Manuscript of the Lecture Course, W7147, University of Bern, Autumn Semester 2008
Deutscher Titel
Christian Mätzler
matzler@iap.unibe.ch
http://www.iap.unibe.ch
Downloads from
http://www.iapmw.unibe.ch/teaching/vorlesungen/remotesensing/
ii
Contents
1 Introduction.........................................................................................................................1
1.1 Remote Sensing ............................................................................................................1
1.2 Key requirements, and a dilemma .................................................................................2
1.3 Active and passive methods ..........................................................................................3
1.4 Examples ......................................................................................................................3
1.5 The significance of system models as a motivation .......................................................8
1.6 Conclusions ..................................................................................................................9
1.7 Literature ......................................................................................................................9
2 Electromagnetic Waves .....................................................................................................11
2.1 From Maxwell's Equations to the Wave Equation .......................................................11
2.2 Plane EM waves .........................................................................................................12
2.3 Polarisation of EM waves ...........................................................................................14
2.4 Interaction between EM waves and homogenous media ..............................................15
2.5 Kramers-Kronig relations, and the Hilbert Transform .................................................16
2.6 The electromagnetic spectrum.....................................................................................17
2.7 Literature ....................................................................................................................19
3 Sensors for EM Waves ......................................................................................................20
3.1 Antenna ......................................................................................................................20
3.2 Radar ..........................................................................................................................25
3.3 Radiometer .................................................................................................................28
3.4 Literature ....................................................................................................................30
4 Effective Medium, and Dielectric Mixing Formulas ..........................................................31
4.1 Maxwell-Garnett Formula...........................................................................................31
4.2 Semi-empirical mixing formulas .................................................................................34
4.3 Literature ....................................................................................................................35
5 EM Waves and Boundaries................................................................................................36
5.1 Boundary conditions ...................................................................................................36
5.2 The Fresnel Equations and Snell's Law of Refraction..................................................36
5.3 Waves in layered media ..............................................................................................43
5.4 Lorenz-Mie scattering .................................................................................................47
5.5 Rayleigh scattering .....................................................................................................51
5.6 Literature ....................................................................................................................53
6 Microscopic View of Matter ..............................................................................................54
6.1 Electric dipole, and polarisation of dielectric media ....................................................54
6.2 Types of polarisability ................................................................................................55
6.3 Electronic polarisation ................................................................................................57
6.4 Resonance absorption .................................................................................................58
6.5 Polar molecules in a static field...................................................................................59
6.6 Debye relaxation in polar liquids.................................................................................61
6.7 Space-charge polarisation ...........................................................................................63
6.8 Summary ....................................................................................................................63
6.9 Literature ....................................................................................................................64
7 Spectra of Matter that Matter .............................................................................................65
7.1 Recapitulation and Introduction ..................................................................................65
7.2 The earth atmosphere..................................................................................................66
iv
1 Introduction
1.1 Remote Sensing
Remote Sensing is understood as the collection of information relating to objects without
being in physical contact with them. Thus our eyes and ears are remote sensors, and the
same is true for cameras and microphones and for many instruments used for all kinds of
applications.
The term, Remote Sensing (Télédétection, in French, Fernerkundung, in German), arose
around the year 1900 when balloons (and later airplanes) became carriers of people to
altitudes well above the surface. These platforms allowed unprecedented views of the
environment, and especially of the earth surface. The bird's eye view enabled an accelerated
progress in Earth Sciences.
The impact of elevated platforms was most pronounced in areas with flat horizons where
natural viewpoints are missing. Therefore masts on ships are remote-sensing platforms. An
even less stable platform was used by Inuit hunters: They threw a man or a child up in the air
to search for seals. Thus "remote sensing" has been essential for survival.
Remote sensing originated from (1) human vision on special platforms, complemented by (2)
the recording of vision information, and (3) photogrammetry, the quantitative exploitation of
image records. Later the methods were extended to spectral ranges beyond the human eye.
For this purpose special sensors and instruments had to be developed. This process is still
ongoing.
Remote Sensing is not a scientific discipline in the classical sense; it is rather a collection of
a large variety of diagnostic methods, mainly using electromagnetic waves covering the
spectrum from radio waves (wavelength > 1 m) to gamma rays ( < 10-12 m). In some
cases sound waves or other elastic waves are also in use, especially where electromagnetic
methods fail. It is obvious that very different techniques and skills are required in the different
parts of remote sensing. Not only the techniques are multidisciplinary, the applications cover
a wide range of human disciplines, e.g. archaeology, botany, climatology, geology,
hydrology, meteorology, security aspects, etc.
2
2) Active methods (Figure 1.1: c) sense artificially produced waves after they interacted with
the objects to be sensed. Examples are radar, sonar, lidar, GPS, but also a photo camera
with a flashlight. An active method is called monostatic if transmitter and receiver are
collocated, otherwise the method is called bi-static, or even multi-static if several receivers at
different locations are used.
Figure 1.1:
Illustration of
examples with
passive (a, b) and
active (c) methods
in remote sensing
(from Schanda
1986).
1.4 Examples
Fig. 1.2: Partial panorama (direction SE) from the Zimmerwald Observatory (Dec. 12, 2006).
4
*Comment: The wave path is determined by the Principle of Fermat, which states that the
wave chooses the path with the fastest propagation.
Table 1.1: Speed of sound (m/s) in pure water and in sea water at P=0.1 MPa (sea surface)
and at 100 MPa (10 km depth), from http://www.akin.ru/spravka_eng/s_i_svel_e.htm
Temperature (C) Pure water Sea water S=3.5%
(surface, 10 km) (surface, 10 km)
0 1402, 1578 1449, 1623
10 1447, 1618 1490, 1659
20 1483, 1650 1522, 1687
30 1511, 1677 1546, 1710
http://modis.gsfc.nasa.gov/
Figure 1.5: Water-vapour column (g/cm2) above the alpine region as derived from MERIS on
ENVISAT. Note the disturbances in areas where clouds occur.
7
Figure 1.6:
European Network of
lightning stations
http://www.sferics.physik.uni-
muenchen.de/
Future re-analysis projects will encompass several components of the Earth System. In the
not-too-distant future, the Earth Science community will have the capacity to produce
coupled re-analyses of the Earth System, including weather, atmospheric composition, state
of the ocean, amount of moisture in the continental soils, hydrology of large rivers, and the
state of the biosphere and cryosphere. This, in turn, will open the way to an objective
verification of the predictional capacity of Earth System models. Special challenges of Earth-
System Science are listed in ESA (2006). For all these reasons it is important for remote
sensing to develop accurate and physical forward models that can be applied to the system
models for required comparison with observations.
The above statements are the motivation for following the present lecture and, well beyond,
for studying the physics of remote sensing.
9
1.6 Conclusions
This introduction gave an overview on the properties and principles of remote sensing. It was
stated that for any successful application, four requirements are to be fulfilled: (1) technical
and logistic feasibility (instrument and platform), (2) transparency of the propagating medium
from the object to be sensed to the observer, (3) interaction between the sensing wave and
the object, and (4) a signature, to allow the retrieval of the requested information from the
observed signal. It was found that a balance between transparency and interaction is needed
to get optimum results.
Several examples were discussed to illustrate the methods and the meaning of the 4
requirements. For certain tasks, several methods were identified, each of which having
different properties. Whereas satellite observations are optimal to unveil large-scale features,
surface-based observations are best to monitor dynamic processes.
The motivation for the physical approach is based on the required link with Earth-System
models allowing the application of data assimilation, leading to predictions and to a deeper
understanding of the Earth or its major components as complex systems.
In the further chapters we will concentrate on the physical properties of nature relevant to the
requirements for remote sensing, including an introduction to remote-sensing instruments.
More information can be found elsewhere, see e.g. the list of previous lectures:
http://www.iap.unibe.ch/content.php/teaching/. Scripts are also available at the ExWi library.
1.7 Literature
The following list is a small selection of monographs mainly from the inventory of the ExWi
Library (BEWI):
Introductory books
A booklet written by
E. Schanda (1986), Physical Fundamentals of Remote Sensing, BEWI: XKA 118 etc. is still very useful today.
Similar are:
W.G. Rees (2001), Physical Principles of Remote Sensing, BEWI: XKA 212, and
C. Elachi and J. van Zyl (2006), Introduction to the Physics and Techniques of Remote Sensing, 2nd Ed. BEWI:
XKA 214.
G.W. Petty (2006), A First Course in Atmospheric radiation, 2 nd Ed. Sundog Publishing, Madison, Wisconsin.
Excellent introduction also for the physics of remote sensing, BEWI: XJX 205
In-depth studies
A Manual of Remote Sensing has been published by the American Society of Photogrammetry. The two volumes
of the second edition from 1983 are available at our library BEWI: XKA 116, 117.
