* Corresponding author. Tel.: #31-40-247-3673.

E-mail address: j.g.boelhouwer@tue.nl (J. G. Boelhouwer).

Chemical Engineering Science 56 (2001) 2605}2614
The induction of pulses in trickle-bed reactors by
cycling the liquid feed
J. G. Boelhouwer*, H. W. Piepers, A. A. H. Drinkenburg
Department of Chemical Engineering, Eindhoven University of Technology, Den Dolech 2, PO Box 513, 5600 MB Eindhoven, Netherlands
Received 28 January 2000; received in revised form 15 September 2000; accepted 7 November 2000
Abstract
The operation of a trickle-bed reactor in the pulsing #ow regime is well known for its advantages in terms of an increase in mass and
heat transfer rates. However, fairly high gas and liquid #ow rates necessitate the operation in the pulsing #ow regime, resulting in
relatively short contact times between the phases. By means of the periodic operation of a trickle-bed reactor it is possible to obtain
pulsing #ow at average throughputs of liquid usually associated with trickle #ow during steady-state operation. This feed strategy to
force pulse initiation is termed liquid-induced pulsing #ow. The advantages associated with pulsing #ow may then be utilized to
improve reactor performance in terms of an increase in capacity and the elimination of hot spots, while interfacial contact times are
comparable to trickle #ow. An additional advantage of liquid-induced pulsing #ow is the possibility to tune the pulse frequency and
therefore the time constant of the pulses. During the periodic operation of a trickle bed, continuity shock waves are initiated in the
column due to the step-change in liquid #ow rate. This results in the division of the column into a region of high liquid holdup and
a region of low liquid holdup. At high enough gas #ow rates, the inception of pulses takes place in the liquid-rich region. Analysis of
the performed experiments indicates that besides gas and liquid #ow rates, an additional criterion for pulse inception is the available
length for disturbances to grow into pulses. For self-generated pulsing #ow this results in the upward movement of the position of the
point of pulse inception with increasing gas #ow rate. With liquid-induced pulsing #ow this means that higher gas #ow rates are
necessary to induce pulses as the length of the liquid-rich region decreases. For both self-generated and liquid-induced pulsing #ow
this relationship between the gas #ow rate and the available length for pulse formation is identical. 2001 Elsevier Science Ltd.
All rights reserved.
Keywords: Multi-phase reactors; Hydrodynamics; Cycled liquid feed; Continuity shock waves; Induced pulsing #ow; Pulse frequency
1. Introduction
A trickle-bed reactor is a commonly used type of
three-phase catalytic reactors in which a gas and a liquid
phase #ow cocurrently downward through a "xed bed of
catalyst particles. A trickle-bed reactor is usually oper-
ated in the trickle #ow regime, which means that the
interaction between the phases is rather poor. The overall
reaction rate is often governed by mass transfer resist-
ances. Changing the feed strategy may reduce mass trans-
fer resistances and thus enhance the performance of
a trickle-bed reactor. One of most simple feed strategies is
cycling the liquid feed. With this mode of operation,
signi"cant increases in conversion can be obtained
(Haure, Hudgins, & Silveston, 1989; Lee, Hudgins, & Sil-
veston 1995; Castellari & Haure, 1995). Performance
improvement is due to reduction of mass transfer resist-
ance, the formation of controlled hot spots and the
appearance of a gas-phase reaction over an almost dry
catalyst (Gabarain, Castellari, Cechini, Tobolski, &
Haure, 1997).
Pulsing #ow can be considered as a spontaneously
arising unsteady-state behavior of the reactor. Studies
of Wu, McCready, and Varma (1995, 1999) showed an
increase in selectivity for the hydrogenation of
phenylacetylene to styrene due to the change in #ow
regime from trickle to pulsing #ow. Boelhouwer, Piepers,
and Drinkenburg (1999) used the concept of cycling the
liquid feed to induce pulses at average liquid #ow rates
usually associated with trickle #ow. When the duration
of the high liquid #ow rate decreased, higher gas #ow
0009-2509/01/$- see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 0 ) 0 0 5 2 7 - 3
Table 1
Properties of the packed beds used in this investigation
Packing material Nominal diameter (mm) Packed height (m) Porosity Speci"c area (m\)
Glass spheres 6.0 1.20 0.36 640
Glass spheres 3.0 1.04 0.40 1200
Fig. 1. Schematic view of the experimental equipment (1: packed col-
umn; 2: liquid storage tank; 3: pump; 4: liquid rotameters primary feed
line; 5: liquid rotameters secondary feed line 6: gas rotameters; 7: air
bu!er tank; 8: compressor; 9: conductivity probes; 10: pressure taps; 11:
magnetic valve and 12: electronic timer).
rates were necessary to induce pulses in the column. This
study makes clear that not a combination of the gas and
liquid #owrate as such determines the existence of pulses,
but at least one or the other parameter must be
accounted for.
