21.

Green Functions for First Order Equations

L[y ] y + p(x)y = f (x) for x > a (21.2)

Consider the rst order inhomogeneous equation

subject to a homogeneous initial condition, B [y ] y (a) = 0. The Green function G(x| ) is dened as the solution to L[G(x| )] = (x ) subject to G(a| ) = 0.

We can represent the solution to the inhomogeneous problem in Equation 21.2 as an integral involving the Green function. To show that G(x| )f ( ) d y (x) =
a

is the solution, we apply the linear operator L to the integral. (Assume that the integral is uniformly convergent.) L G(x| )f ( ) d = L[G(x| )]f ( ) d a a = (x )f ( ) d
a

= f (x)

The integral also satises the initial condition. B G(x| )f ( ) d = B [G(x| )]f ( ) d a a = (0)f ( ) d
a

=0

Now we consider the qualitiative behavior of the Green function. For x = , the Green function is simply a homogeneous solution of the dierential equation, however at x = we expect some singular behavior. G (x| ) will have a Dirac delta function type singularity. This means that G(x| ) will have a jump discontinuity at x = . We integrate the dierential equation on the vanishing interval ( . . . + ) to determine this jump. G + p(x)G = (x ) + G( + | ) G( | ) + p(x)G(x| ) dx = 1

G( | ) G( | ) = 1

(21.3)

The homogeneous solution of the dierential equation is yh = e

p(x) dx

Since the Green function satises the homogeneous equation for x = , it will be a constant times this homogeneous solution for x < and x > . c1 e p(x) dx a<x< G(x| ) = p ( x ) d x <x c2 e 661

In order to satisfy the homogeneous initial condition G(a| ) = 0, the Green function must vanish on the interval (a . . . ). 0 a<x< G(x| ) = <x c e p(x) dx The jump condition, (Equation 21.3), gives us the constraint G( + | ) = 1. This determines the constant in the homogeneous solution for x > . 0 a<x< x G(x| ) = e p(t) dt <x We can use the Heaviside function to write the Green function without using a case statement. G(x| ) = e
x

p(t) dt

H (x )

Clearly the Green function is of little value in solving the inhomogeneous dierential equation in Equation 21.2, as we can solve that problem directly. However, we will encounter rst order Green function problems in solving some partial dierential equations.

Result 21.6.1 The rst order inhomogeneous dierential equation with homogeneous initial condition L[y ] y + p(x)y = f (x) has the solution y=

for a < x

y (a) = 0

G(x| )f ( ) d
a

where G(x| ) satises the equation L[G(x| )] = (x ) The Green function is G(x| ) = e
x

for a < x

G(a| ) = 0.

pt) dt

H (x )

21.7

Green Functions for Second Order Equations

L[y ] = y + p(x)y + q (x)y = f (x) for a < x < b (21.4)

Consider the second order inhomogeneous equation

subject to the homogeneous boundary conditions B1 [y ] = B2 [y ] = 0. The Green function G(x| ) is dened as the solution to L[G(x| )] = (x ) subject to B1 [G] = B2 [G] = 0.

The Green function is useful because you can represent the solution to the inhomogeneous problem in Equation 21.4 as an integral involving the Green function. To show that y (x) =
b

G(x| )f ( ) d
a

662

is the solution, we apply the linear operator L to the integral. (Assume that the integral is uniformly convergent.)
b b

G(x| )f ( ) d =

L[G(x| )]f ( ) d (x )f ( ) d

= f (x) The integral also satises the boundary conditions. b G(x| )f ( ) d = Bi

a

b a

Bi [G(x| )]f ( ) d
a b

=0

[0]f ( ) d
a

One of the advantages of using Green functions is that once you nd the Green function for a linear operator and certain homogeneous boundary conditions, L[G] = (x ) B1 [G] = B2 [G] = 0

you can write the solution for any inhomogeneity, f (x). L[f ] = f (x) B1 [y ] = B2 [y ] = 0

