Chiral Separation of Propranolol Hydrochloride Through an SMB Process Integrated with Crystallization

J. Ind. Eng. Chem., Vol. 12, No. 6, (2006) 868-876

Xin Wang†, Yue Liu, Hong Wei Yu, and Chi Bun Ching

Division of Chemical and Biomolecular Engineering, School of Chemical and Biomedical Engineering, Nanyang Technological University Singapore 637722

Received February 27, 2006; Accepted July 6, 2006

Abstract: Resolution of propranolol hydrochloride was studied in self-packed columns of perphenyl carbamoylated beta-cyclodextrin (beta-CD). Both the bed voidage and linear equilibrium constants were evaluated by moment analysis of a series of linear elution chromatograms. A modified h-root method was used to determine the competitive Langmuir isotherm of propranolol hydrochloride in the nonlinear region. Continuous separation of the target enantiomer from its racemic mixture was studied using Simulated Moving Bed (SMB) chromatography in both the linear and nonlinear region. Desired (S)-propranolol hydrochloride was produced in the raffinate product in high purity. The solubility of propranolol hydrochloride was determined experimentally in methanol at different temperatures. The crystallization of propranolol hydrochloride from solutions of different initial composition in a mixed solvent of methanol and acetone was also investigated with different product purity and yield. The SMB productivity was further increased at the sacrifice of decreasing the product purity. The obtained solution was further purified by crystallization. Compared with direct crystallization, which is suitable only for the racemic conglomerate, the integrated process is especially suitable for the majority of chiral drugs, which belong to racemic compounds as long as suitable and economic chiral stationary phases (CSPs) are available in the SMB separation.

Keywords: separation, propranolol hydrochloride, simulated moving bed (SMB) chromatography, crystallization, chiral drugs

Introduction 1) The chirality of drugs is an important issue from phar- macological, pharmacokinetic, toxicological, and regula- tory points of view [1,2]. Nowadays, most research efforts concentrate on the production of optically pure products because of the increasing demand that such drugs be administered in optically pure form [3]. Pro- pranolol is one of the most important beta-blocker drugs Figure 1. Molecular structure of propranolol hydrochloride. because a variety of analogous compounds have been developed based on it. It is mainly used in the treatment shown in Figure 1. of hypertension and cardiac arrhythmias and it has been The simulated Moving Bed (SMB) process has been reported that its desired activity resides in the S-(-)- applied extensively to the separation of chiral drugs and enantiomer. Propranolol hydrochloride has one chiral intermediates over the last decade [4-7]. Because of a center and is supplied in its hydrochloride from, as continuous countercurrent contact between the liquid and solid phases, the SMB process requires less desorbent

