Power Series Solutions of Linear Differential Equations with Variable Coefficients SECOND-ORDER EQUATIONS A second-order linear differential equation bylwy" + dyooy' + boldy = g00) ern has variable evefficients when by(x), by(x), and by(x) are not all constants or constant multiples of one another. IFhy(x) is not zero in a given interval, then we can divide by it and rewrite Eq. (27.1) as y+ Play’ + QI = 9.09 27.2) where P(x) = 6)(x)/b260), Olx) = by(xvibax), and 4x) = gCo)/bsCx). In this chapter and the next, we describe procedures for solving many equations in the form of (27.1) or (27.2). These procedures can be generalized {ina straightforward manner to solve higher-order linear differential equations with variable cvetlicients. ANALYTIC FUNCTIONS AND ORDINARY POINTS. A function fx) is anatytic at xp if its Taylor series about xo, x)=) ‘converges to f(x) in some neighborhood of x, Polynomials, sin x, cos.x, and e* are analytic everywhere; so (00 are sums, differences, und products of these functions. Quotients of any two of these functions are analytic at all points where the denominator is not zero. The point xp is an ordinary point of the differential equation (27.2) if both P(x) and Q(x) are analytie at I either of these functions is not analytic at xo, then xo is 4 singular point of (27.2). 262 Copyright @ 2006, 1004, 1973 by The McGraw-Hill Companies, Ine. Click here for terms of useCHAR 27) LINEAR DIFFERENTIAL EQUATIONS WITH VARIABLE COEFFICIENTS. 263 SOLt |ONS AROUND THE ORIGIN OF HOMOGENEOUS EQUATIONS Equation (27.1) is homogeneous when g (%)= 0, in whieh y"+ Play’ + Oey =0 27.3) we Eq. (27.2) specializes 10 ‘Theorem 27.1. If.x=0 is an ordinary point of Eq, (27.3), then the general solution in an interval containing this point has the form Ler 90) + a,y3(0) 74) where dg and ay a analytic at x= 0 e arbitrary constants and y,(1) and y3(x) are linearly independent functions To evaluate the coeificients «,, in the solution furnished by Theorem 27.1, use the following five-step procedure known as the power series method. Step 1. Substitute into the left side of the homogeneous differential equation the power series ve Dar aa tartas sae! tay Fa baggy, gE together with the power series for y'=ay + Daye + Bay?+ aye? tran + + Didgg + (Ot Dy aX" (27.6) and + 6agx-+ 12ay+ + (= Taga2+ (+ I dy "N+ (1+ 2)01+ De eat ern Step 2. Collect powers of x and set the coefficients of each power of x equal to zero Step 3. The equation obtained by setting the coefficient of «to zero in Step 2 will contain 4; terms for a finite number of j values, Solve this equation for the a, term having the langest subscript. The resulting equation js known as the recurrence formula for the given differential equation, Step 4. Use the recurrence formula to sequentially determine a (j= 2,3, 4,...) in terms of ag and ay Step 5. Substitute the coefficients determined in Step 4 into Eq. (27.5) and rewrite the solution in the form of Eq. (27.4). The power series method is only applicable when x—0 is an ordinary point. Although a differential ‘equation must be in the Form of Eq, (27.2) to determine whether 1° = 0 isan ordinary point, once this condition is verified, the power series method can be used on either form (27.1) or (27.2). If P(x) or QC) in (27.2) ane quotients of polynomials, it is often simpler first to multiply through by the lowest common denominator thereby clearing fractions, and then to apply the power series method to the resulting equation in the form of Eq. 27.1). SOLUTIONS AROUND THE ORIGIN OF NONHOMOGENEOUS EQUATIONS If § (2) in Bg, (27.2) is analytic at x=0, it hay a Taylor series expansion around that point and the power series method given above can be modified to solve either Eq. (27./) or (27.2). In Step 1, Eqs. (27.5) through (27.7) are substituted into the left side of the nonhomogencous equation; the right side is written as a Taylor series around the origin, Steps 2 and 3 change so that the coefficients of each power of x on the left side of the264 LINEAR DIFFERENTIAL BQUATIONS WITH VARIABLE COEFFICIENTS. [CHAP 27 ‘equation resulting from Step 1 are set equal to their counterparts on the right side of that equation. The form of the solution in Step 5 becomes yt agyy(a) + aya) + yar) which has the form specified in Theorem 8.4, The first (wo terms comprise the general solution to the associated homogeneous differential equation while the last function is a particular solution to the nonhomogeneous equation, INITIAL-VALUE PROBLEMS Solutions to initial-value problems are obtained by first solving the given differential equation and then applying the specified initial conditions. An alternate technique that quickly generates the first few terms of the power series solution to an initial-value problem is described in Problem 2723 SOLUTIONS AROUND OTHER POINTS When solutions are required around the ordinary point x90, it generally simplifies the algebra if xy is translated to the origin by the change of variables ¢=.x~ %o, The solution of the new differential equation that results can be obtained by the power series method about = 0. Then the solution of the original equation is easily obtained by back-substitution Solved Problems 27.1. Determine whether x= (0 is an ordinary point of the differential equation yaa +2y=0 Here H() =—.xand 2(x) = 2 are both polynomials; henee they are analytic everywhere, Therelore, every value of, in particular = 0, is an ordinary point 27.2. Find a recurrence formula for the power series solution around = 0 for the differential equation given in Problem 27.1 {It follows from Problem 27.1 that x=0 is an ordinary point of the given equation, so Theorem 27.1 holds Substituting Eqs. (27.5) through (27.7) into the left side of the differential equation, we Find (a+ 6ase + ayy? + (n= Daya"? (+ Diag oad E + + DU Diy 0+) May Dage+ 3a? + aye? +t ng" (+ Dit + (+ Diy ak] Hay ae] 20 + Dlapt at ase tase + agit to tage dy"! Combining terms tat contin lke powers of x, we hve (2as+ 2a) +.M6ay+ay) #37024) +392005— as) SENT D+ Dit 2 Rat 2) + = 040+ Oe + HOE ‘The last equation holds if and only if each coefficient in the left-hand side is zero, Thus, a+ & | 6a+a=0, 12ay=0, 20a,-a,=0, In general, (n+ 2m Ddy 22-2),CHAP 27] LINEAR DIFFERENTIAL EQUATIONS WITH VARIABLE COEFFICIENTS. 265 (n-2) werd Which is the recurrence formula for this problem, 27.3. Find the general solution near x= 0 of y"—xy' + 2y=0. Successively evaluating the recurrence formula obiained in Problem 27.2 for n=0,1, 2, ..., we caleulate o 0, i follows from the recurtence formula that all the even coefficients beyond ay ae also zee Substituting (7) into Ba. (27.5) we have ya, taxa, 2 eau ra If we define iyi) +a then the general solution (2) can be revritlen as 27.4. Determine whether x= 0 is an ordinary point of the differential equation yy= are both constants; hence they are analytic everywhere. Therefore, every value of xin particular x= 0, is an ordinary point, 27.8. Find a recurrence formula for the power series solution around = 0 for the differential equation given in Problem 27.4. It follows from Problem 27.4 that r= 0 is an ordinary point of the given equation, so Theorem 27.1 holds Substituting Bas (27.5) through (27.7) into the left side of te differential equation, we find ass 6aye + Day? + ml Diage? (UF Dy 14 N+ DONE Digs +]266 LINEAR DIFFERENTIAL BQUATIONS WITH VARIABLE COEFFICIENTS. [CHAP 27 Fag age wt aya or (Cay a9) + Gay 44) + 22a tay) +38 20a +a) E+ 2)n4+ Daya) + Oe HOE HO Equating each coefficient to zero, we have | 6a +a)=0, Tay 20+ In general (#2) Diag gat hich is equivalent ro “1 Geared) ‘This equation isthe recurrence formula for this problem, 0. Since this equation has constant coeficients its solution is obtained easily by either the characteristic equation ‘method, Laplace transforms, oF matrix methods as y= ¢) €08.x+ ¢ Sin Solving by the power series method, we successively evaluate the recurrence formula Found in Problem 27.5 11, Zyoin 27.6. Use the power series method to find the general solution near x=0 of y"+ y= fon Oo ac oe 1 (KS) \ ne" Recall that fora posi 1 integer n,n factorial, which is denoted by nis defined by n= n(n 12) BMD. and 0! is defined as one. Thus, 4 n= nD) Now substi 120. In_ general, Ww ButCHAP 27] LINEAR DIFFERENTIAL EQUATIONS WITH VARIABLE COEFFICIENTS. 