The physical mechanism responsible for pulse incep-
tion often encountered in literature is the occlusion of
packing channels by the liquid and subsequently blowing
o! the liquid slug by the gas #ow. Based on this concept
several models are proposed in literature to describe the
transition to pulsing #ow (Sicardi et al., 1979; Sicardi
& Ho!mann 1980; Blok, Varkevisser, & Drinkenburg
1983; Ng, 1986; Cheng & Yuan, 1999). These models
attempt to explain the transition on the basis of a micro-
scopic view of two-phase #ow in a individual packing
channel. It is not clear how these microscopic occlusions
of various packing channels lead to the macroscopic
non-uniform behavior of pulsing. The "rst to adapt
a macroscopic view to interpret the transition from trick-
ling to pulsing #ow were Grosser, Carbonel, and Sun-
daresan (1988). They demonstrated that a loss of stability
of the uniform state occurs and identi"ed this loss of
stability as the transition boundary. According to Krieg,
Helwick, Dillon, and McCready (1995), travelling
waves of high liquid holdup comparable to pulses are
already present in the trickle #ow regime. A sta-
bility analysis predicts the conditions of onset of these
travelling disturbances, which may or may not evolve
into pulses. These travelling wave instabilities grow
eventually into pulses only if su$cient column length is
available.
The objective of this study is essentially to improve
reactor performance in terms of an increase in capacity,
selectivity and the elimination of initial hot spots. This
may be achieved by altering the feed strategy from
a steady into a cycled liquid feed. With this mode of
operation it is possible to induce pulses at throughputs of
liquid that would only lead to trickle #ow during steady-
state operation (Boelhouwer et al., 1999). The advantages
associated with pulsing #ow may be utilized to increase
reactor performance in terms of enhanced mass and heat
transfer rates, while interfacial contact times remain com-
parable to trickle #ow. The aim of this paper is to
examine the hydrodynamic e!ects concerning the peri-
odic operation of a trickle bed and subsequently the
circumstances in which eventually pulses are generated in
the column.
2. Experimental setup and procedures
A schematic view of the experimental set-up is
presented in Fig. 1. The experiments were performed in
a Plexiglas column of 0.11 m inner diameter. The packing
material consisted of 3.0 and 6.0 mm glass spheres of
which the packing characteristics are listed in Table 1.
The packing was supported at the bottom of the column
by a stainless-steel screen. Air and water were uniformly
fed at the top of the column by means of distributors. For
the air supply, a bu!er tank was kept on 7 bar by a com-
pressor to minimize #uctuations in the gas #ow rate due
to pressure #uctuation in the column. The experiments
were carried out at room temperature and atmospheric
pressure.
For the liquid feed, two di!erent feed lines were ap-
plied. The primary feed line was used to introduce
a steady liquid feed to the column while the secondary
feed line provided an additional liquid feed for a certain
time interval. In this manner a square-wave cycled liquid
feed was achieved. The cycled liquid feed is characterized
by four parameters, schematically shown in Fig. 2.
A magnetic valve in the secondary feed line activated by
an electronic timer was used to regulate the feed times of,
respectively, the high and low liquid feed rates. High and
low liquid feed rates were controlled by calibrated
rotameters.
2606 J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614
Fig. 2. Schematic view of the characterization of the cycled liquid feed
(t
N
: high-liquid-feed-time; t
@
: low-liquid-feed time; ;
JN
: super"cial high
liquid feed velocity; ;
J@
: super"cial low liquid feed velocity).
Fig. 3. Liquid holdup in the trickle #ow regime as a function of
super"cial gas and liquid velocities (packing material: 3.0 mm spheres).
Preliminary experiments showed a slight increase in
pressure at the top of the column when the additional
liquid feed was fed to the column. This increase in pres-
sure drop somewhat lowers the super"cial gas #ow rate
at the top of the column. This variation in gas #ow rate
due to pressure #uctuations is however very small. The
reported super"cial gas velocities in this study are cal-
culated by using the pressure at the top of the column at
the moment the additional liquid feed is ended. It must be
mentioned that pressure #uctuations in non-steady state
operated trickle-bed reactors are inherent to this mode of
operation. Along the column axis, super"cial gas vel-
ocities will vary due to #uctuations in liquid holdup and
pressure drop.
A conductance technique, successfully implemented by
Tsochatzidis, Karapantsios, Kostoglou, and Karabelas
(1992) and Tsochatzidis and Karabelas (1995), was used
to provide instantaneous measurements of cross-section-
ally averaged liquid holdup. The column was provided
with "ve sets of conductance probes, separated 0.2 m
from each other, to measure liquid holdup at various
axial positions. The conductivity probes were calibrated
by salt-tracer injections and by the stop-#ow method.
Both calibration methods proved to be very reproducible
and no signi"cant di!erence existed between the calib-
ration outcome. Pressure drop was measured by pressure
transducers which could be connected to several pressure
taps separated by distances of 0.2 m. Both liquid holdup
and pressure data were simultaneously recorded and
stored in a computer. These data were collected with
a sampling rate of 100 Hz for a period of at least 5 min.