† and the improves the productivity per unit time and unit To whom all correspondence should be addressed. mass of the stationary phase. The SMB process is (e-mail: [email protected], [email protected]) Chiral Separation of Propranolol Hydrochloride Through an SMB Process Integrated with Crystallization 869 believed to be able to achieve high-purity separation, linear and nonlinear regions. The solubilities of the even when the resolution exhibited by an individual racemate and single enantiomer of propranolol hydro- column is not efficient for a batch preparative process, chloride in methanol was determined experimentally at which is often the case in chiral separations. One of the different temperatures. The crystallization of propranolol key issues in operating the SMB process is to determine hydrochloride at different initial compositions in mixed the zone flow rates and the column switching time. solvents of methanol and acetone was investigated with Developed in the frame of equilibrium theory, which different product purity and yield. To increase the pro- neglects the effect of axial mixing and mass transfer ductivity of the desired (S)-enantiomer, the SMB resistances, triangle theory is currently widely applied in experiment was run at higher feed concentrations and SMB design approaches [8,9]. In this method, the zone flow rates using the partially resolved product development of the SMB resorts to its corresponding obtained in the raffinate stream. The obtained solution hypothetical true counter-current (TCC) process; the was concentrated and purified through the crystallization most important parameters required are those of the bed process. voidage (or total porosity) and equilibrium isotherms of the enantiomers to be separated. The TCC operation parameters can then be converted to the SMB unit based Theoretical Background on the geometric and kinematic equivalence between the two processes [10,11]. Column Physicochemical Properties and Adsorption However, the high cost of enantioseparation process, Isotherm especially the use of chiral stationary phases (CSPs), The bed voidage can be evaluated from the zero reten- which usually demonstrate good enantioseparation abil- tion time of a non-adsorbed component to the stationary ities toward specific compounds/drugs, makes the large- phase. For a component that enters the pore system but scale application of SMB in chiral separation less does not adsorb on the surface of the stationary phase, favorable. Crystallization techniques, on the other hand, the retention time is given by remain important and economic processes for industrial- scale production and purification of enantiomers [12]. ε ε    (1) Racemate crystals can be divided into racemic com-    pounds, racemic conglomerates, and pseudoracemates (solid solutions). Although diastereomer crystallization, For packing materials with two pore systems (i.e.,) which is often referred to as classical resolution, has been micropores and macropores), the column total porosity εT studied in detail for more than a hundred years, the and bed voidage are related by the equation selection of the resolving agent is still a matter of trial and error. Preferential crystallization is more attractive, ε   ε (2) but it can be accomplished directly only for conglomerates. Unfortunately, only 5～10 % of all racemates are It is well known that chromatographic separation de- conglomerates: the majority of chiral substances belong pends primarily on the adsorption isotherms, which relate to racemic compounds. Only partially resolved solution the solute concentration in the mobile phase to that of the enantioenriched by another technique, whose composi- stationary phase over the concentration range of interest. tion is over the eutectic composition, can be separated In the diluted region, the linear isotherm is expressed as using this technique. The coupling of liquid chromatography, especially the * (3) SMB process, with crystallization has been investigated q i = K i ․C i recently for efficient enantioseparation [13-15]. In this study, resolution of the racemate of propranolol hy- The method of moments is used to determine the drochloride was achieved on a column packed with adsorption equilibrium of the column. For a linear perphenyl carbamoylated β-cyclodextrin (β-CD) immo- isotherm model, the first moment is expressed as [16] bilized onto silica gel. Both the bed voidage and linear equilibrium constants were evaluated from a series of   ε        (4) linear elution chromatograms conducted at different    ε   interstitial velocity. A modified h-root method was used to determine the competitive Langmuir isotherm of The first moments of the enantiomers to be separated propranolol hydrochloride in the nonlinear region. Com- can be plotted against the inverse interstitial velocity of plete separation of a racemic mixture of propranolol the mobile phase and the linear equilibrium constants can hydrochloride by the SMB was achieved in both the be determined readily from the slopes of the lines. 870 Xin Wang, Yue Liu, Hong Wei Yu, and Chi Bun Ching

It is well known that SMB is preferably conducted in Langmuir competitive adsorption coefficients bi. Thus, n the nonlinear region to achieve higher productivity; equations can be used to determine the unknown bi (i = therefore, it is more important to determine the com- 1,2,…n). petitive adsorption behavior among the feed species. In particular, the non-stoichiometric Langmuir isotherm is SMB Separation of Propranolol Hydrochloride important in SMB development because constraints on In the frame of equilibrium theory, which neglects mass the flow rate ratios (i.e., m1, m2, m3, and m4) in the SMB transfer resistances and axial dispersion, a true counter- unit can be determined explicitly on the frame of current (TCC) adsorption model was employed in a equilibrium theory [8]. It can be expressed as series of efforts to obtain explicit expressions of the fluid-

