Week 02 : Method of C haracteristics From now on we will study one by one classical techniques of obtaining solution formulas for

PDEs. The rst one is the method of characteristics, which is particularly useful when solving rst order equations. 1 . C lassication of rst-order equations. The general form of rst-order PDE ( in R2 ) : F( x , y , u , u x , u y ) = 0, or in R3 : ( x , y) D R2 . (1) ( 2) ( 3) ( 4)

F( x , y , z , u , u x , u y , u z ) = 0. Often the following notation is used in 2D: p = ux , q = u y thus the equation can be written as F( x , y , u , p, q) = 0. The linear equations can be classiced into the following cases, from easier to more dicult: 1 . Linear: 2. Semi-linear: 3. Quasi-linear: 4. General case. F( x , y , u , u x , u y ) = a( x , y) u x + b( x , y) u y + c( x , y) u d( x , y) . F( x , y , u , u x , u y ) = a( x , y) u x + b( x , y) u y c( x , y , u) . F( x , y , u , u x , u y ) = a( x , y , u) u x + b( x , y , u) u y c( x , y , u) .

( 5) ( 6) ( 7)

2. Method of characteristics. We try to nd a method to solve the general rst-order quasi-linear equation a( x , y , u) u x + b( x , y , u) u y = c( x , y , u) . ( 8)

Let s start by thinking geometrically. Consider the 3 dimensional space with coordinates ( x , y , u) . Assume that u = u( x , y) is a solution to the equation, it is clear that it represents a surface in the ( x , y, u) space. Now we explore the geometrical meaning of the equation. Observe that the equation can be written in the form of an inner product a( x , y , u) a( x , y , u) ux ux b( x , y , u) u y . b( x , y , u) u y = 0 ( 9) c( x , y , u) 1 c( x , y , u) 1 Therefore all we need to do is to understand the relation between the vector

First introduce a new function : R

R through

( x , y , u) = u u( x , y) .

Note that in the RHS of the above, the rst u is a variable, the second u is a function. For example, suppose u( x , y) = x 2 + y 2 , then the corresponding ( x , y, u) = u x 2 + y 2 . Now we easily see that x ux uy = y = . (11) u 1

ux uy 1

and the solution. ( 1 0)

Recall ux uy 1 means

that geometrically,

is perpendicular to the solution surface u = u( x , y) .

( and also

) is a normal vector of the surface = 0. As a consequence

On the other hand, from the equation we know that must be tangent to the surface u = u( x , y) .
a ( x , y , u) b ( x , y , u) c( x , y , u)

is perpendicular to the vector

ux uy 1

which

that the vector curve

Now we summarize. We have shown that the equation is equivalent to the geometrical requirement that is any
X( s) Y( s) U( s)

is tangent to the solution surface u = u( x , y) . As a consequence, any integral satisfying

dX = a( X , Y , U) ( 1 2) ds dY = b( X , Y , U) ( 1 3) ds dU = c( X , Y , U) ( 1 4) ds must be contained in one of the solution surfaces. Conversely, any surface woven by such integral curves is a solution surface. The above understanding leads to the following method of characteristics due to Lagrange. Theorem 1 . The general solution of a rst- order, quasi- linear PDE a( x , y , u) u x + b( x , y , u) u y = c( x , y , u) satises F( , ) = 0, ( 1 6) where F is an arb itrary function of ( x , y, u) and ( x , y , u) , and any intersection of the level sets of and is a solution of the characteristic equations dx d y du = = . a b c ( 1 7) ( 1 5)

Remark 2. As we will see soon, , are obtained through solving the characteristic equations. And each F gives a solution to the original equation. Remark 3. The curves mentioned above are called the families of characteristic curves of the equation. Remark 4. As we will see, the main technique in getting and is

a c = b d

ac a c = = . bd b d

( 1 8)

In the following, we will show how to apply this method. We start with the simplest case. 3. Solving linear rst-order equations. 3. 1 . Equations with constant coecients. We start from the simplest case, where a , b , c and d are just constants. Example 5. ( 2. 8, 3( b) ) Find the general solution of the equation a u x + b u y = 0; Solution. The characteristic equations are dx d y du = = . b 0 a ( 20) a , b are constant. ( 1 9)

What we need are two functions ( x , y , u) and ( x , y , u) such that d = 0, d = 0 along the characteristics. Obviously we can take = u. For , notice that thus we can take d( a y b x) = a d y b dx = 0, = a y b x. F( a y b x , u) = 0 u = f( a y b x) . ( 21 ) ( 22) ( 23) ( 24)

As a consequence, the solution satises for any function F. This means for an arbitrary function f.

