Chem. Eng. Technol.

15 (1992) 401 -405 40 1

Crystallization of Potash Alum in an MSMPR Crystallizer

P. Synowiec, R. Sangl and A. Mersmann"

New data are presented on the crystallization of potash alum (KA1(SOJ2. 12H20) from
solution, in a continuously operating MSMPR cooling crystallizer, 6 1 in volume. Popula-
tion density distributions were determined by sieve analysis. Highly polydispersed product
, the range between 264 and 607 Vm. Suspen-
was obtained, with median crystal size, L f O in
sion density was varied between mT = 7 and 98 kg/m3. The results were evaluated accord-
ing to an attrition model and the obtained kinetic efficiency should be valid for other types
of crystallizers.

1 Introduction v
F1 =-
d 3n
The correct design of crystallizers requires kinetic data such
as the rates of secondary nucleation, BO,and of growth, G . Mean residence time was varied from 630 to 2010 s and the
It is as yet not possible to predict these parameters without specific power imput E ranged from 0.09 to 0.3 W/kg. The
special experiments. The usual method, proposed by Ran- product was removed isokinetically in order to obtain
dolph and Larson [I], consists in carrying out experiments
in MSMPR crystallizers (Mixed Suspension Mixed Product
Removal). For evaluating these kinetic data, a supersatura-
tion, 0,has to be established (here by cooling below the
Heod
saturation temperature), which leads to crystal growth and tank
nucleation. Heat and mass transfer, attrition, agglomera-
tion, crystallizer geometry and fluid dynamics in the
crystallizer [2 - 41 exert a strong influence on the crystalliza-
tion process and the crystal size distribution which can be Fiitcr
produced.

This paper presents one method of determining the kinetic

data by sieve analysis and regression of the population den-
sity data. The experiments were performed in a 6x m3
MSMPR-model crystallizer. The kinetic efficiency was
calculated in order to present information which is valid for
different types of crystallizers.
Fig. 1. Flow sheet of pilot plant.

2 Measuring Equipment
coolina
The flow sheet of a continuously operating MSMPR cool-
ing crystallizer is shown in Figs 1 and 2. The suspension is
agitated using a 3-blade marine type propeller. The power
number of the stirrer and the flow number were determined
as Po = 0.28 and F1 = 0.50 for turbulent flow, i.e., for an
Re number larger than 5 x lo3, from the following equa-
tions: ' )

VE
P o = -n3d5 ' and

* Dipl.-Ing. P . Synowiec, Silesian Technical University, Institute of

Chemical Engineering and Plant Design, ul. Kuczewskiego 7, 44-101
Gliwice, Poland, and Dr.-Ing. R. Sangl, Prof. Dr.-Ing. A. Mers-
mann, Lehrstuhl B fur Verfahrenstechnik, T U Miinchen, Arcis-
str. 21, D-8000 Miinchen 2.
1 ) List of symbols at the end of the paper. Fig. 2. Cooling crystallizer.

c VCtI Verlagsgesellschaft mbH, D-6940 Weinheim, 1992 0930-7516/92/0612-0401 $03.50 + .25/0

