Experimental
Measurements and Kinetic
Modeling of CO/H
2
/O
2
/NO
x
Conversion at High Pressure
CHRISTIAN LUND RASMUSSEN,
1
JØRN HANSEN,
1
PAUL MARSHALL,
1,2
PETER GLARBORG
1
1
Department of Chemical Engineering, Technical University of Denmark, DK-2800 Kgs. Lyngby, Denmark
2
Department of Chemistry, University of North Texas, Denton, TX 76203–5070
Received 23 June 2007; revised 10 October 2007, 10 January 2008; accepted 14 January 2008
DOI 10.1002/kin.20327
Published online in Wiley InterScience (www.interscience.wiley.com).
ABSTRACT: This paper presents results from lean CO/H
2
/O
2
/NO
x
oxidation experiments con-
ducted at 20–100 bar and 600–900 K. The experiments were carried out in a new high-pressure
laminar flow reactor designed to conduct well-defined experimental investigations of homoge-
neous gas phase chemistry at pressures and temperatures up to 100 bar and 925 K. The results
have been interpreted in terms of an updated detailed chemical kinetic model, designed to
operate also at high pressures. The model, describing H
2
/O
2
,CO/CO
2
,andNO
x
chemistry,
is developed from a critical review of data for individual elementary reactions, with supple-
mentary rate constants determined from ab initio CBS-QB3 calculations. New or updated rate
constants are proposed for important reactions, including OH + HO
2
H
2
O + O
2
,CO+ OH
[HOCO] CO
2
+ H, HOCO + OH CO + H
2
O
2
,NO
2
+ H
2
HNO
2
+ H, NO
2
+ HO
2
HONO/HNO
2
+ O
2
, and HNO
2
(+M) HONO(+M). Further validation of the model perfor-
mance is obtained through comparisons with flow reactor experiments from the literature
on the chemical systems H
2
/O
2
,H
2
/O
2
/NO
2
,andCO/H
2
O/O
2
at 780–1100 K and 1–10 bar.
Moreover, introduction of the reaction CO + H
2
O
2
→ HOCO + OH into the model yields an
improved prediction, but no final resolution, to the recently debated syngas ignition delay
problem compared to previous kinetic models.
C
2008 Wiley Periodicals, Inc. Int J Chem Kinet
40: 454–480, 2008
The work is part of the CHEC (Combustion and Harmful
Emission Control) research program.
Correspondence to: Peter Glarborg; e-mail: pgl@kt.dtu.dk.
Contract grant sponsor: Technical University of Denmark.
Contract grant sponsor: Danish Technical Research Council.
Contract grant sponsor: Otto Mønsted Fond.
Contract grant sponsor: Robert A. Welch Foundation.
Contract grant sponsor: B-1174.
Contract grant sponsor: UNT Faculty Research Fund.
The kinetic model of CO
2
/H
2
/O
2
/NO
x
conversion at high
pressure is available as supplementary material at http://www.
interscience.wiley.com/jpages/0538-8066/suppmat/.
c
2008 Wiley Periodicals, Inc.
INTRODUCTION
In contrast to empirical models, detailed chemical
kinetic models are complex mechanistic models de-
veloped from an understanding of the conversion of
reactants and formation of products as they actually
take place through a chain of elementary reaction steps.
Ideally, the nature of a detailed kinetic model may al-
low extrapolation to reaction conditions outside the
range of experimental verification, with an expected
accurate response. Throughout decades, this has en-
couraged a continuous development and refinement of
KINETIC MODELING OF CO/H
2
/O
2
/NO
x
CONVERSION 455
detailed kinetic models to unravel the complexities of
chemical reactions across a wide range of conditions
and ultimately to close in on the final solution. The task
relies on the availability of well-defined experimental
data. Data that fall in the outskirts of previously re-
ported results and, hence, extend the boundaries of
model development and verification are of particularly
high value.
