Professional Documents
Culture Documents
Problems 1041
22.2 Using the higher-order terms in the series (22.3.23), show that the input impedance Z
in
=
R+jX of a small dipole is given as follows to order (kl)
4
, where L = ln(2a/l):
R =
2
_
1
12
(kl)
2
+
1
360
(kl)
4
_
, X =
2
_
4(1 +L)
kl
1
3
_
L +
2
3
_
(kl)
1
180
_
L
11
30
_
(kh)
3
_
22.3 Consider a small dipole with a linear current given by Eq. (22.3.25). Determine the radiation
vector, and the radiated electric and magnetic elds at a far distance r from the dipole. Cal-
culate the radiated power P
rad
by integrating the radial Poynting vector over a large sphere.
Then identify the radiation resistance R through the denition:
P
rad
=
1
2
RI
0
2
and show R is the same as that given by Eq. (22.3.24)
23
Appendices
A. Physical Constants
We use SI units throughout this text. Simple ways to convert between SI and other
popular units, such as Gaussian, may be found in Refs. [123126].
The Committee on Data for Science and Technology (CODATA) of NIST maintains
the values of many physical constants [112]. The most current values can be obtained
from the CODATA web site [1433]. Some commonly used constants are listed below:
quantity symbol value units
speed of light in vacuum c
0
, c 299792458 ms
1
permittivity of vacuum
0
8.854187817 10
12
F m
1
permeability of vacuum
0
410
7
H m
1
characteristic impedance
0
, Z
0
376.730313461
electron charge e 1.602176462 10
19
C
electron mass m
e
9.109381887 10
31
kg
Boltzmann constant k 1.380650324 10
23
J K
1
Avogadro constant N
A
, L 6.022141994 10
23
mol
1
Planck constant h 6.62606876 10
34
J/Hz
Gravitational constant G 6.67259 10
11
m
3
kg
1
s
2
Earth mass M
5.972 10
24
kg
Earth equatorial radius a
e
6378 km
In the table, the constants c,
0
are taken to be exact, whereas
0
,
0
are derived
from the relationships:
0
=
1
0
c
2
,
0
=
0
=
0
c
The energy unit of electron volt (eV) is dened to be the work done by an electron
in moving across a voltage of one volt, that is, 1 eV = 1.602176462 10
19
C 1 V, or
1 eV = 1.602176462 10
19
J
B. Electromagnetic Frequency Bands 1043
In units of eV/Hz, Plancks constant h is:
h = 4.13566727 10
15
eV/Hz = 1 eV/241.8 THz
that is, 1 eV corresponds to a frequency of 241.8 THz, or a wavelength of 1.24 m.
B. Electromagnetic Frequency Bands
The ITU
divides the radio frequency (RF) spectrum into the following frequency and
wavelength bands in the range from 30 Hz to 3000 GHz:
RF Spectrum
band designations frequency wavelength
ELF Extremely Low Frequency 30300 Hz 110 Mm
VF Voice Frequency 3003000 Hz 1001000 km
VLF Very Low Frequency 330 kHz 10100 km
LF Low Frequency 30300 kHz 110 km
MF Medium Frequency 3003000 kHz 1001000 m
HF High Frequency 330 MHz 10100 m
VHF Very High Frequency 30300 MHz 110 m
UHF Ultra High Frequency 3003000 MHz 10100 cm
SHF Super High Frequency 330 GHz 110 cm
EHF Extremely High Frequency 30300 GHz 110 mm
Submillimeter 300-3000 GHz 1001000 m
An alternative subdivision of the low-frequency
bands is to designate the bands 330 Hz, 30300 Hz,
and 3003000 Hz as extremely low frequency (ELF),
super low frequency (SLF), and ultra low frequency
(ULF), respectively.
Microwaves span the 300 MHz300 GHz fre-
quency range. Typical microwave and satellite com-
munication systems and radar use the 130 GHz
band. The 30300 GHz EHF band is also referred to
as the millimeter band.
The 1100 GHz range is subdivided further into
the subbands shown on the right.
Microwave Bands
band frequency
L 12 GHz
S 24 GHz
C 48 GHz
X 812 GHz
Ku 1218 GHz
K 1827 GHz
Ka 2740 GHz
V 4075 GHz
W 80100 GHz
Some typical RF applications are as follows. AM radio is broadcast at 5351700
kHz falling within the MF band. The HF band is used in short-wave radio, navigation,
amateur, and CB bands. FM radio at 88108 MHz, ordinary TV, police, walkie-talkies,
and remote control occupy the VHF band.
Cell phones, personal communication systems (PCS), pagers, cordless phones, global
positioning systems (GPS), RF identication systems (RFID), UHF-TVchannels, microwave
ovens, and long-range surveillance radar fall within the UHF band.
X-rays and -rays are best described in terms of their energy, which is
related to frequency through Plancks relationship, E = hf. X-rays have typical energies
of the order of keV, and -rays, of the order of MeV and beyond. By comparison, photons
in the visible spectrum have energies of a couple of eV.
The earths atmosphere is mostly opaque to electromagnetic radiation, except for
three signicant windows, the visible, the infrared, and the radio windows. These
three bands span the wavelength ranges of 380-780 nm, 1-12 m, and 5 mm20 m,
respectively.
Within the 1-10 m infrared band there are some narrow transparent windows. For
the rest of the IRrange (11000 m), water and carbon dioxide molecules absorb infrared
radiationthis is responsible for the Greenhouse effect. There are also some minor
transparent windows for 1740 and 330370 m.
1 THz = 10
12
Hz, 1 PHz = 10
15
Hz (petahertz), 1 EHz = 10
18
Hz (exahertz).
C. Vector Identities and Integral Theorems 1045
Beyond the visible band, ultraviolet and X-ray radiation are absorbed by ozone and
molecular oxygen (except for the ozone holes.)
C. Vector Identities and Integral Theorems
Algebraic Identities
A
2
B
2
= A B
2
+A B
2
(C.1)
(A B)C = (B C)A = (C A)B (C.2)
A (B C) = B(A C)C(A B) (BAC-CAB rule) (C.3)
(A B)(C ) = (A C)(B )(A )(B C) (C.4)
(A B)(C ) =
_
(A B)
_
C
_
(A B)C
_
(C.5)
A = n (A n)+( n A) n = A
+A
|
(C.6)
where n is any unit vector, and A
, A
|
are the components of A perpendicular and
parallel to n. Note also that n (A n)= ( n A) n. A three-dimensional vector can
equally well be represented as a column vector:
a = a
x
x +a
y
y +a
z
z a =
_
_
_
a
x
a
y
b
z
_
_
_ (C.7)
Consequently, the dot and cross products may be represented in matrix form:
a b a
T
b = [a
x
, a
y
, a
z
]
_
_
_
b
x
b
y
b
z
_
_
_ = a
x
b
x
+a
y
b
y
+a
z
b
z
(C.8)
a b Ab =
_
_
_
0 a
z
a
y
a
z
0 a
x
a
y
a
x
0
_
_
_
_
_
_
b
x
b
y
b
z
_
_
_ =
_
_
_
a
y
b
z
a
z
b
y
a
z
b
x
a
x
b
z
a
x
b
y
a
y
b
x
_
_
_ (C.9)
The cross-product matrix A satises the following identity:
A
2
= aa
T
(a
T
a)I (C.10)
where I is the 33 identity matrix. Applied to a unit vector n, this identity reads:
I = n n
T
N
2
, where n =
_
_
_
n
x
n
y
n
z
_
_
_,
N =
_
_
_
0 n
z
n
y
n
z
0 n
x
n
y
n
x
0
_
_
_, n
T
n = 1 (C.11)
This corresponds to the matrix form of the parallel/transverse decomposition (C.6).
Indeed, we have a
|
= n( n
T
a) and a
= ( n a) n = n ( n a)=
N(
Na)=
N
2
a.
Therefore, a = Ia = ( n n
T
N
2
)a = a
|
+a
.
