Contents Preface Preface to the Second Edition R 1 First Order P.D.E, 1.1 Curves and Surfaces 12 Genesis of First Order PDE. 13 Classification of Integrals 14 Linear Equations of the First Order 1.5 Pfaffian Differential Equations... 1.6 Compatible Systems ee 1.7 Charpit's Method . 1.8 Jacobi's Method 19 Integral Surfaces Through a Givon Cure |... 1.10 Quasi-Linear Equations : ae, AL Non-linear First Order P.D.B. 2 Second Order P.D.E. 21 Genesis of Second Order PDE... 2... 22 Classification of Second Order PDE. 23 One Dimensional Wave Equation 23.1 Vibrations of an Infinite String... | 2.3.2 Vibrations of a Somb-infinite String... 1. 88 2.3 Vibrations of a String of Finite Length |. |. 90 23.4 Riemann’s Method a 2353 Vibrations of a String of Finite Length (Method of Separation of Vacate) 24 Laplace's Equation... . 241 Boundary Value Problems... . 24.2 Maximam and Minimur Principles . SS BOBBED oon fgaaa 101 = 102CONTENTS ‘The Cauchy Problem . ‘The Dirichlet Problem forthe Upper Half Plane - 106 ‘The Neumann Problem for the Upper Half Plane 107 ‘The Dirichlet Problom for a Circle. . 108 ‘The Dirichlet Exterior Problem for a rele... 112 ‘The Newmann Problem for a Circle»... .-. 2 ‘The Dirichlet Problem for a Rectangle... . . 11d 2.4.10 Hamack’s Theorem +. AS 2.4.11 Laplace's Bauation - Green's Function. 16 2.4.12 The Ditichlet Problom for a Half Plane us 24.13 The Dichlet Problem for a Circle se 19 2.5. Heat Conduction Problem i, ele 25:1 Heat Conduction - Infinite Rod Case 2.120 | 26.2 Heat Conduction - Finite Rod Case . 12d 2.6 Duhamel’s Principle ae | 26.1 Wave Equation ge 262 Heat Conduction Equation - 2.7 Classification in the Case of r-Variables . | 28 Families of Equipotential Surfaces . 29 A. Pourier Transforms and Integrals 139 Kelvin’s Inversion Theorem : ‘AL Fourier Integral Theorem... = 139 ‘A2 Properties of the Fourier Transform : 139 ‘AS Exempkes ...- 40 ‘Ad Convolution Theorem pecme eee ui AS Vibrations of an Infinite String «2.22... Mi B Additional Problems Suggested Reading Chapter 1 First Order Partial Differential Equations 1.1. Curves and Surfaces - ‘This section briefly describes two geometrical objects, viz., curves and surfaces. ‘They play a very important role in the study of partial dif ferential equations (Curves in space : A curve may be specified by means of parametric equations. Sappoce fi, fa and fy ere continuous functions of a con tinuous variable ¢ which varies in an interval J.C JR, then the three equations x = fa(t),y = Jal), 2 = fa(0) represent the parametric equa- tions of a curve in three-dimensional space, ‘A mandard parameter i the length ofthe curve mesoured rom some fixed point on the curve. In sueh cases, at times, the symbol #13 used instead oft. a Note : The condition that the parameter t i the length of the curve is) Hi + Jf + J2 =1, where prime denotes differentiation. Bxample 1.1.1 ; The simplest example of @ curve in space isa straight line with direction cosines (1,2, 2) passing through a point (25, 9) with parametric equations etls.y=ttmsz=s4ns, 9€R. Bxample 1.1.2 ; A right circular hele is space curve lying on a circular eyinder and is given by the following parametric equations em acosuty = asinwt,z=kt, CEM, 12 CHAPTER 1. FIRST ORDER PD.B, ‘where 6, and k are constants. a Surfaces : A point (2,y, 2) in space is enid to lie on a surface ifthe coordinates 9 and 2 satisty F(ay,2)=0, (Qa) where F is a continuously differentiable funetion defined on a domain in RE, Thereforo a eurface is the locus of a point moving in space with “two degrees of freedom”. Consider a sot of relations of the form (0). a2) ‘Then for exch pair of values of w and v there coreepond theo numbers y and x and hence a point (s,y,2) in space. Tt i, however, not true that int in space corresponds to a pair wand v. Ifthe Jacobian Biel PAE Pn an vs Micah = Dede ~ dep #9 thon tho frst two equations of (1.13) can be solved to express u and v ss functions of» and y lecally, say, w= Mey) and v = p(e,y) by the inverse funetion theorem. Then ‘Wand # are determined once x and y are known and the third equation in (1.1.2) gives a value of >for theso values of 2 and y. ‘That ia PMG, y), ula) ‘hich is thus a functional elation between the co-ordinates x, y and 28 in (11.1), ‘Thus any point (x,y, 2) determined from Equations (1.1.2) always lies on a fixed surface. For this reason, equations of the type (11.2) are calle parametric equstions of the curface, How ‘ever, paremetric equations of a surface ate not unique as the following ‘example demonstrates. Example 1.1.3 : The parametric equations i(uey = (uv) z P= asinwoosy, y = asinusine, 2=ac0su, and a-e) (ev) 200 Trey P= MeN Fm TTT where a is a constant, both represent the surface 22 4 y? + 2? = o? which is a sphore. a Lets now go back to the equation of the curve (), v= Salt), 2 = Jolt) GENESIS OF FIRST ORDER P.D.E. 3 On stntmting een fad fw obi rate $2) = 0. Similarly, from fi and fe we get U(s, 2) = 0. Hence a curve can be thou caste plat sence el oa fe parm ‘in he eqn of cup eg fe ste eae hee tue fed pe, then (28,22) ae ta ton in te tage oi ce a ut Suppose that the curve ©: (1(s),y(s), 2(s)) lies om the surface $ ‘whose equstion is F(z, y,2) = 0. Then F(x(s), u(s), =(3)) (On ciflerentiating with respect to 2, we get oF ds ae ‘This isthe condition thet, the tangent to the curve Cat the point (6,12) is perpendieular to the line with direction ratios (9E, 3°, $2), Ie may be noted thas the curve C's arbitrary excopt for Une fate that it passes through tho pint P and lies on the given surface. There- fore we can conclude that the lino whose direction ratios are given by (3E. 3°, $F) is perpendicular to the tangent Lo every curve lying on S and passing through P, and therefore they give the diection ratios of the normal tothe surface $ atthe point P 1.2 Genesis of First Order P.D.B. By « partial differential equation (p.d.e), we mean an equation of the form LUE Yat, 8, Bey Boy > Oa 0, (12.1) involving two or more independent variables 2, y, t ete., one dependent variable 0(r,y,¢,-++) € O° in some domain D and its partial dorivatives Be, Oyy => Des, re, +, where C* denotes a set of functions possessing. continuous partial derivatives of ordor n. A p.d.e. is thus a relation between the dependent variable @ and some of iss pertial derivatives at covery point (x,y, t,-:) in D. Definition 1.2.1 : We define the ordor of s partial differential equation to be the order of the derivative of the highest order n occurring in the equation.CHAPTER 1. FIRST ORDER PDE. A pide, is aid to be quasi-lincor if the derivatives of the highest order which oceur in the equation are linear. A quaki-linear p.d.e. is said to be semi-tincarif the coeficents of the highest order derivatives do not contain either the dependent variable or its derivatives. A semi- linear p.d.e. is said to be linear if it is linear in the dependent variable and its derivatives (please refer to pages 7 and 75). A pd. which is not @ quasilinear pd. is said to be non-linear ‘We are concerned in this chapter with pertial differential equations of first order with one dependent variable 2 and mostly two independent variables 2 and y. We follow the notation p = 2, and q =. We ean ‘then write the most general frst order p.d.e. in the form Fleue,4) =O. Partial diflerential equations ari in a large variety of subjects : in geometry, physics, mathematics ete Example 1.2.1 : Surfaces of revol Many surfaces of revolution with z-sxis as the axis of revolution are of the form 2=F(), r=@ + yh, ‘where F is an arbitrary continuously differentiable funetion. On differentiating = F{(r) with respect to 2 and y respectively, we obtain P= OF), a= HPO. After eliminating the function #(), me get yp-aq= which is a pe of first order satisfied by the surfaces of revolution of tho form given above. 0 Bxample 1.2.2 : Consider the surfaces of the form Plu, v) = 0 where uw = u(e,y,2) and v = v(z,y,2) are known functions of =, y and 2, and F is an arbitrary function of tw and v having fist ordor partial derivatives with respect to u and v. (On differentiating F(u,») = 0 with respect to 2 and y, treating 2 as a function of 2 and y, we get respectively oF, oF, By le tt) + Fo (ve + 7%) = 0, 1.2. GENESIS OF FIRST ORDER P.D.E. 5 Fwy teu) +H Bu ty + aus) +S lty + 90) On aiminating S° and SE trom thew equation, we cbt > (ue) 4 Aleve) _ Hus0) Po(2) * "Ata,2) ~ Bleu)" which is a p.de. of first order satisfied by the surface F(w,») where SO ato Oy) re - is the Jacobian, a Example 1.2.9 : Functional dependence Let » = (x,y) be a known function of z and y and let w= u(s,y) be ‘a function of w alone, ie., not involving 2 and y explicitly. That is u= Hi), where / is arbitrary. Let Hf € Ct On differentiating the above equation with respect to 2 and y respee- Heely, we got "(oe y ty = Heyy On eliminating H'(v) from these equations, we get tes which isa first order p.de. for u. 5 Note : Hence if there isa functional relation between two functions 1u(2,y) and v(z, y) not involving «and y explicitly, then deve) ea) ~° ‘The converse of this result is proved in Lemma 1.5.1 Example 1.2.4 : Buler’s equation for a homogeneous function A function f(z, 9) is said to be « homogeneous function of x and y of degree 1 if it satistes 102, Ay) = A*F(zu)-6 CHAPTER 1. FIRST ORDER PDE. Tat # = J(s\y) be homogeneous function of # and y of degte n ‘Then by Euler's theorem, the function f(z, y) satisfies the first order ie : ¥ thet Uh =f a Note :n cach of Examples 1.2.1 101.28, we have produsad a pac. by eliminating an artery function, Now consider twoparamter family of urtacs = Fley0,)). (22) On diferentiating with respect to x and y, we obtain respectively Pm Blanes), (123) = Fr.ne,d) (124) We can solve two of the above three equations to find a and & im terms of 2,y,p and 9. This would be possible provided that the matrix (i 8%): aa9 's of rank two by the implicit function theorem, Substituting @ and & ef to obtained in the third ecuation vil usually lad to relation ofthe = form f(2,3,2,9,4) ~ 0 which iso fist order poe ; Example 1.2.5 : Consider f a + azty? +0, (26) (On differentiating (1.26) with respect to 2 and y, we got respectively p=1+2an/, (127) a= tasty, (128) (On eliinating a between (1.2.7) and (1.2.8), we obtain = y-2=0. 