Projects

Project 3.7
Road Mirages

(Solutions were provided by the project author.)

1. If n(y) is increasing then

dn d2 y
>0 and so > 0.
dy dx2
This means the graph of y(x) is concave up.
2. (a) The general solution is y(x) = c1 x + c2 which is the equation of a straight line.
(b) The concavity of a straight line is zero.
(c) This solution is physically reasonable since light will travel in a straight line in a uniform medium.
3. (a) Now
1 dn 1
=
n dy a
so the differential equation becomes

 2 
d2 y dy 1
= 1+ .
dx2 dx a

(b) The new differential equation is

du 1 + u2
=
dx a
which is separable. Separating variables we get
du dx
2
= .
1+u a
After integrating we get
x + c1 dy x + c1
tan−1 u = so = tan .
a dx a
(c) This equation has the solution  
x + c1
y = −a ln cos + c2
a
where c1 and c2 are arbitrary constants. Choosing new constants x0 and y0 the general solution can be
written  
x − x0
y − y0 = −a ln cos .
a
The graph of y(x) is concave up and has a vertex at the point (x0 , y0 ).

944
 Project 3.7 Road Mirages

(d) Here x − x0 = −50 and y0 = 1.19 so y = 1.315 meters. Now find dy/dx:
 
dy x − x0
= tan = 5 × 10−3
dx a
so the distance to the shiny spot is 1.315/5 × 10−3 = 263 meters.
4. (a) Now
1 dn 1
=
n dy 2y
so the differential equation becomes

 2 
d2 y dy 1
= 1+
dx2 dx 2y

which does not have the independent variable x in it. Consequently we make the substitution
dy d2 y du
=u and 2
=u .
dx dx dy

(b) The new differential equation becomes

du 1 + u2
u =
dy 2y
which is also separable. Separating variables we get
u du dy
=
1 + u2 2y
whose solution is ln(1 + u2 ) = ln y + ln c = ln cy for arbitrary c > 0.
(c) Exponentiating both sides we get
dy 
u2 = cy − 1 or ± = cy − 1
dx
for cy − 1 > 0 or c > 1 since y ≥ 1. This is also separable: we get
dy
±√ = dx.
cy − 1
After integrating we have
2
± cy − 1 = x − d
c
which simplifies to
1 c
y= + (x − d)2
c 4
where c and d are arbitrary constants. Let 1/c = y0 and d = x0 . We then get
(x − x0 )2
y − y0 =
4y0
which is the equation of a parabola opening upward with vertex at (x0 , y0 ) for y0 ≥ 1.
(d) Here x − x0 = −3 and y0 = 1.19 so y = 3.08 meters. Now find dy/dx:
dy x − x0 3
= =− = −1.261
dx 2y0 2.38
so the distance to the shiny spot is 3.08/1.261 = 2.44 meters.

945
Project 3.10
Project 3.7 Road Mirages Pendulum
The Ballistic

Project 3.10
The Ballistic Pendulum

(Solutions were provided by the project author.)

1. The auxiliary equation is m2 + g/
= 0, so the general solution of the differential equation is
 
g g
θ(t) = c1 cos t + c2 sin t.




The initial condtion θ(0) = 0 implies c1 = 0 and θ
(0) = ω0 implies c2 = ω0
/g . Thus,
 

g
θ(t) = ω0 sin t.
g


2. At θmax , sin g/
t = 1, so
 

mb vb
mb v
θmax = ω0 = = √b
g mw + mb
g mw + mb
g
and
mw + m b 
vb =
g θmax .
mb
3. We have cos θmax = (
− h)/
= 1 − h/
. Then
1 2 h
cos θmax ≈ 1 − θ =1−
2 max

and 
2 2h 2h
θmax = or θmax = .



Thus 
mw + m b  2h m w + mb 
vb =
g = 2gh .
mb
mb
4. When mb = 5 g, mw = 1 kg, and h = 6 cm, we have
1005 
vb = 2(980)(6) ≈ 21, 797 cm/s.
5

946
 Project 8.1 Two-Ports in Electrical Circuits

Project 8.1
Two-Ports in Electrical Circuits

(Solutions were provided by the project author.)

1. Since the terminals are connected directly, V1 = V2 . Voltage drop across the resistance is V1 . The curent through
the resistance is I1 − I2 . Thus V1 = (I1 − I2 )R. We have two equations, V2 = V1 + 0I1 and I2 = −V1 /R + I1 .
In matrix form     
V2 1 0 V1
= .
I2 −1/R 1 I1
 
1 0
The transmission matrix is .
−1/R 1

2. Current through resistance R1 is I1 − I2 . Drop in voltage across R1 is V1 . Ohm’s law gives V1 =

(I1 − I2 )R1 , I2 = I1 − V1 /R1 . Current through resistance R2 is I2 . Drop in voltage across R2 is V1 − V2 . Thus
V1 − V2 = I2 R2 , V2 = V1 − I2 R2 = V1 − (I1 − V1 R1 )R2 = V1 (1 + R2 /R1 ) − I1 . Combine V2 = V1 (1 + R2 /R1 ) − I1
and I2 = I1 − V1 /R1 .
    
