PARTICLE FORMATION BY RAPID EXPANSION OF SUPERCRITICAL SOLUTIONS

Background. This body of work is intended to serve as a proof of concept for the application of supramolecular chemistry in drug development. More specifically, this work is designed to evaluate crystal doping by recrystallization from supercritical media. The rapid nucleation and growth implicit to supercritical fluid based crystallizations were tested in doping drug crystals with structurally related impurities. The ultimate motive was to tailor the physicochemical properties of active pharmaceutical ingredients (APT) through crystal doping. This, in tum provides the ability to tie functionality to APl 's at early stages of drug discovery and synthesis. Methods. The rapid expansion of supercritical solution (RESS) process was evaluated for these purposes. Pure and co-solvent modified supercritical fluid C02 was used as the recrystallizing solvent. The supercritical region investigated for these studies included pressures from I 07 l-9000psi and temperatures ranging from 31-100°C. The pharmaceutical solids studied included a-naphthalene acetic acid, aspirin, benzoic acid, caffeine, chlorpropamide, indomethacin, naproxen, phenytoin, piroxicam, salicylic acid, theobromine, theophylline, tolbutamide and urea. For comparison purposes, model chlorpropamide+urea system was also recrystallized from three liquid organic solvents using evaporative crystallization. The composition, morphology and the energetics of the crystals thus produced are characterized utilizing techniques such as microscopy (polarizing optical, SEM), thennal analysis (DSC, mDSC, TGA and thennomicroscopy) and HPLC. Results. Selective extraction and a reduction in crystallinity were consistently seen in all of the drug-impurity mixtures cocrystallized by RESS process. ln addition, a number of interesting phenomena were revealed. These include habit modification, solubility enhancement, particle size reduction, eutectic formation, amorphous conversion, hydrate formation and polymorph conversion. In viewing each of these phenomena from an application standpoint, this work serves as proof of concept for enhancing the physicochemical and mechanical attributes of APl's using supercritical fluid crystal doping. Comparative evaluation studies indicated RESS to be superior to organic so lventbased recrystallizations in crystal doping. In summary, RESS offers great promise as a hybrid technique to control both the crystalline and the particle morphologies of API's in a single stage. Conclusions. The presence of an impurity in the crystallization medium exhibits varied effects depending on the phase in which it is present prior to nucleation and its affinity to the host relative to the crystallizing solvent. This in turns dictates the rate at which it nucleates and grows in relation to that of the host. The domain of effects that these kinetics dictate on one extreme includes the fomrntion ofa solid solution or a solid dispersion of the impurity in the host lattice. On the other hand, selective extraction of each of the components with respect to time can also occur, the extent of which primarily depends on the resolution factor of the recrystallizing so lvent. While the former mechanism is largely aided by the rapid nucleation and growth implicit to supercritical fluid recrystallizations, the latter forms the scope of supercritical fluid chromatography. An optimal compromise between these extremes can be reached by utili zing the adjustable solvent power of supercritical fluids. ACKNOWLEDGEMENTS Many have helped and influenced me during the course of my study whose nan1es are impossible to list here and by no means limited to the following. To them all ! am ever grateful. To begin, l would like to thank my major professor, Dr. Thomas Needham for his guidance and support throughout this seemingly never ending process. His style of writing is one quality that I have always admired and attempted to pick up. I would also like to thank my co-advisor Dr. Matthew Mollan who made himself available throughout this work notwithstanding his Pfizer activities. He is a man of alternatives, if not solutions and never can arise a complex situation to bog one down under his direction. Thanks are also due to Dr. Hossein Zia and Dr. Arijit Bose for their time and effort spent serving on my dissertation committee. I am also grateful to Pfizer Inc. that provided me the facility to perform dissertation work in an industrial setting. In particular, I would like to thank Dr. KS Murthy, Dr. David Pope and Dr. Isaac Ghebre-Sellassie who instituted this joint research program and for being instrumental in monitoring the progress. Thanks are due to the members of Technology Development group who have been very supportive and encouraging throughout. I fee l fortunate in having had the support and project-related cynicism of Jeremy Salter. His meticulous lab conduct had the greatest impact on my cGLP skills. I wish to thank Dr. Meb Fessehaie and Dr. Zeri Teweldemedhin for their useful discussions. I am perhaps the greatest benefactor of the ad hoc electrical and machining services of Alex Mircea and Ben Azzaro. I am


I. INTRODUCTION
While organic solvents are extensively used in the processing of pharmaceuticals, there has been a growing concern of late in view of the potential health hazards caused by their emissions and residues in the product. Research aimed at eliminating or reducing their use is an area of particular interest both to the industry and the regulatory agencies. Towards this goal, environmentally safe supercritical fluids (SCF's) appear to be logical alternatives to traditional organic solvents.
While significant advances have been made in such fields as extraction, ceramics, separation science, polymer processing etc, it was not until the recent past [Krukonis I 984,McHugh I 994] that pharmaceutical SCF applications have been realized. Since then, the technology has rapidly progressed as reflected by the number of SCF related publications and patents in pharmaceutical literature.
Supercritical fluids have not only established a place in the series of conventional GRAS solvents, but also possess other distinguishing features that make them attractive in a gam ut of pharmaceutical applications. SCF technology accordingly holds an immense potential, although the progress to date is limited only to the research laboratories. Table I   This section introduces the reader to the origin of SCF material processing and various modifications thereof. Various processes to be discussed hereunder (Table   3) can be best envisioned as the different permutations of contacting the solvent, solute(s), cosolvent, antisolvent and precipitating the solute thereafter. The state of solute prior to the precipitation and the mechanism of solute precipitation is what distinguishes one process from the other. In principle, the basic advantages of supercritical fluid particle formation such as rapid and uniform nucleation remain the same in all the various processes, although the mechanism of particle precipitation varies with the process. From a processing standpoint, the simplest of the processes ( Figure I) involves exposing the solute or a mixture of solutes to supercritical fluid for a fixed period of time. Rapid venting of the SCF leaves a solid product with enhanced attributes [Sand 1986]. In this process, SCF is used as a solvent or as a swelling agent, that may be rapidly removed via depressurization. An extension to this process that gives additional control over product morphology is called supercritical fluid nucleation [Krukonis 1984] or rapid expansion of supercritical solutions (RESS, Figure 2) ].
Herein, the solute(s) of interest is dissolved in a SCF at fixed extraction conditions. The solution is then rapidly depressurized through a restrictive device that is designed to tailor the product morphology. Solute interaction with SCF leading to its dissolution/swelling is a primary requirement to process solids using the above methods. However, the solubility of most drugs and pharmaceutical polymers [Alessi 1997, Subramaniam 1998] in SC C0 2 is prohibitively low, 6  Loss of SCF so lvent power a fl.er rap id evaporation PGSS Solution/Dipserion ofx l in x2 rapidly expanded Phase change in x i +Jou le Thompson Coo ling GAS AS bubbled through solution of x I +x2 Volumetric expansion of solvent by gas PCA, x I +x2 sprayed into AS (or) x2 evaporation into AS ASES Same as above Same as above SAS Sarne as above Same as above SEDS x I +x2 and AS nowed through coaxial nozzle  [Dobbs 1986]. The solubility enhancement factors although high (up to 500%) with certain cosolvents, still do not provide a realistic means to produce mass powders and mixtures on pilot scales. Further, the advantages of pure SCF crystallization may be comprom ised with the use of cosolvents. This stems from the fact that solute recrystallization occurs from the condensed cosolvent and not from the supercritical phase as observed by Larson &King [Larson 1986].
Another method of particle formation by recrystallization from pure SCF is particles from gas saturated solutions (PGSS, Figure 3) [Weidner 1995]. Unlike dissolving the solid in SCF (see RESS), PGSS involves dissolution of SCF in a soli d melt and subsequent expansion through a restriction device. The process is based on the fact that gases have higher solubility and diffusivities in liquids than in solids. Thermal stability of drug compounds and significant solubility of SCF in the melt are the primary requirements for PGSS.
However, the poor solubilities of many drug compounds in SC C0 2 propelled research efforts towards the use of SCF' s as antisolvents. This approach formed the basis for gas antisolvent precipitation [Gallagher 1989] (GAS, Figure 4), wherein a gas is bubbled through an organic drug solution.
Large volumetric expansion of the organic solution as a result of dissolved gas coupled with solvent extraction by SCF leads to high supersaturation and hence solute precipitation. Solubility requirements in this case include (i) miscibility of (iii) partial solubility of drug in the organic solvent. Realizing the fact that solvent-SCF mass transfer rates are higher from a fine spray than from the bulk solvent, modifications have been made that constitute processes like precipitation using compressed antisolvent (PCA) [Bodmeier 1995], supercritical antisolvent process (SAS) [Bertucco 1996] and aerosol solvent extraction system (ASES) [Bleich 1993]. These processes are schematically shown in Figure 5. Herein, an organic drug solution is sprayed into a compressed gas (PCA) or a supercritical fluid (SAS or ASES) that selectively extracts the solvent and thereby causes precipitation of solute.
With the objective of improving extraction conditions, simultaneous introduction of the drug solution and SCF has also been evaluated. This formed the basis for continuous PCN AS ES/SAS processes [ Figure 6] as well as solution enhanced dispersion by supercritical fluids process [York 1995] (SEDS, Figure   7). In SEDS, the mechanical energy of rapidly expanding gases is streamlined into dispersing a drug solution by passing through a coaxial nozzle. This enhances the extractive capability of the SCF's, thereby precipitating microparticles with desired attributes. Studies have shown that supercritical C02 has the ability to swell a number of glassy polymers [Wissinger 1987, Wissinger 1991, Shieh 1996b, Kazarian 1996. Sorption of C02 into an amorphous polymer matrix weakens the intermolecular forces that bind the polymer chains together, leading to an increase in the molecular motion. As a result, a depression in T g and plasticization of the polymer are observed upon C0 2 sorption. Theoretical and practical considerations of the sorption of supercritical fluids and gases in polymers have been extensively 14 dealt with by a number of authors. [McHugh 1994, Sefcik l 986a, Sefcik l 986b, Hirose 1986, Wissinger 1991Wissinger 1987, Condo 1992, Shieh 1996aShieh l 996b, Kazarian 1997, Kazarian 1996

2 Process:
The solution of drug in SCF is brought into contact with the polymer for a time sufficient to permit sorption into the polymer. Rapid decompression of the system results in the loss of SCF' s solvent power lead ing to solute deposition within the polymer matrix. Also, SCF becomes gaseous and rapidly diffuses out of the polymer, leaving a solute laden polymer. System requirements to carryout this process include a high pressured vessel capable of withstanding the operating pressure and temperature ( Figure 1, p8). Mixing of the contents in the vessel while under pressure is shown to be important for homogenous infusion [Juvekar 1994, Muth 2000]. Stirring of the contents in the pressurized vessels has an inherent difficulty arising from the moving parts that cause the leaks. This can be circumvented by the use of magnetic mixers. Typical operating conditions are T= -55 to 60°C, P=600-4300psi, t= l 5-300min, v=0.01-5 ft/sec [Lindsay 1992]. 15

Applications:
Typical applications utilizing this process employ the impregnation/infusion of solutes in porous supports, forming intimate mixtures of actives & excipients, solid dispersions, polymerization and micronization. Reported applications of thi s technique are summarized in Table 4. As can be seen from the table, a number of small drug molecules have been impregnated into porous supports and polymers using the above technique. A preferred drug molecule is one that shows some degree of solubility (at least 0.1 wt%) in the SCF. Also, some degree of interaction between the SCF and the polymer is necessary for polymer dilation (at least 2 vol%) [Berens 1989]. Partitioning of the solute in a swollen polymer is controlled by adjusting the concentration of solute in SCF and the rate of venting.
The degree of loading and the form of the resulting mixture therefore depend on the temperature (T), pressure (P), mixing conditions, time of exposure(t), and venting rate(v) as well as the properties of drug, SCF and polymer discussed above.
The degree of success achieved using this technique is limited, considering the poor interaction of many pharmaceutical compounds and excip ients with SC C0 2 . While a compound ' s solubility in SC C02 limits the level of loading that may be achieved, it is not the only factor in determining the efficacy of the process. 16 T a ble 4 '"' 12.Smol/molinclusionor Berens 1989 Van  [Zia 1997, Lindsay 1992, Howdle 1998]. Specific intermolecular interactions that come into play need to be addressed for the system under study in evaluating this process.