0 Example 1.2.6 : Consider (ea + ye? (1.2.9) On differentiating (1.2.9) with respect to x and y, we get respectively. p= —(2—a), NESIS OF FIRST ORDER PD, = Wb) Therefore #04 P+¢)=1, a Exercise 1.2.1; Bliminate the stbitrary functcn F from cach of following equations and find the corresponding pd, () Fe—syztt+y)=0. Ans: py— gray? a? i) Fleyrty-2)=0. Ans: pe—qyae oy, Gi) z= F(Z). Ans. () P@+i2~ yao. F 22) 09) Fepns-+y42)=0 Peay =O. Ans: p-9= 2h : =2(a+y), Ans: ay~2)p+y2—a)y=xe-) Faureite 1.2.2 : Eliminate tho parameters « and b rom each of the following equations and find the corresponding pd.e. () 2=@+aly+), dae 36 Gi) 2e=(ersyF +, ‘Ans. Gi) s=artly Ans. () A0+e)=Se+0n40% Ane: pee, Classification of first order partial differential equations 1. Linear equation : A first order pe, is said to be a linear tuation if is linear in p,q and ie, if it of tho form Pla, vip+ O(a wa = Renae + S(a,y) Example : yp —.2q = 2ys+-2, 2. Bemi-HInear equation : A fist order pe. is aaid to be a semi Kear cquation Ii i linear in p and y and the cootients of p and ¢ are functions of = and y onl, ic, if itis of the form PE YP + Qey)e~ Rley,2). Example : ep — yrq = 224 % Quasi-linear equation : A first order pd.e. is said to be a ftasHlinear equation if i is linear in p and g, Le, if itis of the PCsuy2le 4 Qlou, 2a = Ray, 2) Brample : (2° + 2?)p— ryq~ 282 +92,10 CHAPTER 1. FIRST ORDER P.D.B. and is called the singular integral of (1.3.1). Lemma 1.3.2 : The singular integral is also a solution. i Proof : Let z = F(x,y,a,b) be a complete integral. We will show ‘that the envelope of this two - parameter family, if it exist, is also a Hels that the envelope is obtained by eliminating o and b between 22 Fle.v.0)), (a9) 0=Fx(2,040,2), (1319) =e, ya,0) (an) Hence the envelope will be given by 2 = Clay) = Flasyalzy), 2.9), ‘where a(z,y) and 8(2,) are obtained from (1.3.10) and (1.8.11) by solving for a and b in terms of 2 and y, ‘The envelope will satisfy the given partial differential equation. For, Gy = Ft Fate + Fibs = Fe, Gy = Fy + Faty + Fiby = Fy, since #0, =O. ‘That is, the envelope will hove the come paetial derivatives ao a méimber of the family and for the same reasons as in Lerama 1.3.1, this envelope, if it exists, is also a solution. a ‘The singular integral can, however, be found from the pe. itself without knowing any complete integral. Lemma 1.5.3: The singular integral of f(c, y,2,p,q) = 0 satisies the following equations. Hla.y37.9) =0, Ht. 2.000 (31) Sle, 2,700) Proof: Since z = F(x,y,0,2) is @ completo integral of (13.1), the cxustion flau.Fla 100), Flz,.0,0).F(e, 4.00) =0, (18.13) ‘which holds ential for all a and bean be diferentiate! with respect ‘to @ and 6, and hence leads to LeEat IpPon + IeBon LE Et fie + asi 13. CLASSIFICATION OF INTEGRALS 0 On the singular intogrl, F, = 0 and F, = 0, Therefore the equations 'n (13.14) simplify to Sect fFvo Jefe + SeBo On thie surface, FeFip ~ FaFye #0 (ince F, = 0, = 0) and hence fy = 0, fg = 0. Otherwise the matrix B he fel F Fa Fy) vill nt have rank two contradicting the fact that 2 = 2, yja,8) is a complete integral. Hence the lemma, a ‘Special intogral : Usually (but not always) the three classes (a), (b) and (c) given above include all the intogrle of the fist order p.de. (1.3.1). However, there are some solutions of certain frst ordor partial 1. o Exercise 14.1; Find the general intograls of 14. LINEAR EQUATIONS OP THE FIRST ORDER Ww (i) =(op— va) ; Ans. : P(ay,(2—y)? +24) =0. (i) vp —249 = (2 — 29). Ans. : Fl? +1, uy~ 9) 0. (ii) (2? ~ aye — Ap + sly +2) = aly — 2) Ans. 2 F249 + 23,9 — tye 2) () yaptaze=sty. Ans. : F(z? — ys —2(¢-+y)) =0. ©) G+ let @ += Ans. : F(a! +25 —2y,2(2~ 4) =0. (i) (2+ yp + 209g = (2+ y)2. Ans. : P{(a+4)/z,(2°/y)-y) =0. » (Vil) 2(y + 2*)p + yay + 22)q = oF Ans.: P(z/yz, (2? — 2y) (yz) = 0. (vill) (2? + Say") 4 (VP 4307p) = 222 + yz. // Ans.: sy/2? = G((2* — y*)/2). () 2(y — 2)p + ule - 2) = x( - y), Ans. 2y2=Gle+y-+2). ‘We will now state the theorem for the case when the number of independent variables is more than two. This result would be required in Section 1.6 a2) = Cr (é = 1,2,--+,n) ate independent solutions of the equations “Py ond R are continuously differentiable functions of and 2, not simultancously zero, then the relation tn) = 0 where @ is an arbitrary differentiable function General solution of the quasilinear p.d.e. az | pO a: Pigg t Piggy t+ Page = Re + Similar to the proof given in Theorem 1.4.1 a18 CHAPTER 1, FIRST ORDER PDE. 1.5 Pfaffian Differential Equations By a Pfaffian differential equation, we mean a differential equation of i! the form Filsise-sta)den + Paley 24)dtg bos + BCs 5)dey = 0, (sa) where F's (é = 1,-+-n) are continuous functions, ‘The expression on the left hand side of Equation (1.5.1) is called a Pfafhan differential ion 1.5.1 : A Pfaffian differential form is said to be exact if we ‘can find a continuously differentiable function «(2,,-+-, 24) euch thet du = F(a, +++, t)dry + Fa(tays-+stn)daa ++ Definition 1.5.2 : A Pfaffian differential equation is seid to be integrable if there exists a non-zoro differentiable function y(t, 2a) such that the Pfaffan differential form + Fal2iy--- tad lA, Bala, toot Fa(ty'*+sTn)Atn), is exact, The function s(x, +-,.4) is ealled the intagrating factor and ‘u(@iy:*"s4n) = ¢ where cis an arbitrary constant, is called the integral of the corresponding Pfafan differential oqustion. ‘Note : A Pfaffian differential equation (1.5.1) is said to be exact if the fafian differential form on the left: hand side of Equation (1.5.1) is exact. Note : Observe that u(s,+--,4) =e is a surface in JR" such that at evety point on it the normal has direction ratios (FF, ->- Fy) Theorem 1.5.1 : Thore always exists an integrating factor for a Pfaffian diferential equation in two variable. Proof : Consider P(c, ybdr + Q(x, }dy =0. I(x, 4) #0, then { y __ Plow) &~ "Wa From the existence theorem for a first order ordinary differential equae sion, the above equation has a solution F(,y) = ¢,. Then OF OF, Bit Gano. 1.5. PFAFFIAN DIFFERENTIAL EQUATIONS 19 ‘This exact Pfaffian form differs from Pds + Qdy = 0 by a factor as they have the se solution. This fartor i nothing but ihe integrating facto: a mn geeral, Paton ferential equation in more thaa two variables say nol be integrable. Tn the next Uhorem, we shall derive aocosary ‘and sufficient condition for the integrability of a Pfaffian differential ‘equation in three variables. We shell first prove two lemmas which are vsed in the totem. ‘At was shown in Example 1.2.3 that if there is a functional depen- dence between u(z,y) and v(z,y) not involving x and y explicitly, then Sis4) =o, We sal prove th cones in the floving emma, emma 1.5.1 : Let u(r,y) and v(a,y) be two fonctions of z and y such that z #0. (052) a uss then there exists a relation Fluv)=0, (154) Detween u and v not involving 2 and y explicitly Proof : Since 5” # 0, we can eliminate y between uw = u(z,y) and v= o(r,y) and obtain the rolation Fluea) =0. (155) We vill now show that F does not depend on 2, leading to (1.5.4). On diferentiting (1.6.6) with respect to x and y, we get, respectively OF , aF du | OF av Bs * Gade * Bede 7 ee) aF Ou , OF bv Bu dy + ‘Oe dy (1.5.7) On siminating FF from these ecuations, whichis posible it 2 4 6 ‘ve find that a2) OF Bee. 0) Bu ar. Gi , (2.58)2 (CHAPTER 1. FIRST ORDER PD.E. aR au le, eo =o, by virtue of (1.5.3) ince 2 i. or However, since 5° #0, this implin that [= 0. ou , thea Fe (1.5.6) implies that SF = 0, Hence Fis independent of Lemma 1.5.2: if X oul = 0 whore X = (P,Q,R) and y ie en arbitrary diferentiabe function of y and, then 18 -curl(uX) = 0. Note: By X= (P,Q, R), wemean the vector ¥ = P7+Qj+RE, where 7 J and are the unit vectors in the positive x,y and z directions respectively Proof : Consider uk -curl(ut) Suppose = 0 (why 2). In this case also, Equation a zon (ene (2 eur) ‘The lemma flloys from the above result Convoresly, i aX cur(.X) —0 then X -cual —0 for p 7 0. ‘Theorem 1.5.2: A necessary and sufcient condition that the Pfaffian differential equation ar = Pls, y. 2a + Qa, y, 2)dy + R(a,y.2)de = 0, (159) ‘ be integrable is that i (Fem) (1510) Proof : We first show that the condition is necessary. For, if Bquation (1.5.9) is intograble, then there exit diferentable functions plz, y 2) and u(2,y,2) such that du = plxs4,2)IP(e, v.2)de + Q(z, v.e)dy + R(x, y,2)¢2), (1.6.11) 1.5. PFAPFTAN DIFFERENTIAL BQUATIONS bere u(x, y,2) # 0. However, Ou, du= Ou, , Ou dot Glut Bde (2.5.12) Be Comparing Equations (1.8.11) and (1.5.12), we got eu me that is, wX = Vu. Since curl(u) = 0, we heve curl) = 0. Hence uk curls) = 0 ‘Therefore, from the converse of the previous lemma, we have X-eulk ~0. ‘Now wo shall prove that the condition is sulfclont. Suppase = i seated as ¢ constant, the difleential equation (1.5.9) becomes P(x, y2)dx + Q(z, y, 2)dy = 0. Since a Pfaffian differential equation in two variables is always inte arable, this has a solution of the form U(z,u,2) = 1, where cy may Involve 2. In addition, there must exist a non-zor0 differentiable func- ton such that. ou ie oe P= eee, ulm “ o=F, Hl On multiplying (1.6.9) by and substituting the above, we get Uap Way, W, au, Fatt t Bye geet HR Gelds = 0 oe ‘This implies that (1.5.13) where Kae We are given shat (curl) =0. Hence by Lemma 1.52, X curl.) ~ 0, Obeotve that ae = Fe Fy gt OOK), = V+ (0,0,%).2 CHAPTER 1. FIRST ORDER P.DE. Hence - : = (QU BU U.K OK eR veatlak) = (Be Bi Be +X): (Gy Be WeK. OU oK * Ge Oy” Oy = 9}, follows th It follows that wie ae since wX-curl{u) = 0 by Lemma 1.5.2. Hence K can be expresed a8 ‘2 function of U and z alone as Equation (1.5.14) indicates that there is ‘a relation between U and X independent of z and y, but not necessarily. of x (from Lemma 15.1), Therefore Equation (1.513) becomes =0, (18.14) w SeKuaqo Tk possesses a solution #(U,2) = c, where ¢ is an arbitrary const ‘This solution can now be expressed in the form u(z,y,2) = ¢ by using the expression for U in terms of z,y and 2. ‘Thetefore the Pfaffian differential equation (1.59) is integrable. 5 Note: The Pfaffian differential equation (59) is, in fact, exact if and only if curl = 0. Example 1.5.1: Show that the following Pfaffian differential equation is integrable and find its integral ude + ady + 2xdz =0. Sol. : Observe that curl = 0, where X = (y, 1,22). Therefore the Pfoffian differential equation is exact. In fact, it ean be written as de + dy + Dade = dry +2, Hence the intogral is wla,y,3) say +2= o Example 1.5.2: Find tho integral of yade + 2ezdy — Sryds ~ 0. Sol. : Observe that X-curL¥ = 0 where X = (yz, 2x2, —3ry), and U = nif = eis a solution of yds + 22dy =0. Then w= ©. Further, nw) _ _3U COMPATIBLE SYSTEMS fore 1 ww, ae 5 whose solution ie U = 6. Theeoe the naga of he glen Phan form is a Exercise 1.5.1: Verify that the Pfaffian differential equations are texact/integrable and find the corresponding integrals. Gy + ya)de + (ar + ae + (y* ~ sy)de = 0. Ans. : (r+ 2) @ Grualde rst ate Cane + (L+ay)(L-bey) = (+ y) a) part eeu Ans. | 2y Cin) yd + (Py 2a) (es < ilds~ 0. Ans. : Day —a(y2 + 24 () (6x-+ye)de + (2 — 2y)dy+ ae + 2z)dz = 0. Ans. : 82? y? tat + zyme. (Wi) e-ylde+ x(e+ )dy +a(e+y)ds =0. Ans. : a(t +1) = clr+ 2) (vil) (22-4 y? + 22e)de + 2eydy + 28d = 0. Ans. = 24 +ay! 4282 =e (wil) (1+ y2)de + 2(2—2)dy — (1+ aya Ans. (1+ y2)(1 + ey) =(1+ 20). 