to-solid flow rate ratios, m j (j =1,…4), for complete a c separation of binary mixtures [8,9,20-23]. The operation * j j (5) q j = n conditions of the SMB were then determined based on 1+ ∑ b ic i i =1 the equivalence between the SMB and TCC processes by maintaining the liquid velocity constant relative to the where the values of ai are measures of the intrinsic solid velocity in the two processes. In particular, the affinities of the respective species for the sorbent, and the desorbent is usually nonadsorbable (or it is so weak that values of bi are characteristic of the nature and strength its adsorptivity is negligible) for enantiomeric separation; of the interference produced by the species. It is worth explicit criteria have been obtained [8] to determine the noticing that the linear isotherm can be seen as a boundaries of the complete separation region in the space particular case of the nonstoichiometric Langmuir and spanned by m j (j =1,…4). It should be noted that the linear equilibrium constants Ki being equal to the purity and yield of both components are 100 % in theory Langmuir coefficients ai. within the complete separation region. The h-root method without the introduction of dummy The fluid phase flow rate over the solid phase flow rate species has been applied to determine the non-linear of a TCC unit can be defined as competitive Langmuir isotherms of nadolol, a three- chiral-center beta-blocker drug [17]. In this method, the TCC   individual isomers of interest, which are often not Q j =  (8) m j =   commercially available, are not required and only a very Q S  small amount of the racemic mixture is needed. This situation facilitates the determination of the isotherms for which can be converted to the flow rate ratios of the racemic drugs. This method divides the determination of equivalent SMB unit using the conversion equation the Langmuir parameters into two parts. The intrinsic   affinity coefficients were obtained from linear elution      (9) chromatography, and competitive interference coeffi-   cients were obtained from non-linear frontal chromatography. The parameters m j (j = 1,…4) define a four-dimen- The equations used to determine the competitive Lang- sional space divided into different regions, and it is muir coefficients of racemic mixture are given as follows useful to consider the projection of the four-dimensional [18,19]: regions onto the ( m 2,m 3) plane. The boundaries between the different separation regions depend only on the n  f  c i adsorption isotherm of the mixtures to be separated and ∑  ' b i=1 (6) i =1   k i  the feed concentration and composition. Having decided  ' -1   K n  the values of m j (j=1, …4) and t* (or Q1), Equation 9 is often used to determine the liquid flow rate in the four n  c f  sections of the SMB and, thus, the flow rates of the inlet  i  (7) ∑  ' ' b i= 1 j = 1,2,⋯n -1 and outlet streams. The advantage of this approach are i =1 k i K j +1   -1  that the flow rate ratio is a dimensionless group bringing  ' '  k j +1 K j together information regarding the column volume (V), unit flow rates (Qi), and switch time (t*), and, thus, it can where the values of f are feed concentrations and ' C i k i be applied whatever the configuration, size, and produc- ' and K i are the elution capacity factors and frontal tivity of the SMB unit in both linear and non-linear capacity factors, respectively. systems. In equations (6) and (7), all of the terms are known or can be determined experimentally, except those of the Chiral Separation of Propranolol Hydrochloride Through an SMB Process Integrated with Crystallization 871

SMB unit. The flow rates of the two inlet streams, i.e., the feed and eluent, as well as two of the three outlet streams, e.g., the extract and raffinate, were controlled, thus leaving the recycled eluent stream free and determined by the overall material balance of the SMB unit. An online vacuum degasses (SUPELCO) degassed all of the liquid being pumped into the system. The concentrations of the extract and raffinate streams were analyzed using a Shimadzu SCL-10 AVP chromatographic system. The samples of products were collected at the middle of the switch times at different cycle and switch times. An analytical column (25 cm × 0.46 cm I. D.) packed with perphenyl carbamoylated β- CD bonded onto 5-µm silica gel was used to analyze the Figure 2. Schematic diagram of the SMB unit: 8 columns, concentration of samples based on calibration lines 2-2-2-2 configuration, open looped. obtained previously from external standard solutions. The absorbance wavelength was set at 220 nm. All chromatographic experiments were conducted at room Experimental temperature (ca. 23 oC).

