<<

SUBCRITICAL AS A TUNABLE FOR PARTICLE ENGINEERING

By

Adam Carr, B.E. (Chemical Engineering)

A thesis submitted to the School of Chemical Engineering in partial fulfillment of the requirements of the degree:

Doctor of Philosophy (PhD)

The University of New South Wales

May 2010

ABSTRACT

In the research described in this thesis, an environmentally friendly technique for engineering the morphology of particles was developed. Active pharmaceutical ingredients (APIs) were used as model compounds to explore the potential of the new particle engineering technology. APIs were chosen because tuning the size and shape of drug particles can have pharmacokinetic benefits. The new technology used subcritical water (SBCW) as a solvent to dissolve APIs. By rapidly cooling a SBCW-API solution, the dissolved drug was rapidly precipitated, often with a narrow particle size distribution. Furthermore, it was shown that the morphology of the precipitated particles can be changed by altering several process variables.

In order to develop a new precipitation technology, fundamental solubility data were required. Solubility data were collected for the model APIs; budesonide, griseofulvin, naproxen and pyrimethamine in SBCW from 100°C to 200°C. To ensure that the solubility results were reliable, data were also collected for anthracene, which were compared to published SBCW solubility data. The solubility of budesonide in SBCW was low. Organic were added to the SBCW-budesonide solution to increase the solubility of the API. A model that correlated the solubility of the solute in SBCW with the dielectric constant of the solvent was developed. Model outputs were within 5% of the experimental solubility values.

The solubilities of the APIs were also modelled using a state of the art model available in the published literature. The Modified UNIversal Functional Activity Coefficient (M-UNIFAC) model was used to predict the solubility of budesonide, griseofulvin, naproxen and pyrimethamine in SBCW. It was shown that some of the parameters in literature were not sufficient to describe the data. Model error is reduced significantly when the model parameters were optimized.

The potential of the SBCW micronization process to produce particles for drug delivery via inhalation has been assessed. Budesonide was micronized under a model precipitation condition, the suspension of API in water was then spray dried, and the aerodynamic particle size was established using an Anderson Cascade Impactor. It has been shown that, when coupled with an effective drying process, particles suitable for inhalable drug delivery may be produced.

i

“I keep asking that the God of our Lord Jesus Christ, the glorious Father, may give you the Spirit of wisdom and revelation, so that you might know Him better.”

Ephesians 1:17

“I, wisdom, dwell together with prudence; I possess both knowledge and discretion”

Proverbs 8:12

For my father and mother

ii

ACKNOWLEDGEMENTS AND THANKS

I wish to thank my principle academic supervisor, Professor Neil Foster, for his guidance and support throughout my PhD. He has been very generous with his professional and personal time, and has encouraged me to grow as a researcher through many different hurdles. His ability to see opportunity in what most people would see as a dead end continues to surprise and inspire me.

I would also like to thank my co-supervisor, Dr. Raffaella Mammucari. Her day-to-day insight, knowledge and critical thinking kept me on my toes, and helped me to develop many of the research skills I now possess. Perhaps most of all, I have enjoyed the way that Raffaella has made herself available to chat about both academic and personal at any time of the day, making working in the lab a delight.

I would like to thank all of my lab colleagues; particularly Roshan, Danh and Wendy. Their friendship through what can often be a tough environment has been liberating and very much appreciated. Roshan’s good humor was especially appreciated, particularly when the outlook seemed bleak.

I would very much like to thank my family: Greg, Lynda, Jason and Brittany and my fiancée Jessica for both the emotional and directional support that they have given me throughout the course of my PhD. A special thanks goes to my mother for praying for me continually, and my father whose wisdom guided me into doing a PhD in the first place.

