chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

journal homepage: www.elsevier.com/locate/cherd

Mixing of secondary gas injection in a bubbling

fluidized bed

Tingwen Li, Konstantin Pougatch, Martha Salcudean ∗ , Dana Grecov

Department of Mechanical Engineering, University of British Columbia, 6250 Applied Science Lane, Vancouver, BC, Canada V6T 1Z4

a b s t r a c t

In this work, three-dimensional numerical simulations with the aim of investigating the mixing of secondary gas in a
bubbling fluidized bed are performed. Single and multiple horizontal gas jet injections into a small scale rectangular
bubbling fluidized bed are studied. A tracer gas is introduced through the jet orifice to study the gas mixing in the sys-
tem. Both transient and time-averaged results are analyzed. The effect of gas injection velocity and jet arrangement
on the gas–solid contact is evaluated. It is found that the tracer distribution is non-uniform at the injection level.
Within a short distance above the injection, the tracer becomes uniformly distributed. Gas back-mixing is observed
in all simulations, which is prominent near the wall due to the downward flow of solids. For the cases studied, the gas
back-mixing tends to decrease as the secondary gas flow rate increases. For the same secondary gas flow rate, it has
been demonstrated that a better mixing between the tracer gas and solid particles is achieved when the secondary
gas is injected through distributed jet arrangement.
© 2009 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Gas mixing; Secondary gas; Bubbling fluidized bed

1. Introduction tributing gaseous feed along a bench-scale bubbling fluidized

Gas–solid fluidized beds are among the most important reac-
tor systems in the chemical industry because of their excellent
gas–solid contact and favorable heat and mass transfer char-
acteristics. In many industrial processes, the gas or liquid
is often added into fluidized bed reactors through a series
of horizontal or inclined nozzles. The feed injection is used
for different purposes, for example, to enhance the mixing,
stimulate solids flow, and control NOx emission and bed tem-
perature (Rajan and Christoff, 1982; Varol and Atimtay, 2007).
The mixing resulting from the secondary feed injection can
have a considerable effect on the conversion and selectivity of
the processes involving consecutive reactions.
Gas mixing in fluidized beds, as in any other type of reac-
tors, is a major concern in design (Van Deemter, 1985). For
this reason, gas and solids mixing has been extensively stud-
ied with a variety of techniques (Kunii et al., 1991; Bi et al.,
2000). However, only limited work on the mixing of secondary
feed injection in bubbling fluidized beds is reported in the
literature. Al-Sherehy et al. (2004) investigated the effect of dis-
bed with a diameter of 152 mm. In their experiment, all sec-
ondary gas was injected through a single nozzle mounted on
the side wall. Both steady- and unsteady-state experiments
were conducted. It was reported that the secondary gas con-
centration profile at the injection level was almost uniform
for high secondary and low primary gas velocities and the
secondary gas readily mixed within a short distance above
the injection level. In addition, back-mixing of the secondary
gas was detected and it was primarily influenced by the pri-
mary gas through the bottom distributor. Song et al. (2005)
investigated gas mixing in the reactor section of a fluid coker
cold model by using helium as tracer gas. In this experiment,
the secondary gas was injected into the bed through six feed
rings at different levels, each of which contained 10–18 noz-
zles to achieve relatively uniform injection around the outer
perimeter. The mean residence time of tracer gas from the
feed nozzles indicated intensive gas mixing inside the system.
Christensen et al. (2008a,b,c) conducted both experiments
and numerical simulation to study the distributed secondary
gas injection via a fractal injector submerged in fluidized