1046 23. Appendices
Differential Identities
() = 0 (C.12)
(A) = 0 (C.13)
(A) = A + A (C.14)
(A) = A +A (C.15)
(A B) = (A )B +(B )A +A (B)+B (A) (C.16)
(A B) = B (A)A (B) (C.17)
(A B) = A( B)B( A)+(B )A (A )B (C.18)
(A) =( A)
2
A (C.19)
A
x
B
x
+A
y
B
y
+A
z
B
z
= (A )B +A (B) (C.20)
B
x
A
x
+B
y
A
y
+B
z
A
z
= (B )A +B (A) (C.21)
( n )A = n (A)+( n )A n( A) (C.22)
( n )E E( n )=
_
( n )(E)+ n
_
(E)
_
n (E)
_
+
_
n E ( n E) n (E)( n E)
_
(C.23)
With r = x x +y y +zz, r = r =
_
x
2
+y
2
+z
2
, and the unit vector r = r/r, we have:
r =r , r
2
= 2r ,
1
r
=
r
r
2
, r = 3, r = 0, r =
2
r
(C.24)
Integral Theorems for Closed Surfaces
The theorems involve a volume V surrounded by a closed surface S. The divergence or
Gauss theorem is:
_
V
AdV =
_
S
A n dS (Gauss divergence theorem) (C.25)
where n is the outward normal to the surface. Greens rst and second identities are:
_
V
_
2
+
_
dV =
_
S
n
dS (C.26)
_
V
_
_
dV =
_
S
_
n
_
dS (C.27)
C. Vector Identities and Integral Theorems 1047
where
n
= n is the directional derivative along n. Some related theorems are:
_
V
2
dV =
_
S
n dS =
_
S
n
dS (C.28)
_
V
dV =
_
S
ndS (C.29)
_
V
2
AdV =
_
S
( n )AdS =
_
S
A
n
dS (C.30)
_
S
( n )AdS =
_
S
_
n (A)+( n )A n( A)
_
dS = 0 (C.31)
_
V
AdV =
_
S
n AdS (C.32)
Using Eqs. (C.23) and (C.31), we nd:
_
S
_
E
n
E
n
_
dS =
=
_
S
_
n E ( n E) n (E)( n E)
_
dS
(C.33)
The vectorial forms of Greens identities are [1243,1240]:
_
V
(A B A B) dV =
_
S
n (A B) dS (C.34)
_
V
(B A A B) dV =
_
S
n (A B B A) dS (C.35)
Integral Theorems for Open Surfaces
Stokes theorem involves an open surface S and its boundary contour C:
_
S
n AdS =
_
C
A dl (Stokes theorem) (C.36)
where dl is the tangential path length around C. Some related theorems are:
_
S
_
n A ( n A)
_
dS =
_
C
A dl (C.37)
_
S
_
() n A
_
( n A)
_
_
dS =
_
C
()A dl (C.38)
_
S
n dS =
_
C
dl (C.39)
1048 23. Appendices
_
S
( n )AdS =
_
S
_
n (A)+( n )A n( A)
_
dS =
_
C
dl A (C.40)
_
S
ndS =
1
2
_
C
r dl (C.41)
Eq. (C.41) is a special case of (C.40). Using Eqs. (C.23) and (C.40) we nd:
_
S
_
E
n
E
n
_
dS +
_
C
E dl =
=
_
S
_
n E ( n E) n (E)( n E)
_
dS
(C.42)
D. Greens Functions
The Greens functions for the Laplace, Helmholtz, and one-dimensional Helmholtz equa-
tions are listed below:
2
g(r)=
(3)
(r) g(r)=
1
4r
(D.1)
_
2
+k
2
_
G(r)=
(3)
(r) G(r)=
e
jkr
4r
(D.2)
_
2
z
+
2
_
g(z)= (z) g(z)=
e
jz
2j
(D.3)
where r = r. Eqs. (D.2) and (D.3) are appropriate for describing outgoing waves. We
considered other versions of (D.3) in Sec. 21.3. A more general identity satised by the
Greens function g(r) of Eq. (D.1) is as follows (for a proof, see Refs. [134,135]):
j
g(r)=
1
3
ij
(3)
(r)+
3x
i
x
j
r
2
ij
r
4
g(r) i, j = 1, 2, 3 (D.4)
where
i
= /x
i
and x
i
stands for any of x, y, z. By summing the i, j indices, Eq. (D.4)
reduces to (D.1). Using this identity, we nd for the Greens function G(r)= e
jkr
/4r :
j
G(r)=
1
3
ij
(3)
(r)+
_
_
jk +
1
r
_3x
i
x
j
r
2
ij
r
3
k
2
x
i
x
j
r
2
_
G(r) (D.5)
This reduces to Eq. (D.2) upon summing the indices. For any xed vector p, Eq. (D.5)
is equivalent to the vectorial identity:
_
pG(r)
_
=
2
3
p
(3)
(r)+
_
_
jk +
1
r
_3r(r p)p
r
2
+k
2
r (p r)
_
G(r) (D.6)
The second term on the right is simply the left-hand side evaluated at points away
from the origin, thus, we may write:
_
pG(r)
_
=
2
3
p
(3)
(r) +
_
_
pG(r)
_
_
r}=0
(D.7)
D. Greens Functions 1049
Then, Eq. (D.7) implies the following integrated identity, where is with respect to r :
_
V
P(r
)G(r r
) dV
=
2
3
P(r)+
_
V
_
_
P(r
)G(r r
)
_
_
r
}=r
dV
(D.8)
and r is assumed to lie within V. If r is outside V, then the term 2P(r)/3 is absent.
Technically, the integrals in (D.8) are principal-value integrals, that is, the limits as
0 of the integrals over VV
(r), where V
(r)
has shape other than a sphere or a cube. See Refs. [1282,486,498,624] and [129133]
for the denitions and properties of such principal value integrals.
Another useful result is the so-called Weyl representation or plane-wave-spectrum
representation [22,26,1282,27,541] of the outgoing Helmholtz Greens function G(r):
G(r)=
e
jkr
4r
=
_
e
j(kxx+kyy)
e
jkzz
2jk
z
dk
x
dk
y
(2)
2
(D.9)
where k
2
z
= k
2
k
2
, with k
=
_
k
2
x
+k
2
y
. In order to correspond to either outgoing
waves or decaying evanescent waves, k
z
must be dened more precisely as follows:
k
z
=
_
_
_
_
k
2
k
2
, if k
k, (propagating modes)
j
_
k
2
k
2
, if k
G(x, y, z)e
j(kxx+kyy)
dxdy =
e
jkzz
2jk
z
G(x, y, z) =
_
g(k
x
, k
y
, z)e
j(kxx+kyy)
dk
x
dk
y
(2)
2
(D.11)
Writing
(3)
(r)= (x)(y)(z) and using the inverse Fourier transform:
(x)(y)=
_
e
j(kxx+kyy)
dk
x
dk
y
(2)
2
,
we nd from Eq. (D.2) that g(k
x
, k
y
, z) must satisfy the one-dimensional Helmholtz
Greens function equation (D.3), with k
2
z
= k
2
k
2
x
k
2
y
= k
2
k
2
, that is,
_
2
z
+k
2
z
_
g(k
x
, k
y
, z)= (z) (D.12)
whose outgoing/evanescent solution is g(k
x
, k
y
, z)= e
jkzz
/2jk
z
.
A more direct proof of (D.9) is to use cylindrical coordinates, k
x
= k
cos , k
y
=
k
= k
2
x
+k
2
y
and
2
= x
2
+y
2
. It follows that
1050 23. Appendices
k
x
x +k
y
y = k
r
2
z
2
:
g(k
x
, k
y
, z) =
_ _
e
jkr
4r
e
j(kxx+kyy)
dxdy =
_ _
e
jkr
4r
e
jkcos()
r dr d
=
1
2
_
z
dr e
jkr
_
2
0
d
2
e
jkcos()
=
1
2
_
z
dr e
jkr
J
0
_
k
_
r
2
z
2
_
where we used the integral representation (17.9.2) of the Bessel function J
0
(x). Looking
up the last integral in the table of integrals [1402], we nd:
g(k
x
, k
y
, z)=
1
2
_
z
dr e
jkr
J
0
_
k
_
r
2
z
2
_
=
e
jkzz
2jk
z
(D.13)
where k
z
must be dened exactly as in Eq. (D.10). A direct consequence of Eq. (D.11)
and the even-ness of G(r) in r and of g(k
x
, k
y
, z) in k
x
, k
y
, is the following result:
_
e
j(kxx
+kyy
)
G(r r
)dx
dy
= e
j(kxx+kyy)
e
jkzzz
2jk
z
(D.14)
One can also show the integral:
_
0
e
jk
z
z
e
jkzzz
2jk
z
dz
=
_
_
e
jk
z
z
k
2
z
k
2
z
e
jkzz
2k
z
(k
z
k
z
)
, for z 0
e
jkzz
2k
z
(k
z
+k
z
)
, for z < 0
(D.15)
The proof is obtained by splitting the integral over the sub-intervals [0, z] and
[z, ). To handle the limits at innity, k
z
must be assumed to be slightly lossy, that is,
k
z
=
z
j
z
, with
z
> 0. Eqs. (D.14) and (D.15) can be combined into:
_
V+
e
j k
G(r r
) dV
=
_
_
e
j k
r
k
2
k
2
e
j kr
2k
z
(k
z
k
z
)
, for z 0
e
j kr
2k
z
(k
z
+k
z
)
, for z < 0
(D.16)
where V
+
is the half-space z 0, and k, k
, k
= k
x
x +k
y
y k
z
z
k
= k
x
x +k
y
y +k
z
z
(D.17)
where we note that k
2
k
2
= (k
2
x
+k
2
y
+k
2
z
)(k
2
x
+k
2
y
+k
2
z
)= k
2
z
k
2
z
.