402
representative samples. The storage and dissolution tanks
were kept at 5 to 8 "C above the saturation temperature.
The feed flowing from the head tank was cooled in a heat
exchanger to approximately 2 "C above the saturation tem-
perature and partially recirculated to the dissolution and
storage tanks. A filter was installed between the storage and
head tanks in order to purify the solution. The analysis was
performed when steady-state conditions had been attained
in the crystallizer after 10 residence times.
3 Data Processing
Dry sieving in the size range between 150 and 2000pm
yielded directly the mass distribution of particles in the
crystallizer. A sample was filtered, the adhering solution
was removed from the crystals with an oily fluid, which was
later itself removed by washing with acetone. The mass,
mi, of crystals on a single sieve was calculated as the dif-
ference between the mass of the sieve with dry crystals, md,
and that of the empty sieve, m,
mi=md-m, . (2 a)
The cumulative mass undersize distribution can be express-
ed by
The median crystal size, L,,, was determined from the
cumulative mass undersize data (Fig. 3).
The population density n , i.e. the number of crystals in an
interval AL per unit suspension volume, was calculated
from the sampled sieve data according to Eq. (3):
m(Li)
n(Li)= (3)
Vsa Q,L: ALi '
The volumetric shape factor a = 1/2/3 was assumed to be
independent of crystal size.
Crystal size L
Fig. 3. Cumulative mass undersize distribution, with specific
power input, e, as parameter.
Chem. Eng. Technol. 15 (1992) 401 -405
The suspension density in the crystallizer, mT, was
calculated from the sample as the ratio of crystal mass in
the sample to sample volume.
m T =m
-C . (4)
VS
Theoretically, a semilogarithmic plot of population density
should result in a straight line if MSMPR conditions prevail
during the experiment. Deviations from this straight line are
often observed, caused e.g. by classified product removal,
agglomeration, attrition and abrasion, and kinetic phenom-
ena associated with some materials such as growth rate
dispersion or size dependent growth.
For this reason, it was decided to calculate the kinetic data,
crystal growth rate G and the effective rate of secondary
nucleation Bo,efffrom a regression analysis in the range
from 250 to 1060 pm, as follows (see Fig. 4):
- the growth rate G was determined from the slope of the
regression line,
- the rate of secondary nucleation Bo,effwas calculated
from the intercept on the ordinate at particle size L =0
where G and Bo,,ff are the mean effective rates of crystal
growth and secondary nucleation, respectively.
Influence of tip speed on crystal size distribution (CSD):
The experiments demonstrated the influence of stirring in-
tensity and residence time on the crystal size distribution
and median crystal size of the product. This is explained by
some experimental results.
With increasing power input, E , the median crystal size,
L,,, decreases and the size distribution becomes narrower
(Fig. 3). It was also found that there is a limiting residence
time (Figs5 and 6), in excess of which the median crystal
size, LS0,decreases. This may be explained by attrition ef-
fects. The attrition rate, which reduces crystal size, can ex-
ceed the crystal growth rate if the specific power input is too
Fig. 4. Semilogarithmic population density plot
Chem. Eng. Technol. 15 (1992) 401 -405
‘p = 0 021 l o 0 023
MSMPR-V =5.7dm3
300
600 1000
Propeller
5 2000
- m,:(7-
Residence time i:
Fig. 5. Median crystal size L,, versus mean residence time 5.
t specific power inputs i.
I plained this deviation by a size dependent crystal growth
0.03 0.20 W/kg 0.30
Mean specific power input 5
Fig. 6. Median crystal size L,, versus mean specific power input.
high. The maximum crystal size is then limited by attrition
and the median crystal size decreases with increasing
residence time, i.e. at lower growth rates. Therefore, in
practice, the specific power input should not be higher than
required for sufficient mixing.
4 Crystal Growth Rate
It is generally accepted that the crystal growth process oc-
curs in two stages [5, 61, i.e. bulk diffusion and surface
reaction. The “overall” growth rate applies to the combina-
tion of these two processes.
The dependence of the “overall” growth rate on relative
supersaturation r~ can be described by a semi-empirical
equation formulated by Mullin [5].
G = kggn .
The growth rates calculated from the slope of the regression
line are plotted against the relative supersaturation in
Fig. 7.
In our case, it was not possible to correlate experimental
data according to Eq. (6) since a strong effect of the specific
power input, E, on the effective crystal growth rate G was
confirmed. With increasing specific power input, E, the
crystal growth rate G decreases. A precise determination of
the effect of crystal size on the growth rate is not possible
403
‘Q =
98) kg/m3
0.OOL to 0.053
-Bo l
10 10‘ - loo
Relative supersoturation G~~~
Fig. 7. Crystal growth rate G versus supersaturation u at different
with only the population balance of our experiments
(Fig. 7).
The deviations from straight line in the population density
plots may be attributed to different phenomena such as
growth dispersion, attrition, abrasion, agglomeration and
also to classified product removal [7]. Other workers ex-
rate.
5 Secondary Nucleation
The main factors influencing the rate of secondary nuclea-
tion in MSMPR crystallizers are the supersaturation, r ~ ,
magma density, mT,and fluid dynamics in the reactor, ex-
pressed by the specific power input, E. Therefore, it is usual
to describe the relevant effects by an empirical correlation
of these factors:
The exponents i, I and r must be determined experimentally.
Usually, there is a linear relationship between the rate of
secondary nucleation and suspension density [8]. Since it
had not been possible to determine the precise relationship
between suspension density and rate of secondary nuclea-
tion from our experiments, a linear relationship was also
assumed. The other exponents in Eq. (7) were determined
experimentally using the method of least squares, which
yields the following expression:
(6) Bo,effa (E)0.73mT G . (8)
It has been mentioned in the literature that this equation
cannot be transferred to other types or scales of
crystallizers. The exponents in this equation are influenced
by different fluid mechanics at different experimental con-
ditions. Hence, Eq. (8) cannot be universally valid and
should be used only as a first approximation. Fig. 8 il-
lustrates the relationship between the effective rate of sec-
ondary nucleation and crystal growth rate. It is shown that
values of Bo,effare not constant but depend on the specific
power input and mechanical stress. An increase of either the
404 Chem. Eng. Technol. 15 (1992) 401 -405

5- W/kg
m(i 7- 98 lkp/m3 .I- E=O 20 ,"

107 -
1IT8
Growth role G
mls 1

Fig. 8. Rate of secondary nucleation, B,/(o, versus crystal growth rate,

G.
Growth rate G
m/s

Fig. 9. Kinetic efficiency Ekinversus crystal growth rate G.