Chemicals HPLC-grade methanol was obtained from Fisher Scien- Results and Discussions tific (Leics, UK). Glacial acetic acid and triethylamine were obtained from Merck (Germany). HPLC water was Elution Order of the Enantiomers of Propranolol Hy- prepared in the laboratory using a Millipore ultra-pure drochloride water system. The racemate mixture of propranolol To determine the elution order of enantiomers of hydrochloride was purchased from Sigma (St. Louis, propranolol hydrochloride, samples of the two stereoi- MO, USA). All purchased products are used without somers of propranolol hydrochloride were injected into further purification. the column separately under the same chromatographic The empty column (25 cm × 1 cm I. D.) assembly was conditions as those used for the racemic mixture of purchased from Phenomenex (USA). The columns were propranolol hydrochloride. We found that (S)- and packed with perphenyl carbamoylated beta-cyclodextrin (R)-propranolol hydrochloride correspond to the first and bonded onto silica gel using an Alltech pneumatic liquid second peaks of the racemate propranolol hydrochloride, pump (Alltech, USA) and the slurry packing method. respectively. Thus (S)- and (R)-propranolol hydrochloride The silica gel was supplied by Eka Chemicals AB are enriched in the raffinate and extract streams in the (Sweden) with a particle size of 16 µm (KR100-16-SIL). SMB experiments, respectively. The eluent (desorbent) used was a binary mixture containing 60 % aqueous buffer solution (1 % TEAA, pH Determination of Bed Voidage = 4.5) and 40 % methanol. The feed solution was pre- 1,3,5-Tri-tert-butyl benzene (TTBB) has been widely pared by dissolving the racemate propranolol hydro- used for the determination of column dead time tOR for chloride in the desorbent at certain concentrations. The various CSPs [24]. Although the sorption to the eluent and feed solution were degassed in a model LC 60 perphenyl carbamoylated β-cyclodextrin is strongly sup- H ultrasonic bath prior to running the experiment. ported by a phenyl group, this group is surrounded and shielded by the three tert-butyl groups in the case of SMB Separation System TTBB. Furthermore, an exclusion mechanism is not In the SMB unit, the countercurrent contact between the likely to occur due to the relatively small molecular size solid and mobile phase was achieved by periodically of TTBB. Therefore, TTBB is believed not to be retained shifting the inlet (feed, desorbent) and outlet (raffinate, in the stationary phase and was chosen to determine the extract) ports in the direction of the fluid flow. In this total porosity εT of the column in this study. study, the SMB separation unit was open-looped and The total porosity, εT, was determined from the re consisted of eight columns (25 cm × 1 cm I. D.) sponse to a pulse injection of TTBB. The retention time arranged in a 2-2-2-2 configuration, i.e., two columns per of TTBB in the column was corrected by deducting the section (see Figure 2). Five flows (feed, eluent, extract, retention time of the TTBB peak measured when the raffinate, and recycled eluent) are needed to handle the injector was connected directly to the detector. 872 Xin Wang, Yue Liu, Hong Wei Yu, and Chi Bun Ching

Figure 3. Plot of mean retention time of TTBB against the mobile phase inverse flow rate.

Figure 5. Different separation regions in SMB experiments. Feed concentration: (1) 0.15 mg/mL; (2) 0.75 mg/mL; (3) 1.5 mg/mL.

species, was applied to determine the non-linear competitive Langmuir isotherms of the two enantiomers. Although ideally only one frontal experiment is nec- essary to determine the competitive Langmuir coefficients b , the possibility of experimental error and the Figure 4. Retention time of propranolol hydrochloride plotted i against the inverse superficial velocity of the mobile phase. difficulty to determine Ti accurately necessitates other confirming frontal experiments, which may be conducted at different concentrations of the step changes of the The zero retention time of TTBB was given by Equa- solutes and at different flow rate of the mobile phase. In tion 1. From the plot of the mean retention time against this study, the experiments were conducted at concen- the inverse flow rate in Figure 3, the total porosity ε T trations of propranolol hydrochloride of 0.754 and 1.004 was found to be 0.64. From Equation 2, the bed voidage mg/mL, respectively and flow rates of the mobile phase for the column was found to be 0.34. of 3 and 4 mL/min, respectively. The competitive Langmuir coefficients of the two Determination of Equilibrium Isotherm components of propranolol hydrochloride were evaluated The linear isotherm was valid only in the linear concen- at the average of b1 and b2; and the final isotherms at the tration range. Thus, all pulse experiments must be concentration range studied were given as performed under dilute conditions. Dilute propranolol hydrochloride samples were used in the chromatog- 4.357c raphic experiment. With a continuous decrease of the q * = 1 1 1+1.484c +3.495c amount of samples injected, there was only very slight 1 2 difference in the first moments of the two peaks. According to the experimental results, the concentration * 6.307c 2 q 2 = of propranolol hydrochloride solution at 0.104 mg/mL is 1+1.484c 1 +3.495c 2 believed to be in the linear isotherm region. The first moments of the two components of propran- SMB Separation of Propranolol Hydrochloride olol hydrochloride are plotted against the inverse su- In the design of SMB experiments, one is mostly con- perficial velocity of mobile phase in Figure 4. Straight cerned with the projection of the four-dimensional space, lines were fitted to the experimental points. According to m j (j=1,…4), onto ( m 2,m 3) plane, i.e., the plane in the Equation 4, the equilibrium constants were determined operating parameter space spanned by the flow rate ratios from the slopes of the lines, which we found to be 4.36 of the two key sections of the SMB unit. From and 6.31 for (S)-propranolol hydrochloride and (R)-pro- adsorption isotherm determined previously and the feed pranolol hydrochloride, respectively. concentration, complete separation regions for propran- The h-root method, without the introduction of dummy olol hydrochloride separation was constructed in the Chiral Separation of Propranolol Hydrochloride Through an SMB Process Integrated with Crystallization 873