267 279. ‘Substituting these two results into (/) and letting ¢1= a and c,a,, we obtain, as before, cos ++ 6, sin Determine whether x is an ordinary point of the differential equation 2ety" + Tate Dy! = Dividing by 2+, we have Tees yates Pon Oe 0 (both denominators ate zero there), ‘As neither function is analytic at ‘a singular point. is not an ordinary point but, rather, Determine whether x=0 is an ordinary point of the differential equation, By" + 2y’4ay=0 Here P(x) = 2x? and Qex) but, rather, singular point, 1x Neither ofthese functions is analytic at x= 0, so.x= Oi not an ordinary point Find a recurrence formula for the power series solution around = 0 for the differential equation dy o Bra-yBsar-3 Get DE et-3) Both PW) =1~ 1 and Q(9 = 27 ~3 are polynomials; hence every point, in particular '=0, is an ordinary pont ‘Substituting Eqs. (27.5) through (27.7) into the left side of the differential equation, with £ replacing x, we have (2a 6ast + Dag + + nln Nag (+ Lymayt8$ (0+ 2K Diy t+ (C= Dla, + 2agt + Bay + dah +o mage (Diag PE OE Dhy at oo] FOU-Blaytayt tag tat tagtt + tag" ay, Aa, E+] =O or (2a, a, ~3ay) + 645+ a) ~2ay+ 2ay~ 3a) + P20, + 2a) ~ 345+ 2a —3ay) + FIO 2) Dy 2 Mg (+ Diy 1+ 2g 13) + 0004 OF eo 400" + Equating each coeflicient to zero, we obtain ee oO no (nt 2+ Daeg 2— (N+ Dy 5 1+ 3) + 2a = 0 ceenerenens ‘Equation (2) is the recurrence formula for this problem. Note, however, that it is not valid for n= 0, because a isan undeined quanity: To obtain an equation for n=O, we use the fist equation in (J), which gives a =ta, 43a,268 LINEAR DIFFERENTIAL BQUATIONS WITH VARIABLE COEFFICIENTS. [CHAP 27 27.10. Find the general solution near ‘=0 for the differential equation given in Problem 27.9, We have from Problem 23.9 that 13, masatse ‘Then evaluating recurrence formula (2) in Problem 27.9 for successive integer valves of » beginning with n= 1, we Find that Substituting these values into Eg (27.5) with x replaced by 1, we obtain as the general solution fo the given dffer- ential equation 27.11. Determine whether x= or.x= Lis an ordinary point of the differential equation (1-24 )y" = 29" nln Dy= 0. for any positive integer n, ‘We frst transform the dilferential equation into the form of Eq (27.2) by dividing by 4°— 1. Then Pw Both of these functions have Taylor series expansions around '=0, so both are analytic there and x0 is an ‘ordinary point. In contrast, the denominators of both function 0 mether function is defined there and, therefore, nether funetion is analytic there. Consequently, x= 1 isa singular point 27.12, Find a recurrence formula for the power series solution around x =0 for the differential equation given in Problem 27.11 ‘Toavoid fractions, we work with the differential equation in its current form, Substituting Bys. (27.5) through (27.7), with the dummy index 1 replaced by &, into the left side ofthis equation, we have that (=D + Gaye ayy? +AU Dawe? (e+ Dac, ae! 406+ 2M Day ae] Dafa + Daye rays kant" (e+ Day +O 2a, hE de kUk+ Dlagt apt ane? tag -=taarh Fagen tay atd tt] =O Combining terms that contain like powers of x, we obtain (2054 ma] 605+ 2 + m—2)a) + st at[k+ 2k + Day 2+ OP + n-—Bay) += 0CHAP 27] LINEAR DIFFERENTIAL EQUATIONS WITH VARIABLE COEFFICIENTS. 269 27.13. 27.14. Noting that 2+: 2 k= (n~R)(n+ + 1), we obtain the recurrence formula Yn ket), +2 o Show that whenever 1 is positive integer, one solution near x=0 of Legendre’s equation (1-22) 2a + nl + Dy isa polynomial of degree n. (See Chapter 29.) ‘The recurrence formula for tis equation is given by Fg (J) in Problem 27.12. Because ofthe factor n~B, we find, upon letting k=, that a .2=0. It follows at once that 0 dy. 4= ay = Ayo g= -»- Thus, if mi odd all odd coeflicients ay (K> n) are zero, whereas if is even, all even coefficients a, (> n) are 210. Therefore, either yy(x) for yas) in Eq. (27-4) (depending on whether n is even or odd, respectively) will contain only a finite number of nonzero terms up to and including a term in x"; hence, itis a polynomial of degree m Since dy and ay, are arbitrary, it is customary to choose them so that yi(x) or yx(2), whichever isthe polyno ‘nial, will satisfy the condition y(l)= 1. The resulting polynomial, denoted by Px), is known as the Legendre polynomial of degree n. The first Few of these are Rw Re no=ter = i Por=LSe —3n Pior= hase! —300' +3) Find a recurrence formula for the power series solution around x= 0 for the nonhomogeneous differential equation (2-+-)y" +xy=4+2 Dividing the given equation by 2°44, we see that x= 0 an owdinary pint and that 2) = + 2)(02 +4) is analytic there. Hence, the power series method! is applicable tothe entire qition, which, furthermore, we may leave in the form originally given to simplify the algebra. Substtating Eys. 27.5) through (27.7) into the given differential equation, we find that (P+ 4)120;+ 6ayx+ 12ay3?+ + MON Liar > Fn Dy. w+ (+ 2) Diy ead] Fala + age a? ta! eb yO or (Bag) + (2s + ag) +a + 480g + ay) + 6a + 805+ 05) + FANN Dag 4+ + Day 2% dy ad + 2+ (xe OPH Oats o Eaquating coeicients of like powers of x, we have 8a)=2, May tay= 1, 2as+48ay+a)=0, Gay + 800s ay Q In general, N= Day Mn+ 20+ Digs de 1= 0 oo) Which is equivalent ro fo 1 = —1 4, 3 2 Fas Dnt Mere BMD w (=7.4,_) Nove that the menemenee formula (3) isnot valid for n= 0 or n= I, since the coefficients of YP andl o270 LINEAR DIFFERENTIAL BQUATIONS WITH VARIABLE COEFFICIENTS. [CHAP 27 27.18. Use the power series method to find the general solution near x=0 of Cr dy ayaa? Using the results of Problem 27.14, we have that a, and a, ae given by (4) and a, for (n by (3) ILfollows from this recurrence formula that 6...) piven ‘Thus, nm ‘The ti series is the paticulr solution The the associated homogeneous equation (+ y+ ‘stand second series together represent the general solution of Find the recurrence formula for the power series solution around ¢=0 for the nonhomogeneous differential equation (d*y/dt®) + ty = e*" Here P(t) =0, QC) =, and 9()=e'*? are analytic everywhere, so 1=0 is an ordinary point, Substituting Eqs (27.5) through (27.7), wit replacing x nto the given equation, we find that Das Gast + agP bn 2000 Diya] last ate atta, Pleat Recall that e"* has the Taylor expansion qf ("about r= 0. Thus, the last equation can be rewritten as 2a.) + 46a, +4,) +P (2a, +4,) +--+ 0+ Day, ba, “on Equating coelficients of like powers of f, we have pata cnet o =. ba + Te general (n+ 2)in+ Vag #1 =eiA For N= 1, 2a a z pueecauue| Q re 2 0D re Dn Dt which isthe recurrence formula for 1 so Using the first equation in (1), we ean solve for a 27.17, Use the power series method to find the general solution near = 0 for the differential equation given in Problem 27.16, Using the results of Problem 27,16, we have a3=e'2 anda recurrence formula given by Bq. (2). Using this formula, we determine thatCHAP 27] LINEAR DIFFERENTIAL EQUATIONS WITH VARIABLE COEFFICIENTS. mn Find the general solution near x= of ¥”— (= Dy 429= ‘To simplify the algebra, we Fist make the change of variables cowresponding transformations of the derivatives of y 2. Brom the chain rule we find the ay dy dt dy ty dy dt ayy a dx dt dx a” dt fy ddr) d(dy)_ afar\de ay dy de dx\ de) dev dt) dtl dr Jax at? ‘Substituting these results into the differential equation, we obtain ty_jdy ae at tnd this equation is to be solved near £= 0, From Problem 27.3, with x replaced by 1, we see that the solution is we) Substituting ¢=.x— nfo this last equation, we obtain the solution to the original probfem as De eee eee eee {1-(x-'}- fe ee 27 ao" rs) Tao * 2) | a) 27.19. Find the general solution near x =—1 of y" +.’ + (2x— y= 0. ‘To simplify the algebra, we make the substitution *=y~(-1)=x4 1. Then, as in Problem, 27.18 (dyide)= (dvd and (d'y'de) = (yd), Substituting these results into the differential equation, we obtain a ai dy +4-p 2+ ar-3 av S4ar-3y ‘The power series solution to this equation is found in Problems 27.9 and 27.10 as (rede ede bie lea(rebe ede sorte ccas a5LINEAR DIFFERENTIAL BQUATIONS WITH VARIABLE COEFFICIENTS. [CHAP 27 Substituting back 1= + 1, we obtain asthe solution tothe original problem te Loeeay stot yt + iegeearedeory sdecenre] safecenedece nt edevr at race's] “ 27.20, Find the general solution near x= 1 of y+ (= 1) tence 41 Asin bien 2718, £°2 2 2°22 te given dierent equation maybe ad ts solution is (see Problems 27.16 and 27.17) vealigrt Jea(e Substituting back ¢=. 1, we obtain as the solution tothe original problem [i-deno's Jeafe-» xt ] Lgeptta nips hee yt-herep? oflecotebacvedecnt-Luie] 27.21, Solve the initial-value problem Y—(e-2y+2y=0; 92) y¥(2)=60 Since the inital conditions are prescribed at x=2, they are most easily satisfied ifthe solution tothe differential equation is obtained as a power series around this point, This has already been done in Eq, (J) of Problem 27.18. Applying the intial conditions directly to this solution, we find that ay= 5 und a, = 60, Thus, the solution is y sea eae (x =2) =x =2)" = 10(4— 27.22, Solve y" +.’ + 2x— Dy=0; MD) =2, ¥CL Since the initial conditions are prescribed at x= ~ 1 its advantageous to obtain the general solution to the ifferential equation near x =~ 1. This has already been done in Eq, (1) of Problem 27.19. Applying the inital conditions, we find that ay= 2 and a, = ~2. Thus, the solution is =A 1+2es story story! yeafpdeeseteentdaente-] Airey sdiven' eLivsty sonen'e-| peor ee x+h+2e4)? Zoey shot pt+ FD +24 DF Fat 30RDCHAP 27] LINEAR DIFFERENTIAL EQUATIONS WITH VARIABLE COEFFICIENTS. 273 27.23. Solve Problem 27.22 by another method, TAYLOR SERIES METHOD, An altemative method for solving initial-value problems rests on the assumption that the solution can be expand in & Taylor series about the initial pointy; 1, Peony o 2 en Lira ‘The terms y(¥g) and yx) are given as inital conditions the other terms 8%) ( ) can be obtained by successively’ differentiating the differential equation. For Problem 27.22 we have x= 1, ym) =») =2, and ¥G80)= YI) == 2. Solving the diferential equation of Problem 23.22 for y", we find that ye by @ We obtain y"0)= "CH by substituting =~ 1 into (2) and using the given intial conditions, Thus, YED== DYED =RED=IyCD= 1-9-1) =4 @ “To obtain 9-1), we diferente (2) and then substitute x)= ~1 into the resulting equation Thus, (2x- ty’ wo and YED==¥ED=Chy" CD21) BEY= Hyd SH 2) 44-22) 3D) o To obtain y¢-1), we eitferetate(# and then substitute xy=—1 into the resulting equation, Thus, yay =—ay"— (e+ Dy"—4y" 6 and YON SED D- BED + HYD aye) Fad 1) = 4-2) =8 ” “This process can be kept up indefinitely. Substituting Eq. (3), (9, (7), and the intial conditions into (J), we obtain, asbefore, 7 4 a 8 2+2eensdery + tory See's Gee D eet! +h e+ Tot (x +24 DF eaten! wep ehetn'+ yee) (One advantage in using this alternative method. as compared fo the usual method of first solving the differential equation and then applying the initial conditions. is thatthe Taylor series method is easier to apply when only the fies few terms of the solution are required. One disadvantage is thatthe recurrence formula cannot be found by the ‘Taylor series method, and, therefore, a general expression forthe nth term of the solution cannot be obtained. Note that this alternative method is also useful in solving differential equations without initial conditions, In such cases, we set ys) = ay and y'U%) =a}, Where dy and a, are unknown constants, and proceed as before, 27.24, Use the method outlined in Problem 27.23 to solve y” —2xy=0; 2)= 1, '@) Using Fa. (2) of Problem 27.23, we assume a solution ofthe form 09-224 Dey 9) LOa ay 2D ig ay wip FO 2) 4 ee 4 MP2) w From the differential equation, WHE 2e Y= 2VHDY, wR) =dy +2", Substituting into these equations and using the initial conditions, we find that 20)92)= 4 22) #200) =4¥@)+22W'2)= 40) +44) = 1624 LINEAR DIFFERENTIAL BQUATIONS WITH VARIABLE COEFFICIENTS. [CHAP 27 Substituting these results into Eq. (1), we obtain the solution as 2 a denny Zee ay! 1426-27 + heap Zee ayte pod ag 3 (x=) 27.28, Show that the method of undetermined coefficients cannot be used to obtain a particular solution of yi txy=2, By the method of undetermined coefficients, we assume a particular solution of the form y,= Age", where might be zero if the simple guess y,= Ay does not roquire modification (see Chapter 11), Substituting, into the iferential equation, we find (in — I)dox-7+ Agx™*E Ww Regardless of the value of iis impossible to asian Ay any constant valu that will stisty (J). It follows thatthe ‘method of undetermined coefficients is not applicable. ‘One limitation on the method of undetermined coefficients is that itis only valid for linear equations with constant coefficients, Supplementary Problems In Problems 27.26 through 27.34, determine whether the given values of are ordinary points or singular points ofthe given differential equations. 27.26. PUM. v=2, (0 2y" 4308s y+ V=D) y+ nan saogreny yen mi, s=-Kereby by ey 273. x= 0,8y"4y BBL x= 0:57" bay=0 2732. x=0.e'y"+Ginay ta9=0 2733, x=—L:Gr+ Dy" +62 Nort by +o y= 0 BIBL x=2,4(8— ay" (ee Dy +E —3e+ = 0 27,38, Find the general solution near x= (0 of y"~ y'=0. Check your answer by solving the equation by the method of (Chapter 9 an then expanding the result in power seties about x= 0. In Problems 2736 through 27.47, find (a) the recurrence formula and (b) the general solution of the given differential ‘equation by the power series method around, the given value of x IIA ¥=0; 4 wy=0 WH. x20, FI -Py=0 273K x=0; WHY +209 2B. x20; Wo 27.40. x=0; 74 2¢y=0 IMAL x=0; G2 Dy"bay'-y=0 2742 x=0; ¥¥—ay=0 2B x21; oy a7. =x8y"+ (042) 27.45. 