Before performing any experiments, the column was
operated in the pulsing #ow regime for at least 1 h in
order to ensure a perfectly prewetted bed. Liquid holdup
during trickle #ow was measured for six di!erent liquid
#ow rates and a wide range of gas #ow rates. The
transition to self-generated pulsing #ow was established
at several axial positions in the bed to examine the
upward movement of the point of pulse inception with
increasing gas #ow rate. This was accomplished by visual
observation and by monitoring of the signals from two
neighboring conductivity probes. Transition to pulsing
#ow was acknowledged when the lower conductivity
probe clearly showed large #uctuations in liquid holdup
while the upper probe showed an almost unvarying
liquid holdup.
The e!ect of cycling the liquid feed on the liquid
holdup distribution in the column was examined for
a broad range of "xed cycled feed characteristics. In these
experiments, the gas #ow rate was gradually increased
until eventually pulses were observed. With this proced-
ure, the e!ect of cycling the liquid feed on the hydrodyn-
amics at gas #ow rates not high enough for pulse
inception was examined and subsequently, the minimum
gas #ow rate required for pulsing to occur was deter-
mined. The pulse frequency was determined through an
analysis of the #uctuations in the liquid holdup data.
3. Steady-state hydrodynamics
As will be revealed later, accurate liquid holdup data in
the trickle #ow regime are essential to explain the phe-
nomena observed during cycled liquid feed. In Fig. 3, the
liquid holdup, de"ned as the fraction of the empty col-
umn occupied with liquid, is plotted as a function of
super"cial gas and liquid velocities for 3.0 mm glass
spheres as packing material. Liquid holdup increases
with increasing liquid #ow rate and decreases with in-
creasing gas #ow rate. Liquid holdup data regarding
6.0 mm glass spheres as packing material are somewhat
lower, since the liquid #ow is less supported by the
packing because of the smaller speci"c area of the pack-
ing. A comparison of the measured liquid holdup versus
a number of literature correlations proved to be in good
agreement.
In Fig. 4, the transition boundary from trickling to
pulsing #ow is plotted for both packing materials. The
axial position in the column at which this transition is
established is located at approximately 0.1 m from the
bottom of the column. There is no noticeable di!erence
between the transition boundaries for 3.0 and 6.0 mm
spheres.
J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614 2607
Fig. 4. Transition boundary from trickle to self-generated pulsing #ow
located at approximately 0.1 m from the bottom of the column.
Fig. 5. In#uence of super"cial gas velocity on the axial position of the
point of pulse inception (packing material: 6.0 mm spheres).
Fig. 6. Example of the measured liquid holdup during cycled liquid
feed (;
J@
"0.0035 m s\; ;
JN
"0.0102 m s\; ;
E
"0.10 m s\; t
@
"
20 s; t
N
"5 s; packing material"3.0 mm spheres).
A general observation reported in literature is the
upward movement of the point of pulse inception with
increasing gas #ow rate. In Fig. 5, the e!ect of increasing
gas #ow rate on the axial position of the point of pulse
inception is presented. In literature, this upward move-
ment of the position at which pulses are initiated with
increasing gas #ow is explained by the higher volumetric
gas #ow rate with increasing distance from the top of the
column due to the pressure drop over the bed. In Fig. 5, it
is seen that a relatively large increase in gas #ow rate is
needed to shift the point of pulse inception upwards,
much larger than pressure drop can account for. Visual
observations illustrate that even though pulses are gener-
ated at the bottom of the column, traveling disturbances
are present above the point of pulse inception. These
"ndings are in harmony with the observations of Krieg
et al. (1995) who claim that traveling disturbances are
also present in the trickle #ow regime. They consider that
the transition to pulsing #ow corresponds to only
a quantitative change in the strength of these waves.
Convective disturbances, if unstable, will grow with dis-
tance until they become observable and thus recognized
as pulses, at a position that depends on #ow conditions.
Apparently, with increasing gas #ow rate, the required
column length for disturbances to grow into detectable
pulses decreases. Hence the location of the point of pulse
inception moves upward in the column. Available col-
umn length may be identi"ed besides gas and liquid #ow
rates as a fundamental parameter for pulse generation.
4. Unsteady-state hydrodynamics
Before performing liquid-induced pulsing #ow experi-
ments, it is meaningful to investigate the e!ect of cycling
the liquid feed on the hydrodynamics at gas #ow rates
not high enough for pulse initiation. Only then the condi-
tions at which pulses are generated during cycled liquid
feed can be con"dentially examined. An example of the
liquid holdup determined at two conductivity probes
separated by a distance of 0.2 m, during cycled liquid feed
is shown in Fig. 6. The introduction of the additional
liquid feed results in an almost instantaneous increase in
liquid holdup. The back of the liquid-rich region is char-
acterized by a more gradual decrease in liquid holdup.