Table 1. Operating Conditions and Separation Results of SMB Experiments Product Flow rate ratios Switch Flow rates (mL/min) Raffinate Run time Purity (%) Productivity

m1 m2 m3 m4 t* (min) Q1 QF QR QE Raf Ext (mg/day) C 7.19 4.76 4.96 3.58 15 6.33 0.16 1.12 1.97 99.9 99.8 15.91 D 7.19 4.92 5.12 3.60 15 6.33 0.16 1.23 1.84 99.6 99.2 17.47 F 7.48 4.60 4.70 2.90 17 6.06 0.07 1.29 2.33 99.6 99.4 39.05 G 7.48 4.79 4.88 2.47 17 6.07 0.07 1.83 2.04 92.5 91.8 34.56 H 7.48 4.67 4.76 2.37 17 6.09 0.07 1.82 2.14 80.2 77.8 55.56

( m 2,m 3) plane, as shown in Figure 5. It is worth noting that for proper operation of the SMB to obtain desired complete separation, the adsorbent and fluid should be regenerated in sections 1 and 4 respectively. At the SMB’s theoretical optimum operating state, the unit has the highest possible productivity and enrichment of products and the lowest desorbent consumption. However, the performance of the SMB under these conditions is not robust and is very sensitive to various kinds of disturbances. Basically, the SMB operation points should be close to the theoretical optimal point in order to achieve a high production rate, yet far away from it within the boundaries of the operating area to assure robustness. Because (S)-propranolol hydrochloride is the desired enantiomer, which is enriched in the raffinate stream, productivity Figure 6. Solubility of propranolol hydrochloride in methanol. 󰋪 based on raffinate rather than on the feed to SMB is more • (R,S)-Propranolol hydrochloride; (S)-propranolol hydro- useful. From Equation 9, raffinate productivity based on chloride. unit CSP volume can be deduced as follows: resolved products were obtained, indicating less robustness of this run. Run H was performed at a higher c R Q c R (m -m ) P = B R = B 3 4 (10) concentration (1.5 mg/mL), which exceeded the concen- Raf (1-ε)VN * c t N C tration range within which the Langmuir isotherm was determined. The raffinate product with the highest To increase raffinate productivity, one can either in- productivity and 80 % purity was obtained. crease the difference of ( ) or decrease the m 3 - m 4 We found that the SMB can separate both enantiomers switching time. in high purity, e.g., in Runs C and D, if the operation Various SMB experiments were run at different opera- points were chosen inside the complete separation region tion conditions. The operating parameters and separation and one does not seek high productivity of the desired performance, such as purity and productivity, are shown product. It is also suggested that the SMB can be in Table 1. operated to achieve partially separated products of Runs C and D were performed in the linear isotherm interest with higher productivity. This process can be range and m3 in run D was increased (i.e., the operation followed by a simple crystallization step to obtain the condition was changed along the operation line toward pure enantiomer. the pure extract region). It was found that the product It is worth noting that some of the experimental results purities in both product streams were nearly 100 %, do not agree well with the theoretical predictions. This which is consistent with the complete separation regions. result could stem from the different chemico-physical The productivity in Run D was slightly higher because parameters of the columns in the SMB unit and the the operation point was moved along the operation line difficulty in controlling the flow rates accurately in the in the direction of increasing the difference of ( ). m 3 - m 4 SMB experimental studies. Runs F and G were performed in the nonlinear range at a concentration of 0.754 mg/mL, while (m 3 - m 2) was Solubility Phase Diagram of Propranolol Hydrochlo- further increased at Run G with the attempt to increase ride System the raffinate productivity. However, only partially For the study of the crystallization from solution, it is 874 Xin Wang, Yue Liu, Hong Wei Yu, and Chi Bun Ching

