**3.1 Parameters influence on hydrodynamic**

Mass transfer is one of the key factors that control the particle size in the SAS process. This is influenced by both the spray hydrodynamics of the organic solution and the thermodynamic properties of the supercritical fluid phase.

In the last years, the hydrodynamic of the SAS process has been the subject of several papers. Most authors face up to this problem considering that the jet of organic solvent behaves like a liquid jet injected into a gas, allowing to apply the classic theory of jet breakup. This theory could be applied successfully at subcritical conditions, below the mixture critical point solvent-CO2, where there is surface tension. The mixture critical point denotes the limit of the two-phase region of the phase diagram. In other words, this is the point at which an infinitesimal change in some thermodynamic variable such as temperature or pressure will lead to separation of the mixture into two distinct phases.

However, in supercritical conditions, above the critical point of the mixture organic solvent and CO2, it is not possible to distinguish droplets nor interfaces between the liquid solution and the phase of dense CO2 gas. Surface tension decreases to zero in a shorter distance than characteristic break-up lengths. Thus, the jet spreads forming a gaseous plume and will be characterized by the degree of turbulence associated with the vortices produced in the SC CO2 (Chehroudi et al., 2002; Kerst et al., 2000; Reverchon et al., 2010). Lengsfeld et al. were the first group that investigated fluid dynamics of the SAS process, studying the evolution and disappearance of the liquid surface tension of fluids injected in supercritical carbon dioxide. They concluded that a gas-like jet is formed after the jet break-up (Lengsfeld et al., 2000). In this way, Kerst et al. determined the boundaries between the different modes and they noted a strong interdependence between mass transfer and fluid dynamics (Kerst et al., 2000).

In the SAS related literature there is a general agreement about the flow regimes observable when a liquid is injected in a vessel. The way in which the liquid solution is dispersed in the CO2 when the operating conditions are below the mixture critical point (MCP), which is strongly influenced by the operating pressure and the flow rate of liquid solution at fixed temperature, can be described according to one of the following four regimes: 1) the dripping mode, which requires lower flow speed so that drops can detach themselves from the orifice, 2) the Rayleigh break up regime, which is characterized by a rupture of the jet in the form of monodisperse droplets, 3) the sine wave break up regime, in which a helicoidal oscillation of the jet occurs, leading to its rupture into droplets with a polydisperse distribution, and 4) atomization, in which the jet is smooth when it leaves the orifice, until it reaches the zone of highly chaotic rupture where a cone of atomized liquid is formed.

Hydrodynamics Influence on Particles Formation Using SAS Process 175

al., 2011; Bouchard et al., 2008). Petit-Gas et al. found that for the lowest capillary internal diameter studied, there were particles with differences morphologies according to the jet velocity. For the lowest jet velocity, irregular morphology was obtained, and for highest jet velocity spherical morphology was obtained (Petit-Gas et al., 2009). However, for the highest capillary internal diameter experiments, particles morphology difference was less important. Particles were quasi-spherical, to a lesser extent for the smallest jet velocity. Once more time it was demonstrated the parameters interrelation in SAS process and its great complexity. Not only the kind of nozzle but also the nozzle relative position to CO2 inlet must be taken into account. In this way, Martin & Cocero studied the differences on hydrodynamics and mixing when CO2 is not introduced through the concentric annulus, but through a different nozzle, which is placed relatively far from the nozzle of the organic solution. Since the inlet velocity of CO2 is much lower than the inlet velocity of the solution, this flow has a relatively small influence on hydrodynamics and mixing. However, if CO2 is not introduced through the annulus, the fluid that diffuses into the jet is no longer almost pure CO2, but fluid from the bulk fluid phase, which has some amount of organic solvent. This greatly reduces the supersaturation and bigger particles are formed (Martin & Cocero,

