Download Report

Proceedings
of the
Proceedings of the Combustion Institute 30 (2005) 2519–2527
Combustion
Institute
www.elsevier.com/locate/proci
Homogeneous ignition of CH4/air and H2O and
CO2-diluted CH4/O2 mixtures over Pt; an experimental
and numerical investigation at pressures up to 16 bar
Michael Reinke, John Mantzaras*, Rolf Schaeren, Rolf Bombach,
Andreas Inauen, Sabine Schenker
Paul Scherrer Institute, Combustion Research, CH-5232 Villigen PSI, Switzerland
Abstract
The homogeneous ignition of CH4/air, CH4/O2/H2O/N2, and CH4/O2/CO2/N2 mixtures over platinum
was investigated experimentally and numerically at pressures 4 bar 6 p 6 16 bar, temperatures
1120 K 6 T 6 1420 K, and fuel-to-oxygen equivalence ratios 0.30 6 u 6 0.40. Experiments have been
performed in an optically accessible catalytic channel-ﬂow reactor and included planar laser induced
ﬂuorescence (LIF) of the OH radical for the determination of homogeneous (gas-phase) ignition and
one-dimensional Raman measurements of major species concentrations across the reactor boundary layer
for the assessment of the heterogeneous (catalytic) processes preceding homogeneous ignition. Numerical
predictions were carried out with a 2D elliptic CFD code that included elementary heterogeneous and
homogeneous chemical reaction schemes and detailed transport. The employed heterogeneous reaction
scheme accurately captured the catalytic methane conversion upstream of the gaseous combustion zone.
Two well-known gas-phase reaction mechanisms were tested for their capacity to reproduce measured
homogeneous ignition characteristics. There were substantial diﬀerences in the performance of the two
schemes, which were ascribed to their ability to correctly capture the p–T–u parameter range of the
self-inhibited ignition behavior of methane. Comparisons between measured and predicted homogeneous
ignition distances have led to the validation of a gaseous reaction scheme at 6 bar 6 p 6 16 bar, a pressure
range of particular interest to gas-turbine catalytically stabilized combustion (CST) applications. The
presence of heterogeneously produced water chemically promoted the onset of homogeneous ignition.
Experiments and predictions with CH4/O2/H2O/N2 mixtures containing 57% per volume H2O have shown
that the validated gaseous scheme was able to capture the chemical impact of water in the induction zone.
Experiments with CO2 addition (30% per volume) were in good agreement with the numerical simulations
and have indicated that CO2 had only a minor chemical impact on homogeneous ignition.
2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Keywords: High-pressure homogeneous ignition of CH4/air over Pt; Eﬀect of H2O and CO2 on ignition; LIF and Raman
1. Introduction
*
Corresponding author. Fax: +41 56 310 21 99.
E-mail address: ioannis.mantzaras@psi.ch (J.
Mantzaras).
The application of catalytically stabilized combustion (CST) to large-scale gas turbines has been
actively pursued over the last years [1] as a means
1540-7489/$ - see front matter 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.proci.2004.08.054
2520
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
to mitigate NOx emissions with direct combustion
strategies rather than with indirect exhaust-gas
aftertreatment techniques. The knowledge of the
heterogeneous (catalytic) and of the low-temperature homogeneous (gas-phase) kinetics of methane is key to the development of advanced
numerical CST models [2,3]. Validation of the heterogeneous reaction scheme of Deutschmann
et al. [3] for the complete oxidation of CH4 over
Pt at pressures up to 16 bar was recently reported
in Reinke et al. [4]: the catalytic reactivity was assessed with in situ Raman measurements of major
species concentrations over a channel-ﬂow boundary layer. The onset of homogeneous ignition is
detrimental to the catalyst integrity and, therefore, the availability of validated gaseous reaction
schemes is crucial in CST reactor design; such
schemes can also be used to ﬁne-tune analytical
CST homogeneous ignition criteria [5,6], so as to
provide a fast—albeit more restrictive—alternative to detailed computations. To this direction,
gaseous reaction schemes in CST have been validated only at low-to-moderate pressures. Reinke
et al. [7] have shown, encompassing earlier atmospheric-pressure studies [8], the applicability of the
gaseous scheme of Warnatz and Maas [9] in CST
of CH4/air over Pt at pressures up to 6 bar, a
range of interest to microreactors.