You do not need to do any extra work to obtain the solution for a dierent inhomogeneous term. Qualitatively, what kind of behavior will the Green function for a second order dierential equation have? Will it have a delta function singularity; will it be continuous? To answer these questions we will rst look at the behavior of integrals and derivatives of (x). The integral of (x) is the Heaviside function, H (x). x 0 for x < 0 (t) dt = H (x) = 1 for x > 0 The integral of the Heaviside function is the ramp function, r(x). x 0 for x < 0 r(x) = H (t) dt = x for x > 0 The derivative of the delta function is zero for x = 0. At x = 0 it goes from 0 up to +, down to and then back up to 0. In Figure 21.2 we see conceptually the behavior of the ramp function, the Heaviside function, the delta function, and the derivative of the delta function. We write the dierential equation for the Green function. G (x| ) + p(x)G (x| ) + q (x)G(x| ) = (x ) we see that only the G (x| ) term can have a delta function type singularity. If one of the other terms had a delta function type singularity then G (x| ) would be more singular than a delta function and there would be nothing in the right hand side of the equation to match this kind of singularity. Analogous to the progression from a delta function to a Heaviside function to a ramp function, we see that G (x| ) will have a jump discontinuity and G(x| ) will be continuous.

663

Figure 21.2: r(x), H (x), (x) and

d dx (x)

Let y1 and y2 be two linearly independent solutions to the homogeneous equation, L[y ] = 0. Since the Green function satises the homogeneous equation for x = , it will be a linear combination of the homogeneous solutions. G(x| ) = c 1 y1 + c 2 y2 d 1 y1 + d 2 y2 for x < for x >

We require that G(x| ) be continuous. G(x| )x = G(x| )x+

We can write this in terms of the homogeneous solutions.

c1 y1 ( ) + c2 y2 ( ) = d1 y1 ( ) + d2 y2 ( ) We integrate L[G(x| )] = (x ) from to +.

+

[G (x| ) + p(x)G (x| ) + q (x)G(x| )] dx =

(x ) dx.

Since G(x| ) is continuous and G (x| ) has only a jump discontinuity two of the terms vanish.
+

p(x)G (x| ) dx = 0

and
+

q (x)G(x| ) dx = 0

G (x| ) dx =

(x ) dx

G (x| )

G ( + | ) G ( | ) = 1

+ = H (x )

We write this jump condition in terms of the homogeneous solutions.

d 1 y1 ( ) + d2 y2 ( ) c1 y1 ( ) c2 y2 ( ) = 1

Combined with the two boundary conditions, this gives us a total of four equations to determine our four constants, c1 , c2 , d1 , and d2 .

664

Result 21.7.1 The second order inhomogeneous dierential equation with homogeneous boundary conditions L[y ] = y + p(x)y + q (x)y = f (x) has the solution y=
b

for a < x < b

B1 [y ] = B2 [y ] = 0

G(x| )f ( ) d
a

where G(x| ) satises the equation L[G(x| )] = (x ) for a < x < b B1 [G(x| )] = B2 [G(x| )] = 0.

G(x| ) is continuous and G (x| ) has a jump discontinuity of height 1 at x = .

Example 21.7.1 Solve the boundary value problem y = f (x) y (0) = y (1) = 0

using a Green function. A pair of solutions to the homogeneous equation are y1 = 1 and y2 = x. First note that only the trivial solution to the homogeneous equation satises the homogeneous boundary conditions. Thus there is a unique solution to this problem. The Green function satises G (x| ) = (x ) The Green function has the form G(x| ) = c1 + c2 x d1 + d2 x for x < for x > . G(0| ) = G(1| ) = 0.

Applying the two boundary conditions, we see that c1 = 0 and d1 = d2 . The Green function now has the form cx for x < G(x| ) = d(x 1) for x > . Since the Green function must be continuous, c = d( 1) From the jump condition, d d c (x 1) cx =1 dx 1 dx x= x= c c=1 1 c = 1. Thus the Green function is G(x| ) = ( 1)x (x 1) for x < for x > . d=c . 1

The Green function is plotted in Figure 21.3 for various values of . The solution to y = f (x) is

665

0.1 -0.1 -0.2 -0.3 0.5 1

0.1 -0.1 -0.2 -0.3 0.5 1

0.1 -0.1 -0.2 -0.3 0.5 1

0.1 -0.1 -0.2 -0.3 0.5 1

Figure 21.3: Plot of G(x|0.05),G(x|0.25),G(x|0.5) and G(x|0.75).

y (x) = y (x) = (x 1)
x

G(x| )f ( ) d
0

f ( ) d + x
0

( 1)f ( ) d.
x

Example 21.7.2 Solve the boundary value problem y = f (x) y (0) = 1 y (1) = 2.

