Scribd
Upload
99 views34 pages

Chapter 2

Numerical Analysis

Uploaded by

All Research Real Information
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
99 views34 pages

Chapter 2

Numerical Analysis

Uploaded by

All Research Real Information
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 34
CHAPTER 2 ed SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS (Non-Linear equations) Non-Linear and Transcendental cquations occur very frequently in practice. By the term ‘non-Linear equations’ Wwe mean those equations that involve powers more than onc i.c. x* - 33 x x +5 =0 is quartic equation as it involves x*, Transcendental equations are those which involve functions other than just powers of x. ie. ¢*= cosx+ x’ is Transcendental equation. There are a-very few non-Linear equations that are readily solved . 1 . . such as x - Sv + 6 =0,sinx => ore*= 1, but there are majority of non- Linear equations which are not solvable by direct methods or analytical methods. For example P-x-10=0 ore’ + 2v+9=Oorlogx+ sins + 15=Octe. There are two types of methods for solution of non-Linear equations. (i) Analytical methods (i) Numerical (Iterative) methods 2.1. Analytical methods: ; Take nth degree algebraic equation with real co-elticients Paar - ee tag PO dg BF taywt a= 0 wee (1) Equation (1) has #7 roots, which may be (i) real and distinct (ii) real and repeated (iii) complex and repeated. § all the roots may be obtained by For equations of degrce 4 or less a analytical methods. 13 . , = . q aw INTRODUCTION TO NUMERICA | 14 Re (a) Roots of aq uadratic equation The two roots 1 72 ofa quadratic equation Py(x) saae tae + =O are given by a a ma 2a, £ @ Example 1 . i Find the roots of" S#6=0 using co | Solution: x x Example 2 Find the roots of x? +x + 1=0 using + Q) Solution: 1 1 Li yv3 w12 = —t q7! = 1,48 _ 1,8 = at 2! 1B. x sgt yi (b). Roots of cubic equation Consider the cubic equation ayy? + ax? + ayy + ay = 0 (2A) the co-efficients ao, a1, % at ~ eal so that the above equation has a real root, then we can use this 0 reduce the cubic to a quadratic equation and can find the remsiti"é roots. | men \ | [CHAPTER 2] SOLUTION OF ALGEDRAIC AND TRANSCENDENTAL EQUATIONS is The given cubic can be reduced to a cubic without the second degree term by the substitution aie rE Bay and it reduce to Z+3PZ+2Q=0 sees (3) a aay, 2a; Where 2Q= 7, ~3as! * 2705 2 a and 3P=7, a Bat . Aroot of (3) is always be given by are (carden’s Rule) 2=WFFO-? -NPFO+ OP “@ there are five cases a) When P? + Q? <0 TOF Z=-2y-P cosa (ee) 1 +Q) When P? +Q?=0 Z=~-203 @G) P’+Q'>0,P<0 z= Las FFE]? -[o-VFFe}? 1 (4) P?+Q?>0,P=0,Z=-(20)? (5) P?+Q?>0,P>0 1 z=[yP+@-Q]? -[VWP++Q] Then we can find a ~ 3ay Factoring he root x; from (3) by synthetic division we can obtain the quadratic equation for x and x3. Example 1: Determine the roots of the cubic equation. P' (x) =x? - 3x? + dy -2=0 \ODUCTION TO NUM: 16. AN INTR' ERA A £ Soluti Step 1:Given cubic is paar? +dr-2=0 Dezsi + Put x= 2-54 (Z+ir-3 (2+ IP HAZ + W-2=0 2432 +3Z+1-3(2' + + 1) 442 4+4-2=9 Z+zZ=0 es () Step I: Compare (1) with ‘ Z)+3PZ+2Q=0 sone Q) : 1 Q=0 3P=l > PaG Piagt=s5 >0 and P>O (Itis case (5), 4 1 then Z= [VP>+Q'-Q]? - [VP+@+Q]? » of. ye TI! : YEP LBs _ Root of (1) is Z=0 x= Zt1=0+1=) Step III: By synthetic division 1 3 4 2 1 41. 