Moreover, these different unstable modes (Rayleigh break up, sine wave break up and atomization) are controlled by several competing effects: capillary, inertial, viscous, gravity and aerodynamic effects (Petit-Gas et al., 2009). The predominance of each effect has been discussed in several works (Badens et al., 2005; Carretier et al., 2003; Kerst et al., 2000). Reynolds number gives a measure of the ratio of inertial forces to viscous forces. For the lower Reynolds numbers, Rayleigh regime is observed and surface tension is the chief force controlling the break-up of an axisymmetrical jet. For higher Reynolds numbers, the inertial forces compete with the capillary forces. There is a lateral motion in the jet break-up zone which leads to the formation of an asymmetrical jet, which can be either sinuous or helicoidal. Finally, when the flow rate goes beyond a certain value, the aerodynamic effects become quite strong and the jet is atomised. Another dimensionless number frequently used to describe jet fluid dynamics is the Ohnesorge (Oh) number that relates the viscous and the surface tension force by dividing the square root of Weber number by Reynolds number

In this way, taking into account the critical atomization velocity defined as the velocity corresponding to the boundary between the asymmetrical mode and the atomization mode, it is possible to tune the process towards one or another regime. Moreover this critical velocity seems to be dependent on CO2 density. Badens et al. observed a decrease in this critical jet velocity when the CO2 continuous phase density increases (Badens et al., 2005). Badens et al. and Czerwonatis et al. found out the predominant effect of the continuous phase properties on jet break-up, especially in the asymmetrical and direct atomization modes because of the aerodynamic forces preponderance (Badens et al., 2005; Czerwonatis et al., 2001). However Petit-Gas et al. concluded that variations of the continuous phase properties had no effects on the transition velocity in the studied conditions (Petit-Gas et al.,

Some authors attempted to connect the observed flow or mixing regimes to the morphology of the precipitated particles. Lee et al. injected a solution of dichloromethane (DCM) and poly lactic acid (PLA) at subcritical conditions in the dripping and in the Rayleigh

(Badens et al., 2005; Czerwonatis, 2001; Kerst et al., 2000).

2004).

2009).

**3.2 Morphology** 

When SAS is performed at supercritical conditions a transition between multi-phase and single-phase mixing is observed by increasing the operating pressure. Single-phase mixing is due to the very fast disappearance of the interfacial tension between the liquid solvent and the fluid phase in the precipitator. The transition between these two phenomena depends on the operating pressure, but also on the viscosity and the surface tension of the solvent. Reverchon et al. demonstrates that in the case of dimethyl sulfoxide (DMSO) at pressures larger than the MCP a progressive transition exists between multi-phase and single-phase mixing, but is not observed, even for pressures very close to the MCP, in the case of acetone (Reverchon et al., 2010). In the dripping mode, the droplet size decrease with increase in pressure operation due to a corresponding decrease in the interface tension, so the initial droplet size can be manipulated by small changes in the pressure of CO2 (Lee et al., 2008).

However, in the Rayleigh disintegration mode, the droplet size is weakly dependent on the interface tension of the system and is proportional to the diameter of the jet. In the dripping mode, the size and shape of the drops become highly dependent on the nozzle exit condition.

Sometimes, the transition between multi-phase (formation of droplets after jet break-up) and single-phase mixing (no formation of droplets after jet break-up) could not be located at the pressure of the mixture critical point. Dukhin et al. (Dukhin et al., 2003) and Gokhale et al. (Gokhale et al., 2007) found that jet break-up into droplets still takes place at pressures slightly above the MCP. Due to the non-equilibrium conditions during mixing, there is a dynamic (transient) interfacial tension that decreases between the inlet of the liquid and its transformation to a gas-like mixture. The transition between these multi-phase and singlephase mixing depends on the operating pressure, but also on the viscosity and the surface tension of the solvent.

Not only the thermodynamics but also the nozzle device or liquid solution flow rate will influence on the observed regime. The kind of injection device and its orifices diameter will determine the chosen liquid solution flow rate to get a successful jet break up. In this way, in a previous work, when the 200 µm diameter nozzle was used with a liquid flow rate of 1mL/min, the solution was not atomized, and we did not obtain any precipitation (Tenorio et al., 2009).

A lot of parameters control the precipitation process and many particle morphologies have been observed. As it was commented before, the kind of injection device used (and its efficiency), can strongly influence the precipitation process. The objective of these devices in SAS processing is to produce a very large contact surface between the liquid and the fluid phase, to favour the mass transfer between the antisolvent and the liquid solvent inducing jet break-up and atomization of the liquid phase.