Ulaby, Moore and Fung (1981, 1982, 1986), Microwave Remote Sensing, Vol. 1,2,3 (BEWI: XKA 129, 130).
H. Sauvageot (1992), Radar Meteorology, BEWI: TEF 202. Lecture notes on this topic are also available, see
http://www.iapmw.unibe.ch/teaching/vorlesungen/radar_meteorologie/
C.L. Rodgers (2000), Inverse Methods for Atmospheric Sounding (BEWI: MAF 206).
Journals
Many journals deal with remote sensing, the following ones being fully dedicated to the topic:
1) IEEE Transaction on Geoscience and Remote Sensing (at BEWI)
2) Remote Sensing of Environment (at BEWI)
3) International Journal of Remote Sensing (library of the Geographical Institute)
Additional reference
T. J. Hewison and C. Gaffard (2006), Combining data from ground-based microwave radiometers and other
instruments in temperature and humidity profile retrievals, WMO Technical Conference on Meteorological and
Environmental Instruments and Methods of Observation, TECO-2006, Dec. 4-6, Geneva, Switzerland.
11
2 Electromagnetic Waves
2.1 From Maxwell's Equations to the Wave Equation
The electromagnetic (EM) fields, i.e. the electric field E, the displacement field D, the
magnetic field H, and the magnetic induction B, are governed by Maxwell's Equations (Kong,
1986):
D
H = + j; D = e (2.1)
t
B
E= ; B=0 (2.2)
t
where = , , is the Nabla Operator (here applied as rotation, , and divergence,
dx dy dz
), j is the electric current density and e the electric charge density. To understand how
EM waves propagate, we simplify the situation to homogeneous and isotropic media far
away from regions with sources (no transmitters, i.e. no isolated charges: e=0). We will
consider complex time-harmonic fields with the time dependence, exp(it) = cos(t) -
isin(t), where i is the imaginary unit, i = 1 , t is time and the angular frequency.
Complex fields are chosen as usual convention for easier computation. The physical fields
( )
are their real parts, e.g. Re eit = cos(t) . Based on Ohm's Law,
j=E (2.3)
a (complex) current density j will be excited in a conducting medium by the electric field E of
the wave, where is the conductivity of the medium. Now, Maxwell's Equations become
H = iD + E ; D = e = 0 (2.4)
E = iB ; B=0 (2.5)
We eliminate B and D by the linear Constitutive Relations
D = 0E + P = '0E and B = μ0H + M = μμ0H (2.6)
where ' is the (relative) dielectric constant (also called relative electric permittivity),
0=8.85410-12As/V/m the vacuum permittivity, μ the relative magnetic permeability, μ0=410-7
Vs/A/m the vacuum permeability, and P and M, respectively, are called electric and magnetic
polarisability. We can further simplify the right-hand side of the first equation in (2.4)
iD + E = (i'0 + )E to i0E , by defining a complex relative dielectric constant
= '+i" ; "= (2.7)
0
The real part is the original relative dielectric constant, and the imaginary part is related to
the conductivity by (2.7). The complex has a full physical meaning. Conductivity and
imaginary permittivity are different representations of the same effect. It turns out that also
the magnetic permeability can be complex μ = μ'+iμ" . The final form of Maxwell's Equations
for harmonic EM waves in homogeneous media then read as follows
H = iD = i0E ; E=D=0 (2.8)
E = +iB = +iμμ0H ; B=H =0 (2.9)
with the generalised constitutive relations (note the change of D which now includes E)
D = ( )0E and B = μ( )μ0H (2.10)
12
Remarks:
1) In chiral and in bi-isotropic media, also the electric and magnetic fields are linearly related
(Kong, 1986; Sihvola, 1999).
2) In anisotropic media and μ are tensors. Plane waves propagate independently for
special eigenmodes (dichroism).
1 c 0 2 1
2
c (0μμ0 ) = ; c0 = =2.99793108m/s (2.12)
n 0μ 0
It turns out, see Equation (2.32), that the real part of c is the phase velocity of the wave in the
medium, c0 is the speed of light in vacuum, and n is the refractive index of the medium:
n = μ ; for μ = 1 n = . (2.13)
Equation (2.16) allows to relate the amplitudes of the fields, defining the wave impedance
13
Problems
1) Proof Equation (2.21). Hint: Express E = (E'+iE")eit , H = (H'+iH")eit and note that
the real parts of these quantities are the physical fields. Express these parts, multiply them
and average over time to confirm (2.21). For further reading, see e.g. Schanda (1969), p. 30-
31.
2) Explain why the wave equation (2.11) does not apply for inhomogeneous media. Hint:
Show what additional terms appear in the derivation from (2.8) and (2.9) if depends on the
location.
14
x x
E(t=0)
E1
E(t=0) E(t)
t
z y z y
E2 E(t=/2)
Fig. 2.2a: E field at z=0 for E1=E2. The two Fig. 2.2b: E field at z=0 for real E1, E2=iE1
components oscillate in phase. Polarisation is at 3 different times. Polarisation (E) is
linear, but rotated by 45°. E is shown for t=0. rotating clockwise with time. This is called
circular polarisation.
Each component alone describes the E field of a linearly polarized wave. But together, the
situation is more complex. In Figure 2.2a, the two components oscillate in phase, and thus a
linear polarisation results again. In Figure 2.2b at t=0, only an x component exists, because
Ey is purely imaginary. With increasing time the physical y-component first increases as
E1 sin t , while the x component decreases as E1 cost . The two components oscillate with
a phase difference of 90°. This is the motion of a circle with constant radius E1.
In the general situation the polarisation is elliptical, that is, the tip of the E vector describes an
ellipse. The actual behaviour depends on the relationship between the complex amplitudes,
E1 and E2. The rotation can be clockwise or anti-clockwise.
Stokes Parameters
Instead of dealing with phase angles and complex numbers, an easier way to describe the
polarisation of a wave is by the Stokes Parameters I, Q, U, V . All parameters have the
dimension of an intensity, and they are defined by
1 2 2 1 2
I= Ex + Ey or I1 = Ex (2.24a)
2Z 0 2Z 0
1 2 2 1 2
Q= Ex Ey or I2 = Ey (2.24b)
2Z 0 2Z 0
1
U= Re[ E x E y* ] (2.24c)
Z0
1
V= Im[ E x E y* ] (2.24d)
Z0
Here the brackets mean averaging over time (usually many periods). Alternative
conventions are used for the quantities of (2.24a, b), where I1 and I2 are called modified
Stokes parameters. Note that Q and U depend on the coordinate system used, but the
degree of polarisation
15
Q2 + U 2 + V 2
p = (2.24e)
I
is independent of the coordinate system used. For unpolarised radiation p = 0 . Linear
polarisation is described by Q and U, whereas circular polarisation is described by V.
Problem: Express the Stokes Parameters for the 2 examples in Figure 2.2.
'+ '2 +"2 ', "<< ' " '2 +"2 '
n'= ; n"= = (2.26)
2 "/2, '<< " 2n' 2
The meanings of n' and n" become apparent from the properties of a plane wave propagating
parallel to an arbitrarily chosen r axis. Then
E = E 0 exp(ikr it) (2.27)
where k of (2.17) is complex
k = k'+ik"= ±nk 0 ; k'= ±n' k 0 , k"= ±n"k 0 ; k0 (2.28)
c0
Inserting (2.28) into (2.27), we get for the + sign: E = E 0 exp(ik' r it)exp(k"r) ;
The physical field (for real E0) is given by
Re(E) = E 0 cos(k' r t)exp(k"r) (2.29)
For the - sign we get
Re(E) = Re(E 0 exp(ik' r it)exp(k"r)) = E 0 cos(k' r + t)exp(k"r) (2.30)
Whereas (2.29) is a wave propagating in the positive r direction, (2.30) applies for waves in
the negative direction, and both waves are exponentially damped along their path (Figure
2.3). The damping is due to Ohmic currents, which transform the wave energy into heat. This
is called wave absorption. The distance,
1 c
ds = = 0 , (2.31a)
k" n"
after which the field is reduced by a factor e1 , is the field-penetration depth (or skin depth).
2
Since the wave intensity is proportional to E , its spatial variation is an exponential decay
exp(2k"r) . The damping coefficient, 2k", of the intensity is called absorption coefficient
16
Problem: Determine the wavelength and skin depth of a medium with = 2 + 0.3i , μ=1, at the
frequency =2.4 GHz.
+
1 '( ')
"( ) = PV d ' = Hi{'( ) } (2.34)
'
where PV means the Cauchy Principal Value of the integral (due to the singularity at '= ).
