Axial Dispersion in a Pulsed Plate Column

n i . H . I . BAZRD
Chemical Engineering Department, McMaster University, Hamilton, Ontario

A technique involving the indicator colour change in an acid- On a employ6 une mhthode baste sur le changement de
base reaction has been used to measure axial dispersion coeffi- couleur d'un indicatenr dans une rhction entre acide et base
cients in a 15 cm diameter pulsed column. Data have been pour dhterminer les coefficients de dispersion adale dans une
obtained mainly for single phase (aqueous) flow with two colonne pulste de 15 centimetres de diamhtre. On a obtenu des
different types of plate at two different spacings. With coarsely informations, surtout en ce qui a trait i un 6couIement en
perforated plates, the dispersion coefficient is proportional to phase simple (aqueuse) impliquant deux sortes de plaques et des
(amplitude)* times frequency, but semicircular unperforated kcartements diffbrents. Dans le cas des plaques i grosses ouver-
baffle plates show a dependence on amplitude times frequency. tures, le coefficient de dispersion s'est avtrt proportionnel
These results are critically compared with published data, and l'amplitude au c a r d rnultiplihe par la frkquence; cependant, on
two types of flow regime for axial dispersion are indicated. a constatk que les plaques en chicane semi-circulaires et non
perforkes dhpendaient de I'amplitude multiplik par la frhquence.
On fait une comparaison critique entre les r h l t a t s obtenus et
ceux qui ont dtji &t publib, et Yon indique deux types de
A xial mixing tends to reduce the effectiveness of
any countercurrent mass transfer process, by
flattening the axial concentration gradients in each
systhme d'kcoulement pour une dispersion axiale.

phase and thus reducing the overall driving force

available. This effect has been known generally for point, and the distance between the neutralisation
about 20 years and has led to a growing volume of zone and the injection point is a simple function of
research activity. In one hand, theoretical treatments the axial dispersion coefficient.
of countercurrent processing, allowing f o r backmixing, The acid-base-phenolphthalein system was first used
have been proposed. On the other hand, many experi- as a tracer by Northrup"z' in photographic studies of
mental measurements of backmixing in columns have underwater vortices. Fox and Gex(I3' have measured
been carried out. These two topics have been thorough- "mixing times" in stirred tanks using the same sys-
ly reviewed respectively by Misek and Rod"' and by tem. Tyr01er"~' has recently examined mass transfer
Ingham'*' with reference to liquid-liquid extraction of acetic acid into drops of sodium hydroxide contain-
columns, which are particularly prone to backmixing ing phenolphthalein. However, the present investiga-
effects because of the relatively low superficial velo- tion appears to be the f i r s t to apply the acid-base-
cities of the flowing streams. phenolphthalein system to the measurement of axial
This paper is concerned with the measurement of dispersion in a column.
axial mixing in a pulsed extraction column. The stand-
ard methods for such measurements involve the in-
jection into one of the phases of a tracer which can Theory - steady state
readily be analysed and is insoluble in the other phase.
In the steady-state technique, the tracer solution is The method can be modelled as a one-dimensional
continually injected and its concentration in the transport process in the axial direction, with rapid
column at different positions upstream of the injection chemical reaction, as shown in Figure 1. The alkali
point is determinedis5'. Unsteady state techniques in- concentration is considered to be uniform at czs at all
volve a single pulse, or a step change or a sinusoidally positions downstream of the injection plane ( z = 0).
varying injection of tracer, with monitoring of the Upstream of the injection plane, the net flux of alkali
Concentration a s a function of time a t a position is equal and opposite to the flux of acid, assuming
downstream of the The theory of these that one mole of acid reacts with one of alkali.
techniques is well understood and they give reasonably
consistent results, but they do require special arrange-
ments for either sampling or monitoring the tracer-
containing streams. Care has to be taken that t h e
radial position of the probe o r sampling tube is such This equation cpplies only between the injection plane
that a representative analysis for the whole stream and the reaction plane x = -2, a t which the solution
is obtained. passes through neutrality (i.e. ce =: 0 ) . All the reaction
The method used in the present work requires occurs a t x = -z,~.The boundary conditions are thus:
neither a probe nor a sampling device and consequently atz=O c2 = c23 1 ...................... .(2)
it offers some advantages in rapidity and simplicity.
It also allows the flow pattern to be observed visually. z = -zn cz=o f ....................... (3)
However, it is restricted to aqueous streams. The The solution of Equation (1) is:
principle is that the entering aqueous stream is ren-
dered slightly acidic, but an alkaline side stream is cz = - c10 + Kexp(uloz/&).. . . . . . . . . . . . . . . . . . . . . . . (4)
injected a t an intermediate point in the column. The
flow and concentration of the alkali stream are great Two values of the integration constant K can be ob-
enough to render the main flow alkaline and the tained from the boundary conditions (2 ) and (3) :
neutralisation zone can be observed visually if a suit- K = ~ 2 3+ C ~ O .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ' (5)
able indicator is present. If there is back-mixing,
neutralisation occurs upstream of the alkali injection K = clo exp(uloz,/E). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)