Introduction :
Supercritical fluids exhibit remarkable solvent properties compared to gases. In the vicinity of a fluid's critical point, solvent strength was found to be a very sensitive function of pressure and temperature. Increased solvent strengths of SCF' s compared to gases are attributed to the liquid-like densities of the fluids [Kumar 1988]. C0 2 for example, has a density of 0.47 glee at its critical point, close to the density of Hexane (0.66g/cc at 25°C) [Dixon 1997]. On the other hand, SCF's also exhibit gas-l ike behavior (eg. high diffusivity, low viscosity and zero surface tension) [McHugh 1994]. These intermediate features of SCF's have been utilized in a variety of applications including RESS.

Mechanism:
In RESS, the solute of interest is dissolved in a supercritical fluid. The high solvent power of the SCF allows formation of a homogenous solution. Nucleation of solute is then induced by rapidly reducing the solution density through expansion to atmospheric conditions. A rapid decrease in solvent strength results in high supersaturation that leads to very high nucleation rates [Mohamed l 989b).
The time for crystal formation and growth is very limited (typically 10· 5 to 10-6 seconds) resulting in very small particles [Debenedetti l 993b, Turk 1999). In add ition, the rapid decompression of SCF generates mechanical perturbation within the solution that travels at the speed of sound. Consequently, very uniform conditions are reached within the nucleating media. Narrow size distributions typical of RESS processed materials is attributed to the above mentioned behavior [Tom 199lb). Thus, RESS provides a means of forming microparticles with a unimodal particle size distribution.
The morphology of the particles formed essentially depends on the phase of solution from which the so lute is precipitated. The thermodynamic factors that control the phase behavior of solutions are pre-expansion temperature and pressure (T,P), solution composition(x) and post-expansion temperature and pressure. While solute vapor pressure behavior and chemical interaction dictate solid solubility in the SCF, solvent physicochemical/transport properties need to be considered for precipitation. Preliminary solid solubility studies in the SCF not only help to choose the extraction conditions, but also identify the conditions ideal for solid precipitation. Researchers have attempted to correlate product morphology with the T,P,x conditions of the system and form a theoretical basis for particle growth [Mohamed 1989a, Helfgen 2000, Tom JW 1991b. Most of such attempts to date have only resulted in qualitative models. To this end, density of the solution from which precipitation occurs, the time scale allowed for t9 ( precipitation and growth, as well as agglomeration were shown to have a major effect on particle morphology. A comprehensive model should take into account the combined effects ofT,P,x at each stage of RESS process on the above factors.
Among the process variables that play an important role in particle tailoring, the geometry of the restriction device through which expansion occurs merits a special mention. A restriction device is designed to support the large pressure drop that occurs across it, while maintaining suitable conditions for precipitation. Various configurations have been evaluated to date, for example capillaries, nozzles, laser drilled discs and valves. The Joule-Thompson cooling effect, observed as a result of large volumetric expansion in tum produces a drop in the temperature of the nozzle. This leads to supersaturation and premature precipitation of the solute. Plugging of nozzles that is commonly observed in a continuous RESS process results from premature precipitation of solutes in the expansion line. The restriction devices are therefore heated to prevent clogging.
Typical aspect ratios of the restriction devices evaluated to date are in the range of 6 to 20, with orifices from 20 to 1600µ in diameter. The effects of different geometrical configurations of the restriction devices on the morphology of particles have also been investigated , Mohamed 1989b, Kim 1996.
As a result of these studies, theoretical models addressing the fluid dynamics during expansion and particle growth therein are developed , Lele 1992, Debenedetti l 993a]. 20 Poor yields owing to the low solubility of many pharmaceuticals in SCF' s is a major limitation of the RESS process. Use of co-solutes and co-solvents to improve the solubility of solutes in SCF ' s have been investigated. Tavana and Randolph [Tavana 1990] have shown that the solubility of salicylic acid in SC C02 can be enhanced by an order of magnitude in the presence of a more volatile co-solute like benzoic acid . Similar observations were reported by Kurnik [Kurnik 1982] and Pennisi [Pennisi 1986]. The enhancement of solubility has been attributed in all these cases to the vapor pressure effects of the more volatile cosolute. For further details about solubility of mixtures in supercritical fluids, the reader is advised to refer to a recent review by Lucien and Foster [Lucien 2000].
Various research groups have evaluated the design and synthesis of COi-philic polymers and surfactants that aid in the solubility enhancement of solutes in SC C02 [McClain I 996, Ghenciu EG 1997, Ghenciu 1998, Yazdi AV 1997, Super 1997]. To date, the success with such enhancement aids has been marginal, although the potential for such research using pharmaceutical polymers and surfactants is tremendous.
The practical and theoretical aspects of using co-solvents to enhance solid solubility in SCF' s (for the most part SC C02) have been extensively evaluated by a number of researchers [Dandge 1985, Wong 1986, Dobbs 1986, Larson 1986, Ting 1993. Various mechanisms of solubility enhancement by co-solvents have been postulated (Ekart 1993]. for screening co-solvents with two model compounds griseofulvin and digoxin [Tavana 1989]. The co-solvents are ranked based on the GC retention time with SCF+co-solvent being the mobile phase and solute as the stationary phase. Given the restrictions on the choice of organic solvents in pharmaceutical processing, co-solvent use in RESS processing has been very limited. Jn addition, the presence of a co-solvent can sometimes adversely effect the product characteristics by condensing in the precipitation vessel [Larson 1986]. Jn such instances, recrystallization occurs from the condensed co-solvent and not from the supercritical phase, thereby losing the very attributes of RESS [Larson 1986, Mohamed 1989b. It is therefore important to select pre-expansion conditions that allow precipitation from a single fluid phase. Also, pre-expansion temperature and conditions in the precipitation vessel should be chosen such that the co-solvent stays in vapor phase after expansion. With proper choice of conditions, the removal of SCF and co-solvent should leave solid dry particles within the precipitation vessel. This however, is not an easy task considering the complex phase behavior exhibited by a three component SCF-solute-cosolvent system. 22 RESS has been evaluated in many of the processing areas where organic solvents can be replaced by a SCF. The pivotal role of solvents in crystallization, forming intimate mixtures of different substances that requires mixing at a molecular level, coating, etc requires no special mention. SCF solvents not only aid in serving these objectives, but also possess other distinguishing features that make the RESS process unique. One of the major advantages in RESS is the ease of solvent removal, in contrast to solvent evaporation. While the former is triggered by the thermal perturbations, the latter occurs due to mechanical decompression at supersonic velocities [Debenedetti l 993b]. The primary requirement for a solute to be processed by RESS is a significant solubility in SCF (>0.5 wt%) [Alessi 1996). Among the pharmaceutical class of compounds, steroids with a basic perhydrocyclopentanophenanthrene ring have shown significant solubility in SC C0 2 and are particularly suitable for RESS processing.

Process:
The basic components of a RESS apparatus consist of a pump to deliver the SCF, preheater, extraction vessel, preexpansion chamber, throttling device and a precipitation vessel ( Figure 2). Liquid C02 from a tank is fed to the preheater at a controlled flow rate using the pump. Typically, an air driven pump is used to pressurize the C0 2 prior to delivery to the preheater. The function of the preheater is to bring the temperature of the pressurized liquid to the supercritical region.
The preheater is typically a lengthy stainless steel tube immersed in a temperature 23 controlled bath. Pressure and temperature in the preheater are read using pressure transducers and resistive temperature devices (RTD) that have a sensitivity of 0.05-0.1 %. SCF from the preheater then flows into the extraction vessel that contains the material to be processed. Packing the extractor with alternate layers of glass wool and solute has been shown to improve extraction efficiency by providing better fluid contact with solute. Alternatively, a mixing device may be used that agitates or stirs the contents of the vessel. The pressure and temperature in the extractor are recorded using a pressure transducer and RTD respectively.
The saturated solution from the extractor flows into the pre-expansion chamber that has independent temperature control and a line connecting it to the preheater.
The saturated supercritical fluid solution can be diluted with fresh solvent from the preheater, allowing control over the composition of the solution prior to expansion. Interfacing a HPLC system at this point helps determine the exact composition of solution prior to expansion. The fluid from the pre-expansion chamber is rapidly expanded through a heated restriction device into the precipitation vessel. For the most part, the effects of post expansion pressure and temperature on product morphology are inconclusive or relatively insignificant.
Excepting situations where post expansion conditions have been shown significant [Mohamed 1989a], or where fluid recompression costs are a factor, the precipitation vessel in most instances is maintained at atmospheric conditions.
Typical extraction conditions are T=40-80°C, P=2000-5000psi . The preexpansion temperature is generally maintained at about 50°C higher than the 24 extraction temperature to prevent premature precipitation, which in tum leads to plugging of lines. Solute throughput in a RESS process depends on its solubil ity in the supercritical fluid and is typically up to 1 g/hr with so lvent flow rates ranging between 20-80 standard liters per hour [Tom 1991 b).

