FOURIER TRANSFORM INFRARED SPECTROMETRY

Last Change: April 7, 2008

Tutors: Ronald Macatangay (macatang@iup.physik.uni-bremen.de) and Janina
Messerschmidt (messerschmidt@iup.physik.uni-bremen.de)
Room No.: U3230 Phone: (0) 421 218 – 89 89
Lab: U5080 (Restricted – Go to my office first)

Objectives
1. To learn how a Fourier transform spectrometer (FTS) works.
2. To determine the relation between the absorbance and concentration of water vapor.
3. To determine atmospheric emission using the total power method.

Molecular Spectroscopy and Fourier Transform Infrared Spectrometry

Atoms and molecules cannot exist in all energy states, rather they can only have a series
of discrete values. Energy levels (E) are divided into electronic (Eelec), rotational (Erot),
vibrational (Evib) and translational (Etrans) energy states.

E= Eelec+ Erot + Evib+ Etrans

The electronic energy is due to the electron movement in the potential due to the
arrangement of protons and neutrons in the nucleus and it is dependent on the distance
between the atomic nuclei. This is depicted in Figure 1. It is in the order of 1-10 eV.

Even though the atoms are bound together at the distance R0, where the potential energy is
minimized, the nuclei will vibrate with the center of vibration at this equilibrium state.
The vibrational energy (Evib) results in fine structure in the order of 0.1 eV. The simplest
model to describe the vibrational motion is a harmonic oscillator, i.e. the nuclei are
connected by a spring. The model implies that the potential relation is approximated by a
parabola. As seen in Figure 2 this approximation holds when vibration is close to R0.
 Atomic and Molecular Spectroscopy 2006

Figure 2 Quantized energy levels due to the vibrating atoms. The dotted line
shows how the potential can be approximated with a second order
relation (harmonic oscillator) close to the equilibrium distance.

For a given electronic configuration and a given vibrational level (n value) the
nuclei can finally also rotate around the common center of gravity. This yields a
further energy contribution (Erot) in the range of magnitude 0,001 eV. Figure 3
summarizes the energy structure in a diatomic moelcule.

Figure 3 Energy structure in a diatomic molecule

A quantum mechanical treatment where the molecular vibration is approximated

5
by a harmonic oscillator and where the rotation is described by a rigid rotator
results in following expression for the total energy:
1 !2
E = E elec + E vib + E rot = E elec + !ω 0 (n + ) + ["(" + 1)] .
2 2 µR02

The vibration is described by the quantum number n = 0, 1, 2, 3... and the

resonance frequency ω0. The rotational motion is described by the quantum
number " = 0, 1, 2, 3..., µ = m1⋅ m2/(m1 + m2) is the reduced mass. Table 1 states
the magnitude of the different contributions in energy, wavelength and
temperature.
In a gas in thermal equilibrium the average kinetic energy of a molecule is:

Ek = 3 kT .
2
The last row in Table 1 thus states at what temperature a collision between two
molecules might excite the different energy levels. This, in turn, explains why
the heat capacity (Cv) for gases is different within different temperature ranges.
Table 1 Energy contributions in a diatomic molecule.

Electron Vibration Rotation

-1
E / cm 40000 1500 10
E / eV 5 0,2 0,001
λ / µm 0,3 6 1000
T/K 40000 1500 10

(Note that it happens to be that the energies stated in cm-1 and as temperature
approximately have the same numerical value, i.e. 1/(100⋅h⋅c) = 5.0⋅1022 ≈
2/(3⋅k) = 4.8⋅1022 !)

Elastic rotator

To describe the rotational energy in a molecule it is convenient to introduce the

rotational constant B:

!2 !2
B≡ ≡ where I = µR02 is the moment of inertia of the molecule.
2 µR02 2I

In Figure 2 we see that in the harmonic-oscillator approximation B will depend

on n since R0 is constant in the symmetric parabolic potential. In the real,
asymmetric potential on the other hand B should decrease for increasing n since
R0 increases. Figure 3 show that B decreases in excited states due to the same
reason.