750 The C a n d m Journal of Chemical Enginetdng, Vo1. 52, December, 1974

 I
t
t=o

1 1 1
alkali

alkali mKn,CZ3

Figure 1 -Model for steady state deter-

mination of dispersion coefficient. Figure 2 -Apparatus - schematic diagram.
Figure 3 - Movement of colour change
after alkali injection.
Plates: Perfmated, H = 2.73 cm
Symbol: 0 X
Elimination of K from Equations ( 5 ) and (6) gives Freq. f, Hz 0.798 0.805 0.790
a simple relationship for xn: Ampl. A, cm 0.75 1.02 1.53
Function F 1.45 1.45 1.45
E from eq. (18)
z. = ( E / u I o )In [I + c23/c101.. ...................... (7) cm2/s 0.41 0.90 1.54
The exit concentration cW can be calculated stoichio-
metrically from the inflows and concentrations of acid
and alkali:
-
Czs = ( ~ 2 0 ~ 2 0 uioC1o)/(Uio + U Z O ) .. . . . . . . . . . . . . . . . .(8) (iii) 1 >0 z = - = CI = 00.. . . . . . . . . . . . . . . . . . .(16)
Hence Equation (7) becomes :
(iv) i > 0 z =0 c2 = c z 0 . . . . . . . . . . . . . . . . . . . . . . .(17)
zn = (E/uio) In [(I + cza/cio)/(l + U I O ~ Z ~ Z O.) .~ .. . . . . . (9)
This type of problem has been s o l ~ e d "~ 'f o r an
In practice, z, can be measured and the flows and unsteady-state diffusion-limited rapid second order
inlet concentrations of acid and alkali a r e known, chemical reaction occurring near an interface. The
so E can be calculated as follows: general s o l ~t i o n " ~'can be simplified in the present
case because both reacting species (acid and alkali)
E = UIOZ, Iln [(I + Czo/clo)/(l + ~io/~zo)ll-'. . . . . . . ..(lo) have the same dispersion coefficient in turbulent con-
ditions. The position of the reaction zone is given by

Theory - unsteady state Z, = (4 El)+ F . ................................. .(18)

where
An analytical solution is also obtainable for an un-
steady process in which the acid in the column is
erf ( F ) = C~O/(CZO + cl0).......................... .(19)
considered as stagnant and semi-infinite and is ex-
posed a t times t > 0 to a constant concentration czo
of alkali a t the plane x = 0. In this case, the neutrali-
zation zone moves out from the plane x = 0, rather EXPERIMENTAL
than maintaining a steady position. Two governing
Equations can be written, one for free acid and one Choice of plates
for free alkali. Considerable information is availableca' on disper-
sion in packed columns (pulsed and unpulsed), rotary
-acl
= E
a2cl
for z < -zn.. ................... agitated columns and pulsed plate columns. In the
latter categoryr3-"' however, the emphasis has been
mainly on plates with relatively small holes ( d 2:
2mm) and small free area (a =: 0.1). I n the K a r r
reciprocating-plate coIumn(lE"'' the plates have a much
more open structure with hole diameters in the order
The boundary conditions are: of 12 mm and free areas of u p to 60%. These columns
could be built in diameters up to 75 ern"'), however
(i) i =0 z <0 CI = CIO . . . . . . . . . . . . . . . . . . . . . . ..(13) there seem to be no data on dispersion coefficients
for the open type of perforated plate used. In this
(ii) 1 >0 z = -Zn CI = cz = 0 . . . . . . . . . . . . . . . . . .(14)
work, perforated plates similar to those of the K a r r
column have been studied. A segmented baffle plate
..................... has also been examined. This latter plate appeared to
have excellent radial mixing characteristics and could