In Example 21.7.1 we saw that the solution to u = f (x) is u(x) = (x 1) Now we have to nd the solution to
0 x

u(0) = u(1) = 0
1

f ( ) d + x

( 1)f ( ) d.
x

v = 0 The general solution is

v (0) = 1 v = c1 + c2 x.

u(1) = 2.

Applying the boundary conditions yields v = 1 + x. Thus the solution for y is y = 1 + x + (x 1) Example 21.7.3 Consider y = x Method 1. y (0) = y (1) = 0.
x

f ( ) d + x
0

( 1)f ( xi) d.
x

Integrating the dierential equation twice yields y= 1 3 x + c 1 x + c2 . 6

Applying the boundary conditions, we nd that the solution is y= 1 3 (x x). 6

666

Method 2.

Using the Green function to nd the solution, 1 x 2 d + x ( 1) d y = (x 1) 0 x 1 1 1 3 1 2 1 x + x = (x 1) x3 + x 3 3 2 3 2 y= 1 3 (x x). 6

Example 21.7.4 Find the solution to the dierential equation y y = sin x that is bounded for all x. The Green function for this problem satises G (x| ) G(x| ) = (x ). The homogeneous solutions are y1 = ex , and y2 = ex . The Green function has the form for x < c1 ex +c2 ex G(x| ) = x x for x > . d1 e +d2 e Since the solution must be bounded for all x, the Green function must also be bounded. Thus c2 = d1 = 0. The Green function now has the form c ex for x < G(x| ) = d ex for x > . Requiring that G(x| ) be continuous gives us the condition c e = d e d = c e2 .

G(x| ) has a jump discontinuity of height 1 at x = . d x d 2 x e ce e c =1 dx dx x= x= c e2 e c e = 1 1 c = e 2 ex 1 2 ex+ 1 2

The Green function is then G(x| ) =

for x < for x >

1 G(x| ) = ex . 2 A plot of G(x|0) is given in Figure 21.4. The solution to y y = sin x is 1 y (x) = ex sin d 2 x 1 x x+ = sin e d + sin e d 2 x 1 sin x + cos x sin x + cos x = ( + ) 2 2 2 y= 1 sin x. 2

667

0.6 0.4 0.2 -4 -2 -0.2 -0.4 -0.6 2 4

Figure 21.4: Plot of G(x|0).

21.7.1

Green Functions for Sturm-Liouville Problems

L[y ] = (p(x)y ) + q (x)y = f (x) subject to B1 [y ] = 1 y (a) + 2 y (a) = 0 B2 [y ] = 1 y (b) + 2 y (b) = 0.

Consider the problem

This is known as a Sturm-Liouville problem. Equations of this type often occur when solving partial dierential equations. The Green function associated with this problem satises L[G(x| )] = (x ) B1 [G(x| )] = B2 [G(x| )] = 0.

Let y1 and y2 be two non-trivial homogeneous solutions that satisfy the left and right boundary conditions, respectively. L[y1 ] = 0 B1 [y1 ] = 0 L[y2 ] = 0 B2 [y2 ] = 0.

The Green function satises the homogeneous equation for x = and satises the homogeneous boundary conditions. Thus it must have the following form. c1 ( )y1 (x) for a x G(x| ) = for x b c2 ( )y2 (x) Here c1 and c2 are unknown functions of . The rst constraint on c1 and c2 comes from the continuity condition. G( | ) = G( + | ) c1 ( )y1 ( ) = c2 ( )y2 ( ) We write the inhomogeneous equation in the standard form. G (x| ) + (x ) p q G (x| ) + G(x| ) = p p p

The second constraint on c1 and c2 comes from the jump condition. 1 p( ) 1 c2 ( )y2 ( ) c1 ( )y1 ( ) = p( ) G ( + | ) G ( | ) =

668

Now we have a system of equations to determine c1 and c2 . c1 ( )y1 ( ) c2 ( )y2 ( ) = 0

c1 ( )y1 ( ) c2 ( )y2 ( ) =

1 p( )

We solve this system with Kramers rule. c1 ( ) = y2 ( ) p( )(W ( )) c2 ( ) = y1 ( ) p( )(W ( ))

Here W (x) is the Wronskian of y1 (x) and y2 (x). The Green function is G(x| ) = y
1 (x)y2 ( ) p( )W ( ) y2 (x)y1 ( ) p( )W ( )

for a x for x b.

