i

LECTURE NOTES IN PARTIAL DIFFERENTIAL EQUATIONS Fourth Edition, February 2011 by Tadeusz STYS

University of Botswana

ii

Contents
1 Solution of Partial Dierential Equations 1.1 The General Solution of PDE . . . . . . . . . . . . . . . . . . 1.2 First Order PDE with Constant Coecients . . . . . . . . . . 1.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Classication of Partial Dierential Equations of the Second Order 2.1 Hyperbolic, Elliptic and Parabolic Equations . . . . . . . . . . 2.2 The Standard Form of Hyperbolic, Elliptic and Parabolic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Transformation of a Hyperbolic Equation into a Standard Form 2.4 Transformation of an Elliptic Equation into the Standard Form 2.5 Transformation of a Parabolic Equation into the Standard Form 2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Hyperbolic Equations 3.1 The Initial Value Problem for Wave Equation . . . . . . . . . 3.1.1 DAlemberts Solution . . . . . . . . . . . . . . . . . . 3.1.2 The Initial Boundary Value Problem for Wave Equation 3.2 Solution to the Finite Vibrating String by Separation of Variables 3.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Parabolic Equations 4.1 Initial Boundary Value Problem . . . . . . . . . . . . . . . . . 4.2 Solution by Separation of Variables . . . . . . . . . . . . . . . 4.3 Transformation of Non-homogeneous Boundary Conditions to Homogeneous . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 More Developed Heat Equation . . . . . . . . . . . . . . . . . 4.5 Non-homogeneous Heat Equation . . . . . . . . . . . . . . . . 4.6 Fundamental Solution for the Heat Equation . . . . . . . . . . 4.7 Fundamental Formulae . . . . . . . . . . . . . . . . . . . . . . 4.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iii

1 1 3 7 9 9 11 12 18 21 25 27 27 27 31 31 40 43 43 48 50 53 54 57 58 60

iv

5 Elliptic Equations 63 5.1 Laplace Equation . . . . . . . . . . . . . . . . . . . . . . . . . 63 5.2 Boundary Value Problems for Laplace Equation . . . . . . . . 65 5.3 The Maximum Principle for Laplace Equation . . . . . . . . . 66 5.4 The Maximum Principle for Poisson Equation . . . . . . . . . 70 5.5 The Maximum Principle for Helmholz-Poisson Equation . . . . 73 5.6 Boundary Value Problem for Laplaces Equation in a Rectangle 74 5.7 Boundary Value Problem for Laplaces Equation in a Disk . . 76 5.7.1 Fundamental Solution of Laplace Equation . . . . . . . 78 5.7.2 Theorem on representation of harmonic functions . . . 80 5.7.3 Greens Function . . . . . . . . . . . . . . . . . . . . . 83 5.8 Helmholz Equation and Eigenvalue Problem. . . . . . . . . . . 88 5.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

PREFACE
These lecture notes are designed for undergraduate students as a complementary reading text to an introductory course on Partial Dierential Equations. It is assumed that the students have basic knowledge in Real Analysis. The notes have been used for teaching the course MAT426 (PDE), Partial Dierential Equations at the Faculty of Science, University of Botswana.

University of Botswana Department of Mathematics, February 2011

Tadeusz STYS

Chapter 1

Solution of Partial Dierential Equations

1.1 The General Solution of PDE

The general solution of a partial dierential equation (PDE) is considered as a collection of all possible solutions of a given equation. Example 1.1 Consider the following rst order linear PDE equations ux (x, y) = 2 x + y, < x, y < (1.1) (1.2)

uy (x, y, z) = x + 2 y + z, < x, y, z <

Solution. Let the variable y in (1.1) be xed, and let us integrate both sides of equation (1.1) with respect to the variable x. Then, we obtain u(x, y) = x2 + y x + f(y) (1.3)

for arbitrary dierentiable function f(y). Thus, all solutions of equation (1.1) are of the form (1.3), where f(y) is any dierentiable function. Similarly, let us integrate both sides of the equation (1.2) with respect the variable y, when the variables x and z are xed. Then, we obtain all solutions of equation (1.2) in following form: u(x, y, z) = xy + y 2 + z y + g(x, z), for arbitrary dierentiable function g(x, z). Example 1.2 Consider the following second and third order linear PDE equations uxy (x, y) = 2 x + y, < x, y < (1.5) (1.6) (1.4)

uxyz (x, y, z) = x + 2 y + z, < x, y, z <

1

Solution. Let us integrate both sides of equation (1.5) with respect to the variable y, when the variable x is xed. Then, we obtain 1 ux(x, y) = 2 x y + y 2 + f(x) 2 (1.7)

for arbitrary dierentiable function f(x). Next, we integrate equation (1.7) with respect to the variable x, when the variable y is xed. Then, we obtain 1 u(x, y) = x2y + y 2 x + F (x) + g(y) 2 (1.8)

where F (x) is an antiderivative to the function f(x) and g(y) is an arbitrary dierentiable function of the variable y. Thus, all solutions of equation (1.5) are of the form (1.8), where F (x) is an antiderivative to the arbitrary function f(x), and g(y) is any dierentiable function of the variable y. Now, let us integrate both sides of equation (1.6) with respect to the variable z, when the variables x and y are xed. Then, we obtain 1 uxy (x, y, z) = x z + 2y z + z 2 + f(x, y) 2 (1.9)

Next, we integrate equation (1.9) with respect to the variable y, when the variables x and z are xed. Then, we obtain 1 ux(x, y, z) = x y z + y 2 z + z 2 y + F (x, y) + g(x, z) 2 (1.10)

where F (x, y) is an antiderivative to f(x, y) with respect to the variable y, and g(x, z) is an arbitrary dierentiable function of the variables x and y. Finally, we integrate equation (1.10) with respect to the variable x, when the variables y and z are xed. Then, we obtain 1 1 u(x, y, z) = x2 y z + x y 2 z + x yz 2 + F F (x, y) + G(x, z) 2 2 (1.11)

where F F (x, y) is an antiderivative to the antiderivative F (x, y) with respect to the variable x, and G(x, z) is an antiderivative to g(x, z) with respect to the variable x. Example 1.3 Find all solutions of the following non-linear PDE equation: (uxx)2 + (uyy )2 = 0, < x, y < (1.12)

Solution. We note that uxx = 0 and uyy = 0 . All solutions of the equation uxx = 0 are in the form u(x, y) = f(y)x + g(y) (1.13)

and all solutions of the equation uyy = 0 are in the form u(x, y) = q(x)y + r(x), (1.14)

for arbitrary dierentiable functions f(y), g(y), q(x) and r(x). Then, all solutions of equation (1.12) have common part which is included in both (1.13) and (1.14). So that, the solutions which have both forms are in the following form: u(x, y) = a x y + b x + c y + d, for arbitrary constants a, b, c and d. Example 1.4 Consider the following rst order PDE equation x ux 2 x uy = u, Show that u(x, y) = xf(2 x + y) (1.16) is the solution of equation (1.15) when f is any dierentiable function. Find the solution within the family of solutions (1.16), which satises the condition u(1, y) = y 2, < y < . < x, y < . (1.15)

Solution. Let us note that f(2x + y) is the function of one variable t = 2x + y, < t < . By dierentiation, we nd ux (x, y) = f(2x + y) + 2 x f (2x + y), Hence, we compute x ux2 x uy = x f(2x+y)+2 x2 f (2x+y)2 x2 f (2x+y) = x f(2x+y) = u(x, y). Now, applying the condition u(1, y) = y 2 to the solution (1.16), we nd function f. Thus, u(1, y) = 1, f(2 1 + y) = y 2. Let t = 2 + y and y = t 2. Then, f(t) = (t 2)2 . We can choose f(2x + y) = (2x + y 2)2 . Let us note that the solution u(x, y) = x(2x + y 2)2 satises the condition u(1, y) = y 2. uy (x, y) = x f (2x + y).

1.2

First Order PDE with Constant Coecients

a2 + b2 > 0.

Let us consider the following equation a ux + b uy + c u = f(x, y), (1.17)

where a, b, and c are constant coecients, and f(x, y) is a given continuous function. Let us consider the case when b = 0. Then, we shall transform the equation a ux + b uy + c u = f(x, y),

given in x, y coordinates to the equation b vz + c v = f( in the new coordinates w, z w = b x a y, z = y Hence, we nd w+az , y = z. b In terms of the new coordinates, we compute x= a ux + b uy = a(vw wx + vz zx ) + b(vw wy + vz zy ) = (a b b a)vw + b vz = b vz . Thus, in the new variables, equation (1.17), takes the form b vz + c v = f( Now, we shall solve the equation b vz + c v = g(w, z), for g(w, z) = f( (1.19) w+a z , z) b (1.18) w + az , z) b

w+az , z) b In order to nd the general solution of equation (1.19), we divide the above cz equation by b and multiplying by the factor e b , to obtain cz cz cz c 1 e b vz (w, z) + e b v(w, z) = g(w, z)e b , b b cz cz 1 [e b v(w, z)] = g(w, z)e b . (1.20) z b Integrating both sides of equation (1.20) with respect to z, and multiplying by cz the factor e b , we obtain the following general solution of equation (1.19) cz 1 v(w, z) = e b [ b cz g(w, z)e b dz + C(w)], (1.21) or

where C(w) is an arbitrary dierentiable function of the variable z. In the case when b = 0, we have already the equation in the form (1.19), so that aux + c u = f(x, y).

The new function v(w, z) u(x, y) = u( w+a z , z). b

Now, we can solve equation (1.19) by formula (1.21) to get the solution v(w, z), and then to obtain the solution u(x, y) = v(b xa y, y). Below, we shall present some examples following the above solution of the rst order linear equation with constant coecients. Example 1.5 . (1a) Find all solutions of the equation u u +u=1 x y (1b) Find the solution of equation (1.22) which satises the condition u(x, y) = 2 for y=x (1.22)

Solution 1a. Note that the coecients a = 1, b = 1, c = 1 and the function f(x, y) = 1. First, we transfer the equation to the form vz + c v = g(w, z), by the mapping So that we have w = x y, x = w z, z=y y = z.

We consider the new unknown

v(w, z) = u(x, y) = u(w z, z), for which, we compute the expression ux uy = (vw wx + wz zx) (vw wy + wz zy ) = vz . Since g(w, z) = f(w z, z) = 1, therefore, we obtain the equation vz (w, z) + v(w, z) = 1. Multiplying the above equation by ez , we have ez vz (w, z) + ez v(w, z) = ez or z [e v] = ez . z

By integration with z ez v = ez + C(w) or Hence, we nd the solution v(w, z) = u(w z, z) = 1 C(w) ez and u(x, y) = 1 C(x y)ey , v(w, z) = 1 C(w)ez

for arbitrary dierentiable function C(w). Solution 1b. For y = x, we nd u(x, x) = 1 C(2x)ex = 2. So that C(2x) = ex t Let t = 2x. Then, we have C(t) = e 2 and the solution yx x y y u(x, y) = 1 + e 2 e = 1 + e 2

satises the condition u(x, x) = 2. Example 1.6 Find the general solution of the equation 3 ux 2 uy + u = x. Solution. We consider the new variables w = 2 x + 3 y, Hence w3z , 2 Then, we introduce the unknown x= v(w, z) = u(x, y) = u( Now, we compute 3 ux 2 uy = 3(vw wx + vz zx) 2(vw wy + vz zy ) = 2vz . Equation (1.23), in the new variables becomes 1 2vz + v = (w 3 z). 2 z Dividing by 2 and multiplying by the factor e 2 , we obtain z z 1 [e 2 v(w, z)] = e 2 (w 3 z). z 4 (1.24) z = y. y = z. (1.23)

w 3z , z). 2

(1.25)

Integrating both sides of (1.25) with respect z, when w is xed, we nd z 1 e 2 v(w, z) = w 4 z z 3 e 2 dz + ze 2 dz + C(w) 4 z z z 3 1 = we 2 + [ze 2 (2) e 2 (2) dz] + C(w) (1.26) 2 4 z w 3z = e 2[ 3] + C(w), 2 2

where C(w) is an arbitrary dierentiable function of the variable w. Hence, we nd the solution z 1 v(w, z) = [w 3 z 6] + e 2 C(w). 2 and coming back to the original variables, we obtain the general solution of equation (1.23) in the following form y y 1 u(x, y) = [2x + 3y 3y 6] + e 2 C(2x + 3y) = x 3 + e 2 C(2x + 3y). 2 Let us observe that choosing the function C(2x + 3y), we obtain a particular solution. For example, the particular solution is y u(x, y) = x 3 + e 2 . for C(2x + 3y) = 1, Indeed, we have y y 3ux 2uy + u = 3 + e 2 + x 3 + e 2 = x. Also, for C(2x + 3y) = 2x + 3y, we have the particular solution y u(x, y) = x 3 + e 2 (2x + 3y).

1.3

Exercises

Question 1. Find the general solution of the equations (a) ux = 3x + 2y, (b) uxy = x y, (c) uxyz = x + y + z.

Question 2. Find all solutions of the equations (a) ux 2uy + u = x + y, (b) ux + 2uy + 3u = x + y, (c) ux uy + u = 0. Question 3. 1. Find all solutions of the equation ux + uy 2u = y 2. Find the solution of the equation which satises the condition u(x, 1) = x Question 4. 1. Find the general solution of the equation ux + 4uy + 2u = 5 2. Find the solution u(x, y) of the equation which satises the condition u(1, y) = 2 for <y < for <x<

Chapter 2

Classication of Partial Dierential Equations of the Second Order

2.1 Hyperbolic, Elliptic and Parabolic Equations
2u 2u 2u + 2b(x, y) + c(x, y) 2 + x2 xy y u u +d(x, y) + e(x, y) + g(x, y)u = f(x, y), x py

We shall consider the following form of partial dierential equations: Lu a(x, y)

(2.1)

where u(x, y) is an unknown function and the coecients a(x, y), b(x, y), c(x, y), d(x, y), e(x, y), g(x, y) and the right side f(x, y) are given functions of the variables (x, y) in the domain . For the classication purpose, we consider the following dierential operator of the second order associated with the main part of equation (2.1) L0 u = a(x, y) 2u 2u 2u + 2b(x, y) + c(x, y) 2 x2 xy y u u + e(x, y) + g(x, y)u. x y

We note that the dierential operator L1 u = d(x, y)

of order one does not eect the type of the equation. The type of an equation is determined by the operator L0 of the sedcond order. Classication. All the equations of the general form (2.1) are divided in three the following classes pending on the sign of the discriminant b2 a c.
9

10

1. (2.1) is called hyperbolic equation if the discriminant b2 a c > 0, for all (x, y) , 2. (2.1) is called elliptic equation if the discriminant all (x, y) , 3. (2.1) is called parabolic equation if the discriminant all (x, y) . Also, 1. the operator L is called hyperbolic operator if the discriminant b2 a c > 0, 2. the operator L is called elliptic operator if the discriminant b2 a c < 0, 3. the operator L is called parabolic operator if the discriminant b2 a c = 0. Example 2.1 . The wave equation 2u 2 u 2 = 0, t2 y is the hyperbolic equation, since the discriminant b2 a c = 02 1(1) = 1 > 0. Laplaces equation 2u 2u + = 0, x2 y 2 b2 a c < 0, for b2 a c = 0, for

is the elliptic equation, since the discriminant b2 a c = 02 1 1 = 1 < 0. The heat equation u 2u 2 = 0, t x

is the parabolic equation, since the discriminant b2 a c = 02 1 0 = 0.

