Online Detection of Water Vapor in an
Industrial Gasifier Using Terahertz
Spectroscopy
Master’s Thesis in Physics and Astronomy

JENS NORDMARK

Department of Microtechnology and Nanoscience
Terahertz and Millimeter-wave Laboratory
Chalmers University of Technology
Gothenburg, Sweden 2013



Online Detection of Water Vapor in an Industrial Gasifier Using Terahertz Spectroscopy
Jens Nordmark

c©Jens Nordmark, 2013

Terahertz and millimeter wave laboratory
Department of Microtechnology and Nanoscience
Chalmers University of Technology
SE-412 96 Gothenburg
Sweden



Abstract

A terahertz spectrometer was set up for the monitoring of product gas in a thermal gasi-
fier. The goal was to measure temperature and H2O concentration, and preferably also
CO concentration. This environment presents several difficulties such as high tempera-
ture, toxic gases, high H2O content and a high concentration of particulate pollutants.
Spectroscopy using IR Lasers tend to be obstructed by the high absorption of H2O as
well as the scattering by the particulates, and so does not produce satisfactory results.
It was hoped that these problems could be avoided by using terahertz radiation, with
a lower absorption by water and a longer wavelength possibly avoiding scattering by
particulates. A system operating in the range 300-500 GHz was built and tested in lab-
oratory as well as in an industrial gasifier at Chalmers Power center, yielding promising
results.

Our results indicate that monitoring of H2O, and probably other gases, in the rugged
environment of gasifier rawgas can be done with response times on the order of a few
minutes. This has applications in biomass gasification, where variable quality of the fuel
results in a need for continous monitoring and tuning of the gasification process.

Keywords: Terahertz, Gas spectroscopy, Gasification





Acknowledgements

I wish to thank my supervisor Sergey Cherednichenko for his help. I also wish to thank
Hosein Bidgoli and Martin Seeman at the department for energy and the environment
with whom the project was carried out. Also thanks to Johannes Öhlin for assistance
during experiments at the Chalmers power center.

Jens Nordmark, Gothenburg July 1, 2013





Contents

Contents i

List of Tables iv

List of Figures vi

Nomenclature x

1 Introduction
1

1.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Theory
5

2.1 Gasification of biomass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Terahertz technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2.1 H2O and CO absorption at Terahertz frequencies . . . . . . . . . . 9
2.2.2 Detectors and sources . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Waveguides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.3 Molecular spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1 Energy levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

The vibration-rotation spectrum of CO . . . . . . . . . . . . . . . 15
2.3.2 Line shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Natural broadening . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Doppler broadening . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Collisional broadening . . . . . . . . . . . . . . . . . . . . . . . . . 17
Combined broadening . . . . . . . . . . . . . . . . . . . . . . . . . 17
Temperature dependence of line strength . . . . . . . . . . . . . . 18
Line-by-line models . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.3.3 Collisional line mixing . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.4 The HITRAN database . . . . . . . . . . . . . . . . . . . . . . . . 19

i



CONTENTS

2.3.5 LINEPAK vs custom algorithm . . . . . . . . . . . . . . . . . . . . 19

3 The setup
20

3.1 Gas cell design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.1.1 Window properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.2 Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.1 Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.2 Detector: Golay cell . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.3 Humidity meter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Spectrum analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.4.1 THz absorption by H2O vapors . . . . . . . . . . . . . . . . . . . . 32
3.4.2 THz absorption by Carbon Monoxide . . . . . . . . . . . . . . . . 32

4 Experimental methodology
35

4.1 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Laboratory H2O measurements . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Monitoring of H2O and CO at the gasifier . . . . . . . . . . . . . . . . . . 40
4.4 Noise and drifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5 Discussion
48

5.1 Detection of H2O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.2 Detection of CO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.3 Noise and drifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.4 Other issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5.4.1 Temperature gradient and thermocouples . . . . . . . . . . . . . . 54
5.4.2 Line-by-line model vs LINEPAK . . . . . . . . . . . . . . . . . . . 54

5.5 Usability of the system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Bibliography 57

A Line-by-line algortihm
60

B Line data from HITRAN2008
62

C SpectralCalc simulations
64

D Calculations of T and VMR from S380 and S448

70
D.1 Determination of T from measured VMR and S . . . . . . . . . . . . . . 70

ii



CONTENTS

D.2 Determination of VMR from calculated T and measured S . . . . . . . . 72
D.3 Simultaneous determination of T and VMR from S . . . . . . . . . . . . 73

E Humidity control
75

F Alternative waveguides
76

Index 77

iii



List of Tables

2.1 The expected contents of gasifier rawgas at Chalmers power center, based
on gas chromatography and H2O separation by condensation. All VMR
entries except H2O refer to dry gas composition. (Hosein Bidgoli, private
communication) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2 The expected contents of gasifier fluegas at Chalmers power center, based
on gas chromatography and H2O separation by condensation. All VMR
entries except H2O refer to dry gas composition. (Hosein Bidgoli, private
communication) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.3 Detector types for far IR and THz. A summary of Rogalski and Sizov [2011] 12
2.4 Sensitivity for some detector types for far IR and THz. A summary of

Rogalski and Sizov [2011], last entry is a bolometer made at TML, MC2,
Chalmers. Described in Cherednichenko et al. [2011]. . . . . . . . . . . . . 12

3.1 Amplitudes of frequency swings due to modulation with various ampli-
tudes. ∆fin is the maximum deviation from the mean frequency at any
particular value of the control voltage Uctrl, and is due to the value of
the FM voltage UFM. ∆fout is the output frequency of the AMC after
multiplication. The relation between FM modulation voltage and output
frequency swing varies slightly across the operating range of our source,
as can be seen in the difference in output from these two control voltages
when all other parameters are kept constant. . . . . . . . . . . . . . . . . 27

3.2 Coefficients for the Antoine equation for H2O, from DDBST [2013]. . . . 31

4.1 Transmissions obtained at different humidity levels in two separate mea-
surements of the type shown in figure 4.2. Errors are shown as one stan-
dard deviation. Measurement number indicates which of the two separate
dataseries the entry comes from. . . . . . . . . . . . . . . . . . . . . . . . 37

iv



LIST OF TABLES

5.1 Temperature estimations using least-squares from the data in table 4.1.
The calculation is shown in appendix D. Errors are given as one standard
deviation. The last column shows how VMR is calculated assuming the
calculated temperature and measured transmissions. . . . . . . . . . . . . 50

5.2 The result when fitting both T and VMR using the program in appendix
D.3, and the same data as in the table above. . . . . . . . . . . . . . . . . 50

B.1 Spectral lines of CO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
B.2 Spectral lines of water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

F.1 Transmission at three distances with and without waveguide, normalized
to the 10 cm air transmission. . . . . . . . . . . . . . . . . . . . . . . . . . 76

v



List of Figures

1.1 Scattering cross-section as function of frel = R/λ for a spherical particle.
Calculated with MiePlot 4.3 by Laven [2013], implementing the BHMIE
algorithm by Bohren and Huffman [2007]. [data/mie] . . . . . . . . . . . . 2

1.2 The spectra of two gas mixtures at room temperature (296 K). 10% H2O
(blue) and 10% H2O+10% CO (red). [data/introspectrum] . . . . . . . . 2

1.3 The simulated spectra of H2O at VMR = 20% and varying temperatures.
[data/AppendixC] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.4 The simulated spectra ofH2O at T = 573K and varying humidity. [data/AppendixC] 3

2.1 The device measuring H2O content by condensation. H2O and tar con-
dense while bubbling up from the bottom of the isopropanol pool, and
mix with the liquid. The mass increase of the isopropanol container gives
the amount of H2O and tar in a volume of gas, ∆m = mH2O +mtar. To
give a good accuracy, the mass increase must be measured over periods
of several hours. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2 The operating principle of the Chalmers gasifier . . . . . . . . . . . . . . . 8
2.3 Top: Spectra of 10% H2O and 10% CO + 10% H2O in the range 100GHz

to 1 THz at 296 K. Bottom: The same restricted to 300-500 GHz. [data/simulation
full] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.4 vibrational modes and rotational axes of CO and H2O . . . . . . . . . . . 15
2.5 Sketch of typical rovibrational spectrum for a diatomic molecule with a

vibrational transition at 1000, and rotational constant of 1/2, in arbitrary
units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.6 The type of transitions giving rise to line mixing. . . . . . . . . . . . . . . 19

3.1 A sketch of the gas cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 The gas cell and its connections to the rest of the system. . . . . . . . . . 21
3.3 The gas cell set up at the gasifier . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 The gas cell in the lab within the closed oven, tubing and electronics. . . 22
3.5 The gas cell inside the open oven . . . . . . . . . . . . . . . . . . . . . . . 23

vi



LIST OF FIGURES

3.6 Transmission through teflon(top) and quartz(bottom) windows at normal
incidence relative to air. [data/windowproperties/] . . . . . . . . . . . . . 25

3.7 The setup of the instruments around the gas cell . . . . . . . . . . . . . . 26
3.8 A nitrogen spectrum in terms of control voltage vs output voltage from

the Golay cell. The voltage span corresponds to a full frequency sweep
over 300-500 GHz. Note the very high sensitivity to small variations
in the contol voltage. This sensitivity motivated the use of frequency
modulation. [data/ctrl-to-signal] . . . . . . . . . . . . . . . . . . . . . . . 27

3.9 Sketch of the setup of the source, from VDI manual. . . . . . . . . . . . . 28
3.10 Source power output, from VDI manual. . . . . . . . . . . . . . . . . . . . 28
3.11 Sketch of the Golay cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.12 Left: Golay cell. Right: The microwave source . . . . . . . . . . . . . . . 30
3.13 The Amplifier multiplier chain, concealed within a black box. . . . . . . . 30
3.14 Line positions and intensities for H2O. Upper left: 0-40 cm−1 with cutoff

at intensity < e−23. Upper right: 0-40 cm−1 with no cutoff. Lower left:
10-16.7 cm−1 with cutoff at intensity < e−23 Lower right: 10-16.7 cm−1

with no cutoff. Compare figure 2.3. . . . . . . . . . . . . . . . . . . . . . . 33
3.15 Line positions and intensities for CO. Upper left: 0-40 cm−1 with cutoff

at intensity < e−23. Upper right: 0-40 cm−1 with no cutoff. Lower left:
10-16.7 cm−1 with cutoff at intensity < e−23 Lower right: 10-16.7 cm−1

with no cutoff. Compare figure 2.3. . . . . . . . . . . . . . . . . . . . . . . 34

4.1 Top: Two raw measurements, one of N2 which lacks spectral lines in this
frequency range to be used for calibration, and one of gasifier rawgas. Bot-
tom: The transmission spectrum formed by dividing the rawgasspectra
by the calibration spectra. VMR ≈ 65% and T ≈ 430oC. [data/rawspectra] 36

4.2 Top: the two water lines are plotted as functions of time, where the
H2O content is changed in steps. Bottom: the humidity meter gives the
corresponding VMR. [data/twolines] . . . . . . . . . . . . . . . . . . . . . 38

4.3 Laboratory measurement of transmission at 449 GHz over time for five
humidity levels. The impulse-like objects are products of the humidifier,
it seems to release steam in short pulses. Note that measurements with
a difference in VMR as small as 0.4% can be clearly distinguished from
each other. [/data/20130204] . . . . . . . . . . . . . . . . . . . . . . . . . 39

4.4 The spectrum of the rawgas, 300-500 GHz. Note the CO line at 460 GHz.
Also included is a fluegas measurement at 410-470 GHz. Spectralcalc
simulations for the rawgas are also included, compare table 2.1 of rawgas
contents. [data/rawgasplot] . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4.5 Raw and fluegas from the same measurement as above, zoomed in on the
H2O-line at 448 GHz.[data/rawgasplot] . . . . . . . . . . . . . . . . . . . 41

4.6 H2O (41%) compared to H2O (41%) and CO (20%). [/data/coandh2o] . 42

vii



LIST OF FIGURES

4.7 Measurement over time of the 448 GHz transmission at the gasifier. Vary-
ing the steam input of the process: 160, 200, 240, 160, 120 kg/h. [data/gasifier-
humiditylevels] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.8 N2 spectra taken over a period of four hours for the gas cell at room
temperature and normalized to the first one of them. Moving average
over 15 GHz.[data/drifts/book3-20130311] . . . . . . . . . . . . . . . . . . 45

4.9 N2 spectra taken over a period of eight hours at high temperature (oven
at 600 C) and normalized to the first one of them. The legend shows
the time that the corresponding measurement was taken. Moving average
over 15 GHz.[data/drifts/book1-20130312] . . . . . . . . . . . . . . . . . . 46

4.10 N2 spectra taken over a period of five hours at high temperature (oven
at 600 C) and normalized to the first one of them, with better thermal
isolation (10 cm more air between source and gas cell) than in figure 4.9.
Moving average over 15 GHz.[data/drifts/book2-20130327] . . . . . . . . . 46

4.11 Noise plot in the lab obtained by division of two subsequent N2-spectra.
Integration time 100 ms. Note that signal strength approaches zero at
the ends of the spectrum, causing signal-to-noise ratio to decrease. In
this plot no smoothing has been used. [data/noise] . . . . . . . . . . . . . 47