Example 6. ( C auchy problem, 2. 8, 5( a) ) Oftentimes, the value of the solution along some specic curve in the plane is prescribed. For example, solve 3 u x + 2 u y = 0, u( x , 0) = sin x. ( 25)

Solution. From the above example we know that the general solution takes the form u = f( 2 x 3 y) . Now substituting this into the initial condition, we obtain Therefore

( 26)

f( 2 x) = sin x

x f( x) = sin . 2 2x3y . 2

( 27)

u( x , y) = sin

As a consequence and gives

Example 7. ( c, d 0) What happens when the c, d are not 0? The method still works. We write the characteristic equations dx dy du = = . a b d cu We have dx dy = d( b x a y) = 0, a b 1 1 dx du du = = a 1 d a 1 c u u = C e a c x + c 1 d d e a c x u c 1 d = 0 a d cu dx = b x a y, = ea
1

down ( 28) ( 29) ( 30)

cx

u c 1 d

F( , ) = 0 u = c 1 d + e a
1

cx

f( b x a y) .

Remark 8. In the above, one may be tempted to conclude d u C e a

cx

=0

and try to use u C e a c x as . This is obviously wrong as d somehow disappeared. One should keep in mind that neither or can involve arbitrary constants. It is their values that are arbitrary constants. Example 9. Some times people use the following method of characteristics:

( 31 ) ( 32) ( 33)

( 34)

First solve then solve

dx = a, ds

dy =b ds

( 35) ( 36)

us + c u = d

to obtain the solution u in the form u = u( s , t) . Finally represent s , t by x , y and obtain the solution. This is equivalent to our method but considerably more complicated to use. Anyone who does not believe this should try using this method to the following examples with non-constant coecients. Remark 1 0. One may wonder, how to nd out and eciently? Unfortunately there may not be any short-cut. One way to systematically nd and is the following. The characteristic equations consists of three equations. Pick any two of them. If you can nd general solutions, then you have and . But this fails when any coecient involves all other variables. 3. 2. Equations with non-constant coecients. Example 1 1 . ( 2. 8, 3( h) ) Find the general solution of Solution. The characteristic equations are Using we obtain Thus = x y; On the other hand, from we obtain du = d log y As a consequence we can take Putting these together we obtain which gives

y u y x ux = 1 .

( 37)

dx d y du = = . x y 1 dy dx = y x

( 38) ( 39)

y dx + x dy = 0

d( x y) = 0.

( 40) ( 41 ) ( 42) ( 43) ( 44) ( 45) ( 46)

d y du = 1 y d( u log y) = 0.

= u log y. F( x y , u log y) = 0 u = log y + f( x y) .

Example 1 2. ( 2. 8, 5( c) ) Find the solution of the following Cauchy problem: x ux + y u y = 2 x y , with u = 2 on y = x 2 . ( 47)

Solution. We need to rst nd the general solution, then using the value on y = x 2 to determine the arbitrary function involved. Find the general solution The characteristic equations are Using

dx d y du = = . x y 2xy dx d y = y x

( 48) ( 49)

we obtain d On the other hand, we have therefore The general solution satises F

= u x y.

Determine the solution. We have u = 2 along y = x 2 , that is

y , u xy =0 x

u=xy+ f

u x , x 2 = 2. Using the formula for the general solution, we have As a consequence

x 3 + f( x) = 2

f( x) = 2 x 3 . y x
3

u( x , y) = x y + 2 4. Solving semi-linear rst-order equations.