11

2.2

The Standard Form of Hyperbolic, Elliptic and Parabolic Equations

The following standard or canonical forms of hyperbolic, elliptic and parabolic equations are considered: 1. The rst standard form of a hyperbolic equation 2u u u = f (t, x, u, , ), tx t x 2. the second standard form of a hyperbolic equation 2u 2u u u 2 = f (t, x, u, , ), t2 x t x 3. the standard form of an elliptic equation 2 u 2u u u + 2 = f (t, x, u, , ), 2 x y x y 4. the standard form of a parabolic equation u 2u u = k 2 2 + f (t, x, u, ), t x x Here, f is a function independent of the second derivatives. In order to transform equation (2.1) into its standard form, we consider the new variables = (x, y), = (x, y). For the composed function u(, ) = u((x, y), (x, y)), we compute the following derivatives: u u u = + , x x x and 2u 2u 2 2u 2u 2 u 2 u 2 = 2( ) +2 + ( ) + + x2 x x x 2 x x2 x2 2u 2u 2u 2u 2u = 2 + + + + xy x y x y y x 2 x y u 2 u 2 + + xy xy 2u 2 2u 2u 2 u 2 u 2 2u = 2( ) +2 + ( ) + + y 2 y y y 2 y y 2 y 2 u u u = + y y y

12

Now, substituting the above relationship to equation (2.1), we obtain the following equation in terms of the variables and : A where 2u 2u u u 2u + 2B +C 2 +D +E + Gu = F 2 A = a( (2.2)

2 ) + 2b + c( )2 x x y y + b( + )+c B=a x x x y y x y y 2 2 C = a( ) + 2b + c( ) x x y y 2 2 2 D = a 2 + 2b +c 2 +d +e x xy y x y 2 2 2 E = a 2 + 2b +c 2 +d +e x xy y x y G = g, F = f. It may be veried that 2 B 2 A C = (b2 a c)( ) , x y y x where x y J (, ) = = = 0, x y y x x y is Jacobian of the mapping = (x, y), = (x, y), (2.3) Thus, the type of the equation remains the same in the new coordinates and , provided that Jacobian J (, ) = 0. We note that, every equation of the general form (2.1) can be transformed by a transformation = (x, y), = (x, y). (2.4) to a standard form.

2.3

Transformation of a Hyperbolic Equation into a Standard Form

2u = f (t, , , u), 2u 2 u 2 = f (t, , , u), 2 t x

The standard forms of a hyperbolic equation in the two variables , are: or

13

The wave equations as the representatives 2u =0 or 2u 2 u 2 = 0. 2

To obtain a standard form put in (2.2) the conditions: 2 ) + 2b + c( )2 = 0, x x y y (2.5) 2 2 C = a( ) + 2b + c( ) = 0, x x y y We shall call the curves determined by equations (2.5) as characteristics of equation (2.1). These characteristic curves are solutions of equations (2.5) given in the implicit form A = a( (x, y) = constant and (x, y) = constant Then, along of the characteristic curves the dierentials d = 0 and d = 0, that is dx + dy = 0. x y (2.6) dx + dy = 0, x y dy Solving the above equations for , we nd the two ordinary dierential equadx tions dy x = , dx y (2.7) dy x = , dx y dy Substituting relation (2.7) between and x , x, y , y , to equation (2.5), we dx obtain the following ordinary dierential equation dy dy a( )2 2 b + c = 0. (2.8) dx dx dy Again, solving equation (2.8) for = , that is the quadratic equation dx a2 2b + c = 0,

we obtain the following two ordinary dierential equations for characteristics curves: b(x, y) b(x, y)2 a(x, y)c(x, y) dy = 1 = dx a(x, y) (2.9) b(x, y) + b(x, y)2 a(x, y)c(x, y) dy = 2 = dx a(x, y)

14

where 1 and 2 are roots of the quadratic equation. a2 2b + c = 0. and x = 1 , y x = 2 y (2.10)

(2.11)

Example 2.2 . (a) Find the characteristics for the equation: utt utx 6uxx 5u = 2, (b) Transform equation (2.12) into its standard form. Solution To (a). The coecients of the equation are: 1 a = 1, b = , c = 6, d = 0, e = 0, g = 5, f = 2 2 (see formula ( 2.1)) The equation is hyperbolic since the discriminant = b2 ac = 6.25 > 0 is positive. The equation for characteristics is (see formula ( 2.8)): ( dx 2 dx ) + 6=0 dt dt dy , we nd the two equations dx dx b + b2 ac = 2 = = 3 dt a
1 4

(2.12)

+6 =

Solving the above equation for = dx b = 1 = dt b2 ac = 2, a

(2.13)

Hence, we nd the solutions x = 2t + constant, x = 3t + constant

In order to transform equation (2.12) to its standard form in terms of the variables , , we apply the mapping = x 2t, = x + 3t

15

Evaluate the coecients A, B, C, D, E, G, F by the formulae (see (2.2)) 2 ) + 2b + c( )2 = (2)2 + 2 6 = 0, t t x x 1 13 B=a + b[ + ]+c = 6 [2 + 3] 6 = , t t t x x t x x 2 2 2 2 C = a( ) + 2b + c( ) = (3)2 3 6 = 0, t t x x D = 0, E = 0, G = 5, F = 2 A = a( Thus, the rst standard form is 13 2u 5u = 2 tx or 2u 1 = (5u + 2) tx 13 (2.14)

Note that one can obtain from the rst standard form the second standard form by the mapping = t x, =t+x Indeed, we evaluate, u u u u u = + = + t t t Hence, we nd 2u 2 u 2u 2u 2u 2u 2u = + + + = 2 + tx 2 x x x 2 x 2 By the rst standard form (2.14), we obtain the second standard from 2u 2u 1 = (5u + 2) 2 2 13 Example 2.3 . (a) Find equations of the characteristic for the following hyperbolic equation: y 2uxx x2uyy = 0, x > 0, y > 0. (2.15)

(b) Transform equation (2.15) into the canonical form Solution To (a). The two characteristic equations are dy b b2 ac x dy b = b2 ac x = = , = = dx a y dx a y

(2.16)

16

Note that this is equivalent to setting A = C = 0. Solving the above ordinary dierential equations by the method of separating variables, we nd the equations for the characteristics in the implicit form y 2 x2 = constant, y 2 + x2 = constant.

To (b). We nd the standard form of the hyperbolic equation in the new variables = y 2 x2 , = y 2 + x2 . Then, we compute the coecients A, B, C, D, E, F, G in the equation Au + 2Bu + Cu + Du + Eu + Gu = F. (2.17)

Having the coecients a = y 2, b = 0, c = x2, d = 0, e = 0, f = 0, g = 0 in the equations au + au + au + au + eu + fu = g. we compute (see (2.2)) A = a(x )2 + 2 b x y + c (y )2 = y 2(2x)2 x2(2y)2 = 0 B = ax x + b(x y + y x ) + cy y = y 2 (2x)(2x) x2 (2y)(2y) = 8x2 y 2 C = a(x )2 + 2 b x y + c (y )2 = y 2(2x)2 x2(2y)2 = 0 D = axx + 2bxy + cyy + dx + ey = 2((x2 + y 2 ) E = axx + 2bxy + cyy + dx + ey = 2(y 2 x2 ) F = 0, G = 0.

Substituting the above coecients to the equation (2.17), we obtain the following equation (x2 + y 2)u + (y 2 x2)u u = . (2.18) 8x2 y 2 Hence, by the equations of the characteristics, we nd the rst standard form u = u u 2( 2 2 )

We can nd the second canonical form of equation (2.17), introducing the new variables = + , = . Now, we can rewrite equation (2.18), given in variables , in terms of the variables , to obtain the second canonical form.

17

Then, we compute u = u + u = u + u u = u + u = u u u = u + u + u + u = u u Hence, by substituting, we obtain the second canonical from equation (2.15). u u = Example 2.4 . Consider th equation y 2u 2u 2u + (x + y) + x 2 = 0, x2 xy y (2.19) u + u . 2

1. Find the range of x and y for which the equation is hyperbolic. 2. Transform the equation to a canonical form. 1 Solution. We have a = y, b = (x + y) c = x, d = e = g = f = 0. The dis2 criminant 1 b2 a c = (x y)2 > 0, 4 is positive for all real x = y . Then, equation (2.19) is hyperbolic on the whole x, y plane, except the line y = x, where (2.19) becomes the parabolic equation. The equation for the characteristics is: y( Then, we nd the roots x , (x + y) |x y| r1 = = = y 2y 2y 1, x 2 4xy (x + y) (x + y) (x + y) + |x y| y , r2 = = = 2y 2y 1, (x + y) (x + y)2 4xy

dy 2 dy ) (x + y) + x = 0. dx dx

(2.20)

if x y if x > y, if x y if x > y,

x and 2 = 1. y Hence, we obtain the following two equations for characteristics So that, we consider 1 = dy x = , dx y dy =1 dx (2.21)

18

Solving the above equations, we nd (x, y) = y 2 x2 = constant, Now, we consider the mapping = y 2 x2 , = y x. (2.23) (x, y) = y x = constant. (2.22)

To transform the hyperbolic equation to the standard form, we compute A = a(x)2 + 2bx y + c (y )2 = y(2x)2 + (x + y)(2x)(2y) + (2y)2 = 0 B = axx + b(xy + y x ) + cy y = = y(2x)(1) + 1 (x + y)(2x 2y) + 2xy = (x y)2 = 2, 2 C = a(x)2 + 2bx y + c(y )2 = y (x + y) + x = 0 D = axx + 2bxy + cyy + dx + ey = 2(y x) = 2, E = axx + 2bxy + cyy + dx + ey = 0, F = 0. In the new variables and , the hyperbolic equation (2.19) takes the standard form 2u u + = 0. (2.24)

2.4

Transformation of an Elliptic Equation into the Standard Form

Let us recall that the equation a(x, y) 2u 2u 2u + 2b(x, y) + c(x, y) 2 + x2 xy y u u +d(x, y) + e(x, y) + g(x, y)u = f(x, y), x py

(2.25)

is elliptic when the discriminant b2 ac < 0. Then the roots 1 and 2 of the quadratic equation (2.10) are complex and hence the solution (x, y) and (x, y) of equation (2.8) for characteristic will be also complex. Therefore, mapping (2.3) is determined by the conjugate roots of quadratic equation (2.10), that is, by i and + i. Then , the functions (x, y) and (x, y) are complex solutions of the dierential equation (2.8) for the chracteristics This characteristic equation is a( dy 2 dy ) 2 b + c = 0. dx dx

19

Soloving the quadratic equation a2 2b + c = 0

for =

dy , we get two equations dx

dy = (x, y) i(x, y), dx

dy = (x, y) + i(x, y) dx

(2.26)

Thus, if (x, y) = (x, y) + i(x, y) = constant is the complex solution of the characteristic equation then the conjugate (x, y) = (x, y) i(x, y) is also the complex solution of the characteristic equation. We consider the complex mapping (x, y) = (x, y) + i(x, y), (x, y) = i(x, y).

In order to obtain a real mappipng which leads the original equation to the standard form, we apply the mapping (x, y) = (x, y), (x, y) = (x, y)

as the real and the imaginary parts of the complex solution. By the real mapping, we transform the equation in x, y variables to the equation in the new variables , using formulae (2.2) for the evaluation of the coecients A, B, C, D, E, G, F . Then we arrive to the equation A( 2u 2 u u u + 2) + D +E + Gu = F 2 (2.27)

with A = C and B = 0. Dividing by A, we obtain the standard form 2u 2 u D u E u G F + 2+ + + u= , 2 A A A A with the Laplaces equation 2u 2u + 2 = 0, 2 as the representative. Example 2.5 . Determine type of the equation 2u 2u 2u 2 + 2 2 = 0. x2 xy y Transform the equation into the standard form. (2.29) (2.28)

20

Solution. We have data: a = 1, b = 1, c = 2, d = 0, e = 0, g = 0, f = 0. The discriminant b2 ac = 1 < 0. So that, it is the elliptic equation. The characteristics of this equation are dened by the equation ( Solving the equation for = dy 2 dy ) 2 + 2 = 0. dx dx dy , we have the two equations dx dy = 1 + i. dx

dy = 1 i, dx

Integrating the above equations, we nd the following complex mapping (x, y) = y + x + ix, (x, y) = y + x ix.

Apply the real mapping = x + y and = x. Then, by formulae (2.2), we nd A = a(x )2 + 2bx y + c(y )2 = 1 B = ax x + b(x y + y x ) + cy y = 0 C = a(x)2 + 2bx y + c(y )2 = 1 D = 0, E = 0, G = 0, F = 0. In terms of , we get 2u 2u 2u 2u 2u 2 + 2 2 = 2 + 2. x2 xy y Thus, the standard form of the equation in the new variables is: 2u 2u + 2 = 0. 2 Example 2.6 . Determine type of the equation 2u 2u + y 2 = 0, x2 y y > 0, (2.31) (2.30)

Transform the equation into the standard form. Solution. We have data: a = 1, b = 0, c = y, d = 0, e = 0, g = 0, f = 0. The discriminant b2 ac = y < 0 for y > 0. So that, it is the elliptic equation. The characteristics of this equation are dened by the equation ( dy 2 ) + y = 0, dx y > 0.

21

Solving the equation for =

dy , we have the two equations dx dy = i y, dx y > 0.

dy = i y, dx

Integrating the above separate variables equations, we nd the following complex mapping = (x, y) = x + 2i y, = (x, y) = x 2i y. y > 0 Let = x and = 2 y. Then, we nd u u 2u 2u = , = 2 x x2 2 u 1 u u 1 2u 1 u = , = 3 2 2 y y y y 2 y In terms of , we get 2u 2u 2 u 2 u 1 u +y 2 = 2 + 2 = 0, x2 y y > 0.

(2.32)

Thus, the standard form of the equation in the new variables is: 2u 2u 1 u + 2 = 0, 2 y > 0.

