You are on page 1of 90

Mie Scattering in the X-Ray Regime

Diploma Thesis
vorgelegt von

Nandan Joshi
aus

Mumbai, Indien

angefertigt im

Institut für Röntgenphysik


der Georg–August–Universität zu Göttingen

2007
To my late father,

whose smile was delight in lucidity,


whose presence was comfort with security,
whose advice was guidance with conformity,
whose work was the inspiration of eternity,
whose love was with no other similarity,
whose life was worth more than infinity....
Acknowledgements

Understanding the subject of electromagnetic scattering from its fundamental level to the
application could not have been possible without the help of others.
First of all I would like to express my gratitude to Prof. Dr. Tim Salditt, for giving
me an opportunity to explore such quasi virgin field, showing confidence in my success as
well as failure and, of course, giving me important inputs, whereby bringing me onto the
right path to proper understanding.
My many thanks goes to the colleagues from the institute. I would like to show my
special gratitude to my colleague Aram Giahi for timely discussion whenever I get stuck
into difficulties. I am very glad to have got prompt assistance in the correction of my thesis
from Christian Fuhse and Simon Castorph. I am very thankful to them. On the verge of
submission, I have received great help from Michael Reese. His free nature has given me
great moral support working late and working very hard. I’m absolutely thankful to him.
I would also like to thank Klaus Giewekemeyer, Sebastian Kalbfleich, Sebastian Panknin
for their efforts to guide me towards the understanding of X-ray physics. I acknowledge
with many thanks to all my colleagues from the institute.
My gratitude is also to my girlfriend Karolina Wagrodzka, who showed great support
and understanding to my work schedule. Without her presence and readiness, it wouldn’t
have been possible to get moral boost to complete my thesis on time.
I owe my deepest gratitude to my parents, my mother and late father. It was their pri-
mary intention, effort and support that I had a chance to study in Germany, subsequently
write this thesis.

5
Table of Contents

1 Introduction and motivation 1


1.1 Scattering by a single particle . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Resume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 X-ray scattering: fundamentals 9


2.1 Overview of scattering processes . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 X-ray interactions with matter . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Elementary scalar theory of scattering . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Outgoing Green function . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.2 Integral-equation formulation . . . . . . . . . . . . . . . . . . . . . . 14
2.3.3 First Born approximation . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.4 Scattering by a sphere . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.5 Scattering cross section . . . . . . . . . . . . . . . . . . . . . . . . . 18
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3 Mie Scattering 21
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.1 Maxwell’s equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.2 Time-harmonic fields . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.3 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.4 Solution of differential equations with boundary conditions . . . . . 25
3.2.5 Series expansions for the scalar potentials . . . . . . . . . . . . . . . 27
3.2.6 Mie coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.7 Field vectors of scattered wave . . . . . . . . . . . . . . . . . . . . . 32
3.2.8 Intensity and polarization of the scattered light . . . . . . . . . . . . 32
3.2.9 Amplitude functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.10 Scattering cross section . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.1 Debye potentials and preceeding attempts . . . . . . . . . . . . . . . 35
3.3.2 Vector representations in spherical polar coordinates . . . . . . . . . 37
3.3.3 Associated Legendre functions . . . . . . . . . . . . . . . . . . . . . 37
3.3.4 Cylindrical functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.5 Scattering matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

7
8 TABLE OF CONTENTS

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4 Approximation for X-rays 41


4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2 Analytical Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.1 Limiting cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.2 Anomalous diffraction theory . . . . . . . . . . . . . . . . . . . . . . 43
4.2.3 Diffraction patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Numerical approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3.1 Using Mathematica . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3.2 Using IDL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5 Comparison 53
5.1 Comparing theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2 Errors and problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2.1 Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2.2 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.4 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.4.1 Term Rewriting System . . . . . . . . . . . . . . . . . . . . . . . . . 60

6 Outlook 65
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

7 Numerical calculation 69
7.1 Mathematica code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
7.2 IDL code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
List of Figures

1.1 Scattering by an obstacle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2


1.2 Interaction of light with a single particle . . . . . . . . . . . . . . . . . . . . 3

2.1 Monochromatic plane waves are incident from the left. Two ray paths are
shown, one associated with unscattered beam, in the absence of the volume
and other with scattered beam, in the presence of the scatterer. . . . . . . . 12
2.2 Illustration to the derivation of (2.25) . . . . . . . . . . . . . . . . . . . . . 15
2.3 Illustration to the notation in the given Born approximation . . . . . . . . . 16

3.1 Scattering of a incident wave from a spherical target located at the origin. . 23

4.1 Limiting cases in which m → 1 . . . . . . . . . . . . . . . . . . . . . . . . . 43


4.2 Ray passing through the sphere . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.1 Relative intensity graph for the sphere with radius 0.5 nm . . . . . . . . . . 55
5.2 Relative intensity graph for the sphere with radius 1.0 nm . . . . . . . . . . 55
5.3 Relative intensity graph for the sphere with radius 2.5 nm . . . . . . . . . . 56
5.4 Relative intensity graph for the sphere with radius 5.0 nm . . . . . . . . . . 57
5.5 Relative intensity graph for the sphere with radius 10.0 nm . . . . . . . . . 58
5.6 Scattering cross section graph for the sphere with radius 0.5 nm . . . . . . . 59
5.7 Scattering cross section graph for the sphere with radius 1.0 nm . . . . . . . 60
5.8 Scattering cross section graph for the sphere with radius 2.5 nm . . . . . . . 61
5.9 Scattering cross section graph for the sphere with radius 5.0 nm . . . . . . . 62
5.10 Scattering cross section graph for the sphere with radius 10.0 nm . . . . . . 62
5.11 Ratio of differential scattering cross sections of Mie theory to kinematical
theory at Q = 0 against radius . . . . . . . . . . . . . . . . . . . . . . . . . 63

9
Chapter 1

Introduction and motivation

1
2 Chapter 1. Introduction and motivation

No matter how old the subject of the scattering of plane waves by a sphere may look,
it has not yet been fully explored for different wavelengths.
The phenomena of scattering of light are ubiquitous, and quite naturally one of the
most important in investigative disciplines like science and engineering. Many optical phe-
nomena, like colors of the sunset, rainbow, the glory around the sun, the corona observed
in our atmospheres to the scattering by interstellar dust causing the starlight to scatter
before reaching us, are subjected to thorough investigation using the theory of scattering.
The strong dependence of the scattering interaction on particle size, shape, and refractive
index makes measurements of electromagnetic scattering a powerful noninvasive means of
particle characterization in different fields of science.
The optical properties of a medium is characterized by its refractive index. As long as it
is uniform, the ray passing through the medium will traverse the distance undeflected. But
if any obstacle, which may be a single electron, an atom or molecule, a solid or even liquid
particle, is illuminated by an electromagnetic wave, electric charges in the obstacle are set
into oscillatory motion by the electric field of the incident wave, i.e. there is small scale
density fluctuation, and the accelerated electric charges radiate electromagnetic energy in
all directions, this secondary radiation is called radiation scattered by the very obstacle.
Apart from reradiation of electromagnetic energy, the excited elementary charges may
transform part of the incident electromagnetic energy into other forms, a process called
absorption. In my investigation in this thesis, absorption shall be playing an important
role.

INCIDENT

OBSTACLE

SCATTERED

Figure 1.1: Scattering by an obstacle

The classic treatise in the subject has long been van de Hulst’s Light Scattering by Small
Particles[1], which was later supplemented in later years by more elaborative, chronicled,
at the same time, more application oriented work of Kerker, The Scattering of Light and
Other Electromagnetic Radiation[2], and much later by Bohren and Huffman, Absorption
and Scattering of Light by Small Particles[3]. But the scientific study of light scattering
may be said to have commenced by Tyndall[4]. The problem is to relate the properties
of the scatterer, like its size, shape, and refractive index, to the angular distribution of
the scattered light. In our case, where the scatterer is absorptive, part of the light will
be absorbed within it as heat, another part will be scattered, and the remainder will
be transmitted unperturbed along the incident direction. A complete description of the
scattered light entails a knowledge of wavelength, amplitude, phase and polarization of
1.1. Scattering by a single particle 3

the radiation from the scatterer. Scattering is hardly restricted to the optical part of
the spectrum, i.e. the scattering laws apply with equal validity to all wavelengths. It
is a quite interesting fact that these depend upon the ratio of a characteristic dimension
of the particle to the wavelength rather than explicitly upon the size. Thus, there is a
built-in scaling factor, called size parameter, which we shall come across in the up-coming
chapters.
In this thesis, I shall be dealing with a single particle of the size of a sphere and elastic
scattering. The complete mathematical solution to the problem of light scattering from a
sphere was obtained over a century ago as the well-known infinite series of particle waves.
This series, which is generally known as the Mie solution, contains in principle all the
physics of the problem for any size of particle and all wavelengths. For particles with
smaller radii than the wavelength, the few terms of the series need to be retained and one
acquires the satisfying knowledge. As the particle size increases, the series becomes slowly
convergent, and extracting the important physics from it becomes a daunting task. For
these reasons much of the theoretical work concerning Mie solution was of computational
nature. Although it can provide insightful graphical representation, it is still less effective
to uncover the physical origins of many of the interesting phenomena.

1.1 Scattering by a single particle


We can qualitatively understand the physics of scattering by a single particle without going
into computation. Consider an arbitrary particle, which we can divide in our gedanken
experiment into small regions. As applied oscillating field, like in our case, e.g. an incident
electromagnetic wave, induces a dipole moment in each region. These dipoles oscillate at
the frequency of the field and therefore scatter secondary radiation in all directions. In
a particular direction, the total scattered field is obtained by superposing the scattered
wavelets, like at the point P in the figure (1.2). The scattering by the dipoles is coherent.
The phase relations change for a different scattering direction; we therefore expect the
scattered field to vary with scattering direction. If the particle is small compared to the
wavelength of the incident wave, all the secondary wavelets are approximately in phase,
i.e. we don’t expect much variation in scattering with direction. But as the particle
size increases, the number of possibilities of interference effects of the scattered wavelets
increases. Shape is also important, if the particle is distorted, all the phase relations are
different.
P

INCIDENT

SCATTERED
WAVELETS

OBSTACLE

Figure 1.2: Interaction of light with a single particle

The phase relations among the scattered wavelets depend on geometrical factors, like
4 Chapter 1. Introduction and motivation

scattering direction, size and shape. But the amplitude and phase of the induced dipole
moment for a given wavelength depend on the material of which the particle is composed
of. In such case, it is important to know how the matter responds to oscillatory electro-
magnetic fields.
Although the methods for calculating scattering by a single particle is physically the
same as outlined above, but the mathematical form may cast a shadow on the underlying
physics. For certain classes of particles, the scattered field may be approximated by
subdividing the particle into dipole scatterers and superposing the scattered wavelets,
which is normally done in known Rayleigh-Gans theory, where interaction among the
dipoles are ignored. But in more general Mie theory, the interactions among the dipoles
are accounted for, which shall be explained mathematically in Chapter 3.

1.2 Resume
In this thesis, I shall be describing the scattering of waves, both scalar and electromagnetic,
from dielectric spheres in the X-ray regime. As mentioned above, we are treating here only
single particle scattering and elastic scattering. The latter condition means that there is no
shift of frequency between the incident and the scattered radiation. This excludes quantum
mechanical phenomena such as Raman effect and fluorescence or Brillouin scattering which
arises from the Doppler shifts associated with the motion of the scattering particles. The
restriction to single particle scattering implies that the scattering particle is unaffected by
the presence of neighboring particles.
Chapter 2 is a review of X-ray scattering. Here, the overall scattering process along
with the fundamentals of X-ray interactions with the matter is briefly explained. The
detailed discussion is available in the popular literature like Attwood[5], Jens-Nielson[6].
The X-ray scalar theory of scattering is developed by starting with the monochromatic
wave field. The whole development is restricted to the single particle scattering, which
leads to the solution for a dielectric sphere.
Chapter 3 deals with the theory of scattering by spheres. We are very fortunate in
having this exact theory in the sense of classical physics. The only restrictions are that the
substance of which the particle is composed of is isotropic and that any variations in the
refractive index are radially symmetric. This theory, as the title of this thesis is designated
after it, is frequently called the Mie theory, owing to the work of Gustav Mie[7]. Mie’s
paper was preceded by the independent solutions of a number of workers, starting with the
elegant work of Lorenz[8] to the work of Debye[9], which can more historically be studied
in the appendix of this chapter. This chapter gives the rigorous solution on the basis of
the explanation by Born & Wolf[10] with more of a inclination towards the results derived
by van de Hulst[1]. These are the results on which the modern computing technique is
mostly based on. The alternative rigorous derivation is due to Bohren & Huffman[3].
Chapter 4 gives attempts towards the rationalization of the Mie theory for X-rays.
The more adventurous, speculative, but yet possible analytical explanation towards the
approximation for X-ray region has been put forward, known as anomalous diffraction
theory. It is formerly due to the work of van de Hulst[1]. Since the elementary scalar
theory is pretty analogous to the Rayleigh-Gans-Theory, which is valid for particles of
smaller sizes, there is yet no proper analytical theory for the larger particle sizes, though
not large enough to be considered in geometrical optics region. Anomalous Diffraction
Theory may fill up this gap, maybe qualifying for the problems we are considering here.
The other half of the chapter deals with numerical solution, first with Mathematica, and
later with IDL. The respective algorithmic approach has been discussed.
Chapter 5 compares the result produced by Mathematica, which compares the Mie
theory with scalar kinematical theory. The results are discussed extensively in the chapter.
Chapter 6 gives the outlook in studying the exact theory for the X-ray region. And in
the last chapter the source code of the written program is attached.
1.3. Notation 5

1.3 Notation
The selection of notation is always a big problem, when it comes to the writings on
electromagnetic scattering. The attempt is made to make this thesis self-consistent, I have
used right-handed coordinate systems and the angle of rotation is positive if the rotation is
in clockwise direction when one is looking in the positive direction of the rotation axis, and
I have adapted the time-harmonic factor exp(−iωt), which is the most preferred choice
in various literature and publications, and implies a non-negative imaginary part of the
relative refractive index. The primes on a function always denote differentiation, whereas
primes on variables distinguish them from each other. An asterisk on a variable or function
always denotes complex conjugation and vectors are always denoted by boldface type. The
choices made for Bessel, spherical Bessel, and Ricatti-Bessel functions are described in the
appendix of Chapter 3. The comprehensive list of conventions, notations and results on
Mie theory prior to 1970 can be found in Kerker[11]. Whenever necessary, alternative
convention is given, but I have shown a reluctance to abandon widely used and recognized
symbols.
Bibliography

[1] H.C. van de Hulst. Light Scattering by Small Particles. Dover Publications Inc., New
York, 1981.

[2] Milton Kerker. The Scattering of Light. Academic Press, 1969.

[3] Craig F. Bohren and Donald R. Huffman. Absorption and Scattering of Light by
Small Particles. Wiley-VCH, 2004.

[4] J. Tyndall. On the blue color of the sky, the polarization of skylight and the polar-
ization of light by cloudy matter generally. Phil. Mag., 37:384–394, 1869.

[5] David Attwood. Soft X-Rays and Extreme Ultraviolet Radiation: Principles and
Applications. Cambridge University Press, 1999.

[6] Jens Als-Nielson and Des McMorrow. Elements of Modern X-Ray Physics. John
Wiley & Sons, 2001.

[7] Gustav Mie. Beiträge zur Optik trüber Medien, speziell kolloidaler Metallösungen.
Ann. d. Phys., 25:377–445, 1908.

[8] L. V. Lorenz. Upon the light reflected and refracted by a transparent sphere. Videknsk.
Selesk. Shrifter, 6:1–62, 1890.

[9] P. Debye. Der Lichtdruck auf Kugeln von beliebigem Material. Ann. d. Phys., 30:57–
136, 1909.

[10] Max Born and Emil Wolf. Principles of Optics. Cambridge University Press, 7 edition,
2002.

[11] Milton Kerker. The Scattering of Light, chapter 3, page 60. Academic Press, 1969.

7
Chapter 2

X-ray Scattering

9
10 Chapter 2. X-ray scattering: fundamentals

Diffraction by X-rays has been of great success in crystal structure determination, e.g.
structures of DNA, reconstructed silicon surfaces, etc. For a perfectly ordered crystal,
diffraction results in arrays of sharp Bragg reflection spots periodically arranged in re-
ciprocal space. Analysis of the Bragg peak locations and their intensities gives various
information of the crystal lattice, like lattice type, symmetry group, unit cell dimensions,
etc. On the other hand by crystals containing lattice defects such as dislocations, diffuse
intensities are produced in addition to Bragg peaks. The distribution and magnitude of
diffuse intensities are dependent on the type of imperfection present and the X-ray energy
used in a diffraction experiment. Diffuse scattering is usually weak, and thus more difficult
to measure, but it is rich in structure information that often cannot be obtained by other
experimental means.
In this chapter we are going to address the fundamental principles of diffraction based
upon the kinematic diffraction theory for X-rays. We are not going to discuss dynamical
diffraction since we are concentrating on single particle scattering. Dynamical diffraction
is applied to diffraction from single crystals of high quality so that multiple scattering
becomes significant and kinematic diffraction theory becomes invalid. But in practice,
most X-ray diffraction experiments are carried out on crystals containing a sufficiently
large number of defects that kinematic theory is generally applicable.