∗
Corresponding author. Tel.: +1 604 822 2732; fax: +1 604 822 2005.
E-mail address: msal@interchange.ubc.ca (M. Salcudean).
Received 3 February 2009; Received in revised form 15 April 2009; Accepted 29 April 2009
0263-8762/$ – see front matter © 2009 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.cherd.2009.04.012
1452 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465
Nomenclature
C concentration (vol.%)
d particle mean diameter (␮m)
D diameter (m)
e particle–particle restitution coefficient
E exit age distribution (s−1 )
F cumulative distribution function
g gravitational acceleration (m s−2 )
H bed height (m)
L length (m)
P pressure (Pa)
Q volumetric gas flow rate (m3 s−1 )
t time (s)
U, V velocity (m s−1 )
y mass fraction
X, Y, Z coordinates (m)
Greek letters
ˇ fitting parameter
ε volume fraction
ϕ fitting parameter
 viscosity (Pa s)
 granular temperature (m2 s−2 )
 density (kg m−3 )
 standard deviation of residence time (s)
 residence time (s)
Indices
0 initial state
bubble bubble
g gas phase
j jet
s solids or particulate phase
tracer tracer gas
beds. Their results indicated improved gas–solid contact in
the fluidized bed due to secondary gas injection. However,
the fractal injector used in their experiment is different from
secondary feeding systems investigated before (Koksal and
Hamdullahpur, 2004; Song et al., 2004).
Due to limitations of the available experimental techniques
and inherent complexity of gas–solid fluidization systems,
the mechanisms of mixing have only been studied to a lim-
ited degree. With significant improvements in computational
power and numerical algorithms, numerical modeling has
become an attractive tool for studying fluidization problems.
Despite the popularity of CFD modeling in fluidization, only
a few numerical studies of horizontal jet penetration into flu-
idized beds were reported. In our previous work, we carried
out a three-dimensional numerical simulation of a single hor-
izontal gas jet into a laboratory scale cylindrical fluidized bed
(Li et al., 2008). The jet penetration lengths and jet expansion
angles were predicted by numerical simulations and a good
agreement with the experimental findings reported in the lit-
erature was attained. Recently, we extended our numerical
simulation to the hydrodynamics of multiple horizontal gas
jets in a rectangular bubbling fluidized bed (Li et al., 2009).
The jet penetrations as well as the interactions between the
jet and surrounding gas, solids, bubbles, and other jets were
investigated. The effect of the secondary gas injection on the
flow hydrodynamics in the bed was examined for multiple jets
with different jetting velocities and arrangements. Our numer-
ical predictions agreed well with empirical correlations and
experimental observations in the literature. Ultimately, a bet-
ter understanding of the jet behavior is achieved through these
numerical simulations. The objective of the current study is
to further extend our previous work by introducing a tracer
gas to the secondary gas injection to investigate the mixing
of secondary gas with bed materials. Both transient and time-
averaged numerical results are analyzed to understand the
mixing behavior. The effects of gas injection velocity and jet
arrangement on the gas and solids mixing are evaluated.
2. Numerical models
In the current study, the Multiphase Flow with Interphase
eXchanges (MFIX) CFD code, available from the U.S. Depart-
ment of Energy’s National Energy Technology Laboratory
(NETL) at http://www.mfix.org, is used to solve the partial
differential equations system of the Eulerian–Eulerian model
(Syamlal et al., 1993). MFIX is a general-purpose computer code
for modeling the hydrodynamics, heat transfer, and chemical
reactions in fluid–solids systems, which has been successfully
used for describing bubbling and circulating fluidized beds and
spouted beds (McKeen and Pugsley, 2003; Syamlal and O’Brien,
2003; Benyahia et al., 2005; Das Sharma et al., 2006). In MFIX,
the governing equations for the solids phase are closed by the
kinetic granular theory (Gidaspow, 1994). In the kinetic granu-
lar theory, it is assumed that the random motion of particles is
analogous to the motion of molecules in a gas. Therefore, a so-
called granular temperature, proportional to the mean square
of the random particle velocity based on the Maxwellian veloc-
ity distribution, is defined to model the fluctuating energy of
the solid phase. Constitutive relations for the solids phase
stress tensor can be derived based on the kinetic theory (Lun et
al., 1984; Gidaspow, 1994). For concision, the details on the gov-
erning equations and closure correlations are not given here.
A brief summary of governing equations and constitutive cor-
relations is given in Appendix A. Reader is referred to Syamlal
et al. (1993) and Benyahia et al. (2007) for full details on models
in MFIX.
For dense gas–solid flows such as bubbling fluidized beds in
the current study, the turbulence of the carrier phase is not of
primary concern as particle–particle collisions dominate the
flow (Crowe et al., 1996). Furthermore, it has been reported
that the inertia of particles damps out the turbulence in the
carrier phase, and the turbulence of the carrier phase can
be ignored (Enwald et al., 1996; Portela and Oliemans, 2006).
However, in dilute flows, the turbulent fluid–particle interac-
tion becomes important and complex. A significant amount
of work has been done to modify the turbulence of the gas
phase by taking interfacial turbulent momentum transfer into
account (Reeks, 1991; Cao and Ahmadi, 1995; Simonin, 1996;
de Bertodano, 1998; Xu and Subramaniam, 2006). So far, a gen-
erally applicable turbulence model for two-phase flow is not
available, and the multiphase turbulence is still a challenging
problem in CFD. For simplicity, the sub-grid scale (SGS) model
proposed by Deardorf (1971) is employed to model the gas
phase turbulence. This model has been applied extensively
in numerical simulations of fluidized beds (Samuelsberg and
Hjertager, 1996; Lu and Gidaspow, 2003; Hansen et al., 2004;
Liu et al., 2004). To study the gas mixing in the system, a
species transport equation for the tracer gas is solved, with
an effective diffusivity calculated from turbulent viscosity.
 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465 1453

terize the mixing performance. Unless noted otherwise, these

Fig. 1 – Schematic diagram of the column: (a) bed geometry;
(b) arrangement of jets.
Another important closure for the Eulerian–Eulerian model
is the drag force between the gas and solid phase, which char-
acterizes the momentum exchange between two phases. To
date, several drag models have been developed to predict the
inter-phase drag coefficient, and there is no substantial devia-
tion between them for monodisperse systems (Li et al., 2008).
In our simulation, the well-known Gidaspow drag model is
employed (Gidaspow, 1994). More detail about the numeri-
cal models employed in the simulation can be found in Li
(2009).
parameters are fixed throughout this study.
A non-uniform structured grid with 139,000 grid points
is employed in the simulation. It is relatively coarse in the
freeboard region and fine in the bed region, with some extra
refinement around the injection area. The grid is believed to
be sufficient after a grid independence study (Li, 2009). The
lateral wall of the bed is modeled using a non-slip boundary
condition for the gas phase and a partial-slip boundary con-
dition for the solid phase (Johnson and Jackson, 1987). At the
top boundary, constant pressure is assumed and particles are
free to leave the system. For the bottom distributor and the
nozzle inlet, uniform inlet gas velocities are specified with no
particles entering the domain.
Each simulation is performed in four steps. Initially, the bed
is partially filled with stationary particles to a depth of 0.4 m
with a volume fraction of 0.5. Then, the bed is fluidized by the
primary gas through the bottom distributor at a superficial
gas velocity of 0.052 m s−1 . After a fully developed bubbling
fluidization is achieved, secondary gas is introduced via the
orifices on the wall. To eliminate the startup effect, simulation
results are saved for the analyses of flow hydrodynamics after
5 s injection. At the last step, the tracer gas is injected through
all the orifices to study the mixing behavior of secondary feed
in the reactor. The results and discussion presented in the fol-

