Anda di halaman 1dari 23

3

Laplaces Equation
u = 0

We now turn to studying Laplaces equation and its inhomogeneous version, Poissons equation, u = f. We say a function u satisfying Laplaces equation is a harmonic function.

3.1

The Fundamental Solution


u = 0 x Rn .

Consider Laplaces equation in Rn , Clearly, there are a lot of functions u which satisfy this equation. In particular, any constant function is harmonic. In addition, any function of the form u(x) = a1 x1 + . . . + an xn for constants ai is also a solution. Of course, we can list a number of others. Here, however, we are interested in nding a particular solution of Laplaces equation which will allow us to solve Poissons equation. Given the symmetric nature of Laplaces equation, we look for a radial solution. That is, we look for a harmonic function u on Rn such that u(x) = v (|x|). In addition, to being a natural choice due to the symmetry of Laplaces equation, radial solutions are natural to look for because they reduce a PDE to an ODE, which is generally easier to solve. Therefore, we look for a radial solution. If u(x) = v (|x|), then xi uxi = v (|x|) |x | = 0 , |x | which implies x2 x2 1 i uxi xi = v (|x|) 3 v (|x|) + i 2 v (|x|) |x | = 0 . |x | |x | |x | Therefore, n1 u = v (|x|) + v (|x|). |x | Letting r = |x|, we see that u(x) = v (|x|) is a radial solution of Laplaces equation implies v satises n1 v (r) + v (r) = 0. r Therefore, 1n v = v r 1n v = = v r = ln v = (1 n) ln r + C C = v (r) = n1 , r 1

which implies v (r) = c1 ln r + c2 c1 + c2 (2n)rn2 c1 ln |x| + c2 c1 + c2 (2n)|x|n2 n=2 n 3.

From these calculations, we see that for any constants c1 , c2 , the function u(x) n=2 n 3. (3.1)

for x Rn , |x| = 0 is a solution of Laplaces equation in Rn {0}. We notice that the function u dened in (3.1) satises u(x) = 0 for x = 0, but at x = 0, u(0) is undened. We claim that we can choose constants c1 and c2 appropriately so that x u = 0 in the sense of distributions. Recall that 0 is the distribution which is dened as follows. For all D, (0 , ) = (0). Below, we will prove this claim. For now, though, assume we can prove this. That is, assume we can nd constants c1 , c2 such that u dened in (3.1) satises x u = 0 . Let denote the solution of (3.2). Then, dene v (x) = (3.2)

4n

(x y )f (y ) dy.

Formally, we compute the Laplacian of v as follows, x v = = =

4n 4n

x (x y )f (y ) dy y (x y )f (y ) dy

4n

x f (y ) dy = f (x).

That is, v is a solution of Poissons equation! Of course, this set of equalities above is entirely formal. We have not proven anything yet. However, we have motivated a solution formula for Poissons equation from a solution to (3.2). We now return to using the radial solution (3.1) to nd a solution of (3.2). Dene the function as follows. For |x| = 0, let (x) = 21 ln |x|
1 1 n(n2)(n) |x|n2

n=2 n 3,

(3.3)

where (n) is the volume of the unit ball in Rn . We see that satises Laplaces equation on Rn {0}. As we will show in the following claim, satises x = 0 . For this reason, we call the fundamental solution of Laplaces equation. 2

Claim 1. For dened in (3.3), satises x = 0 in the sense of distributions. That is, for all g D,

4n

(x)x g (x) dx = g (0).

Proof. Let F be the distribution associated with the fundamental solution . That is, let F : D R be dened such that (F , g ) =

4n

(x)g (x) dx

for all g D. Recall that the derivative of a distribution F is dened as the distribution G such that (G, g ) = (F, g ) for all g D. Therefore, the distributional Laplacian of is dened as the distribution F such that (F , g ) = (F , g ) for all g D. We will show that (F , g ) = (0 , g ) = g (0), and, therefore, (F , g ) = g (0), which means x = 0 in the sense of distributions. By denition, (F , g ) =

4n

(x)g (x) dx.