The solution of the Sturm-Liouville problem is b y= G(x| )f ( ) d.

a

Result 21.7.2 The problem L[y ] = (p(x)y ) + q (x)y = f (x) subject to B1 [y ] = 1 y (a) + 2 y (a) = 0 B2 [y ] = 1 y (b) + 2 y (b) = 0. has the Green function G(x| ) = y
1 x)y2 ) p )W ) y2 x)y1 ) p )W )

for a x for x b

where y1 and y2 are non-trivial homogeneous solutions that satisfy B1 [y1 ] = B2 [y2 ] = 0, and W (x) is the Wronskian of y1 and y2 .
Example 21.7.5 Consider the equation y y = f (x) y (0) = y (1) = 0.

A set of solutions to the homogeneous equation is {ex ex }. Equivalently, one could use the set {cosh x sinh x}. Note that sinh x satises the left boundary condition and sinh(x 1) satises the right boundary condition. The Wronskian of these two homogeneous solutions is sinh x sinh(x 1) W (x) = cosh x cosh(x 1) = sinh x cosh(x 1) cosh x sinh(x 1) 1 1 = [sinh(2x 1) + sinh(1)] [sinh(2x 1) sinh(1)] 2 2 = sinh(1).

The Green function for the problem is then G(x| ) = sinh x sinh(1)
sinh(1) sinh(x1) sinh sinh(1)

for 0 x for x 1.

669

The solution to the problem is y= sinh(x 1) sinh(1)

x

sinh( )f ( ) d +
0

sinh(x) sinh(1)

sinh( 1)f ( ) d.
x

21.7.2
Consider

Initial Value Problems

L[y ] = y + p(x)y + q (x)y = f (x) for a < x < b

subject the the initial conditions y ( a ) = 1 The solution is y = u + v where u + p(x)u + q (x)u = f (x) and v + p(x)v + q (x)v = 0 Since the Wronskian v (a) = 1 v (a) = 2 . u(a) = 0 u (a) = 0 y (a) = 2 .

is non-vanishing, the solutions of the dierential equation for v are linearly independent. Thus there is a unique solution for v that satises the initial conditions. The Green function for u satises G (x| ) + p(x)G (x| ) + q (x)G(x| ) = (x ) The continuity and jump conditions are G( | ) = G( + | ) G ( | ) + 1 = G ( + | ). G(a| ) = 0 G (a| ) = 0.

W (x) = c exp p(x) dx

Let u1 and u2 be two linearly independent solutions of the dierential equation. For x < , G(x| ) is a linear combination of these solutions. Since the Wronskian is non-vanishing, only the trivial solution satises the homogeneous initial conditions. The Green function must be 0 for x < G(x| ) = for x > u (x) where u (x) is the linear combination of u1 and u2 that satises u ( ) = 0 u ( ) = 1.

Note that the non-vanishing Wronskian ensures a unique solution for u . We can write the Green function in the form G(x| ) = H (x )u (x). This is known as the causal solution. The solution for u is b u= G(x| )f ( ) d
a

