Challenge of the supercritical antisolvent technique SAS to prepare cocrystal-pure powders of naproxen-nicotinamide

Chemical Engineering Journal 303 (2016) 238–251

Contents lists available at ScienceDirect

Chemical Engineering Journal journal homepage: www.elsevier.com/locate/cej

Challenge of the supercritical antisolvent technique SAS to prepare cocrystal-pure powders of naproxen-nicotinamide C. Neurohr a, A. Erriguible b, S. Laugier b, P. Subra-Paternault a,⇑ a b

Université de Bordeaux, CNRS, Bordeaux INP, Laboratoire CBMN-UMR5248, Allée Geoffroy St Hilaire, 33600 Pessac, France Université de Bordeaux, Bordeaux INP, CNRS, I2M-UMR5295, site ENSCBP, 16 avenue Pey-Berland, 33607 Pessac Cedex, France

h i g h l i g h t s
Cocrystals-pure powders was produced using CO2 as antisolvent.
Providing a feed flows combination between 2 and 11.
Whereas a flow ratio of 36 yielded mixtures of co- and homo-crystals.
Propensity of forming homocrystals seems correlated to discrepancy in supersaturation profiles.
Cocrystals were bright plate-like particles up to 1 mm in size.

a r t i c l e

i n f o

Article history: Received 9 March 2016 Received in revised form 20 May 2016 Accepted 25 May 2016 Available online 28 May 2016 Keywords: Cocrystal Supercritical antisolvent Simulation Supersaturation Naproxen:nicotinamide

a b s t r a c t Crystallization assisted by supercritical CO2 has been long ago developed for micronization purposes, but its application to fabricate cocrystals is in its infancy. In this work, the cocrystallization of naproxen and nicotinamide was investigated using CO2 as an antisolvent. In the so-called SAS technique (Supercritical AntiSolvent) a solution of the species dissolved in acetone was injected into a continuous flow of CO2. The conditions of temperature and pressure were constant and set at 37 °C and 10 MPa whereas the CO2 and solution feed rates were varied. For CO2/solution flow ratios between 2 and 11 in wt basis yielding an overall mixture composition between 75 and 93 mol% in CO2, the powders were almost cocrystalspure with a cocrystal content ranging between 94 and 100 wt%. The yield of precipitation ranged between 60 and 70 wt%. The cocrystal had a 2:1 naproxen to nicotinamide molar ratio and exhibited the same hydrogen bond interactions and crystalline structure than cocrystals obtained by conventional or GAS techniques. The cocrystal particles exhibited a thin plate-like morphology and a size distribution ranging between 20 lm and 1 mm. When the flows ratio was increased to 36 (wt basis), heterogeneous powders made of cocrystals and homocrystals of naproxen were produced. The simulation was developed to get insight of the mixing scale and of the supersaturation field within the injection zone and it enabled to construct reasonable assumptions about the appearance of naproxen homocrystals. Since the solution and the CO2 were well mixed at micrometric levels, the evolution of the ternary phase diagram with the CO2/acetone composition was considered. A richer environment in solvent seems then to be favorable conditions for the formation of naproxen-nicotinamide cocrystals. Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction A cocrystal is a molecular arrangement at solid state of two different molecules packed in periodic order through non-covalent bonds like hydrogen bonds, ion pairing, van der Waals forces, hydrophobic interactions, etc. [1]. The formation of a cocrystal enables to modify the physico-chemical properties of a given

⇑ Corresponding author. E-mail address: [email protected] (P. Subra-Paternault). http://dx.doi.org/10.1016/j.cej.2016.05.129 1385-8947/Ó 2016 Elsevier B.V. All rights reserved.

molecule as for instance its stability or its bioavailability in case of a pharmaceutical drug [2–4]. It becomes one strategy to tailor material properties along with salts formation, polymorphism control, amorphization or particles nanosizing [5,6]. The understanding of the interactions between the functional motifs is well developed and the analysis of hierarchical hydrogen propensity is a useful guide to anticipate a successful fabrication of cocrystal. However, their preparation remains fundamentally an experimental science. The cocrystal fabrication is mostly based on conventional crystallization techniques, such as evaporative

239

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

and cooling crystallization [7,8], grinding [7,8] slurrying [9] although more original ways such as continuous melt extrusion [10,11], high pressure homogenization [12] or coupling chemical synthesis and cooling [13] are now considered. Some of these techniques are carried out for screening purposes but for large scale production, cocrystallization from solution as spray-drying [14,15], solutions mixing [16], isothermal suspension conversion [17] or antisolvent precipitation [18] may offer more viable routes of fabrication. In this work, we investigated a new route of cocrystallization based on the use of compressed CO2 as an antisolvent and more specifically using the so-called SAS technique. In its concept, SAS can be considered as a hybrid crystallization technique combining spray drying (injection of the solution in a heated vessel where the evaporation of the solvent takes place) and antisolvent precipitation (addition of a liquid that reduces the solubility of the compound in the initial solvent). In the SAS technique, the solution is sprayed into a vessel filled with compressed CO2. Being miscible with the solvent, the CO2 acts as the solvent eliminator (as temperature does in spray drying) and as an antisolvent for the solute. The use of compressed CO2 to generate particles has been investigated for more than twenty years, in particular for micronization and formulation of drugs [19–21] or for elaboration of hybrid and composite materials [22]. The application of CO2 to control the crystalline structure is more marginal. In such applications, CO2 is used either as a solvent allowing the polymorphic conversion to be realized in a completely dry route [23,24] or as an antisolvent [25–29]. In that latter approach, the greatest advantage of compressed antisolvent compared to liquid antisolvents lies in the one-pot process design that allows for removing the initial solvent and drying the particles almost concurrently with the particles formation. It therefore eliminates the fastidious steps of filtration, washing and drying of the resulting slurries encountered in classical solution crystallization, which is an advantage for the process sustainability in regards of energy inputs, volatile compounds emissions and waste streams production. The use of CO2 to elaborate cocrystals has been only recently investigated and less than twenty drug-coformer systems have been processed so far. Some of them were produced by CO2 as antisolvent [30–34]. Solution atomization assisted by CO2 [30,35–37], contacting the two solids into neat CO2 [30,38] or expanding the {CO2 + dissolved species} solution [39,40] complete the methods panel. Generally speaking, cocrystals obtained by the supercritical routes were found to exhibit the same stoichiometry and crystalline phase than the cocrystals obtained by conventional methods. Comparing specifically cocrystals prepared by CO2 antisolvent to those produced by n-heptane as traditional liquid antisolvent, Ober et al. showed that both techniques did produce the cocrystals but together with an unquantified amount of amorphous material [32] or homocrystals [31]. To authors’ knowledge, naproxen (NPX) cocrystallization with nicotinamide (NCTA) has never been reported by CO2 assisted techniques excepted in our previous work [41,42]. Naproxen is a widely used non-steroidal anti-inflammatory drug of poor solubility in water. Castro et al. [43] reported that naproxen formed cocrystals with the water-soluble nicotinamide at the molar ratio of 2:1 by the Kofler contact method and the liquid-assisted ballmilling process. Later, Ando et al. [44] prepared the cocrystal by solvent drop grinding, clarified the crystal structure and characterized the physicochemical properties. The NPX2:NCTA cocrystals showed improved dissolution properties compared to single NPX or NPX-NCTA physical mixture, and was less hygroscopic than the sodium salt of NPX. Let us mention that the drug-coformer system does not have to be crystalline to exhibit superior properties but can be a co-amorphous blend [45].