4.12 Noise plot at the gasifier obtained by division of two subsequent N2-
spectra. Integration time 100 ms. In this plot no smoothing has been
used. [data/noise] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.1 Transmission as function of temperature calculated by the line-by-line-
model for the VMR of 73% detected in the measurements of figure . The
black lines show where the measured transmissions intersect the simula-
tion, giving the temperature. The values of T should coincide, but do not.
Precise estimates from a least-squares fit for several measurements of this
type are shown in table 5.1. [data/temp humidity] . . . . . . . . . . . . . 49

5.2 Transmission as function of temperature calculated by Spectralcalc for the
VMR of 73% detected in the measurements of figure . The black lines
show where the measured transmissions intersect the simulation, giving
the temperature. The values of T should coincide, but do not. The
temperatures differ much more than above. [data/temp humidity] . . . . 49

5.3 An N2 measurement at 383 GHz with 300 millisecond integration time
and one data point taken per second, in the lab.[data/rmsd] . . . . . . . . 52

5.4 Allan variance of the above signal.[data/rmsd] . . . . . . . . . . . . . . . . 52
5.5 The 449 GHz line from rawgas at the gasifier (200 kg/h of H2O), mon-

itored for 1200 seconds. 100 millisecond integration time and one data
point taken per 300 milliseconds.[data/rmsd] . . . . . . . . . . . . . . . . 53

5.6 Allan variance of the above signal. [data/rmsd] . . . . . . . . . . . . . . . 53
5.7 A comparison of rawgas measurment, spectralcalc and line-by-line simu-

lation. [data/comparison] . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

viii



LIST OF FIGURES

C.1 The simulated spectra of H2O at VMR = 20% and varying temperatures.
[data/AppendixC] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

C.2 The simulated spectra ofH2O at T = 573K and varying humidity. [data/AppendixC] 65
C.3 Ratio of H2O peak transmission at 448 and 383 as function of temperature

when VMR is kept constant at 20%. [data/AppendixC] . . . . . . . . . . 66
C.4 Ratio of H2O peak transmission at 448 and 383 as function of VMR when

T is kept constant at 573 K. [data/AppendixC] . . . . . . . . . . . . . . . 66
C.5 The simulated transmission of H2O at 380 GHz for various T and VMR.

[data/AppendixC] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.6 The simulated transmission of H2O at 448 GHz for various T and VMR.

[data/AppendixC] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.7 The simulated transmission at 380 GHz, dependence on T for 10 different

values of H2O VMR. [data/AppendixC] . . . . . . . . . . . . . . . . . . . 68
C.8 The simulated transmission at 448 GHz, dependence on T for 10 different

values of H2O VMR. [data/AppendixC] . . . . . . . . . . . . . . . . . . . 68
C.9 The simulated transmission at 380 GHz, dependence on H2O VMR for

12 different values of T . [data/AppendixC] . . . . . . . . . . . . . . . . . 69
C.10 The simulated transmission at 448 GHz, dependence on H2O VMR for

12 different values of T . [data/AppendixC] . . . . . . . . . . . . . . . . . 69

ix



Nomenclature

CH4 Methane
CO Carbon monoxide
CO2 Carbon dioxide
H2 Hydrogen
RH Relative Humidity, ratio of H2O partial pressure to saturated vapor

pressure.
VMR Volume Mixing Ratio
AMC Amplifier-/Multiplier Chain
EM Electromagnetism, Electromagnetic
GPIB General purpose interface bus, same as IEEE 488
HITRAN high-resolution transmission molecular absorption database
IEEE Institute of Electrical and Electronics Engineers
IEEE 488 General purpose interface bus, GPIB
IR Infrared
LINEPAK Software library for spectral simulation described in Gordley et al. [1994].
NEP Noise Equivalent Power
RMSD Root Mean Square Deviation
SNG Substitute Natural Gas, methane
tar A loosely defined viscous black substance consisting mainly of large

hydrocarbons, often aromatic, polycyclic and containing various sub-
stituents.

THz TeraHertz
TTL Transistor-Transistor Logic, a binary logic scheme using 0-0.8 V signals

for low and 2.2-5 V for high

x



1
Introduction

Gasification of biomass is a technology expected to become more important as fossil
fuels are to be replaced by carbon neutral bioenergy. The company WSP has made
projections based on three scenarios, showing a potential for up to 12 TWh in biogas
production by thermal gasification up to 2030 under favorable conditions. Distribution
of the combustible gas produced (typically SNG, i.e. methane) could take place using
the present natural gas infrastructure (Dahlgren et al. [2013]).

The properties of the fuels used are however such that the producer gas, also called
rawgas in the stage immediately after it is generated, is full of char particles at sizes up
to about one micron. At the same time its gas contents vary in time, and need to be
tracked to tune the process. The particulate contents as well as the high H2O content
disturb the IR-spectroscopy that would typically be used to monitor gas content. H2O
simply has to high absorption in the IR for such radiation to be used successfully. The
wavelength of IR is also at the order of the particulates diameter, giving rise to scatter-
ing that disrupts measurements. Gas chromatography is another technique traditionally
used, but it cannot handle gas mixtures containing H2O, which condenses in the chro-
matograph and destroys it. It is possible to measure the dry gas composition by first
drying and cleaning the gas. This also provides a way to determine the sum of H2O and
tar contents, but only the total content and not both independently. In addition, it only
works with a time delay of tens of minutes.

As an alternative solution, this project aimed to use terahertz spectroscopy in the
hope to escape the aforementioned problems. Our aim was to measure content of H2O
as well as CO, one of the dry gases present. Our results indicate that this approach does
work as desired.

There are several reasons why this approach succeeds. THz has a lower absorption
for H2O, avoiding saturation of the spectral lines. THz has longer wavelengths than
IR, which reduces the scattering by particulates. The scattering of radiation by ideal
spherical particles is described by Mie theory, where a threshold in the scattering cross

1



CHAPTER 1. INTRODUCTION

1.00E 11

1.00E 09

1.00E 07

1.00E 05

1.00E 03

1.00E 01

1.00E+01

1.00E+03

0 0.5 1 1.5 2 2.5 3

Cross section

Scatterer!size!(Radius/wavelength)

Figure 1.1: Scattering cross-section as function of frel = R/λ for a spherical particle.
Calculated with MiePlot 4.3 by Laven [2013], implementing the BHMIE algorithm by Bohren
and Huffman [2007]. [data/mie]

0.5

0.6

0.7

0.8

0.9

1

300 320 340 360 380 400 420 440 460 480 500

Tr
an

sm
is
si
on

Frequency [GHz]

H2O+CO
H2O

Figure 1.2: The spectra of two gas mixtures at room temperature (296 K). 10% H2O
(blue) and 10% H2O+10% CO (red). [data/introspectrum]

section appears when the wavelength is comparable to the size of the scatterers, which
can be formulated in terms of the relative frequency frel = R/λ ≈ 1/2 as shown in figure
1.1. In the low frequency limit, there is weak Rayleigh scattering that decreases quickly
with particle size, while in the optical limit of large frel the scattering cross section is
close to 1. By increasing λ we decrease relative frequency (frel) and thus move closer to
the Rayleigh limit. This was one of the motivations for using THz radiation.

Spectroscopy in field environments using terahertz has been demonstrated in several
applications, for instance the remote detection of chemicals in Gopalsami et al. [2008]
or the remote detection of nuclear radiation in Gopalsami et al. [2009]. Currently the
spectra of H2O as well as that of CO in the region 300-500 GHz are well known and
included in the HITRAN database. This spectral range contains a few strong H2O lines
and a few conveniently places CO lines, shown in figure 1.2.

The absorption at the line centers is dependent on both concentration and tem-

2



CHAPTER 1. INTRODUCTION

perature as shown in figures 1.3 and 1.4. Thus it is possible to determine these two
parameters from measured spectra.

Figure 1.3: The simulated spectra of H2O at VMR = 20% and varying temperatures.
[data/AppendixC]

Figure 1.4: The simulated spectra of H2O at T = 573K and varying humidity.
[data/AppendixC]

3



1.1. DATA
CHAPTER 1. INTRODUCTION

1.1 Data

Some of the data and code used for obtaining the figures in the text has been put in the
following Google Drive folder:
https://drive.google.com/folderview?id=0B25vmzARFvvOdnlYNDBocEFQNGM

4

https://drive.google.com/folderview?id=0B25vmzARFvvOdnlYNDBocEFQNGM


2
Theory

2.1 Gasification of biomass

Production of combustible gas from biomass can be achieved using several processess,
including biological and thermal. In the present project, we are concerned with thermal
gasification of biomass, particularly from wood pellets. However, many types of high-
grade biomass are required in other industries and the energy sector will have to settle for
the leftovers. Thus it will be necessary to handle a highly variable input in a process. This
is why online detection of the product is required, since the process must be continuosly
tuned to the variable input.

In thermal gasification the solid biomass is broken down thermally in the presence of
an oxidising agent, in this case H2O. The main products are CO, CO2, H2, CH4, tars
and other hydrocarbons. The dry gases and the water contents expected in the rawgas
of the Chalmers gasifier are given in table 2.1, and for the fluegas data are given in table
2.2. Primary tars are of the form CxHyOz, but are themselves broken down into either
new tars, H2 or the product gases mentioned. The process requires a temperature above
600oC. (Gomez-Barea and Leckner [2010]).

The H2O content is of critical importance in regulating the rate of the process, since
it is involved in most reactions breaking down the large molecules of the fuel into product
gas. The contents of the gas must be very finely tuned for downstream processes to work,
which requires a low-delay monitoring of the most significant species. The intended
downstream process in this case is generation of substitute natural gas (SNG), by the
process of water gas shift:

3H2 + CO → CH4 +H2O (2.1)

This is highly sensitive to the ratio of CO to H2.

5



2.1. GASIFICATION OF BIOMASS
CHAPTER 2. THEORY

Table 2.1: The expected contents of gasifier rawgas at Chalmers power center, based on
gas chromatography and H2O separation by condensation. All VMR entries except H2O
refer to dry gas composition. (Hosein Bidgoli, private communication)

Gas H2 CO CO2 CH4 C2H2 C2H4

VMR [%] 4.64 23.40 35.50 16.35 13.78 0.33

Gas (cont.) C2H6 C3H6 C3H8 He N2 H2O

VMR [%] 0.70 0.47 0.02 1.11 3.42 ca 65

Table 2.2: The expected contents of gasifier fluegas at Chalmers power center, based on
gas chromatography and H2O separation by condensation. All VMR entries except H2O
refer to dry gas composition. (Hosein Bidgoli, private communication)

Gas N2 O2 CO2 H2O

VMR [%] 79.5 3.5 17 ca 20

Figure 2.1: The device measuring H2O content by condensation. H2O and tar condense
while bubbling up from the bottom of the isopropanol pool, and mix with the liquid. The
mass increase of the isopropanol container gives the amount of H2O and tar in a volume of
gas, ∆m = mH2O + mtar. To give a good accuracy, the mass increase must be measured
over periods of several hours.

6



2.1. GASIFICATION OF BIOMASS
CHAPTER 2. THEORY

Other methods than THz spectroscopy to monitor H2O are IR spectroscopy or con-
densation (explained in figure 2.1). IR spectroscopy has the problems of increased scat-
tering by the particulates, due to the shorter wavelength of IR compared to terahertz.
The absorption of water is also very high in the IR, creating a risk of saturation of such
spectrometers. Condensation is more robust: just let the H2O condense and compare its
mass to the volume of gas it came from. But this creates a long delay and also fails to
differentiate between H2O and tar, since tar also condenses. This motivates the attempt
to use THz instead.

A gasifier has a bed of material with high heat capacity to keep the fuel at an even
temperature. There are several designs of these beds, in the Chalmers case it is a fluidized
bed, made of sand that has properties resembling a fluid when under the conditions of
the gasifiers operation. The fuel is mixed with the bed material. Heat is provided
by a separate combustion chamber, this is called external heating to differentiate from
the case where the gasifier itself provides heat. Bed material is continously circulated
between combustion chamber and gasifier to provide heat. Bubbles of steam will move
upward through the bed material, accumulating gasifier products along the way. At this
stage the particulates in the gas are small, submicron size, so they will not affect the THz
radiation much. The temperature of flue- and rawgas leaving the gasifier is anticipated
to be around 850o C, which is the operating temperature of the process, but it cools
down to about 350o C before reaching the gas cell, since it is transported in heating
hoses with a capacity of maximum 350o C. It is then heated again inside the oven to
430o C to avoid condensation in the gas cell.

Figure 2.2 shows a flow diagram of the process at the Chalmers gasifier. Fuel is
fed from above into both chambers. In the combustion chamber, air is injected for
combustion while in the gasification chamber steam, the oxidizing agent, is inserted
instead. Heat is transported from the combustion chamber to the gasification chamber
by circulation of the hot sand of the fluidized bed. The output gas of the combustor is
referred to as fluegas, while the product gas of the gasifier is called rawgas when it is
leaving the gasifier, before being cleaned for downstream use.