6
 Atomic and Molecular Spectroscopy 2006

Even within a given vibrational state the distance of equilibrium, R0, between the
nuclei should increase due to the centrifugal force when the molecule rotates
faster (higher ") so that Erot decreases. Due to this the upper rotational levels are
successively shifted downwards. This improvement, of the model “rigid rotator”,
is called “elastic rotator”. One can show that the rotational energy now instead
can be described by
Erot = B⋅"("+1) - D⋅"2("+1)2
where D is the so called centrifugal distortion constant.

4B3
D≡ where ω0 is the vibration frequency of the molecule.
ω 02
Since D << B the elastic rotator will differ noticeable from the rigid rotator not
until the "-values get reasonably high. This can also be expressed as it is required
very high experimental precision to see the effect.

Figur 4 Rotational levels for rigid and elastic rotator.

7
Molecular spectra

For light to be emitted or absorbed the total angular momentum, of the molecule,
has to change since the photon that is emitted/absorbed at a transition carries one
momentum quanta. Since the vibrational states are not associated with any
angular momentum this implies that the rotational state and/or the electronic
state must change. To be able to state selection rules we therefore have to
separate certain cases.

Pure rotational transitions, i.e. within a vibrational level in a given electronic

state. Wavelengths in the microwave range.
a. Homonuclear molecules (e.g. O2, N2), that lack permanent dipole moment do
not have rotational spectra, due to symmetry reasons.
b. ∆" = ±1

Rotational-vibrational transitions, i.e. within a given electronic state.

Wavelengths in IR.
a. Homonuclear molecules, that lack a permanent dipole moment do not have
rotational-vibrational spectra, due to symmetry reasons.
b. ∆n = ±1 (harmonic-oscillator approximation), ±2, ±3,...
c. ∆" = ±1 (+1 yield R-branch, -1 yield P-branch)

Transitions between electronic states. Wavelengths in visible or UV.

a. Analogous with the situation in atoms selection rules can be stated for
electronic transitions in molecules utilizing the total eletron angular
momentum and spin.
b. ∆n = 0, ±1, ±2, ±3,...
c. ∆" = 0, ±1
∆" = 0 is forbidden in a pure rotational-vibrational transition, i.e. within an
electronic state but permitted in conjunction with electronic transitions since
the total angular momentum of the molecule might be different in the lower
and upper electronic state respectively. ∆" = 0 trasnitions give rise to another
branch, the Q-branch. If B is equal in the two electronic states the Q-branch is
reduced to one spectral line.

Electronic spectra
Transitions where the electronic state in the molecule is changed appear in the
visible or the ultraviolet wavelength interval, as is the case for neutral atoms. If
the absorption or emission spectra is studied with moderate resolution the fine
structure can not be seen due to the rotational levels but instead a number of so
called bands. These bands represent different combinations of the vibration

8
 Atomic and Molecular Spectroscopy 2006

quantum numbers (n1, n2) in the electronic states (1, 2); see Figure 3. If we
neglect the rotational energy contribution to the total energy we can state:
E i = E eleci + !ω 0i (ni + 1 / 2)

where ω0i is the resonance frequency in state i. The frequency f for a transition is
then:
f = (E elec1 + !ω 01 (n1 + 1 / 2) − E elec2 − !ω 02 (n 2 + 1 / 2)) / h =
1
f0 + (n1ω 01 − n 2 ω 02 )
2π
In the harmonic-oscillator approximation ∆n = 0 or ±1. We now can separate the
cases where the resonance frequency is equal or different in the two levels.
ω0
ω 01 = ω 02 = ω 0 Ÿ f = f 0 + ∆n Ÿ f = f 0 − ∆f , f 0 , f 0 + ∆f
2π
In this (unrealistic) case we only get three spectral lines in the band structure.
Now instead let the resonance frequencies be different and write n2 = n1+∆n
which will result in the characteristic band structure:
1
f = f0 + ( n1 ⋅ ∆ω 0 − ∆n ⋅ ω 02 )
2π
that is instead of the three lines ∆n = -1, 0 or 1 we now get several sets of
equidistant transitions 1-0, 2-1, 3-2, 4-3..., 0-0, 1-1, 2-2, 3-3...., and 0-1, 1-2, 2-3,
3-4....