= =

H (x )u (x)f ( ) d u (x)f ( ) d

a x
a

670

Now we have the solution for y , y=v+

x

u (x)f ( ) d.
a

Result 21.7.3 The solution of the problem y + p(x)y + q (x)y = f (x) is y = yh +

x

y (a) = 1 y (x)f ( ) d

y (a) = 2

where yh is the combination of the homogeneous solutions of the equation that satisfy the initial conditions and y (x) is the linear combination of homoge ( ) = 1. neous solutions that satisfy y ( ) = 0, y 21.7.3
Consider L[y ] = y + p(x)y + q (x)y = f (x) subject the the unmixed boundary conditions 1 y (a) + 2 y (a) = 1 The solution is y = u + v where u + p(x)u + q (x)u = f (x) and v + p(x)v + q (x)v = 0 1 v (a) + 2 v (a) = 1 1 v (b) + 2 v (b) = 2 . The problem for v may have no solution, a unique solution or an innite number of solutions. We consider only the case that there is a unique solution for v . In this case the homogeneous equation subject to homogeneous boundary conditions has only the trivial solution. The Green function for u satises G (x| ) + p(x)G (x| ) + q (x)G(x| ) = (x ) 1 G(a| ) + 2 G (a| ) = 0 The continuity and jump conditions are G( | ) = G( + | ) G ( | ) + 1 = G ( + | ). 1 G(b| ) + 2 G (b| ) = 0. 1 u(a) + 2 u (a) = 0 1 u(b) + 2 u (b) = 0 1 y (b) + 2 y (b) = 2 . for a < x < b

Problems with Unmixed Boundary Conditions

Let u1 and u2 be two solutions of the homogeneous equation that satisfy the left and right boundary conditions, respectively. The non-vanishing of the Wronskian ensures that these solutions exist. Let W (x) denote the Wronskian of u1 and u2 . Since the homogeneous equation with homogeneous boundary conditions has only the trivial solution, W (x) is nonzero on [a b]. The Green function has the form c 1 u1 for x < G(x| ) = for x > . c 2 u2 671

The continuity and jump conditions for Green function gives us the equations c1 u1 ( ) c2 u2 ( ) = 0 c 1 u 1 ( ) c2 u2 ( ) = 1. Using Kramers rule, the solution is c1 = Thus the Green function is G(x| ) = The solution for u is u=
a

u2 ( ) W ( ) u

c2 =

u1 ( ) . W ( ) for x < for x > .

1 (x)u2 ( )

W ( ) u1 ( )u2 (x) W ( ) b

G(x| )f ( ) d.

Thus if there is a unique solution for v , the solution for y is b y=v+ G(x| )f ( ) d.
a

Result 21.7.4 Consider the problem y + p(x)y + q (x)y = f (x) 1 y (a) + 2 y (a) = 1 1 y (b) + 2 y (b) = 2 .

If the homogeneous dierential equation subject to the inhomogeneous boundary conditions has the unique solution yh , then the problem has the unique solution b y = yh + G(x| )f ( ) d
a

where

G(x| ) =

1 x)u2 )

W ) u1 )u2 x) W )

for x < for x >

u1 and u2 are solutions of the homogeneous dierential equation that satisfy the left and right boundary conditions, respectively, and W (x) is the Wronskian of u1 and u2 . 21.7.4
Consider L[y ] = y + p(x)y + q (x)y = f (x) subject the the mixed boundary conditions B1 [y ] = 11 y (a) + 12 y (a) + 11 y (b) + 12 y (b) = 1 B2 [y ] = 21 y (a) + 22 y (a) + 21 y (b) + 22 y (b) = 2 . The solution is y = u + v where u + p(x)u + q (x)u = f (x) and v + p(x)v + q (x)v = 0 B1 [v ] = 1 B2 [v ] = 2 . B1 [u] = 0 B2 [u] = 0 for a < x < b

Problems with Mixed Boundary Conditions

672

The problem for v may have no solution, a unique solution or an innite number of solutions. Again we consider only the case that there is a unique solution for v . In this case the homogeneous equation subject to homogeneous boundary conditions has only the trivial solution. Let y1 and y2 be two solutions of the homogeneous equation that satisfy the boundary conditions B1 [y1 ] = 0 and B2 [y2 ] = 0. Since the completely homogeneous problem has no solutions, we know that B1 [y2 ] and B2 [y1 ] are nonzero. The solution for v has the form v = c 1 y1 + c 2 y2 . Applying the two boundary conditions yields v= 1 2 y1 + y2 . B2 [y1 ] B1 [y2 ]

The Green function for u satises G (x| ) + p(x)G (x| ) + q (x)G(x| ) = (x ) The continuity and jump conditions are G( | ) = G( + | ) G ( | ) + 1 = G ( + | ). B1 [G] = 0 B2 [G] = 0.