In this work, we investigate the cocrystallization of NPX and NCTA by the semi-continuous SAS process. Using previously the batch version of the process (the so-called GAS technique) we have demonstrated that the formation of NPX2:NCTA cocrystals was possible with CO2 [41]. The powders were made at almost 98 wt % by cocrystals. However, particles were large and varying the operating parameters did not help at reducing particles sizes below 100 lm. According to the process simulation [42], the cocrystal particles were formed mostly through secondary nucleation. In such pattern, the primary nucleation only forms enough particles and surface area to trigger secondary nucleation whose rate is so large that this secondary event forms more particles than the first one. This dominance of secondary nucleation is consistent with low levels of supersaturation. In attempts of obtaining smaller particles, the SAS process was investigated because of its superiority at producing micro- or nano-metric particles [27,46,47]. However, by working in a richer CO2 environment, SAS might as well impact the cocrystal fabrication or the powder purity as it has affected the crystalline form or the polymorph purity in case of single component processing [29,48]. The SAS process is a rather complicated technique since it combines the hydrodynamic of the injection and the antisolvent mechanism. Simulation was thus carried out in order to get insight of the mixing scale and of the supersaturation field in the injection zone. The computational work is based on our previous model developed for the precipitation of a single species [49], accounting now for the cocrystal formation through the supersaturation definition. 2. Materials and methods 2.1. Materials S-Naproxen ((+)-(S)-2-(6-methoxynaphtalen-2-yl)propanoic acid, 98%, MW = 230.26 g/mol, NPX) and nicotinamide (pyridine3-carboxamide, 99.5%, MW = 122.12 g/mol, NCTA) were supplied by Sigma–Aldrich (France). Formulae are given in Fig. 1. Carbon dioxide (CO2, 99.5%) was supplied from Air Liquide (France) and acetone (99.5%, Scharlau) was purchased from Atlantic Labo (France). 2.2. SAS description The schematic diagram of the SAS setup is provided in Fig. 2. More details can be found elsewhere [50]. Briefly, the vessel is a cylindrical piece of 320 mL equipped with 8 sapphire windows distributed at top, medium and bottom levels to visualize the injection and the precipitation. A feedback loop comprising electrical mantles and thermocouple controlled the vessel temperature. A LEWA pump (LDB1, LEWA, Germany) introduced the CO2 into the vessel at a flow rate checked downstream by a gas meter whereas the solution was injected by means of a dual piston pump (miniPump, Thermo Separation Product, USA). The set conditions in the vessel were 37 °C and 10.0 MPa. The solution admission line comprised a non-return valve and was pre-conditioned with pure solvent to prevent any backflow of CO2 and solute precipitation within the line. The solution and CO2 were flowing co-currently

Naproxen (NPX)

Nicotinamide (NCTA)

Fig. 1. Structural formulae of naproxen and nicotinamide.

240

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

chiller

heater

P

T

P

P CO2

Solution

F

Crystallization vessel

separator

stop valve

needle valve

non-return valve

F

flowmeter

Fig. 2. Schematic diagram of the SAS setup.

and entered the vessel by two separated inlet ports since the vessel top cannot accommodate a coaxial nozzle of concentric tubings. Experiments were thus carried out with a single capillary of 180 lm i.d. as nozzle design. The length of the capillary was adjusted to fit with the upper sapphire window so the injection zone could be viewed and recorded with a mini-camera (QuickCam, Logitech). The CO2 and solution flow rates were varied from 7 to 59 g/min and 2–13 mL/min, respectively, depending on the foreseen conditions. The injected solutions were made of acetone with NPX and NCTA dissolved in a 2:1 molar ratio at overall concentrations of 40 mg/mL unless noticed. Acetone was selected as the liquid solvent because it can dissolve both components in appreciable concentrations, it was previously employed in the preparation of naproxen:nicotinamid cocrystals [41] and thermodynamic data of the quaternary {NPX + NCTA + CO2 + acetone} system were available [51]. During the injection step, the solution and CO2 merge and if conditions of supersaturation are fulfilled, the precipitation of components occurs. The formed particles were kept in the vessel by a filter plate overtopped by a 0.45 lm membrane. After the injection, the vessel was purged with 1 kg of fresh CO2 to remove traces of solvent. After the vessel depressurization, the particles were harvested, weighed and characterized. 2.3. Powder characterization Optical microscopy was used to document crystals morphology (Olympus SZX12 or BX51TF and camera ColorView U-CMAD3). The powder size distribution was measured by laser diffraction using a Mastersizer 2000 (Malvern) equipped with a low volume circulation unit whose stirring was set at 650 rpm and using silicon oil as dispersing medium. Powders were first dispersed in a small volume of silicon oil to de-agglomerate particles. Infrared spectroscopy (ATR-FTIR) was used as routine analysis to assess the presence of naproxen-nicotinamide cocrystals and identify the homocrystals in excess, when any. FTIR results were confirmed when necessary by X-Rays diffraction. For Infrared spectroscopy, powders were deposited and gently pressed on a

diamond crystal (GoldenGate). Spectra were recorded at room temperature with a NEXUS 870 FTIR ESP spectrometer from Nicolet (Madison, USA) equipped with mercury-cadmium-telluride detector cooled by liquid nitrogen. A hundred of scans were recorded between 800 and 4000 cm1 at a resolution of 4 cm1. Powder X-ray diffraction analysis (XRD) was conducted on a PANalytical XPERT-PRO diffractometer equipped with a CuKa radiation source (k = 15418 Å). Powders were placed on stainless steel holders without preliminary grinding. Diffraction patterns were collected in the 2h range of 4–38° using a step size of 0.0167° and a scanning speed of 0.15°/min. The powder composition was determined by dissolving a known mass in acetonitrile and analyzing the naproxen and nicotinamide content by HPLC (see [41] and supplementary data S1 for detailed conditions). Acetonitrile solutions were prepared in triplicate. The powder composition and the identification of the compound that has precipitated as homocrystal, if any, allowed for quantifying the content in cocrystals of the produced powders thanks to mass balance equations ([41] and supplementary data). 2.4. SAS modeling The model is based on our earlier works dealing with single species precipitation [49] and cocrystallization in the GAS technique [42]. Equations are detailed in Appendix B. The main aspects are summarized hereafter. Cocrystal solubility and supersaturation expressions were derived from Rodriguez-Hornedo’s work [52], whose basis is to represent cocrystallization as an equilibrium reaction in which a binary cocrystal AaBb dissociates in solution to A and B according to:

Aa Bbsolid $ aAsolution þ bBsolution

ð1Þ

By writing the equilibrium constant of this reaction and assuming that activities could be approximated by concentrations, the cocrystal solubility can be described by the solubility product:

K sp ¼ ½Aa ½Bb

ð2Þ

241

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

The supersaturation for a cocrystal takes the general form of [52]:


S¼

P

1 PC ti i = K sp m

m i i

ð3Þ

where PC i i is the product of concentration of cocrystal components in the supersaturated solution when the activity coefficients are unity, and mi the stoichiometric number of the cocrystal components (i.e. a or b). For naproxen and nicotinamide system, a and b are equal to 2 and 1 since the cocrystal has a 2:1 stoichiometry (NPX2:NCTA). In the case of a precipitation induced by antisolvent addition, the concentration of each component at equilibrium evolves with the solvent/antisolvent composition. The quaternary data of the {NPX + NCTA + CO2 + acetone} system were taken from Revelli et al. [51] and further correlated by an empirical equation. For the fluid dynamics topic, the equation of continuity and conservation of momentum was described by the Navier-Stokes equation taking into account turbulent phenomena by using a LES (Large Eddy Simulation) turbulence model. The Peng Robinson equation of state with quadratic mixing rules accounted for the density of the acetone–CO2 mixtures. The species continuity equations were expressed by taking into account the diffusion according to the Fick’s law and considering the cocrystal as a unique solute. To simulate the particles formation, a general form of population balance was selected assuming a growth rate independent of particle size and neglecting agglomeration and breakage. The non-spherical shape of the particles produced experimentally was taken into consideration through the volume shape factor. The population balance equation was solved with the standard method of moments assuming that the particle size distribution (PSD) fits a log-normal distribution. The particle generation was considered to be due to both primary and secondary nucleations and the main mechanism for growth was assumed to be diffusion [42]. Nucleation and growth kinetics were written for the cocrystal only, implying thus that for solutions in which NPX and NCT were dissolved in stoichiometric ratio, the two components did not precipitate independently. Parameters of the secondary nucleation rate were taken from our previous study [42] whereas the unknown solid-fluid interfacial tension in the primary nucleation rate equation was obtained by fitting simulated particles size with experimental results. The set of equations was numerically solved to estimate the velocities, composition and moments of the PSD at each time step.