Finally, I would like to thank my God and Heavenly Father, who has directed me onto a path I would never have thought likely, and for blessing me with this insight into His creation. I thank him for the hope that He gives me through my Lord Jesus Christ, which enables me to overcome barriers and hardships, to endure suffering, and to continue to walk on the road He has laid before my feet.

iii

LIST OF PUBLICATIONS

Patents

Subcritical water Processing of Hydrophobic Pharmaceuticals, Combined patent with Professor Neil Foster and Dr. Raffaella Mammucari (Filed for provisional patent on 25th March 2010)

Articles

Carr, A., Mammucari, R. and N. Foster, The Solubility and Micronization of Griseofulvin using Subcritical Water. Industrial & Engineering Chemistry Research, 2010 (accepted February 2010)

Carr, A., Mammucari, R., and N. Foster, The Solubility, Solubility Modeling and Precipitation of Naproxen from a Subcritical Water Solution. Industrial & Engineering Chemistry Research, 2010 (Accepted August 2010)

Carr, A., Mammucari, R., and N. Foster, The Solubility and Solubility Modelling of Budesonide in Pure and Alcohol-Modified Subcritical Water Solutions, Journal of Supercritical Fluids, 2010 (Accepted July 2010)

Carr, A., Mammucari, R., and N. Foster, Micronization of Budesonide using Pure and Alcohol Modified Subcritical Water Solutions, (Submitted to the Journal of Pharmaceutical Science, August 2010)

Carr, A., Mammucari, R., and N. Foster, Subcritical Water as a Tunable Solvent for Organic Compounds (Submitted to the Journal of Chemical Reviews, August 2010)

Conference Papers

Adam G. Carr, R. Mammucari, N.R. Foster, “Subcritical Water as a Universal Solvent and Antisolvent for Active Pharmaceutical Ingredients”, International Symposium of Supercritical Fluids, Arcachon, France, May 2009

Adam G. Carr, R. Mammucari, N.R. Foster, “Subcritical water as an Innovative Green Fluid for Manufacturing Micronized Biomaterials”, ICONN 2008 meeting, Melbourne, Australia, January 2008

Adam G. Carr, R. Mammucari, N.R. Foster, “Subcritical Water as a Novel Fluid for Pharmaceutical Processing Medium”, AICHE Annual Meeting, Salt Lake City, Utah, USA, Nov. 2007