Equation (2.33) is also known as Hilbert Transform, Hi{ } , and (2.34) as inverse Hilbert
Transform (which is equal to the negative Hi Transform). They mean that if either the
complete real or the complete imaginary spectrum is known, the other spectrum follows
(apart from a constant high-frequency limit ) from the above relations. They are useful to
check the physical correctness of model functions and of experimental data. Similar
equations also hold for μ and n. Note that the Hilbert Transform of a constant is zero.
17
Therefore a constant 1 is subtracted from '. Some Hilbert Transforms are shown in
Table 2.1.
sin x cos x 1
x x
1 x
1+ x 2 1+ x 2
(x) 1
x
The fact that sin and cos functions are Hilbert Transforms allows us to formulate the following
rule: If the real part of the dielectric constant can be expressed by a Fourier series of cos
functions, then the imaginary part is the respective series of sin functions. This is always the
case because ( ) = * (*) ; thus '( ) is a symmetric function of the real frequency axis
whereas ''( ) is antisymmetric.
Fig. 2.4: The electromagnetic spectrum in terms of wave number, wavelength (both in
vacuum), photon energy and frequency ( f ), and some of the nomenclature of spectral
bands used in engineering (from Kong, 1986).
19
Visible spectrum
Fig. 2.5: Visible spectrum on a wall
created by sunlight after crossing a glass
prism.
The wavelength ranges (in nm) of the visible colours are, according to Petty (2006):
Ultraviolet (UV)
Extreme UV 10 – 100
UV- C 100 – 280 absorbed in mesosphere (>50 km) by O2
UV- B 280 – 320 reduced by O3, responsible for sun burn
UV- A 320 – 390 99% of solar UV at sea level, not dangerous for living tissue
2.7 Literature
R. Bracewell (1965), The Fourier Transform and its Applications, New York, BEWI: GQE 119.
D. J. Griffiths, Introduction to Electrodynamics, (1999) 3rd Ed., BEWI: OGA 151. Good introductory book, includes
introduction to mathematical concepts (e.g. vector analysis.
A. von Hippel, Dielectrics and Waves, 1st Ed. (1954) BEWI: TEA 149, 2nd Ed. (1995) BEWI: VTZ 201.
J.A. Kong, Electromagnetic Wave Theory, New York 1st Ed. (1986) BEWI: TEA 150, 2nd Ed. (1990) ETH Library.
G.W. Petty (2006), A First Course in Atmospheric radiation, 2 nd Ed. Sundog Publishing, Madison, Wisconsin.
Excellent introduction also for the physics of remote sensing, BEWI: XJX 205
In German:
G. Eder, Elektrodynamik, BI Hochschultaschenbuch 233 (1961), BEWI: OGA 130.
E. Schanda (Ed.), Theorie der elektromagnetischen Wellen, Birkhäuser Verlag, Basel (1969), BEWI: OGF 120,
TEA 152, 156. Dieses Buch kann als kompakte Einführung in Vektoranalysis (Kapitel von H. Carnal), ins
Verständnis der Maxwellgleichungen (Elektrizitätslehre und die Maxwellsche Theorie von E. Schanda) und
Wellen, Antennen (weitere Kapitel) empfohlen werden.
20
3.1 Antenna
Transmitters and receivers need elements that allow the transition between the propagating
radiation in free space and the guided radiation in the sensor, and vice versa. The transition
is realised by the antenna. This term is derived from Latin for 'sail' in analogy to wind and
surface waves and is well known for radio- and microwaves. Although not often used in other
domains, the term, antenna, is of relevance to other wavelength ranges, but may be hard to
realise. Special to the antenna is that a single, polarised wave mode is exited from a given
feed point, meaning that there is an unambiguous phase and field relationship between the
feed point of the antenna and any point in space. Thus antennas radiate fully polarised
radiation. Incoherent radiators and detectors, on the other hand, cannot provide this property.
The directivity D(, ) describes the directional distribution of the radiated power P1. In the
far field (Figure 3.2), that is at distances r larger than
2d 2
r> (3.1)
where d is the maximum diameter of the antenna, the radiation intensity S(, ) is
proportional to 1/r 2 . Therefore we can define a quantity that is independent of distance by
S(, )
D(, ) = 4 r 2 (3.2)
P1
22
Here D(, ) is the directivity in direction (, ) . Integration of S(, ) over a sphere with
radius r around the antenna must give the total radiated power P1. Therefore the integral over
directivity must give
2
Also used is the antenna gain G(, ) . It is similar and proportional to D(, ) , with the
difference that in (3.2) the radiated power in the denominator is replaced by the total power
Pin fed into the antenna. Thus,
G(, ) = r D(, ) (3.4)
where r = P1 /Pin is the radiation efficiency of the antenna. Ideal antennas are lossless,
meaning that gain and directivity are the same.
The quantity shown in Figure 3.3 is the directivity normalised to the maximum
p= D(, ) / Dmax (, ) and expressed in decibel (dB).
Special cases:
1) Isotropic antenna: D is independent of direction. Then, with the requirement in Equation
(3.3) we find that D=1. Small antennas in comparison to the wavelength are nearly isotropic.
Therefore their directivity is always on the order of 1.
1200
Figure 3.4: Directivity of a boxcar
1000
antenna with Dmax =1000. From
Equations (3.5) and (3.6) we get
800 max = 0.0632 , corresponding to a full
beam width of 7.2°.
600
D
400
200
2) Boxcar antenna: D=Dmax inside the antenna beam defined by a solid angle e, and D=0
elsewhere. Insertion into Equation (3.3) leads to
Dmax = 4 /e (3.5)
This expression relates the solid angle of the antenna beam with the directivity. Therefore
Dmax gets very large for narrow beams. For a circular-symmetric beam around the pole =0,
the effective solid angle becomes
max
e = 2 sind = 2 (1 cos max
2
) max (3.6)
0
where the last equation is valid for narrow beams ( max << 1). Note that the beam diameter is
given by = 2 max .
In analogy to Special Case 2, an effective solid angle e can be defined for any antenna by
23
1 4 1
b = D(, )d
4
e = Dd = Dmax ; (3.7)
Dmax 4 e
The second equation defines the fraction b of the radiation transmitted in directions within
the beam e ; b is called beam efficiency. The boxcar antenna described above is ideal with
b =1. High beam efficiency is important in remote sensing to relate signals to a well-defined
direction and position.
A wave is radiated through the aperture of a horn. At the horn aperture the radiation is
diffracted. From scalar diffraction theory the directivity can be related to the electric field in
the aperture plane:
2
4
E(x, y)exp(ik x + ik y)dxdy
x y
A
D(, ) = 2 2 (3.10)
E(x, y) dxdy
A
where the nominator contains the diffraction integral over the aperture area A. This integral is
the two-dimensional Fourier Transform of the aperture field E(x,y,z=0) if E is set to 0 outside
the aperture. The denominator of Equation (3.10) represents the radiated power passing
24
through the aperture. Using spherical coordinates of Figure 3.2 we have for the wave vector
components kx and ky perpendicular to the z direction:
kx = k sin cos , ky = k sin sin , k = 2 / (3.11)
Example: Let E be the scalar electric field amplitude of a plane wave in z direction, and let
x = y = a , and A = a 2 . Then E is a constant within the aperture, leading to
2 +a / 2 2 +a / 2 2
4 4
D(, ) = 2
A
exp(ikx x + iky y)dxdy = A2 exp(ik x)dx exp(ik y)dy
x y , and thus to
A a / 2 a / 2
Example: The radiation pattern for Dmax = 1000 is realised if the antenna size a = 8.9206 .
Inserting this value in Equation (3.12) for = 0 , we get X=28.025sin. The directivity is
shown in Figure 3.6.
Figure 3.6: Directivity in the plane = 0 of a square antenna with homogeneous aperture
field for Dmax=1000.
Problem: Compute and plot the directivity of this antenna for = 45° . Furthermore compute
e and b (limit integration to 0<<90°), and compare the results with the boxcar antenna of
Section 3.1.2.
Comments: The presented diffraction model is limited to radiation in the forward direction,
and we must assume that no radiation is transmitted in the backward hemisphere
(180>>90°). We must also assume that the far-field condition of Equation (3.1) applies. This
latter condition can be relaxed by quasi optics (see Lecture Microwave Physics and Quasi
Optics). Polarisation has been neglected here, but must be included when needed.
25
3.2 Radar
Transmitter
scatterer
scattered wave
Figure 3.7: Bistatic radar configuration with separate transmitter and receiver.
