University of Groningen
Experimental study of the combustion properties of methane/hydrogen mixtures
Gersen, Sander
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to
cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2007
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Gersen, S. (2007). Experimental study of the combustion properties of methane/hydrogen mixtures s.n.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.
Download date: 16-06-2017
RIJKSUNIVERSITEIT GRONINGEN
Experimental study of the combustion properties of
methane/hydrogen mixtures
Proefschrift
ter verkrijging van het doctoraat in de
Wiskunde en Natuurwetenschappen
aan de Rijksuniversiteit Groningen
op gezag van de
Rector Magnificus, dr. F. Zwarts,
in het openbaar te verdedigen op
vrijdag 7 december 2007
om 14:45 uur
door
Sander Gersen
geboren op 2 oktober 1976
te Gouda
Promotor:
Copromotor:
Prof. dr. H.B. Levinsky
Dr. A.V. Mokhov
Beoordelingscommissie: Prof. dr. ir. R. Baert
Prof. dr. H.C. Moll
Prof. dr. ir. Th.H. van der Meer
ISBN 978-90-367-3254-3
ISBN 978-90-367-3255-0 (electronic version)
Experimental study of the combustion
properties of methane/hydrogen
mixtures
Sander Gersen
The work described in this thesis was performed in the Laboratory for Fuel and
Combustion Science at the University of Groningen, Nijenborgh 4, 9747 AG
Groningen, the Netherlands. This project is supported with a grant of the Dutch
Program EET (Economy, Ecology, Technology) a joint initiative of the Ministries of
Economic Affairs, Education, Culture and Sciences and of Housing, Spatial Planning
and the Environment. The program is run by the EET Program Office, SenterNovem.
S.Gersen,
Experimental study of the combustion properties of methane/hydrogen mixtures,
Proefschrift Rijksuniversiteit Groningen (2007)
Table of contents
Introduction
5
Chapter 1 : Combustion properties of homogeneous reacting gas mixtures
1.1 Motivation to study the combustion properties of CH4/H2 gas mixtures
8
1.2 Governing equations for a homogeneous reacting gas mixture in a closed
gas system
1.3 Laminar premixed flames
1.3.1 Governing equations for a one-dimensional laminar flame
1.4 Chemical mechanisms
16
18
21
22
Chapter 2: The Rapid Compression Machine, Experimental Techniques,
Procedures and Setup
2.1 Background
27
2.2 Design and Operations
38
2.2.1 Experimental System
32
2.2.2 Gas filling system and filling procedure
33
2.2.3 Instrumentation and data acquisition
34
2.3.4 Determination autoignition delay time
35
2.3.5. Temperature determination
36
Appendix A.1
40
Appendix A.2
41
Appendix A.3
42
Appendix A.4
43
Appendix A.5
44
Chapter 3: High-pressure autoignition delay time measurements in
methane/hydrogen fuel mixtures in a Rapid Compression Machine
3.1 Introduction
49
3.2 Experimental approach
51
3.3 Numerical simulation and analysis of experimental data
52
I
3.3.1 Chemical mechanisms
52
3.3.2 Numerical simulations
53
3.4 Results and discussion
56
3.5 Comparison of experimental results with numerical simulations
63
3.6 Summary and conclusions
68
Chapter 4: One-dimensional laminar flames, Experimental Techniques,
Procedures and Burner Setup
4.1 General introduction
73
4.2 Burner
76
4.3 Gas handling system
77
4.4 Extractive probe sampling system
79
4.5 Estimate of the conversion of C2H2 and HCN during sampling
80
4.6 Laser absorption spectroscopy
82
4.6.1 Theory
82
4.6.2 Wavelength Modulation Absorption Spectroscopy (WMAS)
83
4.7 Experimental setup for Tunable Diode Laser Absorption Spectroscopy
87
4.7.1 Experimental procedure TDAS measurements of acetylene
88
4.8 Experimental procedure WMAS with second harmonic detection
90
4.8.1 HCN measurements
90
Chapter 5: Extractive Probe Measurements of acetylene in atmospheric pressure
fuel-rich premixed methane/air flames
5.1 Introduction
100
5.2 Experimental
101
5.3 Results and discussion
102
5.4 Conclusions
107
II
Chapter 6: HCN formation and destruction in atmospheric pressure fuel-rich
premixed methane/air flames
6.1 Introduction
110
6.2 Experimental
112
6.3 Results and discussion
113
6.4 Conclusions
120
Chapter 7: The effect of hydrogen addition to rich stabilized methane/air flames
7.1 Introduction
124
7.2 Experimental
126
7.3 Results and discussion
127
7.3.1 HCN profiles
128
7.3.2 C2H2 profiles
131
7.4 Conclusion
132
Summary
136
Samenvatting
140
Dankwoord
144
III
IV
Inroduction
Introduction
Combustion is mankind’s oldest technology. Nowadays the combustion of fossil
fuels provides more than 80% of the world’s energy, and is used for electric power
generation, domestic heating, transportation and many other processes. A negative
aspect of fossil fuels is that during combustion not only heat is generated, but also
pollutants such as soot and NOx. Moreover, the combustion of fossil fuels disturbs the
atmospheric CO2 balance, which is believed to contribute to global warming.
Stringent emission regulations and the expectation that the known fossil fuel
reserves will be exhausted within this century, forces combustion researchers to find
methods to reduce pollutant emission, improve the efficiency of combustion
equipment and to utilize renewable energy sources, such as biogas and hydrogen, as
alternative fuels. However, the currently available renewable energy sources are
insufficient to satisfy the world’s energy consumption. In a sustainable economy,
hydrogen, either from electrolysis of water from sustainable generated electricity
(wind, water) or from biomass, can fulfill a role as energy carrier. Yet at present, there
is no sustainable hydrogen production, nor is there widespread energy conversion
technology to utilize hydrogen as a fuel. To avoid the necessity of large investment in
new hydrogen utilization equipment, the addition of hydrogen to natural gas could be
a first step towards the wide-scale introduction of hydrogen into the energy
infrastructure. However, since the combustion properties of hydrogen differ in many
respects from those of natural gas, the allowable fraction of hydrogen in natural gas
may be limited by the deteriorating performance of gas combustion equipment such as
spark-ignited engines, burners and turbines to hydrogen-enriched natural gas. For
example, increased knock in gas engines, causing extensive damage to the machines,
or unacceptable increases in NOx formation from combustion equipment, both caused
by the presence of hydrogen in the fuel, are clearly unwanted side-effects and must be
avoided.
To investigate these practical consequences of the changes if fuel composition
effectively, it is necessary to study the changes in the underlying physical and
chemical processes that are responsible for the combustion behavior of natural gas
when hydrogen is added. Gaining fundamental insight into these consequences using
practical combustion devices is difficult, since the experimental conditions are
generally poorly defined, complicating the interpretation of the data. For this reason,
5
Introduction
devices as shock tubes, rapid compression machines (RCM) and one-dimensional
flame burners have been developed to enable the study of combustion under welldefined conditions. The insights gained from such studies permit the analysis of the
behavior of broad groups of practical combustion equipment, and are also
indispensable for the design of new combustion equipment. Furthermore, data from
these well-defined studies aid the development of methods for modeling complex
combustion phenomena.
The objective of this thesis is to investigate potential changes in the combustion
properties of methane caused by the addition of hydrogen to the fuel. Specifically, the
ignition properties of methane/oxygen and methane/hydrogen/oxygen mixtures are
studied by measuring auto ignition delay times in a rapid compression machine
(RCM) at conditions relevant to knock in gas engines (950<T<1100K and
10<P<70bar). In addition, insight into changes in soot and NOx formation in methane
flames is gained by measuring the spatial profiles of C2H2 and HCN in atmospheric
pressure, one-dimensional CH4/air and CH4/H2/air flames. All measurements are
compared with the results of numerical calculations designed to predict the behavior
of these experimental systems.
6
Chapter 1
CHAPTER 1
Combustion properties of homogeneous reacting
gas mixtures
7
Chapter 1
1.1. Motivation to study the combustion properties of CH4/H2 gas mixtures
Consider a homogeneous (premixed) fuel/oxidizer gas mixture. A premixed
mixture is characterized by the equivalence ratio, ϕ, which expresses the ratio of fuel
and oxidizer the unburned mixture. This is given by,
ϕ=
[ Fuel ]
1
.
,
[Oxidizer ] f st
(1.1)
where the amounts [fuel] and [oxidizer] can be expressed in molar, volume or mass
units, and fst is the ratio of fuel to oxidizer under stoichiometric conditions using the
same units. Here we will generally use moles or mole fraction as units. A mixture is
said to be stoichiometric (ϕ=1) when fuel and oxidizer are present in the ratios
prescribed by the balanced chemical reaction for combustion:
n
n
C x H n + ( x + )O2 → xCO2 + H 2 O ,
4
2
(R1.1)
If the oxidizer in the unburned mixture is in excess, the mixture is said to be fuel-lean
(ϕ<1), while the mixture is called fuel-rich (ϕ>1) when an excess of fuel in the
unburned mixture is present. The gas mixture can remain unreacted, such as in fuel-air
mixtures at room temperature in the absence of an ignition source, or the fuel and
oxidizer react (combust) to form products. Combustion can take place either in a
flame (a reaction front propagates subsonically through the mixture) or in a non-flame
mode (“homogeneous combustion”, reaction occurs simultaneously everywhere in the
mixture). To understand which mode of combustion takes place under a given set of
conditions (temperature, pressure, equivalence ratio), it is necessary to study the
chemical processes in the system in detail. The overall combustion process can be
described by reaction (R1.1). However, it is unrealistic to think that combustion
proceeds via this single reaction because it would require breaking high-energy bonds,
which at room temperature makes this reaction extremely slow. Instead, combustion
occurs in a sequential process involving many reactive intermediate species.
To illustrate this process, we first consider a H2-O2 gas mixture. The conversion of
hydrogen and oxygen to water starts with the formation of reactive species (radicals)
8
Chapter 1
to initiate a chain of reactions [1]. The reactions in which radicals are formed from
stable species are called chain-initiation reactions, and an example of a chaininitiation reaction is the (endothermic) dissociation reaction:
H 2 + M = 2 H + M − 436kJ / mole ,
(R1.2)
The H radicals formed in reaction (R1.2) can react further with oxygen molecules,
forming two new radicals, OH and O,
H + O2 = OH + O − 70.6kJ / mole .
(R1.3)
This reaction (R1.3), in which two radicals are created for each radical consumed is
called a chain-branching reaction and is crucially important in combustion processes.
The formation of the radicals OH and O can lead to further chain branching via
O + H 2 = OH + H − 8.2kJ / mole .
(R1.4)
In addition, there are reactions in which the number of radicals does not change, such
as
OH + H 2 = H 2 O + H + 63.21kJ / mole ,
(R1.5)
which are called chain-propagating reactions. The reactions in which radicals react to
stable species without forming another radical are called chain-terminating reactions,
as in
H + OH + M = H 2 O + M + 499.2kJ / mole .
(R1.6)
Although not strictly chain terminating, since HO2 is a radical, the reaction
H + O2 + M = HO2 + M + 208.25kJ / mole
(R1.7)
9
Chapter 1
is often considered chain terminating because, compared with the “flame radicals”, H,
O and OH, the HO2 radical is relatively unreactive. At low pressures, reaction (R1.7)
is an important chain-terminating reaction because the mildly reactive HO2 radicals
diffuse to the wall, where they react at the surface. Summing the chain-branching and
chain-propagating reactions (R1.3+R1.4+R1.5+R1.5) results in
H + 3H 2 + O2 → 3H + 2 H 2 O + 47.62kJ / mole ,
(R1.8)
from which we can see that starting with one radical, three radicals are formed from
the reactants in this simplified mechanism.
For quantitative description of chemical processes the rate of change of the
species concentrations (formation and consumption) should be determined.
For
species A in an arbitrary bimolecular elementary reaction,
kf
aA + bB ⇔ cC + dD,
kr
(R1.9)
this is expressed as:
dA
= − k f [ A] a [ B ]b + k r [C ]c [ D] d ,
dt
(1.2)
where A,B,C,D denote the different species in the reaction, a,b,c,d are the
stoichiometric coefficients of species A,B,C,D, respectively, and kf and kr represent
the forward and reverse rate coefficient of the reaction. For example, the rate of
change of the oxygen radical in reaction (R1.3) is expressed as,
d [O]
= − k Rf1.3 [ H ][O2 ] + k rR1.3 [OH ][O] .
dt
(1.3)
The rate coefficients kf and kr are connected through the equilibrium constant Kw. The
reaction rate constant k of a reaction is generally assumed to have a modifiedArrhenius temperature dependence,
⎛ E ⎞
k = AT b exp ⎜ − A ⎟ ,
⎝ RT ⎠
10
(1.4)
Chapter 1
where A is the pre-exponential factor, b the temperature exponent, T the temperature,
R the universal gas constant and Ea the activation energy, which corresponds to the
energy barrier that has to be overcome during reaction. Here should be mentioned that
the activation energy is always higher than the heat of the formation. Generally, the
more exothermic a reaction is, the smaller the activation energy.
As can be seen from reactions (R1.3-R1.5), the rate of formation of the
important free radicals ( H + OH + O) is proportional to the concentration [n] of the
radicals with some coefficient α. The rate of consumption of radicals (chain
termination) is proportional to the concentration [n] as well (R1.7), with some rate β;
the rate of chain initiation, denoted as γ, is independent of the concentration [n]
(R1.2). Thus, in generalized form, the rate of change of the concentration of free
radicals can be expressed as,
d [ n]
= γ + (α − β )[n] ,
dt
(1.5)
From equation (1.5) three different scenarios can be derived, which are presented
schematically in figure 1.1 a [2].
Figure 1.1a) Schematic of the growth of the concentration of free radicals [n] in time.
b) Schematic of the growth of free radicals [n] in time for the cases with and without
heat release.
11
Chapter 1
For the condition α>β the concentration [n] increases exponentially in time, and
ignition takes place. When α<β, d[n]/dt becomes zero, and no exponential growth of
free radicals occurs (no ignition). The condition α=β results in a linear growth of the
concentration of free radicals, and defines the ignition limit. Equation (1.5) shows that
the growth of the concentration of free radicals is determined by the competition
between the chain branching (R1.3) and chain terminating reactions (R1.7). A very
important parameter in this competition is the temperature, since the chain branching
reaction (R1.3) has large activation energy Ea [3] while that of the chain terminating
reaction (R1.7) is small [4]. Thus, the value of α (chain branching) is strongly
dependent upon the temperature, and β (chain terminating) is more or less
independent of the temperature. At low temperatures the endothermic chain branching
reaction will not proceed rapidly, so the value of α is much smaller than β (α<β), and
ignition does not occur. Increasing the temperature results in an increase in α, while β
remains unchanged; at sufficiently high temperatures α will be larger than β and
ignition occurs. The temperature during the early period of the ignition process
remains more or less constant because the heat release from the branching and
propagating reactions (R1.8) is small, but as the radical concentration grows,
exothermic reactions such as (R1.6) and (R1.7) will produce substantial quantities of
heat. If the heat produced by the exothermic reactions in system exceeds the rate of
heat loss to the surrounding, the temperature in the system will rise. Since the rate of
reaction, and thus the rate of heat release, grows exponentially with temperature, the
overall reaction will auto-accelerate, that is, the system will “explode”. The time
before explosion takes place is called autoignition delay time (figure 1.1b). In this
example, if no heat accumulates in the system (T=constant), no auto-acceleration of
the reaction rate takes place; in this case we speak of “non-explosive” reactions, both
situations are shown in figure 1.1b.
As described above, the dominance of chain branching reaction (R1.3)
characterizes the high temperature regime, while at low and intermediate temperatures
the in essential chain terminating reaction (R1.7) competes effectively with reaction
(R1.3) [5]. Since the rate of a three-body reaction increases with pressure much faster
than the rate of a two-body reaction, there exists a pressure above which reaction
(R1.7) exceeds the rate of the competing reaction (R1.3). Reaction (R1.7) is only a
chain terminating reaction when the produced HO2 radicals will diffuse to the wall
12
Chapter 1
and recombine to stable molecules without having undergone a reaction. At high
pressures, species collide much more frequently. Therefore, at sufficiently high
pressures the HO2 radicals will be frequently interrupted in its path to the walls by
reacting with H2 molecules to produce H and H2O2 (R1.10) [5],
HO2 + H 2 = H 2 O2 + H − 72.45kJ / mole .
(R1.10)
The H radicals so produced can contribute to chain branching via (R1.3) or will
generate more HO2 via reaction (R1.7) and H2O2 itself can contribute to branching via
[5],
H 2 O2 + M = 2OH + M − 214.6kJ / mole .
(R1.11)
Thus at high pressure and moderate temperatures reaction (R1.7) will dominate over
(R1.3), and the number of active centers will grow (α>β) via the sequence of
reactions (R1.7), (R1.10), and (R1.11).
Instead of heating the “cold” gas mixture by the heat release of exothermic
reactions as in this example of a closed homogeneous system, in flames, the mixture
is heated by conduction from the hot flame gases. Furthermore, radicals needed to
decompose the fuel are also transported from the high temperature region of the
flame. As in the closed system described above, the rate of formation of radicals
controls the overall rate of reaction in flames. For flames, however, the chain
initiation
and
chain-terminating
reactions
are
less
important
in
the
formation/destruction of radicals (α>>β), and the reactions (R1.3)-(R1.5), responsible
for the growth of the radical pool, dominate the overall reaction rate in H2-O2 flames
[6].
The combustion chemistry of hydrocarbon fuels is much more complicated than
that of hydrogen. As an example, we consider a CH4-O2 mixture. Under flame
conditions (α>>β), the most important chain branching reaction in the oxidation of
methane is also reaction (R1.3) [6]. After radicals are transported into the unburned
gas mixture, methane is attacked by the radicals, as in
CH 4 + H → CH 3 + H 2 − 13kJ / mole .
(R1.12)
13
Chapter 1
As can be seen from figure 1.2, the rate coefficient of reaction (R1.12) [7] is much
larger than that for reaction (R1.3) [3]. Thus reaction (R1.12) competes effectively
with reaction (R1.3) for H atoms and reduces the chain branching rate.
Figure 1.2. Reaction rate expressions for reaction R1.3 [3] and reaction R1.12 [7].
Furthermore, the very reactive H radical is replaced in reaction (R1.12) by the
unreactive CH3 radical. These two processes result in a slow conversion of methane
and contributes to the low burning velocity of methane (40 cm/s) as compared to
hydrogen (340 cm/s) [8].
If the CH4-O2 mixture under consideration is at moderate or low temperature (T
below roughly 1100K), chain branching reaction (R1.3) is too slow to provide a
sufficient branching rate for autoignition, and a different reaction path dominates [9].
These paths are extremely complex and strongly dependent upon temperature and
pressure [10]. At sufficiently high pressures, reactions involving the radical HO2
become important in the low temperature regime, for example [9],
CH 4 + HO2 → CH 3 + H 2 O2 − 85.5kJ / mole .
(R1.13)
The oxidation of the CH3 formed, and the development of the radical pool, is
complicated and slow. This process dominates most of the ignition delay period and is
14
Chapter 1
characterized by the accumulation of significant amounts intermediate species such as
H2, C2H6, CH2O and H2O2. The decomposition of H2O2 via reaction (R1.11) and the
oxidation of CH2O via a sequence of reactions lead to a sharp increase in the
concentration of free radicals [10] and ultimately ignition occurs (α>β) [10].
The H (R1.12) and HO2 (R1.13) scavenging reactions compete effectively with
R1.3 and R1.10 respectively, and thus effectively reduce the chain branching rate in
the CH4-O2 system. This, together with the formation of the relatively unreactive CH3
radical, contributes to the fact that CH4-O2 mixtures tend to auto ignite slower than
H2-O2 mixtures [6], illustrated in figure 1.3.
Figure 1.3. Computed autoignition delay times for stoichiometric H2/air and CH4/air
mixtures at P=30 bar. Calculations were made using the GRI-Mech 3.0
mechanism [11].
Besides their effects on the combustion properties, such as burning velocity and
ignition, the differences in the combustion chemistry of methane and hydrogen have
significant consequences for pollutant formation. One of the main consequences is
that during the combustion of methane (and other hydrocarbons) carbon containing
pollutant species like soot, HCN and CO are formed, while the only pollutant from
hydrogen combustion is NOx. Moreover, the formation of NO in methane combustion
is different than that from H2 combustion; in methane flames an additional mechanism
exist that produces NO via the hydrocarbon intermediate CH [12]:
15
Chapter 1
CH + N 2 ⇔ products ⇔ NO .
(R1.14)
( NCN ?, HCN ?)
This mechanism is particularly important under fuel rich conditions.
A challenging task is to understand the possible changes in the combustion
chemistry caused by addition of hydrogen to methane and how this affects
combustion properties like pollutant formation and ignition delay. Since there is a
clear distinction between the chemistry in flames and ignition, it is necessary to study
both. To gain understanding in the underlying chemical kinetics and physical
processes involved in these two kinds of combustion process, it is necessary to
analyze the combustion processes quantitatively by solving the governing equations
with detailed chemical mechanisms.
1.2 Governing equations for a homogeneous reacting gas mixture in a closed
system
The time-dependent behavior of a closed system containing a reacting gas
mixture is described by the system of the conservation equations for mass and energy.
Since no mass can be formed or destroyed by chemical reaction, the total mass of the
closed systems remains constant over time:
d ( ρV ) d K
=
∑ ρVYk =0,
dt
dt
k =1
(1.6)
where ρ is the overall mass density, V is the system volume and Yk is the mass fraction
of k-th component in the gas mixture. The mass fractions of individual species change
in time according to
ρ
16
dYk
= ω k Wk , k = 1…….K
dt
(1.7)
Chapter 1
where ωk and Wk are the molar chemical production rate per unit of volume and
molecular weight of the k-th species, respectively. The density is related to
temperature T, pressure p and composition through the ideal gas equation of state:
ρ=
p
W,
RT
(1.8)
⎛ K Y ⎞
where W = 1/ ⎜ ∑ k ⎟ is the average molecular weight of the mixture. The system
⎝ k =1 M k ⎠
(1.6) – (1.8) consists of (K+1) linearly independent equations and contains (K+3)
unknown parameters: ρ, p, T, V and Yk. Since the system contains more unknowns
than equations, it can be solved only when two unknown parameters (for example, the
measured temperature and pressure) are used as input. The number of input
parameters can be decreased to one if the energy conservation equation is added to the
system. The energy conservation equation can be derived from the first law of
thermodynamics, which states that heat δQ added to the system is equal to the sum of
the change of its internal energy dU and the work PdV of the system against an
external force:
dU + PdV = δ Q .
(1.9)
Equation (1.9) can be rewritten as
dV dQ •
dU
+P
=
= Q loss .
dt
dt
dt
(1.10)
The internal energy of the mixture is given by
U=
K
∑ ρVYk uk ,
(1.11)
k =1
where uk is a specific energy of the k-th component. After differentiating expression
(1.11) and substituting in (1.10) one receives
17
Chapter 1
⎛1⎞
d⎜ ⎟
.
ρ
dT
1 K
+ p ⎝ ⎠ + ∑ uk ωkWk = qloss ,
Cv
ρ
dt
dt
k =1
(1.12)
•
where qloss = Q loss /(ρV) is the heat loss per unit of mass and Cv is the specific heat of
system at constant volume.
For an adiabatic mixture of inert gases (ωk = 0 and qloss = 0), equation (1.12) can be
easily integrated, resulting in the following expression
T
∫
T0
⎛ ρ ⎞
C vW dT
⎟⎟ .
= ln⎜⎜
R T
⎝ ρ0 ⎠
(1.13)
It is common to use the ratio of molar heat capacities at constant pressure and constant
volume, γ, and a specific volume v, instead of Cv and ρ. In this case, equation (1.13)
can be rewritten as
T
1
∫ (γ − 1) T
T0
dT
⎛V ⎞
= ln⎜ 0 ⎟ .
⎝V ⎠
(1.14)
Several simulation programs have been developed to solve the set of
governing equations. The program used in this study is SENKIN [13], and runs in
conjunction with pre-processors from the CHEMKIN library [14], which incorporate
the chemical mechanism and thermodynamic properties.
1.3 Laminar premixed flames
Flat laminar premixed flames are ideally suited for combustion research, since
the one-dimensional character offers great advantages for modeling and unambiguous
model-experiment comparison. Moreover, the structure of these flames is
representative for many practical flames. The structure of laminar premixed flames
18
Chapter 1
can be divided in three zones (figure 1.4), the preheat zone, the flame front (reaction
zone) and the post-flame zone (burned-gas zone). In the preheat zone the unburned
gas mixture is heated by conduction and diffusion of species from the flame front; this
zone can be considered as chemically inert. The flame front, located downstream of
the preheat zone is a thin zone (in the order of 1 mm at atmospheric pressure) in
which the fuel is rapidly oxidized by radicals from the post-flame zone as described
above, leading to a steep gradients in temperature and species concentrations. The
flame front is rich in radicals and intermediate species. Although the temperature and
major species in the post-flame zone are close to their equilibrium value, the
concentrations of minor species can differ substantially from their equilibrium value.