2.5

Transformation of a Parabolic Equation into the Standard Form

For a parabolic equation there is one repeating root of equation (2.10) equal b to . Then, we nd only one solution (x, y) = constant of the equation a dy b = dx a In this case, we consider the mapping = (x, y), = (x, y),

where (x, y) = constant is an arbitrary family of curves such that Jacobian J (, ) = x y y x = 0. Because x b = = y a

22

therefore ax + by = 0. Multiplying this equation by b and using the equality b2 = ac, we nd abx + b2y = 0 abx + ac2y = 0 a(bx + cy ) = 0 Hence, we have B = axx + b(x y + y x ) + cy y = x (ax + by ) + y (bx + cy ) = 0. Therefore, the term with 2u in the canonical form is absent. bx + cy = 0

Example 2.7 Transform the following equation: utt 2utx + uxx ut ux + u = 0, to its standard form. Solution. From the general form of a linear PDE of the second order (2.1), we nd the coecients a = 1, b = 1, c = 1, d = 1, e = 1, g = 1, f = 0. In order to determine the type of the equation, we compute the discriminant b2 a c = (1)2 1 = 0. Since the discriminant equals to zero, the equation is parabolic one. Then, there is one family of characteristics determined by the ordinary dierential equation dx b = = 1 dt a Hence, we obtain the solution t + x = constant. Now, we choose the mapping = t + x, = t. t 0, x .

Let us note that for , we are free to choose any function for which the Jacobian t x x t = 0.

23

Now, we compute the coecients A = a(t )2 + 2bt x + c(x )2 = 1 2 + 1 = 0, B = at t + b(t x + x t ) + c x x = 1 (0 + 1) + 0 = 0 C = a (t )2 + 2b t x + c(x )2 = 1 2 0 + 0 = 1, D = a tt + 2b tx + c (x )2 + d t + e x = 0 2 0 + 0 1 1 = 2, E = a tt + 2b tx + c (x )2 + d t + e x = 0 2 0 + 0 1 0 = 1, G = 1, F = 0.

Hence, by the general form (2.2), we obtain the standard form of the equation u 2u u + u = 0. or 1 1 1 u = u + u u. 2 2 2

Example 2.8 . Determine type of the equation x2 2u 2u 2u + 2xy + y 2 2 = 0, x2 xy y (2.33)

Transform the equation into the standard form. Solution. The discriminant b2 ac = x2 y 2 x2y 2 = 0, so that it is a parabolic equation for all real x and y. In order to nd the canonical form of the equation, we solve the characteristics equation dy y = . dx x It is easy to nd the solution y = kx, where k is a generic constant. Then, we consider the mapping = y kx, = (x, y),

Here, (x, y) is an arbitrary function such that Jacobian x y y x = 0. Let (x, y) = x. Then, we nd the coecients
2 2 A = x2x + 2xyx y + y 2y = (kx y)2 = 0,

B = x2x x + xy(xy + y x ) + y 2y x = 0,
2 2 C = x2x + 2xyx y + y 2y = x2 ,

24

Hence, in the new variables and equation (2.33) takes the canonical form 2u = 0. 2 Applications of the canonical form of elliptic, parabolic and hyperbolic equations 1. The three major classications as elliptic, parabolic and hyperbolic equations, in fact classify physical problems into three basic physical types: steady-state problems, diusion and wave propagation. The mathematical solutions of these three types of equations are very dierent. 2. Much of the theoretical work on the properties of solutions to hyperbolic problems assume the equation has been written in the canonical form u u = (, , u, u , u ). 3. Many computer programs have been written to nd the numerical solution of the canonical form. Having the numerical solution in the new variables, we can always come back to the original variables.

25

2.6

Exercises

Question 2.1 Classify the following equations: 1. 9uxx + 12uxy + 4uyy + ux = 0. 2. uxx 8uxy + 2uyy + xux yuy = 0, 3. 4uxx + 2uxy + uyy uy = 0. 4. Find characteristics curves of the above three equations. 5. Transform the above three equations into their canonical forms. Question 2.2 Transform the equation utx + ut + ux + u = f(t, x), into the equation u, u + u + u + u = g(, ). Question 2.3 Find characteristics of Tricomis equation y uxx + uyy = 0, in the lower half-plane y < 0. Transform Tricomis equation into the canonical form in the upper-half of the plane when y > 0. Question 2.4 Show that all linear partial dierential equations of the second order in two variables x and y of elliptic type with constant coecients can be transformed into the canonical form uxx + uyy + gu = f(x, y) Question 2.5 Show that all linear partial dierential equations of the second order in two variables t and x of hyperbolic type with constant coecients can be transformed into the canonical form utt uxx + gu = f(t, x)

26

Chapter 3

Hyperbolic Equations
3.1 The Initial Value Problem for Wave Equation
utt = k 2 uxx, u(0, x) = 0 (x), < x < , ut (0, x) = 1 (x), 0 < t < .

Find the solution u(t, x) of the initial value problem (3.1)

for give functions 0 , 1 and the coecient k. This problem describes the motion of an innite string under initial conditions. The problem was solve by French mathematician DAlembert. The solution u(t, x) is given by the formula 1 1 u(t, x) = [0(x kt) + 0 (x + kt)] + 2 2k
x+kt xkt

1 ()d.

(3.2)

3.1.1

DAlemberts Solution

We solve initial value problem (3.1) in four steps. Step1. Note that the characteristic equations for the wave equation (3.1) are dx 2 dx dx ) k 2 = 0, = k, = k. dt dt dt The solutions of the equations for characteristics are: ( x kt = constant = x kt, and x + kt = constant.

Let us write equation (3.1) in terms of the new variables = x + kt.

to obtain the rst standard form of equation (3.1). u = 0.

27

(3.3)

28

Simple application of the chain rule gives ux = u + u ut = k(u + u ) uxx = u + 2u + u utt = k 2 (u 2u + u ) By substitution to equation (3.1), we nd the rst standard form of equation (3.1). This completes step 1. Step 2. In step 2, we integrate equation (3.3), rstly with respect to the variable , to obtain the solution u (, ) = (). Secondly, we integrate the equation with respect to the variable to obtain u(, ) = () + (), () = ( d,

Thus, all solutions of equation (3.3) have the following form: u(, ) = () + (), where () and () are any dierentiable functions of the variables and . For example, one can easily check that the functions u(, ) = sin + 2 u(, ) = 2 + e u(, ) = 1 + tan ,

are solutions of the equation u, = 0. This completes step 2. Step 3. In the step 3, we transform the solution u(, ) given in terms of the variables and to the original variables t and x To nd all solutions in terms of the original variables t and x, we substitute = x kt, to u(, ) = () + (), to obtain u(t, x) = (x + kt) + (x kt). (3.4) = x + kt,

29

The general solution (3.4) (all solutions) represents the sum of any two moving waves. Each wave moves in opposite directions with velocity k. For example, the functions u(t, x) = sin(x kt), u(t, x) = (x + kt)2 , one right moving wave one left moving wave

u(t, x) = sin(x kt) + (x + kt)2 , two oppositely moving waves This ends step 3. Step 4. In the step 4, we choose from all solutions given by formula (3.4) the solution u(t, x) which satises the initial-value conditions. Thus, among all solutions of the form u(t, x) = (x + kt) + (x kt) (3.5)

with any dierentiable functions and , we choose that one which satises the initial value conditions u(0, x) = 0 (x), ut(0, x) = 1(x)

This means to nd functions and such that (x) + (x) = 0 (x), k (x) k (x) = 1(x). 1 k
x 0

(3.6)

We now integrate the second equation of (3.6) from 0 to x to obtain (x) (x) = for a constant K. Solving the two equations (x) + (x) = 0(x) (x) (x) = we nd 1 k
x 0

1() d + K.

(3.7)

1() d + K. K 2 0 x K 1() d . 2 0
x

1 1 (x) = 0 (x) + 2 2k 1 1 (x) = 0 (x) 2 2k Now, we substitute to formula (3.5),

1 () d +

(3.8)

1 1 (x + kt) = 0 (x + kt) + 2 2k 1 1 (x kt) = 0(x kt) 2 2k

K 2 0 xkt K 1 (s) ds 2 0
x+kt

1 (s) ds +

30

Hence, the solution of the initial-value problem is 1 1 u(t, x) = [0(x kt) + 0 (x + kt)] + 2 2k Examples 1. Let us consider the initial value problem utt k 2 uxx = 0, u(0, x) = sin x, t > 0, < x < ,
x+kt xkt

1 ()d.

(3.9)

ut(0, x) = 0.

For 1 (x) = 0, by DAlemberts formula the solution is 1 u(t, x) = [sin(x kt) + sin(x + kt)] 2

2. Let us consider the initial problem with the initial velocity utt k 2 uxx = 0, u(0, x) = 0, t > 0, < x < ,

ut(0, x) = sin x.

For 0 (x) = 0, by DAlemberts formula the solution u(t, x) is u(t, x) = 1 2k

x+kt xkt

sin d =

1 [cos(x + kt) cos(x kt)]. 2k

Example 3.1 . Solve the following initial value problem: utt 25uxx = 0, with the initial conditions u(0, x) = cos 3x, ut (0, x) = sin 3x. x < , t 0,

31

Solution. Apply DAlemberts formula for k = 5, 0 (x) = cos 3x and 1 (x) = sin 3x to obtain 1 1 x+kt u(t, x) = [0(x kt) + 0 (x + kt)] + 1()d. 2 2k xkt 1 1 x+5t = [sin 3(x 5t) + sin 3(x + 5t)] + cos 3)d 2 10 x5t 1 1 = [sin 3(x 5t) + sin 3(x + 5t)] + [sin 3(x + 5t) sin 3(x 5t)]. 2 30 3.1.2 The Initial Boundary Value Problem for Wave Equation utt = k 2uxx ,

Let us consider the following initial boundary value problem: 0 < x < L, 0 < t < , (3.10)

u(0, x) = 0 (x), ut (0, x) = 1 (x).

When variable x [0, L], the following three kind of boundary conditions are considered: 1. Controlled end points boundary conditions u(t, 0) = 0(t), u(t, L) = 1(t). 2. Force specied on the boundaries ux(t, 0) = 0(t), ux(t, L) = 1(t). 3. Elastic attachment ux (t, 0) 0 u(t, 0) = 0(t), ux (t, L 1 u(t, L) = 1 (t).

3.2

Solution to the Finite Vibrating String by Separation of Variables

utt = k 2 uxx, u(0, x) = 0 (x), u(t, 0) = 0, k 2 > 0, 0 < x < L, 0 < t < , (3.11)

To solve the initial boundary value problem

ut (0, x) = 1 (x), u(t, L) = 0, t 0.

32

we start by seeking standing wave solutions to the wave equation, that is, solutions of the following form: u(t, x) = X(x)T (t) Substituting this expression to the wave equation 1 utt = uxx k2 we nd 1 T (t)X(x) = X (x)T (t), k2 or 1 T (t) X (x) = = 2 T (t) k X(x)

where is a constant. Hence, we obtain the two ordinary dierential equations T (t) k 2 T (t) = 0, X (x) X(x) = 0. (3.12)

In order to solve these ordinary dierential equations, we nd roots of the quadratic functions P2 () = 2 k 2 = 0, Q() = 2 = 0.

Next, we consider the following three cases: Case 1. < 0. If < 0 then there are two complex roots 1 = ik and 2 = ik . So that, for = 2, the solutions are T (t) = A sin(kt) + B cos(kt), (3.13) X(x) = C sin(x) + D cos(x). where A, B, C and D are constants to be determined by the initial and boundary conditions. Case 2. = 0 If = 0 then there are linear solutions to the equations (3.12) T (t) = At + B, X(x) = Cx + D.

In this case the solution can be trivial (u(t, x) 0) or unbounded and feasible because of the initial value conditions. Case 3. > 0. If = 2 > 0 then the solutions of equations (3.12) take the form T (t) = Aekt + Bekt, X(x) = Cex + Dex .

So, in this case, the solution either it is trivial (u(t, x) = 0) or unbounded because of initial boundary conditions.

33

Let us consider the solution given by formula (3.13), when < 0 . Now, we apply the homogeneous boundary conditions plugging into u(t, 0) = u(t, L) = 0, t 0. Then, we obtain u(t, 0) = T (t)X(0) = D[A sin(kt) + B cos(kt)] = 0, D = 0, u(t, L) = T (t)X(L) = C sin(L)[A sin(kt) + B cos(kt)] = 0, sin(L) = 0. The constant has to satisfy the equation sin(L) = 0. So that, we nd n n = , n = 1, 2, .... L We note that for Tn (t) = An sin(kn t) + Bn cos(kn t), Xn (x) = sin(n x) un (t, x) = sin(n x)[An sin(kn t) + Bn cos(kn t))] un (t, x) is the solution of the wave equation which satises the homogeneous boundary conditions for arbitrary constants An , Bn , n = 1, 2, ...; Because the wave equation is linear one, therefore every linear combination of u1(t, x), u2(t, x), ...; is also a solution of the wave equation which satises the homogeneous boundary conditions. So that, the function

(3.14)

u(t, x) =
n=1

sin

nkt nkt nx [An sin + Bn cos ] L L L

(3.15)

satises the wave equation and homogenous boundary conditions. Substituting sum (3.15) into the initial conditions u(0, x) = 0(x), gives the two equations

ut(0, x) = 1 (x),

Bn sin
n=1

nx = 0(x), L

An
n=1

nk nx sin = 1 (x) L L

(3.16)

Using the orthogonality condition 0, m = n, mx nx sin sin dx = L L L 0 , m=n 2 we can nd the coecients L 2 nx An = 1 (x) sin dx, nk 0 L
L

2 Bn = L

L 0

nx 0 (x) sin dx. L

(3.17)

34

for n = 1, 2, ...; Finally, the solution u(t, x) of the initial boundary problem is given by formula (3.15) with the constants An and Bn , n = 1, 2, ...; determined by the formulae (3.17).

35

We present these three cases below on the following diagram :

Solution u(t, x)

< 0, = 2
c

$$$ $$ $$$ $$$ c c c

Possible values of

=0

T (t) = A sin(kt) + B cos(kt) X(x) = C sin(x) + D cos(x)

d d

> 0, = 2
c
(k2)2t

(k2 )2 t

T (t) = Ae + Be X(x) = Cex + Dex

d d

d d d

T (t) = At + B X(x) = Cx + D
c

u(t, x) = X(x)T (t)

We shall now make the following observations: 1. Let us note that the solution takes the following form:

u(t, x) =
n=1

Bn sin

nx nkt cos L L

(3.18)

if the initial velocity 1(x) = 0 with the homogeneous boundary conditions, when the initial position of the vibrating string u(0, x) = 0 (x) is present. Let the function nx 0 (x) = Bn sin . L n=1 Then, simple sine vibration of a string is given by the term Bn sin nx nkt cos . L L

Thus, adding the terms of each simple vibration, we obtain the solution of the initial boundary problem. Namely, consider the initial state of a

36

vibrating string with xed end points

m

0 (x) =
n=1

bn sin

nx , L

and with zero initial velocity 1(x) = 0. Then, the solution of such initial boundary value problem is
m

u(t, x) =
n=1

bn sin

nx nkt cos . L L

(3.19)

We can obtain formula (3.19) from the solution given by (3.18). Indeed, from the initial conditions, the coecients are bn for n = 1, 2, ..., m; and bn = 0 for n > m. So, we compute the coecients Bn = nx 2 L 0(x) sin dx L 0 L L 2 m sx nx = bs sin sin dx L s=1 L L 0 = bn .