The high liquid holdup endures for a period equal to the
high liquid feed time, from now on abbreviated to hlf-
time.
4.1. Liquid holdup
Due to cycling the liquid feed, the liquid holdup varies
between a low and a high value. To examine whether
there is an e!ect of the hlf-time on the liquid holdup
realized as a consequence of the high liquid feed rate, the
relative high liquid holdup is de"ned as follows:

N
"

N
(t
N
(10 s)

N
(t
N
"10 s)
. (1)
Regarding this de"nition, the high liquid holdup at a hlf-
time of 10 s serves as a reference to examine the e!ect of
decreasing hlf-time on the high liquid holdup. A plot of
2608 J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614
Fig. 7. E!ect of the hlf-time on the relative high liquid holdup
(;
J@
"0.0035}0.0077 m s\; ;
JN
"0.0059}0.0128 m s\; ;
E
"0.03}
0.30 ms\; t
@
"20 s).
Fig. 8. Comparison of liquid holdup during high liquid feed rate with
liquid holdup during trickle #ow (;
J
"0.0059}0.0128 m s\; ;
E
"
0.03}0.30 m s\; t
N
"2}10 s; t
@
"20 s).
Fig. 9. Velocity of the moving liquid front (t
@
"20 s; t
N
"1}10 s; pack-
ing material"6.0 mm spheres).
the relative liquid holdup as a response to decreasing
hlf-times is provided in Fig. 7. In this "gure it is noticed
that, except for hlf-times of 1 s and less, no e!ect of the
hlf-time is recognized. The liquid feed is probably not
perfectly square-wave cycled since some delay in the
secondary feed line is inherently present due to the time
needed for the acceleration of the liquid. For very short
hlf-times, the high liquid feed is ended before it is fully
developed. Hence for hlf-times of 1 s and less, the high
liquid holdup which should be reached during cycled
liquid feed is not entirely accomplished.
As seen in the foregoing paragraph, a certain high
liquid holdup is reached as the result of the additional
liquid feed. This high liquid holdup is probably identical
to the liquid holdup during steady-state operation at
a steady liquid #ow rate, equivalent to the high liquid
feed rate. A comparison of the high liquid holdup during
cycled liquid feed with the liquid holdup during constant
liquid feed is made in Fig. 8. Due to cycling the liquid
feed, the liquid holdup varies between two values resem-
bling the liquid holdup obtained during steady-state op-
eration at comparable liquid feed rates.
4.2. Continuity shock waves
By evaluating the time lag between the two signals
from neighboring conductivity probes, it is possible to
evaluate the velocity of the moving liquid front, resulting
from the step-change in liquid feed rate. The experi-
mentally determined linear velocity of the liquid front is
plotted in Fig. 9 as a function of the super"cial gas
velocity for 6.0 mm spheres as packing material. The
velocity of this moving liquid front increases both with
increasing gas #ow rate and with increasing di!erence
between pulse and base liquid #ow rate. The velocities of
the moving liquid front, altering roughly between 0.1 and
0.2 m s\, are much higher than the linear liquid vel-
ocities (super"cial liquid velocity divided by liquid hold-
up) which vary between 0.02 and 0.08 m s\. Hence it can
be concluded that due to cycling the liquid feed some
kind of waves are initiated in the column.
To examine the e!ect of the hlf-time on the velocity of
the moving liquid front, similar to the relative high liquid
holdup, the relative wave velocity can be de"ned as
<
U
"
<
U
(t
N
(10 s)
<
U
(t
N
"10 s)
. (2)
A plot of the relative wave velocity versus the hlf-time is
shown in Fig. 10. Analogous to liquid holdup during high
liquid #ow rate, it is noticed that for hlf-times of 1 s and
less, the velocity of these waves is somewhat less than the
values corresponding to hlf-times above 1 s.
It seems useful to investigate the possibility of the
formation of continuity shock waves resulting from the
step change in liquid #ow rate. Continuity waves occur
whenever the #ow rate of a substance depends on the
amount of that substance which is present. Continuity
waves often emerge in systems where gravity and pres-
sure drop are balanced against drag forces, as is the case
with liquid #ow in packed beds. One steady-state value
simply propagates into another one and there are no
J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614 2609
Fig. 10. E!ect of the hlf-time on the relative shock wave velocity
(;
J@
"0.0035}0.0077 m s\; ;
JN
"0.0059}0.0128 m s\; ;
E
"0.03
}0.30 m s\; t
@
"20 s).
Fig. 11. Comparison of the experimentally determined and calculated
shock wave velocity (;
J@
"0.0035}0.0077 m s\; ;
JN
"
0.0059}0.0128 m s\; ;
E
"0.03}0.30 m s\, t
@
"20 s; t
N
"2}15 s).