2.1 Overview of scattering processes

When a stream of radiation strikes matter, various interactions can take place, one of
which is the scattering process, that may well be described using the wave properties of
radiation. Depending on the wavelength of the incident radiation, scattering may occur
on different levels like atomic, molecular or microscopic level. Although some scattering
events are noticeable in daily life, others are more difficult to be taken true by naked eyes,
esp. those involving X-rays or neutrons.
X-rays are electromagnetic waves with wavelengths ranging from a few hundredths of
an angstrom (Å) to a few hundred angstroms. The conversion from wavelength to energy
for all photons is given in the following equation with wavelength λ in angstroms and
energy in kilo-electron volts (keV ):

c 12.40 keV
λ(Å) = = Å (2.1)
ν E(keV )

where c is the speed of light and ν the frequency. It is important to classify X-rays with
a wavelength longer than a few angstroms as ”soft X-rays” whereas ”hard X-rays” with
shorter wavelengths (≤ 1Å) and higher energies (≥ 1keV ).
It is however important to understand what it means by ”scattering” and ”diffrac-
tion”. The word ”scattering” refers to a deflection of a beam from its original direction
by the scattering centers that could be electrons, atoms, molecules, voids, dislocations,
etc. Whereas, the word ”diffraction” is generally defined as the constructive interference
of coherently scattered radiation from regularly arranged scattering centers such as grat-
ings, crystals, superlattices, and so on. Diffraction generally results in strong intensity in
specific, fixed directions in reciprocal space, which depends on the translational symmetry
of the diffracting system. Scattering, on the other hand, often generates weak and diffuse
intensities that are widely distributed in reciprocal space. E.g., interaction of radiation
with an amorphous substance is a ”scattering” process that shows broad and diffuse in-
tensity maxima, whereas with a crystal it is a ”diffraction” event, as sharp and distinct
peaks appear.
2.2. X-ray interactions with matter 11

2.2 X-ray interactions with matter


X-rays are electromagnetic waves with wavelengths in the region of an Å. It is a well-
known fact that when X-rays traverse through the medium, photons are absorbed and
intensity of the beam decreases exponentially over the distance x
I(r) = I0 e−µx (2.2)
where µ is the linear absorption coefficient. This expression is known as Lambert-Beer
law. In this process, an X-ray photon is absorbed by the atom, and the excess energy is
transferred to an electron, which is expelled from the atom, leaving the atom ionized. This
process is usually known as photoelectric absorption. Apart from that, the phase velocity
of the electromagnetic wave changes. The real term of refractive index n of a material
is given as Re(n) = c0 /cm , where c0 is the velocity in the vacuum and cm is the phase
velocity in the medium. Likewise the wave vector in the medium changes with the similar
relationship Re(n) = km /k0 . Then the monochromatic plane electromagnetic wave in the
medium can be written as
E(x, t) = E0 e−i(ωt−nk0 x) (2.3)
where ω is the frequency of the light. Since we are here concerned only with the amplitudes
as function of x, the time-dependent term e−iωt will be skipped. So we can write
E(x) = E0 eink0 x (2.4)
The refractive index for electromagnetic waves displays resonant behavior at frequen-
cies corresponding to electronic transitions in atoms and molecules. The binding energy
of most of the electrons in an atom lies way below than that of 10keV . The electric vector
of the plane X-ray wave excites the electrons way above the resonance to small forced
oscillations. This way the phase velocity of the X-ray wave is a bit more than the velocity
of light, hence the refractive index is less than 1. X-ray frequencies are usually higher than
all transition frequencies, except those involving the inner K- or maybe L-shell electrons.
In the end the X-ray region Re(n) turns out to be less than unity.
The refractive index is usually given in the following form
n = 1 − δ + iβ (2.5)
where δ is the dispersive term and β is the absorption term.
The dispersive term δ, in fact, the deviation of Re(n) from unity is related to the
scattering properties of the medium. Each electron scatters a X-ray beam with Thomson
scattering amplitude r0 , which is, in fact, the classical electron radius of e2 /(4π0 mc2 ) ≈
2.8 · 10−15 m. It turns out that δ is proportional to the product of r0 and the electron
density ρ. Hence, it can be given as
2πρr0
δ= (2.6)
k02
δ is normally of the order 10−6 . . . 10−5 .
To get the imaginary term, we substitute the definition of the refractive index in (2.4),
then we get
E(x) = E0 ei(1−δ+iβ)k0 x = E0 ei(1−δ)k0 x e−βk0 x (2.7)
For the intensity we get
I(x) = |E(x)|2 = I0 e−2βk0 x (2.8)
Comparing this equation with (2.2), we get for the absorbing term of the refractive index
µ
β= (2.9)
2k0
β is normally of the order less than of δ. These optical constants varies from material to
material.∗

the optical constants for X-ray interactions with matter can be obtained here:[1]
12 Chapter 2. X-ray scattering: fundamentals

2.3 Elementary scalar theory of scattering


We consider a monochromatic electromagnetic plane wave incident on a linear, isotropic,
nonmagnetic scattering medium of volume V , as shown in figure (2.1) as a series of parallel
arrows. Although knowing that we are strictly working within the context of wave optics,
nevertheless ’ray paths’ such as the one which is shown in the figure can be visualized. If
we were working within the framework of geometric optics, these lines would represent the
trajectory of a certain ray in the absence of the scatterer. However, within the framework of
scalar wave optics, this path may be defined such that the phase gradient of the unscattered
field is everywhere parallel to the trajectory. Here we assume that the scatterers are
sufficiently weak so as to negligibly perturb the ray paths which would have existed in the
volume occupied by the scatterer had the scatterer been absent. In such kind of situation
the phase and amplitude of the disturbances at the surface, where rays are exiting, can
be expressed in terms of the phase and amplitude shifts accumulated as the disturbance
traverses a given ray path connecting the entry and exit surfaces, with the ray paths
corresponding to those that would have existed if the scattering volume were replaced by
the surrounding medium.

INCIDENT

Figure 2.1: Monochromatic plane waves are incident from the left. Two ray paths are
shown, one associated with unscattered beam, in the absence of the volume and other
with scattered beam, in the presence of the scatterer.

We assume that there are no sources in V and that the space-dependent part of the
complex electric field will satisfy the following equation, specialized to a monochromatic
wavefield[2]:
~ [E(r, ω) · ∇
∇2 E(r, ω) + k 2 (r, ω)E(r, ω) + ∇ ~ ln (r, ω)] = 0 (2.10)

where k = ω/c, E is the electric vector and (r, ω) is the dielectric constant (permittivity).
We note that the last term on the left of the equation couples the Cartesian components
of the electric field, which makes the treatment of scattering based on this equation quite
complicated. It can be simplified to the inhomogeneous Helmholtz equation with the
following assumptions:
• the medium is non-magnetic
• the electrical permittivity, hence the refractive index, is time-independent and slowly
varying over spatial length scales comparable with the wavelength of the radiation
with which it interacts
• we may work with a single complex quantity to quantify the disturbance due to the
monochromatic electromagnetic field
• the effects of inelastic scattering may be ignored
2.3. Elementary scalar theory of scattering 13

Under these circumstances the last term of the left-hand side of the equation may be
neglected, delivering the reduced form of the equation:

∇2 E(r, ω) + k 2 n2 (r, ω)E(r, ω) = 0 (2.11)

with Maxwell formula (r, ω) = n2 (r, ω), where n(r, ω) denotes the refractive index of
the medium. Better understanding of the general behaviour of the scattered field may
be obtained by studying the behaviour of the solution of the equation (2.10) for a single
Cartesian component of E(r, ω), i.e. the problem concerning the single point scatterer.
We shall denote this component by U (r, ω), there we obtain

∇2 U (r, ω) + k 2 n2 (r, ω)U (r, ω) = 0 (2.12)

The same equation can be rewritten in the following way

∇2 U (r, ω) + k 2 U (r, ω) = −4πF (r, ω)U (r, ω), (2.13)

where
1 2 2
F (r, ω) = k [n (r, ω) − 1] (2.14)

The function F (r, ω) is usually called the scattering potential of the medium. We shall
express U (r, ω) as the sum of the incident field and the scattered field

U (r, ω) = U (i) (r, ω) + U (s) (r, ω) (2.15)

The incident field, being the plane wave, satisfies the Helmholtz equation throughout
the space
(∇2 + k 2 )U (i) (r, ω) = 0 (2.16)
Whereas we get for the scattered field

(∇2 + k 2 )U (s) (r, ω) = −4πF (r, ω)U (r, ω) (2.17)

Our ultimate aim is to convert this differential equation into an integral equation, which
we shall do after introducing Green functions in the next section.

2.3.1 Outgoing Green function


Green functions are a powerful and widely applicable tool in mathematical physics. Here
we shall use them to study the single Cartesian component of the field defining the Green
function to be the field that is scattered from the volume. We can decompose the total
field, in the presence of the scatterer, as the sum of the two fields: the field which would
have existed in the absence of the scatterer, and the scattered field. The outgoing Green
function G(r − r0 ) is by definition equal to the scattered field due to a point scatterer
located at r0 , as a function of the position coordinate r in three-dimensional space. So let
G(r−r0 ) be the Green function of the Helmholtz operator, obeying the following equation:

(∇2 + k 2 )G(r − r0 , ω) = −4πδ (3) (r − r0 ) (2.18)

where δ (3) (r − r0 ) is the three-dimensional Dirac delta function.


Now we have to obtain free-space Green function of the Helmholtz operator. Since free
space is homogeneous and isotropic, the scattered field observed at a given point depends
only on the relative position of the observation point with respect to the scatterer∗ . we
shall now introduce polar coordinates centered at r, with radial distance from the origin

For rigorous method of obtaining Green function using complex analysis can be found in [3]
14 Chapter 2. X-ray scattering: fundamentals

denoted by r. Using Laplacian in spherical polar coordinates and assuming Green function
being rotationally symmetric, we can get for equation (2.18)
 
1 d 2 d
r G(r) + k 2 G(r) = −4πδ(r) (2.19)
r2 dr dr

After multiplying
R∞ both sides by r, thereby eliminating Dirac delta function using the
relation −∞ f (x)δ(x − a)dx = f (a), we get
 
1 d 2 d
r G(r) = −k 2 G(r) (2.20)
r2 dr dr
d2
[rG(r)] = −k 2 G(r) (2.21)
dr2

It shows that rG(r) is the oscillatory eigenfunction of d2 /dr2 , so that rG(r) = exp(±ikr).
Which yields the explicit form of the free-space Green function:

exp(±ikr)
G(r) = (2.22)
r
In our context, we should choose positive sign in the exponent, since we are concerned with
a outgoing spherical wave. So we can not define the outgoing free-space Green function as

exp(ik|r − r0 |)
G(r − r0 , ω) = (2.23)
|r − r0 |

2.3.2 Integral-equation formulation


In this section, we shall rewrite the differential equation (2.17) as an integral equation. It
is mandatory to note that the integral equation is more restrictive than the differential
equation from which it is obtained, in the sense that the latter must be supplemented by
appropriate boundary conditions, whereas the former has those conditions within itself.
This integral equation will be used to derive approximate expressions later in the chapter.
Let us multiply the equation (2.17) by G(r−r0 , ω) and the equation (2.18) by U (s) (r, ω)
and subtract them from each other, which gives[4]

U (s) (r, ω)∇2 G(r − r0 , ω) − G(r − r0 , ω)∇2 U (s) (r, ω) =


(2.24)
4πF (r, ω)U (r, ω)G(r − r0 , ω) − 4πU (s) (r, ω)δ (3) (r − r0 , ω)

We now integrate both sides of (2.24) with respect to r0 throughout a volume VR ,


bounded by a large sphere SR of radius R, centered on the origin O in the region of the
scatterer. After applying Green’s theorem to convert the volume integral on the left into
a surface integral, we obtain
Z
1
(s)
U (r, ω) = F (r0 , ω)U (r0 , ω)G(r − r0 , ω) −
V 4π
Z " 0 (s) (r0 , ω)
# (2.25)
(s) 0 δG(r − r , ω) 0 δU
U (r , ω) − G(r − r , ω) d SR
SR δn0 δn0

where δ/δn0 denotes the differentiation along the normal n0 to SR . The first integral
is taken only over the scattering volume V , since the scattering potential F vanishes
throughout the exterior of V , as can be seen from the equation (2.14)
Now we shall substitute the G(r − r0 , ω) with the free-space Green function obtained
in the last section. It seems plausible that sufficiently far away from the scatterer, the
scattered field will behave like an outgoing spherical wave. So we can conclude that in the
2.3. Elementary scalar theory of scattering 15

O n'

VR
SR

Figure 2.2: Illustration to the derivation of (2.25)

limit for R → ∞, the surface integral will not contribute to the total field. So we get for
scattering field
exp(ik|r − r0 |) 3 0
Z
U (r, ω) = F (r0 , ω)U (r0 , ω)
(s)
d r (2.26)
|r − r0 |
Supposing that the incident field on the scatterer is a monochromatic plane wave of
unit amplitude and frequency ω, the time-independent part of the incident field is given
as
U (i) (r, ω) = eik0 ·r , (2.27)
so we get the final integral equation
exp(ik|r − r0 |) 3 0
Z
ik0 ·r
U (r, ω) = e + F (r0 , ω)U (r0 , ω) d r (2.28)
V |r − r0 |
This equation determines the total field U (r, ω), including the incident and scattered field
and is usually called as the integral equation of potential scattering. Note that the desired
field U (r, ω) appears on the both sides of the equation, which is why we call it as an
integral equation rather than an integral expression for U (r, ω).

2.3.3 First Born approximation


It is not easy to obtain solutions of the integral equation of potential scattering in a closed
form. These, then, should be solved by some approximate technique. The most common
among different approximate techniques is perturbation. In this technique, the successive
terms in the perturbation expansions are obtained by iteration from the previous ones,
provided the medium scatters weakly.
From expression (2.14) it can said that a medium will scatter weakly if its refractive
index differs only slightly from unity, which is exactly our case. If one can assume that
the X-ray perturbation inside the scattering volume is only slightly different from the
perturbation that would have existed at each point r0 in that volume in the absence of the
scatterer, then one can replace U (r0 , ω) by the unscattered perturbation U (i) (r0 , ω), inside
the integral. From now on we shall not display the dependancy of the various quantities
on the frequency ω for convenience. The expression looks like,
0
0 exp(ik|r − r |) 3 0
Z
U (r) = eik0 ·r + F (r0 )eik0 ·r d r (2.29)
V |r − r0 |
Now this is no longer an integral equation, but rather an approximate expression for the
total wave-field, which is popularly known as the first Born approximation.
16 Chapter 2. X-ray scattering: fundamentals

It can be noticed that the first Born approximation corresponds to a single-scattering


theory in which the incident wave-field is either not scattered at all, or scattered only once
by a single point within the sample. Such single-scattering approximations are known
as kinematical theories. If the incident wave-field may be scattered multiple times, the
devised theories on this scenario are known as dynamical theories.

r'|
|r-

r
Q

r'
N
O

Figure 2.3: Illustration to the notation in the given Born approximation

Now we treat the special case of the first Born approximation which corresponds to
the scattered radiation very far from all scatterers. The treatment given here is in direct
mathematical correspondence to the treatment of non-relativistic potential scattering in
quantum mechanics. From the figure (2.3), let Q be a typical point in the scattering
volume V and P a point far away from it. Let r0 be the position vector of Q and r = rr̂
be the position vector of P , where r̂ is the unit vector in the direction of r. Let N be the
foot of the perpendicular dropped from Q onto the line OP . Like in our case, when r is
large enough, then
|r − r0 | ∼ r − r̂ · r0 (2.30)

Hence we obtain the following approximation form for the spherical wave scattered from
a point r0 within the scattering volume:
p
exp(ik|r − r0 |) (r − r0 ) · (r − r0 )]
exp[ik
= (2.31)
|r − r0 | |r − r0 |
p
exp[ik (|r|2 − 2r · r0 + |r0 |2 ]
= (2.32)
|r − r0 |
p
exp[ik (|r|2 − 2r · r0 ]
≈ (2.33)
√ r
exp[ikr 1 − 2r−2 r · r0 ]
= (2.34)
r

After making the binomial approximation 1 − 2r−2 r · r0 ≈ 1 − r−2 r · r0 , we get

exp(ik|r − r0 |) exp[ikr]
0
∼ exp[−ikr̂ · r0 ] (2.35)
|r − r r

Substituting it in the equation (2.29), we get the approximation

exp[ikr]
U (r) ∼ eik0 ·r + f (Q) (2.36)
r
2.3. Elementary scalar theory of scattering 17

where,
Z
0
f (Q) ≡ F (r0 )e−iQ·r dr0 (2.37)
Q ≡ kr̂ − k0 . (2.38)
The function f (Q) is known as scattering amplitude and it has important physical impli-
cations. Because the scattered radiation is given by a distorted spherical wave originating
from the point, but still behaving like an outgoing spherical wave at very far distance, and
the form of distortion is quantified by the envelope f . f multiplies the undistorted outgo-
ing spherical wave exp(ikr)/r in order to form the distorted spherical wave emerging from
the scatterer. The value of f is the three-dimensional Fourier transformation of F (r0 ). So
it can be stated that within the accuracy of the first Born approximation, the scattering
amplitude f in the far zone of the scatterer depends entirely on one and only one Fourier
component of the scattering potential, namely the vector
Q = k − k0 (2.39)

2.3.4 Scattering by a sphere


In order to find a differential scattering cross section of the sphere, we shall calculate the
scattering amplitude of a sphere using the equation (2.37). We can rewrite the equation
for a sphere as Z
f (Q) = F (r)e−iQ·r (2.40)
Vsph
Here is the integration over the volume of the scatterer. The scattering potential F (r) is
defined like before in (2.14). Let the sphere be homogeneous, then we can redefine the
scattering potential as
(
1 2 2
k (n − 1) r ≤ R ∵ n(r ≤ R) = n,
F (r) = 4π (2.41)
0 r≥R ∵ n(r ≥ R) = 1.
where R is the radius of the sphere.
Thus for f (Q) in spherical coordinates we get
k2 ∞ 2 2
Z Z 2π Z +1
f (Q) = r (n (r) − 1)dr dφ d(cos θ)ei|Q||r| cos θ (2.42)
4π 0 0 −1
2 Z ∞ 2
k r
= (n2 (r) − 1)(eiQr − e−iQr )dr (2.43)
2iQ 0 r
Z R
k2 2
= (n − 1) r sin(Qr)dr (2.44)
Q 0
sin(QR) − QR cos(QR)
= k 2 (n2 − 1) (2.45)
Q3
3 2 2 sin(QR) − QR cos(QR)
= k (n − 1)V (2.46)
4π (QR)3
Now we can approximate this expression for the refractive index in X-ray region. With
the help of the discussion in the section sec. 2.2, we can write
n2 ≈ 1 − 2δ + δ 2 ' 1 − 2δ (2.47)
Substituting it in (2.46), we get
3 2 sin(QR) − QR cos(QR)
f (Q) = k · 2δV (2.48)
4π (QR)3
Replacing δ with (2.6) from the section sec. 2.2, we finally get for scattering amplitude
sin(QR) − QR cos(QR)
f (Q) = 3ρr0 V (2.49)
(QR)3
18 Chapter 2. X-ray scattering: fundamentals

2.3.5 Scattering cross section


The power radiated in the scattered direction, per unit solid angle, per unit incident flux
(power per unit area) in the incident direction, is a quantity with dimensions of area
per unit solid angle, which is known as differential scattering cross section. The proper
definition is the following way:

dσ Energy radiated/unit time/unit solid angle


= (2.50)
dΩ Incident energy flux/unit area

So in this example of scattering by a sphere, the differential scattering cross section is


given by

= |f (Q)|2 (2.51)
dΩ
After substituting (2.49) in the equation above, we get

dσ (sin(QR) − QR cos(QR))2
= r02 ρ2 V 2 9 (2.52)
dΩ (QR)6
Bibliography

[1] Eric Gullikson. X-ray interactions with matter. http://www-cxro.lbl.gov/.