27.46. yi-onby 27aT v=0, Yay ae 27.48, Use the ‘Taylor series method described in Problem 27.23 t0 solve y" 0; 0) (0) 27.49. Use the Taylor series method described in Problem 27.23 to solve y" = 2xv=2% x(1) =0, yA) =2,CHAPTER 28 Series Solutions Near a Regular Singular Point REGULAR SINGULAR POINTS The point x; is a regular singular point of the second-order homogeneous linear differential equation y+ Pony’ + Oly =0 (28.1) if.apis not an ordinary point (see Chapter 27) but both (x —x9)PGx) and (x ~x4)°QLx) are analytic at xp, We only consider regular singular points at = 0: if this isnot the case, then the change of variables 1 =x—xp will translate 4 t0 the origin. METHOD OF FROBENIUS Theorem 28.1. Wx=Oisan the form sular Singular point of (28.1), then the equation has at least one solution of = ax where A and a, (n=O, 1, 2, ...) are constants, This solution is valid in an interval 0.x R for some real number R. To evaluate the coefficients a, and A in Theorem 28.1, one proceeds as in the power series method of Chapter 27. The infinite series Sox =Sax Fay ta el tah ba, he (nay with its derivatives Daas?" A+ Day + + Dae + HOM Day EEO Mage OEM Dye Ph (283) 275 Copyright @ 2006, 1004, 1973 by The McGraw-Hill Companies, Ine. Click here for terms of use.276 ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT [CHAP 28 and Y= MA Dato? (+ Dane + Ot 2+ Da HO bn DE MDa, P84 OE MOE = Nar FOE MEME Maye we 8s are substituted into Eq, (28.1). Terms with like powers of x are collected together and set equal to zero. When this is done for x" the resulting equation is a recurrence formula. A quadratic equation in 2, called the indicial equation, arises when the coefficient of 2° is set to zero and ap is left arbitrary. ‘The two roots of the indicial equation can be real or complex. If complex they will occur in a conjugate pair and the complex solutions that they produce can be combined (by using Euler’s relations and the identity 4! = x%e*# 8) to form real solutions. In this book we shall, for simplicity, suppose that both roots of the 12 Aa, the method of Frobenius indicia equation are rea. Then, if} is taken as the larger indicial root, 2- always yields a solution WO= Sayre (285) to Fg, (28.1), [We have written ¢,(2,) to indicate the coefficients produced by the method when A= Ay.] If PG) and Q(x) are quotients of polynomials, it is usually easier first to multiply (28.1) by their lowest common denominator and then to apply the method of Frobenius to the resulting equation. GENERAL SOLUTION ‘The method of Frobenius always yields one solution (0 (28.1) of the form (28.5). ‘The general solution (see ‘Theorem 8.2) has the form 8) + cqys(x) where ¢, and ¢ are arbitrary constants and y3(x) is a second solution of (28.1) that is linearly independent from y,(3). The method for obtaining this second solution depends ‘on the relationship between the two roots of the indlicial equation, Case 1. If hy — 2g is not an integer, then y.@)= 09 Sa)" (286) where y3(+) is obtained in an identical manner as v(x) by the method of Frobenius, using Ain place of Py, Case 2, If Ay = Aa, then septa ea $ bo ean ‘To generate this solution, keep the recurrence formula in terms of & and use it to find the coefficients, (1 1) in terms of both 2 and do, where the coefficient dp remains arbitrary. Substitute these a, into Eq, (28.2) to obtain a function y(A, x) which depends on the variables % and x. Then 30) (238) Case 3. If Ay ta =N, a postive integer, then yO)=d yO)lnx + Fd,Q)e" (289) “To generate this solution, first try the method of Frobenius. with IFit yields a second solution, then this solution is y,(4), having the form of (28.9) with d_, ~ 0. Otherwise, proceed a in Case 2 to generate ¥Q.,2), whence a Va(2)= L(A — Ag) MAM, 0, (28.10) WO= FIA A a, 28.10)CHAP 28] ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT an 28.3. Solved Problems Determine whether x= 0 is a regular singular point ofthe differential equation yaa’ +2y=0 [As shown in Problem 27.1, x=0 is an ordinary pont of # singular point and Qay=t- ae Tar RF Neither of these functions is defined at x= 0, so this point 4 singular point. Furthermore, both spi= and x% =F ‘at -» 8 are analytic everywhere: the firs is @ constant and the second a polynomial. Hence, both are analytic at this point isa regular singular point.288. ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT [CHAR 28 Find a recurrence formula and the indicial equation for an infinite series solution around x = 0 for the differential equation given in Problem 28.4. {follows trom Problem 28.4 that v= 0/is a regular singular point ofthe differential equation, so Theorem 241 holds. Substituting Eqs. (28.2) through (28.4) into the left side of the given differential equation and combining coefficients of like powers of x, we obtain SBA ~ Day + 10Ray — a +98 + They + 10+ Day + a5 a4) + PTB MA m= Diy + 10+ My +g ag) = 0 Dividing by.x* and simplifying, we have [82 +2A— Tay +Xf(8H? + 18K + May tag) + +2804 +204 m)— ay tay} = 0 Factoring the coeflicient of a, and equating the coefficient of each power of x to ze70, we find (92+22— Ny o and, for n> 1, HA+m) ~ 112040) + ta, +a, =0 on - ae « * [40+ n)— 20+ 41)" @ Equation (2) isa recurrence formula for this differential equation From (), ether ay 82 42h-1=0 @ It is convenient fo keep ap arbitrary; therefore, we must choose 2. to satisfy (3), which isthe inicial equation, Find the general solution near x=0 of 8@y" + 10xy’ + (r= Dy= 0. The roots of the indicia equation given by (3) of Problem 28.5 are 2,=4, and 2, =—4. Since 2y~Ay=. jon is given by Eqs. (28.5) and (28.6). Subst othe recurrence Formula (2) of Problem 28.5 ‘and simplifying, we obtain the sol ‘Thus, Lyte and wior=ayel{ hee et a (: a a6 ) + into recurrence formula (2) of Problem 28.5 and simplifying, we obtain ‘Thus,CHAP 28] ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT 279 288, ‘The general solution is na tome) (ifr tos Jotpre(ictr bee (cata erp) here ky = eyap and k= cay, Find a recurrence formula and the indicial equation for an infinite series solution around x= 0 for the differential equation Dey" + Tate + Dy =3y=0 It follows from Problem 282 that x= 0 isa regular singular point of the differential equation, so Pheorem 28.1 holds. Substituting Eqs. 28.2) through (28.4) into the left side of the given differential equation and combining coefficients of like powers of x, we obtain PEMA 1+ Thay ETHIE mn Day +704 n= Diy 9 #704 wy Bag) 0 Bay] +28 [2R+ Ika, + Thay + 70+ Vay — 3a) + Dividing by x and simplifying, we have OR 4 Sh—3)ay 4 ON + RA May + Thal + FAM2O+ a9? + S04) 3+ TA = Diy 3+ Factoring the coefficient of ay and equaling each coefficient t zero, we find H+ 5h-3)ay=0 o and, for n= 1, 24m) = 1+ 43a, + WE Dy ar n=» 7 Rem —Ma+ meat 2 Equation (2) is a recurence formula or this differential equation, From (1) either ay =0 oF o is convenient fo keep ay arbitrary; therefore, We require 2. to satisfy the indicial equation (3). Find the general solution near x=0 of 2x¢y"+7xex+ Ly’ —3y The oot ofthe indicia equation given by (3 of Problem 28.7 are 2, solution is given by qs. (28.5) and (28.6). Substituting A=+ into (2) of Problem 287 and simplifying, we obtain ~2n-1) iene) ‘Ths, and )28.10. ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT [CHAR 28 Substituting 2=—3 into (2) of Problem 28.7 and simplifying, we obtain n= 4) a ay (2D) nQn—7) ‘Thus, ao sine a4=0, 4, = Ofer n2 4 Ta, norrar"{t ‘The general solution is Yee Hev00) 7,47 = ha"(i- dae is" 792 where ky = cya and = ead, Find the general solution near v=0 of 31°)" — a’ + ere Pa) =-1/G) and Qe) = 1/622} hence, x= Os epular singular point ad the method of Frobenius is applicable. Subsituting Pgs. (28.2 tough (25-4 no the ere equation and Simplifying, we have BEAR Nay HE BE 42a, + OBA NA+ Nog =O Dividing by x* and equating all coefficients to zor, we find EAL 1yay=0 o and A+? =4h+m-+ 14-0 (nz) @ rom (2), we conclude that the indcil equation is 3224+ 1=0, which has rots 2y = 1 and y= 4 Since 2, 7, =4, the solution is given by Egs. (28.5) and (28.6). Note that for either value of 2, (2) is satisfied by simply choosing a, =0, 2 1, Thus, ny=s and the general solution is si) Heaysee)= hye he where ky =eyay and k= ea Use the method of Frobenius to find one solution near x= 0 of x2" ay! +2 Here Pla) = Lx and Q(4)= 1, 80 ‘regular singular point and the method of Frobenins is applicable, Substituting Bgs. 28.2) through (28.4) into the left side of the differential equation, as given, and combining coefficients of like powers of x, we obtain SRB +22 1A Dad 42 1 Pa ag +o bP ML ya] Thos, o @ and, for 122, QL nay ayCHAP 28] ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT 281 28.11. ‘The stipulation n 22 is required in (3) because cy isnot defined for n= 0 or n= 1. From (1), the indicia equation =0. Thus, we will obtain oly ane solution ofthe form of (28.5); the second solution, is? =0, which has 008, 2y = Ay 3:09, wl have the form of O8 7 Substituting into (2) and (3), we find that ay=0 and dy=—(n? jay» Since a, =O, it follows that Furthermore, tnd, in general, a, wo Find the general solution near x=0 to the differential equation given in Problem 28.10, ‘One solution is given by (4) in Problem 28 10. Because the roots ofthe indicial equation are equal, we use Eq (28.8) {o generate a second linearly independent solution. Ihe recurrence formula (3) of Problem 2.10, augmented Dy (2) of Problem 210 forthe special case n= 1. From (2) a =0, which implies that O= a = ‘Then, from (3), Gey" “Grae asaya Salting these values into Fy (282), we have eer eas erent eaten eee cee meer Oy) ving ¢ 2 yt gt ae on "| (he 2V (+27 -—2 ys A+ arae ey ht aya 2y +o te OF 4D? ws o ‘hich isthe Form claimed in Eq, (28.7). The general solution is y= cyyy(a) 4 c2ys(0)28.12, 28.13. 