* log 2x) ek eM lop 2h) log 2x: = evil log * =e [eto fist (Vlog) ou (A 48) | el) Z| | ~suttl a ee Py osl ow (ne log 2 & +h) — eM log ax 4 yr log 2x ~ &* log 2x xh + (6°) log 2x {eos 2 (x + /) ~ cos pe ‘cos 2 (x +h):- A cos 2 = [cos 2(v + 2h) ~ cos ee + AIL eos Qin + si e =—2sin (2x + 8h) sinh + 2 sin (2v +h) sink 2 sin h [sin (2x + 8h) ~ sin Qe 1 hd) we (Ae \ ca -@%, »(Aeh) yanne WCE AL METHODS INE inner 1S6 NUMERE g ain h [2 co" (ww + 2h) eit ni gain? h cos (r+ ah) ox 12 pe cab’) Exam 2 poaluate ()/ alg Spee ya ponene': being unily wre gx +12 \eA? Bele pLyene es 3) | ple 6: ee uy, B Tee h., 2001 §) s ——_- ) (x + 4)e + 5) (w+ 3) +4] i 6 : fi = Ga De + +3e@+4) *@+ Bye + 4) + 5) “ A pe FIG) ha = GQ) + a+ Aya + 5)” . Gi) A(ab*) = 2 AO") A) ( oa : ~albt— bY) = ab" - Y) a2 (ab?) = ALA) =a(b- 1) Ae’) pt_w) = ab - 1b Ee sy Ae -1)¢ (ui) pet = eto = (= Det A2e* = A(Ag*) = ait e-Ve'l 2(e-1) Ae"=(e- 1) (e-1 f - ig pee ’ imilarly Me = Blin 4 and Nie = .. wie det = (0 - Ihe ( 6.3, Differ polynomial ofthe t thai of a polynornial. The nth differe" p ee eeareconstant and all higher order "hi ie * aré zero. a ae—_ DIEFERENCES: pit ve Let the polynomial me i NN degre ny, be aeelaitien Lan thee Mv) = fie +A) Ay) allv + Ay Ph I Olive pyre vl, thh , na , + by ox ' th’ a | (1) + & are the new constant co. efficients the first differences ofa polynomial of the nth degree is a ee (m2 — 1), Ax) = AlAx + h) —fix)| = ARE +h) — any Sanhl(e shyt _ yey Bila + Ayr? _ xray ee r th'h Boe eas bees, Crea +h", by (1) - Thesecond differences re] Present a polynomial of degree (n ~ 2), Continuing this Process, for the nth differences we get ap mial of degree zero i.e. APA) =antn—Dr—2).1 42 > L424 olyno- ~ =an'h® +(2) which is a constant. Hence the (n + 1th and higher differences of a polynomial of nth degree will be zero- - ‘nomial of degree n. This fact is important in numerical ani les us to approximate a function by er differences become nearly constant, ane 6.3. Evaluate AL ~ ax\C1 — bx2)(2 — 043)(4 ox9)(1 — det) alysis as it @ polynomial of nth degree, if its nth (Madras B.E., 1996 S) AME — ax)(1 —bx2)(1 — ex*)(1- dx) = AMabed x34 ()494 (x84. 4 u = abed Al (19) Co Ae") = 0 forn <10 = abed (10!) ‘[by (2) above. / : Problems KS Oe + 1. Write forward difference table if oe 20 30 40 yt y: Li 2.0 44 1.9 (Gulbarga B.B., 1993) — iilVT_NUMERICAL METHODS IN ENGINEERING AND SCIEN¢ 1 158 2. Cons ruct the (able of differences for the data below : 0 1 2 3 4 fo) ey 15 2.2 3.1 46 Evaluate A® fi2). Ifug = 3, uy = 12, up = 81, Ug = 2000, t44 = 100, calculate Aty,, (Kerala B. Tech., 1990) 4, Show that Ay; = yiga— 8yi42 + Bad Jie (Madras B. Tech., 1996) 5. Form the table of backward differences of the function fio) = x9 - 8x? = 5x — 7 forx = - 1, 0, 1, 2, 8, 4, 5. 6. Forma table of differences for the function Qw) = x8 + 5x - 7 fora =—1, 0,1, 2,3, 4, 5. Contii 1¢ the table to obtain f(6). 