Академический Документы
Профессиональный Документы
Культура Документы
Problem 1 (50 pts.) Solve the heat equation in a rectangle 0 < x < , 0 < y < , 2u 2u u = + t x2 y 2 subject to the initial condition u(x, y, 0) = (sin 2x + sin 3x) sin y and the boundary conditions u(0, y, t) = u(, y, t) = 0, u(x, 0, t) = u(x, , t) = 0.
Conversely, if functions , h, and G are solutions of the above ODEs for the same values of and , then u(x, y, t) = (x)h(y )G(t) is a solution of the heat equation. Substituting u(x, y, t) = (x)h(y )G(t) into the boundary conditions, we get (0)h(y )G(t) = ( )h(y )G(t) = 0, (x)h(0)G(t) = (x)h( )G(t) = 0.
It is no loss to assume that neither nor h nor G is identically zero. Then the boundary conditions are satised if and only if (0) = ( ) = 0, h(0) = h( ) = 0. To determine , we have an eigenvalue problem = , (0) = ( ) = 0.
This problem has eigenvalues n = n2 , n = 1, 2, . . . . The corresponding eigenfunctions are n (x) = sin nx.
Hence the eigenvalues are m = m2 , m = 1, 2, . . . . The corresponding eigenfunctions are hm (y ) = sin my . The function G is to be determined from the equation G = ( + )G. The general solution of this equation is G(t) = c0 e(+)t , where c0 is a constant. Thus we obtain the following solutions of the heat equation satisfying the boundary conditions: unm (x, y, t) = e(n +m )t n (x)hm (y ) = e(n
2 +m2 )t
n, m = 1, 2, 3, . . .
u(x, y, t) =
n=1 m=1
cnm e(n
2 +m2 )t
where cnm are constants. To determine the coecients cnm , we substitute the series into the initial condition u(x, y, 0) = (sin 2x + sin 3x) sin y :
It is easy to observe that c2,1 = c3,1 = 1 while the other coecients are equal to 0. Therefore u(x, y, t) = e5t sin 2x sin y + e10t sin 3x sin y.
Problem 2 (50 pts.) Solve Laplaces equation inside a quarter-circle 0 < r < 1, 0 < < /2 (in polar coordinates r, ) subject to the boundary conditions u(r, 0) = 0, Solution: u(r, ) = u(r, /2) = 0,
n=1
|u(0, )| < ,
u(1, ) = f ().
cn =
f () sin 2n d,
0
n = 1, 2, . . .
Laplaces equation in polar coordinates (r, ): 2 u 1 u 1 2u + + = 0. r2 r r r2 2 We search for the solution of the boundary value problem as a superposition of solutions u(r, ) = h(r)() with separated variables of Laplaces equation that satisfy the three homogeneous boundary conditions. Substituting u(r, ) = h(r)() into Laplaces equation, we obtain h (r)() + 1 1 h (r)() + 2 h(r) () = 0, r r
Since the left-hand side does not depend on while the right-hand side does not depend on r, it follows that r2 h (r) + rh (r) () = = , h(r) () where is a constant. Then r2 h (r) + rh (r) = h(r), = .
Conversely, if functions h and are solutions of the above ODEs for the same value of , then u(r, ) = h(r)() is a solution of Laplaces equation in polar coordinates. Substituting u(r, ) = h(r)() into the homogeneous boundary conditions, we get h(r)(0) = 0, h(r)(/2) = 0, |h(0)()| < .
It is no loss to assume that neither h nor is identically zero. Then the boundary conditions are satised if and only if (0) = (/2) = 0, |h(0)| < . To determine , we have an eigenvalue problem = , (0) = (/2) = 0.