Table 2. Preferential Crystallization of (R,S)-Propranolol hydrochloride in methanol show an obvious increasing Hydrochloride trend as the temperature increases with the solubility curve of the racemate having a deeper slope than that of Initial Initial Product Seed Yield Run Quantities R:S e.e enantiomer. Because of stability concerns, solubility data (mg) (%) o (mg) Ratio (%) higher than 30 C were not determined. The solid-state properties of propranolol hydrochloride 1 300 50:50 15 0 28.2 with respect to the relationship between the racemic mix- 2 300 65:35 15 78.5 25.5 3 300 70:30 15 90.8 18.6 ture and (S)-enantiomer, have been reported previously 4 300 75:25 15 91.2 16.7 [25]. The shape of a ternary phase diagram can be deduced theoretically from the respective binary phase useful to determine the solid/liquid equilibrium solubility diagrams. Similar to the results of the binary melting diagram of the racemic species of interest. The ternary point phase diagram, the ternary phase diagram shows solubility diagram is helpful to understand the nature of the shape of a typical conglomerate-type compound [25]. the racemic mixture. In fact, the feasibility and yield of However, the two eutectic points are so close to each enantioseparation of a partially resolved mixture is de- other that the exact position of the eutectic points is not pendent on the shape of the phase diagram and the likely to be determined precisely. position of the eutectic points. In consideration of the solvent used in the chromatography separation process, Crystallization of Propranolol Hydrochloride System methanol was selected as the crystallization solvent in Propranolol hydrochloride was identified as a racemic these experiments. The solubility of propranolol hy- compound, although it possesses the phase diagram of drochloride in methanol was measured using a classical conglomerate shape. The eutectic points are close to the visual-polythermal method; the results are shown in racemic mixture, which means that resolution might be Figure 6. In the polythermal method, the solvent and successful by crystallization of a solution at low enan- solute were weighed into a small closed glass vessel in tiomeric excess (e.e). The favorable temperature range to suitable proportions. The contents were heated gently be identified for the crystallization operation is the region with agitation until all of the crystals had dissolved. The within which the solubility of the racemate is much clear solution was first cooled until it nucleated. The higher than that of the enantiomer. Crystallization reso- temperature was then increased slowly (lower than 0.2 lution of (R,S)-propranolol hydrochloride was performed o oC/min) until the last crystal dissolved. At this point the under a constant temperature (15 C) in a 1:2 (V/V) equilibrium saturation temperature has been achieved. methanol and acetone mixture (the mixture of methanol The procedures were repeated by adding solute or and acetone, rather than pure methanol was employed as solvent to obtain the solubility data in the desired tem- the crystallization solvent here because of the suitable perate range. solubility of propranolol hydrochloride). Dissolving a o The ternary solubility phase diagram of (S)- and (R)- certain quantity of racemate in the solvent at 30 C and propranolol hydrochloride in a mixed solvent of metha- then slowly cooling the solution to the desired experi- o nol and acetone was measured using the isothermal mental temperature (15 C), was followed by thoroughly method [25]. For the isothermal method, a sufficient collect the crystals and analyzing the product purity. amount of powder, namely 100 ± 0.1 mg, was dissolved Crystallization results are shown in Table 2. in the solvent (methanol) in a test tube. Saturated Preferential crystallization attempts performed on a solution samples were carefully withdrawn and filtered, racemate solution (Run 1) failed to obtain the enantiomer and the concentrations of the solutes were analyzed using as a pure product, which might be due to the lower lattice the HPLC system and the above-mentioned self-packed energy for the two enantiomers packed orderly in one column. single crystal in a racemic compound system. Starting The solubility data helped us to choose the most suita- from a higher initial purity, for example 70 %, relatively ble conditions for the crystallization operation. In a high purity crystals were obtained. The 91.2 % product binary chiral system, a solubility phase diagram is e.e. (Run 4), rather than pure crystals of one enantiomer, essential for identifying the region for crystallization was due to the difficulty of separating the crystals from resolution. Because of thermodynamic constraints, for the mother liquor. The successful removal of the mother almost 95 % of chiral substances that exist as racemic liquor is crucial for obtaining a higher product e.e compounds, crystallization separation is likely to succeed because the mixture