3. Simulation setup

A small-scale bubbling fluidized bed is simulated. The cross-

section of the column is a square of 0.1 m × 0.1 m, and the
total height is 0.65 m, as shown schematically in Fig. 1(a). Four
orifices, one on each side wall, are located 0.17 m above the dis-
tributor for secondary gas injection. The square orifice with an
equivalent diameter of 1.5 mm is used to simulate a round tube
nozzle, since the nozzle orifice shape has only a minimal effect
on the jet behavior (Chyang et al., 1997). The jets arrangement
and coordinate system used in our numerical simulations are
shown in Fig. 1(b). For single jet injection, secondary gas is
introduced only through orifice A. For multiple-jet injections,
combinations of jets AB, AC, ABC, and ABCD are considered in
our study.
The material properties and operating conditions used in
our simulations are summarized in Table 1. The initial bed
height is 0.4 m with a solid volume fraction of 0.5. Air at ambi-
ent condition is used for both primary and secondary gas
feeds. Tracer with the same properties as the fluidizing gas
is introduced through the secondary gas injection to charac-

Table 1 – Bed properties and operating conditions.

Property Value

Particle average diameter (ds ) 150 ␮m

Particle density (s ) 1400 kg m−3
Superficial gas velocity from distributor (Up ) 0.052 m s−1
Initial solid volume fraction (ε0 ) 0.5
Static bed height (H0 ) 0.4 m Fig. 2 – Snapshot of voidage (a) and tracer concentration (b)
Nozzle diameter (Dj ) 1.5 mm
in XZ plane for two opposing jet injections with
Restitution coefficient (e) 0.95
Uj = 50 m s−1 (a: Z = 0.15 m; b: Z = 0.20 m; c: Z = 0.30 m; d:
Gas viscosity (g ) 1.8 × 10−5 Pa s
Z = 0.40 m; e: Z = 0.45 m).
1454 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465

lowing paragraphs are mainly based on the last simulation

step as mentioned above.

4. Results and discussion

In this section, the numerical results concerning the mixing of

secondary gas inside the bubbling fluidized bed are analyzed.
The reader is referred to our previous work for the full detail
on flow hydrodynamics inside the system resulting from sec-
ondary gas injection (Li et al., 2009). In our analyses, the bed
is divided into two zones with different superficial gas veloc-
ities due to the secondary gas injection, namely the primary
zone below the injection level with a superficial gas velocity Fig. 4 – The fitted F(t) curve and original scaled
concentration signal for two-jet injection Uj = 50 m s−1 .

Fig. 5 – The E(t) curve for two-jet injection with Uj = 50 m s−1 .

denoted as Up and a secondary zone above the injection with

a total superficial gas velocity denoted as Ug .

4.1. Tracer measurement

Gas mixing in the bed is studied by introducing tracer gas

through jet orifices. Take two opposing jet injections with
Uj = 50 m s−1 as an example, Fig. 2 shows the voidage con-
tour as well as the concentration of tracer gas in the XZ
plane. According to this figure, the tracer gas is released into
the flow mainly in form of bubbles from the jets. The gas is
well distributed in a short distance above the injection level,
indicating an intensive mixing in the bed. Also, due to the

Fig. 6 – Mean residence time and standard deviation of RTD

Fig. 3 – Tracer concentration at different points for two-jet of the tracer gas for single jet injection at different
injection with Uj = 50 m s−1 . velocities.
 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465 1455

Fig. 7 – Residence time and standard deviation of RTD of Fig. 10 – Gas velocities at Z = 0.3 m in XZ and YZ planes for
the tracer gas for four-jet injection at different velocities. the same Up and Ug with different jet arrangements.

downflow of solids near the wall, back-mixing is observed in

the primary zone. Similar phenomena can be observed for the
other cases.
The tracer gas concentrations at different points, as shown
in Fig. 2, are reported to characterize the transient mixing
behavior. The tracer gas concentrations at different measure-
ment points are plotted against time in Fig. 3. Upstream of
the injection level (a: Z = 0.15 m), the tracer concentration is
detected to be around 0.05 though it is fluctuating. Imme-
diately downstream (b: Z = 0.20 m), the tracer concentration
shows strong peaks. These peaks correspond to tracer-
enriched bubbles detaching from the jets when jets penetrate
deep into the bed. With increasing distance downstream from
the injection (c and d: Z = 0.30 and 0.40 m), the fluctuations
gradually decay due to intense mixing. In the freeboard region

Fig. 8 – Gas velocities at Z = 0.3 m in XZ plane for four-jet

injections at different jet velocities.

Fig. 11 – Contours of the time-averaged tracer concentration

Fig. 9 – Residence time and standard deviation of RTD of the in the XZ plane for single jet injection at different velocities
tracer gas for different jet velocities and jets arrangements. (Uj = 25, 50, 75, and 100 m s−1 from left to right).
1456 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465

Fig. 12 – Scaled mean tracer concentration profiles at different heights in the XZ and YZ planes for single jet injection at
different velocities.

(e: Z = 0.45 m), the fluctuations become negligible, which tend for F(t) curve is as follows:
to diminish with time, indicating that good radial and axial
mixing has been achieved. Assuming a uniform tracer concen- 1
F(t) = (1 − erf(y)) (3)
tration at the bed surface, the curve corresponding to point e 2
(in the freeboard) can be used to determine the residence time
distribution (RTD) in the bed. The smoothed F(t) curve is differentiated to calculate the
exit age distribution, E(t). Then the mean residence time and
standard deviation can be calculated using the following
4.2. Residence time
equations:

The tracer concentration measured at the bed surface, as  ∞

shown in Fig. 3, is subject to fluctuations and needs to be = tE(t) dt (4)
smoothed to obtain the cumulative distribution function, F(t). 0