Fig. 12. Enlargement of the pulsing #ow regime by cycling the liquid
feed (t
@
"20 s; t
N
"0.5}15 s; packing material: 3.0 mm spheres).
dynamic e!ects of inertia or momentum (Wallis, 1969).
Waves can either propagate continuous changes in some
variables or can involve a step change or "nite discon-
tinuity. The later are called shock waves. According to
Wallis (1969), the velocity of a shock wave, derived from
continuity considerations obeys the following equation:
<
U
"
;
JN
!;
J@

N
!
@
. (3)
According to Eq. (3), the shock wave velocity is directly,
related to the di!erence between liquid feed rates and the
di!erence between the resulting values of the liquid hold-
up.
To calculate the shock wave velocity, the experi-
mentally determined liquid holdup data during trickle
#ow are used. This is justi"ed by the fact that the liquid
holdup varies between the values, which are obtained
from steady-state experiments, as shown in Fig. 8. The
solid lines in Fig. 9 depict the calculated velocity of the
moving liquid front conform Eq. (3). A good agreement
exists between experimentally and calculated values, as
well qualitatively as quantitatively. A comparison of all
the experimental data with calculated values is shown in
Fig. 11. Shock wave velocities corresponding to hlf-times
of 1 s and less are not shown here for reasons presented
earlier. The overall agreement is satisfactory for the
whole range of shock wave velocities. The accuracy of the
calculated values increases with increasing di!erence be-
tween high and low liquid holdup. Since this di!erence
varies in a narrow range between 0.03 and 0.1, it is very
important to use very accurate values of the liquid hold-
up, because this has a large e!ect on the calculated shock
wave velocity.
In summary, due to the step-change in liquid feed rate,
continuity shock waves are initiated in the column. As
a result, the column can be split up into two regions of
di!erent liquid holdup, both corresponding to steady-
state conditions. In other words, at the same time two
di!erent steady-state conditions are present in the col-
umn. These steady-state conditions are separated by
a moving boundary of which the velocity can be cal-
culated using accurate liquid holdup data obtained dur-
ing steady-state operation at equivalent liquid feed rates.
5. Liquid-induced pulsing 6ow
5.1. Enlargement of the pulsing yow regime
Fig. 12 provides the results of the enlargement of the
pulsing #ow regime by cycling the liquid feed for 3.0 mm
glass spheres as packing material. Similar results are
found for the 6.0 mm glass spheres as packing material.
The solid line, denoting the transition boundary to self-
generated pulsing #ow, is taken at approximately 0.1 m
from the bottom of the column. In Fig. 12, for each series
of data, the hlf-time increases in the downward direction
2610 J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614
Fig. 13. The for induced pulsing #ow necessary super"cial gas velocity
as a function of the hlf-time (t
@
"20 s; packing material: 3.0 mm
spheres).
Fig. 14. Example of the measured liquid holdup during liquid-induced
pulsing #ow (;
J@
"0.0035 m s\; ;
JN
"0.0077 m s\; ;
E
"
0.29 m s\; t
@
"20 s; t
N
"3 s; packing material"6.0 mm spheres).
denoted by the arrow. This means that with decreasing
hlf-time, higher gas velocities are necessary to obtain
pulsing. It is possible to induce pulses at average liquid
#ow rates, generally associated with trickle #ow during
steady-state operation. Hence, it can be concluded that
although throughputs of liquid are equal, the prevailing
#ow regime is pulsing instead of trickle #ow. The advant-
ages associated with pulsing #ow may be utilized to
enhance mass and heat transfer rates resulting in im-
proved reactor performance. Moreover, with liquid-in-
duced pulsing #ow longer contact times can be achieved
compared to self-generated pulsing #ow, while average
liquid #ow rates are reduced.
The observation that higher gas #ow rates are required
to induce pulses when hlf-times decrease, is presented
more pronounced in Fig. 13. In this "gure, the necessary
gas velocity is plotted as a function of the pulse feed time
for some of the results presented in Fig. 12. At relative
long hlf-times, the required gas velocity equals the velo-
city needed for transition to self-generated pulsing #ow at
the bottom of the column. However for relatively short
hlf-times, a higher gas velocity is necessary to initiate
pulsing #ow. This gas #ow rate increases with decreasing
hlf-times, although the cycled liquid feed rates remain
constant. These results indicate that not a combination of
gas and liquid #ow rates as such determines whether
pulses are initiated or not, but some other parameter has
to be included.
5.2. The process of liquid-induced pulsing yow
Due to the step-change in liquid feed rate, continuity
shock waves are initiated in the column. This results in
the appearance and downward movement of a liquid-rich
region. In Fig. 14, an example of the liquid holdup during
liquid-induced pulsing #ow at two neighboring conduct-
ivity probes is plotted. In the signal taken at 0.5 m from
the top of the column no pulse is present, while in the
signal taken at 0.7 m from the top evidently a pulse is
visible. Hence it can be concluded that pulses are in-
itiated in the liquid-rich region.