The Cmodim,Journal of Chemical Engineering, Vol. 52, December, 1974 751

create a fine dispersion a t relatively low applied am- possible (3.5 mm) to minimise by-passing effects at
plitudes and frequencies, so its axial dispersion charac- the wall. In this case, a fraction 0.55 of the cross-
teristics were considered worthy of investigation. sectional area is left free for flow. The centre-to-centre
spacing can be either 2.60 or 5.15 em depending on
the positioning of the sleeves.
Apearatus
A schematic diagram of the apparatus and ancil-
laries is given in Figure 2, in the form used for
Systems used
steady-state tests. The aqueous solutions were made u p a few hours
The main extraction column is made of 15 cm before each experiment, using tap water a t room tem-
nominal diameter glass pipe with an overall height of perature (26-28°C). An entering hydrochloric acid
3 m. The mean cross-sectional area was determined concentration clo of about 0.01N and a sodium hy-
volumetrically to be 179.3 cm’. The plates extend from droxide concentration czo of about 0.3N were used
30 to 180 cm measured from the base of the main with accurate determination by titration after each
section, with alkali injection 72 cm from the base. series of tests. The flow of alkali entering the system
The alkali is injected through four 6 mm stainless was in most cases 0.41 l/min, considerably less than
steel tubes distributed around the circumference of the flow of acid to the top of the column. The phenol-
the column and fed by a circular manifold connected phthalein indicator was added with the acid stream,
to a rotameter, valve and overhead supply tank. The as its red colour is known to fade after prolonged
distributor tubes project about 3 cm in from the wall contact with alkali.
of the column. The dispersed phase (kerosene) is
pumped from a storage tank to a valve and rotameLer, The organic phase was commercial kerosene with a
and enters the base of the main column via a circular, density of 0.7906 g/ml., a viscosity of 6.83 cp and an
perforated-tube distributor. The continuous aqueous interfacial tension of 34 dynfcm with respect to
phase, consisting of dilute hydrochloric acid with a water (determined by ring tensiometer).
trace of phenolphthalein, enters the top of the column
with no special distributor.
Procedures
Both phases are removed from the column by means
of overflow chambers, which can be independently In the steady-state method, the column was f i r s t
vertically adjusted in order to control the levels of filled with dilute hydrochloric acid and the air pulsa-
the aqueous-kerosene and the kerosene-air interfaces tion system was started up. The flows of acid, alkali
a t the top of the column. and kerosene (if used) were set a t the required levels.
The main column is connected via 5 cm nominal The pulse amplitude could be measured accurately by
diameter glass pipe to the air pulsing system which observing the liquid displacement in the 5 cm diameter
is shown on the left of Figure 2. The air pulsing side arm (Figure 2 ) and then applying a factor to
is carried out on the self -triggering principle”!” allow for the ratio of cross-sectional areas. The am-
whereby the supply and exhaust of air is controlled plitude was adjusted by altering the settings of the
by the liquid level itself. The operating range of the air supply and exhaust valves. The frequency of puls-
present column is between a maximum amplitude of ing depended mainly on the volume of the pulsing
3 cm at 0.55 Hz and a maximum amplitude of 1 cm arm. I t was measured by timing 10 or 20 cycles.
a t 2.5 Hz. The operating characteristics and ad-
vantages of self-triggered air-pulsed columns have A s soon a s the alkali flow to the column was started,
been discussed more fully elsewhere‘20’. a red zone appeared in the region of the injection
The plates are supported within the main column level. The colour spread rapidly downwards (the di-
by means of four vertical 6 mm diameter stainless rection of the aqueous phase flow) but more slowly
steel rods on a 5.4 cm radius circle. Stainless steel upwards until it reached a steady limiting position.
sleeves around these rods act as spacers for the plates. When the position of colour change as measured to
The plates are made of stainless steel which is pre- the nearest k 1 cm had remained the same f o r 3
ferentially wet by water. The entire rod-and-plate minutes, the reading of zn was taken. The pulse am-
assembly can be removed by lifting it from the open plitude was then changed and Zn was read again after
top of the column. i t had become steady. This usually took about 10
The first type of plate to be evaluated is a 1.8 mm minutes with single phase flow, but somewhat longer
thick 13.7 cm diameter perforated plate provided with in the presence of kerosene drops. Data were collected
14.3 mm holes on a 17.5 mm triangular pitch. It was at different amplitudes f o r a single narrow range of
necessary to leave a gap (7 mm in this case) between frequency (frequency varied very slightly with am-
the outer edge of the plate and the wall of the column plitude). Major changes in the frequency range were
t o permit easy removal of the plate assembly. The made by shutting down the air pulser and altering the
fraction of the overall cross-sectional area of the volume of the surge chamber.
coldmn free of obstruction is 0.61. The plates are The unsteady state method was used mainly for
“staggered” in order to avoid correspondence between measuring E at low amplitudes and only for single
hole positions in adjacent plates. By the use of one phase conditions. As before, the column was initially
or two sleeves between plates, the centre-to-centre filled with dilute acid, but the acid feed was then
spacing H can be set at 2.73 cm or 5.30 cm. stopped. The pulser was started and after adjustment
The other plate is a segment baffle formed by simply of amplitude the alkali flow to the injector tubes was
cutting in half a 14.3 cm diam. circle of 0.8 mm thick started. As before, the red colour began to spread,
stainless steel. These half circle plates a r e mounted faster downwards than upwards. The rate of down-
on the support rods in alternate directions, forcing wards spread was disregarded because it was affected
the liquid phases and the pulsations t o pursue a zig- by the bulk flow to the overflow tube. The upwards
zag path in t h e column. The average gap between the spread of colour was measured a s Zn a t different times
plate edge and the column wall is made as small a s after the start of alkali flow.