1.6 Compatible Systems of First Order Partial Differential Equations Definition 1.6.1 : The equations Hla.uz.704) = 0, and 92,942,744) = 9, are compatible on a domain D if 9 (0.1= LB goo 0, (i) p= Ole, v2). a= ony, 2),Et CHAPTER 1. PIRST ORDER PD. obtained by solving (1.6.1) and (1.6.2) (in view of (i)), render dz = 92,9, 2)d2 + W(e,y,2)dy, (1.6.4) integrable, ‘Theorem 1.6.1 : A necessery and sufficient condition for the integra: bility of (16.4) is U8) 4 ha) , had , Ais) We Bean) *Pacz9) * Buna) * Bla) Proof: (1.64) i ilagrable snd only if X-cunF =0, whore X= [4,,-1), ie, He) + U6.) — (He — by) = 0, ie, Yr + bls = dy + Uy (65) On substituting # and y for p and g respectively in (1.6.1) and difer- entiating it with respect to 2 and z, we obtain Jet fbe+ fe Jet hybet ibs = ‘On multiplying the second equation by ¢ and adding it to the first, we st Jot Ble fle 661) + lia +40.) =0 Similarly me may desluce from Equation (1.6.2) that De + $0. (bs + $6) + 9y( Ye + Os) = 0. Solving these equations, we find that vet onan 5 (AED vopthalh. (169) If wo differentiate the given pair of equations with respect to y and 2, we obtain, after a similar analysis as above 2-3 (athe) , tha) tt vb {SS + oe 1.6. COMPATIBLE SYSTEMS 5 ‘On substituting from (1.6.6) and (1.6.7) into (1.6.5) and replacing ¢ and by pane a sepectively, me ae thatthe condition fr tegrebty of (164) ie that [f,9] = Note ‘okt a 6.4) is of the form Flayae) (1.6.8) where ¢ ie an arbitrary parameter. Hence we assert that if Equations (1.6.1) and (1.6.2) are compatible then they have a one-parameter family of common solutions. Example 1.6.1 : Show that the equations J=s-m-2 (189) g=ptq-se% (2.6.10) ae compatible and find a one-parameter family of common solution. So. + Observe that SES} — (1 +29) #0 on the domain D where does not contain the points (r,y) such that 2 = 0 or 1+ 2y = 0. Then on D, we obtain (+y) Cray)" _ 2-2) Gay)" and Equation (1.64) becomes alte, He-2) Tray * Trey a(e—s) ay dz— de mee Hea de ee + ney aoa tay Hence (1.6.8) and (1.6.10) are compatible on D as (1.6.1) is integrable 25% = ia one-parameter family of common solutions, 9 Note tility ean ako be shown by verifying the conditions aha) pad FE 0m tal =0 Srsredl4n),26 (CHAPTER 1. FIRST ORDER PDB. Exercise 1.6.1: Show that the equations far te-1=0, (1.6.12) g= (P+ e)e-pr=0, (1.6.13) are compatible and find the ono-parameter family of common solutions, Ans. : 22+ (y+, ‘Note : For the compatibility of f(z,u,,7,4) = 0 and g(x,y, 2,P,9) =0 is not necessary that every solution of f(s, , zp, q) =0 bea solution ‘of g(2y4yp,4) = 0 oF vice-versa as is genorally believed. This can be seen fom the following observations : (@) In Bxample 16:1, «= 2(y-+1) isa solution of (1.6.9) and not of (1.610). (b) In Brercise 1.6.1, 2 = (2+0)/V2 is a solution of (1.6.12) and not of (1.6.13). 1.7 Charpit’s Method In this section, we present a method to find # complete integral of & frat order pd. Let Le.wep.) =0, an) be the partial differential equation whose complete integral is being sought. ‘A family of partial differential equations: aleyzbaa) 72) is said to be a one-porometor femily of partial differential equations compatible with (1.7.1) if (1.7.2) is compatible with (1.71) for each value of a Since f = O and g=0 are compatible, we have [f,4]=0, ie, IL-1 oho ah ge Unt d g(a 9e =O. 173) ‘This is a quasi-linear first order pe. for g with x, y,z,p and q as the Independent variables. 17, CHARPIT'S METHOD ” Our problem, ie, finding a one-parameter family of parti differ- ential equations (1.7.2) which is such that each member of the family is compatible with the given p.d.e. (1.7.1) then is to find a solution of thie Equation (1.7.8), in ae simple a form a6 possible, involving p or q or both and an arbitrary constant o. This we do by finding an integral of the auxiliary equations (refer to Theorem 1.4.2) 8 = ae hot Phtth fetPhe Stabe” CCharpit’s mothod consists in choosing a one-parameter family of partial differential equations (1.7.2) which is such that each member of the family is compatible with the given equation (1.7.1), Then, solving for and q from (1.7.1) and (1.7.2), we get ara) P=9emea), = HEY 4) Then ds = ode + dy, 78) isintograble by virte of he fact that (1.7.1 and (1.7.2) are compatible ‘An integral of (1.78) willbe ofthe form Fanaa) = ‘As this ie & ovo-paramoter family of solutions of Equation (1.7.1), it willbe a complete in'gral of (17.1) Example 1.7.1 : Find a complete integral of f ‘Charpit’s method. Sol. |The auraliary equations (refer to Equations (1.7-4))are i es qey pay Qpqay 2ep—pqy 22q— par’ ds _ pds + ody t-xdp + ude 2” ‘2a(pa + qu) . de _ dep+va) 2” apty = pasy = 0 by ie, = a(ep + ug), where o is an arbitrary constant. Hence the tequred one parameter amily of pd, which iscompatiblowith 7 =O is o€z.n,2,,¢,0) = #~ ole +4) = Further, solving for p and q from f = amid iu od obtaisCHAPTER 1, FIRST ORDER PDE. On integrating, we obtain 2 = brly®, Hence F(2,u,2,b,0) = 2— bely? = 0, ‘is a complete integral of f = 0. ‘At can easily be verified thet the matrix (BB Be) FP Fa)! is of rank two. Example 1.7.2 : Find o complete integral of fa Qh +ey—a=0. Sol. : The auxiliary equations are (2 112.9,4,8 Solving for p and 9, me get g = —2. Hence (1.7.8) becomes ten SPE TF se ay 2dz— oydy Jaa Po atyt= (art bP. ‘Hence the required complete integral is atm aly (ar tb). Hence it follows thet de. ‘This implies that 17, CHARPIT’S METHOD Example 2.7.8 : Find a complete integtal of f= apa ty? -1=0. Sol. : The auxiliary equations (1.7.4) are ad se ft aa p+ Tame eo)” =o0 =e” ‘Then Equation (1.7.5) becomes dz (e+)? =4las +0), is the roquired complete intogral. Escumple 1.7.4: Find a complete incegral of the p.d.e. we nyhed Sol. : The auxiliary equations are eg Baty ~ Dytg ~ Dep +P) — Dap? ~ “Og Suppose we consider the equationCHAPTER 1. FIRST ORDER PD.E. = VIR Flog + alogy +6, is a complete integral of f = 0. Instead, if we consider de __dp wp oand p= a Via v ize v ‘Then Equation (1.7.5) ten tees Bay =clogz + VT—@logy +, is another complete integral of f = 0. Note : A first order p.d.c. can havo sevoral complete integrals. Example 1.7.5 Consider the pe. F+P +2) ‘The above equation as (-oP +(y-oF +2 =1, 76) and (y-ma—o? = (1+m'yQ- 2), as its complete integrals. Observe that these two complete integrals are not equivalent, ic. ‘we cannot obtain one from another merely by a change in the choice ‘of arbitrary constants. In fact, these two families aro entirely different types of surfaces. Note that the two complete integrals in Example 1.74 are, however, equivalent. ‘When, however, one complete integral has been obtained, every ‘other solution, including every other complete integral can be obtained (We shall explain a procedure for this in Section 1.9). Consider a sub- family of (1.7.6) by putting 6 = ma +e where m and c are fixed. Its envelope which is found by eliminating'@ between ar (2-0)? + [y—ma—o? + 2 1.7. CHARPIT'S METHOD and (2-4) +(y~ma-om= ives the other complete integral (y-ma~ 0? = (14 mh(1~ 2), Some Standard type Type + she given p.d.e, does not involve x, y and 2 (P.9) = ceoplicitly. ‘The aueliary equations are ay Sa” th +h 0-0" Solving the last equation, we get cither p= a (or g =). pees we solve f(a,q) = 0 (or f(p,a) = 0) for g Q(a) (or p= Pla) ds= ode + Qle)dy > x=02 + Q(aly +8, (or ds = Pla) + ady =} «= Piolo +oy +0 Example 1.7.6: Find a complete integral of fine) = a -m=0. Sol. = Puta=a Then p= —*. te oqnalion (1.7.5) becones tady, ‘which implies that Which is the required complete integral a Type II: f(2,p,4) = 0 (ic, the given pad.e. does not involve 2 and y explicitly), ‘The auxiliary equations are fi | tie a Ih” Ph +h Ph PueM ye yg re Therofore $2,009) = 0 0: g = Qla,2) and also p= aQ(a,2),CHAPTER 1, FIRST ORDER PDE. dz = pds + ody = Q(a,2)(adz + dy). ‘The complete integral is Sagyretet® Example 1.7.7 : Finda comploto integral ofthe equation m-p-9=0. Sol. + Putting p= ag in the above equation, wo got sq a~1=0, oF [p=0,9=0in which case 2 = constan!] 10, then Now dz = pde-+ aly = **%(as-+ 2dy), and hence ie WD ee +ytb Type IIT: o(z,9) = Hye), (separable type) ‘The auxiliary equations are See mh 9H - ay Note that ged + gydp = 0. ‘Thorefore g(2,p) =a Hence f(y. a) = a. Prom these wo can salve for p and g 82 p=G(a,2) and g= F(e,y). ‘Then dz = pd + qdy implies z= [Glazide+ | Havidy +6. Example 1.7.8: Find a complete integral of Br@acty. LT. CHARPIT'S METHOD 3 Sol. : Observe that the given equetion can be put in the form p-2=-@-y) =a. Then patvete and g=+V¥—6, 00 that dz = (vEFa)de+ (vu=a)dy. 2a Fetes ya +d ype 1V (Clataut Form) : = 92-+49+ 916.) een campus eaegeali ary z= orthy + old), for, iti a solution and the matrix (ete 39) yea 01 is of rank two, Example 1.7.9: Find a completo integral of the p.de z= petay+ lope Sol. : ‘Phe complete integral is fart by+loged Example 1.7.10 : Find a complete integral of the p.d.e. pas = P(g +P) + ap +a) ata or etme. Sol. : The complete integral is Example 1.7.11: Find a complete integral of the p.de Sol. : We haveEY CHAPTER 1. FIRST ORDER P.D.E. ving for ¢, we get g = == Solving for we get = ST ‘which implies that » — ar = Bly +a), i, a+ by+ab, o Exercise 1.7.1 : Find complete integrals ofthe following partial i ferential equations (4 ap -hyt - 2 pg = 0 Ans. © 22 say+K(1—ar) (i) pa? — daa? + 64s ~2 = 0. Ans. 2 = Bey +a) be + gh (iv) Pata Ans. + {(1-+a)3]!=(a2)2 +9? 44 (v) et pt bay +29? = 0. Ans.: 922+ (a—2)*+97)=0 (i) pry + pa + ay = we. Ans. :2= 07 + bel(y +a) (vi) Bale? +1) =e Ans. : = 2(02 + ja +2ey +6. (i) 2mm = Pale + 0) +000 +0) Ans. ; 2 = ar +by + ab. (ix) 2? +) + pe + ay = 0. Ans. «+ (ax-+y)2/(1-+ 68) = 6. (x) PP + aty? = a(n? + 9) Ans: 2 4a? +a)? 4 (Pa 40, JACOBI'S METHOD Gi) 20 +0) ot 4 1p? ‘haanege Bat 1.8 Jacobi’s Method ‘Let us consider the following p.d.e. Flas 2,85, (181) Here 2,y and 2 are the independent variables and the dependent vari- able w does not appear in the equation, A function u = F(z, v,2,a,b,¢) is said to’be a complete integral of (18:1) if it estistis the p.d.e. and the associated matrix Fa Be Fy Box Py Fiz Fy For} Fe Fea Fay Fee is of rank three. Jacobi's method : For a given pds. ofthe type (1.8.1), we consider tno one-parameter fans of pati diffeensal equations Te(2, sz) Urs Ups thay) = 0, (18.2) Fale.v tn ty td) = 0, aaa) . 