power input or supersaturation and therefore of the crystal

growth rate causes an increase in the Bo,eff.
Potash alum is very sensitive to attrition effects. Hence, the
conditions of our MSMPR experiments, namely magma
density mT of up to 98 kg/m3, specific power input C<
0.3 W/kg, volumetric hold-up p50.055% and residence
times 7 5 2012 s, all affect the secondary nucleation rate.
According to [lo, 111, a kinetic efficiency factor E k i n can
be calculated from the experimentally determined rates of
secondary nucleation and the calculated attrition volume in
a crystallizer:
BO/P
Ekin =
d Va/( Vp) dt
The term dVJ( Vpdt) is the volumetric attrition rate of
crystals in a suspension [12], based on the crystal volume
Vp. The value dVa/(Vpdt) represents the sum of rate con-
tributions due to collisions with the stirrer (index R) and
other crystals (index p).
The volumetric attrition rate d Va/( Vpdt) is independent of
supersaturation and kinetics in the crystallizer.
The kinetic efficiency can provide a scale-up rule for dif-
ferent crystallizers because its value should be independent
of the crystallizer type as long as the same growth rate is
achieved. The volumetric attrition rate must be calculated
for the geometric and fluid dynamic conditions of the
respective crystallizer. Fig. 9 presents the ratio of the mea-
sured rate of secondary nucleation to the calculated attri-
tion rate plotted versus the crystal growth rate. The faster
the growth rate, the higher is the kinetic efficiency.
(9) Fig. 10 shows the relationships between secondary nuclea-
tion rate Bo,effbased on the volumetric hold-up p and the
crystal growth rate G according to own and literature data
for a constant power input. It can be observed that own
results are in good agreement with the data of other
workers.
Acknowledgement

One author (P.S.) wishes to thank the Alexander von Hum-

boldt Foundation for the financial support of this research
Finally, the volumetric rate of attrition of crystals in a sus- work and the Lehrstuhl B fur Verfahrenstechnik,
pension may be calculated as follows [2, 91:

-
edge
impeller - crystal - width

crystal-crystal

where K e , b are geometrical functions including the

volumetric shape factor, stirrer geometry and the probabili-
ty of crystals colliding with the stirrer. qT is the target effi- , , , ,,,€=(009;26!$kg] L--...--
/
-1-
10-8 m/s 10 10% 10" m/s
ciency which depends on crystal size, fluid velocity and on Growth rote G Growth role G
other factors. Fig. 10. Results of this investigation compared with literature data.
Chem. Eng. Technol. 15 (1992) 401 -405 405

Technische Universitat Munchen, for the opportunity to v, [m31 volume of sample

carry out the investigations. a shape factor
e specific power input
'pv volumetric hold-up
Received: February 6, 1991 [CET 3651 suspension density
QS
Qc crystal density
0 relative supersaturation (hydrate) ( = ( c - c,)/c,)
Symbols used 5 residence time
'IT target efficiency
a number of stirrer blades
BO rate of secondary nucleation
Bo,eff effective rate of secondary nucleation
c concentration References
cs solubility
d propeller diameter [l] Randolph, A.D., Larson, M.A., Theory of Particulate Pro-
4 draft tube diameter cesses, Academic Press, New York 1971.
D tank diameter [2] Pohlisch, J., Mersmann, A,, Chem. Eng Technol. 11 (1988)
Ekin kinetic efficiency factor pp. 40 - 49.
G growth rate [3] Mersmann, A., Sangl, R., Kind, M., Pohlisch, J., Chem. Eng
H liquid height Technol. 11 (1988) pp. 80-88.
'CP growth rate constant [4] Beer, W., Mersmann, A., Ger. Chem. Eng3 (1980) pp. 333 - 341.
L average crystal size [5] Mullin, J. W . , Crystallization, 2nd ed., Butterworths, London
Li average crystal size in a given size interval 1972.
L50 median crystal size [6] Karpinski, P.H., Chem. Eng Sci. 40 (1985) pp. 641 -651.
AL width of a given size interval [7] Mersmann, A., Chem.-Zng.-Tech.54 (1982) No. 7, pp. 631 -643.
mc total mass of crystals 181 Garside. J., Davey, R.J., Chem. Eng Commun. (1980) No. 4,
m, mass of dry crystals on a single sieve pp. 349- 358.
ms mass of sieve [9] Mersmann, A., AngerhOfer, M., Gutwald, T., Sangl, R., Wang,
mT magma density S . , Sep. Technol. 2 (1992) pp. 85-97.
F1 volumetric flow number ( = v / ( s d 3 ) ) [lo] Mersmann, A., Chem. Eng Process. 23 (1988) pp. 213-223.
n population density [ l l ] Sangl, R., Mersmann, A., Proceedings of the 11th Symposium on
Po power number ( = Ve(s3ds)) Znd. Crystallization, Garmisch-Partenkirchen 1990.
Q3 cumulative mass fraction undersize [12] Pohlisch, J., Thesis, TU Miinchen 1987.
S stirrer speed [13] Zableka, M., Thesis, TH Ziirich 1979.
t draft tube thickness [14] Schachtl, M., Internal Report, TU Munchen 1984.
V crystallizer volume [15] Jancic, S. J., Thesis, Univ. London 1976.
va crystal volume produced by attrition [16] Ottens, E.P.K., Thesis, TU Delft 1973.