We distinguish the following configurations:
• Monostatic Radar (collocated transmitter and receiver, most common type)
• Bistatic Radar (as in Figure 3.7)
• Multistatic Radar (using more than one receiver)
In order to locate a wave train it is necessary to modulate the transmitted wave, either by
creating sufficiently short pulses of duration in the μs (10-6s) or even in the ns (10-9s)
range, or by a (linear) frequency modulation. With such means the total travel time t = l /c
(path length l , speed of light c) between transmitter and receiver can be measured to an
accuracy (error t ), which is limited by the uncertainty relation
t 1 (3.13)
where is the frequency bandwidth of the radar signal. The equal sign can be reached if a
so-called matched filter is used (Ulaby et al. 1982). From Equation (3.13) we get the path-
length error
l = c t c / (3.14)
In monostatic radar the path length is two times the range, r (distance between transmitter
and scatterer); then
r = 0.5c t 0.5c / (3.15)
Furthermore radar types distinguish between coherent and incoherent systems, types with a
single polarisation for both transmission and reception, with multiple polarisation, and fully
26
polarimetric radars that can transmit any state of polarisation (transmit Stokes Parameter)
and detect any polarisation upon reception (receive Stokes Parameter). A coherent radar
measures the phase relationship between transmit and receive signals (Figure 3.8) to
determine travel time t. For this purpose a very stable oscillator is required especially for long
propagation paths. In non-coherent radars t is measured with a clock triggered by the
transmit pulse.
Oscillator Antenna
U1
It
U2
Frequency Qt
Control
90°
In a coherent monostatic radar, the phase difference 2kr between a transmitted voltage
U1 = U 0 cos(t) (3.16)
(with the circular frequency = 2 ) and the received echo voltage U2 after travelling the
distance 2r is to be measured
U2 = L U 0 cos(t 2kr) (3.17)
where L << 1 describes the signal loss during propagation. The detection is accomplished
with a so-called I-Q mixer. Its outputs consist of It = U1 U 2 and Qt, which is the same
product, but after a phase delay in U1 of 90°. The low-frequency parts of these products are
registered. With the frequency control, the signal can be modulated, for instance a linear
sweep over a time T, (=TR/2 in Figure 3.9).
(t) = 0 + at , for 0 < t < T (3.18)
This sweep has a modulation bandwidth ( = f m in Figure 3.9) :
= aT /(2 ) (3.19)
It and Qt contain the difference frequency between transmission and reception (Figure 3.9):
f = 2ar /c (3.20)
Radar wind profiler (surface based): Measures Doppler velocity of the turbulent atmosphere
to determine horizontal wind velocity.
Altimeter: Radar operated from satellites and aircraft to measure the altitude above the
surface, but also to observe surface properties, especially over the ocean (wind, waves,
currents).
LIDAR is the optical equivalent of radar. It is used to measure the surface shape with high
spatial resolution, and for atmospheric sensing of water vapour, aerosols, clouds, wind, etc.
Synthetic Aperture Radar (SAR) operated from satellite and aircraft to image the backscatter
with high spatial resolution (a few m). See lecture notes on ENVISAT (Mätzler, 1998), SAR
Basics from Gamma Remote Sensing (2008).
3.3 Radiometer
The radiometer collects thermal radiation from an object at temperature T. Part of this
radiation passes through the filter into the absorbing box. If we wait until thermal equilibrium
has been reached, the measured temperature inside the box will be T. The radiance can
then be quantified by one half (one polarisation only) of the Planck function for temperature T
at the specified frequency. See Chapter 8. This type of radiometer is called a bolometer,
because the detection uses a thermal property of the radiation.
Often the radiative heating power is too weak, requiring other detection methods.
TA U
TD
Antenna Dicke Switch Filter Isolator Amplifier synch. Power Det Integrator
Figure 3.12: Block diagram of a Dicke radiometer using synchronous switching and detecting
between the signals of the antenna (TA) and of a reference (Dicke load at temperature TD).
The following figure shows a set of Dicke radiometers constructed and operated at the IAP:
More sophisticated IAP radiometers are operated on the ground for dedicated observations:
MIAWARA: spectrometer for water-vapour profiling of the middle atmosphere (altitude range:
20 to 70 km)
ASMUWARA: for all-sky scanning of the troposphere for clouds, water vapour and
temperature profiling
TROWARA: for monitoring integrated water vapour and cloud liquid water of the atmosphere.
Imaging radiometers have been operated from satellites for global earth observations for
more than 30 years; consult literature and internet for SSMR, AMSU-A, AMSU-B, SSM/I,
AMSR, Windsat Radiometer, and others. An example is shown by the cover image.
3.4 Literature
Gamma Remote Sensing, SAR Basics, Documentation Theory, Version 1.4, Gümligen, Schweiz (2008).
Mätzler C. ENVISAT: der neue Europäische Fernerkundungssatellit für die Umweltforschung Universität Bern,
1998, BEWI SK1 PH 1998: 1.
Ulaby, Moore and Fung (1981, 1982, 1986), Microwave Remote Sensing, Vol. 1,2,3 (BEWI: XKA 129, 130).
Information about IAP radiometers: s. IAP Homepage, Microwave Physics: Research Projects and Publications.
31
From (4.3) or (4.3a) it appears that the problem of the mixing formula is solved if we find the
correct value of K or Kj. It turns out that the K factors mainly depend on particle shape. In
general K has to be determined either experimentally or theoretically. We will give
expressions for some examples.
32
Sphere
For a single sphere in an infinite host volume the electrical field E2 inside the sphere is
homogeneous and can be expressed by (4.2) where K is given by the electrostatic
expression (4.4).
31
K= (4.4)
2 + 21
Ellipsoid
Electrostatic field expressions exist also for ellipsoids. For an ellipsoid in an infinite host
volume with the electrical field parallel to a principal axis, the electrical field E2 inside the
spheroid is again homogeneous and can be expressed by (4.2), where K now depends on
the axis ( i = a,b,c ) along the E field, and is given by
1
Ki = (4.6)
1 + Ai (2 1 )
and Ai is called the depolarisation factor of the ellipsoid along the i-axis. These factors can be
computed from an elliptical integral. For i = a we have
abc ds
Aa = (4.7)
2 0 (s + a )
2
(s + a 2 )(s + b 2 )(s + c 2 )
and for i = b, c the integral follows by respective replacements. In the integral a, b, c are
the 3 semi axes of the ellipsoid. The sum of all factors equals one:
+
Aa + Ab + Ac = 1 (4.8)
Problem: Show, by integrating (4.7) for the special case of a sphere ( a = b = c ), that A = 1/3 .
33
Spheroid
In the case of spheroids (ellipsoids with one axis of symmetry), the integral (4.7) can be
solved analytically. The results are given in Table 4.1, and the values for Ab = Ac are plotted
versus the axial ratio in Figure 4.2.
Tab. 4.1: Depolarisation factors of spheroids with eccentricity e and semi axes a, b = c .
Type of spheroid Aa Ab = Ac
a 1 1 e 2 1 Aa
Oblate b=c = arcsine
1 e 2
e2 e3 2
Prolate b = c = a 1 e
2 1 e 2 1 1+ e 1 Aa
ln 1
e 2 2e 1 e 2
1
K= (K a + K b + K c ) (4.11)
3
has to be inserted in Equation (4.3).
2
Now since for non-magnetic media, = n 2 and i = n i for all i = 1,... N , the refractive mixing
formula reads, in terms of the dielectric constant
35
N 2 N
= f i i ; where f i =1 (4.14)
i=1 i=1
A physical motivation of the refractive formula is based on the fact that the real part n' of the
refractive index is proportional to the propagation time of waves travelling through the
material. An average propagation time in a mixed material can be represented by an average
n' as expressed by (4.13), see Figure 4.3. The same reasoning applies to the imaginary part
n". The average absorption of the wave can be expressed by the average n".
Problem: Show that the same result is obtained for the imaginary part n", again assuming
that the wave propagates through the three volumes along path s from left to right.
The imagination of a wave travelling first through one particle followed by another particle
implies a model such as geometrical optics where the particles are much larger than the
wavelength. Nevertheless, (4.14) is also used in situations with particles being smaller than
the wavelength. Note that the refractive model fails if the propagation direction is vertical in
the example of Figure 4.3. It only works in a serial way.
A generalisation
A generalisation of (4.14) is the following form
N
1/ N
= f i (i ) ; where f i =1 (4.15)
i=1 i=1
The refractive formula is represented by =0.5, the linear formula by =1, and a physically-
based formula for spherical particles called after Looyenga (1965) by =1/3.
4.3 Literature
H. Looyenga, "Dielectric constants of heterogeneous mixtures", Physica 31, 401-406 (1965).
J.C. Maxwell Garnett, "Colors in metal glasses and metal films", Trans. Royal Soc. (London), CCIII, pp. 385-420
(1904).