2ibuta). (1.84) Ble) and the Pfaffian form du = ued + wydy + uate, (185) is integrable, where ug(1,4,4,8),tp(t,9,2,0,8) and s(t, ys 2008) ‘are obtained by solving (1.8.1), (1.8.2) and (1.8.3) by virtue of (1.8.4). Such hy = 0 and h; = 0 are said to be compatible with f = 0. ‘Observe that (1.8.5) is either exart or not integrable ab all (why 7) ‘The conditions for (1.85) to be exact are uy _ One Dun _ Ble Bey! be ae ep)36 CHAPTER 1. FIRST ORDER PD.B. ‘The integral of (1.8.5) gives the complete integral of (1.8.1). ‘Theorem If h; = 0 and hy = 0 are compatible with f = 0, then fa and fi satisfy AN), OA) Ba,22) * yi vy) where h = A,(i= 1,2) Proof : On differentiating (1.6.1) with respect to x, y and 2, we get Paras tet Bete + Fe tn = (188) 2, at, oF tr Ba ee Ait) au) (89) & (1.8.10) Consider Fa, 1,25 ey ty te) = 0, as) where h= A; (i = 1,2). On differentiating (1.8.11) with respect to z,y and 2, we obiain 1.8.12) (2.8.13) (1.8.14) On mulling (188) with 2* ond (1.812) by 22 uy weet af an af de _af am, (i; Bue Buz Buy) S08 tay = tt = yt te fon (LB) ‘On multiplying (1.8.9) with x and (1.8.13) by be and subtracting, wwe eet 29h Of haf ah Of hy, 2y Buy Oy By Dus Oy Ou Bes and subtracting, ed GES 1.8. JACOBI'S METHOD a7 (On multiplying (1.8.20) with 2* mend a3) ty 2 z and subtracting, weet 9F OK _9F 9%, (OF On OF Oh. ap ob i Bs Bu, (On adding (1.4.18), (1.8.16) and (18.17), we got HAN) , AGN, Bf) Bawa) ~ Bly.) * He, ue) which isthe roquived result. ‘The above equation can also be rewritten ae (818) an, oh Iuige t lage + Su ‘which is ¢ semi-linear p.c.e. for h. ‘That is, for a given f if hy and hy «satisfy (1.8.4) and (1.8.18), then hy = 0 (# = 1,2) are compatible with {= 0. The following example illustrates Jacnbi's method. Example 1.8.1: By Jacobi's method, solve the equation Sol. : This oqustion is of the form (1.8.1). ‘The equation for h and hy (Le, (1-8.18)) in this example is ‘The auxiliary equatione (refer to Theorem 1.4.2) are __ due “Tw Tm ‘The wo independent solutions of the above system of equations are ‘te =@ and wy =>, and then from the given equation @+R-2 —: Now du = adz+bdy + [(a? +6 — 2%)/2}ds is exact. On integrating, we obtain ae! ont by + (2 +P logs — 52? +6,38 CHAPTER 1. FIRST ORDER P.D.B. hich is @ compte intogral ofthe given equation. o Example 1.8.2 : Show that a complete integral of f(t, ty.) = 0 is u= ar + by +cz+d where f(a,b,c) = 0. Hence find the complete integral of te + Uy + Us — tatyts = 0 Sol. * The ainiliary equations are Therefore te = 0, = 6,0 usarthy+atd, (1.8.19) where o,b,¢ and d are constants, will satisfy the given equation if S{e,b,¢) =0. (1.8.20) ‘This equation determines e in terms of @ and b, Hence (1.8.19) hes cree arbitrary constants a, band d, and it isa complete integral. Por example, a complete integral of tue + ty + us — Wetytts = 0 is (1.8.19) if a+b+c—abe=0, ie, e= (@+1)/(ab- 1}. a Exercise 1.6.1 : Solve the following equotions by Jacob's method. () 242m, - (a, +1)? =0, Ans.: user+be+ var? — uf — au = 0, Ana: v= E49) yay se 46 where ois «fixed constant. true Ans.: u=an+ly+(1-a?-0)2+0 Gis) rust uty = 08, Ans. ; u=blogz + (a? — b)logytaz+e. where @, and ¢ are arbitrary constants. ‘We will now apply Jacobi’s method to find a complete integral for a first order p.d.e. in two independent variables. Consider the following pao. flen2pa) = (1.821) 18. JACOBIS METHOD 39 ‘The solution of (1.8.21) is a relation betwoon 2,y and z. If this relation is u(zy2) =o then (On substituting the above in (1.8.21), we obtain (6, 1924, y, ts) = 0, (1822) which is in the form given in (1.8.1) and can be solved by Jacobi's: ‘method discussed earlier, which yields Um F(a 9, z,0,8) +e In this, if we choose w FG, v.2.0,8) = 0. Example 1.8.9 ; Pind a complete integral of the equation pPr-tqty = z by Jacobs method, pay Sol. : 1 we get a complete integral of (1.8.21) 28 peatt, get ‘The given equation becomes ail + yd — 2i = ‘The auxiliary equations sre a ty ou; ~ Byuy ‘The two eolutions of these equations are witb, whence wu. = {6/2)"/*,u, = (t/a)! and then wu, = [(a-+b)/2)) hb ci) ty = (b/y)"/? and then wu, = [(a+)/2) d= Ge + Cay + S=2)00a,CHAPTER 1. FIRST ORDER P.D.E. w= Zaz)? + 2(by)! + 2((a+ Be)? +6 Whiting u =o, we get the complete integral of the given pide. as (en"-(T Exercise 1.8.2: Solve the following equations by Jeccbis method. @ Gta hy=az. (iil) 8 =naav. G) r=Gtay (w) rete = () g-pe-p=0. 1.9 Integral Surfaces Through a Given Curve : The Cauchy Problem ‘This section deals with finding an integral surface passing through a given curve. Let us first discuss the problem for a quasilinear p.d.e. Tiere we will obtain the required integral eurface from the goneral inte. gral. ‘Lot (4,2) =0 (or » = G(u)) be the general integral of she partial differential equation Pp + Qq = R, where Fis an arbitrary function of u and v (or G is an arbitrary function of u), where u(z,¥,2) = ¢1, ‘v(z,y2) = en aro the solutions of Equation (1.4.4). ‘Lat C be the given curve whose parametric equations are given by 2=a9(s),u =0ls),2 = als)s where 5 is a parameter (not necessarily the arc length from a fixed point). We want to find a particular F such thet the surface F(u, ) = 0. contains the given curve C. This is done as follows. Consider the ‘equations u(r,y.2) Substituting 2 = 2o(6),y = yo(s) 2 = 20(s) in these equations, we get ‘u(9(8),¥o(8),20(8)) = €1 and x(r0(s), vals), 20(¢)) = c2- Eliminating $ Detween ther, we get a relation botween ¢ and cp, Let the relation between them be given by F(ci,ca) = 0. Then Flu,u) = 0 ie the 19. INTEGRAL SURFAQBS THROUGH A GIVEN CURVE 41 required solution as P(u,) = 0 is an integral surface. Moreover the {initial curve lies on it. For, Fuels), ans) zl), ozs) ws, 205)) = Feo Sometimes the solution ean also be oblsined by secuming v = Gu) and determining G. Suppeee we assume the frm of the integral surface containing the givma curve C at v= Gu). Since th given curve Clie on the surface, we ind that substituting = = 24s), y = y(6), 2 = 9s) in the relation u(z,u,2) = G(u(z,y,2)) may sometimes enable us to find the explicit form of the function G at G = G(s) as ilstrated in the folowing example. However, it may not always be possible to do so. Example 1.9.1: Find the integral surface of the equation (2ey ~ Up (2-229 = 2(0~ v2), which passes through the line zp(s] | yo(s) = 0 and a9(s) = s. Sol. : The ausdiary equations (1.4) ore de da ~ Ue=w) ” sdet dy + ede , _ Bae + Dud +e a 0 ‘Therefore vy }ar—o,u=at +P temo. ‘The general integrals Pty te= Clytaz), where G is arbitrary. We want to choose G such thet the given curve lies on the surface 1?-+y?+2 = G(y~n2). This happens if G(s) = 1+s. ‘Therefore the required integral sucface is Prpss=Gy+z2)=14y+e2, ie, Baytaee-ytinl je 1.9.2 : Find the integral surface of the equation ar a p+ u(ds" + y)a= 2(25* +9),42 CHAPTER. 1. FIRST ORDER PDB. which passes through the curve 2 = 1, th Sol. : The auxiliary equations are @__w as at yBa2 ty) 2(2a? +y) ‘Thorofore y= a(l+e). 2 Bettye _ dy _ (Ox + ldo + dy Wey ey (st +u)ds+dytady dv, ,_ (tubo), tye ¥ v 8, 22 (1+5)inw =o andy =o, = a. Eliminating s between them gives F(o.) = 1a e1 ~ 62 +2 =0. Thovefore F (1, ») = 0 implies that we = Petme-ry-¥, (P+vler—9), ‘which i the required integral surface. o ‘Let us now discuss the same peoblom for & non-linear p.d.e. Here we will obtain the solution from a complete integral. Let F(e,9,2,8)=0, (92) be a complete integral of Hlaysa.r@) =0. (192) ‘We are interested in finding 2 solution of (1.9.2) which passes through the curve O: x= 2a(s),y = yo(s),2 = za(s), 5 being a parameter ‘We expect this solution to be an envelope of « one-parameter sub- family of (1.9.1). Let be this envelope which contains the eurve C. Let $ be the sub-family. Then B, the envelope of the sub-farily, will touch each member of the sub-family. As C lies entirely on Bit will touch each member of the sub-family for some s. This requires the relation F (o(5), yo(s), 208); 0 6) (1.9.3) 1.9. INTEGRAL SURPACES THROUGH A GIVEN CURVE 43 to be satisfied for each @ and b belonging to the corresponding sub- family, for some s (this is the condition for C to intersect e member of the sub-family) and also the relation 3 (eal), sn(sy2ls).0,8) = (9.4) to be satisfied forthe same s. (This i the condition that at thie point of intersection, the curve is actually a tangent to that member). On eliminating ¢ between (1.9.3) and (1.9.4), we get the required relation between a and b. This defines the sub-family we ate looking for, However there could be many solutions. Since on eliminating 3 we set (0,8) (198) Equation (1.9.5) may be factored into a set of alternative equations b= $:(a),6 = ¢2(a) ete. Each one of them defines a sub-family and the envelope ofeach of these sub-fmilies, i text, isa solution of the problem. Hence a solution, if it exists, may not be unique. Observe that this does not happen in quasi-linear equations, if the cane Cis propery choer. ‘Example 14 Find a complete integral of the equation (+a? = ps, ‘and the integral surface containing the eurve C : a = Oy = 34, ap _ dy Fata eteed. ae eee ‘which implies4 OHAPTER 1. FIRST ORDER P.D.E. (On intograting, the complete integral is found to be to? + (ay +)? (1.96) Equation (1.98) in this case becomes 4st = (at +07 ‘Then on differentiating (1.9.7) with ropec to 2, wo got 2 = a(as? +5). aor) (198) Further, on eliminating ¢ between (1.9.7) and (19.8), we get ab = 1 ‘Therefore, on substituting b = 1/a in (1.9.6), the one-parameter sub- family of the complete integral is : at 1s 22 + (ay + 2, ofa? + 7) +alQy~ a4) +1 =, 99) ‘and its envelope is obtained by eliminating « between (1.9.9) and tats +9?) + (2y- #7) =0 (1910) ‘The envelope is (2y— 27)? = A(z" +9"), Le, 2? = Ay VP), Since ¥a7-F9? > y , the minus sign isto be discarded and then ParytyPre), which is the required integral surface. Example 1.0.4 ; Find the complete intogral ofthe equation Pr+gy , and derive the equation ofthe integral surface containing the line etr=d, ‘Sol: : The auxiliary equations are dz dy ___ dz dp _ do ey ~ Beta wtp) OD ge aad pe (*29)” 19. INTEGRAL SURFACES THROUGH A GIVEN CURVE 45 (On substitating for p and g in ds = pds + ody, we obtain damody de Vinay ye" (On integrating, we obtain Vim W=tye+vb, le, (oy— 24245) = Ate, 1s the complete integral. ‘The parametric form ofthe given line ie'= = ,y = 1,2 =~», Equation (19.3) in this case takes the following form (91) (a+8-+ 2s)? = Abe. ‘Then, on differentiating with respect to &, we