iv

TABLE OF CONTENTS

Abstract ...... i

Acknowledgements and Thanks ...... iii

List of Publications ...... iv

1. Thesis Introduction ...... 1

1.1. The Pharmacokinetics and Therapeutic Action of Selected APIs ...... 2

1.1.1. Pharmaceutical Action and Micronization ...... 2

1.2. Micronization Techniques ...... 5

1.2.1. ‘Top-down’ Methods...... 5

1.2.1. ‘Bottom-up’ Methods ...... 7

1.2.2. Micronization ...... 8

1.2.3. Limitations of Rapid Precipitation Techniques ...... 9

1.2.4. Subcritical Water as an Alternative Solvent for Micronization ...... 10

1.3. Summary ...... 11

1.4. References ...... 12

2. Water as a Tunable Solvent for Hydrophobic Organic Compounds ...... 16

2.1. Introduction ...... 17

2.2. The Solubility of Organic Compounds in Subcritical Water ...... 19

2.2.1. The Influence of Solvent Conditions ...... 22

2.2.2. The Influence of the Solute ...... 24

2.3. Modeling the Solubility of Organic Compounds in Subcritical Water ...... 30

2.3.1. Empirical Models ...... 30

2.3.2. Hansen Solubility Parameters ...... 32

2.3.3. Solubility Model Based on Pure Component Properties ...... 32

2.3.4. Modified UNIversal Functional Activity Coefficient (M-UNIFAC) Model ...... 33

2.3.5. Discussion on Modelling SBCW Solubility Data ...... 34

2.4. Subcritical Water Extractions of Therapeutic Compounds ...... 35

v

2.4.1. Extraction ...... 35

2.4.2. Optimum Extraction Conditions ...... 40

2.5. Subcritical Water as a Solvent for Rapid Precipitation of Active Pharmaceutical Ingredients ...... 42

2.6. Conclusions ...... 43

2.7. References ...... 45

3. Solubility Studies ...... 51

3.1. Introduction ...... 52

3.2. Experimental Method ...... 52

3.2.1. Addition of Organic Solvents as Modifiers to SBCW ...... 57

3.3. Experimental Considerations ...... 57

3.3.1. Equilibration Time ...... 57

3.3.2. Drying Method ...... 58

3.3.3. Solute Quantification ...... 59

3.3.4. Oven Fluctuations ...... 59

3.4. Analytical Techniques ...... 60

3.4.1. Infrared Spectroscopy ...... 60

3.4.2. Differential Scanning Calorimetry ...... 60

3.5. Results...... 60

3.5.1. Chemical Stability of the Solutes in Subcritical Water ...... 60

3.5.2. Solubility ...... 62

3.5.3. The Influence of Dielectric Constant on Solubility...... 66

3.6. Conclusions ...... 69

3.7. References ...... 70

4. Solubility Modelling ...... 73

4.1. Introduction ...... 74

4.2. Thermodynamic Framework ...... 75

4.2.1. Estimation of Activity Coefficients ...... 80

4.2.2. Division of Functional Groups for the Solutes ...... 80 vi

4.3. UNIFAC Model Results and Discussion ...... 84

4.3.1. Anthracene ...... 84

4.3.2. Naproxen ...... 85

4.3.3. Griseofulvin ...... 87

4.3.4. Pyrimethamine ...... 87

4.3.5. Budesonide ...... 89

4.4. UNIFAC Model Parameter Optimization ...... 90

4.4.1. Naproxen ...... 91

4.4.2. Griseofulvin and Pyrimethamine ...... 95

4.4.3. Budesonide and 9-anthracenemethanol ...... 96

4.5. Alternative Model Based on the Dielectric Constant of Water ...... 98

4.5.1. Dielectric Constant Model Fits for Other APIs ...... 100

4.6. Modelling Conclusions ...... 102

4.7. References ...... 103

5. Particle Formation ...... 106

5.1. Particle Engineering Using SBCW as a Solvent ...... 107

5.2. Methods ...... 108

5.2.1. Materials ...... 108

5.2.2. Injection into Cold Water Particle Formation Method ...... 108

5.2.3. Precipitation in the Presence of Excipients ...... 110

5.2.4. SBCW-Solute Spray Particle Formation Method ...... 111

5.2.5. SEM and LS Methods ...... 112

5.2.6. XRD Method ...... 113

5.2.7. DSC Method ...... 113

5.2.8. Spray Drying Method ...... 113

5.2.9. Aerodynamic Particle Sizing Method ...... 114