In the post-flame zone, the system goes to equilibrium predominantly via radicalrecombination reactions such as (R1.6).
Figure 1.4. Schematic illustration of the structure of a premix one-dimensional flame.
Premixed flat flames can be characterized by the “free-flame” laminar burning
velocity, vL. In the laboratory system, where the cold gas moves with velocity vu, the
flame front propagates with velocity vu-vL. We can consider three situations regarding
the stability of a idealized one-dimensional flame. If the cold gas velocity is larger
than the laminar burning velocity, vu>vL, the flame front propagates upstream. When
19
Chapter 1
the burning velocity, vL is equal to the velocity of the unburned gasses, (v=vL) the
flame front is stationary in space, and if vu<vL the flame front will be convected
downstream. The laminar burning velocity and the temperature of the burned gas are
completely determined by the properties of the unburned mixture, such as the
equivalence ratio, temperature and the identity of the fuel [2]. Figure 1.5 shows the
interaction of the idealized 1-D flame with a porous-plug burner. Figure 1.5a
illustrates a flat flame stabilized on a burner where the unburned gas velocity is set
equal to the free-flame laminar burning velocity (vu=vL). In this situation, all heat
generated during combustion is transferred completely into the gas mixture and the
flame is essentially adiabatic (neglecting flame radiation). Lowering the unburned gas
velocity vu causes propagation of the flame front towards the burner surface. Since the
porous plug is too dense to allow propagation of the flame into the burner, the flame is
stopped in its upstream propagation. In this case, the flame transfers heat to the burner
by conduction, lowering the flame temperature and thus lowering the actual burning
velocity of the flame vL'. The flame transfers enough heat to the burner to reach a
stationary situation (vu= vL'), illustrated in figure 1.5b. This type of flame is called a
burner-stabilized flame.
Figure 1.5. a) Adiabatic flat flame (freely propagating flame)
b) Burner-stabilized flat flame
Further decrease in the unburned gas velocity results in increasing heat loss and drop
in temperature; ultimately the temperature drops to such a level that α<β and the
flame extinguishes.
20
Chapter 1
1.3.1. Governing equations for a one-dimensional laminar flame
The description of one-dimensional laminar flames is based on the conservation
equation for mass, species mass fraction and energy. Using the assumptions that onedimensional laminar flames are: (1) stationary (all flame parameters are independent
of time), (2) the system is at constant pressure and (3) effects due to viscosity,
radiation and external forces are negligible [2,15], the conservation equations
governing the behavior of these flames can be summarized as follows:
overall conservation of mass
d ( ρv) dM
=
= 0,
dx
dx
(1.15)
where v is the mass averaged flow velocity, x is the distance along the line normal to
the burner surface and M is called the mass flux.
conservation of species
d ( ρYk (Vk + v))
= ωkWk ,
dx
k=1….K
(1.16)
where Yk is the mass fraction and Vk is the diffusion velocity, which accounts for the
effect of molecular transport due to concentration gradients of the kth species [2,16].
Since mass can neither be destroyed nor formed in chemical reactions it follows from
(1.15) and (1.16) that,
K d ( ρY (V + v)) K
d ( ρ v)
k k
= ∑ ωkWk =
= 0 k=1…K.
dx
dx
k =1
k =1
∑
(1.17)
Addition of the ideal gas equation of state (1.8) to the system of equations (1.15, 1.16)
results in a system containing (K+1) linear independent equations. Assuming that the
diffusion velocity Vk is a known function of temperature and species concentrations,
21
Chapter 1
the system contains (K+2) unknown parameters (T, ρ, v and Yk). Therefore an
additional equation should be introduced to solve the system of equations:
conservation of energy
d ⎛
dT
⎜ ∑ ρYk (v + Vk ) H k − λ
dx ⎜
dx
⎝ k
⎞
⎟ = 0,
⎟
⎠
(1.18)
where Hk the specific enthalpy of species k and λ the thermal conductivity coefficient.
With the proper choice of the boundary conditions for one-dimensional flames, it is
possible to solve the governing equations [2,16].
Various software packages have been developed, which are able to calculate the
one-dimensional flame structure in only a few minutes by solving the set of governing
equations. The simulation program used in this study is the PREMIX code [17]. This
code is included in the CHEMKIN II simulation package [13]. This package operates
using a reaction mechanism data file as input, along with thermal and transport
properties of the species involved in the mechanism. The program is able to calculate
temperature- and mole fraction profiles in both burner-stabilized and free flames.
1.4 Chemical mechanisms
In the last decades chemical kinetic mechanisms have been developed to model
combustion of hydrogen (for example, see [18]) and hydrocarbon mixtures ([11,19],
among many others). These mechanisms, used to describe the transformation of
reactants into products, may contain hundreds of species and thousands of elementary
reactions. Improvement of the mechanisms currently in use is necessary, since none of
them can be regarded as comprehensive [20], i.e. accounting for all combustion
phenomena and the predictive power is only accurate for a small range of parameters.
In order to improve the existing chemical mechanisms, they should be validated
against experimental data, where parameters are varied in a well-defined manner.
Sensitivity and rate-of-production analyses are used to design and optimize models.
Using these methods rate-limiting steps and characteristic reaction paths can be
identified [2].
22
Chapter 1
The experimental data obtained in this study have been modeled using different
mechanisms. One of these mechanisms is GRI-Mech 3.0 [11], which is widely used
and has arguably become the “industry standard” for methane in the research
community. This mechanism is optimized to model natural gas combustion and
contains 325 reactions and 53 species, including reactions that describe NO formation
and reburn chemistry.
23
Chapter 1
Literature
1.
G. Dixon-Lewis., D. J. Williams., Comprehensive Chem. Kin. 17, (1977).
2.
J. Warnatz, U. Maas, R. W. Dibble, Combustion, (Springer, Berlin, 1996).
3.
C.-L. Yu, M. Frenklach, D. A. Masten, R. K. Hanson, C. T. Bowman, J.
Phys. Chem. 98 (1994) 4770-4771.
4.
M. Frenklach, H. Wang, M. J. Rabinowitz, Prog., Energy Combust. Sci. 18:
(1992) 47-73.
5.
B. Lewis, Q. von Elbe, Combustion Flames and Explosions of Gases, (Third
edition 1987).
6.
C. K.Westbrook, F. L. Dryer, Prog. Energy. Combust. Sci. 10 (1984) 1-57.
7.
J. M. Rabinowitz, J. W. Sutherland, P. M. Patterson, R. B. Klemm, J. Phys.
Chem. 95 (1991) 67-681.
8.
B. E. Milton, J. C. Keck, Combust. Flame 58 (1984) 13-22.
9.
C. K.Westbrook., Proc. of the Combust. Inst. 28 (2000) 1563-1577.
10. J. Huang, P. G. Hill, W. K.Bushe, S. R. Munshi, Combust. Flame 136 (2004)
25-42.
11. G. P. Smith, D. M. Golden, M. Frenklach, N. W. Moriarty, B. Eiteneer, M.
Goldenberg, C. T. Bowman, R. K. Hanson, S. Song, W.C. Gardiner, V.
Lissanski, Z. Qin, http://www.me.berkeley.edu/gri_mech/.
12. C. P. Fenimore, Proc. Combust. Inst. 13 (1971) 373-379.
13. A. E. Lutz, R. J. Kee, J. A. Miller, SENKIN: A FORTRAN program for
predicting homogeneous gas phase chemical kinetics with sensitivity
analysis. Sandia Report SAND87-8248. Sandia National Laboratories,
(1987).
14. R.J. Kee, F.M. Rupley, J.A. Miller, CHEMKIN II: A Fortran Chemical
Kinetics Package for the Analysis of Gas-Phase Chemical Kinetics., Sandia
National Laboratories, (1989).
15. R. M. Fristrom and A. A. Westenberg, Flame Structure, (McGraw-Hill, New
York, 1965).
16. R. J. Kee, F. M. Rupley, J. A. Miller, M. E. Coltrin, J. F. Grcar, E. Meeks, H.
K. Moffat, A. E. Lutz, G. Dixon-Lewis, M. D. Smooke, J. Warnatz, G. H.
Evans, R. S. Larson, R. E. Mitchell, L. R. Petzold, W. C. Reynolds,
24
Chapter 1
M. Caracotsios, W. E. Stewart, P. Glarborg, C. Wang, and O. Adigun,
CHEMKIN Collection, Release 3.6, Reaction Design, Inc., San Diego, CA,
(2000).
17. R. J. Kee, J. F. Grcar, M. D. Smooke, J. A. Miller, Fortran program for
modelling steady one-dimensional premixed flames. Sandia Report SAND858240. Sandia National Laboratories, (1985).
18. O. M. Conaire, H. J. Curran, J. M. Simmy, W. J. Pitz, C. K. Westbrook, Int.
J. Chem. Kin. 36 (2004) 603-622.
19. http://www.chem.leeds.ac.uk/Combustion/Combustion.html
20. J. M. Simmie, Prog. Energy Combust. Sci. 29 (2003) 599–634.
25
Chapter 2
Chapter 2
The Rapid Compression Machine
Experimental Techniques, Procedures and Setup
26
Chapter 2
2.1 Background
Over the years, several facilities have been used to investigate autoignition under
strictly controlled experimental conditions, including flow reactors, shock tubes and
rapid compression machines (RCM). While each of these facilities has its merits, their
utility is restricted to certain ranges of pressure, temperature and ignition time.
Flow reactors: In a flow reactor, fuel is injected into a flowing air stream at high
temperature and/or pressure. The combustible mixture propagates through the reactor
and, depending on the velocity ignites at some distance downstream the fuel injector
location. Because the reaction zone is spread over a large distance, the flow reactor
offers the advantage of relatively simple measurements of the evolution of species
concentrations during the ignition process. The main drawback is that pressures
achievable in flow reactors are relatively low; further, since flow reactors makes use
of electric heaters, the maximum air temperature is limited, on the order of 1000K.
One of the most advanced flow reactors was developed at Princeton University, and
provides pressures up to ∼20 bar and temperatures up to ∼1200K [1,2].
Shock tubes: A shock tube uses the compressive heating of a shock wave to
bring a premixed combustible mixture to high temperature and pressure in a very
short time. A shock tube is ideal for studying ignition phenomena with short
characteristic times (order of tens of microseconds) under the conditions obtained. A
limitation of this technique is that the well-controlled test conditions persist for less
than 5 ms [3].
Rapid Compression Machines (RCM): The operating principle of the RCM is to
compress a homogeneous fuel/oxidizer mixture to
moderate
temperatures
(Tmax ≈ 1200K) and high pressures (Pmax ≈ 70bar) in a cylinder by the motion of a
piston. The RCM offers the advantage that the temperature and pressure of the
compressed mixture can be sustained for times longer than 10ms [4]. Moreover, it
provides a simple method of simulating the processes that take place in practical
devices such as spark engines and Homogeneous Charge Compression Ignition
(HCCI) engines. The time needed to compress the test gas mixtures limits the
minimum characteristic time of investigation to ∼1ms.
Several rapid compression machines have been developed and used to study
autoignition. The RCM developed at the University of Science and Technology at
27
Chapter 2
Lille is a right angle dual piston design RCM [5]. One of the pistons is air driven and
is connected by way of a cam to the other piston that compresses the mixture. The
cam controls the length of the stroke, the initial and final position of the compressing
piston, and prevents piston rebound after ignition. The maximum compression ratio
achievable with this machine is 10. Maximum pressures and temperatures after
compression are reported around 17 bar and 900K, with total times of compression of
20-60 ms. Minetti et al. used this RCM to study autoignition and two-stage ignition of
several hydrocarbon fuels [6-8]. In addition, this group performed measurements of
the temperature distribution in their RCM [4] and found that the gas temperature is
homogeneous for ∼15ms after compression, which is then distorted due to heat loss to
the wall. Griffiths et al. (University of Leeds) studied autoignition behavior of several
fuels [9-12] using an RCM that consists of a pneumatically driven piston. In support
of their experimental work they examined the development of the temperature field in
the combustion chamber of their machine [13,14] and observed that piston motion
during compression causes a roll-up vortex that moves “cold” gas from the wall into
the core. This resulted in a region with adiabatically heated gas directly after
compression containing a plug of colder gas. The Leeds RCM has a maximum
compression ratio around 15 and is able to compress the mixture in 22 ms. Final
pressures up to 20 bar and temperatures up to 1000K have been reported in this
machine. Park and Keck (MIT) developed an RCM [15,16] that consists of a
hydraulically operated piston-cylinder assembly. They also used a piston head with a
special crevice, designed to capture vortices created during compression [15,16], to
improve the homogeneity of the core gas. Lee and Hochgreb (MIT) optimized this
piston design for the suppression of the vortices [17,18], using results of detailed
modeling. Compression ratios of ∼19, maximum peak temperatures of ∼1200K and
maximum peak pressures of ∼70 bar can be achieved in this RCM. The gas mixture
can be compressed within 10 to 30 ms. Several autoignition experiments have been
performed with this machine [19,20]. Simmie and coworkers (National University of
Ireland, Galway) used an RCM, originally built at Shell laboratories [21] that uses two
horizontally opposed pneumatically driven pistons to rapidly compress the gas
mixture. For this machine, the maximum compression ratio reported is 13, the
compression time is ∼10 ms, the maximum peak compression pressure is 44 bar and
temperatures up to 1060 K have been reported. The machine has been used to study
28
Chapter 2
methane ignition [22], among other fuels. Also, this group confirmed the importance
of the crevice in the piston head [22,23]. Recently, a free-piston RCM had been
developed by Donovan et al. (University of Michigan). Compression ratios between
16 and 37, peak pressures around 20 bar and peak temperatures of 2000K have been
reported [24] using this RCM.
Given the wide range of pressures and temperatures achievable, we chose the
MIT design to study autoignition of CH4/H2/oxidizer mixtures. For this reasons a
replica of the MIT RCM was built in our laboratory and used to study autoignition.
29
Chapter 2
2.2 Design and Operation
Conceptually, the rapid compression machine simulates a single compression
event of an internal combustion engine. It is designed to compress a gas mixture in a
short time to high temperature and pressure while maintaining a well-defined uniform
core temperature in the reaction chamber (adiabatic core). Fast compression is
necessary to prevent substantial heat losses and radical build up before the end of the
compression. The piston is driven pneumatically and is decelerated smoothly by a
special hydraulic damper to reduce the impact velocity at the end of the compression.
The machine contains only two moving parts, the “fast acting valve” and the piston.
Both are made of aluminum, while all other parts are made from steel. The piston
used is hollow, to reduce its mass and to have uniformly distributed stress throughout
the body [17]. The piston head is removable and can be replaced by heads with other
crevice configuration.
Figure 2.1. Sketch of the cross sectional view of the RCM.
30
Chapter 2
The RCM, shown schematically in figure 2.1, and in detail in Appendix A.1,
includes a nitrogen-filled driving chamber, a speed-control oil chamber, an oil
reservoir chamber, a fast acting valve, a combustion chamber and a piston. The speedcontrol oil chamber and the oil reservoir chamber, alternately connected and separated
by the fast acting valve are part of the hydraulic system that controls the movement of
the piston. The detailed sequence of operation for the RCM is given in Appendix A.2.
Briefly, after the RCM is triggered, the piston is in the “down” position and the fast
acing valve is in the “up” position, thus connecting the speed control oil chamber and
the oil reservoir chamber. To prepare for the next run, the piston is moved up until it
hits the stroke stop by pressurizing the oil reservoir chamber with ∼3 bar nitrogen.
The fast acting valve is than moved down by pressurizing the chamber above the fast
acting valve with 7 bar nitrogen, and locked in down position using 70 bar oil
pressure. The speed control and the oil reservoir chamber are now no longer
connected hydraulically. By pressurizing the speed control oil chamber with ∼48 bar
high-pressure oil, the piston is firmly locked in place against the stroke stop. After
loading the combustion chamber with the test gas mixture, the driver chamber is
pressurized with ∼35 bar nitrogen. The force on the piston created by the 35 bar
nitrogen pressure in the driving chamber is lower than the opposing force of oil on the
hydraulic piston, and hence the piston assembly is held in position by the stroke stop.
By opening the solenoid valve, the 70 bar oil pressure on the fast acting valve is
released and the fast acting valve will be pushed up by the 48 bar oil pressure in the
speed control chamber. The forces between the driving chamber and speed control
chamber are no longer balanced, and the pressure in the driving chamber causes the
piston to accelerate downward, compressing the test gas in the combustion chamber.
Subsequently, the piston’s acceleration slows, and the piston moves with constant
velocity until it is smoothly decelerated by a hydraulic damper [16]. In the final stage,
the deceleration force and velocity are reduced to zero, so that the final stop of the
piston at the bottom plate occurs without rebound. The piston is held firmly by the
force of driving nitrogen, which is greater than the force of the compressed gas
mixture in the reaction chamber. This allows combustion to take place at constant
volume. Since the area ratio of the piston on the driving side compared to the side of
combustion chamber is 4:1, the pressure inside the combustion chamber may be a
factor of 4 larger than the maximum pressure in the driving chamber before the piston
31
Chapter 2
will move. In the present construction, gas mixtures can be compressed with total
compression times of 15-30 ms up to pressures around 70bar. The piston speed can be
controlled, by varying the pressure in the driving chamber. The characteristics of the
RCM are presented in Table 2.1.
To cover a wide range of compression ratios, the rapid compression machine
was designed with adjustable piston stroke and clearance height. The piston stroke,
which is determined by the initial position of the piston, can be varied by turning the
stroke adjustment screw, see appendix A.1.. The clearance height can be changed by
replacing the clearance ring in the combustion chamber. To simulate temperatures and
pressures realistic for gas engines, different combinations of compression ratios,
initial pressures of the test gas and heat capacity of the diluent gases are used in this
study.
Table 2.1 RCM Characteristics
Cylinder bore
50.8 mm
Maximum stroke
∼160 mm
Maximum compression ratio
∼25
Clearance height
∼6.2-13 mm
Piston length
∼172 mm
Maximum driving pressure
∼35 bar
Maximum compression pressure
∼70 bar
Compression time
∼10-30 ms
2.2.1 Experimental System
Appendix A.3 shows the overall diagram of the RCM experimental system,
containing all main lines, pressure meters, oil drums, valves and the oil reservoir. All
lines are made from stainless steel with an inner diameter of 11 mm. The orifice
diameter of the solenoid valve has the same inner diameter (11 mm) as the lines, to
allow maximum throughput of oil. The high-pressure oil was supplied to the speed
control chamber (∼48 bar) and to the fast acting valve (∼70 bar) from two high-
32
Chapter 2
pressure oil accumulators. These oil accumulators contain oil and a bladder filled with
nitrogen to prevent mixing of nitrogen with oil. After ∼25 runs the accumulators were
refilled by oil from the main oil reservoir. Compressed nitrogen from five fifty-liter,
200 bar nitrogen bottles provided the required pressures in all parts of the system. The
operation pressures used, recommended by Park [16], are given in Appendix A.4
2.2.2 Gas filling system and filling procedure
All test gas mixtures were prepared in advance in a 10-liter gas bottle and used
to charge the combustion chamber at the required pressure. The gas filling system is
shown in figure 2.2. Before preparing the gas mixture, the gas bottle and the gas lines
connected to it are evacuated to less than 0.5 mbar using a vacuum pump. After
adding an individual component the bottle is closed. Subsequently, the gas lines are
again evacuated and the next mixture component is added to the bottle. The test
mixtures thus prepared were allowed to mix ∼24 hours to ensure homogeneity. Before
filling the combustion chamber, the poppet valve and the solenoid valve are opened,
and the whole system is evacuated to a pressure below 0.5 mbar. The combustion
chamber was filled with the gas mixture to the desired initial pressure, by opening the
bottle that contains the test gas. The poppet valve is then closed and the mixture is
ready for compression. After each run, the compressed gases in the RCM were vented
to the outside air, and the chamber was evacuated again before preparing the next run.
The solenoid valve in the gas-filling line was included for safety purposes, and
electrically connected such that when the solenoid valve used to trigger the fast-acting
valve is open, the solenoid in the gas-filling line is always closed. This prevents flame
propagation back to the gas-mixture bottle when the poppet valve is not properly
closed.
All test gases used in this study have purity greater than 99.5%. The composition
ratios of the gas mixtures are calculated from the measured partial pressures of the
individual gases.
33
Chapter 2
Figure 2.2. Gas handling system
2.2.3 Instrumentation and data acquisition
An MKS baratron diaphragm pressure gauge (type 722A) was used for
measuring all partial pressures of the components in the test mixture, and all other
pressures in the gas filling system. This pressure meter has an operating range from 01300 mbar with an accuracy of 0.5% of reading. The dynamic pressures in the
combustion chamber during compression and throughout the post-compression period
were measured using a Kistler 6025B piezoelectric pressure transducer (range 1-250
bar, linearity ± 0.1%) placed at the bottom of the combustion chamber. The signal
from the transducer was amplified by a 5010B Kistler charge amplifier, recorded
digitally by an oscilloscope with a sample rate of 500 kHz and 16-bit resolution, and
processed by a PC. The initial temperatures of the mixture were measured by a Pt-Rh
thermocouple with an accuracy of ± 0.2K, located at the wall of the combustion
chamber. The data acquisition was triggered simultaneously with the opening of the
solenoid for the fast-acting valve.
34
Chapter 2
2.2.4. Determination autoignition delay time
A typical measured pressure trace is presented in figure 2.3. The gas mixture is
compressed in ∼20ms, to a peak pressure that indicates the end of the compression
event. The majority of the pressure rise in the compression period takes place in a
very short time (<3 ms). During this rapid compression heat losses and radical built up
are not substantial. After the peak compression pressure is reached, the pressure drops
gradually due to heat transfer to the walls. Subsequently, heat release due to
exothermic reactions causes a slight increase in pressure, followed by a sharp increase
in pressure indicating ignition.
Figure 2.3. Typical measured pressure trace for a stoichiometric CH4/H2//O2/N2/Ar
mixture, with the definitions of autoignition delay time, peak pressure and
compression time. The dotted line shows the calculated adiabatic peak
pressure for the given compression ratio.
The auto ignition delay time is defined in this study as the time interval between
the peak pressure Pc that marks the end of compression and the time of maximum
pressure rise during ignition.
35
Chapter 2
2.2.5 Temperature Determination
To study autoignition behavior, one must know the instantaneous temperature in
the combustion chamber, since chemistry is very sensitive to temperature. However,
measuring the temperature in the reaction chamber directly by optical methods is
problematical, given the characteristic test times (order of a millisecond) and the
difficulty of making the reaction chamber optically accessible. The use of
thermocouples also permits in-situ temperature measurements, but the presence of the
thermocouple in the combustion chamber can significantly influence the
measurements, since the test gas can interact with the surface of the thermocouple.
For example, the surface of the thermocouple may act as a catalyst for chemical
reactions, and unintentionally induce ignition. The most straightforward method is to
calculate the temperature from the instantaneous measured pressure, assuming the
existence of an adiabatic core in the RCM chamber [18].
When an adiabatic system goes from one state (P1,T1,V1) to another state (P2,T2,V2),
initial and final parameters are related to each other by the isentropic relation (1.13) of
an ideal gas, which can be rewritten as,
V
ln( 1 ) =
V2
T2
1
T2
1
∫ γ − 1 d ln T ,
(2.1)
T1
∫ γ − 1 d ln T = ln P12 ,
P
(2.2)
T1
where γ(T) is the ratio of temperature-dependent heat capacities of the mixture at
constant pressure and constant volume, γ(T)=Cp(T)/Cv(T). The heat capacities used in
this study are taken from [25]. In figure 2.4, the temperature dependence of γ for two
inert gases, N2 and Ar, and the mixture H2/CH4/O2/Ar/N2 (0.95/0.05/1.85/3/4.3) used
in our experiments is illustrated (for a pressure trace for this mixture, see figure 2.3).
The figure shows that γ for Ar is much larger than that for both N2 and the
combustible mixture. Moreover, as expected for a monatomic gas, γ for Ar is constant
while those for N2 and the combustible mixture show a slight dependence upon the
temperature.
36
Chapter 2
Figure 2.4, Temperature dependence of γ for the gas mixtures N2, Ar and
CH4/H2/O2/N2/Ar.