28.14. ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT [CHAR 28 Use the method of Frobenius to find one solution near x= 0 of x2)" —xy by =0. ae Py) =—ix and Q(2)= Lv, s0.x= 0s a eaulr singular point and the method of Frobenius is applicable Sobsitung Eqs. 28.2) through (28-)nt het ie ofthe diferential equation, as given, and combining ceicients of ke powers of x, we obtain AAA Dag + Bay) + a MT 20+ + Ny Thus, (Pag =0 wy ane in general [d+ 20+ + 1a, =0 o From (J), the indicia equation is (.~ 1)? =0, which has soots 2; = = L. Substituting A= 1 into (2), we obiain ‘yO, whieh ingples thal dy =O, #2 1. Thus, 9) =a Find the general solution near x=0 to the differential equation given in Problem 28.12. (One solution is given in Problem 28.12. Because the roots of the indcial equation are equal, we use Fy. (28.8) {to generate a second linearly independent solution. The recurrence formula is (2) of Problem 28.12. Solving it for in terms of A, we find that a, =0 (n2 1), and when these values are substituted into iq. (28.2), we have yx) =agx Thus, and \hich is precisely the form of Eq. (28.7), where, fortis particular differential equation, 6,04) = 0.0 ‘The general solution is = ey Heys) = AC) +e In say, and k= 630, Use the method of Frobenius to find one solution near x= 0 of xy" + (x? — 2x)y’ + 2y Here so x=0 is a regular singular point and the method of Frobenius is applicable. Substituting, Eqs. (28.2) through (28.4) into the left side of the differential equation, as given, and combining coelicients of like powers of x, we obtain YQ? = 3A + Dag] +. YP — Ray + ag] + +e 30+) + 2]a,+ ht Day} Dividing by x, factoring the coefficient of a,, and equating the coefficient of each power of to zero, we obtain (2-34 2Dq=0 o tnd, in general, [0.++n) ~ 20.4 0) = ag + QF M= Diy 1 =0, 0% 1 m2 2 dewges 2d @ From (J), the indicial equation is A? ~3A-+2=0, which has roots 2 =2 and Ay. Since Ay — Rs integer, the solution is given by Eqs. (28.5) and (28.9), Substituting 4=2 into (2), we have a, 1a positive Vig 16CHAP 28] ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT 283 28.15. from which we obiain a Find the general solution near x=0 to the differential equation given in Problem 28.14, ‘One solution is given by (3) in Problem 28.14 for te indicial, root Ay =2. If we try the method of Frobenius withthe indicial root 2 = 1, recurrence formula (2) of Problem 28.14 becomes hich leaves aj, undefined because the denominator is zero when n= 1. Instead, we must use (28,10) to generate a second linearly independent solution. Using the recurrence formula (2) of Problem 28.14 to solve sequentially for fy (= 1, 2,3, im tems of A, we find Substituting these values into Eq, (28.2) we obtain afr -—toey Lyt sional. a5" *ia-n" apa | and, since R= yA eee eee OA WO. 2) = [o Dy x Jaen ] Then Fur wa.01-0[x +(h-Da* Inx and af ero-eimen eae Ine sctnig( so 5 “4 Jeafee eds ) soatevan{ioe ede] ‘This is the form claimed in Bq, (28.9), with lad + 0, 1, do= ay dh ‘The general solution is284 ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT [CHAP 28 28.16. Use the method of Probenius to find one solution near.x=0 of 2y" +ay’ + (2 = Dy=0. Here 1 P amt 009 s0.x=0 isa regular singular point and the method of Frobenius is applicable, Substituting Eas, (28.2) through (28.4) into the left side of the differential equation, as given, and combining coefficients of like powers of x, we obtain IE Dag) 104+ DE Hay +24 P= Maa + a) + SPM (Gem May tay a} t= ‘Thus, 02 ~1ay=0 o [04 DF May =0 Q and, For m2, [04+ m)?= I]t, + y-3=0, 0%, na (N22) o From (0), the iniial equation is 22— 1 positive integer, the solutions given by (28.5) and (289). Substituting and, in genera (e1 ) 5" Thus, 10 0D DT @ Le 28.17. Find the general solution near x =0 to the differential equation given in Problem 28.16. (One solution is given by (4) in Problem 28.16 forthe indicial root y= 1. If we try the method of Frobenius withthe indicial 1oot 2 =—1, recurence formula (3) of Problem 28.16 becomes which fails to define as because the denominator is zero when n= 2. Instead, we must use Eq. (28.10) to generate 4 scond linearly independent solution. Using Eqs. (2) and (3) of Problem 2816 to solve sequentially for ay(n=1,2,3, i terms of we find 0= a =as= a5= "and tt _. aaa” 1 aaa ‘Thus, yx) oo 1 : cafe ay of Graney” *ar5a4 OED | Since A= 2y=2e+ 1 -Aynxr= [acne lesa aE | Oa SASFCHAP 28] ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT 285 2 1 * Ing -—————"* 43) A+ SF 43) ao ne +543" ht A438) o inthe form of (28.9) with lA 40. “The general solution is 28.18. Use the method of Frobenius to find one solution near.x=0 of 22)" + (a? + 2x)" —2v= 0. Here so.r= 0 sa gular singular point and the method of Frobenius is applicable. Substituting Fys. (28.2) through (28-4) Into the left side ofthe differential equation, as given, and combining coefficients of like powers of x, we obtain MUO +h EA mT my? + (hm) ~ 2a, Qt Hyg) +o = 0 J 108 + 3A, + Ral + Dividing by x, factoring the coefficient of a,, and equating to zero the coefficient of each power of x, we obtain 2 +2-2ay=0 o and, for n> 1 [Q+m) +214) ~ Hay+ (+n Tay, =0 whit is equivalent Toeretes 2D a From (J) the indicial equation is 22+ A—2=0, which has roots Ay integer, the solution is given by Eqs. (28.5) and (28.9). Substituting i Which in turn yields 2 Since hy = 2g =3, a positive we obtain a, =[-14+ 304,1628.19. ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT [CHAR 28 and, in general, Hence, veomasfiny Which can be simplified to a Find the general solution near x= 0 to the differential equation given in Problem 28.18, (One solution is given by (3) in Problem 28.18 forthe indicial root 4, = 1. If we try the method of Frobenius with the indicial root 4g =~2, recurrence formula (2) of Problem 28.18 becomes * o hich does define all a(n 1), Solving sequentially, we obtain and, in general, ‘This is precisely i the Form of (28.9), with d_, =0 and d, = (-1)"agin!. The general solution is 9100) + e290) Find a general expression for the indcial equation of (28./), Since x=0 is a regular singular point; xP(x) and x°Q(x) are analytic near the origin and can be expanded in Taylor series here. TH, seo= Spat=p.t exten za Dividing by and 2°, respectively, we have ¥QU tae tax + PO)= porter tpt Oe)=quP tax tat Substituting these two results with Eqs. (28.2) through (28.4) into (28.1) and con AINA I)ay + Ragpy + ayqul + =0 which can hold only if ag? + (py D+ gol = Since ay s40 (ay isan arbitrary constant, hence ean be chosen nonzero), the indicial equation is 22 + y= Dat as oCHAP 28] ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT 287 28.21. Find the indicial equation of sy" +.xe"y’ + (8 — Dy = Of the solutio s required near x=0, Hse Po and weave woyee ateee® FOO) =x" +0x+0x7 +107 40x" + from which p= 1 and qy=—1. Using (1) of Problem 2820, we obtain the indicial equation as 2? ~ 1= 0. 