The Greens function results (D.8)(D.17) are used in the discussion of the Ewald-
Oseen extinction theorem in Sec. 14.6.
E. Coordinate Systems 1051
A related Weyl-type representation is obtained by differentiating Eq. (D.9) with re-
spect to z. Assuming that z 0 and interchanging differentiation and integration (and
multiplying by 2), we obtain the identity:
2
z
_
e
jkr
4r
_
=
_
e
jkxx
e
jkyy
e
jkzz
dk
x
dk
y
(2)
2
, z 0 (D.18)
This just means that the left-hand side is the two-dimensional inverse Fourier trans-
form of e
jkzz
with k
z
given by Eq. (D.10). Replacing r by r r
, and r by R = r r
,
and noting that
z
=
z
, we also obtain:
2
z
_
e
jkR
4R
_
=
_
e
jkx(xx
)
e
jky(yy
)
e
jkz(zz
)
dk
x
dk
y
(2)
2
, z z
(D.19)
This result establishes the equivalence between the Kirchhoff-Fresnel diffraction for-
mula and the plane-wave spectrum representation as discussed in Sec. 17.17. For the
vector diffraction case, we also need the derivatives of G with respect to the transverse
coordinates x, y. Differentiating (D.9) with respect to x (or with respect to y), we have:
2
x
_
e
jkr
4r
_
=
_
k
x
k
z
e
jkxx
e
jkyy
e
jkzz
dk
x
dk
y
(2)
2
, z 0 (D.20)
E. Coordinate Systems
The denitions of cylindrical and spherical coordinates were given in Sec. 14.8. The
expressions of the gradient, divergence, curl, Laplacian operators, and delta functions
are given below in cartesian, cylindrical, and spherical coordinates.
Cartesian Coordinates
= x
x
+ y
y
+z
z
2
=
2
x
2
+
2
y
2
+
2
z
2
A =
A
x
x
+
A
y
y
+
A
z
z
A = x
_
A
z
y
A
y
z
_
+ y
_
A
x
z
A
z
x
_
+z
_
A
y
x
A
x
y
_
=
x y z
z
A
x
A
y
A
z
(3)
(r r
)= (x x
)(y y
)(z z
)
(E.1)
1052 23. Appendices
Cylindrical Coordinates
=
+z
z
(E.2a)
2
=
1
_
+
1
2
+
2
z
2
(E.2b)
A =
1
(A
+
1
+
A
z
z
(E.2c)
A =
_
1
A
z
z
_
+
_
A
z
A
z
_
+z
1
_
(A
_
(E.2d)
(3)
(r r
)=
1
)(
)(z z
) (E.2e)
Spherical Coordinates
=r
r
+
1
r
1
r sin
(E.3a)
2
=
1
r
2
r
_
r
2
r
_
+
1
r
2
sin
_
sin
_
+
1
r
2
sin
2
2
(E.3b)
A =
1
r
2
(r
2
A
r
)
r
+
1
r sin
(sinA
+
1
r sin
A
(E.3c)
A =r
1
r sin
_
(sinA
_
+
1
r
_
1
sin
A
r
(rA
)
r
_
(E.3d)
+
1
r
_
(rA
)
r
A
r
(3)
(r r
)=
1
r
2
sin
(r r
)(
)(
) (E.3e)
Transformations Between Coordinate Systems
A vector A can be expressed component-wise in the three coordinate systems as:
A = xA
x
+ yA
y
+z A
z
= A
+
A
+z A
z
=r A
r
+
+
A
(E.4)
The components in one coordinate system can be expressed in terms of the compo-
nents of another by using the following relationships between the unit vectors, which
F. Fresnel, Exponential, Sine, and Cosine Integrals 1053
were also given in Eqs. (14.8.1)(14.8.3):
x = cos
y = sin
= xcos + ysin
= xsin+ ycos
x = cos
sin
y = sin+
cos
(E.5)
= r sin
z = r cos
r = z cos + sin
= z sin + cos
z =r cos
sin
=r sin +
cos
(E.6)
x = r sincos
y = r sinsin
z = r cos
r = xcos sin + ysinsin +z cos
= xsin+ ycos
(E.7)
x =r sincos +
cos cos
sin
y =r sinsin+
cos sin+
cos
z =r cos
sin
(E.8)
For example, to express the spherical components A
, A
=
A =
( xA
x
+ yA
y
+z A
z
)= (
x) A
x
+(
y) A
y
+(
z) A
z
A
=
A =
( xA
x
+ yA
y
+z A
z
)= (
x) A
x
+(
y) A
y
+(
z) A
z
The dot products can be read off Eq. (E.7), resulting in:
A
= cos cos A
x
+sincos A
y
sinA
z
A
= sinA
x
+cos A
y
(E.9)
Similarly, using Eq. (E.6) the cylindrical components A
, A
z
can be expressed in terms
of spherical components as:
A
= A = (r A
r
+
+
A
)= sinA
r
+cos A
A
z
= z A = z (r A
r
+
+
A
)= cos A
r
cos A
(E.10)
F. Fresnel, Exponential, Sine, and Cosine Integrals
The Fresnel functions C(x) and S(x) are dened by [1401]:
C(x)=
_
x
0
cos
_
2
t
2
_
dt , S(x)=
_
x
0
sin
_
2
t
2
_
dt (F.1)
They may be combined into the complex function:
(x)= C(x)jS(x)=
_
x
0
e
j(/2)t
2
dt (F.2)
C(x), S(x), and (x) are odd functions of x and have the asymptotic values:
C()= S()=
1
2
, ()=
1 j
2
(F.3)
1054 23. Appendices
At x = 0, we have (0)= 0 and
2
x
2
_
S(x) =
1
2
1
x
cos
_
2
x
2
_
(F.4)
Associated with C(x) and S(x) are the type-2 Fresnel integrals:
C
2
(x)=
_
x
0
cos t
2t
dt , S
2
(x)=
_
x
0
sint
2t
dt (F.5)
They are combined into the complex function:
2
(x)= C
2
(x)jS
2
(x)=
_
x
0
e
jt
2t
dt (F.6)
The two types are related by, if x 0:
C(x)= C
2
_
2
x
2
_
, S(x)= S
2
_
2
x
2
_
, (x)=
2
_
2
x
2
_
(F.7)
and if x < 0, we set (x)= (x)=
2
(x
2
/2).