of two enantiomers retained in the only when the initial solution composition is above the mother liquor of the crystal product will work as eutectic point. From Figure 6, we observe that impurities and, thus, decrease the final product purity. In propranolol hydrochloride is highly soluble in methanol addition to the initial solution purity, the separation and the solubility data of both (R,S)- and (S)-propranolol process is controlled by another essential factor: the Chiral Separation of Propranolol Hydrochloride Through an SMB Process Integrated with Crystallization 875 degree of supersaturation. A highly supersaturated so- of methanol and acetone resulted in different product lution most likely leads to deposition of the racemate, purities and yields. Furthermore, crystallization of the even when seeded with a pure enantiomer. On the other concentrated enantioenriched solution obtained from the hand, a lower supersaturation will suffer the difficulty in SMB process, when the composition was above the increasing the product yield. eutectic point composition, led to crystals with high purity being obtained. The integrated process is feasible Crystallization of Propranolol Hydrochloride from SMB and promising for the resolution of racemic-compound- Products forming chiral systems. Although the eutectic points of propranolol hydrochloride are close to the racemic mixture, crystallization of a racemate solution or a solution at a low e.e. failed to Symbols Used provide the pure enantiomer as the product. The SMB Intrinsic affinity coefficients (dimensionless) process on the other hand can be operated to produce the a i optically pure enantiomer, e.g., in Runs C, D, and F at b i Langmuir competitive interference coefficient productivities of 15.9, 17.5, and 39 mg/day. A certain (mL/mg) amount of the solution from SMB Run H was c i Mobile phase concentration based on fluid concentrated and crystallized using the method discussed volume (mg/mL) previously; the final product of (S)-propranolol hydro- F c i Feed concentration (mg/mL) chloride was obtained with 92.5 % e.e.. The integrated k’ Elution capacity (retention) factor of the solute SMB and crystallization process thus, theoretically, (dimensionless) calculated from linear elution could give a productivity of 53.5 mg/day [pure (S)-    chromatography ( ′  ⋅ ) enantiomer], which is higher than that produced by the    SMB process alone. It should be mentioned that upon Ki Equilibrium constant (dimensionless) further increasing the SMB productivity, more crystals ' Frontal capacity factor (dimensionless) calcu- can be obtained from crystallization, which facilitates the K i lated from non-linear frontal chromatography process of washing off the mother liquor. This process could give a higher e.e. product and, thus, increase the ' T i -T 0 ( k i = ) final amount of the desired enantiomer. In a future study, T 0 SMB experiments could be performed at higher feed L Column length (cm) concentration, a larger product flow rate, and a higher m j Fluid phase flow rate over sold phase flow rate enrichment for the desired component. It is worth noting in j section of TCC and SMB unit that the solvent selection is difficult and important. It NC Total number of columns in SMB should provide good separation capacity because it is q i Concentration of component i on stationary used as the mobile phase and desorbent batch chro- phase (mg/mL) matography and SMB separations, respectively. It should * q i Equilibrium concentration of component i on also have suitable solubility for the sample of interest stationary phase (mg/mL) because it is also the crystallization solvent. In the future QF Feed flow-rate fed to SMB process study, the integrated process will be investigated in a Qj Liquid phase flow rate in j section of TCC or normal phase that facilitates the removal of the solvent to SMB process obtain a pure crystalline product. Qs Solid phase flow rate in TCC process t* Switching time in SMB process (min) Conclusions t0R Mean retention time for an unretained compound (min), when compound can enter the Based on the column physicochemical properties and pore system of the stationary phase adsorption equilibrium isotherm determined, the contin- T0 Column hold up time in frontal experiments uous separation of the target enantiomer of propranolol (min) hydrochloride from its racemate mixture was studied Ti Breakthrough time of the waves in frontal using SMB chromatography in both the linear and experiments (min) nonlinear regions. The solubilities of the racemate and u Superficial velocity (cm/s) enantiomer of propranolol hydrochloride in methanol v Interstitial fluid velocity of the mobile phase were determined experimentally at different temperatures. (cm/s) Interstitial fluid velocity of the fluid phase in Crystallization of propranolol hydrochloride from dif- v L ferent initial-composition solutions in the mixed solvent SMB process 876 Xin Wang, Yue Liu, Hong Wei Yu, and Chi Bun Ching