The numerical tool is the home-made CFD code Thetis developed at the I2M/Trefle Department. The simulation aimed at focusing on the injector zone to visualize the time- and space-scales of the solution mixing and the supersaturation field, so the dimensions of the computational domain were of 5 cm  3 cm  3 cm in regards of the experimental merging zone between CO2 and the solution. The mesh, constituted by more than 10 millions of nodes, was refined near the injector to improve the accuracy of the micromixing description. Calculations were performed in 3D by MPI parallel programming on 96 processors. More details about the numerical methods can be found in Erriguible et al. [49]. 3. Results 3.1. Influence of solution/CO2 flow rates The variation of solution and CO2 feed rates has been reported to influence the particle sizes [53] and the polymorphism [29,25]. Moreover, cocrystallization equilibria are sensitive to solvent properties [54,55]. Hence, by varying the composition of the fluid mixture one can expect to influence the characteristics of the produced powders. In this set of experiments, the flow rates of CO2 and solution were changed so that their ratio ranged from 2.5 (run 1) to 11 (run 7) in weight basis and the CO2:solution mixture varied from 75 to 93 mol% in CO2. Table 1 summarizes the process parameters

Table 1 Effect of the solution/CO2 flow rates combination on cocrystallisation by SAS, yielding compositions of 75–93 mol% in CO2. Solution of 2:1 NPX:NCTA molar ratio. T = 37 °C, P = 10.0 ± 0.5 MPa. FCO2, Fsol: CO2 flow rate and solution feed rate; vsol: solution velocity in the capillary nozzle of 180 lm i.d., XCO2: composition of the CO2:solvent mixture calculated from flow rates; Yield: precipitation yield; Excess: percentile of homocrystals in products. Solution

Processing

CNPX CNCTA FCO2

Run Run Run Run Run Run Run Run

1 2 3 4 5 6 7 8

Product

Fsol

vsol

XCO2

6 2 6 13 2 7 2 6

4.1 1.2 3.8 8.8 1.2 4.5 1.2 4.0

75 86 86 86 91 93 93 93

wt%

wt%

Yield Cocrystal Excess content g/min mL/min m/s mol% wt% wt% wt%

4.0 4.0 4.0 4.0 4.0 4.0 4.0 5.1

1.1 1.1 1.1 1.1 1.1 1.1 1.2 1.3

12 7 21 52 10 56 17 51

63 65 70 60 66 63 67 66

98 94 100 100 100 99 100 95

2 6 0 0 0 1 0 5

Fig. 3. Optical microscopy of NPX2:NCTA cocrystals obtained in run 7 (scale bar of 1 mm) and particle size distributions of CO2-processed powders (runs 1–8).

242

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

Fig. 4. XRD patterns of powders produced in conditions of run 3 and run 9 compared to patterns of naproxen and nicotinamide.

and some of the powders characteristics. The specific effect of one flow rate can be analyzed through runs 2–5–7 (fixed solution flow rate of 2 mL/min and increased FCO2 from 7 to 17 g/min), runs 1–3– 6 (fixed solution flow rate of 6.5 ± 0.5 mL/min, increased FCO2 from 12 to 56 g/min) or runs 5–1 and 6–4 for increasing the solution flow rate at fixed CO2. Before discussing results, it is important to analyze the process in terms of hydrodynamics by considering the CO2 and solution velocities. For CO2, although the range of the flow rate variation was large (7–56 g/min), there was no strong modification of the flow regime in the vessel since the Reynolds number was still below 500 when the velocity changed from 0.1 mm/s (FCO2 of 7 g/min) to 1 mm/s (FCO2 of 56 g/min). For the solution and because of the small injector diameter, the solution velocities at the capillary exit varied significantly with the applied flow rate, from 1.2– 8.8 m/s. Nevertheless, based on simulation [49] or experimental results [56,57], this range does not affect significantly the jet breakup regime when SAS is carried out in conditions above the fluid critical pressure (7.5–8 MPa at 35–45 °C [58]) since the mixing takes place in a single-phase flow. The increase of velocity tends to increase the length of the jet core that remains rich in solvent, so the area of precipitation is possibly displaced upstream but this does not impact notably the overall supersaturation behavior [49]. Within the ranges of velocities investigated in this work, it is then reasonable to assume that neither the jet behavior nor the hydrodynamics in the vessel were modified when the conditions were changed. In the range of flows investigated and by processing solutions of 32 and 8 mg/mL of NPX and NCTA respectively (run 1–7), the produced powders were all made of large and bright particles with a plate-like morphology (Fig. 3) similar to cocrystal particles produced by the GAS technique [41]. The XRD pattern (Fig. 4) evidences the formation of the cocrystal by the presence of characteristic peaks at 5.4°, 12.3°, 17.3° that are distinct from the specific peaks of pure naproxen and pure nicotinamide. The strong anisotropy of the produced plate-like particles is responsible for the difference of the peaks intensity compared to that in published patterns [44,41] because of preferred orientations effects. The ATR-

FTIR analyses (Fig. 5) confirmed the hydrogen bonding scheme with broad bands at 2525 cm1 and 1982 cm1 that are specific of the O–Hcarboxylic acid. . .Naromatic hydrogen bond between NPX and NCTA. The distinctive band of the C@O functional group visible at 1700 cm1 is due to formation of the acid-amide heterosynthon. Furthermore, the anti-symmetric mas and symmetric ms stretching of the NH2 group in NCTA are now respectively at 3367 cm1 and 3194 cm1 instead of 3350 cm1 and 3148 cm1 for single NCTA. The particles morphology, XRD patterns and ATR-FTIR spectra thus confirm that the naproxen-nicotinamide cocrystals obtained by this SAS process correspond well with the cocrystals produced in the batch version [41]. However, we cannot say if the cocrystal powders contained a small amount of amorphous material as it was reported in the case of itraconazole cocrystals produced by CO2 antisolvent [31,32]. For these seven experiments, the precipitation yield ranged between 60 and 70 wt%, and the cocrystal content was above 94 wt%. The small amounts of homocrystals were calculated from HPLC analysis and were attributed to NPX, although not detected by ATR-FTIR. Taking into consideration inherent fluctuations of operating parameters, particles harvesting and standard deviations of the quantification, the variations of yield and cocrystal content were not deemed to be significant. Therefore within this range of CO2-solvent flow rates, neither the CO2 nor the solution velocities appear to affect significantly the cocrystal precipitation and neither does it affect the particles size distributions (Fig. 3). The particle sizes of the various powders were indeed identical in that they were mostly mono-modal and centered around 300–400 lm. In attempts of reducing particle size, the overall concentration of the solution was increased to 50 mg/mL (run 8), still keeping the NPX:NCTA molar ratio to 2:1. This run produced a powder with characteristics of cocrystal content, precipitation yield and particle size distribution (Fig. 6) that are similar to those of powders previously produced. The obtaining of large particles might come from low supersaturation values. In order to evaluate the supersaturation profile created by the mixing between the solution and CO2, a computational work was carried out by simulating conditions of run 6, i.e. at a

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

243

Fig. 5. ATR-FTIR spectra and hydrogen bondings in the cocrystal. (a) NPX, (b) NCTA, (c) NPX2:NCTA cocrystal; labels 1–8 are referring to the run number, as provided in Table 1.

Fig. 6. Optical microscopy of particles produced in run 6 (scale bar of 1 mm) and comparison between experimental and calculated size distribution.

244

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

Fig. 7. Instantaneous fields in the meridian plane of the vessel. (a) Mass fraction of acetone, (b) supersaturation ratio, (c) concentration of particles, in volume. T = 37 °C, P = 10 MPa, vinj = 4.5 m/s. Length of the domain: 5 cm.