2 42 2 #2 00 ‘The remaining quadratic equation is 2x +2=0 / =1+] | The three roots are xy = |,x2=1+ hx = 1-7 [CHAPTER 2) SOLUTION OF ALGEORAIC AND TRANSCENOENTAL EQUATIONS 17 eee. \UAMSEENDENTALEQUATIONS 17 (o) The Roots of the Biquadratic equation Vacant aan even? - - # ay x24 ay" dy" = Oas the quartic Equation in s made equal to | by dividing throughout by Pye) = ay which the co-eflicient of x 1 ay . a’yasx* + aye! 4 asx + aww + aay = 0 (1) where ey =F te, Using Brown's 5 method and writing (1) as the difference of Iwo squares (HAVA BY (Cet DP=0 ee (2) Comparing Co-eMlicients of each power of x in (1) and (2) we get co-efli of x” 2A=ay co-efli of x? A 42B-Cmm fo ) co-cfli ofx 2AB - 2CD = a Constt. Term B?-D?=ay | Eliminating A, C and D from (1) shows that 2 B is a root of the cubic equation Z- Where by = a3 a — dean } by = a (day - a3") - a)? Choosing an 2B the alyebraically largest real root Z of (3) Now using the method of cubic equation onc finds Pe btb=0 00 tees GQ) vee ID) 3 z : a-F B-G p=\B ow Sap forD <0 D C=YA?~a; +2, forD=0 With these values of A, B, C and D equation (2) and ultimately equation (1) reduces two quadratic equations Y4(A-C) x4 (B-D)=Oands 4 (A4 C)x + (B+ D)=0 which can be casily solved Example 1. Determine the roots of the biquadratic equation xi tdx' 4 7x4 4 Gv 43 =0 TO NU! AN INTRODUCTION MERICAL ay ie the co-efficient of xis already | i erearig the given equation a at + ayy + tu = Say ay=4 me7 a =6 y= 3 ‘ Now by Browns" method consider the cubic equation 7+ by = 0 ++-(1) where tra = 7, by = even ~ 4a, 7 by =4xX6-4x3=99 dg = ay (4aq - as) - a? =3(4x 7-44) —@ =3 (28-16) -36=9 Step I. Ci ‘ + al tae Deas th equation (1) takes the form: . Bu + 22205 2-724 12)=z - 3) - 4) thy largest real root is 4. m4 a Step Il, A=G=5 =2, B=5=2 D=\BP—a =V4-3 =! SinceD #0. 2. C= YA? =i tar =V4-744 = Equation (A) reduces to two quadratic equations. ¥+(A-C)x+(B-D)=0>r+e4+1=0 P+(A+C)xt(B+D)=0543x43=0 the roots of (2) x, - taf in 3M . Relationship between the coefficients and roots of the Non - Linear Algebraic equations (a) Quadratic Equation Suppose the quadratic Equation is a@itaxta=0 °° 0 Let o and B be the roots of (1) = 2 to a Then a +B = and a+fP= [CHAPTER 2] SOLUTION OF ALGEDIAIC AND TRANSCENDENTAL EQUATIONS 19 en —eeEEeeeeeeecw_ arrose (b) Cubic Equation Suppose the cubic cquation is aye! ¢agx? + ayy + ay = 0 Let c, fi and y be the roots of (2) a then a+ BP+y= af + By + ya = 7, and aBy= z (©) —_ Biquadratic Equation Suppose the Biquadratic Equation is , ay! ayy’ + ayy? + aye + dy =0 Let a, B .y and 8 be the roots of (3) Then +B+y+5 == a+ By +y5 +502 = apy + be +60 +6ap=—— orfys = EXA MPLES Example 1. Solve x’ - 1217 + 39x - 28 =0 Provided the'roots are in A P. Solution: Step 1. Let a, i, y be the roots of - 127 + 39x -28 =0 Since the roots are in arithmetic progression a=a-d, B=aandy=atd AN INTRODUCTIO; F " 20 NTO NUMERIEy, be at pros daz l2 =>=4 Also aft pytya=39 or afa-dta(atdy +(e -d)=39 pur aed thon F29 2d 43 Also ofy= 28 OR a(a’-d’)=28 a=4 2. &=9 ad=43 Step 11. The foots of the eubie are (for d3) a=a-d=4-3=1 Bea. 74 yeatd =7 The three roots are 1, 4 and 7. Example 2. . Solve the cubic equation. . x 7x) + 14x-8=0 if the roots are in GP, Solution: Step I. Let c., B, y be the roots of x’ - 7x” + 14g- 8 =0 since the roots are in G.P , so. a=a, B=arandy=ar Nowa+B+y=7 atartar-8 ates Q) Of + By + y= 14 arta tara ig apy=s or ar =y or ar=2 2 fi or