V1 1 + R2 /R1 −1 V1
= .
I2 −1/R1 1 I1
 
1 + R2 /R1 −1
The transmission matrix is .
−1/R 1

3. Current through resistance R1 is I1 . Drop in voltage across R1 is V1 − V2 . Thus V1 − V2 = I1 R1 so

V2 = V1 −I1 R1 . Current through resistance R2 is I1 −I2 . Drop in voltage across R2 is V2 . Thus V2 = (I1 −I2 )R2
and I2 = I1 − V2 /R2 = I1 − (V1 − I1 R1 )/R2 = −V1 /R2 + I1 (1 + R1 /R2 ). Combine V2 = V1 − I1 R1 and
I2 = −V1 /R2 + I1 (1 + R1 /R2 ):
    
V1 1 −R1 V1
= .
I2 −1/R2 1 + R1 /R2 I1
 
1 −R1
The transmission matrix is .
−1/R2 1 + R1 /R2

4. (a) The transmission matrix is

     
1 −1 1 0 3 −1 1 −1
= .
0 1 1 1 −1 1 2 0
    
1 −1 3 1
(b) Outputs = ; V = 1 volt, I = 6 amps.
2 0 2 6

947
Project 8.2
Project 8.1 Traffic
Two-Ports
Flowin Electrical Circuits

Project 8.2
Traffic Flow

1. Using conservation of flow at each intersection we have

A: x1 + x4 = 300 B: x1 + x2 = 250
C: x2 + x3 = 100 D: x3 + x4 = 150.
The corresponding system of equations is described by the augmented matrices
   
1 0 0 1 300 1 0 0 1 300
1 1 0 0 250   0 1 0 −1 −50 
  row  
  −−−−−−→  .
0 1 1 0 100  operations  0 0 1 1 150 
0 0 1 1 150 0 0 0 0 0
Thus
x1 = −x4 + 300
x2 = x4 − 50
x3 = −x4 + 150.
We see that x4 must be at least 50 (so that x2 is nonnegative) and at most 150 (so that x3 is nonnegative).
Taking x4 = 100 we have the solution x1 = 200, x2 = 50, x3 = 50, and x4 = 0. Taking x4 = 50 we have the
solution x1 = 250, x2 = 0, x3 = 100, and x4 = 50. The minimum flow along AB corresponds to the smallest
possible value of x1 , which occurs when x4 is the largest it can be, namely, 150. Thus, the minimum value of
x1 is −150 + 300 = 150.

2. Using conservation of flow at each intersection and setting the number of vehicles out of each intersection equal
to the number into that intersection we have
A: x1 = 100 + x4 B: 200 + x2 = x1
C: x3 = x2 + 150 D: 50 + x4 = x3 .
The corresponding system of equations is described by the augmented matrices
   
1 0 0 −1 100 1 0 0 −1 100
 1 −1 0 200   0 1 0 −1 −100 
 0  row  
  −−−−−−→  .
 0 −1 1 0 150  operations  0 0 1 −1 50 
0 0 1 −1 50 0 0 0 0 0
Thus
x1 = x4 + 100
x2 = x4 − 100
x3 = x4 + 50.
We see that x4 must be at least 100, in which event x2 will be 0. When x4 = 100 we have x1 = 200 and
x3 = 150.

3. Starting with the right, upper intersection we label the intersections counterclockwise A, B, . . . , H. Using
conservation of flow at each intersection and setting the number of vehicles out of each intersection equal to the

948
 Project 8.15 Temperature Dependence of Resistivity

number into that intersection we have

A: x1 = x8 + 100 B: x2 + 160 = x1
C: x3 = x2 + 80 D: x4 + 100 = x3
E: x5 = x4 + 130 F: x6 + 90 = x5
G: x7 = x6 + 120 H: x8 + 80 = x7 .
Since each xn , n = 1, . . . , 8 must be nonnegative, we see from A and B that x1 ≥ 160, from C and D that
x3 ≥ 100, from E and F that x5 ≥ 130, and from G and H that x7 ≥ 120. Now, if x1 = 160 then x2 = 0 from
B, so x3 = 80 from C. But x3 ≥ 100 so if x3 = 100 then x2 = 20 from C and x1 = 180. Thus, x1 is at least
180. When x1 = 180 we find that x2 = 20 (from B), x3 = 100 (from C), x4 = 0 (from D), x5 = 130 (from
E), x6 = 40 (from F), x7 = 160 (from G), and x8 = 80 (from H and A). Since these values are all valid, the
minimum possible flow along x1 is 180 vehicles.