For this purpose, the response curve of the tracer concen-  ∞

tration is fitted using the error function (Al-Sherehy et al., 2 =
2
(t − ) E(t) dt (5)
2004): 0

 y To obtain the mean residence time of tracer gas in the dense

2 2
erf(y) = √ e−t dt (1) region of the bed, the average tracer concentration at XY
0
cross-section of Z = 0.45 m is measured at different times to
fit F(t). For example, the fitted F(t) curve and E(t) curve for
where
two-jet injection with Uj = 50 m s−1 are shown in Figs. 4 and 5,
ˇ − ϕt respectively.
y= √ (2)
t Fig. 6 shows the mean residence time, , and standard
deviation of RTD, , of the tracer gas as a function of jet
In these equations, ˇ and ϕ are the fitting parameters, y is the velocity for single jet injections. The superficial gas velocity
error function parameter and t is time. The fitting equation in the secondary zone increases with increasing jet velocity.
 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465
As expected, the mean residence time of tracer gas decreases
as the jet velocity increases.
A faster decrease in mean residence time with increasing
jet velocity is reported for the four-jet injection into the sys-
tem, as shown in Fig. 7. In addition, it can be noted that the
standard deviation of RTD tends to decrease as the jet velocity
increases. To further investigate this trend, the RTD dimen-
sionless standard deviation / is also plotted in Fig. 7. /
or the dimensionless variance  2 / 2 can be considered as a
measure of the axial mixing (Missen et al., 1999). The two
important values of / are 0 and 1.0, which correspond to plug
flow and perfectly mixed flow, respectively. In Fig. 7, a slight
increase in / towards 1.0 can be observed as the jet velocity
is increased from 25 to 100 m s−1 , which reveals that the flow
is far from plug flow and tends to deviate further from it with
increasing secondary gas injection. This is attributed to the
increasing non-uniformity in gas velocity profiles as the sec-
ondary gas injection increases. As the velocity profile becomes
blunter, the flow is closer to plug flow. The gas velocity profiles
above the injection are shown in Fig. 8 for different jet veloc-
ities. As a consequence of a symmetric jet arrangement, only
the profiles in XZ plane is shown, the profiles in YZ plane are
very similar. The slight asymmetry in the profiles is an indi-
cation that the time interval used to average the results is not
long enough. However, the simulation time is sufficient for
analysis of the effect of secondary gas injection on the bed
hydrodynamics (Li, 2009).
Fig. 13 – Scaled mean tracer concentration profiles at different heights in the XZ and YZ planes for four-jet injection at
different velocities.
1457
When the same amount of secondary gas is injected
through, single-, two-, and four-jet, different jet arrangements,
the mean residence times of tracer gas in the dense bed are
calculated and shown in Fig. 9. Here, the same amount of
tracer gas is injected into the bed through one, two or four
nozzles (arrangement: A, AC, and ABCD in Fig. 1) with veloci-
ties of 100, 50, and 25 m s−1 , respectively. The mean residence
time is almost the same for all jet arrangements, about 4.2 s,
since the primary and secondary gas flow rates are identical
for these cases. However, a decrease in / is demonstrated in
the figure when secondary gas is injected through more jets
with lower velocity, i.e. through four jets with Uj = 25 m s−1 ,
instead of one jet with Uj = 100 m s−1 . To analyze the reason for
this trend, the gas velocity profiles at Z = 0.3 m in the XZ and
YZ planes for the same Up and Ug with different jet arrange-
ments are shown in Fig. 10. As can be seen, the velocity profiles
of four-jet injection are blunter than the others since the gas is
more uniformly distributed in the flow (Li et al., 2009). Hence
it behaves closer to plug flow than the others.
4.3. Time-averaged concentration
At the statistical steady state, time-averaged tracer concen-
trations at different levels downstream and upstream of the
secondary gas injection are analyzed to study the radial mix-
ing and back-mixing inside the reactor. In the current study,
30 s after the start of continuous tracer injection are simulated
1458 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465
and results for the last 10 s are used to analyze the mixing of
tracer gas at statistical steady state. Contours of time-averaged
tracer concentration in the XZ plane are shown in Fig. 11 for
single jet injection at different jetting velocities. According to
this figure, the tracer concentration decreases quickly away
from the jet. Since the tracer is introduced through the jet, the
tracer concentration contours reflect the jet penetration depth
into the bed to some extent. It is obvious that the jet penetra-
tion has a significant effect on the radial mixing of tracer gas,
with more radial mixing for the jet with deeper penetration.
Fig. 12 shows the mean tracer concentration profiles at dif-
ferent heights in the XZ and YZ planes for single jet injection at
various velocities. The tracer gas concentration is scaled with
C0 , which is defined as
Qtracer
C0 = (6)
Qg
where Qtracer is the volumetric flow rate of tracer gas and Qg is
the total gas flow rate.
For all cases shown in Fig. 12, there exists a bell-shaped
peak at Z = 0.2 m in XZ plane, which is 0.03 m above the injec-
tion. This high tracer concentration above the jet corresponds
to the tip of the curved jet or the tracer-rich bubbles gener-
ated by the jet. As the jet velocity increases, the peak migrates
toward the centre of the bed. This pattern is in accordance with
Fig. 14 – Contours of the scaled mean tracer concentration at different heights for single-, two-, and four-jet injections at
different velocities.
our previous results of the jet behaviors, and the jet penetra-
tion based on the measured concentration profiles compares
favorably with the jet penetration depths reported in our pre-
vious work (Li et al., 2009). In the far downstream region, the
tracer concentration profiles become flatter due to the radial
diffusion. It can be seen from the figure, the tracer gas is
uniformly mixed at Z = 0.40 m, which corresponds to the bed
surface region. At 0.02 m upstream of the injection (Z = 0.15 m),
the tracer gas is at a low concentration but greater than 0,
which indicates back-mixing at that level.
Fig. 13 shows the mean tracer concentration profiles at dif-
ferent heights in the XZ and YZ planes for four-jet injections
at various jet velocities. In the figure, two peaks above the
injection at Z = 0.20 m corresponding to the opposing jets are
observed for low jet velocities (25 and 50 m s−1 ). The peaks
move toward the centre of the bed and finally merge into one
with increasing jet velocity (75 and 100 m s−1 ). Similarly, flat
concentration profiles are observed for Z = 0.30 and 0.40 m,
indicating good radial mixing there. In Fig. 13, rough sym-
metry of the concentration profiles can be observed, which
corresponds to the symmetric arrangement of jets. Again,
back-mixing is observed for all the cases, as discussed in detail
in the next section.
When the same amount of secondary gas is injected
through different jet arrangements, it is not straightforward
to compare the concentration profiles at certain lines as done
 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465
Fig. 15 – Scaled mean tracer concentration profiles at different upstream levels in the XZ plane for single jet injection at
different velocities.
in the previous analyses. The contours of averaged tracer
concentration scaled with C0 are shown in Fig. 14 at dif-
ferent levels. Due to the brief time interval studied in our
current steady state analysis (compared with experimental
measurements which usually last a few minutes), the sym-
metry in bed geometry and jet arrangement is not completely
reflected in the concentration contours for four- and two-jet
configurations. Similar to the above discussion, high tracer