A. Sihvola, Electromagnetic mixing formulas and applications, IEE Electromagnetic Waves Series 47, London UK
(1999), BEWI: OGE 201.
36
Maxwell's Equations lead to the following boundary conditions for the fields at the interface:
1) Tangential components of the E fields are the same on both sides of the interface:
E1 = E 2 (5.1a)
2) Tangential components of the H fields are also the same:
H1 = H 2 (5.1b)
3) Normal components of B are the same on both sides of the interface:
μ1H1 = μ2 H 2 (5.1c)
4) Normal components of D are also the same:
1 E1 = 2 E 2 (5.1d)
Exceptions occur if one medium has infinite conductivity: then electrical surface charges (for
D) and surface currents (for H) must be taken into account (Kong, 1986).
incidence. The orthogonal situation to Figure 5.1 is called vertical polarisation, also called
transverse magnetic (TM) or parallel. It is obtained by exchanging the E and H fields, more
exactly by replacing E by H and H by –E. This procedure follows from the principle of duality
(Kong, 1986). A general EM wave can be decomposed into a fraction in vertical polarisation
and another fraction in horizontal polarisation.
Based on Figure 5.1 the electrical fields of the 3 waves only have y components, which can
be written as
E i = E 0 exp(ik1x x ik1z z it) (5.2a)
E r = Rh E 0 exp(ik1x x + ik1z z it) (5.2b)
E t = Th E 0 exp(ik2x x ik 2z z it) (5.2c)
where k1x = k1 sin 1; k1z = k1 cos1; k 2x = k2 sin 2 ; k 2z = k2 cos 2 (5.3)
and Rh and Th are reflection and transmission coefficients at horizontal polarisation to be
determined. Note that the incident and reflected waves have the same value of k1x and
opposite values of k1z. This choice is necessary to get a solution. It means that the reflection
angle is equal to the incidence angle 1.
The total field on the upper side of the interface at z=0 consists of the sum Ei+Er, and on the
lower side of Et: The two are required to be equal according to (5.1a), giving
(1+ Rh )exp(ik1 x sin 1 ) = Th exp(ik2 x sin 2 ) (5.4)
This equation can only be true for all x if the following equations are both valid
1+ Rh = Th (5.5)
k1 sin 1 = k 2 sin 2 (5.6)
38
n i
and since ki = ; i = 1,2 , Equation (5.6) leads to Snell's Law of refraction (5.7). The law
c0
determines the direction (angle 2) of the refracted wave, given the refractive indices and 1:
n1 sin 1 = n 2 sin 2 (5.7)
Figure 5.2:
Graphical demonstrations of
Snell's Law of refraction (a), and
of the law of reflection (b) for a
water surface. The wave is
incident from medium 1, here
called medium i, transmitted in
medium 2, here called medium t.
From Petty (2006).
In remote sensing, Medium 1 is usually the atmosphere where n1 is very close to 1, and n2 is
the refractive index of the surface material. Note that if n2 is complex, Equation (5.7) tells that
the refracted angle 2 is complex, too.
To solve for Rh and Th, a second equation is needed, which is found from Equation (5.1b) for
the case of Figure 5.1 and from the relation (2.18) between the E and H fields:
E0 E0
H0 = . This leads to
Z Z0 μ
1
(1 Rh )cos 1 = 2 Th cos 2 (5.8)
μ1 μ2
Solving Equations (5.5) and (5.8) for Rh gives the Fresnel Equation for h polarisation:
1 2
cos1 cos 2
μ1 μ2 Z cos1 Z1 cos 2
Rh = = 2 ; Th = 1+ Rh ; (5.9)
1 2 Z 2 cos1 + Z1 cos 2
cos1 + cos 2
μ1 μ2
39
In analogy to the derivation found above, the following results are obtained for vertical
polarisation:
μ1 μ2
cos1 cos 2
1 2 Z cos1 Z 2 cos 2
Rv = = 1 ; Tv = 1 Rv ; (5.10)
μ1 μ2 Z1 cos1 + Z 2 cos 2
cos1 + cos 2
1 2
For non-magnetic media, the reflection coefficients of Equations (5.9) and (5.10) can be
expressed by the more familiar refractive indices:
n1 cos1 n 2 cos 2 n 2 cos1 n1 cos 2
Rh = ; Rv = (5.11)
n1 cos1 + n 2 cos 2 n 2 cos1 + n1 cos 2
Although the two formulas look very similar, the results are different. The Fresnel Formulas
can be expressed in various ways, e.g. by eliminating 2 by Snell's law. For cos 2 we can
write, using (5.7),
n12
cos 2 = 1 sin 2 2 = 1 2
sin 2 1 (5.12)
n2
Equation (5.12) also helps to understand the meaning of the complex angle.
The fraction of reflected power (or intensity) is called reflectivity, or reflectance, denoted by rh
and rv. Furthermore, the fraction of transmitted power is called transmissivity, or
transmittance, denoted by th and tv. Since the intensity is related to the fields by (2.22):
2
I = 0.5 E 0 Re(1/Z *) , the rh and rv are simply given by
2 2
rh = Rh ; rv = Rv (5.13)
The transmissivities have to include the change in Z from Medium 1 to Medium 2. But since
energy is conserved, the transmissivities also follow from
t h = 1 rh ; t v = 1 rv (5.14)
Of special interest is the Brewster Effect, a situation with completely vanishing reflection for
real n1 and n2. By inserting (5.12) into (5.11) it can be confirmed that for nonmagnetic
materials, the Brewster Angle occurs at vertical polarisation, and it is given by
Brewster = arctan(n 2 /n1 ) (5.15)
For a medium with an imaginary refractive index n2, the reflectivity is a minimum at the
Brewster angle defined by the real part of n2 in (5.15).
Total reflection is another phenomenon to be mentioned. This occurs for upwelling waves in
Medium 2 at the interface if 2 is larger than a critical value 2,c with
n1
sin 2,c = (5.16)
n2
Then there is no real solution for 1 according to Snell's law. The only solution is total
reflection back to Medium 2.
40
Numerical computations
We will use MATLAB to illustrate results. A general function for non-magnetic media is
fresnel.m; it computes all quantities defined above, and in addition a few more. Some more
functions are available as explained below.
Examples
For small angles of incidence, the Fresnel reflectivities show small angular variation. The
behaviour changes towards large angles, always leading to r h =rv =1 at grazing incidence
(2=90°). The difference between small and large dielectric constants can be seen in Figure
5.4 showing the reflectivity of water at 22 GHz and in the visible range. Note that in the
microwave range the reflectivity is nonzero at the Brewster minimum because of the large
imaginary dielectric constant.
42
In remote sensing the incidence angle is often not a freely selectable parameter because it is
either a fixed value (for conically scanning sensors), or else it changes with the scan angle
and thus with the position of the observation point. Therefore it may be useful to see what
type of information can be obtained from the measurement of the reflectivity at h and v
polarisation, by assuming a fixed value of 1.
Figure 5.5 shows the behaviour at the incidence angle of 57° (1 radian). Both Fresnel curves
(x and o) are very similar, indicating that the reflectivities are unable to give independent
information on the real and imaginary parts of . This property is useful as signature of
Fresnel surfaces. Rough surfaces, on the other hand, are characterised by very small
polarisation, i.e. by rv = rh, as represented by the dashed curve.
Figure 5.6: Geometry of the layered medium with n( j) = n j '+in j " the complex refractive
index of Layer j and d(j) its thickness.
Recursion formula
The reflection and transmission coefficients (R and T), and the reflectivity and transmissivity
of a multi-layer sandwich can be computed either by a matrix method (Dobrowolski, 1995,
Born and Wolf, 1975), or by a recursive method called invariant embedding (Adams and
Denman, 1966). The recursion formula will be presented here. However, note that for m>>1
the matrix method is much faster.
Let us consider (Figure 5.6) a multi-layer medium with plane-parallel boundaries (planes with
constant z). The bottom medium has Number 1 with propagation angle 1 to the z axis,
relative dielectric constant 1 and relative magnetic permeability μ1. Furthermore we call F1
the Fresnel reflection coefficient and G1 the transmission coefficient at the boundary from
Medium 2 to Medium 1. Medium 2 is a layer of thickness d2 above Medium 1.
