Chemical Engineering Science, 1969, Vol. 24, pp. 1083-1095. Pergamon Press. Printed in Great Britain.

Residence time distribution in the liquid phase of

trickle flow in packed columns

W. P. M. VAN SWAAIJt, J. C. CHARPENTIER and J. VILLERMAUX

Ecole des Industries Chimiques, UniversitC de Nancy, France

(Firstreceived 17September 1968; in revisedform 17 February 1969)

Abstract-The residence time distributions in packed columns under trickle-flow conditions have been
measured with by injection of a tracer. Analysis of the results obtained in columns packed with Ras-
chig rings of 6.4, 10.3 and 22 mm indicated the occurrence of two different elementary mechanisms
causing a spread in residence time, viz. an eddy diffusion process and mass exchange with stagnant
areas. The rate of this mass transfer was expressed in terms of the height of a transfer unit, which was
found to be about O-4-0.6 m.

INTRODUCTION in counter-current two-phase operations.

THE SIMULTANEOUS passage of liquid and gas Miyauchi, et a/.[11 considered the dispersion in
through a packed column is an important opera- each of the phases in a two-phase process.
tion in the chemical industry. It takes place in Sternerding and Zuiderweg[2] published a no-
gas absorption towers, distillation columns and mogram showing with reasonable accuracy
chemical reactors (e.g. trickle-flow hydrodesul- the influence of the dispersion in two-phase
furization reactors). The principal aim of a flow. They found that if one wants to employ
process of this type is to distribute the liquid over a larw number of real extraction stages, axial
a large area-formed by the surface of special dispersion soon becomes a limiting factor.
packing elements such as Raschig rings -and The extent to which axial mixing occurs is
thus to enhance mass and heat transfer. usually found by residence time distribution
The present investigation deals especially measurements. Axial mixing at trickle flow condi-
with the residence time distribution in the liquid tions has been the subject of a number of investi-
in a packed column. The different elements gations [3-81. However, the results reported
of a liquid entering a column generally will not differ widely. De Waal[6] measured residence
remain in it for exactly the same time, as liquid time distributions in a column packed with
elements will mix up and overtake each other. 25 mm Raschig rings. He found symmetrical
As in many cases it is desirable for all the liquid distribution curves and concluded that the spread
elements to be given the same treatment, this, in residence time was very small, One layer of
axial, intermixing will nearly always have an Raschig rings was equivalent to one ideal mixer,
adverse effect. In cylindrical vessels it is often irrespective of liquid and gas flow rates. Hoogen-
convenient to distinguish between the disper- doorn and Lips [4]$on the other hand, found for
sions in the axial and the radial direction, as they Raschig rings of 12.5 mm highly skewed
usually differ in order of magnitude. As regards residence time distributions showing pro-
good radial mixing, this is essential to effect mass nounced “tailing”. This indicates the existence of
and heat transport to the wall of the process relatively stagnant regions in the liquid phase.
vessel and to diminish the influence of radial The present investigation was undertaken in
velocity differences. view of the divergence in published data and in
Axial mixing is particularly disadvantageous order to search into the basic phenomena.

tPresent address: Koninklijke/Sbell-Laboratorium, Amsterdam (Shell Research N.V.), The Netherlands.

1083
 W. P. M. VAN SWAAIJ, J, C. CHARPENTIER and J. VILLERMAUX

MEASURING METHODS AND PROPERTIES In 1953 Otake and Okada[l2] proposed for
OF THE PACKINGS the dynamic hold-up for packed columns with-
Residence time distributions were measured out gas flow:
in two columns containing different packings of
dumped Raschig rings (see Fig. 1). Packing and
column dimensions are presented in Table 1.
Only the system water-air was studied.
This correlation was found to hold for the wet-
table packings used in our investigation for water
and other liquids, accuracy being within 20 per
cent [ 131. Non-wettable packings had lower
hold-ups than indicated by Eq. (1).
The static hold-up or capillary hold-up, PC,
was found by weighing dry and wet packing in a
separate column element. A packing had a
somewhat higher static hold-up after prolonged
use than during the first hours of water contact.
The static hold-up for wettable packings was
correlated with the Eiitvos number[ 141 (see
Fig. 2).
Fig. 1. General arrangement of the equipment. 1. Column; In most of the earlier work, the residence time
2. Calming section; 3. Static pressure measuring point; distribution was measured outside the column.
4. Packing; 5. Grid; 6. Calming section; 7. Liquid distributor;
This means that either the liquid flow from the
8. Static pressure measuring point; 9. Roots compressor; 10.
Water ring pump; 11. Orifice plates; 12. Gas flow rifuter; 13. column had to be mixed (De Waal[6]) or the
Filter; 14. Wet bulb thermometer; 15, 50 1 container; lb
Centrifugal pump; 17. Rotameters; 18. Sergot quick closing --____._
pfqc)cJ