Hence, by formula (3.18), we get the solution (3.19). For example, suppose that the initial string position is 0(x) = sin x 3x 5x + 0.5 sin + 0.25 sin . L L L

The overall response to this initial condition would then be the sum of the responses to each term, that is u(t, x) = sin kt 3x 3kt 5x 5kt x cos + 0.5 sin cos + 0.25 sin cos . L L L L L L

2. The n-th term in the solution (3.15) sin nx nkt nkt [An sin + Bn cos ] L L L

is called n-th mode of vibration or n-th harmonic. This harmonic can be rewritten in the following form Rn sin nx nk cos (t n ), L L (3.20)

37

where n is the phase angle and Rn is the amplitude. Indeed, we have [An sin nkt nkt + Bn cos ]= L L An nkt Bn nkt 2 A2 + Bn [ sin + cos ]= n 2 2 L L A2 + Bn A2 + Bn n n nkt nkn nkt nkn 2 A2 + Bn [sin sin + cos cos ]= n L L L L nk 2 A2 + Bn cos (t n ) = Rn cos n (t n ), n L
2 A2 + Bn is the amplitude and n is n

nk is frequency, Rn = where = L the phase angle. Example 1.

1. By using separation variables and Fourier cosine series, solve the following problem for nite string with xed ends for appropriate initial data 0(x) and 1 (x) utt = 4uxx , 0 x L, 0 < t < , u(t, 0) = 0,
m

u(t, L) = 0, ut(0, x) = 0.

(3.21)

u(0, x) =

1 nx sin , n L n=1 2

2. Determine the frequency , the amplitude Rn and the phase angle n . 3. Graph the solution u(t, x) for m = 1, L = 2 and t = 1, 0, 1 Solution. To (a): By the formula (3.19), we nd the solution u(t, x) = 1 nx 2nt sin cos . n L L n=1 2
m

2n To (b). From the above formula, we nd that the frequency is = , the L 1 amplitude is Rn = n and the phase angle n = 0. 2 To (c): For m = 1, L = 2, the solution is u(t, x) = 1 x sin cos t. 2 2

38

. Example 2. Solve the following initial boundary problem by the method of separation of variables: utt = 36uxx , t > 0, 0 x 1, u(0, x) = 0, u(t, 0) = 0, ut (0, x) = 4, 0 x 1, u(t, 1) = 0, t 0. (3.22)

Solution. We note that 0(x) = 0. Therefore, by formula (3.17), the coecint Bn = 0 and we compute the coecients An = 1 3n
1 0

4 sin nx dx =

4 [1 (1)n ]. 3n2 2

Hence, by formula (3.15), the solution is: u(t, x) = 4 [1 (1)n ] sin(nx) sin(6nt). 3n2 2 n=1

Example 3. What is the solution to the simple supported at the ends beam with initial conditions u(0, x) = sin x, ut (0, x) = sin x, , 0 x 1.

Solution. We note that the solution u(t, x) satises the wave equation utt = k 2 uxx, 0 x 1,

with the homogeneous boundary conditions u(t, 0) = 0, u(t, 1) = 0, t 0.

39

and the initial value conditions u(0, x) = sin x, ut (0, x) = sin x, 0 x 1.

By method of separation variables, the solution is given by the formula

u(t, x) =
n=1

sin

nkt nkt nx [An sin + Bn cos ], L L L

Hence, for 0(x) = sin x, 1 (x) = sin x, and L = 1, we compute the coecients using formulae (3.17) 2 An = nk
1 1 0

sin x sin nxdx = =

Bn = 2

sin x sin nxdx

1 , n = 1, k 0 n = 1, 1, n = 1, 0 n = 1,

Thus, the solution u(t, x) of the supported beam problem is u(t, x) = sinx[ 1 sin(kt) + cos(kt)], k

for 0 x 1 and t 0. Example 4. A gitar string of length L = 1 is pulled upward at middle so the it reaches heigt 0.5 and satises the wave equation utt = 9uxx , Assuming the initial position of the string u(0, x) =

0 x 1. 0 x 0.5,

x,

1 x, 0.5 x 1, 0 x 1.

and the initial speed of the string ut(0, x) = 1,

Find the position u(t, x) of the string at time t and point x. Solution. We note that the solution u(t, x) satises the wave equation utt = 9uxx , 0 x 1,

with the homogeneous boundary conditions u(t, 0) = 0, u(t, 1) = 0, t 0.

40

and the initial value functions u(0, x) = 0 (x) =

x,

0 x 0.5,

1 x, 0.5 x 1,

ut(0, x= 1(x) = 1.

(3.23)

By method of separation variables the solution is given by the formula u(t, x) =

n=1

sin

nx nkt nkt [An sin + Bn cos , L L L

Hence, for the given functions 0 (x), 1 (x) by (3.23) and for L = 1, we (icients using formulae (3.17) An = 2 nk
0 1 0 1/2

sin nxdx =

2 1 cos n 2 ( )= [1 (1)n ], kn n n kn2 2

1 1/2

Bn = 2

x sin nxdx + 2

n n n n n n cos + n cos 2 sin 4 sin 2 2 + 2 2 = 2 = n2 2 n2 2 n2 2 Hence, the solution u(t, x) of the gitar string problem is n 4 sin 2sin(nx) 2 cos(3nt), u(t, x) = ((1 (1)n ) sin(3nt) + kn2 2 n2 2 n=1 2 sin for 0 x 1 and t 0.

(1 x) sin x dx =

3.3
1.

Exercises

Example 3.2 Solve the following initial value problems: utt = 4uxx , t > 0, < x < ,

u(0, x) = cos 3x, ut (0, x) = x. 2. utt = uxx, t > 0, < x < ,

u(0, x) = sin 3x, ut (0, x) = cos 3x. 3. utt = uxx , u(0, x) =

t > 0, < x < , sin 2x, x , 0, |x| > , ut (0, x) = 0,

41

Example 3.3 For the following equation: yutt 16xuxx = 0, (a) Determine type of the equation (3.24) (b) Find characteristic curves of the equation (3.24). (c) Transform equation (3.24) into canonical form. Example 3.4 For the following equation: uxx + 3uxy + 2uyy = 0, (a) Determine type of the equation (3.25) (b) Find characteristic curves of the equation (3.25). (c) Transform equation (3.25) into its canonical form. Example 3.5 For the following equation: utt 4utx + 4uxx ut = 0, t geq0, < x < . (3.26) < x, y. (3.25) t > 0, x > 0, (3.24)

(a) Determine the type of the equation (3.26) (b) Find the characteristic curves of the equation (3.26). (c) Transform the equation (3.26) into its canonical form. Example 3.6 Solve the initial value problem by the DAlembert method. utt 9uxx = 0, t 0, <x< (3.27)

u(0, x) = cos 4x, ut(0, x) = sin 4x, < x < . Example 3.7 Solve the initial boundary value problem by the method of separation of variables. utt 9uxx = 0, u(t, 0) = 0, t 0, u(t, 2) = 0, 0x4 (3.28) t 0.

u(0, x) = x(4 x), ut (0, x) = 1, 0 x 2,

42

Example 3.8 Consider the telegraphic equation utt + ut + u = c2 uxx , t 0, 0 x L.

Find the solution u(t, x) of the telegraphic equation which satises the initial condition u(0, x) = x(L x), u(t, 0) = 0, ut(0, x) = 0, 0 x L, t 0.

and the homogeneous boundary value conditions u(t, L) = 0,

Hint: Apply the method of separation of the variables t and x.

Chapter 4

Parabolic Equations
4.1 Initial Boundary Value Problem

We shall consider the heat equation ut = k 2 uxx + f(t, x), with the initial condition u(0, x) = 0(x), and with the boundary conditions u(t, 0) = 0(t), u(t, L) = L (t), t 0. (4.3) 0 x L, (4.2) t 0, 0 x L, (4.1)

Here, k 2 is a constant and the given functions f(t, x), 0(x), 0(t), L(t) are continuous for 0 x L, t 0. Let us establish some of the properties of the solution u(t, x). Firstly, we shall state the weak maximum principle for the heat equation ut = k 2 uxx (4.4)

in the closed rectangle R = {(t, x) : 0 x L, 0 t T }, with the boundary

R =

(t, x) : 0 x L, when t = 0, (t, x) : 0 t < T

43

when x = 0 or x = L

44

Let us note that the points on the interval (T, x), when 0 x L, are not included in the boundary R. t T T u = 0 u(t, x) in R u = L

L u(0, x) = 0 (x)

The initial boundary conditions at R

The maximum principle 4.1 Let u(t, x) be a solution of the heat equation (4.4) in the rectangle R. Then u(t, x) assumes its maximum value on whole closed rectangle R at a point on the boundary R. Also, u(t, x) attains its minimum at a point on the boundary R of R. Proof. We know that u(t, x) attains its maximum M at the closed rectangle R. Also, we know that u(t, x) attains its maximum MR at the closed boundary R. To proof the thesis of the maximum principle, we shall show that M = MR . That is, the maximum on the boundary does not exceeds the maximum on the closed rectangle, so that MR M. Suppose that M MR = , then we choose a point (t0 , x0) R interior to R, such that u(t0, x0) = M. Since > 0 and (t0, x0) is not on the boundary R, therefore 0 < x0 < L and 0 < t0 < T . Dene the auxiliary function (x x0)2 . 4L2 Then, consider w(t, x) at points on R. We note that w(t, x) = u(t, x) + w(0, x) = u(0, x) + M + 4L2 (x x0 )2 MR + 4L2 (x x0 )2

3 L2 = M < M. 4L2 4 In similar way, we arrive at the inequalities w(t, 0) < M, and w(t, L) < M, 0 t T.

45

Indeed, we have w(t, 0) = u(t, 0) + M + and w(t, L) = u(t, L) + M + 4L2 (L x0)2 MR + 4L2 (L x0 )2 4L2 (0 x0)2 MR + 4L2 (0 x0)2

3 L2 = M < M. 4L2 4

3 L2 = M < M. 4L2 4

But w(t0 , x0) = u(t0, x0 ) = M. Therefore the maximum of w(t, x) on R is at least M and it is attained at a point (t1 , x1) R, not on the boundary R. Because 0 < x1 < L, 0 < t1 < T, then wt (t1, x1) = 0, Hence wt (t1, x1 ) k 2 wxx (t1, x1) 0. But k2 k2 = < 0. 2L2 2L (4.7) Thus, we have arrived at the contradiction, the inequality (4.6) against the inequality (4.7). Therefore, u(t, x) attains its maximum value on the boundary R of the rectangle R. Similarly, we can prove for minimum of u(t, x), taking u(t, x) instead of u(t, x). Then, we conclude that M = MR . End of the proof. Conclusion. From the weak maximum principle, it follows that every solution u(t, x) of the initial boundary value problem ( 4.1), (4.2),(4.3) satises the inequality wt(t1 , x1) k 2wxx (t1, x1 ) = ut (t1, x1) k 2 uxx (t1, x1) k 2 |u(t, x)| max |u(t, x)|,
(t,x)R

wxx(t1, x1 ) 0.

(4.5) (4.6)

t 0,

0 x L.

(4.8)

As a consequence of the maximum principle, we can state the following theorems Theorem 4.1 (Uniqueness) The initial boundary value problem (4.1),(4.2) and (4.3) has at most one continuous solution. Proof. Assume that there are two solutions u1 (t, x) and u2(t, x). Then, it is easy to show that the dierence w(t, x) = u1(t, x) u2 (t, x) is the solution of

46

the homogeneous heat equation with the homogeneous initial boundary conditions. By the maximum principle, we conclude that such a homogeneous initial boundary value problem has only trivial solution, that is w(t, x) 0 for all (t, x) R. Indeed, because w(t, x) attains its total maximum on R at the boundary R, not greater than zero, therefore w(t, x) 0 for all (t, x) R. But, also w(t, x) attains its total minim on R at the boundary R. So that w(t, x) 0 for (t, x) R. Hence w(t, x) 0 for all (t, x) R. End of the proof. Theorem 4.2 Let u(1)(t, x) and u(2)(t, x) be two solutions of the two initial boundary value problems ut = k 2 u(1) + f(t, x), 0 x L, xx u(1)(0, x) = 0 (x), u(1)(t, 0) = 0 (t), and
(2) (1) (1) (1)

t 0, (4.9)
(1)

0 x L, u(1)(t, L) = L (t), t 0, t 0,

ut = k 2 u(2) + f(t, x), 0 x L, xx u(2)(0, x) = 0 (x), u(2)(t, 0) = 0 (t),

(2) (2)

0 x L, u(2)(t, L) = L (t), t 0,
(1)

(4.10)

for the same sours of energy f(t, x). Suppose that the distance of the initial boundary conditions is less than > 0, so that |0 (x) 0 (x)| < , |0 (t) 0 (t)| < , |L (t) L (t)| < , (4.11) Then, the inequality |u(1)(t, x) u(2)(t, x)| < , holds for all t 0, and 0 x L. Proof. Let us note that the dierence v(t, x) = u(1)(t, x) u(2)(t, x) is the solution of the heat equation vt = k 2 vxx, which satises the initial condition v(0, x) = 0 (x) 0 (x),
(1) (2) (1) (2) (1) (2) (1) (2)

t 0,

0 x L,

0 x L,

47

and the boundary conditions v(t, 0) = 0 (t) 0 (t),

(1) (2)

v(t, L) = L (t) L (t),

(1)

(2)

t 0.

By the assumption, |v(0, x)| < , |v(t, 0)| < , |v(t, L)| < . Then, by the maximum principle |v(t, x)| < , for all t 0 and 0 leqx L. Question 1. Consider the following initial value problem: ut = k 2uxx , < x < , t 0,

u(0, x) = 0(x), < x < , Assume that the given function 0 (x) is continuous and bounded for all x (, ) and u(t, x) 0 when x . Using the maximum principle show that |u(t, x)| max |0 (x)|.
<x<

for all t 0 and < x < . Solution. Let x [a, a] for a positive a > 0. Then, by the weak maximum principle (see (4.8)) |u(t, x)| max |u(t, x)|,
(t,x)R

t 0,

a x a.

where R = {(0, x), (t, a)), (t, a)}. By the assumption u(t, x) 0, when x , so that, for suciently large a, we have |u(t, x)| max |0(x)|,
x[a,a]

for all t 0 and |x| a. Therefore, the inequality |u(t, x)| for all t 0 and < x < . Example 4.1 Let us apply the maximum principle to give an estimate of the solution u(t, x) of the initial boundary value problem ut = 36uxx , 0 x 1, t 0, 2t u(t, 1) = 2 + t2
x[,]

max |0 (x)|,

2x u(0, x) = , 1 + x2

(4.12)

u(t, 0) = 0,

48

By the maximum principle, we observe that the maximum of all values of the solution u(t, x) in the domain R = {(t, x) : 0 x 1, t 0} is on the boundary R =

(t, x) : 0 x 1, when t = 0, (t, x) : 0 t < T when x = 0 or x = 1.