The length of the liquid-rich region can readily be
calculated by
l
U
"<
U
t
N
. (4)
For relatively long hlf-times, the column will eventually
be "lled entirely with the liquid-rich region, as schemati-
cally shown in Fig. 15a. In this case the necessary gas
velocity for pulse formation equals the velocity needed to
obtain self-generated pulsing #ow at the bottom of the
column. For short enough hlf-times, the length of the
liquid-rich region is less than the column height. As
a result the column can now be divided into two zones of
di!erent liquid holdup. With decreasing hlf-time, the
length of the liquid-rich region decreases, as schemati-
cally shown in Fig. 15b}d. With decreasing hlf-times and
hence decreasing length of the liquid-rich region, increas-
ing gas #ow rates are necessary to obtain pulsing #ow.
A similar phenomenon was observed for self-generated
pulsing #ow where the point of pulse inception was found
to move upwards with increasing gas #ow rate. As well
for self-generated pulsing #ow as for liquid-induced
pulsing #ow, there seems to be a relationship between
necessary gas #ow rate and available length for pulse
formation.
It is also noticed in Fig. 14, pulses are initiated at the
front of the liquid-rich region. This is a general observa-
tion during liquid-induced pulsing #ow. In Fig. 15 the
point of pulse inception during liquid-induced pulsing
#ow is visualized. At relative large hlf-times, just as soon
as the column is entirely "lled with the liquid-rich region,
pulses are initiated at the bottom of the column. For
relative short hlf-times, the point of pulse inception
moves upwards in the column with decreasing length of
the liquid-rich region. Pulses are observable just as soon
as the entire liquid-rich region is present in the column.
J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614 2611
Fig. 15. Schematic representation of the in#uence of the hlf-time on the
liquid distribution in the packed bed and the point of pulse inception
during liquid-induced pulsing #ow.
Fig. 16. Example of the measured liquid holdup during liquid-induced
pulsing #ow (;
J@
"0.0035 m s\; ;
JN
"0.0128 m s\; ;
E
"
0.26 m s\; t
@
"20 s; t
N
"2 s; packing material"6.0 mm spheres).
Considering the fact that su$cient length for pulse
formation must be available, pulses are always initiated
in the lower part of the liquid-rich region, at the moment
the entire liquid-rich region is present in the column. This
is also observed visually. Once the liquid-rich region
enters the column disturbances are present in the liquid-
rich region. However these disturbances do not growinto
detectable pulses. This occurs only as the high liquid feed
is ended, when the liquid-rich region is completely pres-
ent in the column.
Because the velocity of the initiated pulses is much
higher than the shock wave velocity, pulses will eventual-
ly move out of the liquid-rich region. Occasionally the
initiated pulses fade away when moving out of the
liquid-rich region, but in most cases the pulses remain
stable. An example of the later case is shown in Fig. 16, in
which the liquid holdup during liquid-induced pulsing
#ow at the two neighboring conductivity probes is plot-
ted. It is clearly seen that the initiated pulses have moved
out of the liquid-rich region, but remain stable. However
while the length of the column is rather short, it cannot
be assured at this moment if pulses remain stable in
columns of longer length.
5.3. Necessary gas yow rate
As described previously, as well for self-generated puls-
ing #ow as for liquid-induced pulsing #ow, there is a rela-
tionship between necessary gas #ow rate and available
length for pulse formation. For self-generated pulsing
#ow this results in the upward movement of the point of
pulse inception with increasing gas #ow rate. Neverthe-
less, this point of pulse inception never reaches entirely
the top of the column. The upper part of the column does
not actively participate in the process of pulse formation.
It is likely to expect that this part of the column is needed
to reach the necessary distribution of the phases. How-
ever, to accomplish a relationship between available
length for pulse formation and gas #ow rate, it is required
to know the length of this distribution zone. With liquid-
induced pulsing #ow, the available length for pulse
formation is externally controlled by the hlf-time. This
provides the possibility to determine the length of the
distribution zone. Referring back to Fig. 13, we can
identify a certain critical hlf-time. The critical hlf-time is
de"ned as the hlf-time at which higher gas velocities
compared to those needed to obtain self-generated puls-
ing #ow at the bottom of the column, are required for
pulse induction. The length of the liquid-rich region
corresponding to the critical hlf-time can be readily cal-
culated by multiplying the critical hlf-time with the shock
wave velocity. A plot of this critical length of the liquid-
rich region is given in Fig. 17. The critical lengths are
roughly constant and 0.15 m less compared to the height
of the packed bed for both packing materials. It seems
that the upper 0.15 m of the packed column is not active-
ly participating in the process of pulse formation.