valve; 19. Shock wave eliminator; 20. Beaudoin valve. WATER WITH SURFACE-ACM/E AGENT
----SHULMAN

EPC’ m3/m3
As a complete description of residence time
distributions in packed columns cannot be given 1 l OTAKE
0 SMLMAN
without some knowledge of the hydrodynamic
4- 0 PROST
properties, we also measured the liquid hold-up, 05- +DE WAAL
WATER

pressure drop etc. v MORTON

Fenske et a1.[9] stated that the liquid hold-up A PRESENT INVESTIGATION
can be split up into a dynamic and a static part 01 -
(free-draining and non-free-draining hold-up). In .
d 0
the present investigation the dynamic hold-up Q-v”
was found by collecting liquid after interruption
0.04 9
of the flow. Even a few hours after the liquid flow ‘10
OaO5 +
.had been interrupted an extremely small quan- w5:Y
tity of liquid was still flowing from the packing. ‘\\
Therefore, for practical reasons the dynamic 0.004. I I I \
4 40 ioo mm
hold-up was defined as the quantity of liquid
that drained from the packing during a ten-minute d2
Ea=PL
period following the interruption of the flow (see 71
also Shulman et aL[lO] and Mayo, Hunter and Fig. 2. Correlation of the static hold-up with the Eiitvos
Nash[ll]. number.

1084
 Table 1. Properties of the packings
La
8
Height of the b
measuring Specific
section Porosity surface Number of
Wetting particles s.
m (4 (S&J s
dk m (void fraction) 8, properties m-l per m3
:
0 5.
I 10.0 10.6 -1.7 10.3 4.41 2+8 0.10 1.567 0.69 1.34 non- 1360 721,890 s
wettable g
*
silicone-
B
coated a
0 g
II 10.0 10.6 -1.7 IO.3 4.41 2.08 0.10 1.567 0.69 5.0 wettable 1360 721,890
E
0
III 6.5 6.3 ~0.8 6.4 2.13 2.10 0.10 1.63 0.70 5.0 wettable 2818 3,2O8,O93 B
p 5
IV 21.5 22.5 ==3.3 22 8.72 2.72 0.193 2.47 0.733 2.2 wettable 687.8 62,194 g
 W. P. M. VAN SWAAIJ, J. C. CHARPENTIER and J. VILLERMAUX