The relations (2.1) and (2.2) show that the pressure and temperature of the gas
only depend on the volumetric ratio and the specific heat capacities and the initial
conditions (Pi, Ti) of the gas mixture. Thus for an ideal RCM, the temperature (Tc)
and pressure (Pc) after compression can easily be calculated from equation (2.1) and
(2.2) by measuring only the initial temperature (Ti) and pressure (Pi), and the
mechanical compression ratio CR, which is defined as the ratio of the initial volume
(Vi) to the final volume (Vc) of the reaction chamber. As an example, the pressures
and temperatures are calculated, based on the mechanical compression ratios that can
be obtained in our RCM, for the three aforementioned gases/mixtures using the
relations (2.1) and (2.2), and presented in figure 2.5a and 2.5b. The calculations show
that the larger γ(T) for Ar, in comparison to N2 and the combustible mixture (figure
2.4), results in a much higher temperature and pressure after compression at identical
compression ratio. In reality, the temperatures and pressures will be lower than those
calculated due to heat losses during compression. As an example, for the compressed
CH4/H2/O2/N2/Ar mixture with CR=22.5 and Pi=0.49 bar, we measured a peak
pressure of ∼32 bar (figure 2.3). However, simple adiabatic calculations show that at
CR=22.5 the calculated temperature is 1050K (figure 2.5a) and Pc/Pi=80 (figure 2.5b),
this results in a calculated compressed pressure of ∼40 bar at Pi=0.49bar.
37
Chapter 2
Figure 2.5a) Temperature calculated as function of compression ratios for different
gas mixtures at Ti=295K, using the relation (2.1). b) Pressure ratio
(Pc/Pi) calculated as function of temperature for different gas mixtures
at initial temperature Ti=295K, using the relation (2.2).
It is more realistic to make the assumption of the existence of an isentropically
compressed core region [11] that is unaffected by heat and mass transfer. In this case,
the relations (2.1) and (2.2) are valid for the pressures, temperatures and volumes in
the adiabatic core. To avoid the difficulty of calculating an effective compression ratio
based on the unknown volume of the core gas within the combustion chamber, the
ratio of measured pressures is used to calculate the temperature of the adiabatic core
gas using equation (2.2). The uncertainty of the calculated core gas temperatures (Tc)
is less than ±3.5K for all measurements. (Appendix A.5).
38
Chapter 2
Several studies [26,17] have indicated that vortices created by the motion of the
piston causes unwanted mixing of cold boundary-layer gas into the compressed core
gas, destroying the adiabatic core. It has been shown, both numerically [19,23,27]
using CFD calculations and experimentally [27] by temperature mapping using the
planar laser-induced fluorescence of acetone, that the incorporation of a specially
designed crevice on the piston head successfully suppresses the vortex formation, and
preserves the well-defined homogeneous core region intact. In this study, the creviced
piston head based upon the best design from the MIT RCM [17,18] was used.
39
Chapter 2
Appendix A.1
Stroke
adjustment
Driving chamber
Speed control
chamber
Piston lock
chamber
Piston
Oil reservoir chamber
Fast acting
valve
Combustion
chamber
Poppet valve
Pressure
transducer
Figure A.1. Cross Sectional view of the Rapid Compression machine (RCM).
40
Chapter 2
Appendix A.2
The numbers in the operation sequence table below are referring to the numbered
parts (valves, reducers etc.) in appendix A.3.
Step
Operation
Manometer Pressure (bar)
Result
1
2
3
4
Close valve 1
Open valve 2
Open Valve 5
Reducer R2
M4
2
Venting air out of
chambers and lines
5
6
7
8
9
10
M4
48
Pressurizing oil drum
M10
70
Pressurizing oil drum
Connected to solenoid
M2
0
Venting driving
chamber
M5/M8
1Æ3
Lifting piston
15
Close valve 2
Reducer R2
Close valve 4
Open valve 7
Reducer R5
3-way valve 4A in up
position
3-way valve 1A in up
position
3-way valve 2A in
down position
Reducer R3
3-way valve 3A in
down position
Reducer R4
M9/M3
7
Lowering fast acting
valve with N2
16
17
Open valve 4
Open solenoid
M7
70
18
19
Open valve 2
3-way valve 3A in up
position
3-way valve 2A in up
position
Open poppet valve
3-way valve 1A in
down position
Reducer R1
Close poppet valve
Close solenoid S1
(close = no current)
Close valve 4
3-way valve 4A in
down position
Check amplifier
Close valve 2
Open solenoid S1
M1
M3
48
0
M5
0
Locking fast acting
valve
Locking piston
Venting N2 fast acting
valve
Drain oil/N2
M6
30
Pressurizing
11
12
13
14
20
21
22
23
24
25
26
27
28
29
30
Reset
Fire the RCM
41
Chapter 2
Appendix A.3
42
Chapter 2
Appendix A.4
Table 3.4 Operating pressures
Operation
System
Operating Pressure (bar)
Piston up
Pneumatic
3.5
FAV*) down
Pneumatic
7
Lock FAV
Hydraulic
70
Lock piston
Hydraulic
48
Driving pressure
Pneumatic
35
*) Fast-acting valve
43
Chapter 2
Appendix A.5 Uncertainty analyses
As described above, the peak temperatures of the core gas at the end of
compression Tc were calculated from the equation (2.2). Since ignition chemistry is
very sensitive to temperature, it is important to estimate the uncertainty in the
calculated peak temperature caused by the uncertainties in the measured parameters.
Assuming that Ti, Pi, Pc and γ are uncorrelated, then the uncertainty in the calculated
peak temperature (ΔTc) can be determined from the following equation:
2
⎛ ∂T
⎞
∂T
∂T
∂T
ΔTc = ⎜
ΔTi +
ΔPi +
ΔPc +
Δγ ⎟ ,
∂Pi
∂Pc
∂γ
⎝ ∂Ti
⎠
(A.2)
As mentioned in 2.2.3,
ΔTi=±0.2K
ΔPi/P=0.5%
ΔPc/P=0.1%
and Δγ=0.04, based on the accuracy of the determined Cp values [25]. Figure A.2
shows the calculated uncertainty as function of the peak pressure (Pc), by using
equation A.2.
44
Chapter 2
Figure A.2. The calculated uncertainty in the peak temperature.
The figure shows that the uncertainty in the peak temperature is better than ±3.5K in
the range of pressures of interest (10-70 bar).
45
Chapter 2
Literature
1.
M. L. Vermeersch, T. J. Held, Y. Stein, F. L. Dryer, SAE paper, 912316
(1991).
2.
T. J. Kim, R. A. Yetter, F. L. Dryer, Proc. Combust. Inst. 25 (1994) 759-766.
3.
A. G. Gaydon, I. R. Hurle, The Shock Tube in High-Temperature Chemical
Physics, Reinhold, New York, 1963.
4.
P. Desgroux, R. Minetti, L. R. Sochet, Combust. Sci. Technol. 113 (1996)
193-203.
5.
M. Carlier, C. Corre, R. Minetti, J.F. Pauwels, M. Ribacour, L. F. Socket,
Proc. Combust. Inst. 23 (1990) 1753-1758.
6.
R. Minetti, M. Ribaucour, M. Carlier, L.R. Sochet, Combust. Sci. Technol.
113-114 (1996) 179-192.
7.
R. Minetti, M. Carlier, M. Ribaucour, E. Therssen, L. R. Sochet, Combust.
Flame. 102 (1995) 298-309.
8.
R. Minetti, M. Carlier, M. Ribaucour, E. Therssen, L. R. Sochet, Proc.
Combust. Inst. 26 (1996) 747-753.
9.
J. F. Griffiths, P. A. Halford-Maw, D.J. Rose, Combust. Flame. 95 (1993)
291-306.
10. P. Beeley, J. F. Griffiths, P. Gray, Combus.t Flame 39 (19980) 269-281.
11. J. F. Griffiths, Q. Jiao, W. Kordylewski, M. Schreiber, J. Meyer, K. F.
Knoche, Combust. Flame, 93 (1993) 303-315.
12. A. Cox, J. F.
Grifiths, C. Mohamed, H.J. Curran, W. J. Pitz, C. K.
Westbrook, Proc. Combust. Inst. 26 (1996) 2685-2692.
13. J. Clarkson, J.F. Grifiths, J. P. Macnamara, B. J. Whitaker, Combust. Flame
125 (2001) 1162-1175.
14. J.F. Grifiths, J.P. Macnamara, C. Mohamed, B.J. Whitaker, J. Pan, C.G.W.
Sheppard, Faraday Discussion, 119 (2001) 287-303.
15. P. Park, J. C. Keck, SAE Paper 900027 (1990).
16. P. Park, ‘Rapid Compression Machine Measurements of Ignition Delays for
Primary Reference Fuels’, Ph.D. Thesis, MIT, 1990.
17. D. Lee, S. Hochgreb, Combust. Flame, 114 (1998) 531-545.
18. D. Lee, ‘Autoignition Measurements and Modeling in a Rapid Compression
Machine’, Ph.D. Thesis, MIT, 1997.
46
Chapter 2
19. D. Lee, S. Hochgreb, J.C. Keck, SAE paper 932755 (1993).
20.
D. Lee, S. Hochgreb, Int. J. Chem. Kin. 30 (1998) 385-406.
21. W. S. Affleck, A. Thomas, Proc. Inst. Mech. Eng. 183 (1969) 365-385.
22. L. Brett, J. Macnamara, P. Musch, J. M. Simmie, Combust. Flame, 124
(2001) 326-329.
23. J. Wurmel, J. M. Simmie, Combust. Flame, 141 (2005) 417-430.
24. M. T. Donovan, X. He, B. T. Zigler, T. R. Palmer, M. S. Wooldridge, A.
Atreya, Combust. Flame 137 (2004) 351-365.
25. B. J. McBride, S. Gordon, M. A. Reno, 'Coefficients for Calculating
Thermodynamic and Transport Properties of Individual Species', NASA
Report TM-4513, October 1993
26. R. J. Tabaczynski, D. P. Hoult, J.C. Keck, J. Fluid Mech., 42 (1970) 249255.
27. G. Mittal, C. J. Sung, Combust. Flame 145 (2006) 160-180.
47
Chapter 3
Chapter 3
High-pressure autoignition delay time
measurements in methane/hydrogen fuel mixtures
in a Rapid Compression Machine
48
Chapter 3
3.1 Introduction
Increasingly stringent regulations regarding CO2 emissions, and the knowledge
that fossil fuel reserves will be exhausted within this century, have called attention to
the possible use of admixtures of hydrogen in natural gas as an alternative fuel in
combustion devices. Experimental results [1] have shown that addition of small
amounts of hydrogen to methane, the principal component of natural gas, enhance the
performance of a gas-powered spark-ignited engine. In the same paper, numerical
simulations were used to indicate that hydrogen addition to methane significantly
increases the tendency to knock at hydrogen fractions larger than 20%. Knocking
combustion in spark-ignited engines is closely related to autoignition of the unburned
end gas, and should be avoided at all cost since it can physically damage the engine
and increase pollutant emissions. Of course, the methane number [2], used to
characterize the knock tendency of natural gases, takes the difference in knock
behavior between methane and hydrogen to define the extremes of the scale, with
pure methane as most knock resistant and pure hydrogen as the least resistant.
Understanding autoignition behavior is also important for designing gas turbines [3]
and Homogeneous Charge Compression Ignition (HCCI) engines [4] Moreover,
autoignition delay times are used as targets for the development and benchmarking of
chemical kinetic models for combustion.
While a large number of studies of the ignition of methane and hydrogen have
been reported, most of them have been conducted under diluted conditions (fuel mole
fraction < 10%) at relatively low pressures (< 5bar). Autoignition studies under
conditions relevant to engines are scarce. Undiluted H2/O2/Ar/N2 [5] and H2/O2/Ar [6]
mixtures have been studied in a rapid compression machine (RCM) at temperatures
ranging from 950 to 1100K and pressures lower than 50 bar. Shock tube
measurements of the autoignition delay times in slightly diluted CH4/O2 mixtures
(fuel + oxidizer ∼30%) at high pressures (40-240 bar) and intermediate temperatures
(1040 - 1500 K) have also been reported [7]. Ignition delay times in CH4/O2/Ar
mixtures using an RCM [8] have been measured at 16 bar between 980 and 1060K.
Ignition delay times obtained in non-diluted lean (ϕ=0.5) methane/air mixtures behind
reflected shock waves between 3 and 450 bar and at temperatures from 1300 to
1700 K [9] were compared with those calculated using the GRI-Mech 3.0 chemical
mechanism [10] and showed good agreement. To our knowledge, only three studies of
49
Chapter 3
autoignition in hydrogen/methane fuel mixtures have been reported [11-13]. In [11]
the influence of small additions of hydrogen (2 and 15% of the fuel by volume) to
highly diluted methane/air mixtures at high temperatures (1500 – 2150 K), moderate
pressures (2 - 10 bar) and equivalence ratios ranging from 0.5 to 2.0 was studied using
shock tubes. A thermal-based promotion theory was proposed to account for the effect
of hydrogen addition. A very extensive shock tube study of hydrogen/methane
mixtures (temperatures and pressures ranging from 800 to 2000 K and from 1 to 3 bar,
respectively) has been reported in [12]. The ignition delay time τ of dilute H2/CH4
mixtures was related to the ignition delay times of pure gases through the empirical
relation
(1− β ) β
τ = τ CH
τH ,
4
where τH
2
(3.1)
2
and τCH
4
are the ignition delay times of hydrogen and methane,
respectively, and β is the mole fraction of hydrogen in the fuel. Recently, the
autoignition delay times of two stoichiometric CH4/H2/air mixtures at pressures from
16 to 40 atm and temperatures between 1000 and 1300 K have been measured in a
shock tube [13]. Because the well-controlled test conditions in a shock tube persist
only for a few milliseconds, the combination of pressure and temperature in this study
was chosen to give ignition delay times <3 ms. Interestingly, while a relatively large
amount of hydrogen was added to the fuel (35%), only a relatively small reduction in
ignition delay time was observed compared to that observed for pure methane. The
experiments also show that the ignition-enhancing effect of hydrogen decreases with
decreasing mixture temperature, and decreases significantly upon increasing pressure
from 16 to 40 atm. The experimental results were compared with calculations
performed using a mechanism [14] that was a modified version of that taken from [7];
the comparison showed substantial disagreement, prompting the authors [13] to
recommend additional experimental and kinetic studies aimed at the autoignition
behavior of methane/hydrogen mixtures.
In this Chapter, we report the autoignition delay times of stoichiometric
methane/hydrogen mixtures using oxygen/nitrogen/argon oxidizers at high pressures
(10 – 70 bar), and temperatures from 950 to 1060 K. The pressures and temperatures
50
Chapter 3
of the unburned mixtures were chosen to give ignition delay times ranging from 2 to
50 ms. The measurements were performed in an RCM and compared with numerical
simulations using different chemical mechanisms, taking into account heat loss
occurring in the period between compression and ignition.
3.2. Experimental approach
The measurements have been performed in an RCM. The RCM construction,
operating procedure, gas handling system and pressure measurements are described in
detail in Chapter 2. The H2/CH4 mixtures containing 0, 5, 10, 20, 50 and 100% of
hydrogen are used as fuel. The compositions of gas mixtures used are given in
Table 3.1.
Table 3.1
Compositions of mixtures used in ignition experiments
Mixture
[H2]
[CH4]
[O2]
[N2]
[Ar]
A
1
0
0.5
0
2.5
B
1
0
0.5
1.05
0.95
C
0.5
0.5
1.25
2.18
2.83
D
0.2
0.8
1.7
2.85
3.95
E
0.1
0.9
1.85
3.07
4.34
F
0.05
0.95
1.93
3.18
4.53
G
0
1
2
3.3
4.7
H
0.5
0.5
2.49
3.36
6.6
The fuel and oxygen concentrations in the mixtures A-G are in stoichiometric
proportions and mixture H is a fuel lean mixture (ϕ=0.5). The total concentration of
diluting inert gases are close to that of nitrogen in air, while the N2/Ar ratio is chosen
to provide similar temperatures after compression for all fuels. For comparison with
the results of a previous RCM study [6], the measurements are also performed for
pure hydrogen without nitrogen (flame A in Table 3.1). As mentioned above, the
pressures were varied between 10 and 70 bar, and temperatures in the range 9501060 K. In addition, a substantial number of measurements were performed along an
isotherm at a peak temperature after compression of 995 ± 4 K between ~25 and ~65
51
Chapter 3
bar (see below), allowing an examination of the pressure dependence of ignition. For
measurements under identical conditions (composition, initial/final pressure), the
reproducibility of the measured ignition delay times is ∼5% and the uncertainty in
deriving the ignition delay time from the measurements is ∼0.3ms. For the data taken
along the isotherm, the variation in temperature (± 4 K) causes a scatter in the results
of ~10-20%.
3.3. Numerical simulation and analysis of experimental data
3.3.1. Chemical mechanisms
In this work we compare different chemical mechanisms for the calculation of
the ignition delay times. In the discussion of the mechanisms it will be convenient to
refer to them either by acronym (e.g., GRI-Mech) or by author. While one does not a
priori anticipate good performance from the GRI-Mech 3.0 chemical mechanism [10]
since it was optimized to model natural gas combustion over the ranges 1000 2500 K, 0.013 – 10 atm and equivalence ratios from 0.1 to 1.5, the large popularity of
this mechanism compels us to evaluate its predictive power under the experimental
conditions discussed here. Better predictions at high pressures can be expected from
the RAMEC mechanism[7], which includes 190-reactions involving 38 species based
on the GRI-Mech 1.2 mechanism [15], with additional reactions important in methane
oxidation at lower temperature. This mechanism emerged from the kinetic study
based on the high-pressure shock tube measurements referred to above [7], and
showed the increased importance of reactions involving HO2, CH3O2 and H2O2 at high
pressures and low temperatures. The comprehensive chemical mechanism for
methane oxidation, developed at the University of Leeds [16], is also used in the
present study. This mechanism consists of 351 chemical reactions between 37 species,
and is built on the same experimental base as GRI-Mech 3.0. Recently, two revised
mechanisms for hydrogen oxidation have been reported. A mechanism for hydrogen
oxidation consisting of 19 reversible elementary reactions has been developed by
O’Connaire et al. [17] and evaluated for temperatures ranging from 298 to 2700K,
pressures from 0.05-87 bar and equivalence ratios in the range from 0.2-6. The
hydrogen mechanism developed by Li et al. [18] is close to that of O’Connaire et al.
52
Chapter 3
and also includes 19 chemical reactions. The mechanism of O’Connaire is included in
the comprehensive kinetic model of methane/propane oxidation of Petersen et al.[19],
which consists of 663 chemical reactions between 118 species. In this mechanism, the
methane oxidation chemistry incorporates recent theoretical and experimental data for
the reaction rates. Clearly, the assumption is that a mechanism that performs well for
both pure hydrogen and pure methane will adequately describe H2/CH4 mixtures as
well.
3.3.2. Numerical simulations
As mentioned in Chapter 1, the system of governing equations describing the
time evolution of the adiabatic core is not closed. Therefore, for meaningful
comparison between measurements and numerical simulations, the experiment should
provide sufficient information to specify the system for the simulation. In RCM
experiments, pressure can be measured relatively easily with a high degree of
accuracy, and is generally used for this purpose. To our knowledge, all efforts thus far
have been directed to estimating the specific volume (equation 1.14) from the pressure
trace. As was mentioned in Chapter 2, estimating the specific volume of the adiabatic
core based on the geometrical size of the combustion chamber is very inaccurate. A
more accurate approach is to calculate the specific volume directly after compression
assuming that the gas mixture is chemically inert (no heat release) and neglecting heat
losses between compression and ignition. For ignition delay times that are longer than
the time of compression, neglecting heat release from chemical reactions is a
reasonable assumption during compression. However, significant cooling of the
compressed gas mixture before ignition (as illustrated in Figure 2.3) is expected under
typical RCM conditions. To overcome this difficulty, the system of governing
equations (1.6)-(1.9) is supplemented by additional equations taking into account heat
and mass exchange in the RCM [20,21]. Unfortunately, the complex geometry of the
RCM requires many assumptions, and the implementation of this approach ultimately
requires the use of parameters derived from fitting the pressure trace obtained in an
inert gas mixture. In an alternative approach [5], the specific volume is calculated
from the measured pressure trace in an inert mixture with the same values of heat
capacity, initial pressure and temperature as those in the reactive mixture under
investigation. Clearly, this method accurately predicts the specific volume of the
53
Chapter 3
reactive mixture during the initial stage of the ignition process, when heat release is
marginal. Just prior to ignition, the specific volume can be substantially different from
that of the inert mixture as result of the pressure and temperature increase in the
adiabatic core due to chemical reactions. The error arising from the neglect of heat
release in the estimate of the specific volume on the calculated ignition delay time is
small if the duration of this phase is also small. We anticipate that this effect will be
larger for complex alkanes showing multistage ignition [21] than for the simple fuels
used here, for which only negligible heat release occurs prior to ignition.
In light of these considerations, in the present work we determine the specific
volume from the measured pressure in the period between compression and ignition,
and extrapolate the time dependence derived in this fashion to the region in which
substantial heat release begins. This method bypasses time consuming measurements
in inert gas mixtures, and, as will be seen below, yields fits to the pressure trace that is
on par with those obtained by the other methods [5,21]. While our approach is also
not free from the potential uncertainties due to heat release during ignition, the good
fit to the pressure trace indicates an accuracy of the same order that obtained by
determining the specific volume in the inert mixture.
The implementation of the method is illustrated in figure 3.1, where the
measured pressure trace in mixture B (Table 3.1) is presented together with the
specific volume derived from the pressure. The pressure in the interval from tmax
(point during compression at which the pressure reaches its maximum, Pmax) to tA (at
which heat release is still negligible) is approximated by an exponential time
dependence with a characteristic time τc:
−(τ −τ max) /τ c
+ P0 ,
P(t ) = ( Pmax − P0 ) exp
(3.2)
where P0 is the pressure in the combustion chamber at room temperature. The specific
volume, v(t), is calculated from this relation assuming adiabatic expansion of the
“inert” mixture in the core
1/ γ
⎛ P(t ) ⎞
⎟⎟
v(t ) = v max ⎜⎜
,
⎝ Pmax ⎠
54
(3.3)
Chapter 3
where vmax is the specific volume at tmax, and γ is Cp/Cv; v(t) used as an input in
SENKIN for all times t > tmax. The choice of tA is important for the accuracy of this
method. If tA is chosen too large, heat release due to chemical reactions will be
sufficient to substantially slow the pressure drop by heat loss, and the specific volume
thus derived will be inaccurate. To avoid this error, we determine tA using an iterative
procedure. In first approximation, tA is chosen to lie halfway between tmax and the time
of ignition. If the SENKIN calculations using v(t) yield a temperature T(tA), which is
2 K higher than that for the adiabatic core calculated from the measured pressure
trace, tA is decreased (shifted towards tmax); if T(tA) is within 2 K, tA is increased. The
iterations continue until increasing tA increases the temperature by more than 2 K.
Usually, 3 – 4 iterations were performed when analyzing the experimental results. It
should be pointed out that this iteration procedure is coupled to the specific chemical
mechanism being used in the simulations, which itself is the subject of investigation.
Therefore, a computed ignition delay time that is too short by more than a factor of
two or three can result in a value for tA that is too short for a reliable extrapolation.
Fortunately, the chemical mechanisms used here predict the ignition delay with
sufficient accuracy (within a factor of two) to implement the iteration procedure, and
the temporal profiles of specific volume used in the simulations are independent of the
choice of tA.
55
Chapter 3
Figure 3.1. Pressure (a) and specific volume (b) traces in mixture B (Table 3.1) at initial
temperature of 995 K. Curves 1, 2 and 3 denote the results from the measurements, the
calculations assuming constant specific volume after compression, and the calculations using
the specific volume derived from the measured pressure trace, respectively.
The simulated pressure trace using the mechanism of O’Connaire et al. [17] are
presented in figure 3.1. Here we see that the simulated trace agrees with the
experimental trace as well as that reported by the other methods [5,21]. For
comparison, the pressure trace calculated assuming negligible heat loss (specific
volume is constant after compression) is also shown. As can be seen, neglecting heat
loss in the RCM can lead to the (erroneous) conclusion that this chemical mechanism
underpredicts the ignition delay time. However, the more realistic input of an
increasing specific volume shows opposite result – the mechanism actually
overpredicts the ignition delay time.
3.4. Results and discussion
To assess the quality of the experimental data, we compare the ignition delay
times obtained here with the results of previous RCM studies of the autoignition of
pure hydrogen [5,6]. The ignition delay times from different data sets are scaled by
56
Chapter 3
the oxygen number density (mol/cm3) at the peak pressure after compression [6], and
are presented in figure 3.2 as a function of the reciprocal peak temperature after
compression. As mentioned above, the measured ignition delay time depends
substantially upon the heat losses in the RCM combustion chamber. Based upon the
simulations, we estimate that heat loss in our RCM results in as much as a 35%
increase in the ignition delay time as compared with an ideal adiabatic RCM. As can
be seen from figure 3.2, all experimental results are within an interval of ±35% of our
measurements. Taking into account the assumption of the validity of the scaling
method, and that heat loss can vary significantly between the physically different
machines, we consider the agreement of the results obtained here and the data in the
literature to be excellent.