get dat b+2s) = ab. Eliminating s between them gives 8 + 2ab =0, ie, b= ‘The case 6 = 0 does not lead to a solution. (On substituting b= —2a ia (1.0.11), we got the eub-family orb= (ay— 142-20)? = ~Bar. (19.12) ‘The envelope of (1.9.12) is obtained on eliminating a between (1.9.12) and (w-2)(ay — 24.22) = 42, ‘The envelope of (1.9.12) is (1.9.13) ay 29-2), and thsi the required intagral eurfce o ‘We now show that we can derive any other completa integral from a. siven one in a similar way. Let Pz,y,2,0,b) =, de a complete integral of f(x, u,4,m,9) = 0. Suppose (1.9.14) Gle,y,2,hyk) =0, (1915)6 OHAPTER 1. FIRST ORDER PDE. be any other complete integral of the some p.de. involving two arbi- trary constants h and k. We will now chow that it is possible to derive (1.9.18) as an envelope of a one-parameter subcfamily of (1.9.14). We first choose a curve C’on the surface (1.9.15) uch thet the constants fh and are independent parameters in ite equations. Then we find the envelope of a one~ parameter subfamily of (1.9.14) touching the curve . This solution containing two arbitrary constants hand &, is thus a complete integral ‘Example 1.9.5; Show that the differential equation Qua +9? = x(xp + ya), hhas a coraplete integral etd's = ary tbe", (1.9.16) (2 ~ ka”) is also a complete integral + It can be verified that Equation (1.9.16) satisfes the gives partial differential equation. The corresponding matrix (1.2.5) (aap re 9 2% 0. is of rank two ifr # 0. Hence (1.9.16) is 2 complete integral in domain that does not contain the line x = 0. Let us now take a cur ‘on the surface 2(y + fer)? = 4(2 ~ kx?) involving h and k as B= 9 mms, 2 = ks? ‘Therefore we want to find « cubSamily of (1.9.16) which touches curve. Hence ive On difforontiating with respect to 5, we get k = —ah +b. ‘Therefor b= k+ah. Hence the eub-farmly of (1.9.16) is son = any + (b+ ah)s* ‘To find the envelope of (1.9.17] we have to eliminate a between (1.9.11 and 2a = ay + hot (9g QUASHLINBAR BQUATIONS a Equation (1.9.18) givesa = (y+ha)/2. On substituting fora in (1.9.17), ‘we get the requized envelope as Me = kx”) which Is the required completo integral o Exercise 1.9.1 : Find the integral surface of the differential equation (e=v)o+(u~2~2)a = », passing through the circle 2 = 1,22 49? = 1. Exercise 1.9.2 : Find tho intogral surface ofthe diffecential equation 2(z + 2)p + (rz + 2y2 + 2y)q = 2(2 + 1), passing through the curve 9 = 8%, uo =O and 2) = 2s. Exorcise 1.9.3 : Find the generol integral of the differential equation (e—vhy’e + (@~)e4g= (24+y2)x and the particular soltion through tz= a, y=0. Exorciso 1.9.4 : Find the integral surface of the differential equation pPr+ pry = 2pe +2, passing through the line y= 1, = =. Exercise 1.9.5 : Find a solution of p +4" = 0, passing through the line 2(8) = 0, y(s) =, 20(8) = Bs Bxerelse 1.9.6 ; Find a solution of z = p? — q? which passes through the curve C with the equation 42+? = 0,9 =0 Exercise 1.9.7 : Find the integral surface of the pd. pg containing the eurvo x = 0,2 — 92 wy + hey, 1.10 Quasi-Linear Equations : Geometry of Solutions ‘We sll fst discuss the semilinear case Somi-linear equations : CGoosider a semilinear equation Plawite + Q[e, way = Rlay.e), (10.1) where P,Q and R are continuously differentiable functions and both P(z,y) and Q(x,y) do not vanish simultaneously. Note that the ex- pression on the let hand side of (1.101) isthe directional dociative of (ay) in the direction (P(r, y),Q(x,y)) atthe paint (x,y) Lot us now consider the one-peremeter family of curves in the 2,5 + plane defined by the ordinary differential equation dy _ Qu) de Pay)"CHAPTER 1. FIRST ORDER PDE. 6 110. QUASI-LINBAR EQUATIONS 49 or the eystem of ordinary differential equations Bi $= rev, Z =n (1192) having the solution 2? 44? = c?. Along such a curve, « satisfies the ordinary differential equation ‘These curves have the property thet along them 2(r,y) will satlafy the ordinary differential equation a dy _ Paws + Un) et ae Pay) = Hoje", ‘where the arbitrary function k may depend on c, Hence we have the general solution de dz dy Faget GH HP) +4Q(ey) = Alen). (1.103) Ree ‘The one-parameter family of cures Cj, defined by Equations (1.10.2]} Op applying tho intial conditions, we eet are called the characteristic curves of the partial differential equation (1103). M02) = (2)? . Let (seine) be a point in the x y-plane. By the existence and (2) = (Ve? uniqueness of the solution of the initial value problem for the ordinary ep uaye! differential equations, Equation (1.10.2) will define ¢ unique character- | Hence the required solution is MOO _ Matai Gate “Ca Any) =f rae, o it Ae) = ay a) — in apt ven sig i AM acta ogunttons. + Wenn conser gaa wai for s(9) at (enn). Then the Bqution (1.30.4) determines a ungua See Sieve eet ‘ Pleswaiert Qn} = Rens) (L406) 2 = 2(0, 404), (.105)] vere P,Q and R are continuously differentiable functions of 2,y and ‘such that 2(to, 0,0) = 29- That is, 2(2,9) is niquoly determined along | z and P,@ and R do not vanish simultaneously. Its solution 2(z, y) the whole characteristic passing through (zo, yo) if we assign a value for | defines an integral surface 2 ~ 2(a,y) inthe x,y, -gpace. ‘The normal at (20,40) to this surface has direction ratios (2, 2,,~1). Hence Equation (1.10.6) Hoeace if z(z,y) is known on a curve To in the x,y-plane and ifthe] states the condition thot the integral surface is such that at each point, curve issuch that it intersects the one-parameter family of characteristic] the line with direction ratios (P,Q, R) is tangont to the surface at that | curves Chg, we can determine 2(2,y} uniquely in the rosion covered by point. In fact, any surface : = 2(z,y) is an integral surface if and Gr only f the tangent plane contains the characteristic direction (P,Q, R) ‘Observe that the curve Tp cannot be chosen arbitrarily. We will] whieh is direction field defined by the pd. ot each point deal with this point later (refer to Theorem 1.10.1). We consider below the integral curves of this field called the charac- Example 1.10.1 : Solve rz, ~pzz = 2 with the intial condition) teristic curves. ‘They area family of space curves whose tangent at each 20,0) = fla), +20. + | point coincides with the characteristic direction (P,Q, R) at that point Sol. : The chacacteristic curves are given by the equation ‘ane are given by the following system of ordinary differential equations wee ae ay a ay Ple.w.2) ~ Qe,v.2) ~ Rley.2) at (say), (2.10.7)CHAPTER 1. FIRST ORDER P.D.B. or & B i Qn), F = Playa) By the existence and uniqueness ofthe solution of the initial value prob- lem fore aystem of ordinary diflerential equetions, there passes & char- acteristic curve 2 = 220,30, 20,t),y = veo, Yos 20st), # = a(t, 80,2016) through each potnt (ro, 4,29). ‘The system of ordinary differential ‘equations (1.108) being an autonomous sysiem, there will be 2 two Parameter family of integral curves. Hence there is a two-parameter family of cherscteristic curves. Every surface generated by 2 one-parameter family of characteristics 1s an integral surface, For, if we consider aay point on such a surface, then the tangent to the characteristic carve passing through that point lice on the tangent plane to the surface. ‘Thus the tangent plane to the surface at each point contains the Line with direction ratios (P,Q, 22). Hence the surface is an integral surface. In addition, the converse is also true, ie. evecy integral surface is generated by a family of charactoristie curves. For, let us consider an integral surface 5 given by 2 = 2(2,y). Consider any point (ey, yp, 29) ‘on S, Let the solution of Plan sta) = Olen slo = Ple.v,2), (220.8) ‘with the initial conditions x = zg, y = to at ¢ = 0 be given by x = (0) and y = y(t). Consider the corresponding curve in three dimensions z= al), v=alt), 2= 2la(t), y(t) Observe that this curve lies on the given integral surface. Moreover, dz de dy Gna tag = Pale + Qe aden, = R(2,y,2)- ‘Therefore the curve satisfies Bavation (1.10.8) for characteristic curves ‘and it passes through the point (zo, tm, 29). Therefore $ is generated the characteristic curves as it contains the characteristic curve thro 110. QUASI-LINEAR EQUATIONS a ach point (ro, 4,20) on S. Further, if a characteristic curve has one point in common with the integral surface, ities entirely on the integral surface since only one characteristic curve passes through e given point, In addition, if two integral ourfaces intersect at a point, then they intersect along the entire characteristic curve through this point. Hence ‘tho curve of intersection of two intagral surfaces must be a charactetiatic curve. Note : The two-parameter familly of characteristics is nothing but the curves of intezecction ofthe surfaces u =, and v = «, given by (1.4.3). ‘To find an integral surface passing through a given curve, as in Section 19, we find a relation between c, and ¢,, which means we are actuslly shooting a one-parameter sub-family of characteristics, which in turn generates the required integral surfece. ‘Theorem 1.10.1: Consider the frst order quasilinear partial differ- ctal equation Playa + Qlevizly = Rley2), (24108) ‘where P,Q and have continuous partial derivatives with respect to zy and 2 and they do not vanish simultancouely, Let the value #~ s4(c) be proscribed along the initial eueve To ' 2 = 26(6), y= ols), 4%, ys, and 29 being continuovsly differentiable functions (a < s < }) Further, for a< $<}, if SE Plats), sls) t0) ~ BQ) whsha)) #0, (13020) then there axists a unique solution z(z,y) defined in some neighbour- hood ofthe initial curve Fo, which satises the pide. and the initial ondition 2(20(9),wls)) = als), ie, the integral surface x = =(z,y) contains the inital data curve G5 = m0lsh.y=wls),2 = 206). Note : The initio carve I is the projection of the initial data eurve Con the 2, »-plane Proof ; The ordinary differential equations (1.10.8) can be solved to obtain a unique family of characteristics (sit), os.0), 9 =vl20, v0, @uon) 220, wz © = 20, Yo 50,4) = 2(5,t), |32 CHAPTER 1. FIRST ORDER PDB. having continuous derivatives with respect to the parameters » and ¢, satinfying the initial conditions 2(s,0) = ze(s),(6,0) = yo(e) and 2(6,0) = (2) (from the existence and uniqueness theorem for the crdinary differential equations). Observe that the Jacobian aCa.0) 3.8) le ” from (1.10.10). Hence wo ean solve the rst two equations of (1.10:11), for s and ¢ in terms of x and y in a neighbourhood of the initial curve t= O,ie, om olzaht= teu) Lat £ = 6(s,y) = =(0(2,y), Hay9))- Nove that, #(z,2) satis the initial condition 38 4(¢e, 06 (8). Moreover, 2 = #(2,0) ‘satisfies the p.de. (1.10.9). For, consider Pb. + Qty = P(zs8s + sete) + Weis) + 2ety), = 4(Pss + Q5y) + 2(Pts + Qty), coat + syy) + alles + ty) tenth = Hence Poet Qhyn a= Re ‘The uniqueness of (2, ») follows from the following argument. Su pose 6(z,y) is not unique, then there are tno integral surfaces whi intersect along the given intial data curve C. ‘Then through each p on the initial date curve, there pastes one and only one char tic curve, ‘Therefore this characeristc curve has to lie on both surfaces. Hence the same family of characteristic curves which through each point ofthe initial data curve lie on both the sur ‘Hence both the surfaces must coincide as both are generated by same family of characteristic curves. Thus ¢(z,y) is unique. Note : The condition (1.10.10) ica tobe the admissibility cond and any initial data curve C satisfying this condition is aid to be 10. QUASE-LINEAR EQUATIONS admissible curve Fig 1.0.1 ‘rence the integral surface (the solution) is generated by the one parant- te fomily of characteristics Oy issuing from each point of the initial data curve C (refer to Fig. 1104). Example 1.10.2 : Solve the initial value problem for the quasilinear equation 22, + 2 = I containing the intial data curve CG : 2 = w=s = 4s for0 1 and since = has different values on them, the solution is not defined in the aquadeant 2 2 1y 21 ° Brample 1.10.4 : Solve the Cauchy problem for 22,4 yay ~ s forthe fnitial data curve C: 2) =2, y= 5? 