M412 Practice Problems for Final Exam

1. Solve the PDE u t + t3 u x = u u(t, 0) = t, t>0 x > 0. u(0, x) = 1 ex , 2. Solve the PDE utt = c2 uxx ; x > 0, t > 0 u(0, x) = f (x); x > 0 ut (0, x) = g (x); ux (t, 0) = t; 3. Solve the PDE uxx + uyy = 0 u(x, 0) = 0, u(x, 2) = 0 u(0, y ) = 0, u(1, y ) = 2. 4. Solve the PDE ut =uxx + et sin 3x u(t, 0) =0, u(t, 1) = 0 u(0, x) = sin x. 5. For the PDE in Problem 4, nd an equilbrium solution and show that it matches the limit as t of your solution to Problem 4. 6. For the PDE ut = uxx + t sin x ux (t, 0) = 1 ux (t, ) = 0 u(0, x) = cos x, nd the total energy
0

x>0

t > 0.

u(t, x)dx. 7. Use separation of variables to show that solutions to the quarter-plane problem ut = uxx ; ux (t, 0) = 0 |u(t, +)| bounded u(0, x) = f (x) 1 t > 0, x > 0

can be written in the form u(t, x) =

C ( )e t cos xd,

for some appropriate constant C ( ). 8. Use the method of Fourier tranforms to solve the rst order equation u t = ux u(0, x) = f (x). 9. [This question appeared on Exam 3.] Use Fouriers Theorem to prove that if a function f (x) is piecewise smooth on an interval [0, L], then the Fourier cosine series for f (x) converges for all x (0, L) to (i) : f (x) if f is continuous at the point x 1 (ii) : (f (x ) + f (x+ )) if f has a jump discontinuity at the point x 2 What does the Fourier cosine series converge to at the endpoints x = 0 and x = L? 10. We have seen in the homework that if a function f (x) is piecewise smooth on an interval [0, L], then the Fourier sine series for f (x) converges for all x (0, L) to (i) : f (x) if f is continuous at the point x 1 (ii) : (f (x ) + f (x+ )) if f has a jump discontinuity at the point x. 2 Use this and Problem 9 to prove that if f (x) is continuous on [0, L] and f (x) is piecewise smooth on the same interval, then the Fourier cosine series for f (x) can be dierentiated term by term.

Solutions
1. For x
t4 4,

we have t4 dx = t3 ; x(0) = x0 x(t) = + x0 dt 4 du = u; u(0) = 1 ex0 u(t) = (1 ex0 )et , dt

from which we conclude For x

u(t, x) = (1 e(x 4 ) )et . we have t4 t4 dx = t3 ; x(t0 ) = 0 x(t) = 0, dt 4 4 du tt0 = u; u(t0 ) = t0 u(t) = t0 e , dt from which we conclude u(t, x) = (t4 4x)1/4 et(t 2
4

t4

4x)1/4

Combining these, u(t, x) = 2. We write solutions in the form (1 e

(t4 4x)1/4 et(t

4x)1/4

x x

)et ,

t4 4 t4 4.

u(t, x) = F (x ct) + G(x + ct), where for x > 0, we have 1 F (x) = f (x) 2 1 G(x) = f (x) + 2 1 2c 1 2c
x

g (y )dy
0 x

g (y )dy.
0

This entirely determines the solution for x ct > 0. For x ct < 0, we need to evaluate F at negative numbers. In order to do this, we notice that our nal condition gives t = F (ct) + G (ct). Setting x = ct, we nd F (x) = We compute, now,
x x

x G (x). c

F (y )dy =
0 0

y x2 G (y )dy F (x) F (0) = + G(x) G(0). c 2c

Its clear from our expressions for F and G that (assuming our solution is continuous) F (0) = G(0), from which we conclude x2 F (x) = + G(x). 2c In this we, for x ct < 0, (x ct)2 + G(ct x). F (x ct) = 2c We have, then u(t, x) =
x+ct 1 1 2 [f (x ct) + f (x + ct)] + 2c xct g (y )dy, 2 x+ct ct) 1 +1 g (y )dy (x 2c 2 [f (ct x) + f (x + ct)] + 2c 0 1 2c ctx 0

x ct > 0 g (y )dy, x ct < 0.

3. Since we have a bounded domain, we proceed by separation of variables, letting u(x, y ) = X (x)Y (y ), for which we nd uxx + uyy = 0 X (x)Y (y ) + X (x)Y (y ) = 0 Y (y ) X (x) = = . X (x) Y (y )

Observe here in particular that we have chosen the sign in front of so that the variable with both boundary conditions 0 (Y in this case) will have the standard eigenvalue equation, Y + Y = 0. We have, u(x, 0) = 0 Y (0) = 0, u(x, 2) = 0 Y (2) = 0, and u(0, y ) = 0 X (0) = 0. We have, then, the two ODE Y + Y =0; X X =0; Y (0) = 0, Y (2) = 0 X (0) = 0. 3

 For the Y (y ) equation, we take Y (y ) = C1 cos y + C2 sin y , and use the boundary conditions to conclude Yn (y ) = sin For X (x), we have n y, 2 n = 1, 2, 3....

n n x + C4 sinh x, 2 2 for which our boundary condition X (0) = 0 determines C3 = 0, eliminating one constant of integration. We nally have our general expansion for u(x, y ), X (x) = C3 cosh

u(x, y ) =
n=1

An sinh

n n x sin y. 2 2

Finally, we employ our last boundary condition, u(1, y ) = 2 to obtain the Fourier sine series

2=
n=1

An sinh

n n sin y. 2 2

We have, then An sinh 2 n = 2 2

2 sin
0

4 n n ydy = cos y 2 n 2
2 2

2 0

4 [(1)n 1], n
2 H.

where I have explicitly written the fraction

as a reminder that it comes from

Our solution is

u(x, y ) =

4 n [(1)n 1] n nx sin y. sinh n sinh 2 2 2 n=1

4. Due to the non-homogeneous term, we must proceed here by eigenfunction expansion. First, we construct eigenfunctions, Xn (x), for the homogeneous problem. Substituting u(t, x) = T (t)X (x) into ut = uxx , and considering our boundary conditions, we determine X + X = 0; X (0) = 0, X (1) = 0,

for which we have Xn (x) = sin nx. We now look for a solution as an expansion of these eigenfunctions

u(t, x) =
n=1

cn (t) sin nx.