Table 2 Cocrystallisation by SAS in conditions of flows yielding composition of 98 mol% in CO2. Solution of 2:1 NPX:NCTA molar ratio. T = 37 °C, P = 10.0 ± 0.5 MPa. Solution

Run 9 Run 10

Processing

Product

CNPX wt%

CNCTA wt%

FCO2 g/min

Fsol mL/min

vsol

4.04 ± 0.02 5.07 ± 0.01

1.08 ± 0.01 1.29 ± 0.01

54 ± 2 55 ± 1

2 ± 0.1 2.2 ± 0.3

solution flow rate of 7 mL/min and a mass fraction of cocrystal NPX2:NCTA of 0.056 g/gsolvent. As mentioned earlier, the fitting of the calculated sizes distribution to the experimental profile allowed for optimizing the simulation. Fig. 6 shows the comparison between the experimental and calculated PSD. The simulation fits the data over a large range of sizes but tends to underestimate the smaller sizes of the distribution because of the log-normal assumption in the modeling. Moreover, although the particle shape factor that appears in the cocrystal mass balance was calculated for an elongated particle whose length and width were assessed from optical microscopy images, the experimental population was composed of non-uniform particles (Fig. 6). These heterogeneities in both dimensions are likely to contribute as well to the discrepancies of the PSD fitting. The mixing behavior of the injected solution and ambient CO2 is illustrated in Fig. 7 through the instantaneous spatial field of acetone mass fraction. A video is given as supplementary information to provide a real-time view of the solution injection. In the monophasic conditions generated by the operating pressure and temperature, the solution and CO2 merge without forming droplets. Close to the capillary exit (<0.5 cm), the inlet core of the jet is mostly composed of acetone (Fig. 7a), and, as the merging proceeds in time and in space through convection and diffusion, the acetone content decreases in favor of CO2. The depletion in acetone creates conditions of supersaturation because of the decrease of the solute solubility with increasing antisolvent concentration [51]. Therefore, supersaturation areas are localized where the acetone content is low, i.e. above the 0.5 cm from the nozzle tip (Fig. 7b). From 0.5 to 1.3 cm upstream the tip, the supersaturation level is maximum (1.3–1.5) but the concentration of solids created in this area is low (<0.025 m3/m3, Fig. 7c), suggesting that this area is the center of nucleation events rather than of growth ones. Upstream the jet (above 1.3 cm), and because the image given is an instantaneous image capture of the on-going crystallization course, part of the solute has been consumed by earlier precipitation events. This area appears thus to be less supersaturated than

m/s

XCO2 mol%

Yield wt%

Cocrystal content wt%

NPX excess wt%

1.2 ± 0.1 1.4 ± 0.2

98 ± 1 98 ± 1

55 ± 2 62 ± 5

8–25 4–86

92–75 96–14

before because of the lower concentration of solute that remains in solution. Conversely, the concentration of solids is the highest, comprised between 0.025 and 0.050 m3/m3. In the conditions investigated here, the maximum attainable supersaturation ratio was below 1.5, which is a very low level that can explain the large sizes of the produced particles. This result is consistent with previous results on mefenamic acid, a solute that exhibited a supersaturation ratio in this range and precipitated as needle shaped particles of several hundreds of micron in length [59]. In order to reach higher CO2 content in the mixture to promote higher supersaturation level, two runs were carried out at 2 mL/min and 52 g/min for solution and CO2 feed rates, respectively, yielding a CO2/solution flow ratio of 36 (weight basis) and an overall composition of 98 mol% in CO2 (Table 2). NPX and NCTA concentrations were of 4.0 wt% and 1.1 wt% as before (run 9) or increased to 5.1 wt % and 1.3 wt% respectively (run 10). Results were quite different from previously. The obtained powders resembled small cotton balls made of aggregated particles of plate-like crystals and needles (Fig. 8). The needles were analogous to the particles obtained when processing pure NPX, whilst the plates were similar to particles obtained in run 1–8. The particle size distributions depicted that powder heterogeneity with the apparition of three modes centered near 15 lm, 100 lm and 300 lm. The population around 300 lm coincided with the plate-like population produced in runs 1–8, whereas the second mode likely corresponded to the needles population and the 15 lm-population came from smaller plate-like or needles particles. The suspicion of a powder made by a mixture of NPX particles and NPX:NCTA cocrystals was confirmed by ATR-FTIR analyses (Fig. 9), XRD pattern (Fig. 4) and HPLC quantification. Besides the cocrystal signature (double band at 2526 cm1 and broad band around 1982 cm1 for OHcarboxylic acid. . .Naromatic cocrystal bond, and shifted NH2 stretching bands at 3367 cm1 and 3193 cm1), additional bands specific to NPX were evidenced, such as the C@O stretching at 1723 cm1 and the band at 2940 cm1 associated with the CH3 stretching of

245

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

Absorbance (au)

Fig. 8. Characteristics of powders produced in conditions of 9 and 10, that proved to be physical mixtures of NPX2:NCTA cocrystals + NPX crystals. Left: optical microscopy of run 9 (scale bar of 1 mm). Right: volume particle size distributions of runs 9 and 10.

3500

3300

3100

2900

2700

2500

2300

2100

1900

1700

1500

Wavelength (cm-1) Fig. 9. ATR-FTIR spectra of powders produced in conditions of run 9 and 10, compared to spectra of (a) single NPX, (b) single NCTA, (c) NPX2:NCTA cocrystal.

the propanoic acid chain. The XRD pattern showed also additional peaks at 2h of 6.6°, 12.6° and 13.3° corresponding to the crystalline naproxen. None of the nicotinamide characteristic bands in FTIR spectra or peaks in XRD pattern were observed, suggesting that powders were only made of cocrystals and NPX homocrystals. The powder composition was confirmed by HPLC quantification that evidenced a large excess of NPX compared to the 2:1 molar ratio obtained in case of pure cocrystals. Experiments were repeated for reproducibility estimation. While the global precipitation yield was in the same range and while the massive NPX homocrystals precipitation was confirmed, the cocrystal content was very different from one run to the other,

indicating a very difficult control of the cocrystallization in such condition. This behavior suggests that despite a favorable ratio of 2:1 in the processed solution, local conditions might be created allowing for the NPX to precipitate preferably as a single species instead of coprecipitating with NCTA. The injection process was thus simulated in the conditions of run 9 and compared to run 6. From the hydrodynamic standpoint, there was no difference in the jet formation or in its dimensions since the velocities were not thoroughly modified. The mixing of the solution and CO2 was still achieved, as illustrated by the acetone evolution given in Fig. 10. The calculation grid is provided as well to detail the scale at which the simulation was performed. With a grid mesh of few

246

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

Fig. 10. Instantaneous field of acetone (expressed as mass fraction) in simulated conditions of run 9 (v = 1.2 m/s) and details of the calculation grid. The acetone mass fraction represented by the color scale is ranging from 0.05 to 0.85.

micrometers and a time step for calculations of microsecond, the high spatial resolution provided by the simulation enables to visualize precisely the quality of the mixing between the two fluids. Hence, it is reasonable to assume that the appearance of uncocrystallized naproxen alongside the cocrystals is not due to the quality of the mixing but rather to thermodynamics through the solubility levels of components. For a fine quantitative analysis, the supersaturation profiles were described as a volume distribution of supersaturation levels (Fig. 11). A striking result is the difference of the three species profiles when the solution flow rate increased from 2 mL/min (run 9) to 7 mL/min (run 6). In conditions of run 9, the maximum supersaturation level for the cocrystal stood in the range of 2.2 instead of the 1.6 value obtained in run 6. However, these higher levels contributed only to a half of an overall volume that was significantly smaller than before, so they were not numerous enough to induce a significant effect on the cocrystals size. In run 9, naproxen exhibited now high levels of saturation distributed over numerous and large volumes, as several 108 m3 volumes were occupied by supersaturation levels up to 2 with a significant contribution of levels up to 3.4. It was assumed that volumes below 1014 m3 were negligible since they corresponded to a spatial area of 1 lm  100 lm  100 lm. For the cocrystal, the volumes occupied by supersaturation levels higher than 1 and that were not negligible, i.e. higher than 1014 m3, were rather limited in numbers and it was even worse in case of nicotinamide. In those conditions of flow, the three species exhibited thus very distinct distribution profiles. The configuration of higher saturation levels spread out over larger volumes observed for naproxen seem then to be conducive to its single crystallization rather than the cocrystal formation. For the conditions of run 6, the distribution profiles of the three species were more alike, with volumes of 108 m3 occupied by all of them. Experimental results suggested that in such situation where all species exhibit the same profiles of large volumes associated to low levels of supersaturation, the cocrystal formation is favored. Conversely, more discrepancies in the supersaturation distributions favor homoprecipitation. A plausible hypothesis is that low levels of saturation keep species in interactions as a complex that further crystallizes as so when conditions of precipitation are met. 3.2. Influence of NPX/NCTA ratio in solution In conditions of run 9 and 10, only naproxen was found as the homoprecipitate. It suggested that the NCTA coformer was still soluble in the CO2-acetone mixture and was continuously flushed out with the mixture. The concentration of NCTA in the processed solution was thus increased to 2.1 wt%, keeping the NPX content unchanged (Run 11, Table 3). Large and bright plates were recovered and their sizes were similar to those obtained in the first set