4. (a) Using conservation of flow at each intersection and setting the number of vehicles out of each intersection
equal to the number into that intersection we have
A: x1 + x2 = 200 B: x3 + x4 = x1
D: 200 = x4 + x5 E: x5 = x2 + x3 .
(b) When k = 4,

kx1 + 2kx2 + kx3 + 2kx4 + kx5 = k(x1 + 2x2 + x3 + 2x4 + x5 ) = 4(x1 + 2x2 + x3 + 2x4 + x5 ).

Expressing the system of equations in part (a) in reduced echelon form we obtain
x1 + x2 = 200
x2 + x3 + x4 = 200
x4 + x5 = 200.
Adding these equations we see that x1 + 2x2 + x3 + 2x4 + x5 = 600. Thus, the total time for all 200 vehicles
to be in the network is 4 × 600 = 2400 minutes or 40 hours. Since there are 200 vehicles, this is an average
of 12 minutes per vehicle.

Project 8.15
Temperature Dependence of Resistivity

(Solutions were provided by the project author.)

   
5.6 0 0 1
 5.65   400 20 1
   
   
 5.7   3600 60 1
   
1. Y =  
 7.82  and the matrix A = 
 3.24e4 180 1.
   
 11.1   2.30e5 480 1
   
 20.2   4.62e5 680 1
30.5 9.60e5 980 1

949
Project 8.15 Temperature Dependence of Resistivity

 
2.2e − 5
 
2. The column vector X∗ =  0.00458 
5.57

3. Using the values in X∗ above, the resistivity at 300◦ C is

5.57 + 0.00458(300 − 20) + (2.2e − 5)(300 − 20)2 = 8.55 ohm-m.

4. The resistivity has increased by a factor 8.55/5.6 = 1.53, so the resistance becomes 5 × 1.53 = 7.63 ohms.

5. The RMS error is 0.90 ohm-m.

6. The RMS error of 0.90 ohm-m roughly means that the least squares quadratic differs from the data by about
1 ohm-m.

7. The predicted resistivity at 2000◦ C is

5.57 + 0.00458(2000 − 20) + (2.2e − 5)(2000 − 20)2 = 101 ohm-m.
Extrapolation outside the domain of the data is very unreliable.

Project 9.16
Solutions for Minimal Surfaces

(Solutions were provided by the project author.)


1 d
1. F
(t) =  ||∇u + t∇w||2 dA
2 1 + ||∇u + t∇w||2 dt
 R
1 d
=  [(∇u + t∇w) · (∇u + t∇w)] dA
R 2 1 + ||∇u + t∇w|| dt
2

1 d
=  [||∇u||2 + 2t∇w · ∇u + t2 ||∇w||2 ] dA
R 2 1 + ||∇u + t∇w|| dt
2

1
=  [2∇w · ∇u + 2t||∇w||2 ] dA
R 2 1 + ||∇u + t∇w||
2

∇u
F
(0) =  · ∇w dA
R 1 + ||∇u||2
  
2. (hF · n) ds = div (hF) dA = (h div F + (grad h) · F) dA
c R R

∇u
3. Applying the identity from Problem 2 with F =  and h = w yields
1 + ||∇u||2
   
   
∇u ∇u ∇u
w  · n ds = w div  + (∇w) ·  dA
c 1 + ||∇u||2 R 1 + ||∇u||2 1 + ||∇u||2
  
∇u
but w = 0 on C implies w  · n ds = 0 so that
c 1 + ||∇u||2
     
∇u ∇u
F
(0) = (∇w) ·  dA = − w div  dA = 0.
R 1 + ||∇u||2 R 1 + ||∇u||2

950
 Project 9.16 Solutions for Minimal Surfaces

4. Using the fact that ||∇u||2 = u2x + u2y and the Chain Rule we have
 
∂ ux (1 + u2x + u2y )uxx − (u2x uxx + ux uy uyx )
 =
∂x 1 + ||∇u||2 (1 + u2x + u2y )3/2

and  
∂ uy (1 + u2x + u2y )uyy − (uy ux uxy + u2y uyy )
 =
∂y 1 + ||∇u||2 (1 + u2x + u2y )3/2
so that  
∇u (1 + u2y )uxx + (1 + u2x )uyy − 2ux uy uxy
div  = .
1 + ||∇u||2 (1 + u2x + u2y )3/2