concentration region for each jet can be seen 0.03 m above the
jet injection (Z = 0.20 m). For Z = 0.30 m, relatively good mixing
between the tracer gas and the primary gas has been achieved.
However, the effect of jet arrangement is still visible from the
concentration contour. For regions farther downstream, the
tracer concentration becomes more uniform and the influ-
ence of jet arrangement fades away. Close to the bed surface
(Z = 0.40 m), the influence of jet arrangement disappears.
4.4. Back-mixing
In fluidized bed chemical reactors, the axial gas back-mixing
will significantly decrease conversion and selectivity; and it is
consequently undesirable in most applications. In most stud-
ies (e.g. Nguyen et al., 1977, 1981; Li and Weinstein, 1989;
Deshmukh et al., 2007), gas back-mixing is due to the dense
solids downflow, in which gas is dragged downwards as the
solid descending velocity exceeds the interstitial gas velocity.
1459
As already shown in the tracer concentration plots at
Z = 0.15 m in previous analyses, back-mixing takes place in
all the cases simulated. To further study this phenomenon,
the tracer concentration profiles at different upstream levels
Z = 0.05, 0.10, and 0.15 m, which are 0.12, 0.07, and 0.02 m below
the injection, respectively, are examined. Similar to before, the
tracer concentration is scaled with C0 .
For single jet injection, as shown in Fig. 15, the concentra-
tion of tracer gas is high right below the injection point, and it
becomes uniform and low in farther upstream region. Gener-
ally, it can be observed from Fig. 15 that back-mixing of tracer
gas seems stronger for low jet velocities (25 and 50 m s−1 ) than
for high jet velocities (75 and 100 m s−1 ). However, no obvious
trend can be obtained from these plots.
For four-jet injections shown in Fig. 16, the high tracer
concentration close to the wall indicates that back-mixing
near the wall is prominent compared to that in the central
core region. This is mainly because of the downflow of solids
near the wall, in which gas is entrained. This finding agrees
well with an experimental observation of Gilliland and Mason
(1952) that back-mixing increased when a tracer was injected
at the wall compared to when it was injected at the centre of
a column. However, the prominence of back-mixing close to
the wall is damped as the jet velocity increases and the con-
centration profile becomes more uniform. In addition, we can
observe a decrease in back-mixing for high jet velocities as the
1460 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465
Fig. 16 – Scaled mean tracer concentration profiles at different upstream levels in the XZ plane for four-jet injections at
different velocities.
tracer gas is transported deep into the bed. This is consistent
with the results of Christensen et al. (2008a) that back-mixing
decreases with the increasing secondary gas feed. Further-
more, back-mixing is stronger in the corners, where the wall
effect is most significant for this rectangular column, as shown
in the tracer concentration contour at Z = 0.15 in Fig. 14.
By comparing the tracer concentration profiles at different
upstream level for single-, two-, and four-jet injections with
same flow rate shown in Fig. 17, it can be concluded that the
tracer back-mixing is more significant for the four-jet injection
than for the other two cases. For single jet injection with high
velocity, the jet penetration depth is about 0.04 m. Most sec-
ondary gas is transported deeply into the upward flow at the
centre of column. However, the average jet penetration depth
is only 0.015 m for four-jet injection with the same flow rate.
The back-mixing is promoted as more tracer gas is entering
the downward solids flow close to the wall due to the limited
jet penetration into the bed.
4.5. Tracer/solids contact
The contact between the secondary reactant feed and catalyst
particles has a significant impact on the chemical reactions
taking place in reactors. A good contact is essential to the
overall performance of these processes in achieving high con-
version and selectivity. To characterize the contact between
the solids and secondary gas, the dependence between the
tracer gas volume fraction and solid volume fraction at dif-
ferent heights above the injection is statistically analyzed. A
plot of the tracer volume fraction versus the solid volume frac-
tion is shown in Fig. 18, for four-jet injection with Uj = 50 m s−1 .
The data points are collected from different control volumes
in Z = 0.3 m plane for a 10 s simulation with a data recording
frequency of 100 Hz. The tracer volume fraction averaged in
classes of different solid volume fractions is shown in the fig-
ure with the standard deviation as an error bar to indicate the
dispersion of our data.
Fig. 19 shows the plots of the averaged tracer volume
fraction versus solid volume fraction at different heights for
four-jet injection with various velocities. These plots can be
used to reveal the effect of tracer mixing in the gas phase
on tracer/solids mixing. Generally, the volume fraction of the
tracer decreases with increasing solid volume fraction. At
Z = 0.2 m, most tracer gas exists in the form of bubbles detach-
ing from the tip of gas jet. This leads to a high tracer volume
fraction in the relative dilute region (εs < 0.25) and a low value
in the dense region. Higher above the injection level, the pro-
files become straight and the standard deviation decreases
greatly because of the fast mixing of tracer with primary gas
feed. This indicates that the contact between the tracer and
solids is totally governed by the gas–solid mixing inside the
bed at high levels. Moreover, the pattern mentioned above
 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465
Fig. 17 – Scaled tracer concentration profiles at different upstream levels in XZ plane for single-, two-, and four-jet injections
with same total flow rate and different velocities.
is more evident for injection with high jet velocity than for
injection at low jet velocity.
Similarly, the plots of tracer volume fraction versus the
solid volume fraction at different heights for single-, two-, and
four-jet injections with identical secondary gas flow rate are
Fig. 18 – Tracer volume fraction versus solid volume
fraction at Z = 0.3 m for four-jet injection with Uj = 50 m s−1 .
1461
presented in Fig. 20. Although the primary and secondary gas
flow rates are identical for all cases, there exists obvious dif-
ference in tracer/solids contact as can be seen in the figure. It
can be inferred that the tracer gas tends to enter the emulsion
phase for distributed injections with low jet velocities. This
conclusion is consistent with the finding recently reported by
Christensen et al. (2008b,c). Christensen et al. studied the dis-
tributed secondary gas injection in a 2D bubbling fluidized
bed via a fractal injector and concluded that the secondary
gas tends to stay in the dense phase, resulting in improved
gas–solid contact.
The contact between the tracer gas and solids is fur-
ther investigated by comparing the bubble hold-up (volume
fraction of bubbles) at Z = 0.3 m for different jet arrange-
ments, as shown in Fig. 21. To aid the analysis, the tracer
volume fraction versus the solid volume fraction for dif-
ferent jet arrangements is also presented in this figure. As
already discussed, less tracer gas exists in the dilute region
and more in the dense region when it is introduced into
the system through four jets with Uj = 25 m s−1 instead of
one jet with Uj = 100 m s−1 . From the bubble hold-up plot in
Fig. 21, it is evident that less gas is in the form of bub-
bles for the distributed secondary gas injections (two- and
four-jet injections). Summing up these two phenomena, it
can be concluded that the gas and solids contact is greatly
improved for four-jet injection when compared to single jet
injection.
1462 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465