Applications:
One of the potential applications of rapid expansion of supercritical solution process is in the area of particle size reduction. The RESS process, in principle, offers the advantage of growing the crystals to a desired size unlike most other high energy comminution techniques like wet milling, spray drying etc. Most of these processes commonly involve energizing the particles to bring about size reduction. The implications of imparting energy into the system are pronounced when dealing with proteins, peptides, and other unstable compounds.
Accordingly, particles generated using RESS process frequently retain their crystallinity and do not carry static charge. Particles with various morphologies like microspheres, needles, fibres, dendrites, etc. were produced by changing the process conditions. Production of micron and submicron particles with a narrow size distribution has been demonstrated with a range of compounds (Table 5). A universal model relating process conditions to product morphology is yet to be developed, but definitive trends have been observed in each of these compound specific studies. Agglomeration of particles was prominent, as observed with cyclosporine [Henriksen 1997] and lovastatin [Mohamed l 989b ], when the 25   [Tom 1993, Kim 1996]. Coprecipitation of drug and polymer is then carried out with an optimal balance of process conditions to form microparticles or microspheres. A complete understanding of the 3-component phase behavior is necessary to prevent independent precipitation of solutes.
Microcapsules (10-100µ) of lovastatin needles embedded in DL-PLA have been produced [Tom JW 1993]. The authors proposed that the readily so luble lovastatin was extracted first and then acted as a nucleating . site for the later precipitating polymer. Similar results with a better dispersion of drug within polymer were reported for pyrene and naproxen in L-PLA [Tom 1994, Kim 1996].
In comparison with the conventional solvent evaporation and coacervation techniques of solid formation , RESS offers an effective means of producing micron and submicron particles with unimodal size distribution. However, the limited solubility of many pharmaceuticals in the most widely used SCF viz. C0 2 29 restricts its application to a few low molecular weight lipophilic compounds.
Also, theoretical understanding of the process is limited and further complicated by the introduction of a third component such as a co-solvent or a polymer. Future research should aim at identifying potential approaches to improve solute solubility in SCF and generalize the effect of process conditions on RESS product morphology.

fntroduction :
The solubility of solid solutes in SCFs and vice versa have been explored in RESS and the basic supercritical fluid processes discussed above. Considering the scarcity of interaction observed leading to low solubilities and poor yields, a logical alternative is to exploit the solubility of SCFs in solid melts (liquids). This approach forms the basis for yet another particle formation process called PGSS.
The solubility of SCFs in liquids is about three to four orders of magnitude higher than the typical solid solubilities in SCF [Weidner 1996]. Accordingly, the product yield of PGSS process is significantly high in comparison to poor RESS throughputs. Also, the fluid consumption in PGSS is considerably reduced compared to RESS. A distinguishing feature of PGSS, in contrast to the other SCF techniques of particle formation is the complete absence of co-solvent use. 30

Mechanism:
The schematic of a typical POSS process is shown in Figure 3. In POSS, the solid(s) to be processed is melted to form a single liquid phase prior to saturation with a compressed gas or a supercritical fluid.  [Weidner 1996]. Reported studies using a divergent class of chemical compounds (Table 6) support the theory that the properties of a compound have a weak influence on the SCF solubility, often outweighed by the P,T effects of the process. The PGSS process involving dispersion of a compressed gas in a solid melt is also shown to form powder particles [Mura G 1995]. A clear understanding of the solubility influence on product morphology remains to be established.

Process:
The basic components used in producing particles from gas saturated solutions are a pump to pressurize carbon dioxide, a saturation vessel capable of withstanding 32 ~  [Weidner 1996] to pilot scale designs producing 1000 Kg solids/hr [Mura 1995].

Applications:
Reported applications of PGSS process are in the grinding of difficult-tocomminute PEG polymers, microni zation of drugs, formation of solid dispersions aimed at improving aqueous dissolution rate of hydrophobic drugs, etc. (Table 6) .
Weidner et al have processed polyethylene glycols with molecular weights ranging between 1500 to 35,000 and formed unimodal microparticles with different morphologies, such as fibers, spheres and sponges by varying the process conditions [Weidner 1996]. Micronization of phenacetin to produce homogenous 5µ particles has been demonstrated by Mura and Pozzoli using a pilot scale PGSS process with product yields up to 1000 kg/hr [Mura 1995].
PGSS, as a means of forming solid dispersions has been evaluated using a number of hydrophobic drugs. Improving the aqueous dissolution rate of a series of drugs has been evaluated in view of the micronization and solid dispersion capabilities of particles from gas saturated solution process. Up to a 15 fold enhancement in dissolution rate is achieved with felodipine and nifedipine when the agglomeration of produced particles is kept to minimum [Kerc 1999 Further, the milling step associated with all the latter methods of solid processing is avoided in the single-step PGSS process. In comparison to other SCF techniques of powder processing, PGSS operates at much higher energy conditions. The feasibility of the PGSS process for the compounds under consideration should be critically evaluated along the lines described above, prior to selecting the process.

Introduction:
Poor solubilities of many pharmaceutical compounds and polymers in SC C0  Review of literature reveals that some of these terms are loosely and interchangeably used, with no rigid definitions distinguishing one from the other.
A general conception of the terminology is stated unambiguously by Subramaniam et al and is followed here [Subramaniam 1997]. GAS is generall y used for a batch process ( Figure 4) wherein a gas/subcritical or supercritical fluid is bubbled through a stationary bulk of solute laden organic solvent. Decrease in solvent density as a result of large volumetric expansion leads to a rapid loss of solvent power and therefore the solute precipitates out instantly. Particle formation occurs in the liquid phase and a secondary solvent removal process is required to produce dry particles. A modification of this process with the objective of enhancing mass transfer by spraying organic solution into compressed fluid is broadly called PCA. Due to the improved transfer rates of organic solvent into the compressed fluid and vice versa, rapid evaporation of organic solvent and droplet expansion take place respectively, leading to the precipitation of fine particles. The process is tem1ed ASES or SAS when the state 37 of compressed fluid used as the antisolvent is supercritical. The PCA process has been investigated both under batch and continuous modes. The former involves spraying organic solution into a vessel containing compressed fluid ( Figure 5) [Bodmeier 1995), whi le concurrent administration of compressed fluid and organic so lution at predetermined flow rates in a continuous manner constitutes the latter process ( Figure 6) [Yeo 1993) .
6.2 Mechanism of particle formation: The mechanism of particle formation is by solvent-antisolvent precipitation. The solute(s) to be processed is dissolved or dispersed in an organic solvent that has preferential affinity to the compressed fluid rather than the sol ute. When brought in contact with the compressed fluid , the organic solvent instantly throws out the solute owing to the loss of its solvent power. Particle precipitation is postulated to occur through two different mechanisms [Tom 1993). The influx of compressed fluid into the bulk of organic solution or the spray droplets brings about a large volumetric expansion of the solvent. This is followed by a loss of solvent power and very high supersaturation within the organic solution. The degree of supersaturation in the organic solution and particle growth are controlled by the rate and extent of antisolvent addition . Preliminary studies to determine the nature of solvent expansion caused by a compressed gas, as a function of pressure and temperature allows selection of appropriate process conditions [Gallagher 1989, Yeo 1993. 38 On the other hand, solvent flux into the compressed fluid causes rapid evaporation of the solvent, thereby supersaturating the solution. This is influenced by the relative affinity of solvent to the compressed fluid versus the solute. Also, other conditions such as the solute concentration in the organic solvent, relative rates of flow of organic solution and compressed fluid , pressure and temperature conditions of the compressed fluid etc. affect solvent evaporation. The rate of solvent evaporation and the degree of antisolvent penetration in the droplets have been shown to have a major effect on the porosity of the particles formed [Dixon 1993, Werling 1999] .
While solvent expansion by the gas is shown to potentially influence particle morphology in the GAS process, solvent evaporation and other spraying conditions mostly affect particle formation and growth in spray processes. Mass transport rates and the dynamics of jet breakup dictate particle morphology in spray processes. The former is found to have greater influence on particle morphology compared to the latter. Werling and Debenedetti have developed an integrated model of the effects of these mechanisms on particle morphology [Werling 1999]. The choice of process conditions should take into account the effects of these two different mechanisms on optimal particle formation. In addition, the phase of the medium where particle formation occurs is another factor affecting particle morphology. Particle formation in the GAS process occurs in the liquid organic phase and involves secondary solvent removal and drying steps. On the other hand, spray processes offer particle formation in the 39 supercritical phase in which the solvent is instantly extracted, leaving dry microparticles. In a recent development, spraying organic solution into a two phase vapor over a liquid antisolvent has also been evaluated [Young 1999]. It is postulated that particles formed in the vapor phase are later solidified in the liquid antisolvent beneath it. Preliminary phase behavior studies of the ternary system will form a basis for understanding the site of particle formation and thereby allow control of particle morphology.
Compounds most suitable for antisolvent processing should have negligible interaction with the SCF and sparing solubility in the solvent used . In the presence of interactions between the compounds and the antisolvent, the solute is extracted along with the solvent. These interactions not only effect the ease of solvent removal, but also result in low overall yields. While visual inspection of compound behavior in SCF is one way of determining compound suitability for antisolvent processing [Bodmeier 1995], Steckel et al. have attempted to rationalize it based on partition coefficients of compounds [Steckel 1997]. The authors have shown for glucocorticoids that a log P (octanol/water) value of less than four eliminates the possibility of compound extraction by SC C0 2 . Ideal solvents for use in supercritical antisolvent processes should have a significant affinity for SCF and a high vapor pressure. Most common solvents used to date with C0 2 as the antisolvent include methylene chloride, dimethyl sulfoxide, methanol and ethanol. Minimal solvent residues that are an order of 40 magnitude below the pem1itted levels have been achieved through proper choice of operating conditions.

Process:
The basic components of a GAS system ( Figure 4) are a precipitator with ends fitted with filters, pumps to precisely deliver compressed gas and organic solution into the precipitator and optionally, a post expansion vessel to separate the compressed gas from organic solvent for reuse. The solute(s) to be processed is dissolved in the organic solvent, typically in the concentration range of 0.1-5 mg/ml and is introduced into the precipitator using a pump. In the particle fomiation step, a predetemiined amount of compressed gas flows through the organic solution at a regulated rate. Owing to the volumetric expansion of the solvent, particle precipitation occurs. The morphology of particles fomied depends on expansion path followed, regulated by the rate of addition of compressed gas and the solute concentration in the organic solution. Particle precipitation is followed by an extended drying step where generous amounts of compressed gas are bubbled through the precipitator. During this process, particles are restored in the precipitator using filters at both ends. The pressure and temperature conditions within the precipitator are controlled and recorded using a pressure transducer and a RTD . Typical operating conditions are 30-40°C and l OOO-l 500psi and antisolvent flow rates of 17-18 SLPM. In comparison to RESS, the gas antisolvent process operates at milder conditions and produces 41 higher yields. On the other hand, particle agglomeration and the additional drying step in the GAS process owing to low rates of mass transfer prompted spraying organic solution into the compressed fluid.
Spray processes require an atomizing device in addition to the components described above. Various spray devices that range from simple capillaries and nozzles to vibrating, energized nozzles have been evaluated. The influence of nozzle configuration on the final particle morphology has been found to be rather insignificant compared to other operating conditions like relative rates of flow of organic solution and antisolvent, pressure, temperature, etc. This is explained by the fact that mass transport rates between the solvent and SCF have greater influence on particle size relative to the dynamics of jet break up and initial droplet size [Werling 1999].
In a batch spray process ( Figure 5), organic solution is sprayed into the precipitator containing a compressed fluid where the particles are formed.
Additional compressed antisolvent is then swept through the precipitator to remove the organic solvent completely. Typical flow rates of organic solution are between 0.1 -1 ml/min. ln the continuous mode ( Figure 6), organic solution and the compressed fluid are simultaneously administered at predetermined flow rates into the precipitator. Typical flow rates of organic solvent and compressed fluid during particle precipitation are 0. l-3ml/min and 6-20 SLPM respectively. At the end of particle precipitation, spraying of organic solution is stopped while additional amounts of compressed fluid is passed through the precipitator to 42 remove the organic solvent. With proper choice of operating conditions, dry particles containing very low levels of residual volatile organic content can be produced.