Well-defined experimental results obtained at high
pressure fall into this category. Within recent years,
some of the most significant contributions have
emerged from flow reactor experiments by Dryer and
coworkers, e.g. [1–5], and Dagaut and coworkers, who
use a jet-stirred reactor, e.g. [6–8]. These systems are,
however, limited to operations within the range of mod-
erately high pressures, up to 20 and 10 bar, respec-
tively. The use of rapid compression machines [9–14]
and shock tubes [15–21] has extended this pressure
range considerably. However, flow reactor results at
very high pressures are still missing despite the rele-
vance to a number of important applications includ-
ing engines and gas turbines. High pressure is also a
driving force in the partial oxidation of natural gas to
oxygenated hydrocarbons that play an important role
as fuels as well as feedstock in a range of industrial
processes.
Because they exhibit relatively high surface to vol-
ume ratios, laboratory-scale reactors designed to in-
vestigate homogeneous gas phase chemistry typically
face the challenge of avoiding heterogeneous inter-
ference from the surface material. As a consequence,
reactor materials with minimal surface activity are re-
quired. Quartz or Pyrex glass are typically preferred
even though experiments [22,23] indicate that these
are not zero surface activity materials. The brittle na-
ture of these materials, however, imposes a problem
when they are subjected to large pressure gradients.
As part of this work, a new high-pressure laminar
flow reactor setup was established to conduct well-
defined experimental investigations of homogeneous
gas phase chemistry at high pressures from 10 to
100 bar. The reaction takes place in a simple tubu-
lar quartz reactor with a constant reactor temperature
up to 925 K. Experimental results from the conversion
of CO/H
2
/O
2
/NO
x
mixtures are presented. Based on a
thorough review, a detailed chemical kinetic model of
the CO/H
2
/O
2
/NO
x
system is presented. It includes
improved estimates of selected rate constants from
ab initio CBS-QB3 calculations. The CBS-QB3
methodology [24] employs geometries and frequen-
cies computed with density functional theory, at the
B3LYP/6-311G(d,p) level of theory for the species
considered here. The energy is then obtained by ex-
trapolation of coupled-cluster results to the complete
basis set limit. Where the ground electronic state is
split by spin–orbit coupling, such as in OH and NO,
an empirical energy correction was included. These
data are employed in transition state theory to derive
rate constants, which in the case of hydrogen trans-
fer reactions include an Eckart tunneling correction.
The kinetic model is validated across a pressure range
of 1–100 bar using data obtained from the new high-
pressure flow reactor as well as previous results from
the literature, and the most important reaction path-
ways are discussed based on the modeling predictions.
Finally, the paper discusses the model application to
predictions of synthesis gas (CO/H
2
) ignition times at
low temperatures and elevated pressure.
HIGH-PRESSURE FLOW REACTOR
The experimental setup is a new laboratory-scale high-
pressure laminar flow reactor designed to approximate
plug flow. The system enables well-defined investiga-
tions of homogeneous gas phase chemistry at pressures
from 10 to 100 bar at temperatures up to 925 K and
flow rates of 1–5 NL/min (“N” refers to “normal” con-
ditions at 273.15 K and 1 bar). The reactions take place
in a tubular reactor made of quartz (i.d. 8 mm, o.d. 10
mm, lg. 1545 mm). The reactor is enclosed in a TP347
stainless steel tube (i.d. 22 mm, o.d. 38 mm) that acts
as a pressure shell. A pressure control system con-
sisting of two thermal mass flow pressure controllers
(model 5866 from Brooks Instruments) automatically
delivers N
2
to the shell-side of the reactor to obtain a
pressure similar to that inside the reactor, thus avoiding
devastating pressure gradients across the fragile quartz
glass. The steel tube is placed horizontally in an Entech
tube oven with three individually controlled electrical
heating elements that produce an isothermal reaction
zone (±5 K) of approximately. 50 cm with steep tem-
perature gradients toward both the inlet and outlet of
the reactor tube. This is verified by the measured tem-
perature profiles shown in Fig. 1. The reactor temper-
ature is monitored by type K thermocouples (±2.2 K
or 0.75%) positioned inside two steel thermo-pockets
placed in the void between the quartz reactor and the
steel shell.
