(187c) Design and Characterization of a Continuous Stirred Tank Crystallizer

Design and Characterisation of a Continuous Stirred
Tank Crystallizer

G. Morris, G. Hou, M. Barrett, B. Glennon

Solid State Pharmaceutical Cluster, The School of
Chemical and Bioprocess Engineering, University College Dublin, Belfield,
Dublin 4, Ireland

Keywords: Mixed Suspension Mixed
Product Removal, Benzoic Acid, Cooling Crystallization, Process Analytical
Technology

Introduction

This
study presents the development of a continuous crystallization platform consisting
of a single stirred tank Mixed Suspension Mixed Product Removal (MSMPR)
crystallizer. A photo and schematic drawing of the MSMPR setup appear in
Figures 1 and 2 respectively. A recirculation loop is incorporated such that
the withdrawn product is recycled to a dissolution vessel which serves as the
feed tank to the crystallizer. Two OptiMax^TM workstations
(Mettler-Toledo) are used as both the feed/dissolution tank and the MSMPR
crystallizer. Difficulties can often arise in the continuous removal of
representative, non-classified product from lab-scale MSMPRs as a result of the
relatively low volumes involved and consequently slow flowrates and removal
velocities observed [1]. Intermittent withdrawal, whereby a certain
percentage volume of the MSMPR is periodically removed at a high velocity so as
to aid the potential for iso-kinetic removal and mitigate classification has
been applied previously and is adopted in this work [2, 3]. This
quasi-continuous transfer of the MSMPR product to the feed/dissolution vessel
is achieved using vacuum with 7.1% of the suspension volume removed per
transfer.

Figure
1: Image of the MSMPR setup showing feed/dissolution vessel on the left and
MSMPR on the right with FBRM in-situ

Figure
2: Schematic drawing of the MSMPR setup showing application of vacuum in
recycling material from MSMPR to feed/dissolution vessel

In-situ
process analytical technologies (FBRM, PVM), together with periodic sampling of
the withdrawn product at steady-state and during the dynamic period of the process,
have been applied in order to characterize the continuous cooling
crystallization of benzoic acid from water and ethanol. Comparison is drawn
between high magma density crystallizations (feed saturated at 40^oC
with the crystallizer at 18^oC) and low magma density
crystallizations (feed saturated at 15^oC with the crystallizer at 0^oC).
For the low magma density runs the influence of residence time on affecting the
steady-state particle size was investigated. For each of these studies,
continuous operation was started with a suspension in the MSMPR at time zero by
first batch cooling the saturated feed to the desired MSMPR temperature. In
addition, a further set of experiments was carried out to investigate whether
the mode of start-up bears any influence on the final steady-state product
obtained and the route to steady-state.

Characterization
with FBRM and Periodic Sampling

The
crystallization is monitored in-situ using FBRM and PVM. This allows real time
tracking of the changes in the particle size distribution and attainment of
steady-state. Figure 3 shows the dynamic response in the FBRM total counts and
mean chord length trends for a high-magma density crystallization proceeding to
steady-state.

Figure
3: Tracking MSMPR to steady-state with FBRM for a high magma density run of 60
minute residence time. Showing regions: (A) wash-out period, (B) nucleation
response and (C) attainment of steady-state

Figure
4: MSMPR solute concentration and magma density trends from start-up to
steady-state for a 15 minute residence time high magma density run

As
the crystallization proceeds initially there is a wash-out period (A) where
particles are withdrawn in the absence of nucleation as the supersaturation
builds. Over this period crystal growth is observed which is confirmed by an
increase in the FBRM size trends and in-situ PVM images (Figure 5). Supersaturation
increases to a maximum just before region (B), where the kinetics of the system
force a period of constant nucleation which steadily consumes some of the
supersaturation and slightly shifts the CSD back towards the fine end.
Eventually the rate at which new crystals are formed by nucleation balances the
rate at which they are withdrawn from the MSMPR and the system tends towards a
steady-state (C).