7



2.1. GASIFICATION OF BIOMASS
CHAPTER 2. THEORY

Figure 2.2: The operating principle of the Chalmers gasifier

8



2.2. TERAHERTZ TECHNOLOGY
CHAPTER 2. THEORY

2.2 Terahertz technology

The terahertz region of the electromagnetic spectrum is defined as wavelengths from
3mm to 30 µm. It starts at the upper edge of millimeter waves and ends at the lower
edge of the infrared. Generation and detection of radiation in this range is more difficult
than in the microwave region of the EM spectrum, and THz equipment is currently an
area of intense research.

Several applications of THz radiation exploit its ability to penetrate tissue, fabric and
plastics to a significant depth, while being partially reflected depending on density and
contents of objects. This makes possible a variety of imaging applications. In addition,
it is non-ionizing1 and thus harmless to living organisms unless used with very high
power output, which tends to be neither needed nor available. In the present project,
we attempted to exploit THz radiation in the monitoring of water vapor content in dirty
gasifier rawgas, where IR laser spectroscopy fails due to the particulate matter in the
gas. THz was expected to pass through the particulates. We also face the issue of
high temperatures. Temperature affects spectral lines both by changing their relative
strengths and by broadening of lines. This is further described in section 2.3.2.

2.2.1 H2O and CO absorption at Terahertz frequencies

We need to detect H2O under various concentrations and temperatures. It is important
to avoid saturation of the spectrometer. In figure 2.3 we show the spectra of H2O as
well as CO. Three of the H2O lines are very strong and risk to saturate our system. We
want several weak lines in the frequency range used. The band 300-500 GHz suits this
purpose for both gases considered, and equipment for sweeping over these frequencies
exists. The danger of saturation is exacerbated by high temperatures, which increase
absorption significantly. See appendix C for plots of the temperature dependence of
absorption at various concentrations of H2O.

1Photon energies vary from 0.413 meV at 3 mm to 41.3 meV av 30 µm

9



2.2. TERAHERTZ TECHNOLOGY
CHAPTER 2. THEORY

0
0.
1

0.
2

0.
3

0.
4

0.
5

0.
6

0.
7

0.
8

0.
91

10
0

20
0

30
0

40
0

50
0

60
0

70
0

80
0

90
0

10
00

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

H2
O
+C

O
H
2O

0.
9

0.
91

0.
92

0.
93

0.
94

0.
95

0.
96

0.
97

0.
98

0.
991

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

H2
O
+C

O
H2

O

F
ig
u
re

2
.3
:

T
op

:
S

p
ec

tr
a

of
10

%
H

2
O

an
d

10
%
C
O

+
1
0
%
H

2
O

in
th

e
ra

n
g
e

1
0
0
G

H
z

to
1

T
H

z
a
t

2
9
6

K
.

B
o
tt

o
m

:
T

h
e

sa
m

e
re

st
ri

ct
ed

to
30

0-
50

0
G

H
z.

[d
at

a/
si

m
u
la

ti
on

fu
ll
]

10



2.2. TERAHERTZ TECHNOLOGY
CHAPTER 2. THEORY

2.2.2 Detectors and sources

Detectors for THz can be broadly categorized into thermal or photon detectors, as Ro-
galski and Sizov [2011] describe. Thermal detectors absorb the radiation and gain heat,
which induces change in some physical property of the material that can be easily mea-
sured. For instance, metals or semiconductors can be used that change their electrical
conductivity in response to temperature differences. This is a common category of de-
tectors called Bolometers. Another example would be to measure the thermal expansion
of a gas, which is the method we use. That type of detector is called a Golay cell and
is described in detail in section 3.2.2. Thermocouples are devices that generate voltage
over a junction of two different materials in response to temperature changes. They are
used in many applications and can be used as IR detectors as well as for temperature
measurements in general.

Photon detectors rely on excitations in semiconductors. When a photon is absorbed
its energy gives rise to an electron-hole pair. These gives rise to electrical conduction, in
a variety of ways for different detector types. They have a frequency-dependent response,
unlike thermal detectors. Their advantage is that they tend to have very fast response
times and good signal-to-noise ratio. A major inconvenience is that they require cooling
to very low temperatures to avoid thermally generated noise. A thermal detector is for
this reason the favoured option provided that its precision is acceptable. The Golay cell
was used because it was available. We list some detector types in table 2.3. A measure
of detector sensitivity in common use is Noise Equivalent Power, NEP. It is defined as
the power required of an input signal to generate a signal-to-noise ratio of 1 in an output
signal with 1 Hz bandwidth. The unit is W /

√
Hz. This should of course be as low as

possible for good detectors. A few examples from Rogalski and Sizov [2011] are collected
in table 2.4. A higher sensitivty of a detector lessens the noise and shortens the required
integration time at a particular noise level.

11



2.2. TERAHERTZ TECHNOLOGY
CHAPTER 2. THEORY

Table 2.3: Detector types for far IR and THz. A summary of Rogalski and Sizov [2011]

Category Detector type Mechanism

Thermal Bolometers Temperature dependence of conductivity

Thermocouple Voltage generated by temperature change

Pyroelectric Charged plane conductors changing polarity

Golay cell Thermal expansion of gas

Photon Photoconductors Photon absorption in energies above the band gap cre-
ates carrier/hole pair in a uniform piece of some semi-
conductor, raising conductivity.

p-n junction diodes Photon-induced pair production creates a photocur-
rent across a junction between p- and n-doped semi-
conductors

Schottky barrier diodes Photocurrent across Schottky barrier

Table 2.4: Sensitivity for some detector types for far IR and THz. A summary of Rogalski
and Sizov [2011], last entry is a bolometer made at TML, MC2, Chalmers. Described in
Cherednichenko et al. [2011].

Detector type NEP [W /
√

Hz]

Golay cell 10−9 − 10−10

Schottky diode 10−10, increasing with ν

Bi bolometer 1.6× 10−10, increasing with ν

Nb microbolometer 5× 10−11

Ti microbolometer 4× 10−11

Ni microbolometer 1.9× 10−11

YBCO bolometer 3.7× 10−10

12



2.2. TERAHERTZ TECHNOLOGY
CHAPTER 2. THEORY

2.2.3 Waveguides

Waveguides are conducting tubes used to transfer radiation with as little attenuation as
possible. Their dimensions are chosen with respect to the wavelength of the radiation
considered, and longer wavelengths require broader waveguides. Each waveguide geom-
etry has a cutoff frequency below which decay of the transmitted wave will occur. All
waveguides that are not perfect conductors will experience attenuation, which is lower
for higher frequencies. The minimum attenuation naturally occurs for an infinitely large
waveguide, but detector apertures tend to be quite small and one needs to balance these
two requirements. Conical horns at the end of the waveguide is one way of balancing
these needs. Waveguides at which the intended wavelength is small enough to allow
”overmoding”, where many additional modes of propagation can occur, are called over-
sized waveguides. These give a low attenuation, but less mode purity. Such a waveguide
was used in this project. It is oversized since it has an internal diameter of 26 mm while
the longest wavelength used is 1 mm. In our case only the total transmitted energy is
measured, the Golay cell treats all modes of propagation equally, so overmoding is not
a big problem.

13



2.3. MOLECULAR SPECTROSCOPY
CHAPTER 2. THEORY

2.3 Molecular spectroscopy

When EM radiation passes through a sample, its rate of transmission will depend on
various properties of the sample. The Beer-Lambert law formulates this as:

T = T0e
−Al (2.2)

where l is the length of the sample along the beam and A is an absorption coefficient, de-
pending of a number of parameters. Since these parameters are specific to the substances
in a sample, as well as to some other conditions such as pressure and temperature, one
can infer facts about the sample by studying its absorption spectrum. Calculating the
absorption in relation to the pressure, temperature and composition is the subject of the
following sections.

2.3.1 Energy levels

In an atom there are only electronic energy levels, which usually have optical transitions.
Some fine structure transitions are in the terahertz, though. When forming a molecule,
valence electrons of the constituent atoms will inhabit molecular orbitals instead of
the atomic orbitals, with somewhat different energies. But there are also new degrees
of freedom present. Molecules do not have rotational symmetry and will thus have
rotational energy. The rotation must be treated quantum mechanically by rewriting the
expression for rotational energy with quantum mechanical operators, and this gives rise
to quantized rotational states. This will be demonstrated for CO below. Each molecular
bond will in addition introduce vibrational modes, giving a spectrum of vibrational
states. The number of vibrational modes depends on the molecule. A diatomic molecule
such as CO has only one vibrational mode, while water is more complicated and has
three vibrational modes. A sketch of the various modes of the two molecules is given in
figure 2.4.

To calculate electronic energy levels in molecules one uses the Born-Oppenheimer
approximation: nuclear and electronic motion are assumed to be completely separate.
This is justified since electron mass is about two thousandths of the mass of a nucleon,
thus moving much quicker. In effect, electron wavefunctions depend only on the position
of the nuclei and the motion of the nuclei depends only on the time-averaged distribution
of electrons. Another approximation that might be used is to consider the molecule a
rigid rotor when calculating rotational energy levels. Without this assumption there is
some vibration-rotation coupling. This is however true to a significant degree at higher
energy levels and then the rigid rotor approximation has to be abandoned. Vibration
can be modeled as a harmonic oscillator for low energy levels, but at high energy levels
it must be replaced by an anharmonic oscillator that couples to the rotation.

While rotation-vibration in diatomic molecules can be handled analytically with ease
the difficulty increases enormously for even triatomic molecules such as water. Theses
are written on this subject, for instance Lori [2008]. A basic description of the vibrational
levels would be a linear superposition of single-mode vibrations, but calculating exact
levels requires handling nonlinear effects. We instead describe the diatomic molecule.

14



2.3. MOLECULAR SPECTROSCOPY
CHAPTER 2. THEORY

Figure 2.4: vibrational modes and rotational axes of CO and H2O

The vibration-rotation spectrum of CO

Vibration of CO is modeled as a harmonic oscillator for low energy levels (McQuarrie

[2008], chapter 5), so the energy levels are of the form Evib = (ν + 1
2)h̄ω where ω =

√
k
µ

is determined empirically. µ is the reduced mass and k is the bond force constant. These
lines generally are in the infrared. For CO, ν = 1→ 2 is at 269 meV or 65 THz.

Rotation of CO is modeled as a rigid rotor with length re, the interatomic distance
at equilibrium. Its moment of inertia, with respect to a line normal to the bond and
intersecting it at the center of mass, is then given by I = µr2

e where µ is the reduced
mass of the two-body system. We can insert this in the Schrödinger equation expressed
in terms of the angular momentum operator J :

J2

2I
ψ = Erotψ (2.3)

with eigenvalues Erot = h̄2

2I j(j+1). In the case of CO we do not have to worry about the
other rotational axes, one is identical and the other is a symmetry axis. Inserting the
parameters for CO (from Hush and Williams [1974]) gives Erot = 0.24j(j+1)meV . This

15



2.3. MOLECULAR SPECTROSCOPY
CHAPTER 2. THEORY

Figure 2.5: Sketch of typical rovibrational spectrum for a diatomic molecule with a vibra-
tional transition at 1000, and rotational constant of 1/2, in arbitrary units.

is a multiple of about 58 GHz, starting in the microwave region. It is not surprising that
rotational lines are also present in the neighboring THz region. The energy of the tran-
sitions has a clear hierarchy, electronic>vibrational>rotational. Rotiational transitions
are on the order of 1-10 cm−1 compared to about 1000 cm−1 for vibrational transitions.
For each vibrational level one can thus superimpose rotational transitions and get a
spectrum such as that sketched in figure 2.5. Selection rule ∆J 6= 0 suppresses the pure
vibrational transition, which is surrounded by two branches of rotational transitions,
the lower energy P-branch for ∆J = −1 transitions and the higher energy R-branch of
∆J = 1.

Similar principles apply for the water molecule, but the calculations are very involved.
Both CO and H2O have rotational lines in the region 300-500 GHz, where we have
measured. See figure 3.15 for the rotational lines of CO. They can be easily understood,
while the H2O lines in figure 3.14 are more complex. In practical applications one uses
semiempirical rather than calculated spectra.

2.3.2 Line shape

The line shape is affected by pressure- and temperature induced broadening, tempera-
ture dependences in line strength, and line mixing. There are three main types of line
broadening (Liou [2002], chapter 1):

• Natural broadening: due to the time-energy uncertainty principle.

• Doppler broadening: due to doppler shifts of the radiation from sources moving in
all directions.

• Pressure broadening: due to several pressure-dependent effects including collisions
(shortening excited state lifetimes) and perturbations of the energy levels of atoms
by the electric fields of nearby atoms.

We briefly describe these types to show which are important in our case.

16



2.3. MOLECULAR SPECTROSCOPY
CHAPTER 2. THEORY

Natural broadening

This type of broadening is the most fundamental, existing due to the Heisenberg uncer-
tainty principle:

∆E∆t ≥ h/2π (2.4)

The lifetime of a state is thus inversely proportional to its uncertainty in energy. Unless
states in the gas are extremely short-lived, this is completely negligible.

Doppler broadening

The velocity of molecules in a gas follow a Maxwell distribution:

p(v) =
1√
πv0

ev
2/v0 (2.5)

where v0 is the mean speed given by v0 =
√

2kBT/m. This gives rise to a Gaussian
lineshape:

fD(ν − ν0) =
1

αD
√
π
exp
(
− (ν − ν0)2

α2
D

)
(2.6)

where αD = ν0

√
2kBT
mc2

. This type of broadening is significant at very high temperatures

or at low pressures, and the temperatures in our case might be high enough.