Keeping in mind, that the "rst 0.15 m do not partici-
pate in the process of pulse formation, it is possible to
establish the interdependence between gas #ow rate and
available length for pulse formation for self-generated
pulsing #ow. The data in Fig. 5, simply have to be shifted
0.15 m towards lower distances from the top of the col-
umn. Additionally, the x-coordinate from Fig. 13 can be
transferred from hlf-times to length of liquid-rich regions
by multiplying the hlf-time with the calculated shock
wave velocity. Subsequently, the relationship between the
necessary gas velocity and available length for pulse
formation for both self-generated and liquid-induced
pulsing #ow is established. A comparison of these rela-
tionships for 3.0 mm glass spheres as packing material is
presented in Fig. 18. The data of self-generated and
liquid-induced pulsing #ow coincide. Note that in
the legend of Fig. 18, only high liquid #ow rates are
mentioned because pulses are initiated in the liquid-rich
region. Low liquid #ow rates solely participate in the
resulting shock wave velocity and thus in the length of
the liquid-rich region.
2612 J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614
Fig. 17. Critical length of the liquid-rich region.
Fig. 18. Comparison of the relationships between necessary gas velo-
city and available length for pulse formation for self-generated and
liquid-induced pulsing (solid points: self-generated pulsing #ow; open
points: liquid-induced pulsing #ow).
Fig. 19. Comparison between necessary super"cial gas velocity for
self-generated and liquid-induced pulsing #ow at equivalent available
lengths for pulse formation (;
J@
"0.0035}0.0077 m s\; ;
JN
"
0.0059}0.0128 m s\; t
@
"20 s; t
N
"2}8 s).
Fig. 20. Number of pulses generated during one liquid feed cycle as
a function of the length of the liquid-rich region (packing mater-
ial"3.0 mm spheres).
In Fig. 19 a comparison of the super"cial gas velocities
required for self-generated respectively liquid-induced
pulsing #ow at equal available lengths for pulse forma-
tion is presented for all the performed experiments. The
solid line denoting the necessary gas velocity for self-
generated pulsing #ow is obtained by intrapolation of the
modi"ed results of Fig. 5, considering the fact that the
point of pulse inception during self-generated pulsing
#ow could only be measured at "xed points along the
column axis. From the data of Fig. 19, it is concluded
that the interdependence between necessary length for
pulse formation and super"cial gas velocity is equivalent
for both self-generated and liquid-induced pulsing #ow.
5.4. Pulse frequency
Another feature of liquid-induced pulsing #ow is the
possibility to tune the pulse frequency and therefore the
time constant of the pulses. In Fig. 20, the number of
pulses generated during one liquid feed cycle is plotted as
a function of the length of the liquid-rich region for
3.0 mm glass spheres as packing material. Apparently,
only one pulse is generated when the length of the liquid-
rich region is less than approximately 0.5 m. For lengths
of the liquid-rich region above 0.5 m, the number of
pulses increases linearly with increasing length. A compa-
rable plot is encountered for 6.0 mm spheres as packing
material. In this case, the critical length beneath which
only one pulse is generated during a liquid feed cycle is
approximately 0.6 m.
6. Concluding remarks
In this contribution it is made clear that by means of
the periodic operation of a trickle-bed reactor it is pos-
sible to obtain pulsing #ow at throughputs of liquid
associated with trickle #ow during steady-state opera-
tion. Enhanced mass and heat transfer rates are then
considered to be due to the change in #ow regime. More-
over, the fairly high #ow rates required to achieve natural
pulsing #ow are generally considered to be a drawback
since they are associated with rather short contact times
between the phases. Thus applicability of self-generated
pulsing #ow appears to be restricted to chemical
J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614 2613
reactions with fast kinetics. By means of liquid-induced
pulsing #ow, longer contact times can be achieved. An
additional advantage of liquid-induced pulsing #ow is
the opportunity to control the pulse frequency. This
tuning of the pulses is an important issue in the optimiza-
tion of selectivity in catalytic reactions. Higher selectiv-
ity's can be achieved when the time constant of pulsing is
comparable to the time constant of important physical
and chemical processes (Wu et al., 1995, 1999).
The formation of detectable pulses is the result of the
growth of convective instabilities, which grow with dis-
tance into pulses. With increasing gas #ow rate, the
necessary length for pulse formation decreases. This phe-
nomenon is responsible for the upward movement of the
point of pulse inception for self-generated pulsing #ow.
This same phenomenon is found to control the process of
liquid-induced pulsing #ow. By cycling the liquid feed,
continuity shock waves are initiated in the column. As
a result, for relatively short hlf-times two regions of
di!erent liquid holdup are present in the column. At high
enough gas #ow rates, pulses are initiated in the liquid-
rich region of which the length is a linear function of
the hlf-time. With decreasing length of the liquid-
rich region, higher gas velocities necessitate the induction
of pulses. For both self-generated and liquid-induced
pulsing #ow, the relationship between the available
length for pulse formation and the required gas #ow rate
is equivalent.