The Fresnel function
2
(x) can be evaluated numerically using Boersmas approx-
imation [1259], which achieves a maximum error of 10
9
over all x. The algorithm
approximates the function
2
(x) as follows:
2
(x)=
_
_
e
jx
_
x
4
11
_
n=0
(a
n
+jb
n
)
_
x
4
_
n
, if 0 x 4
1 j
2
+e
jx
4
x
11
_
n=0
(c
n
+jd
n
)
_
4
x
_
n
, if x > 4
(F.8)
where the coefcients a
n
, b
n
, c
n
, d
n
are given in [1259]. Consistency with the small- and
large-x expansions of (x) requires that a
0
+jb
0
=
8/ and c
0
+jd
0
= j/
8. We
have implemented Eq. (F.8) with the MATLAB function fcs2:
F2 = fcs2(x); % Fresnel integrals
2
(x) = C
2
(x)jS
2
(x)
The ordinary Fresnel integral (x) can be computed with the help of Eq. (F.7). The
MATLAB function fcs calculates (x) for any vector of values x by calling fcs2:
F = fcs(x); % Fresnel integrals (x) = C(x)jS(x)
In calculating the radiation patterns of pyramidal horns, it is desired to calculate a
Fresnel diffraction integral of the type:
F
0
(v, )=
_
1
1
e
jv
e
j(/2)
2
2
d (F.9)
F. Fresnel, Exponential, Sine, and Cosine Integrals 1055
Making the variable change t = v/, this integral can be computed in terms of
the Fresnel function (x)= C(x)jS(x) as follows:
F
0
(v, )=
1
e
j(/2)(v
2
/
2
)
_
_
v
+
_
_
v
__
(F.10)
where we also used the oddness of (x). The value of Eq. (F.9) at v = 0 is:
F
0
(0, )=
1
_
()()
_
= 2
()
(F.11)
Eq. (F.10) assumes that }= 0. If = 0, the integral (F.9) reduces to the sinc function:
F
0
(v, 0)= 2
sin(v)
v
(F.12)
From either (F.11) or (F.12), we nd F
0
(0, 0)= 2. A related integral that is also
required in the theory of horns is the following:
F
1
(v, )=
_
1
1
cos
_
2
_
e
jv
e
j(/2)
2
2
d (F.13)
Writing cos(/2)= (e
j/2
+e
j/2
)/2, the integral F
1
(v, s) can be expressed in
terms of F
0
(v, ) as follows:
F
1
(v, )=
1
2
_
F
0
(v +0.5, )+F
0
(v 0.5, )
_
(F.14)
It can be veried easily that F
0
(0.5, )= F
0
(0.5, ), therefore, the value of F
1
(v, )
at v = 0 will be given by:
F
1
(0, )= F
0
(0.5, )=
1
e
j/(8
2
)
_
_
1
2
+
_
_
1
2
__
(F.15)
Using the asymptotic expansion (F.4), we nd the expansion valid for small :
_
1
2
:
_
=
1 j
2
2
e
j/(8
2
)
, for small (F.16)
For = 0, the integral F
1
(v, ) reduces to the double-sinc function:
F
1
(v, 0)=
_
1
1
cos
_
2
_
e
jv
d =
1
2
_
F
0
(v +0.5, 0)+F
0
(v 0.5, 0)
_
=
sin
_
(v +0.5)
_
(v +0.5)
+
sin
_
(v 0.5)
_
(v 0.5)
=
4
cos(v)
1 4v
2
(F.17)
From either Eq. (F.16) or (F.17), we nd F
1
(0, 0)= 4/.
The MATLAB function diffint can be used to evaluate both Eq. (F.9) and (F.13) for
any vector of values v and any vector of positive numbers , including = 0. It calls
fcs to evaluate the diffraction integral (F.9) according to Eq. (F.10). Its usage is:
1056 23. Appendices
F0 = diffint(v,sigma,0); % diffraction integral F
0
(v, ), Eq. (F.9)
F1 = diffint(v,sigma,1); % diffraction integral F
1
(v, ), Eq. (F.13)
The vectors v,sigma can be entered either as rows or columns, but the result will
be a matrix of size length(v) x length(sigma). The integral F
0
(v, ) can also be
calculated by the simplied call:
F0 = diffint(v,sigma); % diffraction integral F
0
(v, ), Eq. (F.9)
Actually, the most general syntax of diffint is as follows:
F = diffint(v,sigma,a,c1,c2); % diffraction integral F(v, , a), Eq. (F.18)
It evaluates the more general integral:
F(v, , a)=
_
c2
c1
cos
_
a
2
_
e
jv
e
j(/2)
2
2
d (F.18)
For a = 0, we have:
F(v, , 0)=
1
e
j(/2)(v
2
/
2
)
_
_
v
c
1
_
_
v
c
2
__
(F.19)
For a }= 0, we can express F(v, , a) in terms of F(v, , 0):
F(v, , a)=
1
2
_
F(v +0.5a, , 0)+F(v 0.5a, , 0)
_
(F.20)
For a = 0 and = 0, F(v, , a) reduces to the complex sinc function:
F(v, 0, 0)=
e
jvc2
e
jvc1
jv
= (c
2
c
1
)
sin
_
(c
2
c
1
)v/2
_
(c
2
c
1
)v/2
e
j(c2+c1)v/2
(F.21)
Stationary Phase Approximation
The Fresnel integrals nd also application in the stationary-phase approximation for
evaluating integrals. The approximation can be stated as follows:
_
f(x)e
j(x)
dx =
2j
(x
0
)
f(x
0
)e
j(x0)
(F.22)
where x
0
is a stationary point of the phase (x), that is, the solution of
(x
0
)= 0,
where for simplicity we assume that there is only one such point (otherwise, one has a
sum of terms like (F.22), one for each solution of
(x
0
)(x x
0
)+
1
2
(x
0
)(x x
0
)
2
= (x
0
)+
1
2
(x
0
)(x x
0
)
2
F. Fresnel, Exponential, Sine, and Cosine Integrals 1057
Making this approximation in the integral and assuming that f(x) is slowly varying
in the neighborhood of x
0
, we may replace f(x) by its value at x
0
:
_
f(x)e
j(x)
dx =
_
f(x
0
)e
j
_
(x0)+
(x0)(xx0)
2
/2
_
dx
= f(x
0
)e
j(x0)
_
e
j
(x0)(xx0)
2
/2
dx
The last integral can be reduced to the complex Fresnel integral by the change of
variables (x x
0
)=
_
/
(x
0
) u:
_
e
j
(x0)(xx0)
2
/2
dx =
(x
0
)
_
e
ju
2
/2
du =
(x
0
)
_
()()
_
Using
_
()()
_
= 2
()= 1 +j =
_
2j, we obtain
_
e
j
(x0)(xx0)
2
/2
dx =
2j
(x
0
)
Normally, the phase depends on a positive parameter in the form (x)= (x),
and the stationary-phase approximation is justied in the limit .
Exponential, Sine, and Cosine Integrals
Several antenna calculations, such as mutual impedances and directivities, can be re-
duced to the exponential integral, which is dened as follows [1401]:
E
1
(z)=
_
z
e
u
u
du = e
z
_
0
e
t
z +t
dt (exponential integral) (F.23)
where z is a complex number with phase restricted such that arg z < . This range
allows pure imaginary zs. The built-in MATLAB function expint evaluates E
1
(z) at an
array of zs. Related to E
1
(z) are the sine and cosine integrals:
S
i
(z)=
_
z
0
sinu
u
du (sine integral)
C
i
(z)= +lnz +
_
z
0
cos u 1
u
du (cosine integral)
(F.24)
where is the Euler constant = 0.5772156649... . A related cosine integral is:
C
in
(z)=
_
z
0
1 cos u
u
du = +lnz C
i
(z) (F.25)
For z 0, the sine and cosine integrals are related to E
1
(z) by [1401]:
S
i
(z)=
E
1
(jz)E
1
(jz)
2j
+
2
= Im
_
E
1
(jz)
_
+
2
C
i
(z)=
E
1
(jz)+E
1
(jz)
2
= Re
_
E
1
(jz)
_
(F.26)
1058 23. Appendices
while for z 0, we have S
i
(z)= S
i
(z) and C
i
(z)= C
i
(z)+j. Conversely, we have
for z > 0:
E
1
(jz)= C
i
(z)+j
_
S
i
(z)
2
_
= ln(z)+C
in
(z)+j
_
S
i
(z)
2
_
(F.27)
The MATLAB functions Si, Ci, Cin evaluate the sine and cosine integrals at any
vector of zs by using the relations (F.26) and the built-in function expint:
y = Si(z); % sine integral, Eq. (F.24)
y = Ci(z); % sine integral, Eq. (F.24)
y = Cin(z); % sine integral, Eq. (F.25)
A related integral that appears in calculating mutual and self impedances is what
may be called a Greens function integral:
Gi(d, z
0
, h, s)=
_
h
0
e
jkR
R
e
jksz
dz, R =
_
d
2
+(z z
0
)
2
, s = :1 (F.28)
This integral can be reduced to the exponential integral by the change of variables:
v = jk
_
R+s(z z
0
)
_
s
dv
v
=
dz
R
which gives
_
h
0
e
jkR
R
e
jksz
dz = se
jksz0
_
v1
v0
e
u
u
du, or,
Gi(d, z
0
, h, s)=
_
h
0
e
jkR
R
e
jksz
dz = se
jksz0
_
E
1
(ju
0
)E
1
(ju
1
)
_
(F.29)
where
v
0
= ju
0
, u
0
= k
_
_
d
2
+z
2
0
sz
0
_
v
1
= ju
1
, u
1
= k
_
_
d
2
+(h z
0
)
2
+s(h z
0
)
_
The function Gi evaluates Eq. (F.29), where z
0
, s, and the resulting integral J, can be
vectors of the same dimension. Its usage is:
J = Gi(d,z0,h,s); % Greens function integral, Eq. (F.29)
Another integral that appears commonly in antenna work is:
_
0
cos(cos )cos
sin
d = S
i
(2)sinC
in
(2)cos (F.30)
Its proof is straightforward by rst changing variables to z = cos , then using
partial fraction expansion, and nally changing variables to u = (1 + z), and using
the denitions (F.24) and (F.25):
_
0
cos(cos )cos
sin
d =
_
1
1
cos(z)cos
1 z
2
dz
=
1
2
_
1
1
cos(z)cos
1 +z
dz +
1
2
_
1
1
cos(z)cos
1 z
dz =
_
1
1
cos(z)cos
1 +z
dz
=
_
2
0
cos(u )cos
u
du = sin
_
2
0
sinu
u
du cos
_
2
0
1 cos u
u
du
G. Gauss-Legendre Quadrature 1059
G. Gauss-Legendre Quadrature
In many parts of this book it is necessary to perform numerical integration. Gauss-
Legendre quadrature is one of the best integration methods, and we have implemented
it with the MATLAB functions quadr and quadrs. Below, we give a brief description of
the method.