Fig. 19 – Tracer volume fraction versus solid volume fraction at different heights for four-jet injections at different velocities.

Fig. 20 – Tracer volume fraction versus solid volume fraction at different heights for single-, two- and four-jet injections.
 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465 1463

Fig. 21 – Tracer volume fraction versus solid volume fraction plot (left) and average bubble hold-up (right) at Z = 0.30 m for
single-, two- and four-jet injections.

5. Conclusion Appendix A.

Single and multiple horizontal gas jets in a small-scale rect- A.1. Governing equations
angular bubbling fluidized bed were numerically simulated.
The mixing of the secondary gas with bed materials was stud- (a) Continuity equations:
ied by introducing a tracer gas into the secondary injections.
Both transient and time-averaged results were analyzed to ∂  g) = 0
Gas phase : (εg g ) + ∇ · (εg g V
understand the mixing behavior. The effect of jet velocity ∂t
and jet arrangement on the mixing behavior was also eval-
uated. ∂
Solids phase :  s) = 0
(εs s ) + ∇ · (εs s V
In our transient studies, the mean residence time and stan- ∂t
dard deviation of RTD of the tracer gas were obtained and
analyzed. It was found that the dimensionless variance of RTD (b) Momentum equations:
increased with the jet velocity for four-jet injection. In addi-
∂  g ) + ∇ · (εg g V
 gV
 g)
tion, the time-averaged tracer gas distribution inside the bed Gas phase : (εg g V
∂t
was examined. At the injection level, the tracer distribution
was non-uniform. Within a short distance above the injection, = ∇ · ¯¯ g − εg ∇P + εg g g − Igs
the tracer became uniformly distributed. Gas back-mixing was
observed in all simulations, which was prominent near the
wall and tended to decrease as the secondary gas flow rate ∂  s ) + ∇ · (εs s V
 sV
 s)
Solids phase : (εs s V
increased. For the same secondary gas flow rate, better con- ∂t
tact between the tracer gas and solid particles was achieved = ∇ · ¯¯ s − εs ∇P + εs s g + Igs
when the secondary gas is injected through distributed jet
arrangement. Observations from our numerical simulations
showed reasonable agreement with those in the literature. ∂  g ygn ) = ∇ · (Dgn ∇ygn )
Gas species : (εg g ygn ) + ∇ · (εg g V
Findings in the current study will be very helpful in interpret- ∂t
ing experimental observations and measurements. Numerical
simulation of mixing constitutes a very useful tool in the
design and optimization of advanced, complex industrial sys- A.2. Constitutive equations
tems.
(a) Gas stress tensor:

Acknowledgements ¯¯ g = 2ge S̄¯ g

1   T 1
S̄¯ g = (∇ V  ¯
g + (∇ Vg ) ) − ∇ · Vg Ī
The financial support of the Natural Sciences and Engineer- 2 3
ing Research Council (NSERC) of Canada and of Syncrude ge = Min(max , g + t )
Canada Ltd. are gratefully acknowledged. The computational 
t = εg g (0.1)
2
D̄ ¯
¯ : D̄
resource funded by the Canada Foundation for Innova- g g

tion and the Canadian Institute for Advanced Research  = (x y z)