[2] Max Born and Emil Wolf. Principles of Optics, chapter 1, page 11. Cambridge Uni-
versity Press, 7 edition, 2002.

[3] David Paganin. Coherent X-Ray Optics, chapter 2, page 80. Oxford University Press,
1 edition, 2006.

[4] Max Born and Emil Wolf. Principles of Optics, chapter 13, page 697. Cambridge
University Press, 7 edition, 2002.

19
Chapter 3

Mie Scattering

21
22 Chapter 3. Mie Scattering

3.1 Introduction
In 1902, Gustav Mie moved to Greifswald, where he assumed a special professorship post
(Extraordinarius) at the University of Greifswald, where, in fact, he wrote the famous
69-page paper on particle light scattering, published in 1908 in the Annalen der Physik,
bearing the title ”Beiträge zur Optik Trüber Medien, speziell kolloidaller Metallösungen”.
It appears that Mie’s 1908 paper represents a single major involvement with the subject
of particle light scattering and absorption. In these years in Greifswald, having been
confounded with the idea of developing a comprehensive Theory of Matter, it was the ex-
perimental investigations on colloidal gold suspensions by a student (Walter Steubing) at
the Greifswald Institut, which actually triggered his interest to develop a rigorous theoret-
ical interpretation of the empirical results of the student. This theoretical interpretation
was based on Maxwell’s equations, on whose ramification Mie had concentrated his at-
tention since at least 1896. He proceeded further beyond the original intention of merely
explaining the colors of colloidal gold observed by Steubing. The thorough understanding
of the theory using the behavior of ellipsoidal particles, as mentioned and wished by Mie
in conclusion, was unfortunately left behind, having been entangled in his other theoret-
ical works. The reason was that this great work was underestimated by its author and
contemporary scientists. The study lies in the trap of various cylindrical functions and
Legendre polynomials. It was not until the dawn of computational machines, when the
subject was taken for serious investigation. In this chapter we shall be voyaging through
the original analytical work done on light scattering by small particles.

3.2 Derivation
The Mie solution to the scattering of light at particles of any size is a classical electro-
dynamics problem. As will be seen in the following sections, the scattering of light by a
homogeneous sphere is treated in a general way by the formal solution of Maxwell’s equa-
tion with the appropriate boundary conditions. Although I shall be, to certain extent,
employing the procedure followed by Gustav Mie[1], I shall also be applying the similar
technique developed quite independently by Debye[2] a year later. It is, in fact, the so-
lution of Maxwell’s equation with certain boundary conditions using Debye potentials for
the electromagnetic field[3]. The definition of the Debye potential and historical perspec-
tive to arrive at this idea, subsequently the solution given by Gustav Mie, is explained
explicitly in the appendix (see 3.3.1).

3.2.1 Maxwell’s equation


The treatment of scattering of light by small particles is a problem in electromagnetic
theory. We are adopting here a macroscopic approach to the problem, i.e. the Maxwell
equations for the macroscopic electromagnetic field at interior points in matter, which can
be written as,

4π 1 dD
curl H = I+ (3.1)
c c dt
1 dH
curl E = − (3.2)
c dt
div D = 4πρ (3.3)
div H = 0 (3.4)

where E is the electric field, H is the magnetic field and D is the electric displacement.
The charge density ρ and current density I are associated with free charges. Although it
is shown convincingly by Purcell[4] that it is not possible to unambiguously distinguish
between free and bound charges in matter, we shall assume that the ambiguity in the
3.2. Derivation 23

meanings of free and bound leads to no observable consequences in the concerned problem
in the following sections.
The Maxwell’s equations must be supplemented with constitutive relations, which has
the form in our disscussion

D = E (3.5)
I = σE (3.6)

where  is the permittivity of free space and σ is the conductivity. These coefficients , σ
depend on the medium under consideration, but here it will be assumed to be independent
of the fields, position and direction. Now we have complete set of Maxwell’s equations.
As per the method of Mie, we are going to write them in polar coordinates, but the little
more discussion of periodic phenomena, i.e. nature of time-harmonic fields is required.

Es2
θ Es1

H
Ei1 Y
Ei2
E

Φ
X

Figure 3.1: Scattering of a incident wave from a spherical target located at the origin.

3.2.2 Time-harmonic fields


Let A be the general time-harmonic field, then it takes the form

A = X cos ωt + Y sin ωt, (3.7)

where ω is the angular frequency. But in this chapter, we shall be dealing with complex
representation of the this field and it is convenient to work with complex representation.
Then, we can say A = Re{Ac }, where

Ac = Ze−iωt , Z = X + iY (3.8)

It might be quite important to notice that the same real field A can also be written
as A = Re{A?c }, where A?c = Z? eiωt is, in fact, the complex conjugate of Ac . It says that
24 Chapter 3. Mie Scattering

there are two different choices of representing time-harmonic fields, namely e−iωt and eiωt .
We take the usual convention, seen in many popular books, i.e. that of e−iωt .
Now let us assume here the time-dependency is taken care by the term exp(−iωt), and
we substitute the constitutive relations (3.5) in the Maxwell’s equations (3.1). Then we
can say that the both time independent parts of the electric and magnetic vectors outside
and inside the scattering medium satisfy the following relationship

curl H = −k1 E (3.9a)


curl E = k2 H (3.9b)

where
 
iω 4πσ
k1 = +i (3.10a)
c ω

k2 = (3.10b)
c
Now we write the same equations in polar coordinates using the relations explained in
3.3.2, and we get

 
1 ∂(r sin θHφ ) ∂(rHθ )
−k1 Er = 2 − (3.11a)
r sin θ ∂θ ∂φ
 
1 ∂Hr ∂(r sin θHφ )
−k1 Eθ = − (3.11b)
r sin θ ∂φ ∂r
 
1 ∂(rHθ ) ∂Hr
−k1 Eφ = − (3.11c)
r ∂r ∂θ

 
1 ∂(r sin θEφ ) ∂(rEθ )
k2 Hr = − (3.12a)
r2 sin θ ∂θ ∂φ
 
1 ∂Er ∂(r sin θEφ )
k2 Hθ = − (3.12b)
r sin θ ∂φ ∂r
 
1 ∂(rEθ ) ∂Er
k2 Hφ = − (3.12c)
r ∂r ∂θ

These equations will now be our point of departure in scattering problems.

3.2.3 Boundary conditions


The scattering medium we are considering is a sphere. Now let us give prefix I to the
medium surrounding the sphere and prefix II to the sphere. If the refractive index of both
the media is infinite, then surface current exist, else there shall be no surface current to
consider. So we can put the boundary condition that the tangential component of E and
H shall be continuous across the surface of the sphere:
I II
Etang = Etang (3.13a)
I II
Htang = Htang (3.13b)

which can be interpreted in spherical polar coordinates as


I II I II
Eθ = Eθ , Eφ = Eφ (3.14a)
I II I II
Hθ = Hθ , Hφ = Hφ (3.14b)
3.2. Derivation 25

3.2.4 Solution of differential equations with boundary conditions


The equations (3.11) and (3.12) with boundary conditions (3.14) are to be solved. What we
are trying to solve is the vector wave equation. So we will start with the some explanation,
which will help us determine Debye Potentials for our problem.
Let a complex-valued vector field Π is defined in a domain D and satisfies

∇2 Π + k 2 Π = ∇φ (3.15)

where φ is an arbitrary function, normally called a Hertz vector for D. Now the equation
(3.15) may be written as

∇ × (∇ × Π) − k 2 Π = ∇(∇ · Π − φ) (3.16)

from which each Hertz vector gives rise to a pair of associated vector wave functions,

M = k(∇ × Π), N = ∇ × (∇ × Π). (3.17)

These functions M and N have all the required properties of an electromagnetic field.
They satisfy the vector wave equation, they are divergence-free, the curls are proportional
to each other. This way the problem of finding solutions to the field equations reduces
to the comparatively simpler problem of finding solutions to the scalar wave equation.
Now, we shall consider a scalar function ψ, a generating function for the vector harmonics
M and N for the domain D and r a position or guiding vector relative to an origin but
outside D, then we can write
Π = ψr (3.18)
which defines a radial Hertz vector for D

M = k(∇ × ψr) (3.19)

Now, M is a solution to the vector wave equation in spherical polar coordinates. In


problems involving spherical symmetry, we shall take M given in (3.19) and the associated
N as the fundamental solution to the field equations. The scalar wave equation in spherical
polar coordinates is

1 ∂ 2 (rψ) ∂2ψ
 
1 ∂ ∂ψ 1
+ sin θ + + k2 ψ = 0 (3.20)
r ∂r2 r2 sin θ ∂θ ∂θ r2 sin2 θ ∂φ2
But before that, we must define these vector wave functions M and N. From (3.18),
we can write
∇2 Π + k 2 Π = 2∇ψ (3.21)
The scalar wave function ψ is, in fact, called a Debye potential for the vector wave functions
defined by (3.17). To be more precise, ψ is called an electric Debye potential for M and
a magnetic Debye potential for N.
These vector wave functions can be represented the following way

M ≡ (e E,e H) and N ≡ (m E,m H) (3.22)

One can divide all the solutions of the equations in three groups:

• The first one represents the waves, which are produced through electrical oscillation
of the sphere, i.e.
e e
Er = Er , Hr = 0 (3.23)

• The second group represent waves, which are produced through magnetic oscillation
of the sphere, i.e.
m e
Er = 0, Hr = Hr (3.24)
26 Chapter 3. Mie Scattering

• The third group contains the integrals of Maxwell’s equations, which shows the
regular periodic oscillations, which, in fact, can be attained through addition of
integrals of first two groups, i.e. addition of all polar components respectively

Let us consider a first case. For Hr = 0 equations (3.11b) and (3.11c) become

1 ∂ e
k1e Eθ = (r Hφ )
r ∂r (3.25)
1 ∂ e
k1e Eφ = − (r Hθ )
r ∂r
Substituting it back in (3.12b) and (3.12c) we get
 2 
∂ e k1 ∂ e Er
(r H θ ) = −
∂ r2 sin θ ∂φ
 2  (3.26)
∂ e ∂ e Er
(r H φ ) = +k 1
∂ r2 ∂θ

Equation (3.26) along with the equation (3.11a) constitute a system of equations for
e E ,e H ,e H .
r θ φ The only solutions satisfying the condition ’div e H = 0’ represent physical
fields. In spherical polar coordinates, taking into consideration the assumption e Hr = 0,
this condition becomes
∂ ∂ e
(sin θe Hθ ) + ( Hφ ) = 0 (3.27)
∂θ ∂φ
On the substitution from (3.25), the equation (3.12a) becomes
 
1 ∂ ∂ e ∂ e
0= 2 2 (r sin θ Hθ ) + ( Hφ ) (3.28)
k1 r sin θ ∂r ∂θ ∂φ

It is well satisfied from the relation (3.27). Analog to the method above, one can obtain
the solutions to the second case. Now, I shall write the generating function ψ with respect
to vector wave functions M and N as Debye potentials e Π and m Π, respectively.
Like done above, we shall calculate for the scalar potential e Π in the first case of
e E = E and e H = 0. Since e H = 0, e E and e E can be represented as
r r r r φ θ

e 1 ∂2r eΠ e 1 ∂2r eΠ
Eφ = , Eθ = (3.29)
r sin θ ∂r∂φ r ∂r∂θ

So the equation (3.25) may be satisfied by

e ∂ eΠ k1 r∂ e Π
Hφ = k1 =
∂θ r ∂θ
(3.30)
e k1 ∂ e Π k1 r∂ e Π
Hθ = − =−
sin θ ∂φ r sin θ ∂φ

After substituting this equation in (3.11a) we get

∂ eΠ 1 ∂2 eΠ
   
e 1 ∂
Er = − sin θ + (3.31)
r sin θ ∂θ ∂θ sin θ ∂φ2

Substitution from (3.30) and (3.31) into (3.26) gives two equations, the first of which
expresses the vanishing of the φ derivative, the second the vanishing of the θ derivative of
the same expression. So the best way is to equate them to zero which leaves us with the
following expression,

1 ∂ 2 (re Π) ∂ eΠ ∂2 eΠ
 
1 ∂ 1
+ sin θ + 2 + k2 e Π = 0 (3.32)
r ∂r2 r2 sin θ ∂θ ∂θ 2
r sin θ ∂φ 2
3.2. Derivation 27

That means the equation (3.31) becomes

e ∂ 2 (re Π)
Er = + k2 r e Π (3.33)
∂r2
After substituting all these equations in main Maxwell’s equations with polar coordi-
nates, in (3.11) and in (3.12), we obtain a solution of our set of equations.
In a similar way, one can find that (3.32) is also valid in case of potential m Π. The
complete solution of the field equations can be obtained by adding the two fields.
As we have seen, both the potentials are solutions of the differential equation (3.32),
which, in fact, is the following wave equation, but written in polar coordinates,

∇2 Π + k 2 Π = 0 (3.34)

For the components Eθ , Eφ , Hθ , Hφ to be continuous over the spherical surface r = a,


it is sufficient that the following four quantities shall be continuous over the same surface,
∂ e ∂ m
k1 r e Π, k2 r m Π, (r Π), (r Π) (3.35)
∂r ∂r
So the boundary condition splits into examining independent condition for the potentials
e Π and m Π. That means, the whole problem is reduced to the problem of finding two

mutually independent solution of the wave equation with boundary conditions given above.

3.2.5 Series expansions for the scalar potentials


Let us consider that the solution of the wave equation comprises of undetermined coeffi-
cients, which can be determined by using boundary conditions. So the solution takes a
form
Π = R(r)Θ(θ)Φ(φ) (3.36)
After substituting in (3.32), one can see that these coefficients must satisfy the following
ordinary differential equations

d2 (rR)  2 α
+ k − rR = 0 (3.37a)
 dr2  r2 
1 d dΘ β
sin θ + α− Θ=0 (3.37b)
sin θ dθ dθ sin2 θ
d2 Φ
+ βΦ = 0 (3.37c)
dφ2
where α and β are integration constants.
Taking into account that the field E, H is a single valued function of position, Π must
also be single valued, imposing conditions on Θ and Φ. E.g., one can write down the
general equation for the last ODE in (3.37c)
p p
a cos( βφ) + b sin( βφ) (3.38)

But due to the condition of single-valuedness demanding

β = m2 , (m = integer) (3.39)

we get the solution


Φ = am cos(mφ) + bm sin(mφ) (3.40)
The equation (3.37b) is the equation of spherial harmonics. In this case, a necessary
and sufficient condition for a single valued solution is

α = l(l + 1), (l > |m|, integer) (3.41)


28 Chapter 3. Mie Scattering

with the introduction of the following variable ‡

ξ = cos θ (3.42)

The equation then takes the form

m2
   
d 2 dΘ
(1 − ξ ) + l(l + 1) − Θ=0 (3.43)
dξ dξ 1 − ξ2

The solution to this differential equation is the well-known associated Legendre func-
tions
Θ = Plm (ξ) = Plm (cos θ) (3.44)
In an attempt to find a solution to the remaining first equation (3.37a) with the
following substitutions
1
kr = ρ, R(r) = √ Z(ρ) (3.45)
ρ

we obtain the Bessel equation §

( )
d2 Z 1 dZ (l + 12 )2
2
+ + 1− Z=0 (3.46)
dρ ρ dρ ρ2

The solution of this equation is the general cylindrical function Z = Zl+ 1 (ρ) of the
2
order l + 21 , which gives us the complete solution to be

1
R = √ Zl+ 1 (kr) (3.47)
kr 2

Each cylindrical function may be expressed as a linear combination of two cylindrical


functions of standard type, e.g. the Bessel function Jl+ 1 and the Neumann function Nl+ 1 .
2 2
Let us introduce the convenient form of functions
r r
πρ πρ
ψl (ρ) = Jl+ 1 and χl (ρ) = − N 1 (3.48)
2 2 2 l+ 2

The functions χl (ρ) have singularities at the origin ρ = 0, whereas the functions ψl (ρ)
are regular in every finite domain of the ρ-plane, including that of origin. That is why we
will use the functions ψl (ρ) for representing wave inside the sphere.
So the general integral of the equation (3.37a) may be given as

rR = cl ψl (kr) + dl χl (kr) (3.49)

Taking cl = 1, dl = −i, we have


r
(1) πρ (1)
rR = ζl (ρ) = ψl (ρ) − iχl (ρ) = H 1 (ρ) (3.50)
2 l+ 2

where H (1) is the Hankel function. This has been acquired through immediate con-
sequences of the well known relationship between the Bessel, Neumann and Hankel func-
(1)
tions, that Jp + iNp = Hp . These Hankel functions have distinguishing property among
cylindrical functions that they vanish at infinity in the complex plane. This very Hankel
function obtained vanishes in the half-plane of the positive imaginary part of ρ, hence
suitable for the representation of the scattered wave.

Cf. [5]
§
Cf. [6]
3.2. Derivation 29

Now we put all the solutions of the equations (3.37) in the equation (3.36), to obtain
the solution of the wave equation:
∞ X
l
(m)
X
rΠ = r Πl
l=0 m=−l
(3.51)
∞ l n o
(m)
X X
= {cl ψl (kr) + dl χl (kr)} Pl (cos θ) {am cos(mφ) + bm sin(mφ)}
l=0 m=−l

where am , bm , cl , dl are arbitrary constants, commonly known as Mie Coefficients.