Figure 3.2. Scaled ignition delay times in pure hydrogen fuel as a function of
reciprocal temperature after compression. Solid lines denote ±35% interval around
the measurements in present work approximated by the relation (3.4), see text.
As can be seen from figure 3.2, the ignition delay time in the pure hydrogen fuel
is exponential function of the reciprocal temperature. The same dependence is
observed for pure methane. Incorporating the pressure dependence using the power
function for number density, we obtain an Arrhenius-like empirical relation for the
functional dependence of the ignition delay time upon pressure and temperature after
compression (Pc and Tc, respectively) for pure hydrogen and methane fuels:
57
Chapter 3
⎛P
τ = A⎜ c
⎜T
⎝ c
n
⎞
⎛ E ⎞
⎟ exp⎜ a ⎟ .
⎟
⎜ RT ⎟
⎠
⎝ c⎠
(3.4)
The magnitudes of A, n, and Ea (units: s, Pa, mole, kJ, K) derived for the
stoichiometric mixtures of hydrogen and methane with oxygen are given Table 3.2.
Table 3.2
Fit coefficients
A
Ea
n
H2
2.82E-13
336
-1.3
CH4
3.23E-2
192
-2.1
It should be pointed out that the negative value of the power n for both gases means
decreasing ignition delay time with increasing pressure. We remark in passing that the
apparent activation energy for hydrogen is in excellent agreement with that observed
in Ref. [6]; while for methane Ea is significantly lower than that obtained in recent
studies [7,14].
To assess whether the recommended mixing expression (3.1) can also be used
for methane/hydrogen mixtures studied in the RCM, ignition delay times have been
measured at constant peak pressure (Pc = 33.5 ± 1 bar) and peak temperature
(Tc = 995 ± 4 K) as a function of hydrogen mole fractions in the fuel. As can be seen
from the results, presented in figure 3.3, replacing methane by hydrogen decreases the
ignition delay time, as reported in shock tube studies [11-13]. Moreover, within the
limits of the experimental uncertainty, the logarithm of the ignition delay time appears
to be linear function of the hydrogen mole fraction, suggesting the utility of mixing
expression (3.1). Anticipating the discussion below, we note that the computed
ignition delay times, using the mechanism from Ref. [19] and accounting for heat loss
as described in Section 3.3.2, predict this trend within the limits of experimental error.
For further analysis, we rewrite the mixing expression (3.1) by using Equation (3.4)
for the ignition delay time in pure fuel:
58
Chapter 3
(1 − β ) ⎛ Pc ⎞
τ = AHβ . ACH
⎜ ⎟
2
4
⎜T ⎟
⎝ c⎠
nH 2 β + nCH 4 (1 − β )
⎛ EH 2 β + ECH 4 (1 − β ) ⎞
exp ⎜⎜
⎟⎟ . (3.5)
RT
⎝
⎠
From this relation, at fixed hydrogen mole fraction we expect a linear dependence of
the logarithm of ignition time divided by number density to the power
n = nΗ2β + nCH4(1-β) upon the reciprocal of the temperature. As can be seen from
figure 3.4, which shows the results for different hydrogen mole fractions, the mixing
relation (3.5) approximates the experimental data very well. For the mixtures with H2
content ≤ 20% the effect of hydrogen addition on the ignition delay time is relatively
small, but becomes substantial when the hydrogen fraction is more than 50%. It is
interesting to note that the slope of the lines in figure 3.4 increases with increasing
hydrogen content in the mixture, reflecting the differences in the “overall” activation
energy Ea between the two pure fuels; for hydrogen Ea is two times larger than for
methane (Table 3.2). Thus, at high temperatures, effect of the added hydrogen on the
ignition delay time is more pronounced than at low temperatures, as also observed in
[13].
Figure 3.3. Measured and calculated ignition delay times versus hydrogen mole
fraction in fuel at Pc = 33.5 ± 1 bar and Tc = 995 ± 3 K). The simulations were
performed accounting for heat loss using the mechanism of Petersen et al. [19]. The
solid line is obtained from the mixing relation Equation 3.5, see text.
59
Chapter 3
Figure 3.4. Measured ignition delay times scaled according to equation 3.4 as a
function of reciprocal temperature (symbols), and the calculated results using the
mixing relation (3.5) (lines).
As can be seen in figure 3.4, all measurements obtained along the isotherm at peak
compression temperature ~995 K and fixed hydrogen mole fraction collapse to a small
cluster, which demonstrates that equation (3.5) correctly predicts the pressure
dependence of the ignition delay time. Presenting the isotherm data in figure 3.5 on a
linear scale, as a function of pressure for different volume fractions of hydrogen, we
observe some scatter around the lines from equation (3.5), which is caused by day-today variations in temperature (± 4 K) in the measurements, as mentioned in section
3.2. As observed above, the ignition delay times decrease with increasing pressure for
all hydrogen fractions measured, extending the observations of the recent shock tube
study [13].
60
Chapter 3
Figure 3.5. Measured (symbols) and calculated (lines) autoignition delay times as a
function of pressure at fixed peak compression temperature Tc=995 ± 4 K and
different hydrogen mole fractions in fuel. The calculated curves were obtained using
mixing relation (3.5),
3.5. Comparison of experimental results with numerical simulations
Figures 3.6 to 3.8 show the experimental and calculated autoignition delay times
using different chemical mechanisms. To avoid clutter in the figures, the simulated
data are presented as polynomial trend lines through the calculated points. The
measured and calculated ignition delay times are presented in two sets. The first set
(figures 3.6a-3.8a) includes logarithms of the ignition delay times scaled by (P/T)n as
a function of the reciprocal temperature to eliminate the density dependence and
highlight the expected Arrhenius behavior with temperature. The second set of figures
presents the ignition delay times measured along the 995 ± 4 K compression isotherm.
61
Chapter 3
Figure 3.6. Measured (diamonds) and calculated (lines) autoignition delay times for
pure hydrogen (mixture B in Table 3.1). (a) Scaled delay times vs. reciprocal
temperature; (b) delay time vs. pressure at fixed temperature Tc = 995 ± 4 K.
The results for pure hydrogen (mixture B, Table 3.1) are presented in Fig 3.6. As
can be seen, the calculations using the mechanism from Petersen et al. show excellent
agreement with the measurements over the entire range of pressure and temperature
studied, while the mechanism of Li et al. and the Leeds and RAMEC mechanism
systematically overpredict the measured ignition delay times. Calculations using the
62
Chapter 3
Leeds mechanism and GRI-Mech 3.0 (not shown in figure 3.6) give identical results
for all experimental conditions, which considering their similarity is perhaps not
surprising. Based on the decidedly better agreement obtained using Petersen et al. we
suggest the use of this mechanism for ignition delay studies of hydrogen combustion
under gas turbine conditions, similar to the recommendation made in a recent study
[22].
Figure 3.7 presents the measured and calculated ignition delay times for pure
methane (mixture G, Table 3.1). As can be seen from figure 3.7a, the calculations
using the Leeds and Petersen et al. mechanisms are in excellent agreement with the
experimental results for all conditions measured. The predictions of the RAMEC
mechanism are in reasonable agreement with the experiments at the low temperatures
but substantially underpredict (up to a factor of two) the ignition delay times at high
temperatures. The results of the calculations with GRI-Mech 3.0 are more than a
factor of two higher than the scaled measured ignition delay times for all data in
figure 3.7a. The unscaled data for ignition delay times along the isotherm at
Tc = 995 K presented in figure 3.7b also show excellent agreement between
measurements and calculations with the Leeds and Petersen et al. mechanisms. At this
temperature the RAMEC mechanism slightly underpredicts the measurements at
pressures below 45 bar, but improves with increasing pressure. GRI-Mech 3.0, while
following the experimental trend well, substantially overpredicts the ignition delay in
this range.
63
Chapter 3
Figure 3.7. Measured (diamonds) and calculated (lines) ignition delay times for pure
methane (mixture G in Table 3.1). (a) Scaled delay times vs. reciprocal temperature;
(b) delay time vs. pressure at fixed temperature Tc = 995 ± 4 K.
As observed above, the mechanism proposed by Petersen et al. predicts the ignition
delay time in both pure hydrogen and methane very well, while the predictions of the
other mechanisms considered are poorer. Furthermore, this mechanism also yields the
best agreement with the experiments performed on the hydrogen/methane mixtures.
Consequently, in the following comparisons we only show these computational
results. As can be seen from figure 3.8, the computed ignition delay times are in
64
Chapter 3
excellent agreement with the experiments in the hydrogen/methane mixtures. Over the
entire range of pressure and temperature studied, the agreement between the
calculations and measurements is better than 25%.
Figure 3.8. Measured (symbols) and calculated (lines) autoignition delay times for
hydrogen/methane fuel (mixtures D, E and F in Table 3.1) as a function of reciprocal
temperature (a) and pressure at fixed temperature Tc = 995 ± 4 K (b). The
calculations were performed using Petersen et al.[19].
65
Chapter 3
Given the agreement with the experimental results presented here, we also
compare the predictions of this mechanism with the experimental data reported in Ref.
[13], where a significant disparity between the experimental and numerical results
was observed. Figure 3.9 reproduces the experimental data for the hydrogen/methane
mixtures from Ref. [13], and the computations using Petersen et al.; as was done in
the original report [13], we have simulated the results at constant density. At 16 bar
(figure 3.9a), the current mechanism gives a better reproduction of the experimental
data at T > 1150K, the computations for 15 and 35% now bracketing the experiments;
in this region, the agreement is substantially better than a factor of 2. However, at
lower temperature the agreement is poorer than in the original model [13]. At 40 bar
(figure 3.9b), we observe a similar trend, albeit with a similar agreement to the
original model at lower temperatures. Particularly vexing is the relatively large effect
for the addition of hydrogen predicted by the computations, but which is apparently
much less manifest in the shock tube results, even at high temperature. We remark
that the good predictive power shown in figure 3.8, particularly in reproducing the
trends with pressure and hydrogen addition as well as the magnitude of the results in
the low temperature region, supports the use of the mechanism under these conditions.
66
Chapter 3
Figure 3.9. Comparison of measured (points) and calculated (lines) ignition delay
times for shock tube measurements taken from Ref. [13]. Calculations performed
using mechanism from Petersen et al. [19]; dashed lines calculations at 15% H2, solid
lines calculations at 35% H2 in mixture. Figure (a) at 16 bar, figure (b) at 40 bar.
While the discussion above has focused entirely on stoichiometric mixtures, we
have also obtained results under lean conditions (ϕ = 0.5), for the 50/50
hydrogen/methane mixture. The composition used in these measurements is shown in
Table 3.1 as mixture H. The ignition delay times measured along the 995 K isotherm
are presented in figure 3.10, together with the stoichiometric results. Interestingly, we
see no change in the measured delay times between the lean and stoichiometric
mixtures within the experimental uncertainty (dominated by the ± 4 K temperature
uncertainty discussed above). In addition, as seen in the previous figures, the
computational results using Petersen et al. predict both trends and magnitude of the
results excellently. Under the conditions of the experiments the predicted differences
for the two equivalence ratios is less than 1 millisecond. We are currently extending
these measurements to other hydrogen/methane ratios.
67
Chapter 3
Figure 3.10. Measured (symbols) and calculated (lines) ignition delay times for 50%
H2 in the fuel at ϕ = 1.0 and ϕ = 0.5 (Mixtures C and H, respectively, in Table 3.1) as
a function of pressure at fixed temperature Tc = 995 ± 4 K. The calculations were
performed using Petersen et al. [19].
3.6 Summary and Conclusions
Autoignition delay times of methane/hydrogen mixtures at high pressure (1070 bar) and moderate temperatures (960 – 1060 K) have been measured in a rapid
compression machine. The experimental results obtained under stoichiometric
conditions show that replacing methane by hydrogen reduces the measured ignition
delay time. Both measured and computed ignition delay times in the fuel mixtures are
shown to be related quantitatively to the hydrogen mole fraction in fuel according to
the mixing relation proposed in the literature [12]. At low mole fractions (≤20%),
hydrogen addition has a modest effect on the measured ignition time under the
experimental conditions presented here. At 50% hydrogen mole fraction in fuel a
substantial reduction in ignition delay time is observed. The measurements show that
the effects of hydrogen in promoting ignition increases with temperature and
decreases with pressure. Interestingly, results for 50% hydrogen in the fuel at
ϕ = 0.5 are essentially identical to those at ϕ = 1.0. These results suggest that the
adverse affects of hydrogen addition to natural gas on engine knock may be limited
for hydrogen fractions of only a few tens of percent. Very good agreement between
the measurements and calculations using the mechanism proposed by Petersen et al.
68
Chapter 3
[19] is observed for all fuel mixtures studied. Over the entire operational range of
temperatures and pressures used in the present study, the differences between the
measured and calculated values of the ignition delay time are less than 10% for pure
fuels and better than 25% for the hydrogen/methane mixtures.
69
Chapter 3
Literature
1.
G. A. Karim, Int. J. hydrogen energy 28 (5) (2003) 569-577.
2.
M. Leiker, W. Cartelliere, H. Christoph, U. Pfeifer, M. Rankl, ASME paper
72-DGP-4, April 1972.
3.
J. Y. Ren, F. N. Egolfopoulos and T. T. Tsotsis, Combust. Sci. Technol 174
(4) (2002) 181-205.
4.
M. Richter, R. Collin, J. Nygren, M. Alden, L. Hildingsson and B.
Johansson, JSME Int. J. Series B-Fluids and Thermal Engineering 48 (4)
(2005) 701-707.
5.
G. Mittal, C. J. Sung and R. A. Yetter, Int. J. Chem. Kin. 38 (8) (2006) 516529.
6.
D. Lee and S. Hochgreb, Int. J. Chem. Kin. 30 (6) (1998) 385-406.
7.
E. L. Petersen, D. F. Davidson and R. K. Hanson, Combust. Flame 117 (1-2)
(1999) 272-290.
8.
L. Brett, J. MacNamara, P. Musch and J. M. Simmie, Combust. Flame 124
(1-2) (2001) 326-329.
9.
V. P. Zhukov, V.A. Sechenov and A.Y. Starikovskii, Combustion Explosion
and Shock Waves 39 (5) (2003) 487-495.
10. G. P. Smith, D. M. Golden, M. Frenklach, N. W. Moriarty, B. Eiteneer, M.
Goldenberg, C. T. Bowman, R. K. Hanson, S. Song, W. C. Gardiner, V.
Lissanski and Z. Qin, http://www.me.berkeley.edu/gri_mech/..
11. A. Lifshitz, K. Scheller, A. Burcat and G. B. Skinner, Combust. Flame 16 (3)
(1971) 311-321
12. R. K. Cheng and A. K. Oppenheim, Combust. Flame 58 (2) (1984) 125-139.
13. J. Huang, W. K. Bushe, P. G. Hill and S. R. Munshi, Int. J. Chem. Kin. 38 (4)
(2006) 221-233.
14. J. Huang, P. G. Hill, W. K. Bushe and S. R. Munshi, Combust. Flame 136
(1-2) (2004) 25-42.
15. C. T. Bowman, R. K. Hanson, D. F. Davidson, W. C. Gardiner, V. Lissanski,
G. P. Smith, D. M. Golden, M. Frenklach and M. Goldenberg,
http://www.me.berkeley.edu/gri_mech/.
16. K. J. Hughes, T. Turanyi, A. R. Clague and M. J. Pilling, Int. J. Chem. Kin.
33 (9) (2001) 513-538.
70
Chapter 3
17. O. Conaire, H. J. Curran, J. M. Simmie, W. J. Pitz, C. K. Westbrook, Int. J.
Chem. Kin. 36 (11) (2004) 603-622.
18. J. Li, Z. W. Zhao, A. Kazakov and F. L. Dryer, Int. J. Chem. Kin. 36 (10)
(2004) 566-575.
19. E. L. Petersen, D. M. Kalitan, S. Simmons, G. Bourgue, H. J. Curran and J.
M. Simmie, Proc. Combust . Inst .31 (2007) 447-454.
20. D. Lee and S. Hochgreb, Combust. Flame 114 (3-4) (1998) 531-545.
21. S. Tanaka, F. Ayala and J. C. Keck, Combust. Flame 133 (4) (2003) 467481.
22. J. Strohle and T. Myhrvold, Int. J. hydrogen energy 32 (2007) 125-135.
71
Chapter 4
Chapter 4
One-dimensional laminar flames
Experimental Techniques, Procedures and Burner
Setup
72
Chapter 4
4.1 General introduction
In this chapter, the experimental procedures for the measurement of temperature
and concentrations of HCN and C2H2 in one-dimensional flames are described.
The flame temperature is measured using Coherent Anti-Stokes Raman
Scattering (CARS) for N2 thermometry. This in-situ technique has been demonstrated
to provide an accuracy of few tens degrees with high spatial and temporal resolution,
see for example [1,2], and is one of the best methods for flame thermometry [3]. The
experimental setup and measurement procedure was essentially identical to that
reported elsewhere [4].
In contrast to the “standard” method of CARS for temperature measurements,
the low concentrations of HCN and C2H2 in flames are more difficult to measure.
Various methods have been developed to detect HCN and C2H2 in combustion
systems, among them intrusive techniques (extractive probe sampling) and in-situ
laser diagnostic techniques. Extractive probe sampling and subsequent analysis of the
sample, using infrared spectroscopy or mass spectrometry, offer the advantage of
flexibility and relatively simple and accurate methods to analyze the flame samples.
One of the drawbacks of extractive probe sampling in flames is the potential distortion
of species profiles [5]. This problem can be avoided by the use of in-situ laser
diagnostic approaches such as laser-induced fluorescence (LIF), spontaneous Raman
scattering, direct absorption spectroscopy and many others. Each of these techniques
has its strengths and weaknesses. Spontaneous Raman scattering and LIF have both
been successful at detecting C2H2 in flames without significant interference from
other flame species [6,7]. However, the current state of these techniques is such that
they are not yet generally applicable for measurements under flame conditions. The
sensitivity of the Raman technique is low [6] and therefore only applicable in
environments of high C2H2 concentration. Measurement of C2H2 using LIF also
suffers from low signal levels and requires further research to quantify the measured
signal [7]. Although acetylene absorbs strongly in the infrared (IR) spectral range,
selective detection using a direct absorption technique is difficult, due to interference
from other flame molecules such as CO2, CO, and H2O. As an illustration of possible
interferences in the IR spectral range, figure 4.1a shows a part of the calculated
absorption spectrum of a gas mixture of acetylene and water at typical flame
conditions, using data from refs. [8-11]. The figure shows that the acetylene
73
Chapter 4
absorption lines are impossible to resolve due to the large number of strong and
overlapping water lines.
Figure 4.1. Calculated absorption spectra for a gas mixture of 0.3% C2H2 and 18%
H2O in N2 at 1800K (a) and 300K (b). Line positions, strengths and broadening
coefficients are taken from [8-10] and [11], respectively.
The in-situ detection of low concentrations of HCN in flames is difficult as well,
since it only absorbs strongly in the infrared spectral range. To our knowledge, the
only in-situ technique used to measure HCN in a low pressure flame is fiber laser
74
Chapter 4
intracavity absorption spectroscopy (FLICAS) [12], applied in the IR-region around
1.5 μm. The flame study was doped with NH3, which increased the HCN
concentration in the flame substantially. The concentration of native HCN in
hydrocarbon flames is much lower, and, similar to acetylene, the absorption lines of
HCN in these flames will be difficult to distinguish from other flame molecules.
Consequently, the concentrations HCN and C2H2 in the flames we are interested in are
too low to detect accurately using in-situ absorption techniques. At room temperature,
where the HCN and C2H2 absorption features are much stronger and H2O and other
interfering lines are less intense, this results in several HCN and C2H2 absorption
peaks with negligible interference. This is illustrated for the gas mixture of C2H2
(0.3%) and H2O (18%) at room temperature in figure 4.1b. To take advantage of this
situation, we thus developed an extractive probe sampling system with room
temperature analyses, using tunable diode laser absorption spectroscopy (TDLAS) at
∼1.5 μm to measure concentrations of HCN and C2H2. We analyze probe
perturbations to minimize and correct for distortion of the species profile arising from
the presence of the probe sampling system in the flame.
75
Chapter 4
4.2 Burner
The measurements described in this thesis were performed in one-dimensional,
atmospheric-pressure premixed flames stabilized above a McKenna products burner
of 60 mm diameter, shown schematically in figure 4.2. The burner consists of a
sintered bronze plug, which serves as a flame holder, and is surrounded by another
sintered section for a shroud gas. In this study a nitrogen shroud was used to prevent
air entrainment in the combustion products. A cylindrical chimney with a 60 mm
inner diameter was positioned approximately 30 mm above the burner to stabilize the
column of post-flame gases. The burner surface is cooled by water flowing through
coils imbedded in the sintered flame holder.
Figure 4.2. Schematic of the McKenna Products burner.
The burner was affixed to a positioner that moves the burner with a precision of
0.1 mm, to allow measurement of axial profiles of species concentration and
temperature.
76
Chapter 4
4.3 Gas handling system
Figure 4.3. Experimental set-up used for transport, mixing, and analysis of the fuelair mixture.
Figure 4.3 shows the flow scheme for the burner feed gases and the system for
analyzing the composition of the unburned fuel-air mixture. The flow rates of all
gases were measured by calibrated mass-flow meters (Bronkhorst, EL-FLOW),
digitized by an analog-digital converter and processed by a PC. The flow ranges of the
meters were selected to provide an accuracy of better than 5%. All fuels used in this
study were supplied in cylinders with purity better than 99.99% and dry, filtered air
was supplied by an oil-free compressor. The gases were mixed in a tube 50 cm long,
and a small part was diverted to a Maihak Unior 610 gas analyzer for measuring the
methane and oxygen mole fractions in the fuel-air mixture, while the rest is supplied
to the burner system. The by-pass burner makes it possible to vary the flow rate
through the burner during the experiment without changing the choked flows of the
individual gases.
The equivalence ratio ϕ of the fuel-oxidizer mixtures was calculated according
the following expression (described in detail in chapter 1),
77
Chapter 4
ϕ=
[CH 4 ] 1
[H ] 1
.
.
+ 2 .
[O2 ] f st ,CH 4 [O2 ] f st , H 2
(4.1)
Whereas it is possible to determine the equivalence ratio through measuring the flow
rates, to increase accuracy we determined ϕ based on measuring CH4 and O2 mole
fractions in the unburned mixture using the Maihak analyzer. In CH4/air flames,
measuring only the CH4 mole fraction in the cold mixture is sufficient to calculate the
complete mixture composition, since the oxygen/nitrogen ratio in the mixture known.
This procedure provided accuracy better than 2% for all equivalence ratios used. To
determine the composition of the cold unburned mixtures of the CH4/H2/air flames,
the CH4 and O2 mole fractions are both measured using the Maihak gas analyzer, the
N2 mole fraction is calculated from the measured O2 mole fraction and the known
ratio [O2]/[N2] in air, and the H2 mole fraction is calculated from the balance,
⎛ [N ] ⎞
[ H 2 ] = 1 − [CH 4 ] meas − [O2 ]meas. − [O2 ]meas. * ⎜⎜ 2 ⎟⎟ .
⎝ [O2 ] ⎠ air
(4.2)
Based on this method, the accuracy of the equivalence ratios determined was better
than 5% in all CH4/H2/air flames measured. Detailed information on the uncertainty
analyses can be found in [13]. The mass flux is calculated from the measured mass
flow, assuming the surface area of the burner is equal to that of the flame. The
accuracy of the mass flux through the burner surface was estimated approximately
10%, determined by the uncertainties in the measured mass flow rate and the
uncertainty in the flame area.
78
Chapter 4
4.4 Extractive-probe sampling system
Figure 4.4. Schematic of the microprobe sampling system.
Figure 4.4 shows a schematic of the probe sampling system. All flames were
sampled by a quartz microprobe having a similar design to that described in [14], with
an orifice diameter of about 100 μm. For the acetylene measurements, with exception
of the probe tip, the quartz probe was cooled over a distance of 35 cm by water at
12οC. After passing an ice-cooled water trap, the sampled gas flowed through a 1 mlength stainless-steel tube sealed on both ends by quartz windows to form a
measurement cell. The pressure in the tube was monitored by an electronic pressure
transducer and kept constant at 100 mbar by a vacuum pump installed in the exit of
the sampling system; this provided rapid removal and quenching of the gas sample.
Because HCN is soluble in water, measurement of this component was performed
using an uncooled probe and water at room temperature in the cooling trap. In this
case no condensation of water from the combustion gases in the cooling trap was
observed during sampling.