5.3. Results...... 116

5.3.1. Precipitation into a Cold Water-Filled Vessel ...... 116

vii

5.3.2. Precipitation with Co-solvents ...... 125

5.3.3. Precipitation into a Solution Containing Excipients ...... 131

5.3.4. Precipitation into a Solution with Excipients and Co-solvents ...... 135

5.4. The Drying of Therapeutic Formulations ...... 141

5.4.1. Precipitation by Spraying a SBCW-Solution into a Vacuum Chamber ...... 141

5.4.2. Spray-drying of Budesonide Formulations ...... 143

5.4.3. Spray Drying Results ...... 146

5.4.4. Overall Process Efficiency ...... 149

5.5. Conclusions ...... 151

5.6. References ...... 152

6. Conclusions and Recommendations ...... 157

6.1. Conclusions ...... 158

6.1.1. Development of A Novel, Environmentally Friendly Particle Engineering Technique ...... 158

6.1.2. Establishment of Chemical Stability, Solubility Data and Solubility Models for APIs in SBCW ...... 159

6.2. Recommendations ...... 160

6.2.1. Fundamental Studies ...... 160

6.2.2. Particle Formation Application Studies ...... 161

7. Appendix A ...... 163

8. Appendix B...... 180

8.1. Thermodynamic Development ...... 181

9. Appendix C ...... 182

9.1. Introduction ...... 182

9.2. Naproxen Solubility data ...... 182

9.3. Extrusion Experiment ...... 183

9.3.1. Result ...... 183

viii

LIST OF FIGURES

Figure 1.1: Pharmaceutical pathway from dosage to clinical response. Based on Figure 4-1 of [2] ...... 3 Figure 1.2: Dissolution rate of micronized (median size 2µm) vs. non-micronized ibuprofen in phosphate buffer solution at pH 6.3. Reproduced from [5] ...... 4 Figure 1.3: Unit operations involved in synthesizing a pharmaceutical, including primary and secondary processes ...... 6 Figure 1.4: Comparison of steps involved in a conventional micronization process and a supercritical fluid micronization process. Image reproduced from [17] ...... 8 Figure 1.5: Schematic of micronization process using subcritical water as a solvent for the compound to be micronized and water at ambient temperature as the antisolvent...... 11 Figure 2.1: Comparison of the temperature-sensitive dielectric constant of water with the dielectric constants of different solvents at room temperature at saturated . Data was taken from the public dielectric constant release from the International Association for the and Steam (IAPWS)[9] ...... 18 Figure 2.2: Solubility of anthracene (◊)[17], p-terphenyl (□)[17] and 1,8-Cineole (Δ)[13] in subcritical water. Trend lines are added as a guide to the eye. The dielectric constant(-) was constructed from literature data[9] ...... 23 Figure 2.3: SBCW solubility of anthracene, experimental results from Miller et al.[12] using the MF-UNIFAC model, modified by Fornari et al.[28]. Model constructed in Excel 2007 using literature data[28, 34-35]...... 34 Figure 2.4: The solubilization and precipitation of griseofulvin from subcritical water solutions[56] ...... 42 Figure 3.1: Schematic of the solubility apparatus ...... 53 Figure 3.2: Filter setup at the ends of SV ...... 55 Figure 3.3: Design of the solubility apparatus a) stirrer crankshaft, b) internal disc magnet and c) overall vessel design showing the crankshaft with the ring magnet held outside SV ...... 56 Figure 3.4: TGA curve of naproxen samples produced at 160°C and subjected to drying in the oven ...... 58 Figure 3.5: Comparison of the solubility data produced from measuring the mass of budesonide and water using gravimetry (■) and UV analysis (♦) ...... 59 Figure 3.6: Comparative FTIR spectra of a) griseofulvin, b) naproxen, c) pyrimethamine and d) budesonide in both raw and processed states ...... 61