Now, we would like to apply the divergence theorem, but has a singularity at x = 0. We get around this, by breaking up the integral into two pieces: one piece consisting of the ball of radius about the origin, B (0, ) and the other piece consisting of the complement of this ball in Rn . Therefore, we have (F , g ) = =
B (0, )

4n

(x)g (x) dx (x)g (x) dx +

4n B(0,)

(x)g (x) dx

= I + J.

We look rst at term I . For n = 2, term I is bounded as follows,


B (0, )

1 ln |x|g (x) dx C |g |L 2
2

ln |x| dx
B (0, )

C
0 0

ln |r|r dr d ln |r|r dr
0

C ln | | 2 . For n 3, term I is bounded as follows, 1 1 g (x) dx C |g |L n(n 2)(n) |x|n2

B (0, )

B (0, )

1 dx |x|n2 dr

C
0 B (0,r)

1 dS (y ) |y |n 2 dS (y ) dr

=
0

1 rn2 1 rn2
B (0,r)

=
0

n(n)rn1 dr

= n(n)
0 +

r dr =

n(n) 2 . 2

Therefore, as 0 , |I | 0. Next, we look at term J . Applying the divergence theorem, we have

4n B(0,)

(x)x g (x) dx =

4n B(0,)
+
(

x (x)g (x) dx
(

4n B(0,))

(x)

g dS (x)

4n B(0,))

g (x) dS (x)

4n B(0,))

g (x) dS (x) +

4n B(0,))

(x)

g dS (x)

J 1 + J 2. using the fact that x (x) = 0 for x Rn B (0, ). We rst look at term J 1. Now, by assumption, g D, and, therefore, g vanishes at . Consequently, we only need to calculate the integral over B (0, ) where the normal derivative is the outer normal to Rn B (0, ). By a straightforward calculation, we see that x . x (x) = n(n)|x|n The outer unit normal to Rn B (0, ) on B (0, ) is given by = 4 x . |x |

Therefore, the normal derivative of on B (0, ) is given by = x n(n)|x|n x |x | = 1 . n(n)|x|n1

Therefore, J 1 can be written as


B (0, )

1 1 g (x) dS (x) = n 1 n(n)|x| n(n) n1

g (x) dS (x) =
B (0, ) B (0, )

g (x) dS (x).

Now if g is a continuous function, then g (x) dS (x) g (0) as 0.

Lastly, we look at term J 2. Now using the fact that g vanishes as |x| +, we only need to integrate over B (0, ). Using the fact that g D, and, therefore, innitely dierentiable, we have (x)
B (0, )

g g dS (x) C

|(x)| dS (x)
L (B (0, )) B (0, )

|(x)| dS (x).
B (0, )

Now rst, for n = 2, |(x)| dS (x) = C


B (0, ) B (0, )

| ln |x|| dS (x) dS (x)


B (0, )

C | ln | ||

= C | ln | ||(2 ) C | ln | ||. Next, for n 3, |(x)| dS (x) = C


B (0, ) B (0, )

1 dS (x) |x|n2 dS (x)

C n2 C n2
B (0, )

n(n) n1 C.

Therefore, we conclude that term J 2 is bounded in absolute value by C | ln | C Therefore, |J 2| 0 as 0+ . 5 n=2 n 3.

Combining these estimates, we see that

4n

(x)x g (x) dx = lim I + J 1 + J 2 = g (0). +


0

Therefore, our claim is proved. Solving Poissons Equation. We now return to solving Poissons equation u = f x Rn .

From our discussion before the above claim, we expect the function v (x)

4n

(x y )f (y ) dy

to give us a solution of Poissons equation. We now prove that this is in fact true. First, we make a remark. Remark. If we hope that the function v dened above solves Poissons equation, we must rst verify that this integral actually converges. If we assume f has compact support on some bounded set K in Rn , then we see that

4n

(x y )f (y ) dy |f |L
K

|(x y )| dy.