5. If u is a function of x only, then uy = uyy = 0 so the minimal surface equation reduces to uxx = 0. Antidiffer-
entiating once yields ux = f (y) where f is an arbitrary function. But since ux is a function of x only, ux = a
where a is a constant. Antidifferentiating again yields u(x, y) = ax + b where b is a constant. If u is a function
of y only then a similar argument yields u(x, y) = cy + d where c and d are constants.
 ∂r x ∂r y
6. If u = f (r) then, since r = x2 + y 2 , ux = f
(r) = f
(r) and uy = f
(r) = f
(r) . Then we have
∂x r ∂y r
 
x2 1 x2
uxx = f

(r) 2 + f
(r) − 3
r r r
2
 


y
1 y2
uyy = f (r) 2 + f (r) − 3
r r r
 
xy −xy
uxy = f

(r) 2 + f
(r) .
r r3
Substituting these results into the minimal surface equation yields
     
y2 x2 y2 x2 y2 x2
1 + [f
(r)]2 2 f

(r) 2 + f
(r) 3 + 1 + [f
(r)]2 2 f

(r) 2 + f
(r) 3
r r r r r r
xy  xy xy 
− 2[f
(r)]2 2 f

(r) 2 − f
(r) 3 = 0.
r r r
Algebraic rearrangement yields
  2 2    4 



x y x2 y 2 2x2 y 2
1
y x4 2x2 y 2
f (r) 1 + [f (r)]2
+ 4 − + f (r) + [f (r)]2
+ 5 + = 0,
r4 r r4 r r5 r r5
so  
1 [f
(r)]2
f

(r) + f
(r) + =0 and rf

(r) + f
(r)(1 + [f
(r)]2 ) = 0.
r r

dg
7. Letting g = f
(r) we have r + g(1 + g 2 ) = 0 and separating variables yields
dr
 
1 g 1
− dg = − dr.
g 1 + g2 r
Antidifferentiation yields
1
ln g − ln(1 + g 2 ) = − ln r + k, where k is a constant,
2
or  
g c
ln  = ln , where ln c = k.
1 + g2 r

951
Project 9.16 Solutions for Minimal Surfaces

Solving for g we obtain



c 1 1
f (r) = g(r) = √ = so u = f (r) =  dr.
r −c
2 2 r /c2 − 1
2 r2 /c2 −1
 
u−d
The substitution r = c cosh t results in u = ct + d where d is a constant, so that finally r = c cosh .
c
8. If u = f (θ) then, since tan θ = y/x, ux = f
(θ)(−y/r2 ) and uy = f
(θ)(x/r2 ). Then we have
 


y2
2xy
uxx = f (θ) 4 + f (θ)
r r4
 
x2 −2xy
uyy = f

(θ) 4 + f
(θ)
r r4
   2 


−xy
2y 1
uxy = f (θ) + f (θ) − 2 .
r4 r4 r
Substituting these results into the minimal surface equation yields
       

2x
2


y2
2xy
2y
2


x2
−2xy
1 + [f (θ)] 4 f (θ) 4 + f (θ) + 1 + [f (θ)] 4 f (θ) 4 + f (θ)
r r r4 r r r4
    2 
xy −xy y − x2
+ 2[f
(θ)]2 4 f

(θ) + f

(θ) = 0.
r r4 r4
1


Algebraic rearrangement yields f (θ) = 0, so f

(θ) = 0 and f (θ) = cθ + d, where c and d are constants.
r2

Project 14.3
The Hydrogen Atom

(Solutions were provided by the project author.)

1. The total mechanical energy for an electron of charge −e and mass m moving in a circular orbit of radius r
with velocity v is given by
mv 2 e
E= + (−e) ,
2 4π0 r
where we used the fact that the electric potential due to the proton is V = e/4π0 r. To obtain an expression
for the kinetic term mv 2 we use the force balance
mv 2 e2
= , (1)
r 4π0 r
where we used the formula for the Coulomb force between the proton and the electron. Inserting this into the
expression for the energy gives
e2 e2 e2
E= − =− . (2)
8π0 r 4π0 r 8π0 r
The classical angular momentum is given by L = mvr. We obtain from (1) that
  
e2 e2 me2 r
v= , which leads to L = m r= .
m4π0 r 4π0 r 4π0

952
 Project 14.3 The Hydrogen Atom

2. Letting

me2 r
nh̄ =
4π0
leads to the following expression for the orbital radius:

4π0 n2 h̄2
r= .
me2
Substituting this into (2) gives
me4
En = − . (3)
(4π0 )2 2h̄2 n2

3. The energy difference for a transition from the energy level Ek to the energy level En , k > n, is
 
me4 me4 me4 1 1
Ek − En = − + = − .
(4π0 )2 2h̄2 k 2 (4π0 )2 2h̄2 n2 (4π0 )2 2h̄2 n2 k2

Substituting this into Plank’s formula ∆E = h̄ν and using the relation λν = c gives
 