The present study undertakes a combined
experimental and numerical investigation of
CH4/air CST over Pt, with the main objective
of providing validated homogeneous reaction
schemes at gas-turbine-relevant conditions. Particular objectives were to assess the CST applicability of gaseous schemes in the presence of large
H2O or CO2 dilution (an issue of interest in gasturbines with exhaust gas recycle [10]), and to
study the hetero/homogeneous chemistry coupling at high pressures. Experiments were performed in an optically accessible catalytic
laminar channel-ﬂow reactor at pressures
4 6 p 6 16 bar. The onset of homogeneous ignition was assessed with planar laser induced ﬂuorescence (LIF) of the OH radical, and the
catalytic processes preceding homogeneous ignition were investigated with one-dimensional Raman
measurements
of
major
species
concentrations. The numerical predictions included an elliptic two-dimensional CFD code
with elementary heterogeneous and homogeneous chemical reaction schemes and detailed
transport.
2. Experimental
2.1. High-pressure test-rig
The test-rig (Fig. 1) consisted of a rectangular
reactor, that formed a liner inside a high-pressure
cylindrical vessel [7]. The reactor comprised of
Fig. 1. Schematic of the test-rig and the LIF/Raman
set-up.
two horizontal Si[SiC] ceramic plates (300-mm
long (x), 110-mm wide (z), 9-mm thick, and positioned 7-mm (y) apart) and two 3-mm thick vertical quartz windows [11]. The inner Si[SiC]
surfaces were coated via plasma vapor deposition
with a 1.5 lm thick non-porous Al2O3 layer, followed by a 2.2 lm thick Pt layer. Measurements
of the total and active catalyst areas with BET
(Kr-physisorption)
and
CO-chemisorption,
respectively, veriﬁed the absence of a porous surface structure. The surface temperature along
the x–y symmetry plane was measured by S-type
thermocouples (12 for each plate) embedded
0.9 mm beneath the catalyst, through holes eroded
from the outer Si[SiC] surfaces. The plate temperatures were controlled by two resistive heaters
positioned above the ceramic plates.
Air was preheated and mixed with CH4 in
two sequential static mixers. The CH4/air premixture was driven into the reactor through a
50-mm long inert rectangular honeycomb section that provided uniform inlet velocity. A
thermocouple positioned at the downstream
end of the honeycomb measured the reactor inlet temperature. Optical accessibility from both
reactor sides was maintained by two 350-mmlong and 35-mm-thick quartz windows on the
high-pressure tank (see Fig. 1). Two additional
quartz windows, one located at the rear ﬂange
of the high-pressure tank and the other (not
shown in Fig. 1) at the exhaust section of the
reactor, provided a counterﬂow streamwise optical access for the LIF experiments. In the tests
with H2O dilution, superheated steam was supplied by an AWTEC-DLR steam generator. In
both H2O- and CO2-dilution tests, the oxidizer
was pure oxygen; nitrogen was separately added
as a balance.
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
2.2. Laser diagnostics
The LIF/Raman set-up is depicted in Fig. 1.
The 532-nm radiation of a frequency-doubled
pulsed Nd:YAG laser (Quantel YG781C20) was
directed to the Raman- or to the LIF-set-up by
a traversable mirror. As the experimental conditions were steady, simultaneous acquisition of Raman and LIF was not necessary. For the OH-LIF,
the 532-nm radiation pumped a tunable dye laser
(Quantel TDL50); its frequency-doubled radiation (285 nm) had a pulse energy of 0.5 mJ, low
enough
to
avoid
saturation
of
the
A (v = 1) ‹ X (v0 = 0) transition. The 285 nm
beam was transformed into a laser sheet (by a
cylindrical lens telescope and a 1-mm slit mask)
that propagated counterﬂow, along the x–y symmetry plane (Fig. 1). The ﬂuorescence of both
(1–1) and (0–0) transitions at 308 and 314 nm,
respectively, was collected at 90 (through the
reactor and tank side-windows) with an intensiﬁed
CCD camera (LaVision FlameStar 2F, 576 · 384
pixels). A 120 · 7 mm2 section of the combustor
was imaged on a 576 · 34 pixel CCD-area. The
camera was traversed axially to map the 300 mm
reactor extent; at each measuring location, 400
images were averaged. The LIF was calibrated
with absorption measurements performed with
the laser beam crossing the reactor laterally
through both side windows, as in [7].
Compared to previous homogeneous ignition
studies [7,8] that employed only OH-LIF, the Raman measurements have eliminated uncertainties
with regard to the heterogeneous pathway: the
high surface temperatures could, potentially, lead
to a partial catalyst deactivation, and hence to a
near-wall fuel excess that could, in turn, promote
homogeneous ignition and complicate the assessment of the gaseous reactivity. In the Raman tests,
the 532-nm beam was focused through the tank
and reactor side-windows into a vertical line
(0.3 mm thick) by an f = 150 mm cylindrical
lens. The focal line spanned the 7-mm channel
separation and was oﬀset laterally (z = 15 mm)
to increase the collection angle and minimize thermal beam steering, as in [11]. Two f = 300 mm
lenses collected the scattered light at a 50 angle
with respect to the sending optical path and focused it to the entrance slit of a 25-cm imaging
spectrograph (Chromex-250i) equipped with an
intensiﬁed CCD camera identical to that of the
LIF-set-up. The 576- and 374-pixel-long CCD
dimensions corresponded to wavelength and
transverse distance, respectively; in the latter
dimension, 250 pixels resolved the 7-mm gap.