A maximum of four different reactant gases are
premixed before entering the reactor. The flow rates
are regulated by high-pressure digital mass flow con-
trollers (model 5850S from Brooks Instruments). All
gases used in the experiments are high purity gases
or mixtures with certificated concentrations (±2% un-
certainty supplied by Linde Gas AGA). The system
is pressurized from the feed gas cylinders. The re-
actor pressure is monitored before the reactor by a
International Journal of Chemical Kinetics DOI 10.1002/kin
456 RASMUSSEN ET AL.
Figure 1 Measured temperature profiles across the reaction
zone. The vertical dashed lines delimit the isothermal section
of the reactor: L
iso
= 43 cm.
differential pressure transducer (DPharp EJX from
Yokogawa) and controlled by a pneumatic pressure
valve (Flowserve K
¨
ammer) positioned after the reac-
tor. The system employs two pressure valves designed
for steady operation above and below 60 bars, respec-
tively. They are installed in parallel and can be manu-
ally selected through a three-way valve. A schematic
overview of the system is provided in Fig. 2.
The pressure valves reduce the system pressure to
atmospheric level prior to product analysis, which is
conducted by an online 6890N Agilent gas chromato-
graph (GC-TCD/FID from Agilent Technologies) and
aNO
x
chemiluminescence gas analyzer (CLD 700 EL
model from Eco Physics).
All tubing is 1/8 in. stainless steel with Swagelok fit-
tings. The entire downstream section is gently heated
Figure 2 Schematic overview of the system. Reactant gases are premixed from up to four different digital mass flow controllers
(MFC) before the reactor inlet. N
2
is supplied to the pressure shell through two thermal mass flow pressure controllers (MFPC).
The steel shell with the tubular quartz reactor inside is positioned in an electrically heated oven with three heating elements.
Reduction of the pressure to atmospheric level is obtained in the downstream section through one of two pneumatic pressure
valves. A pressure transducer (PT) provides the signal for the acting pressure control loops. The simplified control loops are
indicated with dashed lines. Manually operated purge valves are used during startup and shutdown.
to avoid condensation of potential condensible com-
ponents before product analysis. This is obtained by
covering all tubings and exposed reactor parts with
heating cables. The GC has three operational columns
(DB1, Porapak N, and Molesieve 13×). Using helium
as carrier gas, the GC is used for detection of O
2
,CO,
and CO
2
, while the current configuration does not al-
low a sufficiently accurate determination of H
2
.The
overall relative uncertainty of the GC measurements is
typically in the range ±5% depending on the applied
calibration gases. A similar accuracy is obtained for
measurements of NO and NO
2
using the NO
x
chemi-
luminescence gas analyzer.
A unique feature of the system design is the mount-
ing of the quartz reactor inside the steel shell, which
prevents the reactant gases from having any contact
with surfaces other than the quartz wall during the
entire heating, isothermal, and cooling section of the
reactor. This is facilitated by two AISI 316 stainless
steel flanges positioned at each end of the stainless
steel tube. The design is shown in Fig. 3. The reac-
tor enters the flanges through holes that are slightly
larger than the quartz tube to make space for the ther-
mal expansion of the steel during heating. The holes
turn into small compartments that enclose both ends of
the reactor before the 1/4 in. Swagelok connectors that
define the reactor inlet and outlet. Inside each com-
partment, two small AISI 316 steel plates are mounted
around the end of the reactor and tightly bolted to
the flanges. These plates carry two Viton O-rings that
press against the quartz tube and the steel surface of
the flanges, respectively, thereby sealing the reactor
interior from the pressure shell compartment without
damaging the glass during the thermal expansion of the
metal. The N
2
inlet to the pressure shell compartment is
International Journal of Chemical Kinetics DOI 10.1002/kin
KINETIC MODELING OF CO/H
2
/O
2
/NO
x
CONVERSION 457
Figure 3 Principle design of the two AISI 316 stainless steel flanges (light hatched parts) with the two small AISI 316
steel plates and the tubular quartz reactor mounted inside. Positions of bolts are indicated by white silhouettes. Gray hatched
parts illustrate the TP347 stainless steel pressure shell. Black circles are Viton O-rings, and black triangles indicate welded
connections.
incorporated in one of the flanges and so is the access to
the two steel thermo-pockets that contain the thermo-
couples in the space between the reactor and the steel
shell. The thermo-pockets are sealed from any contact
with the high-pressure area by welding. The flanges
are bolted together using a third Viton O-ring to seal
the high pressure inside. The bolts are easily removed
to gain access to the reactor tube or the Viton O-rings
in case they need to be changed.