tracer concentration in part of the liquid had

to be considered representative of the entire
liquid outflow. Schiesser and Lapidus [5] showed
that in some cases of trickle flow axial mixing MAGNET VAIYE
might vary with the radial position, it then being
difficult to obtain a sample representative of the
whole column.
In the present investigation the tracer con- ----I
LIQU D DISTRIBUTOR
centration was measured in the packing. For this
purpose two layers of silvered Raschig rings
were placed in the column, which were separated
by one or two layers of non-conducting rings.
Contact with outside was provided by special
H’ Z
supports (see Prost [15]>. The electrical resis-
tance due to contact betweer the individual
rings in the conducting layer was found to be
negligible. The conductivity of the column sec-
tion between the two layers was measured to
obtain a quantity proportional to the tracer t
concentration in a specific cross section of the \
WEASURING ELEMENT
column. The two electrodes were placed in a
t
Wheatstone bridge fed with a 1000 Hz alterna- \
-SUPPORTING GRID
ting current to avoid polarization. A recorder Fig. 3. Experimental set-up.
was connected in the circuit as a measuring
instrument. A 3 per cent solution of NaCl was the Dirac’s 6 function. De Waal[6] used a step
injected in the centre of the liquid distributor by function injection. He found that the transient
means of an injection system consisting of a behaviour when water was, displaced by a 10
magnetic valve and a pressure container (Fig. 3). per cent salt solution differed from that when the
This arrangement made it possible to obtain a salt solution was displaced by water. This would
short injection time (< 1 set), which was indi- mean that the physical properties of the tracer
cated on the recorder paper with a special pen. change the flow pattern. Under our experimental
Also, it almost always gave a stable base line on conditions we found no difference between the
the recorder when no salt was injected. Devia- transient responses to the two step functions. A
tions due to fluctuations of the hold-up were pulse function injection was made directly after
extremely small. The displacements registered the two step function injections. The integrated
by the recorder were proportional to the tracer responses to a pulse function injection and the
concentration in the liquid. An approx. 9 per two step function injections are shown in Fig. 4.
cent salt solution injected into the column gave They are essentially the same. In our experi-
the same reduced residence time distribution ments c,, was about O-3 g/l. and so Ap/p = 3 X
curve as the 3 per cent solution. It was ensured 10V4. This value is much lower than that in the
that the injected quantities were always so small experiments carried out by De Waal, which may
that the conductivity did not change when pure explain the difference in the observations con-
water was injected at the same pressure for the cerning the transient behaviour.
same time. So the change in hold-up due to the Finally we would mention here the experi-
tracer injection could be neglected. ments of Rulkens and Rietema[l6], who
For nearly all the measurements a pulse in- compared the spread in residence times measured
jection was used as a physical approximation of inside and outside the packing. They found the
1086
 Residence time distribution in packed columns

- INTEGRATED PULSE RESPONSE G=O

V SALT SOLUTION DISPLACES WATER
E PACKING
F
R ASCHIG RINGS. 403 mm
l WATER DISPLACES SALT SOLUTION
z = 4.57m
L, kg/m*.sEc

n 1.8 46,6
l 5.3 72.9
A 9.2 72.3
74.7

0 4-O 4.5 8’

Fig. 4. Comparison of the responses to a &Dirac pulse

function injection and step function injections.
Fig. 5. Influence of the liquid flow rate on the residence time
distribution fora non-wettable packing.
results obtained by the two methods to be in
good agreement.
A G=O
87 RASCHIG RINGS, 03 mm
RESIDENCE TIME DISTRIBUTIONS FOR
THE DIFFERENT PACKINGS; INFLUENCE 4% Z =1.57m ,,7
C’ L, kg/mZ.sEc Pe =E
OF THE “WETTABILITY”
2- . 2.3 155
For each packing used in our experiments the v 7.9 30.9
first measurements were made at trickle flow AiI.4 39.8
without gas flow. The first packing (I) was made 017.5 78.A

non-wettable with a silicone resin (SISS SI

1107) by a method developed by Prost [ 171
i-
and Charpentier [ 181. The walls of the column
were also treated with silicone to avoid preferen-
tial wetting of the wall. The packing treated in
this way had a low static hold-up and the dy-
namic hold-up was lower than that of a non- 0
treated packing. The non-wetting properties of
the packing, however, did not last: after some Fig. 6. Infhrence of the liquid flow rate on the residence time
time of liquid contact the packing became wet; distribution for a wettable packing.
table, which caused an increase in the static and
dynamic hold-ups. packing (I) except for its wetting properties.
Figure 5 gives the residence time distribution When comparing Figs. 5 and 6 we see that:
as a function of the water flow rate for a non- (1) A wettable packing gives strongly asym-
wettable packing. The measurements concerned metric curves for low and moderate liquid flow
were made as soon as the column was filled, so rates (“tailing”). With a non-wettable packing
that the packing was as little wettable as pos- the reduced curves are practically the same for
sible. Figure 6 shows the influence of the liquid all liquid flow rates; only at extremely low flow
flow rate on the residence time distribution for a rates do the residence time distribution curves
wettable packing (II). This packing was equal to become asymmetric.
1087
 W. P. M. VAN SWAAIJ, J. C. CHARPENTIER and J. VILLERMAUX