If we additionally assume that f is bounded, then |f |L C . It is left as an exercise to verify that |(x y )| dy < +
K

on any compact set K . Theorem 2. Assume f C 2 (Rn ) and has compact support. Let u(x)

4n

(x y )f (y ) dy

where is the fundamental solution of Laplaces equation (3.3). Then 1. u C 2 (Rn ) 2. u = f in Rn . Ref: Evans, p. 23. Proof. 1. By a change of variables, we write u(x) =

4n

(x y )f (y ) dy =

4n

(y )f (x y ) dy.

Let ei = (. . . , 0, 1, 0, . . .) be the unit vector in Rn with a 1 in the ith slot. Then u(x + hei ) u(x) f (x + hei y ) f (x y ) = (y ) dy. h h 4n Now f C 2 implies f (x + hei y ) f (x y ) f (x y ) as h 0 h xi uniformly on Rn . Therefore, f u ( x) = (y ) (x y ) dy. xi xi 4n Similarly, 2u 2f ( x) = (y ) (x y ) dy. xi xj xi xj 4n This function is continuous because the right-hand side is continuous. 2. By the above calculations and Claim 1, we see that x u(x) = =

4n 4n

(y )x f (x y ) dy (y )y f (x y ) dy

= f (x).

3.2
3.2.1

Properties of Harmonic Functions


Mean Value Property

In this section, we prove a mean value property which all harmonic functions satisfy. First, we give some denitions. Let B (x, r) = ball of radius r about x in Rn B (x, r) = boundary of ball of radius r about x in Rn (n) = volume of unit ball in Rn n(n) = surface area of unit ball in Rn . For a function u dened on B (x, r), the average of u on B (x, r) is given by
B (x,r )

u(y ) dy =

1 (n)rn 7

u(y ) dy.
B (x,r)

For a function u dened on B (x, r), the average of u on B (x, r) is given by


B (x,r)

u(y ) dS (y ) =

1 n(n)rn1

u(y ) dS (y ).
B (x,r )

Theorem 3. (Mean-Value Formulas) Let Rn . If u C 2 () is harmonic, then u(x) =


B (x,r)

u(y ) dS (y ) =
B (x,r )

u(y ) dy

for every ball B (x, r) . Proof. Assume u C 2 () is harmonic. For r > 0, dene (r) =
B (x,r )

u(y ) dS (y ).

For r = 0, dene (r) = u(x). Notice that if u is a smooth function, then limr0+ (r) = u(x), and, therefore, is a continuous function. Therefore, if we can show that (r) = 0, then we can conclude that is a constant function, and, therefore, u(x) =
B (x,r)

u(y ) dS (y ).

We prove (r) = 0 as follows. First, making a change of variables, we have (r) =


B (x,r)

u(y ) dS (y ) u(x + rz ) dS (z ).

=
B (0,1)

Therefore, (r) =
B (0,1)

u(x + rz ) z dS (z ) u(y ) yx dS (y ) r

=
B (x,r)

u (y ) dS (y ) B (x,r) 1 u = (y ) dS (y ) n(n)rn1 B (x,r) 1 = (u) dy (by the Divergence Theorem) n(n)rn1 B (x,r) 1 = u(y ) dy = 0, n(n)rn1 B (x,r) = 8

using the fact that u is harmonic. Therefore, we have proven the rst part of the theorem. It remains to prove that u(x) =
B (x,r)

u(y ) dy.

We do so as follows, using the rst result,


r

u(y ) dy =
B (x,r) 0 r B (x,s)

u(y ) dS (y ) n(n)sn1
B (x,s) r

ds u(y ) dS (y ) ds

=
0

=
0

n(n)sn1 u(x) ds
r 0

= n(n)u(x)

sn1 ds

=r = (n)u(x) sn |s s=0 = (n)u(x)rn .

Therefore, u(y ) dy = (n)rn u(x),


B (x,r)

which implies u(x) = as claimed.