Let us nd the maximum values on the boundary R

(t,x)R

max u(t, x) = max{ max u(0, x), max u(t, 0), max u(t, 1)}
0x1 t0 t0

2x 2t , max 0, max } 2 t0 0x1 1 + x t0 1 + t2 = max{1, 0, 1} = 1 = max{ max Hence, the maximum of all values of the solution u(t, x) in the domain R do not exceed 1, that is
(t,x)R

max u(t, x) max u(t, x) = 1

(t,x)R

4.2

Solution by Separation of Variables

ut = k 2 uxx , t 0, 0 x L, (4.13)

. Let us consider the following initial boundary value problem

u(0, x) = 0 (x), 0 x L, u(t, 0) = 0, u(t, L) = 0, t 0. X T = 2 = , X k T where is the separation constant. Hence, we get the equations X + X = 0, T + k 2 T = 0. Substituting u(t, x) = T (t)X(x), to the heat equation, we obtain

(4.14)

Solving the boundary value problem for the ordinary dierential equation X + X = 0, we arrive at the solution Xn (x) = sin nx , L n = n2 2 , L2 n = 1, 2, ... (4.16) X(0) = X(L) = 0, (4.15)

49

Now, the equation for T (t) becomes T + with the solution k 2n2 2t L2 Tn (t) = e , n = 1, 2, ... (4.17) k 2 n2 2 T = 0, L2 n = 1, 2, ...

Hence, we nd the solution of the heat equation k 2n2 2t nx L2 sin , n = 1, 2, ... un (t, x) = Tn (t)Xn (t) = e L which satises the homogeneous boundary conditions. The function k 2 2n2 t nx L2 sin Bn e u(t, x) = , L n=1

(4.18)

(4.19)

is also the solution of the heat equation and satises the homogeneous boundary conditions for any choice of the coecients Bn , n = 1, 2, ... In order to determine the coecients Bn , n = 1, 2, ..., we expand in the Fourier series of sines the function 0 (x) given in the initial condition. Then, we have

u(0, x) = 0(x) =
n=1

Bn sin

nx . L nx dx, L

Having Fourier series of the initial function

0(x) =
n=1

bn sin

nx , L

bn =

2 L

L 0

0(x) sin

we arrive at the solution of the initial boundary value problem

L

u(t, x) = where Bn = bn and

n=1 0

n2 2k 2 t nx L2 Bn e sin dx L

(4.20)

2 L nx 0(x) sin dx, n = 1, 2, ...; L 0 L Example 4.2 Solve the following initial boundary value problem: Bn = ut = k 2 uxx, u(t, 0) = 0, t 0, 0 x 1, u(t, 1) = 0, t 0, 1 sin 3x, 2 (4.21)

u(0, x) = 0 (x) = sin x +

50

Solution. We note that the coecients of the Fourier series of the initial value function 1 0 (x) = sin x + sin 3x, 2 1 are B1 = 1, B2 = 0, B3 = , B4 = B5 = ... = 0. 2 Therefore, the solution is
2 2 2 2 1 u(t, x) = e k t sin x + e9 k t sin 3x. 2

4.3

Transformation of Non-homogeneous Boundary Conditions to Homogeneous

Let us consider the heat equation with constant temperature at the end of a rod. So, we consider the following initial boundary value problem: ut = k 2uxx , u(t, 0) = 0, u(0, x) = 0(x), where this time 0 , L are constants. In order to transform the non-homogeneous boundary conditions to homogeneous ones, we introduce the new unknown function v(t, x) by the formula u(t, x) = v(t, x) + 0 + x (L 0). L (4.23) t 0, 0 x L, (4.22)

u(t, L) = L , t 0,

Clearly, the unknown v(t, x) satises the homogeneous boundary conditions v(t, 0) = v(t, L) = 0. So that v(t, x) is the solution of initial boundary problem vt = k 2 vxx, v(t, 0) = 0, t 0, 0 x L, x (L 0)], L t 0, (4.24) 0 x L.

u(t, L) = 0,

v(0, x) = 0(x) [0 +

By the formula (4.20), we nd the solution

v(t, x) = where

n=1 0

n2 2k 2 t nx L2 Bn e sin dx, L

(4.25)

0 (x) = 0 (x) [0 +

x (L 0 )], L

Bn =

2 L

L 0

0 (x) sin

nx dx. L

51

Finally, in terms of original unknown u(t, x) = v(t, x) + [0 + x (L 0)]. L 0 x 2, t 0, (4.26)

Example 4.3 Solve the following initial boundary value problem: ut = 4uxx , u(t, 0) = 1, t 0, (4.27) (4.28)

u(t, 2) = 4,

u(0, x) = 1,

0 x 2.

Solution. By introducing the new unknown function v(t, x) by the formula

x u(t, x) = v(t, x) + 0 + L (L 0) = v(t, x) + 1 + x (4 1) 2

= v(t, x) + 1 + we note that vt(t, x) = ut(t, x), vt = k 2 vxx, and vxx = uxx (t, x),

3x , 2

(4.29)

Therefore v(t, x) is the solution of the heat equation. So that t 0, 0 x L.

u(t, 0) = v(t, 0) + 1 = 1, u(t, 2) = v(t, 2) + 1 + 2 (4 1) = v(t, 2) + 4 = 4. 2 v(t, 0) = 0, v(t, 2) = 0, t 0.

Hence, we obtain

Then, the new unknown function v(t, x) satises the initial condition x v(0, x) = 0 (x) = 0 (x) [0(0) + (L(t) 0(t)) 2 x 3x = 1 [1 + (4 1)] = . 2 2 Hence, by the formula (4.20), we obtain the solution n2 2k 2 t 2 n nx L2 v(t, x) = [ 0 () sin d]e sin L n=1 0 L L 2 2 4n t 2 3 n nx 4 = [ sin d]e sin 2 2 2 n=1 0 nx 3 n2 2t = e sin 2 n=1 n
L

(4.30)

52

Now, coming back to the original unknown function u(t, x), by the formula (4.29), we nd the solution u(t, x) = 1 +
3x 3 n2 2t nx + e sin , 2 2 n=1 n

t 0,

0 x 2.

Example 4.4 . Solve the initial boundary value problem for the diusion equation: ut = 4uxx , t 0, 0 x 2, u(t, 0) = 0, u(t, 2) = 1,

t 0,

(4.31)

u(0, x) = 0 (x) =

x, 0 x 1, 1, 1 x 2

Write rst three terms of the solution u(t, ).

Solution. First, we transform the equation with non-homogeneous boundary condition to the equation with homogeneous boundary conditions. For this purpose we introduce new unknown by formula (4.23) for L = 2, 0 = 0, L = 1 x x v(t, x) = u(t, x) + 0 (L 0) = u(t, x) . L 2 Clearly, the new function v(t, x) satises the homogeneous boundary conditions, i.e. v(t, 0) = 0 and v(t, 2) = 0. Now, we solve the initial boundary problem vt = 4vxx , v(t, 0) = 0, t 0, 0 x 2,

v(t, 2) = 0, t 0,

v(0, x) = u(0, x) 0 (x) = 0 (x) =

x 1 , 2

x , 2

0 x 1, 1x2

(4.32)

We compute the coecients Bn , n = 1, 2, ... of the Fourier series of the initial value function 0 (x) for k = 2, L = 2 Bn = 2 L
L 0

0 (x) sin

2 nx 1 1 nx x nx dx = x sin dx + (1 ) sin dx L 2 0 2 2 2 1 1 n 2 n = cos + 2 2 sin n 2 n 2 2 n 1 n 2 + 2 2 sin cos 2 2 sin n n 2 n 2 n 4 n = sin n2 2 2

53

Hence, by the formula (4.20), we get the solution

v(t, x) =
n=1

4 n2 2

sin

n n2 2t nx e sin 2 2

(4.33)

Coming back to the original unknown, we nd the solution u(t, x) =

x x 4 n n2 2t nx + v(t, x) = + sin e sin 2 2 n=1 n2 2 2 2

4.4

More Developed Heat Equation

ut = k 2 uxx u, 0 x L, u(0, x) = 0 (x), u(t, 0) = 0, 0 x L, u(t, L) = 0, t 0 t 0, (4.34)

Let us consider the following initial boundary value problem:

Here, the term u, > 0, represents heat ow across the lateral boundary. Let us note that by the substitution u(t, x) = etv(t, x), ut = etv(t, x) + etvt (t, x), uxx = etvxx (t, x) we can transform the initial boundary problem (4.34) into simple one for the new unknown v(t, x). vt = k 2 vxx , 0 x L, t 0, (4.35) v(t, L) = 0, t 0

v(0, x) = 0 (x), 0 x L, v(t, 0) = 0,

Solving the initial boundary value problem (4.35), by the formula (4.20), we obtain the solution n2 2 k 2t nx L2 u(t, x) = etv(t, x) = et Bn e sin (4.36) L n=1 2 L nx int0 0 (x) sin dx, n = 1, 2, ...; L L Example 4.5 . Let us consider the initial boundary value problem: Bn = ut = uxx u, u(0, x) = sin x + u(t, 0) = 0, 0 x L, t 0, 1 sin 3x, 0 x L, 2 v(t, L) = 0, t 0, (4.37) where

54

Solution. We apply the substitution u(t, x) = et v(t, x), to eliminate the term u. Then, v(t, x) satises the heat equation with the initial boundary conditions vt = uxx, 0 x L, t 0, (4.38)

v(0, x) = sin x + v(t, 0) = 0,

1 sin 3x, 0 x L, 2 0, u(t, L) = 0, t 0, 1 92 t e sin 3x, 2 1 92 t e sin 3x], 2

The solution of the problem (4.38) is v(t, x) = e t sin x +

2

Hence, coming back to the original unknown, we nd the solution u(t, x) = et [e t sin x +
2

4.5

Non-homogeneous Heat Equation

Let us consider the non homogenous heat equation with the initial boundary value conditions ut = k 2uxx + f(t, x), t 0, u(0, x) = (x), u(t, 0) = 0, 0 x L, u(t, L) = 0, t 0. 0 x L, (4.39)

In the previous section (see (4.19), we have found the solution k 2 2n2 t nx L2 sin u(t, x) = Bn e . L n=1

(4.40)

where Bn =

2 L nx (x) sin dx, and f(t, x) 0, 0 (t) 0, l (t) 0. L 0 L Now, we shall nd the solution u(t, x) of the non homogeneous heat equation, when f(t, x) = 0. Assume that the given function as the heat sourse f(t, x) possesses the following series presentation: f(t, x) = f1 (t) sin x 2x nx + f2 (t) sin + ... + fn (t) sin + ... L L L (4.41)

55

In order to nd the coecients fn (t), n = 1, 2, ...; we multiply both sides of mx (4.41) by sin , and integrate from zero to L with respect to the variable L x. Then, we obtain nx mx L sin dx = fm (t). (4.42) L L 2 0 0 nx Hence, by the orthogonality of the sequence {sin , n = 1, , 2, ...; } we nd L the coecients nx 2 L f(t, x) sin dx, n = 1, 2, ...; (4.43) fn (t) = L 0 L Replacing the heat sourse function f(t, x) by its decomposition (4.41). we nd the solution nx u(t, x) = Tn (t) sin (4.44) L n=1 of the initial boundary value problem
L L

f(t, x) sin

nx dx = fn (t) L n=1

sin

ut = k 2 uxx +
n=1

fn (t) sin

u(0, x) = (x), u(t, 0) = 0,

nx , L 0 x L,

t 0,

0 x L, (4.45)

u(t, L) = 0,

where the functions Tn (t), n = 1, 2, ...; are to be determined. By substitution (4.44) to the equations (4.45), we have the following equations:

t 0.

Tn (t) sin
n=1 n=1

nx n2 2 nx nx = k 2 Tn (t)sin + fn (t) sin 2 L L L n=1 L n=1

Tn (t) sin 0 = 0,
n=1

Tn (t) sin n = 0, 0 x L.

t 0,

(4.46)

Tn (0) sin
n=1

nx = (x), L

Hence

k 2n2 2 nx Tn (t) fn (t)] sin = 0, L2 L n=1 nx Tn (0) sin = (x). L n=1 [T (t) +

(4.47)

Then,we nd the coecients Tn (t), n = 1, 2, ...; solving the ordinary dierential equations Tn (t) + k 2 2n2 Tn (t) = fn (t), t 0, L2 2 L nx Tn (0) = (x) sin = Bn . L 0 L n = 1, 2, ..., (4.48)

56

By integrating factor method, we nd the solution k nt + L2 Tn (t) = Bn e

2 2 2 t 0

k 2 2n2 (t ) L2 e

fn ( ) d.

(4.49)

Finally, the solution of the initial boundary value problem (4.39) is given by the following formula: k 2 2n2 t nx nx L2 sin Bn e + sin u(t, x) = L L n=1 n=1

t 0

k 2 2n2 (t ) L2 e fn ( ) d. (4.50)

Example 4.6 Solve the following initial boundary value problem: ut = uxx + sin x + sin 2x, 0 x 1, t 0, u(0, x) = sin x, u(t, 0) = 0, u(t, 1) = 0, 0 x 1, t 0. (4.51)

Solution. We shall nd the solution u(t, x) in the series form

u(t, x) =
n=1

Tn (t) sin nx,

0 x 1,

t 0,

where the coecients Tn (t), n = 1, 2, ...; are determined by the ordinary differential equation T (t) + n2 2Tn (t) =
1 1 0

1 [sin s + sin 2s] sin nx ds = fn (t), 2

0 x 1, (4.52)

t 0,

Tn (0) = 2

sin s sin ns ds = Bn .

We compute
1

fn (t) = 2 Bn = 2
1 0

[sin s + sin 2s] sin nx ds =

1, n = 1, 2 0, otherewise,

1, n = 1,

sin s sin ns ds

0, otherewise

57

Solving the ordinary dierential equation (4.52) for n = 1, 2, ...; we nd n = 1, T1 (t) + T1(t) = 1, n = 2, T2 (t) + 4 T2 (t) = 1,
2 2

2 + e t 1 T1(0) = 1, T1 (t) = 2 2 e
2 1 e4 t T2(0) = 0, T2 (t) = , 4 2

(4.53)

n 3, Tn (t) + n2 2Tn (t) = 0, T1(0) = 0, Tn (t) = 0. Finally, we obtain the solution

u(t, x) =
n=1

Tn (t) sin nx = T1 (t) sin x + T2(t) sin 2x

2 2 2 + e t 1 1 e4 t = sin x + sin 2x. 2 4 2 2 e

4.6

Fundamental Solution for the Heat Equation

ut = uxx, (4.54) (x )2 e 4(t ) ,

Let us consider the heat equation

The fundamental solution of the heat equation is given by the formula 1 U(x , t ) = t (4.55)

Let us note that the fundamental solution is the function of two points P = (t, x) and Q = (, ) given for < t. Also, we note that U(t, x; , ) satises the heat equation as the function of the variables t and x at xed and . Indeed, we nd ut = [ and uxx (x )2 4(t )5/2
2

(x )2 1 ]e 4((t ) 2(t )3/2

So that ut = uxx. Also, one can check that the fundamental solution U(t, x, , ) satises the conjugate heat equation u + u = 0, as the function of the variables and , at xed t and x.