Rotational-vibrational spectrum

In the infrared wavelength region rotational-vibrational transitions occur, that is

transitions between different levels within a given electronic state. During the
laboration you will study the transitions 0 - 1 and 0 - 2 in the ground state of CO.
With the Fourier transform spectrometer you will be able to measure the wave
numbers with high precision giving the possibility to analyze the energy
structure with different approximations.

9


Figure 5 Energy level diagram of the lowest rotational-vibrational levels in a
diatomic molecule with absorption transitions from n = 0 to n = 1.
The spectrum has two branches. The R-branch with ∆" = +1 and the
P-branch with ∆" = -1. In the lower part of the figure a spectrum is
shown with the corresponding spectral lines. The rotational constants
are set to equal (B"=B'=B) which leads to equidistant spacing
between the lines, except the void where σ0 should have occured.

Approximation 1
If the distortion term D is neglected and if the B-constants are assumed to be
identical in the two vibrational states the wave number (σ) for a transition
between two vibrational states n’ och n’’ (n’ > n’’) is given by:
σ = σ 0 + B"' ("'+1) − B"' ' ("' '+1) (1)
where
σ 0 = !ω 0 ( n '−n" )
Utilizing the selection rule ∆" = ±1 results in
∆" = +1, σ = σ0 + 2B⋅("’’+1) "’’ = 0,1,2… (2a)
∆" = -1, σ= σ0 - 2B⋅"’’ "’’ = 1,2… (2b)

10
 Atomic and Molecular Spectroscopy 2006

If ("’’+1) and -"’’ is replaced by m the equations (2a) and (2b) can be combined
to:
σ = σ0 + 2Bm, m = ±1, ±2… (3)
Lines with positive m-values (∆" = +1) are called the R-branch, while those with
negative m- values (∆" = -1) are called the P-branch, see Figure 5.

Approximation 2
According to previous discussions we know that the distance of equilibrium (R0)
should increase slightly with increasing vibration energy which leads to that B
decreases for increasing n. If the rotational constants are denoted B’ and B” in
the upper and lower state respectively, instead of equation 3 we get:

σ = σ 0 + ( B' + B" ) m + ( B' − B" ) m 2 (4)

m = "’’+1 =1, 2, 3, ... R-branch

m = -"’’ = -1, -2, -3, ... P-branch
Note that if B’ and B” are equal the quadratic term vanish and equation 3 still
holds. Since B’ < B” the distance between the spectral lines will decrease
successively in the R-branch while the distances will increase in the P-branch.
For sufficiently high " a conversion occur of the R-branch which results in the
formation of a so called band head. Equation 4 has a parabolic shape and is
called a Fortrat parabola, see Figure 6.

Figure 6 Fortrat parabola.

Approximation 3
If the distortion term cannot be neglected, instead of equation 4 we get:

σ = σ 0 + ( B'+ B" ) m + ( B'− B"− D '+ D" ) m 2 − 2( D'+ D" )m 3 − ( D '− D" )m 4 (5)
Equation 5 show that σ is given by a fourth order polynomial in m.

11
Relative intensities – temperature measurements
At thermodynamic equilibrium the population of the different rotational levels,
within a given vibrational level, is given by the Boltzmann distribution.

N " = ( 2" + 1) e − B"( " +1) / kT

We have here regarded that each rotational state has 2"+1 magnetic sublevels
and assigned the population of the lowest rotational level (" = 0) to 1.
The relative intensity of a spectral line is generally given by:
I = ε (λ ) ⋅ A ⋅ N
where ε(λ) is the relative detection efficiency and A is the transition probability.
If the A value is assumed to be the same for all rotational transitions and ε is
constant it is seen that the relative intensity is a direct measure of the relative
population of different rotational levels, which in turn depends on the
temperature. The observed relative intensities in the R- and P-branches can
therefore be utilized for non-contact temperature measurements. The highest
populated rotational level in the lower vibrational state at a given temperature is
approximately given by:

µ ⋅ R02 ⋅ k ⋅ T 1
" max = −
! 2
In fact the probability for absorption depends on both the "-value and ∆". One
can show that a better model for the relative intensity is:

I ∝ (""+ "'+1) ⋅ e − B""( ""+1) / kT

where "" is the rotational quantum number in the lower state and "' is the
rotational quantum number in the upper state. This results in different relative
intensities in the R- and P-branch.