a= [CHAPTER 2] SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS 21 Putting in (1), © 2 we get 22,2, reece? “or art -Sr+2=0 2 or r=2, W2 Nowa=7 =! or a=4 Step Hl. fora =1,r=2 fi=ar=2, yzarn4 a=a=l, fora=4 r= 1/2 a=a=4,P=ar=2, y=ar=1 ‘The three roots are 1, 2 and 4. Example 3. Solve the Biquadratic equation yf 82) + 14x? - Bx - 15 = 0 when the roots are in AP. Solution: Step I. Let a, B,¥ and 6 be the roots of xt 8x" + 14? - 8x- 15=0 since the roots are in A.P. so a=a-3d, Bp=a-d yzatd, B=a=3d Nowa+P+y+6=4a=8 + (1) a=2 af + Py +7 +50 = or (a-3d) (a- d)t+(a- paras rd (#3 sees (2) + (a —3d) (at 3d) = 14 ay + fyS + ya + Saf = -8 apy6 = -15 AN INTRODUCTION TO NUMERICAL ANALYS1g + (4) pq (@-s) 2-5 or (a a= 2then (4) becomes a-@) (4-94) 2-15 or 6d? + 40d? + 9d" = -15 or dt tS +31=0 od! = 9d - 31d + 31= od - 31) (P-1)=0 +1 dtl Step I. For a= 2andd=1 The four roots are. _ aaan 3d=-1 . Pea-d=l -y=atd=3 S=at3d=5 2.2 Numerical (Iterative) Methods Iteration: 1 R Consider the formula = 3 (« + 5) te the value of when x = 40.0 by putting in 97. 1 equation (A), it gives = 4 40 +B) 1 (40 + 49,3) = 44.6 Now ted for instead a in (A), the answer this time for 1 (A) Take = 1973 and calcula suppose 44.6 is substitu _ is 1 1973 1 -2 3 [4 6 +74. aa] eo [44.6 + 44.24) = 44,42 If x is substituted as 44.42 in (A) we get 1973 1 : [a 42+ | -2 [44.42 + 44.4169] 44.42 = 44,4186 Again putting ¥ = 44.4184 in (A) we get t= 44.4186 [CHAPTER 2) t R.2) SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS 23 You will notice that x and = at x i have been getting closer and closer to cach other as we have proceeded. If th i i il become cqual to an even higher degree ofaccurasy. ‘scomimue, whe “il As this is a repetiti A petitive it it 5 programming purposes. Process, it is very convenient for The process of finding successive’ mati 4 approximation to a ity i called an ITERATIVE PROCESS. Thus equation (A) would lead 10 the successive equations 1 R n=D (+8) at(a+4) ' D R In general Xe +1 = ( +§), k=0, 1, 2, ..... which evaluates the square’ root of a real number R. To calculate the cube root of a Real number R, the iterative process is 1 R nz [ane] Linear equations by means of the Iteration Solution of Nou formula x4 +1 =f 0%). Let us take the Equation? - ay +x-10=0 wee (1) rearrange (1) so that it is in the form x = f @). There is evidently more than one way to arrange it. Some of the forms are yo-+4x (ii) xy Ve +x- 10) @ 4vextl0 | etx 10 Gi) x= @y . . . ined b Now we need an jnitial ‘starting point, which can be bins y various methods such as / Q Pexn drawn, “(@)_ A rough sketch of the curve “8 = ar’ +x oe xp would be taken that value which crosses the + ‘ ; 1 aacxt land y= es of y= 48 -* (b) Rough sketches on the sama a ve ae ction at these x are drawn and Xo is curves AN INTRODUCTION 10 NUMeRIcy 24 Lan 2 ues of ¢ andy vey (Ong vatties of x, to be taken near that value woe ofy-chanke ‘The talc is ‘ the sign changes between 4 and 5 ie.xr4 Now setling the iteration , for (i) teu = 1O-Ka + Ae xy = 10ry = 590 x3 = — 203986590 n=0,x0=4- The process is diverging and not coming closer to the true sq dv! ox, +10 for (ii) Xa+t ee tion forn=0 x=4 m=44 = 429 454.310 y= 43063 x3 =4.3070 3069 x7 = 4.3069 The last two values are the same indicating that iteration Bro may’be stopped and the root x = 4.3069 is correct to 4 decimal places Remark: Why