Applications:
The advantages of using supercritical antisolvent crystallization in the micronization of drugs and pharmaceutical excipients are numerous. Energy requirements for the process are low and the technique offers the ability to process compounds under mild conditions. As summarized in Table 7, a number of pharmaceutical actives and excipients have been processed and various particle morphologies have been achieved. While the process has mostly been evaluated in the production of dry particles for nasal administration, it can also be extended to tailor particle morphology for any desired situation. Schmitt patented the miconization of a number of API's including alprazolam, triazolam, ibuprofen, erythromycin, penicillin, ampicillin, glyburide, dexamethasone, hydrocortisone etc. [Schmitt 1990]. The technique offers the ability to form discrete microparticles with a tight size distribution using mild process conditions while preserving the activity of sensitive molecules [Yeo 1993, Young 1999    of magnitude lower than FDA regulated limits have been achieved [Ruchatz 1997, Steckel 1997]. Compared to solvent based crystalli zation, the levels of solvent waste can be significantly reduced by reusing the solvent. Removal of C0 2 from the solvent after particle formation can be affected by simple depressuri zation and both the solvent and antisolvent can be recirculated for reuse.
Another potential pharmaceutical application in the area of controlled release by microencapsulation of drugs is reflected in a number of reported publications. The efficiency of encapsulation can be improved by mixing at the molecular level, which is only possible with solvent based microencapsulation techniques. Supercritical microencapsulation, in principle, combines the advantages of solvent based techniques while providing a number of other advantages. These include the ability of controlling particle morphology and the ease of solvent removal. As can be seen in the Table 7, a number of pharmaceutical actives have been microencapsulated in various polymers using supercritical microencapsulation. The effect of the lipophilicity of drugs on the efficiency of loading into L-PLA has been investigated by Bleich et al. [Bleich 1996]. The least hydrophilic among the compounds studied showed maximum loading in L-PLA, while lipophilic piroxicam was found to be extracted by the antisolvent. Preliminary studies of the ternary phase behavior for the selected system should help in choosing appropriate solvent and process conditions to form the desired microcapsules. 47 Typical polymers evaluated to date include L-PLA, HYAFF-11, PGLA, etc. The thermal and crystal attributes of these polymers that make them particularly suitable for particle formation using the antisolvent process was reported recently by Engwicht et al. [Engwicht 1999]. Due to the fact that the thermodynamic and phase behavior of only a few polymers in supercritical C02 are well documented, a majority of the supercritical microencapsulation studies are restricted to a selective few polymers. It remains to be seen how other pharmaceutical polymers will perform in supercritical microencapsulation.
Rapid dissolution and absorption of actives of thermodynamically unstable forms of drugs has been a subject of interest in the recent past. In this direction, a nearly amorphous form of prednisolone has been produced by Steckel et al. [Steckel 1997]. The processing of drug mixtures to alter crystallinity has been reported by Weber et al. [Weber A 1997]. Varying degrees of crystallinity of aspirin and chloramphenicol were achieved by coprecipitation with ascorbic acid and urea using supercritical conditions. Although the technique in principle seems to offer the ability to alter crystallinity, more investigation needs to be performed before definitive conclusions can be made.
Utilizing this technique, Subramaniam et al patented the process of coating actives with excipients and forming free flowing microparticles in a single step [Subramaniam 1998] . Possible applications are in the taste masking, controlled release and enhancing dissolution rates of pharmaceutical actives. With the objective of improving wetting and thereby dissolution, Steckel et al. [Steckel 48 1997) have coprecipitated a series of steroids with phosphitidyl choline and observed a significant decrease in contact angle with water. If this study could be extended to other poorly soluble pharmaceutical actives, this approach may provide a convenient way to forn1 free flowing discrete microparticles with enhanced solubility attributes in a single stage processing.
The supercritical antisolvent technique has the potential for use in a multitude of applications for particle formation . However, the current level of understanding of a ternary phased supercritical mixture is rather primitive, restricting its application to a few excipients and actives. Pharmaceutical applications utilizing this technique have mostly been restricted to processing drugs with a few excipients. Extension of the technique to new molecules requires better understanding of the physicochemical properties of compounds that make them amenable to antisolvent processing as well as the phase behavior of ternary system under consideration.

Introduction:
With an ever-increasing need to tailor the particle morphology of pharmaceutical powders and to overcome the limitations of above described particle formation methods, alternate combinations of particle precipitation techniques have been explored. A more recent development among these supercritical fluid processes is what is known as ' Solution enhanced dispersion by supercritical fluids' (SEDS) 49 [York 1995]. ln SEDS, the solute(s) of interest is dissolved or suspended in an organic and/or aqueous solvent that is brought into contact with pure or modified SCF (antisolvent) using a coaxial nozzle. Mixing of the two fluids takes place in the nozzle just prior to the expansion through a restriction. The efficiency of particle precipitation by SCF is enhanced in SEDS by utilizing the energy of the rapidly expanding gas in dispersing the solvent. This feature of SCF's coupled with their solvent-extraction capability is believed to enhance the mass transport between fluids . Improved mass transfer between the solvent and the SCF assists in complete removal of solvent, which in tum aids in the formation of nonagglomerated powders [York 1999]. Among the several SCF particle formation methods evaluated to date, SEDS offers a convenient means of forming nonagglomerated powders under mild processing conditions while placing fewer restrictions on the solubility properties of the compounds.

Mechanism of Particle Formation:
Similar to the supercritical antisolvent processes, antisolvent-induced precipitation of solute from a solution or a suspension forms the basis for particle formation in SEDS. Refer to Section 4.2 for details about precipitation by supercritical antisolvents. A major limitation to the processes discussed in Section 4 arises from poor mass transfer between the fluids, leading to incomplete solvent removal and hence agglomerated particles. While the morphology of newly formed primary particles depends on such factors as pressure, temperature, 50 density of SCF, initial droplet size, nucleation rates, spray velocities etc, incomplete solvent removal leads to growth and agglomeration of primary particles. To retain the characteristics of the primary particles, it is therefore important to remove the solvent immediately upon particle formation . From a theoretical standpoint, so lvent flux from the droplet into the SCF is higher from a finely dispersed mist of solution. Dispersing the solution in SEDS is brought about by the use of a coaxial nozzle. In principle, SEDS is an extension of the supercritical antisolvent spray process and operates similar to the continuous PCA process. The major difference from the PCA process lies in the use of a coaxial nozzle with multiple passages for different fluids. The nozzle not only helps in reproducibly contacting the fluids at a specific site of interest, but also helps in streamlining the mechanical energy of the rapidly expanding SCF to disperse the solvent.
The solute of interest is dissolved or suspended in a solvent, which can be organic or aqueous. Precipitation of solute from its solution is caused by contact with pure or modified supercritical fluid. A multi-channeled nozzle with a mixing chamber allows convenient contact between the fluids at the site of interest, prior to dispersion and extraction of solvent and particle formation. Complete understanding of the mode of particle growth and the effects of process variables on particle morphology requires knowledge of the phase behavior of the system under study. The importance of fluid phase behavior is addressed in a recent review by Palakodaty and York, in which the authors addressed U1e fundamentals 51 of binary and ternary phase behavior involving SCFs ]. The literature on the phase behavior of solvents and solutes that are routinely used in pharmaceuticals, however is scarce and therefore are limited to predictive calculations to date. One such study, characterizing the crystallization mechanisms of paracetamol by SEDS was recently published by Shekunov et al. [Shekunov 1999] . With rapidly increasing SCF applications in the field of pharmaceutics, it remains to be seen how new developments would aid m understanding the process better.

Process:
A general schematic of the SEDS process is shown in Figure 7. The basic components of SEDS process include pumps to deliver the fluids at desired rates, a co-axial nozzle, and a particle formation vessel. Solute(s) to be processed is Specific designs of various coaxial nozzles can be found in York' s patent [York 1995]. In the particle formation step, the fluids are expanded through the nozzle where solvent dispersion and extraction take place instantly. Particles are retained by frits placed at the outlet end of particle formation vessel. Temperature and pressure of the particle formation vessel are controlled using an oven and a back pressure regulator respectively. At the end of the run, pure SCF is flowed through the system for 10-15 min to remove any remaining solvent.

Applications:
One of the potential areas of SEDS application is in the crystallization of pure drugs. Compared to conventional crystallization, SEDS has been shown to generate particles that have better attributes such as crystallographic purity, 53 uniformity in size and size distribution, lower solvent residues, etc. For delivery purposes, conventional crystallization methods often require secondary processing of the material that may effect the activity of the molecules, besides adding to the economics of production. SEDS, on the other hand offers the ability to combine the processes such as crystallization, purification, micronization, etc. into one unit operation, while providing better control over particle morphology [York P 1999].
As can be seen from Table 8, the technique is particularly attractive in producing dry particles (pure, 1-5µ, static free, non-cohesive, free flowing, freely dispersible) intended for insufflation. Pulmonary delivery of such compounds places demanding requirements on the particle size, size distribution and other aerodynamic properties of the powders. Studies involving comparative evaluation of various processes in producing such powders have revealed that SEDS generated particles have better attributes [York 1996, Palakodaty 1997].
Utilizing the selective nature of supercritical fluids and the rapidity of extraction in the SEDS process, it is possible to separate polymorphs [Beach 1999] and enantiomers [Koordikowski I 999]. This provides a convenient way of cleaning up active phannaceutical ingredients (AP!s) and thereby producing pure drug substances. To the same objective, the amorphous and metastable domains in AP!s that would otherwise compromise their stability can be removed by selective SEDS based crystallization. Highly crystalline forms of salmeterol xinafoate, lactose and fluticasone propionate have been produced using SEDS process.
Conversely, crystal to amorphous conversion of pharmaceutical actives can be 54 I "' 1"00 IM"""'"''ioo 1 06" "'""'"  [York 1995] to control the release and enhance the fluidi zation efficiency of the drug. Polyn1er processing using SEDS technology has recently been reported by Ghaderi el al. Various morphologies of polymers like L-PLA, DL-PLA, polycaprolactone have been produced using the SEDS process.
Processing of pharmaceutical sol ids frequently involves optimizing particle size, purity, crysta!linity, flow, static charge, cohesiveness, solvent-57 content, stability besides other features specific to the delivery system. A particle formation process that effectively combines all these steps into one unit operation is of particular interest in the context of integrating chemical synthesis and formulation development. With increasing emphasis on reducing the time scales of different phases of drug development, there is a growing attention to such techniques with feasibility for scale up. Understanding the basic mechanism of particle nucleation and growth is essential in reproducibly producing the powders and in scale-up. Such understanding to date is limited and future work in this field should aim at forming a general basis for processing a wider variety of compounds. It is noteworthy that SEDS and PGSS are the only two processes, among several SCF particle formation methods that approached commercialization on a pilot scale.

8.SUMMARY:
In the reality of growing competition and emphasis on reducing the drug    shortcoming to the technological progress. It is therefore the purpose of this article to provide the information and resources necessary for startup research involving particle formation using supercritical fluids. The various stages of supercritical particle formation can be broadly classified into Delivery, Reaction, Pre-expansion, Expansion and Collection. The importance of each of these processes from the standpoint of tailoring the particle morphology is discussed in this article, while also providing various alternatives to perform these operations.

