18.303 Problem Set 6 Solutions
Problem 1: (10 points)
We just plug u + s into F, using the facts that Âu = f and that  is self-adjoint, collecting the terms that have 0 or
1 or 2 factors of s, to obtain:
F {u + s} = hu + s, Â(u + s)i − hf, u + si − hu + s, f i
=
0
F {u}
+ hs, Âu − f i + hÂu − f, si + hs, Âsi
| {z }
|
{z
}
| {z }
factors of s 1 factor of s, using Â=Â∗ 2 factors of
= F {u} +
s
0 + 0 +hs, Âsi
| {z }
using Âu=f
which is > F {u} for s 6= 0 by the positive-definiteness of  (or ≥ F {u} if  is semidefinite).
Problem 2: (10 points)
We already
´ showed in class that c∇ · κ∇u is self-adjoint for appropriate boundary conditions under the inner product
hu, vi = Ω ūv/c, and hence (as shown in the notes) the (real) Green’s function obeys a reciprocity relation G(x0 , x) =
c(x0 )
0
c(x) G(x, x ).
If x is at end 2 of the rod and x0 is at end 1, G(x, x0 ) gives the temperature rise at end 2 from a source
at end 1, which Sal claims is 6= 0. Reciprocity tells us, however, that G(x0 , x) is then also 6= 0, so a heat source at
end 2 must give a temperature rise at end 1, contrary to Sal’s claims, for any distribution c(x) > 0 and κ(x) > 0 of
materials.
Problem 3: (5+5+10+10+5 points)
Recall that the displacement u(x, t) of a stretched string [with fixed ends: u(0, t) = u(L, t) = 0] satisfies the wave
2
2
equation ∂∂xu2 + f (x, t) = ∂∂t2u , where f (x, t) is an external force density (pressure) on the string.
2
2
(a) Suppose that ũ solves ∂∂xũ2 + f˜(x, t) = ∂∂t2ũ and satisfies ũ(0, t) = ũ(L, t) = 0. Now, consider u = Re ũ =
Clearly, u(0, t) = u(L, t) = 0, so u satisfies the same boundary conditions. It also satisfies the PDE:
"
#
∂2u
1 ∂ 2 ũ ∂ 2 ũ
=
+ 2
∂t2
2 ∂t2
∂t
"
#
1 ∂ 2 ũ
∂ 2 ũ
˜
˜
+f +
+f
=
2 ∂x2
∂x2
=
¯
ũ+ũ
2 .
∂2u
+ f,
∂x2
˜ ¯
˜
f
˜
since f = f +
2 = Re f . The key factors that allowed us to do this are (i) linearity, and (ii) the real-ness of the
PDE (the PDE itself contains no i factors or other complex coefficients).
(b) Plugging u(x, t) = v(x)e−iωt and f (x, t) = g(x)e−iωt into the PDE, we obtain
∂ 2 v −iωt
+ g
= −ω 2 v
,
e e−iωt
e−iωt
∂x2
and hence
and  = −
∂2
− 2 − ω2 v = g
∂x
∂2
− ω 2 . The boundary conditions are v(0) = v(L) = 0, from the boundary conditions on u.
∂x2
Since ω 2 is real, this is in the general Sturm-Liouville form that we showed in class is self-adjoint.
1
Subtracting a constant from an operator just shifts all of the eigenvalues by that constant, keeping the eigenfunctions the same. Thus  is still positive-definite if ω 2 is < the smallest eigenvalue of −∂ 2 /∂x2 , and positive
semidefinite if ω 2 = the smallest eigenvalue. In this case, we know analytically that the eigenvalues of −∂ 2 /∂x2
with these boundary conditions are (nπ/L)2 for n = 1, 2, . . .. So  is positive-definite if ω 2 < (π/L)2 , it is
positive-semidefinite if ω 2 = (π/L)2 , and it is indefinite otherwise.
(c) We know that ÂG(x, x0 ) = 0 for x 6= x0 . Also, just as in class and as in the notes, the fact that ÂG(x, x0 ) =
δ(x − x0 ) meas that G must be continuous (otherwise there would be a δ 0 factor) and ∂G/∂x must have a jump
discontinuity:
∂G ∂G −
= −1
∂x x=x0+
∂x x=x0−
for −∂ 2 G/∂x2 to give δ(x − x0 ). We could also show this more explicitly by integrating:
ˆ
ˆ
x0 +0+
x0 +0+
δ(x − x0 ) dx = 1
ÂG dx =
x0 −0+
x0 −0+
x0 +0+ ˆ x0 +0+
∂G 2
−
=−
ω G dx,
∂x x0 −0+
0 −0+
x
which gives the same result.