1 (n)rn

u(y ) dy =
B (x,r) B (x,r)

u(y ) dy,

3.2.2

Converse to Mean Value Property

In this section, we prove that if a smooth function u satises the mean value property described above, then u must be harmonic. Theorem 4. If u C 2 () satises u(x) =
B (x,r )

u(y ) dS (y )

for all B (x, r) , then u is harmonic. Proof. Let (r) =


B (x,r )

u(y ) dS (y ).

If u(x) =
B (x,r )

u(y ) dS (y )

for all B (x, r) , then (r) = 0. As described in the previous theorem, (r) = r u(y ) dy. n B (x,r)

Suppose u is not harmonic. Then there exists some ball B (x, r) such that u > 0 or u < 0. Without loss of generality, we assume there is some ball B (x, r) such that u > 0. Therefore, r (r ) = u(y ) dy > 0, n B (x,r) which contradicts the fact that (r) = 0. Therefore, u must be harmonic.

3.2.3

Maximum Principle

In this section, we prove that if u is a harmonic function on a bounded domain in Rn , then u attains its maximum value on the boundary of . Theorem 5. Suppose Rn is open and bounded. Suppose u C 2 () C () is harmonic. Then 1. (Maximum principle) max u(x) = max u(x).

2. (Strong maximum principle) If is connected and there exists a point x0 such that u(x0 ) = max u(x),

then u is constant within . Proof. We prove the second assertion. The rst follows from the second. Suppose there exists a point x0 in such that u(x0 ) = M = max u(x).

Then for 0 < r < dist(x0 , ), the mean value property says M = u(x0 ) =
B (x0 ,r)

u(y ) dy M.

But, therefore,
B (x0 ,r)

u(y ) dy = M,

and M = max u(x). Therefore, u(y ) M for y B (x0 , r). To prove u M throughout , you continue with this argument, lling with balls.

10

Remark. By replacing u by u above, we can prove the Minimum Principle. Next, we use the maximum principle to prove uniqueness of solutions to Poissons equation on bounded domains in Rn . Theorem 6. (Uniqueness) There exists at most one solution u C 2 () C () of the boundary-value problem, u = f x u=g x . Proof. Suppose there are two solutions u and v . Let w = u v and let w = v u. Then w and w satisfy w = 0 x w=0 x . Therefore, using the maximum principle, we conclude max |u v | = max |u v | = 0.

3.2.4

Smoothness of Harmonic Functions

In this section, we prove that harmonic functions are C . Theorem 7. Let be an open, bounded subset of Rn . If u C () and u satises the mean value property, u(x) =
B (x,r )

u(y ) dS (y )

for every ball B (x, r) , then u C (). Remarks. 1. As proven earlier, if u C 2 () C () and u is harmonic, then u satises the mean value property, and, therefore, u C (). 2. In fact, if u satises the hypothesis of the above theorem, then u is analytic, but we will not prove that here. (See Evans.) Proof. First, we introduce the function such that (x) Ce |x|2 1 0
1

|x | < 1 |x | 1

where the constant C is chosen such that 4n (x) dx = 1. Notice that C (Rn ) and has compact support. Now dene the function (x) such that (x) 1 n x .

11

Therefore, C (Rn ) and supp( ) {x : |x| < }. Further,

4n
Now choose such that

(x) dx = 1.

< dist(x, ). Dene u (x) =

(x y )u(y ) dy.

Now we claim 1. u C 2. u (x) = u(x). First, for (1), u C because C . We prove (2) as follows. Using the fact that supp (x y ) {y : |x y | < }. Therefore, u ( x) =
B (x, )

(x y )u(y ) dy 1
n B (x, )

= = = = = =

|x y |

u(y ) dy u(y ) dS (y ) dr dr

1
n 0 B (x,r )

|x y | r

1
n 0 B (x,r )

u(y ) dS (y ) u(y ) dS (y ) dr

1
n

r
B (x,r)

1
n 0

r r
0

n(n)rn1
B (x,r)

u(y ) dS (y ) dr

1
n

n(n)rn1 u(x) dr r
B (0,r)

= u(x) = u(x) = u(x)

1
n

dS (y ) dr |y | dy

1
n B (0, )

(y ) dy
B (0, )

= u(x).