(2) At high flow rates the residence time

distribution curves are the same for a wettable
and a non-wettable packing.
Strong asymmetry (“tailing”) was also ob-
served with other wettable and nearly wettable
packings.
To find a measure for the spread in residence
time we compared the residence time distribu-
tion curves with those of the axially dispersed
plug flow model of Danckwerts [ 191 and
Levenspiel and Smith[20]. For the simple case
of a measuring section in an infinite tube
Levenspiel found:

~=~)112e~p[-Pe(41~e’2], (2) Fig. 8. Axial dispersion in the liquid phase in a packed

column.

where: 8 = ut/Z, u = mean real fluid velocity, different packings as functions of the ratio of the
Pe = P&let number = (uZ/D) , D = diffusion static to the dynamic hold-up. For packings with
coefficient. d 3 10 mm the experimental Bodenstein num-
We calculated the P&let numbers by compar- bers are grouped around a single curve. For
ing the place and the height of the peak of the /3& > 8 the Bodenstein numbers reach a con-
experimental curves with the theoretical values stant and no longer depend on the hold-up
derived from Eq. (2) (see for example Fig. 7). ratio. The spread in residence time for &/PC > 8
In order to compare the results obtained with is probably caused by elementary mechanisms,
such as a division of rivulets by the particles,
remixing in the cavities between the particles
C
t (C DIAGRAM FOR THE MODEL) etc. These mechanisms are the same as those

‘1 A L =2.3kg/m”.sEc
PeE15.5
which cause the spread in residence time in
single-phase flow. In Fig. 9 a comparison is made
between data on single-phase flow dispersion
collected by Levenspiel and Bischoff[21] and
those of trickle flow for &/PC > 8. The disper-
sions for trickle flow and single-phase flow are
essentially the same under these conditions.
For &/PC > 8 an additional mechanism
Fig. 7. Comparison of the experimental curve with the
“axially dispersed plug flow” model; experimental curve increases the spread in residence time. The
fitted to the model. higher dispersion values belonging to these
conditions are not due to a diffusion-like process
different packings we introduced the P&let superimposed on the flow. This we concluded
number based on the nominal diameter of the from backmixing tests with a coloured tracer.
packing elements, which is also called Boden- For trickle flow without gas flow backmixing
stein number: was always very low.
From the shape of the residence time distri-
bution curves it may be deduced that there
occurs a slow mass exchange with stagnant
Figure 8 gives the Bodenstein numbers for the regions, causing “tailing”. This mass transfer
1088
 Residence time distribution in packed columns

FROM LlTEmwE REsuus)

-PHASE Fq lOnnn FPMING
m TRICKLE FLOW, 64 mm PPCKING
m TRICKLE FLCW, 1Omm RAKING
eza TRICKLE Row. 22mm PpcKlt’JG
I TRlCKLE FLOW, 26mm PACKING(DE WAAL)
Rd,_ FOR TRICKLE FLOW

0 20 50 (00 200 500 -

R% FOR SINGLE-PHASE FLOW

Fig. 9. Comparison of dispersion at single-phase flow and

trickle flow for pcJ/3e > 8.

takes place by a change of the direction of the p1 (CALCULATED)

revulets, rupture and coalescence in the local l G=O
flow, turbulent diffusion etc. The molecular V COUNTERCURRENT
diffusion coefficient of the tracer has no influence I V COCURRENT
on this tailing phenomenon, as has been shown
by Rulkens and Rietema[l6]. From a compari-
son between the directly measured total hold-up
in the measuring section, j3j (measured), and the
hold-up deduced from the mean residence time
and liquid flow rate it can be concluded that
practically no completely stagnant regions exist
which are inaccessible to the tracer (see Fig. 10).
Figures 2, 9 and 11, together with the equation
of Otake and Okada [ 121 for the dynamic hold-up,
enable a reasonable estimation to be made of
axial dispersion in the liquid phase under trickle
flow conditions.
0
INFLUENCE OF THE GAS FLOW RATE 0 10 20 30
Figure 12 shows for a few conditions the Bt (MEASURED)

influence of the gas flow rate on the Bodenstein Fig. 10. Measured hold-up and hold-up calculated from mean
number for countercurrent flow. Before the load- residence time and liquid flow rate (packing IV).
ing point no influence of the gas flow can be
detected. In the loading region the Bodenstein
number increases and the residence time dis- the liquid induced by the countercurrent gas
tribution curves become more symmetrical. In flow gives rise to lower Bodenstein values.
this region the dynamic hold-up increases and At cocurrent flow we also found an influence
apparently there is better contact between the of the gas flow rate. But generally the effects
stagnant regions and the dynamic phase. When were smaller and the relation with the gas flow
the flooding point is approached, backmixing of rate rather complicated [ 141.
1089
 W. P. M. VAN SWAAIJ, J. C. CHARPENTIER and J. VILLERMAUX