INTRODUCTION
The central role of solvents m the processing of pharmaceutical materials is widely accepted since the origin of modem pharmaceutical processing. It is only in the recent past that the adverse effects of the residual solvents from both processing and environmental standpoints have been recognized. Strict regulations on the use of organic solvents and their content in the end products forms a major limitation to the traditional techniques. In an effort to eliminate or reduce the use of volatile organics, search for alternative techniques of material processing developed as a new facet to pharmaceutical research. Supercritical fluid (SCF) technology is a recent outcome of such research with particular emphasis in the green synthesis and particle formation. Particle formation using supercritical fluids involves negligible or no use of organic solvents, while the processing conditions are relatively mild. ln contrast to the conventional particle formation methods where a larger particle is originally formed and then comminuted to the desired size, SCF technology involves growing the particles in a controlled fashion to attain the desired morphology. The adverse effects originating from the energy imparted to the system to bring about size reduction can thus be circumvented. This feature makes supercritical fluid technology amenable to processing biomolecules and other sensitive compounds.
Particle engineering using supercritical fluids is a relatively recent development in the pharmaceutical arena. Growing demands on the particle and crystalline morphologies of pharmaceutical actives and excipients, coupled with 72 the limitations of current methods, brought wide attention to this technology [York 1999]. The technology is rapidly evolving, as reflected by the number of modified processes reported since its inception. These include Static Supercritical Fluid process [Lindsay 1992], Rapid Expansion of Supercritical Solutions (RESS) ], Particles from Gas Saturated Solutions (PGSS) [Weidner 1995], Gas Antisolvent process (GAS) [Gallagher 1989], Precipitation from Compressed antisolvent (PCA) [Bodmeier 1995], Aerosol Solvent Extraction System (ASES) [Bleich 1993], Supercritical Antisolvent process (SAS) [Bertucco 1996] and Solution Enhanced Dispersion by Supercritical fluids (SEDS) [York 1995]. Refer Table 3 and Figures  Carbon dioxide is regarded as an ideal processing medium [Subramaniam 1997] for a number of reasons. It is generally regarded as safe (GRAS), nonflammable, inexpensive; has a low critical temperature and pressure and exhibits solubilization and plasticization effects that can be varied continuously by moderate changes in pressure and temperature. The solvent properties of supercritical carbon dioxide are reported to resemble those of hexane, toluene, isopentane and methylene chloride depending on the pressure and temperature conditions of the fluid (Hyatt 1984, Dandge 1985, Dobbs 1987, Ting 1993 The majority of the off-the-shelf SCF instrumentation currently available is designed for extraction purposes. Only a selective few vendors are in the early stages of manufacturing equipment specific to particle formation (Table 9). 74  A general practice however, as reflected from the reported publications and patents, is to reconfigure a commercially available system speci fie to the end use.
It is the purpose of this article to provide the information and resources necessary for startup research involving particle fonnation using supercri tical fluids. The various stages of supercritical particle formation can be broadly classified into Delivery, Reaction, Pre-expansion, Expansion and Collection. The importance of each of these processes from the standpoint of tailoring the particle morphology is discussed in the following sections while also providing various alternatives to perform these operations.

SUPERCRITICAL FLUID DELIVERY
The critical point for any pure substance is defined by the temperature and

REACTION
A reaction vessel is where the supercritical fluid is brought in contact with the material(s) to be processed. Essential requirements for a reaction vessel are chemical inertness, ability to withstand the operating temperature and pressure conditions and ASME specified design. Several designs of the pressure vessels are currently available and in general are distinguished by the type of closures. Mixers are available as off-the-shelf items (Table 9).

PRE-EXPANSION
The composition and phase of the supercritical solution from which particles are precipitated is found to have a major effect on the morphology of particles in RESS and PGSS processes and is controlled during the pre-expansion stage [Helfgen 2000, Weidner 1996]. Independent control of the temperature and pressure during the pre-expansion stage is therefore critical in these processes.
Additionally, the phase changes in the supercritical solutions, which often lead to plugging of the lines, can be eliminated through the use of a controlled pre-

S. SPRAY CONFIGURATI ONS
In supercritical fluid particle formation, the fluids are expanded through a restriction device in a controlled fashion. A restriction device is designed to support the large pressure drop that occurs across it, while maintaining suitable conditions for precipitation. The geometry of the restriction device has been shown to influence the morphology of the particles to varying degrees and by different mechanisms , Debenedetti 1993a, Subrahmaniam 1998]. In RESS and PGSS processes, the device controls the growth of particle after the nucleation process by affecting the dynamics of jet expansion. On the other hand, the restriction device in antisolvent processes affects particle morphology by controlling the initial droplet size and also the rate of solvent extraction by the SCF. Various configurations have been used to date, namely capillaries, nozzles, 83 laser-drilled discs and valves. For investigative purposes, capillaries are preferred to other specialized designs owing to their availability, cost and the ease of changing the geometry of the device in house [Kim 1996]. Typical aspect ratios of the restriction devices evaluated to date are in the range of 6 to 20, with orifices from 20 to 1600µ in diameter. Joule-Thompson cooling, resulting from the large volumetric expansion across the restriction device, causes a drop in temperature, thereby affecting a phase change and subsequently leads to plugging of the device. The restriction devices are therefore heated to compensate for such effects. While stainless steel nozzles are most frequently used owing to their strength to withstand the large pressure differential, they are limited by their poor thermal conductivities. Wherever necessary, they can be replaced with sapphire nozzles that provide better heat transfer to the fluid while also maintaining the material strength. The devices for the most part are custom designed according to the specific needs of the researcher. Off the shelf devices with standard configurations can also be obtained from selective supercritical fluid vendors (Table 9). Other coaxial nozzles that are specific to the SEDS process are regulated by the stringent patent protection and can be purchased for purposes notwithstanding the claims of the patent [Hanna 1999].

PARTICLE COLLECTION
Retaining the original characteristics of the particles produced by supercritical fluid process is as critical as forming the particles and constitutes the particle 84 ( collection step. This step is critical in that the distinct characteristics of the particles can be completely lost owing to a poor collection technique [Turk 1999] .
In rapid expansion of supercritical solution and particles from gas saturated solution processes, the rapid ly expand ing supercritical fluids impart high kinetic energies to the particles produced. Insufficient path for expansion can therefore result in the agglomeration of particles. The agglomeration is even worse in the presence of residual amounts of co-solvent in RESS process or uncongealed portions in PGSS process. Design of particle collection vessel in these processes should be such that agglomeration is kept to a minimum by providing a sufficient path of expansion for the supercritical fluids. While a logical solution is to make the collection vessel very large, the collection of small amounts of material from a relatively larger vessel can be difficult, resulting in low yields. This problem can be circumvented in part by inserting detachable baskets inside the vessel. The baskets can be taken apart at the end of the process to collect the particles. While precipitating the solutes into a non-solvent containing a surfactant is another solution to agglomeration, it adds one more step to an otherwise continuous unit operation. An optimum balance between the ease of collection and the expansion path of the SCFs should be reached in designing the particle collection vessel.
Other design factors that merit consideration include: surface finish of the inside of the baskets/vessel, shape of the vessel, alignment etc. , Debenedetti l 993b, Turk 1999

SUMMARY
Current advances in pharmaceutical research have not only contributed to the discovery of various new technologies, but also identified the potential limitations of the conventional techniques of material processing (York 1999] . Among the different nascent technologies currently under investigation, supercritical fluid aided particle formation is reported to operate under relatively mild conditions making the process amenable to sensitive molecules, enzymes, proteins and other macromolecules [Yeo 1993, Moshashaee 2000] . Volatile organic solvents can be 86 reused making their usage minimal. Different SCF processes have been demonstrated to produce particles with residual organic content of an order below the permitted levels [Steckel 1997]. Further, control over the morphology and crystallographic purity of the particles is shown to be better than several other conventionally used processes [Beach 1999]. The potential for SCF technology in the pharmaceutical realm manifests from all the above-mentioned features combined with the feasibility of producing particles under cGMP conditions in a unit operation. The information provided in this article is intended to assist investigative researchers in evaluating such potential either through setting up a particle formation system in house or by contracting the work to established supercritical fluid consultants. concluded from these studies that RESS offers great promise as a hybrid technique to control both the crystalline and the particle morphologies of API's in a single stage. In addition, a number of interesting phenomena were revealed.

LIST OF REFERENCES
These include habit modification, solubility enhancement, particle size reduction, eutectic formation, reduction in crystallinity, amorphous conversion, hydrate formation, polymorph conversion and selective extraction. In viewing each of these phenomena from an application standpoint, this manuscript serves as proof of concept for enhancing the physicochemical and mechanical attributes of API's using supercritical fluid crystal doping.

I. INTRODUCTION
Imperfections prevai l in virtually all solids to varying degrees, resulting in a wide range of material s from almost-perfect crystals to amorphous substances. While the nature of these imperfections can be studied in crystalline substances, the effects on an already disordered amorphous state are rather difficult to isolate [Suga 1997, Suga 1999. The extent and nature of imperfections largely depend on the structural properties of solids, kinetics of crystallization and impurity levels, as well as other crystallization conditions [Weissbuch 2001]. The defects in crystals impart higher localized energies as a result of the elastic strain arising from the reduction in symmetry [Burt 1981, Weisinger 1989 . The higher energy of the system contributed by such pockets, although slightly compensated by increased entropy, is what renders higher free energy to imperfect crystals.
Increased chemical potential and thermodynamic instability of such crystals can have profound implications in a wide variety of pharmaceutical applications.
The utility of impurities in causing crystal disruption is evaluated in this work by controlled co-crystallization of AP! and impurity from supercritical media. Besides modifying the energy of crystals, impurities are also reported to elicit a broad range of effects on the polymorphism, habit, size, true density and surface area of host crystals [Zhang 1999] . The combined effects on the morphology and energetics of the host crystals can be advantageously used in tailoring crystals to pharmaceutical needs and forms the scope of this research.
Such research is of both fundamental and practical relevance. From a theoretical 92 standpoint, the role of impurities on crystal disruption can be studied and can be extended to tailoring additives for specific purposes. From an application perspective, the bulk properties of crystalline pharmaceutical actives can be modified according to their functional utility at early stages of chemical synthesis.