Collisional broadening

Collisional broadening follows a Lorentz line shape:

f(ν − ν0) =
α/π

(ν − ν0)2 + α2
(2.7)

where

α = α0

( p
p0

)(T0

T

)n
(2.8)

Where α0 is the halfwidth at reference temperature T0. The half-width is composed of
the self- and foreign-broadened halfwidths, that are added with weighting in accordance
with the volume mixing ratio (VMR) of the gases present. This type of broadening
is very significant in atmospheric pressures and temperatures, and in the temperature
range we are to operate in.

Combined broadening

If more than one type of broadening is relevant the lineshape functions can be convoluted
to get a realistic shape. When convoluting the Lorentz and Gaussian functions, one gets
the Voigt function, a function for which there is no closed form expression. This could
possibly improve our modelling algorithm.

17



2.3. MOLECULAR SPECTROSCOPY
CHAPTER 2. THEORY

Temperature dependence of line strength

The line strength parameter Si(T ) is necessary to determine the shape of the spectrum.
It has a temperature dependence given by:

S(T ) = S(T0)
e−hcEL/kBTQT (T0)[1− e−hcν0/kBT ]

e−hcEL/kBT0QT (T )[1− e−hcν0/kBT0 ]
(2.9)

where T0 is a reference temperature and T is the temperature considered. QT is the
thermodynamic partition function and is specific to each chemical species. EL is the
energy of the lower state in the transition and ν0 is the frequency of the line center.

Line-by-line models

To calculate the transmission profile one must sum the contribution of all nearby lines,
one at a time. We describe how to do this in the most straightforward way, although
highly optimized schemes exist for saving computational power. The direct approach
is to calculate the contribution of each spectral line at each frequency and sum over
all spectral lines in the vicinity of the spectral range of interest. The transmission at
frequency ν can be written as

T (ν) =
∑
i

[
Si(T )f(ν − νi;αi)l

]
(2.10)

there T is total transmission, Si is line strength of line i, f is a shape function (Lorentz
under atmospheric and similar condition), α is the Lorentz half-width and l is the length
of the radiations passage through the medium. The α is itself affected by pressure,
temperature and VMR of gases as described above.

2.3.3 Collisional line mixing

An effect that is significant in IR gas spectroscopy is pressure-induced collisional mixing
of lines. Adjacent lines will interfere with each other to a significant degree, which
will create line profiles the differ significantly from the ones calculated by a line-by-line
Lorentz-profile model. To see why this is so, we refer to a toy example by Hartmann
et al. [2008], chapter four, pp. 148. The example assumes we have a pair of transitions
i → f and i′ → f ′, both in the optical range. If the transitions i → i′ or f → f ′

are very small, they might be triggered by collision, thus introducing processes such
as i → i′ → f ′ that mix the two optical lines. An illustration is shown in figure 2.6.
While we obviously don’t get mixing between optical states of different molecules, the
lineshapes of a molecule will still be affected by molecules of other species as well as
its own kind. This is due to the low energy transitions triggered in the collisions, that
transfer energy from one molecule to another in a way that is specific for each pair of
molecules. To calculate the effects of line mixing one thus needs calculated or empirical
mixing coefficients for each combination of gases. There is a large number of models for
handling line mixing with varying precision and computational complexity, from simple
classical models to full quantum scattering models.

18



2.3. MOLECULAR SPECTROSCOPY
CHAPTER 2. THEORY

Collisional transition

Collisional transition

Optical transitions

E

Figure 2.6: The type of transitions giving rise to line mixing.

2.3.4 The HITRAN database

The HITRAN (high-resolution transmission molecular absorption) database contains
empirical data on a large number of spectral lines for a number of molecules, including
CO and H2O. It contains intensities and positions as well as lineshape parameters over
a wide range of frequencies. The foreign-broadened halfwidths are given with respect to
air. The database contents as of 2008 are described by Rothman et al. [2009].

2.3.5 LINEPAK vs custom algorithm

Since the effects described above are quite complex to simulate, much work has been
done by researchers on optimizing algorithms for this task. In this project we used the
webservice SpectralCalc, provided by GATS, Inc and powered by the LINEPAK library
of GATS founder L. Gordley, described in Gordley et al. [1994]. The simulation takes all
of the effects mentioned above into account and automatically collects needed data from
HITRAN. It proved less tedious to extrapolate from a dataset obtained using Spectral-
Calc rather than perfecting a custom algorithm for spectral simulation. The primary
difficulty in simulating spectra in the range considered in this project is the complexity of
line mixing, which is very significant for H2O. The line-by-line Lorentz-profile algorithm
is included in appendix A. The dataset downloaded from Spectralcalc.com is described
in appendix C.

19



3
The setup

3.1 Gas cell design

The gas cell was a sealed steel tube (with inner diamter 26 mm and outer diameter
33 mm) with dual windows at both ends. The compartment for the sample gas was
1 m long, and in total the tube was 1.8 m long. The inner windows were made of
crystalline quartz, 5mm thick, and the outer ones were made of teflon, 2mm thick. Both
window materials are mostly transparent in the frequency range used, the transmission
as function of frequency is plotted in figure 3.6. The windows where inserted at a 60o

angle to the beam to minimize standing waves and other unwanted reflections. The gas
cell was fully enclosed, except for the ends, by an oven (Entech ESTF 50/11-11) to keep
the temperature stable. A sketch of the cell is shown in figure 3.1 and its connections to
the other system components is shown in figure 3.2. A photo of the gas cell is shown in
figure 3.5.

In the laboratory setting, nitrogen gas was injected at a rate of 5L/min using a
Bronkhorst EL-FLOW mass flow controller, and pre-heated using heating hoses from
Hillesheim, model H3406-030-08-250C-83 and temperature controllers HT-40. Steam
was injected into the nitrogen stream before entering the gas cell, using a steam genera-
tor from Cellkraft (E-1000 precision evaporator). The gas flowed continuously through
the gas cell and exited at the opposite end into the open air. Thus a temperature gra-
dient formed when equilibrium was reached, and this was approximately measured with
thermocouples at inlet and outlet. Figure 3.4 shows a photo of the setup. The volume
between the windows was continously flushed with N2, with mass flow controllers set to
1 L/min, to avoid contamination with H2O or other gases.

In the gasifier setting, the input rawgas came directly from the gasifier, and exited
into the gas cleaning apparatus of the facility. Fluegas was also measured, directly from
the combustor. Several times the input gas was also switched to bottle gas for testing
of CO detection. Figure 3.3 shows a photo of the setup.

20



3.1. GAS CELL DESIGN
CHAPTER 3. THE SETUP

Gas inlet
Gas outlet 

+ thermocouple

insertion

N2 N2

Heating zone 1

Heating zone 1 Heating zone 2

Heating zone 2

Gas pre-heating

Thermocouple

insertion

1.0m

1.8 m

33mm

26mm
W1 W1W2 W2

W1: Te!on windows

W2: Quartz windows

Figure 3.1: A sketch of the gas cell

Water

Humidity

meter

Heating hoses 200 C

Steam generator

Cellkraft E-1000

Precision Evaporator

Into 

open air

Mass !ow

controllers

Bronkhorst

EL-FLOW

1 L/min

1 L/min

5 L/min

N2 wall outlet

 7 atm

Digital scale 

Ohaus Scout Pro

Gas inlet

Gas

outlet 
N2 N2

Gas pre-heating

Oven Entech

ESTF 50/11-11

Thermocouple

Thermocouple

Hillesheim H3406-030-08-250C-83

Figure 3.2: The gas cell and its connections to the rest of the system.

21



3.1. GAS CELL DESIGN
CHAPTER 3. THE SETUP

Figure 3.3: The gas cell set up at the gasifier

Figure 3.4: The gas cell in the lab within the closed oven, tubing and electronics.

22



3.1. GAS CELL DESIGN
CHAPTER 3. THE SETUP

Figure 3.5: The gas cell inside the open oven

23



3.1. GAS CELL DESIGN
CHAPTER 3. THE SETUP

3.1.1 Window properties

The transmission through the windows of the gas cell was measured. The two window
types were quartz (5 mm) and teflon (2 mm), and their transmissivity as function of
frequency is given in figure 3.6. The refractive index is obtained by the equation (Hecht
[2001], chapter 9.6).

λ

n
=

2L

m
(3.1)

where m is the number of standing wave nodes between the window surfaces and has
to be an integer. When it is, we will have a peak in the transmission due to a standing
wave. We could set m = 1 and let ν = c/λ be the frequency difference between two of
our observed peaks in the transmission. Then

ν =
c

2Ln
(3.2)

in our cases we can se in figure 3.6 that ν is 16 GHz for quartz, corresponding to n = 1.88
and for teflon ν is about 7 GHz corresponding to n = 1.43.

24



3.1. GAS CELL DESIGN
CHAPTER 3. THE SETUP

0

0.
2

0.
4

0.
6

0.
81

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

Te
flo

n

0

0.
2

0.
4

0.
6

0.
81

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

Q
ua
rt
z

F
ig
u
re

3
.6
:

T
ra

n
sm

is
si

on
th

ro
u
gh

te
fl
on

(t
o
p
)

a
n
d

q
u
a
rt

z(
b

o
tt

o
m

)
w

in
d
ow

s
a
t

n
o
rm

a
l

in
ci

d
en

ce
re

la
ti

ve
to

a
ir

.
[d

at
a/

w
in

d
ow

p
ro

p
er

ti
es

/]

25



3.2. SPECTROMETER
CHAPTER 3. THE SETUP

3.2 Spectrometer

The spectrometer setup consisted of a source, a detector and a reference detector as
shown in figure 3.7. All instruments were controlled from a PC with LabVIEW using
GPIB (IEEE 488).

Figure 3.7: The setup of the instruments around the gas cell

3.2.1 Source

Since we required a tunable source in the range 300-500 GHz, we used first a tunable mi-
crowave source over a lower frequency range, then an amplifier/multiplier chain (AMC)
multiplying the frequency in several steps up to the THz range. A block diagram of
the devices is shown in figure 3.9. The microwave source was a Virginia Diodes Tx 195,
driving a Virginia Diodes WR2.2AMC with input frequency range 9.0-13.9 GHz and
output frequency range 325-500 GHz. The source is shown in figure 3.12 and the AMC
in figure 3.13. We used the TTL output on the lock-in amplifier for amplitude modula-
tion of the source, set to a frequency of 18 Hz. The frequency of the output signal from
of the microwave source was controlled by a DC control voltage, for which we used the
auxiliary outputs of the lock-in amplifier. The detected signal, an 18 Hz AC voltage,
was then read out using the same lock-in amplifier. The correspondence between DC
control voltage Uctrl and output frequency fout [GHz] was given by

Uctrl = (fout/36− 8.0074)/1.1998 (3.3)

26



3.2. SPECTROMETER
CHAPTER 3. THE SETUP

Where the upper limit on Uctrl was 5 V. Since there is a large power variation in the
frequency setting(see figure 3.8), we used frequency modulation to have some built-in
signal averaging. FM was controlled by an AC output on the lock-in set to a frequency
of 333 Hz. The amplitude of the FM control signal was set to 5 V which gave rise to
a slightly nonlinear frequency swing in the final output signal, given for two different
control voltages in table 3.1. The output power of the source specified by VDI is shown
in figure 3.10.

Table 3.1: Amplitudes of frequency swings due to modulation with various amplitudes.
∆fin is the maximum deviation from the mean frequency at any particular value of the
control voltage Uctrl, and is due to the value of the FM voltage UFM. ∆fout is the output
frequency of the AMC after multiplication. The relation between FM modulation voltage
and output frequency swing varies slightly across the operating range of our source, as can
be seen in the difference in output from these two control voltages when all other parameters
are kept constant.

Uctrl[V] UFM[V] ∆fin[V] ∆fout[GHz] Uctrl[V] UFM[V] ∆fin[V] ∆fout[GHz]

2.2 1 0.004 0.144 4 4.8 1 0.005 0.18

2 0.008 0.288 2 0.009 0.324

3 0.012 0.432 3 0.013 0.468

4 0.016 0.576 4 0.017 0.612

5 0.02 0.72 5 0.021 0.756

-1 -0.005 -0.18 -1 -0.004 -0.144

-2 -0.008 -0.288 -2 -0.008 -0.288

-5 -0.021 -0.756 -5 -0.021 -0.756

0

5

10

15

20

25

30

0.272 0.772 1.272 1.772 2.272 2.772 3.272 3.772 4.272 4.772

D
et
ec
te
d 
si
gn
al
 [m

V]

Control voltage [V]300 GHz 500 GHz

Figure 3.8: A nitrogen spectrum in terms of control voltage vs output voltage from the
Golay cell. The voltage span corresponds to a full frequency sweep over 300-500 GHz. Note
the very high sensitivity to small variations in the contol voltage. This sensitivity motivated
the use of frequency modulation. [data/ctrl-to-signal]

27



3.2. SPECTROMETER
CHAPTER 3. THE SETUP

Figure 3.9: Sketch of the setup of the source, from VDI manual.

Figure 3.10: Source power output, from VDI manual.