AXIALLY DISPERSED PWG FLOW

I

3
0 RAS4JilG RINGS, 64 mm, WETTABLE C4 c2
K
A RASCHIG RINGS, iO.3 mq WETTABLE
l RASCHIG RINGS, 22 mm, WETTABI
A RASClilG RINGS, 40.3 mm, NON-WETTABLE
V RASC-llG RbJGS, 25.4 mm, DE WAAL
0 RASGilG RINGS, 12.7 mm, HooGENMw)RN STAGNANT REGION
AND LIPS
I III I I
t 2 3 45 40 20 pd lOBILE PHASE

Fig. 13. Model comprising “axially dispersed plug flow with

Fig. 11. Correlation for axial dispersion in the liquid phase at mass exchange with stagnant-areas” (model PDE).
trickle flow.

Bo l L = 17.4 kg/mz.s
2t RASCHIG RINGS, 22 mm
V L = 7.9 kg/mz.s I
‘I L = 7.9 kg/m’s
l L = 2.3 !q/m2a 1 RASCHIG RINGS, 10 mm
x+m C,andC,+ 0
where

volume of mobile phase

’ = volume of mobile phase + stagnant phase *

Q, kg/m’.% The moments of the distribution function for

Fig. 12. Influence of the gas flow rate on the Bodenstein C, at x = 1 can be calculated directly from the
number at countercurrent flow. Laplace transforms of (4) and (5) by:

MODEL COMPRISING “PISTON FLOW WITH akC

(uk= (-I)k lim. 1 (6)
AXIAL DISPERSION AND MASS TRANSFER S+0 ask -
WITH DEAD-WATER AREAS” (MODEL PDE) We found:
A picture of the mixing mechanism at trickle
flow is given schematically in Fig. 13. In the pL1= 1 +b (7)
mobile phase there is axially dispersed plug flow
with mass exchange with the stagnant regions. 4 4 2(11--(o)*
From the mass balance for this system we can P2 = F+p+ N (8)
deduce the following differential equations for
transient behaviour:

=+z
~+$cY~+N(C,-C2) (4)
+V-_(P)~
N*

& cc*--Cl) =o, (5)

from which it follows
=P2- (P 1 )2=2+2_+2(1-1p)*
that the variance ti

PP* N
with initial conditions 8 = 0, C1 = C2 = 0 and
boundary conditions for a pulse injection: As this static liquid is situated mainly on the
1090
 Residence time distribution in packed columns

contact points of the packing, there will be no Figure 14 gives the height of a transfer unit:
axial mixing within this phase.

Calculation of the mass exchange rate Htr = $7 (14)

The stagnant regions observed at trickle flow
are associated with water retained in the packing W RASCHIG RINGS. 64 mm
by capillary forces. If the packing is poorly A RASCHIG RINGS. IO.3 mm
0 RASCHIG RINGS. 22 mm
wetted by the liquid (packing I), stagnant zones
are absent. The static hold-up for such a packing
is very low. As, in addition, the spread in resi-
dence time is clearly related with the ratio be-
tween dynamic hold-up and static hold-up it
seems reasonable as a first approximation to
identify the static hold-up with the stagnant
regions.
so: 041 I I I III
dynamic hold-up (0 50 (00 500
(10)
’ = dynamic hold-up + static hold-up’ Fig&14. Height of a transfer unit for mass exchange between
mobile phase and stagnant phase.
The P&let number for the dynamic phase P is
supposed to be constant. Indeed we found ex- which seems to be independent of the Reynolds
perimentally that if the influence of the static number.
hold-up can be neglected (non-wettable packing) Hoogendoom and Lips[4] neglected the dis-
the overall P&let number is constant (see persion in the mobile phase and found for Qin.
Fig. 5). So P can be found from the value of the Raschig rings a height of a transfer unit of about
Bo number for high values of Pd/Pc (Fig. 8). The 50-60 cm. Under the experimental conditions
number of mass transfer units for mass transfer they used (very low liquid flow rates, cp very
between the dynamic and the static hold-up can small) this is a reasonable approximation and
now be estimated by comparing the relative agreement with our results is satisfactory.
standard deviation around the mass u’, of the
experimental curve with that given by the model. MIXING CELL MODEL
Let us define Mixing can often also be described by mixing
cell cascades. The elementary mixing pattern for
Mk = fomtky dt, (11) trickle flow can be represented by the element
given in Fig. 15. The liquid is mixed stage-wise.
where y is the recorder deflection t seconds after
injection of the tracer. Then we find:

d =J(5$- 1). (12)

For the model we can deduce:

Fig. 15. Elementary mixing unit for the mixing cell model.
1
Cr’=-Cr. (13)
k A small quantity (Y passes through a stagnant
zone V,. The moments of the residence time
From Eqs. (12) and ( 13) N can be calculated. distribution function for a cascade of j elements
1091
 W. P. M. VAN SWAAIJ, J. C. CHARPENTIER and J. VILLERMAUX

can be deduced from the mass balances: from the standard deviation of the model dis-
tribution function and the experimental standard
deviation. Figure 16 gives the exchange flow
C(i- 1) = C,(i) +vy (15) fraction (Yas a function of the Reynolds number.

C(i- 1) = C,(i) ++$ (16) I RASCHG RNS, 6*4mm

A RASCHG RINGS, 10~3mm
’ RASCHIG RINGS, 22 n-m
aC,(i) + (1 -a)C,(i) = C(i), (17) at AA
where a .

e=$,7=j[LX71+(1-(Y)72],
_+!_--!--_;-m l

TI= WbQ,, 72 = O’,/(l -a>Q>. 7

With a pulse injection to the first unit: “t

o-oil I I I
lo 50 iO0 500
C(0) =S@). (18) R.+!$

Fig. 16. Exchange coefficient for mass exchange between

BY applying Laplace transforms there is mobile phase and stagnant phase (mixing cell model).
obtained:
Although the models described above pro-
UC) (19) bably reflect essentially the basic mechanism of
the mixing process in packed columns it is not
certain that the dynamic and the static hold-up
are completely identical with the mobile phase
and the stagnant regions.
The best value of the ratio of mobile phase to
stagnant phase, together with the exchange
From Eq. (20) it follows that the moments for parameter, can be found by comparing the ex-
(Y> Oare: perimental residence time distribution curves
with the distributions given by the models. This
Pt = 1 (21) gives more information than only the first two
moments. For the Model PDE this has been
2 l+cr(A*-1) +j-1
solved by Villermaux; the results will be
‘*=J’ [l+a(A-l)]* j
published later[22].
6 1+&13- 1) 6(j- 1) In Table 2 the skewness of the experimental
CL3=jG[1+o(A-l)]3+ j curves, defined as:

(1 +a@*- ‘)I+ (j- lj:j-2! (23) R _ (PP (24)

‘[1+&C-l)]* mc8.
t/-G3’
where:
Vl’l(l---cu) is compared with the skewness of the model
A =
_- . distribution curves:
v2a

The number of stages, j, can be found from R Cal. _ (P3)*

(p2)3’
these experimental conditions where the stag-
nant areas can be neglected. For this case
ti = 1/j. The exchange parameter cythen follows Within the accuracy of the experiments no
1092
 Residence time distributionin packed columns

Table 2. Mass exchange parameters

Packing L, kg/m*. secQ’)* Model PDE Mixing cell model

cp P N R,,. R-1. j u V,lV, &a.