For Medium 2, we define the respective parameters as d2, 2, 2, μ2, F2 and G2, and so on for
all layers to the uppermost layer (No. m) of the sandwich with parameters, dm, m, m, μm, Fm
and Gm. An EM wave is incident from the halfspace above the layered medium with
parameters m+1, m+1, μm+1. Now the recursive formulas read for j=2,... m
2
F j + R j1Q j G j T j1Q j
Rj = ; Tj = ; j=2, 3, ..., m (5.17)
1+ F j R j1Q j 2
1+ F j R j1Q j 2
and for j=1, the values are simply given by
R1 = F1; T1 = G1 (5.18)
44
The factor Qj describes the complex phase change of the wave on its way through Layer j:
Q j = exp(ik j d j cos j ) (5.19)
We have to remember that Fj and Gj depend on polarisation. Then the final reflection and
transmission coefficients R and T at the top of the layer are the values of (5.17) for j=m, but
now indicating also the polarisation (h, and v)
Rv,h = Rm ; Tv,h = Tm (5.20)
Part of the wave intensity is reflected, another part is transmitted to the bottom of the
sandwich, and the remaining intensity is absorbed by the sandwich. The respective fractions
are the reflectivity r, transmissivity t, and absorptivity a at v and h polarisation, respectively,
of the sandwich, and they are determined from (assuming that the top and bottom media are
identical):
2 2
rv,h = Rv,h ; t v,h = Tv,h ; av,h = 1 rv,h t v,h (5.21)
Matlab functions
function result = layers(freq, thetai, e, d)
% General multilayer program using recursive formula of R.N.Adams,
% & E.D.Denman, Wave Propagation and Turbulent Media,
% American Elsevier, New York, 1966
% Input parameters:
% freq: frequency [GHz]
% thetai [rad]: Incidence angle of incident beam (top at j=m+1)
% e: vector with complex dielectric constants
% d: vector with thicknesses [m] of layers
% d and e are both of length m+1 (dummy values for d at j=1 and j=m+1)
% Lorenz Martin, 2005.
Problem
Show that for n j "<< n j ' , the one-way phase k j d j cos j of Q j can be written as
in j "
k0 n j ' d j cos j,eff + where j,eff is the real refracted angle resulting for a layer with
cos j,eff
real refractive index n j ' , i.e. for n j "=0. Use Snell's Law of refraction.
Note that k j = n j k 0 where k0 = /c 0 is the vacuum wave number, and c0 is the speed of light
in vacuum.
Thus the real part of the phase through the layer decreases with increasing incidence angle,
whereas the opposite is true for the imaginary part.
Examples
1) Reflectivity of a frozen lake at a frequency of 18 GHz: Ice layer (2=3.188 + 0.0017i) on top
of water (1=21.7 + 32.5i). The following example was computed for an ice thickness of 20
cm. With increasing , the phase factor between the ice and water surface decreases, a
strong interference phenomenon appears. It disappears at the Brewster angle of ice near 60°
(v pol). The method could be used to measure the lake-ice thickness.
2) If a surface has a gradual transition layer from air to soil, the dielectric constant changes
over a certain depth range from air (=1) to soil. The reflectivity of such a transition can be
modelled by a number of layers with gradually changing . We want to find out under which
condition this effect is relevant. For simplicity we represent the transition region by a single
layer of thickness d, thus m=2, and we assume loss-less media. Then 1 is the real dielectric
constant of the bottom halfspace, 2 is the real dielectric constant of the transition layer and d
is its thickness. The upper halfspace has 3=1. The value of 2 is intermediate: 1< 2 < 1.
From (5.17-9) for m=2 we get directly the solution
2
F + F1e i2
2
r= R = 2 ; k2 d cos 2 (5.22)
1+ F2 F1e i2
For loss-less (and approximately also for low-loss) media, the Fresnel reflection coefficients
are real. Then we get, after some transformations
46
1 2 sin 2 2 F1F2
r = r0 2 2
; (5.23)
1 r0 sin F1 + F2
where r0 is the Fresnel reflectivity without the transition layer (i.e. from 3 to 1), and is the
ratio of the geometrical to the arithmetic mean values of F1 and F2. If F1 = F2, =1, otherwise
<1. The reflectivity r changes with phase angle . For =1, r can be suppressed completely
for sin=1. This happens, for instance, if n 2 d cos 2 = (1/4) 0 .
The variation r /r0 with increasing layer thickness, expressed by , is shown in Figure 5.8 for
3 values of r0, all for =1. For <10° the reduction of r with respect to r0 is negligible.
However, for =30° ( d = /12 ) the reduction is already significant, especially if r0 < 0.4. The
reduction leads to zero reflection at 90°, that is when d is a quarter wavelength. This effect is
used in optics to suppress reflections of glasses and lenses by anti-reflection coatings.
With still higher values of r0, the reflectivity in Figure 5.8 approaches the properties of a
Fabry-Perot resonator.
5.4.1 Introduction
Scattering of electromagnetic waves at homogeneous and layered spheres can be computed
in the analogous way as we did for plane surfaces and plane layers, i.e. by fitting the fields of
the incident plane wave to the sum of a scattered wave and internal fields at the surface
boundaries of the sphere. The method is complicated by the fact that the boundary
conditions cannot easily be expressed when plane waves meet a sphere. The main task is to
express a plane wave by a superposition of spherical waves. An excellent description can be
found in the textbook of Bohren and Huffman (1983), in short BH. The theory was developed
by Ludvig V. Lorenz in 1890 and by Gustav Mie (1867-1957) in 1908. Here the basic results
are presented, making use of BH and MATLAB functions developed and described in IAP
reports and lecture notes Mätzler (2002-2004). For scattering by coated spheres, and for
more details on the physics and numerical problems, see the mentioned references.
exp(ikr it)
A spherical wave is described by . BH use the convention of Equation (5.24)
kr
for the scattered electric far field components, where the time factor is omitted, and where
parallel and perpendicular field components are defined with respect to the scattering plane
as shown in Figure 5.8.
E s S S E i exp(ikr)
= 2 3 (5.24)
E s S4 S1 E i ikr
x2 Ei Es
Es
s
x3 , i scatterer
x1 , Ei
Figure 5.8: Scattering geometry with incident fields on the left scattered by a sphere in (x2, x3)
plane showing parallel and perpendicular components of the incident (Index i) and scattered
(Index s) electric fields, and scattering angle .
In general the scattering matrix is composed of 4 complex numbers Sj. In case of spheres,
for reasons of symmetry, only S1 and S2 are different from 0. They turn out to be given by
2n + 1
S1 (cos ) = (an n + bn n );
n=1
n(n + 1)
(5.25)
2n + 1
S2 (cos ) = (an n + bn n )
n=1
n(n + 1)
The functions n and n describe the angular scattering patterns of the spherical harmonics
used to describe S1 and S2 and follow from the recurrence relations
2n 1 n
n = cos n1 n2 ; n = n cos n (n + 1) n1 (5.26)
n 1 n 1
starting with
0 = 0; 1 = 1; 2 = 3cos ; 0 = 0; 1 = cos ; 2 = 3cos(2 ) (5.27)
48
The elements Si are scattering amplitudes; they describe the angular and polarisation
dependence of the scattered wave.
jn ( z ) = J n + 0.5 ( z ) (5.31)
2z
yn ( z ) = Yn + 0.5 ( z ) (5.32)
2z
Here, J and Y are Bessel functions of the First and Second Kind; for n=0 and 1 the
spherical Bessel functions are given (BH, p. 87) by
j0 ( z ) = sin z / z; j1 ( z ) = sin z / z 2 cos z / z
(5.33)
y0 ( z ) = cos z / z; y1 ( z ) = cos z / z 2 sin z / z
and the recurrence formula can be used to obtain higher orders
2n + 1
f n 1 ( z ) + f n +1 ( z ) = fn ( z) (5.34)
z
where fn is any of the functions jn and y n. Power-series expansions for small arguments of jn
and yn are given on p. 130 of BH. The Spherical Hankel functions are linear combinations of
jn and yn. Here, the first type is required
hn(1) ( z ) = jn ( z ) + iyn ( z ) (5.35)
49
To describe the internal field we need additional coefficients as described in BH; see also the
lecture notes on Strahlungstransport (Radiative Transfer) with MATLAB functions and many
examples.