J. Stoer and R. Burlisch, Introduction to Numerical Analysis, Springer, NY, (1980); and, G. H. Golub and
J. H. Welsch, Calculation of Gauss Quadrature Rules, Math. Comput., 23, 221 (1969).
1060 23. Appendices
[w,x] = quadrs([1,1.5,2],N); % or by, [w,x] = quadrs(1:0.5:2, N);
which has a percentage error of 1.2810
9
. Next, we discuss the theoretical basis of
the method.
The interval [a, b] can be replaced by the standardized interval [1, 1] with the
transformation from a x b to 1 z 1:
x =
_
b a
2
_
z +
_
b +a
2
_
(G.3)
If w
i
and z
i
are the weights and nodes with respect to the interval [1, 1], then those
with respect to [a, b] can be constructed simply as follows, for i = 1, 2, . . . , N:
x
i
=
_
b a
2
_
z
i
+
_
b +a
2
_
w
x
i
=
_
b a
2
_
w
i
(G.4)
where the scaling of the weights follows from the scaling of the differentials dx =
dz(b a)/2, so the value of the integral (G.1) is preserved by the transformation.
Gauss-Legendre quadrature is nicely tied with the theory of orthogonal polynomials
over the interval [1, 1], which are the Legendre polynomials. For N-point quadrature,
the nodes z
i
, i = 1, 2, . . . , N are the N roots of the Legendre polynomial P
N
(z), which
all lie in the interval [1, 1]. The method is justied by the following theorem:
For any polynomial P(z) of degree at most 2N1, the quadrature formula (G.1) is
satised exactly, that is,
_
1
1
P(z) dz =
N
_
i=1
w
i
P(z
i
) (G.5)
provided that the z
i
are the N roots of the Legendre polynomial P
N
(z).
The Legendre polynomials P
n
(z) are obtained via the process of Gram-Schmidt or-
thogonalization of the non-orthogonal monomial basis 1, z, z
2
, . . . , z
n
. . . . Orthogo-
nality is dened with respect to the following inner product over the interval [1, 1]:
(f, g)=
_
1
1
f(z)g(z)dz (G.6)
The standard denition of the Legendre polynomials is:
P
n
(z)=
1
2
n
n!
d
n
dz
n
_
(z
2
1)
n
_
, n = 0, 1, 2, . . . (G.7)
The rst few of them are listed below:
P
0
(z) = 1
P
1
(z) = z
P
2
(z) = (3/2)
_
z
2
(1/3)
_
P
3
(z) = (5/2)
_
z
3
(3/5)z
_
P
4
(z) = (35/8)
_
z
4
(6/7)z
2
+(3/35)
_
(G.8)
G. Gauss-Legendre Quadrature 1061
They are normalized such that P
n
(1)= 1 and are mutually orthogonal with respect
to (G.6), but do not have unit norm:
(P
n
, P
m
)=
_
1
1
P
n
(z)P
m
(z)dz =
2
2n +1
nm
(G.9)
Moreover, they satisfy the three-term recurrence relation:
zP
n
(z)=
_
n
2n +1
_
P
n1
(z)+
_
n +1
2n +1
_
P
n+1
(z) (G.10)
The Gram-Schmidt orthogonalization process of the monomial basis f
n
(z)= z
n
is
the following order-recursive construction:
initialize P
0
(z)= f
0
(z)= 1
for n = 1, 2, 3, . . . , do
P
n
(z)= f
n
(z)
n1
_
k=0
(f
n
, P
k
)
(P
k
, P
k
)
P
k
(z)
A few steps of the construction will clarify it:
P
1
(z)= f
1
(z)
(f
1
, P
0
)
(P
0
, P
0
)
P
0
(z)= z
where (f
1
, P
0
)= (z, 1)=
_
1
1
zdz = 0. Then, construct P
2
by:
P
2
(z)= f
2
(z)
(f
2
, P
0
)
(P
0
, P
0
)
P
0
(z)
(f
2
, P
1
)
(P
1
, P
1
)
P
1
(z)
where now we have (f
2
, P
1
)= (z
2
, z)=
_
1
1
z
3
dz = 0, and
(f
2
, P
0
)= (z
2
, 1)=
_
1
1
z
2
dz =
2
3
, (P
0
, P
0
)= (1, 1)=
_
1
1
dz = 2
Therefore,
P
2
(z)= z
2
2/3
2
= z
2
1
3
Then, normalize it such that P
2
(1)= 1, and so on. For our discussion, we are going
to renormalize the Legendre polynomials to unit norm. Because of (G.9), this amounts
to multiplying the standard P
n
(z) by the factor
_
(2n +1)/2. Thus, we re-dene:
P
n
(z)=
2n +1
2
1
2
n
n!
d
n
dz
n
_
(z
2
1)
n
_
, n = 0, 1, 2, . . . (G.11)
Thus, (G.9) becomes (P
n
, P
m
)=
nm
. In particular, we note that now
P
0
(z)=
1
2
(G.12)
1062 23. Appendices
By introducing the same scaling factors into each term of the recurrence (G.10), we
nd that the renormalized P
n
(z) satisfy:
zP
n
(z)=
n
P
n1
(z)+
n+1
P
n+1
(z) ,
n
=
n
4n
2
1
(G.13)
This relationship can be assumed to be valid also at n = 0, provided we dene
P
1
(z)= 0. For each order n, the Gram-Schmidt procedure replaces the non-orthogonal
monomial basis by the orthonormalized Legendre basis:
_
1, z, z
2
, . . . , z
n
_
_
P
0
(z), P
1
(z), P
2
(z), . . . , P
n
(z)
_
Thus, any polynomial Q(z) of degree n can be expanded uniquely in either basis:
Q(z)=
n
_
k=0
q
k
z
k
=
n
_
k=0
c
k
P
k
(z)
with the expansion coefcients calculated from c
k
= (Q, P
k
). This also implies that if
Q(z) has order n 1 then, it will be orthogonal to P
n
(z).