Substituting this expansion back into the full non-homogeneous equation, we nd

cn (t) + n2 2 cn (t) sin nx = et sin 3x.

The key observation we make here is that this is simply a Fourier sine series with fancy constants, Bn = cn (t) n2 2 cn (t). Consequently, we have cn (t) + n2 2 cn (t) =2
1 0

et sin(3x) sin(nx)dx =

e t , n = 3 0, n = 3.

For initial conditions, we take our initial data

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

cn (0) sin nx,

for which cn (0) = 2

sin(x) sin(nx)dx =

1, n = 1 0, n = 1.

We have now an ODE to solve for each n = 1, 2, 3..., but we observe that if both cn (t) + n2 2 cn (t) and cn (0) are 0, the cn (t) 0. In this case, the only two expansion coecients that are not identically 0 are c1 (t) and c3 (t). For c1 (t), we have 2 c1 + 2 c1 = 0; c1 (0) = 1 c1 (t) = e t . For c3 (t), we have c3 + 9 2 c3 = e t ; c3 (0) = 0,

which we solve by the integrating factor method. (Recall that for a general linear rst order equation y (t) + p(t)y (t) = g (t), the integrating factor is e p(t)dt , where the constant of integration can be dropped.) 2 In this case, the integrating factor is simply e9 t , and we have (e9 t c3 ) = e9 t et e9 t c3 (t) = According to our intial condition c3 (0) = 0, we have C= We conclude that c3 (t) = with then u(t, x) = e t sin(x) + 5. Our equilibrium equation for u (x) is u xx = 0 u (0) = 0 u (1) = 0, which is solved by u (x) 0. Taking a limit as t of our solution to Problem 4, we see that they agree. 6. Integrating the full equation, we have
0 0
2 2 2 2

9 2

2 1 et(19 ) + C. 1

1 . 1 9 2

2 1 (e9 t et ), 2 1 9 2 1 (e9 t et ) sin(3x). 1 9 2

ut dx =

uxx dx +

t sin xdx d dt

d dt

u(t, x)dx = ux (t, ) ux (t, 0) t cos x .

It follows that

u(t, x)dx = 1 + 2t.

Integrating,
0

u(t, x)dx = t + t2 + C.

In order to nd C , we use u(0, x) = cos x to compute

cos xdx = C C = 0.
0

We conclude

u(t, x)dx = t + t2 .

7. Separate variables with u(t, x) = T (t)X (x), and set X T = = , T X from which we have the eigenvalue problem X + X = 0 X (0) = 0 X (+) bounded. In this case, all 0 are eigenvalues, with associated eigenfunctions X (x) = cos x. Since the eigenvalues are continuous, we integrate rather than summing, obtaining a general solution of the form A()et cos xd. u(t, x) = Finally, set = to get u(t, x) =
0 0

A( 2 )e t cos x2d.

The stated result follows from the choice C ( ) = 2A( 2 ). 8. Taking the Fourier transform of this equation, we have u t = i u ( )eit . u (t, ) = f Inverting, we compute
+

u(t, x) =

( )eit d = eix f

( )d, ei(x+t) f

, evaluated at x + t. That is, where this last expression is the inverse transform of f u(t, x) = f (x + t). 6

9. Since f (x) is only dened on the interval [0, L], we are free to extend it in any way we like to the full interval [L, L], where Fouriers theorem is valid. We extend it as an even function, so that the extension fE (x) is dened by f (x), 0xL fE (x) = f (x), L x 0. If f (x) is piecewise smooth on [0, L], then fE (x) is piecewise smooth on [L, L], and Fouriers Theorem states that fE denitely has a convergent Fourier series,

fE (x) = A0 +
n=1

An cos

nx nx + Bn sin . L L

We now compute A0 , An , and Bn , keeping in mind that fE (x) is an even function. We have A0 = An = 1 2L 1 L
+L L +L L +L L

fE (x)dx = fE (x) cos

1 L

L 0

fE (x)dx
L 0

2 nx dx = L L

fE (x) cos

nx dx L

1 Bn = L

nx dx = 0. fE (x) sin L

In this way, we see that the series for fE (x) is a Fourier cosine series that converges on [L, L]. If it converges on [L, L], it must converge on [0, L], and since f (x) and fE (x) agree there, it converges to f (x). Last, since fE (x) is an even extension, we have
x0 xL

lim fE (x) = lim+ fE (x)

lim fE (x) =

lim fE (x),

so that the Fourier cosine series of f (x) converges at x = 0 to

lim f (x), lim f (x).

and at x = L to
xL

10. First, under these assumptions, f (x) has a convergent Fourier cosine series (by Problem 9),

f (x) = A0 +
n=1

An cos

nx . L

Moreover, f (x) has a convergent sine series

f (x) =
n=1

Bn sin

nx , L
L 0 L

with Bn = 2 L
L

f (x) sin
0

2 nx nx dx = f (x) sin L L L

n L

f (x) cos
0

nx dx L

n An , L which gives precisely the series that arises by dierentiating the Fourier cosine series of f (x) term by term. = 7