12

3.2.5

Liouvilles Theorem

In this section, we show that the only functions which are bounded and harmonic on Rn are constant functions. Theorem 8. Suppose u : Rn R is harmonic and bounded. Then u is constant. Proof. Let x0 Rn . By the mean value property, u( x 0 ) =
B (x0 ,r)

u(y ) dy

for all B (x0 , r). Now by the previous theorem, we know that if u C 2 () C () and u is harmonic, then u is C . Therefore, u = 0 = uxi = 0 for i = 1, . . . , n. Therefore, uxi is harmonic and satises the mean value property. Therefore, uxi (x0 ) = = uxi (y ) dy uxi (y ) dy ui dS (y ),
B (x0 ,r)

B (x0 ,r)

1 (n)rn 1 = (n)rn

B (x0 ,r)

by the Divergence theorem, where = (1 , . . . , n ) is the outward unit normal to B (x0 , r). Therefore, |uxi (x0 )| 1 (n)rn ui dS (y )
B (x0 ,r)

|u|L (B (x0 ,r)) |i |L |u|L (4n ) Therefore, |uxi (x0 )| n(n)rn1 (n)rn

1 (n)rn

dS (y )
B (x0 ,r)

n |u|L (4n ) . r

n |u|L (4n ) r n C , r

by the assumption that u is bounded. Now this is true for all r. Taking the limit as r +, we see that |uxi (x0 )| = 0. Therefore, uxi (x0 ) = 0. This is true for i = 1, . . . , n and for all x0 Rn . Therefore, we conclude that u constant.

13

As a corollary of Liouvilles Theorem, we have the following representation formula for all bounded solutions of Poissons equation on Rn , n 3. Theorem 9. (Representation Formula) Let f C 2 (Rn ) with compact support. Let n 3. Then every bounded solution of u = f has the form u(x) = x Rn (3.4)

4n

(x y )f (y ) dy + C

for some constant C , where (x) is the fundamental solution of Laplaces equation in Rn . Proof. Recall that the fundamental solution of Laplaces equation in Rn , n 3 is given by (x) = where K = 1/n(n 2)(n). As shown earlier, u(x) = K |x|n2

4n

(x y )f (y ) dy > 0. Then,

is a solution of (3.4). Here we show this is a bounded solution for n 3. Fix we have |u(x)| =

1 f (y ) dy 4n |x y |n2 1 1 f (y ) dy + K f (y ) dy = K n 2 B (x, ) |x y | 4n B(x, ) |x y |n2 1 dy + C |f (y )| dy. |f (y )|L n2 4n B(x, ) B (x, ) |x y | = K It is easy to see that the rst term on the right-hand side is bounded. The second term on the right-hand side is bounded, using the assumption that f C 2 (Rn ) with compact support. Therefore, we conclude that u(x) =

4n

(x y )f (y ) dy

4n

(x y )f (y ) dy

is a bounded solution of (3.4). Now suppose there is another bounded solution of (3.4). Let u be such a solution. Let w(x) = u(x) u(x).

14

Then w is a bounded, harmonic function on Rn . Then, by Liouvilles Theorem, w must be constant. Therefore, we conclude that u(x) = u(x) + C = as claimed.

4n

(x y )f (y ) dy + C,

3.3
3.3.1

Solving Laplaces Equation on Bounded Domains


Laplaces Equation on a Rectangle

In this section, we will solve Laplaces equation on a rectangle in R2 . First, we consider the case of Dirichlet boundary conditions. That is, we consider the following boundary value problem. Let = {(x, y ) R2 : 0 < x < a, 0 < y < b}. We want to look for a solution of the following, (x, y ) uxx + uyy = 0 u(0, y ) = g1 (y ), u(a, y ) = g2 (y ) 0<y<b (3.5) u(x, 0) = g3 (x), u(x, b) = g4 (y ) 0 < x < a. In order to do so, we consider the following simpler example. From this, we will show how to solve the more general problem above. Example 10. Let = {(x, y ) R2 : 0 < x < a, 0 < y < b}. Consider (x, y ) uxx + uyy = 0 u(0, y ) = g1 (y ), u(a, y ) = 0 0<y<b u(x, 0) = 0, u(x, b) = 0 0 < x < a. We use separation of variables. We look for a solution of the form u(x, y ) = X (x)Y (y ). Plugging this into our equation, we get X Y + XY = 0. Now dividing by XY , we arrive at X Y + = 0, X Y which implies Y X = = Y X