3.2.6 Mie coefficients


Definitely we would like to know how the various observable quantities vary with the size
and optical properties of the sphere and the nature of surrounding medium. To have some
numerical examples in hand, we should obtain explicit expressions for these coefficients
am , bm , cl , dl . These are four unknown coefficients, i.e. pure mathematically speaking, we
need four independent equations, which are best obtained from four boundary conditions
(3.14). These four expressions have to have equal values at either side of the boundary
surface r = a. These boundary conditions can be written as:

∂ ∂
{r(e Π(i) + e Π(s) )}(r=a) = {r e Π(w) }(r=a) (3.52a)
∂r ∂r
∂ ∂
{r(m Π(i) + m Π(s) )}(r=a) = {r m Π(w) }(r=a) (3.52b)
∂r ∂r
1
k1 {r(e Π(i) + e Π(s) )}(r=a) = II k1 {re Π(w) }(r=a) (3.52c)
1 m (i) m (s) II m (w)
k2 {r( Π + Π )}(r=a) = k2 {r Π }(r=a) (3.52d)
where the superscripts (i), (s), and (w) represents incident wave, scattered wave, and
inside (within sphere) wave, respectively.
We must express the potentials of incident, scattered and inside wave in a series form
of (3.51). We shall first start with a incident wave, where we write the incident wave into
spherical polar coordinates with help of 3.3.2:
I kr cos θ i I k i I kr cos θ
Er(i) = ei sin θ cos φ, Hr(i) = 1k
e sin θ sin φ, (3.53a)
2
(i) I kr cos θ (i) i Ik I
Eθ = ei cos θ cos φ, Hθ = 1 ei kr cos θ cos θ sin φ, (3.53b)
k2
(i) I kr cos θ (i) i Ik I
Eθ = −ei sin φ, Hθ = 1 ei kr cos θ cos φ, (3.53c)
k2
(3.53d)
From the previous section we have Er = e Er + m Er , and with the use of (3.33) we
can write:
I ∂ 2 (r e Π(i) )
ei kr cos θ sin θ cos φ = + I k 2 r e Π(i) (3.54)
∂r2
Using Bauer’s formula

i I kr cos θ
X ψl ( I kr)
e = il (2l + 1) I kr
Pl (cos θ) (3.55)
l=o

and using the following identities


I kr cos θ 1 ∂ i I kr cos θ
ei sin θ ≡ − I kr
(e ) (3.56)
i ∂θ
∂ (1) (1)
Pl (cos θ) ≡ −Pl (cos θ); P0 (cos θ) ≡ 0 (3.57)
∂θ
30 Chapter 3. Mie Scattering

we can finally write the left hand side of the equation (3.54)

i I kr cos θ 1 X l−1
e sin θ cos φ = I i (2l + 1)ψl ( I kr)Pl (cos θ) cos φ. (3.58)
( kr)2
l=o

Let us take similar series as a supposed solution of the equation (3.54)



e (i) 1 X
r Π = I 2 αl ψl ( I kr)Pl (cos θ) cos φ. (3.59)
k
l=o

Now substituting these two series back in the equation (3.54), we obtain the following
relation
∂ 2 ψl ( I kr) ψl ( I kr)
 
I 2 I l−1
αl k ψl ( kr) + = i (2l + 1) (3.60)
∂r2 r2
Interestingly we have encountered a similar kind of equation in the previous section,
which we can write appropriately for this situation with the help of the equation (3.49)
and with cl = 1, d = 0, which yields

ψl ( I kr) = rR, (3.61)

a solution of the equation:

d2 ψl  I 2 α
+ k − ψl ( I kr) = 0, (3.62)
dr2 r2
provided that α = l(l + 1). Now we shall compare (3.62) with (3.60), giving

2l + 1
αl = il−1 . (3.63)
l(l + 1)

The similar calculation follows for the magnetic potential of the incident wave m Π(i) .

Thus, we can write the two potentials of the incident wave as



e (i) 1 X l−1 2l + 1
r Π = I 2 i ψl ( I kr)Pl (cos θ) cos φ (3.64a)
k l(l + 1)
l=o
∞ 1k
1 X 2l + 1
r m Π(i) = I k2
il 1 ψl ( I kr)Pl (cos θ) sin φ (3.64b)
k2 l(l + 1)
l=o

As we can see from the equation above that the equations (3.52) can only be satisfied
if the terms with m = 1 occurs in the expansion (3.51) for the remaining potentials Π(s)
and Π(w) and if a1 = 0 for the magnetic and b1 = 0 for the electric potential.
As explained in the previous section, only functions ψl are appropriate for Π(w) , then
we get

1 X l−1 2l + 1 e
r e Π(w) = II k 2
i Al ψl ( II kr)Pl (cos θ) cos φ (3.65a)
l(l + 1)
l=o
∞ 1
1 X
l k 2l + 1
r m Π(w) = II k II k
i 1 k l(l + 1)
m
Al ψl ( II kr)Pl (cos θ) sin φ (3.65b)
2 2
l=o

We have also seen in the previous section that the function ζ (1) is appropriate to
represent the potential for a scattered wave. It can be seen from the equation (3.50) that
√ (1) (1)
for large ρ, H (1) behaves as eiρ / ρ, which says ζl behaves as eiρ , i.e. R = ζl /r as
I
ei kr /r. It says simply that at larger distances from the sphere, the scattered wave is
3.2. Derivation 31

spherical. For the simplicity in the following discussion, we skip the superscript (1) from
(1)
ζl . So we set the potentials for scattered wave

e (s) 1 X l−1 2l + 1 e
r Π = I 2 i Bl ζl ( I kr)Pl (cos θ) cos φ (3.66a)
k l(l + 1)
l=o

1 X l 1 k 2l + 1
r m Π(s) = Ik Ik
i1 m
Bl ζl ( I kr)Pl (cos θ) sin φ (3.66b)
2 k2 l(l + 1)
l=o

Now we substitute (3.64), (3.65) and (3.66) into the boundary conditions (3.52),

1k
2 e
ψl (1 ka) − e Bl ζl (1 ka) = 2k
Al ψl (2 ka), (3.67a)
2
1k
ψl0 (1 ka) − e Bl ζl0 (1 ka) = 2
e
Al ψl0 (2 ka), (3.67b)
k
1k
ψl (1 ka) − m Bl ζl (1 ka) = 2 m Al ψl (2 ka), (3.67c)
k
1k
2
ψl0 (1 ka) − m Bl ζl0 (1 ka) = 2 m Al ψl0 (2 ka). (3.67d)
k2
In fact, these are linear independent equations between the coefficients, exactly what
we were searching for. But we are only interested in the coefficients e Bl and m Bl , since they
characterize the scattered wave. With the elimination of other unimportant coefficients
we can get the required one:
II k I kψ 0 ( II ka)ψ ( I ka) − I k II kψ 0 ( I ka)ψ ( II ka)
e 2 l l 2 l l
Bl = II k I kψ 0 ( II ka)ζ ( I ka) − I k II kζ 0 ( I ka)ψ ( II ka)
(3.68a)
2 l l 2 l l
I k II kψ ( I ka)ψ 0 ( II ka) − II k I kψ 0 ( I ka)ψ ( II ka)
m 2 l l 2 l l
Bl = I k II kζ ( I ka)ψ 0 ( II ka) − I k I kψ ( II ka)ζ 0 ( I ka)
(3.68b)
2 l l 2 l l

Let us simplify the expression given above. In this case, we have again to recall
the meaning of the constants. We are assuming that the surrounding medium is non-
conducting, i.e. I σ = 0. Let us write only σ instead of I σ. Let λ0 be the wavelength
of the light in the vacuum and I λ in the surrounding medium. Now we have
I iω I 2π I iω 2π
k1 = =i , I k2 = =i (3.69a)
c λ0 c λ0
p 2π √
I I  = f rac2π I λ
k = − I k1 I k2 = (3.69b)
λ0
   
II iω II 4πσ 2π II 4πσ II iω 2π
k1 = +i =i +i , k2 = =i (3.69c)
c ω λ0 ω c λ0
r
II
p 2π II 4πσ
k = − II k1 II k2 = + (3.69d)
λ0 ω
Let us introduce now a complex refractive index relative to each medium
II
n2 II 2
k II
 4πσ II
k1
m2 = I n2
= I k2
= I
+i = , (3.70)
ω I Ik
1

and a dimensionless parameter x and argument y



x= Iλ
a, y = mx. (3.71)

It is interesting to see that x equals to the ratio of the circumference of the sphere
to the wavelength, which is, in fact, very important factor in further calculations. It
can also reduce the argument to Rayleigh scattering, which is valid for the diameter of
the object is one tenth of the wavelength.
32 Chapter 3. Mie Scattering

Now with the given simplification, we can rewrite the Mie coefficients as follows∗

e ψl0 (y)ψl (x) − mψl0 (x)ψl (y)


Bl = (3.72a)
ψl0 (y)ζl (x) − mζl0 (x)ψl (y)
m mψl (x)ψl0 (y) − ψl0 (x)ψl (y)
Bl = (3.72b)
mζl (x)ψl0 (y) − ψl (y)ζl0 (x)

3.2.7 Field vectors of scattered wave


The components of the field vectors of the scattered wave can now be obtained by using
Mie coefficients from (3.68) as follows


1 cos φ X (1)
Er(s) = I 2 2 l(l + 1) e Bl ζl ( I kr)Pl (cos θ), (3.73a)
k r
l=1
∞  
(s) 1 cos φ X e 0 I (1)0 m I (1) 1
Eθ =−I Bl ζl ( kr)Pl (cos θ) sin θ − i Bl ζl ( kr)Pl (cos θ) ,
k r sin θ
l=1
(3.73b)
∞  
(s) 1 sin φ X (1) 1 (1)0
Eφ = − Ik
e
Bl ζl0 ( I kr)Pl (cos θ) m I
− i Bl ζl ( kr)Pl (cos θ) sin θ ,
r sin θ
l=1
(3.73c)


i sin φ X (1)
Hr(s) = I k II k 2
l(l + 1) e Bl ζl ( I kr)Pl (cos θ), (3.74a)
2 r
l=1
∞  
(s) 1 sin φ X e I (1) 1 m 0 I (1)0
Hθ =−I Bl ζl ( kr)Pl (cos θ) + i Bl ζl ( kr)Pl (cos θ) sin θ ,
k2 r sin θ
l=1
(3.74b)
∞  
(s) 1 sin φ X (1)0 (1) 1
Hφ = Ik
e
Bl ζl ( I kr)Pl (cos θ) sin θ + i m Bl ζl0 ( I kr)Pl (cos θ) ,
r sin θ
l=1
(3.74c)

This is the final solution of the boundary value problem.

3.2.8 Intensity and polarization of the scattered light


We shall now examine the intensity and the polarization of the scattered light. As being
only interested in relative values of the intensity, we may take as a measure of the intensity
the square of the real amplitude of the electric vector. Here we shall only consider the
distant field, i.e. r  λ, which allows us to replace the functions ζl and ζl0 by their
asymptotic approximations

ζl (x) ∼ (−i)l+1 eix and ζl0 (x) ∼ (−i)l eix (3.75)



Mie coefficients e Bl and m Bl are denoted as an and bn respectively in the various literature on this
topic: Cf. H.C. van de Hulst, Light scattering by small particles (New York, Dover publications, 1981),
p.123 and Cf. Bohren and Huffman, Absorption and scattering of light by small particles (Weinheim,
Wiley-VCH Verlag, 2004), p.101
3.2. Derivation 33

we obtain for intensity


∞ (1)
!
I λ2 P (cos θ)
0(1)
X
Ipar = | (−i)l e
Bl Pl (cos θ) sin θ − m Bl l |2 , (3.76a)
4π 2 r2 sin θ
l=1
∞ (1)
!
I λ2 P (cos θ) 0(1)
X
l e m
Iperp = | (−i) Bl l − Bl Pl (cos θ) sin θ |2 (3.76b)
4π 2 r2 sin θ
l=1

Then
(s) (s)
|Eθ |2 = Ipar cos2 φ, |Eφ |2 = Iperp sin2 φ (3.77)
We define here the plane of observation as the plane that contains the direction of
propagation of the incident light and the direction (θ, φ) of observation. φ represents
the angle between this plane and the direction of vibration of the electric vector of the
(s) (s)
incident wave. Since either Eθ or Eφ vanishes when φ = 0 or φ = π/2, the scattered
light is linearly polarized when the plane of observation is parallel or perpendicular to the
primary vibrations, whereas for any other direction (θ, φ) the light is normally elliptically
(s) (s)
polarized, since the ratio of Eθ /Eφ is complex. But in case of Rayleigh scattering, it
is always real, so that the scattered light is then linearly polarized for all directions of
observation.

3.2.9 Amplitude functions


To simplify the numerical calculations of intensity, we shall look into the amplitude func-
tions, which for our approximation gives clear understanding of the phenomena. For the
amplitude functions, we have to analyze again field potentials for scattered wave given as
(3.66). The problem we are studying is about the very large distances from the sphere to
the observer compared to l, i.e. l  | I kr|. In this case, we can take asymptotic expression
for ζl ( I kr)
I
ζl ( I kr) ∼ (−i)l+1 ei kr (3.78)
so we get for the scattered field

e (s) i I
X 2l + 1
r Π = − I 2 2 ei kr cos φ e
Bl Pl (cos θ) (3.79a)
k r l(l + 1)
l=o

m (s) i i I kr
X 2l + 1
m
r Π =−I I e sin φ Bl Pl (cos θ) (3.79b)
k k2 r2 l(l + 1)
l=o

Let us introduce the following expressions


dPl (cos θ)
πl (cos θ) = (3.80a)
d cos θ
dπl (cos θ)
τl (cos θ) = cos θπl (cos θ) − sin2 θ (3.80b)
d cos θ
Resulting field components can be put simply as
i I kr
Eθ = H φ = − I k2 r2
ei cos φS2 (θ) (3.81a)
i I kr
−Eφ = Hθ = − I k2 r2
ei sin φS1 (θ) (3.81b)

where

X 2l + 1 e
S1 (θ) = il−1 { Bl πl (cos θ) + m
Bl τl (cos θ)} (3.82a)
l(l + 1)
l=o

X 2l + 1 m
S2 (θ) = il−1 { Bl πl (cos θ) + e Bl τl (cos θ)} (3.82b)
l(l + 1)
l=o
34 Chapter 3. Mie Scattering

S1 (θ) and S2 (θ) are the elements of the scattering matrix (see sec. 3.3.5).
The radical components Er and Hr tend to zero with higher powers of 1/r. With these
amplitude functions we can get intensity and the state of polarization of the scattered wave.
The scattered light is generally elliptically polarized, even if the incident light has linear
polarization, since S1 (θ) and S2 (θ) are complex numbers with different phase.
Due to spherical symmetry, S3 (θ) and S4 (θ) are zero (see sec. 3.3.5), one can obtain
the solution to Mie problem with known i1 = |S1 (θ)|2 and i2 = |S2 (θ)|2 , which is described
in the Chapter 04.
If Iin is the initial flux, then in terms of different polarization, the scattered fluxes are:

i1
Iperp (θ) = Iin (3.83)
k r2
2
i2
Ipar (θ) = Iin (3.84)
k2 r2
i1 + i2
Iunpol (θ) = Iin . (3.85)
2k 2 r2

3.2.10 Scattering cross section


We can recall from the last chapter, that the differential scattering cross section is defined
as the ratio of the scattered flux to the incident flux per unit area:

dσ Isc
= . (3.86)
dΩ Iin

Following from the last subsection, one can write the differential cross section:

dσ 1
= 2 |S2 (θ)|2 cos2 φ + |S1 (θ)|2 sin2 φ

(3.87)
dΩ k

And the total scattering cross section is found by integrating over all solid angles, but the
integral is rather difficult to solve. There is another easier expression for the differential
cross section:

dσ 1  ∗ ∗
= 2 |S+ (θ)|2 + |S− (θ)|2 + (S+ (θ)S− (θ))(cos2 φ − sin2 φ)

(θ) + S− (θ)S+ (3.88)
dΩ 2k

where

S+ (θ) ≡ S1 (θ) + S2 (θ) (3.89)


S− (θ) ≡ S1 (θ) − S2 (θ) (3.90)

Then one can write for the total scattering cross section:
Z   Z +1
dσ π
σsca = dΩ = (|S+ (θ)|2 + |S− (θ)|2 ) d(cos θ) (3.91)
dΩ 2k 2 −1

As one can see, σsca does not provide intensity and phase information like the differential
scattering cross section.
3.3. Appendix 35

3.3 Appendix
3.3.1 Debye potentials and preceeding attempts
Def. If u and v are functions that satisfy the Helmholtz equation

∇2 u + k 2 u = 0 (3.92)

in some domain D and if r is a position vector from the origin, but lying outside D, then
the vector fields

A = ∇ × (∇ × ur) + k∇ × vr, B = ∇ × (∇ × vr) + k∇ × ur (3.93)

satisfy the following Maxwell’s equations

∇ × A = kB, ∇ × B = kA (3.94)

for a time-harmonic electromagnetic field in D. The functions u and v are called Debye
potentials for the electromagnetic field (A, B). Of course, these potentials were invented
by Debye[2] to solve Maxwell’s equation. The procedure is quite peculiar, in the sense, by
assuming that the unknown electromagnetic field has a representation by Debye potentials,
using the condition of the problem to determine what u and v must be, and then verifying
that the field A and B generated by this pair of potentials actually solve the original
problem.
To understand the preceding endeavor to solve such Maxwell’s explanation, it makes
sense to start with an example of the theorem that every electromagnetic field defined in
a spherical domain can be expanded in a series of electromagnetic multipoles. The field is
given as
An = ∇ × (∇ × un r), Bn = k∇ × un r (3.95)
where
Zn+ 1 (kr)
un (r) = √2 Sn (θ, φ) (3.96)
r
Sn is a surface spherical harmonic, and Zn+ 1 is a Bessel function of order n + 12 .
2
The first such attempt to do the expansions in spherical harmonics was due
P∞to A.Clebsch(1863),
where he showed that the vector field A can be expanded into a series A = n=0 An where

vn (r)
An (r) = αn (r)∇un (r) + βn (r)r2n+1 ∇ + γn (r)r × ∇wn (r) (3.97)
r2n−1
where αn , βn , and, γn (r) are functions of the spherical coordinate r alone, while un , vn , and, wn (r)
are solid harmonics of degree n. If A is P the electric or magnetic part of a time-harmonic

electric field then the expansion A = n=0 An may be shown to be equivalent to a
multipole expansion of A.
A very similar method was developed by C.W.Borchardt(1873), where he considered
a vector fields A satisfying

∇ × (∇ × A) = 0, ∇ · A = 0 (3.98)

showing that every field could be given as

A = ∇u + r × ∇v (3.99)

where u and v are solutions of Laplace’s equation. In fact, these solutions, expanded in
solid harmonics, are substituted in (3.99) to obtain the quite similar expansion of A as
that of Clebsch expansion.
36 Chapter 3. Mie Scattering

Another closely related method was due to H.Lamb(1881). He considered the following
equations
∂u ∂v ∂w
(∇2 + h2 )u = 0, (∇2 + h2 )v = 0, (∇2 + h2 )w = 0, + + = 0, (3.100)
∂x ∂y ∂z
which are, in fact, equivalent to the vector equation

∇ × (∇ × A) − h2 A = 0 (3.101)

for the vector field A with u, v, and w as rectangular components. After seeing carefully,
it is nothing but the more generalized case of (3.98) devised by Brochardt, and actually it
was so, when
P Lamb calculated it in his way. He said: every solution of A may be expanded
as A = ∞ n=0 (A n + A 0 ) where
n