did the first formula (i) for xq + 1 lead to a ridiculous rather to no result? Let us answer this question by doing the error analysy of the above problem. Error Analy We have to solve the equation x = J (x) and we have used the iteration formula x, . | =f(,) Now suppose x =a is the actual value of the root then f (a) =0 and also a=f (a). If x, is approximation for the actual root a then €,=a-% Ifxq is a reasnaly good approximation close toa then €, is small thusx= 4 ~-€, and the iteration process x). 1 =f (v,) putting x, . =a eq 10M expanding the RIS by Taylor's series we wet a> Gn =f (d)~ caf (a) since f(a) =a then €,.)= af (4) where ny) < ey iff) | <| ‘Which is the criterion for convergence, [CHAPTER 2] SOLUTION OF ALGEORAIC AND TRANSCENDENTAL EQUATION is 25 Nowin (i) f(r) = 10 J(s) 3x? + 16x | process divergent + dx? ac4 in Gig SEH SPH 10 r LO) =3 2 <|for.x=4 process is convergent Remark: It is not always possible to find an interval centered on the root such that [/(x)| <1, this can be demonstrated by the following example. Example 2.” 06 Solve the cquation x + log x = 0 when the root lies between 0.5 and Solution: Step I. We write the equation as x =— log x compare with x= f(x) SQ) =-logx 1 L@)A-¥ JU’ @lel : Hence [f' (x) | 2 in the interval (0, 1] and in particular for x € (0.5, 0.6) Lf" (x)|>1 . This means that we have failed to prove that this functional iteration will converge for any interval carried on the root. In fact this functional iteration will diverge for any starting value xo € [0, 1]. as Step IT. Now given equation may also be expressed xee™ ° Now f(x) =e sof’ (x) =-e* and |’) =] -e 71 forallx>0, [f'@)I<1 In fact the functional iteration x = f(x) any starting value xy > 0 OF Xpey = e7*" will converge for 26 AN INTRODUCTION To. Nun ; Example 3. Find the root of x’ - 5x” - 29“ O correct t0 3 decimay Play Solution: Given equation is fxn -x7-29=0 Step I. Calculate of the range of the root when the sign of fx) chars yPol,r,2[3 [4 [5 Fe | -29 | 33 | a1 | -47 | -45 | -29 |G The root lies in between 5 and G xo=6 Step II. Sctling an iteration as in the first example. _ te +29 ene for n=0 %y=6 n=l x=58 x= 5.86 2 yp=5860 xy = 5.845 ; 3 Ky=5845 x, =5 849 4 xy=5849 9 x5=5 484 ol ” n u 7 n=5 — xs=S848 x, = 5 B48 Then root x = §.848 Bisection method or Method of halving the interval Bolzano method Given a continuous function fix) of 2 real variable x, f (x) bet given function whose root (s) we wish to find and suppose that we ket been able to find (By tabulation of f (x)) 1wo points ay and by (bs? which are such that f (cis) f(6,) < 0 Now: for this condition to be setiss, (40) and f (bo) must be of opposite signs and consequently we can sy there is at lest one root of f (x) = lying in the interval (as, bol. SUP there is only one root. The method of bisection aims to generes * sequence of successively smaller intervals in which the root is know? lie. Take the mid point Mo = (1, + by)/2 of the initial interval, J (Mo) and form f (a). f (Mc). If this quantity is negative the root ! [eto, Mo} jourttR 2] SOLUTION OF ALGEDRAIC AND TRANSCENDENTAL EQUATIONS 27 Let (a1, 41] be the new interval containin, i i we generate a sequence of intervals [(,, Bq] a= OTs te is ia rane size comaining the rot of f(x) = 0 123s nF decreasing By repeating this Bisection (halving the int come arbitrarily hear to the zero (root) of f(x) = 0, eral) role we can size comtaiing the root of f(x) =0, 1 ee Fe OF CCCTCASIN Remark: : The method of Biscction is of interest and importance in its own right and sometimes it provides an initial approximation for some other method which is expected to give more rapid convergence. We illustrate the process by the following examples Example 1. Consider the non-Linear Equation fQ@)=e-¥ -3y-3=0 Solution: At x=] f(x)=-4 and at x=2_ f(x) = 43; since the function is continuous and the change in sign of the function between x'= 1 and x = 2 guarantees at least cone root in the interval (1, 2). Now suppose we compute the value of f (x) = 0 at x= LS f(x) = -1.475, again there is a change of sign between the interval (1.5, 2.0) at x= 1.5 the sign is negative and at x = 2.0 the sign is tive. In this way go on squcezing the interval by halving it each time, and nothing down the change of sign in the interval Continuing this cess we have x = 1/3 = 1.732051 as a positive root of the given non- re Linear equation. The iterative process is illustrated in Table 1. TABLE I weraion? sy [om Tegan | £00 | fe) rT Poo} 15s | -40 | 30 | -1475 Ts | 20 | 175 | -1a7s | 3.0 | v IIs? rs_| 7s |_1.625_| -1.75 017187 | -094335 | 0.40942 1.75 1.6875. | -0.94335 | 0. 17887 1,6250 0.17187 | -O/12478 1.6875 | 1.75 | 1.71875 | + | 40942 100 . 1.731 SL wfafa}olol— Example 2 Find a positive real root of S(x)=ef-3x=90 Solution: The bisection method works all right on the Transcen equations also. Given f (x) = e* - 3x for x=1 f(x)is negative for x=2 f(x)is positive Therefore the root lies betws can be seen from the Table 2. cen x = 1 and x = 2 the other y; TABLEZ = = Tteration x= No | ™! | 7 | @itey2 L 1 l 2.0 15 +1.38906 2 15 2.0 1.75 -0,01831 | 1.38906 3 1s + 1.75 1.625 |_-0.01830 0.50470 4 1.5 | 1.625 | 1.5625 | -0.018298 | 0.20342 5 1.5 | 1.5625 | 1.53125 | -0.018298 | 0.03020 2 | i 100 1.51209 x = 1.51209 is the required root j Example 3, 4 Solve the Transcendental Equation. Six) Solution: sin (mx/2) = 0 to a positive real root by Bisection met! Given function is S (x)= e* = sin (2) =0 at F(& ) is positive at J (x) is negative [CHAPTER 2} SOLUTION OF ALCEDRATC AND TRANSCENDENTAL EQUATIONS 29 Therefore the root of the non-Lincar equation lies in the interval (0,1). Now at x= 0.5 f(x) =-0 1006 Now the sign changes at 0 x=0and x= 0.5, Squeezing the in to a desired accuracy, noting th and 0.5, therefore the root lies between terval (0, 0.5) to get a Positive real root acy, ¢ change of sign in each interval the various steps are given in the Table 3, interval (0, 0.5) Where f (x) = o sin (m x/2) and intervi A TABLE3 tion . . tera to x Sin &) ee f(s) 1 0.50 0.70710681_| 0.606530 | - 0.100576 | 2 0.25 0.38268344 | 0.778801 0.396117 [3 0,375 0.55557026 | 0.687289 0.131719 {4 0.4375 0.634393 0.645748 0.011255 {5 0.46875 | 0.67155898 | 0.625784 | -0045775 |_6 | 0.453125 | 06531728 0.635638 | -0.017534 7 04453125 | 0.6483156 0.640624 | -0.003207 so nearly 445125 is a positive real root of f (x) = 0. Example 4. Find by simple iteration to five decimal places the root near 0.5 of the equation sin x = 5x - 2 Solution: Step I. Let us write the given Equation. Sin x = Sx 2 in the for x = sin’! (x 2) S0) = sin"! (x-2) (x= flr )is the formula) J) f' 05)=95 >I (1) will not lead to convergence. Step Il, Let us we tho given equation in some other form say; at =(sinx +2) pooped const rand OO zon (0.5)< 1 1 f° (05) 5 298 gence, Let us set up the i iteratio Step TIT. Which will lead to conver! Q). nantt (sin X42) with x9 = 0.5 for n=l +} (2479) = 0.496 n=2 ms (oars) = (0.4982 ne3 net (2.475208) = 0.495042 . n=4 uy (2.475061) 0.495014 " n=5 Xs =} (2.475044) 0.495009 Hence the rot near 0.5 is x = 0.29501 Example 5. Find to 4 decimal place the root near 3 of the equation. 