28



3.2. SPECTROMETER
CHAPTER 3. THE SETUP

3.2.2 Detector: Golay cell

The detectors were Golay cells with beam collimators, from Tydex. The Golay cell
is a detector for IR, also usable in the THz range. Its principle of operation was first
described by Golay [1947], hence its name. The device contains a material absorbing EM
radiation within the relevant frequency range, producing heat. The absorber is placed
in a sealed compartment of gas, sealed at one end by a transparent window and at the
other by a flexible membrane. A mirror is attached to the outside of the membrane,
and a photocell observes it. When the absorber heats up, heat is transferred to the gas,
which expands. The flexible membrane moves in response to the pressure change, and
the photocell detects this by measuning the change in reflected light from the membrane.
The signal of the photocell is the output signal. Due to its thermal-mechanical nature
the Golay cell is sensitive to mechanical vibrations as well as thermal noise and has a
modest response time, on the order of tens of milliseconds (Rogalski [2010], pp 157-158)
and a modulation frequency of about 20 Hz. It is however cheap and robust. A sketch
is shown in figure 3.11.

Figure 3.11: Sketch of the Golay cell

29



3.2. SPECTROMETER
CHAPTER 3. THE SETUP

Figure 3.12: Left: Golay cell. Right: The microwave source

Figure 3.13: The Amplifier multiplier chain, concealed within a black box.

30



3.3. HUMIDITY METER
CHAPTER 3. THE SETUP

Table 3.2: Coefficients for the Antoine equation for H2O, from DDBST [2013].

T range [C] A B C

1-100 8.07131 1730.63 233.426

99-374 8.14019 1810.94 244.485

3.3 Humidity meter

To get a reference value for the humidity, a humidity meter was needed. The device
chosen for this purpose was made in-house at the department of energy and the envi-
ronment, and is described by Hermansson et al. [2011]. It could only be used in the
lab environment where there was no tar in the gas, or for fluegas. The device monitors
both temperature and relative humidity (RH), although the temperature readings are
of little direct use in our case since the gas must be cooled before entering the device.
They are however necessary to convert relative humidity to VMR (Volume mixing ra-
tio, the ratio of the gas by volume). The RH measurement is done by use of chemically
modified polymers that change capacitance depending on RH. The detector materials
work at temperatures up to about 180o C. To avoid condensation in the device it is
kept at about 110o C. Relative humidity is temperature dependent so VMR is extracted
by multiplying the measured RH with the factor Psat/P , where P is total pressure in
the gas and Psat saturation pressure. P is 750 mmHg and Psat is given by the Antoine
equation (Reid et al. [1987], chapter 7.3.1):

Psat/P = 10(8.14019−(1810.94/(244.485+T )))/750 (3.4)

where pressure is in mmHg and temperature in Celsius. The Antoine equation is derived
from the Clausius-Clapeyron equation and has two sets of coefficients, one below boiling
point and one above so that RH can be defined at any temperature. The general form
is:

log10Psat = A− B

C − T
(3.5)

with coefficients for H2O in the relevant units given in table 3.2.

31



3.4. SPECTRUM ANALYSIS
CHAPTER 3. THE SETUP

3.4 Spectrum analysis

3.4.1 THz absorption by H2O vapors

We have described the theory of spectral lines in section 2.3.2. We assumed a Lorentzian
lineshape as a first guess, but at the high temperaure it might be necessary to use the
Voigt function to get better results. Line parameters for the spectral range 0− 40cm−1

were extracted from HITRAN and their contributions to the spectrum summed, since
they might affect the transmission within in our spectral range. The thermodynamic
partition function of water was obtained using piecewise polynomial interpolation of
the values given by Vidler and Tennyson [2000]. MATLABs nlinfit was used to fit
the parameters T and VMR in a least squares sense to the experimental data. The
MATLAB function for the theoretical transmission is included as appendix A.

An alternative approach also pursued was to download a dataset from spectral-
calc.com, which was also done. This data included the effects of line mixing and would
be used to determine parameters via least-squares fitting.

According to HITRAN, water has 505 lines in the range 0-40 cm−1. Most of these
are very weak and can be disregarded. Values for the 48 strongest lines are provided in
appendix B. We use only the two strongest lines at 383 and 448 GHz (12.68 and 14.44
cm−1), but other lines contribute to the background so some of them must be considered.
Figure 3.14 shows the H2O lines in the intervals 10-16.7 cm−1 and 0-40 cm−1 with and
without a cutoff intensity set at e−23.

3.4.2 THz absorption by Carbon Monoxide

According to HITRAN, CO has 103 lines in 0-40 cm−1. As for water, most are very
weak and can be disregarded. Values for the 11 strongest lines are provided in appendix
B. The line at 460 GHz is clearly visible and distinguishable from the nearby water lines
and is the primary candidate for determining levels of CO. It is further commented in
sections 4 and 5. Figure 3.15 shows the CO lines in the intervals 10-16.7 cm−1 and 0-40
cm−1 with and without a cutoff intensity set at e−23.

32



3.4. SPECTRUM ANALYSIS
CHAPTER 3. THE SETUP

F
ig
u
re

3
.1
4
:

L
in

e
p

os
it

io
n

s
an

d
in

te
n

si
ti

es
fo

r
H

2
O

.
U

p
p

er
le

ft
:

0
-4

0
cm

−
1

w
it

h
cu

to
ff

a
t

in
te

n
si

ty
<
e−

2
3
.

U
p
p

er
ri

g
h
t:

0
-4

0
cm

−
1

w
it

h
n
o

cu
to

ff
.

L
ow

er
le

ft
:

10
-1

6.
7

cm
−
1

w
it

h
cu

to
ff

a
t

in
te

n
si

ty
<
e−

2
3

L
ow

er
ri

g
h
t:

1
0
-1

6
.7

cm
−
1

w
it

h
n
o

cu
to

ff
.

C
o
m

p
a
re

fi
gu

re
2.

3.

33



3.4. SPECTRUM ANALYSIS
CHAPTER 3. THE SETUP

F
ig
u
re

3
.1
5
:

L
in

e
p

os
it

io
n

s
an

d
in

te
n
si

ti
es

fo
r
C
O

.
U

p
p

er
le

ft
:

0
-4

0
cm

−
1

w
it

h
cu

to
ff

a
t

in
te

n
si

ty
<
e−

2
3
.

U
p

p
er

ri
g
h
t:

0
-4

0
cm

−
1

w
it

h
n
o

cu
to

ff
.

L
ow

er
le

ft
:

10
-1

6.
7

cm
−
1

w
it

h
cu

to
ff

a
t

in
te

n
si

ty
<
e−

2
3

L
ow

er
ri

g
h
t:

1
0
-1

6
.7

cm
−
1

w
it

h
n
o

cu
to

ff
.

C
o
m

p
a
re

fi
g
u
re

2.
3.

34



4
Experimental methodology

4.1 Calibration

For obtaining a transmission spectrum, the backgrund transmission needs to be deter-
mined, then the transmission through the sample can be isolated from all other phe-
nomena affecting absolute transmission. In figure 4.1 we show an example of how one
calibration spectra and a raw measurement are used to produce the expected H2O spec-
trum. In this case it is the spectrum of rawgas, where the H2O-lines are the most
prominent features.

Another type of measurement that was done is monitoring of specific frequencies
over extended periods of time. In those cases, calibration was done by monitoring a
pure N2-spectrum for some time before the sample was let in, then dividing subsequent
measurements with the average N2 transmission at each frequency.

35



4.1. CALIBRATION
CHAPTER 4. EXPERIMENTAL METHODOLOGY

02040608010
0

12
0

14
0

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Signal strength [mV]

Fr
eq

ue
nc
y 
[G
H
z]

N
2

Ra
w
ga
s

0.
6

0.
650.
7

0.
750.
8

0.
850.
9

0.
951

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Signal strength (ratio)

Fr
eq

ue
nc
y 
[G
H
z]

N
or
m
al
ize

d
ra
w
ga
s

S

S

38
3

44
8

F
ig
u
re

4
.1
:

T
op

:
T

w
o

ra
w

m
ea

su
re

m
en

ts
,

o
n
e

o
f
N

2
w

h
ic

h
la

ck
s

sp
ec

tr
a
l

li
n

es
in

th
is

fr
eq

u
en

cy
ra

n
g
e

to
b

e
u
se

d
fo

r
ca

li
b

ra
ti

o
n

,
an

d
on

e
of

ga
si

fi
er

ra
w

ga
s.

B
ot

to
m

:
T

h
e

tr
a
n

sm
is

si
o
n

sp
ec

tr
u

m
fo

rm
ed

b
y

d
iv

id
in

g
th

e
ra

w
g
a
ss

p
ec

tr
a

b
y

th
e

ca
li
b

ra
ti

o
n

sp
ec

tr
a
.

V
M
R
≈

65
%

an
d
T
≈

43
0
o
C

.
[d

at
a/

ra
w

sp
ec

tr
a]

36



4.2. LABORATORY H2O MEASUREMENTS
CHAPTER 4. EXPERIMENTAL METHODOLOGY

4.2 Laboratory H2O measurements

Two ways of measuring H2O were employed. A sweep of the entire spectrum, that can
be used by fitting spectralcalc simulations to extract temperature and VMR. The other
method was continous monitoring of a chosen set of frequencies at the line centers that
should be enough to do the determination of VMR and T . These measurements where
performed in parallell with humidity meter measurements using the device described
in section 3.3. Measurements were made at 383 and 448 GHz, one at a time and also
switching between the two lines for measuring them simultaneously. The transmissions
and humidity measurements are summarized for several measurements of this type in
table 4.1. One measurement of two lines simultaneously is shown in figure 4.2, this is
the data called ”1” in table 4.1. There was also a long timeseries taken where humidity
was kept at a few closely spaced levels to determine if they could be differentiated from
each other. The result is plotted in figure 4.3.

Table 4.1: Transmissions obtained at different humidity levels in two separate measure-
ments of the type shown in figure 4.2. Errors are shown as one standard deviation. Mea-
surement number indicates which of the two separate dataseries the entry comes from.

Measurement Transmission Transmission Humidity meter

number at 448 GHz at 383 GHz VMR [%]

1 0.684± 0.00870 0.784± 0.00428 95± 1.52

1 0.721± 0.00921 0.8± 0.00519 73± 0.99

1 0.751± 0.00836 0.813± 0.00502 57± 1.10

1 0.774± 0.00838 0.822± 0.00527 48± 1.10

1 0.781± 0.00830 0.825± 0.00509 41± 0.77

2 0.786± 0.00692 0.824± 0.00366 49± 1.19

2 0.76± 0.00979 0.819± 0.00259 57± 0.84

2 0.73± 0.00723 0.792± 0.00261 73± 0.90

37



4.2. LABORATORY H2O MEASUREMENTS
CHAPTER 4. EXPERIMENTAL METHODOLOGY

0.
650.
7

0.
750.
8

0.
850.
9

0.
951

0
50
0

10
00

15
00

20
00

25
00

30
00

Transmission

Ti
m
e 
[s
]

38
3 
G
Hz

44
8 
G
Hz

10
0 
pe

r. 
M
ov
. A

vg
. (
38

3 
G
H
z)

10
0 
pe

r. 
M
ov
. A

vg
. (
44
8 
G
H
z)

02040608010
0

0
50

0
10

00
15

00
20

00
25

00
30
00

VMR [%]

Ti
m
e 
[s
]

VM
R 
fr
om

 h
um

id
ity

 m
et
er

F
ig
u
re

4
.2
:

T
op

:
th

e
tw

o
w

at
er

li
n
es

ar
e

p
lo

tt
ed

a
s

fu
n
ct

io
n

s
o
f

ti
m

e,
w

h
er

e
th

e
H

2
O

co
n
te

n
t

is
ch

a
n

g
ed

in
st

ep
s.

B
o
tt

o
m

:
th

e
h
u

m
id

it
y

m
et

er
gi

ve
s

th
e

co
rr

es
p

on
d
in

g
V
M
R

.
[d

a
ta

/
tw

o
li
n

es
]

38



4.2. LABORATORY H2O MEASUREMENTS
CHAPTER 4. EXPERIMENTAL METHODOLOGY

23
20
0

23
40
0

23
60
0

23
80
0

24
00
0

24
20
0

24
40
0

0
50

0
10

00
15

00
20

00
25

00
30

00

Signal (power units)

Ti
m
e 
(s
ec
)

44
9 
G
H
z w

at
er
 v
ap

or
 li
ne

  (
20
13

02
04
)

VM
R0

.1
03

VM
R0

.1
13

VM
R0

.1
55

VM
R0

.1
59

VM
R0

.1
71

F
ig
u
re

4
.3
:

L
ab

or
at

or
y

m
ea

su
re

m
en

t
of

tr
an

sm
is

si
o
n

a
t

4
4
9

G
H

z
ov

er
ti

m
e

fo
r

fi
v
e

h
u
m

id
it

y
le

ve
ls

.
T

h
e

im
p
u
ls

e-
li
ke

o
b

je
ct

s
a
re

p
ro

d
u

ct
s

of
th

e
h
u

m
id

ifi
er

,
it

se
em

s
to

re
le

a
se

st
ea

m
in

sh
o
rt

p
u
ls

es
.