79
Chapter 4
4.5 Estimate of the conversion of C2H2 and HCN during sampling
To estimate the degree of conversion of C2H2 and HCN to other products during
sampling, the probe is modeled as a closed system containing a homogeneous gas
mixture, described in detail in chapter 1. The model uses a prescribed cooling rate,
based on characteristic time scales in the experimental set-up to calculate the mixture
composition history profile. The probe modeling was divided in two parts: rapid
adiabatic expansion in the probe tip (τ1) followed by cooling of the sample at constant
pressure (τ2). During expansion in the probe tip the gas sample undergoes a rapid
decrease in pressure and temperature, described by,
P = ( Pmax . − P0. ) exp
−t / τ
1
.
(4.3)
The temperature drop associated with the pressure drop is calculated in the program
using equation (2.2), by assuming isentropic flow conditions. The characteristic
cooling time in the probe tip used was τ1 =10-4 s, which is one order longer than
estimated by assuming critical flow in the probe orifice. The initial composition and
temperature were varied in the model using the results of one-dimensional flame
calculations, described in detail in chapter 1. The temperature (T0) of the gas flow in
the probe immediately after expansion is taken from the temperature calculations in
the probe tip at a probe backpressure (P0) of 0.1 bar and put in the model. Further
downstream of the probe tip the gas is cooled at constant pressure (P0=0.1 bar) to
room temperature (Tmin) by heat transfer to the coolant, according to,
T = (T0 − Tmin ) exp −t / τ 2 + Tmin .
(4.4)
In the model we used τ2=0.5s; this value is one order higher than that based on heat
transfer estimations. GRI-Mech 3.0 [15] is used as chemical mechanism in the
calculations.
80
Chapter 4
Figure 4.5. a) Calculated temperature and concentration curves during expansion in
the probe tip, using τ1=10-4s model. The initial temperature and composition are
taken from a simulated adiabatic (free) premixed methane/air flame with φ=1.4 and
mass flux 0.014g/cm2s, at 3mm from the burner surface. b) Calculated temperature
and concentration curves downstream the probe tip (P=0.1 bar), using τ2=0.5s.
As an example, figure 4.5a presents the predicted C2H2 conversion during
expansion in the probe tip and 4.5b shows the predicted C2H2 conversion further
downstream the probe tip during sampling, for the case in which the conversion was
maximal. The calculations show that in the present experimental setup, for all flames
under investigation, the conversion of C2H2 in the probe is less than 15%, and less
than 10% for HCN, at axial distances greater than 2.5 mm from the burner surface.
We shall return to complications of sampling in Chapter 5.
81
Chapter 4
4.6 Laser absorption spectroscopy
4.6.1 Theory
If laser radiation of intensity I0(v) at frequency v passes through an absorbing
medium over a length l, the intensity of the transmitted laser beam I(v) can be
expresses according to the Lambert-Beer law:
⎛ I (v ) ⎞
ln⎜⎜ 0 ⎟⎟ = α (v)l ,
⎝ I (v ) ⎠
(4.5)
where α(v) is the absorption coefficient. The absorption coefficient is proportional to
the concentration of the absorbing molecules and is dependent upon the temperature;
it can written as [16]
α (v) = S (T ) f (v) Px m ,
(4.6)
where S(T) is the line intensity or integrated absorption coefficient per unit pressure
(expressed in cm-2atm-1), f(v) the spectral line function (normalized such that
∞
∫ f (v)dv = 1),
P the pressure and xm the species mole fraction. The line function f(v)
−∞
depends on temperature and the collisional environment of the molecule. The shape of
f(v) depends upon the type of broadening. Doppler broadening is due to thermal
motion of molecules and collisional broadening is a result of perturbations in the
energy levels of absorbing molecules caused by collision with the other molecules.
Since the parameters of the line function f(v) are often unknown, it is better to
integrate the left and right parts of equation (4.5) over the entire line profile and
substitute (4.6) into (4.5) in order to determine the species mole fraction:
∞
∫ ln
x m = −∞
82
I 0 (v )
dv
I (v )
S (T ) Pl
.
(4.7)
Chapter 4
The smallest detectable absorption is limited by the minimum difference in I and I0
that can be measured, being determined by noise in the measurement system. Typical
absorption sensitivities using direct absorption allow the detection of absorbances
( ln
I
) of the order of 10-3. For smaller absorbance it is difficult to reach an adequate
I0
signal-to-noise ratio to distinguish between Io and I.
Whereas the mole fraction of acetylene is expected to be large enough to be
measured by direct absorption (using TDLAS, see below), we anticipate that this will
not be the case for HCN. One-dimensional flame calculations (chapter 1) using GRI
Mech 3.0 [15] indicate ppm level mole fractions of HCN. For such mole fractions the
calculated absorbance, using line positions, strengths and broadening coefficients
from [17] and [10] respectively is on the order of 10-4. It is clear that the detection
sensitivity should be increased to resolve low concentrations of HCN with reasonable
accuracy. To increase the sensitivity, several detection techniques have been
developed over the past decades. These include signal enhancement methods such as
multipass cells [18,19] and cavity-enhanced spectroscopy [20], and noise-reduction
methods like wavelength- and frequency-modulation absorption spectroscopy [21]. In
this study, modulation spectroscopy with second harmonic detection is implemented
to increase the detection sensitivity. Wavelength-modulation absorption spectroscopy
(WMAS) is well described in the literature [21,22] and is only briefly described
below.
4.6.2 Wavelength Modulation Absorption Spectroscopy (WMAS)
The measured intensities always contain a certain amount of noise, which limits
the smallest detectable absorption. The dominating source of noise is often 1/f noise
[23], which is larger at low frequencies and will decrease at higher frequencies.
Wavelength-modulation absorption spectroscopy (WMAS) is an effective technique
to reduce the noise, and thereby increase the detectability. The principle of WMAS is
smooth modulation of the wavelength at a certain frequency followed by detection of
the modulated laser radiation at the original modulation frequency or at one of its
harmonics. One of the advantages of modulation spectroscopy is that only the noise
centered on the detection frequency will influence the measurements. The detection is
83
Chapter 4
shifted to higher frequencies where the 1/f noise is smaller. The basic approach is to
modulate the wavelength of the laser radiation around its center (here using the
frequency of the radiation, vc, in the derivation) at frequency fm with a modulation
amplitude va. The instantaneous frequency v(t) of the laser radiation can than be
represented as,
v(t ) = vc + v a sin 2πf m t .
(4.8)
After absorption in the medium, the intensity of the modulated, transmitted laser
beam I(v(t)) and the reference laser beam I0(v(t)) are related by the Lambert-Beer law,
already introduced in (4.5). WMAS is usually used for samples that have low
absorptions. This allows expanding the exponent in equation. (4.5) in a Taylor series.
By neglecting the higher order terms in the Taylor series one obtains:
I (v(t )) = I 0 (v(t ))[1 − α (v(t )) L] .
(4.9)
The instantaneous laser intensity I(vc+vasin2πfmt) can be expanded in a Fourier sine
series;
I (vc + v a sin 2πf m t ) =
∞
∑ H n (vc , va ) sin 2πnf mt .
(4.10)
n =0
The nth harmonic component of the intensity of the transmitted laser beam Hn(vc,va),
for n≥1 is expressed as
1
H n (vc , va ) =
2π
t
∫ I (vc + va sin 2π fmt ) sin 2π fmntdt .
(4.11)
−t
To simplify the analyses we assume that the laser intensity is independent of the
frequency. In this case Hn can be written as,
84
Chapter 4
LI
Hn = 0
2π
t
∫ −α (v + va sin 2π fmt ) sin 2π fmntdt .
(4.12)
−t
It can be seen from equation (4.12) that Hn is directly proportional to the
absorption coefficient α(vc+vasin2πfmt) and thus each harmonic component is directly
proportional to the species concentration in the absorption layer. Figure 4.6 shows a
schematic illustration of a broadened absorption line and the calculated corresponding
first (n=1) and second (n=2) harmonics (Fourier components).
Figure 4.6 Principle of WMAS with harmonic detection.
Any of the harmonics (n=1,2…) can be used; however optimum signals are
generated using second-harmonic detection [21]. The increased sensitivity of the
second-harmonic detection technique arises in part from the elimination of featureless
background signals. For detecting weak absorptions it is important to maximize the
harmonic signals by optimizing the modulation amplitudes. The simulated second
harmonic signals shown in figure 4.7 are calculated for different modulation
85
Chapter 4
indices m =
va
, using an exact solution to the integral in equation (4.11) for a
Δv1 / 2
Lorentzian line shape [24]. Here, Δν1/2 is the half width at half maximum (HWHM) of
the absorption line.
Figure 4.7 Typical 2f-lineshape for different modulation indices.
As can be seen from figure 4.7, the 2f-line shapes becomes broader with
increasing modulation index and reaches the maximum peak height at a modulation
index m∼2. The maximum peak height occurs at m∼2 for all line shapes [24,25]. In
environments where the absorption is very low, the modulation index is often set to a
value close to m∼2; for cases in which it is necessary to reduce interferences from
nearby transitions, the modulation index is decreased to minimize the broadening of
the 2f-line shape.
86
Chapter 4
4.7 Experimental setup for Tunable Diode Laser Absorption Spectroscopy
Figure 4.8. General schematic of the experimental TDLAS set-up.
Figure 4.8 shows the basic schematic of the experimental setup for Tunable
Diode Laser Absorption Spectroscopy (TDLAS) used for direct absorption
measurements of acetylene. The radiation from a New Focus 6326 tunable diode
laser, with a linewidth less than 300 kHz, was directed through the absorption cell.
Before entering the cell, a part of the laser radiation was split off to produce the
reference signal. The powers of the reference and sample beams were measured by
New Focus 2033 large area photodiodes with internal amplifiers. The photodiode
signals were digitized and processed by a PC. The laser wavelength was swept over
the span of 30 GHz with a scan rate 150 MHz/s by applying a voltage to the
piezoceramic plate, which moves the tuning end-mirror inside the laser. The
measurements were performed in the region around 1530 nm, where the P (9)
absorption line of the ν1 + ν3 band of C2H2 is located [8]. This line was selected
because of its relatively high oscillator strength [9] and lack of interference from
transitions of other flame molecules.
87
Chapter 4
4.7.1 Experimental procedure TDAS measurements of acetylene
Whereas it is possible to derive absolute values of acetylene concentration from
the measured integral absorption coefficient (equation 4.7), direct calibration by a gas
with known acetylene concentration was used in present work to enhance the
accuracy. This method avoids the uncertainties arising from the non-infinite limits of
integration when deriving the integrated absorption coefficient, and from converting
the voltage applied to the piezoceramic plate to the wavelength shift. The calibration
procedure is performed by measuring the absorption coefficient of a known amount of
acetylene under the same experimental conditions as those existing for the sampled
gas. As a typical example, Figure 4.9 shows the dependence of the logarithm of the
ratio of the reference to transmitted signal (absorbance) upon the laser wavelength
(expressed as the applied voltage) when the measurements were performed in a
mixture containing 5000 ppm C2H2 in N2 at 0.1 bar.
Figure 4.9 experimental absorption profile obtained in a mixture containing
5000 ppm C2H2 in N2 in the vicinity of the P(9) line of ν1+ν3 band.
When comparing the measurements performed in the calibration gases with those of
the combustion products, no differences were found in absorption line shapes. This
implies that the spectral line function in equation (4.6) is independent of the gas
composition in the current experiments. This result is perhaps to be expected, since
88
Chapter 4
the broadening efficiencies of C2H2 and N2 for the acetylene lines are comparable
[10], while the concentrations of the other gases with unknown broadening efficiency
are low.
Taking into account that the spectral line function is the same in the calibration
gas and in combustion products one derives from equation. (4.7),
v2
⎛ I (v ) ⎞
∫ ln⎜⎜⎝ I0(v) ⎟⎟⎠dv
⎛ I (v ) ⎞
∑ ln⎜⎜⎝ I0(vii) ⎟⎟⎠
v1
,
x M = xcal.
= xcal. i
v2
⎛ I 0cal (vi ) ⎞
⎛ I 0call. (v) ⎞
⎟
ln⎜⎜
⎟dv
ln⎜
I cal (vi ) ⎟⎠
⎜ I
⎟
⎝
i
cal
⎝
⎠
v
∫
∑
(4.13)
1
where the subscripts cal and i refer to the calibration gas and i-th measured point,
respectively. In the current experimental setup, for acetylene mole fractions above
1000 ppm, the accuracy of the measured C2H2 concentrations is ~5%, and is
determined mainly by the uncertainty in the calibration gas concentration. At low
acetylene mole fractions, the uncertainties in the measured integral absorption
coefficient become dominant, which results in deteriorating accuracy, up to 15% at
100 ppm C2H2. While equation (4.7) only requires one calibration point to determine
the measured acetylene concentration, to verify the possible influence of non-linearity
in the detection system, the integral absorption coefficient was measured in
calibration gases with different C2H2 concentrations. As can be seen from figure 4.10,
showing the results of these measurements, the detection system possesses excellent
linearity: deviation of the fitted and experimental values is less than 5%.
89
Chapter 4
Figure 4.10. Integrated absorption coefficient of the C2H2 P(9) line of the ν1+ν3 band
as a function of acetylene mole fraction diluted in N2. Solid line is a linear fit.
The measured C2H2 mole fractions were recalculated to flame conditions by
taking into account the partial removal of water in the ice-cooled trap. The residual
water concentrations in the absorption tube were determined from the integrated
absorption coefficient of the water absorption line at 6755.02 cm-1 [11] using equation
(4.7). The line intensity S(T) of the water line was taken from [11].
4.8 Experimental procedure WMAS with second harmonic detection
4.8.1 HCN measurements
The experimental sep-up used for the WMAS experiments with second harmonic
detection is almost identical to that presented schematically in figure 4.8. The only
difference is that we do not use the reference beam. The laser frequency v is
modulated around the center frequency vc by applying a sinusoidal voltage from an
external generator to the piezoceramic plate which moves the tuning end mirror inside
the laser. The modulation depth va, and frequency fm, in equation (4.8) are chosen to
be ∼0.67 GHz and 0.5 kHz, respectively, to provide maximal sensitivity for the
present experimental setup [21]. The power of the transmitted laser beam is digitized
90
Chapter 4
by an Agilent 54830B oscilloscope with sampling rate fs = 50 kHz and processed by a
PC. The second harmonic amplitude I2f(vc) is calculated from the sampled signal
series using the following expression:
I 2 f (vc ) =
1
N
N
∑ I (v ) ⋅ sin(2π
i =1
i
c
2 fm
i) .
fs
(4.14)
The summation (4.14) is performed with the number of samples N ~ 10000, which
corresponds 0.2 s total sampling time. Further increasing the number of samples does
not result in substantial improvement in the signal-to-noise ratio. The HCN
measurements were performed in the region around 1545 nm where the P(13)
absorption line of the ν2 band of HCN is located [17]. This line was selected because
of its relatively high oscillator strength and lack of interference from other flame
molecules. To cover the entire absorption line, the center frequency vc is tuned in
steps of 0.12 GHz over span of ∼10 GHz by changing the mean voltage applied to the
piezotransducer. The improvement in signal-to-noise ratio obtained using WMAS as
compared to direct absorption is illustrated in figure 4.11. The measured direct
absorption profile measured in a gas mixture containing 90 ppm HCN in N2 is shown
in figure. 4.11a. Figure 4.11b shows a typical HCN second harmonic spectrum in the
same gas mixture. As can be seen, the signal-to-noise ratio is improved by more than
one order by using WMAS with second harmonic detection.
91
Chapter 4
Figure. 4.11a.) Direct absorption scan of 90 ppm HCN in nitrogen in the vicinity of
the P(13) line of ν2 Σ-Σ band, b.) Second harmonic signal in the same mixture for the
same absorption transition.
Although it is possible to derive absolute HCN concentrations by fitting the
measured second harmonic absorption profiles, we use signal calibration in gases with
known HCN mole fractions, as done for acetylene above. In this approach, knowledge
of the absorption line shape is not required, which substantially improves the accuracy
92
Chapter 4
of measured HCN concentrations. Below, we give a short description of the
calibration procedure, starting by presenting the absorption coefficient α(v) as the sum
of the absorption coefficient αM of the HCN molecules and background absorption
αBG.
α (vl ) = α m (v) + α BG (v) = S (T ) g (v) Px m + α BG (v) .
(4.15)
Substituting (4.15) into the Lambert-Beer law for optically thin absorption layers (4.9)
and taking the Fourier transform, while taking into account that according (4.8) v is a
function of time, we receive the following expression for the Fourier transform of the
transmitted signal at frequency f,
I% ( f ) = I 0 ( f ) − I 0α BG ( f )l − I 0 g ( f ) S (T ) PxM l ,
(4.16)
where the tilde above the variable denotes its Fourier transform. As we can see from
this expression, the measured signal consists in general of three terms. The first two
terms in (4.16) are due to the frequency dependence of the laser power and
background absorption coefficient. In the case of sinusoidal modulation and second
harmonic detection (f = 2fm), these terms will vanish only when laser power and
background absorption coefficient are constant or one of them is linearly dependent
upon the laser frequency v. The term I 0 (2 f m ) − I 0α BG (2 f m )l can be determined by
measuring the second harmonic signal in media without HCN molecules. The
measurements, performed at a height of 10 mm above the burner surface in a
methane-air flame with equivalence ratio ϕ = 1.3 and mass flux ρν=0.015g/cm2s
where HCN concentration is expected to be very low, showed that in the spectral
region used in the present work the background I 0α BG (2 f m ) does not exceed the
signal from HCN at mole fraction of 5 ppm. Because the concentrations of main flame
components that contribute to the background absorption coefficient (H2O, CO and
CO2) vary only slightly with equivalence ratio, the value of I 0 (2 f m ) − I 0α BG (2 f m )l
was used for deriving HCN concentrations in all CH4/air flames studied here. To
determine the background for the CH4/H2/air flames, a similar reference flame (ϕ=1.3
93
Chapter 4
and ρν=0.015g/cm2.s) was used with the same percentage of hydrogen added as for
the flames under investigation.
A typical second harmonic spectrum of HCN, measured in a flame at ϕ=1.4, at 2
mm above the burner surface, is presented in figure 4.12a. The second harmonic
spectrum of HCN ( I 0 g (2 f m ) S (T ) Pxml ), shown in figure 4.12c, was obtained by
subtracting the measured second harmonic signal of the reference flame
( I 0 (2 f m ) − I 0α BG (2 f m )l ), given in figure 4.12b, from that shown in figure 4.12a. In
figure 4.12c, we can see the characteristic shape expected for the second derivative of
the bell-shaped spectral line profile.
Using the second harmonic signal to derive quantitative information on species
concentrations is rather difficult. Direct fitting according to the expression (4.16)
requires knowledge of functional dependence of both absorption spectral line and
laser intensity upon frequency. Moreover, direct integration of the second harmonic
signal would yield a value close to zero. To overcome these difficulties, the extracted
second harmonic signal is integrated over a sufficiently large region [v1, v2] in the
vicinity of the HCN spectral line. From the expression (4.16) it follows that
v2
xm = C
∫ I0 (2 fm ) − I% (2 fm ) − I0α BG (2 fm )l dv ,
(4.17)
v1
where the coefficient C is expressed as
1
C=
v2
S (T ) Pl
.
(4.18)
∫ I0 g (2 fm) dv
v1
It should be pointed out that, in general, the coefficient C depends upon
composition of the sampled gas. When comparing the measurements performed in the
calibration gases with those from the combustion products, no differences were found
in the second harmonic line-shapes. This implies that the spectral line function is
independent of the gas composition in the present experiments, and therefore the
coefficient C can be regarded as constant.
94
Chapter 4
Figure 4.12a) 2f-signal measured in a flame with ϕ=1.4 and ρν=0.01g/cm2.s at 2mm
from the burner surface, b) 2f signal measured in the reference flame (ϕ=1.3,
ρν=0.15g/cm2.s at 10mm from the burner surface), c) Extracted second harmonic
spectrum of HCN.
95
Chapter 4
Equation (4.17) was verified by measurements of the second harmonic signals at
different HCN concentrations in N2. The results of these measurements, presented in
figure 4.13, show a linear dependence between the integrals of the absolute value of
( I% (2 f m ) − I 0 (2 f m ) and the HCN mole fraction. Clearly visible is the offset at zero
HCN concentration; this is mainly due to the noise in the measured signals. When the
second harmonic signal is close to zero, the integration is performed over absolute
values of the noise, which results in a non-zero integral. In the current experimental
setup, when HCN mole fractions are above ~10 ppm, the accuracy (~10%) of the
measured HCN concentrations is determined mainly by the uncertainty in the
calibration gas concentration. At lower mole fractions the uncertainties in the
measured integral absorption coefficient become dominant, which results in
deteriorating accuracy, up to 30% at 3 ppm HCN.
Figure 4.13. Integrated absolute value of the second harmonic signal in the vicinity of
the P(13) line of ν2 Σ-Σ band as a function of HCN concentrations in N2.
96
Chapter 4
Literature
1.
A. V. Mokhov and H. B. Levinsky, Proc. Combust. Inst. 26 (1996) 21472154.
2.
A. C. Eckbreth, Laser Diagnostic for Combustion Temperature and Species,
(2nd Edition, Gordon and Breach, Amsterdam, 1996).
3.
"Applied Combustion Diagnostics," K. KohseHoinghaus and J. B. Jeffries,
eds., (Taylor & Francis, New York, 2002).
4.
V. V. Toro, Experimental study of the structure of laminar axisymetric H2/air
diffusion flames, Ph.D. Thesis, RUG, 2006 (ISBN 90-367-2703-0).
5.
E. L. Knuth, Combust. Flame 103 (3) (1995) 171-180.
6.
A. V. Mokhov, S. Gersen, H. B. Levinsky, J. Chem. Phys. Let., 403 (2005)
233-237.
7.
B. A. Williams., J. W. Fleming., Appl. Phys .B. 75 (2002) 883-980.
8.
Q. Kou, G. Guelachvili, M. A. Temsamani, M. Herman, Can. J. Phys. 72
(11-12) (1994) 1241-1250.
9.
R. El Hachtouki, J. Vander Auwera, J. Mol. Spectrosc. 216 (2) (2002) 355362.
10. A. S. Pine, J. Quant. Spectrosc. Radiat. Transfer 50 (2) (1993) 149-166.
11. R. A. Toth., Appl.Opt., 33 (1994) 4852.
12. A. Goldman., I.Rahinov., S. Cheskis., B. Lohden., S. Wexler., K. Sengstock.,
V. M. Baev, Chem.Phys.Let., 423 (2006) 147-151.
13. A. Sepman, Effect of burner stabilization on nitric oxide formation and
destruction in atmospheric pressure flat premixed flames, Ph.D. Thesis,
RUG, 2006 (ISBN 90-367-2702-2).
14. R. M. Fristrom and A. A. Westenberg, Flame Structure, (McGraw-Hill, New
York, 1965).
15. G. P. Smith, D. M. Golden, M. Frenklach, N. W. Moriarty, B. Eiteneer, M.
Goldenberg, C. T. Bowman, R. Hanson, S. Song, W. C. Gardiner, V.
Lissanski, Z. Qin, http://www.me.berkeley.edu/gri_mech/.
16. A. C. Eckbredt, Laser diagnostic for combustion temperature and species,
(2nd Edition, Gordon and Breach, Cambridge, 1996).
17. A. M. Smith, S.L. Coy, W. Klemperer, K.K. Lehmann, J. Mol. Spectros.
134 (1) (1989) 134-153.
97
Chapter 4
18. R. G. Pilston, J. U. White: A long path gas absorption cell, J. Opt. Soc. Am.
44, 572–573 (1954).
19. J. U. White: Long optical paths of large aperture, J. Opt. Soc. Am. 32, 285–
288 (1942).
20. A. O’Keefe, D. A. G. Deacon: Cavity ring-down optical spectrometer for
absorption measurements using pulsed laser sources, Rev. Sci. Instrum. 59,
2544 (1988).
21. J.A. Silver, Applied Optics, 31 (6) (1992) 707-717.
22. P. Kluczynski, J. Gustafsson, A. M. Lindberg, O. Axner, Spectrochimica
Acte Part B, 56 (2001) 1277-1354.
23. C. Th. J. Alkemade, T. J. Hollander, W. Snelleman, P. J. Th. Zeegers,
International Series in Natural Philosophy 3 (1982) 272 .
24. R. Arndt, Appl. Opt. 36 (8) (1965) 2522-2524.
25. J. Reid, D. Labie, Appl. Phys. B 26 (1981) 203-210.
98
Chapter 5
Chapter 5
Extractive Probe Measurements of Acetylene in
Atmospheric-Pressure Fuel-Rich Premixed
Methane/Air Flames
99
Chapter 5
5.1 Introduction
As discussed in previous chapters, one of the major advances of the last decades
in the combustion science is the prediction of flame structure by numerical
simulations using detailed transport and chemical mechanisms. Because of the
complexity of these mechanisms and uncertainties in the rates of the key chemical
reactions, the predictive power of the numerical simulations can be tested only by
comparing calculated and measured flame properties under well-defined experimental
conditions. The comparison of the spatial profiles of intermediate species is
particularly important for testing the adequacy of chemical mechanisms. One of the
key intermediates in many high temperature processes is acetylene (C2H2), which
plays important role in the formation of polycyclic aromatic hydrocarbons and soot in
hydrocarbon combustion [1-3] and in the chemical vapor deposition of diamond [4].