I R ∝ ( 2""+2) ⋅ e − B""( ""+1) / kT , "" = 0,1, 2,... "' = ""+1

I P ∝ 2""⋅e − B""( ""+1) / kT , "" = 1, 2, 3,... "' = ""−1

12
 Atomic and Molecular Spectroscopy 2006

Figure 7. Relative intensities in the R- and P-branch in CO. Note that the R-
branch is slightly more intense.

Car exhaust

Air pollution contributes to deteriorated health for many. In densely populated

areas the traffic is believed to cause between 300 and 2000 cancer incidences in
Sweden each year.
The most common pollutions in car exhaust are:

Carbon monoxide (CO)

Carbon monoxide is a gas without smell or colour that is formed when the
combustion in an engine cylinder is incomplete. When engine temperature is low
the content of CO in the exhaust is especially high.
The health risks due to CO are that it decreases the oxygen transporting ability of
the blood. CO also affects the heart and artery system.
The problem with too high concentrations of CO has been reduced with
improved emission control.

Hydrocarbons (CxHy)
During combustion of petrol and diesel oils the exhaust contains hydrocarbons
due to incomplete combustion. Other harmful substances, such as ozone, are
created when the hydrocarbons spread in the air. Hydrocarbons are also
deposited in crop fields and through the food accumulated in the human
digestive system.
Hydrocarbons can yield irritation in eyes and throat. Certain types are directly
carcinogenic while others are suspected to increase the risk for developing
cancer. The ozone decreases the physical performance something which is
serious for especially asthmatic people.

13
Nitrous oxides (NOx)
Nitrous oxides are formed during combustion, especially in engines that are
subject to high load. The car traffic contributes largely to the total emission of
nitrous oxides in urban areas. Nitrous oxides forms, together with hydrocarbons,
ozone.
Nitrous oxides can result in decreased breathing capability, especially difficult
for asthmatic people.

Particles
The combustion of fuel results in paticles, in particular soot. Emission from
diesel engines is the major part of the particle emission from vehicles.
Increased doses affect breathing.

Emission control by catalytic converter

Most vehicles on the market today have either a Otto- or a Diesel engine. Most
of the Otto engines have a so called three way catalytic converter that reduces
the emission of NOx, CxHy and CO. The condition for optimal catalytic
performance is that the engine runs with a constant air/fuel-mixture. This
condition is not fulfilled in a Diesel engine since this engine runs with an
abundance of air. In a three way catalytic converter hydrocarbons and nitrous
oxides are oxidized to carbon dioxide and water. At the same time the carbon
monoxide is transformed to nitrous gas, carbon dioxide and water
2 CO + O2 Ÿ 2 CO2 Oxidation of carbon monoxide
C2H4 + 3O2 Ÿ 2 CO2 + 2 H2O Oxidation of hydrocarbon
2 NO + 2 CO Ÿ N2 + 2 CO2 Reduction of nitrous oxide

In spite of the catalytic emission control present in almost all cars of today the
emission of nitrous oxides hardly decreases. There are several reasons to this:
1. Cold engine starts and short trips. A cold catalytic converter does not operate
optimally.
2. The catalytic converter is aging fast which reduces the efficiency.
3. The emission control is not working properly at high engine loads and high
rpm:s (number of revolutions).

14
Appendix A: Fourier Transform Spectrometer

A Fourier trasnform spectrometer is a Michelson interferometer where one of the

mirrors can move with a constant speed. Hence the optical system is simple, the
interferometer consists of two mirrors, one detector and a beam splitter. In
Figure A1 it is shown how a beam from A hits the beam splitter C. Some part of
the light (ray 1) is reflected by the beam splitter towards the mirror M1. The ray
is reflected from mirror M1 back through C to the detector. The rest of the light
(ray 2) is transmitted through the beam splitter and is reflected on mirror M2.
Ray 2 is then reflected at C towards the detector.