28.22. Solve Problem 28.9 by an alternative method ‘The given ciferental equation, 229" — ay" + y a special case of Euler's equation DV by xt IM 4 bye" + bay + by = 600) o where 5 «7 is .constant, Euler's equation can always be transformed into a linea differential equation With constant coefcients by the change of variables zsinx or xs6 2 It follows from (2) and from the chain rule and the product rule of differentiation that a ee Using the method of Chapter 9 we find that the solution ofthis last equation is y= ce" + "95 Then using (2) and noting that = (€)!°, we have as before, 28.23. Solve the differential equation given in Problem 28.12 by an alternative method, ‘The given ciferental equation, xy" ~ ay +-»=0, ia special ease of Fulers equation, () of Problem 2822 Using the transformations (2), (3), and (4 of Problem 28,22, we reduce the given equation to fy “The soliton to this exatinn is (sen Chapter 9) y= yet estet Then, using (2) of Profem 98 79, we have fr the solution ofthe original differential equation yeepebegelnx as before.288 ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT [CHAP 28 28.24. Find the general solution near x=0 of the hypergeometric equation x oy IC IB E Del’ ABy= where A and B are any real numbers, and Cis any real nonintegral number. Since x= (isa regular singular point, the methex! of Frobenius is applicable, Substituting, Eqs. (28.2) through (28.4) into the differential equation, simplifying and equating the coefficient of each power of x 10 zer0, We obtain 22+(C-DA=0 o as the indicial equation and G+ mAtn+A+ B+ AB, Genehhensc) ea asthe recurence formula, The roots of (J) are 2, = Oand Ay = 1 C, hence, by Re = C= 1. Since Cis not an integer, the solution of the hypergeometiic equation is given by Eqs. 28.5) and 286), Substituting A= 0 into (2), we have N+ AB) + AB iebneey Which is equivalent ro em Bem, ‘Thus Ae DB 2+ A+ 2B +2) BIC +2) and y\(2) = agFtA, B, C:2), where 4 AB, MAT DAB +D 2 ie mac+h AA +A 2)B(B + KB +2) ~SICIe+C+2) FIA BC.) ‘The series FUA, B,C.) is known as the hyporgeometic series; it canbe shown that this series converges foe “I €v #1 Itis customary to assign the arbitrary constant ay the value 1. Then yi(8)= VA, B; C;x) and the hypergeometric series isa solution ofthe hypergeometric equation “To find y3(x), we substitute X= 1 — C into 2) and obtain Cin th AtB-C)s AB 4 2-Cynd follows that Solving fora, in terms of ap, and again setting ay y20) cay ‘The general solutionCHAP 28] ‘SERIES SOLUTIONS NEAR A REGULAR SINGULAR POINT 289 Supplementary Problems In Problems 2825 through 2833, ind wo inal independent solutions to the given ferential equations 28.25. 2y”—ay’ + (1 —xjy=0. 2826, Dry" + (2 —xyy ty=0 28.27. 28.28, xy"+y'-y=0 28.29, 2830, 2y"+ (yy 2A3l. ay’ (e+ Dy-y=0 2A32,_ Avy" + (Ar +2 +GN— by 2838 y+ (A —Aeyy"—(r—4ayy=0 In Problem 2834 through 2838, ind the general solution othe given equations using the method described in Problem 28.22 BH. Avy" +ay’—y=0 2835, vy"-any'+4y=0 2836, 2eV"+ Lay’ H4y=0 2837, xy" 0 2838. y"—6y’=0CHAPTER 29 Some Classical Differential Equations CLASSICAL DIFFERENTIAL EQUATIONS Because some special differential equations have been studied for many years, both for the aesthetic beauty of their solutions and because they lend themselves to many physical applications, they may be considered classical. We have already scen an example of such an equation, the equation of Legendre, in Problem 27.13. ‘We will touch upon four classical equations: the Chebyshev differential equation, named in honor of Pafnuty Chebyshey (1821-1894): the Hermite differential equation, so named because of Charles Hermite (1822-1901): the Laguerre differential equation, labeled after Bdmond Laguerte (1834-1886); and the Legendre differential ‘equation, so titled because of Adrien Legendre (1752-1833). These equations are given in Table 29-1 below Table 29-1 (Note: n=0, 1.2.3, 00) Chebyshev Differential Equation (1-2) y*aay ty=0 Hermite Differenial Equation y= ey’ +2ny=0 Laguerre Differenial Equation wy Hay taps 0 Legendre Differential Equation | (1—x3)y" ~2xy+ntn + y= POLYNOMIAL SOLUTIONS AND ASSOCIATED CONCEPTS ‘One of the most important properties these four equations possess, is the Fact that they have polynomial solutions, naturally called Chebyshev polynomials, Hermite polynomials, etc There are many way’ to obiain these polynomial solutions. One way is to employ series techniques, as discussed in Chapters 27 and 28. An alternate way is by the use of Rodrigues formulas, so named in honor of (, Rodrigues (1794-1851), a French banker. This method makes use of repeated differentiations (see, For example, Problem 29.1). These polynomial solutions can also be obtained by the use of generating functions. In this upproach, infinite series expansions of the specific function “generates” the desired polynomials (see Problem 29.3), It should be noted, from a computational perspective, that this approach becomes more time-consuming the further along \we go in the series 290 Copyright @ 2006, 1604, 1973 by The McGraw-Hill Companies, Ine. Click here for tems of use.CHAP 29] SOME CLASSICAL DIFFERENTIAL EQUATIONS. 291 ‘These polynomials enjoy many properties, orzhogonality being one of the most important. This condition, which is expressed in terms of an integral, makes it possible for “more complicated” functions to be expressed in terms of these polynomials, much like the expansions which will be addressed in Chapter 33. We say thatthe polynomials are orthogonal with respect to a weight fnction (see, for example, Problem 29.2) We now list the first five polynomials (n= 0, 1,2, 3, 4) of each type: + Chebyshev Polynomials, T(x) © Te)=1 © Tex © Tyay= 2-4 © TG) =48 38 © Tya)=88- 82 +1 '* Hermite Polynomials, H,(x): ° HQ)=1 © Hy@)=2¢ © AQ) =42-2 © Ha) =80— 12 © Ha(x)= Wort 482 + 12 + Laguerre Polynomials, 1,0) © L)=1 © Laj=-x+1 ° L4Q)=8-4e+2 © LQ) =-8 +98 — 18046 © Lya)=xt~ 1608 +7242 960+ 24 + Legendre Polynomials, Pa): © Poxy= 1 © Pax L Po) = 360° -) © Rey=16r -3y P= 2 a 829.2. 298. SOME CLASSICAL DIFFERENTIAL EQUATIONS. [CHAP 29 Solved Problems Let n=2 in the Hermite DE. Use the Rodrigues formula to find the polynomial solution, ‘The Hermite DE becomes y"~ 2xy/ + 4y = 0. The Rodrigues formula for the Hermite polynomials, Hy(x), is aiven by Letting n=2, we have Hy(x)=(-D'e" 2 This agrees with our listing above and via direct substitution into the DE, we see thar 4X {s indeed a sotution, ‘Notes: 1) Any non-zero multiple of 4x? ~2 is also solution. 2) When = 0 in the Rodrigues formula, the “O-th Derivative” is defined as the function itself. That i, acne & Ho=cyte’ Se Given the Laguerre polynomials L(x) = -x-+ 1 and L(x) =? ~4+ 2, show that these two functions are orthogonal with respect to the weight function * on the interval (0, i Jics+n o8 —41+2€%ae=0.This mega is inde zero, a is ved by negation by pars and applying Lvospta’s Rate Using the generating function for the Chebyshev polynomials, 7,(x), find Ty(x), 7y(3), and 736). ‘The desired generating function is given by Using long division on the left side of this equation and combing like powers of # yields (e+ ont! +28 Det. Hence, Z(2)= 1. T(x) = x, and 73(x) = 2°— 1, which agsees with our list above, We note that, due tothe nature of the computation, the use of the generating Function docs not provide an efficient way to actually obtain the Chebyshev polynomials. Tet n= in the Legendre DI yi tt (9) =}: a +3) aon ‘The DE becomes (1-7) y"—21y’+20y=0. Taking the first and second derivatives of Py(x) we obiain (105x" ~15). Direct substitution into the DE, followed by collecting like 2 =X )P%0) 240) + 20P “The Hermite polynomials, f(x), satisfy the recurrence retarion ysl) Verify this relationship for n =3. uty (x) ~ 2H, 0).CHAP 29] SOME CLASSICAL DIFFERENTIAL EQUATIONS. 