Artemisinin (purity>99%) was obtained from Xiang Xi Holley Pharmaceutical Co. Ltd. in China. The impurity, quercetin hydrate (purity≥95%) was purchased from Sigma-Aldrich, Germany. Acetonitrile, acetone, and methanol were of HPLC grade from VWR Prolabo, Denmark. Ethanol (96% purity) was obtained from Kemetyl A/S, Denmark. Water was purified using an SG Ultra Clear Basic UV system (Holm & Halby, Germany).

Dionex Ultimate 3000 high performance liquid chromatography (HPLC) system equipped with a photodiode array (PDA) detector (Agilent, Waldbronn, Germany) was used to determine the concentration of artemisinin and quercetin in the solution during anti-solvent crystallization experiments. It was also used to determine the purity of crystals obtained at the end of the experiment. Separations were carried out on a reverse phase Eclipse XDB-C18 column (5 μm particle size, 150 mm × 4.6 mm i.d., Agilent, USA). The column temperature was maintained at 35°C, and the mobile phase consisted of water and acetonitrile with a gradient elution: starting with 50% acetonitrile and ending with 100% acetonitrile at a flow rate of 0.5 mL/min. Sample injection volume was 10 μL and UV wavelength of 254 nm was used for detection. The chromatograms of standard artemisinin and quercetin analyzed by this method are shown in Fig. 1. It is evident from the Fig. 1 that retention times of quercetin and artemisinin are 3.61 and 12.2 minutes, respectively. The height of peaks indicates that quercetin is highly sensitive to UV wavelength of 254 nm while artemisinin is less sensitive. The system was calibrated by measuring the peak areas for known concentrations of artemisinin and quercetin.

Fig.1

	Figure Option ViewDownloadNew WindowDownload As Powerpoint Slide
	Fig.1 HPLC chromatograms of standard artemisinin and quercetin obtained by using the described method

Qualitative analysis of crystals obtained during solubility measurement and anti-solvent crystallization experiments was performed with MultiRam FT-Raman spectrometer from Bruker Optics GmbH. The spectrometer was equipped with liquid nitrogen cooled Germanium detector. Laser wavelength of 1064 nm and laser power of 500 mW was used for the measurements. The samples were analyzed with 50 scans and 4 cm^–1 of resolution.

The crystalline materials were characterized on a Siemens D5000 Crystalloflex diffractometer with CuKα radiation at 35 mA and 40 kV. The scanned 2 θ region was from 5° to 30° with a step size of 0.02° and a counting time of 1 sec per step. XRPD was used to determine the crystal form of artemisinin obtained during anti-solvent crystallization as artemisinin is known to form two different polymorphs [ 11].

The solubility of pure artemisinin in organic solvents or their binary mixtures with water was measured at room temperature. The organic solvents used include acetonitrile, acetone, methanol, and ethanol. The binary mixtures consisted of water and 90, 80, 70, 60, 50, 40, 30, 20, and 10 percent organic solvent, respectively. Solvent (10 mL) and excess artemisinin were added to a 25 mL flask. The resulting liquid-solid suspension was then kept to attain equilibrium under mixing in a water bath at room temperature for 5 h. The mixture was then filtered and the solubility was determined by gravimetric method. Solid phase was analyzed with FT-Raman spectroscopy to ascertain the form of crystals.

Anti-solvent crystallization experiments were designed with the help of fundamental process data, in this case the solubility of artemisinin in pure organic solvents and their binary mixtures of varying composition with water. Experimental setup used to perform experiments consisted of 1l capacity crystallizer equipped with stirrer, sampling probe and constant rate addition funnel. The experiments were started with slightly undersaturated solution of artemisinin in pure organic solvent at room temperature. Anti-solvent (water) was added to the clear solution of artemisinin at constant rate of 15 mL/min at a constant stirring speed of 300 rpm. The volume of water added was determined on the basis of the solubility of artemisinin in organic solvent-water mixtures of varying composition. Accordingly, enough volume of water was added to reach the organic solvent-water composition at a point when artemisinin was less than 0.1 mg/mL. During the addition of water, samples were taken from the solution at regular time of intervals. The samples were then analyzed with the HPLC method described in section 2.2 to determine the concentration of artemisinin. The experiments were conducted separately for artemisinin without impurity and artemisinin with impurity as a solute while keeping other operating parameters constant. Two different concentrations of impurity, 10% and 50% of its solubility in respective organic solvent were used. Crystals obtained at the end of the experiments were analyzed with FT-Raman spectroscopy and XRPD to ascertain the polymorph of artemisinin.