Various injection devices to produce liquid jet break-up have been proposed in the literature. Yeo et al. (Yeo et al., 1993) proposed the adoption of a nozzle and tested various nozzle diameters ranging from 5 to 50 μm. Moussa et al. (Moussa et al., 2005) showed that the pressure distribution during the expansion of the supercritical fluid is a function of the nozzle length and diameter. Other authors used small internal diameter capillaries (Dixon et al., 1993; Randolph et al., 1993). Coaxial devices have also been proposed: in the SEDS process (solution enhanced dispersion by supercritical fluids) a coaxial twin-fluid nozzle to co-introduce the SCF antisolvent and solution is used (Bałdyga et al., 2010; He et al., 2010; Mawson et al., 1997; Wena et al., 2010). Complex nozzles geometries have also been tested carrying out a comparative study of the nozzle by computational fluid dynamics (Balabel et

When SAS is performed at supercritical conditions a transition between multi-phase and single-phase mixing is observed by increasing the operating pressure. Single-phase mixing is due to the very fast disappearance of the interfacial tension between the liquid solvent and the fluid phase in the precipitator. The transition between these two phenomena depends on the operating pressure, but also on the viscosity and the surface tension of the solvent. Reverchon et al. demonstrates that in the case of dimethyl sulfoxide (DMSO) at pressures larger than the MCP a progressive transition exists between multi-phase and single-phase mixing, but is not observed, even for pressures very close to the MCP, in the case of acetone (Reverchon et al., 2010). In the dripping mode, the droplet size decrease with increase in pressure operation due to a corresponding decrease in the interface tension, so the initial droplet size can be manipulated by small changes in the pressure of CO2 (Lee et

However, in the Rayleigh disintegration mode, the droplet size is weakly dependent on the interface tension of the system and is proportional to the diameter of the jet. In the dripping mode, the size and shape of the drops become highly dependent on the nozzle exit

Sometimes, the transition between multi-phase (formation of droplets after jet break-up) and single-phase mixing (no formation of droplets after jet break-up) could not be located at the pressure of the mixture critical point. Dukhin et al. (Dukhin et al., 2003) and Gokhale et al. (Gokhale et al., 2007) found that jet break-up into droplets still takes place at pressures slightly above the MCP. Due to the non-equilibrium conditions during mixing, there is a dynamic (transient) interfacial tension that decreases between the inlet of the liquid and its transformation to a gas-like mixture. The transition between these multi-phase and singlephase mixing depends on the operating pressure, but also on the viscosity and the surface

Not only the thermodynamics but also the nozzle device or liquid solution flow rate will influence on the observed regime. The kind of injection device and its orifices diameter will determine the chosen liquid solution flow rate to get a successful jet break up. In this way, in a previous work, when the 200 µm diameter nozzle was used with a liquid flow rate of 1mL/min, the solution was not atomized, and we did not obtain any precipitation (Tenorio

A lot of parameters control the precipitation process and many particle morphologies have been observed. As it was commented before, the kind of injection device used (and its efficiency), can strongly influence the precipitation process. The objective of these devices in SAS processing is to produce a very large contact surface between the liquid and the fluid phase, to favour the mass transfer between the antisolvent and the liquid solvent inducing

Various injection devices to produce liquid jet break-up have been proposed in the literature. Yeo et al. (Yeo et al., 1993) proposed the adoption of a nozzle and tested various nozzle diameters ranging from 5 to 50 μm. Moussa et al. (Moussa et al., 2005) showed that the pressure distribution during the expansion of the supercritical fluid is a function of the nozzle length and diameter. Other authors used small internal diameter capillaries (Dixon et al., 1993; Randolph et al., 1993). Coaxial devices have also been proposed: in the SEDS process (solution enhanced dispersion by supercritical fluids) a coaxial twin-fluid nozzle to co-introduce the SCF antisolvent and solution is used (Bałdyga et al., 2010; He et al., 2010; Mawson et al., 1997; Wena et al., 2010). Complex nozzles geometries have also been tested carrying out a comparative study of the nozzle by computational fluid dynamics (Balabel et

al., 2008).

condition.

tension of the solvent.

jet break-up and atomization of the liquid phase.

et al., 2009).