The eﬀective Raman cross-sections, which included transmission eﬃciencies, were evaluated
by recording the signals of pure CH4, air, and
completely burned gases of known composition.
Spectroscopic data for the CH4 and H2O Raman
cross-sections were taken from Steiner [12]. Ra-
2521
man data were acquired at diﬀerent positions by
traversing axially an optical table that supported
the sending and collecting optics (Fig. 1). The
250-pixel-long 7-mm distance was binned to 63
pixels, providing a resolution of 0.11 mm. Raman
data points closer than 0.6 mm to both walls were
discarded due to low signal-to-noise ratio.
3. Numerical
Simulations were carried out with an elliptic,
2D CFD code [2,11]. The elementary heterogeneous scheme of Deutschmann et al. [3] (24 reactions, 11 surface, and 9 gaseous species) was
used; the surface site density was 2.7 · 109 mol/
cm2, simulating polycrystalline platinum [2,3].
For gaseous chemistry, the C1/H/O schemes of
Warnatz et al. [13], further denoted as Warnatz
(81 reversible reactions, 27 irreversible reactions,
and 25 species, with appropriate pressure dependencies for 3 reactions), and GRI-3.0 [14] (131
reversible reactions, 6 irreversible reactions, and
26 species) were employed. The scheme of Warnatz [13] should not be confused with the earlier
C1/H/O scheme of Warnatz and Maas [10] (48
reversible reactions, 10 irreversible reactions, and
18 species) that has been validated over the range
1 6 p 6 6 bar previously [7,8]. Gas-phase and surface reaction rates were evaluated with CHEMKIN [15] and Surface-CHEMKIN [16],
respectively. Gaseous and surface thermodynamic
data were taken from CHEMKIN [17] and Warnatz et al. [18], respectively. Mixture-average diffusion including thermal diﬀusion [19] was
considered in the species transport.
An orthogonal staggered grid of 420 · 120
points (in x and y, respectively, over the
300 · 7 mm2 domain) was suﬃcient to produce a
grid independent solution. The inlet conditions
were uniform proﬁles for the temperature, the axial velocity, and the species mass fractions. Fitted
curves through the individual thermocouple measurements provided the bottom- and top-wall
temperature proﬁles, which were used as energy
boundary conditions at y = 0 and 7 mm, respectively. No-slip was applied for the velocity at the
walls and zero-Neumann conditions for all scalars
at the outlet.
4. Results and discussion
The experimental laminar-ﬂow conditions are
presented in Table 1. Cases 1–6 pertain to CH4/
air, whereas Cases 7 and 8 to CH4/O2 mixtures
with CO2/N2 and H2O/N2 dilution, respectively.
Comparisons between measured and predicted
OH maps are illustrated in Fig. 2 for the CH4/
air cases. The ﬂames of Fig. 2 exhibited a
slight-to-moderate asymmetry due to temperature
2522
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
Table 1
Experimental conditionsa
Case
p (bar)
u
TIN (K)
UIN (m/s)
ReIN
1
2
3
4
5
6
7
8
4
6
8
10
14
16
10
14
0.3
0.36
0.36
0.4
0.4
0.4
0.4
0.4
548
567
587
572
581
643
602
558
0.4
0.43
0.38
0.3
0.26
0.45
0.35
0.39
473
717
797
824
962
1606
1038
1477
a
Pressure, equivalence ratio, inlet temperature,
velocity, and Reynolds number. Cases 1–6 are CH4/air
mixtures. In Case 7 the inert is 30.2% CO2 and 39.9% N2
and in Case 8 the inert is 57.1% H2O and 14.0% N2 (per
volume).
diﬀerences between the two channel walls. Typical
wall-temperature proﬁles are depicted in Figs. 3A
and B. The onset of homogeneous ignition, shown
with the green arrows in Fig. 2, was deﬁned by
intersection with the wall of a line ﬁtted through
the stronger ﬂame tail (the one pertaining to the
hotter wall). This deﬁnition was consistent with
the rise in OH concentration (see the computed
streamwise proﬁles of the y-averaged OH mole
fraction in Figs. 3A and B). The lower-pressure
ﬂames (Cases 1 and 2) had the highest OH levels
and their peak values relaxed rapidly in the
post-ﬂame zones. The higher-pressure ﬂames had
lower peak OH levels which, however, were maintained farther downstream (see Fig. 2 and Figs.
3A and B).