4 using simple iteration. Solution: Step I. Let us write the given equation in the form “x= loge(20+log) sees qQ) F(X) = loge (20 + log.*) ‘y= _ f O° SGT 8H J’ ad <1 [cHAPTER 2] SOLUTION OF ALGEDRATC AND TRANSCENDENTAL EQUATIONS 31 Step Il. Since (1) leads to the convergence, therefore setti Iteration to (1) ting up an Xn+1 = loge (20 + log, 1) sree (2) where Xo = 3 for n=0 x, =log, 21.099 = 3.0493 for m=] X2=loge 21.115 = 3.0500 for n=2 x)= log, 21.115 = 3.0500 Hence the required root is x = 3.0500. 2. Acceleration of convergence : Aitkens A? process. Convergence of the first order iterative process for solving /(x) = 0, can be accelerated, Suppose we have obtained three successive. approximations to the desired root x say Xn-1%m X91 then X¥—Xy-1=En-1 X-X, = En X—Xne1 = Ent Since En-1, Enr 1 arin G.P. so we write tem as Enr1__Gn En Enel or ‘on solving for x we get 2 Kane | XI Xn ————Eoe Xnvt 7 2%ntXn-1 2 Xn TM Xe =Xnr a a OF ERI ga 1 Wnt Xn _ Pasi Bait] PHL 4 = ent Xan d (Xp01=%n) ves (A) THntl yyy = Din t Xa 32 Now in finite difference relations xn = Ani itn Alener = A(A%e=1) = Bln 0-2) Exe yn What Xa-d (A) because Ax)" XH NCE) Which is the required 4” process Example: Calculate a positive real root of f (x) = v-2k-3=9 Aitken’s process and the formula x =f (X), Solution: “Step I. The given equation is x* -2x-3=0. where f (x) =x? - 2x -3 Let xo =4 be he initial iterate Pemt3 or x=y2rt3 To set up the iteration we have Xai y= \Y2xqt3 for .n=0 x04 = 3.316 for n=1 X=3.316 x2 = 3.104 Step II. Table of differences, x x ae x= 4.000 sees 0.684 M336 eee 0.472 sess 0.212 x2 = 3.104 ‘The root isr= x, - 4% x= 3,009 [CHAPTER 2) SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS 33 TON RAPHSON FORMULA TO FIND A P. REAL ROOT OF EQUATION 10) = 0 OSITIVE Statement: If x, is an approximation to an exact root a of the equation f(x) =0 then a better approximation is in general given by xq +1 where LG). kno Xan ra) s of establishing this result we shall obtain it in well NEW Proof: There are many way: by the following argument which although scems simple but fits i wit the theory of iterative processes already developed Let the non-Linear equation be f(x) = 0 further Let xo be the initial approximation 10 the exact root a, the second approximation be xy=xoth Where Ais tive and very very small Then using ¥ =f (x) (a functional iteration) we have Sef @ot hy=0 expanding (2) using Taylor's se higher powers of Awe have Set HH Loo) * Af (xu) ~ 0 fo where ' (x0) #0) ries and since h is small ignoring If and he F) putting in()) i) @) HOF) Likewise x2 =*1 +t BENE) in general {Wa @) Bev Xa F (x) (4) is the Newton Raphson. method. AN, snTRODUCTION ‘TO NUMERICAL AN CONVERGENCE 34 ORDER OF Definition: Given a non-Linear Equation f(x) = 0, we have iteration formula ne v= (tn) Letxo, Xt gy Xn bE sequence of approximaya, = 0, Set En= ~x; Which converges to the exact root & Ol f(s) = 0 Set n= ~ x, ip | Lim