N
o
te

th
a
t

m
ea

su
re

m
en

ts
w

it
h

a
d

iff
er

en
ce

in
V
M
R

a
s

sm
a
ll

as
0.

4%
ca

n
b

e
cl

ea
rl

y
d
is

ti
n
gu

is
h
ed

fr
om

ea
ch

o
th

er
.

[/
d
a
ta

/
2
0
1
3
0
2
0
4
]

39



4.3. MONITORING OF H2O AND CO AT THE GASIFIER
CHAPTER 4. EXPERIMENTAL METHODOLOGY

4.3 Monitoring of H2O and CO at the gasifier

Several spectra were taken where bottled CO and H2O were introduced into the gas cell.
H2O and CO+H2O is displayed in the figure 4.6. These give an indication of what one
could expect to see in the rawgas, whose main features should be H2O- and CO peaks.

A few spectra of the rawgas and fluegas were taken and are displayed in figures 4.4
and 4.5. The rawgas was kept at an underpressure of 0.8 atmospheres to prevent toxic
leaks. This has some effect on the shape of lines, which must be accounted for when
simulating the spectrum. The transmission at 448 GHz was monitored over time while
changing the humidity just as in the laboratory case, the result is shown in figure 4.7.
At the gasifier humidity is regulated by changing the input of steam to the process.
Note the relatively fast oscillations in humidity level. This cannot be detected using the
presently used, slower methods.

40



4.3. MONITORING OF H2O AND CO AT THE GASIFIER
CHAPTER 4. EXPERIMENTAL METHODOLOGY

0.
6

0.
650.
7

0.
750.
8

0.
850.
9

0.
951

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

Ra
w
ga
s

Fl
ue

ga
s

H
2O

65
%
 +

CO
23
%

sim
ul
at
io
n

65
 %
 H
2O

sim
ul
at
io
n

F
ig
u
re

4
.4
:

T
h
e

sp
ec

tr
u
m

of
th

e
ra

w
ga

s,
3
00

-5
0
0

G
H

z.
N

o
te

th
e

C
O

li
n
e

a
t

4
6
0

G
H

z.
A

ls
o

in
cl

u
d

ed
is

a
fl
u

eg
a
s

m
ea

su
re

m
en

t
a
t

41
0-

47
0

G
H

z.
S
p

ec
tr

al
ca

lc
si

m
u

la
ti

on
s

fo
r

th
e

ra
w

g
a
s

a
re

a
ls

o
in

cl
u
d
ed

,
co

m
p
a
re

ta
b

le
2
.1

o
f

ra
w

g
a
s

co
n
te

n
ts

.
[d

a
ta

/
ra

w
g
a
sp

lo
t]

0.
6

0.
650.
7

0.
750.
8

0.
850.
9

0.
951

41
0

42
0

43
0

44
0

45
0

46
0

47
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

Ra
w
ga
s

Fl
ue

ga
s

H2
O
65
%
 +

CO
23

%
(r
aw

ga
s)

sim
ul
at
io
n

65
 %
 H
2O

(r
aw

ga
s)

sim
ul
at
io
n

F
ig
u
re

4
.5
:

R
aw

an
d

fl
u
eg

as
fr

om
th

e
sa

m
e

m
ea

su
re

m
en

t
a
s

a
b

ov
e,

zo
o
m

ed
in

o
n

th
e
H

2
O

-l
in

e
a
t

4
4
8

G
H

z.
[d

a
ta

/
ra

w
g
a
sp

lo
t]

41



4.3. MONITORING OF H2O AND CO AT THE GASIFIER
CHAPTER 4. EXPERIMENTAL METHODOLOGY

0.
6

0.
650.
7

0.
750.
8

0.
850.
9

0.
951

44
0

44
5

45
0

45
5

46
0

46
5

47
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

H2
O
 (4

1%
)

H2
O
 (4

1%
) +

 C
O

(2
0%

)

H2
O
 (4

1%
) ‐

sp
ec
tr
al
ca
lc

H2
O
 (4

1%
) +

 C
O

(2
0%

) ‐
 sp

ec
tr
al
ca
lc

F
ig
u
re

4
.6
:
H

2
O

(4
1
%

)
co

m
p

a
re

d
to
H

2
O

(4
1
%

)
a
n
d
C
O

(2
0
%

).
[/

d
a
ta

/
co

a
n

d
h

2
o
]

42



4.3. MONITORING OF H2O AND CO AT THE GASIFIER
CHAPTER 4. EXPERIMENTAL METHODOLOGY

27
00
0

27
50
0

28
00
0

28
50
0

29
00
0

29
50
0

30
00
0

0
50

0
10
00

15
00

20
00

25
00

30
00

35
00

40
00

45
00

Transmission, 448 GHz

Ti
m
e 
[s
]

Tr
an
sm

iss
io
n

Av
er
ag
ed

25
0 
pe

r. 
M
ov
. A

vg
. (
Tr
an
sm

iss
io
n)

16
0 
kg
/h

20
0 
kg
/h

24
0 
kg
/h

16
0 
kg
/h

12
0 
kg
/h

F
ig
u
re

4
.7
:

M
ea

su
re

m
en

t
ov

er
ti

m
e

of
th

e
4
48

G
H

z
tr

a
n

sm
is

si
o
n

a
t

th
e

g
a
si

fi
er

.
V

a
ry

in
g

th
e

st
ea

m
in

p
u
t

o
f

th
e

p
ro

ce
ss

:
1
6
0
,

2
0
0
,

24
0,

16
0,

12
0

k
g/

h
.

[d
at

a/
ga

si
fi
er

-h
u
m

id
it

y
le

ve
ls

]

43



4.4. NOISE AND DRIFTS
CHAPTER 4. EXPERIMENTAL METHODOLOGY

4.4 Noise and drifts

The stability of the source over time was tested at room temperature, and at high
temperature, to determine how heat leaking from the gas cell affects it. Significant drift,
seemingly linear in time, was found at the higher temperature. More isolation was then
achieved by moving the source further (10 cm) from the gas cell. The measurements were
conducted by periodic sweeps of the entire frequency range, and without the reference
detector. Figure 4.8,4.9 and 4.10 show the results. Noise in the lab environment was
mostly below 2% and often below 1% of total signal strength at an integration time of
100 ms, and apparently consisted of white noise.

Figure 4.11 shows a representative plot of noise over the entire frequency range, in the
lab. The noise level at the gasifier was about 2% of signal strength at integration time
100 ms, as shown in figure 4.12. At the ends of the operating range of the source, signal
strength is close to zero and thus noise dominates. This can be clearly seen in the graphs.
Numerous long duration scans at a single frequency were conducted for determining the
precision and required integration time for the vapor measurement, both at gasifier and
lab.

44



4.4. NOISE AND DRIFTS
CHAPTER 4. EXPERIMENTAL METHODOLOGY

0.
98

0.
98

5

0.
99

0.
99

51

1.
00

5

1.
01

1.
01

5

1.
02

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission relative to 1145

Fr
eq

ue
nc
y 
[G
H
z]

11
45

12
45

13
45

15
00

16
00

Ti
m
e

hh
:m

m

F
ig
u
re

4
.8
:
N

2
sp

ec
tr

a
ta

ke
n

ov
er

a
p

er
io

d
of

fo
u
r

h
o
u
rs

fo
r

th
e

g
a
s

ce
ll

a
t

ro
o
m

te
m

p
er

a
tu

re
a
n
d

n
o
rm

a
li
ze

d
to

th
e

fi
rs

t
o
n

e
o
f

th
em

.
M

ov
in

g
av

er
ag

e
ov

er
15

G
H

z.
[d

at
a/

d
ri

ft
s/

b
o
o
k
3
-2

0
1
3
0
3
1
1
]

45



4.4. NOISE AND DRIFTS
CHAPTER 4. EXPERIMENTAL METHODOLOGY

0.
99

1.
01

1.
03

1.
05

1.
07

1.
09

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission relative to 0740

Fr
eq

ue
nc
y 
[G
H
z]
 

74
0

93
0

10
00

11
10

12
10

13
20

14
40

15
30

Ti
m
e 

hh
:m

m

F
ig
u
re

4
.9
:
N

2
sp

ec
tr

a
ta

ke
n

ov
er

a
p

er
io

d
of

ei
g
h
t

h
o
u

rs
a
t

h
ig

h
te

m
p

er
a
tu

re
(o

ve
n

a
t

6
0
0

C
)

a
n
d

n
o
rm

a
li
ze

d
to

th
e

fi
rs

t
o
n
e

o
f

th
em

.
T

h
e

le
ge

n
d

sh
ow

s
th

e
ti

m
e

th
at

th
e

co
rr

es
p

o
n
d

in
g

m
ea

su
re

m
en

t
w

a
s

ta
k
en

.
M

ov
in

g
av

er
a
g
e

ov
er

1
5

G
H

z.
[d

a
ta

/
d
ri

ft
s/

b
o
o
k
1
-

20
13

03
12

]

0.
95

0.
96

0.
97

0.
98

0.
991

1.
01

1.
02

1.
03

1.
04

1.
05

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission relative to 1210

Fr
eq

ue
nc
y 
[G
H
z]

12
10

13
00

14
00

15
00

16
10

17
00

Ti
m
e

hh
:m

m

F
ig
u
re

4
.1
0
:
N

2
sp

ec
tr

a
ta

ke
n

ov
er

a
p

er
io

d
o
f

fi
ve

h
o
u

rs
a
t

h
ig

h
te

m
p

er
a
tu

re
(o

v
en

a
t

6
0
0

C
)

a
n

d
n

o
rm

a
li
ze

d
to

th
e

fi
rs

t
o
n
e

of
th

em
,

w
it

h
b

et
te

r
th

er
m

al
is

ol
at

io
n

(1
0

cm
m

o
re

a
ir

b
et

w
ee

n
so

u
rc

e
a
n
d

g
a
s

ce
ll
)

th
a
n

in
fi
g
u
re

4
.9

.
M

ov
in

g
av

er
a
g
e

ov
er

1
5

G
H

z.
[d

at
a/

d
ri

ft
s/

b
o
ok

2-
20

13
03

27
]

46



4.4. NOISE AND DRIFTS
CHAPTER 4. EXPERIMENTAL METHODOLOGY

0.
95

0.
96

0.
97

0.
98

0.
991

1.
01

1.
02

1.
03

1.
04

1.
05

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

N
oi
se
 (l
ab
)

F
ig
u
re

4
.1
1
:

N
oi

se
p
lo

t
in

th
e

la
b

ob
ta

in
ed

b
y

d
iv

is
io

n
o
f

tw
o

su
b

se
q
u

en
t
N

2
-s

p
ec

tr
a
.

In
te

g
ra

ti
o
n

ti
m

e
1
0
0

m
s.

N
o
te

th
a
t

si
g
n
a
l

st
re

n
gt

h
ap

p
ro

ac
h
es

ze
ro

at
th

e
en

d
s

of
th

e
sp

ec
tr

u
m

,
ca

u
si

n
g

si
g
n
a
l-

to
-n

o
is

e
ra

ti
o

to
d
ec

re
a
se

.
In

th
is

p
lo

t
n

o
sm

o
o
th

in
g

h
a
s

b
ee

n
u
se

d
.

[d
at

a/
n
oi

se
]

0.
95

0.
96

0.
97

0.
98

0.
991

1.
01

1.
02

1.
03

1.
04

1.
05

30
0

32
0

34
0

36
0

38
0

40
0

42
0

44
0

46
0

48
0

50
0

Transmission

Fr
eq

ue
nc
y 
[G
H
z]

N
oi
se
 (G

as
ifi
er
)

F
ig
u
re

4
.1
2
:

N
oi

se
p
lo

t
at

th
e

ga
si

fi
er

o
b

ta
in

ed
b
y

d
iv

is
io

n
o
f

tw
o

su
b

se
q
u
en

t
N

2
-s

p
ec

tr
a
.

In
te

g
ra

ti
o
n

ti
m

e
1
0
0

m
s.

In
th

is
p
lo

t
n
o

sm
o
ot

h
in

g
h
as

b
ee

n
u
se

d
.

[d
at

a/
n
oi

se
]

47



5
Discussion

5.1 Detection of H2O

We have seen in figure 4.3 that the system can detect small differences in H2O VMR in
the lab, as small as 0.5% and probably smaller with sufficient time. Figure 4.7 indicates
that the system detects changes in humidity at the gasifier as well. To determine the
humidity level from the detected transmission, we need to simulate line strengths as
described previously.

The effects of line mixing seem to be significant enough to make our line-by-line model
ineffectice for full spectrum simulation. Instead, we downloaded simulated spectra from
SpectralCalc.com to get an array of transmission values for temperatures in the range
373 K to 1073 K in steps of 20 K, VMRs in the range 0% to 100% in steps of 5% and
wavelengths from 600 to 1000 µm in steps of 0.4 µm. MATLABs interp3 interpolation
function could then be used to extract values between these points.