1 me4 1 1
= 3 2 − 2 . (4)
λ 8h̄ 0 c n2 k

me4
If we now set n = 2 in the formula above we obtain exactly the Balmer series with RH = .
8h̄3 20 c
A literature search provides the following physical constants:

m = 9.109534 · 10−31 kg
e = 1.6021892 · 10−19 C
h̄ = 6.626176 · 10−34 J · s
C2
0 = 8.854187918 · 10−12
N · m2
c = 2.99792458 · 10 m/s.
8

Inserting these values in the formula for RH gives

RH = 1.0973731 · 107 m−1

which is comparable to the experimental value 10967757.6 ± 1.2m−1 , but lying outside the error range. To
improve the accuracy we can replace the mass m by the reduced mass mM/(m + M ), where

M = 1.6726485 · 10−27

is the mass of the proton. With this correction, we obtain the value

RH = 10967757.78m−1 ,

which is now remarkable close to the best available empirical value.

4. Expressing the Laplacian in spherical coordinates, the Schrödinger equation with a Coulomb potential assumes
the form
 
h̄2 ∂ 2 Ψ(x) 2 ∂Ψ(x) 1 ∂ 2 Ψ(x) cot θ ∂Ψ(x) 1 ∂ 2 Ψ(x) e2
− 2
+ + 2 2
+ 2
+ 2 2
− Ψ(x) = EΨ(x).
2m ∂r r ∂r r ∂θ r ∂θ r sin θ ∂φ
2 4π0 r

953
Project 14.3 The Hydrogen Atom

If we now substitute the separated solution Ψ(x) = R(r)Θ(θ)Φ(φ) into this equation, divide each term by Ψ
and multiply by r2 we find
   
h̄2 r2 R

2 R
2me2 2mE 1 Φ

Θ
Θ


− + − − + + cot θ + = 0.
2m R r R 4π0 r3 h̄2 r2 sin2 θ Φ Θ Θ
Since the first term above is a function of r alone while the second depends only on θ and φ, they must each be
equal to a constant, which we call k. Therefore the radial equation is
 
h̄2 r2 R

2 R
2me2 2mE
− + − − 2 = k,
2m R r R 4π0 r h̄
which after some rearrangement of terms becomes
 2 


2
2m e 2m
R + R + 2 + E R = −k 2 2 . (5)
r h̄ 4π0 r h̄ r

5. With k = 0 and r → ∞, the radial equation becomes

2mE
R

+ R = 0.
h̄2
Recalling that E < 0, we see that the general solution to this equation is a linear combination of the functions
eCr and e−Cr where 
2mE
C= − 2 .
h̄
Since the solution cannot explode to infinity as r increases, we discard the term eCr .

6. If we insert a solution of the form R(r) = f (r)e−Cr into (5) (with k = 0) we find that f must satisfy
    2 


2
2C 2m e
f + − 2C f + C − 2
+ + E f = 0. (6)
r r h̄ 4π0 r
Using the expansion
f (r) = a0 + a1 r + a2 r2 + · · · ,
we see that
f
(r) = a1 + 2a2 r + 3a3 r2 + · · ·
and
f

(r) = 2a2 + 6a3 r + · · · .
Inserting these into (6) and collecting the coefficients of the terms proportional to 1/r, 1, r, r2 , . . . leads to the
recurrence relation
jC − B
aj = 2 aj−1 , j = 1, 2, . . . ,
j(j + 1)
where B = me2 /4π0 h̄2 (notice that the coefficient a0 remains undetermined).

7. We can rewrite the recurrence relation above as

C − B/j
aj = 2 aj−1 , j = 1, 2, . . . ,
j+1
which, for large values of j gives
2C 2C
aj = aj−1 ≈ aj−1 .
j+1 j
But this is the power series for the function e2Cr , which then implies that the function R(r) would behave like
e2Cr e−Cr = eCr . Since this is unbounded for large values of r we conclude that the power series for f (r) must

954
 Project 15.4 The Uncertainty Inequality in Signal Processing

terminate at a finite number of terms. It is then easy to see from the original recurrence relation that the only
way for this to happen is if B = nC for some integer n.

8. Since 
me2 2mE
B= and C = − ,
4π0 h̄2 h̄2
the condition B = nC implies that 
me2 2mE
=n −
4π0 h̄2 h̄2
or
m2 e4 2mE
= −n2 2 .
(4π0 )2 h̄4 h̄
We conclude that the allowed energies for the hydrogen atom in a state with zero angular momentum are
me4
En = − .
(4π0 )2 h̄2 n2

Project 15.4
The Uncertainty Inequality in Signal Processing

(Solutions were provided by the project author.)