Lim |Sa21 exists a number P and a constant C #0 such that Lim Lim P| a | then p is called the order of convergence of the sequence and Ci al the asymptotic for constant. Theorem: Prove that N methad or is a second order method Proof: Newton Raphson Formula is Lea. a) Tae =¥a~ FF (y,) Let the exact root of f (x) = 0 bex=a and the approximations to a, yield ewton Raphson is a quadratically Converge Xn-@= En Nye TO Ent From relations (1) and (2) we get _. La OF En Ea Gs c,) 2 [reo ef +7" 0).. Ene) = Eq i (a) + enf" (a +77" @).. since f(a) = 0 and €, is small we get S'(e)| 14 S2£@) ' rolite; [CHAPTER 2) SOLUTION OF ALGEDRAIC AND TRANSCENDENTAL EQUATIONS 35 oo IESILENTAL EQUATIONS 35. | (ignoring & etc) Enel = (a) Eni = e493) where k=4 ae Relation (3) signifies that each error is proportional to the square of the previous error showing there by that Newton Raphson is a quadratically convergent method or second order process. Remarks: 1, Newton Raphson isa rapidly convergent iterative method, 2. the evaluation of the derivative /’ (x,) at each step makes the method unpopular computationally. 3. if the root is very near to the maximum or minimum value of the function at tht point, Newton Raphson method fails. 4. _ if two or more roots are nearly equal then the method is not fastly convergent. Iffa nymber p and a constant C #0 such that Lim | en+1 1/1 & | =C, then pis called the order of na " convergence of the sequence and C is called the asymptotic error constant. Example 1. Using Newton Raphson method find to 4D, the root near 0.5 of the equation e* —sin x = 0. Solution: Given equation is I(x) =e7 -sinx =0 x= 0.5 Sf’ (8) =— (€7 + 005 x) Formula x01 = %)-T Xne =X GC ~—~—SF RICAL popuction TO NUMERICAL ANALY, AN INT 36 9 0705 For 4 OS sat 586 x 705 7485 0.0035 _ 9.5885 =| y= 0.586 +7,3897 0.0005. _ 9 5885 x = 0,5885 + 7 38693 0.5885. ua? Hence the required root is * = Example 2. Find the root of x7.~ X 2 Newton Raphson method, that lies near * y-3 = 0 to four decimal places, Using Solution: Step J. Given f@)=v-2-3=0 S'@)=3"-3 xy=2 Formula to be used v,- 3-3 i Xe =Iq— Step 11, n=0 x20] n=) 42 = 2.10383 n=2 3 = 2103834032 Thus to four decimal places, the root is 2.1038, Example 3, Using the Ni the root of the e&-3x=0 ston Raphson m : \ethod evaluate t quation which lies be 1 0 two decimal places” ween O and 1, the equation is f(@) [CHAPTER 2) SOLUTION OF ALGEORAIC AND TRANSCENDENTAL EQUATIONS 37 EET EQUATIONS 7 Solution: Step I. Given equation is IW) =e ~ 3x %o=0 I’ @)HE-3)=e-3 Formula to be used oe — 3x, Meu in~ Gg Step II. . n=O0,x%=0 x, =0.50 n=l 120.6! n=2, x3 = 0.62 The root is x = 0.62. The Secant Method: One disadvantage of Newton Raphson formula is the need to evaluate the derivative f’ (x) at cach stage, which is sometimes difficult when the function f (x) is a complicated one. One way to avoid this is replace f' (xn) by the gradient of the secant joning the points (x, f (%-)) and (%n-1,J (in-1)) the gradient of the secant is fn) =f Gin- Xn—Xn-t L1G) putting this value in sent ea fe we get LQ) n= Xn-t Bnet ¥n— Fy) =f (r-1) (vq) + Xn— 1 J Or) Xaf (Xe) — Sef %n-1) = Xn . F (i) -f Gn) Xan f (Xn) =X f Ono Sen) -f Xn=1) Relation (B) is known as secant formula. AN INTRODUCTION TO NUMERICA, Note: The two estimates x0, x4 for the ci root ‘x’ are "eOireg edo not dema We vey oes the nule of false position discusseg