Fig. 11. Volume distribution of supersaturation levels in the whole domain. T = 37 °C, P = 10 MPa. (a) Conditions of run 9: FCO2 = 56 g/min, vinj = 1.2 m/s (98 mol% CO2); (b) conditions of run 6: FCO2 = 56 g/min, vinj = 4.5 m/s (93 mol% CO2).

of experiments (runs 1–8) as a mono-modal distribution centered near 300 lm (Fig. 12). ATR-FTIR characterized the powder as NPX2: NCTA cocrystal and did not evidence NPX or NCTA homocrystals. Quantification by HPLC allowed for concluding that the powder was cocrystal pure. Increasing the NCTA in solution enabled thus to increase the purity in cocrystal of the powder and provided a better control of the cocrystal precipitation as viewed by the much better reproducibility. The precipitation yield was increased as well from 55% (run 9) to 63% (run 11). It is worth noting that, thanks to the process, the powder was cocrystal pure despite the processing of a non stoichiometric solution, i.e. a solution in which the molar ratio of NPX and NCTA was set to 2:2 instead of the 2:1 ratio encountered in the cocrystal. An extra run was carried out increasing further the NCTA ratio to 4 (in conditions of 94 mol% CO2). Besides cocrystals, the produced powder was now containing NCTA homocrystals. Offering more NCTA to NPX that had the propensity to precipitate as single species is thus an interesting option for promoting the cocrystal recovery.

4. Discussion In the applied conditions of 10 MPa, 37 °C and for the concentrations tested (32 and 8 mg/ml for NPX and NCTA, respectively),

247

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251 Table 3 Processing a 2:2 NPX:NCTA solution in conditions of flow yielding composition of 98 mol% in CO2. T = 37 °C, P = 10.0 ± 0.5 MPa. Solution

Run 11

Processing

Product

CNPX wt%

CNCTA wt%

FCO2 g/min

Fsol mL/min

vsol m/s

Yield wt%

Cocrystal content wt%

NPX excess wt%

4.02 ± 0.01

2.14 ± 0.01

58 ± 2

2 ± 0.1

1.3 ± 0.1

63 ± 2

99 ± 1

±1

Fig. 12. Size distribution of NPX2:NCTA cocrystals produced in run 11.

the cocrystallization of the two species was effective provided that the CO2 and solution flow rates ratio ((wt basis) stayed between 2 and 11 yielding an overall mixture of composition between 75 and 93 mol% in CO2. The obtained powders were made of NPX2:NCTA cocrystals by at least 94 wt%. The sizes of particles were centered in the 300–400 lm range whatever the conditions. Moreover, the size distribution profiles were not significantly altered by the variation of the CO2 and solution flow rates underlining that the crystallization kinetics (nucleation and growth rates) were rather unaffected by hydrodynamic conditions. The numerical study pointed out very low supersaturation levels, directly related to the species solubility in CO2. Indeed, with solubility in the range of 1.0  105 mol/mol and 2.5  104 mol/mol for NPX and NCTA respectively at 37 °C and 10 MPa [51], a maximum supersaturation ratio of 1.5 was attained. The antisolvent effect of CO2 was then rather weak for this system. Although the cocrystals growth was not affected since the PSDs related to cocrystals populations were always in the same range, a very different crystallization behavior was observed when the ratio of CO2 to solution was increased up to 36 by decreasing the solution flow rate (conditions of run 9 and 10, overall composition of 98 mol% in CO2). Contrary to the previous results, small amounts of cocrystals were obtained while massive crystallization of single NPX occurred yielding physical mixtures of NPX2:NCTA cocrystals and NPX homocrystals. Furthermore, the reproducibility was critical for these experiments. These results suggest modified phase equilibria as function of the CO2 content and a possible contribution of composition heterogeneities not only in the mixing zone close to the nozzle but within the whole vessel as well. Indeed, cocrystal solid-liquid equilibria are strongly influenced by the nature of the solvent [54,55,60]. The SAS technique is based on the rapid modification of solvent properties to induce solute precipitation. Therefore the environment of the solutes changes from pure acetone, where they were completely solubilized, to

Fig. 13. Schematic ternary phase diagrams at 10 MPa and 37 °C for (a) 93 mol% of CO2, (b) 93 mol% of CO2 equilibria in blue and 98 mol% of CO2 equilibria in red. Dashed line represents the trajectory line for a 2:1 NPX:NCTA initial acetone solution. Black cross stands for global composition of the system at crystallization.

CO2-enriched mixtures where they are forced out into the solid state. The phase equilibria encountered in this antisolvent process of cocrystallization is a quaternary system of NPX/NCTA/CO2/acetone. Phase equilibria of quaternary system can be represented by prismatic diagram. The z-axis of the prism describes the variation of the fluid phase composition, i.e. the various CO2 + acetone mixtures, whereas the sections are triangles whose two apices are the solutes and the third one is a given fluid. The section of the prism corresponds thus to a classical ternary phase diagram that delimits the existence zones of the solid cocrystal, single

248

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

species and mixtures of them. Fig. 13a displays a schematic ternary diagram of the NPX/NCTA/CO2/acetone system, considering that the fluid phase is a given {CO2 + acetone} mixture. Equilibrium lines are hypothetical since we do not have thermodynamic equilibrium data for setting accurately the boundaries of the various zones. The cocrystal zone is drawn from the cocrystallization results obtained in this study, i.e. the obtaining of a powder made of only cocrystals or a powder made of cocrystals plus NPX homocrystals. Nevertheless, the NPX and NCTA solubility ratio and their trends with an increasing content in CO2 are representative of Revelli’s data [51]. For instance, in a mixture of 93 mol% of CO2, NPX and NCTA exhibit the same range of solubility whereas at 98 mol%, NCTA is more soluble. Since the acetone solution prepared with a 2:1 molar ratio yielded cocrystal powders, the trajectory line from the fluid apex to the 2:1 NPX:NCTA solid would only cross the NPX2:NCTA cocrystal zone. Thermodynamic conditions would therefore be fulfilled for the cocrystal-only precipitation. On Fig. 13b, the first diagram is overlaid with a ternary diagram for a fluid phase richer in CO2. Since solubilities are decreasing with increasing amounts of CO2, the existing zones of the solid phases enlarge while the miscibility zone diminishes. In this new diagram, the trajectory line from the fluid apex to the 2:1 NPX:NCTA solid crosses the single NPX zone and the {cocrystal + NPX} zone. Whatever the fluid quantity at crystallization, i.e. wherever crystallization starts on the trajectory line, single NPX crystallization is then thermodynamically possible. Hence, based on the evolution of boundaries with the CO2/solvent composition, the precipitation in conditions of 93 mol% and 98 mol% of CO2 will occur in two different existence zones yielding a cocrystal-pure powder in one case or a NPX + cocrystal physical mixture in the second case. Previous thermodynamic considerations do not take into account kinetic pathways, i.e. scales at which the CO2 and the solution merge to create suitable conditions for precipitation. In other words, assuming that coexistence zones of cocrystal and homocrystal are dependent of the CO2/solvent composition, any local heterogeneity in the fluid composition could induce the precipitation of a different phase. SAS is a fast process in which the injected solution, by being miscible with the surrounding CO2, experiences disintegration and mixing on scales within centimeters and few seconds. Nucleation and growth are initiated within this merging zone (numerical results, Figs. 7 and 10), but the very low attained supersaturation levels questioned about the completeness of the crystallization events above that mixing zone, especially if heterogeneities in composition arise. 5. Conclusions In this work, the fabrication of cocrystals using CO2 as an antisolvent and acetone as solvent was attempted in the so-called SAS technique that consists in an injection of the solution into a continuous flow of compressed CO2. The experimental investigations were focused on the effect of CO2 and solution feed rates, and more specifically on their ratio that was varied between 2 and 36 (weight basis). Main results are summarized below:
For flows ratios between 2 and 11, yielding an overall mixture of composition between 75 and 93 mol% in CO2, the powders were almost cocrystals-pure, made of thin plate-like particles whose sizes distribution ranged between 20 lm and 1 mm.
For the ratio of 36 (FCO2 of 55 g/min, Fsol of 2 mL/min), a substantial apparition of NPX crystals arose leading to heterogeneous powders made of cocrystals and homocrystals, whose purity in cocrystals was poorly reproducible.