We write the Green function as the sum of the causal solution and the two homogeneous solutions G(x| ) = H (x )y (x) + c1 y1 (x) + c2 y2 (x) With this form, the continuity and jump conditions are automatically satised. Applying the boundary conditions yields B1 [G] = B1 [H (x )y ] + c2 B1 [y2 ] = 0 B2 [G] = B2 [H (x )y ] + c1 B2 [y1 ] = 0

(b) + c2 B1 [y2 ] = 0 B1 [G] = 11 y (b) + 12 y (b) + c1 B2 [y1 ] = 0 B2 [G] = 21 y (b) + 22 y

G(x| ) = H (x )y (x)

(b) (b) 21 y (b) + 22 y 11 y (b) + 12 y y1 (x) y2 (x). B2 [y1 ] B1 [y2 ]

Note that the Green function is well dened since B2 [y1 ] and B1 [y2 ] are nonzero. The solution for u is b G(x| )f ( ) d. u=
a

Thus if there is a unique solution for v , the solution for y is y=

b

G(x| )f ( ) d +
a

2 1 y1 + y2 . B2 [y1 ] B1 [y2 ]

673

Result 21.7.5 Consider the problem y + p(x)y + q (x)y = f (x) B1 [y ] = 11 y (a) + 12 y (a) + 11 y (b) + 12 y (b) = 1 B2 [y ] = 21 y (a) + 22 y (a) + 21 y (b) + 22 y (b) = 2 . If the homogeneous dierential equation subject to the homogeneous boundary conditions has no solution, then the problem has the unique solution b 2 1 y= G(x| )f ( ) d + y1 + y2 B 2 [ y1 ] B 1 [ y2 ] a where
21 y (b) + 22 y (b) G(x| ) = H (x )y (x) y1 (x) B 2 [ y1 ] 11 y (b) + 12 y (b) y2 (x) B 1 [ y2 ]

y1 and y2 are solutions of the homogeneous dierential equation that satisfy the rst and second boundary conditions, respectively, and y (x) is the solution ( ) = 1. of the homogeneous equation that satises y ( ) = 0, y

21.8

Green Functions for Higher Order Problems

L[y ] = y (n) + pn1 (x)y (n1) + + p1 (x)y + p0 y = f (x) on a < x < b

Consider the nth order dierential equation

subject to the n independent boundary conditions Bj [y ] = j where the boundary conditions are of the form B [y ]
n 1 k=0

k y (k) (a) +

n 1 k=0

k y (k) (b).

We assume that the coecient functions in the dierential equation are continuous on [a b]. The solution is y = u + v where u and v satisfy L[u] = f (x) and L[v ] = 0 with Bj [v ] = j with Bj [u] = 0

From Result 21.5.3, we know that if the completely homogeneous problem L[w] = 0 with Bj [w] = 0

has only the trivial solution, then the solution for y exists and is unique. We will construct this solution using Green functions.

674

First we consider the problem for v . Let {y1 . . . yn } be a set of linearly independent solutions. The solution for v has the form v = c 1 y1 + + c n yn where the constants are determined by the matrix equation B1 [y1 ] B2 [y1 ] . . . B1 [y2 ] B2 [y2 ] . . . Bn [y2 ] .. . B1 [yn ] B2 [yn ] . . . c1 1 c2 2 . = . . . . . . cn n

Bn [y1 ]

Bn [yn ]

To solve the problem for u we consider the Green function satisfying L[G(x| )] = (x ) with Bj [G] = 0.

Let y (x) be the linear combination of the homogeneous solutions that satisfy the conditions y ( ) = 0 ( ) = 0 y . . . . =. . y
(n2)

( ) = 0 = 1.

(n1) y ( )

The causal solution is then yc (x) = H (x )y (x). The Green function has the form G(x| ) = H (x )y (x) + d1 y1 (x) + + dn yn (x) The constants are determined by the matrix equation B1 [y1 ] B2 [y1 ] . . . B1 [y2 ] B2 [y2 ] . . . Bn [y2 ] .. . B1 [yn ] B2 [yn ] . . . d1 B1 [H (x )y (x)] d2 B2 [H (x )y (x)] . . = . . . . . dn Bn [H (x )y (x)]

Bn [y1 ]

Bn [yn ]

The solution for u then is u=

b

G(x| )f ( ) d.
a

675