2.A. THEORY OF CRYSTAL DOPING:
Doping is defined as the deliberate addition of an impurity (guest) into the crystallizing medium of the host drug substance. Depending on molecular size and shape, stereochemistry, solubility and chemical affinity towards the host, impurities can profoundly alter the kinetics of nucleation and growth of the host crystals [Rauls 2000, Weissbuch 2001]. To date, various mechanisms have been proposed that typify the impurity-induced effects on the host crystals at both molecular and bulk levels. Firstly, the impurity can function as a co-solute in either enhancing or reducing the solubility of host crystals in the crystallization media. As a consequence of altered supersaturation, the induction time for nucleation and the metastable zone width are modified leading to changes in crystal size, size distribution and habit as observed with acetaminophen doped with p-acetoxyacetanilide [Prasad 2001].
Other means of crystal modification by impurities involves stereoselective adsorption of impurity onto specific faces of a crystal, causing differential inhibition of growth [Addadi l 982]. Inhibition of growth in a specific direction 93 manifests in increased surface area of the face perpendicular to that direction.
Such selective inhibition might lead to modified aspect ratios, a change in habit and in some instances, crystallization of selective isomers and polymorphs. This phenomenon has been illustrated in several host-guest systems such as racemic glutamic acid mixtures, adipic acid+n-alkanoic acids, Benzamide+Benzoic acid, Sucrose+Raffinose, [Weissbuch 2001] Triglycine sulfate + L-alanine [Aravazhi 1997], Phenytoin + 3-acetoxymethyl-5,5-diphenyl hydantoin [Chow 1991], and to report that the guest molecules most commonly exist in a solid solution rather than in liquid inclusions [Zhang 1999].
Depending on the differential rates at which the host and guest are  [Chow 1995b] and reactivity [Duddu 1995] among other biopharmaceutical properties. Given the complexity of the crystallization process and the inadequacy of current analytical techniques to specify the exact location of impurity within lattices, the selection of doping agents has mostly been by trial and error. A direct correlation between the nature of the impurity and its role on the crystallization process is yet to be established, although significant inroads have been made towards this goal [Weissbuch 2001 ]. Habit modification in the host crystals as a result of the incorporation of an impurity translates into changes in such properties as particle size, aspect ratio, density, specific surface area and surface roughness. The effects on particle size are based on the impurity effects on supersaturation and hence the crystallization kinetics, besides being habit related. A decrease in the particle size of phenytoin crystals is reported when doped with 3-propanoyloxymethyl-5,5-diphenyl hydantoin, which the authors attributed to habit thinning [Gordon 1992). Changes in the aspect ratio of host crystals are frequently observed owing to the differential inhibition of growth in specific directions by the impurity. Chow and Grant investigated the influence of p-acetoxyacetanilide on the aspect ratios of acetaminophen [Chow I 989a) and further correlated such influence to the aqueous dissolution rates of acetaminophen [Chow 1989b). True density of the crystals was found to be sensitive to the presence of impurities and is claimed to 97 be a sensitive indicator in quantifying the extent of crystal disruption. The influence of impurities on crystal densities has been experimentally verified using adipic acid/o leic acid and acetaminophen/p-acetoxyacetanilide as host/guest systems. [Duncan-Hewitt 1986] Another habit related property that was shown to be largely influenced by the doping process is specific surface area. Significant increases in the surface areas of acetaminophen [Chow 1985] and phenytoin [Gordon 1992, Chow 1995a were observed when doped with impurities. Part of this enhancement has been attributed to the surface irregularities arising from the dislocation sites during measurements by gas adsorption techniques [Chow 1985].
As a result of crystal doping, surface irregularities have also been reported in few instances that significantly contribute to enhanced dissolution rates [Chow 1991].
A majority of the above mentioned properties are habit dependent. Inducing changes in habit, for exan1ple from acicular prisms to long thin plates as observed in phenytoin [Chow 1991] and from columnar to plate-like in acetaminophen [Prasad 2001] can have potential implications in processes such as wetting, dissolution, compaction etc. Another means of altering the properties of pharmaceutical actives is though the conversion of polymorphs & isomers [Kopp 1989, Laihanen 1996, Bosela 1997, Badawi 1997]. In theory, polymorphism arising from differences in conformation and packing can both be controlled using tailor made impurities. Reported proof of concept studies substantiating this fact include impurity induced crystallization of the polar polymorph of N(2acetamido-4-nitrophenyl)pyrolidene (PAN) [Staab 1990] and a-form of L-98 ( glutamic acid [Sano 1997 in the crystals are stated to be associated with higher localized energies compared to the regions of normal configuration [Burt 1981]. These high-energy pockets are composed of the excess energy resulting from lattice strain and the core potential energy stored in the dislocation sites [Burt 1981, Weisinger 1989. The higher energy of the system contributed by such pockets is slightly compensated by increased entropy of the disordered solids. In effect, impurities thereby render higher free energy to the imperfect crystals. Consequently, an increase in chemical potential and thermodynamic instability results in such crystals. The combined effects of loss of symmetry and increased activity lead to increased wettability, intrinsic dissolution rates and crystal reactivity in general [Chow l 995b]. This has been unambiguously proven using model systems like phenytoin [Chow 1995b) and adipic acid [Chan 1989] as hosts and a number of structurally related impurities as guests. Dissolution enhancement utilizing such subtle crystal modifications appears particularly attractive in the wake of recent amorphization efforts of a number of active pharmaceuticals [Yu 200 l]. 99

2.C. CHARACTERIZATION OF CRYSTAL DOPING:
The various techniques for evaluating the nature and magnitude of crystal disruption can be broadly classified into ones that characterize modifications in crystal morphologies and others that quanify the crystal energetics. Among the spectroscopic and microscopic techniques that study the primary morphological changes following crystal doping include optical microscopy [Burt 1981,Chow 1985,Prasad 2001, SEM [Chow 1991, Gordon 1992, Chow 1995a, Prasad 2001, atomic force microscopy [Li 2000], single crystal x-ray diffraction, [Bettinetti 2000, Williams-Seton 2000, Prasad 2001, Lynch 2000, Atencio 2000, Foxman 2001) powder x-ray diffraction [Burt 1981, Chow 1985, Gordon 1992, neutron X-ray diffraction, [Weisinger-Lewin 1989], IR [Aravazhi 1997, Bondar 2000 and solid state NMR [Yatsenko 1997, Bauer 2001, Gustafsson 1998] . Secondary manifestations that are sensitive to morphology changes such as density [Duncan-Hewitt 1986, Chow 1991) and thermal expansivity [Duncan- Hewitt 1986] are also used as indicators in evaluating crystal disruption. As reported by Burt [Burt 1981 ), Chow [Chow 1985] and Prasad [Prasad 2001] in their studies involving doped potassium perchlorate and acetaminophen crystals, optical microscopy aids in characterizing the aspect ratios, habit and dislocation sites such as etch pits in the doped crystals. In addition, changes in birefringence of the doped crystals can also be studied using polarizing optical microscopy.
While gross structural changes are easily detectable using this technique, subtle crystal modifications are rather difficult to study and require further sensitive 100 techniques such as x-ray topography, scanning electron microscopy and atomic force microscopy. By contributing to the sensitivity, these techniques can also aid in locating the impurity in the doped crystals. Differentiation of surface adsorption from lattice incorporation of impurities is clearly demonstrated utilizing these techniques [Chow 1991, Gordon 1992, Chow !995a, Prasad 2001, Li 2000].
Whenever growth of sufficiently large crystals is attainable, single crystal X-ray diffraction is most frequently used in typifying the structure of doped  [Pikal 1987). This prompted academicians to develop scales to measure disruption based on the thermodynamic analysis of crystalline solids. Accordingly, two indices were defined namely disruption index (d.i) [York 1986) and excess entropy index (e.e.i). [Pikal 1987).
The dimensionless disruption index compares the disorder created in the solid with that created in the liquid host by incorporation of guest molecules. It is defined as rate of change of the difference between the entropy of the solid and that of the liquid, with respect to the ideal entropy of mixing. For impurity mole fractions (x2) less than 0.05, a plot of the entropy change following solid to liquid transition of the doped crystals (l'>S) versus the ideal entropy of mixing (l'>Smideol) was experimentally found to give a straight line according to the equation: where (b-c) is defined as the disruption index. This behavior has been experimentally verified in several host guest systems [York 1986] and used it to Although the validity of these assumptions is arguable to a degree, the concept of disruption index is simple, of practical interest and experimentally substantiated by various host/guest systems at x2 < 0.05 [Pikal 1987). The values of disruption index in the experimental systems evaluated thus far were found to range from 5 104 to 800 [Duddu 1995]. A correlation between the d.i values and the dissimilarty in the properties of the host and the guest has also been established [York 1986]. strength results in high supersaturation that leads to very high nucleation rates [Mohamed 1989]. The time for nucleation and growth is very limited {typically 10· 5 to 10· 6 seconds), resulting in very small particles [Debenedetti 1993, Turk 1999]. Also, the rapid nucleation and growth aids in locking the impurities into the interest here is in the crystal morphology of the pharmaceutical actives rather than their particle size, a 40ml particle collection vial is best suited for these purposes. Use of a 40 ml glass vial also improved yields by preventing losses from the particle collection typically observed with larger vessels. C02 gas after deposition of the solids was exhausted through a custom filter and passed through lengthy tubing (5meters) prior to feeding to the thermo mass flow meter (Porter Instruments, Hatfield, PA). The gas flow rates were further measured (Infinity Rate Totalizer, Newport Electronics, Santa Ana, CA) over the course of the experiment to get a more reliable estimate of the average C0 2 flow rates though the system. Typical flow rates of C0 2 through the system were between 5-10 SLPM. At the end of each run, yields of the recrystallized materials were recorded and the vials stored in low humidity plastic bags at ambient temperature until further use.
Following the above method, a number of drug-impurity mixtures (Table 10) were recrystallized and the efficiency of SCF aided crystal doping was evaluated.
The supercritical region investigated in these studies included a temperature regime of 45-100°C and pressures between 2000-8000 psi. 3

3.B. 3. Thermogravimetric Analysis:
Thermal decomposition, moisture and residual solvent contents of the recrystallized materials were investigated using Perkin-Elmer TGA-7 at a heating   Tables 12 and 14. Calibrated HP 1100 series LC system equipped with a diode array detector was used in these analyses.       The effects on the crystallinity of pure salicylic acid recrystallized by RESS were investigated using DSC and XRPD. The results of DSC analysis of SCF recrystallized salicylic acid are shown in Figure 14 and summari zed in Table 15.
As can be seen from Table 15  124 expected be the cause for improved yields. In a similar study, the effect of trace amounts of methanol and acetone on the solubility of theophylline+caffeine was also investigated. The results of these studies are summarized in Figure 16. As can be seen from figure 16, acetone appears to significantly enhance the solute uptake by SC C0 2 at lower temperatures. Higher temperatures, on the other hand reduced the solute uptake, perhaps because of a reduction in the solvent density.
The results of the effects of methanol were inconclusive owing to the difficulty in preventing MeOH from condensing in the collection vial during particle formation.

4.B. Eutectic Formation:
Addition of aspirin as an impurity to the crystallization media of salicylic acid resulted in the formation of a low melting mixture. As can be seen from Figure   17, recrystalllization from pure SC C02 at 75°C and 4000 psi formed a low melting mixture that melted at l l 5°C. On the other band, use of SC C0 2 + ethanol at 45°C and 3000 psi as the solvent system resulted in the formation of a similar low melting mixture as a minor component and pure salicylic acid as the major component. Selective salicylic acid crystallization is evident in the latter case.
Although no eutectic formation between salicylic acid and aspirin has been reported to date, similar melting point depressions were observed in these mixtures by Mroso et al. [Mroso 1982

4.D. Hydrate Formation:
In RESS, the saturated supercritical solution in the pre-expansion chamber at a significantly high pressure is rapidly expanded through a micrometering valve into a collection vial at atmospheric conditions. Owing to the large pressure drop across the micrometering valve, Joule-Thompson cooling occurs that has the potential to plug the valve and the lines downstrean1 of it. The micrometering valve is therefore maintained at 1OO-l50°C to compensate for the cooling effect.
The effect of the temperature of the micrometering valve on the particles formed is often disregarded so long as the flow of supercritical solution through it is uniform. An extreme case where this norm does not hold was identified while dealing with the RESS of theophylline+caffeine mixtures. Figure 33 shows the XRPD patterns oftheophylline +caffeine co-crystals produced by the RESS Increase in the temperature of the micrometering valve to l 50°C (Figure 36) also did not allow the conversion to a hydrate, perhaps by raising the temperature of the particles to the extent where no condensation from the atmosphere occurred.

4.E. Polymorph Conversion:
The commercially available polymorph ofTolbutamide is the orthorhombic form I that crystallizes as rectangular prisms [Leary 1981  al's study [Kimura 1999) where polymorph II was produced from a spray dried intennediate (fonn IV).
It is noteworthy that polymorphs III and IV closely resemble fonns I and 11 respectively, with negligible free energy differences within each pair [Rowe 1984) . It is therefore possible that reversible transfonnations between these fonns may occur during the analytical characterization. The XRPD patterns of Tolbutamide+Urea mixtures recrystallized from SC C02 exemplifies this fact where a mixture of fonns II and IV resulted at few extraction conditions (see Figure 38 and Section C, Appendix B). Conversion of fonn I to II following RESS recrystallization was evident even in the presence of urea as impurity. Figures 39 and 40 validate the conversion to fonn 11. In addition, a reduction in the crystallinity of tolbutamide was seen followin g doping with urea (refer to section C for details). Among the other mixtures that also exhibited polymorphic conversions upon recrystallization from SC C0 2 included Tolbutamide+Chlorpropamide. The XRPD results of these mixtures are summarized in Figure 42. As can be seen, both the sulfonylrea compounds in this case were found to undergo polymorphic changes, making the study rather complex. Interestingly, DSC analyses of these mixtures revealed the formation of a low melting composition between these two hypoglycemic agents. (Figure 41). A rational extension to this study would be to test the bioavailability of the low melting composition of this metastable mixture and forms the scope for future research.     The above observations from XRPD analysis are consistent with the thermal behavior of the mixtures, as can be seen from Figure 47B. Interestingly, an intermediary condition was found at [45°C, 8000 psi] where significant amounts of both the components are extracted as can be seen from its diffraction pattern.