4 =5, 1S 9S 2 Sol. : We cbserve that dep _ 09 45g? Hp - B29 = 45-9 40, frigss2 ‘Therefore C is admissible for 1 < # <2. Wo solve se ee e tai ‘ich that 2(s,0) = s,y(s,0) = 52, 2(5,0) =». The sclutions of the system of ordinary differential equations which determine the characteristics through the initial deta curve are Bo 642 poet, ease, The solution is obtained by eliminating s and t, ie., F = yeao|(ze ~ v)/22). a Example 1.10.5 : Find the solution of the initial vale problem for the quasilinear equation 2, ~ 22, +2 = 0 forall y aad z > 0, for the itil data curve C : 29 = 0,49 = 8,2) = ~25,-00 < # <00, Sol. : The admissibility condition i satisfied since tym dt BP - BQ=140 vs.oT CHAPTER 1. FIRST ORDER P._D.B|NON-LINDAR FIRST ORDER PD.E, ar iti curves which generate the suas ar obtsinad fom formation to hive unique elution fr the Cauchy probe in the oo non-linear ease. We discuss this problom in the next sestion, ow, ecrcne 1.10.1: Find the integral aac pasing trough he ini at Idata curve C’ 0 = —1,% = 8,2 = of the equation ‘satisfying the following conditions (o+2)p-+2yq = 22, tafying the following condi Fe ccm ynievn (8,0) =0, u(s,0) = cercise 1.10.2 : = with the Initial condition ‘The solutions are found to be gre ie—9) Ta y= —20e' 499, 2 = —2se" (-9f@—y) togral surface passing through the initial yP +y of the equation Pee + y(Bc? + y)z, = 2(207 +4). ‘+ u)(r2—y) . ‘The parameters s and 1 in terms of @ and y are found to be @ intogal surface for the differential equation vf —2 passing through tho initial data curve (28,3), ne. 2? Sry — a? a2, log(2/3) > 2 >0. 3" ‘The solution breaks down at ‘ Consider a non-linear frst order pide problem is usique ifthe initia! curve is smooth and isnot parallel to characteristic curve (ie, it satisfies the condition (1.10.10)). However, Soy. 59,9) =0. aun) the same is not true ia the ease of non-linear equations as seen in th following example. Measeume that f has continuous second order derivatives with respect ‘Example 1.10.6 ; Find the solution of the equation fo its variables z,y, 2,» and ¢ and either f, or jy i non-zero al every 1 Without loss of generality, lot us assume that jy 9 0, eo that 22+) tle), ean solve (1.11.1) at each point (z,y,2) for ¢ a8 ¢= a(x, y.,p), Of Jneed not alays be asinglevahied function ofp, but it ey be which passes through the = ~ axis. sumed that one branch of a possible set of salutions q = g(a: y, 2.9) Ans. Lz =y(da~ 9)/2. Ans, 2:2= 7/2 ‘Observe thatthe above two surfaces passthrough the 2-axib and louge Cone : Let (2, ys) be some point in space. We consider & the given equation. Hence this problem has no unique solution. R ible integral surface 2 = 2(z,y) and the direction ratios (p, 9,1) to Bxample (1.11.2). hot the normal to the tangent plane ‘Non-linear equations unlike quas-linar equations may have many so lutions oF no solution for the Cauchy problem. One ntads some 1 2-9 =Ple— 29) +aly— im), (ua)58 CHAPTER 1. FIRST ORDER P.D-E. to the integral surface ct that point. ‘Tho equation (1.11.1) which is equivalent to a= (2, 10.2059), (13) indicates that p and q are not independent at (20,26)20). Thi im- plies that the integral surfaces are those surfaces having tangent planes belonging to a one-parameter (p) family Consider tho family of planes given by Equation (1.11.2) passing thro (Zo, 40,20) 88 p varies and q is determined by Bauation (1.11.3). planes envelope a cone called the Monge cone having its vertex (Zo,t0,40). ‘Tus if : = 2(2,y) is a solution of (1.11.1) it mast b tangent to the Monge cone at each point (x,y, 2) on the surface, the differential equation (1.11.1) characterizes a fed of cones auch th ‘a surface will be an integral surface if and only if it is tangent to th ‘Monge cone at each point (refer to Fig.1.11.1). The line of cont of the Monge cone with the tengent plane $0 the integral surface ‘och point gives a field of directions. ‘These directions are called tl Characteristic directions at that point and lie along the generators dl the Monge cone. ‘The integral curves of this fold of directione on th surface define a family of curves called the characteristic curves, “Analytic expression for the Monge cone at (2»,Ye,) : ‘As the Monge cone at (5, y, 2) is the envelope of the one ~ param eter family of planes 4% =—(r— 20) +91), where ¢ is given by 9 = (0, %.*0,P), 1.11. NON-LINEAR FIRST ORDER P.D.E. jt can be obtained by eliminating p from the following equations — z= (2 20) + 9(t0,¥,70.9)(¥— to), (tata) jad = (e-20)+y- wt (tts) Cn diferentiting (111.1) with respect op, we got 58. funetion ofp. fone aliminaes & from (1.11.5) and the above ‘equation, then the equations describing ge Monge cone can be written : 4 = aes 200) rom =p(e—0)+4l¥-W)s am y= pi ey “The ast two equations define generator ofthe cone fora given yond the corresponding q given by Equation (1.11.3), ie., the line of contact between the fangont pane andthe con. Example 1.11.1 : Consider p* + q* Let (xo, yo, 20) = (0,0,0). ‘hen the Monge cone is obtained by cliinaingp and fom (1.8) q=yl-y, sapetay, ee 2p” 2g" which is 2? +y? = 2?. This is cone with vertex at (0,0,0) (as every tecond degree homogeneous equation in 2,y and = represents a cone with vortex at (0,0,0)) 3 Note : Observe that in the quasi-linear case Pp + Qq = 2, every tangent plane (1.11.2) a (20,80) contains the folloming line (why?) Po Qk _ Heace the Monge cone degenerates into the above straight line at cach Point.6 CHAPTER I. FIRST ORDER PD.B. [Lot us now find tho equations which give the characteristic curves. ‘As the line of contact between the tangent plane to an integral surface ‘and the Monge cone at that point is the generator of the Monge cone, it can be written from (1.11.6) as 2-% _ yw _ 2% to i Phy take Hence the characteristic directions which lie along the generators of tha ‘Monge cone are given by Jp, fa:Pfp + af) ‘Tho characteristic curves are thus the solutions of de ty ds he Se Phy tthe” (aur) ai, wl, a Gn gate gate Since these three ordinary differential equations for determining the characteristic curve (z(t) y(t),2(t)) cannot be solved because they ia- volve the functions p and g, we require more information regarding the variation of p and q along a characteristic curve, Along such a curve on the given integral surface dp_ de, dy Teas + Bae = at Further, by differentiating Equation (1.11.1) as y, we get respectively (2.1.8) raf + Pubor ey aE tgp = tele tle respect to 2 98 fet Laat poe fete =0, Lyt Lat Sy Sal =. Since py = ¢s, Equations (1.11.9) may now be remitien as _ 4 Be-1-tng ‘Thus for a given integral surface 2 = =(z,y), there corresponds a fam archaea sete on We wah oeouubed ik tek ‘the co-ordinates of the eurve («(¢),v(t),<(¢)) and the numbers p(t) and =r @an.19 LIL NON-LINEAR FIRST ORDER P.D.B, 61 a(t) along the curve are related by a system of five ordinary differential equations given by (1.11.8) and (1.11.10). ‘These five ordinary differen- tial equations are called the characteristic differential equations related to the given pd, (1.11.1), Jn order to determine the integral surface, by the previous discus- sion, we consider the p.d.e. (1.11.1) along with the systom of cherac- ‘teristic equations (1.11.8) and (1.11.10) as a sysiem of six equations, LG,u.4,7,4) Ses hi Quinny fe SPs IS 81S S/F RIS 8/8 $e -h- ta, forthe Sve unknown functions x(t), y(t), a(t) p(t) and g(t). This system is not over-determined. Consider a solution of the last five oquations of (1.1.11), Along such a solution we find that pd, dy ds, dp, dy Bg hg t hg tbe + ne = Soba t Sule * hale + Uefa ~ byle— Inhed— boby— Iba s =o Hence f = constant, is an integral ofthe systomn of ordinary differential uations. Thus we cen conclude that the equation f(r, y,2,9,¢) = 0 is not much ofa restriction. Suppose that f = 0 at (xe, oy 0,200) at #0. Then the unique solution of the characteristic equations pans through this point is such that f = 0 will be satisfied forall ¢ along the solution Characteristic strip : A solution (=(0),u(t), 2(1),r(t), 9(¢)) of the system of ordinary diferen-a CHAPTER 1. FIRST ORDER PDE. tial equations (1.11.11) can be interpreted as a strip. eae o ooo eee Fig. L112 ‘The fit three functions, viz. (2(t) y(t), #(0)) determine a space curve. ‘At cach point of the space curve, p(t) and g(t) define a tangent plane with (p,q, ~1) as the normal vector. The curve along with these tangent planes at each point is called o characteristic strip (refer to Fig.1.11.2) ‘and the curve is called the characteristic curve. ora fixed t given by to, (to, 0,2 Po, 9) is said to define an element of the sip, ie, the point (70, ¥o, 40) on the curve and the correspon tangent plane whose normal is given by (te, qo)—1) at that point. Note : Any set of five functions z(t) u(t), z(@ap(t) and (@) can be interpreted as a strip only ifthe following condition called the ‘strip condition’ a, te dy, gt = POG + ZO, (1.11.12) jsfiod, This isthe conditicn thet at each point on the space curve 'a(0),y = u(t), = z{0), the tangent to the space curve lies on the plane whose norraal hes direction ratios (p(t) q(t), —1) “Any solution (2(1),y(t), 2(t),p(@), (0) of (1-11.11) automatically satisfies the condition (1.11.12) by virtue ofthe second, third and fou equations of (11.11). Lemma 1.11. If an element (70, ye, 20,7042) is common to both an integral suf z= 2(z,y) and a characteristic strip, then the corresponding char teristic curve lies completely on the surface Proof : Let 2 = 2(t,y) be an integral surface, Consider the two ord nary differential equations £0 = plenste vials v) less Hee) = Slasarelaadesl29) efe9b 111. NON-LINBAR FIRST ORDER PD. «a for x(t), y(t) with the initial conditions 2(0) aniquely determine a curve 2 = a(t) and y Consider the curve = = x(t), y= u(t), #= 210) on the integral surface. Then on such a curve 1%, (0) = yp which y(t) in the 2,y- plane. (z(t), ¥(@) im space refy + Ayben (las) = teafy + Ayhoy (Lal44) di at = aly + Aah a (uanas) 2(0) = (20,40) 4(0) = 22(0, 00) 24(0) = 2y(t0, 46) (2, y) 38 an integral surface of the given p.d.e., we have H(2,y, 2(e 0). 