ix

Figure 3.7: Solubility of anthracene in SBCW from our results (◊) and Miller et al. (▪)[1]. The lines serve as guides to the eye...... 62 Figure 3.9: Solubility of budesonide in SBCW with different volume fractions of . Room temperature solubility data was from [28] ...... 67 Figure 3.10: Solubility of budesonide in pure and ethanol-modified SBCW...... 67 Figure 4.1: Chemical structure of anthracene ...... 82 Figure 4.2: Chemical structure of naproxen ...... 82 Figure 4.3: Chemical structure of griseofulvin ...... 83 Figure 4.4: Chemical structure of pyrimethamine ...... 83 Figure 4.5: Chemical structure of budesonide ...... 84 Figure 4.6: Anthracene solubility and solubility prediction using the M-UNIFAC the MF-UNIFAC and the AF-UNIFAC models ...... 86 Figure 4.7: Naproxen solubility and solubility predictions using the M-UNIFAC, MF-UNIFAC and AF-UNIFAC models ...... 86 Figure 4.8: Solubility and solubility predictions for griseofulvin using the UNIFAC, MF-UNIFAC and AF-UNIFAC models ...... 88 Figure 4.9: Solubility and solubility predictions for pyrimethamine using the M-UNIFAC, MF- UNIFAC and AF-UNIFAC models ...... 88 Figure 4.10: Solubility and solubility predictions of budesonide using the UNIFAC and MF- UNIFAC models ...... 89 Figure 4.11: Solubility of naproxen (Δ) and the MF-UNIFAC model (-) with optimized carboxylic interaction parameters ...... 92 Figure 4.12: Solubility of fatty in SBCW as published by Khwujitjaru et al. [24] and predicted by the MF-UNIFAC model with original COOH interaction parameters ...... 94 Figure 4.13: Solubility of fatty acids in SBCW as published by Khwujitjaru et al. [24] and predicted by the MF-UNIFAC model with updated COOH interaction parameters ...... 94 Figure 4.14: Solubility of griseofulvin and pyrimethamine and the MF-UNIFAC model with optimized chlorine-water interaction parameters ...... 96 Figure 4.15: Solubility of budesonide and the MF-UNIFAC model with optimized OH-water interaction parameters ...... 97 Figure 4.16: Chemical structure of 9-anthracenemethanol ...... 98 Figure 4.17: M-UNIFAC, MF-UNIFAC and Dielectric Constant models of budesonide ...... 99 Figure 4.18: Experimental and predicted solubility data for budesonide. Solubility predictions were calculated from Equation 4.20 with the parameters A and B fitted from 0% to 20% ethanol/SBCW solubility data. The line for the solubility model was added as a guide to the eye ...... 100 x

Figure 4.19: Solubility data and Dielectric Constant model of solubility for a) naproxen, b) griseofulvin, c) pyrimethamine, d) anthracene and e) 9-anthracenemethanol ...... 101 Figure 5.2: Subcritical water particle formation apparatus ...... 109 Figure 5.3: Spray apparatus for the formation of dry particles from SBCW solutions ...... 112 Figure 5.4: Standard setup of the Andersen Cascade Impactor ...... 114 Figure 5.5: Particle size distribution of griseofulvin produced by SBCW micronization from saturated SBCW solutions at different ...... 117 Figure 5.6: Griseofulvin precipitated in water at 20bar from saturated SBCW solutions at: a) 140°C b) 160°C and c) 170°C...... 118 Figure 5.7: Griseofulvin precipitated from SBCW solutions at 170°C and different concentrations: a) 5.3 × 10-4 mol/mol, b) 2.6 × 10-4 mol/mol and c) 1.3 × 10-4 mol/mol...... 120 Figure 5.8: X-ray diffraction of: a) bi-pyramidal crystals from SBCW processing, b) needle-like crystals from SBCW processing and c) raw griseofulvin...... 121 Figure 5.9: DSC diagrams for griseofulvin crystals: a) bi-pyramidal crystals from SBCW processing b) needle-like ...... 122 Figure 5.10: Precipitates of naproxen at experimental conditions a) N1 and b) N3 ...... 124 Figure 5.11: SEM of raw budesonide (a), budesonide processed using conditions B1 (b), B2 (c), B3 (d) and B4 (e) ...... 128 Figure 5.12: Particle size distributions of budesonide processed using experimental conditions B1, B2, B3 and B4 according to Table 5.4 ...... 129 Figure 5.13: DSC curves of raw and processed budesonide ...... 130 Figure 5.14: XRD patterns of raw budesonide and budesonide processed in 20% ethanol/SBCW solutions ...... 131 Figure 5.15: Precipitates of naproxen at experimental conditions a) N4 and b) N6 ...... 132 Figure 5.16: DSC of naproxen precipitated from SBCW without lactose (- - -) and with lactose (-) ...... 133 Figure 5.17: XRD patterns of naproxen produced at different processing conditions and unprocessed naproxen...... 134 Figure 5.18: Budesonide precipitated according to experimental conditions a) B3A, b) B3B, c) B2A, d) B2B , e) B1A and f) B4A described in Table 5.6 ...... 136 Figure 5.19: Budesonide precipitated according to the experimental conditions summarized in Table 5.6 ...... 137 Figure 5.20: Budesonide precipitated according to conditions a) B2C, b) B3D, c) B4C and d) B1D as described in Table 5.6 ...... 139 Figure 5.21: Budesonide precipitated according to the experimental conditions summarized in Table 5.6 ...... 140 xi