Prior to the evaluation of the gaseous schemes,
an assessment of the catalytic processes preceding
homogeneous ignition was undertaken. Comparisons between Raman-measured and predicted
(Deutschmann/Warnatz schemes) transverse proﬁles of the CH4 and H2O mole fractions are
shown in Fig. 4; for clarity, 28 of the 63 transverse
points are presented. The ﬁrst two positions
x = 15.5 and 43.5 mm of Case 2 (Figs. 4A and
B) were far upstream of the homogeneous ignition
location (xig). Consequently, the computations in
Figs. 4A and B were totally unaﬀected by the
inclusion of gaseous chemistry (Warnatz, GRI3.0, or no gaseous scheme), and solely reﬂected
the contribution of the catalytic pathway. The
very good agreement between measurements and
predictions in Figs. 4A and B demonstrated that
the scheme of Deutschmann accurately captured
the underlying upstream heterogeneous processes.
These processes were, in turn, crucial for determining the amount of fuel available for followup homogeneous combustion. In Case 2, for
example, at the point of homogeneous ignition
73% of CH4 was already converted (Fig. 3A),
whereas in Case 6 the corresponding conversion
was 49% (Fig. 3B). The measured and predicted
proﬁles of Case 6 at x = 15.5 mm (Fig. 4D) were,
likewise, dictated exclusively by the catalytic pathway and were in good agreement with each other.
The above comparisons showed that the scheme
of Deutschmann realistically reproduced the catalytic CH4 consumption over the entire pressure
range. The near-wall CH4 concentrations are very
low in Figs. 4A, B, and D, indicating an operation
close to the mass-transport-limit; for this reason,
the species transverse proﬁles exhibited a far-less
pronounced asymmetry compared to the OH
maps of Fig. 2. In Figs. 4C, E, and F, the gaseous
pathway had a non-negligible impact as will be
discussed next. It is ﬁnally clariﬁed that in our earlier heterogeneous reactivity studies [4] (wherein
the scheme of Deutschmann was validated) the
wall temperatures were suﬃciently low as to assure a kinetically controlled CH4 conversion.
The homogeneous ignition characteristics are
discussed next. The Deutschmann/Warnatz
schemes captured the measured ﬂame shapes and
the OH levels over the pressure range
6 6 p 6 16 bar (Fig. 2(2–6)) very well and underpredicted only to a small degree (by 4% to 14%)
the measured xig. At p = 4 bar (Fig. 2(1)), however, there was an appreciable underprediction
of xig. Additional simulations (using the Deutschmann/Warnatz schemes) of earlier [7,8]
low-pressure ﬂames illustrated that the xig-underpredictions were even more pronounced at
1 6 p < 4 bar. The aforementioned studies had
established the validity of the homogeneous
scheme of Warnatz and Maas [9] over the range
1 6 p 6 6 bar; at p > 6 bar, however, this scheme
signiﬁcantly overpredicted xig. Since the present
study
focuses
on
the
pressure
range
6 6 p 6 16 bar, the diﬀerences between the Warnatz and Maas [9] and Warnatz [13] schemes will
not be elaborated; it suﬃces to say that the CH4
consumption in the former scheme followed only
the route CH4ﬁCH3ﬁCH2O whereas in the latter
scheme both CH4 ﬁ CH3 ﬁ CH3O ﬁ CH2O and
CH4 ﬁ CH3 ﬁ CH2O routes were present. Computations indicated that the downstream species
proﬁles of Figs. 4C, E, and F were aﬀected, to a
lesser-or-greater degree, by gaseous chemistry.
The position x = 93.5 mm of Fig. 4F, for example,
was far downstream of xig, and this was manifested by the absence of CH4 in extended zones
near both walls.
GRI-3.0 signiﬁcantly underpredicted the measured homogeneous ignition distances (by 56%
in Case 2c and 74% in Case 6c of Fig. 2); such
underpredictions were also typical of any intermediate pressure. To understand the origin of the differences between the two gaseous schemes, a
systematic study of ignition characteristics was
undertaken. The SENKIN code [20] was used to
compute ignition delay times of fuel-lean CH4/
air mixtures, at a ﬁxed pressure and temperature.
The ﬁxed temperature mimicked the presence of
the heterogeneous pathway that supplied heat to
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
2523
Fig. 2. LIF-measured (1a–6a) and numerically predicted OH color-maps. The predictions 1b–6b refer to the
Deutschmann/Warnatz schemes, while 2c and 6c refer to the Deutschmann/GRI-3.0 schemes. The indicated OH levels
(ppm vol.) pertain to both the measurements and the Deutschmann/Warnatz predictions. The arrows deﬁne the onset of
homogeneous ignition.