The solubility of artemisinin was measured in pure organic solvents and their binary mixtures of varying composition with water. The solubility of quercetin was measured only in pure organic solvents. Table 1 shows the solubility of artemisinin in organic solvents and the binary mixtures at room temperature and the solubility of quercetin is shown in Table 2. It is clear from Table 1 that artemisinin has the highest solubility in acetonitrile followed by acetone, methanol and ethanol. The solubility ratio of artemisinin to quercetin was highest in acetonitrile as compared to other solvents, so acetonitrile may serve as an attractive solvent to purify artemisinin from quercetin. It is also evident that the solubility of artemisinin decreases as the content of water in the mixture increase. Apparently, artemisinin solubility becomes less than 0.1 mg/mL in acetonitrile-water (20∶80), acetone-water (30∶ 70), methanol-water (30∶70), and ethanol-water (40∶60). This trend of decreasing solubility with increasing water content indicates the potential of water as an anti-solvent. The solid phase analysis by FT-Raman spectroscopy and XRPD confirmed that the same orthorhombic polymorph of artemisinin used for solubility measurements crystallized from all solvents and their binary mixtures with water.

Tab.1

Tab.1

Tab.1 Solubility of artemisinin in four different organic solvents and their binary mixtures with water at room temperature Solvent composition /vol-% Solubility /(mg∙mL–1) Methanol Acetone Acetonitrile Ethanol 100 37 175.6 206.5 24.9 90 22.1 112 187.8 18 80 11.6 60 133.6 12 70 6 30 67 3.7 60 4.2 12 36.4 2.5 50 1.1 4 17 0.9 40 1 1 6.4 <0.1 30 <0.1 <0.1 1.9 <0.1 20 <0.1 <0.1 <0.1 <0.1 10 <0.1 <0.1 <0.1 <0.1

Tab.1 Solubility of artemisinin in four different organic solvents and their binary mixtures with water at room temperature

Tab.2

Tab.2

Tab.2 Solubility of quercetin in organic solvents at room temperature Solvent Solubility /(mg∙mL–1) Methanol 5 Acetone 67 Acetonitrile 1 Ethanol 13

Tab.2 Solubility of quercetin in organic solvents at room temperature

Anti-solvent crystallization experiments of artemisinin without impurity from four organic solvents were designed based on the solubility of artemisinin in these solvents and the binary mixture of varying composition with water. Accordingly, the experiments were started with a solution (100 mL) of artemisinin in methanol, ethanol, acetonitrile and acetone with concentrations 27 mg/mL, 18 mg/mL, 180 mg/mL and 150 mg/mL, respectively. Experiments were started with a slightly undersaturated solution to make sure that a clear solution is obtained. The volumes of anti-solvent (water) added to the artemisinin solution in methanol, ethanol, acetonitrile and acetone were 233, 150, 400 and 233 mL respectively. Figure 2 shows exemplary de-supersaturation profile of artemisinin observed during anti-solvent crystallization from methanol along with the equilibrium solubility profile of artemisinin in methanol-water mixture. It is clear from Fig. 2 that as adding the anti-solvent addition, the crystallization profile travels from undersaturated region to supersaturation region crossing equilibrium solubility curve and eventually crystallization occurs when the metastable zone limit is approached. The width of the metastable zone is system specific and depends on the addition rate, mixing, primary and/or secondary nucleation rate, and growth rate, as well as the amount and type of impurities present in solution [ 12]. After the crystallization commences, the system tends to reach equilibrium as it is clear from the Fig. 2 that the de-supersaturation profiles and equilibrium profile of artemisinin are gradually close to each other. Similar de-supersaturation profiles were observed for other solvents too. The yield and purity of artemisinin crystals obtained are shown in Table 3. Exemplary FT-Raman spectrum and XRPD pattern of the crystals obtained from methanol solution are shown in Fig. 3 and Fig. 4 respectively. XRPD pattern of the samples are compared with the calculated XRPD patterns of two polymorphs of artemisinin, a triclinic form (QINGHAOSU01) and orthorhombic form (QINGHAOSU10) obtained from Cambridge Crystal Structure database. It is clear from Fig. 4 that the XRPD pattern of the samples resembles to that of the orthorhombic form of artemisinin, which is a thermodynamically stable form of artemisinin. XRPD results confirm that the same orthorhombic form of artemisinin has been obtained for all anti-solvent crystallization experiments.

