Page 165 - Chiral Separation Techniques

5.3 Membrane-Assisted Chiral Separations 143

Fig. 5-14. The optical purity (R/S)
of the outflows as a function of the
number of transfer units (NTU) in
the apparatus at equal flow rates.
Different lines are given for differ-
ent α values.

of transfer units can now be calculated. This is shown in Fig. 5-14. For a 99 % pure
product (R/S = 100), about 190 NTU are required at an enantioselectivity of 1.05, a
number decreasing to approximately 30 when α increases to a value of 1.20.
Once the required number of transfer units for a given degree of separation is
known, the height of a transfer unit (HTU) must be determined in order to design the
column. Membrane modules containing well-spaced Cuprophan regenerated cellu-
lose hollow fibers with an internal diameter of 200 µm and a dry wall thickness of 8
µm were used. The walls of the fibers are filled with the aqueous “membrane liquid”
and the two chiral liquids flow through the lumen and the shell side of the module,
respectively. Typical values for HTU were found to be in range of 2 to 6 cm [65].
These extremely low values are mainly due to the high specific surface area that can
2
3
be obtained in the hollow-fiber modules (typically in the order of 10 000 m m– )
[62]. The total length of the membrane column used for the separation (L) is now
given by
L = NTU · HTU (5)
Using the calculated and measured values for NTU and HTU, typical column
lengths of 2–5 m are expected.
To demonstrate the potential of the process in obtaining both enantiomers at a
high purity, experiments were performed using racemic norephedrine as the com-
pound to be separated. Two columns of seven small membrane modules were used.
The enantiomer ratios in the outflows during start-up are shown in Fig. 5-15. It can
be concluded that the system reaches equilibrium within approximately 24 h, and
that both enantiomers are recovered at 99.3–99.8 % purity.
To evaluate the economics of this process, a cost model has been developed to
estimate the separation costs for a specific racemate [68, 69]. For this purpose, the
sensitivity of the separation costs for several key process parameters have been
established as compared to a base-case separation in which a purity of 99 % is
required at an enantioselectivity of 1.15. The maximum solubility of the drug is set