Qa = Qe Qs (5.38)
The infinite series are truncated after nmax terms. Furthermore, the asymmetry parameter
g= cos indicates the average cosine of the scattering angle with respect to power; it is
used in radiative transfer, and it is related to the efficiency Qpr of radiation pressure:
Qpr = Qe Qs cos (5.39)
4 n(n + 2)
2n + 1
2
* *
Qs cos = Re(a a
n n +1 +b b
n n +1 ) + Re(an bn* ) (5.40)
x n=1 n + 1 n=1
n(n + 1)
Finally, the backscattering efficiency Qb, applicable to monostatic radar, is given by
2
1 n
Qb = 2
x
(2n + 1)(1) (a
n =1
n bn ) (5.41)
5.4.4 Examples
MATLAB functions:
mie_ab(m, x) computes an and bn for n=1 to nmax
mie(m, x) computes Qext, Qsca, Qabs, Qb, g=<costeta>, for non-magnetic spheres
mie2(eps1, mu1, x) computes Qext, Qsca, Qabs, Qb, <costeta>, for magnetic spheres
mie_xscan(m, nsteps, dx) and Mie2_xscan(eps1, mu1, nsteps, dx) are used to compute and plot the efficiencies
versus size parameter x in a number (nsteps) of steps with increment dx from x=0 to x=nstepsdx.
k2
f(s,i) =
4 E i
{s [s E(r')]} {(r') 1} exp(ikr's)dV ' (5.43)
Vs
The integral requires knowledge on the electric field E(r') inside the scatterer. Since the
incident and scattered intensities Ii and Is are proportional to the squared absolute value of
the respective fields, we get
2
f
Is = Ii 2 (5.44)
r
2
The nominator f =d is the differential scattering cross section d. And the bistatic scattering
cross section bi, is defined by
2
bi4d = 4 f (s, i ) (5.45)
The total-scattered power Ps is the integral of Is over a closed surface A around the scatterer.
Ps = s s
2
I dA = I r d = I i d d = Ii s (5.46)
A 4 4
The last quantity s is the scattering cross section, which can also be expressed by
Ps 1
s = =
Ii 4
bi d (5.47)
4
In analogy, the absorption cross section a uses the absorbed power Pa:
Pa
a = (5.48)
Ii
A formal expression for Pa, again requiring the internal E field, was given by Ishimaru (1978)
2
Pa = 0.50 "(r') E(r') dV ' (5.49)
V
This power is just the Ohmic loss due to the electrical current density j=E excited in the
particle.
52
Scattering amplitude
Inserting (5.50) in Equation (5.43) for a point-like scatterer, and noting that the volume is
given by Vs = 4 a 3 /3, the scattering amplitude becomes
1
f(s,i) = a 3 k 2 {s [s e i ]} (5.51)
+2
where ei is the unit vector describing the direction of the incident electric field. Equation
(5.51) can also be regarded as the scattering amplitude from an electric dipole with dipole
3 1
moment De = 0e i for a unit-amplitude incident field and polarisability = 4 a of
+2
the sphere:
k2
f(s,i) = {s [s De ]} (5.52)
4 0
This is a more general expression, which is valid for all kinds of small particles.
Figure 5.11: Dipole radiation pattern for Rayleigh scattering: solid lines for parallel, dashed
line for perpendicular field. The lengths of the vectors s are proportional to the scattered field
strength.
Problems
1) Express S1 and S2 for Rayleigh scattering. Hint: Compare the definitions (5.24) and (5.42)
for the scattered field and use the result (5.51) for the scattering amplitude.
2) Compare the results of Lorenz-Mie scattering with Rayleigh scattering by doing your own
computations for Qs and Qa, and/or see Section 4.9 of the lecture notes on Radiative
Transfer.
3) Show that Pa of (5.49) represents the power loss due to the induced Ohmic current in the
scatterer.
5.6 Literature
R.N. Adams, E.D. Denman (1966), Wave Propagation and Turbulent Media, American Elsevier, New York, BEWI:
TEE 123.
C.F. Bohren and D.R. Huffman, Absorption and Scattering of Light by Small Particles, John Wiley, New York, NY
(1983). BEWI: TDD 122.
M. Born and E. Wolf, Principles of Optics, Pergamon Press (1975).
J.A. Dobrowolski, Optical properties of films and coatings, Ch. 42 in M. Bass et al. (Eds) Handbook of Optics, Vol.
1 (1995)
A. von Hippel, Dielectrics and Waves, 1st Ed. (1954) BEWI: TEA 149, 2nd Ed. (1995), BEWI: VTZ 201.
A. Ishimaru, Wave propagation and scattering in random media, Vol. 1, Academic Press, Orlando (1978), BEWI:
TDD 111.
C. Mätzler, “MATLAB Functions for Mie Scattering and Absorption”, IAP Res. Rep. No. 02-08, Institute of Applied
Physics, University of Bern, June (2002).
C. Mätzler, “MATLAB Functions for Mie Scattering and Absorption, Version 2”, IAP Res. Rep. No. 02-11, Institute
of Applied Physics, University of Bern, August (2002).
C. Mätzler and L. Martin, “Advanced Model of Extinction by Rain and Measurements at 38 and 94 GHz and in the
Visible Range”, IAP Res. Rep. No. 2003-1, February (2003).
C. Mätzler, “Mie Scattering With and Without Diffraction”, IAP Res. Rep. 2004-02, April (2004).
F.T. Ulaby, Moore and Fung (1981), Microwave Remote Sensing, Active and Passive, Vol. 1. BEWI: XKA 124,
125, 205, 207.
54
Figure 6.1: Model of a water molecule as a dipole aligned in an electric field (left), charges
and field of an electric dipole (middle), and schematics of the effect of polarisation of matter
by electric dipoles in a dielectric medium of a capacitor (right). Charges on its plates are due
to an applied voltage. Bound charges result from the neutralisation of dipoles, thus
enhancing the charge-storing capacity (from Mike Schwank, and A. von Hippel).
We will focus on because most media are essentially non magnetic (a similar consideration
would apply to μ for magnetic media). The ability of matter to become polarised by an electric
field is the basis of the dielectric behaviour. If matter gets polarised positive and negative
charges of atoms and molecules are separated, forming electric dipoles with the electric
dipole moment D e = qd where q is the positive electric charge and d the displacement vector
from the negative (–q) to the positive (+q) charge centre. The electric polarisation vector P is
the density of electric dipole moments in a given volume element V and is related with the
dielectric displacement vector by
a) induced polar. in
free atoms (positive
nucleus is displaced
versus negative
electron cloud)
b) in free molecules
(atmosphere)
c) orientation of
permanent molecular
dipoles and dipole
chains in liquids (also
in gases and solids)
d) space-charge
polarisation of mobile
charges in crystals
(conductors and semi
conductors)
56
Figure 6.3:
Polarisability versus
frequency of a
complex medium
(from Schanda,
1986).
The fastest response is due to the electron cloud in an atom or in a molecule. The electronic
polarisability e reacts to the wide frequency range from the static up to the ultraviolet (UV)
range where resonances occur due to electronic transitions. Atomic spectra are governed by
quantum theory.
A slower response is due to the deformation of single molecules where atomic or ionic
motions are involved (molecular vibrations and rotations). The corresponding polarisability is
denoted by i . Since atoms are heavier than electrons (typically by a factor 104), the
resulting resonances occur at much lower frequencies, namely in the infrared (IR) and mm
wavelength range. Quantum theory and observations show that vibrations of polar molecules
result in absorption bands and lines in the wavelength range from about 1 to 100 μm,
whereas rotations result in absorption lines at still longer wavelengths extending to about 1
cm.
Electronic transitions also occur in condensed matter. However, individual vibrations and
rotations of molecules are strongly hindered in liquids and solids. The absorption features
lack sharp resonances. In ordered media, such as crystals, the electric field can interact with
elastic deformations (transverse and longitudinal modes). Insight is provided by solid-state
physics.
In polar liquids, like water or alcohols, the electric polarisability d is due to the orientational
force of the electric field on the dipoles of the polar molecules. Thermal motion and friction
forces counteract the electric field. The resulting response is a relaxation phenomenon called
after Debye, who studied the process in detail (Debye, 1929). Debye relaxation dominates
the interaction in the microwave and radio-frequency range.
At the low-frequency end, freely moving charges (ions or electrons) are the dominant
contributors s to . They are displaced if electric field is applied. The limitations are given
by geometric boundaries of the conductor. Space charges of the limited conductors result in
dipoles. They are called Maxwell-Wagner effect. The build up of space charges is often slow
due to the properties of capacity and resistance.
The low-frequency response in molecular crystals (ice) is due to Bjerrum Defects, i.e.
imperfect crystals due to variable orientations of the molecules. The resulting response is a
very slow Debye relaxation.
The total polarisability of the medium is the sum of all contributions, thus
= e + i + d + s (6.6)
If there are no freely moving charges, s = 0 , and if there are no permanent dipoles, d = 0 .
Since the terrestrial atmosphere mostly consists of molecules, the first two terms are usually
non zero.
To get a physical idea on we will discuss some models based on classical physics. More
details can be found in the book of von Hippel (1954).