Offering more NCTA coformer to NPX that had the propensity to homo-precipitate (i.e. processing 2:2 solution instead of 2:1) allowed for shifting to a cocrystal-pure powder and provided a much better reproducibility.
Thanks to the simulation, an innovative representation of supersaturation mapping through histograms of fluid volumes/saturation levels was developed aiming at comparing levels for the three species that might precipitate, i.e. the drug, the coformer and the cocrystal. Confronted to experimental results, it was suggested that conditions that create discrepancies of supersaturation levels and volumes between the three species favored the homocrystallization (run 9), whereas more similar supersaturation distribution profiles tend to favor cocrystallization (run 6).
The jet disintegration and the efficiency of the mixing were not modified by the explored conditions of flow rates. Therefore the obtaining of cocrystals versus {cocrystals + homocrystals} mixtures is mostly tributary of coexistence phase boundaries, which, by analogy of their sensitivity to solvent or solvent composition in conventional crystallization, evolve with the fluid composition. Moreover, and because of the continuous running, the composition of the fluid phase is not homogenous neither in space neither in time, and this may certainly contribute to the difficulty at controlling the purity in cocrystals of the produced powders.
Whatever, the results demonstrate that a richer environment in solvent provides better conditions for the cocrystal to be produced. This conclusion is consistent with the results of GAS processing for which cocrystal-pure powders were produced [41]. The SAS process is a complex hybrid technique combining spray drying in regards of the injection step and antisolvent precipitation. The fabrication of cocrystals is another complex task by itself, so coming up with an interpretation that takes into account all phenomena is not easy. By describing the mixing between the solution and the ambient CO2 and by mapping the supersaturation levels thanks to existing solubility data, the simulation allowed for constructing reasonable assumptions about the mechanisms that might favor the co-crystallization rather than the appearance of homocrystals alongside the cocrystals. Indeed, since the solution and the CO2 were well mixed at micrometric levels, the appearance of homocrystals seems to be driven by thermodynamics and more specifically by the higher solubility of the coformer that induces a less symmetrical coexistence phase diagram. Changing acetone for another solvent that will keep the diagram more symmetrical would be an interesting option. The drawback of the dependency of the phase diagram with the fluid composition is in the process robustness since any heterogeneity of composition in the vessel might induce the precipitation of the unwanted homocrystals. It would be interesting to run experiments in various SAS set-ups that differ by the vessel dimensions, flows capacities, nozzle size or modus operandi to enlarge the investigated conditions and further improve the understanding of this challenging control of the crystalline phase at CO2-rich conditions. Acknowledgements The financial support of the french ANR Agency (Project ANR11-BS09-41, 2012–2016) is greatly acknowledged. C. Neurohr especially thanks the Agency for her PhD grant. M. Marchivie is thanked for the XRD analysis and for the hydrogen bonding scheme.

249

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

Assuming that the distribution is log-normal, the following relationship is used to build the simulated particle size distribution with the calculated moments:

Appendix A. List of abbreviations in text

Techniques ATR-FTIR GAS HPLC PSD SAS XRD Compounds NPX NCTA NPX2:NCTA Model Ksp S C mi i

attenuated total reflectance Fourier transform infrared spectroscopy gaseous antisolvent high performance liquid chromatography particle size distribution (in vol%) supercritical antisolvent X-rays diffraction

Appendix B. SAS modeling B.1. The population balance equation

ðB1Þ

In this equation n represents the number density function which depends on internal coordinates (particle size L) and external coordinates (space coordinates X (x, y, z)). V and q are respectively the velocity and the density of the fluid and G is the growth rate of the particles. The population balance equation is solved with the standard method of moments assuming that the particle size distribution follows a log-normal distribution. In our case, only the four first moments are solved, the system of equations to be solved remains:

dmj  mj Þ ¼ qð0 j B0 þ jGmj1 Þ for j ¼ 0; 1; 2; 3 þ r  ðqu dt

ðB2Þ

where mj is the jth moment defined by:

Z

mj ¼

1

nðL; X; tÞL j dL

A with ln ðrÞ ¼ ln pffiffiffiffiffiffiffi @ L lnðrÞ 2p 2ln2 ðrÞ 2

  m0 m2 m1 ðB4Þ

B.2. Nucleation and growth parameters

ðB5Þ

It is assumed in the following that only cocrystals are generated by the crystallization process, which was evidenced experimentally. Therefore, nucleation and growth kinetics are written for only one species, the cocrystal, and not for three species, naproxen, nicotinamide and their cocrystal. In the previous equation, Bpn stands for the primary nucleation rate and is calculated by the classical crystallization theory:



7=3 Bpn ¼ 1:5Db C cc sat SN a

rffiffiffiffiffiffiffiffi

c

kb T

V m exp 

!
3 16p c V 2m 3 kb T ln2 ðSÞ

ðB3Þ

0

where mj is the jth moment of the distribution. B0 and G represent the nucleation and the growth rates. (m0 = total particle number; m1 = total particle length; m2 = total particle area; m3 = total particle volume).

ðB6Þ

with Db being the solute diffusion coefficient estimated by Wilke and Chang correlation, C cc sat the cocrystal concentration at saturation, Vm the cocrystal molecular volume, kb the Boltzmann constant, Na the Avogadro number and c he solid-fluid interfacial tension. S is the supersaturation, driving force of the crystallization which is detailed in the Section 2.4. The secondary nucleation rate is commonly represented for industrial crystallization by a phenomenological law that depends on the supersaturation S and on the concentration of solid in suspension Cs: n Bsn ¼ ASN C m s ðS  1Þ

The general form of the population balance assuming a growth rate of particles independent of particle size and neglecting agglomeration and breakage takes the following form:

q

 1

2 L L

B0 ¼ Bpn þ Bsn

solubility product (in (mol/L)a+b) supersaturation concentration of the cocrystal component i (mol/L) to the power of its stoichiometric number in the cocrystal (mi) large Eddy Simulation turbulence model computational fluid dynamics

@nðL; X; tÞ @nðL; X; tÞG  nðL; X; tÞÞ þ q þ r  ðqu ¼0 @t @L

NðLÞ ¼

ln

Particle formation is supposed to be due to both primary and secondary nucleations, so that the overall nucleation rate can be defined as the sum of the two contributions:

naproxen nicotinamide cocrystal of naproxen and niotinamide of 2:1 stoichiometry

LES CFD Data in tables CNPX, CNCTA concentration of naproxen or nicotinamide in (wt%) the prepared acetone solution (wt%) FCO2 CO2 flow rate (g/min) Fsol solution flow rate (mL/min) XCO2 composition of the CO2 + solvent mixture (mol %) calculated from flow rates vsol solution velocity in the capillary nozzle (m/s)

q

0 1

ðB7Þ

ASN determines the order of magnitude of the secondary nucleation. The parameters ASN, m and n was determined in our previous study in the case of batch antisolvent synthesis of the co-crystal system naproxen/nicotinamide. The growth rate, assuming that the main mechanism for growth is diffusion, is calculated by:

G ¼ kg C cc sat ðS  1Þ

ðB8Þ

where kg represents the mass transfer coefficient deduced of the local Sherwood number. This one is estimated by the Froessling equation. B.3. Hydrodynamic and thermodynamical contributions In the conditions investigated in this work, the pressure is above the critical pressure of the CO2-acetone mixture so the two fluids are fully miscible and merge as a monophasic flow. The flow is considered as incompressible since the variation of density is only due to the variation of the mixture composition. The equations of continuity and momentum conservation are described by the Navier-Stokes model:

¼0 ru   

 @u   ru  ¼ rp þ qg þ r  l þ lt ru  þ rt u  q þu @t

ðB9Þ ðB10Þ

250

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251

in which p is the pressure, q the density of the fluid, g the gravity  the filtered velocity vector, l the dynamic viscosity, t the time and u vector. The turbulence is modeled through a deterministic Large Eddy Simulation approach and lt the turbulent viscosity is calculated with a structural mixed scale model. The Peng Robinson equation of state (PR-EOS) with quadratic mixing rules is chosen to calculate the density of the fluid. The species continuity equations are expressed by taking into account the diffusion of species according to the Fick’s law:

q

@xa   qðDa þ Dt Þrxa Þ ¼ 0 þ r  ðqxa u @t

ðB11Þ

q

@xcc   qðDcc þ Dt Þrxcc Þ ¼ 3qp kv Gm2 þ r  ðqxcc u @t

ðB12Þ

with xa and xcc the mass fractions of solvent and solute, respectively, Da the diffusion coefficient of the solvent in CO2, Dcc the diffusion coefficient of the solute in the solvent-CO2 mixture, and Dt is the turbulent diffusion coefficient. qp is the density of the solid cocrystal and kv the volume shape factor of the particles that is based on the morphology of produced cocrystal. Nomenclature B0 Bpn Bsn C D g G kb kg L mk n Na p S T t u Vm x Greek letters

l c q

nucleation rate (1/(m3 s)) primary nucleation rate (1/(m3 s)) secondary nucleation rate (1/(m3 s)) concentration (mol/m3) diffusion coefficient (m2/s) gravity vector (m/s2) growth rate (m/s) Boltzmann constant (m2 kg s1 K1) mass transfer coefficient (m4/(mol s)) length of the particles (m) kth order moment (mk/m3) particles density function (1/m3) Avogadro number (1/mol) pressure (Pa) supersaturation temperature (K) time (s) velocity vector (m/s) molecular volume (m3) mass fraction viscosity (Pa s) surface tension (N/m) density (kg/m3)

Appendix C. Supplementary data Supplementary data associated with this article (HPLC quantification of naproxen and nicotinamide in produced powders and a video showing the distribution of acetone upon the injection of the solution) can be found in the online version, at http://dx.doi. org/10.1016/j.cej.2016.05.129. References [1] G. Desiraju, Crystal engineering: from molecule to crystal, J. Am. Chem. Soc. 135 (2013) 9952–9967. [2] S. Chow, M. Chen, L. Shi, A. Chow, C. Sun, Simultaneously improving the mechanical properties, dissolution performance, and hygroscopicity of ibuprofen and flurbiprofen by cocrystallization with nicotinamide, Pharm. Res. 29 (2012) 1854–1865. [3] D. McNamara, S. Childs, J. Giordano, A. Iarriccio, J. Cassidy, M. Shet, R. Mannion, E. O’Donnell, A. Park, Use of a glutaric acid cocrystal to improve oral bioavailability of a low solubility API, Pharm. Res. 23 (2006) 1888–1897.

[4] N. Schultheiss, A. Newman, Pharmaceutical cocrystals and their physicochemical properties, Cryst. Growth Des. 9 (2009) 2950–2967. [5] A. Newman, Specialized solid form screening techniques, Org. Process Res. Dev. 17 (2013) 457–471. [6] S. Morissette, O. Almarsson, M. Peterson, J. Remenara, M. Read, A. Lemmo, S. Ellis, M. Cima, C. Gardner, High-throughput crystallization: polymorphs, salts, co-crystals and solvates of pharmaceutical solids, Adv. Drug Deliv. Rev. 56 (2004) 275–300. [7] N. Tumanova, N. Tumanov, K. Robeyns, Y. Filichunk, J. Wouters, T. Leyssens, Structural insight into cocrystallization with zwitterionic co-formers: cocrystals of S-naproxen, Cryst. Eng. Comm. 16 (2014) 8185–8196. [8] S. Skovsgaard, A. Bond, Co-crystallisation of benzoic acid derivatives with N-containing bases in solution and by mechanical grinding: stoichiometric variants, polymorphism and twinning, Cryst. Eng. Comm 11 (2009) 444– 453. [9] T. Kojima, S. Tsutsumi, K. Yamamoto, Y. Ikeda, T. Moriwaki, High-throughput cocrystal slurry screening by use of in situ Raman microscopy and multi-well plate, Int. J. Pharm. 399 (2010) 52–59. [10] K. Boksa, A. Otte, R. Pinal, Matrix-assisted cocrystallization (MAC) simultaneous production and formulation of pharmaceutical cocrystals by hot-melt extrusion, J. Pharm. Sci. 103 (2014) 2904–2910. [11] D. Daurio, K. Nagapudi, L. Li, P. Quan, F.-A. Nunez, Application of twin screw extrusion to the manufacture of cocrystals: scale-up of AMG 517–sorbic acid cocrystal production, Faraday Discuss. 170 (2014) 235–249. [12] M. Fernandez-Ronco, J. Kluge, M. Mazzotti, High pressure homogenization as a novel approach for the preparation of co-crystal, Cryst. Growth Des. 13 (2013) 2013–2024. [13] H. Lee, T. Lee, Direct co-crystal assembly from synthesis to co-crystallization, CrystEngComm 17 (2015) 9002–9006. [14] S. Patil, S. Modi, A. Bansal, Generation of 1:1 carbamazepine: nicotinamide cocrystals by spray drying, Eur. J. Pharm. Sci. 62 (2014) 251–257. [15] A. Alhalaweh, S. Velaga, Formation of cocrystals from stoichiometric solutions of incongruently saturating systems by spray drying, Cryst. Growth Des. 10 (2010) 3302–3305. [16] S. Kudo, H. Takiyama, Production method of carbamazepine/saccharin cocrystal particles by using two solution mixing based on the ternary phase diagram, J. Cryst. Growth 392 (2014) 87–91. [17] D. Croker, Å. Rasmuson, Isothermal suspension conversion as a route to cocrystal production: one-pot scalable synthesis, Org. Process Res. Dev. 18 (2014) 941–946. [18] I.-C. Wang, M.-J. Lee, S.-J. Sim, W.-S. Kim, N.-H. Chun, G. Choi, Anti-solvent cocrystallization of carbamazepine and saccharin, Int. J. Pharm. 450 (2013) 311– 322. [19] E. Reverchon, I. De Marco, Mechanisms controlling supercritical antisolvent precipitate morphology, Chem. Eng. J. 169 (2011) 358–370. [20] N. Esfandiari, Production of micro and nano particles of pharmaceutical by supercritical carbon dioxide, J. Supercrit. Fluids 100 (2015) 129–141. [21] P. Subra-Paternault, C. Domingo, ScCO2 techniques for surface modification of micro-and nano-particles, in: V. Mittal (Ed.), Surface Modification of Nanoparticle and Natural Fiber Fillers, Wiley-VCH, Weinheim, Allemagne, 2015, pp. 109–150. [22] P. Subra-Paternault, Methods for particles production: antisolvent techniques, in: C. Domingo, P. Subra Paternault (Eds.), Supercritical Fluid Nanotechnology: Advances and Applications in Composites and Hybrids Nanomaterials, Pan Stanford Publishing Pte. Ltd., Singapore, 2015, pp. 103–129. [23] R. Bettini, L. Bonassi, V. Castoro, A. Rossi, L. Zema, A. Gazzaniga, F. Giordano, Solubility and conversion of carbamazepine polymorphs in supercritical carbon dioxide, Eur. J. Pharm. Sci. 13 (2001) 281–286. [24] Y. Tozuka, D. Kawada, T. Oguchi, K. Yamamoto, Supercritical carbon dioxide treatment as a method for polymorph preparation of deoxycholic acid, Int. J. Pharm. 263 (2003) 45–50. [25] Á. Martín, K. Scholle, F. Mattea, D. Meterc, M.-J. Cocero, Production of polymorphs of ibuprofen sodium by supercritical antisolvent (SAS) precipitation, Cryst. Growth Des. 9 (2009) 2504–2511. [26] M. Rossmann, A. Braeuer, A. Leipertz, E. Schluecker, Manipulating the size, the morphology and the polymorphism of acetaminophen using supercritical antisolvent (SAS) precipitation, J. Supercrit. Fluids 82 (2013) 230–237. [27] C. Roy, D. Vrel, A. Vega-González, P. Jestin, S. Laugier, P. Subra-Paternault, Effect of CO2-antisolvent techniques on size distribution and crystal lattice of theophylline, J. Supercrit. Fluids 57 (2011) 267–277. [28] Y.-M. Chen, M. Tang, Y.-P. Chen, Recrystallization and micronization of sulfathiazole by applying the supercritical antisolvent technology, Chem. Eng. J. 165 (2010) 358–364. [29] A. Kordikowski, T. Shekunov, P. York, Polymorph control of sulfathiazole in supercritical CO2, Pharm. Res. 18 (2001) 682–688. [30] L. Padrela, M. Rodrigues, S. Velaga, H. Matos, E. de Azevedo, Formation of indomethacin–saccharin cocrystals using supercritical fluid technology, Eur. J. Pharm. Sci. 38 (2009) 9–17. [31] C. Ober, R. Gupta, Formation of itraconazole-succinic acid cocrystals by gas antisolvent cocrystallization, AAPS Pharm. Sci. Tech. 13 (2012) 1396–1406. [32] C. Ober, S. Montgomery, R. Gupta, Formation of itraconazole/L-malic acid cocrystals by gas antisolvent cocrystallization, Powder Technol. 236 (2013) 122–131. [33] A. Shikhar, M. Bommana, S. Gupta, E. Squillante, Formulation development of carbamazepine–nicotinamide co-crystals complexed with c-cyclodextrin using supercritical fluid process, J. Supercrit. Fluids 55 (2011) 1070–1078.