Experimental data are obtained as mole fractions
as a function of the reactor temperature measured at
intervals of typically 25 K. Hence, each measurement
represents the steady-state concentration at a constant
temperature, pressure, and flow rate. This makes the
residence time τ throughout an experimental series
depend solely on the temperature in accordance with
τ =
V
F [m
3
/s]
=
π
4
D
2
L
iso
F
Nm
3
/s
T
273.15K
1bar
P
−1
(1)
where V is the volume of the isothermal reaction zone.
D is the inner diameter of the reactor tube, and L
iso
is the isothermal length (±3 cm) as determined below.
The volumetric flow rate F (±2%) is measured dur-
ing experiments in units of ”normal” cubic meters per
second (Nm
3
/s). Conversion to m
3
/s is based on the
application of the ideal gas law, which has been verified
at the current high-pressure conditions by calculations
of compressibility factors (Z) for representative gas
mixtures of N
2
,O
2
,H
2
,H
2
O, CO, CO
2
,CH
4
, and
CH
3
OH, using the Peng–Robinson cubic equation of
state [25] with customary mixing rules based on a ran-
dom mixing approximation [26]. Calculations at high
pressure from 10 to 100 bar and medium temperatures
from 600–900 K, which encompass the typical opera-
tional range of the system, yield maximum deviations
of Z from unity of 4%. This result validates the ideal
gas assumption.
Calculations indicate that the reactor operates in
the laminar flow regime with a steady, fully developed
laminar flow profile. Hence, 75 < Re < 564 for rep-
resentative gas mixtures within the operational range
of the system. The flow pattern has been confirmed by
a CFD calculation (Fluent 6.2, Fluent Inc.), which has
also verified that potential vertical upward components
of the velocity profile as a result of natural convection
following the horizontal positioning of the reactor are
negligible.
It is reasonable to approximate the laminar flow field
to plug flow and reduce the mathematical description
from a 2D to a 1D problem if the gas is premixed and
the radial velocity gradients are sufficiently small to
allow fluid elements to exhibit similar residence times.
The former is readily obtained in the present setup due
to a very long mixing section before the reactor in-
let. A useful measure of the radial velocity gradients
in laminar flow is the longitudinal or axial dispersion
[27,28] that characterizes the spreading, or overtak-
ing, of fluid elements as a result of different local flow
velocities and molecular diffusion. This process is rep-
resented by the dispersion coefficient D
disp
(= m
2
/s),
where large values of D
disp
indicate rapid spreading
and hence, mixed flow, whereas lower values indicate
International Journal of Chemical Kinetics DOI 10.1002/kin
458 RASMUSSEN ET AL.
slower spreading until D
disp
= 0 that corresponds to
ideal plug flow. The numeric value of D
disp
can be de-
termined either from flow experiments with a single
injection of a trace species at the inlet and subsequent
measurement of the outlet concentration or from the
correlation
D
disp
=
AB
+
u
2
D
2
192
AB
(2)
from Levenspiel [27,28] that was derived from the
early work of Taylor [29] and Aris [30]. This correla-
tion illustrates that the dispersion coefficient depends
strongly on the molecular diffusion
AB
.Itpromotes
dispersion at low linear flow rates u, while it has an
opposite effect at high flow rates, where dispersion
is instead facilitated by axial convection with radial
diffusion. Consequently, there is an optimum relation
between values of
AB
, and u and D in terms of low
axial dispersion; here expressed by the dimensionless
group D
disp
/uD. If the reactor length L is chosen as
characteristic length instead of the diameter D,the
dimensionless group can be referred to as the vessel
dispersion number that exhibits an upper critical limit
of 0.01 to delimit “small” deviations from plug flow
Figure 4 Intensity of axial dispersion (D
disp
/uD) correlated with the Peclet number, Pe (here denoted the Bodenstein number,
Bo). The bottom of the curve (D
disp
/uD = 0.14 and Pe = 13) defines the optimum conditions for low axial dispersion in steady
laminar pipe flow, and hence, the desired conditions where the plug flow assumption is most accurate. Marked line segments
denote ranges of D
disp
/uD and Peobtained for representative gas mixtures within the operational range of the system (see text).