Figure 5.22: Precipitate collected from SBCW-naproxen solution spray into a vacuum vessel . 142 Figure 5.23: Particle size distribution of naproxen precipitated by the SBCW-spray technique at 160°C ...... 143 Figure 5.24: FTIR spectra of a) budesonide processed at 200°C with multiple injections, b) budesonide processed at 200°C with a single injection and c) raw budesonide ...... 144 Figure 5.25: Particle size distribution of budesonide particles from a budesonide-SBCW solution injected into the precipitation chamber in single and triple injections ...... 145 Figure 5.26: Budesonide precipitated from a 200°C injection temperature into cold water and the dried using the Buchi 290 spray-drier...... 146 Figure 5.27: Raw (◊) and spray-dried (▲) lactose aerodynamic particle sizes...... 147 Figure 5.28: Aerodynamic particle diameter of raw (◊) and processed (⧠) budesonide ...... 148 Figure 5.29: Aerodynamic particle sizes of the raw and spray-dried budesonide from the budesonide-lactose mixture ...... 149

xii

LIST OF TABLES

Table 2.1: Generalized effects of temperature change in water in relation to the polarity of water (as measured by the dielectric constant) and solubility of organic compounds in water ...... 17 Table 2.2: Compounds that have quantified solubilities in subcritical water (between 25°C and 280°C) ...... 20 Table 2.3: Solubility of linear aromatic compounds in SBCW at a between 45 and 50 bar ...... 26 Table 2.4: Solubility of clustered aromatics in SBCW at a pressures between 45 and 50 bar ...... 27 Table 2.5: Solubility of other aromatics in SBCW at pressures between 45 and 55 bar ...... 28 Table 2.6: Solubility and structure of some oxygenated aromatic compounds at pressures between 50 and 55 bar [19] ...... 29 Table 2.7: Errors in modeling the solubility of HOCs in SBCW [12]. Errors were calculated using Equation 2.3 ...... 31 Table 2.8: Therapeutic substances extracted using SBCW ...... 37 Table 3.1: Solubility of budesonide in SBCW at 100°C at 10 and 30 minutes exposure times ...... 58 Table 3.2: Solubility of griseofulvin from 25°C to 170°C ...... 63 Table 3.3: Solubility of naproxen from 25°C to 170°C ...... 63 Table 3.4: Solubility of pyrimethamine from 25°C to 180°C ...... 63 Table 3.5: Solubility of budesonide from 25°C to 170°C ...... 63 Table 3.6: Solubilities and structures of APIs and anthracene in subcritical water, from 25°C to 200°C ...... 65 Table 3.7: Dielectric constant values for temperatures between 0°C and 200°C and 0% and 20% (v/v) ethanol in water solutions ...... 68 Table 4.1: Heat of fusion and temperature for the studied APIs ...... 77 Table 4.2: Subgroup designation for the modelled solutes ...... 81 Table 4.3: Average errors of the solubility models for anthracene in SBCW ...... 85 Table 4.4: Average errors of the solubility models for naproxen in SBCW ...... 85 Table 4.5: Average errors of the solubility models for griseofulvin in SBCW ...... 87 Table 4.6: Average errors of the solubility models for pyrimethamine in SBCW ...... 87 Table 4.7: Average errors of the solubility models for budesonide in SBCW ...... 89 Table 4.8: MF-UNIFAC model of xanthenes, anthrone, xanthene and 9, 10 anthraquinone ...... 91 Table 4.9: Original and fitted water-carboxylic (side-group 11,18) and carboxylic acid- water (side-group 18,11) interaction parameters. The letters a, b and c represent M-UNIFAC interaction parameters ...... 92