202, u)s (2, ¥)) On differentiating the above equation with respect to z and y, we get fet fete Santee fyi = 9 fat fey + fester + Snyty = 0. Equations (1.11.14) and (1.11.15) therefore reduce to (anis) B= - hy ity @auty Therefore the Give functions x = 2(0), y (=), } x(x(t). 910), P= 2(z(1),y(t)) and q = z,(x(¢),y(#)) determine a characteristic ‘trip as they satisfy tho characteristic equations (1.11.11), In addition, ce ‘Saeed determine the unique charaeteristic strip with the initial lement 2, ¥0, }, and gy. However, this curve lies on surf ‘definition. Hence the result. aOHAPTER 1. FIRST ORDER PD. ot J. NON-LINEAR FIRST ORDER PDE. 65 We shall make use of this lemma while proving the uniqueness of ‘the solution of the Cauchy problem for the non-linesr partial differential ‘equations ‘Stop 1 : Given the pds. (1.111) and an inital data curve 2 = 29(0), y= wls), z = 2(s), determine the functions po(s) and go(s) such that the ive Functions 39(¢), 1n(¢) (4), (4) and ole) satay Bquations Initial strip (4.11.18) and (1.11.19). ‘There could be several chotces for po(s) and Let = x0(s), y = yo(s),2 = 20(8) be some arbitrary initial data curve. qo(s). One may expect to find a unique solution for each such choice. Suppose that along this curve we can opecfy the functions po(a) ond | S'ep 2: Once a choice is made for po(4) and qs) i. te initial stip 4o(2} such that together with tho iitet data curve (z(s),2a(s), (8), | B-hosen, we can golve the last five equations of Equation (1-11.11) subs they satisfy the equation ject to the initial conditions x = xo(s), y = y(s), 2 = 208), P= Pols) and ¢ =qo(s) at t = 0. Hence we have the characteristic strips issuing (1.11.18) | from each point of the initial data curve. The corresponding charac- teristic curves generate the coqured integral surface. We demonstrate ‘the method in the following examples, We shall later prove the exis- hi fence and uniquenses ofthe solution in Theorem 1.1.1 whieh actualy +t, (1.11.19) | justifies the above method. * ‘Example 1.11.2 : Pind the solution of the equation F(z0(s), vols), 20(8), 08), a0(s)) and the strip condition dap as as Thon euch an initial element 24(s),ao(6), (4) tos)a9s) 58 sald to , define an initial strip for the initial data curve (za(s), vo(s), zo(s}). The 2=5(F° +0) +(P-2)(¢-9), integral surface may be constructed with the help of the characteristic 2 strip suing from each point of the inital data eurve by pleeing UD) aaay , passes through the axis aon a ea ea sinying | 80h: Ext os fn the possible inital sips. The intel dota curve is tho two equations (L121 18) and (111-10), there can be several initial f £4 * #0 ~ Gs #0 ~ Gh and Equations (111.18) and (1-11-19) respec. strips associated with the same initial data curve. Hence there can’ e bbe more than oae integral surface passing through a given initil data 1 carve Om 58 +0) + o-oo), Note: In the quasi-linear case, we note that both Equations (1.11.18) 0 = py 1+q°0. and (111.19) reduce to P(xo(s), vo(s],20(8))P0 + Q(Zo(2), vol), #0(8))a = (0, 80s 40), and ‘The second equation implies that py = 0. On substituting in the first equation, we get 1 diy de, de 58+ (90) =0, a @=0 or m=2s which ate Linear inp and gp. Further, if PP — ‘and qo are determined uniquely. (Refer to Theorem 1.10.1), Therofore there are two initial strips ‘The above discussion suggests a method of selving the Cauchy probe tes con ivin ab O A tem for the nonlinear pd.e. (1.11.1). ‘The method involves the follo o = ae ing steps. Cis Che = ab (1120) (uanan) 2 = 0,0) = 0, = 0.66 CHAPTER 1. FIRST ORDER PDE. ‘The characteristic equations of this pile. are given by de Gray dy gre ta-s, PE +9-9) +4(P+ 9-5), Ge-a=0 aa 46-920. 1m Case (i), Equations (1.11.22) givez=v+p, y=q—2v, Observe that 4 GOre-2) =pta-z, ptgaraed. Similarly p+ ¢~ y= 2ve'. Hence t=o(2e!-1), y=v(et—-1), p= tvle'~1), g=0(e +3). ‘On substituting in the equation for 2 and integrating, we get = feet) - a=) (On solving for ¢ and v from (1.11.23), we got to yoe de -%y a 2y, (On substituting for t and v in terms of x and y in (1.11.25), we get #= jolts ~ 59). 111, NON-LINEAR FIRST ORDER P.D.E. ‘In Case (ji), Bquations (1.11.22) give 2 = 5+, Observe that fete—w =+e- G@te-s) =o+0-y, pto-a= se Similarly p+ y=, Therefore p ) B=, gay =s—sel, Hence ulu-2), = "2 ‘On integrating and using the initial conditions, we get -yF a Example 1.11.3 : Find by the method of characteristics, the integral ‘surface of pq = zy which passes through the line 2 = 2, y= 0. Sol. : Let us find the inital strips. ‘The initial data curve is (8) = 8; wls)=0, z0(s) {In this case, Bquations (1.11.18) and (1.11.19) respectively become Pol)ao(s) t= py1+q@-0 Po= 19 =0, (unique initial strip),68 CHAPTER 1. FIRST ORDER PDE. ‘The characteristic equations are dr) dy) de op Bah Bh Ea PG Hence a= Ad+ Be“, g= Ae — Bet ‘Similarly y=Cel+Det, p=Cet— Der & apg u DAC + 2BDe“™ = (BC + AD), 2= ACE — BDe™ — Using the initial conditions 2(BC + AD)t +B. A+B=s,A-B=05A=B= C4+D=0,0-D=190=-D AC-BD4+E=o, Finally, r= scosht,y = sinht,z = scosh*t, p = cosht,7 = esinht. Hence te sufoce ping Hhvough the Ital daca curve gen by Farts). Example L1La p+ yq—pg = 0 and obtain the equ: the cune @: z= 2/2,y=0, Sol. : The initial data eurvo is ad 1h carat tip’ of thn exutial of the integral surface through (0) = 28, als) =0, 20(s) ‘The Equations (1.11.18) and (1.11.19) become respectively 2spo ~ po = 0 and 1 = which implies that 1 =A and q =2s To = 5 and ge JIL NON-LINEAR FIRST ORDER PD.E, o ‘Since only one initial strip is possible through the given initial data curve, we can find only one solution, ‘The characteristic equations are ge Bag a ‘The characteristic strip issuing from the initial element at ‘e’ on the ‘Therefore the integral surface through C is 2? — 42? + Bryz ‘Exercise 1.11.1 ; Find an integral surface of p’x + gy — ‘containing the initial line a9(s) = s, yo(s) 08) dns: (2 (eto Bsample 104) ‘Exercise 1.11.2: Find the solution of 2 = p*—. the curve C': 9 = $,4o = 0,29 = —}s? Hint : (Two initial strips are possible) = 1.11.3: Determine the to eoltios of th equation 7a =1 ? which passes through ng through tho stzaght line Cy ~ 20,44 = 2ey20 = 50- cise 1.11.4 : Find the characteristic strips of the equation 2+ px tay = 1+ pany’, assing through the intial date curve C's zy = 9,9) = 1yzy =~. Exercise 1.11.5 : Pind the integral surface of the equation pq Passing through C: a= 0, 5,40 = 57. Bie, «= E49) fEcercise 1.11.05 Find the ndegral surface ofthe equation 2 = p!-39%, ‘Passing through C': 29 = 9, yy = 0,2) = s*. Show that there are two. aside intial strips = 2s, qo ~ ‘Ans. : For the initial strip (¢,0,s?,2s,s), the characteristie strip is (ets. =6s{c!—1),s%e™, se! set). ‘The integral surface is given 2r4y)?/A. (Pind the integral surface forthe other initial strip).10 CHAPTER 1. FIRST ORDER PDB. jn conditions, the We will now state and prove that under certain conditions, solution to the Canchy problem for @ non-linear p.die, exists and is unique. ‘Theorem 1.11.1 : Consider the pd. H(a4,2,2,9) pose that the initial values * = zo(s) are specified along the initial cu vega seed = Wl. 2 55 boone ee devas’ Suppee (and ws) Bae determined such that ‘Fl 20(2), aoe) 208), Pos), (8) = 0, and. where py and gy are continuously differentiable functions of s. Tf, addition, the five Functions 2, 20, Po, and ge satisty jal curve 1 then in some neighbourhood of each point of the init exists one and only one solution 2 = z(2,y) of (1.11.26) such that slee(0)w(0) aloes). 09) sy(oels). es) = int surface z = 2(x,y) contains the initial strip. Peta in ucomider th oyiem of arcterti oats de, yl, ie Hug, Ban, Sorta ® (+08), Grants +P) (fy + a4), with the initial conditions © als), v = wls), 2 = a0(8) P= (s) at t= 0. We can solve this system of ordinary Lil, NON-LINEAR FIRST ORDER P.D.E. n ‘equations to obtain (from the existence and uniqueness theorem for the initial value problems for a system of ordinary differential equations) EHX!) VAY, v= 20, p= Pls, 9= Als), where X,¥; Z,P and Q have continuous derivatives with respect to s ‘and ¢ and such that thay satisfy the initial conditions X{8,0) = 20(s), ¥(5,0) = ols), Z(s,0) = ao(0), P(2,0) = pols), Q(8,0) = an(s) Proposition 1.11.1 : The characteristic curves 2 = X(s,t), y= Y(a¢),z = Z(s,t) indeed form an integral surface. That is, ifs and ¢ are solved in terms of x and y from the first two equa- tions, then the last equation gives the solution expressed in the form 2 = Z(slz,y),i(z,9)) = 2(2,9). Let (xo, to) be a point on the initial curve. Since the Jacobian AY)! p deo _ 5 don Fe [ag de es at evry penton te intial curve there exists neighbourhood N(ze,4) such that the Jacobian STF? 40, In this aveNbouood Ne.) we have the following wall dnd functions. 9 = (2,9), (o(z,4), (x,y), t= tzu), y= ¥ (s(z.y),t(a,9)), B(stz,y),t(a,y)) = 2(0,2), P(s(2.y),t(2,9)) = Pez9), Qs(x,y).t2,y)) = g(2,4), We will now show that 2(2,y) satisfies (uan.27) S(9.7(2,9), 2612 y), H12,4)) = 0. 