 ∞  ∞  ∞
1. fˆ(−α) = f (x)e−iαx dx = f (x)eiαx dx = f (x)eiαx dx = fˆ(α)
−∞ −∞ −∞

2. Note that  ∞
fk (α) = f (x − k)eiαx dx
−∞
so letting u = x − k yields
 ∞  ∞  ∞
fk (α) = f (u)eiα(u+k) du = f (u)eiαu eiαk du = eiαk f (u)eiαu du = eiαk fˆ(α).
−∞ −∞ −∞

3. Note that  ∞
fc (α) = f (cx)eiαx dx
−∞
so letting u = cx yields
 ∞  ∞
1 1 1ˆ
fc (α) = f (u)e iα(u/c)
du = f (u)ei(α/c)u du = f (α/c).
c −∞ c −∞ c

4. Note that ∞ ∞
−∞
x2 [fc (x)]2 dx α2 |fc (α)|2 dα
[D(fc ) · B(fc )] = 2
∞ · −∞
∞
[f (x)]2 dx
−∞ c −∞
|fc (α)|2 dα
∞ ∞
−∞
x2 [f (cx)]2 dx −∞ α2 (1/c2 )|fˆ(α/c)|2 dα
= ∞ · ∞
−∞
[f (cx)]2 dx −∞
(1/c2 )|fˆ(α/c)|2 dα

955
Project 15.4 The Uncertainty Inequality in Signal Processing

and with the substitutions u = cx and β = α/c this becomes

∞ ∞ ∞ 2 ∞
(1/c3 ) −∞ u2 [f (u)]2 du c3 −∞ β 2 |fˆ(β)|2 dβ −∞
u [f (u)]2 du −∞ β 2 |fˆ(β)|2 dβ
∞ · ∞ = ∞ · ∞ .
(1/c) −∞ [f (u)]2 du c −∞ |fˆ(β)|2 dβ −∞
[f (u)]2 du −∞
|fˆ(β)|2 dβ

2
5. From Example 3 of Section 15.3 in the text it follows that for f (x) = e−|x| , fˆ(α) = . It is easy to check
1 + α2
that   
∞ ∞ ∞
[f (x)]2 dx = e−2|x| dx = 2 e−2x dx = 1.
−∞ −∞ 0

Integrating by parts (twice) shows that

 ∞  ∞
1
x2 [f (x)]2 dx = 2 x2 e−2x dx = .
−∞ 0 2
Finally the substitution α = tan θ, followed by a partial fraction decomposition, yields
 ∞  ∞  π/2
ˆ 2 1
[f (α)] dα = 4 2 2
dα = 4 cos2 θdθ = 4(π/2) = 2π.
−∞ −∞ (1 + α ) −π/2

The fact that  ∞

∞
1 
dα = arctan α  =π
−∞ 1 + α2 −∞

yields
   
∞ ∞
α2 ∞
1 1 π
α [fˆ(α)]2 dα = 4
2
dα = 4 − dα = 4 π − = 2π.
−∞ −∞ (1 + α)2 −∞ 1 + α2 (1 + α2 )2 2
So finally
∞ ∞
−∞
x2 [f (x)]2 dx α2 |fˆ(α)|2 dα (1/2) 2π 1
[D(f ) · B(f )] = 2
∞ · −∞
∞ = · = .
[f (x)]2 dx ˆ
|f (α)| dα
2 1 2π 2
−∞ −∞

6. (a) By Problem 4 above and (6) of Section 15.4 in the text, if {f (x)} = F (α) and {g(x)} = G(α) then
 ∞  ∞
−1 1
f (τ )g(x − τ )dτ = (F (α)G(α)) = F (α)G(α)e−iαx dα.
−∞ 2π −∞

Taking g(x) = f (−x), and noting that in this case G(α) = F (α), it follows that
 ∞  ∞
1
f (τ )f (τ − x)dτ = F (α)F (α)e−iαx dα.
−∞ 2π −∞

The result follows from taking x = 0 in this last equation.

(b) Define a real valued quadratic function by

 b
 
 
 
b b b
2
q(λ) = [λh1 (s) − h2 (s)] ds = [h1 (s)]2 ds λ2 − 2 h1 (s)h2 (s)ds λ + [h2 (s)]2 ds .
a a a a

Since q(λ) ≥ 0 for all λ, the discriminant of q(λ) must be less than or equal to 0. So
 b 2  b  b
4 h1 (s)h2 (s)ds − 4 2
[h1 (s)] ds [h2 (s)]2 ds ≤ 0
a a a

from which the inequality follows immediately. The case of equality occurs when the discriminant of q(λ) is
 b
2
0, which means that there is exactly one real root c of q(λ). For that c, q(c) = [ch1 (s) − h2 (s)] ds = 0
a
so ch1 (s) − h2 (s) = 0 for all s, that is, h2 = ch1 .