φn (r)
An (r) = ψn−1 (hr)∇φn (r) + ψn+1 (hr)r2n+3 ∇ ,
r2n+1 (3.102)
A0n (r) = ψn (hr)r × ∇χn (r)

χn and φn are solid harmonics whereas ψn (x) is a solution of

d2 ψn 2(n + 1) dψn
+ + ψn = 0. (3.103)
dx2 x dx
This expansion is equivalent to the Clebsch expansion for solutions of (3.101). He remarked
on the fundamental role of radial components

φ = r · A, ψ = r · (∇ × A) (3.104)

He discovered that φ and ψ determine A uniquely, are solutions of the Helmholtz equation
and the harmonics are simply multiples of the harmonics occurring in expansions of φ and
ψ. Which leads to the conclusion that any problem involving the vector equations (3.101)
can be reduced to a pair of scalar problems.
J.J.Thomson(1893) was the first to use this method in the context of electromagnetic
problems to find the wave scattered by a perfectly conducting sphere when it is illuminated
by a plane electromagnetic wave. It was in 1908, when G.Mie extended this problem to
spheres having arbitrary conductivity and dielectric constant. Unlike the earlier writers,
Mie used the unknown functions as the components of the field A relative to a system of
spherical coordinates rather than the rectangular components. Like Lamb, Mie had taken
the radial components r · A and r̂ · B (where kB = ∇ × A) as the principal unknown of
his problem. Of course, in such cases, the spherical components of A and A0 are simpler
to describe than rectangular ones.
Borchardt’s theorem differs from that of Clebsch and Lamb in the sense that it repre-
sents the general solution of his system of differential equations (3.98) in terms of a pair
of scalar functions, with absolutely no explicit reference to spherical harmonics. And it
was in 1909, that the analogous representation for solutions of Maxwell’s equations was
obtained by P.Debye in a paper concerning the pressure exerted by electromagnetic waves
on spheres of arbitrary material. But he, like G.Mie, used the spherical components of the
electromagnetic fields. On adding the fields A and A0 one obtains Debye’s representation
of an electromagnetic fields in terms of a pair of potentials u and v. In fact, these fields A
and A0 can be written in the simpler vector representation in relations to the potentials∗

A = ∇ × (∇ × ur), A0 = k∇ × vr (3.105)

in compliance with the definition mentioned at the beginning. This is how the history
promoted Debye at the helm of idea of using the scalar potentials to simply the quite
often problems involving Maxwell’s equations.

thanks to the work of Sommerfeld[7]
3.3. Appendix 37

3.3.2 Vector representations in spherical polar coordinates


The curvilinear coordinates can be represented in spherical polar coordinates as

x = r sin θ cos φ (3.106a)


y = r sin θ sin φ (3.106b)
z = r cos θ (3.106c)

The components of vector A are transformed from cartesian to polar coordinates as

Ar = Ax sin θ cos φ + Ay sin θ sin φ + Az cos θ (3.107a)


Aθ = Ax cos θ cos φ + Ay cos θ sin φ − Az sin θ (3.107b)
Aφ = −Ax sin φ + Ay cos φ (3.107c)

Applying these formulae to the vector curl A


 
1 ∂(r sin θAφ ) ∂(rAθ )
(curl A)r = 2 − (3.108a)
r sin θ ∂θ ∂φ
 
1 ∂Ar ∂(r sin θAφ )
(curl A)θ = − (3.108b)
r sin θ ∂φ ∂r
 
1 ∂(rAθ ) ∂Ar
(curl A)φ = − (3.108c)
r ∂r ∂θ

3.3.3 Associated Legendre functions


The Legendre polynomials are the polynomials

l/2
X (2l − 2m)!
Pl (x) = (−1)m xl−2m (3.109)
2l m!(l− m)!(l − 2m)!
m=0

and the associated Legendre functions of the first kind are defined by

dm
Plm (x) = (1 − x2 )m/2 Pl (x) (3.110)
dxm
Notice that there is an arbitrariness in phase associated with the definition and some
authors may insert a factor (−1)m on the right-hand side of equation (3.110). We shall
adopt the convention to insert the phase factor into the definition of the spherical har-
monics below, in fact, either way the same phase factor will arise in the latter. We shall
need the following relations

(1) 1
Pl (cos θ) = [Pl−1 (cos θ) − cos θPl (cos θ)], (3.111)
sin θ  
(1) cos θ (1) Pl (cos θ)
Pl (cos θ) = P (cos θ) − l(l + 1) (3.112)
sin2 θ l sin θ

3.3.4 Cylindrical functions


Small values of argument x For small values of x, one can write for ψl (x)

xl+1
ψl (x) = fl (x) (3.113)
1 · 3 · . . . · (2l + 1)

where
2  x 2
fl (x) = 1 − + ... (3.114)
2l + 3 2
38 Chapter 3. Mie Scattering

(1)
For ζl (x), one has

(1) 1 · 3 · . . . · (2l − 1) ix
ζl (x) = −i e [hl (x) − ixgl (x)], (3.115)
xl
where hl (x) and gl (x) are power series of which the first term is unity and the second term
is quadratic. Analogous, one can express the functions ψl0 (x) and ζl0 (x):

(l + 1)xl
ψl0 (x) = f + (x) (3.116)
1 · 3 · . . . · (2l + 1) l
1 · 3 · . . . · (2l − 1) ix +
ζl0 (x) = il e [hl (x) − ixgl+ (x)] (3.117)
xl+1
where fl+ (x), h+ +
l (x) and gl (x) are power series of the same kind as before.

Large values of argument x For larger values of x, provided that l is small compared
with |x|, then the following asymptotic relations can be used:
1 l+1 −ix
ψl (x) ∼ [i e + (−i)l+1 eix ], (3.118)
2
(1)
ζl (x) ∼ (−i)l+1 eix , (3.119)

and
1 l −ix
ψl0 (x) ∼[i e + (−i)l eix ], (3.120)
2
0(1)
ζl (x) ∼ (−i)l eix , (3.121)

3.3.5 Scattering matrix


The linearity of Maxwell’s equations implies that the scattering process mixes the compo-
nents E1 and E2 linearly (see figure 3.1). Such a process is described in a matrix context,
allowing us to write for the far field amplitudes

eikr
    
E2 S2 S3 E2
= (3.122)
E1 sc ikr S4 S1 E1 in

which defines the scattering amplitude matrix S.†


But a simplification for spherical symmetry gives the scattering-amplitude matrix the
following way:
eikr
    
E2 S2 0 E2
= (3.123)
E1 sc ikr 0 S1 E1 in
It can be explained as follows: the reflection of the scatterer through the scattering plane
is equivalent to reflecting everything but the scatterer; then (E1 )in and (E1 )sc change sign,
but (E2 )in and (E2 )sc do not; hence, S3 and S4 must change sign. But, if the scatterer
is completely symmetric under reflection, S3 and S4 can not change sign, and this way,
S3 = S4 = 0.


S is not what is usually termed the ’scattering matrix’, but rather a matrix of scattering amplitudes.
Bibliography

[1] Gustav Mie. Beiträge zur Optik trüber Medien, speziell kolloidaler Metallösungen.
Ann. d. Phys., 25:377–445, 1908.

[2] P. Debye. Der Lichtdruck auf Kugeln von beliebigem Material. Ann. d. Phys., 30:57–
136, 1909.

[3] Calvin H. Wilcox. Debye potentials. 1957.

[4] E.M.Purcell. Electricity and Magnetism. McGraw-Hill, New York, 1963.

[5] Arnold Sommerfeld. Partial Differential Equations of Physics, page 127. Academic
Press, 1949.

[6] Arnold Sommerfeld. Partial Differential Equations of Physics, page 86. Academic
Press, 1949.

[7] Arnold Sommerfeld. Elektromagnetische Schwingungen, volume Die Differential- und


Integralgleichungen der Mechanik und Physik. Vieweg, Braunschweig, 1927.

39
Chapter 4

Approximation for X-rays

41
42 Chapter 4. Approximation for X-rays

4.1 Introduction
The practical importance of approximate theories of electromagnetic scattering by small
particles fades away as various exact techniques mature and become even more powerful.
This is absolutely true in case of Mie theory, which provides the exact solution to particles
of any size. The computational power has developed in last few decades and will be the best
tool to analyze the theory in the future. Nonetheless, approximate theories still remain
a valuable source of physical insight into the processes of scattering of electromagnetic
radiation. In this chapter, we are going to devote ourselves to both of these methods.
The last two chapters have given comprehensive overview of two different scattering
explanations available on the ground for small particles. What we are going to do here
is to develop those theories in the context of X-ray region. The analytical approximation
matching to our prerequisites will be mainly dealt with in the next section. In the latter
part of this chapter, the computational methods will be elaborated, where the powerful
mathematical tools like Mathematica and IDL are used.

4.2 Analytical Approximation


In the context of Mie theory, which is an exact theory in itself, various approximations
have been proposed, because the numerical techniques consume a lot of computational
time. Mainly they were divided with respect to the refractive index and size parameter.
Traditionally X-ray scattering can be confined to the idea of Rayleigh-Gans scattering,
since this approximation in optical problems is nothing but the Born approximation, which
we have studied extensively in kinematic scattering theory. But in this subsection, we will
work with a different scattering approximation which gives more accurate result with
respect to the exact Mie theory.
In the last chapter we have derived amplitude functions (see sec. 3.2.9), from which
the intensities are calculated. The best way to attain approximation is to approximate
these amplitude functions to find solutions for intensities in X-ray regions, which we shall
be doing in the subsequent discussion.

4.2.1 Limiting cases

As discussed above, we are going to consider a complex refractive index with m → 1.


In this case the scattering is small. If m − 1 is very small, it makes no difference in the
scattering pattern whether m − 1 is < 0 or > 0. Involving m → 1, there are two limiting
cases, one is Rayleigh-Gans scattering and other the anomalous diffraction[1]. The limiting
cases for m → 1 can very well be illustrated in the figure 4.1
Let us write ρ = 2x(m − 1). For the very small values of x and ρ we get Rayleigh-Gans
theory, where x is kept fixed and transition is made for m → 1. On the other hand,
the theory of anomalous diffraction is based on keeping ρ fixed and making the same
transition. Here x → ∞ gives definite values for cross-sections and the scattering patterns
runs through a set of ”homologous diagrams”, which will be discussed shortly.
As could be seen in the figure 4.1, Rayleigh-Gans domain can be described by inter-
ference of light scattered independently by all volume elements, where the anomalous-
diffraction domain can be described as straight transmission and subsequent diffraction
according to Huygen’s principle.
Although for x → ∞ the anomalous diffraction is the limiting case, there could be two
different limiting cases within, namely ρ → 0 and ρ → ∞. What we are interested in is
the case for ρ → 0, so-named intermediate case, whereas ρ → ∞ describes geometrical
optics.
4.2. Analytical Approximation 43

x small x large x large

ρ small ρ small ρ large

Rayleigh Intermediate Geometrical


scattering case optics plus
diffraction

Rayleigh-Gans scattering Anomalous diffraction

Figure 4.1: Limiting cases in which m → 1

4.2.2 Anomalous diffraction theory


The anomalous diffraction theory is a brain-child of van de Hulst [1], which was primarily
developed to obtain the extinction of particles in interstellar space. Later the theory has
been applied to describe the scattering properties of many different types of particles.
The anomalous diffraction theory is based on the premise that the extinction of light by
a particle is primarily a result of the interference between the rays that pass through
the particle with those that do not. This interference takes place at some point very far
from the particle and gives rise to extinction. It is based on two assumption as described
in previous sub-subsection. The assumption |m − 1|  1 states that the refraction and
reflection of the ray passing through the particle are negligible. And the assumption x  1
allows one to trace the wave, in the form of ray, through the particle.
These assumptions imply that the ray cannot be traced through the sphere nor one
can observe deviations at two boundaries it crosses, i.e. the ray is virtually straight. The
energy reflected at the boundaries is negligible, since the Fresnel reflection coefficient goes
to 0, when m → 1. That means the field is not changed in amplitude, but only in phase.
From the figure 4.2, it can be seen that the path traveled in the medium wih refractive
index m is 2a sin τ . The phase lag observed at the point Q is
2a sin τ · (m − 1) · 2π/λ = ρ sin τ (4.1)
where ρ = 2x(m − 1). The physical meaning of ρ is the phase lag by the central ray that
passes through the sphere along a diameter.
Now the field in the entire plane V is known. If we put it to 1 in all points outside the
geometrical shadow circle, it is e−iρ sin τ at a point, which is at a distance of a cos τ from
the center of the circle, for phase lag corresponding to a negative imaginary exponent. So
the field added to the field of the original wave only inside the shadow circle is e−iρ sin τ − 1.
This added field determines the scattered wave. For an opaque body only the term −1 is
present and gives according to [2]
k2
S(0) = G (4.2)

where Z Z
G= dx dy = πa2 (4.3)
area
44 Chapter 4. Approximation for X-rays

n+½

Incident Ray

Figure 4.2: Ray passing through the sphere

G is the area of geometrical shadow and generalizing this term for our case it gives

k2
Z Z
S(0) = (1 − e−iρ sin τ )dx dy (4.4)
2π circle

The same result can be obtained using Babinet’s principle


Now we use polar coordinated inside the shadow circle.

−2πa cos τ d(a cos τ ) = 2πa2 cos τ sin τ dτ (4.5)


Z π/2
2 2
S(0) = k a (1 − e−iρ sin τ ) cos τ sin τ dτ (4.6)
0

with
Z π/2
K(w) = (1 − e−iρ sin τ ) cos τ sin τ dτ (4.7)
0
1 e−w e−w − 1
= + + (4.8)
2 w w2

So the result is[2]


S(0) = x2 K(iρ) (4.9)

and∗
4
Qext = Re[S(0)] = 4Re[K(iρ)] (4.10)
x2

Qext is the extinction factor, i.e. how much energy has been lost from the incident flux to the area of
the surface. It is widely used tool to analyze Mie scattering. More about this and the same absorption
factor Qabs and scattering factor Qsca can be found in the book from van de Hulst[2]
4.2. Analytical Approximation 45

Complex refractive index


Let us denote the refractive index by m = n − in0 . Since m → 1, both n − 1 and n0 must
be  1. We write,
n0
= tan β (4.11)
n−1
Now the phase shift of a ray passing through the center of the sphere is

ρ∗ = 2x(m − 1) = ρ(1 − i tan β) (4.12)

Here, the real part denotes an actual phase shift and imaginary the decay of amplitude.
From the above discussion, we can rewrite for the efficiency factor for extinction

Qext = 4Re[K(iρ + ρ tan β)] (4.13)


∗ 0
The phase shift factor e(−iρ sin τ ) contains a factor e−2xn sin τ for the decrease in am-
0
plitude, which consequently leads to decrease in intensity by e−4xn sin τ and the fraction
0
of 1 − e−4xn sin τ of the original ray is absorbed within the sphere. The absorbed fraction
of the total energy incident on the sphere is found to be

Qabs = 2K(4xn0 ) (4.14)

where the argument can also be written as 4xn0 = 2ρ tan β = 2aγ with radius a and γ as
the absorption coefficient of its material per unit length.
When Qabs approaches 1, hardly any light gets through the sphere and nothing inter-
feres with the ordinary diffraction and we have simply Qext = 2, which holds for any large
opaque body.
The scattered fraction of the total incident energy, i.e. Qsca can be easily calculated
taking subtraction of Qabs from Qext . So here we get

Qsca = 2(2Re[K(iρ + ρ tan β)] − K(2ρ tan β)) (4.15)

But in our case, we are considering small values of ρ, then we can use the series
expansion of K(w) for small w

w w2 w3
K(w) = − + − ... (4.16)
3 8 30
So we get
4 1
Qext = ρ tan β + ρ2 (1 − tan2 β) (4.17)
3 2
4
Qabs = ρ tan β + ρ2 tan2 β (4.18)
3
and for scattered fraction
1
Qsca = ρ2 (1 + tan2 β) (4.19)
2

4.2.3 Diffraction patterns


Not only the amplitude function S(0) is affected due to the interference between diffracted
and transmitted light, but also the function S(θ) for which the diffraction is appreciable
for all values of θ. To derive this function, we shall make use of rigorous Mie theory and
we shall know in the end that the result can well be acquired from Huygen’s principle.
We know the Mie coefficients e Bl and m Bl from last chapter (3.72), which we shall
denote as an and bn in the upcoming discussion for simplicity. These coefficients can be
written as
1 1
an = (1 − e−2iαn ), bn = (1 − e−2iβn ) (4.20)
2 2
46 Chapter 4. Approximation for X-rays

They consist of two terms: one independent of the nature of particle and another dependent
on it. The term 1 gives the Fraunhofer diffraction pattern and the term e−2iαn and e−2iβn
the scattering by reflection and refraction. This separation can be useful only if αn and
βn are large and should be made only for the terms with n + 12 < x (localization principle
[3]).
In the limit of x → ∞ and mx → ∞, αn can be approximated as[4]

1 − ir2 exp(2ix0 f 0 )
exp(−2iαn ) = exp[2i(xf − x0 f 0 )] (4.21)
1 + ir2 exp(−2ix0 f 0 )
where
sin τ − m sin τ 0 m sin τ − sin τ 0
r1 = , r2 = (4.22)
sin τ + m sin τ 0 m sin τ + sin τ 0
and
f = sin τ − τ cos τ, f 0 = sin τ 0 − τ 0 cos τ 0 , x0 = mx (4.23)
The angle of incidence τ is related to τ 0 and n through the equations
1
x cos τ = x0 cos τ 0 = n + (4.24)
2
The phase (xf − x0 f 0 ) may the be expressed as

xf − x0 f 0 = −xη + x(τ 0 − τ ) cos τ, (4.25)

where η = m sin τ 0 − sin τ .


If u = (n + 21 )θ is fixed and n goes to infinity, the asymptotic formulae for spherical
harmonics appearing in amplitude functions in the last chapter are
1
πn (cos θ) ' n(n + 1)[J0 (u) + J2 (u)], (4.26)
2
1
τn (cos θ) ' n(n + 1)[J0 (u) − J2 (u)], (4.27)
2
which for u  1 or θ  1/x may further be approximated as
1
πn (cos θ) ' τn (cos θ) ' n(n + 1)J0 (u) (4.28)
2
For very large x one can make the following substitution
X Z π/2
=x sin τ dτ (4.29)
n 0

After all the calculation the scattering function for near forward scattering assumes
the following form
Z   
2 exp(−2iα(τ )) + exp(−2iβ(τ ))
S(θ) = x 1− J0 (xθ cos τ ) cos τ sin τ dτ (4.30)
2

where S1 (θ) ' S2 (θ) ≡ S(θ).