752 The Canadian J o u m l of Chemical Engineering, Vol. 52, December, 1974

 TABLE1 from the set rates.
Some typical unsteady state results are shown in
EFFECT
OF FLOW
VELOCITY UPON DISPERSION
COEFFICIENT
Figure 3. Although the plots of 2% versus t& a r e lar-
gely straight lines as predicted by Equation (18),
czo U ~ O Z, E they do not pass through the origin. This is because
Conditions the boundary condition a t z = 0 (Equation (17)) is
impossible to achieve perfectly without an infinite,
Perf. plates, H = 2.73 cm 0.2145 0.3663 0.283 14 3.52 well-mixed region a t x > 0. In the apparatus, the
A = 2.39 cm, f = 0.677Hz ,, 0.204 19 2.74
uz0 = 0.0381 cm/s
”
”
0.153 35 3.24 injection of alkali a t z = 0 occurs a t a finite rate
”
0.097 74 3.62 and therefore some time elapses before the boundary
condition cg = czo is approached closely. However, the
0.00592 0.3169 0.520 18 7.10 slope of the straight-line portion of the plots in Figure
Semicircular plates 0.00780 ” 0.395 22 6.65
H = 2.60 cm 0.00780 ” 0.297 32 7.34 3 can still be used to estimate E by means of Equation
A = 0.817 cm, (18). The overall duration of an unsteady-state test
f = 1.35 HZ
u, = 0.0381 cm/s
I, 0.0151 ” 0.204 46 7.55 was typically 15 minutes and the start-up effect dis-
cussed above occupied only the first two minutes or
so.

Comparison of steady-state and

RESULTS unsteady-state results
Most of the results presented refer to single phase Data for perforated plates (H = 2.73 cm) and
flow. The steady state method has been used more single-phase conditions has been collected, using both
than the unsteady state method, though i t has been the steady-state and unsteady-state methods. The latter
found that both methods give consistent results. was found to be more suited for measuring low values
Table 1 shows the detailed results of some steady of E because the movement of the colour change zone
state tests of the effect of velocity uro upon zn was slow enough to be measured accurately. The
and E, a t constant pulsing conditions. The time-aver- steady-state method became rather tedious a t low
age distance zn could be measured only to the nearest values of E because of the longer time required to
cm because the colour change zone was oscillating and establish steady conditions. Some of the data, for two
somewhat diffuse due to turbulence. However, in all different narrow ranges of frequency, is plotted in
the single-phase and most of the two-phase cases, the Figure 4. There is no significant difference between
colour change zone was essentially horizontal, indicat- the results from the steady and unsteady methods,
ing that the overall effects were one-dimensional a s despite the fact that the liquid velocity ulo is zero in
assumed in the theory. the latter case. This confirms the evidence in Table 1
The right-hand column of Table 1 gives the values that the velocity - a t least up to 0.5 cm/s - does not
of E as calculated from the data using Equation (10). effect E significantly.
No significant effect of ulo upon E can be seen: the
scatter in the values of E for a particular pulsing
condition is about + l o % which is consistent with
Effects of amplitude and frequency
the errors in measuring zn and t h e other variables, (perf orated plates)
and the occasional tendency of the flows to drift away Figure 4 suggests an approximate square-law de-

8, I I

x steady stab method

0 unsteady state method
A “contaminated

1
03 0.5 I.0 2.0 50
amplitude A, cm A’ f , c m2/s

Figure 4 - Axial dispersion coefficient Figure 5 -Correlation of data for per-

versus amplitude for perforated plates forated plates with H = 2.73 cm (aqueous
Frw. r a m Hz: (076+ phase only).
0.81) (1.38+ 1.45)
Method used’: steady ksteady stady unsteady Figure 6 - Data for semicircular baffles,
Method used: H = 2.60 cm.
X f = 0.578 - 0.596 HZ
steady unsteady steady unsteady
Symbol A A 0 .
-
0 f = 1.36 1.46 HZ
f = 1.35 Hz, A = 0.817 cm