(x )2 (x ) 1 =[ ]e 4((t ) 5 2(t )3/2 2 4(t )

58

4.7

Fundamental Formulae

Greens Formula. Below, we present Greens formula in its simplest form for two continuosly dierentiable functions P (x1 , x2) and Q(x1, x2)

Q P ]dx1 dx2 = x1 x2

P dx1 + Q dx2 ,

(4.56)

where the rectangle

R = {(x1, x2) : a x1 b, c x2 d}. with the boundary R. Proof. We shall show that

P dx1 dx2 = P dx1 x2 R Q dx1 dx2 = Q dx2 x1 R

Indeed, we compute (see picture)

P dx1 dx2 = x2

b a

dx1
b a

d c

P dx2 = x2
a

b a b

[P (x1, d) P (x1 , c)]dx1

R

=
R

P (x1, d)dx1
b a

P (x1 , c)dx1 =
d

P dx1

Q dx1 dx2 = x1 =

d c d c

dx2

Q dx1 = x1
d c

[Q(b, x2) Q(c, x2)]dx2

R

Q(b, x2)dx2

Q(c, x2)dx2 =

Q dx2

Hence, we obtain Greens formula (4.56)

59

t d

T '

R
T

R
E

c 0 a

The rectangle R with the boundary R

For Q = u v and P = 0, from, Greens formula, by the identities uv dx1 dx2 = x1 uv dx1 dx2 = x2 v u +v dx1 dx2 = x1 x1 v u u +v dx1 dx2 = x2 x2 u uv ds (4.57) uv ds

we obtain the formula of integration by parts in two variables

R

v = x1 v u = x2

uv ds uv ds

u dx1 dx2 x1 u v dx1 dx2 x2 v

(4.58)

Here the line integral along the boundary R of the rectangle R is

R

u v ds =

u(x1(), x2 ())v(x1(), x2 ()) (x1 ())2 + (x2())2 d

d b a d

which becomes
R

u v ds = +

b a c b

u(x1(), c)v(x1(), c)d + u(x1(), d)v(x1 (), d)d +

u(b, x2())v(b, x2())d u(a, x2())v(a, x2())d

The rst fundamental formula for the heat equation Let us note that the following identity holds: vF (u) uG(v) = u v (uv) [v u ] , x x x t

60

2u u 2v v F (u) = , G(v) = + x2 t x2 t Integrating by parts both sides of the above identity (see formula integration by parts (4.57)), in the rectangle R = {(t, x) : 0 t T, 0 x L} we obtain the rst fundamental formula for the heat equation
R

for

[vF (u) uG(v)]dt dx =

u v dx + (v

u v u ) dt x x

(4.59)

Hence, if u and v satisfy the equations u 2 u 2 = 0, t x v 2 v + = 0, t x2 u v u ) dt = 0. x x

then the rst fundamental formula for the heat equation becomes
R

u v dx + (v

(4.60)

The second fundamental formula for the heat equation Substituting into the rst fundamental formula, = t , v(, ) = U(t, x; , ) for < , we obtain (x )2 d 4 = u(, t )e [u (, )U(t, x; , ) u(, )U t, x; , )]d R R + u(, )U(t, x; , ) d. (4.61) Hence, in the limit when > 0, we obtain the second fundamental formula
R

[u (, )U(t, x; , ) u(, )U (t, x; , )]d+ 2 u(t, x), (t, x) R, +u(, )U(t, x; , )d = 0, (t, x) out of closed R

(4.62)

4.8

Exercises

Question 1. Solve the initial boundary problem: ut = 9uxx , 0 x 4, t 0, (4.63) v(t, 4) = 2, t 0

u(0, x) = sin x + 2 sin 5x, 0 x 4, u(t, 0) = 1,

61

Question 2. Solve the initial boundary problem: ut = 16uxx 3u, u(0, x) =

0 x 2,

t 0, (4.64)

x, 0 x 1, 1, 1 < x 2, u(t, 2) = 0, t 0

u(t, 0) = 1,

62

Chapter 5

Elliptic Equations
5.1 Laplace Equation

. Laplaces equation takes the following form: 1. In two variables x, y u = 2u 2u + = 0, x2 y 2 u = uxx + uyy = 0, 2 2 + 2. x2 y (5.1)

where the Laplaces operator

2. Laplaces equation in the polar coordinates (r, ), r = 0, x = r cos , y = r sin , 1 1 u = urr + ur + 2 u = 0. r r y where r2 = x2 + y 2 and = arctan , x = 0, = , x = 0, x 2 yT r
s

(5.2)

(r, )

6 E

0 x Polar Coordinates x = r cos , y = r sin Indeed, we can transform Laplaces equation from Cartesian coordinates
63

64

to Laplaces equation in polar coordinates by the following computations: x = r cos y = r sin , u(x, y) = u(r cos , r sin ) u u u = cos + sin r x y 2u 2u 2u 2u = cos2 + 2 sin2 + 2 sin cos r2 x2 y xy u u u = r( cos sin ) y x u 2u 2u 2u = r sin + r2 2 cos2 r2 sin cos 2 y y xy u 2u 2u r cos + r2 2 sin2 r2 sin cos x x xy Now, we nd 1 1 2u 2u 2u urr + ur + 2 u = ( 2 cos2 + 2 sin2 + 2 sin cos ) r r x y xy 1 u u ( cos + sin ) + r x y 1 u 2u 2u + 2 (r sin + r2 2 cos2 r2 sin cos ) r y y xy 1 u 2u 2u + 2 (r cos + r2 2 sin2 r2 sin cos ) r x x xy 2u 2u = + x2 y 2

3. Laplaces equation in three variables x, y, z u = 2u 2u 2u + + = 0, x2 y 2 z 2 u = uxx + uyy + uzz = 0. (5.3)

4. Laplaces equation in the three cylindrical coordinates r, , z, with r = 0, x = r cos , y = sin , z = z 1 1 u = urr + ur + 2 u + uzz = 0. r r . (5.4)

65

z T

(r, , z)

d d

d d

y
d

Cylindrical coordinates x = r cos , y = r sin , z = z

y where r2 = x2 + y 2, r = 0, = arctan , x = 0, z = z. x 5. Laplaces equation in the spherical coordinates r, , , with x = r sin cos , y = r sin sin , z = r cos . 1 cot 1 2 u = urr + ur + 2 u + 2 u + 2 2 u = 0. r r r r sin . (5.5)

z T r
d d

(r, , )

y
d d d

Spherical coordinates x = r sin cos , y = r sin sin , z = r cos .

z y where r2 = x2 + y 2 + z 2 , cos = , tan = r x

5.2

Boundary Value Problems for Laplace Equation

The following three types of the boundary value problems are considered: 1. Dirichlet boundary value problem Find the solution u(x, y) of Laplace equation uxx + uyy = 0, (x, y) (5.6)

66

in the domain , which satises the Dirichlets condition (BDC) u(x, y) = (x, y), (x, y) . (5.7)

Throughout this chapter, we shall denote by a bounded domain with the boundary . Here (x, y) is a given function on the boundary of . 2. Neumann boundary problem Find the solution u(x, y) of Laplace equation uxx + uyy = 0, (x, y) in the domain , which satises the Neumanns boundary condition (BNC) u(x, y) = (x, y), (x, y) , n du where denotes normal derivative internal to the boundary of the dn domain . Here (x, y) is a given function on the boundary . 3. Third kind boundary problem Find the solution u(x, y) of Laplace equation uxx + uyy = 0, (x, y) in the domain , which satises the third kind boundary condition A(x, y) u(x, y) + B(x, t)u = (x, y), n (x, y) ,

where A2(x, y) + B 2(x, y) > 0 and (x, y) are given functions on the boundary .

5.3

The Maximum Principle for Laplace Equation

Every solution of Laplaces equation is called harmonic function. Below, we shall give some of properties of the harmonic functions. 1. The maximum principle. Let u(x, y) be a continuous function in the bounded and closed domain . If u(x, y) is a harmonic function in the domain , then the function u(x, y) attains its maximum and minimum values on the boundary of . Proof. Firstly, we note that u(x, y) attains its maximum in the bounded and closed domain as a continuous function. By contradiction to the thesis, assume that the maximum value of u(x, y) is not on the boundary . Then, the maximum value of u(x, y) is at some interior point

67

(x0, y0) , say M = u(x0 , y0) > Mb , where Mb is maximum of u(x, y) on the boundary . Let us introduce the auxiliary function v(x, y) = u(x, y) + [(x x0 )2 + (y y0)2 ], for some > 0. Then v(x0, y0) = u(x0, y0) = M, and the maximum of v(x, y) on the boundary of is equal at most Mb + d2 , where d is the diameter of . For suciently small > 0, we have M > Mb + d, i.e. 0 < < (M Mb )/d2 . For such , the maximum of v(x, y) cannot occur on the boundary of , since the value M of v(x, y) at (x0, y0) is larger than the value of v(x, y) at any boundary point. There may, however, be points in , where v(x, y) is larger than M. Let the maximum of v(x, y) be attained at (x1 , y1), which, as we have seen, must be in . At (x1, y1), we must have vxx 0 and vyy < 0, since the graph of v(x, y) cannot be concave up in the x or y direction at (x1 , y1). Thus, at (x1 , y1), we have vxx + vyy 0. However, by the denition of v(x, y), we have vxx + vyy + 2 + 2 = 4 > 0. Here, we have used the assumption that u(x, y) is harmonic on . The above two inequalities contradict one the other. So that, the assumption that u(x, y) attains its maximum of u(x, y) at an interior point, and not on the boundary, leads to the contradiction. In order to prove that minimum value of u(x, y) is attanable at the boundary , we repeat the proof for the maximum of u(x, y). So that, the minimum of u(x, y) must be also attainable on the boundary of . There is also strong maximum principle for harmonic function which we present below. The Strong maximum principle. Let u(x, y) be a harmonic function on the domain . Suppose that the function u(x, y) attains its maximum or minimum at some interior point of . Then u(x, y) must be constant throughout . . Conclusion From the maximum principle it follows that every harmonic function u(x, y) satises the inequality
(x,y)

min (x, y) u(x, y) max (x, y),

(x,y)

Also, we have |u(x, y)| max |(x, y)|,

(x,y)

(x, y) , (x, y) ,

(5.8)

or max |(x, y)| u(x, y) max |(x, y)|,

(x,y) (x,y)

68

where u(x, y) = (x, y) on the boundary of . . Question 1. Consider the following boundary value problem: uxx + uyy = 0, (x, y) = {(x, y) : 0 < x < 2, 0 < y < 1},

(5.9) Find the range of the values of the solution u(x, y), for (x, y) . Solution. By the maximum principle, the range of the values of the solution u(x, u) is determined by the inequality
(x,y)

u(x, y) = sin x + cos y, (x, y) ,

min (x, y) u(x, y) max (x, y).

(x,y)

where (x, y) = sin x + cos y. Now, we compute

(x,y)

min [sin x + cos y],

and

(x,y)

max [sin x + cos y]

Clearly, the minimum and maximum of (x, y) = sin x + cos y are at points when the partial derivatives are equal to zero, so that x (x, y) = cos x = 0, and y (x, y) = sin y = 0.

1 3 We nd two boundary points ( , 0) and ( , 1) Thus, the maximum is at 2 2 1 the point ( , 0) equal 2 max[sin x + cos y] = 2,

3 and the minimum is at the point ( , 1) equal 2 min[sin x + cos y] = 2.

Hence, the values of the solution u(x, y) are in the range from 2 to 2, so that the following inequality holds: 2 u(x, y) 2. for all points (x, y) . . The maximum principle implies uniqueness and continuous dependence of solutions on boundary data. Namely, we shall prove the following theorems:

69

Theorem 5.1 (Uniqueness). The Dirichlet boundary value problem has at most one solution. Proof. Assume that there are two solutions u1 and u2 of the Dirichlet boundary problem (5.1). Then, the function v = u1 u2 is continuous in the closed domain and harmonic in the open domain . Since v = 0 on , the maximum principle implies that v 0 and v 0 on , so that v 0 on and u1 u2. Now, we shall state the theorem on continuous dependence of a harmonic function on its boundary values. Theorem 5.2 Let u1 and u2 be the solutions of the Dirichlet boundary value problems u1 = 0 u2 = 0 Then where M = max |1 (x, y) 2(x, y)|.

in , u1 = 1 in , u2 = 2

on on

|u1(x, y) u2 (x, y)| M,

Proof. Let v = u1 u2 . Then, we have max |1(x, y) 2 (x, y)| v(x, y) max |1 (x, y) 2(x, y)|.

Hence, we obtain the inequality |u1(x, y) u2 (x, y)| M, for all (x, y) . Example 1. Suppose that u(x, y) is a continuous function on the closed disk r 1, and harmonic in the open disk r < 1. If u(cos , sin ) sin + cos 2, then show that u(x, y) y + x2 y 2,

for all x2 + y 2 1. . Solution. Note that v(x, y) = y + x2 y 2 is a harmonic function with v(cos , sin ) = sin + cos 2. By the assumption, u v on the boundary of the disk r 1. Then, the maximum of the harmonic function u v on the boundary r = 1 must be less than or equal to zero. The maximum principle then implies that u v 0 throughout the disk.