Again here the term 1 give the Fraunhofer diffraction and the remaining terms give
the scattering by reflection and refraction. The later part of S(θ) can also be expressed as


x2
Z
(r1 + r2 ) X
A(θ) = − exp[2i(xf − x0 f 0 )]  − (1 − r12 ) (−r1 ν)p−1 − (1 − r22 )
2 ν
p=1
 (4.31)
X∞
(−r2 ν)p−1  J0 (xθ cos τ ) cos τ sin τ dτ
p=1
4.3. Numerical approximation 47

with ν = iexp(−2ix0 f 0 ).
Although it looks more complicated, it is one of the best approximations available
yet. The original idea is from Debye[5]. The extension of which provides interesting
insight into Rainbow theory, which investigates particles of very large size. The validity
of anomalous diffraction theory with Mie theory has been scrutinized by Farone[6] and
Sharma[7]. The algorithms are developed in this direction for numerical calculations,
mostly in the applications in environmental physics or astrophysics[8]. Direct calculation
of anomalous diffraction theory in the context of X-ray scattering, to my knowledge, has
not been done yet.

4.3 Numerical approximation


4.3.1 Using Mathematica
Mathematica is a computer algebra system of high-level programming language emulating
multiple paradigms on the top of term-rewriting (more in sec. 5.4.1). In our calcula-
tion of the exact Mie theory, it comes very handy, knowing that the solution contains
Bessel, Ricatti-Bessel functions and Legendre polynomials. Mathematica has these func-
tions built-in, which makes writing the algorithm presumably very straight forward.
Since, in the X-ray, the analysis is confined to reciprocal space, there was a need to
rewrite the all those Ricatti-Bessel functions and Legendre polynomials in reciprocal space.
Although these functions are selected functions of the family of hypergeometric functions,
it was unfortunately not possible to write them in simple readable forms. For this reason,
they were kept dependent on the scattering angle θ, whereas θ is defined as the function
of the momentum-transfer vector Q, which is itself defined in sec. 2.3.3. So the relation
between θ and Q is given as:  
Q
θ = 2 arcsin (4.32)
2k
where k is the wave vector in the vacuum.
The relative intensity that we are interested in are squares of the amplitude functions
(3.82), which can be obtained from Mie coefficients (3.72) and angle-dependent functions
(3.80). Mie coefficients are composed of Ricatti-Bessel functions, which can be given
symbolically in Mathematica, so is the case with angle-dependent functions πn and τn ,
which are nothing but Legendre polynomials and its derivatives, respectively. Legendre
polynomials are in-built functions in Mathematica as well. The detailed algorithm with
the accompanying theory has been reproduced extensively in Chapter 7, precisely in the
section sec. 7.1.

4.3.2 Using IDL


The algorithm used to calculate with IDL is due to Bohren & Huffman[9]. In this pro-
gramme, unlike in Mathematica, the convergence of series has to be justified. After the
extensive study by Wiscombe[10], the stopping term for the series has been introduced,
which is, in fact, the integer closest to

x + 4x1/3 + 2 (4.33)

where x is the largest size parameter among the size parameters calculated for all the
medium coming into questions. This stopping term has been used for all the series calcu-
lations concerned.
Fortunately, recursive techniques are available for the calculations of Mie coefficients
(3.72) and angle-dependent functions (3.80), required to calculate the amplitude functions
(3.82). In next sub-subsections the respective recursive techniques will be put into details.
48 Chapter 4. Approximation for X-rays

Recursion for Mie coefficients


Like in the section before, here I shall stay attached to the popular convention of repre-
senting Mie coefficients of an and bn for e Bl and m Bl respectively detailed in the last
chapter(3.72). Mie coefficients are rather complicated functions of the spherical Bessel
functions and their derivatives, the argument of which are generally complex. But the
spehrical Bessel function satisfy the recurrence relations

2n + 1
zn−1 (ρ) + zn+1 (ρ) = zn (ρ) (4.34)
ρ
d
(2n + 1) zn (ρ) = nzn−1 (ρ) − (n + 1)zn+1 (ρ) (4.35)

where zn is either a Bessel function of first or second kind. So in this algorithm, we are
tempted to start with n = 2. Moreover, these functions have simple trigonometric first
orders, which makes the calculation easier. The logarithmic derivative was first introduced
by Aden[11] in the context of computing coefficients for a sphere.

d
Dn (ρ) = lnψn (ρ) (4.36)

Now, we may rewrite the Mie coefficients as

[Dn (mx)/m + n/x]ψn (x) − ψn−1 (x)


an = (4.37)
[Dn (mx)/m + n/x]ξn (x) − ξn−1 (x)
[mDn (mx) + n/x]ψn (x) − ψn−1 (x)
bn = , (4.38)
[mDn (mx) + n/x]ξn (x) − ξn−1 (x)

using the recurrence relation

nψn (x) nξn (x)


ψn0 (x) = ψn−1 (x) − , ξn0 (x) = ξn−1 (x) − (4.39)
x x
And the logarithmic derivative in itself satisfies the recurrence relation

n 1
Dn−1 = − (4.40)
ρ Dn + n/ρ

which are the consequences of recurrence relation (4.34). There are two possible schemes
for calculating Dn (mx), but in this programme we shall concentrate only on downward
recurrence, i.e. lower order generated from higher order. Beginning with an estimate for
Dn , where n is larger than number of terms required for convergence, successively more
accurate lower-order logarithmic derivatives can be generated. So here we shall begin with
DNMX . Let NMX be sufficiently greater than stopping term and |mx|. Thus, NMX is
taken to be M ax(NSTOP, |mx|) + 15 and the recurrence begins with DNMX = 0.0 + i0.0.
Both ψn and ξn satisfy

2n + 1
ψn+1 (x) = ψn (x) − ψn−1 (x) (4.41)
x
where ξn = ψn − iχn , with the initial values for recurrence of

ψ−1 (x) = cos x, ψ0 (x) = sin x, (4.42)


χ−1 (x) = − sin x, χ0 (x) = cos x. (4.43)

The downward recurrence is computationally stable method than forward recurrence,


which was shown by Kattawar and Plass[12].
4.3. Numerical approximation 49

Recursion for angle-dependent functions


The angle-dependent functions πn and τn can as well be calculated using recurrence rela-
tionship, given as
2n − 1 n
πn = µπn−1 − πn−2 (4.44)
n−1 n−1
τn = nµπn − (n + 1)πn−1 , (4.45)

where µ = cos θ. In case of angle-dependent functions, the upward recurrence is enough


for calculation, unlike in case of Mie coefficients, where downward recurrence was the time
saving solution. The initial values for recurrence will be

π0 = 0, π1 = 1. (4.46)

Due to the following relationship, these functions should only be calculated between 0◦
and 90◦ ,
πn (−µ) = (−1)n−1 πn (µ), τn (−µ) = (−1)n τn (µ). (4.47)
Bibliography

[1] H.C. van de Hulst. Light Scattering by Small Particles, chapter 11, page 172. Dover
Publications Inc., New York, 1981.

[2] H.C. van de Hulst. Light Scattering by Small Particles, chapter 4, page 30. Dover
Publications Inc., New York, 1981.

[3] H.C. van de Hulst. Light Scattering by Small Particles, chapter 12, page 208. Dover
Publications Inc., New York, 1981.

[4] H.C. van de Hulst. Light Scattering by Small Particles, chapter 12, page 210. Dover
Publications Inc., New York, 1981.

[5] P. Debye. Der Lichtdruck auf Kugeln von beliebigem Material. Ann. d. Phys., 30:57–
136, 1909.

[6] W.A.Farone and M.J.Robinson III. The range of validity of the anomalous diffraction
approximation of electromagnetic scattering by a sphere. Applied Optics, 7(643-645),
1968.

[7] S.K.Sharma. On the validity of the anomalous diffraction approximation. Journal of


modern optics, 39(11):2355–2361, 1992.

[8] B.T.Draine and Khosrow Allaf-Akbari. X-ray scattering by nonspherical grains.i.


oblate spheroids. The Astrophysical Journal, (652):1318–1330, 2006.

[9] Craig F. Bohren and Donald R. Huffman. Absorption and Scattering of Light by
Small Particles. Wiley-VCH, 2004.

[10] W. J. Wiscombe. Improved Mie scattering algorithms. Applied Optics, 19:1505–1506,


1980.

[11] A.L. Aden. Electromagnetic scattering from spheres with sizes comparable to the
wavelength. Journal of Applied Physics, 22:601–605, 1951.

[12] G. W. Kattwar and G. N. Plass. Electromagnetic scattering from absorbing spheres.


Applied Optics, 6:1377–1382, 1967.

51
Chapter 5

Comparison

53
54 Chapter 5. Comparison

In this up-coming discussion, the exact Mie theory and the scalar kinematical theory
will be compared with the help of graphs generated by Mathematica, using the code given
in sec. 7.1. This is the direct comparison between the relative intensity graphs generated
by the Mie solution and the solution given by the scalar kinematical theory. For the sake of
uniformity, only the relative intensity with perpendicular polarization has been considered
in graphs concerning Mie theory.
As we have seen in the preceding chapters, the size parameter x is the most crucial
factor in Mie theory. The size parameter is given as
2πa
x= (5.1)
λ
where a is the radius of the sphere and λ is the wavelength in the vacuum. For achieving
clearer understanding, I shall be analyzing only with fixed wavelength and only one type
of material, but with varying radii. It shall be interesting to know how far the scalar
theory matches the exact vector theory. It is always useful to keep in mind that scalar
theory, i.e. kinematical theory, views a sphere as a disk, whereas the vector theory, i.e.
Mie theory, encompasses three-dimensional structure. So it is interesting to know to what
size of matter the assumption of taking sphere as a disc is plausible.

5.1 Comparing theories


In this analysis, the wavelength in vacuum is kept constant at 1Å. We shall be considering
the gold spheres of various radii, giving us different size parameters to compare. The
refractive index n = 1 − δ + iβ, we have for the gold:

δ = 1.9 · 10−5
β = 2.5 · 10−6

The density appearing in the equation (2.49) is the electron density. It can be calcu-
lated using the relation between δ and ρ given in (2.6). It turns out to be ρ = 4233.351/nm3
for Gold.
We shall compare here the relative intensity given by the kinematical theory to the
relative intensity given by Mie theory. In case of graphs with Mie theory, only the perpen-
dicular polarization has been taken into account. It is usual in X-ray theories to compare
intensities against the scattering wave vector Q, and we shall stick to this tradition. The
relative intensity for the kinematical theory is given by the relation

ikin (Q) = |f (Q)|2 (5.2)

where f (Q) is the scattering amplitude, in our case, of a sphere (see the equation (2.49)).
Whereas the relative intensity for the Mie theory is given by

imie (Q) = |S(Q)|2 (5.3)

where S(Q) is the amplitude function described in sec. 3.2.9∗ .


We shall also compare the differential scattering cross section given by both theories
versus Q. We shall be taking the azimuthal angle φ = π2 † for a convenience. To put it in
a simple form, the differential scattering cross section by the kinematical theory given by,

= r02 |f (Q)|2 (5.4)
dΩ

In the subsection of sec. 3.2.9, the amplitude function has been given in the dependancy of scattering
angle θ, and θ relates to the scattering wave vector Q with Q = 4π λ
sin(θ/2), where λ is the wavelength in
the medium.

see sec. 3.2.10 for more information on azimuthal angle in the scattering
5.1. Comparing theories 55

and by the Mie theory,


dσ 1
= 2 |S(Q)|2 . (5.5)
dΩ k
The reference to graphs are tabulated with respect to taken radius in the following:

Radius [nm] Intensity graphs Scattering cross section graphs


0.5 fig. 5.1 fig. 5.6
1.0 fig. 5.2 fig. 5.7
2.5 fig. 5.3 fig. 5.8
5.0 fig. 5.4 fig. 5.9
10.0 fig. 5.5 fig. 5.10

1e+10

Ikin
Imie

1e+05

1e-05

1e-10

1e-15
0 50 100
Q [1/nm]

Figure 5.1: Relative intensity graph for the sphere with radius 0.5 nm

1e+10

I kin
I mie

1e+05

1e-05

1e-10

1e-15
0 20 40 60 80 100 120
Q [1/nm]

Figure 5.2: Relative intensity graph for the sphere with radius 1.0 nm
56 Chapter 5. Comparison

1e+15

Ikin
Imie

1e+10

1e+05

1e-05

1e-10
0 20 40 60 80 100
Q [1/nm]

Figure 5.3: Relative intensity graph for the sphere with radius 2.5 nm

It can be seen from the differential scattering cross section graphs that they match
almost on each other, except for the graphs with size parameter of 100, where some
internal calculation error occurs in Mathematica (see sec. 5.2.2). Although, the graphs
looks quite the same, there is small difference between maxima, which is almost of the same
order for each graph. It was tested taking the differential cross section for kinematical and
Mie graphs for the lowest possible value of Q. It means that we have to find the function
Imie (Q → 0) and Ikin (Q → 0) which gives us the differential cross section for Q = 0.

Imie (Q → 0) For Q to be 0, θ should be 0. We have to find amplitude function S1 (θ = 0).


It is known from the last chapter (see 3.2.9) that for cos θ = 1, we get
n(n + 1)
πn (1) = τn (1) = (5.6)
2
Then we can write simply for S1 (0)

1X
S1 (0) = (2l + 1)(an + bn ) (5.7)
2
l=o

It gives us the differential scattering cross section for Q = 0 as


2


dσ 1 1 X
= 2 (2l + 1)(an + bn ) (5.8)

dΩ k 2
l=o

It solves the problem with long calculation time required in Mathematica, since the most
time-consuming term has been angle-dependent functions, which are absent in the expres-
sion above.

Ikin (Q → 0) In case of Ikin (Q), the lim has been applied on the Q-dependent term. It
gives
(sin(QR) − QR cos(QR))2 1
lim 6
= 0.1111 = (5.9)
Q→0 (QR) 9
The final expression for the differential scattering cross section looks like
 2
dσ 2 2 4 3
= r0 ρ πr (5.10)
dΩ 3
5.2. Errors and problems 57

1e+15

Ikin
Imie
1e+10

1e+05

1e-05

1e-10
0 10 20 30 40 50
Q [1/nm]

Figure 5.4: Relative intensity graph for the sphere with radius 5.0 nm

dσ dσ
 
The figure 5.11 gives the ratio between dΩ kin
and dΩ mie
. It can be seen from the
figure 5.11 that the ratio decreases with increasing radius, i.e increasing size parameter.
But the linear regression
 on it dσ shows that it has slope of 2.5 · 10−5 , i.e. hardly any


difference between dΩ kin and dΩ mie . The absolute values of the differential scattering
cross section of lower radii are way smaller than for higher radii, which means that this
slope is the result of linearly increasing computational anomaly as the radius increases.
So we can say that


dΩ mie

 '1 (5.11)
dΩ kin

We can substitute (5.5) and (5.4) in the ratio given above, giving us

imie (Q)
' k 2 r02 (5.12)
ikin (Q)

where r0 is the classical electron radius and k is the wave vector in the surrounding
medium. It gives us interesting result that the ratio between relative intensities of two
different theory is (kr0 )2 . Analogous to the size parameter, where x = kR, the same way,
kr0 can be termed as the classical electron size parameter. It should be noticed that this
discussion is based on the results acquired upto the size parameter as large as 500. It seems
to be valid in this range of size parameter. Unfortunately, due to problems discussed later,
it was not possible to calculate for larger size parameters than of 500.
In fact, the most interesting part of the Mie theory is the intensity graphs for the higher
Q, which was unfortunately not possible. Mie theory takes care of all the phenomena like
internal reflection, backscattering, etc. It might have been possible to see some differences
with the kinematical theory. In the up-coming sections, it will be explained why the
calculation for single value of Q for larger particles takes enormous time.

5.2 Errors and problems


5.2.1 Errors
The possible error in the calculation can be observed in the graph with the size parameter
of 100. No direct statement can be made, regarding the error observed. The most possible
58 Chapter 5. Comparison

1e+15

Ikin
Imie

1e+10

1e+05

1e-05

1e-10
0 5 10 15 20 25
Q [1/nm]

Figure 5.5: Relative intensity graph for the sphere with radius 10.0 nm

explanation would be the limit of the algorithms used lying beneath the infinite series
containing Ricatti-Bessel functions and Legendre polynomials. The problem underlying
the internal algorithm has been discussed to some extent in sec. 5.2.2

5.2.2 Problems
With the given algorithm, there are constraints for Mathematica, which doesn’t allow
calculating spheres with larger size parameters. To my personal knowledge, there has been
no attempt to calculate the intensity graphs for such a large parameter for a wavelength so
small as considered here. The problems encountered in Mathematica have been shortlisted
in the following discussion.

Runtime
The runtime for calculating intensity graphs for kinematical theory was a fraction of a
second, whereas the runtime for calculating intensity graphs for Mie theory increases
with increasing size parameter and decreasing wavelength. The following table gives the
exact time required to calculate particular intensity graphs for Mie theory for particular
Q-domain. It can be determined using the command Timing[expr]

Radius [nm] Q-Range Time required in seconds


0.5 0 . . . 100 382.49
1.0 0 . . . 110 1648.88
2.5 0 . . . 100 9541.1
5.0 0 . . . 50 41102.6
10.0 0 . . . 25 206513.0

After step-by-step investigation it has been found that the calculation for the Legendre
polynomials takes enormous time. Since Mathematica selects randomly its x-value in
plotting a graph for f (x), it has not been possible to temporarily save the values of
Legendre polynomials, which are evoked as many times as the termination term, described
in sec. 4.3.2. The termination term is larger for large size parameters. Nonetheless, there
is a possibility of using an alternative algorithm, though with problems, which shall be
explained later.
5.2. Errors and problems 59
GRAPH
I_kin
1e−5 I_mie

1e−6

1e−7

Differential scattering cross section [nm*nm]


1e−8

1e−9

1e−10

1e−11

1e−12

1e−13

1e−14

1e−15

1e−16

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140


Q [nm]

Figure 5.6: Scattering cross section graph for the sphere with radius 0.5 nm

Internal algorithm

Mathematica does have internal built-in functions, which may have eased the calcula-
tion time. The one, which is widely used in the command like Plot[expr] Plot[expr], is called
Evaluate[expr]
Evaluate[expr]. Evaluate[expr] causes expr to be evaluated even if it appears as the ar-
gument of a function whose attributes specify that it should be held unevaluated. This
command is good, when the graphs cannot be calculated for particular values, which stops
the further calculation(see sec. 5.4.1 for details). It speeds up the whole process, but un-
fortunately, it hasn’t been of use in our Mathematica programme. For the size parameters
greater than 40, the initial values started showing greater discrepancies, which is not to
be seen, when using Plot[expr] without using this command. The internal algorithm is
not accessible, so no further comments can be made on its functioning.