al., 2011; Bouchard et al., 2008). Petit-Gas et al. found that for the lowest capillary internal diameter studied, there were particles with differences morphologies according to the jet velocity. For the lowest jet velocity, irregular morphology was obtained, and for highest jet velocity spherical morphology was obtained (Petit-Gas et al., 2009). However, for the highest capillary internal diameter experiments, particles morphology difference was less important. Particles were quasi-spherical, to a lesser extent for the smallest jet velocity. Once more time it was demonstrated the parameters interrelation in SAS process and its great complexity. Not only the kind of nozzle but also the nozzle relative position to CO2 inlet must be taken into account. In this way, Martin & Cocero studied the differences on hydrodynamics and mixing when CO2 is not introduced through the concentric annulus, but through a different nozzle, which is placed relatively far from the nozzle of the organic solution. Since the inlet velocity of CO2 is much lower than the inlet velocity of the solution, this flow has a relatively small influence on hydrodynamics and mixing. However, if CO2 is not introduced through the annulus, the fluid that diffuses into the jet is no longer almost pure CO2, but fluid from the bulk fluid phase, which has some amount of organic solvent. This greatly reduces the supersaturation and bigger particles are formed (Martin & Cocero, 2004).

Moreover, these different unstable modes (Rayleigh break up, sine wave break up and atomization) are controlled by several competing effects: capillary, inertial, viscous, gravity and aerodynamic effects (Petit-Gas et al., 2009). The predominance of each effect has been discussed in several works (Badens et al., 2005; Carretier et al., 2003; Kerst et al., 2000). Reynolds number gives a measure of the ratio of inertial forces to viscous forces. For the lower Reynolds numbers, Rayleigh regime is observed and surface tension is the chief force controlling the break-up of an axisymmetrical jet. For higher Reynolds numbers, the inertial forces compete with the capillary forces. There is a lateral motion in the jet break-up zone which leads to the formation of an asymmetrical jet, which can be either sinuous or helicoidal. Finally, when the flow rate goes beyond a certain value, the aerodynamic effects become quite strong and the jet is atomised. Another dimensionless number frequently used to describe jet fluid dynamics is the Ohnesorge (Oh) number that relates the viscous and the surface tension force by dividing the square root of Weber number by Reynolds number (Badens et al., 2005; Czerwonatis, 2001; Kerst et al., 2000).

In this way, taking into account the critical atomization velocity defined as the velocity corresponding to the boundary between the asymmetrical mode and the atomization mode, it is possible to tune the process towards one or another regime. Moreover this critical velocity seems to be dependent on CO2 density. Badens et al. observed a decrease in this critical jet velocity when the CO2 continuous phase density increases (Badens et al., 2005). Badens et al. and Czerwonatis et al. found out the predominant effect of the continuous phase properties on jet break-up, especially in the asymmetrical and direct atomization modes because of the aerodynamic forces preponderance (Badens et al., 2005; Czerwonatis et al., 2001). However Petit-Gas et al. concluded that variations of the continuous phase properties had no effects on the transition velocity in the studied conditions (Petit-Gas et al., 2009).

## **3.2 Morphology**

Some authors attempted to connect the observed flow or mixing regimes to the morphology of the precipitated particles. Lee et al. injected a solution of dichloromethane (DCM) and poly lactic acid (PLA) at subcritical conditions in the dripping and in the Rayleigh

Hydrodynamics Influence on Particles Formation Using SAS Process 177

The ability to identify and characterize these small formations drives future system improvements, including lighting enhancements laser-induced fluorescence, and higher spatial resolution cameras. In this way Reverchon et al. used light scattering technique to clearly differentiate between an atomized very droplet laden spray and a dense "gasplume", limitation which cannot be gained by applying optical techniques due to the fact that both the droplet laden spray and the dense "gas-plume" result in a dark shadow

 On the other hand, extensive research has been done using scanning electron microscopy (SEM) to evaluate the size and morphology of particles formed under supercritical conditions (Armellini& Tester, 1994; Bleich et al., 1994; Mawson et al. 1997; Randolph et al., 1993; Shekunov et al., 2001;). A limitation of SEM analysis is that it is applied to particles

In our research group a study was carried out to establish a correlation between the morphologies of the particles obtained in the ampicillin precipitation assays and the estimated regimes. This correlation would be an ideal tool to establish the limiting hydrodynamic conditions for the success of the test in order to define the successful experiments; that is, the appropriate conditions to orientate the process toward the formation of uniform spherical nanoparticles instead of irregular and larger-sized particles,