70

5.4

The Maximum Principle for Poisson Equation

We shall state the maximum Principe for Poissons equation u = f(x, y), or 2u 2u + = f(x, y), (x, y) , x2 y 2 where f(x, y) is given continuous functions in the bounded domain . The following maximum principle holds: Maximum principle. If f(x, y) 0, for (x, y) , then the solution u(x, y) attains non-negative maximum M at a boundary point, or if f(x, y) 0 then u(x, y) attains its non-positive minimum m at a boundary point. This maximum principle can be proved in the same way as maximum principle for harmonic functions. As a consequence of the above maximum principle, we can state the theorem on stability of the boundary problem for equation (5.10) with Dirichlets boundary condition. Theorem 5.3 (Stability Theorem). If the given function f(x, y) is continuous in the closed bounded domain , then the the boundary value problem u = f(x, y), (x, y) , (5.11) (x, y) , (5.10)

u(x, y) = (x, y) (x, y) , is stable in the maximum norm, that is, the solution u(x, y) satises the following inequality: |u(x, y)| max |u(x, y)| + M max |f(x, y)|],
X (x,y)

(x, y) ,

(5.12)

where M = ea 1 for 0 x a. Proof. Without any restriction, we can assume that the domain is on the right side of y axis, that is, 0 x a. We shall prove the theorem on stability using the following lemma: Lemma 5.1 .(Comparison Principle.) If a function v(x, y) is a regular solution of the Poissons equation, and if there exists a regular function g(x, y), which satises the following conditions: 1. g(x, y) max |v(x, y)|,
(x,y)

(x, y)

71

2. g(x, y) max |f(x, y)|,

(x,y)

(x, y) .

then for all (x, y) . |v(x, y)| g(x, y)

Proof of lemma. In order to prove the lemma, we shall show that the functions z1(x, y) = v(x, y) g(x, y), and z2(x, y) = v(x, y) + g(x, y),

satisfy the inequalities z1(x, y) 0 and z2(x, y) 0 for all (x, y) . We note that, by the denition, z1(x, y) 0, on the boundary . By assumption 2, z1(x, y) = v(x, y) g(x, y) f(x, y) + max |f(x, y)| 0 z2(x, y) = v(x, y) + g(x, y) f(x, y) max |f(x, y)| 0 z2(x, y) 0,

for all (x, y) . Hence, by the maximum principle

z1(x, y) = v(x, y) g(x, y) 0, hold for (x, y) . Thus, we have the inequalities g(x, y) v(x, y) g(x, y),

z2(x, y) = v(x, y) + g(x, y) 0,

or

|v(x, y)| g(x, y)

for all (x, y) . End of lemma proof. In order to prove the theorem, we assume that the domain lies in the right side of x axis, that is, x 0. This assumption can be satised by a linear translation of in x direction. We consider the function g(x, y) = max |u(x, y) + [ex ex ] max |f(x, y)|,
(x,y)

(x, y) .

where x x. Here, x is a xed value on x axis. We shall show that the function g(x, y) satises assumptions 1 and 2 of the lemma. Indeed, we estimate

72

1. From the denition of g(x, y), it is clear that g(x, y) max |u(x, y)|
(x,y)

for (x, y) . 2. g(x, y = ex max |f(x, y)| max |f(x, y)|.

for 0 x x. Hence, we obtain the required inequality |u(x, y)| max |u(x, y)| + M max |f(x, y)|,
(x,y)

(x, y) .

where dsM = ea 1 is the upper bound of the expression ex ex ,when 0 x a. Let us note that if the solution u(x, y) satises the boundary condition u(x, y) = 0(x, y), (x, y) , (x, y) .

then we get the following the estimate of the solution |u(x, y)| max |(x, y)| + (ea 1) max f(x, y)|],
(x,y) (x,y)

This inequality means stability of the Dirichlet boundary value problem for the Poissons equation. Question 1. Consider the following boundary value problem: 2u 2u + = sin x + cos y, (x, y) = {(x, y) : 0 < x, y < 1, x2 y 2 u(x, y) = sin xy, (x, y) . (5.13)

Show that the boundary value problem is stable and estimate the solution u(x, y) Solution. We note that the Poissons equation (5.13) satises the assumptions of the theorem on stability. By the thesis, we obtain the following estimate of the solution u(x, y): |u(x, y)| max |(x, y)| + M max |f(x, y)|

= max | sin xy| + M max | sin x + cos y)|

1+2M Because, 0 x 1 therefore, the constant M = e 1 and |u(x, y)| 1 + 2(e 1), for all (x, y) .

73

5.5

The Maximum Principle for Helmholz-Poisson Equation

u + c(x, y)u = f(x, y), or (x, y) , (5.14)

We shall state the maximum Principe for Helmholz-Poisson equation

2u 2u + + c(x, y)u = f(x, y), (x, y) , x2 y 2 where c(x, y) 0, and f(x, y) are given continuous functions in the bounded domain . The following maximum principle holds: Maximum principle. If f(x, y) 0 and c(x, y) 0 for (x, y) , then the solution u(x, y) attains non-negative maximum M at a boundary point, or if f(x, y) 0 and c(x, y) 0, then u(x, y) attains its non-positive minimum m at a boundary point. As a consequence of the above maximum principle, we state the theorem on stability of the solution of Dirichlets boundary problem for equation (5.14). Theorem 5.4 (Stability Theorem). If the given functions c(x, y) 0 and f(x, y) are continuous in the closed bounded domain , then the solution u(x, y) satises the following inequality:
(x,y)

max |u(x, y)| max |u(x, y)| + M max |f(x, y)|],

X (x,y)

(5.15)

where M = ea 1 for 0 x a. We can prove this theorem in a similar way as theorem on stability for Poissons equation. Let us note that if the solution u(x, y) satises the boundary condition u(x, y) = 0(x, y), then we get the following priori estimate of the solution u(x, y) |u(x, y)| max |u(x, y)| + M max f(x, y)|],
(x,y) (x,y)

(x, y) ,

(x, y) .

where M = ea 1, for 0 x a. This inequality means stability of the Dirichlet boundary value problem for the Helmholz-Poisson equation. Question 1. Consider the following boundary value problem: 2u 2u + 2u = sin x + cos y, (x, y) = {(x, y) : 0 < x, y < 1, x2 y 2 u(x, y) = sin xy, (x, y) . (5.16)

74

Show that the boundary value problem is stable and estimate the solution u(x, y) Solution. We note that the Helmholz-Poisson equation (5.16) satises the assumptions of the theorem on stability. Indeed, the coecient c(x, y) = 2 0. By the thesis, we obtain the following estimate of the solution u(x, y): |u(x, y)| max |(x, y)| + M max |f(x, y)|

= max | sin xy| + M max | sin x + cos y)|

1+2M Because, 0 x 1, then the constant M = e 1, and the solution u(x, y) satises the inequality |u(x, y)| 1 + 2(e 1), for all (x, y) .

5.6

Boundary Value Problem for Laplaces Equation in a Rectangle

Let us consider the following boundary value problem: 2u 2u + = 0, x2 y 2 u(x, 0) = 0, (x, y) = {(x, y) : 0 < x < L1 , 0 < y < L2 }, 0 x L1 0 x L1 .

u(0, y) = u(L1 , y) = 0, 0 y L2 , u(x, L2) = (x), (5.17) This boundary value problem can be solved by the method of separation of variables. Namely, let u(x, y) = X(x)Y (y). Then, by substitution to Laplaces equation, we obtain X (x)Y (x) + X(x)Y (y) = 0, or X (x) Y (y) = = . X(x) Y (y) X (x) + X(x) = 0, From the boundary conditions u(x, 0) = X(x)Y (0) = 0. Y (y) Y (y) = 0.

Hence, we have two equations

75

So that Y (0) = 0. Also, we have u(0, y) = X(0)Y (y) = 0 and u(L1, y) = X(L1 )Y (y) = 0

therefore X(0) = X(L1 ) = 0. Let us note the solution of the eigenvalue problem X (x) + X(x) = 0, X(0) = X(L1 ) = 0

is known and the eigenvalues and eigenfunctions are given by the formulae n = n2 2 L2 1

X (n) (x) = sin

nx , L1

n = 1, 2, ...,

Then, we solve the corresponding equation for Y (y) Y (y) n Y (y) = 0, which has the general solution Y (y) = C1 e n y + C2 e n y

for arbitrary constants C1 and C2 . By the condition Y (0) = 0, we nd C1 + C2 = 0 and C2 = C1 , so that the solution ny ny ny , n = 1, 2, ...; Y (n)(y) = C1(e L1 e L1 ) = 2C1 sinh L1 Now, we observe that the terms of the sequence un (x, y) = Bn sin nx ny sinh , L1 L1 n = 1, 2, ...;

are harmonic functions and satisfy the homogeneous boundary value condition at the three sides of the rectangle , that is when y = 0 or x = 0, or x = L1, for arbitrary constants Bn , n = 1, 2, ...,. Then, we consider the solution in the form of the following series:

u(x, y) =
n=1

Bn sin

nx ny sinh , L1 L1

(5.18)

where the coecients Bn , n = 1, 2, ...; are determined by the boundary condition nx nL2 u(x, L2) = Bn sin sinh = 0 (x). L1 L1 n=1

76

Multiplying the above identity by sin

mx and integrating from 0 to L1 , and L1 using orthogonality property, we obtain the following formula for the coecients Bn , n = 1, 2, ....; Bn sinh 2 nL2 = L1 L1
L1 L1 0

0 (s) sin

ns ds L1 (5.19)

or Bn =

n = 1, 2, ...; nL2 0 L1 sinh L1 Example. Consider the following boundary value problem: 2u 2u + = 0, x2 y 2 u(x, 0) = 0,

0 (s) sin

ns ds, L1

(x, y) = {(x, y) : 0 < x < 2, 0 < y < 4}, 0x2 0 x 2.

u(0, y) = u(2, y) = 0, 0 y 4, (5.20) Solution. By the formulae (5.18) and (5.19), we compute the coecients Bn = 2 2 sinh n4 2
0 2

u(x, 4) = x(2 x),

s(2 s) sin

ns 16(1 (1)n ) ds = 3 3 , 2 n sinh 2n

Hence, we nd the solution in the form of the following series u(x, y) = 16 (1 (1)n ) nx sin sinh 2ny, 3 3 sinh 2n n1 n 2 0 x 2, 0 y 4.

5.7

Boundary Value Problem for Laplaces Equation in a Disk

1 1 urr + ur + 2 u = 0, 0 < r < R, < < , r r

Let us consider the Laplaces equation in the polar coordinates (5.21)

with the boundary condition u(R, ) = (), It is easy to check that the functions 1, rn cos n, rn sin n, . (5.22)

77

are harmonic in the disk with radius R and the center at the origin, for n = 1, 2, ...; We shall nd the solution u(r, ) in the form of the following trigonometric series: 1 u(r, ) = a0 + an rn cos n + bn rn sin n. (5.23) 2 n=1 Clearly u(r, ) given by formula (5.23) satises the Laplaces equation in the poplar coordinates (r, ). In order to nd the coecients an and bn , we apply the boundary condition
1 u(R, ) = a0 + an Rn cos n + bn Rn sin n = (), 2 n=1

So, an and bn are the coecients of the Fourier series of the given function () for . These coecients are given by the formulae 1 (s) ds, a0 = (5.24) 1 1 an = n (s) cos ns ds, bn = n (s) sin ns ds,, R R for n = 1, 2, ...; Therefore, the solution u(r, ) of the boundary value problem for Laplaces equation in a disk is given by the formulae (5.23) and (5.24). Example. Let us consider the following boundary value problem: 1 1 urr + ur + 2 u = 0, < < , 0 < r < 2, r r (5.25) u(2, ) = cos , , 4 Solution. In order to nd the solution u(r, ), we apply the formulae (5.23) and (5.24). So that, we compute the Fourier coecients of the function () = cos given in the boundary condition. By formulae (5.24), we nd 4 s 4 2 cos ds = , 4 1 s 4(1)n 2 an = n cos cos ns ds = n , 2 4 2 (16n2 1) 1 s bn = n cos sin ns ds = 0, n = 1, 2, ... , 2 4 Hence, the solution 2 2 4 2 (1)n u(r, ) = rn cos n, n=1 2n (16n2 1) 1 a0 =

(5.26)

78

5.7.1

Fundamental Solution of Laplace Equation

We shall nd the fundamental solution of Laplaces equation. Here, we write the Laplace equations in the variables X = (x1, x2 ), and X = (x1 , x2, x3). 2u 2u + = 0, n=2 x2 x2 1 2 2u 2u 2u + + = 0, n = 3, x2 x2 x2 1 2 3

(5.27)

Let n = 2 and let Y = (y1, y2) be a xed point on x1 , x2plane. Then the distance of the point Y = (y1, y2) from a point X = (x1, x2 ) is r(X, Y ) = We note that (x1 y1)2 + (x2 y2 )2.

r x2 y2 = . x2 r 1 Let r(X, Y ) > 0 and let U(X, Y ) = ln . Now, we compute r(X, Y ) 1 1 ln x1 y1 r = r = x2 y2 , , 2 x1 r x2 r2 1 1 2 2 ln 1 2(x1 y1)2 ln r 1 2(x2 y2)2 r = , = 2 . x2 r2 r4 x2 r r4 11 2 ln Hence, we nd 2U 2U + = 0. x2 x2 1 2 1 , r(X, Y ) > 0 is the harmonic function on r(X, Y )) the whole x, y plane except the focus point Y = (y1, y2). This function is called fundamental solution of the two dimensional Laplace equation. . Now, let n = 3 and Y = (y1, y2 , y3) be a xed point on the R3 space. Then the distance of the point Y from the point X = (x1, x2, x3 ) is Therefore, U(X, Y ) = ln r(X, Y ) = We note that r x1 y1 = , x1 r r x2 y2 = , x2 r r x3 y3 = . x3 r (x1 y1)2 + (x2 y2 )2 + (x3 y3)2 .

x1 y1 r = , x1 r

79

Let r(X, Y )) > 0, X = (x1, x2, x3 ), Y = (y1, y2 , y3) and let U(X, Y ) = Then, we compute

1 . r(X, Y )

U x1 y1 U x2 y2 U x3 y3 = , = , = 3 2 x1 r x2 r x3 r3 2U 3(x1 y1 )2 1 2U 3(x2 y2 )2 1 2u 3(x3 y3)2 1 = 3, = 3, = 3 2 2 2 5 5 5 x2 r r x2 r r x3 r r Hence, we nd 2U 2U 2U + + = 0. x2 x2 x2 1 2 3 1 , r(X, Y )) > 0 is the harmonic function on the r(X, Y )) variables x1 , x2, x3 in the whole space except the focus point Y = (y1 , y2, y3). This function is called fundamental solution of the three dimensional Laplace equation. . Greens identities. Let be a bounded domain in R2 with the piecewise smooth closed boundary . Let u, v C 2() be twice continuously dierentiable functions in . Then, the following second Greens identity holds: Therefore, U(X, Y ) = n = 2, n = 3, (vu uv)d

(u

(vu uv)d =

dv du v )ds. dn dn dv du v )ds. (u dn dn

(5.28)

du denotes normal inner derivative to the boundary . Note that for Here dn n = 2, is a curve on x1, x2 plane, and for n = 3, is a surface in the space R3 . Note that the both functions u and v are not assumed to be harmonic. . Proof. The prove of both Greens identities are similar. So, let us prove the Greens identity for n = 2. Integrating by parts, we have 2u v 2 dx1 = x1 2u v 2 dx2 = x2 u v [v u ] cos(n, x1)ds + x1 x1 u v [v u ] cos(n, x2)ds + x2 x2 2v u 2 dx1 , x1 2v u 2 dx2 . x2

Hence, by adding both sides and moving the term with Laplacian from right to left side, we obtain the Greens identity [vu uv]d = [u dv du v ]ds, dn dn