The C d i a n Journal of Chemical Engineering, Vol. 5 2 , December, 1974 753

pendence of E on amplitude, with somewhat less sensi- These cases are shown by vertical bars in Figure 7
tivity to frequency. More data has been obtained with with the ends corresponding t o the values of z,, at
the perforated plates (H = 2.73 em) a t frequencies either side of the tilted colour change zone. The tilting
between 0.53 and 2.68 Hz, and a correlation of E with effect became worse as the pulsing amplitude was
ABf a s shown in Figure 5 is satisfactory. The choice reduced, so that it was impossible to measure E in
of this group is appropriate because its dimensions are the absence of pulsation. I t was first thought that the
the same as those of E itself. At values of A2f up tilting effect was due to uneven alignment of the
to about 3 cm2/s, E is approximately equal to A2f, plates, but no improvement was obtained when the
but a t higher intensities the dispersion coefficient plate assembly was rotated.
rises more slowly. These results are discussed in more An attempt to use the unsteady state method under
detail later in the paper. two-phase flow conditions with perforated plates, was
unsuccessful. The rising colour change zone was ini-
Contamination tially only slightly tilted, but after a few minutes the
red colour appeared to rise preferentially in a narrow,
In one series of single phase experiments a t 1.42 Hz, fast-moving stream up one side of the column. Event-
values of E were obtained 20-30% above the main ually, the upper part of the aqueous phase became red,
correlation (see triangles in Figure 5). This discre- while most of the middle of the column was still
pancy was greater than could be expected from the colourless. This appears to be a circulation phenomenon
measurement techniques. The cause was eventually similar to what has been observed in bubble columns
traced to a new polyethylene bottle that had been in- of large diameter'21'.
troduced for storing the dilute hydrochloric acid. When The semicircular baffle plates were introduced with
the bottle was replaced by a glass carboy, the results the aims of improving the radial mixing and so
reverted to the main correlation. The acid that had preventing the establishment of vertical circulation
been stored in the new polyethene bottle was found, patterns. Typical data appear as crosses on Figure 7,
by diazotisation, to contain a trace of an aromatic indicating that there is no tilting of the colour change
amine. Such material is commonly added as a stabiliser zone and E is less sensitive t o u.d than in the case
to polyethylene. It is thought that it had a hydro- of perforated plates. Some other results (not shown
mechanical effect similar to that of polymers used in Figure 7 for reasons of space) with semicircular
in drag reduction. Mar and Babb'31 have noted that plates a t an amplitude of 0.56 cm and frequency
backmixing increases almost twofold when polye- 1.41 Hz, show that E is 6.63 cmVs independent of
thylene plates are used in place of stainless steel ones U<Lin the range 0 < Ud < 0.48 cm/s.
in a pulsed column. Unfortunately, time did not permit
further study of this interesting effect.
DISCUSSION
Effects of amplitude and frequency Most of the discussion will concern the data on
(semicircular baffles) single-phase dispersion. Unless the factors affecting
The results f o r semicircular baffles with a spacing such dispersion are understood, little progress can be
of 2.60 cm are given in Figure 6. The correlating made towards understanding dispersion in a two phase
group in this case is Af, rather than A2f as in the system.
case of the perforated plates. The scatter of the data Dimensional analysis shows that the dispersion coef-
points is somewhat better than i t was for the per- ficient under turbulent flow conditions can be ex-
forated plates (Figure 5 ) . pressed as the product of a characteristic velocity and
The flow pattern as observed in the colour change a length:
region, was more organised than in the case of the E cc ul . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. (20)
perforated plates. On the upstroke of each cycle, a
vortex formed above the diametral edge of each plate Levich'=) has pointed out that u and 1 can be taken
and then shed itself into the bulk of the liquid. On respectively as the velocity fluctuation and the scale
the downstroke, a vortex formed below the edge of associated with the largest eddies, in which case the
each plate and then shed. A continuous streaming proportionality constant in Equation (20) would be
motion occurred from the column wall, across each approximately unity. Unfortunately u and I as defined
baffle plate to the diametral edge, resulting in ex- in this way are hard to measure or predict, so most
cellent radial mixing. experimental investigations have made use of "charac-
teristic" velocities and lengths. For example, an eddy
Peclet number ullE for steady turbulent flow through
Effect of plate spacing a pipe can be usefully defined in terms of 1 as the
Some data were obtained a t approximately double pipe diameter and u as the flow velocity'23?
the previous plate spacing, i.e. 5.3 cm for the per-
forated plates and 5.15 cm for the semicircular baffles. Characteristic velocity
The increased spacing resulted in a decrease in E by
about 50% in the case of the perforated plates a t a In a pulsed extraction column, with a single liquid
given frequency and amplitude, but an increase in E phase flowing, we may identify a steady and a fluctu-
by a like amount in the case of the semicircular baf- ating contribution to the characteristic velocity. The
fles ! These seemingly contradictory results are con- time-average velocity of the continuous phase (uc)
sidered in more detail in the discussion. is usually an order of magnitude less than the fluctuat-
ing velocity due to pulsation. The pulsation velocity
Two-phase flow can be expressed in terms of its maximum value 2nAf
o r as a root mean square velocity, or more simply as
Figure 7 shows the effect upon E ,of the dispersed Af.
phase (kerosene) superficial velocity Ud. The disper- Several recent w ~ r k e r s ' ~ ~using
~ ~ ' " 'sieve plates with
sion coefficient increased approximately linearly with IT < 0.32, have found that E is approximately propor-
a d but in some cases with perforated plates the colour-
change zone was tilted from the horizontal position.
tional to (Af +
0 . 5 ~ ~ The
) . present results, given in
Table 1, are for continuous-phase velocities up to

7s4 The Canodian lournal of Chemical Engineeringryol. 52, December, 1974

 10

I+
5 x
++ ++ + +

x
+ +
x x x*tX xx
++ +

,','b 8
A
0.5- 0 '
I 0.1 0.2 0.3 0.4 05
,,4J
d h p m d phse vabcity %, c d s

Figure 7 - Effect of dispersed phase velocity at u, = 0.204 cm/s. - 2 . 0 0.5 1.0 2 5