Tab.3

Tab.3

Tab.3 Yield and purity of artemisinin along with volume of anti-solvent added during anti-solvent crystallization of artemisinin from acetone. Sample point Anti-solvent added /mL Yield of artemisinin /wt-% Purity of artemisinin /% Aa) Bb) Aa) Bb) Aa) Bb) 1 46.27 50.78 85.3 57.91 99.15 86.37 2 92.70 64.23 94.7 83.33 97.86 86.23 3 130.8 77.68 97.7 88.77 97.84 89.78 4 183.93 232.6 97.9 98.67 97.67 73.45 a) Quercetin concentration is 10% of its solubility in acetone (6.7 mg/mL);b) Quercetin concentration is 50% of its solubility in acetone (33.5 mg/mL).

Tab.3 Yield and purity of artemisinin along with volume of anti-solvent added during anti-solvent crystallization of artemisinin from acetone.

Fig.2

	Figure Option ViewDownloadNew WindowDownload As Powerpoint Slide
	Fig.2 De-supersaturation profiles of artemisinin along with solubility of artemisinin in methanol and methanol-water mixtures

In many cases, an impurity in the solution changes the course of crystallization and affects the yield and purity of the final product [ 13]. Desired compound from natural products is often crystallized from a solution containing many other compounds. In case of recovery of artemisinin from Artemisia annua, it is crystallized from a complex crude extract containing many compounds, one of which is quercetin. Anti-solvent crystallization experiments with addition of quercetin as an impurity were performed while keeping other operating parameters constant. Figure 2 shows exemplary de-supersaturation profile of artemisinin contaminated with two different concentrations of quercetin in methanol along with the equilibrium solubility profile of artemisinin in methanol-water mixture. It is obvious from Fig. 2 that the de-supersaturation profile of artemisinin contaminated with impurity closely resembles to that of artemisinin without impurity. This illustrates that the presence of impurity does not affect the nucleation and crystal growth kinetics of artemisinin. Also it is clear from the figure that the concentration of artemisinin in the mother liquor at the end of the experiment is almost the same for all experiments with and without quercetin, which may suggest that the presence of quercetin has no effect on the solubility of artemisinin in the solvent mixtures that has been studied in the present work. As shown in Table 3, however, the purity of artemisinin crystals decreased significantly in case of higher concentration of quercetin in the solution. This result implies that with the higher concentration of quercetin, the addition of water also generates supersaturation of quercetin, and thus quercetin crystallized out together with artemisinin crystals. It is confirmed from the FT-Raman spectrum of sample with higher concentration of impurity (50% of its solubility), as shown in Fig. 3. It is evident from the figure that the FT-Raman spectrum shows the peaks from both, the standard artemisinin and quercetin. XRPD patterns of the samples containing impurity are shown in Fig. 4. It is clear from the figure that the same orthorhom-bic polymorph of artemisinin (QINGHAOSU10) is obtained, indicating that there is no effect of quercetin on the polymorphism of artemisinin. Similar results were obtained for anti-solvent crystallization of artemisinin from other solvents.

Fig.3

	Figure Option ViewDownloadNew WindowDownload As Powerpoint Slide
	Fig.3 FT-Raman spectra of crystals obtained from anti-solvent crystallization of artemisinin from methanol: 1 quercetin hydrate; 2 artemisinin with quercetin (50% of solubility); 3 artemisinin with quercetin (10% of solubility); 4 artemisinin without quercetin; 5 artemisinin raw material

Fig.4

	Figure Option ViewDownloadNew WindowDownload As Powerpoint Slide
	Fig.4 XRPD patterns of crystals obtained from anti-solvent crystallization of artemisinin from methnol: 1 quercetin hydrate; 2 artemisinin with quercetin (50% of solubility); 3 artemisinin with quercetin (10% of solubility); 4 artemisinin without impurity; 5 artemisinin triclinic polymorph (QINGHAOSU10); 6 artemisinin orthorhombic polymorph (QINGHAOSU01)