Problems with alternative algorithm

There is a possibility of using an alternative algorithm used for IDL and described in
the last chapter in sec. 4.3.2. These algorithms are based on the recurrence relation
given in (4.34) and (4.41). The attempt was made to introduce recursive algorithms
in Mathematica. Unfortunately, it couldn’t fructify, since for the larger parameters the
recursion depth is larger too. The default Mathematica RecursionLimit is 1000, which can
be increased by setting it using the command $RecursionLimit
$RecursionLimit. But Mathematica can’t
manage larger values. The current Mathematica kernel occupies stack space proportional
to the number of recursive calls, limited by $RecursionLimit
$RecursionLimit. Moreover, every stack frame
is about 10 times the size of equivalent stack frame in a popular compiled language like C,
FORTRAN, etc. So the $RecursionLimit must be kept low to avoid kernel segmentation
fault, causing it to stop functioning when the operating system runs out of stack space.
This is not an inherent or theoretical limitation, just a practical limitation of the current
implementation that may well be fixed at a future date.∗


The theory that underpins this problem can be marginalized by understanding tail recursion in func-
tional languages and continuation passing style. It is covered in the book, OCaml for scientists, by Dr.Jon
D. Harrop
60 Chapter 5. Comparison
GRAPH
1e−3 I_kin
I_mie
1e−4

1e−5

1e−6
Differential scattering cross section [nm*nm]

1e−7

1e−8

1e−9

1e−10

1e−11

1e−12

1e−13

1e−14

1e−15

1e−16

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115


Q [nm]

Figure 5.7: Scattering cross section graph for the sphere with radius 1.0 nm

5.3 Conclusion
Apart from marginal exception of the last graph with size parameter of 100, the theory
seems to be in greater accordance with the scalar kinematical theory. Unfortunately, it has
not been possible, due to reasons mentioned above, to see the breaking point where one
can differentiate the theory better. The technical difficulties encountered for this endeavor
have been provided with technical discussion in subsection sec. 5.2.2. The possible solu-
tion would be to other programming language in entirety using the algorithm techniques
mentioned in sec. 4.3.2.
It is true that the Mie theory gives the explanation based on only two variables, the
size parameter x and the refractive index m. But it should be noticed that refractive index
depends on the material and wavelength. And the size parameter is just the ratio between
the radius and the wavelength. That means there is implicit dependency on the variables
on these variables of radius, wavelength, and the refractive index. One can write pairs
of radius and wavelength giving the same size parameter, but graphs containing these
pairs cannot show the same nature. The point to be made is that it is not possible to
make graphs of one pair from the knowledge of the other, though both have the same size
parameter. Otherwise, it would have been possible to make graphs for higher radii. This
may explain why it takes comparatively more time for shorter wavelength than the longer
ones, irrespective of the size parameter.
The same argumentation can be made to the interesting parameter QR, which appears
in the equation for the kinematical scattering, but also contains size parameter in it (as
QR = 2x sin(θ/2)). Nonetheless, investigation with this parameter may be of greater help
in one way or the other.

5.4 Appendix
5.4.1 Term Rewriting System
Mathematica is a term rewriting system. Whenever an expression is entered, it is evaluated
by term rewriting using rewrite rules. These rules consist of two parts: a pattern on the
left-hand side(lhs) and a replacement text on the right-hand side(rhs). When the lhs of
a rewrite rule is found to pattern-match part of the expression, that part is replaced by
5.4. Appendix 61
GRAPH
I_kin
1e−1
I_mie

1e−2

1e−3

1e−4

Differential scattering cross section [nm*nm]


1e−5

1e−6

1e−7

1e−8

1e−9

1e−10

1e−11

1e−12

1e−13

1e−14

1e−15
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Q [nm]

Figure 5.8: Scattering cross section graph for the sphere with radius 2.5 nm

the rhs of the rule, after substituting values in the expression which match labeled blanks
in the pattern into the rhs of the rule. Evaluation then proceeds by searching for further
matching rules until no more are found.
The evaluation process in Mathematica can be understood with a simple analogy to
daily experiences. One can recount the experience of using handbook of mathematical
formulas, such as the integral tables. In order to solve an integral, one consults the
handbook which contains formulas consisting of a lhs and rhs, separated by an ’equals’
sign. And then one looks for an integration formula in the handbook whose lhs has the
same form as the questioned integral.
While no two formulas in the handbook have the identical lhs, there may be several
whose lhs have the same form as the required integral. When such thing happens, one use
the formula whose lhs gives the closest fit to the integral. Then, one replaces the integral
with the rhs of the matching lhs and one substitutes the specific values in an integral for
the corresponding variable symbols in the rhs. Finally, one looks through the handbook
for formulas, like trigonometric identities, etc., that can be used to change the answer
further.
This depiction provides a better description of the Mathematica evaluation process,
but not entirely. It can be absolutely understood in Mathematica Documentation.
62 Chapter 5. Comparison

GRAPH
1e1
I_kin
I_mie
1e0

1e−1

1e−2
Differential scattering cross section [nm*nm]

1e−3

1e−4

1e−5

1e−6

1e−7

1e−8

1e−9

1e−10

1e−11

1e−12

1e−13
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0
Q [nm]

Figure 5.9: Scattering cross section graph for the sphere with radius 5.0 nm

GRAPH
1e3 I_kin
I_mie
1e2

1e1

1e0
Differential scattering cross section [nm*nm]

1e−1

1e−2

1e−3

1e−4

1e−5

1e−6

1e−7

1e−8

1e−9

1e−10

1e−11
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
Q [nm]

Figure 5.10: Scattering cross section graph for the sphere with radius 10.0 nm
5.4. Appendix 63

1.015

1.01

1.005

0 100 200 300 400 500


Radius [Ang]

Figure 5.11: Ratio of differential scattering cross sections of Mie theory to kinematical
theory at Q = 0 against radius
Chapter 6

Outlook

65
66 Chapter 6. Outlook

In this thesis, the whole attempt was to understand the scattering by a sphere in the
X-ray region with more insight. The two theories taken for comparison have their own
might in the respective expression. This comparison was of the scalar scattering theory
and vector scattering theory. The scalar theory was reduced to the discussion of the first
Born approximation, mostly due to its great proven applicability in the X-ray physics. On
the other hand, the vector theory, first developed by G.Mie, was tried in X-ray region. The
latter theory is the exact theory to the scattering by a sphere, valid for all wavelengths.
This theory has won the ground on applicability quite recently, mostly due to increasing
machine powers for calculations. The physics represented by this theory is very much
embedded in the mathematical jargon its solution carries. Nonetheless, in this thesis, it was
tried to see this theory approximated in X-ray region, analytically as well as numerically.
The aim was to comprehend the scattering for the particles, not so small to be in the range
of Rayleigh-Gans theory, and not so big to be in the region of geometrical optics. To the
explanation of such spheres, the aforementioned anomalous diffraction theory has been
introduced (see 4.1). But more effort was laid on the numerical calculation. Due to its
symbolic and clear representative nature, the mathematical tool Mathematica was used.
The whole straight-forward algorithm was written for the calculation. It provided very
wonderful graphs for comparison, but again the machine calculation power has become
the obstacle. Due to the problems mentioned in sec. 5.2.2, the direct calculation for the
particles larger than size parameter 100 has not been possible. Looking at the graphs
of differential scattering cross section, it can be stated that the scalar theory seemed
in accordance with the vector theory, at least, up to this size parameter. The possible
deviation from the scalar theory in the X-ray region may lie in the particles larger than
that of with size parameter 100.
There are several outlooks that can be mentioned in this subject for the future inves-
tigation. They vary from the computational level to promising analytical level. Perhaps,
the work in other fields where Mie theory has been used successfully, like Meteorology,
Astrophysics, could be of very much use.

Another programming language Using different programming languages for the de-
velopment of Mie code could be very helpful, especially those languages which disallows
the problems encountered in Mathematica. The low-level language is the primary choice
for its reusability and compatibility with different platforms. Other commercially avail-
able languages like OCaml, IDL, etc. can be used extensively, which avoids exactly the
same problem shown in Mathematica. The algorithm provided in the last chapter (see
sec. 4.3.2) is the most widely used till now.

Use of Anomalous Diffraction Theory The possible analytical approximation for


the X-rays, proposed here as anomalous diffraction theory is also the great candidate,
when it comes to understanding the physics behind this phenomenon. There is also a
computational method available for calculating anomalous diffraction theory, and inter-
estingly, for the X-ray scattering by interstellar dust. It is explained extensively in the
paper from Draine and Allaf-Akbari[1]. Although there is less literature available on this
topic, it is a virgin field in the X-ray physics, where its validity can be explored.

Patterns in Mie scattering The novel effort of evaluating Mie scattering intensity
versus the scattering wave vector Q has been due to C.M.Sorensen[2]. To my knowledge, it
was the first such effort. He has used the FORTRAN code given by Bohren and Huffman[3],
where the idea was to create graphs versus QR, where R is the radius of the sphere, rather
than rudimentary against the scattering wave vector Q. With the calculation for larger
QR for different size parameter, certain patterns have been observed. He gives certain
interpretations to the results. His analysis was restricted to a higher refractive index. It
would be interesting to analyze it for the X-ray Mie scattering.
Bibliography

[1] B.T.Draine and Khosrow Allaf-Akbari. X-ray scattering by nonspherical grains.i.


oblate spheroids. The Astrophysical Journal, (652):1318–1330, 2006.

[2] C. M. Sorensen and D. J. Fischbach. Patterns in Mie scattering. Optics Communica-


tions, 173:145–153, 2000.

[3] Craig F. Bohren and Donald R. Huffman. Absorption and Scattering of Light by Small
Particles. Wiley-VCH, 2004.

67
Chapter 7

Numerical calculation

69
70 Chapter 7. Numerical calculation

7.1 Mathematica code

Mie Scattering - X-Ray - Gold

Packages required.
Some of the plot routines require certain graphics packages be loaded. A working directory may be
set for subsequent input/output operations. For example, this directory could contain the optical
dispersion data, n` (l) = n(l)+ik(l), required for any spectral analysis.

H* Needed for the polar plotting routine. *L


H* Needed for the spherical plotting routine. *L
Needs@"Graphics`Graphics`"D

H* Needed for plot legends *L


Needs@"Graphics`ParametricPlot3D`"D
Needs@"Graphics`Legend`"D

Off@General::spell1D;

SetDirectory@"êUsersênandan"D;

Parameters
This section is meant for the known parameters for the problem (i.e., complex indices of refraction, wavelength of the
illuminating radiation, sphere radius, etc.). Subscripts of 1 correspond to the embedding medium and subscripts of 2
represent the sphere. The permeability is put to one, as described in the theory of the thesis. It is important to notice
that the imaginary part of the refractive index is represented by a lower case k, and that capital K represents the wave
number. Input parameters are integers (which have infinite precision in Mathematica) whenever possible. If real num-
bers are required, I would be better to use increased precision (via the backtick mark, `) if high precision output is
required. The increased precision will slow down the code execution. Finally, it would be more efficient to assign the
parameter values using rewrite rules in the actual function calls but I've presented them as constants because this format
saves valuable runtime.

The calculations are extensible to the general case of a sphere of arbitrary radius and known optical
properties, within a medium of known optical properties. In order to end up with a prefered size parameter
as in the reference example in Bohren & Huffman, the chosen wavelength is entered.

H* Real part of refractive index of medium *L


H* Delta part of refractive index of scatterer *L
n1 = 1;

H* Real part of refractive index of scatterer *L


d2 = 1.9 * 10-5 ;
n2 = 1 - d2 ;

H* Imaginary part of refractive index of medium *L


H* Imaginary part of refractive index of scatterer *L
k1 = 0 * 10-2 ;
k2 = 2.5 * 10-6 ;

H* V acuum wavelength in nm *L
r0 := 2.82 * 10^ H-6L; H* The classic electron radius in nm *L
l0 = 10 * 10-2 ;

r = 100 * 10-2 ; H* Sphere radius in nm *L

Required derived quantities for given parameter set.


7.1. Mathematica code 71

H* Complex refractive index of medium *L


H* Complex refractive index of scatterer *L
m1 = n1 + I k1 ;
m2 = n2 + I k2 ;

H* Relative refractive index between scatterer and medium *L


n2 + I k2
mRel = ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ;
n1 + I k1

H* W avelength in external medium in nm *L


l0
l1 = ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ;
n1 + I k1

H* W avelength inside sphere in nm *L


l0
l2 = ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ;
n2 + I k2

H* Wave number in vacuum in nm-1 *L


2p
K0 = ÅÅÅÅÅÅÅÅ ;
l0

H* Wave number in external medium in nm-1 *L


2p
K1 = ÅÅÅÅÅÅÅÅ ;
l1

H* Wave number inside the sphere in nm-1 *L


2p
K2 = ÅÅÅÅÅÅÅÅ ;
l2
rho = Hd2 * K0 2 L ê H2 * p * r0L;H* Electron density in 1êcubic nm *L;

Now create the size parameter. Note there are actually three size parameters defined (as opposed to Bohren and Huff-
man's single size parameter), two of which are now complex and due to the fact that the external medium may be absorp-
tive. They come from the paper by Fu and Sun (Applied Optics, Vol. 40, No. 9, 2001,pp. 1354-1361) which explains
how to treat a particle in an absorbing medium. They serve as arguments to the scattering coefficient equations. They
are set up as functions of particle radius so that this may be varied.

s1 @rD := K1 r
s@rD := K0 r

s2 @rD := K2 r

I use the equation suggested by Wiscombe (Applied Optics, Vol. 19, 1980, pp. 1505-1506) for obtaining the series
solution termination point for a given size parameter. Since there are now three (two possibly complex) size parameters,
I choose the largest of the absolute values of the three as the argument to the equation. The equation was found to be
1
valid for the range 8 < s < 4200. For smaller s values, the suggestion is LastTerm = s + 4 s ÅÅÅÅ3Å + 1. It is not suggested to
try to run the program for very large size parameters due to the resulting long computation times (in fact, anything over
s=100 becomes tediously long).

Largest@rD := Max@s@rD, Abs@s1 @rDD, Abs@s2 @rDDD


LastTerm@rD := CeilingAAbsALargest@rD + 4.05 HLargest@rDL ÅÅÅÅ3 + 2EE ;
1

Required function definitions


We are going to need spherical Bessel, Neumann and Hankel functions to represent the radial field dependence. First

y1 = "####### HrL,
introduced by Debye, the mean was to simply the notations. These can be given as follow:

cn1 = - "######pr #
pr
ÅÅÅÅ
ÅÅ J with Bessel function of 1st kind
ÅÅ Nn+1ê2 HrL, with Bessel function of 2nd kind
r 2 n+1ê2

h1n = "#######
ÅÅÅÅ
ÅÅÅÅ2ÅÅ H H2L n+1ê2 HrL = y1r + i c1n
2
pr
with Hankel function
Boundary conditions at r = 0 and r = ¶ require finite fields and hence the need for all three types of Bessel functions.
We go straight to the Ricatti-Bessel functions here by simply multiplying through by r.

Printed by Mathematica for Students


72 Chapter 7. Numerical calculation

i 1y
y@l_, r_D := r $%%%%%%%%%% jl + ÅÅÅÅ z
ÅÅÅÅÅÅÅÅ BesselJAj z, r E
p
2r k 2{
H* Ricatti-Bessel function of 1 st kind. *L

H* Partial derivative of Ricatti-Bessel function w.r.t. r. *L


yPrime@l_, r_D := Evaluate@D@ y@l, rD, rDD ;

i 1y
ÅÅÅÅÅÅÅÅ BesselYAj
z@l_, r_D := y@l, rD + I r $%%%%%%%%%% jl + ÅÅÅÅ z
z, r E ; H* Ricatti-Hankel function *L
p
2r k 2{

H* Partial derivative of Ricatti-Hankel function w.r.t. r *L


zPrime@l_, r_D := Evaluate@D@ z@l, rD, rDD;

matica. Note that the Mathematica version has a H-1Lm in front of the function definition, which Born and Wolf do not
We also need the Associated Legendre functions for the azimuthal components of the field. These are built into Mathe-

have (but make mention of). Here, m is the separation constant that arises from the separation of variables technique
used to solve the wave equation and must equal one for this problem. Since m = 1, we get a negative sign out front so I
place another one in my definition of the function to conform with what we actually want. Once that's done, we can
construct the "angle function" coefficients pn and tn . They derive from the Legendre functions

P1 HcosqL d P1 HcosqL
as,

pn := ÅÅÅÅÅÅÅÅ
l
ÅÅÅÅÅÅÅÅÅ , and tn := ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
l
ÅÅÅÅÅ
Since, we are interested in writing the Phase function in the dependency of scattering transfer vector ÷q” these "angle
sinq dHcosqL

P1 @cosH2 sin-1 gLD


function" are represented as follows,

è!!!!!!!!!!
pn := ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ! ÅÅÅ and
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
l
2g
è!!!!!!!!!!2!
2 1-g

K1 1-g d P1l @cosH2 sin-1 gLD q


tn := ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
dq
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ , where g := ÅÅÅÅÅÅÅÅ
2 K1
ÅÅ
LegendrePAi, 1, CosA2 ArcSinA ÅÅÅÅÅÅÅÅ 2 K1
q
EEE
pi@i_, q_D := ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ;
2 q $%%%%%%%%%%%%%%%%
q %%%%%%%%
2%%
1-I ÅÅÅÅ 2 ÅKÅÅÅÅÅÅ M
1
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
2K
ÅÅÅÅÅÅÅÅÅÅÅÅÅ
1

i
j
j 1-i
K $%%%%%%%%%%%%%%%% ÅÅ y
q%%%%%%%%
j
j
j j
j ÅÅÅÅÅÅÅÅ z %%
z
j 1
1 2

"################ ########## k 2 K1 {
t@i_, q_D := ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
H2 K1 L2 - q2 k

i
j i + iL LegendrePA-1 + i, 1, CosA2 ArcSinA ÅÅÅÅÅÅÅÅ
j-2 CscA2 ArcSinA ÅÅÅÅÅÅÅÅÅÅ EE j
jH1 ÅÅ EEE -
q q
k 2 K1 k 2 K1
y
i CosA2 ArcSinA ÅÅÅÅÅÅÅÅÅÅ EE LegendrePAi, 1, CosA2 ArcSinA ÅÅÅÅÅÅÅÅÅÅ EEEyyz
z;
zzz
zzz
z
{{z
q q