XRPD and DSC results of RESS mixtures summarized in
This perhaps led to a significant reduction in crystallinity of the co-crystals ( Figure 47A), which upon subjecting to DSC analysis did not exhibit any melting endotherms (Figure 4 7B). Aspirin+benzoic acid is another such system that exhibited selective extraction of benzoic acid at lower extraction temperatures, while aspirin was preferentially extracted at temperatures higher than 62°C (see figures 48A and 48B). Qualitative analysis of the XRPD and DSC data was performed analogous to salicylic acid+benzoic system discussed above. The DSC analysis in this case, however could not be performed above temperatures higher than 130°C as significant sublimation of the mixtures occurred.
Salicylic acid+indomethacin ( Figure 49) and phenytoin+caffeine ( Figure 50) are two other systems that exhibited selective extraction at most of the supercritical extraction conditions investigated. Preferential extraction of salicylic acid and caffeine occurred at a majority of the conditions from these two systems.
Increases in the amounts of second component in these mixtures resulted in amorphous conversions as was discussed in section C. In summary, the composition of the recrystallization media is not only dependent on the RESS extraction conditions, but also on the relative amounts of drug and impurity in the reaction vessel.

S. CONCLUSIONS
The presence of an impurity in the crystallization medium exhibits varied effects depending on the phase in which it is present prior to nucleation and its affinity to the host relative to the crystallizing solvent. This in turns dictates the rate at which it nucleates and grows in relation to that of the host. The domain of effects that these kinetics dictate on one extreme includes the formation of a solid solution or a solid dispersion of the impurity in the host lattice. On the other hand, selective extraction of each of the components with respect to time can also occur, the extent of which primarily depends on the resolution factor of the recrystallizing solvent. While the former mechanism is largely aided by the rapid nucleation and growth implicit to supercritical fluid recrystallizations, the latter forms the scope of supercritical fluid chromatography. An optimal compromise between these 175 extremes can be reached by utilizing the adjustable so lvent power of supercritical fluids. This hypothesis was tested utilizing a number of host/guest systems and SC C02 as the recrystallizing medium. In this process, various interesting phenomena were identified (Table 17).
The presence of aspirin as an impurity was found to alter the habit of salicylic acid crystals from avicular to fibrous form. Supercritical fluid recrystallization herein provided the independent ability to change the crystal habit while not alteri ng the polymorphic form of the APL On the other hand, the polymorphic conversion to a metastable form of tolbutamide was seen upon SCF recrystallization. Doping tolbutamide with urea not only promoted such conversion, but also induced a reduction in the overall crystallinity. Loss of crystallinity in an already existing metastable form can be expected to enhance the dissolution rates, and thereby the bioavailability of the otherwise poorly soluble tolbutamide.
Utilizing a co-solute and a co-solvent to alter the solvent power of SC C0 2 , enhancement in the solid solubility in SC C0 2 was demonstrated in SA+aspirin and theophylline+caffeine systems respectively. In addition, a general reduction in crystallinity was seen in all the doped crystals. This manifested as a reduction in the heat of fusion values, melting point depressions and eutectic formation in salicylic acid+aspirin, salicylic acid+benzoic acid, aspirin+benzoic acid, tolbutamide+ chlorpropamide. In the drug-impurity systems that did not permit the use of thermal analysis, crystallinity was evaluated based on the XRPD studies. Consistent broadening and shifts of XRPD peaks were seen in aspirin+benzoic acid, tolbutamide+urea and naproxen+u-naphthalene acetic acid co-crystals, reiterating a loss in crystallinity. While crystal doping resulted in such reductions in crystallinity for the most part, extreme situations were also identified where major loss of crystallinity and amorphous conversion ensued. For   size reduction of about an order magnitude was seen following RESS processing.
In providing the ability to control both the particle and crystal morphology of AP!s, RESS proved potentially advantageous to crystal engineering.  (Figure 51 ). In addition, doping is more controllable with a small molecule such as urea and in theory will reduce the propensity for segregation and associated stability problems.
Chlorpropamide belongs to the sufonyl urea class of oral hypoglycemics.
It is known to be practically insoluble in water and belongs to class n of biopharmaceutical classification (BCS). Five different polymorphs of CPD are identified to date, of which three are most commonly referred to in the published literature [Burger 1975, Aal-Saieq 1982, Simmons 1973, De Villiers 1999. As is often the case with APls exhibiting multiple conformations, the nomenclature of various fom1s of CPD is very confusing. For the purposes of consistency, the notation defined by Simmons [Simmons 1973] is used in this study. Even after the three decades since it was discovered, it is interesting to note that polymorphism supersaturation that leads to very high nucleation rates [Mohamed 1989). The time for nucleation and growth is very limited (typically 10· 5 to 10-6 seconds), resulting in very small particles [Debenedetti 1993, Turk 1999. Also, the rapid nucleation and growth aids in locking the impurities into the crystal domains of the hosts by not providing sufficient time for the impurities to segregate. Absence of residual liquid solvents in the RESS produced crystals further reduces the possibility for segregation effects in the solid state.
In addition, the rapid decompression of SCF generates mechanical perturbation within the solution that travels at the speed of sound [Debenedetti 1993]. Consequently, very uniform conditions are reached within the nucleating media. Uniform conditions in the nucleation media assist in homogenous dispersion of impurities in the crystal domains of the hosts. The crystal disruption following such uniform and rapid co-crystallization can be expected to be controlled and large [Burt 1981 ]. All the above factors contributed to the special interest in RESS aided crystal doping and formed the rationale for its choice.
Further, the concept is fairly nascent as reflected by the number of SCF aided crystal doping studies reported in the published literature [Weber 1997, York 1995].
The commercially available supercritical fluid extraction equipment (SFT150, Supercritical Fluid Technologies Inc., Delaware) was reconfigured to produce co-crystals of drugs and impurities by the rapid expansion of supercritical solution process. The modified design for the RESS process is schematically represented in Figure 8 and shown in Figure 9  here is in the crystal morphology of the pharmaceutical actives rather than their particle size, a 40 ml particle collection vial is best suited for these purposes. Use of 40 ml glass vial also improved yields by preventing losses from the particle collection typically observed with larger vessels. C02 gas after deposition of the solids was exhausted through a custom filter and passed through lengthy tubing (5 meters) prior to feeding to the thermo mass flow meter (Porter Instruments, Hatfield, PA). The gas flow rates were further totalized (Infinity Rate Totalizer, Newport Electronics, Santa Ana, CA) over the course of the experiment to get a more reliable estimate of the average C0 2 flow rates though the system. Typical flow rates of C0 2 through the system were between 5-10 SLPM. At the end of each run, yields of the recrystallized materials were recorded and the vials stored in low humidity plastic bags at ambient temperature until further use.
Following the above method, Chlorpropamide + Urea mixtures were recrystallized and the efficiency of SCF aided crystal doping was evaluated. The supercritical region investigated in these studies included a temperature regime of 45-100°C and pressures between 2000-8000 psi. The yields of co-crystals extracted at pressure less than 4000psi Were very low to perform further characterization work and hence are not reported in this manuscript.

2.B. 2. Crystallization from Organic Solvents:
The role of the rapid crystallization kinetics in the supercritical fluid aided crystal doping can be best evaluated by comparing the doped crystals from RESS process to the ones crystallized at a much slower rate, for example by evaporative recrystallizing solvents, which are reported to closely correspond to supercritical C0 2 in their solubility behavior [Hyatt 1984, Dandge 1985, Dobbs 1987]. In Sapphire and Indium were used as calibration standards. The instrument was within the calibration period at all times when the reported analyses were performed.  was performed at 5-10 kV, 20 pA probe current, 100-4000x, and a working distance of 6-9 mm. Calibration is performed annually by LEO associates for morphological use only.

2.B. IO . HPLC Analysis:
The amounts of host and guest in chlorpropamide+urea co-crystals were assayed by liquid chromatography (LC). Isocratic, reversed phase LC separation methods were developed and validated using external standards following modification in USP method for the hosts. Specific details of the method are summarized in Table   21, while the representative chromatogram of these mixtures is shown in Figure   52. A calibrated HP 1100 series LC system equipped with a diode array detector was used in this analysis. Absence of chromophores in urea limited its detection by the diode array detector in both the UV and visible ranges. The separation however was accomplished by HPLC method and the interference from urea in the detection of CPD was established to be negligible. In the validation of this method of analysis, stock solutions of CPD and urea in the mobile phase were made at a concentration of 0.05% w/v each. Known aliquots of these solutions were then mixed to obtain varying proportions ofCPD+Urea in the final mixture.
These solutions were then subjected to HPLC analysis and the calibration curve was developed. Similar procedure was repeated at two other dilutions of stock 199 (  The amounts of urea in a mixture can thus be confidently calculated from the CPD levels in the mixture.

3.A. POLYMORPHISM IN CHLORPROPAMIDE:
The commercially available form of CPD is obtained by crystallization from ethano l-water mixture and is called form-A. It is the most thermodynamically stable fom1 at room temperature and has the lowest dissolution rate. Melt recrystallization of this form results in polymorph-B, which is monotropic with fom1-A (Yu 1995). This form is unstable at all the temperatures and is stated to convert into form-A through multiple transformations [De Villerrs 1999).
Polymorph C is obtained by heating form-A at l 20°C for 4 hours. Taken as a pair, fom1s A and C are enantiotropically related, form C being the thermodynamically stable form at higher temperatures while form A is stable at lower temperatures.
The transition temperature for this conversion, however is not reported to date.
DeVilliers and Wurster determined the heats of solution of these fom1s in DMF at 25°C [De Villerrs 1999). The difference in the heats of solution between these forms was found significant (-4kJ/mol) in this study. This did not reflect in the DSC analysis, which did not show any endotherms corresponding to this heat of 203 transition even at heating rates as low as 0.5°C/min ( Figure 54). Conversion fo r form-A to form-C was seen during the analysis, that did not permit calculation of individual melting data for these forms. The thermal behavior of polymorph A was therefore investigated in detail. As can be seen from Figure 54, subjecting form-A to DSC analysis at different heating rates revealed some interesting results. The endotherm at 12 l °C corresponds to the melting of form-A , while the one at l 29°C to that of form-C. The transition from A to C was found to occur gradually with increase in temperature and was most rapid at temperatures nearing the melting point of A. Apparently, the melting endotherms overlap, making the heat of fusion values for these polymorphs indeterminable. Efforts at reso lving these melting endotherms by reducing the heating rates revealed intermediate recrystallization that was hidden before at higher heating rates.
Thermomicroscopy, simulating the heating ramps used in the above DSC analysis revealed that the transition from A to C is rapid around melting point of A, but does not necessarily occur from the melt. Apparent change in the particle morphology was seen upon gradually heating from 100°C to l 20°C. TGA analysis of polymorph A (Figure 55a and 55b) did not indicate any weight loss around this temperature, excluding the possibility of solvent/water mediated transition. From the discussion above, it can be stated that the transition from A to C occurs in a solid state with no change in the composition of the solid. It is hoped that the intermediary recrystallization exotherm can be separated as a kinetic event, utilizing the modulations in heating by mDSC.  indicates that a reversible transformation between forms A and C occurs just before form-A melts. Calculation of the heat of fusion value of form-A , to be utilized in evaluating its crystallinity is therefore not possible by direct DSC analysis. For the purposes of this study, this value was calculated from the LiHr of form C in a manner analogous to Behme and Brooke' s study [Behme 1991 J. Unlike in the case of carbamazepine, the transition from one polymorph to other however did not occur during the DSC analysis of Chip. For the purposes of this study, this heat of transition was estimated to be 4 kJ/mol (or 14.45 Ji g), the difference in the heats of so lution values reported by DeVilliers at 25°C. This estimation was made on the basis that a linear relationship exists between the heat of solution and the heat of fusion for the same polymorph with fixed chemical structure [Yoshihashi 2000]. Such an estimation is further supported by Hess Law that states that the energy associated with a transition depends on the final states and is independent of the path. Assuming that this difference is constant over the temperature range of 30-I 20°C, the heat of fusion of form-A can be estimated to be 85. 77+ 14.45= 100.22 Jig. This assumption was validated when the heat capacity values (Cp) of polymorphs A and C were found to vary similarly in this temperature range (notice the parallel baselines for various polymorphs in     DSC analysis, given that these polymorphs can be produced in pure form and no concurrent phase changes occur during their thermal analysis. The melting data for polymorphs A, B and C, following the above discussion is summarized in Table 22.
Two other means of characterizing the various polymorphs of Chip were also developed. The first used polarizing optical microscopy. As can be seen from  Table 23 . Also, the thermodynamic data useful in evaluating the crystallinity of these polymorphs in the later doping studies is developed.

