Page 266 - Chiral Separation Techniques

244 9 Modeling and Simulation in SMB for Chiral Purification

packing procedure proposed by Nicoud [36] was used to fill eight SMB columns.
The adsorbent bed was compacted until the column reached approximately 10 cm
length. The eight columns were tested with a nonretained compound (1,3,5-tri-tert.-
butylbenzene). Retention times are very reproducible, showing deviations smaller
than 2 % and the total porosity was found to be 0.67.
A breakthrough curve with the nonretained compound was carried out to estimate
the axial dispersion in the SMB column. A Peclet number of Pe = 1000 was found
by comparing experimental and simulated results from a model which includes axial
dispersion in the interparticle fluid phase, accumulation in both interparticle and intra-
particle fluid phases, and assuming that the average pore concentration is equal to the
bulk fluid concentration; this assumption is justified by the fact that the ratio of time
–4
constant for pore diffusion and space time in the column is of the order of 10 .
The competitive adsorption isotherms were determined experimentally for the
separation of chiral epoxide enantiomers at 25 °C by the adsorption-desorption
method [37]. A mass balance allows the knowledge of the concentration of each
F
*
component retained in the particle, q , in equilibrium with the feed concentration, c .
i
i
*
In fact q includes both the adsorbed phase concentration and the concentration in the
i
*
fluid inside pores. This overall retained concentration q is used to be consistent with
i
the models presented for the SMB simulations based on homogeneous particles. The
bed porosity was taken as ε = 0.4 since the total porosity was measured as ε = 0.67
T
and the particle porosity of microcrystalline cellulose triacetate is ε = 0.45 [38].
P
This procedure provides one point of the adsorption isotherm for each component
*
F
(c , q ). The determination of the complete isotherm will require a set of experiments
i i
using different feed concentrations. To support the measured isotherms, a dynamic
method of frontal chromatography is implemented based on the analysis of the
response curves to a step change in feed concentration (adsorption) followed by the
desorption of the column with pure eluent. It is well known that often the selectivity
factor decreases with the increase of the concentration of chiral species and there-
fore the linear + Langmuir competitive isotherm was used:
×
*
.
q = 135 C + . 7 32 0 .087 C A (35a)
A A
1 + . 0 087 C + 0 .163 C B
A
×
*
q = 135 C + . 7 32 0 .163 C B (35b)
.
B
B
1 + . 0 087 C + 0 .163 C
A B
–1
A pulse of a racemic mixture (5 g L each enantiomer) was carried out to check
the adsorption model and to predict the mass transfer coefficient. The other model
parameters used in simulation were ε = 0.4 and Pe = 1000. The mass transfer coef-
ficient used to fit experimental and model predictions in the pulse experiment was
–1
k = 0.4 s . Model and experimental results are compared in Figs. 9-16 and 9-17.
Considering that the separation system is fully characterized, i.e., adsorbent and
mobile phases, column dimensions, SMB configuration and feed concentration, the
optimization of the TMB operating conditions consists in setting the liquid flow rates
in each section and also the solid flow rate. The resulting optimization problem with
five variables will be certainly tedious and difficult to implement. Fortunately, the