Results are given for the specific volumetric flow rates of 1, 3, and 5 NL/min. The figure is courtesy of Octave Levenspiel [28].
Here, “D” denotes the dispersion coefficient (D
disp
), whereas “d
t
” is the diameter of the flow cross section (D in the present
work). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]
[27,28]. The ratio between mass transfer by convection
and diffusion is expressed by the Peclet number (also
known as the Bodenstein number, Bo), which is the
product of the Reynolds’ and the Schmidt numbers:
Pe = Re × Sc =
ρuD
µ
µ
ρ
AB
=
uD
AB
(3)
Figure 4 correlates D
disp
/uD with Pe (shown as Bo)
for steady laminar pipe flow, modified from Levenspiel
[28]. The bottom of the curve defines the optimum
conditions for low axial dispersion in laminar pipe
flow, and hence, the conditions where the plug flow
assumption is most accurate. Considering representa-
tive gas mixtures at temperatures from 600–900 K and
10–100 bar, calculations of D
disp
/uD and Pe have been
conducted at volumetric flow rates from 1 to 5 NL/min
and the resulting function value span depicted in Fig. 4.
Even though this approach does not account for dif-
ferences in diffusion coefficients between species and
a representative value has to be chosen, it provides a
useful measure of the plug-flow approximation. The
highlighted intervals of D
disp
/uD and Pe in Fig. 4 re-
veal that optimum conditions in terms of low axial dis-
persion are nearly obtained at the lowest possible flow
International Journal of Chemical Kinetics DOI 10.1002/kin
![Figure 14 Comparison between experimental data from Mueller et al. [2] and model predictions for the H2/O2/NO2/N2 system at 10 bar. Inlet conditions are 0.98% H2, 0.50% O2, 85 ppm NO2, bal. N2, and Tin = 780 K. Simulations (lines) are conducted as adiabatic plug flow calculations using the present kinetic model and include timeshifts to match the steepest gradient of the fuel conversion.](/figures/figure-14-comparison-between-experimental-data-from-mueller-3046oc56.webp)
![Figure 12 Comparison between experimental data from Mueller et al. [1] and model predictions for the H2/O2/N2 system at stoichiometric conditions and three different pressures. Inlet concentrations are everywhere: 1.01% H2, 0.52% O2, and bal. N2. Furthermore: (a) P = 2.58 bar, Tin = 935 K; (b) P = 3.49 bar, Tin = 933 K; (c) P = 6.1 bar, Tin = 934 K. Simulations (lines) are conducted as adiabatic plug flow calculations and include timeshifts to match the steepest gradient of the fuel conversion. Full lines denote simulations with the present kinetic model from Tables I–IV. Dashed lines represent simulations using the rate constant of HO2 +OH H2O+O2 (R14), k14 = 2.89 × 1013 exp(500/[RT ]) cm3/(mol s), from Baulch et al. [162] (advocated by Mueller et al. [1]) instead of k14 proposed in the present work, cf. Fig. 5.](/figures/figure-12-comparison-between-experimental-data-from-mueller-2n149h45.webp)
![Figure 17 Comparison of ignition delay time data for syngas combustion. Symbols denote experimental results: ×Peschke and Spadaccini [166] (flow reactor); and ◦ Petersen et al. [21] (flow reactor and shock tube, respectively); Walton et al. [11] (RCM). Lines are model predictions using the mechanisms of: -·- Sun et al. [99]; · · ·Li et al. [164]; - - Saxena and Williams [165]; — This work. All calculations were conducted for the mixture: 7.33% H2, 9.71% CO, 1.98% CO2, 17.01% O2, 63.97% N2 at 20 bar. The figure is largely adopted from Petersen et al. [21].](/figures/figure-17-comparison-of-ignition-delay-time-data-for-syngas-1hzrr9ft.webp)