Perforated plates at amplitude A,cm
H = 2.73 cm,
f = 1.45 Hz Figure 8 -Dispersion scale as a function of amplitude.
IJ A = 1.07 cm
A A = 1.45 cm 0 H = 2.73 cm
0 A = 1.93 cm A H = 5.30 cm,
x semicircular baffle, H = 2.60 cm, Perforated f = 1.43 - 1.47 Hz
plates v H = 5.30 cm,
f = 1.35 Hz, A = 0.852 cm
f = 0.585 - 0.591 Hz
x H = 2.60 cm
Semicircular
baffles + H=5.15 cm
__-_- line E/Af = A

about 50% of A f , but no significant effect of u, on spacing (2.73 cm) because eddies formed at one plate
E is observed. The characteristic velocity must also would be broken up by an adjacent plate.
depend on plate geometry because of possible jet or The effect of increasing the spacing between the
vortex formation, but for a given plate geometry we perforated plates is to reduce the apparent scale E I A f .
assume that the character.istic velocity is proportional In the experiments it was noted visually that the
to A f . turbulence seemed less intense midway between the
plates, and i t can be concluded that the eddies formed
Characteristic length a t the plates tended to break down to smaller sizes
There are several factors which could contribute in the poorly agitated regions between plates. The ef-
t o a "characteristic length" for dispersion in a pulsed fective scale E / A f a t H = 5.30 cm is somewhat re-
plate column. By analogy with pipe flow, the column duced at lower frequency, perhaps because the eddies
diameter D could be important, particularly for small have more time in which t o break up between velocity
diameter columns. The plate spacing H , the plate maxima. At a very large spacing, laminar zones could
thickness, and the diameter d and the pitch of the even exist between plates, leading to virtually no
holes could be significant. Finally, the amplitude A mixing. A model incorporating the idea of non-uniform
of pulsation could affect the size of the eddies. mixing has been proposed by N ~ v o t n y ' ~and
' success-
Time did not permit the variation of all six of fully tested with results for plates with small perfora-
these lengths. The nearest approach to a thorough in- tions ( d = 1.5 - 3 mm) and small open areas (a <
vestigation on these lines was made by Mar and Babb'3' 14.6%). However the model postulates an equal effect
who carried out an analysis of variance on the results of either amplitude or frequency on the dispersion
of factorially designed experiments. However, their coefficient, whereas the present results showed a
conclusions have been criticised by subsequent in- stronger effect of amplitude than frequency.
v e s t i g a t o r ~ ' ~The
~ ~ ' .main length variable in the present Several previous i n ~ e s t i g a t o r s ' ~ ~have
~ * ' ~ ' found a
work was the amplitude A , with two plate spacings direct proportionality between the dispersion coef-
(H)used for each type of plate. The single phase ficient and the characteristic velocity ( A f plus a
data is therefore presented in Figure 8 with the dis- small term for u,). For the special case where uc <<
persion scale (defined as E I A f ) plotted against am- A f , the relationships between EIAf and other vari-
plitude for different plate designs and spacings. ables can be simply summarised in Table 2. The cor-
relation of Novotny(J) is also given in its limiting
Perforated plates form for u, -+ 0.
The data for perforated plates (lower portion of Even with the effects of continuous phase flow rate
Figure 8) show an approximately linear relationship omitted, the orr relations'^^^^^^'^^ do not agree well. There
between E / A f and amplitude. The dotted line, drawn is a discrepancy on whether E / A f increases with plate
for E I A f = A , approximately represents most of the spacing H(7,8,10'or de~reases'~). The fraction of free
data a t 2.73 cm spacing. However, at the highest area u is found to have a positive effect on EIAf"'
attainable amplitudes, in the region of 3 cm, the scale or a negative one",". An increase in column diameter
E I A f seems to level off. It is believed that the scale has been claimed to decrease EIAf'", to have no ef-
of the turbulence a t higher amplitudes would reach fect"" or to increase it"', although the latter authors")
a limiting value of the same order as the plate-to-plate just included D in their correlation without varying

The Canadian Journal of Chemical Engineering, Vol. 52, December, 1974 755
 Experimental details
~- --- - Correlation for EIAf

---
Source and
reference

Novotny (5, *
col. dia.

I
D, cm
hole dia.