Next, we turn to the proof of the basic Gauss-Legendre result (G.5). Given a polyno-
mial P(z) of order 2N1, we can expand it uniquely in the form:
P(z)= P
N
(z)Q(z)+R(z) (G.14)
where Q(z) and R(z) are the quotient and remainder of the division by the Legendre
polynomial P
N
(z), and both will have order N 1. Then, the integral of P(z) can be
written in inner-product notation as follows:
_
1
1
P(z)dz = (P, 1)= (P
N
Q +R, 1)= (P
N
Q, 1)+(R, 1)= (Q, P
N
)+(R, 1)
But (Q, P
N
)= 0 because Q(z) has order N 1 and P
N
(z) is orthogonal to all such
polynomials. Thus, the integral of P(z) can be expressed only in terms of the integral
of the remainder polynomial R(z), which has order N1:
_
1
1
P(z)dz = (P, 1)= (R, 1)=
_
1
1
R(z)dz (G.15)
The right-hand side of the integration rule (G.5) can also be expressed in terms of R(z):
N
_
i=1
w
i
P(z
i
)=
N
_
i=1
w
i
P
N
(z
i
)Q(z
i
)+
N
_
i=1
w
i
R(z
i
) (G.16)
and, because we assumed that P
N
(z
i
)= 0,
N
_
i=1
w
i
P(z
i
)=
N
_
i=1
w
i
R(z
i
) (G.17)
Thus, combining (G.15) and (G.17), we obtain the following condition, which is equiv-
alent to Eq. (G.5),
_
1
1
R(z)dz =
N
_
i=1
w
i
R(z
i
) (G.18)
G. Gauss-Legendre Quadrature 1063
Because R(z) is an arbitrary polynomial of degree N1, and has only Ncoefcients,
this condition can be satised with a common set of N weights w
i
for all such R(z). If
we had not assumed initially that the z
i
were the zeros of P
N
(z), and took them to be
an arbitrary set of N distinct points in [1, 1], then (G.18) would read as
_
1
1
R(z)dz =
N
_
i=1
w
i
P
N
(z
i
)Q(z
i
)+
N
_
i=1
w
i
R(z
i
)
In order for this to be satised for all R(z) and all Q(z), then (G.18) must still be
satised by setting Q(z)= 0, which xes the weights w
i
. Therefore, the rst term in
the right-hand side must be zero for all polynomials Q(z) of degree N1, and one can
show that this implies that P
N
(z
i
)= 0, that is, the z
i
must be the zeros of P
N
(z).
Condition (G.18) can be used to determine the weights by expanding R(z) into either
the monomial basis or the Legendre basis, that is, because R(z) has degree N1:
R(z)=
N1
_
k=0
r
k
z
k
=
N1
_
k=0
c
k
P
k
(z) (G.19)
Inserting, for example, the monomial basis into (G.18) and matching the coefcients
of r
k
on either side, we obtain the system of N equations for the weights:
N
_
i=1
z
k
i
w
i
=
_
1
1
z
k
dz =
1 +(1)
k
k +1
, k = 0, 1, . . . , N1 (G.20)
Dening the matrix F
ki
= z
k
i
and the vector u
k
=
_
1+(1)
k
_
/(k+1), we may write
(G.20) in the compact matrix form:
Fw = u w = F
1
u (G.21)
Alternatively, we may use the Legendre basis, which is more elegant. The left hand
side of (G.18) will receive contribution only fromthe k = 0 termbecause P
0
is orthogonal
to all the succeeding P
k
. Indeed, using the denition (G.12), we have:
_
1
1
R(z)dz = (R, 1)=
2(R, P
0
)=
2
N1
_
k=0
c
k
(P
K
, P
0
)=
2
N1
_
k=0
c
k
k0
=
2c
0
The right-hand side of (G.18) may be written as follows. Dening the NN ma-
trix P
ki
= P
k
(z
i
), i = 1, 2. . . , N, and k = 0, 1, . . . , N 1, and the row vector c
T
=
[c
0
, c
1
, . . . , c
N1
] of expansion coefcients, we have,
N
_
i=1
w
i
R(z
i
)=
N1
_
k=0
N
_
i=1
c
k
P
k
(z
i
)w
i
= c
T
Pw
Thus, (G.18) now reads, where u
0
= [1, 0, 0, . . . , 0]
T
:
c
T
Pw =
2c
0
=
2c
T
u
0
1064 23. Appendices
Because the vector c is arbitrary, we must have the condition:
Pw =
2u
0
w =
2P
1
u
0
(G.22)
The matrix Phas some rather interesting properties. First, it has mutually orthogonal
columns. Second, these columns are the eigenvectors of a Hermitian tridiagonal matrix
whose eigenvalues are the zeros z
i
. Thus, the problem of nding both z
i
and w
i
is
reduced to an eigenvalue problem.
These eigenvalue properties follow from the recursion (G.13) of the normalized Leg-
endre polynomials. For n = 0, 1, 2, 3, the recursion reads explicitly:
zP
0
(z) =
1
P
1
(z)
zP
1
(z) =
1
P
0
(z)+
2
P
2
(z)
zP
2
(z) =
2
P
1
(z)+
3
P
3
(z)
zP
3
(z) =
3
P
2
(z)+
4
P
4
(z)
which can be written in matrix form:
z
_
_
_
_
_
P
0
(z)
P
1
(z)
P
2
(z)
P
3
(z)
_
_
_
_
_
=
_
_
_
_
_
0
1
0 0
1
0
2
0
0
2
0
3
0 0
3
0
_
_
_
_
_
_
_
_
_
_
P
0
(z)
P
1
(z)
P
2
(z)
P
3
(z)
_
_
_
_
_
+
_
_
_
_
_
0
0
0
4
P
4
(z)
_
_
_
_
_
and more generally,
z
_
_
_
_
_
_
_
_
_
_
_
P
0
(z)
P
1
(z)
P
2
(z)
.
.
.
P
N2
(z)
P
N1
(z)
_
_
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_
0
1
0 0 0
1
0
2
0 0
0
2
0
3
0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0
N2
0
N1
0 0 0
N1
0
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
P
0
(z)
P
1
(z)
P
2
(z)
.
.
.
P
N2
(z)
P
N1
(z)
_
_
_
_
_
_
_
_
_
_
_
+
_
_
_
_
_
_
_
_
_
_
_
0
0
0
.
.
.
0
N
P
N
(z)
_
_
_
_
_
_
_
_
_
_
_
Now, if z is replaced by the ith zero z
i
of P
N
(z), the last column will vanish and we
obtain the eigenvalue equation:
_
_
_
_
_
_
_
_
_
_
_
0
1
0 0 0
1
0
2
0 0
0
2
0
3
0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0
N2
0
N1
0 0 0
N1
0
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
P
0
(z
i
)
P
1
(z
i
)
P
2
(z
i
)
.
.
.
P
N2
(z
i
)
P
N1
(z
i
)
_
_
_
_
_
_
_
_
_
_
_
= z
i
_
_
_
_
_
_
_
_
_
_
_
P
0
(z
i
)
P
1
(z
i
)
P
2
(z
i
)
.
.
.
P
N2
(z
i
)
P
N1
(z
i
)
_
_
_
_
_
_
_
_
_
_
_
(G.23)
Denoting the above tridiagonal matrix by A and the column of P
k
(z
i
)s by p
i
, we
may write compactly:
Ap
i
= z
i
p
i
, i = 1, 2, . . . , N (G.24)
H. Lorentz Transformations 1065
Thus, the eigenvalues of A are the zeros z
i
and the corresponding eigenvectors are
the columns p
i
of the matrix P that we introduced in (G.22). Because the zeros z
i
are
distinct and A is a Hermitian matrix, its eigenvectors will be mutually orthogonal:
p
T
i
p
j
= d
2
i
ij
(G.25)
where d
i
= |p
i
| are the norms of the vectors p
i
. It follows that the orthonormalized
eigenvectors of A will be v
i
= p
i
/d
i
, and the orthogonal matrix of eigenvectors having
the v
i
as columns will be V = [v
1
, v
2
, . . . , v
N
], or, expressed in terms of the matrix P
and the diagonal matrix D = diagd
1
, d
2
, . . . , d
N
:
V = PD
1
(G.26)
Replacing P in (G.22) by P = VD and using the orthogonality V
T
V = I of the eigen-
vector matrix, or V
1
= V
T
, we obtain the solution:
w =
2D
1
V
T
u
0
w
i
=
2d
1
i
(v
T
i
u
0
) (G.27)
The matrix D can itself be expressed in terms of V by noting that the top entry of p
i
is P
0
(z
i
)= 1/
2d
i
), or, d
1
i
=
2(v
T
i
u
0
). It nally follows from Eq. (G.27) that
w
i
= d
2
i
= 2(v
T
i
u
0
)
2
(G.28)
In MATLAB language, v
T
i
u
0
= V(1, i), that is, the rst row of V. Because the eigen-
vectors of the Hermitian matrix A are real-valued and unique up to a sign, Eq. (G.28)
allows the unique determination of the weights from the eigenvector matrix V.
The above discussion leads to two possible implementations of the MATLAB function
quadr. In the rst, we obtain the coefcients of the Legendre polynomial P
N
(z), nd its
zeros using the built-in function root, and then solve the linear equation (G.21) for the
weights. The second approach, implemented by the function quadr2 and the related
function quadrs2, determines z
i
, w
i
from the eigenvalue problem of the matrix A.