3.B. RESS OF PURE CHLORPROPAMIDE:
Pure Chip was recrystallized from varying RESS conditions shown in Table 24.
As can be seen from Figure

Simmons, Canad ian Joumal of Pharmaceutical Sciences, 8(4), 1973
Burger, Sc i. Pharm 1975Saieq, Phann Acra Helvetica, 57(/). 1982 (Table 25). Of interest here in view of enhancing the dissolution performance is the formation of polymorph C (Figure 64).While complete polymorph conversion from A to C was seen at certain conditions, the original form remained at other extraction temperatures and pressures (Table 25). The polymorphic identity of the RESS recrystallized materials was positively confirmed from their XRPD data ( Figure 65). On the other hand, thermal behavior of RESS recrystallized materials as determined by DSC exhibited inconsistency with the XRPD results in certain cases. The melting temperatures, however exactly match Burgers polymorphs denoted by the Roman nomenclature (Table   25) [Burger 1975]. XRPD data for these polymorphs has not been reported in Burgers study and hence no definitive matches can be made.
The results of RESS recrystallizations of pure Chip indicate the ability of the RESS process to form different polymorphs from the same solvent by mere changes in the temperature and pressure conditions. The polymorphic conversion from form A to C can be explained based on the individual effects of temperature and pressure on the Chip crystallites during their nucleation and growth. The effect of temperature on this conversion was addressed in detail in section A. This conversion upon recrystallization from SC C02 is consistent with the reported effects of temperature and compression pressure during the tabletting of Chlorpropamide [Matsumoto 1995]. It appears from these studies that these forms   67f). While a tabular habitat can be seen in the commercially available material, all the RESS recrystallized samples attained the shape of blades that is typical of form-C. Consistent with the XRPD results, Chlp recrystallized from selective RESS conditions contained both the forms A and C reflected as a mix of tabular and plate like crystals. (Figure 67-b, c, d). Also, the particle size reduction was significant at the 75°C condition, perhaps due to higher supersaturations attained at this temperature versus the 60°C condition. As can be seen from Figures 67a to   67f, submicron to few micron sized particles were produced by RESS recrystallization.

3.C. RESS OF CHLORPROPAMIDE+ UREA:
The presence of urea in the crystallizing medium of Chlorporpamide reduced the yields as can be seen from Table 26. The solubility of Chip appears to be significantly higher than urea in SC C02 at the various conditions studied. An apparent reduction in the overall yields can therefore be expected in the presence   (Ford 1977) and is shown in  (Tables 29 and 30).
Two mechanisms are proposed that caused this consistent broadening and shifts in the XRPD peaks and illustrated in Figure 69. Firstly, as shown in mechanism-I , urea may have been adsorbed onto selective faces of the crystals of Chip that apparently changed the way it packs. This leads to altered symmetry and increased mosaic spread mirroring in the manner in which different planes reflect x-rays. Apparently, peak broadening and a shift in the XRPD peaks is evident.
Another fact that further validates this mechanism is the preferential crystallization of polymorph C in the presence of urea.     Published single crystal data however is only available for form A [Koo 1980] and such studies could not be performed for polymorph C, which is frequently formed in these studies. The evidence of crystal disruption was also confirmed by the lowering of the melting points and the heat of fusion values of the doped crystals compared to pure crystal of the same polymorph (Table 28). Melting point reductions up to 9°C were seen upon doping with urea. Also, significant reductions in the LI.Hr values of Chip up to 50% were seen as a result of doping with urea. By imparting a strain in the lattices of chlorpropamide crystals that was observed in XRPD results, urea may have reduced the symmetry in the original 231 ( crystals and hence a reduction in the heat of fusion values were seen. Such reductions manifest in significant increases in the initial dissolution rates owing to the ease with which the solvent can destroy the crystal structure for subsequent dissolution. Following the log-linear relationship observed between these entities by Yoshihashi [Yoshihashi 2000], projected enhancement in the initial dissolution rates can be expected to be significant. Scanning electron microscopy of the doped crystals indicated surface adsorption of urea onto Chip crystals. Also, the particles appeared severely agglomerated owing to the use of a smaller collection vial. Given that the interest here is in the crystalline morphology of the RESS produced crystals of Chip, no exhaustive attempts were made to restore the microcrystals formed by SCC0 2 from agglomeration. To prove the concept of agglomeration arising from the bouncing of particles coming at high velocities into the collection vessel ( 40 ml), a larger collection vessel was used (I L). Owing to the altered dynamic of jet expansion in this case, agglomeration was significantly reduced (compare Figure   70f versus Figures 70b-e). The SEMS shown in figure 70f indicate that the particle size of the primary RESS produced particles is in the range of 1-2 µ while that of the starting material was around I 0 µ. A particle size reduction of up to an order of magnitude was therefore produced upon RESS processing.
The efficiency of RESS process in doping was evaluated by direct comparison to the doped crystals produced from liquid organic solvents.    66.95(2.22) work. The utility of rapid co-crystallizations using the RESS process was tested for these purposes. Toward this objective, three different means of characterizing the various polymorphs of Chip were developed in this study. Following polarizing optical microscopy, it was found that polymorph A crystallizes in tabular habit, while the metastable forms B and C appear as blades and plates respectively. The major XRPD diffraction peaks distinguishing the various polymorphs were also identified. Thirdly, the melting data useful in evaluating the crystallinity of these polymorphs was developed following thermal analysis by DSC. In summary, the results reported in this manuscript reflect the potential for RESS aided crystal doping in not only controlling the crystallinity levels in AP!s, but also tailor the polymorphism and particle morphology.  4. Burger A. In German. Sci. Phann 1975, 43:pl52-161. 5. Burt HM, Mitchell AG. Crystal defects anddissolution. International Journal of Pharmaceutics 1981, 9: pl37-152. 6

Abstract:
Purpose. To study the phase behavior of selected pharmaceutical sol ids as a function of temperature, pressure and composition of the supercritical solvent. (

INTRODUCTION
Control over supercritical fluid (SCF) based crystallization processes depend on the knowledge of the mechanism of solute nucleation, supersaturation levels and the phase of the solution from which solutes nucleate and grow [Turk 1999

DESIGN OF PHASE MONITOR
A Phase Monitor provides direct, visual observation of materials under SCF conditions, which may be controlled precisely. Depending on the supercritical fluid under consideration and the range of data desired, several designs of the phase monitors can be used, ranging from a simple Jerguson gauge to a complex diamond anvil cell  """'

Sample preparation:
About 50 mg of sample was introduced into the sample holder of the Phase Monitor. C0 2 from the tank was introduced and allowed to equilibrate at the set temperature for 5 minutes. Using the syringe pump, C02 was pressurized to the maximum value. Rapid depressurization of the supercritical solution from this stage allowed formation of crystals onto the quartz window and in the direct view of the camera.

Co-solvent addition:
Predetermined amount of a co-solvent was pumped using a liquid metering pump and co-introduced along with C0 2 . Mixing of the two fluids was affected in a low dead volume-T as they are delivered into the syringe pump. (

Qualitative Observations
The reproducibility of the solvent effects of SCC02 upon repeated pressurization and depressurization was initially confirmed. This was stru1ed with broad sweeps up and down in pressure hundreds of psi to characterize the significant events.
Following this, the oscillat ions were attenuated with each pass until the degree of resolution was attained and the events were then recorded. In instances where the occurrence of an event is gradual , the onset and the endpoints are noted and the event was continuously recorded over a broad pressure range. A similar procedure was then repeated at a different temperature setting.

RESULTS AND DISCUSSION
The phase behavior of salicy lic acid in SC C02 as a function of temperature and pressure is shown in Figure 75. Solid salicylic acid can be seen as needles at 75°C, 2300 psi condition. As can be seen from the figure, continuous pressurization of the supercritical solvent resulted in the dissolution of these crystals. Complete dissolution of the crystals at 75°C was evident when the pressure reached 4300 psi. Similar trend was seen at 100°C, although the pressure needed to completely dissolve the salicylic acid in this case was found to be 2700psi. Figures 76 and 77 represent the results from a simi lar study to identify the temperature and pressure conditions required for complete dissolution of chloramphenicol and chlorpropamide respectively. The optimum conditions for the maximum solubi lity of the various solids studied are summarized in Table 32. [75°C, 2300psi) [75"C, 2700psi) [

5.A. Design of Equipment:
I) The view of the camera is too narrow and appears to represent only about 1 percent of the total volume in the view cell.
2) Areas where solids would tend to settle, dissolve and recrystallize/ reprecipitate typically are outside the scope of the camera. The design of the view cell needs to be corrected so that the camera is directed towards a larger area where solids would tend to be deposited. 2) Introduce the solute into the view cell. The current setup does not allow for accurate placement of quantitative amounts of solute, which is required to get quantitative data out of this instrument. If the system can be redesigned so that an exact weight of solute can be placed in the view of 258 the camera, semi quantitative investigations of solute solubilities as functions of pressure and temperature of SCF solvent can be made.

CONCLUSIONS
By continuously tuning the solvent power of SC C02, regions were located that exhibited enhanced solubilization of the solids tested. Upon depressurization from these regions, a characteristic turbidity was seen that subsequently lead to the recrystallization of solids. A general trend of increased solubilization with increasing pressures was seen in all the cases.
Qualitative observations from such phase behavioral studies can assist in choosing the optimum extraction conditions for subsequent RESS processing.
The Section C compiles the thermal analysis data by differential scanning calorimetry.
The data from XRPD analyses is divided into two subsections, DI and D2. While the XRPD patterns in each analysis are shown in DI, the analyzed data demonstrating peak shifts and peak broadening is tabulated under D2. Broadened peaks in this subsection are shown in bold font and peak shifts in italicized font.