7. Extend the following table to two more terms on either side by constructing the difference table: x: —0.2 0.0 0.2 0.4 0.6 0.8 1.0 ye 2.6 3.0 SA omen (OS aay 14-2 0108 8. Show that yn (hale Aes OEE) “Toots + 1) # yo Afx) (ii) A log f() = vo | 1+ 7M 9. Evaluate (taking interval of differencing as unity) : @ Ate + cos) Gi) A tan? ( aot ) (iit) A (e* log 2x) (iv) AQ*/x !) ™ WA { aa | (Modras, B.E., 2001) vi) a? ( ae ) P2U,, B.Tech 10. Evaluate: (i) A" (09) Gye is (Madras B (iit) A” [sin (ax +b). 11. Prove that (i) yg = 32 + ANd + A%yG 4 A% 9 (S. GuySl” FINITE DIFFERENCES 159 (i) Vg = yg v7 +6 o*y5 = 6 = yrs + Yq (Calicut B. Ton I, 10916 uate: @ Atta -aya 2x)(1 ~ 3x) (1 = 42), (h = 1), (Madras, B.b., 2001) 10 2, GY) AIA <2) ~ 242)(4 _ 3x°)(1 ~ 4x4), if the interval of differ. encing is 2. 13. If u(x) and v(y) be two functions of x, prove that © A [u@)v(Ge)] = U(x) Au(x) + oe 4 Dj Aux) = Ue + 1) Av(x) + U(x) Au(x) (Madras, B.E., 1996) (i) a 2) _ v@) A u(x) — U(x) A v(x) €2) V(X). vole + 1) 6.4. (1) Factorial n, Otation. 4 Product of the form x (x— DL (&-2)..... (e-r+ lis den oted by [x] and is called a factorial. In particular [x] = x, [x]? = x @-D, bel? = xe —~ 1)@— 2) ete. In general [x]? = x(__ D@- 2)... (<—n +1) In case, the interval of differencing is h, then GP =a A) - 2h) The factorial notation is differences. It helps in finding t directly by simple rule of differentiation. Similarly given any difference of a function in factorial notation, we can find the corresponding function by simple integration, he result of differencing [xf is analogous to that of differentiat- ing x’. (2) To show that A” [x]" =n! and a™#l [x}?=0 We have Alx]" = [x +h]”—[x]" es PE TM the +h—2h)..... @+h-aaih) = x0 — A) - 2h) ..... (@—n- 1h) =x@—h)...(«—n-—2h) [x+h-(x-—nh+h)] = nh{xy"1 rd) Similarly A’{x]" = A(vh x]") = nh Alxy160 NUMERICAL METHONIS 166 ESOC Replacing n by n ~ I im (i), WE ges, ; be = nh in - Waist? =n SWOT Proceeding in thie way, we obtaure a {cf" = nin - 2s BEY? = 1600 — 3) ve 2 nin 1). 5 =n" Also S* tcf" =n th nth =0 In particular, when hk = 1, we have Sixl* = niet! ad BY =e Sb Thus we arrive at the [lowing maporizant contusion Tz result of differencing (c! is analogous to that f OLA OLLILZ Was hol Obs. Every colpnomial of degree 2 om 2 A DB 2A, polynomial of the same degree ant viz te Example64. Exprasy =i - +z Wafer ion and hence show that My = 12. Crdivagli, £2909) First method : Let 7 =Ai + Bie +OA2D Ee Using the method of synthetic dovicion tp. $8), we dove z-1x«-2, ce. sseasindy Then “ bs @y =6xBA+6 ty = 12, whieh Ay ate cmatath, wide —ee eeeOrrrEE ee FANTTE DIBEERENCES . ains unchanged while ation Second method (Direct method) Let y + dv= 10 Aviv = Div = 2) + Bele = 1) + Cx + D Putting = 0,- 10 =D. Putting = 12 Cc 8+3-10=C4+D 8-D=-8+10=- Putting 2, 16-12 + 6-10 =2B+2C+D B=3©2C-D)=14+ 10)=3. Hence y = ve - D@-2)4 S8x@e— 1) + 2x- 10 = 2ix]8 + Six? + 2b] — 10 Ay =2x SikP 4 3 x Qe] 4.2 Ay =6 x 2k] +6 =12. Example 6.5. Obtain the function whose first difference is, 9x? + Lx + 5. Let fix) be the required function so that Aftx) = 9x? + 11x + 5. Let 9x? + Llx +5 = 9b}? + AR] +B ta) ae = Ox(x—1) + Ax+B Puttingx = 0,B=5. Putting x = 1, A = 20. a Aftx) = 9[xl? + 20[x] + 5 Integrating, we get fay-9 bE, 2082. stl +0 78 — Dor=2)+ ele 1) + Sr +6 = 8x9 4x24 240 where c is the constant of integration. Example 