Raschig rings 6.4 mm 2 008710 0.624 160 3.8 1.2920 1.2888 81 0.0391 166 1.2955
Z = 1.63 m 2.3 007138 0664 160 3.8 1.2300 1.2348 81 0.0385 198 1.2402
Pc=S% 3 0*05446 0686 160 4.7 1.1919 1.1752 81 00445 2.19 1.1790
5 0.03566 0.756 160 5.1 1.1198 1.1124 81 00442 3.10 1.1144
7.9 O-03234 0.804 160 3.8 I.1062 1.1023 81 0.0338 409 1.1043
10 0.02568 0.816 160 5.0 1.0787 10798 81 00402 4.42 1.0809

Raschig rings 10-3 mm 1*8* 004062* 0*800* 72* 6*1* l-1274* 1.1261* 37* 0*0711* 4*0* 1.1280
Z= 157m 2.3 0.1524 0.541 72 3.3 1.5361 1.5111 37 0*0700 1.18 1.5322
Pc=5% 7.9 0.1054 0.677 72 2.6 1.3966 1.3554 37 0.0530 2.10 1.3710
*pF = 1.34% 11.1 00684 0.737 72 3.3 1.2361 1.2212 37 0.0583 2.80 1.2291

Raschig rings 22 mm 2.3 007284 0.619 140 4.9 1.2210 1.2355 71 0.0540 1.63 1.2413
Z = 2.47 m 2.3 009286 0605 140 3.9 1.3175 1.3068 71 00457 I.53 1.3142
& = 2.2% 5 0.05162 0661 140 6.1 1.1574 1.1623 71 00616 1.95 1.1673
5 0.03916 0.717 140 6.4 1.1198 1.1228 71 0.0598 2.53 1*1251

1805 0.03862
0.02772 0.751
0.770 140 7.8
5.0 1.1223
1.0846 10856
1.1218 71 00485
00623 3.02
3.34 10870
1.1245

= 0.0147 -nF2
I>=000743

*Non-wettable packing.

systematic differences between R,,, and R,,,, of Nancy. We thank him and Professor Rietema (Eindhoven)
can be detected. for the encouragingdiscussions.

CONCLUSIONS
If the static hold-up of a packed column is low
in relation to the dynamic hold-up (as it is with NOTATION
non-wettable packings, high liquid flow rates or
A Column cross section, m2
large packing elements), the spread in residence
Bo Bodenstein number
time for the liquid phase in packed columns is
very small and comparable to the dispersion at C concentration, mole/m3
single-phase flow. The Bodenstein numbers vary CO concentration of the tracer if it was
between 0.5-2. If the ratio &/& is smaller than uniformly distributed over the
8, the dispersion increases because of slow mass measuring section, mole/m3
C’ ymIlm02
transfer between the stagnant regions and the
mobile phase. The height of a transfer unit for e L(C) . Laplace transform of C
this mass transfer is about 06-0~4 m and depends D axial dispersion coefficient, m2/sec
only little on the diameter of the packing ele- Dl axial dispersion coefficient in the
ments used. mobile phase, mVsec
d nominal particle diameter, m
Acknowledgments-The present investigation was carried 4 diameter of sphere having the same
out under the direction of Professor P. Le Gaff, University specific surface as the particle, m
1093
 W. P. M. VAN SWAAIJ, J. C. CHARPENTIER and J. VILLERMAUX

dk column diameter, m z distance from the top of the column, m

Et3Eiitvos number 2 height of the measuring section, m
acceleration due to gravity, m/se?
c”
gas flow rate, kg/m*.sec Greek Symbols
o! liquid fraction passing through the
H' packing height, m
stagnant region
HtT height of a transfer unit
Pt, Pd, PC total, dynamic, static hold-up (fraction
zi
number of model units
mass transfer coefficient, set-l
L liquid flow rate, kg/m*.sec
of void volume), /It = Pd + PC
surface tension, N/m
Mk kth moment around the origin of
Dirac’s 6 function
Y =f (t) void fraction
N number of transfer units dynamic viscosity, kg/m.sec
P P&let number for the dynamic phase = t/r
Pe P&let number (overall) based on 2 = tmo/m,
Q volumetric flow rate, m3/sec kth moment around the origin in the
Re’ Reynolds number, u C diagram
Rei Reynolds number, u,d/v kth moment around the origin in the
Re& Reynolds number, uodg/v C’ diagram
Red, Reynolds number, ud,lv V kinematic viscosity
s Laplace transform variable PL density, kg/m3
&I specific surface’of a particle, m2/m3 u d\/(p2 - l,~~*)standard deviation in the
u Q/A & mean real liquid velocity, C diagram
mlsec u’ v/(&- (pi)*) standard deviation in
uo superficial liquid velocity C’ diagram
Ul Q/AePd mean velocity based on the 7 Z/u, set
mobile phase 71 VJcxQ, set
X z/Z 72 v2/(l--a)Q, set
Y recorder deflection t set after tracer volume of mobile phase
injection, mm cp volume of mobile phase + stagnant phase