Figure
5: In-situ PVM images from a low magma density run of 15 minute residence time
showing the change in particle size during MSMPR start-up. Taken at: (A) time
zero after batch cooling at 1.5^oC/min,
(B) at 1 residence time, (C) at 2 residence times and (D) at 3 residence times

High
& Low Magma Density Crystallizations

Comparing
high and low magma density crystallizations it was observed that the level of
the nucleation response in the dynamic route to steady-state is reduced when
the slurry density in the crystallizer is relatively low (Figure 6), demonstrating
the dependency of the nucleation rate on magma density. Consequently for the
high magma density process the steady-state chord length distribution is
shifted more towards the fine end (Figure 7).

Evolution
of Particle Size with Increasing Residence Time

In
an MSMPR, for a given temperature drop between feed and crystallizer, it is
possible to access different positions on the phase diagram by varying the
residence time and as a result manipulate the steady-state supersaturation. The
speed of the crystallization kinetics and the meta-stable zone width will
ultimately dictate the supersaturation range accessible, and whether changing
residence time will be favourable in terms of manipulating particle size
advantageously. The benzoic acid system used in this study has a small
meta-stable zone width and is more growth dominated. Consequently, the
accessible range of supersaturation is limited but particle size has been shown
to grow with residence time (Figure 8). Figure 9 demonstrates that steady-state
particle size increases whilst MSMPR solute concentration and hence
supersaturation deceases with increasing residence time. 

Figure
8: Steady-state CLDs (un-weighted) for low magma density runs of varying
residence time

Figure
9: Demonstrating increase in particle size with residence time whilst solute
concentration decreases

Influence
of Start-Up Mode on Attainment of Steady-State

Several
start-up regimes for running an MSMPR could potentially be employed. To examine
whether particular choices in this area can have a significant bearing on the
steady-state particle size obtained by the system, three methods of starting
the MSMPR were investigated whilst keeping all other conditions the same:

(i) Batch cooling the
saturated feed to the desired MSMPR operating temperature, therefore allowing
continuous operation to be started with an equilibrium suspension in the MSMPR
at time zero

(ii) Starting with a
clear solution which is saturated at the MSMPR operating temperature

(iii) Seeding a
solution that is saturated at the MSMPR operating temperature with isolated
steady-state product from a previous run, therefore creating a suspension of
the final product at time zero

All
start-up methodologies were assessed through high magma density benzoic acid
crystallizations, where the feed to the MSMPR is saturated at 40^oC
and the crystallizer is operated at 18^oC. The resulting steady-state
chord length distributions from each start-up regime are seen to be similar in
all cases (Figure 10). This suggests that the final steady-state condition
attained by the system is largely unaffected by the start-up methodology
adopted.

Figure
10: Steady-state CLDs (un-weighted) for high magma density crystallizations
with different start-up regimes

Comparison
of the FBRM total counts trends during the dynamic route to steady-state
suggests that starting with saturated solution that is seeded with final
product material may offer the quickest route to steady-state (Figure 11). Unlike
the equilibrium batch start-up, in this case the initial particle size distribution
is the same as the final and as a result the system does not require periods of
significant oscillation in the size distribution to develop the steady-state
CSD for the continuous process. This therefore may represent an optimum
start-up strategy for MSMPRs, as the time required to reach steady-state is
reduced and hence wastage of material whilst the final product is potentially
out of spec is minimised.

Figure
11: Comparison of the dynamic response in the FBRM total counts trends as the
system proceeds to steady-state for all three start-up methodologies
investigated

References

[1] A.D. Randolph, M.A. Larson, Theory
of Particulate Processes, 2^nd edition, Academic Press, New York,
1988.

[2] V.A. Chan, H.M. Ang, Journal of Crystal Growth 166 (1996) 1009-1014

[3] E. Kougoulos, A.G. Jones, K.H.
Jennings, M.W. Wood-Kaczmar, Journal of Crystal Growth 273 (2005) 529?534