Acetylene has been extensively investigated in both atmospheric- and low-pressure
flat premixed flames [5-11]. At atmospheric pressure, large discrepancies have been
observed [8,11] between measured results and those calculated based on the C2H2
submechanism derived from Miller and Mellius [12]. However, the acetylene
measurements in these studies were performed using extractive probe sampling,
which as discussed in Chapter 4 has a serious drawback, i.e., the distortion of the
composition and temperature profiles in the flame. Estimating the magnitude of this
distortion, for example, from chemical reactions on the probe surface or acceleration
of the combustion products into the probe orifice are rather difficult [13]. Moreover,
these estimates (as was done in Chapter 4) require detailed knowledge of the kinetics
of the chemical reactions involving the measured species that itself is the subject of
investigation. These complications necessitate the verification of the results obtained
by the extractive probe by an independent technique. Recently, we have reported the
measurement of native C2H2 in a fuel-rich methane/air flame at equivalence ratio
ϕ = 1.55, using spontaneous Raman scattering [14]. This method thus provides us
with the means to verify the results of extractive probe sampling for acetylene
measurement, and to deliver reliable experimental results regarding C2H2 formation
and destruction in atmospheric-pressure methane/air flames.
Towards this end, we have measured the profiles of C2H2 mole fraction in flat
atmospheric-pressure rich-premixed methane/air flames using both spontaneous
Raman scattering and microprobe gas sampling followed by tunable diode laser
100
Chapter 5
absorption spectroscopy (TDLAS). These measurements are supplemented with
profiles of flame temperature, obtained using coherent anti-Stokes Raman Scattering
(CARS), and the experimental results are compared with the predictions of onedimensional flame calculations.
5.2 Experimental
Here we briefly summarize the experimental method; Chapter 4 discusses the
methods in more detail. The measurements were performed in atmospheric-pressure
methane/air flames stabilized above a McKenna Products sintered bronze burner of
60 mm diameter. To prevent air entrainment in the combustion products a nitrogen
shroud was used. The flame was stabilized by a cylindrical chimney with a 60 mm
inner diameter, which was positioned approximately 30 mm above the burner surface.
The flame temperature was varied by changing the mass flow through the burner and
measured by broadband planar BOXCARS for nitrogen thermometry. Details of the
CARS experiment are described elsewhere [15]. The flow rates of methane and air
were measured by calibrated mass flow meters and the equivalence ratio was
determined by measuring the methane concentration in the unburned fuel-air mixture.
For calibration purposes, nitrogen doped with a known amount of acetylene was
flowed through the burner instead of the methane-air mixture. Measurements were
obtained at different axial positions in the flame by moving the burner with a
precision positioner relative to the laser beams and sampling probe in steps of 1 mm.
The flames were sampled by a cooled quartz micro-probe, and the sampled gas
flowed through an absorption cell and analyzed using TDLAS. As discussed in
Chapter 4, estimates of C2H2 conversion during sampling indicated that in the present
experimental setup conversion of acetylene in the probe is less than 15% when
sampling is made at axial distances greater than 2.5 mm from the burner surface.
These estimates are supported by measurements at different suction backpressures,
which showed no significant changing in the measured HCN concentration when
varying pressure from 0.05 to 0.35 Bar.
101
Chapter 5
5.3 Results and discussion
The measurements were performed in a set of fuel-rich flames with different
equivalence ratios and mass fluxes. The flame parameters (equivalence ratios, mass
fluxes and temperatures at 5 mm above the burner surface) are presented in Table 5.1.
Table. 5.1. Flame parameters
Flame
ϕ
ρv, g/cm2·s)
T, K
A
1.5
0.005
1763
B
1.5
0.007
1835
C
1.5
0.008
1852
D
1.45
0.007
1833
E
1.45
0.0085
1885
F
1.45
0.010
1916
G
1.4
0.005
1762
H
1.4
0.007
1816
I
1.4
0.0085
1850
The temperature measurements showed that all the flames studied had a domain with
constant temperature extending at least 20 mm radially from the centerline, and from
3 mm to 15 mm in the axial direction. As typical examples of the temperature
measurements, the radial profile at height 10 mm above the burner surface and
centerline axial profile in flame A are presented in Figs. 5.1 and 5.2, respectively. The
radial profile shows a core region of ~ 20 mm length of constant temperature
surrounded by a layer where the temperature is higher due to penetrating surrounding
air through the nitrogen shroud. The axial centerline temperature profile is in excellent
agreement with the flame calculations, indicating the robustness of the GRI-Mech 3.0
[16] mechanism in predicting the burning velocities of CH4/air flames and marginal
radiative heat losses in these flames. The excellent agreement between measured and
calculated axial temperature profiles was observed in all flames studied.
102
Chapter 5
Figure 5.1. Radial temperature profile measured in flame A at 10 mm above the
burner surface.
Figure 5.2. Axial centreline temperature profile measured in flame A. Solid line and
diamonds denote flame calculations and measurements, respectively.
Acetylene mole fractions, measured by Raman scattering in flame C, and shown
in figure 5.3, reach a maximum at an axial distance between 2 and 3 mm and then
decrease to ∼ 500 ppm at 9 mm, the detection limit of the current setup [15]. As can
be seen in this figure, the profile obtained with the probe is shifted approximately
1.3 mm farther downstream. A similar shift between probe and optical measurements
103
Chapter 5
was observed in temperature and hydroxyl profiles in other flames [17,18], and is the
result of the acceleration of the combustion products into the probe orifice [13].
Shifting the probe profile results in agreement with the Raman profiles to better than
20% (also observed in Ref. [14] at ϕ = 1.58), which substantiates the extractive probe
technique for the measurements of acetylene presented below.
Figure 5.3. Axial centerline profiles of acetylene mole fraction in methane/air flame,
ϕ = 1.50 and ρv = 0.008 g/(cm2·s). Symbols denote Raman (triangles) and probe
(squares with solid line) measurements; the dashed line denotes the shifted probe
measurements.
At mole fractions below 500 ppm the acetylene Raman spectrum was barely
distinguishable in the noise, while the signal-to-noise ratio of the TDLAS spectrum
remained higher than 10 for mole fractions down to 100 ppm. This difference in the
limit of detectability precludes comparison of probe and Raman data in the flames
with low C2H2 mole fraction in the post-flame zone. However, due to the modest
changes in flame structure upon changing the equivalence ratio from ϕ = 1.5 to
ϕ = 1.4, we do not expect the accuracy of the probe measurements observed at
ϕ = 1.58 and ϕ = 1.5 to deteriorate substantially. The results of the extractive probe
measurements of acetylene at equivalence ratios ϕ = 1.5, 1.45 and 1.4 at different
mass fluxes are presented in figure 5.4-5.6. Consistent with the results in figure 5.3
and those presented in Ref. [14], all experimental acetylene profiles are shifted
104
Chapter 5
1.3 mm towards the burner surface. As can be seen from the figures, at a fixed
equivalence ratio the maximum C2H2 mole fraction depends only slightly on the mass
flux, while C2H2 oxidation in the post-flame zone increases substantially in the flames
at higher the mass flux, caused by the higher gas temperatures (given in Table 5.1). At
the same time, decreasing the equivalence ratio from ϕ = 1.5 to 1.4 decreases the
peak C2H2 mole fraction by nearly a factor of two.
Figure 5.4. Axial profiles of acetylene mole fraction in methane/air flames, ϕ = 1.5.
Symbols denote probe measurements in flames A (squares), B (diamonds) and C
(triangles). The dashed lines denote flame calculations with GRI-Mech 3.0, and the
solid lines are the results of calculations with the increased rate coefficient for
C2H2 + OH Æ CH2CO + H discussed in the text.
105
Chapter 5
Figure 5.5. Axial profiles of acetylene mole fraction in methane/air flames, ϕ = 1.45.
Symbols denote probe measurements in flames D (squares), E (diamonds) and F
(triangles). Solid lines denote flame calculations with the increased rate coefficient
for C2H2 + OH Æ CH2CO + H discussed in the text.
Figure 5.6. Axial profiles of acetylene mole fraction in methane/air flames, ϕ = 1.40.
Symbols denote probe measurements in flames G (squares), H (diamonds) and I
(triangles). Solid lines denote flame calculations with the increased rate coefficient
for C2H2 + OH Æ CH2CO + H discussed in the text.
106
Chapter 5
In addition, we note substantial discrepancies between the measured acetylene
profiles and those obtained from the flame calculations using GRI-Mech 3.0. As can
be seen in figure 5.4, the calculations give substantially higher peak concentrations
and slower decay in the post-flame zone than those measured, well outside the 20%
differences observed between the experimental methods. The computed profiles at the
other equivalence ratios showed discrepancies similar to those presented in figure 5.4
This discrepancy has been observed previously [11], where it was attributed to the
choice of the rate coefficient of the reaction C2H2 + OH Æ CH2CO + H used in GRIMech 3.0. Following the suggestion made in Ref. [11] we increased the preexponential factor of the rate coefficient to 1.7·1012 cm3/mole·s, and these results are
also presented in Figs. 5.4-5.6. The calculated acetylene profiles are now in excellent
agreement for all flames studied here. Although the limited parameter variation in the
present work precludes an unambiguous recommendation regarding increasing the
rate coefficient of this reaction, the agreement between experiment and calculations
favors this recommendation.
5.4 Conclusions
We report the measurements of acetylene in fuel-rich atmospheric-pressure
methane/air flames using spontaneous Raman and extractive probe sampling
techniques. Excepting a shift of approximately 1.3 mm, resulting from the
acceleration of the combustion products in the probe orifice, the axial Raman and
probe profiles are in very good agreement. This result validates using the extractive
probe sampling technique as a diagnostic tool for measurements of acetylene for the
conditions studied. Substantial disagreement is observed between the experimental
profiles of acetylene and those obtained from calculations based on GRI-Mech 3.0,
which predict higher acetylene concentrations and slower decay in the post-flame
zone. Increasing the pre-exponential factor in the rate coefficient for the reaction
C2H2 + OH Æ CH2CO + H to the value of 1.7·1012 cm3/mole·s brings the calculated
acetylene profiles into excellent agreement with those derived experimentally. Further
improvement of the sensitivity of both spontaneous Raman and extractive probe
techniques will provide more information on acetylene chemistry in fuel-rich
methane-air flames. These improvements are currently in progress in our laboratory
107
Chapter 5
Literature
1.
J. Warnatz, H. Bockhorn, A. Mozer, H. W. Wenz, Proc. Combust. Inst. 19
(1982) 197-209.
2.
P. Lindstedt, Proc. Combust. Inst. 27 (1998) 269-285.
3.
H. Richter, J.B. Howard, Prog. Energy Combust. Sci. 26 (4-6) (2000) 565608.
4.
A. Dollet, Surf. Coat. Technol. 177 (2004) 245-251.
5.
C. P. Fenimore, G. W. Jones, J. Chem. Phys. 41 (7) (1964) 1887-1889.
6.
J. Vandooren, P. J. van Tiggelen, Proc. Combust. Inst. 16 (1977) 1133-1144.
7.
D. Bittner, J. B. Howard, Proc. Combust. Inst. 19 (1982) 211-221.
8.
E. W. Kaiser, J. Phys. Chem. 94 (11) (1990) 4493-4499.
9.
I. T. Woods, B. S. Haynes, Combust. Sci. Technol. 87 (1-6) (1993) 199-215.
10. I. T. Woods, B. S. Haynes, Proc. Combust. Inst. 25 (1994) 909.
11. E. W. Kaiser, T. J. Wallington, M. D. Hurley, J. Platz, H. J. Curran, W. J.
Pitz, C. K. Westbrook, J. Phys. Chem. A. 104 (2000) 8194-8206.
12. J. A. Miller, C. F. Melius, Proc. Combust. Inst. 22 (1988) 1031.
13. E. L. Knuth, Combust. Flame 103 (3) (1995) 171-180.
14. A. V. Mokhov, S. Gersen, H. B. Levinsky J. Chem. Phys. Let., 403 (4-6)
(2005) 233-237.
15. A. V. Mokhov, C. E. van der Meij, H. B. Levinsky, Appl.Opt. 36 (1997)
3233-3243.
16. G. P. Smith, D. M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M.
Goldenberg, C. T. Bowman, R.K. Hanson, S. Song, W.C. Gardiner, V.
Lissanski, Z. Qin, http://www.me.berkeley.edu/gri_mech/.
17. R. J. Cattolica, S. Yoon, E. L. Knuth, Combust. Sci. Technol. 28 (5-6) (1982)
225-239.
18. A. T. Hartlieb, B. Atakan, K. Kohse-Hoinghaus, Combust. Flame 121 (4)
(2000) 610-624.
108
Chapter 6
Chapter 6
HCN formation and destruction in atmosphericpressure fuel-rich premixed methane/air flames
109
Chapter 6
6.1 Introduction
To date, the vast majority of the methods developed to lower NOx emissions are
based on decreasing the flame temperature: either through flue-gas recirculation or
fuel-lean combustion. Application of these methods to obtain the lowest NOx possible
ultimately leads to deteriorating flame stability. Decreasing the temperature of fuelrich flames, which can be more stable than fuel-lean flames, is often not considered to
be promising as an NOx control strategy because the key initiation reaction of the
Fenimore mechanism [1]
CH + N 2 = HCN + N
(R6.1)
has a relatively low activation energy [2], and as such should be less sensitive to
flame temperature than the Zeldovich mechanism. However, experiments performed
in premixed fuel-rich natural-gas/air flames [3], showed that the measured NO mole
fractions decrease with increasing upstream heat loss, suggesting that varying the
flame temperature by altering the degree of burner stabilization can influence the NO
production in rich flames. Recent experiments at low pressure [4] have supported
these observations, and analyzed the effects of burner stabilization on Fenimore NO
formation. On the other hand, it is also possible that part of the observed decrease in
NO with flame temperature can be caused by the retardation of the conversion of the
HCN formed in reaction (R6.1), or other fixed-nitrogen species (so-called Total Fixed
Nitrogen or TFN) to NO, as suggested in [5, 6]. In this case, the “residual” TFN either
will be converted to NO in the second stage of combustion or emitted into the
environment with flue gases. To resolve this uncertainty, and to determine the
ultimate low NOx potential of fuel-rich combustion, it is essential to measure HCN in
these flames. The experimental investigations of HCN formation and destruction
performed thus far have seeded flames with bound-nitrogen additives, where the
reaction between CH and N2 is of minor importance. HCN measurements in
methane/air flames are scarce and often contradictory. For example, one study [6]
reported “as ϕ increases the maximum concentration of HCN increases initially, but
falls again in very fuel-rich flames”, while another study [7] observed a strong
increase in HCN with equivalence ratio. Moreover, Ref. [7] states that the HCN
concentration is not very dependent upon the temperature, while the reported HCN
110
Chapter 6
concentrations at T = 2560 K are almost one order of magnitude higher than those at
the same equivalence ratio (ϕ = 1.20) in Ref. [6] in the methane/air flame. Further, we
are not aware of HCN measurements in which the flame temperature was varied at
fixed equivalence ratio.
Although there are numerous experimental observations in fuel-rich flames (for
example, Ref. [8] and references therein) showing correlations between the measured
NO and CH mole fractions, as well as direct high-temperature measurements of the
rate constant of reaction (R6.1) [9,10] that reasonably agree with the results of flame
modeling, controversy about this reaction is not resolved [11]. Reaction (R6.1) as
written is “spin-forbidden”, and to reconcile theory and experiment a near unit
probability of crossing between doublet and quartet potential surfaces had to be
assumed [12]. Recently, ab initio RRKM calculations have been performed for the
reaction CH + N2 Æ products [13,14], which showed that the reaction between CH
and N2 takes place mainly through the “spin-allowed” channel:
CH + N 2 = NCN + H ,
(R6.2)
while reaction (R6.1) is not important. These calculations have been supported by
NCN detection in a low-pressure CH4/air flame [15], although the measurements were
not quantitative. It should be pointed out that theoretical results giving (R6.2) does not
contradict the experimental observations: the rate constant of the reaction (R6.1) was
determined without measurement of the products, while rapid conversion of any NCN
produced in (R6.2) to NO will preserve the correlation between the CH mole fraction
and NO formation. The experimental determination of the concentration profiles of
HCN in flames in which we expect significant Fenimore NO formation will help
resolve the uncertainty as to both the primary products of CH + N2 and the role of
HCN as a stable intermediate in NO formation.
Towards this end we measure the axial profiles of the mole fraction of HCN in
burner-stabilized rich-premixed methane/air flames at equivalence ratios ϕ = 1.3-1.5
at different flame temperatures, similar to the method followed in Chapter 5. In these
flames the HCN mole fractions are of order of a few tens of ppm. Whereas the
sensitivity of the absorption method allows measuring such low concentrations, the
strong background from the hot bands in the water absorption spectrum has frustrated
111
Chapter 6
direct in-situ HCN measurements in flames thus far. To circumvent this problem, we
follow the approach described in Chapter 4, and use microprobe sampling and
wavelength modulation absorption spectroscopy with second harmonic detection,
supplemented with CARS temperature measurements. As was done in Chapter 5 for
acetylene, the experimental observations are compared with one-dimensional flame
calculations.
6.2 Experimental
The premixed atmospheric pressure fuel-rich methane/air flames were examined
using the setup and experimental methodology described in Chapter 4, and briefly
summarized here. The flame temperature was varied by changing the mass flow
through the McKenna Products burner and measured by broadband planar BOXCARS
for nitrogen thermometry, as described elsewhere [16]. Again, the accuracy of the
CARS measurements is estimated at better than 40 K [16]. The flows of all gases
were measured by mass flow meters, while the equivalence ratio was determined by
measuring the methane volume fraction in the unburned gas/air mixture. For
calibration purposes, nitrogen doped with a known amount of HCN was flowed
through the burner instead of the methane-air mixture. In the current set-up, as
described in chapter 4, the accuracy of the measured HCN concentrations in the
calibration gases was better than 30% at mole fractions above 3ppm. Measurements
were obtained at different axial positions in the flame by moving the burner with a
precision positioner relative to the CARS laser beams and sampling probe in steps of
1 mm. The sampled gas mixtures were analyzed using WMAS with detection at the
second harmonic. Estimates of the sample cooling process showed that in the present
setup the conversion of HCN during sampling is less than 10% for all measured
flames. These estimations are supported by measurements at different suction
backpressures, which showed no significant changing in the measured HCN
concentration when varying pressure from 0.05 to 0.15 Bar.
112
Chapter 6
6.3 Results and discussion
The measurements were performed in a set of fuel-rich flames with different
equivalence ratios and mass fluxes. The flame parameters (equivalence ratios, mass
fluxes and temperatures measured at 5 mm above the burner surface) are presented in
Table 6.1.
Table 6.1
Flame
ϕ
ρv, g/cm2·s
T, K
A
1.3
0.007
1775
B
1.3
0.011
1855
C
1.3
0.015
1942
D
1.4
0.008
1842
E
1.4
0.01
1910
F
1.4
0.014
1950
G
1.5
0.005
1763
H
1.5
0.008
1852
The temperature measurements showed that all the flames studied had a domain
with constant temperature extending at least 20 mm radially from the centerline, from
3 mm to 15 mm in the axial direction. Moreover, the measured temperatures in this
domain were in excellent agreement with the flame calculations using GRI-Mech 3.0
[17], indicating marginal radiative heat losses in these flames.
The axial HCN profiles measured in flames with equivalence ratios ϕ = 1.3, 1.4
and 1.5 are presented in figures 6.1, 6.2 and 6.3, respectively. The experimental
profiles of HCN mole fraction were shifted downstream, as in done chapter 5, where
the shift of ~1.3 mm was observed between acetylene concentrations profiles
measured by the Raman and probe techniques. Since the equivalence ratios and mass
fluxes are close to those in chapter 5, we apply the same shift (1.3 mm) to the
measured HCN profiles.
113
Chapter 6
Figure 6.1 Axial profiles of HCN mole fractions in methane air flames, ϕ = 1.3.
Symbols denote probe measurements in flames A (squares), B (diamonds) and C
(triangles). Lines denote flame calculations in flames A (solid), B (dashed) and C
(dotted).
Figure 6.2. Axial profiles of HCN mole fractions in methane air flames, ϕ = 1.4.
Symbols denote probe measurements in flames D (squares), E (diamonds) and F
(triangles). Lines denote flame calculations in flames D (solid), E (dashed) and F
(dotted).
114
Chapter 6
Figure 6.3. Axial profiles of HCN mole fractions in methane air flames, ϕ = 1.5.
Symbols denote probe measurements in flames G (squares) and H (diamonds). Lines
denote flame calculations in flames G (solid) and H (dashed).
As can be seen from figure 6.1-6.3, the measured HCN concentrations peak in
the region of the flame front, and decrease (albeit gradually) in the post-flame zone.
The maximum measured HCN mole fractions are only slightly dependent upon
equivalence ratio and mass flux, and do not exceed 15 ppm in all the flames studied
here. In flames with ϕ = 1.3 and 1.4, increasing the mass flux at fixed equivalence
ratio results in shifting the maxima towards the burner surface and accelerating the
HCN burnout. Both effects are expected based upon the reduced degree of
stabilization and the concomitant increase in flame temperature. At ϕ = 1.5 the
maximum is shifted so far downstream that the HCN burnout region is beyond the
measurement domain. The observation of 5-10 ppm of “residual” HCN downstream
of the flame front that is only slowly oxidized supports the conclusion drawn in earlier
studies of NO formation in the burnout region of rich methane flames [5]. It is
interesting to point out the strong temperature dependence of the HCN burnout. For
example, the change in residence time between flames D and F by nearly a factor of
two should lead to an increase in the HCN mole fraction at the same axial position.
However, we observe a decrease in HCN mole fraction by nearly a factor of two, most
likely caused by the relatively modest (~100 K) increase in the flame temperature.
Also interesting is the strong shift downstream of the maximum in the HCN profiles
115
Chapter 6
and the reduced oxidation rate in the post flame zone with increasing equivalence
ratio at fixed flame temperature, illustrated by comparing the flames B, D and H,
which have almost the same temperature (~1850 K) but increasing the equivalence
ratios from ϕ = 1.3 to 1.5.
Large discrepancies are observed between the measured results and those
calculated using the GRI-Mech 3.0 chemical mechanism. As can be seen from figures
6.1– 6.3, the calculations significantly overpredict the measured HCN mole fractions
for all mass fluxes at equivalence ratios ϕ = 1.3 and 1.4. At ϕ = 1.5, the calculated
HCN profiles are in surprisingly good agreement with the measurements for
ρv = 0.008 g/(cm2s), while the experimental results are underpredicted for ρv = 0.005
g/(cm2s). At the same time, GRI-Mech 3.0 predicts the qualitative trends in the
burnout region reasonably well. This suggests that adjustment of the rates of the
reactions that form and consume HCN in the flame front in GRI-Mech 3.0 would
improve the predictions substantially.
According to the GRI-Mech 3.0 scenario, HCN is formed mainly in reaction
(R6.1) between CH and N2. A substantial part of the nitrogen atoms produced in this
reaction will also be converted to HCN. First they react with CH3 to form H2CN
CH 3 + N = H 2 CN + H
(R6.3)
and then H2CN dissociates into HCN and H; here too, the HCN produced will be
oxidized to NO. Oxidation of HCN in the flame front occurs mainly by
HCN + O = NCO + H .
(R6.4)
To lower the predicted peak HCN mole fraction, either the rate of formation must be
decreased or the rate of consumption increased.
To improve the agreement for the results shown above, the rate of the reaction
(R6.1) must be decreased by a factor 2.5. This reaction, being the first step of the
Fenimore mechanism, ultimately determines amount of NO that will be produced in
fuel-rich hydrocarbon flames [1,2]; changing its rate will result in large disagreement
with numerous experimental and modeling studies of the NO formation, for which
there is currently good agreement [4] en references therein). Alternatively, decreasing
the calculated HCN mole fraction in the flame front by varying the rate constant of the
116
Chapter 6
reaction (R6.4) requires increasing its rate by more than 10 times. The rate constant of
this reaction was determined both experimentally and by transition state calculations
[18-20]. Very good agreement (~30%) was observed between calculations and
measurements in the temperature region from 450 K up to 2400 K. Moreover, the rate
constant of this reaction was used as an optimization variable for the GRI-Mech 3.0
mechanism, where it was already increased by a factor 1.45. Further increase will
bring the rate constant of reaction (R6.4) far beyond the specified margins of
uncertainty.