1 suffices to show that p= sp and q = x since fl, 1.2.4) = 0-05 the characteristic strips are essentilly solutions of (1.1111). Let us consider Zz U(a.t) = 2, - PX, - O%.CHAPTER 1. FIRST ORDER PDB. | Ll. NON-LINEAR FIRST ORDER PD. Ps (6,0) = 2 ~ pott2 — qo = 0, by the strip condition for J fill and oy, Baie: the determinant of the coefficients of these linear (Gb tadial Gimants. We shall show tEle 0 = 0 for allt systems it 7-7 This determinant is non-zero inthe asighbourhood i sGensies 1N(, yo). Hence we have au o ns — PAX, = QW, — PXu~ QYou, o Dy (x9) = 2(2,0), (0,0) = (2,0), — (2(CENTRaLYS P(z,¥) = za(2,0), a(t, v) = 24(2, 9), gion) Op which was to be shown, Hence the result, aw P(s.4) = ae(2(syt), (ait), Qlas4) = ayle(s,t),ula, ER > = Le PR. QH) + Pk + UM Qs PX, S BOF P Ly + Qsfat (fet feP)Xet (fy + .O¥o, by making use of the strip condition for characteristics, This con be remitien es 2X, + HY fest SyPat fiQe fel2s— PX — Wi) fi-£U, =-fU, — (oince f=0 Wo and &) ‘Therefore, for a fixed e (Le. along the characteristic curve), U satisfies the following ordinary differential equation au de Also, the solution 2 = =(z,y) contains the initil strip. For, #(0)¥0) = 2(2(6, 0), (6,0)) = (6,0) = als), #2(20, 4a) = p(t, w) = pols, 0), uls,0)) = Pls, 0 Ay(0,¥0) = a(z0, vo) = a((s,0},v(s,0)) = Q(s,0) = Proposition 1.11.2: 2= 2(z,y) so determined is unique Suppose instead thet there ore two solutions, sey, 2 = 2i(z,y) and = 1(z,) containing the initial strip. Consider any point on the inital deta curve. ‘These two surfaces contain the initial element at that point, and hence the characteristic strip issuing from this initial ement through that point (refer to Lemma 1.11.1). Therefore the cor responding characteristic curve has to lie on both the surfaces. As this Inppens for every point on the initial data eurve, both these sutfaces ae generated by the same set of characteristic curves and must hence coincide; Thus the solution is unique. a $0, which has a solution = U(ero(~ [fait Combined with the condition that 7 = 0 for t = 0 the above expresioa implies that U = 0 for all ¢. Therefore 22 PK + Qh ‘Ch der it uti () 2, = PXs+ QM, i) 2, = 2X, +2¥., (iv) Zr = zaXr + aYe (i) It has been proved above. (ii) It is the characteristic equation, (Gi) and (iv) are obtained by differentiating the identities of Bquati (11127) Observe that (P,Q) satay () and (i) and (zn) satis (8) (iv). Av ayatam of simaltancoas equations () and (i) ae the same (i) = PX AONChapter 2 Second Order Partial Differential Equations 2.1 Genesis of Second Order P.D.B. Definition 2, Partial differential equation is suid to be a second. order semilinear p.d.e, if it can be put in the form Rx )lae + (2,4) oy + Te, yyy + aby 4 tes ty) = 0, (2.2.1) where H+ $?47* #0 and RS and Tare continuous functions of and v. Definition 2.1.2 : A function u = u(a,y) is said to be a regular solution of (2.1.1) in a domain D © Rx R iu € C%D), and the function and its derivatives satisfy Bouation (21.1) identically in 2 and y for (z,u) € D. Example 21-1: If f € C°(R) then w= f(x+t) isa solution of the second onder pd.e Wee te = 0. a ‘We now discuss some exernples of second order partial differontial equations which arise in physics and mathemetics. ‘Transverse vibrations of astring: Let y = y(z,t) be the transverse | displacement from the mean position (x-axis) ofa string at time ¢ at the point z, Consider a small portion of the sting As between the to points P and Q. ‘The focees acting on this part of the string are tensions at P and Q. We neglect the weight ofthe sting, ‘The equations of motion aro (refer to Fig. 2.1.1) 8CHAPTER 2. SECOND ORDER P.D.E. 6 (@) in e-ditection (assuming n0 displacement in the x direction) ‘Tye08 Ys = Ty cosy =T (say), (i) in g-direction (oA2)e = Tsinys—Tysinds , ‘T(tandr— ten vs) . where p is the linoar density of the string. Fig. 211 tandy = (ye)|n) tanva = (ve)le = (us)lo + (Werle, inhere |» and lo indicate the evaluation at the pointe P and Q respes- tively. Hence (ods) yu = Tl(ve) |p + (ves) oz ~ (ue)le] » ‘Taking the limit as Ae + 0 and Ax - 0 after dividing both sides of the above equation by Az, we obtain Th vit lye k€ 1 (ies if the slope is small everywhere), eget nee tee = OHS ‘The above equation is calle the linear one-dimensional wave equation. Uf the conductivity & is constant throughout the body, then ‘which io referred to as the heat conduction In the one-dimensional cee, 2.1. GENBSIS OF SECOND ORDER P. 7” Heat conduction equation : Let us consider a homogeneous, isotropic folid, (Homogoncoue means that the material propertiog are transla- tional invariant and isotropic means that the material properties are Se seme in all directions). et Ve any arbitrary volume inside the Solid boundod by a eurfece S. Let 6V be a volume slement. The heat energy stored in 6V is equal to cpuSV, where c isthe specific heat of ‘the eclid, is its density and u is the temperature which isa function of pestion and time, Therefore the oa at eney in V = ff, out Let 65 be a surface element. ‘The heat ow across 53 = kVu- 5S ‘where @ is the outward drawn normal to the surface S and k is the ‘thermal conductivity of the solid. Therefore, the total flux across using divergence theorem, is given as follows [fp vw-nts= [ff 9-a oo, ‘Tho rate of change of heat ens Sie, ergy in V = the flux of heat energy across al hee = ff f.9- wou, Sih (2-9-0970) av=0 As V is arbitrary and the integrand is continuous, we have ove Va) =0, SKv%y, a equation, where = k/ep. ‘this equation becomes@ CHAPTER 2, SECOND ORDER PDE. Analytic function of z + If f(2) = u(ayy) + #u(ayy) is analytic in (= 2+ iy), then it is well known that u and v satisfy the Cauchy- ‘Riemann equations le = Uy 9 ty = Ys Therefore tert ty =O. Similarly Ue + ty = 0. ‘These equations are called two-dimensional Laplace's equations. ‘Second order partial differential equations widely occur in many other areas of physics, chemistry, mathematics and engineering. 2.2 Classification of Second Order P.D.E. Consider the following second order semi-tinear p.d.e (23) Le + 9(2,95%. te ty) here e b= Raney +stengegy +Trad R447? #0 end R,S and T are continuous functions of 7 and Pieter tsumete thre frelon pena canons paral derivatives with respect to x and y- We change the independent variables (x,y) to (G7) €= Eloy), n=nl2,9): Note : We essume that (Em, ~ &te) # 0 50 that the transformation is invortible at leet locally Tb is easy to 00 that te = uke tte» ty = by tay» May = techaby + Meret + tekey tty + Maye + ley + has = UeeRE + tenSeTte + HeSae + ugerake + toa + tates » yy = they + Ue + Heke Atty + Yon thy 22. CLASSIFICATION OF SECOND ORDER PD.B. ‘Therefore Poles + Sttey + Tiyy = vee PEL + S66, +782) + gy 2REatie + S(Eemy + may) + 20 Evry] + ung Rid + Samy +778) + a function of Gn, ttn). ‘Therefore Equation (2.2.1) becomes Ales y)tgg + 2B (bay Bian MU + Ces a) = GUE, tte tn (222) where A(u,9) = Ru? + Sev +Tv, (2.2.30) Blu viva.) = Rests + ZS(uva + uor) + Tome (2238) Exercise 2.2.1 : Show that AlGer€s)A (Mes) ~ BSS ReyM) = (ART ~ $°)(Gary ~ &yne}*/4. (2.24) ‘he problem now i o choose € and no that Bauation (2.2.2) takes a Case (i) : S*~4RT > 0. (225) ‘Then we eholl show that £ and n can be so chosen that the coeficients of ugg andy in (22.2) vanish Consider Ra! + Sa+T = 0. ‘This equation has two real distinct roots A:(z)y) and Aa(z,y) due to the condition (2.2.8) We choose € and 7 such tha % 4% ena ds aly as Es ‘Those are first order partial differential eq Az,v) = & and fo(z,y) = are the 0. Hence Equation (222) request Ft oem (2.2.6) Bein ~ HE Mot) (2.26) ‘The curves (2,9) = constant and n(z,g) = constant, are called the characteristic curves of Equation (2.21). Case (ii) + 5*— ART = 0. (2.2.7) fi this caso, the roots ofthe equation Ra? + Sa+ T= 0 ecincide (soy 2a, Define € = f(esy), where f(z,2) = 1 is the solution of $+ san) =0. ‘Take 1) as any arbitrary function of = and y independent of € and from (2.2.4) observe that B = 0 since A(Ex,&) = 0 ince n is independent of £, A(ta,my) 0. Therefore Equation (2.2.2) reduces t0 ie oF Case (jil) : 5*— ART <0. (229) ‘This is formally the same as Case (i). However here the roots are Gomploe. Proceeding, as in Case (i), ve find that Bauation (22.2) Seduees to the form (2:26) but that the variables € end 7 are not Teal and ane in fact complex conjugates, Therefore there are no real characteristic curves in this cose, ‘We make te further transformation MEM, 0, WE, Un) (2.2.8) 1 i deen, B= 30-9) Equation (2.2.2) then becomes eu Ou Feat age 7 mPa tentl: (2210) 2.2 CLASSIFICATION OF SECOND ORDER P_D.B. aL Classification of Second Order Partial Differential Equations (@) S@-4RT>0 Hyperbolic type Eg: Wave equati () 8 —4RT=0 Parabolctype By: Heat equation, (ii) S°-4R7 <0 Blliptic type Bg: Laplece’s equation Canonical forms : Equations (2.26), (2.2.8) C ), (2.2.8) and (2.2.10) are said to be the canonical forms ofthe hyperboli poe be he sania oat ofthe Jovi: pnb a lite type t Beample 2.2.1; Redue the equation tag — 24 =0 toa nonin Sok In this 0, Bm 18 = OT = —s?, en S* — ART = ds? > 0 (hyperbolic Rab +Sa+T =O Deca gre Te yo ast 2h=1, b= cna and are sememuimse qgtino. Throne § = 9+ 2 and = 9%, Changing the i wusharituccl ada t= 0 — att ty a= Bg Dae hg te Ag te Bey ag ‘Therefor the given equation becomes -2 L wen = Gas ¥e— M9) = Tey le) a Bxample 2.2.2: Reduce the equation Piss —2attey ty = Lee ay,CHAPTER 2. SECOND ORDER PDB. ONE DIMENSIONAL WAVE EQUATION Example 2.2.4 ; Reduce the equation (n= 1s — yyy = ‘where n is an integer to a cancnical form. Sol. If m= 1, the equation reduce t0 ty = —!uy which ofthe | ‘canonical form (parabolic). Then a jonical form and solve it. Bi alas ae, $48 = 0 (parbalic tyne) ‘The equation Ro? + Sa-+T7'=0 becomes y?a? — 2aya + 2° = 0. (va- 2 =059 ‘A eolution of ie B50 yo iso +9? = ex. Therefore Thus yt, = f(a), where f(s) is an arbitrary function. Henoe ‘ty = Ff (2) and thus u = f(z)logy + 9(z), where (a) is an arbi- teary fnction. ifn > 1, then it is hyperbolic in type. Here € +e ra y= 2—y'", Then the given equation transforms to the canonical gare. Lat us choose a 26) m ‘Observe that 7s independent of for, ges 40. The given equation transforms to tue) = ‘The solution is u(ryy) = fle+ 4") + 9(2— 9"). tig =O: 4 a Note : The arbitrary functions f aud g that oveur in the solution in m= HO Bramples 222 and 2 4 havo to be C2 functions eo that the solution at is regular solution a = aafortae 5 | Bxsrelse 2.2.2: Reduce the following into canonical forms and solve a, ule,o) = S02 +e? - A) tle +9) a ion tes +2%ty =O toa canonizal | (i) toy +2toy + 17ty =0. Bxample 223; Rete ention tt oe (i) ee = Aaa iv) ee ~ (coch) toy = 0. (¥) tee $v, = 0 in the region 2 <0. (Vi) Zilee + 2V Diy + Vy ~ he = 0. (i) uae + LeMay + ey = 0. (ill) (6in*2)tne + 2(c080)tay ” ty = 0. : (3) MY Man = 200? = tay + (Y — Ugg + aye — ty = 0. ‘Sol. : In this case, S*— 4RT = ~42? <0 (elliptic type). ‘The equation Ra” +Sa+T = 0 becomes a* +2? = 0. This implies that ai ei. Here 2 oe gat 5 nat y Therefore 4 a=5 bry ‘The given equation then transforms to the ‘canonical form 2.3 One Dimensional Wave Equation In Sections 2.3.1 to 2.3.3 and Section 23.5, we will use y as a dependent variable with + end é as independent variables, 1 alle ee + 8 =~ 34" :