xiii

Table 4.10: Original and fitted water-chlorine acid (side-group 11,19) and chlorine-water (side- group 19,11) interaction parameters...... 95 Table 4.11: Original and fitted water-OH (side-group 11,10) and OH-water (side-group 10,11) interaction parameters...... 97 Table 4.12: 9-anthracenemethanol solubility from 25°C to 170°C and MF-UNIFAC and budesonide-optimized MF-UNIFAC (MFB-UNIFAC) model errors ...... 98 Table 4.13: Dielectric Model parameters and errors ...... 100 Table 5.1: Precipitation experiment variables and results ...... 116 Table 5.2: Precipitation methods for crystalline griseofulvin micro-particles ...... 118 Table 5.3: Naproxen experimental conditions and results summary ...... 123 Table 5.4: Budesonide particle formation conditions and results summary ...... 127 Table 5.5: Experimental condition and results of SBCW-naproxen solutions injected into a lactose-water environment ...... 131 Table 5.6: Particle formation conditions and summary of budesonide precipitated into lactose and PEG400 environments using SBCW and co-solvents ...... 138 Table 5.7: Experimental conditions and results of naproxen micronization by injection of SBCW solutions into a heated vacuum chamber ...... 142 Table 5.8: Spray drier experimental conditions ...... 145 Table 5.9: Spray dryer results ...... 145 Table 5.10: Process efficiency in terms of mass loss at each process step ...... 150 Table 7.1: Inputs for the calculation of solubility using the UNIFAC-type models ...... 164 Table 7.2: Griseofulvin and water-specific constants used for the calculation of solubility using the M-UNIFAC model ...... 165 Table 7.3: Interaction parameter data ...... 166 Table 7.4: Calculated values of e(k) for griseofulvin and water, where the numbers are the nominal interaction parameters as described by Table 7.2 ...... 175 Table 7.5: Theta values calculated for griseofulvin and water ...... 176 Table 7.6: s-values calculated for griseofulvin ...... 177 Table 7.7: β values calculated for griseofulvin ...... 177 Table 7.8: R values calculated for griseofulvin ...... 178 Table 7.9: Calculation of activity coefficients for griseofulvin at 160°C ...... 179 Table 9.1: Naproxen mixture mole fractions extracted from SV above 150°C ...... 182

xiv

SYMBOLS AND ABBREVIATIONS

Symbols

am,n , bm,n, cm,n - Interaction parameters, UNIFAC model (dimensionless)

A, B - Constants, dielectric constant model (dimensionless)

- Activity coefficient

dHm - Heat of melting/ Heat of fusion (J/mol)

E - Error (%)

- Dielectric constant (or ) f - Fugacity

P - Pressure (bar)

R - Universal constant (8.314 J/K∙mol)

T - Temperature (°C or K)

Tm - temperature (K)

V, v - Volume (mL)

vk - number of subgroups of type k (e.g. aromatic carbon, ArC)

x2 - Solubility (mole fraction, mol/mol)

X - Activity coefficient correction for hydrogen bonding in the A-UNIFAC model

xv

Abbreviations

A-UNIFAC - Association UNIversal Functional Activity Coefficient model

API - Active Pharmaceutical Ingredient

CCA - Crystal-Co-Agglomeration Technique

CFA - Compressed (supercritical) Fluid Antisolvent process

DMSO - Dimethyl Sulfoxide

DSC - Differential Scanning Calorimetry exp - Experimental

EXP - Exponential function

FTIR - Fourier Transform Infrared spectroscopy

HOC - Hydrophobic id - Ideal

ID - Internal Diameter inf - Infinite

LS - Light Scattering

M-UNIFAC - Modified-Universal Functional Activity Coefficient model

MF-UNIFAC - Modified-Fornari corrected Universal Functional Activity Coefficient model