Fixed mirror

Moving
mirror Ray 1
Light
source

Ray 2

Beam splitter

Detector




Figure A1 A collimator is irradiated with monochromatic light yielding a
parallel ray of light. The ray is split into two components in the beam
splitter. Following reflection in the mirrors another passage through
the beam splitter occurs and thereafter the rays are added on the
detector.

All radiation incident on the interferometer is registered by the detector or is

reflected back to the light source. If both mirrors in the interferometer are
positioned at equal distances to the detector the two light rays propagate exactly
the same path length and arrives at the detector in phase, thus interfere
constructively. The two components that interfere at the light source are on the
other hand completely out of phase. In this case all incident energy is transferred
to the detector. If the movable mirror (M2) is translated a quarter of a wavelength
optical retardation of a half of a wavelength is achieved since the light
propagates back and forth. This results in complete destructive interference of

18
 Atomic and Molecular Spectroscopy 2006

the detected signal. So when the moveable mirror is translated the detector
registers darkness followed by brightness.
A Fourier transform spectrometer can be used to study both emission spectrum
from a light source and absorption spectrum from a gas sample placed in the
light path. The latter is the most common modality and this is what we will use.
For absorption measurements a light source with continuous spectrum in a broad
wavelength range is utilized. The gas sample is placed between the beam splitter
and the detector. Since the light path in the interferometer is in air absorption due
to air will occur even outside the sample cell. Therefore a measurement is done
in two steps: first a background is acquired without the sample cell, secondly a
measurement is done with the cell place in the sample compartment. The
difference of the measurements will then be a measure of the absorption in the
sample under investigation.

Specifications for the instrument

Beam splitter Wave number / cm-1 Wavelength / µm
KBr 350 - 7000 1,4 - 28
Quartz 4000 – 25000 0,4 - 2,5


Detector Wave number / cm-1 Wavelength / µm
DTGS 200 - 6000 1,7 - 50,0
InGaAs 5900 – 13300 0.75 - 1,7
Si (Red) 9100 – 15700 0,64 -1,1
Si (Blue) 15900 - 25000 0,40 - 0,63

Fourier analysis
For monochromatic light with the wavelength λ and wave number σ =1/λ the
phase difference δ at the detector is given by
δ = 2π σ x

where x is the path length difference between the two light paths. The refractive
index is here set to 1. According to the formula for interference between two
waves the total intensity Iσ(x) is given below.
Iσ ( x ) = I 01 + I 02 + 2 I 01I 02 cos(δ ).

If I 01 = I 02 = I 0 we can write:

Iσ ( x ) = 2 I 0 (1 + cos(2π σ x ))

19
If one of the mirrors is moving with a constant speed (v/2) we get, since x = vt ,
a time dependent signal which oscillates around a mean value of I σ
I σ ( x) = I σ (1 + cos 2πvσt ) .

Thus we get a signal, the so called interferogram, with an oscillation that is

dependent on the wave number of the incident radiation. For a monochromatic
light source, e.g. a HeNe-laser, this means that the detector registers an
interferogram with cosine-shape (Figure A2).


Relative intensity

Mirror displacement (Ȝ)

Figure A2 Intensity as a function of mirror translation for monochromatic light.

A polychromatic light source emits light in many frequencies. Each frequency

gives rise to a cosine shaped signal and the resulting interferogram is the sum of
all of these. Only where the distance from both of the mirrors to the beam splitter
is equal we get constructive interference for all wavelengths. At all other
distances some wavelengths interfere constructively while others do not. With
Fourier analysis one can convert the intensity as a function of the mirror
displacement (interferogram) to the intensity as a function of the wave number
(the spectrum).
If the light source intensity distribution is I(σ) the mean intensity that reaches the
detector is proportional to I(σ)dσ and a slowly varying function of σ that takes
into account the transmission and reflection coefficient in beam splitter and
mirrors. The modified source function is denoted B(σ). We now instead detect:
∞ ∞ ∞
Iσ ( x ) = ³ B(σ )(1 + cos 2πσx )dσ = ³ B(σ )dσ + ³ B (σ ) cos(2πx )dσ =
0 0 0