REFERENCES

[l] MIYAUCHI T., MCMULLEN A. K. and VERMEULEN T., U.S. Atomic Energy Commission Rep. No. UCRL-
3911 Rev. 1960.
[2] STEMERDING S. and ZUIDERWEG F. J., Chem. Engng 1963 168 CE 156.
[3] KRAMERS H. and ALBERDA G., Chem. Engng Sci. 1953 2 173.
[4] HOOGENDOORN C. J. and LIPS J., Can..l. them. Engng 1965 43 125.
[S] SCHIESSER W. E. and LAPIDUS L.,A.Z.Ch.E.JI 19617 163.
[6] DE WAAL K. J. A., Thesis, Delft 1965.
I71 OTAKE T. and KUNUGITA E., Chem. Engng, Tokyo 1958 22 144.
[8] SATER V. E. and LEVENSPIEL O., Znd. Engng Chem. Fundls 1966 5 86.
[9] FENSKE, TONGBERG and QUIGGLE, Znd. Engng Chem. 1934 26 1169.
UOI SHULMAN H. L., ULRICH C. F., WELLS N. and PROULX A. Z.,A.Z.Ch.E.JI 1955 1247,259.
I1 11 MAYO F., HUNTER T. G. and NASH A. W.,J. Sot. Chem. Znd. Land. 1935 54 375.
[12] OTAKE T. and OKADA K. Chem. Engng, Tokyo 1953 17 176.
[13] CHARPENTIER J. C., PROST C., VAN SWAAIJ W. P. M. and LE GOFF, P., Chim. Znd. GEnie chim. 1968
99 803.
[14] VAN SWAAIJ W. P. M., Thesis, Eindhoven 1967.
[I51 PROST C., Chem. Engng Sci. 1967 22 1283.
[16] RULKENS W. and RIETEMA K., Private communication.
[17] PROST C., Thesis, University of Nancy 1965.
1181 CHARPENTIER J. C., Thesis, University of Nancy 1968.

1094
 Residence time distribution in packed columns

[19] DANCKWERTS P. V., Chem. Engng Sci. 1953 2 1.

[20] LEVENSPIEL 0. and SMITH W. K., Chem. Engng Sci. 1957 6 227.
1211 LEVENSPIEL 0. and BISCHOFF K. B., Patterns of Flow in Chemical Process Vessels. Advances in Chemical
Engineering. Academic Press 1963.
[223 VILLERMAUX J. and VAN SWAAIJ W. P. M., Chem. Engng Sci. 1969 24 1097.

ReSum& La distribution des temps de sejour du fluide en Ccoulement ruisselant dans des colonnes d
gamissage en et6 determinCe par injection de traceur. L’analyse des r&hats obtenus dans des
garnissages d’anneaux Raschig de 6,4; 10,3 et 22 mm conduit B invoquer deux mtcanismes ClCmen-
taires responsables de la dispersion des temps de sejour, B savoir un processus de diffusion turbulente,
et un Bchange de mat&e avec des zones stagnantes. L’importance du transfert de mat&e est carac-
t&see par une hauteur d’unite de transfer& qui a et6 trouvee ici Cgale B environ 0,4-0,6 m.

Zusammenfassung-Die Verweilzeitverteilung in Ftlllkor perkolonnen mit Rieselstrdmung wurde mit

Hilfe eines Spurtoffes gemessen. Die Analyse der an Rascig-Ringen von 6,4 bzw. IO,3 und 22 mm
ermittelten Ergebnisse veranlasst zur Annahme, dass der Verweilzeitverteilung zwei Elemen-
tarmechanismen zugrunde liegen, und zwar ein Prozess turbulenter Diffusion und ein Stoffenaustausch
mit Stillstandzonen. Der Umfang des StolBlbergangs wird von der Hdhe eines Ubertragungseinheits
gekennzeichnet, die sich im vorliegenden Fall aufetwa 0,4-0,6 m belief.

1095

CES Vol. 24 No. 7 E