Using the data shown in figure 4.2, the two known parameters of transmission and
VMR can be used to extract the temperature by use of the plot in figure 5.1, which is
line-by-line simulated transmission as function of temperature for the VMR given in the
middle plot. One can pinpoint the correct ration between peak amplitudes between the
two black lines inserted inte the plot. This is here shown for one of the humidity levels
but is repeatable as made clerar in table 5.1. Very high humidity levels give condensation
problems in the humidity meter but at intermediate levels the procedure seems to work
quite well. In figure 5.2 the procedure for determining temperature is illustrated for
the spectralcalc data, which gives a large inconsistency in temperature between the two
different spectral lines. The error in the calculated temperature can be estimated as
∆T = dT

dS∆S where S is the transmission. These values are included in the table 5.1.
As shown in the figures 5.1 and 5.2 below, spectralcalc data gave inconsistent results

when applied to the measured values of table 4.1. Line-by-line model gave less incon-
sistent results, but we show in figure 5.7 that it does not fit well with the spectra from

48



5.1. DETECTION OF H2O
CHAPTER 5. DISCUSSION

frequency sweeps. Thus there might be an error in the measurement of the two line
centers used here.

For the system to be operational however we need not only determine temperature for
known transmission and humidity, we need to determine both temperature and humidity
from only the transmission data. This was tried by extending the fitting procedure to
the VMR-axis as well. This gave results that deviated a lot from expected values,
probably because of the low resolution used in the fitting due to longer computation
time compared to the fitting of one variable at a time. The MATLAB program is shown
in appendix D.3. Using it to extract T and VMR from the measurements in table 5.1
gives the output shown in table 5.2.

Figure 5.1: Transmission as function of temperature calculated by the line-by-line-model
for the VMR of 73% detected in the measurements of figure . The black lines show where
the measured transmissions intersect the simulation, giving the temperature. The values
of T should coincide, but do not. Precise estimates from a least-squares fit for several
measurements of this type are shown in table 5.1. [data/temp humidity]

Figure 5.2: Transmission as function of temperature calculated by Spectralcalc for the
VMR of 73% detected in the measurements of figure . The black lines show where the mea-
sured transmissions intersect the simulation, giving the temperature. The values of T should
coincide, but do not. The temperatures differ much more than above. [data/temp humidity]

49



5.1. DETECTION OF H2O
CHAPTER 5. DISCUSSION

T
a
b
le

5
.1
:

T
em

p
er

at
u

re
es

ti
m

at
io

n
s

u
si

n
g

le
a
st

-s
q
u
a
re

s
fr

o
m

th
e

d
a
ta

in
ta

b
le

4
.1

.
T

h
e

ca
lc

u
la

ti
o
n

is
sh

ow
n

in
a
p
p

en
d

ix
D

.
E

rr
o
rs

ar
e

gi
ve

n
as

on
e

st
an

d
ar

d
d
ev

ia
ti

on
.

T
h
e

la
st

co
lu

m
n

sh
ow

s
h
ow

V
M
R

is
ca

lc
u

la
te

d
a
ss

u
m

in
g

th
e

ca
lc

u
la

te
d

te
m

p
er

a
tu

re
a
n
d

m
ea

su
re

d
tr

an
sm

is
si

on
s.

se
ri

es
T

ra
n
sm

is
si

on
T

ra
n

sm
is

si
o
n

M
ea

su
re

d
C

al
cu

la
te

d
C

al
cu

la
te

d
C

al
cu

la
te

d
C

al
cu

la
te

d

#
at

4
48

G
H

z
a
t

38
3

G
H

z
V
M
R

T
[K

]
T

[K
]

T
[K

]
V
M
R

[%
]

at
38

3
G

H
z

at
44

8
G

H
z

[%
]

1
0
.6

84
±

0.
0
08

70
0
.7

8
4
±

0.
00

42
8

95
±

1.
52

56
4
±

4
.0

55
3

56
9
±

5
.6

29
5

56
7
±

4.
84

2
92
±

8
.1

1
0
.7

21
±

0.
0
09

21
0
.8
±

0.
0
05

19
73
±

0.
99

60
4
±

5
.8

64
8

61
1
±

7
.2

91
2

60
8
±

6.
57

8
71
±

5
.4

1
0
.7

51
±

0.
0
08

36
0
.8

1
3
±

0.
00

50
2

57
±

1.
10

64
8
±

6
.6

44
2

65
9
±

7
.9

77
8

65
4
±

7.
31

1
56
±

3
.7

1
0
.7

74
±

0.
0
08

38
0
.8

2
2
±

0.
00

52
7

48
±

1.
10

68
3
±

7
.8

81
4

70
1
±

9
.2

12
6

69
2
±

8.
54

7
47
±

3
.2

1
0
.7

81
±

0.
0
08

30
0
.8

2
5
±

0.
00

50
9

41
±

0.
77

71
0
±

8
.0

59
0

72
7
±

9
.8

41
4

71
9
±

8.
95

0
40
±

2
.8

2
0
.7

86
±

0.
0
06

92
0
.8

2
4
±

0.
00

36
6

49
±

1.
19

68
3
±

5
.7

43
7

71
2
±

7
.9

47
7

69
8
±

6.
85

0
48
±

2
.6

2
0
.7

6
±

0.
00

97
9

0
.8

1
9
±

0.
00

25
9

57
±

0.
84

65
6
±

3
.5

90
8

66
8
±

9
.7

66
1

66
2
±

6.
67

9
56
±

3
.5

2
0
.7

3
±

0.
00

72
3

0
.7

9
2
±

0.
00

26
1

73
±

0.
90

59
5
±

2
.9

82
8

61
9
±

6
.3

36
3

60
7
±

4.
66

0
70
±

3
.8

T
a
b
le

5
.2
:

T
h
e

re
su

lt
w

h
en

fi
tt

in
g

b
ot

h
T

an
d
V
M
R

u
si

n
g

th
e

p
ro

g
ra

m
in

a
p
p

en
d
ix

D
.3

,
a
n
d

th
e

sa
m

e
d
a
ta

a
s

in
th

e
ta

b
le

a
b

ov
e.

S
er

ie
s

#
1

1
1

1
1

2
2

2

M
ea

su
re

d
V
M
R

[%
]

95
7
3

57
48

41
49

57
73

C
al

cu
la

te
d
V
M
R

[%
]

85
5
5

35
20

20
10

40
30

C
al

cu
la

te
d
T

[K
]

5
73

63
3

7
13

80
3

81
3

89
3

70
3

70
3

50



5.2. DETECTION OF CO
CHAPTER 5. DISCUSSION

5.2 Detection of CO

While CO is clearly visible in the rawgasspectra of figure 4.4 and 4.5 the precision with
which it can be detected has not been determined.

5.3 Noise and drifts

The measurements in figure 4.8,4.9 and 4.10 indicate that the room temperature drift
is less than 0.25% per hour, small compared to the noise. At higher temperature, the
signal drifts at a rate of up to 0.6% per hour. When the heat was reduced by moving
the source further from the gas cell, the magnitude of drifts has reduced to about 0.4%
per hour, indicating that thermal isolation is appropriate to achieve good stability.

The relation of noise to integration time was tested in the lab with pure nitrogen
at a temperature of 308/325oC as shown by the thermocouples at inlet/outlet and at a
frequency 383 GHz. Measurements were taken with a time constant of 300 ms once per
second for an hour. Figure 5.4 shows the Allan variance of the measurement. The Allan
variance of the 449 GHz line taken at the gasifier, for rawgas, is shown in figure 5.6 and
the raw signal in figure 5.5. That measurment was taken with 100 ms integration time
and one point per 300 ms.

Allan variance is a measure of how RMSD1 relates to integration time. The Allan
M -sample variance is defined as:

σ2
y(M,T,τ) =

1

M − 1

(M−1∑
i=0

[x(iT + τ)− x(iT )

τ

]2
− 1

M

[M−1∑
i=0

[x(iT + τ)− x(iT )

τ

])
(5.2)

Where M is number of samples, x is the data and T is the time. Integration time is then
Mτ . We can see in the figure that the optimal integration time is reached after only
20-30 seconds. If there were no drifts, there would be no optimum but an exponentially
decreasing payoff for increasing integration time.

1

RMSD =

√
1

n

∑
i

[(xi − x̄)2] (5.1)

where x̄ is the signal mean, used as the ”true” signal.

51



5.3. NOISE AND DRIFTS
CHAPTER 5. DISCUSSION

32
00
0

32
10
0

32
20
0

32
30
0

32
40
0

32
50
0

32
60
0

32
70
0

32
80
0

32
90
0

0
50
0

10
00

15
00

20
00

25
00

30
00

35
00

Signal strength

Ti
m
e 
[s
]

F
ig
u
re

5
.3
:

A
n
N

2
m

ea
su

re
m

en
t

at
38

3
G

H
z

w
it

h
3
0
0

m
il
li
se

co
n

d
in

te
g
ra

ti
o
n

ti
m

e
a
n
d

o
n
e

d
a
ta

p
o
in

t
ta

ke
n

p
er

se
co

n
d

,
in

th
e

la
b

.[
d
at

a/
rm

sd
]

0
0.
00
02

0.
00
04

0.
00
06

0.
00
08

0.
00
1

0.
00
12

0.
00
14

0.
00
16

0.
00
18

0
20

40
60

80
10
0

12
0

Allan Variance

In
te
gr
at
io
n 
tim

e 
[s
]

F
ig
u
re

5
.4
:

A
ll
a
n

va
ri

a
n

ce
o
f

th
e

a
b

ov
e

si
g
n
a
l.
[d

a
ta

/
rm

sd
]

52



5.3. NOISE AND DRIFTS
CHAPTER 5. DISCUSSION

27
50
0

27
70
0

27
90
0

28
10
0

28
30
0

28
50
0

28
70
0

28
90
0

0
10
0

20
0

30
0

40
0

50
0

60
0

70
0

80
0

90
0

10
00

11
00

12
00

Signal strength

Ti
m
e 
[s
]

F
ig
u
re

5
.5
:

T
h
e

44
9

G
H

z
li
n

e
fr

om
ra

w
ga

s
a
t

th
e

g
a
si

fi
er

(2
0
0

k
g
/
h

o
f
H

2
O

),
m

o
n
it

o
re

d
fo

r
1
2
0
0

se
co

n
d
s.

1
0
0

m
il
li
se

co
n

d
in

te
gr

at
io

n
ti

m
e

an
d

on
e

d
at

a
p

oi
n
t

ta
ke

n
p

er
30

0
m

il
li
se

co
n
d

s.
[d

a
ta

/
rm

sd
]

0

0.
00
1

0.
00
2

0.
00
3

0.
00
4

0.
00
5

0
20

40
60

80
10
0

12
0

Allan Variance

In
te
gr
at
io
n 
tim

e 
[s
]

F
ig
u
re

5
.6
:

A
ll
a
n

va
ri

a
n

ce
o
f

th
e

a
b

ov
e

si
g
n

a
l.

[d
a
ta

/
rm

sd
]

53



5.4. OTHER ISSUES
CHAPTER 5. DISCUSSION

5.4 Other issues

5.4.1 Temperature gradient and thermocouples

The thermocouples used to determine gas temperature at the inlet and outlet of the
cell proved to be quite unpredictable. This is so for several reasons. First, they couple
radiatively to the hot steel of the waveguide, leading to exaggarated temperatures. On
the other hand, they are not very exactly aligned to measure the temperature at repre-
sentative points, such as the exact center of the gas flow. Patterns of turbulence inside
the cell may have an unknown impact. It is thus not surprising that the thermocouple
readings were only very roughly related to temperatures estimated from the spectra.

One effect seen in the thermocouple measurements is worthy of consideration how-
ever. Gas enters the cell from 200oC heated tubes, coming into contact with a hotter
environment. It starts heating in the pre-heating tube spiralling along the gas cell. Then
it exits, possibly not yet in thermal equilibrium, into the gas cell. Thus a temperature
gradient of significant magnitude can be expected to exist in the cell. Its magnitude
is not in any simple way related to the various parameters of the gas. One could have
expected that higher vapor contents would lower the magnitude of the gradient due to
higher heat capacity, but on the other hand vapor content was regulated by decreas-
ing N2 flow, meaning that the gas had longer time to gain heat from the walls of the
waveguide. Higher mass flow would introduce turbulence at lower humidity levels, which
would promote heat transfer from wall to gas.

A third insertion point for a thermocouple straight into the middle of the gas cell
was tested and gave a temperature varying around 600o C, the temperature of the oven.
This is probably a good reason to believe the measurements at the other two insertion
points were of little relevance. Spectroscopy cannot determine the temperature gradient
but only the average temperature.

5.4.2 Line-by-line model vs LINEPAK

The line-by-line simulation did not perform as good as the Spectralcalc simulations to
reproduce measured spectra. When trying to extract parameters, spectralcalc data gave
inconsistent results while the line-by-line algorithm produced more consistent results. A
comparison of the two methods is shown in 5.7. Since the Spectralcalc data worked well
for fitting to the full spectra, we would expect it to produce consistent results. There
might have been a systematic error in the measurement of two lines simultaneously that
can explain this problem.

5.5 Usability of the system

The system can detect water in a matter of seconds in the environment of the lab.
We have also shown that waterlines in the rawgas can be detected, and the required
integration time is on the order of 2 minutes. The presence of CO can be simultaneously
detected with an as yet unclear accuracy. For these experiments an expensive THz source

54



5.5. USABILITY OF THE SYSTEM
CHAPTER 5. DISCUSSION

F
ig
u
re

5
.7
:

A
co

m
p

ar
is

on
of

ra
w

g
as

m
ea

su
rm

en
t,

sp
ec

tr
a
lc

a
lc

a
n
d

li
n

e-
b
y
-l

in
e

si
m

u
la

ti
o
n
.