293 I'm =3, then we must show that the equation Hy(x) = 2H. (x) ~ GH(x is satisfied by the appropriate Hermite polynomials: Direct substitution gives Jeet — 4802 + 12 = (2H) BE ~ 128) O49, ‘We sce that the right-side does indeed equal the left side, henoe, the recurrence relation is verified, 29.6. Legendre polynomials satisfy the recurrence formula (0+ DP a4 10) ~ Girt DxPy(0) + nP 0) = Use this formula to find Ps Lasting and solving for P30), we have B(x) (Bx? ~ 70x? +150). (9xP,{x) ~ AP). Substituting for Px) and for Pc), we have P(x) 29.7. Chebyshev polynomials, 7, (1), ean also be obtained by using the formula T,(x) = costa cos"'09). Verity this formula for 7364) = 2 — 1 Letting m=2, we have cos(2cox"H), Let a= cos"). Then cos(2a) = cosa ~sin*(a) = cosa ~ (1—c0s*(@)) = 2 cosa) -1, But if = cos (x), then x= cos(a), Hence, cos(2 eos '(x}) = 2° 29.8. The differential equation (I~ x2)y" + Axy’ + By =O closely resembles both the Chebyshev and Legendre coquations, where A and B are constants. A theorem of differential equations states that this differential equation has two finite polynomial solutions, one of degree m, the other of degree 1, if and only if A=m-+n—1 and B=—nn, where m and n are nonnegative integers and n+ m is odd For example, the equation (1 = 2)y" +h’ 6y=0 has polynomial solutions of degree 2 and 3 y= +30? and y=x+4 (these are obtained by using the series techniques diseussed in Chapter 27) We note here that A=4=m-+m—1 and mn necessarily imply that m=2, 1=3 (or conversely), Hence our theorem is verified for this equation Determine whether the three following differential equations have two polynomial solutions: (=x )y" + 6x9" = 12y= 05 b) (1 = YY" +9" + 8Y= 05 ©) (= 24)y" = ay’ + 3y= 0. 4a) Here A=6=n-+m-~ 1, B=-mn=—12 implies m=3, n=; hence we have two Finite polynomial solutions, one of degree 3, the other of degree 4 by. Here A=1 and B=8; this implies m=2, » ‘one polynomial solution, of degree 2.) “4; therefore, we do not have two such solutions, (We will have. ©) Since A=-1, B=3 implies m=, 1=—3, we do not have wo polynomial solutions to the diferent equation Supplementary Problems 29.9, Verify H,(x) and Hy(x) are orthogonal with respect tothe weight function &°F on the interval (~=, =) 29,10, Find Ha) by using the recurrence formula Hy (x) = 2xH,(X) ~2nFl, 0) 29,11, The Rodrigues formula for the Legendre polynomials is given by294 SOME CLASSICAL DIFFERENTIAL EQUATIONS. [CHAP 29 Use this formula to obtain P(x). Compare this tothe results given in Problem 29.6 20.12 Find (2) hy following the provesties given in Problem 20.6 29.13 Tollowing the procedure in Problem 297, show that cos(3 cos'a)) = 48° 29.14 Chebyshev polynomials satisfy the reursion formula Ty s(0) — 2XT (0) + Ty -1(0)= 0. Use this result to obtain 7400). ‘Show that this is tue for Py. 20.18. Legendre polynomials sas the contin f Pye = 29.16. Laguorre polynomials satisfy the condition Je*(L,(x))*de =(n!), Show that this is true for L3(). 29.17. Laguerre polynomials also satisfy the equation L(x) ~mL_y(0) + my t)=0. Show that this is true for L(x) 20.41% Gora 6) by ane she ton et? SALLE 2918. conte te ops ction “(0h tem 180, 2,3 The pom ee fom i equation ae called Associated Laguerre polynomials, and are denoted L(x). Find L3(x) and LQ, 29.20. Determine whether the five following differential equations have two polynomial solutions; if they do, give the degrees of the solutions, a) (1 =a°}y"+50y’—Sy=0, b) (I =a jy" + 8ty’= 18y=0, 0) (1=we)y +207 + 10y=0; (1 -x3)y"+ Mixy — S6y= 0, e) (1—)y"+ Lay’CHAPTER 30 Gamma and Bessel Functions GAMMA FUNCTION ‘The gamma function, U(p), is defined for any positive real number p by roy= [area ou) CConsequenty, (1) = 1 and for any postive real number p Tp+)=0rw) (aoa) BEER wren p=n, positive integer Ft =n! (203) ‘Thus, the gamma function (which is defined on all positive real numbers) is un extension of the factorial function (which is defined only on the nonnegative integers). Equation (30.2) may be rewritten as rin)-4tr+p (30.4) which defines the gamma function iteratively for all nonintegral negative values of p. T(0) remains undefined, because Tip+t) Ti+) > lim Fp) = tim, and tiny Pp) = lim I then follows from Eq. (30.4) that Pp) is Table 30-1 lists values of the gamma function in the interval 1 < p <2. These tabular values are used with Eqs. (30.2) and (30.4) to generate values of Tp) in other intervals. values of p. BESSEL FUNCTIONS Tet p represent any real number. The Bessel function of the frst kind of order p, Jy(t). is (x 1,40) LFourpet (30.8) 295 Copyright @ 2006, 1694, 1973 by The McGraw-Hill Companies, Ine. Click here for tems of use.296 GAMMA AND BESSEL FUNCTIONS [CHAP 30 ‘The function J,(x) isa solution near the regular singular point x= 0 of Bessel’s differential equation of order p: By" ix 1 (2 -phy-0 (306) In fact, J, (4) is that solution of Eq. (30.6) guaranteed by Theorem 28.1. ALGEBRAIC OPERATIONS ON INFINITE SERIES ‘Changing the dummy index. The dummy index in an infinite series can be changed at will without altering the series. For example, L ‘Change of variables. Consider the infinite series ¥, ———. If we make the change of variables j= k++ 1, GSD! or k= f= I, then eo ‘Note that a change of variables generally changes the limits on the summation. For instance, if) + 1, it follows that j= 1 when &=0,j=c: when k=, and, as Kuns from 0 to %, j runs from 1 10 The two operations given above are often used in concert. For example, 1 a 2a he Here, the second series results from the change of variables j=£+2 in the first series, while the third series is the result of simply changing the dummy index in the second series from j to & Note that all three series equal Solved Problems 30.1. Determine T3.5). It follows from Table 30-1 that F(1.5) = 0.8862, rounded to four decimal places, Using Bg. (30.2) with we obiain 76.5) =(2.5)F(2.5) But abo from Eq. (302), with p=1.5, we have F25)=(1.5)TU1.5). Thus, TBS) = 2.5/1.5) M.S) = G.75)0:8862)=3 32: 30.2. Determine (0.5). It follows from Table 30-1 that [(.5) = (18862, rounded to four decimal places, Using Ba. (30.4) with p=! we obtain T(0.5)= 21.5). But also from Eq. (30-4), with p=0.5, we have T(-0.5)=-270.5) Thus, 1-05) = (212) PU.) = HO8862) = 3.5448.CHAP 30] GAMMA AND BESSEL FUNCTIONS ‘Table 30-1 The Gamma Function (1,00 << 1.99) * Te) x Te) * Tw © Te) 1.00 | 1,00000000 | 1.25 | o,90640248 | 1.50 | 0862 2603 | 1.75 | 091906259 ror | 099432585 | 1.26 | ososso712 | 1st | oss6soi69 | 1.76 | 092137488 102 | 098884420 | 1.27 | 0,90250306 | 1.52 | 088703878 | 1.77 | 092376313 103 | 0,98354995 | 1.28 | o.9007 1948 | 1.53 | os7s76s | 178 | 092622731 Log | 097843820 | 1.29 | os900.4159 | 1.54 | osssi 7766 | 1.79 | 092876749 105 | 097350427 | 130 | 089747070 | 1.55 | ossssesss | 180 | 093138377 106 | 096874365 | 131 | 089600818 | 1.56 | 088963920 | 181 | 093407626 Lo? | ase szo4 | 132 | oavdeaod | 157 | oaseday7s | Ls2 | aysos 4sas 108 | 095972531 | 1.33 | 08933 7805 | 1.58 | ogoi4i9ss | 183 | 0.9396 9040 109 | 09584599 | 134 | 080221581 | 1.59 | oge2a2821 | 1s4 | 09426 1236 110 | 09513 5077 | 135 | oso si44 | 1.60 | ogsasisas | 18s | ao4s6 118 Li | 094739550 | 136 | og901 8453 | Ler | ogo¥68oe1 | 186 | ao4868704 112 | o943s9019 | 137 | 08893 1351 | 1.62 | 089592367 | 187 | a9s184019 113 | 093993145 | 138 | ossss37is | 1.63 | 099724423 | 188 | 095507085 L14 | o93641607 | 139 | 688785129 | 164 | 089864203 | 189 | 095837931 115 | 0,98304093 | 1.40 | 0,8872.6382 | 1.65 | 0.9001 1682 | 1.90 | 096176583 116 | 092980307 | 141 | oss67 6466 | 1.66 | 09016687 | 191 | 096523073 117 | 092669961 | 142 | 08863 5579 | 1.67 | 090829630 | 192 | 097877431 Lis | 092372781 | 143 | osso3624 | 1.68 | o9osoo103 | 193 | 0.9723 9692 119 | oszossso4 | 144 | assssasos | 1.69 | 090678182 | 194 | o9760 9801 120 | ogisies74 | Las | assseoi3s | 1.70 | 0.9086 3873 0.9798 8065, 121 | o91ss7e9 | 146 | assseosss | 1.71 | 0.9105 7168 019837 4254 122 | 091310595 | 147 | 088563312 | 1.72 | 091258058 0.9876 8498 123 | asto7sase | Las | oass7aov0 | 173 | ov460ss7 S917 HL 1.24 | 090852106 | 149 | oseso4si3 | 1.74 | 0.9168 2003 0.9958 1326 30.3. Determine F(-1.42). 