∞
I + ³ B(σ ) cos(2πσx ) dσ

0

where I is the constant level with no modulation present. The second term gives
all information about the spectrum. The lower integration limit can be set to -∞

20
 Atomic and Molecular Spectroscopy 2006

since B (σ ) = 0 for all negativeσ. I(x) is defined as the modulated part of the
interferogram which yield
∞
I ( x) = ³ B(σ ) cos(2π σ x ) dσ
−∞

and the spectrum is achieved through the inverse Fourier transform of this
relation
∞
B(σ ) = ³ I ( x ) cos(2π σ x ) dx
−∞

There are though two fundamental differences from these equations in real
applications. First the spectrum is acquired over a finite distance, i.e. x = L. In
this instrument L is 8 cm (maximum). The finite distance determine the
resolution of the instrument so that the smallest structure that can be resolved is
1/L, that is 1/8 = 0,125 cm-1 in this case. Secondly discrete sampling points are
used with the interval ∆x to calculate the Fourier transform which results in a
Fourier sum instead of a Fourier integral. In this instrument the minimum step
size is 0,0625 cm-1.

Fourier transform spectroscopy compared to grating spectroscopy

It is instructive to compare the Fourier transform technique with a more
traditional spectroscopy modality such as spectroscopy using grating
spectrometers.
An interferometer does not separate the light into its frequency components
before it is registered. This implies that each point in the interferogram consists
of information about all wavelengths in the light source. So if 8000 sampling
points is utilized to acquire the interferogram this means that all wavelengths are
registered 8000 times. This yields the same data collection efficiency as a
spectrograph or a CCD detector. On the other if we compare with a
monochromator where the intensity is measured at the exit slit as a function of
the grating angle one achieve a lot with the FT spectrometer since a
monochromator that collects 8000 sampling points only measures the
wavelength once. This is usually referred as the Fellgett advantage.
In a spectrograph the resolution depends in the slit width since the spectral lines
are images of the entrance slit. To achieve a higher resolution a thin slit is
required which reduces the intensity. Furthermore a lot of light is lost in the
spectral orders that are not detected. In an interferometer, on the other hand, no
slits are needed and the only light lost is the portion if the light that is reflected
back to the light source (approximately half). This leads to that more energy can
be registered than what is possible with a dispersive spectrometer which
increases the signal-to-noise ratio in the spectrum. This is referred to as the
Jacquinot advantage..
The Fellgett and Jacquinot advantages imply that a Fourier transform
spectrometer can acquire a high-quality spectrum in the infrared wavelength

21
region in just a fraction of the time required for a dispersive spectrometer to do
the same thing. Additionally in a Fourier transform spectrometer the resolution
of the detected intensity can be increased through the scanning length of the
moving mirror is increased.
In a dispersive instrument the wavelength determination exactitude and precision
depends on
1. Calibration using external reference wavelengths, i.e. one calibrates the
measured spectrum with wavelengths acquired at another time.
2. The mechanical ability of the instrument to repeat grating movements and
aperture translation exactly during and between different measurements.
In comparison the Fourier transform spectrometer has an internal frequency
standard, usually a HeNe-laser (Figure A3). The moving mirror and the detector
sampling interval are timed through the interference fringes that arises from the
monochromatic light from the laser. All wave numbers in the resulting spectrum
is then calculated using the known laser frequency. This is called the Conne
advantage. A Fourier transform spectrometer can reach a precision in wave
number better than 0.01 cm-1. This means that spectra can be compared
irrespective if these are acquired with five minute interval or five year interval.

Fixed retro-reflective
mirror

Moveable retro-
reflective mirror

Figure A3 The ray trace of the HeNe-laser in the FT-spectrometer.

In a FT-spectrometer the resolution is the same for all wave numbers while the
resolution of the grating spectrometer is wavelength dependent. Finally a
disadvantage of the FT-spectrometer is that the light source has to be very stable
since fluctuations in the light source, i.e. if B(σ) is time dependent, affect the
inverse transform.