Fig. 7. LIF-measured (a) and numerically predicted with the Deutschmann/Warnatz schemes (b) OH concentrations
(ppm vol.) for Cases 7 and 8.
the gas. Predicted ignition delay times sig (deﬁned
as the times of 50% CH4 conversion) versus u are
provided in Fig. 5. It is known [21] that CH4 selfinhibits its ignition: sig [CH4]a, the exponent a
being a positive number (a 0.33). This inhibition, however, has been established at u P 0.45
and T P 1300 K [21], which range outside the
CST domain. At 4 bar and T > 900 K, GRI-3.0
always displayed a self-inhibition (as manifested
by the positive slope in the plots of Fig. 5A) down
to u = 0.05, thus resulting in rapid acceleration of
the gaseous reactivity at ultra-lean mixtures. Only
at suﬃciently low temperatures (T = 900 K) and
very low equivalence ratios (u < 0.20), this trend
was reversed. On the other hand, the scheme of
Warnatz (Fig. 5A) displayed self-inhibition only
at higher temperatures (T P 1300 K) and at u
greater than a minimum value that decreased with
increasing temperature. At T = 1400 K, the two
schemes were in good agreement, except for
2524
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
Fig. 5. Predicted ignition delay times versus equivalence
ratio: Warnatz (solid lines) and GRI-3.0 (dashed lines).
Fig. 3. Upper-wall (solid gray lines) and lower-wall
(dashed gray lines) streamwise temperature proﬁles ﬁtted
through the thermocouple measurements (upper: circles;
lower: triangles). Computed (Deutschmann/Warnatz
schemes) species streamwise proﬁles (averaged over y):
CH4 (solid line), OH (dotted-dashed), and H2O (dotted).
The arrows deﬁne the onset of homogeneous ignition.
Fig. 6. Ratios of predicted ignition delay times versus
temperature: Warnatz to GRI-3.0.
Fig. 4. Measured (symbols) and predicted (lines) species
transverse proﬁles: CH4 (solid lines, squares), H2O
(dashed-lines, triangles). The axial positions are: (A,D)
x = 15.5 mm,
(B,E)
x = 43.5 mm,
and
(C,F)
x = 93.5 mm.
u 6 0.2. At p = 16 bar (Fig. 5B), both schemes
exhibited the same trends as at p = 4 bar, with
the diﬀerence being that the shift towards self-inhibition occurred at higher temperatures. The ratio of the Warnatz-to-GRI-3.0 ignition delay
times is illustrated in Fig. 6. At suﬃciently high
temperatures (1400 K 6 T 6 1550 K at p = 4 bar
and 1500 K 6 T 6 1550 K at p = 16 bar), the ratio
was lower than 1.25 for all u, indicating a good
agreement between the two schemes. Over the
CST operational window 1000 K 6 T 6 1300 K
and 0.05 6 u 6 0.4, however, the two schemes
had substantial diﬀerences (ratios up to 4.5 and
3.7 for u = 0.05). It is noted that u as low as
0.05 are relevant in CST due to the signiﬁcant catalytic fuel conversion preceding homogeneous
ignition (see Fig. 3). The performance of the
scheme of Warnatz was, therefore, attributed to
its capacity to correctly capture the p–T–u range
of self-inhibited CH4 ignition. Computations with
the validated scheme of Warnatz indicated that
homogeneous ignition was possible in commercial
honeycomb reactors: for example, a ﬂame was
established in a Pt-coated catalytic channel with
a length of 300 mm, a diameter of 1.2 mm, a wall
temperature of 1400 K, p = 16 bar, UIN = 15 m/s,
TIN = 750 K, and u = 0.40.
Heterogeneously produced major species such
as H2O and CO2 aﬀected the ignition characteristics. The levels of H2O in the CH4/air experiments
were relatively low, for example, 4.8% and 2.7% at
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
xig for Cases 2 and 6, respectively (Fig. 3); the CO2
levels were lower (about 55% of the corresponding
H2O levels). To investigate the eﬀect of catalytic
product formation, additional experiments with
large H2O and CO2 dilution were performed.