956
 Project 15.4 The Uncertainty Inequality in Signal Processing

(c) By the Schwartz inequality we have:

 ∞ 2    
  ∞ ∞
 xf (x)f (x) dx ≤


|xf (x)| dx2
2
(f (x)) dx .

−∞ −∞ −∞

On the left,
  
∞
[f (x)]2 ∞ 1 ∞ 1 ∞
xf (x)f
(x) dx = x  − [f (x)]2 dx = − [f (x)]2 dx.
−∞ 2 −∞ 2 −∞ 2 −∞
On the right, by Parseval’s formula and operational property (11) of Section 15.4 in the text,
 ∞  ∞   ∞  2

1  
2 1  
2
[f (x)] dx = f (α) dα = α2 fˆ(α) dα
−∞ 2π −∞ 2π −∞
so  2  
∞ ∞ ∞  2
1 1  
[f (x)] dx ≤ 2
x [f (x)] dx ·
2 2
α2 fˆ(α) dα.
4
−∞ −∞ 2π −∞
 ∞ 2
Dividing both sides by [f (x)]2 dx we have
−∞
∞ 2 ∞
1 −∞
x [f (x)]2 dx −∞ α2 |fˆ(α)|2 dα
≤ ∞ · ∞
4 −∞
[f (x)]2 dx 2π −∞ [f (x)]2 dx
and the result follows from one more application of Parseval’s formula since
 ∞  ∞
2π 2
[f (x)] dx = |fˆ(α)|2 dα.
−∞ −∞

7. (a) By the solution of Problem 6(c) above, equality occurs in the uncertainty inequality when equality occurs in
the Schwartz inequality used in the first step of the solution. By Problem 6(b), this occurs when f
= cxf (x)
for some constant c. Then
f
(x)
= cx
f (x)
ln |f (x)| = cx2 /2 + k
2 2
f (x) = ±ek ecx /2
= decx /2
.

(b) Taking the Fourier transform of both sides of f
= cxf (x) yields

−iαfˆ(α) = cxf(x)
 
d
−iαfˆ(α) = c −i fˆ(α)
dα
d ˆ 1 ˆ
f (α) = αf (α).
dα c
Then proceeding as in part (a) we obtain
2
fˆ(α) = fˆ(0)eα /(2c) .

957
Project 15.4
Project 15.4 Fraunhofer
The Uncertainty Inequality
Diffraction in SignalAperture
by a Circular Processing

Project 15.4 Fraunhofer Diffraction

by a Circular Aperture

(Solutions were provided by the project author.)

1. LX + M Y = ρw cos θ cos ψ + ρw sin θ sin ψ = ρw cos(θ − ψ) and dX dY = ρ dρ dθ

2. Now  
R 2π
U (P ) = C e−ikρw cos(θ−ψ) ρ dρ dθ.
0 0
In 
i−n 2π
eix cos α einα dα = Jn (x)
2π 0
We choose n = 0 and x = −kρw and α = θ − ψ and dα = dθ to get
 2π
1
J0 (−kρw) = J0 (kρw) = e−ikρw cos(θ−ψ) dθ
2π 0
for any ψ so we choose ψ = 0. Then
 R
U (P ) = 2πC J0 (kρw)ρ dρ
0

as required.

3. Using
d  n+1 
u Jn+1 (u) = un+1 Jn (u)
du
with n = 0 we integrate both sides to get
 x
uJ0 (u) du = xJ1 (x) − 0J1 (0) = xJ1 (x).
0

4. So let u = kρw and since

 R
U (P ) = 2πC J0 (kρw)ρ dρ
0
we get
 kwR
J0 (kρw)ρkw d(ρkw)
U (P ) = 2πC (with kw = constant)
0 (kw)2
kRwJ1 (kRw) 2J1 (kRw)
= 2πC = CπR2 .
(kw)2 kRw

5. 1 since J1 (kRw) behaves like kRw/2 for small kRw.

6. I0 is the maximum intensity in the diffraction pattern and occurs at the location of the geometric image of the
star at point O.

7. The smallest nonzero root of J1 (u) is u = 3.832 approximately. So the angular radius of the central diffraction
disk is w = 3.832/kR = 0.69 arc seconds.

958
 Project 15.4 Fraunhofer Diffraction by a Circular Aperture

8. Here is the amplitude.

0.8

0.6

0.4

0.2

kRw
2 4 6 8 10 12 14 16
-0.2

and below is the intensity which is the amplitude squared.