Example 1 3. ( 2. 8, 3( g) ) Find the general solution of the following equation: y 2 u x x y u y = x ( u 2 y) . dy du dx = = . x y x ( u 2 y) y2 dy dx = y2 xy

Solution. The characteristic equations are

From we have On the other hand, we have

d x2 + y2 = 0 dy du = x y x ( u 2 y)

= x2 + y2 . d( u y) u y = y dy

Thus we take Now gives

= y ( u y) . F x 2 + y 2 , y ( u y) = 0 u = y + y 1 f x2 + y2 .

Example 1 4. ( 2. 8, 5( g) ) Solve x ux + y u y = u + 1 Solution. The characteritic equations are dx dy du = = x y u+1 ( 66) with u( x , y) = x 2 on y = x 2 . ( 65)

du u = +2 dy y

du = 2 x dy = 2 y dx = x dy + y dx = d( x y)

y =0 x

can take =

y . x d( u x y) = 0

( 50) ( 51 ) ( 52) y . x ( 53)

( 54)

( 55) ( 56)

( 57)

( 58) ( 59) ( 60)

d( ( u y) y) = 0.

( 61 )

( 62) ( 63) ( 64)

which easily lead to = Thus Now the Cauchy data implies thus As a consequence u( x , y) = x

y , x

u+1 . x

( 67) ( 68) ( 69) ( 70) x2 1. y ( 71 )

u=xf

y 1. x

x f( x) 1 = u x , x 2 = x 2 f( x) = x + x 1 . y y +x x x
1

1 = y+

5. Solving quasi-linear rst-order equations. Example 1 5. ( 2. 8, 3( f) ) Find the general solution of ( y + u) u x + y u y = x y. dy du dx = = . y x y y+u

( 72)

Solution. The characteristic equations are

( 73) ( 74) ( 75)

From this we have Thus we can take On the other hand, we have

d( y + u) dx = y+u x

d x 2 ( y + u) 2 = 0.

= x 2 + ( y + u) 2 . d( x y) du = x y u

d u 2 ( x y)

= 0.

( 76)

As a consequence, the solution is given by

F x 2 ( y + u) 2 , u 2 ( x y) 2 = 0. Remark 1 6. Note that, in the above we can also use d y d( x + u) = x+u y Thus the formula may not be unique. Example 1 7. ( 2. 8, 5( h) ) solve u u x u u y = u 2 + ( x + y)
2

( 77)

x+u y

= const.

( 78)

with u = 1 on y = 0.

( 79)

Solution. The characteristic equation are

We have Then we have

dy du dx = = . u u 2 + ( x + y) 2 u

( 80) ( 81 ) ( 82)

dx dy = u u

d( x + y) = 0

= x + y.

dy du = u u2 + 2

= e2 y u2 + 2 .

Now from the Cauchy data we have Therefore eectively F has to be and the solution satises which leads to F , 1 + 2 0. F( , ) = 2 + 1 e 2 y u 2 + ( x + y) 2 1 + ( x + y) 2 = 0 u= 1 + ( x + y)
2

( 83) ( 84) ( 85) . ( 86)

e 2 y ( x + y)

2 1 /2

6. Equations with more than two variables. The method of characteristics can be applied to higher dimensional problems with no diculty in principle it indeed becomes more dicult in practice! Example 1 8. ( 2. 8, 8( d) ) Solve the following equation y z u x x z u y + x y x 2 + y 2 u z = 0. dy dz du dx = = = . 0 y z x z x y ( x2 + y2 ) This time we need three invariants, let s denote them by , and . Clearly one can take = u. For the second invariant, we observe

( 87)

Solution. The characteristic equations are

( 88)

( 89)

The last invariant can be obtained through

dx dy = yz xz

dx dy = y x

d x2 + y2 = 0

= x2 + y2 .

( 90)

Therefore the solutions are obtained by setting F u , x2 + y2 , z 2 + y2 x2 + y2 which gives u = f x 2 + y2 , z 2 + y2 x2 + y2 for arbitrary f. 1 ( 93) =0 ( 92)

1 . K eep in mind that this f will be determined once some C auchy data is given.

dy dz = xz xy

dy dz = z y

d z 2 + y2 = 0

= z 2 + y 2 .

( 91 )