1/3

is also acknowledged. Finally, the authors would like to

¯ = 1 (∇ V
D̄  g )T )
 g + (∇ V
thank Pirooz Darabi for his help during preparation of this g
2
manuscript.
1464 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465

(b) Solids stress tensor: Cao, J. and Ahmadi, G., 1995, Gas–particle 2-phase turbulent-flow
in a vertical duct. International Journal of Multiphase Flow,
 s )¯Ī + 2s S̄¯ s
¯¯ s = (−Ps +
b ∇ · V 21(6): 1203–1228.
1  1 Christensen, D., Nijenhuis, J., van Ommen, J.R. and Coppens,
S̄¯ s = (∇ V  T  ¯
s + (∇ Vs ) ) − ∇ · Vs Ī
2 3 M.O., 2008, Influence of distributed secondary gas injection on
the performance of a bubbling fluidized-bed reactor.
Ps = εs s s [1 + 4g0 εs
]
2 + ˛ ∗s 8
  Industrial & Engineering Chemistry Research, 47(10):
s = 1+

g0 εs 3601–3618.
3 g0
(2 −
) 5
 8
 3
 Christensen, D., Nijenhuis, J., Van Ornmen, J.R. and Coppens,
M.O., 2008, Residence times in fluidized beds with secondary
× 1 +
(3
− 2)g0 εs +
b
5 5 gas injection. Powder Technology, 180(3):
∗ εs s s g0  321–331.
s =
εs s s g0 + (2ˇ/εs s ) Christensen, D., Vervloet, D., Nijenhuis, J., van Wachem, B.G.M.,
5  van Ommen, J.R. and Coppens, M.O., 2008, Insights in
= s ds s distributed secondary gas injection in a bubbling fluidized bed
96
256 2 via discrete particle simulations. Powder Technology, 183(3):
b = εs g0 454–466.
5
1+e Chyang, C.S., Chang, C.H. and Chang, J.H., 1997, Gas discharge

= modes at a single horizontal nozzle in a two-dimensional
2
fluidized bed. Powder Technology, 90(1): 71–77.
(c) Granular temperature:
⎡ Crowe, C.T., Troutt, T.R. and Chung, J.N., 1996, Numerical models
for two-phase turbulent flows. Annual Review of Fluid
¯ )
−(K1 εs + s )Tr(D̄
s = ⎣
s Mechanics, 28: 11–43.
2K4 εs Das Sharma, S., Pugsley, T. and Delatour, R., 2006,
Three-dimensional CFD model of the deaeration rate of FCC
 ⎤2 particles. AIChE Journal, 52(7): 2391–2400.
2
(K1 εs ) Tr2 (D̄ ¯ 2 ) + K Tr2 (D̄
¯ ) + 4K ε [2K Tr(D̄ ¯ )]
de Bertodano, M.A.L., 1998, Two fluid model for two-phase
+
s 4 s 3 s 2 s
⎦
2K4 εs turbulent jets. Nuclear Engineering and Design, 179(1):
65–74.
K1 = 2(1 − e)s g0 Deardorf, J.W., 1971, Magnitude of subgrid scale eddy coefficient.
Journal of Computational Physics, 7(1): 120–133.
4 2
K2 = √ ds s (1 + e)g0 εs − K3 Deshmukh, S.A.R.K., Annaland, M.V. and Kuipers, J.A.M., 2007,
3 3