This problem has eigenvalues n = (2n)2 , n = 1, 2, . . . . The corresponding eigenfunctions are n () = sin 2n. The function h is to be determined from the equation r2 h + rh = h and the boundary condition |h(0)| < . We may assume that is one of the above eigenvalues so that > 0. Then the general solution of the equation is h(r) = c1 r + c2 r , where = and c1 , c2 are constants. The boundary condition |h(0)| < holds if c2 = 0. Thus we obtain the following solutions of Laplaces equation satisfying the three homogeneous boundary conditions: un (r, ) = r2n sin 2n, n = 1, 2, . . . A superposition of these solutions is a series
u(r, ) =
n=1
where c1 , c2 , . . . are constants. Substituting the series into the boundary condition u(1, ) = f (), we get cn sin 2n. f () =
n=1
The right-hand side is a Fourier sine series on the interval [0, /2]. Therefore the boundary condition is satised if this is the Fourier sine series of the function f () on [0, /2]. Hence cn = 4
/2
f () sin 2n d,
0
n = 1, 2, . . .
Bonus Problem 3 (40 pts.) Consider a regular Sturm-Liouville eigenvalue problem + = 0, (0) = 0, (1) + h(1) = 0,
In the case = 0, the general solution of the equation + = 0 is a linear function (x) = c1 x + c2 , where c1 , c2 are constants. Substituting it into the boundary conditions (0) = 0 and (1) + h(1) = 0, we obtain equalities c1 = 0, c1 + h(c1 + c2 ) = 0. They imply that c1 = hc2 = 0. If h = 0, it follows that c1 = c2 = 0, hence there are no eigenfunctions with eigenvalue = 0. If h = 0 then (x) = 1 is indeed an eigenfunction.
(ii) For what values of h are all eigenvalues positive? Solution: h > 0.
In the case < 0, the general solution of the equation + = 0 is (x) = c1 cosh x + c2 sinh x, where = > 0 and c1 , c2 are constants. Note that (x) = c1 sinh x + c2 cosh x. The boundary condition (0) = 0 is satised if and only if c2 = 0. Substituting (x) = c1 cosh x into the boundary condition (1) + h(1) = 0, we obtain c1 sinh + hc1 cosh = 0, c1 ( tanh + h) = 0. If tanh = h, it follows that c1 = 0, hence there are no eigenfunctions with eigenvalue = 2 . If tanh = h then (x) = cosh x is indeed an eigenfunction. The function f () = tanh is continuous. It is easy to see that f (0) = 0 and f () > 0 for > 0. Since tanh 1 as +, we have that f () + as +. It follows that f takes all positive values on (0, ). By the above the eigenvalue problem has a negative eigenvalue if and only if h < 0. As shown in the solution to the part (i), = 0 is an eigenvalue only for h = 0. Hence all eigenvalues are positive if and only if h > 0. The fact that for any h 0 all eigenvalues are nonnegative can also be obtained using the Rayleigh quotient. If is an eigenfunction corresponding to an eigenvalue then
1 0 1
+
0 1
| (x)|2 dx
2
|(x)| dx
(iii) How many negative eigenvalues can this problem have? Solution: One negative eigenvalue for h < 0.
Let f () = tanh . As shown in the solution to the part (ii), < 0 is an eigenvalue if and only if f () = h, where = 2 , > 0. Observe that f () = tanh + tanh = tanh + . cosh2
In particular, f () > 0 for > 0. Since f is continuous, f (0) = 0, and f () + as +, it follows that f is a one-to-one map of the interval (0, ) onto itself. Therefore for any h < 0 the eigenvalue problem has exactly one negative eigenvalue.
Positive eigenvalues are found from the equation tan = h/ . The function f1 () = tan is continuous, strictly increasing and assumes all real values on each of the intervals (m /2, m + /2), m = 0, 1, 2, . . . . In the case h > 0, the function f2 () = h/ is continuous and strictly decreasing on (0, ). It follows that the equation f1 () = f2 () has exactly one solution in each of the intervals (0, /2) and (m /2, m + /2), m = 1, 2, . . . . In this case all eigenvalues are positive, hence (n 1) /2 < n < (n 1) + /2. Moreover, since tan n 0 as n , it follows that n (n 1) . 2 If h = 0 then n = (n 1) , n = 1, 2, . . . . If h < 0 then 1 < 0 < 2 . In this case the function f2 () = h/ is negative and strictly increasing on (0, ). The equation f1 () = f2 () has no solution in (0, /2) and exactly one solution in each of the intervals (m /2, m + /2), m = 1, 2, . . . . We conclude that (n 1) /2 < n < (n 1)+ /2 for n 2. It follows that n (n 1) in this case as well.