v s Solid velocity in TCC process Morbidelli, Chem. Eng. Sci., 54, 3735 (1999). V Column volume 7. M. Schulte and J. Strube, J. Chromatogr. A, 906, V Volumetric flow rate of the mobile phase 399 (2001). (mL/min) 8. M. Mazzotti, G. Storti, and M. Morbidelli, J. Chromatogr. A, 769, 3 (1997). Greek symbols 9. G. Storti, M. Mazzotti, M. Morbidelli, and S. Carra, ε Bed voidage AIChE J., 39, 471 (1993). εT Total porosity of column 10. D. M. Ruthven and C. B. Ching, Chem. Eng. Sci., 44, 1011 (1989). Subscripts 11. F. Charton and R. M. Nicoud, J. Chromatogr. A, L Liquid phase 702, 97 (1995) S Solid phase 12. B. G. Lim, C. B. Ching, R. B. H. Tan, and S. C. Ng, 1 First eluted component of propranolol hy- Chem. Eng. Sci., 50, 2289 (1995) drochloride racemic mixture (component 1 or 13. H. Lorenz, P. Sheehan, and A. Seidel-Morgenstern, component B) J. Chromatogr. A, 908, 201 (2001) 2 Second eluted component of propranolol 14. M. Amanullah, S. Abel, and M. Mazzotti, Adsorp- hydrochloride racemic mixture (component 2 tion, 11 893 (2005) or component A) 15. G. Strohlein, M. Schulte, and J. Strube, Sep. Sci. Technol., 38 3353 (2003). Superscripts 16. D. M. Ruthven, Principle of Adsorption and SMB Simulated moving bed chromatography Adsorption Processes, John Wiley and Sons, New TCC True counter-current chromatography York (1984). F SMB Feed stream 17. X. Wang and C. B. Ching, Ind. Eng. Chem. Res., 42, R SMB raffinate product 6171 (2003). E SMB extract product 18. S. C. Jen and N. G. Pinto, J. Chromatogr. A, 662, 396 (1994). 19. F. G. Helfferich, J. Chromatogr. A, 768, 169 (1997). References 20. M. Mazzotti, G. Storti, and M. Morbidelli, AICHE J., 40, 1825 (1994). 1. R. Nation, Clin. Pharmacokinet., 27, 249 (1994). 21. M. Mazzotti, G. Storti, and M. Morbidelli, AICHE 2. H. Y. Aboul-Enein, and I. W. Wainer, The Impact of J., 42, 2784 (1996). Stereochemistry on Drug Development and Use, 22. G. Storti, M. Masi, and S. Carra, Chem. Eng. Sci., John Wiley and Sons, New York (1997). 44, 1329 (1989). 3. J. E. Rekoske, AIChE J., 47, 2 (2001). 23. G. Storti, M. Mazzotti, M. Morbidelli, and R. 4. E. Francotte, P. M. Richert, M. Mazzotti, and M. Zaciocchi, Ind. Eng.Chem. Res., 34, 288 (1995). Morbidelli, J. Chromatogr. A., 796, 239 (1998). 24. C. B. Ching and B. G. Lim, J. Chromatogr., 634, 5. S. L. Pais, M. J. Loureiro, and A. E. Rodrigues, 215 (1993). Chem. Eng. Sci., 52, 245 (1997). 25. X. Wang, X. J. Wang, and C. B. Ching, Chirality, 6. M. Pedeferri, G. Zenoni, M. Mazzotti, and M. 14, 318 (2002).