22

 

 








 

 



 




 






 








 









 

















 




 


 




  

 
!"



#



 

$

%
&



 
%


'



 

-

.
+
%
/0



 
()

*+%
& 


,
























%
&





 
!
' 



% 





 $









 

,  


1 
!,  


$

  


 2
 

 
  
 

 3


 
  





 
 

'









' 


'


' 3


 
' 

'
/




'

/
!/







$

  !   "
#


 4 
*&4


%




 


/
+




 








1













5

  








 
5







 
6



5

6
7




,














'
' 





 

-

 
 






 8
$
 





 9
$
 




 9

$
 




%

'






' 







% 











  '
 




 

5





5


:














1
 



  

5













5












%
&







5




:




-1

!

/<0

$







 
+


















1
5





 
21*41





 





'



+ 






'
;








'









 



-1



 
 

1
5









 

 
! 




  $







































 
2
=

+






5




:



!
/<0

$






2
=



,




-1





()

*+%


& 

,







%
&









 
+


&




-1











1







'








 


  
!








'


 $

1 



 










'

! 





 $





































 




!: 
 $

:
!
 $

:'

 

2' 

>/?




,  

  


'





 


  




 




1   

 
1!$



*


'

 
:!$



 
!$




 

   

 
 



1!$
7
!$
@
:!$

! 
 



*&4$



+


   

 


   

 



/0))/0A)
/
!
'




 $




 






+


 

 



! 
 





*&4$









 



 


/
! 
%
 $


















 

1   


1!$

2' 

>/?







'



   

 




+

+



1


+


 

;






/00BC0C
/

5






5


58









1%

 1

,  
1%

1

1









+
7

/)
!1%$



+%
7

/)
!1$

 1




!

 $



+
7
+
7
+%

+




=



: 
 
1  

:'

,  



1





'


 








 






+
!
 





$




1

 




'





 
  


'

   

1

 




 






,
 
+



8@@ 5@
 @@@










1
:



1 
1%
1
+
+%
+
C
>D?
>E?
>5@ ?
>5@5?

/










<










C










B










0










F










G










A










(










/)













&




&




+
' 


 




/)
 





!
)
 $





 














2





 
  3



2








  3



2 






'



3



<
+
 
 

!


  





2

#
4&

<))<$



:

 








 


  






 






 




% 



5

 






 
 


5






/<0



C)




1
 

 



 '
 







  







1

&

,




1
 




'













  8



 


 

 

1

 



 
 





 






 

 



  

1
 







 



%  

 


 
   

 








 

 8




7

! 
H
   $



!/$





 
 


  





 









  


 

 
  

1
 

 

'

  






  






' 

 



 

 


 



1









 
 

+
 

' 

I

!/$
 
5


1





 



1 







5


  







5



! 

5



$8




7

! 
H
   $



!<$




7

! 
H
   $



!C$



 



 







5






 








5



1


 

 



5





5



5


 
   

4 

I

/
<

C




%  





 


;
 








5



 





 8

 


              

 






J '


'
I

 
!/$
!<$

!C$






































4 
5


:











%
&






()

*+%



& 

,





% 












 
+


&


-1


-1








5




:



!- 
C)




/<0

$












2

!
$9


2


!
$




4 
 




 








 
5











'



















 





 





























  



    !!"#!$%




 




!

7



/<0
$


 





 










!

7
C)
$



   



 
 

&'





 


 
 




J 




I 
8


-1:







 3

2


  


 


 


3


-


5

5







 
! $

 






-
 
  





I


+ 


 







'

!

' 

$

5

 
 





  

 

 














5



+


 






;



K
L




 

 
 


'


 


5
































 
:
 



 


& 



! 


 




5

$

1
I


    
          

 






 
 

 





 

'


 

 




 8

+
 

 


! 

1  $

1

2' 

>/?





1
 

 



 

   

 


 

 



1:+
 





 



+
 

 


! 

1  $

1

1   


1!$
>E?

2' 

>/?