C. Neurohr et al. / Chemical Engineering Journal 303 (2016) 238–251 [34] I. Cuadra, A. Cabañas, J. Cheda, F. Martínez-Casado, C. Pando, Pharmaceutical co-crystals of the anti-inflammatory drug diflunisal and nicotinamide obtained using supercritical CO2 as an antisolvent, J. CO2 Util. 13 (2016) 29–37. [35] J. Tiago, L. Padrela, M. Rodrigues, H. Matos, A. Almeida, E. de Azevedo, Singlestep co-crystallization and lipid dispersion by supercritical enhanced atomization, Cryst. Growth Des. 13 (2013) 4940–4947. [36] L. Padrela, M. Rodrigues, S. Velaga, A. Fernandes, H. Matos, E. de Azevedo, Screening for pharmaceutical cocrystals using the supercritical fluid enhanced atomization process, J. Supercrit. Fluids 53 (2010) 156–164. [37] L. Padrela, M. Rodrigues, J. Tiago, S. Velaga, H. Ma tos, E. de Azevedo, Tuning physicochemical properties of theophylline by cocrystallization using the supercritical fluid enhanced atomization technique, J. Supercrit. Fluids 86 (2014) 129–136. [38] L. Padrela, M. Rodrigues, J. Tiago, S. Velaga, H. Ma tos, E. de Azevedo, Insight into the mechanisms of cocrystallization of pharmaceuticals in supercritical solvents, Cryst. Growth Des. 15 (2015) 3175–3181. [39] C. Vemavarapu, M. Mollan, T. Needham, Crystal doping aided by rapid expansion of supercritical solutions, AAPS PharmSciTech 3 (4) (2002) 17–31. [40] K. Müllers, M. Paisana, M. Wahl, Simultaneous formation and micronization of pharmaceutical cocrystals by rapid expansion of supercritical solutions (RESS), Pharm. Res. 32 (2015) 702–713. [41] C. Neurohr, A.-L. Revelli, P. Billot, M. Marchivie, S. Lecomte, S. Laugier, S. Massip, P. Subra-Paternault, Naproxen–nicotinamide cocrystals produced by CO2 antisolvent, J. Supercrit. Fluids 83 (2013) 78–85. [42] A. Erriguible, C. Neurohr, A.-L. Revelli, S. Laugier, G. Fevotte, P. SubraPaternault, Cocrystallization induced by compressed CO2 as antisolvent: simulation of a batch process for the estimation of nucleation and growth parameters, J. Supercrit. Fluids 98 (2015) 194–203. [43] R. Castro, J. Ribeiro, T. Maria, R. Silva, C. Yuste-Vivas, J. Canotilho, M. Eusébio, Naproxen cocrystals with pyridinecarboxamide isomers, Cryst. Growth Des. 11 (2011) 5396–5404. [44] S. Ando, J. Kikuchi, Y. Fujimura, Y. Ida, K. Higashi, K. Moribe, K. Yamamoto, Physicochemical characterization and structural evaluation of a specific 2:1 cocrystal of naproxen–nicotinamide, J. Pharm. Sci. 101 (2012) 3214–3221. [45] K. Jensen, K. Löbmann, T. Rades, H. Grohganz, Improving co-amorphous drug formulations by the addition of the highly water soluble amino acid, proline, Pharmaceutics 6 (2014) 416–435. [46] Y.-P. Chang, M. Tang, Y.-P. Chen, Micronization of sulfamethoxazole using the supercritical anti-solvent process, J. Mater. Sci. 43 (2008) 2328–2335. [47] F. Fusaro, M. Hanchen, M. Mazzotti, G. Muhrer, B. Subramanian, Dense gas antisolvent precipitation: a comparative investigation of the GAS and PCA techniques, Ind. Eng. Chem. Res. 44 (2005) 1502–1509.

251

[48] G. Weber Brun, Á. Martín, E. Cassel, R. FigueiróVargas, M.-J. Cocero, Crystallization of caffeine by supercritical antisolvent (SAS) process: analysis of process parameters and control of polymorphism, Cryst. Growth Des. 12 (2012) 1943–1951. [49] A. Erriguible, T. Fadli, P. Subra-Paternault, A complete 3D simulation of a crystallization process induced by supercritical CO2 to predict particle size, J. Comput. Chem. Eng. 52 (2013) 1–9. [50] A. Vega-González, C. Domingo, C. Elvira, P. Subra, Precipitation of PMMA/PCL blends using supercritical carbon dioxide, J. Appl. Polym. Sci. 91 (2004) 2422– 2426. [51] A.-L. Revelli, S. Laugier, A. Erriguible, P. Subra-Paternault, High-pressure solubility of naproxen, nicotinamide and their mixture in acetone with supercritical CO2 as an anti-solvent, Fluid Phase Equilib. 373 (2014) 29–33. [52] N. Rodriguez-Hornedo, S. Nehm, K. Seefeldt, Y. Pagàn-Torres, C. Falkiewicz, Reaction crystallization of pharmaceutical molecular complexes, Mol. Pharm. 3 (2006) 362–367. [53] O. Boutin, Lodeling of griseofulvin recrystallization conducted in a supercritical antisolvent process, Cryst. Growth Des. 9 (10) (2009) 4438–4444. [54] R. Chiarella, R. Davey, M. Peterson, Making co-crystals: the utility of ternary phase diagrams, Cryst. Growth Des. 7 (2007) 1223–1226. [55] A. Alhalaweh, A. Sokolowski, N. Rodríguez-Hornedo, S. Velaga, Solubility behavior and solution chemistry of indomethacin cocrystals in organic solvents, Cryst. Growth Des. 11 (2011) 3923–3929. [56] T. Petit-Gas, O. Boutin, I. Raspo, E. Badens, Role of hydrodynamics in supercritical antisolvent processes, J. Supercrit. Fluids 51 (2009) 248– 255. [57] E. Reverchon, E. Torino, S. Dowy, A. Braeuer, A. Leipertz, Interactions of phase equilibria, jet fluid dynamics and mass transfer during supercritical antisolvent micronization, Chem. Eng. J. 156 (2010) 446–458. [58] C.-Y. Day, C. Chang, C.-Y. Chen, Phase equilibrium of ethanol + CO2 and acetone + CO2 at elevated pressures, J. Chem. Eng. Data 41 (1996) 839–843. [59] A. Erriguible, S. Laugier, M. Late, P. Subra-Paternault, Effect of pressure and non isothermal injection on recrystallization by CO2 antisolvent: solubility measurements, simulation of mixing, and experiments, J. Supercrit. Fluids 76 (2013) 115–125. [60] Y. Tong, Z. Wang, L. Dang, H. Wei, Solid–liquid phase equilibrium and ternary phase diagrams of ethenzamide-saccharin cocrystals in different solvents, Fluid Phase Equilib. (2016), http://dx.doi.org/10.1016/j.fluid.2016.02.047.