(3.6)

15

for some constant . By our boundary conditions, we want Y (0) = 0 = Y (b). Therefore, we begin by solving the eigenvalue problem, Y = Y 0<y<b Y (0) = 0 = Y (b). As we know, the solutions of this eigenvalue problem are given by Yn (y ) = sin We now turn to solving n n y , n = b b
2

n 2 X b with the boundary condition X (a) = 0. The solutions of this ODE are given by X = Xn (x) = An cosh n n x + Bn sinh x . b b

Now the boundary condition X (a) = 0 implies An cosh Therefore, un (x, y ) = Xn (x)Yn (y ) = An cosh where An , Bn satisfy the condition An cosh n n a + Bn sinh a = 0. b b n n x + Bn sinh x b b sin n y b n n a + Bn sinh a = 0. b b

is a solution of Laplaces equation on which satises the boundary conditions u(x, 0) = 0, u(x, b) = 0, and u(a, y ) = 0. As we know, Laplaces equation is linear. Therefore, we can take any combination of solutions {un } and get a solution of Laplaces equation which satises these three boundary conditions. Therefore, we look for a solution of the form

u(x, y ) =
n=1

un (x, y ) =
n=1

An cosh

n n x + Bn sinh x b b

sin

n y b

where An , Bn satisfy

n n a + Bn sinh a = 0. (3.7) b b To solve our boundary-value problem (3.6), it remains to nd coecients An , Bn which not only satisfy (3.7), but also satisfy the condition u(0, y ) = g1 (y ). That is, we need An cosh

u(0, y ) =
n=1

An sin 16

n y = g1 (y ). b

That is, we want to be able to express g1 in terms of its Fourier sine series on the interval [0, b]. Assuming g1 is a nice function, we can do this. From our earlier discussion of Fourier series, we know that the Fourier sine series of a function g1 is given by

g1 (y )
n=1

An sin

n y b

where the coecients An are given by An = g1 , sin n y b n sin b y , sin


n y b

where the L2 -inner product is taken over the interval [0, b]. Therefore, to summarize, we have found a solution of (3.6) given by

u(x, y ) =
n=1

un (x, y ) =
n=1

An cosh

n n x + Bn sinh x b b

sin

n y b

where An = and Bn = coth g1 , sin n y b n sin b y , sin


n y b

n a An . b

Now we return to considering (3.5). For the general boundary value problem on a rectangle with Dirichlet boundary conditions, we can nd a solution by nding four separate solutions ui for i = 1, . . . , 4 such that each ui is identically zero on three of the sides and satises the boundary condition on the fourth side. For example, for the boundary value problem (3.5), we use the procedure in the above example to nd a function u1 (x, y ) which is harmonic on and such that u1 (0, y ) = g1 (y ) and u1 (a, y ) = 0 for 0 < y < b, and u1 (x, 0) = 0 = u1 (x, b) for 0 < x < a. Similar we nd functions u2 , u3 and u4 which vanish on three of the sides but satisfy the fourth boundary condition. We now consider an example where we have a mixed boundary condition on one side. Example 11. Let = {(x, y ) R2 , 0 < x < L, 0 < y < H }. Consider the following boundary value problem, (x, y ) uxx + uyy = 0 u(0, y ) = 0, u(L, y ) = 0 0<y<H (3.8) u(x, 0) uy (x, 0) = 0, u(x, H ) = f (x) 0 < x < L. Using separation of variables, we have Y X = = . X Y 17

We rst look to solve

X = X X (0) = 0 = X (L).