5.16
j
1
d, cm

0.15,
0.25,
0.2,
0.3
-I1 --____
spacing
H, cm

2.5 to 15
free area fraction

0.029 to 0.16
_-
(H
(cm)

0.15 Hd
- 4.5) o3Ia
where H > 4.5 cm
- __----
Miyauchic’)

RozenW
__-
3.2,

10
5.4
--
0.15,
-
0.15,
0.3

0.3
9 values from 1 to 10

3.75, 7.5, 15.3

0.095,

0.095, 0.23,
0.19

0.32 6 ( ;:ID r3
1.7 d ( H / D ) 2 / 3/ a

d’’3

KaganP”
-___
1
I
I
I
--__--

0.21 1
I
5 only
5, 10,15
I
-
0.087
__

I
0.98 HO.76

*Reciprocating plate column.

it experimentally. The only points of agreement are Two-phase flow

that EIAf tends to increase with hole size d‘53’,8’and The main emphasis in this work has been on single
that the dispersion coefficient E is itself proportional phase flow, but the limited data in Figure 7 permit
to the product A f . a qualitative comparison between the axial mixing
The present data for perforated plates (Figure 8) characteristics of the two types of plate tested.
differs in a major sense from the data summarised in The axial dispersion coefficient for perforated
Table 2, in that the dispersion scale E I A f varies linear- plates is rather sensitive t o the dispersed phase velo-
ly with amplitude. The hole size d in this case is city. The rising drops of the dispersed phase are
1.43 cm, an order of magnitude greater than in con- broken up to some extent by the small-scale turbulence
ventional perforated plates. This factor and the large produced a t the plates‘24’ but nevertheless they ap-
free area cr give the present plates more of the preciably add to the axial dispersion. The radial mix-
character of a coarse mesh, so that fluid pulsation past ing provided by the perforated plates is rather poor,
them results in a fine-grained turbulence with a so that vertical circulation currents due to radial
scale of the same order as the amplitude. The tur- differences in the dispersion density can be set u p
bulence tends to damp out in the regions between a t low pulsing intensities.
plates, so that E decreases as H increases. The semicircular plates give good radial mixing
Conventional perforated plates with small holes and and so axial dispersion in this case is f a r less sensitive
small free areas produce intense localised jets of to dispersed phase flow. However i t must be noted
liquid, with a large effective scale t h a t is usually that the single-phase mixing scales (Figure 8) a r e f a r
limited by the plate spacing. Hence a n increase in greater for this type of plate than for the perforated
E with H is A decrease in E with H plate.
at large values of H has been reported in some
CONCLUSIONS
The chemical indicator technique has been shown to
Semicircular baffle prates be a rapid and simple method for measuring axial
Turning now to the results for semicircular baffle dispersion in aqueous systems. The accuracy of meas-
plates (Figure 81, it is seen that E I A f is only slightly uring E is limited mainly by the accuracy with which
dependent on amplitude. The dependence is about the the colour-change position can be detected. In the
0.2 power a t a spacing of 5.15 cm and insignificant present work an accuracy of about *lo% is claimed,
at a spacing of 2.6 cm. On t h e other hand, the scale which is comparable with the accuracy of axial dis-
E I A f increases with plate spacing. I n both these re- persion data obtained by other methods‘s-11’. It is
spects, the baffle plates’ behaviour qualitatively re- proposed to extend this technique to the study of
sembles that of the small-hole sieve plates‘7***’”’ whose mass transfer in columns; an acid solute will be trans-
performance is given in Table 2. Visual observation ferred from the organic phase into an aqueous stream
showed that the flow fluctuations around t h e baffle which enters the column alkaline and becomes neutral
plates were intense, owing t o vortex shedding and a t an intermediate position in the column. The height
circulation effects. It is therefore not surprising that of the indicator colour change can, with suitable as-
the dispersion scale E I A f , f o r a given amplitude, is sumptions, be simply related t o the height of a transfer
many times greater than that for the mesh-like per- unit.
forated plates. The order of magnitude of EIAf is With regard to the axial dispemion data in pulsed
similar to the height 2H of the “mixing chambers” columns, simple relationships of the type given in
that are formed between any two of the half-circle Table 2 a r e of limited value. The concept of a dis-
plates on the same side of the column. persion scale E I A f has been found useful, and two

756 The Canadian Journal of Chemical Engineering, Vol. 52, December, 1974
 basic types of flow regime for single phase pulsed K = constant, Equations (5) and (6)
columns are postulated : I = eddy scale
= time
(a) Hindered eddy regime. The size of eddy which u = superficial velocity
would be produced a t a single isolated plate (or piece z = distance from injection plane (Figure 1 )
of packing) is larger than the spacing between the 2, = value of I where colour changes