√   Gas back-mixing studies in membrane assisted bubbling
ds s (3e + 1) 2 fluidized beds. Chemical Engineering Science, 62(15):
K3 = + (1 + e)(3e − 1)g0 εs
2 3(3 − e) 2 5 4095–4111.
 Enwald, H., Peirano, E. and Almstedt, A.E., 1996, Eulerian
8εs two-phase flow theory applied to fluidization. International
+ √ g0 (1 + e)
5 Journal of Multiphase Flow, 22: 21–66.
12(1 − e2 )s g0 Gidaspow, D., (1994). Multiphase Flow and Fluidization: Continuum
K4 = √ and Kinetic Theory Descriptions. (Academic Press, Boston), p. 467
ds
Gilliland, E.R. and Mason, E.A., 1952, Gas mixing in beds of
(d) Inter-phase momentum exchange: fluidized solids. Journal of Industrial and Engineering
Chemistry, 44: 218–224.
Igs g − V
= ˇgs (V  s)
Hansen, K.G., Solberg, T. and Hjertager, B.H., 2004, A
⎧
⎪ ε2 g s − V
εs g |V  g| three-dimensional simulation of gas/particle flow and ozone
⎪
⎨ 150 s 2 + 1.75 if εs > 0.2 decomposition in the riser of a circulating fluidized bed.
εg ds ds
ˇgs = Chemical Engineering Science, 59(22–23): 5217–5224.
⎪
⎪ s − V
εs εg g |V  g|
⎩ 3 Cd ε−2.65
g if εs ≤ 0.2
Johnson, P.C. and Jackson, R., 1987, Frictional collisional
4 ds constitutive relations for antigranulocytes—materials, with
⎧ application to plane shearing. Journal of Fluid Mechanics, 176:
⎨ 24 (1 + 0.15(Re εg )0.687 ) if Re εg < 1000
Re εg 67–93.
Cd =
⎩ 0.44 if Re εg ≥ 1000
Koksal, M. and Hamdullahpur, F., 2004, Gas mixing in circulating
fluidized beds with secondary air injection. Chemical
s − V
g |V  g |ds Engineering Research & Design, 82(A8): 979–992.
Re = Kunii, D. and Levenspiel, o.O., (1991). Fluidization Engineering.
g
(Butterworth–Heinemann, Boston), p. 491
Li, J.H. and Weinstein, H., 1989, An experimental comparison of
gas backmixing in fluidized-beds across the regime spectrum.
References Chemical Engineering Science, 44(8): 1697–1705.
Li, T., 2009. Numerical investigation of the gas/spray jet
Al-Sherehy, F., Grace, J. and Adris, A.E., 2004, Gas mixing and interaction with fluidized bed. Ph.D. Thesis. University of
modeling of secondary gas distribution in a bench-scale British Columbia, Vancouver.
fluidized bed. AIChE Journal, 50(5): 922–936. Li, T., Pougatch, K., Salcudean, M., Grecov, D., 2009, Numerical
Benyahia, S., Syamlal, M. and O’Brien, T.J., 2005, Evaluation of simulation of single and multiple gas jets in bubbling
boundary conditions used to model dilute, turbulent fluidized beds.Submitted to Chemical Engineering Science.
gas/solids flows in a pipe. Powder Technology, 156(2–3): 62–72. Li, T., Pougatch, K., Salcudean, M. and Grecov, D., 2008, Numerical
Benyahia, S., Syamlal, M. and O’Brien, T.J., (2007). Summary of simulation of horizontal jet penetration in a
MFIX Equations 2005-4. From URL three-dimensional fluidized bed. Powder Technology, 184(1):
https://mfix.netl.doe.gov/documentation/MFIXEquations2005- 89–99.
4-3.pdf. Liu, H., Liu, W., Zheng, J., Ding, J., Zhao, X. and Lu, H., 2004,
Bi, H.T., Ellis, N., Abba, I.A. and Grace, J.R., 2000, A state-of-the-art Numerical study of gas–solid flow in a precalciner using
review of gas–solid turbulent fluidization. Chemical kinetic theory of granular flow. Chemical Engineering Journal,
Engineering Science, 55(21): 4789–4825. 102(2): 151.
 chemical engineering research and design 8 7 ( 2 0 0 9 ) 1451–1465
Lu, H.L. and Gidaspow, D., 2003, Hydrodynamics of binary
fluidization in a riser: CFD simulation using two granular
temperatures. Chemical Engineering Science, 58(16):
3777–3792.
Lun, C.K.K., Savage, S.B., Jeffrey, D.J. and Chepurniy, N., 1984,
Kinetic theories for granular flow—inelastic particles in
Couette-flow and slightly inelastic particles in a general
flowfield. Journal of Fluid Mechanics, 140: 223–256 (Mar)
McKeen, T. and Pugsley, T., 2003, Simulation and experimental
validation of a freely bubbling bed of FCC catalyst. Powder
Technology, 129(1–3): 139–152.
Missen, R.W., Mims, C.A. and Saville, B.A., (1999). Introduction to
Chemical Reaction Engineering and Kinetics. (J. Wiley, New York),
p. 672
Nguyen, H.V., Potter, O.E., Dent, D.C. and Whitehead, A.B., 1981,
Gas backmixing in large fluidized-beds containing tube
assemblies. AIChE Journal, 27(3): 509–514.
Nguyen, H.V., Whitehead, A.B. and Potter, O.E., 1977, Gas
backmixing, solids movement, and bubble activities in
large-scale fluidized-beds. AIChE Journal, 23(6): 913–922.
Portela, L.M. and Oliemans, R.V.A., 2006, Possibilities and
limitations of computer simulations of industrial turbulent
dispersed multiphase flows. Flow Turbulence and
Combustion, 77(1–4): 381–403.
Rajan, S. and Christoff, J.D., 1982, Effect of horizontal air jet
penetration on the combustion of coal in a fluidized-bed.
Journal of Energy, 6(2): 125–131.
Reeks, M.W., 1991, On a kinetic-equation for the transport of
particles in turbulent flows. Physics of Fluids A-Fluid
Dynamics, 3(3): 446–456.
1465
Samuelsberg, A. and Hjertager, B.H., 1996, An experimental and
numerical study of flow patterns in a circulating fluidized bed
reactor. International Journal of Multiphase Flow, 22(3):
575–591.
Simonin, O., (1996). (pp. K1–K47). Continuum Modelling of Dispersed
Two-phase Flows. Lecture Series—van Karman Institute for Fluid
Dynamics
Song, X.Q., Bi, H.T., Lim, C.J., Grace, J.R., Chan, E., Knapper, B. and
McKnight, C.A., 2004, Hydrodynamics of the reactor section in
fluid cokers. Powder Technology, 147(1–3): 126–136.
Song, X.Q., Grace, J.R., Bi, H., Lim, C.J., Chan, E., Knapper, B. and
McKnight, C.A., 2005, Gas mixing in the reactor section of
fluid cokers. Industrial & Engineering Chemistry Research,
44(16): 6067–6074.
Syamlal, M. and O’Brien, T.J., 2003, Fluid dynamic simulation of
O-3 decomposition in a bubbling fluidized bed. AIChE Journal,
49(11): 2793–2801.
Syamlal, M., Rogers, W. and Brien, T.J.O., (1993). MFIX
Documentation: Theory Guide U.S. Department of Energy (DOE).
(Morgantown Energy Technology Center, Morgantown, West
Virginia).
Van Deemter, J.J., 1985, Mixing, in Fluidization, Davidson, J.F.,
Clift, R., & Harrison, D. (eds) (Academic Press, New York), pp.
331–355. Chapter 9
Varol, M. and Atimtay, A.T., 2007, Combustion of olive cake and
coal in a bubbling fluidized bed with secondary air injection.
Fuel, 86(10–11): 1430–1438.
Xu, Y. and Subramaniam, S., 2006, A multiscale model for dilute
turbulent gas–particle flows based on the equilibration of
energy concept. Physics of Fluids, 18(3)