[d
a
ta

/
co

m
p

a
ri

so
n
]

55



5.5. USABILITY OF THE SYSTEM
CHAPTER 5. DISCUSSION

has been used, but in an applied system a cheaper set of a few fixed-frequency sources
could be used. This would reduce the cost drastically. It still remains to be determined
how many frequencies need to be monitored, but at least three separate ones for water.
Since temperature will be determined by the water measurement, one peak from each
additional substance would suffice.

56



Bibliography

Craig F. Bohren and Donald R. Huffman. Appendixes: Computer Programs, pages
475–476. Wiley-VCH Verlag GmbH, 2007. ISBN 9783527618156. doi: 10.1002/
9783527618156.app1. URL http://dx.doi.org/10.1002/9783527618156.app1.

S. Cherednichenko, A. Hammar, S. Bevilacqua, V. Drakinskiy, J. Stake, and Alexey
Kalabukhov. A room temperature bolometer for terahertz coherent and incoherent
detection. Terahertz Science and Technology, IEEE Transactions on, 1(2):395–402,
2011. ISSN 2156-342X. doi: 10.1109/TTHZ.2011.2164654.

Stefan Dahlgren, Anna Liljeblad, Jenny Cerruto, Nohlgren Ingrid, and Starberg Kata-
rina. Realiserbar biogaspotential i Sverige år 2030 genom rötning och förgasning. WSP,
Stockholm, 2013.

DDBST. Dortmund data bank saturated vapor pressure online calculator, 2013. URL
http://ddbonline.ddbst.de/AntoineCalculation/AntoineCalculationCGI.

exe?component=Water.

Marcel J. E. Golay. Theoretical consideration in heat and infra-red detection, with
particular reference to the pneumatic detector. Review of Scientific Instruments, 18
(5):347–356, 1947. doi: 10.1063/1.1740948. URL http://link.aip.org/link/?RSI/

18/347/1.

A. Gomez-Barea and B. Leckner. Modeling of biomass gasification in fluidized bed.
Progress in Energy and Combustion Science, 36:444–509, 2010. ISSN 0360-1285. doi:
10.1016/j.pecs.2009.12.002.

Nachappa Gopalsami, Sasan Bakhtiari, Thomas W. Elmer II, and Apostolos C. Rap-
tis. Application of millimeter-wave radiometry for remote chemical detection. IEEE
transactions on microwave theory and techniques, 56(3):700–709, 2008.

Nachappa Gopalsami, Apostolos C. Raptis, H. T. Chien, A. Heifetz, and E. R. Koehl.
Millimeter wave detection of nuclear radiation: An alternative detection mechanism.
Review of scientific instruments, 80, 2009. ISSN 0034-6748.

57

http://dx.doi.org/10.1002/9783527618156.app1
http://ddbonline.ddbst.de/AntoineCalculation/AntoineCalculationCGI.exe?component=Water
http://ddbonline.ddbst.de/AntoineCalculation/AntoineCalculationCGI.exe?component=Water
http://link.aip.org/link/?RSI/18/347/1
http://link.aip.org/link/?RSI/18/347/1


BIBLIOGRAPHY BIBLIOGRAPHY

Larry L. Gordley, Benjamin T. Marshall, and D. Allen Chu. Linepak: Algorithms for
modeling spectral transmittance and radiance. Journal of Quantitative Spectroscopy
and Radiative Transfer, 52(5):563 – 580, 1994. ISSN 0022-4073. doi: 10.1016/
0022-4073(94)90025-6. URL http://www.sciencedirect.com/science/article/

pii/0022407394900256.

Jean-Michel Hartmann, Christian Boulet, and Daniel Robert. Collisional effects on
molecular spectra: laboratory experiments and models, consequences for applications.
Elsevier Science, 2008.

Eugene Hecht. Optics (4th Edition). Addison Wesley, 4 edition, August 2001.
ISBN 0805385665. URL http://www.amazon.com/exec/obidos/redirect?tag=

citeulike07-20&path=ASIN/0805385665.

Sven Hermansson, Fredrik Lind, and Henrik Thunman. On-line monitoring of fuel
moisture-content in biomass-fired furnaces by measuring relative humidity of the flue
gases. Chemical Engineering Research and Design, 89(11):2470 – 2476, 2011. ISSN
0263-8762. doi: 10.1016/j.cherd.2011.03.018. URL http://www.sciencedirect.com/

science/article/pii/S0263876211001419.

N.S. Hush and M.L. Williams. Carbon monoxide bond length, force constant and infrared
intensity variations in strong electric fields: Valence-shell calculations, with applica-
tions to properties of adsorbed and complexed co. Journal of Molecular Spectroscopy,
50(1–3):349 – 368, 1974. ISSN 0022-2852. doi: 10.1016/0022-2852(74)90241-0. URL
http://www.sciencedirect.com/science/article/pii/0022285274902410.

Philip Laven. Mieplot - a computer program for scattering of light from a sphere us-
ing mie theory & the debye series, May 2013. URL http://www.philiplaven.com/

mieplot.htm.

K.N. Liou. An Introduction to Atmospheric Radiation, volume 84 of International
Geophysics. Academic Press, 2002. doi: 10.1016/S0074-6142(02)80015-8. URL
http://www.sciencedirect.com/science/article/pii/S0074614202800158.

Lorenzo Lori. Theoretical Rotational-Vibrational Spectroscopy of Water. PhD thesis,
University of London, 2008.

Donald A. McQuarrie. Quantum Chemistry. University Science Books, Saulito, Califor-
nia, second edition, 2008.

Robert C Reid, John M Prausnitz, and Bruce E Poling. The properties of gases and
liquids. 1987.

A. Rogalski and F. Sizov. Terahertz detectors and focal plane arrays. Opto-Electronics
Review, 19(3):346–404, 2011. ISSN 1230-3402. doi: 10.2478/s11772-011-0033-3. URL
http://dx.doi.org/10.2478/s11772-011-0033-3.

Antonio Rogalski. Infrared Detectors, Second Edition. CRC Press, 2010.

58

http://www.sciencedirect.com/science/article/pii/0022407394900256
http://www.sciencedirect.com/science/article/pii/0022407394900256
http://www.amazon.com/exec/obidos/redirect?tag=citeulike07-20&path=ASIN/0805385665
http://www.amazon.com/exec/obidos/redirect?tag=citeulike07-20&path=ASIN/0805385665
http://www.sciencedirect.com/science/article/pii/S0263876211001419
http://www.sciencedirect.com/science/article/pii/S0263876211001419
http://www.sciencedirect.com/science/article/pii/0022285274902410
http://www.philiplaven.com/mieplot.htm
http://www.philiplaven.com/mieplot.htm
http://www.sciencedirect.com/science/article/pii/S0074614202800158
http://dx.doi.org/10.2478/s11772-011-0033-3


BIBLIOGRAPHY

L.S. Rothman, I.E. Gordon, A. Barbe, D.Chris Benner, P.F. Bernath, M. Birk,
V. Boudon, L.R. Brown, A. Campargue, J.-P. Champion, K. Chance, L.H. Coud-
ert, V. Dana, V.M. Devi, S. Fally, J.-M. Flaud, R.R. Gamache, A. Goldman,
D. Jacquemart, I. Kleiner, N. Lacome, W.J. Lafferty, J.-Y. Mandin, S.T. Massie,
S.N. Mikhailenko, C.E. Miller, N. Moazzen-Ahmadi, O.V. Naumenko, A.V. Nikitin,
J. Orphal, V.I. Perevalov, A. Perrin, A. Predoi-Cross, C.P. Rinsland, M. Rot-
ger, M. Šimečková, M.A.H. Smith, K. Sung, S.A. Tashkun, J. Tennyson, R.A.
Toth, A.C. Vandaele, and J. Vander Auwera. The {HITRAN} 2008 molecular
spectroscopic database. Journal of Quantitative Spectroscopy and Radiative Trans-
fer, 110(9–10):533 – 572, 2009. ISSN 0022-4073. doi: 10.1016/j.jqsrt.2009.02.013.
URL http://www.sciencedirect.com/science/article/pii/S0022407309000727.
<ce:title>HITRAN</ce:title>.

M Vidler and J Tennyson. Accurate partition function and thermodynamic data for
water. J. Chem. Phys., (113):9766–9771, 2000.

59

http://www.sciencedirect.com/science/article/pii/S0022407309000727


A
Line-by-line algortihm

This is the line-by-line algorithm referred to in the text. Function for extracting spec-
tralcalc data provided at the end of the appendix for comparison.

function f = trans(p,vin)

T=p(1);

VMR=p(2);

v = vin*(1/29.9792); % Convert GHz to reciprocal cm

global offset params; % params is a set of HITRAN data

h = 6.63E-34; % Planck constant

c = 3E8; % speed of light

kb = 1.38E-23; % Boltzmann constant

NA=6.022E23; % Avogadro constant

R=8.2e-1; % Molar gas constant

l=1.0; % Length of gas cell

p0=1; % Reference pressure

p=1; % Pressure, atm

T0=296; % Reference temperature

% Loading the HITRAN data

n = params.n;

a0 = params.a0;

v0 = params.v0;

v02 = v0;

s = params.s;

af0 = params.af0;

El = params.El;

pp = params.pp;

60



APPENDIX A. LINE-BY-LINE ALGORTIHM

%summing contributions from 48 lines (probably a lower number would suffice)

f=0;

w=0;

for i=1:1:48

a = (M.*a0(i)+(1-M).*af0(i)).*(p./p0).*(T0./T).^n(i);

st = s(i)*

(exp(-(h*c*El(i))/(kb*T))*

ppval(T0,pp)*

(1-exp(-(h*c*v02(i))/(kb*T))))/(exp(-(h*c*El(i))/(kb*T0))*

ppval(T,pp)*

(1-exp(-(h*c*v02(i))/(kb*T0))));

w = w+(l*NA*1)*(p/(R*T))*st.*((a)./(pi))./((v+offset-v0(i)).^2+(a).^2);

end

f=exp(-w); %final transmission

The function to extract spectralcalc data:

% Interpolates values from spectralcalc data.

% VMR must be given in percent

% T in Kelvin

% freq in gigahertz

function y = spectralcalc(T,VMR,freq)

c=3e8;

nu=freq*10^9; % Frequency in GHz

load(’watersimulation.mat’);

dataset2=dataset(:,:,:);

lambda = 10^6*c/nu; % Get lambda in microns

index = (lambda-600)/0.4+1; % Get index in the array corresponding to lambda

XI = VMR/5+1; % Get the index corresponding to VMR

YI = (T-373)/20+1; % Get the index corresponding to T

ZI = index;

y=interp3(dataset2,XI,YI,ZI); % Interpolation

61



B
Line data from HITRAN2008

Reference temperature T0 for all T -dependent values is 293 K. Reference pressure 1 atm.
Explanations:

• ν0 is the frequency of the line center.

• S is the strength of the line at reference temperature.

• αair foreign-broadened half-width due to collisions with air.

• αself self-broadened half-width, i.e. broadening due to collisions betweenN2 molecules.

• Elower Energy of the lower state involved in the transition generating the line.

• n temperature-dependence exponent used in calculating line broadening.

Table B.1: Spectral lines of CO

ν0[cm−1] S[cm−1/(molecule− cm−2)] αair[cm−1/atm] αself [cm−1/atm] Elower[cm−1] n
7.689917 2.552e-23 .0748 .0820 3.84500 .75
11.534513 8.210e-23 .0709 .0790 11.53490 .74
15.378665 1.820e-22 .0676 .0750 23.06940 .74
19.222222 3.262e-22 .0650 .0730 38.44810 .74
23.065059 5.075e-22 .0629 .0700 57.67030 .74
26.907008 7.117e-22 .0612 .0690 80.73530 .75
30.747929 9.206e-22 .0599 .0670 107.64230 .75
33.067177 1.126e-23 .0589 .0660 132.30500 .75
34.587674 1.114e-21 .0589 .0660 138.39030 .75
36.737059 1.299e-23 .0580 .0640 165.37220 .75
38.426099 1.275e-21 .0580 .0640 172.97790 .75

62



APPENDIX B. LINE DATA FROM HITRAN2008

Table B.2: Spectral lines of water

ν0[cm−1] S[cm−1/(molecule− cm−2)] αair[cm−1/atm] αself [cm−1/atm] Elower[cm−1] n
6.114567 7.785e-23 .0998 .5250 136.16390 .78
10.845940 9.105e-23 .0932 .5070 315.77950 .76
12.682023 8.304e-22 .0961 .6370 212.15640 .76
14.648500 7.145e-23 .0668 .3290 742.07630 .65
14.777502 1.483e-23 .0585 .2700 1045.05800 .62
14.943711 8.679e-22 .0871 .4670 285.41860 .71
15.707169 2.740e-23 .0692 .3460 742.07310 .69
15.833930 1.094e-22 .0764 .3960 488.13420 .70
16.294303 2.216e-23 .0854 .4870 586.47920 .76
18.268530 1.004e-22 .1039 .4860 23.75490 .77
18.413429 1.906e-23 .1039 .4860 23.77350 .77
18.