A possible explanation for the discrepancy between measured and calculated
HCN concentrations can be found in considering the products of the reaction between
CH and N2 asserted in ref. [14]. The NCN formed in this reaction can be converted
directly to NO in the following reactions:
NCN + O = CN + NO
(R6.5)
NCN + O2 = NCO + NO ,
(R6.6)
and
while HCN is formed in reactions
NCN + H = HCN + N
(R6.7)
NCN + OH = HCN + NO .
(R6.8)
and
Reactions (R6.5)-(R6.8) are supposed to be fast [15,21], with the result that the rate of
the reaction between CH and N2 controls NO formation. At the same time, reactions
(R6.5) and (R6.6) produce NO directly while bypassing HCN formation.
Unfortunately, information on NCN kinetics is very scarce. The reactions (R6.5)(R6.8) were included in the modeling study of the NO reburning in a flow reactor
[21], where their rates were estimated. It should be pointed out that the reaction (R6.6)
between NCN and O2 can be important in the fuel-rich highly stabilized flames, where
oxygen is removed relatively slowly, yielding the broad flame front in which the O2
concentrations are one to two orders larger than those of O, H and OH .
117
Chapter 6
To examine the possible improvement achievable using this “NCN-route”, we
perform the calculations for flame F, for which a large discrepancy between the
measurements and numerical prediction is observed (figure 6.4). In these calculations,
we used the same rate constant for the reaction (R6.2) as for the reaction (R6.1).
Whereas this value is approximately 5 times larger than that calculated in Ref. [14],
where this reaction was first proposed, it falls within the expected accuracy of the
calculations. Recent modeling work [22] also suggests using the same rate constant
for the reaction (R6.2) as for (R6.1), and low-pressure experiments and modeling [4]
also support using this value. As can be seen from figure 6.4, using the rate constants
from [21] for the reactions consuming NCN results in decreasing the maximum
calculated HCN mole fraction to 29 ppm, 10 ppm less than that predicted by GRIMech 3.0. Putting the magnitudes of the rates of all NCN removal reactions equal to
1.0·1013cm3/(mol·s) leads to a further 10 ppm reduction in the maximum HCN mole
fraction, and brings the calculated HCN profile even closer to the measured result.
Although not presented to avoid clutter in the figure, the calculations show that
varying the rates of the reactions (R6.5)-(R6.8) has only a minor influence on the NO
formation in these flames, this being determined by the rate of the initial nitrogen
fixation reaction (R6.2).
118
Chapter 6
Figure 6.4. Axial profiles of HCN mole fractions in methane air flame F. Triangles
denote measured HCN concentrations. Lines denote flame calculations using GRIMech 3.0 mechanism with NCN removal reactions [21] (solid – unchanged rate
constants, dashed – the rate constants are 1.0·1013cm3/(mol·s) for all NCN removing
reactions) and dotted - GRI-Mech 3.0 without NCN removal reactions).
It should be pointed out that variation of the rate constants should be performed
very cautiously, remaining inside any expected uncertainty limits. A recent theoretical
study [23] and experiments at room temperature [24] both yielded rates of the reaction
(R6.6) between NCN and O2 that are orders of magnitude lower than that from Ref.
[21]. This large disparity dissuades us from trying optimizing the rate constants of the
reactions (R6.5)-(R6.8) based on the results presented here. In addition, it was
possible to optimize the rates to fit one experiment adequately; however, good
prediction for one experiment resulted in a poorer prediction for other experiments.
Further experimental and theoretical studies of NCN kinetics are needed for a better
understanding of NO and HCN formation in hydrocarbon flames.
119
Chapter 6
6.4 Conclusions
Axial profiles of hydrogen cyanide have been measured in laminar, atmosphericpressure, rich-premixed, methane/air flames at equivalence ratios ϕ = 1.3, 1.4 and 1.5.
The measurements were performed by microprobe gas sampling followed by analyses
using wavelength modulation absorption spectroscopy with second harmonic
detection. In the richest flame under investigation (ϕ = 1.5), very slow removal HCN
is observed in the post flame zone, demonstrating “residual” HCN in the post-flame
gases of fuel-rich methane/air flames. In this flame HCN concentrations of ~ 10 ppm
are measured at 7 mm above the burner surface. In practical combustion systems, this
HCN will most likely be oxidized to NO in a secondary combustion step. Decreasing
the equivalence ratio leads to faster HCN removal in the post flame zone. When
varying the flame temperature at fixed equivalence ratio no significant changes in the
HCN peak concentration is observed, while the HCN removal becomes faster with
increasing gas temperature.
Substantial disagreement is observed between the experimental profiles of HCN
and those obtained from calculations using GRI-Mech 3.0. Changing the rates of key
formation and consumption reactions showed that bringing the calculations using
GRI-Mech 3.0 into agreement with the present results can be done only at the cost of
unreasonable changes in the rates of these reactions. On the other hand, considering
NCN as a primary product of the reaction between CH and N2, based on recent
theoretical studies, allows improvement in the agreement between measured and
calculated HCN mole fractions. The lack of information on the rate constants of the
NCN reactions at high temperatures precludes unambiguous conclusions regarding
this mechanism. To provide this information we are planning to perform
measurements of CH, NO HCN and NCN in low pressure flames.
120
Chapter 6
Literature
1.
C. P. Fenimore, Proc. Combust. Inst. 13 (1971) 373.
2.
J. A. Miller, C. T. Bowman, Prog. Energy Combust. Sci. 15 (4) (1989) 287338.
3.
A. V. Mokhov, H. B. Levinsky, Proc. Combust. Inst. 29 (1996) 2147-2154.
4.
V. M. van Essen, A. V. Sepman, A. V. Mokkov, H. B. Levinsky., Proc.
Combust. Inst. 23 (2007) 329-337.
5.
A. V. Mokhov and H. B. Levinsky., Combust. Flame 118 (1999) 733-740.
6.
B. S. Haynes, D. Iverach, N. Y. Kirov, Proc. Combust. Inst. 15 (1975) 11031112.
7.
C. Morley, Combust. Flame 27 (2) (1976) 189-204.
8.
K. Kohse-Hoinghaus, R. S. Barlow, M. Alden, E. Wolfrum, Proc. Combust.
Inst. 30 (2005) 89-123.
9.
D. Lindackers, M. Burmeister, P. Roth, Proc. Combust. Inst. 23 (1990) 251257.
10. A. J. Dean, R. K. Hanson, C.T. Bowman, Proc. Combust. Inst. 23 (1990)
259-265.
11. J. A. Miller, M.J. Pilling, E. Troe, Proc. Combust. Inst. 30 (2005) 43-88.
12. J. A. Miller, S. P. Walch, Int. J. Chem. Kinet. 29 (4) (1997) 253-259.
13. L. V. Moskaleva, W. S. Xia, M. C. Lin, Chem. Phys. Lett. 331 (2-4) (2000)
269-277.
14. L. V. Moskaleva, M. C. Lin, Proc. Combust. Inst. 28 (2000) 2393-2401.
15. G. P. Smith, Chem. Phys. Lett. 367 (5-6) (2003) 541-548.
16. A. V. Mokhov, C. E. van der Meij, H. B. Levinsky, Appl.Opt. 36 (1997)
3233-3243.
17. G.P. Smith, D.M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M.
Goldenberg, C.T. Bowman, R. Hanson, S. Song, W.C. Gardiner, V.
Lissanski, Z. Qin, http://www.me.berkeley.edu/gri_mech/.
18. P. Roth, R. Lohr, H. D. Hermanns, Berichte der Bunsen-GesellschaftPhysical Chemistry Chemical Physics 84 (9) (1980) 835-840.
19. M. Louge, R. K. Hanson, Proc. Combust. Inst. 20 (1985) 665-675.
20. R. A. Perry, C. F. Melius, Proc. Combust. Inst. 20 (1984) 639-646.
121
Chapter 6
21. P. Glarborg, M.U. Alzueta, K. Dam-Johansen, J. A. Miller, Combust. Flame
115 (1-2) (1998) 1-27.
22. A. El bakali, L. Pillier, P. Desgroux, B. Lefort, L. Gasnot, J. F. Pauwels, I. da
Costa, Fuel 85 (2006) 896-909.
23. R. S. Zhu, M. C. Lin, Int. J. Chem. Kinet. 37 (10) (2005) 593-598.
24. R. E. Baren, J. F. Hershberger, J. Phys. Chem. A 106 (46) (2002) 1109311097.
122
Chapter 7
Chapter 7
The effect of hydrogen addition to rich stabilized
methane/air flames
123
Chapter 7
7.1 Introduction
Stringent emission regulations for greenhouse gases and the drive to conserve
the finite supplies of fossil fuels have resulted in increased interest in the possible
addition of sustainable hydrogen to natural gas. In spite of any potential advantage to
hydrogen addition in this regard, a negative effect on other aspects of combustion
performance, such as increased pollutant emissions (NOx, soot), must be weighed in
the overall considerations. Recently, several studies on the effect of hydrogen addition
to hydrocarbons have been conducted regarding extinction limits [1], burning
velocities [2-4] and engine performance [5]. However, very few flame structure
studies have been performed on hydrogen-hydrocarbon mixtures, and all of these
pertained to non-premixed flames. For example, the effect of hydrogen was studied in
a hydrogen-natural-gas composite fuel in turbulent jet flames [6], and showed that
increasing the hydrogen concentration resulted in an increase in soot, CO and NOx
formation. Another study [7] added natural gas and propane to coflow hydrogen
diffusion flames; a decrease in soot formation and an increase in NOx with increasing
hydrogen content were found. In the same type of flames, intermediate radicals were
measured [8,9] and a decrease in CH radical concentration was observed with
increasing hydrogen content. In these studies, the increase in NOx emission was
ascribed to the increasing flame temperature with hydrogen content, which promotes
thermal NOx production. However, the measured decrease in CH concentration with
hydrogen content suggests that hydrogen addition to hydrocarbon flames reduces the
NOx emissions contributed from the Fenimore mechanism. The analysis of the
chemistry in the coflow diffusion flame studies described above suffers from the
complication of multidimensional transport and the fact that the difference in diffusive
behavior between the relative heavy hydrocarbon fuels and hydrogen is very large.
For this reason, measurements of key intermediates species in pollutant formation
using premixed one-dimensional flames can yield a more unambiguous contribution
to understanding the consequences of hydrogen addition for the chemistry of pollutant
formation. To date, no quantitative studies have been reported on the effect of
hydrogen addition on pollutant formation in one-dimensional flames.
One of the key intermediate species in the formation of soot in rich hydrocarbon
flames is acetylene [10] and understanding the fate of this species is essential to
understand soot inception. Also, acetylene is a precursor of the CH radical [11], which
124
Chapter 7
is an important intermediate in Fenimore NO formation in hydrocarbon flames [12].
As discussed in chapter 6, HCN is an important intermediate in the Fenimore
mechanism as well. Moreover, due to the lack of oxygen radicals in fuel-rich flames
the conversion of HCN to NO is slow and the poisonous HCN can survive in the postflame gases (chapter 6). As mentioned in chapter 5 and 6, and reference therein, both
acetylene and HCN have been investigated in flat, atmospheric-pressure premixed
hydrocarbon flames. However, to our knowledge no quantitative studies have been
reported of the effects of hydrogen addition on the formation and destruction of C2H2
and HCN in premixed CH4/air flames. Since there are substantial concentrations of
HCN and C2H2 in rich premixed flames, we study these effects in flames similar to
those discussed in chapter 5 and 6. Towards this end we measured the profiles of
C2H2 and HCN in rich-premixed H2/CH4/air flames at atmospheric pressure. The
technique used is microprobe gas sampling followed by analyses using tunable diode
laser absorption spectroscopy and wavelength modulation absorption spectroscopy
with second harmonic detection (chapter 4). In addition, the experimental
observations are compared with one-dimensional flame calculations.
125
Chapter 7
7.2 Experimental
The flat, atmospheric-pressure premixed flames of CH4/H2/air have been
stabilized on a 6-cm diameter flat-flame water-cooled Mc-Kenna Products burner and
surrounded by a coflow of nitrogen to prevent from mixing with ambient air (chapter
4). The experimental conditions are summarized in table 7.1.
Table. 7.1. Flame parameters
Flame
ϕ
H2 (%)
ρv, g/cm2·s)
T, K (GRI-3.0)
A
1.5
0
0.005
1760
B
1.5
23
0.005
1740
C
1.3
0
0.005
1763
D
1.3
23
0.005
1743
E
1.4
0
0.008
1854
F
1.4
20
0.008
1833
G
1.4
50
0.010
1834
H
1.4
50
0.008
1797
I
1.5
0
0.008
1848
J
1.5
20
0.008
1829
K
1.5
50
0.008
1796
Measurements were obtained at different vertical positions in the flame by
moving the burner with a precision positioner up to a distance of 10 mm in steps of
1 mm. As was done in the previous chapters, all measured HCN and C2H2
concentration profiles are shifted 1.3 mm upstream to correct for the probe distortion.
The methods for obtaining the C2H2 and HCN mole fractions, via tunable diode laser
absorption spectroscopy are described in detail in chapter 4.
Given the excellent predictive power of Chemkin II [13] with GRI-Mech 3.0
[14] for predicting the flame temperature observed in chapter 5 for methane/air
flames, and in other studies in methane/air [15] and hydrogen/air flames [16],
calculated flame temperatures were used in this study (Table 7.1). Calculations were
126
Chapter 7
performed by using the measured mass flux through the burner surface as an input
parameter.
7.3 Results and discussion
The addition of hydrogen to the unburned methane/air mixture changes the
flame properties. The laminar flame velocity of pure hydrogen is 8 times higher that
that of pure methane [4]; we thus expect that the flame velocity of methane to increase
substantially upon hydrogen addition. When the flame is stabilized above the burner
surface, we expect hydrogen addition to cause a significant temperature reduction; at
constant mass flux through the burner, the higher flame velocity should result in more
heat transfer to the burner. To illustrate this, figure 7.1 presents the calculated flame
temperature as function of the mass flux for methane and a methane/hydrogen
mixture. As can be seen, replacement of 50% methane by hydrogen results, at
constant mass flux, in a substantial decrease in the calculated flame temperature,
caused by increased heat transfer to the burner.
Figure 7.1. Calculated flame temperature as function of the mass flux for a pure
methane (solid line) and a methane/hydrogen, 50/50 mixture (dashed line), both
having an equivalence ratio of ϕ=1.5.
127
Chapter 7
7.3.1 HCN profiles
To study the effect of hydrogen addition to fuel-rich CH4/air flames on the
formation and consumption of HCN, several flames with different equivalence ratios
(ϕ=1,3, 1.4 and 1.5) and hydrogen concentrations in the fuel mixture (0%, 20% and
50% by volume) have been studied (see Table 7.1). For the flames at ϕ=1.5 (flames IK) and ϕ=1.4 (flames E, F, H) the mass flux is kept constant with different hydrogen
content in the mixture. As expected from figure 7.1, hydrogen addition to the
methane/air mixtures while keeping the mass flux constant decreases the flame
temperature (I-K and E, H) slightly (see also Table 7.1). For the flames with
equivalence ratio ϕ=1.4 (flames E-G), the temperature was kept more or less constant
(±20K) by increasing the mass flux when more hydrogen was added.
The experimental profiles of HCN mole fraction in Figs. 7.2 and 7.3, for ϕ=1.4
and ϕ=1.5, respectively, show that as the fraction of hydrogen in the fuel mixture
increases, the HCN mole fraction decreases substantially, well outside the
measurement uncertainty. For example, increasing the hydrogen content in the fuel
from 0 to 50% lowers the HCN mole fraction by more than a factor of two, at both
equivalence ratios. According to reaction (R6.1) in Chapter 6, HCN should be linearly
related to the CH and N2 concentration. Furthermore, one could make the simple
“naïve” assumption that CH is directly proportional to the CH4 concentration in the
fuel/air mixture. Following this simple line of thought, replacing 20% (v/v) methane
by hydrogen in the fuel results in a reduction, by “dilution”, in the CH4 mole fraction
by ∼7% and ∼2% in the N2 mole fraction in the fuel-air mixture; taken together this
would in turn decreases the HCN concentration via reaction R6.1 by ∼10%. When
increasing the hydrogen fraction to 50% a HCN decrease of ∼30% could thus be
expected. That a factor of 2 decrease in HCN is observed for a hydrogen fraction of
50%, suggests an additional effect, besides dilution, on flame structure. Figure 7.2
shows, for the flames with 50% hydrogen addition (flames D and J), that lowering the
temperature by ∼ 60 K does not affect the HCN peak concentration significantly, as
also seen for the “pure” methane/air flames in chapter 6. Interesting to note in figures
7.2 and 7.3 is that at constant mass flux the HCN peak concentrations appears to shift
towards the burner surface as the fraction of hydrogen in the fuel increases from 0 to
50%, indicative of the higher degree of burner stabilisation caused by the increased
burning velocity.
128
Chapter 7
Figure 7.2. Axial profiles of HCN mole fraction in CH4/air flames and CH4/H2/air
flames, ϕ = 1.4. Symbols denote probe measurements in flames E (squares), F
(diamonds), G (circles) and H (triangles). The lines denote flame calculations using
GRI-Mech 3.0 in flames E (dashed), F (dotted) G (bold solid) and H (thin solid).
Figure 7.3. Axial profiles of HCN mole fraction in CH4/air flames and CH4/H2/air
flames, ϕ = 1.5. Symbols denote probe measurements in flames I (diamonds), J
(squares) and K (triangles). The solid lines denote flame calculations using GRIMech 3.0 in flames I (dashed), J (solid) and K (dotted).
129
Chapter 7
Comparison of the calculated and measured HCN concentration profiles at
ϕ=1.4 (figure 7.1) shows for all flames substantial overprediction of the HCN peak
concentration and significant slower HCN decay relative to the experimental results,
similar to those observed in Chapter 6 for the pure methane fuel. Moreover, whereas a
reduction in the measured HCN peak concentration at ϕ=1.4 is related to the amount
of hydrogen addition, with only a modest effect of temperature, the calculations
suggest that the HCN reduction is mainly related to the flame temperature and not due
to a “hydrogen” effect. For example, increasing the hydrogen content from 20%
(flame F) to 50% (flame G) at constant flame temperature does not reduce the
maximum calculated HCN concentration, but lowering the flame temperature while
keeping 50% hydrogen content in the fuel (flame H) results in a substantial reduction
in the calculated HCN peak concentration. Although it is tempting to further interpret
the calculations, we note that these observations are in contradiction with the
measurements. It is thus not prudent to pursue the analysis using GRI-Mech 3.0. A
possible explanation for the large quantitative and qualitative discrepancies between
calculated and measured HCN peak concentrations observed at ϕ=1.4 (figure 7.2) is
that in GRI Mech 3.0 the temperature sensitive reaction CH+N2=HCN+N (R6.1,
chapter 6) is the main source of formation of HCN, while as discussed in Chapter 6
recent theoretical studies [17,18] point to NCN and not HCN as the primary product
of this reaction.
In contrast to the poor agreement at ϕ=1.4, figure 7.3 shows for the flames
studied at ϕ=1.5 moderately good agreement between measured and calculated HCN
profiles. However, as observed above, the calculations are very sensitive to burner
stabilization, as reflected in the flame temperature, and we ascribe the “good”
agreement with the trend observed here for ϕ=1.5 as a spurious effect of the changes
in flame temperature. This “temperature” effect is even more dramatically illustrated
in figure 6.3; where the experimental profiles differ only modestly for a change in
flame temperature of ~100 K, while the calculations change by a factor of 3. Due to
the significant quantitative discrepancies between calculated and measured HCN
profiles observed at ϕ=1.4 (figure 7.2), and for the spurious effect of temperature
observed at ϕ=1.5 in Chapter 6 (figure 6.3), and suspected in figure 7.3, we refrain
from the further mechanistic analyse the effect of hydrogen on the HCN mole
fraction. Based on the results reported here, a reanalysis of the chemical mechanism is
130
Chapter 7
deemed necessary. To do so, detailed information on the NCN kinetics is must be
obtained. We hope that the experimental data presented here will be useful in
performing this task.
7.3.2 C2H2 profiles
The effect of replacing 23% methane by hydrogen, at constant mass flux, on the
C2H2 profiles is illustrated in figure 7.4. The results show for equivalence ratios 1.3
and 1.5 (Flames A-D in Table 1) that H2 addition only slightly decreases the measured
peak C2H2 mole fraction. This small reduction of C2H2 is essentially the same as the
∼15% decrease “naively” expected from dilution when 23% methane is replaced by
hydrogen, here too making the simple assumption that C2H2 formation is directly
proportional to the concentration of hydrocarbons in the fuel. No substantial
difference in the measured C2H2 concentration is observed in the post flame zone for
the flames with (B, D) and without hydrogen addition (A, C).
In addition, the predicted C2H2 profiles obtained using GRI-Mech 3.0 are compared
with the measurements. As can be seen from figure 7.4, the calculated C2H2 profile
(dashed line) shows substantially higher peak concentrations and slower decay in the
post flame zone for flame D. Although not presented, the computed C2H2 profiles for
the other flames shown in figure 7.4 show similar discrepancies, as expected from the
results presented in Chapter 5 and in Ref. [19]. Replacing the rate coefficient of the
reaction C2H2 + OH → CH2CO +H used in the GRI 3.0 mechanism by the expression
4.87 x 1013exp(-12000cal/RT) cm3mol-1s-1 recommended in chapter 5 yields very
good agreement between measurements and calculations for all flames studied, as can
be seen in the figure.
131
Chapter 7
Figure 7.4. Axial profiles of HCN mole fraction in CH4/air flames and CH4/H2/air
flames, ϕ = 1.5 and 1.3. Symbols denote probe measurements in flames A (circles), B
(squares), C (diamonds) and D (triangles). The dashed line denotes flame the
calculation with GRI-3.0, and other lines, flame A (bold solid lines), flame B (thin
solid line), flame C (bold doted line) and flame D (thin dotted line) are the result of
calculations with the increased rate coefficient for C2H2+OH→CH2CO+H.
7.4 Conclusion
In this chapter we examined the effect of hydrogen addition on the formation
and destruction of HCN and C2H2 in rich premixed CH4/air mixtures. The HCN
measurements at equivalence ratios ϕ=1.4 and 1.5 show that the HCN mole fraction
decreases substantially with increasing hydrogen content in the fuel mixture. This
decrease significantly exceeds the reduction in HCN expected from the dilution of the
hydrocarbon fuel and nitrogen when hydrogen is added to the mixture, suggesting that
addition of hydrogen affects the flame structure related to the formation of HCN. In
contradiction to the measurements, calculations, using GRI-Mech 3.0 show at ϕ=1.4
and constant flame temperature no substantial reduction in the HCN peak
concentration when hydrogen is added to the mixture, while a reduction in flame
temperature results in a substantial decrease in the calculated HCN peak concentration
at the same hydrogen fraction in the fuel (50% v/v). Moreover, calculations predict
substantially higher HCN peak mole fractions and slower oxidation in the post flame
132
Chapter 7
zone as compared to the measurements for all flames at ϕ=1.4. Increasing the
hydrogen content in the fuel for the flames at ϕ=1.5 results in roughly the same
systematic decrease in the calculated HCN concentration as measured. However, this
observed reduction is probably caused by the reduction in temperature caused by
increased stabilization, which is not supported by the measurements presented in this
thesis, and not due to a “hydrogen effect”.
The C2H2 measurements show that the addition of 23% hydrogen results in only
a marginal reduction of the C2H2 concentration for the flames measured at ϕ=1.3 and
1.5, suggesting that hydrogen addition does not have any significant effect on the
flame processes responsible for C2H2 formation/consumption in CH4/air flames.
Comparison
between
calculated
and
measured
profiles
shows
significant
overprediction of the maximum C2H2 concentration and a much slower predicted
decay of C2H2 in the post flame zone. As expected from Chapter 5, replacing the rate
coefficient of the reaction C2H2 + OH → CH2CO +H used in the GRI-Mech 3.0
mechanism by 4.87 x 1013exp(-12000cal/RT) cm3mol-1s-1 resulted in good agreement
between measured and calculated C2H2 profiles for all flames studied.
133
Chapter 7
7.5 Literature
1.
I. Wierzba , W. Wang , Int. J. hydrogen energy 31 (2006) 485-489.
2.
M. Ibas , A. P. Crayford, I. Yilmaz , P. J. Bowen, N. Syred, int. J. hydrogen
energy 31 (2006) 1768-1779.
3.
F. Halter, C. Chauveau, N. Djebaili-Chaumeix, I. Gokalp , Proc.Combust.
Inst. 30 (2005) 201-208.
4.
B. E. Milton, J.C. Keck, Combust. Flame 58 (1984) 13-22.
5.
S. O. Bade Shrestha, G. A. Karim, Int. J. Hydrogen Energy 24 (1999) 577586.
6.
F. Cozzi A. Coghe , Int. J. hydrogen Energy 31 (2006) 669-677.
7.