1 follows repeatedly from Fy, (30.4) that re1a)= From Table 30-1, we have F(1.58) = 0.8914, rounded to four decimal places; henee oso Prove that '(p +) pl(p), p> 0. Using (30.1) and integration by parts, we have ‘The result lim, .« #€"=0 is easily obtained by frst writing re" as rie" and then using L’ Hospital's rule, nL) ‘Tan0.42(058) Tes b= [0 ede=linf{ sted wef [ + [reas] =lim(-re" +04 pfx eae= php) 297305. 30.7. GAMMA AND BESSEL FUNCTIONS [CHAP 30 Prove that P(1)= 1 Using Eg (30.1), we find hat Using Problem 30.4 with Prove that if p. ‘The proof is by induction. First we consider we have 1.8 positive integer, then M+ 1 and then Problem 30, T+ 1)=1M)=1d)=1=1! Next we assume that Pn + 1) n! holds for n= K and then try to prove ils validity for TH+ 1) +1]=+ E+ 1) (Problem 30.4 with, =k He) (Grom the induction bypothesis) kent +) ‘Thus, F(n-+ 1) =n is true by induction. "Note dat we eat no use this equality 0 define 0” o=T0+1)= Prove that M(p ++ 1) = (p+ Mp +k-D~(p + 2p + DEP + D. Using Problem 30.4 repeatedly, where first pis replaced by p +k then by p+k~ 1, ete, we obtain Tiptk +0 p++ I=(p+bMP+h = (pe Mp +k Y= (p+ (pe k= DMP +k=1) (p+ kip +k—1)-(p+ 2p + DEP +) Express [ede asa gamma function = 2! und dy Let =; hence Substituting these values int the integral and noting tha as x goes From 0 to so does 2, we have fe va-fe(} eae rede +3) 1 ‘The last equality follows from Bq, (30.1), with the dummy variable x replaced by z and with P= Use the method of Frobenius to find one solution of Bessel’s equation of order p: Ry ay 402 -pay=0 Substituting Es. 28.2) through (28.4) into Bessel’s equation and simplifying, we find that UE pag tat HO + WP — play tt + 2) — plas + a6} + FAM My? play ty a) toe =OCHAP 30] GAMMA AND BESSEL FUNCTIONS 299 ‘Ths, 2=paq=0 (0+ IP =p = o and, in general [0-+ nl? plan + da.9=0, oF 1 4 aos Q) Gaga SD 2 ‘The indicial equation is 22 —p* ‘Substituting % ) which has the roots 2; = p and 2 =—p (p nonnegative). into (1 and (2) and simplifying, we find that = 0 and 1 a, (N22) pen - Top THs aPsy cay =m HED Sipe wipek-D rape” “ nor faror[arar] eae It is customary to choose the arbitrary constant ay a8 1 <1. Then bringing aa” inside the brackets Tipp wing ox and summation in (3), combining, and finally using Problem 30.4, we obtain coy "Eth D, ayes (Sih =J,00) RT perked 30.10. Find the general solution to Bessel’s equation of order zero, For p=0, the equatcn is 2)" +39/+°¥=0, which was solved in Chapter. By (4) of Problem 28 10, one solution is oe vd 1 1 "T0+1) ¥1(8) =Jo(2)A second solution is [see (1) of Problem 28.11, wit Changing n tok, ng Problem 30.6, and letting a, =1 a8 indicated in Problem 309, it follows that 44) again chosen to be 1] x (1), x y= Seine | yA fied Je fiat * iss [sax zor 3) Fant 2300 GAMMA AND BESSEL FUNCTIONS [CHAP 30 which is usually designated by Nox). Thus, the general solution to Bess Y= eyela) + eC. [Anosher common form of the general solution i ahlained when the seen Tneady fvleperulent soliton ie not taken fo be No(x), but a combination of Ny(x) and J(2), In particular, if we define '= equation of order zero is KO= TIN + Ma} o where is the Euler constant defined by yotimfietete+1-ine |=o.s7721565 ate then the general solution to Bessel's equation of order zero can be given as y= eJ(x) + Yo 30.11. Prove that int rene! IT(pe k+l QDR pitt (p+k Writing the k=0 term separately, we have citebe 4, ¢_ciren PORT ps Re” AT FET p RD Which, under the change of variables j= ~ 1 becomes poy + De" +! GHD p+ J1+) AGT p+ I+? DUD T+ Dx ips J+) {he desired result Follows by changing the dummy variable in the fast summation from tok. 30.12, Prove that (eapetirs ip+k+2) yep? PAD (p+k+D) Make the change of variables j= A+ 1 piers Mp + kD) 7j-D\Tp+ a+) Now, multiply the numerator and denominator in the last summation by 2j, noting that jij~1)t=jt and DPQ) = 20", The result is ‘Owing tothe factor jin the numerator, the las infinite series is not altered ifthe lower mit in the sum is changed from j= 1to=0. Once this is done, the desited results achieved by Simply changing the dummy'index fom 10k.char 301 GAMMA AND BESSEL FUNCTIONS 301 30.13. Prove That “fx°'y, al (ole, 00. ‘We may ditlerentate the seres lor the Bessel function term by term. hus, rine) Noting that 20k + p+ 2)=2(k+ p+ Dk + p+ 1) and that the factor 2&-+ p+ 1) cancels, we have Pres ped dont rm 2h 7, 09 U0 Fl os 100) For the particular case p= 0 it follows that a Ape ool=xh(a 1 gel eeOl= al 00) a 30.14, Prove that 2/0) = p/,(4) —AI p41). We have PU Se RSET PEER = piper - 2 rape Using Problem 30.12 on the last summation, we find a Mol = MoT pee yp a THTips Fe Fe rp + ead, 20) For the particular case p = 0, it follows that /§( xh(2), oF J) =e) o 30.18. Prove that x/(4) PIA) + XI p-4().302 GAMMA AND BESSEL FUNCTIONS [CHAP 30 Multiplying the numerator and denominator inthe second summation by 2(p-+) and noting that (p+ AT p+ 4) =Tip+k+ 1), we find cvtcp? "ATip+keD) A 5 ci ie 2p + ky? 2 Prep rk) § MCR DM yey Aare = Cntxpe tx? DFE Tip+ eed s) xd 30.16, Use Problems 30.14 and 30.15 to derive the recurrence formula, 20. Jpal ¥ 40) J, Al) Subtracting the results of Problem 30.15 from the results of Problem 30.14, we find that = 2phf0)—Xly_ (0) Hp A) {Upon solving for J,y(0), we obtain the desired result. 30.17. Show that y =.1/,(x) is a solution of xy" —y’ —2/5(4) =0. First note that (x) is.a solution of Bessel’s equation of order one 22/0) +d} + 02 = Da) =0 o [Now substitute y = x(x) into the left side of the given differential equation: abe GNT = beh OO 2H) = ae +a FEI Wie) +o] 20) But ie h(x) (by (J) of Problem 30.14), so thatthe right-hand side becomes AIO) + BI{C0) — HO) — 20 +2 09 = 2IF 9 +I) + 0 the last equality Following from (1. 30.18. Show that_ y= VirJ,, (2) is a solution of a2)" + (2 — 2)y=0. ‘Observe that J5:(2) is a solution of Bessel's equation of order 3: serv son(e-2)re0 o Now substitute y=5r/, (x) into the left side of the given differential equation, obtaining Everly 96a)" +6? — 205, 964) ox fteny seen PZT] ) UE a0) a! “seo os the at uit Fllosing from (2). Ths (8) satis the ge dif equstonchar 301 GAMMA AND BESSEL FUNCTIONS 303 30.19. 30.20. 30.21. 30.22. 202. 3024. 30.25, 30.26. 30.27 30.29. 30.30. 3031 Supplementary Problems Find P26) Find P14), Find F.14, Find P26), Find F138. Express [fer asa gamma fonction Bahute [eas = planet! cy thal ao Prove hal Foster REET ree @ that Lbs (2-2, Prove that EL (2)=—2% 8) Hint: Use Problem 30.11 Prove tht Jp 08) Jp 10 = 200) (Prove thatthe derivative of 32 (0300) + Tis 0, Hint Use (1) of Problem 30.13 and (1) of Problem 30.14, (H) Evaluate f)x03(x) dr in terms of Bessel functions. ‘Show that y= xJ,(x) is a solution of xy" — xy’ + (14.2 =P} =0. Show that y= 22/3) is a solution of xy" ~3y' 4a =0