A series of ampicillin precipitation experiments by the SAS technique, utilizing N-methylpyrrolidone (NMP) as the solvent and CO2 as the antisolvent, under different operating conditions were carried out. Two nebulizers, with orifice diameters of 100 and 200 μm,

A pilot plant, developed by Thar Technologies® (model SAS 200) was used to carry out all the experiments. A schematic diagram of this plant is shown in Figure 4. The SAS 200 system comprises the following components: two high-pressure pumps, one for the CO2 (P1) and the other for the solution (P2), which incorporate a low-dead-volume head and check valves to provide efficient pumping of CO2 and many solvents; a stainless steel precipitator vessel (V1) with a 2L volume consisting of two parts, the main body and the frit, all surrounded by an electrical heating jacket (V1-HJ1); an automated back-pressure regulator (ABPR1) of high precision, attached to a motor controller with a position indicator; and a jacketed (CS1-HJ1) stainless steel cyclone separator (CS1) with 0.5L volume, to separate the solvent and CO2 once the pressure was released by the manual back-pressure regulator (MBPR1).The following auxiliary elements were also necessary: a low pressure heat exchanger (HE1), cooling lines, and a cooling bath (CWB1) to keep the CO2 inlet pump cold and to chill the pump heads; an electric high-pressure heat exchanger (HE2) to preheat the CO2 in the precipitator vessel to the required temperature quickly; safety devices (rupture discs and safety valve MV2); pressure gauges for measuring the pump outlet pressure (P1, PG1), the precipitator vessel pressure (V1, PG1), and the cyclone separator pressure (CS1, PG1); thermocouples placed inside (V1-TS2) and outside (V1-TS1) the precipitator vessel, inside the cyclone separator (CS1-TS1), and on the electric high pressure heat exchanger to obtain continuous temperature measurements; and a FlexCOR Coriolis mass flowmeter (FM1) to measure the CO2 mass flow rate and another parameters such as total mass,

(Reverchon et al., 2010).

respectively were used.

after they have been removed from the dynamic system.

**4. A particular case: Ampicillin SAS precipitation** 

for the solute-solvent-SC CO2 system studied (Tenorio et al.,2009).

density, temperature, volumetric flow rate, and total volume.

disintegration regimes and observed the formation of uniform PLA microparticles (Lee et al., 2008). Other authors (Chang et al., 2008; Gokhale et al., 2007; Obrzut et al., 2007; Reverchon et al., 2008) did not find relevant differences in the various precipitates obtained. Particularly, PLA morphologies showed to be insensitive to the SAS processing conditions (Randolph et al., 1993). This characteristic fact could be assigned to the high molecular weights and the tendency to form aggregated particles because of the reduction of the glass transition temperature in SC-CO2.

At subcritical conditions the interfacial tension between the injected liquid and the bulk phase never goes to zero and a supercritical mixture is not formed between the liquid solvent and CO2. The droplets formed during atomization are subjected to a very fast internal formation of a liquid/CO2 mixture. Due to a high solubility of CO2 in pressurized organic liquids and a very poor evaporation of organic solvents into the bulk CO2, the droplets expand. During these processes, the interfacial tension allows the droplets to maintain its spherical shape, even when the solute is precipitated within the droplet. Saturation occurs at the droplet surface and solidification takes place with all solutes progressively condensing on the particle internal surface. The final result is the formation of a solid shell.

This kind of particles has also been observed in other SAS works (Reverchon et al., 2008). It has been also obtained expanded hollow particle at same conditions. The different surface morphologies can depend on different controlling mass transfer mechanisms, as suggested by Duhkin et al. (Duhkin et al., 2005).

Operating conditions above the MCP, from a thermodynamic point of view, are characterized by zero interfacial tension. But, the liquid injected into the precipitator, before equilibrium conditions are obtained, experiences the transition from a pure liquid to a supercritical mixture. Therefore, interfacial tension starts from the value typical of the pure liquid and progressively reduces to zero. This fact means that droplets formed after jet break-up (whose presence indicates in every case the existence of an interfacial tension) are formed before the disappearance of the interfacial tension. In other words, the time of equilibration is longer than the time of jet break-up and spherical microparticles instead of nanoparticles can be obtained.