{
2 K1 2 K1

Printed by Mathematica for Students


7.1. Mathematica code 73

ü We now have enough information to construct the Mie coefficients.

an@l_D := Evaluate@Hm2 yPrime @ l, s1 @rDD y@l, s2 @rDD - m1 y@l, s1 @rDD yPrime @ l, s2 @rDDL ê
Hm2 zPrime@l, s1 @rDD y@l, s2 @rDD - m1 z@l, s1 @rDD yPrime@l, s2 @rDD LD

bn@l_D := Evaluate@Hm2 y@l, s1 @rDD yPrime @ l, s2 @rDD - m1 yPrime @ l, s1 @rDD y@l, s2 @rDD L ê
Hm2 z@l, s1 @rDD yPrime@l, s2 @rDD - m1 zPrime@l, s1 @rDD y@l, s2 @rDDLD

cn@l_D := Evaluate@ Hm2 z@l, s1 @rDD yPrime @ l, s1 @rDD - m2 zPrime@l, s1 @rDD y@l, s1 @rDDL ê
Hm2 z@l, s1 @rDD yPrime@l, s2 @rDD - m1 zPrime@l, s1 @rDD y@l, s2 @rDDLD

dn@l_D := Evaluate@Hm2 zPrime@l, s1 @rDD y@l, s1 @rDD - m2 z @ l, s1 @rDD yPrime@l, s1 @rDD L ê


Hm2 zPrime@l, s1 @rDD y@l, s2 @rDD - m1 z@l, s1 @rDD yPrime@l, s2 @rDD LD

Calculate and plot the relative intensities


Next, we construct the formulae for intensity as a function of scattering transfer vector ÷q” polarized in either of the two
linear orthogonal polarization states. Begin by defining the amplitude matrix elements, S1(r, q) and S2(r, q) which
simplify the form of the resulting phase function equations a little

LastTerm @rD
i H2 l + 1L y
‚ j ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ z
j z Han@lD pi@l, qD + bn@lD t@l, qDL;
k l Hl + 1L {
S1@q_D :=
l=1

i H2 l + 1L z y
LastTerm @rD
‚ j
j ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ z H an@lD t@l, qD + bn@lD pi@l, qD L;
k l Hl + 1L {
S2@q_D :=
l=1

Ÿ PHcosHqLL „ W = 1) is sometime used in the literature (e.g., Fu and Sun) and sometimes neglected (e.g Bohren and
Here are the relative intensities and differential scattering cross sections. The normalization factor (to insure

Huffman). We shall consider the latter. Although the formula for both polarization has been given, hereafter, we shall be

ikin HqL := r2 I ÅÅÅÅÅÅÅÅ 3 ÅÅ M 9 ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ


considering only the perpendicular polarization. The kinematical theory can be given as a straightforward formula
HSin HqrL-qrCos Hq*rLL
HqrL 6
3 2 2
4 pr
ÅÅÅÅÅÅÅÅÅÅÅÅÅ

dWÅ = r0 ikin HqL


And the differential scattering cross section for the kinematical theory is
ds 2
ÅÅÅÅÅ

imie HqL
whereas for the Mie theory is
ds
ÅÅÅÅÅ
dWÅ = ÅÅÅÅÅÅÅÅ
k 2
ÅÅÅÅÅÅÅ for the perpendicular polarization.
0

Iperp@q_D := HAbs@S1@qDDL2 ; H* Relative intensity for perpendicular polarization *L


Ipar@q_D := HAbs@S2@qDDL2 ; H* Relative intensity for parallel polarization *L
i 4 pr3 y HSin@qrD - qrCos@qrDL2
Ikin@q_D := HrhoL2 j j z 9 ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ
j ÅÅÅÅÅÅÅÅÅÅÅÅÅ z
z
2

k 3 { HqrL6
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ;

H* Relative intensity for kinematical theory *L

SQperp@q_D := ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ; H* Differential scattering cross section for Mie theory in perpendicular polarization *L
Iperp@qD

SQkin@q_D := r0 Ikin@qD; H* Differential scattering cross section for kinematical theory *L


K0 2
2

If the calculation of whole graphs becomes time-consuming, only the values for q=0 can be considered. The relative
intensity and differential scattering cross section can be reduced to the following expression is calculated at q=0,

Printed by Mathematica for Students


74 Chapter 7. Numerical calculation

LastTerm @rD
‚ H2 l + 1L H an@lD + bn@lD L; H* The amplitude function for q=0 *L
1
S10 := ÅÅÅÅ
2 l=1

i 4 pr3 y
Ikin0 := HrhoL2 j j
j ÅÅÅÅÅÅÅÅÅÅÅÅÅ z
z ; H* Relative intensity for kinematical theory at q=0 *L
z
2

k 3 {
Iperp0 := HAbs@S0 DL2 ; H* Relative intensity for perpendicular polarization at q=0 *L

SQperp0 := ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ; H* Differential scattering cross section for Mie theory at q=0 *L
Iperp0

SQkin0 := r0 Ikin0 ; H* Differential scattering cross section for kinematical theory at q=0*L
K0 2
2

ü Create plots of the relative intensities.

ü Log-linear plots of scattered relative intensity vs. q.

This first cell generates some graphics that make the plots look better.

MyTickLabels = Table@8i, ToString@iD<, 8i, 0.0, 100.0, 10.0<D;


XGridLines = Transpose@MyTickLabelsDP1T;

ImageSize Ø 600, GridLines Ø 8XGridLines, Automatic<,


SetOptions@LogPlot, Frame Ø True, PlotRange Ø All,

FrameTicks -> 8MyTickLabels, Automatic, None, None<, PlotRange Ø All D;

Here are the plots.

Ikin vs.q

Timing@IKinLogPlot = LogPlot@Ikin@qD, 8q, 0, 100<, PlotStyle Ø RGBColor@0, 0, 1D,


FrameLabel Ø 8"q H1ênmL", "Ikin Relative Intensity"<D;D

ListIKinLog = Flatten@Cases@IKinLogPlot, Line@x__D Ø x, InfinityD, 1D;


Export@"dataêMie_Roentgen_Ikin.dat", ListIKinLog, "Table"D;
Export@"dataêMie_Roentgen_Ikin.eps", IKinLogPlot, "EPS"D

Imie vs. q

LogPlot@Evaluate@Iperp@qDD, 8q, 0, 100<, PlotStyle Ø RGBColor@0, 0, 1D,


Timing@IMieLogPlot =

FrameLabel Ø 8"q H1ênmL", "Imie Relative Intensity"<D;D

ListIMieLog = Flatten@Cases@IMieLogPlot, Line@x__D Ø x, InfinityD, 1D;


Export@"dataêMie_Roentgen_Imie.dat", ListIMieLog, "Table"D;
Export@"dataêMie_Roentgen_Imie.eps", IMieLogPlot, "EPS"D

All graphs on the same plot. The first expression creates a legend for the plot.

ILegend = 888RGBColor@1, 0, 0D, "Imie "<, 8RGBColor@1, 0, 1D, "Ikin "<<,


LegendPosition Ø 8.5, 0.3<, LegendSize Ø 80.25, 0.25<, LegendShadow Ø 80, 0<<;

DisplayFunction Ø Identity, FrameLabel Ø 8"q H1ênmL", "Relative intensities"< D,


ICompPlot = ShowLegend@Show@IMieLogPlot, IKinLogPlot, PlotRange Ø All,

ILegend, DisplayFunction Ø $DisplayFunction D;


Export@"dataêMie_Roentgen_IComp.eps", ICompPlot, "EPS"D

Printed by Mathematica for Students


7.1. Mathematica code 75

ü Create plots of the differential scattering cross sections.

ü Log-linear plots of differential scattering cross sections vs. q.

This first cell generates some graphics that make the plots look better.

MyTickLabels = Table@8i, ToString@iD<, 8i, 0.0, 100.0, 10.0<D;


XGridLines = Transpose@MyTickLabelsDP1T;

ImageSize Ø 600, GridLines Ø 8XGridLines, Automatic<,


SetOptions@LogPlot, Frame Ø True, PlotRange Ø All,

FrameTicks -> 8MyTickLabels, Automatic, None, None<, PlotRange Ø All D;

Here are the plots.

J ÅÅÅÅÅÅÅÅÅÅ N vs.q
ds
dW kin

TimingASQKinLogPlot = LogPlotASQkin@qD, 8q, 0, 100<, PlotStyle Ø RGBColor@0, 0, 1D,


i ds z y
FrameLabel Ø 9"q H1ênmL", "j
j ÅÅÅÅÅÅÅ z
k dW {kin
Differential scattering cross section"=E;E

ListSQKinLog = Flatten@Cases@SQKinLogPlot, Line@x__D Ø x, InfinityD, 1D;


Export@"dataêMie_Roentgen_SQkin.dat", ListSQKinLog, "Table"D;
Export@"dataêMie_Roentgen_SQkin.eps", SQKinLogPlot, "EPS"D

J ÅÅÅÅÅÅÅÅÅÅ N vs.q
ds
dW mie

TimingASQMieLogPlot =
LogPlotAEvaluate@SQperp@qDD, 8q, 0, 100<, PlotStyle Ø RGBColor@0, 0, 1D,

j ÅÅÅÅÅÅÅ y
i z
FrameLabel Ø 9"q H1ênmL", "j z
ds
k dW {mie
Differential scattering cross section"=E;E

ListSQMieLog = Flatten@Cases@SQMieLogPlot, Line@x__D Ø x, InfinityD, 1D;


Export@"dataêMie_Roentgen_SQmie.dat", ListSQMieLog, "Table"D;
Export@"dataêMie_Roentgen_SQmie.eps", SQMieLogPlot, "EPS"D

All graphs on the same plot. The first expression creates a legend for the plot.

i ds z y i ds z y
SQLegend = 999RGBColor@1, 0, 0D, "j
j ÅÅÅÅÅÅÅ z "=, 9RGBColor@1, 0, 1D, "j
j ÅÅÅÅÅÅÅ z "==,
k dW {mie k dW {kin
LegendPosition Ø 8.5, 0.3<, LegendSize Ø 80.25, 0.25<, LegendShadow Ø 80, 0<=;
SQCompPlot = ShowLegend@Show@IMieLogPlot, IKinLogPlot,

FrameLabel Ø 8"q H1ênmL", "Differential scattering cross section"< D,


PlotRange Ø All, DisplayFunction Ø Identity,

SQLegend, DisplayFunction Ø $DisplayFunction D;


Export@"dataêMie_Roentgen_SQComp.eps", SQCompPlot, "EPS"D

Printed by Mathematica for Students


76 Chapter 7. Numerical calculation

7.2 IDL code


pro MyMie

; Obtain pi:
pii = 4.D0*atan(1.D0)

; nang = number of angles between 0 and 90 degrees


nang = 1000

n1 = 1 ; Real part of refractive index of medium


delta = 0.000019 ; Delta part of the refractive index of scatterer
n2 = 1 - delta ; Real part of refractive index of scatterer

k1 = 0 ; Imaginary part of refractive index of medium


k2 = 0.00001 ; Imaginary part of refractive index of scatterer

lambda0 = 0.05 ; Vacuum wavelength in "nm"

r = 04.0 ; Sphere radius in "nm"


r0 = 0.00000282 ; The classic electron radius

; Derived quantities for given parameter

m1 = n1 ; Complex refractive index of medium


m2 = n2 ; Complex refractive index of scatterer

mRel = m2/m1 ; Relative refractive index between scatterer and medium

lambda1 = lambda0/m1 ; Wavelength in external medium in nm


lambda2 = lambda0/m2 ; Wavelength inside sphere in nm

K0 = 2*pii/lambda0 ; Wave number in vacuum in nm^-1


K1 = 2*pii/lambda1 ; Wave number in the medium in nm^-1
K2 = 2*pii/lambda2 ; Wave number inside the sphere in nm^-1
rho = (delta*K0^2)/(2*pii*r0) ; The electron density of the scatterer

; According to the paper of Fu and Sun, the size parameters can be given as follows

sz = K0*r ; size parameter


sz1 = K1*r ; external size parameter
sz2 = K2*r ; internal size parameter

LargestTerm = MAX([sz, sz1, sz2]) ;


LastTerm = FLOOR(ABS(LargestTerm + 4.05 * (LargestTerm)^(1/3) + 2))

; ..Parameters..
mxnang = 2100
nmxx= 150000

; Scalar Arguments
s1 = dcomplexarr(2*mxnang-1)
s2 = dcomplexarr(2*mxnang-1)
7.2. IDL code 77

; Local Arrays
mie = fltarr((nang/4)+1)
d = dcomplexarr(nmxx)
amu = dblarr(mxnang)
kin = fltarr((nang/4)+1)
pi=dblarr(mxnang)
pi0=dblarr(mxnang)
pi1=dblarr(mxnang)
tau=dblarr(mxnang)

IF (nang GT mxnang) THEN BEGIN


print, ’ error: nang > mxnang in bhmie’
STOP
ENDIF

IF (nang LT 2) THEN nang = 2

dx = sz1 ; Size parameter as x

y = LastTerm*mRel
ymod = abs(y)

;
; *** Series expansion terminated after NSTOP terms
; Logarithmic derivatives calculated from NMX on down

nmx = max(LastTerm,ymod) + 15
nmx = fix(nmx)
nstop = LastTerm

IF (nmx GT nmxx) THEN BEGIN


print, ’error: nmx > nmxx=’, nmxx, ’ for |m|x=’, ymod
STOP
ENDIF

; *** Require NANG.GE.1 in order to calculate scattering intensities


dang = 0.D0
q = 0.D0
IF (nang GT 1) then dang = .5D0*pii/double(nang-1)

FOR j=1, nang DO BEGIN ; DO I j = 1, nang


theta = double(j-1)*dang
amu(j) = cos(theta)
ENDFOR ; I CONTINUE

FOR j=1, nang DO BEGIN ;DO II j = 1, nang


pi0(j) = 0.D0
pi1(j) = 1.D0
ENDFOR ; II CONTINUE

nn = 2*nang - 1
FOR j=1, nn DO BEGIN ;DO III j = 1, nn
78 Chapter 7. Numerical calculation

s1(j) = dcomplex(0.D0,0.D0)
s2(j) = dcomplex(0.D0,0.D0)
ENDFOR ; III CONTINUE

; Calculate relative intensity for the kinematical theory


FOR j=1, 2*nang-1 DO BEGIN ; DO IV n = 1, 180
theta = double(j-1)*dang
q = 2.D0*K0*sin((theta)/2)
kin(j) = rho^2*(4.D0*pii*r^3)*(((sin(q*r)-q*r*cos(q*r))^2)/((q*r)^6))
ENDFOR ; IV CONTINUE

; Calculating Mie coefficients using methods of logarithmic derivation of


; Bessel Function as explained in Bohren & Hoffman, p.127
; *** Logarithmic derivative D(J) calculated by downward recurrence
; beginning with initial value (0.,0.) at J=NMX

d(nmx) = DCOMPLEX(0.D0,0.D0)
nn = nmx - 1

FOR n=1, nn DO BEGIN ; DO V n = 1, nn


en = nmx - n + 1
d(nmx-n) = (en/y) - (1.D0/ (d(nmx-n+1)+en/y))
ENDFOR ; V CONTINUE

;
; *** Riccati-Bessel functions with real argument X
; calculated by upward recurrence
;
psi0 = COS(dx)
psi1 = SIN(dx)
chi0 = -SIN(dx)
chi1 = COS(dx)
xi1 = DCOMPLEX(psi1,-chi1)
p = -1.D0
FOR n=1, nstop DO BEGIN ; DO VIII n = 1, nstop
en = n
fn = (2.D0*en+1.D0)/ (en* (en+1.D0))

; FOR given N, PSI = psi_n CHI = chi_n


; PSI1 = psi_{n-1} CHI1 = chi_{n-1}
; PSI0 = psi_{n-2} CHI0 = chi_{n-2}
; Calculate psi_n and chi_n
;

psi = (2.D0*en-1.D0)*psi1/dx - psi0


chi = (2.D0*en-1.D0)*chi1/dx - chi0
xi = DCOMPLEX(psi,-chi)

;
; *** Store previous values of AN and BN for use
; in computation of g=<cos(theta)>
IF (n GT 1) THEN BEGIN
an1 = an
7.2. IDL code 79

bn1 = bn
ENDIF
;
; *** Compute AN and BN:
an = (d(n)/mRel+en/dx)*psi - psi1
an = an/ ((d(n)/mRel+en/dx)*xi-xi1)
bn = (mRel*d(n)+en/dx)*psi - psi1
bn = bn/ ((mRel*d(n)+en/dx)*xi-xi1)

; *** Now calculate scattering intensity pattern


; First do angles from 0 to 90
FOR j=1, nang DO BEGIN ; DO VI j = 1, nang
jj = 2*nang - j
pi(j) = pi1(j)
tau(j) = en*amu(j)*pi(j) - (en+1.D0)*pi0(j)
p = (-1.D0)^(n-1)
s1(j) = s1(j) + fn* (an*pi(j)+bn*tau(j))
t = (-1.D0)^n
2(j) = s2(j) + fn* (an*tau(j)+bn*pi(j))

;
; *** Now do angles greater than 90 using PI and TAU from
; angles less than 90.
IF (j NE jj) THEN BEGIN
s1(jj) = s1(jj) + fn* (an*pi(j)*p+bn*tau(j)*t)
s2(jj) = s2(jj) + fn* (bn*pi(j)*t+an*tau(j)*p)
ENDIF
ENDFOR ; VI CONTINUE

psi0 = psi1
psi1 = psi
chi0 = chi1
chi1 = chi
xi1 = dcomplex(psi1,-chi1)

;
; *** Compute pi_n for next value of n
; For each angle J, compute pi_n+1
; from PI = pi_n , PI0 = pi_n-1
;
FOR j=1, nang DO BEGIN ; DO VII j = 1, nang
pi1(j) = ((2.D0*en+1.D0)*amu(j)*pi(j)- (en+1.D0)*pi0(j))/$
en
pi0(j) = pi(j)
ENDFOR ; VII CONTINUE
ENDFOR ; VIII CONTINUE

; Relative intensity for the Mie theory


mie = ABS(s1)^2

; Plot the relative intensities given by both theories


SET_PLOT, ’X’
80 Chapter 7. Numerical calculation

DEVICE, DECOMPOSED=0
DEVICE, RETAIN=2
LOADCT, 39
PLOT, mie, /YLOG, COLOR=255, /NOERASE
OPLOT, kin, COLOR=240

END

You might also like