0.8

0.6

0.4

0.2

kRw
1 2 3 4 5 6 7 8 9 10
-0.2

9. If the radius R is doubled, the angular width of the diffraction pattern is halved.

10. If the wavelength is doubled, the angular width of the diffraction pattern is doubled.

11. If the focal length of the lens is doubled, then the angular width stays the same.
 kwb  
J0 (kρw)ρkw d(ρkw) 2 2J1 (kbw) 2 2J1 (kaw)
12. U (P ) = 2πC = Cπ b −a
kwa (kw)2 kbw kaw
13. Use  
2J1 (kbw) 2J1 (kaw)
U (P ) = Cπ b2 − a2 .
kbw kaw
Let b = a + ∆a. With some rearranging we get
 
2π∆a kw(a + ∆a)J1 (kw(a + ∆a)) − kwaJ1 (kwa)
U (P ) = C .
kw kw∆a
Notice the quantity in brackets is approximately
d(uJ1 (u))
= uJ0 (u)
du
with u = kwa so U (P ) = C(2πa∆a)J0 (kwa) as required.

959
Project 16.2
Project 15.4 Instabilities
Fraunhofer Diffraction by Methods
of Numerical a Circular Aperture

Project 16.2
Instabilities of Numerical Methods

(Solutions were provided by the project author.)

1. Substitution of uij = al,j sin(κl i) into the method (7) results in the equation

[α − 2 cos(κl )]al,j+1 = [2 cos(κl ) − β]al,j .

The factor Ql in the one-step iteration scheme (4) is

1 − λ(1 − cos κl )
Ql = ,
1 + λ(1 − cos κl )
where we have used α = 2(1 + 1/λ) and β = 2(1 − 1/λ). Since λ > 0 and 1 − cos κl > 0 for l = 1, 2, . . . , n, we
have −1 < Ql < 1 for l = 1, 2, . . . , n, such that the stability constraint (5) is met.

2. Substitution of uij = al,j sin(κl i) into the method (8) results in the two-step iteration scheme (9). Let al,j+1 =
Ql al,j and al,j = Ql al,j−1 . The factor Ql then satisfies the quadratic equation

Q2l + 4λ(1 − cos κl )Ql − 1 = 0.

The quadratic equation has two roots:


Q±
l = −2λ(1 − cos κl ) ± 4λ2 (1 − cos κl )2 + 1 .

Since λ > 0 and 1 − cos κl > 0 for l = 1, 2, . . . , n, we have Q±

l > 1 for any l = 1, 2, . . . , n, such that the stability
constraint (5) is violated for any λ > 0.

3. Substitution of uij = al,j sin(κl i) into the method (10) results in the two-step iteration scheme

al,j+1 = 2(1 − λ2 + λ2 cos κl )al,j − al,j−1 .

By using the same method as in Problem 2, we obtain the quadratic equation for the factor Ql :

Q2l − 2(1 − λ2 + λ2 cos κl )Ql + 1 = 0.

The quadratic equation has two roots:


Q±
l = 1 − λ + λ cos κl ±
2 2
(1 − λ2 + λ2 cos κl )2 − 1 .
−
It is clear from the quadratic equation that Q+
l Ql = 1. If the two roots are complex, then they are complex
+ − − 2
conjugates and Ql Ql = |Ql | = |Ql | = 1, such that the stability constraint (5) is met. If the two roots are
+ 2

− −
distinct and real, then either |Q+
l | > 1, |Ql | < 1 or |Ql | < 1, |Ql | > 1, such that the stability constraint (5) is
+

violated. The case of distinct and real roots occur when the discriminant of the quadratic equation is positive,
i.e.     
πl πl
(1 − λ + λ cos κl ) − 1 = 4λ sin
2 2 2 2 2 2
λ sin 2
− 1 > 0.
2n 2n
When λ2 ≤ 1, the above inequality is never true. When λ2 > 1, the above inequality is true for l = n.

960
 Project 16.2 Instabilities of Numerical Methods

4. Substitution of ui,j = al,j eiκl i into the method (11) results in the one-step iteration scheme (4) with the
complex-valued factor Ql :
Ql = 1 − λ + λeiκl = 1 − λ + λ cos κl − iλ sin κl .
The squared absolute value of Ql is

|Ql |2 = 1 − 2λ(1 − λ)(1 − cos κl ).

Since λ > 0 and 1 − cos κl > 0 for l = 1, 2, . . . , n, we have |Ql | ≤ 1 for λ ≤ 1, such that the stability constraint
(5) is met. If λ > 1, we have |Ql | > 1 at least for l = n.

5. Substitution of ui,j = al,j eiκl i into the method (12) results in the one-step iteration scheme (4) with the
complex-valued factor Ql :
2 1
Ql = = ,
2 + λeiκl − λe−iκl 1 + iλ sin κl
such that
1
|Ql |2 = < 1.
1 + λ sin2 κl
2

Therefore, the stability constraint (5) is met for any l = 1, 2, . . . , n.

961