plates (or the packing size). In this regime, the dif- Creek Zetters
fusion scale EIAf is limited by, and approximately
proportional to, the spacing or the packing size. Ex- u = fraction of free area in plate
amples of th i s regime exist in pulsed packed
and in pulsed plate co1umns“~8~’o’ where fast, persistent Subscriet
jets of liquid forming a t each cycle correspond to o = inlet streams
large eddies. = acid
(b) Free eddy regime. This regime is less well- = alkali
known than regime (a) and probably confined to plate 3 = outlet stream
= continuous phase
columns. The eddy produced a t the plate is smaller d = dispersed phase
than the plate spacing, so the diffusion scale is primar-
ily proportional to this eddy size which is in turn
proportional to amplitude for a given plate geometry. References
An increase in plate spacing allows the eddies to (1) Misek, T. and Rod, V. in “Recent Advances in Liquid-Liquid Ex-
traction” (C. Hanson, ed.), p. 197 (Pergamon Press, Oxford,
break up appreciably, so the dispersion scale E I A f 1971).
In&&, J.. ibid. p. 237.
declines with an increase in H. An example of this Mar, B. W. and Babb. A. L., Ind. Eng. Chem., 51, 1011 (1969).
regime is seen in Figure 8 for t h e perforated plates. Sehmel. G. A. and Babb. A. L.. Ind. Eng. Chem. Proc. Des. Dev.,
3. 210 (1964).
Intermediate behaviour between regimes ( a ) and Novotny. P., Prochazka, J. and Landau, J., Can. J. Chem. Eng..
48. 405 (1970).
(b) occurs when the size of the formed eddies is Kagan, S. Z., Aerov, M. E., Lonik, V. and Volkova, T. S.. Intl.
similar to the plate spacing. In such cases, the disper- Chem. Eng.. 5. 656 (1966).
Miyauchi. T. and Oya, K., AIChE Journal, 11 896 (1966).
sion scale EIAf could increase weakly with both am- Rozen, A. M.. Rubezhnyi, Y. G. and Martynov, B. V.. Khim.
Prom, 46 (1) 13” (1970).
plitude and plate spacing. An example of this is seen Barame. S.. Molinier, J. and Angelino. H.. Can. J. Chem. Eng.,
in Figure 8 for the semicircular baffles. 51, 166 (1973).
Kagan, S. Z., Veishein. B. A., Trukhanov. V. G. and Muzychenko.
A general prediction of axial dispersion coeffi- L. A. Intl. Chem. Eng, 13, 211 (1973).
Elenkbv, D., Boyadzhiev, b.. Naplatanova. M. and Atanasova, E.,
cients in pulsed columns, even for the simple case of Isv. Otd. Khim. Nauki. Bulg. Akad. Nauk. 3 (3). 419 (1970).
single-phase flow, will require knowledge of how the Northrup, E. F., Nature, 88. 463 (1912).
Fox, E. A. and Gex, V. E., AIChE Journal. 2. 539 (1966).
initial eddy size depends on plate geometry, column Tyroler. G., Hamielec, A. E., Johnson, A. I. and LeClair. B.P.,
Can. J. Chem. Eng.. 49. 56 (1971).
diameter, etc. as well as amplitude. More data is also Sherwood T. K. and Pigford R. L., “Absorption and Extraction”.
needed on the breakdown of eddies in the spaces be- pp. 332-387, 2nd Edn. ( M c G 1 . i ~Hill, New York, 1962).
Karr, A. E., AIChE Journal, 5, 446 (1969).
tween plates. It is hoped that the acquisition of more La. T. C. and Karr, A. E., Ind. Eng. Chem. Proc. Des. Dev., 11.
data on different types of plate a t different spacings A96 ~-~.-,
-”- 119791
(18) Report on Achema 1973, Chem.-1ng.-Tech., 45. 1343 (1973).
and column diameters, particularly in regime (b) , will (19) Baird, M. H. I., Proc. AIChE 1.Chem.E. Symp.. London, p. 6 5 3
(1966).
eventually lead to a general quantitative correlation. (20) Baird, M. H. I. and Ritcey, G. M., Paper presented a t International
Considerable work is needed because of the large num- Solvent Extraction Conference, Lyon, France (1974).
(21) Towell. G. D. and Ackerman, G. H..European Symp. o n Chemical
ber of variables comprising the plate geometry. Reaction Engineering. 6th Amsterdam, Preprints p. B 3-1, Elsevier

Acknowledgments
This work was carried out a t the Chemical Engineering Department,
University of Queensland, during the author’s sabbatical leave. He is
grateful to the University of Queensland for a Post-Doctoral Research
Fellowship and to the National Research Council of Canada for a
Travel Grant. H e is also grateful to the technical staff of the Chemical
Engineering Department for their skilled assistance in the design and Manuscript received May 15 ; accepted for publication August 17, 1974.
construction of the apparatus, and to the secretarial staff for the
typing of the manuscript for this paper. Finally, he is grateful to
t h e academic staff, notably Dr. R. G. Rice, f o r stimulating discussions.
This paper was presented at t h e 24th Canadian Chemical Engineer-
ing Conference, Ottawa, October 1974.

Nomenclature
A = amplitude
* * *
c = concentration (molar)
d = diameter of holes in plate
D = diameter of column
E = axial dispersion coefficient
f = frequency
F = function, Equation (19)
H = plate spacing

T h e Canadian Journal of Chemical Engineering, Vol. 5 2 , December, 1 9 7 4