0<x<L

As we know, the solutions of this eigenvalue problem are given by Xn (x) = sin Now we need to solve n n x , n = L L
2

n 2 Y L with the boundary condition Y (0) Y (0) = 0. The solutions of this ODE are given by Y = Yn (y ) = An cosh n n y + Bn sinh y . L L

The boundary condition Y (0) Y (0) = 0 implies An Bn Therefore, n = 0. L

n n n cosh y + Bn sinh y . L L L Therefore, we look for a solution of (3.8) of the form Y n ( y ) = Bn

u(x, y ) =
n=1

Bn sin

n x L

n n n cosh y + sinh y L L L

Substituting in the condition u(x, H ) = f (x), we have

u(x, H ) =
n=1

Bn sin

n x L

n n n cosh H + sinh H L L L

= f (x).

Recall the Fourier sine series of f on [0, L] is given by

f
n=1

An sin

n x L

where

f, sin n x L An = n sin L x , sin n x L

where the L2 -inner product is taken over (0, L). Therefore, in order for our boundary condition u(x, H ) = f (x) to be satised, we need Bn to satisfy Bn n n n cosh H + sinh H L L L 18 f, sin n x L = . sin n x , sin n x L L

Using the fact that sin n n x , sin x L L


L

=
0

sin2

L n x dx = , L 2

the solution of (3.8) is given by

u(x, y ) =
n=1

Bn sin

n x L

n n n cosh y + sinh y L L L

where Bn = 2 n n n cosh H + sinh H L L L L


1 0 L

f (x) sin

n x dx. L

3.3.2

Laplaces Equation on a Disk

In this section, we consider Laplaces Equation on a disk in R2 . That is, let = {(x, y ) R2 : x2 + y 2 < a2 }. Consider uxx + uyy = 0 u = h() (x, y ) (x, y ) . (3.9)

To solve, we write this equation in polar coordinates as follows. To transform our equation in to polar coordinates, we will write the operators x and y in polar coordinates. We will use the fact that x2 + y 2 = r 2 y = tan . x Consider a function u such that u = u(r, ), where r = r(x, y ) and = (x, y ). That is, u = u(r(x, y ), (x, y )). Then u(r(x, y ), (x, y )) = ur rx + u x x x y = ur 2 u 2 2 2 1 / 2 (x + y ) x sec sin = ur cos u . r Therefore, the operator
x

can be written in polar coordinates as sin = cos . x r r 19

Similarly, the operator

can be written in polar coordinates as cos = sin + . y r r

Now squaring these operators we have 2 sin = cos 2 x r r 2 sin cos sin cos 2 sin2 sin2 2 = cos2 2 + 2 2 + + . r r2 r r r r r2 2 Similarly, 2 cos = sin + 2 y r r 2 sin cos sin cos 2 cos2 cos2 2 = sin2 2 2 + 2 + + . r r2 r r r r r2 2
2 2 Combining the above terms, we can write the operator x + y in polar coordinates as follows, 2 2 1 1 2 2 + = + + . x2 y 2 r2 r r r2 2 Therefore, in polar coordinates, Laplaces equation is written as 2 2

1 1 urr + ur + 2 u = 0. r r

(3.10)

Now we will solve it using separation of variables. In particular, we look for a solution of the form u(r, ) = R(r)(). Then letting u(x, y ) = u(r(x, y ), (x, y )), we will arrive at a solution of Laplaces equation on the disk. Substituting a function of the form u(r, ) = R(r)(), into (3.10), our equation is written as 1 1 R + R + 2 R = 0. r r Dividing by R, R R + + 2 = 0. R rR r 2 Multiplying by r , we are led to the equations r2 R rR = = R R for some scalar . The boundary condition for this problem is u = h() for (x, y ) . Therefore, we are led to the following eigenvalue problem with periodic boundary conditions, = (0) = (2 ), (0) = (2 ). 20 0 < < 2