When anti-solvent crystallization is used to isolate a target compound from a mixture with an impurity, very often there is a compromise between the purity of the product and the yield of the process, and this is especially true when the target compound and the impurity have similar solubility behaviour in the solvent and the anti-solvent. Since the solubility of the target compound decreases with increasing anti-solvent concentration, the yield of the process will increase with increasing feeding of the anti-solvent. However, the increased feeding of the anti-solvent might also generate high level of supersaturation of the impurity, which would subsequently accelerate the crystallization of the impurity. Consequently, the purity of the final target product would be reduced. Figure 5 depicts the de-supersaturation profiles of artemisinin together with quercetin for both concentrations of quercetin used in the anti-solvent crystallization of artemisinin from acetone. The points 1, 2, 3 and 4 in Fig. 5 represent the sampling points and the total volume of anti-solvent present in the solution at these points is shown in Table 4. It is clear from the Fig. 5(a) that there is only a minor change in the concentration of quercetin from point 2 to point 4, meaning that crystallization of quercetin is slow in this region. However, it is then rapidly crystallized after point 4. This behaviour might be present in the experiment with quercetin concentration of 50% of its solubility in acetone (Fig. 5(b)), but it is not visible due to lack of sampling in that region. This observation implies that the crystallization rate of quercetin is strongly dependent on the supersaturation and there might be a critical supersaturation level above which the crystallization of quercetin is remarkably accelerated. This crystallization behaviour suggests that it is possible to obtain optimized product purity and yield by controlling the concentration of the anti-solvent at the end of the process. Yield and purity of artemisinin crystals calculated from the concentrations of artemisinin and quercetin at sampling points 1-4 is shown in Table 4 while the same obtained at the end of experiment is shown in Table 3. Thus, the purity of artemisinin crystals obtained in the experiment with quercetin concentration of 10% of its solubility in acetone increased from 92.81% to 97.84% by stopping the batch at point 4 shown in Fig. 5(a). In case of the experiment with quercetin concentration of 50% of its solubility in acetone, the purity of artemisinin crystals increased from 73.45% to 89.78% by stopping the batch at point 3 shown in Fig. 5(b). In this case, however, the yield of artemisinin decreased from 98.67 wt-% to 88.77 wt-%.

Fig.5

	Figure Option ViewDownloadNew WindowDownload As Powerpoint Slide
	Fig.5 De-supersaturation profiles of artemisinin and quercetin during anti-solvent crystallization from acetone: (a) quercetin concentration of 10% of its solubility in acetone; (b) quercetin concentration of 50% of its solubility in acetone

Tab.4

Tab.4

Tab.4 Yield and purity of artemisinin crystals obtained in anti-solvent crystallization of artemisinin with and without quercetin from different organic solvents at room temperature Experiment Solvent Quercertin concentration /(mg∙mL–1) Artemisinin product purity /% Artemisinin product yield wt-% 1 Methanol 0 92.24 97.29 2 0.5a 91.27 97.27 3 2.5b 76.18 98.33 4 Acetone 0 98.95 98.92 5 6.7a 92.81 98.67 6 33.5b 73.45 98.67 7 Acetonitrile 0 99.61 98.69 8 0.1a 99.25 98.27 9 0.5b 90.25 98.47 10 Ethanol 0 90.94 83.99 11 1.3a 91.89 87.63 12 6.5b 84.71 86.73 a) Quercetin concentration is 10% of its solubility in respective solvent;b) quercetin concentration is 50% of its solubility in respective solvent

Tab.4 Yield and purity of artemisinin crystals obtained in anti-solvent crystallization of artemisinin with and without quercetin from different organic solvents at room temperature

This supression of the crystallization of quercetin was not observed for other solvents studied. Crystallization of artemisinin and quercetin occurred simulatneously from other solvents. An exemplary de-supersaturation profile of artemisinin and quercetin for methanol solvent is shown in Fig. 6, which clearly indicates that quercetin is crystallizing throughout the experiment along with artemisinin.

Fig.6

	Figure Option ViewDownloadNew WindowDownload As Powerpoint Slide
	Fig.6 De-supersaturation profiles of artemisinin and quercetin during anti-solvent crystallization from methanol: (a) quercetin concentration of 10% of its solubility in methanol; (b) quercetin concentration of 50% of its solubility in methanol