Now, similar to class, we will solve it separately for x < x0 and for x > x0 , and then impose the continuity
requirements at x = x0 to find the unknown coefficients.
2
For x < x0 , ÂG = 0 means that ∂∂xG2 = −ω 2 G, hence G(x, x0 ) is some sine or cosine of ωx. But since G(0, x0 ) = 0,
we must therefore have G(x, x0 ) = α sin(ωx) for some coefficient α.
Similarly, for x > x0 , we also have a sine or cosine of ωx. To get G(L, x0 ) = 0, the simplest way to do this
is to use a sine with a phase shift: G(x, x0 ) = β sin(ω[L − x]) for some coefficient β.
Continuity now gives two equations in the two unknowns α and β:
α sin(ωx0 ) = β sin(ω[L − x0 ])
αω cos(ωx0 ) = −βω cos(ω[L − x0 ]) + 1,
0
0
sin(ω[L−x ])
sin(ωx )
1
which has the solution α = ω1 cos(ωx0 ) sin(ω[L−x
0 ])+sin(ωx0 ) cos(ω[L−x0 ]) , β = ω cos(ωx0 ) sin(ω[L−x0 ])+sin(ωx0 ) cos(ω[L−x0 ]) .
This simplifies a bit from the identity sin A cos B + cos B sin A = sin(A + B), and hence
1
G(x, x ) =
ω sin(ωL)
0
(
sin(ωx) sin(ω[L − x0 ]) x < x0
,
sin(ωx0 ) sin(ω[L − x]) x ≥ x0
which obviously obeys reciprocity.
Note that if ω is an eigenfrequency, i.e. ω = nπ/L for some n, then this G blows up. The reason is that
 in that case is singular (the n-th eigenvalue was shifted to zero), and defining Â−1 is more problematical.
(Physically, this corresponds to driving the oscillating string at a resonance frequency, which generally leads to
a diverging solution unless there is dissipation in the system.)
2
d
T
(d) It is critical to get the signs right. Recall that we approximated dx
2 by −D D for a 1st-derivative matrix D.
T
2
Therefore, you want to make a matrix A = D D − ω I. In Matlab, A = D’*D - omega^2 * eye(N) where N
is the size of the matrix D = diff1(N) / dx with dx = 1/(N+1). We make dk by dk = zeros(N,1); dk(k) =
1/dx. I used N = 100.
The resulting plots are shown in figure 1, for ω = 0.4π/L (left) and ω = 5.4π/L (right) and x0 = 0.5 and
0.75. As expected, for ω < π/L where it is positive-definite, the Green’s function is positive and qualitatively
similar to that for ω = 0, except that it is slightly curved. For larger ω, though, it becomes oscillatory and much
more interesting in shape. The exact G and the finite-difference G match very well, as they should, although the
2
ω = 0.4π/L
ω = 5.4π/L
0.35
0.06
exact x’=0.5
exact x’=0.75
N=100 x’=0.5
N=100 x’=0.75
0.3
exact x’=0.5
exact x’=0.75
N=100 x’=0.5
N=100 x’=0.75
0.04
0.25
0.2
G(x,x’)
G(x,x’)
0.02
0.15
0
−0.02
0.1
−0.04
0.05
0
0
0.2
0.4
0.6
0.8
−0.06
1
0
0.2
0.4
0.6
x
0.8
1
x
2
∂
2
Figure 1: Exact (lines) and finite-difference (dots) Green’s functions G(x, x0 ) for  = − ∂x
2 − ω , for two different ω
0
values (left and right) and two different x values (red and blue).
match becomes worse at higher frequencies—the difference approximations become less and less accurate as the
function becomes more and more oscillatory relative to the grid ∆x, just as was the case for the eigenfunctions
as discussed in class.
[If you try to plug in an eigenfrequency (e.g. 4π/L) for ω, you may notice that the exact G blows up but
the finite-difference G is finite. The reason for this is simple: the eigenvalues of the finite-difference matrix DT D
are not exactly (nπ/L)2 , so the matrix is not exactly singular, nor is it realistically possible to make it exactly
singular thanks to rounding errors. If you plug in something very close to an eigenfrequency, then G becomes
very close to an eigenfunction multipled by a large amplitude. It is easy to see why this is if you look at the
expansion of the solution in terms of the eigenfunctions.]
(e) For small ω, sin(ωy) ≈ ωy for any y, and hence
(
2
ω
x(L − x0 ) x < x0
1
G(x, x ) → 2
,
2
ωx0 (L − x) x ≥ x0
ωL 0
2
d
which matches the Green’s function of − dx
2 from class. (Equivalently, you get 0/0 as ω → 0 and hence must
use L’Hôpital’s rule or similar to take the limit.)
3