H. Lorentz Transformations
According to Einsteins special theory of relativity [461], Lorentz transformations de-
scribe the transformation between the space-time coordinates of two coordinate sys-
tems moving relative to each other at constant velocity. Maxwells equations remain
invariant under Lorentz transformations. This is demonstrated below.
Let the two coordinate frames be S and S
is moving at a constant
velocity v. For example, if v is in the z-direction, the space-time coordinates t, x, y, z
of S are related to the coordinates t
, x
, y
, z
of S
=
_
t
v
c
2
z
_
z
= (z vt)
x
= x
y
= y
, where =
1
1 v
2
/c
2
1066 23. Appendices
where c is the speed of light in vacuum. Dening the scaled quantities = ct and
= v/c, the above transformation and its inverse, obtained by replacing by , may
be written as follows:
= ( z)
z
= (z )
x
= x
y
= y
= (
+z
)
z = (z
)
x = x
y = y
(H.1)
These transformations are also referred to as Lorentz boosts to indicate the fact that
one frame is boosted to move relative to the other. Interchanging the roles of z and x, or
z and y, one obtains the Lorentz transformations for motion along the x or y directions,
respectively. Eqs. (H.1) may be expressed more compactly in matrix form:
x
= Lx , where x =
_
_
_
_
_
x
y
z
_
_
_
_
_
, x
=
_
_
_
_
_
_
_
_
_
_
, L =
_
_
_
_
_
0 0
0 1 0 0
0 0 1 0
0 0
_
_
_
_
_
(H.2)
Such transformations leave the quadratic form (c
2
t
2
x
2
y
2
z
2
) invariant, that is,
c
2
t
2
x
2
y
2
z
2
= c
2
t
2
x
2
y
2
z
2
(H.3)
Introducing the diagonal metric matrix G = diag(1, 1, 1, 1), we may write the
quadratic form as follows, where x
T
denotes the transposed vector, that is, the row
vector x
T
= [, x, y, z]:
x
T
Gx =
2
x
2
y
2
z
2
= c
2
t
2
x
2
y
2
z
2
(H.4)
More generally, a Lorentz transformation is dened as any linear transformation x
=
Lx that leaves the quadratic form x
T
Gx invariant. The invariance condition requires
that: x
T
Gx
= x
T
L
T
GLx = x
T
Gx, or
L
T
GL = G (H.5)
In addition to the Lorentz boosts of Eq. (H.1), the more general transformations
satisfying (H.5) include rotations of the three spatial coordinates, as well as time or
space reections. For example, a rotation has the form:
L =
_
_
_
_
_
1 0 0 0
0
0 R
0
_
_
_
_
_
where R is a 33 orthogonal rotation matrix, that is, R
T
R = I, where I is the 33
identity matrix. The most general Lorentz boost corresponding to arbitrary velocity
v = [v
x
, v
y
, v
z
]
T
is given by:
L =
_
_
_
_
T
I +
2
+1
T
_
_
_
_
, where =
v
c
, =
1
_
1
T
(H.6)
H. Lorentz Transformations 1067
When v = [0, 0, v]
T
, or = [0, 0, ]
T
, Eq. (H.6) reduces to (H.1). Dening = =
_
T
and the unit vector
= /, and using the relationship
2
2
=
2
1, it can be
veried that the spatial part of the matrix L can be written in the form:
I +
2
+1
T
= I +( 1)
T
(H.7)
The set of matrices L satisfying Eq. (H.5) forms a group called the Lorentz group. In
particular, the z-directed boosts of Eq. (H.2) form a commutative subgroup. Denoting
these boosts by L(), the application of two successive boosts by velocity factors
1
=
v
1
/c and
2
= v
2
/c leads to the combined boost L()= L(
1
)L(
2
), where:
=
1
+
2
1 +
1
2
v =
v
1
+v
2
1 +v
1
v
2
/c
2
(H.8)
with = v/c. Eq. (H.8) is Einsteins relativistic velocity addition theorem. The same
group property implies also that L
1
()= L(). The proof of Eq. (H.8) follows from
the following condition, where
1
= 1/
_
1
2
1
and
2
= 1/
_
1
2
2
:
_
_
_
_
_
0 0
0 1 0 0
0 0 1 0
0 0
_
_
_
_
_
=
_
_
_
_
_
1
0 0
1
1
0 1 0 0
0 0 1 0
1
0 0
1
_
_
_
_
_
_
_
_
_
_
2
0 0
2
2
0 1 0 0
0 0 1 0
2
0 0
2
_
_
_
_
_
A four-vector is a four-dimensional vector that transforms like the vector x under
Lorentz transformations, that is, its components with respect to the two moving frames
S and S
= La , where a =
_
_
_
_
_
a
0
a
x
a
y
a
z
_
_
_
_
_
, a
=
_
_
_
_
_
a
0
a
x
a
y
a
z
_
_
_
_
_
(H.9)
For example, under the z-directed boost of Eq. (H.1), the four-vector a will transform as:
a
0
= (a
0
a
z
)
a
z
= (a
z
a
0
)
a
x
= a
x
a
y
= a
y
a
0
= (a
0
+a
z
)
a
z
= (a
z
+a
0
)
a
x
= a
x
a
y
= a
y
(H.10)
Four-vectors transforming according to Eq. (H.9) are referred to as contravariant.
Under the general Lorentz boost of Eq. (H.6), the spatial components of a that are trans-
verse to the direction of the velocity vector v remain unchanged, whereas the parallel
component transforms as in Eq. (H.10), that is, the most general Lorentz boost transfor-
mation for a four-vector takes the form:
a
0
= (a
0
a
|
)
a
|
= (a
|
a
0
)
a
= a
=
1
_
1
2
, = , =
v
c
(H.11)
1068 23. Appendices
where a
|
=
T
a and a = [a
x
, a
y
, a
z
]
T
is the spatial part of a. Then,
a
|
=
a
|
=
(
T
a) and a
= a a
|
= a
a
|
Setting =
0
a
_
=
_
_
I +( 1)
T
_
_
_
a
0
a
_
=
_
(a
0
a
|
)
a
a
|
+
(a
|
a
0
)
_
from which Eq. (H.11) follows.
For any two four-vectors a, b, the quadratic form a
T
Gb remains invariant under
Lorentz transformations, that is, a
T
Gb
= a
T
Gb, or,
a
0
b
0
a
= a
0
b
0
a b, where a =
_
a
0
a
_
, b =
_
b
0
b
_
(H.12)
Some examples of four-vectors are given in the following table:
four-vector a
0
a
x
a
y
a
z
time and space ct x y z
frequency and wavenumber /c k
x
k
y
k
z
energy and momentum E/c p
x
p
y
p
z
charge and current densities c J
x
J
y
J
z
scalar and vector potentials cA
x
cA
y
cA
z
(H.13)
For example, under the z-directed boost of Eq. (H.1), the frequency-wavenumber
transformation will be as follows:
= (ck
z
)
k
z
=
_
k
z
_
k
x
= k
x
k
y
= k
y
= (
+ck
z
)
k
z
=
_
k
z
+
c
_
k
x
= k
x
k
y
= k
y
, c = v,
c
=
v
c
2
(H.14)
where we rewrote the rst equations in terms of instead of /c. The change in
frequency due to motion is the basis of the Doppler effect. The invariance property
(H.12) applied to the space-time and frequency-wavenumber four-vectors reads:
= t k r (H.15)
This implies that a uniformplane wave remains a uniformplane wave in all reference
frames moving at a constant velocity relative to each other. Similarly, the charge and
current densities transform as follows:
c
= (c J
z
)
J
z
= (J
z
c)
J
x
= J
x
J
y
= J
y
c = (c
+J
z
)
J
z
= (J
z
+c
)
J
x
= J
x
J
y
= J
y
(H.16)
H. Lorentz Transformations 1069
Because Eq. (H.5) implies that L
T
= GLG, we are led to dene four-vectors that
transform according to L
T
. Such four-vectors are referred to as being covariant. Given
any contravariant 4-vector a, we dene its covariant version by a = Ga. This operation
simply reverses the sign of the spatial part of a:
a = Ga =
_
1 0
0 I
__
a
0
a
_
=
_
a
0
a
_
(H.17)
The vector a transforms as follows:
a
= Ga
= GLa = (GLG)(Ga)= L
T
a (H.18)
where we used the property that G
2
= I
4
, the 44 identity matrix. The most important
covariant vector is the four-dimensional gradient:
x
=
_
_
_
_
_
z
_
_
_
_
_
=
_
_
(H.19)
Because x
x
. Indeed, we have component-wise:
x
i
=
_
j
x
j
x
i
j
=
_
j
L
ji
j
x
= L
T
x
x
= L
T
x
For the z-directed boost of Eq. (H.1), we have L
T
= L
1
, which gives:
= (
+
z
)
z
= (
z
+
x
=
x
y
=
y
= (
z
)
z
= (
z
x
=
x
y
=
y
(H.20)
The four-dimensional divergence of a four-vector is a Lorentz scalar. For example,
denoting the current density four-vector by J = [c, J
x
, J
y
, J
z
]
T
, the charge conserva-
tion law involves the four-dimensional divergence:
t
+ J = [
,
x
,
y
,
z
]
_
_
_
_
_
c
J
x
J
y
J
z
_
_
_
_
_
=
T
x
J (H.21)
Under a Lorentz transformation, this remains invariant, and therefore, if it is zero
in one frame it will remain zero in all frames. Using
T
x
=
T
x
L, we have:
t
+ J =
T
x
J =
T
x
LJ =
x
J
=
t
(H.22)
Although many quantities in electromagnetism transform like four-vectors, such as
the space-time or the frequency-wavenumber vectors, the actual electromagnetic elds
do not. Rather, they transform like six-vectors or rank-2 antisymmetric tensors.