6.6, Find the missing values it in the following table: 3 45 50 55 60 65 3 30 — 20 x — -24 (nw, BAR001) »162 NUMERICAL METHODS IN ENGINEERING AND Scieyey 62 The difference table is as follows : x y Ay A’y b’y 45 yo=3 y-3 50 vn 5-29; 2-1 3y1+¥32-9 55 yo=2 TIE a y3—2 3.6 —¥1 ~3y, 60 Ys ~24-y; —04—2y, 65 42-24 As only three entries yo, yo, ys are given, the function Y can be represented by a second degree polynomial. A%yo=0 and A%y,=0 Le. 8y, +93=9; 41+ 3y3 = 3.6. : Solving these, we get, = 2.925, ys = 0.225. Otherwise : As only three entries yp = 3, yo =2, yg=—24 are given, the function y can be represented by a second degree polynomial A%yg=0 and © © (eA?) Queue j te (E-1Y'yo=0 “and (E-1),=0 ie (38? , ae = y= 0 ie. Ya~ B24 3y1-yy=0 Y4—3y3 + 38y2—y1 =0 Le. Ya* 871-9; ya Vi Soa Solving these, we get 1 = 2.925, 6.5. Effect of an error ona is an error € in the entry ys of a tal this error spreads out and is consid effects the difference table, 3INIT DIEFERENCES 163 Ys Bis) y9 The above table shows that : (@ The error increases with the order of differences. (ii) The coefficients of e’s in any column are the binomial coeffi- cients of (1 —e)". Thus the errors in the fourth difference column are e, — 4e, 6e, — 4e, ©. (wii) The algebraic sum of the errors in any difference column is zero. (iv) The maximum error in each column, occurs opposite to the entry containing the error i.e, ys. The above facts enable us to detect errors in a difference table. Example 6.7. One entry in the following table is incorrect andy is a cubic polynomial in x. Use the difference table to locate and correct the error. 27300) 0) 1 2 3 4 5 6 7 y: 25 21 18 18 27 45 76 123 i ‘ Re, UMERICAL THOS I OSNPI ING, py 164 ’ The difference table ia as under 4, 4, x y hy Wy Wy Ss 7 123 9 ing 2 phynomial A the third vane, The w31n h the third diffe saad. we 55 iz, B, white can ve written ag £2 6 L Z 2 iB 1 -1 1 =aFINITE DITTERENE ES 105 We construct the difference table from the given date ‘ y Ay A’y Aly A'y 0 youl a 1 yy \ 4 2 4 9 Yoel 4 16 2 4 3 Js \ 4 16 2 4 Yael dys 16 Aya Ay, 5 Yb Ay, 16 AY Ay, 6 6 Ay AY6 7 M7 Since the values ofy belong toa polynomial of degree 4, the fourth differences must be constant. But A4y = 16, *. The other fourth order differences must also be 16. Thus A4y1 = 16 = A%y. ~ A%, ie. A®yp = A°y; + Ay, = 8 + 16 = 24 dys = by, + AYyy = 4 + 24 = 28 | Ay4 = Aya + A*yg = 2 +28 = 30 and | Y5=ya+ Aya= 1430 = 31 | Similarly starting with A‘y» = 16, we get. Ag = 40, AYy4 = 68, Ays = 98, yo = 129. Starting with A4ys = 16, we obtain A*yq = 56, A’ys = 124, Ayg = 222, yz = 351. Problems 1. Bxpress.x°— 2x? + x — 1 into factorial polynomial, Hence show that A4fix) = 0. (P.T.U., B. Tech., 2001) 2, Express u = x4 — 12x! 4 24x” — 30x + 9 and its successive differences in factorial notation, Hence show that A°v = 0 (Kerala B, Tech, 1990)NUMERICAL METHODS IN ENGINEERING jy 166 3. Express Ox! = di" + 6x? + 2x + 1 a8 a factorial polyn, (Karnataha iffer of all orders. r nataka Bp = ~ Find the first and second differences of x4 ~ gy? , j 48 with h = 1, Show that the fourth difference is constant, ay ie ; of 4 We i function whose first difference is 2 Oat ag 5. Obtain the fun em ham 6. Given log 100 = 2, log 101 = 2.0043, log 103 = 2.0128, log = 2.0170, find log 102. ; 7. Estimate the missing term in the following table - 2 il 2 3 4 5 ee 3 9 — . 