80

where the normal derivative u u du = cos(n, x1) + cos(n, x2), dn x1 x2 Here cos(n, x1) cos(n, x2) are directive cosines between normal vector n and x1 and x2 axes, respectively. . corollary 1. For harmonic functions u and v in , from Green;s identities, we obtain the following formulae: [v du dv u ]ds = 0, dn dn n=2 (5.29)

du dv [v u ]ds = 0, n = 3. dn dn

In particular, when v = 1, we obtain next corollary corollary 2. Every harmonic function u satises the following identity: du ds = 0. dn n = 2, (5.30)

du ds = 0, n = 3. dn 5.7.2 Theorem on representation of harmonic functions

Representation Theorem. Every harmonic function u(X) in the bounded domain with a smooth boundary satises the following formula u(X) = u(X) = 1 2 1 4 [u(Y ) dU(X, Y ) du(Y ) U(X, Y ) ]dsY , dn dn X , n = 2

dU(X, Y ) du(Y ) U(X, Y ) ]dsY , X , n = 3. dn dn (5.31) Proof. Let u be a harmonic function and v = U be the fundamental solution of Laplace equation. Then, we cannot apply the formula (5.29) to these functions, since the fundamental solution U(X, Y ) has the singular point X = Y . However, we can apply the formula to the domain 0 = K, where K is a disk when n = 2 or a ball when n = 3. So that, in 0 , we have [u(Y ) [u dU du U ]ds = 0, dn dn n=2 (5.32)

dU du [u U ]ds = 0, n = 3. dn dn 0

81

Now, let us write the above integrals along the boundary of and the boundary K of K. Taking into consideration negative orientation of K with respect , from (5.32), we have [u dU du U ]ds = dn dn [u dU du U ]ds, dn dn n=2 (5.33)

dU du [u U ]ds = dn dn

dU du [u U ]ds n = 3. dn dn K

Next, we compute the normal derivative of the fundamental solution U(X, Y ) to the boundary K 1 ln , n = 2, X = (x1 , x2) 0 , Y = (y1, y2 ) K, r U(X, Y ) = 1 , n = 3, X = (x1 , x2, x3) 0 , Y = (y1, y2 , y3) K r

Then, we compute dU(X, Y ) = dn

1 , n = 2, X = (x1, x2 ) 0 , Y = (y1, y2) K, r 1 , n = 3, X = (x1, x2 , x3) 0 , Y = (y1, y2, y3 ) K r2

Substituting the above formulae for the normal derivatives into (5.33), we obtain

[u

dU du 1 U ] ds = dn dn r

u ds ln

1 r

du ds, K dn

n=2 (5.34)

dU du 1 [u U ]ds = 2 dn dn r Hence, by corollary 2, we have [U

1 u ds r K

du ds n = 3. K dn

u dU 1 u ]ds = dn dn r

u(Y )dsY , u(Y )dsY

n=2 (5.35) n = 3.

du dU 1 [U u ]ds = 2 dn dn r Now, we apply the identity

K

u(Y )dsY =

[u(Y ) u(X)]dsY + u(X)

dsY ,

(5.36)

Because of uniform continuity of the harmonic function u(X), we have |u(Y ) u(X)| <

82

for suciently small r(X, Y ). Then, we have | | [u(Y ) u(X)]dsY | < 2 r [u(Y ) u(X)]dsY | < 4 r
2

for n = 2 (5.37) for n = 3.

Using (5.36) and (5.37), we compute the limits 1 r>0 r(X, Y ) lim
r>0 r 2 (X, Y

u(Y )dsY = 2u(X)

for n = 2,

lim

1 )
K

u(Y )dsY = 4u(X), for n = 3.

Hence by formula (5.35), we obtain equality (5.31). This ends the proof. From the representative theorem, we conclude the following important formula concerning boundary value problems for harmonic functions. Namely, let us substitute to (5.31), u(X) 1. Then, we obtain the formula dU(X, Y ) dsY dn dU(X.Y ) dsY dn = 2, X , n = 2 = 4, X , n = 3.

(5.38)

Gauss Mean Value Formula Let = K be the disk (n = 2) or the ball (n = 3) with the radius R and the center at X. Then, on the surface of K, we have 1 dU 1 U(X, Y ) = ln , = , n = 2, R dn R (5.39) 1 dU 1 U(X, Y ) = , = 2, n = 3 R dn R Hence, by the representative theorem, we obtain Gauss Mean Value Formula for harmonic functions u(X) = 1 2R u(Y )dsY , X = (x1, x2) K, n = 2, (5.40)

1 u(X) = 4R2

u(Y )dsY , X = (x1, x2, x3 ) K, n = 3.

83

5.7.3

Greens Function

. We note that every harmonic function H(X, Y ) of the variable Y , and at xed X, satises the identity (see formula (5.29)) [u(Y ) dH(X, Y ) du(Y ) H(X, Y ) ]ds = 0, dn dn n=2 (5.41)

dH(X, Y ) du(Y ) [u(Y ) H(X, Y ) ]ds = 0, n = 3. dn dn

Let us choose the harmonic function H(X, Y ) = U(X, Y ) for Y at xed X . Then, the Greens function is G(X, Y ) = U(X, Y ) H(X, Y ) By the representative theorem (see formula (5.31)), Greens function satises identity u(X) = u(X) = 1 2 1 4 [u(Y ) dG(X, Y ) du(Y ) G(X, Y ) ]dsY , dn dn X , n = 2

dG(X, Y ) du(Y ) G(X, Y ) ]dsY , X , n = 3. dn dn (5.42) Because G(X, Y ) = 0, for Y , X , therefore, every harmonic function u(X) in , satises the identity [u(Y ) u(X) = 1 2 u(Y ) dG(X, Y ) dsY , dn X , n = 2

1 u(X) = 4

dG(X, Y ) u(Y ) dsY , X , n = 3. dn

(5.43)

Hence, by the formula (5.43), we arrive at the following theorem: Theorem 5.5 If G(X, Y ) is the Greens function for Laplaces equation, then the solution u(X) of the Dirichlets problem u(X) = 0, is given by the formula u(X) = 1 2 (Y ) dG(X, Y ) dsY , dn X , n = 2 X , u(X) = (X), X

1 u(X) = 4

dG(X, Y ) (Y ) dsY , X , n = 3. dn

(5.44)

84

Greens function for a disk with Dirichlets condition. Let r(0, Y ) < R, be the disk with the radius R and the center at the origin. Let us denote by = 0X, = 0X,
X

r = Y X,

r = Y X.

96 X s s r r 0 t rt r R s Y 87

Disk K One can check that the Greens function for the disk K is R 1 r 1 ln( . ) = ln , X = 0, r r Rr G(X, Y ) = ln 1 ln 1 , X=0 r R

ln

(5.45)

Greens function for a ball with Dirichlets conditions. Similarly, we construct Greens function for a ball K(0, R) = {X = (x1, x2, x3 ) : r(0, X) < R}. Then, following the notations given above, we write the Greens function for the ball R 1 1 . , X = 0, r r (5.46) G(X, Y ) = 1 1 , X=0 r R Poissons Integral. Let us note that in the case when the domain is a disk K or a ball K, the Greens function is given by formulae (5.45) and (5.46). Then, the solution u(X) of the Dirichlets boundary value problem, in the case , is given by the Poissons integral (see Representative Theorem, formulae (5.44)). u(X) = 1 2 (Y ) dG(X, Y ) dsY , dn X K, n = 2

1 u(X) = 4

dG(X, Y ) (Y ) dsY , X K, n = 3. dn K

(5.47)

In order to express the solution u(X) in a transparent form, we shall evaluate G(X, Y ) the kernel . Let n = 3 and X = 0. Then, we nd nY dG(X, Y ) 1 dr R 1 dr = 2 + . 2. dnY r dnY r dnY (5.48)

85

Because
3

r=
i=1

(yi xi )2

therefore dr = dnY yi But cos(n, xi ) = , so that R

3 i=1

(yi xi )2 cos(n, xi ).

3 1 3 dr 2 = ( xi yi yi ). dnY rR i=1 i=1 3 3 3 yi , and the expression i=1 i=1

(5.49)

We note that R2 =

xiyi is the inner product of

the vectors 0X and 0Y . Hence, it implies the identity

3

xi yi = R,
i=1

where is the angle between the vectors 0X and 0Y . Then, we can write formula (5.49) in the form dr 1 = ( cos R). dnY r In a similar way, we obtain the derivative Taking into consideration the proportion dr . dnY (5.50)

R r = = , r R we obtain dr 1 R2 1 = ( cos R) = ( cos R) = (R cos ). dnY r Rr r From the formulae (5.48), (5.50) and (5.51), we nd dG(X, Y ) 1 R = 2 ( cos R) + (R cos ) dnY r rr 1 = 3 [( cos R) + (R cos )] r R Hence, w obtain the equality dG(X, Y ) R2 2 = , dnY Rr3 X K, m = 3. (5.51)

(5.52)

(5.53)

86

Therefore, Poissons integral takes the following form: u(X) = and u(X) = 1 4R
K

1 2R

R2 2 (Y )dsY , r2 R2 2 (Y )dsY , r3

X K, n = 2.

(5.54)

X K,

n = 3,

(5.55)

On the gure, we present the parameters r, R, , and of the Greens function for the disk K.

. Let us write the Poissons integral in the polar coordinates y1 = R cos , y2 = R sin

. From the gure and by cosine formula, we nd r2 = R2 + 2 2R cos( ). Therefore, the kernel of the Poissons integral R2 2 R2 2 = 2 . r2 R + 2 2R cos( ) (5.56)

When point Y is moving along the circle K with the radius R, at xed pointX, the angle [0, 2]. Changing the variable of integration Y = (y1, y2) K,

87

to polar coordinates y1 = R cos , y2 = R sin by (5.54) and (5.56), we obtain the solution of the boundary value problem given by the Poissons integral u(, ) = 1 2
2 0

R2 2 ()d. R2 + 2 2R cos( )

(5.57)

We can interpret the Poisson integral solution (5.57) as nding the potential u at (, ) as a weighted average of the boundary potential () weighted by the Poissons kernel (5.56). This tells us something about physical systems: namely that the potential at a point is the weighted average of neighboring potentials. The Poissons kernel tells just how much weight to assign to each point. Let us note that the potential at the center of the circle is given by the formula u(0, 0) = 1 2
2 0

() d.

(5.58)

Example 1. Evaluate the potential at the center of the circle by Poisson integral for the boundary given function () = cos , 0 2. 4 Solution. By formula (5.57), we compute u(0, 0) = 1 2
2 0

()d =

1 2

2 0

cos

2 d = . 4

Example 2. Show that the integral of Poisson kernel is equal to 2, that is

2 0

R2 2 d = 2. R2 + 2 2R cos( )

(5.59)

Solution. Let the boundary given function () 1. Then, the solution of the boundary problem u(X) = 1 for all X K. Thus, by formula (5.57), we get 1 2 R2 2 1= d. 2 0 R2 + 2 2R cos( ) Hence
2 0

Example 3. Solve the following boundary value problem using Poisson integral uxx + uyy = 0, (x, y) K = {x2 + y 2 < 4}, u(x, y) = 2, 1 2
2 0

R2 2 d = 2. R2 + 2 2R cos( )

Solution . By the Poisson formula, the solution in the polar coordinates is u(, ) = 4 2 2 d, 4 + 2 2 4 cos( )

(x, y) K = {x2 + y 2 = 4},

88

Hence, by formula (5.59), we compute u(, ) = 2 2

2 0

4 2 d = 2. 4 + 2 2 4 cos( )

In the Cartesian coordinates, we have also the constant solution u(x, y) = 2 for all (x, y) K.

5.8

Helmholz Equation and Eigenvalue Problem.

We shall solve the following Helmholz equation in the polar coordinates (r, ), x = r cos , y = r sin with homogeneous boundary conditions: u(r, ) + 2 u(r, ) = 0, (5.60) u(1, ) = 0, where 0 2,

1 1 u(r, ) urr + ur + 2 u . r r Applying the method of separation of variables let us substitute to Helmholz equation u(r, ) = R(r)().

Then, we obtain r2 R + rR + (2 r2 n2 ))R = 0, R(1) = 0, + n2 = 0. Bessels Equation Now, we shall solve the ordinary dierential Bessels equation r2 R + rR + (2 r2 n2 ))R = 0, 0 < r < 1, R(1) < , R(1) = 0, + n2 = 0. As we know from the theory of ordinary dierential equations, that Bessels equation has two linearly independent solutions 1. R1(r) = AJn (r), 2. R2(r) = BYn (r), n-th order Bessel function of the rst kind, n-th order Bessel function of the second kind. physical condition, (5.62) Bessel s equation (5.61)

89

Here, there are power series representation of the Bessel functions r n (1)m r ) ( )2m , 2 m=0 m!(n + m)! 2 n1 2 r 1 r (n k 1)! (r)2 k Yn (r) = ln Jn (r) ( )n [ ] . 2 2 k! 4 k=0 Jn (r) = ( Since Yn (r) is unbounded at r = 0, we choose as our solution R(r) = AJn (r). Next, we nd R(r) and using the boundary condition R(1) = 1. Namely, substituting R(1) = 0 into AJn (r), we obtain Jn () = 0. In other words, in order to be R(r) = 0 on the boundary K of the circle K, we must pick the separation constant to be one of roots of the equation Jn (r) = 0, that is = knm , where knm is the m-th root of Jn (r) = 0. Finally, we obtain the solution unm (r, ) = AJn (knm r).

5.9

Exercises

Question 1. Consider the following boundary value problem: uxx + uyy = 0, (x, y) = {(x, y) : 0 < x < 1, 0 < y < 1}, (5.63)

Find the range of the values of the solution u(x, y). which satises one of the boundary condition (a) u(x, y) = sinxy, (b) u(x, y) = cos x y, (x, y) , (x, y) , (x, y) , (x, y) ,

(c) u(x, y) = sin x + sin y, (d) u(x, y) = cos x + cos y,

Question 2. Show that the following boundary value problem is stable uxx + uyy u = f(x, y), (x, y) = {(x, y) : 1 < x < 1, 1 < y < 1}, u(x, y) = (x, y), ((x, y) . (5.64)

90

for given continuous functions f(x, y) and (x, y). Give an priori estimate of 2 2 the solution u(x, y), when f(x, y) = ex y , (x, y) and (x, y) = 0, (x, y) . Question 3. Solve the following boundary value problem: 2u 2u + = 0, x2 y 2 u(x, 0) = 0, (x, y) = {(x, y) : 0 < x < 1, 0 < y < 2}, 0x1 0 x 1.

u(0, y) = u(1, y) = 0, 0 y 2, u(x, 2) = sin x, (5.65)

Question 4. Solve the following boundary value problem: 1 1 urr + ur + 2 u = 0, < < , 0 < r < 4, r r u(4, ) = cos , , 2

(5.66)

Index
Bessels equation, 74 boundary conditions, 25 boundary value problem, 52 canonical form, 9 characteristic equation, 11 characteristics equations, 12 cylindrical coordinates, 52 DAlembert solution, 19 Dirichlet conditions, 53 elliptic equation, 8 rst order PDE, 1 Fourier series, 28 fundamental solution, 64 Gauss mean value theorem, 68 general solution, 1 Green function, 65 Greens function, 69 Greens identity, 65 harmonic functions, 53 heat equation, 8 Helmholtz equation, 59 Hopfs theorem, 58 hyperbolic equation, 8, 19 initial boundary problem, 22 Jacobian, 10, 15 Laplace equation, 8, 51 maximum principle, 53 method of separation of variables, 23 Neumann conditions, 53
91

operator elliptic, 8 operator hyperbolic, 8 operator parabolic, 8 parabolic equation, 8 Poisson equation, 59 Poissons integral, 70 polar coordinates, 51 representation theorem, 66 spherical coordinates, 52 stability theorem, 57 string equation, 27 third kind boundary conditions, 53 wave equation, 8