Because the velocity of the initiated pulses is much
higher than the shock wave velocity, pulses will eventual-
ly move out of the liquid-rich region. In most cases the
pulses remain stable and move to the bottom of the
column. However in some cases, the pulses fade away. At
the present, it is not possible to assure that the induced
pulses remain stable in columns of longer length when
they have moved out of the liquid-rich region. Further-
more, the stability of the initiated continuity shock waves
may be a!ected by the distance these waves travel, al-
though no evidence is found these shock waves are unsta-
ble. In conclusion, it will be worth investigating the
process of liquid-induced pulsing #ow in columns of large
lengths. In the near future, experiments in a column of
3 m in height will be performed in our laboratory.
Notation
f
N
pulse frequency, s\
l
U
length of liquid-rich region, m
n
N
number of pulses during one feed cycle, dimen-
sionless
t
@
low-liquid-feed time, s
t
N
high-liquid-feed-time, s
;
E
super"cial gas velocity, m s\
;
J@
super"cial low liquid feed velocity, m s\
;
JN
super"cial high liquid feed velocity, m s\
<
U
shock wave velocity, m s\
<
U
relative shock wave velocity, dimensionless
Greek letters

@
liquid holdup corresponding to low liquid #ow
rate, dimensionless

N
liquid holdup corresponding to high liquid #ow
rate, dimensionless

N
relative liquid holdup during high liquid #ow
rate, dimensionless
References
Boelhouwer, J. G., Piepers, H. W., & Drinkenburg, A. A. H. (1999).
Enlargement of the pulsing #ow regime by periodic operation of
a trickle-bed reactor. Chemical Engineering Science, 54, 4661}4667.
Blok, J. R., Varkevisser, J., & Drinkenburg, A. A. H. (1983). Transition
to pulsing #ow, holdup & pressure drop in packed columns with
cocurrent gas}liquid down#ow. Chemical Engineering Science, 38,
687}699.
Castellari, A. T., & Haure, P. M. (1995). Experimental study of the
periodic operation of a trickle-bed reactor. A.I.Ch.E. Journal, 41,
1593}1597.
Cheng, Z. M., & Yuan, W. K. (1999). Necessary condition for pulsing
#ow inception in a trickle bed. A.I.Ch.E. Journal, 45, 1394}1400.
Gabarain, L., Castellari, A. T., Cechini, J., Tobolski, A., & Haure, P.
(1997). Analysis of rate enhancement in a periodically operated
trickle-bed reactor. A.I.Ch.E. Journal, 43, 166}172.
Grosser, K., Carbonell, R. G., & Sundaresan, S. (1988). Onset of pulsing
in two-phase cocurrent down#ow through a packed bed. A.I.Ch.E.
Journal, 34, 1850}1860.
Haure, P. M., Hudgins, R. R., & Silveston, P. L. (1989). Periodic
operation of a trickle-bed reactor. A.I.Ch.E. Journal, 35, 1437}1444.
Krieg, D. A., Helwick, J. A., Dillon, P. O., & McCready, M. J. (1995).
Origin of disturbances in cocurrent gas}liquid packed bed #ows.
A.I.Ch.E. Journal, 41, 1653}1666.
Lee, J. K., Hudgins, R. R., & Silveston, P. L. (1995). A cycled trickle-bed
reactor for SO

oxidation. Chemical Engineering Science, 50,

2523}2530.
Ng, K. M. (1986). A model for #ow regime transitions in cocurrent
down-#ow trickle-bed reactors. A.I.Ch.E. Journal, 32, 115}122.
Sicardi, S., Gerhard, H., & Ho!mann, H. (1979). Flow regime transition
in trickle-bed reactors. Chemical Engineering Journal, 18,
173}182.
Sicardi, S., & Ho!mann, H. (1980). In#uence of gas velocity & packing
geometry on pulsing inception in trickle-bed reactors. Chemical
Engineering Journal, 20, 251}253.
Tsochatzidis, N. A., & Karabelas, A. J. (1995). Properties of pulsing #ow
in a trickle bed. A.I.Ch.E. Journal, 41, 2371}2382.
Tsochatzidis, N. A., Karapantsios, T. D., Kostoglou, M. V.,
& Karabelas, A. J. (1992). Aconductance probe for measuring liquid
fraction in pipes & packed beds. International Journal of Multiphase
Flow, 18, 653}667.
Wallis, G. B. (1969). One-Dimensional Two-Phase Flow. (pp. 122}135)
New York: McGraw-Hill Inc..
Wu, R., McCready, M. J., & Varma, A. (1995). In#uence of mass
transfer coe$cient #uctuation frequency on performance of three-
phase packed-bed reactors. Chemical Engineering Science, 50,
3333}3344.
Wu, R., McCready, M. J., & Varma, A. (1999). E!ect of pulsing on
reaction outcome in a gas}liquid catalytic packed-bed reactor.
Catalysis Today, 48, 195}198.
2614 J. G. Boelhouwer et al. / Chemical Engineering Science 56 (2001) 2605}2614