A. R. Choudhui, S. R. Gollahalli, Int. J. Hydrogen Energy 25 (2000) 451462.
8.
A. R. Choudhui, S. R. Gollahalli, Int. J. Hydrogen Energy 25 (2000) 11191127.
9.
A. R. Choudhui , S. R. Gollahalli, Int .J. Hydrogen Energy 29 (2004)12931302.
10. J. Warnatz, H. Bockhorn, A. Mozer, H.W. Wenz, Proc. Combust. Inst 19
(1982) 197-209.
11. J. Warnatz, Ber. Bunsenges. Phys.Chem 87 (1983) 1008-1022
12. C. P. Fenimore, Proc. Combust. Inst. 13 (1971) 373-379.
13. R. J. Kee, F. M. Rupley, J. A. Miller, CHEMKIN II: A Fortran Chemical
Kinetics Package for the Analysis of Gas-Phase Chemical Kinetics., Sandia
National Laboratories, (1989).
14. G. P. Smith, D. M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M.
Goldenberg, C. T. Bowman, R. K. Hanson, S. Song, W. C. Gardiner, V.
Lissanski, Z. Qin, http://www.me.berkeley.edu/gri_mech/.
15. A. V. Sepman, A. V. Mokhov, H. B. Levinsky, Proc. Combust. Inst. 29
(2002) 2187-2194.
16. A. V. Sepman, Effect of burner stabilization on nitric oxide formation and
destruction in atmospheric pressure flat premixed flames, Ph.D. Thesis,
RUG, 2006 (ISBN 90-367-2702-2).
17. L.V. Moskaleva, W.S. Xia, M.C. Lin, Chem. Phys. Let. 331 (2-4) (2000)
269-277.
134
Chapter 7
18. L.V. Moskaleva, M.C. Lin, Proc. Combust. Inst. 28 (2000) 2393-2401.
19. E. W. Kaiser, T. J. Wallington, M. D. Hurley, J. Platz, H.J. Curran, W. J.
Pitz, C. K. Westbrook, J. Phys. Chem. A., 104 (2000) 8194-8206.
135
Summary
Increasingly stringent regulations regarding CO2 emissions, and the expectation
that fossil fuel reserves will be exhausted within this century, have led to interest in
the use of admixtures of hydrogen in natural gas as an alternative fuel in combustion
devices. Combustion equipment is generally tuned for local fuel used, and clearly a
change in fuel must not lead to deterioration in performance. Since the combustion
properties of hydrogen differ in many respects from those of natural gas (Chapter 1),
there are concerns regarding the possible negative response of combustion equipment
such as gas engines, burners and turbines when fuelled with hydrogen-enriched
natural gas. For example, the presence of hydrogen might increase pollutant emissions
from combustion devices and cause knock (uncontrolled ignition) in gas engines. To
understand these practical consequences properly, it is necessary to study the changes
in the underlying physical and chemical processes that are responsible for changes in
combustion behavior when hydrogen is added. The autoignition properties presented
in this thesis provide new insight into the ignition characteristics of methane,
hydrogen and methane/hydrogen fuel mixtures under conditions relevant to knock in
gas engines. In addition, the spatial profiles of C2H2 and HCN (important precursors
of soot and NOx, respectively) measured in atmospheric-pressure, one-dimensional
CH4/air and CH4/H2/air flames provide insight into changes in pollutant formation
upon hydrogen addition. To test the accuracy of different chemical mechanisms,
which could be used to predict the combustion behaviour of natural gas/hydrogen
mixtures, the measurements presented in this thesis are compared with the results of
numerical simulations.
To study autoignition under strictly controlled experimental conditions relevant
to gas engines, a Rapid Compression Machine (RCM) was constructed in our
laboratory based upon a design from MIT (Chapter 2). Test results showed that the
RCM is able to compress the combustible gas-air mixture to final pressures up to ∼70
bar and temperatures up to ∼1100 K, where the majority of the pressure rise in the
compression period takes place in a very short time (<3 ms). The temperature of the
compressed mixture is calculated from the measured pressure by using the isentropic
relations of an ideal gas, the uncertainty in the peak temperature after compression is
better than ±3.5 K in the range of pressures of interest (10-70 bar).
Chapter 3 presents the experimental study of autoignition delay times of
methane/hydrogen mixtures at high pressure (10-70 bar) and moderate temperatures
136
Summary
(960-1100 K). Under stoichiometric conditions the experimental results show that
replacing methane by hydrogen results in a reduction in the measured ignition delay
time. At moderately low concentrations of hydrogen (≤ 20%) only a weak effect on
the measured ignition time is observed, while at 50% hydrogen content in the fuel a
substantial reduction in ignition delay time is seen under all measured conditions.
Moreover, the measurements show that the effects of hydrogen in promoting ignition
increases with temperature and decreases with pressure. Experimental results for 50%
hydrogen in the fuel at equivalence ratio ϕ = 0.5 are essentially identical to those at
ϕ = 1.0. These results suggest that the adverse affects of hydrogen addition to natural
gas on engine knock may be limited for hydrogen fractions of only a few tens of
percent. Comparison between measured and calculated ignition delay times shows
very good agreement for all fuel mixtures using the proposed mechanism of Petersen
et al. (E. L. Petersen, D. M. Kalitan, S. Simmons, G. Bourgue, H. J. Curran and J. M.
Simmie, Proc. Combust. Inst .31 (2007) 447-454.)
Chapter 4 describes the experimental protocols for concentration measurements
of HCN and C2H2 in one-dimensional flames using extractive probe sampling
followed by analysis using tunable diode laser absorption spectroscopy (TDLAS) at
∼1.5 μm. The calibration procedure for acetylene is performed by measuring the
absorption coefficient in a gas sample containing a known concentration of acetylene
under the same experimental conditions as those existing for the flame samples. At
mole fractions above 1000 ppm, the accuracy of the measured C2H2 is ∼5%;
decreasing C2H2 mole fraction results in deteriorating accuracy, to 15% at 100 ppm.
The same calibration procedure is performed for the HCN measurements. However, to
increase the sensitivity wavelength modulation absorption spectroscopy (WMAS)
with second harmonic detection is used. The accuracy of the measured HCN mole
fraction in the sampled flame gases is better than 30% at concentrations above 3 ppm.
Before examining the effects of hydrogen addition on the formation and consumption
of these species, their profiles are first measured and analyzed in flames using pure
methane as fuel.
Chapter 5 presents measurements of C2H2 concentration profiles in onedimensional
atmospheric-pressure
rich
premixed
methane/air
flames
using
spontaneous Raman scattering and an extractive probe sampling technique (Chapter
4). Excepting a shift of approximately 1.3 mm, resulting from the acceleration of the
137
Summary
combustion products in the probe orifice, the axial Raman and probe profiles are in
very good agreement. The measurements show that changing the equivalence ratio
from ϕ = 1.4 to 1.5 results in an increase of the peak C2H2 mole fraction by nearly a
factor two. At fixed equivalence ratio, the maximum C2H2 mole fraction depends only
slightly on the flame temperature, while the C2H2 oxidation in the post flame zone
increases substantially in the flames with increasing flame temperature. Comparison
of measured C2H2 profiles with those calculated, using the GRI-Mech 3.0 chemical
mechanism, shows a much faster post-flame decay in the experimental results.
Increasing the pre-exponential factor in the rate coefficient of reaction
C2H2 + OH → CH2CO + H to 1.7 x 1012 cm3mol-1s-1 in the range 1760 – 1850 K
yields excellent agreement between computed and experimental results.
In Chapter 6 the formation and consumption of HCN in fuel-rich atmospheric
pressure methane/air flames is discussed. Towards this end, axial HCN and
temperature profiles have been measured at equivalence ratios ϕ = 1.3, 1.4 and 1.5.
For the richest flame studied (ϕ = 1.5) very slow oxidation of HCN in the post flame
zone is observed, demonstrating “residual” HCN in the post flame gases of fuel-rich
methane/air flames. The HCN measurements show that increasing the flame
temperature at fixed equivalence ratio does not result in significant changes in the
HCN peak mole fraction, while the HCN oxidation in the post-flame gases increases
substantially. Decreasing the equivalence ratio leads to faster HCN oxidation in the
post flame zone. Large discrepancies are observed between measured and calculated
HCN profiles using GRI-Mech 3.0. Attempts to bring the calculations using GRIMech 3.0 into agreement with the experimental observations by changing the rates of
key formation and consumption reactions within the uncertainties in the literature
were unsuccessful. Consideration of NCN as a primary product of the reaction
between CH and N2, based on recent theoretical studies, allows improvement in the
agreement between measured and calculated HCN concentrations. However, the lack
of information on the rate constants of the NCN reactions at high temperatures
precludes unambiguous conclusions regarding this mechanism.
Chapter 7 is an extension of Chapters 5 and 6 and examines the effect of
hydrogen addition on the formation and consumption of HCN and C2H2 in fuel-rich
stabilized methane/air flames. The HCN measurements at ϕ = 1.4 and 1.5 show that
increasing the hydrogen fraction in the mixture at constant flame temperature results
138
Summary
in a substantial decrease in the HCN mole fraction. This decrease in HCN
significantly exceeds the reduction of HCN caused by the dilution of hydrocarbon fuel
when hydrogen is added to the mixture, indicating that hydrogen addition affects the
chemistry related to the formation of HCN. As observed for pure methane flames
(Chapter 6) at ϕ = 1.4, the calculated HCN profiles using GRI-Mech 3.0 predict
significantly higher HCN peak mole fractions and substantially slower decay in
comparison to the measurements for all flames studied. Moreover, contrary to the
experimental observations, the calculations show no substantial changes in the
calculated peak HCN mole fraction when adding hydrogen to the fuel mixture and a
strong reduction in the HCN mole fraction with decreasing flame temperature at
constant hydrogen fraction. Good agreement between calculations and measurements
is found for the flames at ϕ = 1.5. However, the reduction in the calculated HCN mole
fraction when hydrogen is added to the flame is probably the result of the computed
reduction in flame temperature and not due to a “hydrogen” effect on the chemistry.
No significant changes are observed in the measured C2H2 mole fractions for the
flames with equivalence ratio ϕ=1.3 and 1.5 when hydrogen is added. This suggests
that hydrogen addition does not have a significant effect on the chemistry responsible
for C2H2 in CH4/air flames. Replacing the rate coefficient of the reaction C2H2 + OH
→ CH2CO +H used in the GRI 3.0 mechanism by the rate recommended in Chapter 6
resulted in good agreement between measured and calculated C2H2 profiles for all
flames studied.
139
Samenvatting
De steeds strengere eisen ten aanzien van CO2 emissies en de verwachting dat nog
binnen deze eeuw de reserves van fossiele brandstof uitgeput zullen zijn, heeft geleid
tot toenemende belangstelling in het gebruik van aardgas/waterstof mengsels als
alternatieve brandstof in verbrandingsapparatuur. Deze apparatuur is gewoonlijk
afgesteld voor de locale brandstof, en het zal duidelijk zijn dat veranderingen in
brandstofsamenstelling niet tot verslechterd gedrag mag leiden. Doordat de
verbrandingseigenschappen van waterstof sterk verschillen ten opzichte van aardgas
(Hoofdstuk 1), is de vraag in hoeverre het gedrag van verbrandingsapparatuur zoals
gasmotoren, branders en turbines (negatief) beïnvloed zal worden door het aardgas
met waterstof te verrijken. De aanwezigheid van waterstof kan bijvoorbeeld de
vorming van milieuschadelijke emissies bevorderen bij verbrandingsapparatuur en
leiden tot klopverschijnselen (ongecontroleerde ontsteking) in gasmotoren. Om deze
praktische consequenties adequaat te doorgronden is het essentieel de veranderingen
in de onderliggende fysische en chemische processen die verantwoordelijk zijn voor
het verbrandingsgedrag te onderzoeken, wanneer waterstof wordt toegevoegd. De
zelfontstekingseigenschappen vermeld in dit proefschrift geven nieuwe inzichten in
het
ontstekingsgedrag
van
methaan,
waterstof
en
methaan/waterstof
brandstofmengsels, onder condities relevant voor klopverschijnselen in gasmotoren.
Daarnaast geven de ruimtelijke verdelingen van de molfracties van C2H2 en HCN,
belangrijke tussencomponenten in de vorming van respectievelijk roet en NOx,
gemeten in één-dimensionale CH4/lucht en CH4/H2/lucht vlammen bij atmosferische
druk inzicht in veranderingen in de vorming van milieuschadelijke stoffen als gevolg
van waterstof toevoeging. De metingen die in dit proefschrift worden gepresenteerd
worden ook gebruikt om de nauwkeurigheid te testen van verschillende chemische
mechanismen die gebruikt zouden kunnen worden voor het voorspellen van het
verbrandingsgedrag van waterstof/aardgas mengsels.
Op basis van een ontwerp van MIT is in ons lab een Rapid Compression
Machine (RCM) gebouwd, waarmee onder gecontroleerde omstandigheden die
relevant zijn voor gasmotoren zelfontsteking kan worden onderzocht (Hoofdstuk 2).
Resultaten tonen aan dat de RCM een brandbaar mengsel tot drukken tot ~70 bar en
temperaturen tot ∼1100 K comprimeert, waarbij het merendeel van deze
drukverhoging plaatsvindt in een zeer korte tijd (<3 ms). De temperatuur van het
gecomprimeerde mengsel is berekend aan de hand van de gemeten druk, door gebruik
te maken van de isentropische relaties van een ideaal gas. Binnen het bereik van de
140
Samenvatting
gemeten drukken (10-70 bar) is de onzekerheid in de berekende piektemperaturen na
compressie kleiner dan ± 3.5 K.
Hoofdstuk 3 presenteert de experimentele studie van zelfontstekingstijden van
methaan/waterstof mengsels onder hoge drukken (10-70 bar) en gematigde
temperaturen (960-1100K). Onder stoichiometrische condities tonen de metingen aan
dat het vervangen van methaan door waterstof resulteert in een verlaging van de
gemeten ontstekingstijd. Toevoeging van lage concentraties waterstof (<20%) laat
slechts een bescheiden effect op de gemeten ontstekingstijd zien, terwijl bij de
mengsels met 50% waterstof in de brandstof een substantiële verlaging in de
ontstekingstijd wordt waargenomen. Tevens volgt uit de metingen dat het effect van
waterstof in het bevorderen van ontsteking toeneemt met toenemende temperatuur en
afneemt bij toenemende druk. De experimentele resultaten verkregen bij 50%
waterstof in de brandstof en een equivalentieverhouding ϕ = 0.5 zijn nagenoeg
identiek aan de resultaten verkregen bij ϕ = 1.0. De resultaten suggereren dat de
toevoeging van slechts enkele tientallen procent waterstof aan aardgas waarschijnlijk
een beperkte invloed zal hebben op klopverschijnselen in gasmotoren. De vergelijking
van de gemeten en berekende ontstekingstijden met het chemische mechanisme van
Petersen e.a (E. L. Petersen, D. M. Kalitan, S. Simmons, G. Bourgue, H. J. Curran and
J. M. Simmie, Proc. Combust. Inst. 31 (2007) 447-454.) laten voor alle
brandstofmengsels zeer goede overeenkomsten zien.
Hoofdstuk 4 beschrijft de meetprotocollen voor het bepalen van HCN- en C2H2concentraties in één-dimensionale vlammen. Hierbij worden monsters van de hete
gassen met behulp van een afzuigprobe genomen en geanalyseerd met behulp van
tunable diode laser absortption spectroscopy (TDLAS) bij 1.5 μm. De
calibratieprocedure voor acetyleen wordt uitgevoerd door de absorptiecoëfficiënt te
bepalen van een gasmonster met een bekende concentratie van acetyleen onder
dezelfde condities die gelden voor de vlammonsters. Voor molfracties groter dan
1000 ppm is de nauwkeurigheid van de gemeten C2H2 ca. 5%; de nauwkeurigheid
neemt af bij afnemende molfractie, oplopend tot 15% bij 100 ppm acetyleen. Voor de
HCN-metingen is dezelfde calibratie methode toegepast als voor C2H2. Echter, voor
het verhogen van de gevoeligheid is wavelength modulation absorption spectroscopy
(WMAS) toegepast, met detectie op de tweede harmonische frequentie. De
141
Samenvatting
nauwkeurigheid van de gemeten HCN molfractie in de monstergassen is beter dan
30% voor molfracties >3 ppm.
Hoofdstuk 5 presenteert metingen van de profielen van C2H2-molfracties in
één-dimensionale
brandstofrijke
voorgemengde
methaan/lucht
vlammen
bij
atmosferische druk met behulp van zowel spontane Ramanverstrooiing als de
afzuigprobe-techniek (Hoofdstuk 4). Behalve een verschuiving van ongeveer 1.3 mm,
als gevolg van de versnelling van verbrandingsproducten in de probe orifice, tonen de
Raman- en probe- profielen zeer goede overeenkomst. Uit de meetresultaten volgt dat
verandering van de equivalentieverhouding van ϕ = 1.4 naar 1.5 resulteert in een
verhoging van de piekconcentratie van bijna een factor twee. Bij constante
equivalentieverhouding is de C2H2 piekmolfractie slechts weinig afhankelijk van de
vlamtemperatuur, terwijl de C2H2-oxidatie in de postvlam-zone substantieel toeneemt
voor de vlammen met toenemende vlamtemperatuur. Vergelijking van de gemeten en
de op basis van het GRI 3.0 chemische mechanisme berekende C2H2-profielen laat
een veel snellere afname in C2H2-molfractie in de metingen zien. Verhogen van de
pre-exponentiële factor in de reactiesnelheidscoëfficiënt van de reactie C2H2 + OH →
CH2CO +H naar 1.7 x 1012 cm3mol-1s-1 in het gebied 1760-1850 K leidt tot zeer goede
overeenkomst tussen de berekende en gemeten resultaten.
In hoofdstuk 6 wordt de vorming en afbraak van HCN in brandstofrijke
methaan/lucht vlammen bij atmosferische druk besproken. Hiertoe zijn axiale HCNen temperatuur- profielen gemeten voor de equivalentieverhoudingen ϕ = 1.3, 1.4 en
1.5. Voor de brandstofrijkste vlam (ϕ = 1.5), is een zeer langzame oxidatie van HCN
waargenomen in de postvlam-zone, waarmee “resterend” HCN in de postvlam-gassen
aangetoond is. De HCN-metingen laten zien dat het verhogen van de vlamtemperatuur
bij constante equivalentieverhouding een gering effect heeft op de HCNpiekmolfractie, maar dat door de verhoogde temperatuur de HCN oxidatie substantieel
versnelt in de postvlam-zone. Berekende HCN profielen op basis van het GRI 3.0
mechanisme laten grote verschillen zien met de meetresultaten. Pogingen om de
berekeningen beter in overeenstemming te brengen met de meetresultaten, door
middel van het veranderen van snelheden van belangrijke formatie en consumptie
reacties zijn onsuccesvol gebleken. Door in plaats van HCN, NCN als primair product
van de reactie CH en N2 te beschouwen, is verbetering tussen berekende en gemeten
HCN
142
molfracties
mogelijk.
Echter,
het
gebrek
aan
informatie
over
Samenvatting
reactiesnelheidsconstanten van de reacties van NCN bij hoge temperatuur verhindert
het trekken van duidelijke conclusies over dit mechanisme.
Hoofdstuk 7 verlegt de inhoud van Hoofdstukken 5 en 6 en beschrijft het effect
van de toevoeging van waterstof op de formatie en afbraak van HCN en C2H2 in
brandstofrijke methaan/lucht vlammen. De HCN-metingen bij ϕ = 1.4 en 1.5 laten
zien dat het verhogen van de waterstoffractie in het mengsel bij constante
vlamtemperatuur resulteert in een substantiële verlaging in de HCN molfractie. Deze
verlaging in HCN is significant groter dan de verlaging die verwacht wordt door de
verdunning van het methaan als gevolg van waterstoftoevoeging. Dit duidt aan dat de
vormingschemie van HCN wordt beïnvloed door de toevoeging van waterstof. Net als
waargenomen
voor
pure
methaanvlammen
(Hoofdstuk
6),
voorspellen
de
berekeningen op basis van het GRI 3.0 mechanisme bij ϕ = 1.4 substantieel hogere
piekmolfracties van HCN dan gemeten, en is de voorspelde oxidatiesnelheid
langzamer. Daarnaast, in tegenstelling tot de experimentele waarnemingen, wordt
slechts een geringe invloed voorspeld op de berekende piekmolfractie van HCN bij
het toevoegen van waterstof en ook een sterke verlaging in HCN molfractie bij
dalende vlamtemperatuur bij constante waterstoffractie. Goede overeenkomst tussen
de metingen en berekeningen is gevonden voor de vlammen bij ϕ = 1.5. Echter, de
reductie in de berekende HCN-molfractie door toevoeging van waterstof is
hoogstwaarschijnlijk het resultaat van de berekende reductie in vlamtemperauur en
niet het gevolg van een “waterstofeffect”. Voor de equivalentieverhoudingen ϕ = 1.3
en 1.5 zijn geen significante veranderingen waargenomen in de gemeten C2H2molfracties als gevolg van toevoeging van waterstof. Dit suggereert dat
waterstoftoevoeging geen significant effect heeft op de chemie verantwoordelijk voor
C2H2 in methane/lucht vlammen. Vervangen van de reactiesnelheidscoëfficiënt van de
reactie C2H2 + OH → CH2CO +H in het GRI 3.0 mechanisme door de waarde
voorgesteld in Hoofdstuk 6 levert voor alle bestudeerde vlammen zeer goede
overeenkomsten tussen de gemeten en berekende C2H2 profielen.
143
Dankwoord
Promoveren is in tegenstelling tot wat veel mensen denken geen individualistisch
karwei, maar een proces dat in samenwerking met vele anderen tot stand komt.
Iedereen die een bijdrage heeft geleverd aan dit proefschrift wil ik bij deze bedanken
en een paar mensen in het bijzonder.
Ten eerste wil ik Prof. Dr. Howard Levinsky van harte bedanken voor het feit
dat ik bij hem mijn promotieonderzoek heb mogen uitvoeren. Tijdens mijn doctoraal
onderzoek werd mij duidelijk dat ik mij verder wilde verdiepen in de
verbrandingstechnologie en dit was mogelijk dankzij het promotieonderzoek dat je
mij aanbood. Verder wil ik je bedanken voor de leuke en leerzame discussies tijdens
de wekelijks terugkerende vakgroep besprekingen en de tips die je me hebt gegeven
tijdens het schrijven van artikelen en het proefschrift.
Mijn co-promotor dr. Anatolia Mokhov wil ik bij deze bedanken voor de intensieve
begeleiding op het gebied van onder andere laser diagnostiek en het schrijven van
artikelen. Tolja, je hebt mij versteld doen staan van je vakkennis en ik ben je zeer
dankbaar dat je deze kennis, waar mogelijk, met mij hebt gedeeld. Verder wil ik je
bedanken voor de prettige samenwerking en de leuke tijd op de Universiteit.
Dit onderzoek is tot stand gekomen dankzij de financiësle steun van het
programma
EET (Economie, Ecologie, Technologie), waarvoor dank. De
beoordelingscommissie bestaande uit Prof. dr. ir. R. Baert, Prof. dr. H.C. Moll,
Prof. en dr. ir. Th.H. van der Meer wil ik bedanken voor hun beoordeling van het
proefschrift.
Daarnaast wil Marcel en Edwin van het bedrijf ERMA uitbesteding bijzonder
bedanken voor de constructie van de Rapid Compression Machine. Kees en Ubbel
hebben tijdens mijn promotieonderzoek zeer goed geholpen met onder andere het
leveren van gassen, hiervoor hartelijk dank.
Uiteraard wil ik mijn oud collega’s Alexei, Martijn, Nikolay en Vishal
bedanken voor alle hulp en gezelligheid, hierbij denk ik onder andere aan onze
vrijdagmiddag borrels in de Irish Pub. Especially, I wish to thank Nikolay for his
assistance in performing the autoignition measurements. Martijn, ik denk met veel
plezier terug aan de 4 jaar dat we een kantoortje en lab deelden. Niet alleen de
wetenschappelijke discussies waar ik veel van geleerd heb, maar ook onze befaamde
tafeltenniswedstrijden samen met Alexei waren zeer geslaagd.
144
Dankwoord
Naast werk was er tijd voor ontspanning, daarvoor wil ik mijn vrienden erg
bedanken. Natuurlijk wil ik mijn familie bedanken en in het bijzonder mijn ouders;
Wim en Anneke voor hun steun en vertrouwen in mij.
Verder wil ik Daniëla speciaal bedanken voor het ontwerpen van de prachtige
omslag van mijn proefschrift.
En natuurlijk wil ik Sandra bedanken voor al haar liefde en energie die ze me heeft
gegeven om door te gaan met mijn promotieonderzoek, met name tijdens de
momenten waarin het onderzoek niet verliep zoals ik dat wilde.
145