81 aN (Coimbatore B.E, 1 8. Find the missing values in the following table : x: 0 5 10 15 20 95 ye 6 10 pe 17 — 38 9. Find the first term of the series whose second and sub terms are 8, 3, 0, — 1, 0. (Nagpur B.E., 1 10. Write down the polynomial of lowest degree which Satisfies the following sét of numbers : 0, 7, 26, 68, 124, 215, 842, 511. 11. Find and correct by means of. differences, the error in the data 20736, 28561, 38416, 50625, 65540, 83521, 104976, 13032] 160000. 12. The values of a polynomial of degree 5 are tabulated below. If f(3) is known to be in error, find its correct value. 2y/9 =$(E+E*~2)+(@_E) = 32H -2) =E-1=A (iv) A®yp = (E-1)%yp [Re = (E°— 3H? + 3E— yp = Y5— 8y4 + 8y3 yo , V¥y5 = (1—- Ey, anveet os =(1-8E"4 30?_ Fy, = ¥5— 894 + 8y3—yo From (1) and (2), A8yy = Vy5. 1 Example 6.11. Using the method. of: oar ©) uyx + Wor? + ug? 4 a 7 Tox al ux? uged 2 haoT as sete ux (ii) ug + ae 6 (ot 2tug «2 1 @ LHS. = xu, +x? By, Puy a =a + BoA?FINITE DIFFERENCES 17) ‘Tox taking sum of infinite GP, 7 1 "5 ve | I) Bst+a } = eB yy = elt Duo =e. es Uo i xh x22 x83 we(1t 2828), alee eras =e | uo + 75 Aug + Mug +57 Aug + =RHS. Problems 1. Evaluate taking h as the interval of differencing : @$ sing Gi) (A+ PG? +2), (h =D. (Madurai, B.E., 1991) 2. Evaluate : 2, as @ (i | 2 Bhopal, B.E., 1991) as? Ng A? sin (« +h) ems wi pane ‘Bsin (+h) ” 3. With the usual notations, show that. @A=1-e%?' NUMERICAL METHODS IN ENGINEERING AN em afe (i) D= i sinh ( 2 } (Madray, Wh, (iii) (1 + AYL- 9) = 1 (iv) AV = VA = 0” Madran, By, 1 4. Prove that My, (i) 6 =A + Ay? = va - vy? (Madras, Bp, 2001 9 2 ” y ) wel: . Git) OY + BY) = ait, 4 5. Show that @) 6 = AE = vEY Gi) yb =F (+9) (iii) 1 + 04/2 = VL + 8p?) (Madras, B.h,, 200) 6. Show that @A=pb+ £ (Madras, B.B,, 2001) 1/2 we? =(14 | +2 Delhi, BB, 1994, | a Te Gi) B= (u +50 (Kerala, B. Tech., 1990 8) 7. Prove that @s-V=s9 (Madras, B.E., 2001) Wa+-8_¥ (Madura, B.E., 1996 8) Gii) V = hE" =E* pn =14-¢1 o WwE=> 9; () % = WD? np + HAD — 10 ix we n-1 8. Show that 3 Af, = A fy, - bfy. (Gurukul, M.Sc., 2000) hud 9. Prove with usual notations, that OV fy=h fy, GD) BGP) = fu + fuss) bfy (S, Gujarat B.E., 1999) Using the method of separation of symbols, prove that 10, Ug t+ Uy + Ue 4 1 4 Uy =" yty +" Chun +” CgN ug + san + "Cn (Gurukul, M.Se., 2000) ~FINITE DIFFERENCES Ll. Au, = ty, —" “Wzen +" Cty + * 12. Ye =In—"*Cybgn 5 + 4Copy, 9 + os k 13. y= *CA‘ye, i=0 6.8. Application to summation of seties, The Acsina of finite differences is very useful for finding the sum f 2 given ries. The method is best illustrated by the following examples Example 6.12. Sum the following serizs 2-H Denoting 1%, 27, 3%, required sum S=ugt Uy 4+ Uyt 4 tigg =(+E+ P+. +E, ts = Eng t= Paes ap [tenets ee — FMA la 2 20 ame: =n MED tugs B= UO—D 4... Now Aug =us—up=2?- 17-7, Mug =? — 2; + tg = 3-22 = = 12, 9 = Uz — Bitz + By —ug= 4-29 - 27° _ 12 =