1070 23. Appendices
A rank-2 tensor is represented by a 44 matrix, say F. Its Lorentz transformation
properties are the same as the transformation of the product of a column and a row
four-vector, that is, F transforms like the quantity ab
T
, where a, b are column four-
vectors. This product transforms like a
b
T
= L(ab
T
)L
T
. Thus, a general second-rank
tensor transforms as follows:
F
= LFL
T
(H.23)
An antisymmetric rank-2 tensor F denes, and is completely dened by, two three-
dimensional vectors, say a = [a
x
, a
y
, a
z
]
T
and b = [b
x
, b
y
, b
z
]
T
. Its matrix form is:
F =
_
_
_
_
_
0 a
x
a
y
a
z
a
x
0 b
z
b
y
a
y
b
z
0 b
x
a
z
b
y
b
x
0
_
_
_
_
_
(H.24)
Given the tensor F, one may dene its covariant version through
F = GFG, and its
dual, denoted by
F and obtained by the replacements a b and b a, that is,
F =
_
_
_
_
_
0 a
x
a
y
a
z
a
x
0 b
z
b
y
a
y
b
z
0 b
x
a
z
b
y
b
x
0
_
_
_
_
_
,
F =
_
_
_
_
_
0 b
x
b
y
b
z
b
x
0 a
z
a
y
b
y
a
z
0 a
x
b
z
a
y
a
x
0
_
_
_
_
_
(H.25)
Thus,
F corresponds to the pair (a, b), and
F to (b, a). Their Lorentz transfor-
mation properties are:
= L
T
FL
1
,
F
= L
FL
T
(H.26)
Thus, the dual
F transforms like F itself. For the z-directed boost of Eq. (H.1), it
follows from (H.23) that the two vectors a, b transform as follows:
a
x
= (a
x
b
y
)
a
y
= (a
y
+b
x
)
a
z
= a
z
b
x
= (b
x
+a
y
)
b
y
= (b
y
a
x
)
b
z
= b
z
(H.27)
These are obtained by equating the expressions:
_
_
_
_
_
0 a
x
a
y
a
z
a
x
0 b
z
b
y
a
y
b
z
0 b
x
a
z
b
y
b
x
0
_
_
_
_
_
=
=
_
_
_
_
_
0 0
0 1 0 0
0 0 1 0
0 0
_
_
_
_
_
_
_
_
_
_
0 a
x
a
y
a
z
a
x
0 b
z
b
y
a
y
b
z
0 b
x
a
z
b
y
b
x
0
_
_
_
_
_
_
_
_
_
_
0 0
0 1 0 0
0 0 1 0
0 0
_
_
_
_
_
H. Lorentz Transformations 1071
More generally, under the boost transformation (H.6), it can be veried that the
components of a, b parallel and perpendicular to v transform as follows:
a
= (a
+ b
)
b
= (b
)
a
|
= a
|
b
|
= b
|
=
1
_
1
2
, = , =
v
c
(H.28)
Thus, in contrast to Eq. (H.11) for a four-vector, the parallel components remain un-
changed while the transverse components change. A pair of three-dimensional vectors
(a, b) transforming like Eq. (H.28) is referred to as a six-vector.
It is evident also that Eqs. (H.28) remain invariant under the duality transformation
a b and b a, which justies Eq. (H.26). Some examples of (a, b) six-vector pairs
dening an antisymmetric rank-2 tensor are as follows:
a b
E cB
c H
cP M
(H.29)
where P, M are the polarization and magnetization densities dened through the rela-
tionships =
0
E +P and B =
0
(H+M). Thus, the (E, B) and (, H) elds have the
following Lorentz transformation properties:
E
= (E
+c B
)
B
= (B
1
c
E
)
E
|
= E
|
B
|
= B
|
H
= (H
= (
+
1
c
H
)
H
|
= H
|
|
=
|
(H.30)
where we may replace c = v and /c = v/c
2
. Note that the two groups of equations
transform into each other under the usual duality transformations: E H, H E,
B, B . For the z-directed boost of Eq. (H.1), we have from Eq. (H.30):
E
x
= (E
x
cB
y
)
E
y
= (E
y
+cB
x
)
B
x
= (B
x
+
1
c
E
y
)
B
y
= (B
y
1
c
E
x
)
E
z
= E
z
B
z
= B
z
H
x
= (H
x
+cD
y
)
H
y
= (H
y
cD
x
)
D
x
= (D
x
1
c
H
y
)
D
y
= (D
y
+
1
c
H
x
)
H
z
= H
z
D
z
= D
z
(H.31)
1072 23. Appendices
Associated with a six-vector (a, b), there are two scalar invariants: the quantities
(a b) and (a a b b). Their invariance follows from Eq. (H.28). Thus, the scalars
(E B), (E E c
2
B B), ( H), (c
2
H H) remain invariant under Lorentz
transformations. In addition, it follows from (H.30) that the quantity (E B H) is
invariant.
Given a six-vector (a, b) and its dual (b, a), we may dene the following four-
dimensional current vectors that are dual to each other:
J =
_
a
b
a
_
,
J =
_
b
a
b
_
(H.32)
It can be shown that both J and
J transform as four-vectors under Lorentz trans-
formations, that is, J
= LJ and
J
= L
J, where J
,
J
frame:
J
=
_
_
,
J
=
_
_
(H.33)
The calculation is straightforward but tedious. For example, for the z-directed boost
(H.1), we may use Eqs. (H.20) and (H.27) and the identity
2
(1
2
)= 1 to show:
J
x
=
_
_
x
=
y
b
z
b
x
=
y
b
z
2
(
z
+
)(b
y
a
x
)
2
(
+
z
)(a
x
b
y
)
=
y
b
z
z
b
y
a
x
=
_
b
a
_
x
= J
x
Similarly, we have:
J
0
=
=
x
a
x
+
y
a
y
+
z
a
z
=
x
(a
x
b
y
)+
y
(a
y
+b
x
)+(
z
+
)a
z
=
_
(
x
a
x
+
y
a
y
+
z
a
z
)(
x
b
y
y
b
x
a
z
)
_
= (J
0
J
z
)
In this fashion, one can show that J and
J satisfy the Lorentz transformation equa-
tions (H.10) for a four-vector. To see the signicance of this result, we rewrite Maxwells
equations, with magnetic charge and current densities
m
, J
m
included, in the four-
dimensional forms:
_
c
H
c
_
=
_
c
J
_
,
_
cB
E
cB
_
=
_
c
m
J
m
_
(H.34)
Thus, applying the above result to the six-vector (c, H) and to the dual of (E, cB)
and assuming that the electric and magnetic current densities transform like four-
vectors, it follows that Maxwells equations remain invariant under Lorentz transfor-
mations, that is, they retain their form in the moving system:
_
_
=
_
c
_
,
_
cB
cB
_
=
_
c
m
J
m
_
(H.35)
I. MATLAB Functions 1073
The Lorentz transformation properties of the electromagnetic elds allow one to
solve problems involving moving media, such as the Doppler effect, reection and trans-
mission from moving boundaries, and so on. The main technique for solving such prob-
lems is to transform to the frame (here, S