Lig ro) F(t s Example 1. ‘ei the secant Method to find correct to 4 decimals the between 0.4 and 0.6 of the equation Sinx =5x-2 Solution: In order to apply secant method w ¢ need two starting Value Xo- 0.4 and x, = 0.6. Formula to be used nF (x1) -F n=) yale). (A) I (ei) -f (%0) S(e)= sin x 5x +2 y L(%0) =f (0.4) = 0,329 S(1) =f (0.6) = - 0.435 = 0.494 £(e2) = 0.0042 n=2 x=0.4950 J) = 0.00003 =2 x= 0.4950 . Hen the root is x= 0.4950, Example 2. Using the secant the equation e* ~ 3x = method to evaluate to 4 decimal places the root Starting vac. are X9 = 0.4 and x1 = 0,9 Solution: Formula to be used Fane =X %=04 x =09 [CHAPTER 2] SOLUTION OF ALGEDRAIC AND TRANSCENDENTAL EQUATIONS 39 n=l x2 70.67 n=2 x3=0,624 n=3 xX4=0.620 n=4 xs=0.6192 Remarks: 1. The method is very simple to evaluate the roots of non- Linear equations. 2. the method docs not usually converge as rapidly as Newton - Raphson's method, but avoids the evaluation of f' (x) and is well adapted to computer usage. 3. if the evaluation of f’ (x) is expensive in terms of computer time then the secant method is likely to prove to be more efficient than Newton-Raphson. Result: Calculate the rate of convergence of the secant method (error analysis). : Proof: The secant formula is ~@ 6) Beer En Fe) —f (tn 1) Let xa-1= 4+ En-1 My =ateés Xne = at Ened Then (1) takes the form €,- En f(a+ En at en = a4 En Fat e,)- flat en-i) Using the Taylor’s expansions for fla+ En) and f(a En- 1). We get Lay (en= ene) OD EWS 4D EV @ Ent ay ean S4Z Sel vores @ + en" (A) f(a=0 U0—_=- @ @ ‘AN INTRODUCTION TONUMERICAL Ay “pfeastta yee SOI een eae (ea Gf (Glen ee OL" lear @rters@) =¢,-——,—._.. SFE (Get ene S" (a) 1 i =e 7q (ef (a+4 hf" (a (I-3 Clear ena i =2 ene Gh Let mite Ent? kenGa-t EnetSEnenb te , Relation (2) shows that cach error is proportional to the Prodicty the previous two errors. | Remarks; | It is interesting to compare this result with that obtained for ty Newton Raphson formula, namely Env ieEn From this comparison it is evident that Newton Raphson method is superior to that of secant method. Since it involves the evaluation of hk derivative at cach stage therefore, the secant’ method is a very attractive alternative to the standard Newton Raphson method. Method of False Position (Regula FALSI) The method which is the oldest one to find the real roots ofa equation f(x) = 0 has a close resemblence with the secant method. To find the real root of the Ls) equation Xx) = 0 we consider a sufficiently small interval (x1. , xn) in which the root lies, Since the root lies in the interval (x1, , xn), the curve must cross x — axis between x, and xp and hence yy =f (x1) .f (2) <0. be [CHAPTER 2] SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS 41 ee et The Regula False method is based on replacing the part of the curve between the points and then taking the point of intersection of the chord with x-axis as an approximation to the root, now consider the equation of the chord joining the points (*1, , y1) and (te , y2) ic yopegit (r=) sens (A) The point of intersection of (A) with x-axis is obtained by putting y = in (A) so that we get Xn- 1, ya- x-xL= (1) put =x) yy = f(s) and yo =/Cen) (Xi. (xn — x1) veceee (1) weavers =#.-" 7(¢4)~ (a) which gives the approximation We shall now determine which of the two sub intervals x1, < x < x, or x;

You might also like