Recall from our earlier work that periodic boundary conditions imply our eigenfunctions and eigenvalues are n () = An cos(n) + Bn sin(n), For each n , we need to solve n = n2 n = 0, 1, 2, . . .

r2 Rn + rRn = n Rn . r2 Rn + rRn n2 Rn = 0

That is, we need to solve the second-order ODE,

for n = 0, 1, 2, . . .. Recall that a second-order ODE will have two linearly independent solutions. We look for a solution of the form R(r) = r for some . Doing so, our ODE becomes (2 n2 )r = 0. Therefore, for n 1, we have found two linearly independent solutions, Rn (r) = rn and Rn (r) = rn . Now for n = 0, we have only found one linearly independent solution so far, R0 (r) = 1. We look for another linearly independent solution. If n = 0, our equation can be written as r2 R + rR = 0. Dividing by r, our equation becomes rR + R = 0. A linearly independent solution of this equation is R0 (r) = ln r. Therefore, for each n 0, we have found a solution of (3.10) of the form C rn + Dn [A cos(n) + B sin(n)] n n n rn un (r, ) = Rn (r)n () = A0 [C0 + D0 ln r] . But, we dont want a solution which blows up as r 0+ . Therefore, we reject the solutions 1 and ln r. Therefore, we consider a solution of (3.10) of the form rn

u(r, ) =
n=0

rn [An cos(n) + Bn sin(n)] .

Now in order to solve (3.9), we need u(a, ) = h(). That is, we need

an [An cos(n) + Bn sin(n)] = h().


n=0

Using the fact that our eigenfunctions are orthogonal on [0, 2 ], we can solve for our coefcients An and Bn as follows. Multiplying the above equation by cos(n) and integrating over [0, 2 ], we have An = 1 h(), cos(n) 1 = an cos(n), cos(n) an
2

h() cos(n) d
0

for n = 1, 2, . . .

21

A0 =

h(), 1 1 = 1, 1 2

h() d.
0

Similarly, multiplying by sin(n) and integrating over [0, 2 ], we have 1 h(), sin(n) 1 Bn = n = n a sin(n), sin(n) a
2

h() sin(n) d.
0

To summarize, we have found a solution of Laplaces equation on the disk in polar coordinates, given by

u(r, ) =
n=0

rn [An cos(n) + Bn sin(n)]

where 1 A0 = 2 1 An = n a 1 Bn = n a
2

h() d
0 2

h() cos(n) d
0 2

h() sin(n) d.
0

Now we will rewrite this solution in terms of a single integral by substituting An and Bn into the series solution above. Doing so, we have 1 u(r, ) = 2
2

h() d
0

+
n=1

rn
2

1 an

h() cos(n) d cos(n) +


0

1 an

h() sin(n) d) sin(n)


0

1 = 2 = Now 1 2

h() 1 + 2
0 2 n=1

rn [cos(n) cos(n) + sin(n) sin(n)] an rn cos(n( )) an d.

h() 1 + 2
0 n=1

1+2
n=1

rn cos(n( )) an

1+2
n=1

rn ein() + ein() an 2 rei() a


n

=1+ =1+

+
n=1 i()

re re + i ( ) a re a rei() a2 r 2 = 2 . a 2ar cos( ) + r2 22

n=1 i()

rei() a

Therefore, u(r, ) = 1 2
0

h()

a2 r2 d. a2 2ar cos( ) + r2

We can write this in rectangular coordinates as follows. Let x be a point in the disk with polar coordinates (r, ). Let x be a point on the boundary of the disk with polar coordinates (a, ). Therefore, |x x |2 = a2 + r2 2ar cos( ) by the law of cosines. Therefore, 1 u(x )(a2 |x |2 ) ds u( x ) = , 2 |x |=a |x x |2 a using the fact that ds = a d is the arc length of the curve. Rewriting this, we have u(x) = a2 |x |2 2a u(x ) ds. |x x |2

|x |=a

This is known as Poissons formula for the solution of Laplaces equation on the disk.

23