Comparisons between measured and predicted
(Deutschmann/Warnatz schemes) OH maps are
illustrated in Fig. 7 for CH4/O2 mixtures with
CO2/N2 dilution and H2O/N2 dilution. The H2O
addition was 57.1% per volume in Case 8; at xig,
the H2O levels had risen to 66% (Fig. 3C). The
chemical eﬀect of H2O was numerically investigated by replacing the incoming water in Case 8
with a ﬁctitious species H2O*, which did not participate in any reaction but had the same thermodynamic and transport properties as H2O (including
the same third body eﬃciency). The computed xig
in Case 8 shifted 50 mm (20 mm when the third
body eﬃciency of H2O* was replaced by that of
N2) downstream, clearly showing that H2O promoted kinetically to a signiﬁcant degree the gaseous
combustion; this result was consistent with earlier
atmospheric-pressure studies [2]. The thermal eﬀect
of H2O addition was even more important than the
chemical one; replacing the incoming H2O with N2
shifted the ignition location 70 mm upstream due to
the substantially lower heat capacity of the N2; the
net eﬀect of H2O addition was an inhibition of
homogeneous ignition. The very good agreement
between the measured and predicted homogeneous
ignition distances (Figs. 7(8a,8b)) has shown that
the scheme of Warnatz correctly captured the
chemical promotion of homogeneous ignition due
to H2O addition. In Case7 with CO2 dilution
(30.2% per vol.) there was, again, a good agreement
between measurements and predictions (Figs.
7(7a,7b)). A similar analysis showed that the chemical impact of CO2 was particularly small (it shifted
xig upstream by only 4 mm). Chemical eﬀects for
CO2 have been reported only for equivalence ratios
outside the CST interest, u > 0.6 [22]. The comparisons of Fig. 7 also had an important practical impact: exhaust gas dilutions (up to 60% and 30%
per vol. H2O and CO2, respectively), are currently
investigated for gas-turbines operating with nitrogen-free CH4/O2 mixtures [10].
Finally, the hetero/homogenous radical coupling was assessed. Figure 8 provides a sensitivity
analysis (SA) of the heterogeneous pathway on
homogeneous ignition for the CH4/air cases. The
pre-exponentials of all heterogeneous reactions
were multiplied/divided by a factor of 10, and
the xig were computed anew; the ﬁve most important reactions aﬀecting homogeneous ignition are
shown in Fig. 8 (gray bars: division, black bars:
multiplication). Irrespective of pressure, the most
signiﬁcant reactions were the CH4 and O2 adsorption/desorption. The eﬀect of radical adsorption/
desorption reactions was particularly small, with
only those of OH having a discernible impact.
As long as both the OH adsorption and desorp-
2525
Fig. 8. Sensitivity analysis: ﬁve most signiﬁcant catalytic
reactions aﬀecting homogeneous ignition: (1)
CH4 + 2Pt(s) ﬁ CH3(s) + H(s), (2) O2 + 2Pt(s) ﬁ 2O(s),
(3) OH + Pt(s) ﬁ OH(s), (4) 2O(s) ﬁ O2 + 2Pt(s), and
(5) OH(s) ﬁ OH + Pt(s).
tion reactions were included in the heterogeneous
scheme, even large changes in their kinetic rates
had a minor impact on xig.
5. Conclusions
The homogeneous ignition of CH4/air, CH4/O2/
H2O/N2, and CH4/O2/CO2/N2 mixtures over Pt was
investigated experimentally and numerically at
pressures up to 16 bar, providing a ﬁrst direct validation for homogeneous reaction schemes at gasturbine-relevant CST conditions. The scheme of
Warnatz provided very good agreement to the measured homogeneous ignition distances (xig) whereas
GRI-3.0 underpredicted xig signiﬁcantly. It was
shown that crucial in the performance of the
schemes was their ability to capture the self-inhibition ignition characteristics of CH4 over the low
temperature and equivalence ratios pertinent to
CST. The addition of H2O promoted chemically
homogeneous ignition whereas the addition of
CO2 had a minor chemical impact.
Acknowledgments
Support was provided by the Swiss Federal
Oﬃce of Energy (BFE) and Alstom Power of
Switzerland.
References
[1] R. Carroni, V. Schmidt, T. Griﬃn, Catal. Today 75
(2002) 287–295.
[2] U. Dogwiler, P. Benz, J. Mantzaras, Combust.
Flame 116 (1999) 243–258.
[3] O. Deutschmann, L.I. Maier, U. Riedel, A.H.
Stroemman, R.W. Dibble, Catal. Today 59 (2000)
141–150.
[4] M. Reinke, J. Mantzaras, R. Schaeren, R. Bombach, A. Inauen, S. Schenker, Combust. Flame 136
(2004) 217–240.
[5] J. Mantzaras, P. Benz, Combust. Flame 119 (1999)
455–472.
2526
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
[6] J. Mantzaras, C. Appel, Combust. Flame 130 (2002)
336–351.
[7] M. Reinke, J. Mantzaras, R. Schaeren, R. Bombach, W. Kreutner, A. Inauen, Proc. Combust. Inst.
29 (2002) 1021–1030.
[8] U. Dogwiler, J. Mantzaras, P. Benz, B. Kaeppeli,
R. Bombach, A. Arnold, Proc. Combust. Inst. 27
(1998) 2275–2282.
[9] J. Warnatz, U. Maas, Technische Verbrennung.
Springer-Verlag, New York, 1993, p. 101.