57
To balance the externally applied force F = ZeE on the electrons, a restoring Coulomb
force Fc will be set up in the atom by the separation of the charges. To compute this force,
the contributing charge only consists of the electron cloud inside a sphere with radius d
(Figure 6.4). This charge is Qd = Ze(d /r0 ) 3 , and the restoring force is
ZeQd d d
Fc = 3
= (Ze) 2 3
= F = ZeE (6.7)
4 0 d 4 0 r0
The linearity with d indicates an elastic behaviour. The excited dipole moment is
De = Zed = e E ; (E=E' in gases) (6.8)
From (6.7-8) we get a simple expression for the electronic polarisability
e = 4 0 r0 3 = 30V0 (6.9)
where V0 is the volume of the electron cloud. Inserting (6.9) in the low-density form of (6.5)
we get for the relative dielectric constant of an atomic gas
= 1+ 3NV0 = 1+ 3 f (6.10)
where N is the number density of atoms, and thus NV0 is the fraction of the volume occupied
by atoms. In dielectric mixing models this is called volume fraction f. Equation (6.10) means
that the electric susceptibility, -1, is just 3 times the volume fraction of the atoms in the gas.
Comment: The same equation follows from the Maxwell-Garnett mixing formula (4.5), if the
particle dielectric constant 2 , like for a perfect conductor.
For hydrogen atoms we approximate r0 by the Bohr Radius of 0.053nm. Under standard
conditions (T= 0C, P= 1000 hPa), N = NL= 2.6871025/m3 (Loschmidt Number), and we get
=1.00005. The actual value would be higher (1.0002) because the actual electron cloud
extends beyond the Bohr radius.
58
d 2P dP Ne 2 2
m 2
+ A + B P = Ne E (6.12)
dt dt 30
For a time-harmonic electric field E = E 0eit , also P is harmonic, P = P0eit , and we get
2
P p2 p
1 = = 2 2
= 2 (6.13)
0 E r0 p /3 2 ia r 2 i2a
Ne 2 B
where p = is the angular plasma frequency, r0 = is the angular resonant
m 0 m
p
frequency of the diluted medium ( p << r 0 ), p = , = are the corresponding
2 2
A
frequencies, and a is the damping rate due to friction. Note that (6.13) is a complex
m
susceptibility. The effective resonant frequency
r 1
r = = r0 2 p 2 /3 = r 0 2 p 2 /3 (6.14)
2 2
is lowered by the influence of the internal polarisation interactions in the medium. An
example of the resonance phenomenon is shown in Figure 6.5. For sufficiently low
frequencies the susceptibility is real and approaches the value
p, j 2
1 = 2
(6.15)
j r, j
which corresponds to Equation (6.10) found in the previous section. Below the resonant
frequency, the real part of 1 (and thus of n 1) is positive and increases with frequency.
This is called normal dispersion. It is observed for most transparent media in the optical
wavelength range because the electronic resonances are in the ultraviolet (UV). Through the
resonance region and n strongly decrease (anomalous dispersion), whereas the imaginary
parts show strong positive peaks. This means strong absorption near r . Therefore the
resonance is also called an absorption line.
In the more general case of various oscillator types without mutual coupling, (6.13) may be
generalised to a sum of terms over all oscillator types identified by Index j
59
p, j 2
1 = 2 2
(6.16)
j r, j ia j
This is a system of many absorption lines. The theoretical determination of the resonant
frequencies is the task of quantum theory. Results will be shown later.
Figure 6.5:
Resonance of
1 according
to Equation
(6.13) for
p / r = 0.126
a / r = 0.159
Figure 6.6:
The Langevin Function L(x) .
For x<<1, L(x) = x /3 .
In a typical polar molecule, the distance d between the charges is about 10-10 m, and the
charges correspond to ±1 electron charge. Then De = 1.610-29 Coulombm. For an extreme
electric field of 1 Mio V/m and a temperature of 300 K we get x = 3.810-3. This means that
the thermal motion dominates the orientation of the dipoles, and that L(x) = x /3 . This
linearity also means that does not depend on E'. The average moment due to orientation
becomes xDe/3, thus
De 2
De,E = E' (6.18)
3kb T
Now we can summarise the static polarisability due to electrons (Eq. 6.9), bound ions and
dipolar-orientation (6.18)
De 2
= e + i + (6.19)
3k b T
The constant terms and the last term can be determined experimentally from the temperature
dependence as shown in Figure 6.7.
dP Ne 2 2
A + B P = Ne E (6.21)
dt 30
For time-harmonic fields E = E 0eit , P = P0eit , we get
P Ne 2 /0
1 = = (6.22)
0 E Ne 2
B iA
30
Equation (6.22) corresponds to (6.20) by proper assignment of the constants.
The reason for relaxation is the dynamic behaviour of the permanent dipoles, which feel the
rotational torque by the electric field. At low frequencies the dipole motion follows the field
variation as in the static case. With increasing frequency the friction force more and more
hinders the rotational motion, leading to a reduced dielectric response, and to friction losses
expressed by the imaginary part. The friction corresponds to the dielectric loss as expressed
earlier by the conductivity = "0 = 2"0 " / c , where c = 17.975 GHz if is
expressed in MKS Units (-1m-1). Figure 6.8 shows an example of the Debye Function. It
corresponds to the values of water at T=0C, where the relaxation frequency is about 10 GHz.
There are similarities between the relaxation spectrum and the resonance curve of Figure
6.5. Both show a symmetrical peak of the imaginary part " in the semi-logarithmic
presentation and the real part ' decreases from the low- to the high- frequency end.
However, in Figure 6.8, ' nowhere increases with increasing frequency, and the transition
from high to low values as well as the peak of the imaginary part " are very broad (note the
4 frequency decades here, but only 2 decades in Figure 6.5).
Figure 6.9 shows " versus ' for similar parameters and over the same frequency range as in
Figure 6.8. The curve is a semicircle with its centre on the real axis.
Debye (1929) gave a classical interpretation of the relaxation time =1/ 2 0 of a spherical
molecule rotating in a viscous liquid:
3V
= (6.23)
kbT
where V is the volume of the spherical molecule and is the viscosity (friction parameter) of
the liquid. For more details, see von Hippel (1954). Since strongly decreases with
increasing temperature, strongly decreases with T (Figure 6.10).
Although the Debye Equation (6.20) for the relaxation phenomenon and Equation (6.23) for
give good descriptions of the general behaviour, the absolute values found for polar liquids
are more complex. Other theories have been developed, but they are very complex, and are
still under development. Today, results from empirical measurements are used in models for
remote sensing. Therefore we will also concentrate on such data.
6.8 Summary
It was shown that microscopic effects of dielectric media lead to three different types of
polarisability and thus of dielectric spectra: (1) a very fast response due to the adjustment of
the electron cloud together with similar adjustments with slower response with a frequency-
independent polarisability below a certain limit, (2) resonance phenomena, and (3) relaxation
processes. They were attribute to electronic, ionic, orientation and space-charge polarisation.
Figure 6.11 combines the phenomena for assumed parameters. The combined spectrum has
the form
p2 s p 2 / r 2
= 2 + (6.24)
r 2 i2a 1 i / 0
which is obtained from (6.20) by replacing the dummy parameter by the resonance (6.13)
in the first place and by its low-frequency limit in the second term. The equation is defined by
the 5 parameters given in the title of the figure (f standing for ).
Figure 6.11: Complex dielectric spectrum with one relaxation and one resonance according
to Equation (6.24) with the parameters given in the title.
64
Not discussed before, but of interest is the behaviour of the real and imaginary parts, ' and
" of (6.24):
'=
(
p2 r2 2 ) +
(
s p
2
/ r
2
) (6.25a)
2 2 2
( r
2
2
) + (2a ) 1+ ( / 0 )
"=
2a p
2
+
(
/ 0 s p 2 / r 2 ) (6.25b)
2 2 2
(
r 2 2 + (2a ) ) 1+ ( / 0 )
'= s ; "=
2a p
2
+
(
s p 2 / r 2 ) (6.25c)
4
r 0
Whereas the real part is the constant static value, the imaginary part increases proportionally
with frequency in both terms. This will be important to understand why the low-frequency
absorption of many media increases with increasing frequency in proportion to 2.
Problem
Note that ' decreases when the frequency increases through a resonance, or through a
relaxation phenomenon. For the relaxation process the decrease ' corresponds to twice
the maximum loss '= 2"max , which occurs at the relaxation frequency. Show that a similar
rule exists for the resonance effect:
"max
'= , where Q = r is the quality factor of the resonance, where the maximum
Q
of " occurs, and = 2a is the full width at half maximum of ".
6.9 Literature
Debye, P. "Polar Molecules", first published by Chemical Catalog Company, New York (1929), reprinted by Dover
Publications, New York (1945), German Edition: "Polare Molekeln", Leipzig (1929).
A. von Hippel, Dielectrics and Waves, 1st Ed. (1954) BEWI: TEA 149, 2nd Ed. (1995) BEWI: VTZ 201.