[10] M. Wolf, C. Appel, J. Mantzaras, T. Griﬃn, R.
Carroni, in: Seventh Int. Conference for a Clean
Environment, Lisbon, Portugal, July 7–10, 2003.
[11] C. Appel, J. Mantzaras, R. Schaeren, R. Bombach,
A. Inauen, B. Kaeppeli, B. Hemmerling, A.
Stampanoni, Combust. Flame 128 (2002) 340–368.
[12] B. Steiner, B., Ph.D. thesis, University of Stuttgart
(2002).
[13] J. Warnatz, R.W. Dibble, U. Maas, Combustion,
Physical and Chemical Fundamentals, Modeling and
Simulation. Springer-Verlag, New York, 1996, p. 69.
[14] GRI-3.0, Gas Research Institute, 1999. Available
from http://www.me.berkeley.edu/gri_mech.
[15] R.J. Kee, F.M. Rupley, J.A. Miller, Chemkin II: A
Fortran Chemical Kinetics Package for the Analysis
of Gas-Phase Chemical Kinetics, Report No.
[16]
[17]
[18]
[19]
[20]
[21]
[22]
SAND89-8009B, Sandia National Laboratories,
1996.
M.E. Coltrin, R.J. Kee, F.M. Rupley, Surface
Chemkin: A Fortran Package for Analyzing Heterogeneous Chemical Kinetics at the Solid Surface–Gas
Phase Interface, Report No. SAND90-8003C, Sandia National Laboratories, 1996.
R.J. Kee, F.M. Rupley, J.A. Miller, The Chemkin
Thermodynamic Data Base, Report No. SAND878215B, Sandia National Laboratories, 1996..
J. Warnatz, M.D. Allendorf, R.J. Kee, M.E. Coltrin, Combust. Flame 96 (1994) 393–406.
R.J. Kee, G. Dixon-Lewis, J. Warnatz, M.E.
Coltrin, J.A. Miller, A Fortran Computer Code
Package for the Evaluation of Gas-Phase Multicomponent Transport Properties, Report No. SAND868246, Sandia National Laboratories, 1996.
A.E. Lutz, R.J. Kee, J.A. Miller, SENKIN: A
Fortran Program for Predicting Homogeneous Gas
Phase Chemical Kinetics with Sensitivity Analysis,
Report No. SAND87-8248, Sandia National Laboratories, 1996.
L.J. Spadaccini, M.B. Colket, Prog. Energy Combust. Sci. 20 (1994) 431–460.
F. Liu, H. Guo, G.J. Smallwood, Combust. Flame
133 (2003) 495–497.
Comments
Firooz Rasouli, Philip Morris USA. Please provide
additional information on the advantages of using Raman
versus infrared analyzers? Have you taken into account
the disproportionate reaction between carbon and carbon
dioxide?
Reply. Infrared analyzers (such as FTIR) do not
possess the spatial resolution needed for the determination of the nearwall species boundary layer proﬁles.
The precise shape of the limiting reactant proﬁle is,
in turn, cardinal in assessing the catalytic reactivity.
Furthermore, a probe has to be inserted in the narrow channel to facilitate the measurements. This
intrusion is likely to alter the ﬂow and thermo-scalar
characteristics.
The surface reaction between carbon and carbon
dioxide, C (s) + CO2 (s) ﬁ 2CO (s), has been reported
only for partial oxidation (fuelrich operation) of methane on Pt [1]. Under the very lean conditions of this
study (/ 6 0.40), this reaction plays no role.
Reference
[1] P. Aghalayam, Y.K. Park, N. Fernandes, V. Papavassiliou, A.B. Mhadeshwar, D.G. Vlachos, J. Catalysis 213 (2003) 23–38.
d
J.-Y. Chen, University of California Berkeley, USA.
Is C2 chemistry becoming important at high pressure
especially the ignition delay?
Reply. The relevant C2 chemistry is important, particularly for GRI3.0. In the scheme of Warnatz, ﬁve
reactions involving recombination of C1-radicals to C2
species were considered as an integral part of the C1
mechanism: CH3 + CH3 ﬁ C2H6 (R1), CH2 + CH3 ﬁ
C2H4 + H (R2), CH2 (s) + CH3 ﬁ C2H4 + H (R3),
CH3 + CH3 ﬁ C2H4 + H2 (R4), and CH4 + CH ﬁ
C2H4 + H (R5). The resulting scheme (108 reactions, 25
species) reproduced within 10% (over the entire range
900 K 6 T 6 1400 K, 4 bar 6 p 6 16 bar, 0.05 6 /
6 0.5) the ignition delay times (constant p and T calculations) computed with the full C2 scheme of Warnatz (164
reactions, 33 species). However, exclusion of the ﬁve reactions (R1)–(R5) resulted in ignition delay times shorter (in
comparison to the full C2 scheme) by 20% at 1100 K and
by 80% at 1400 K (p = 4 bar). The last diﬀerences reduced only slightly at p = 16 bar. The most important
reaction in the group (R1)–(R5) was the methyl radical
recombination in R1.
The scheme of GRI3.0 had a stronger sensitivity to C2
chemistry. Twelve reactions involving recombination of
C1 radicals to C2 species were now considered as an integral part of the C1 mechanism (137 reactions and 26 species). This scheme over-predicted the ignition delay times
calculated with the full C3 GRI3.0 mechanism (219 reactions 34 species) by less than 10% at T 6 1100 K and by
less than 20% at 1100 K < T 6 1400 K (p = 4 bar,
0.05 6 / 6 0.5). The over-predictions were reduced
slightly at p = 16 bar. Complete removal of the 12
C1-to-C2 reactions, however, resulted in a very large
under-prediction of the ignition delay times (compared
to the full C3 mechanism) over the range
M. Reinke et al. / Proceedings of the Combustion Institute 30 (2005) 2519–2527
900 K 6 T 6 1200 K: up to 460% for / = 0.5 and 200%
for / = 0.05 (p = 4 bar). At 16 bar, the corresponding under predictions were 350% and 160%.
2527
agreement to the measured homogeneous ignition distances while GRI-3.0 still ignites much earlier.
d
d
Jerry Lee, Sandia National Laboratories, USA.
Regarding the gaseous/surface kinetics pathways leading
to the oxidation of CH4, are the presented results basically controlled by transport? If so, how does this picture
change the range of ﬂow rates (i.e., ReynoldÕs number
relevant to this combustor)?
If transport eﬀects control the oxidation of methane
then the presented results would be speciﬁc to this particular combustion system. Scaling it up to burners used
in ‘‘real’’ applications or burners of diﬀerent geometry
may require separate experiments/simulations.
Reply. Transport is always important in catalytic
reactors, simply because the gas phase/surface interfacial
boundary conditions express a diﬀusion reaction balance. The extent of the heterogeneous and homogeneous
pathway contributions will always bear the inﬂuence of
transport in any realistic combustion system. To remove
transport limitations, the delineation of the regimes of
signiﬁcance for the homogeneous reaction pathway
was computed in an ideal Surface Perfectly Stirred Reactor (SPSR). As far as ‘‘real burners’’ is concerned, this
issue is addressed in the results section: additional
computations are presented for reactor geometries and
spatial velocities typical to those of gas turbines.
d
Serguei Nester, GTI, USA. Could you provide more
details on elementary chemical mechanisms of ignition
for H2O versus CO2?
Reply. As stated in the paper, the dilution with H2O
or CO2 does not alter the performance of the tested
mechanisms: the scheme of Warnatz provides very good
Hai Wang, University of Southern California, USA.
You showed quite large discrepancies between predictions by the GRI-Mech (3.0) and experiments. In addition, the Warnatz model appears to predict closely the
experimental data. Compared with the Warnatz model,
what reaction(s) or reaction set(s) in the GRI-Mech
are responsible for the observed discrepancies?
Reply. In the mechanism of Warnatz, the main route
for CH3 consumption is the formation of CH2O via
CH3 + O2 = CH2O + OH (R1); the route to CH3O (primarily via CH3 + HO2 = CH3O + OH (R2) and to a smaller degree via CH3 + O2 = CH3O + O) is less important.
On the other hand, GRI3.0 exhibits a diﬀerent trend: the
main route is CH3O formation (again via R2) and the
route to CH2O (R1) is much less important. Interestingly,
the older version GRI2.11 provided the same important
pathways as GRI-3.0, although with a strong increase in
the relative importance of (R1). The result was improved
ignition delay time predictions for GRI2.11, which were
between those of Warnatz and GRI-3.0. It is noted that
the reaction rate coeﬃcient in R2 was k = AT b exp(E
RT) with b = E = 0 and A (cm mol s) 1.8 · 1013 in Warnatz, 2.0 · 1013 in GRI2.11 and 3.78 · 1013 in GRI3.0.
An additional factor contributing to the discrepancies between Warnatz and GRI3.0 was the H/O subset:
interchanging the H/O reactions of GRI3.0 to those of
Warnatz improved the homogeneous ignition distance
predictions (or ignition delay times) by 20%. The controlling reaction was the exothermic radical recombination HO2 + OH = H2O + O2: the rate coeﬃcient of this
reaction was about two times larger in the scheme of
Warnatz compared to GRI3.0. This resulted in higher
HO2 levels for GRI3.0 that, in turn, accelerated R2. Detailed analysis for the origin of the discrepancies is currently in progress.