MWT - Molecular Weight

PAH - Polycyclic Aromatic Hydrocarbon

PC - Product collection/precipitation chamber

RP-HPLC - Reversed High Performance Liquid Chromatography

SAS - Supercritical Antisolvent process

SBCW - Subcritical Water

SBCWE - Subcritical Water Extraction xvi

SD - Standard Deviation

SEDS - Solution Enhanced Dispersion by Supercritical fluids

SEM - Scanning Electron Microscope

SV - Solubility Equilibrium Vessel

TGA -Thermogravimetric Analysis

TOC - Therapeutic Organic Compound

UNIFAC - Universal Functional Activity Coefficient model

XRD - X-ray Diffraction

xvii

1. THESIS INTRODUCTION

Modern pharmaceuticals have had a profound impact on human lifestyle and longevity. By micronizing an active pharmaceutical ingredient (API), particularly a hydrophobic drug, dissolution rates in bodily fluids may be enhanced, which could aid in the distribution of API to the site of drug reception. By minimizing particle size, the effect of drug particles may be maximized, which could reduce required the dosage– reducing cost and potentially the risk of side effects. In this chapter, insight into how pharmacokinetics may be improved by simply micronizing an API, and how micronization of APIs may be accomplished using traditional and modern techniques is described. A discussion of how subcritical water may be used to complement the already well- established field of supercritical micronization techniques is presented.

1

1.1. THE PHARMACOKINETICS AND THERAPEUTIC ACTION OF SELECTED APIS

1.1.1. PHARMACEUTICAL ACTION AND MICRONIZATION

The study of the pharmaceutical action of a drug may be broken up into two sections; pharmacokinetics and pharmacodynamics, as shown in Figure 1.1. Pharmacokinetics is concerned with the drug dosage size, the movement of the drug into the blood and concentration of drug at the receptor site. Pharmacodynamics looks at the pharmacological response of the receptor site to the drug (or its metabolite) and the clinical response that is associated with the therapeutic outcome of the drug. In terms of engineering a drug formulation, pharmacokinetics is dependent on the physical properties of the particles, while pharmacodynamics is dependent on the chemical properties of the drug. As this thesis is primarily concerned with the physical modification of pharmaceutical particles, the focus of this chapter will be on pharmacokinetics, rather than pharmacodynamics.

Drug delivery may be divided into two classes; parenteral, and enteral. The main difference between the two routes are that the parenteral mode involves bypassing the gastrointestinal (GI) tract, whereas the enteral makes use of the GI tract [1]. Thus typical parenteral routes of delivery include; intravenous/intramuscular (injectable), topical (transdermal), intravascular (inhalable) and introduction into body cavities. The majority of these routes are invasive because they rely on a physical penetration to deliver the drug to the body. Enteral routes are considered non- invasive because typically a pill is taken like food and absorbed through the GI tract.

The biggest disadvantage of delivering drugs via the enteral route is the drug delivery variability. The variability is not only from person to person, but also from day to day within the same person [2]. The ability of a drug to be absorbed through the GI tract, or before it, is affected by:

1. The pH/pKa of the drug and/or the environment, whereas the stomach is acidic, the GI tract is nearly neutral 2. The ability of the drug particles to pass through membranes, which in a function of lipid/water partition coefficient, molecular radius/size and the degree of 3. The presence of materials that modulate the drug absorption, including food and alcohol [3] 2

Figure 1.1: Pharmaceutical pathway from dosage to clinical response. Based on Figure 4-1 of [2]

3

It is important to be able to minimize the variability in the absorption of an API so that the therapeutic response of a drug formulation is more predictable. While the pKa and partition coefficient of a drug will be dependent upon the chemical properties of the drug, the ability of a drug to pass through membranes and, depending on the type of drug and the delivery site, the rate of drug absorption can be engineered by manipulating the physical properties of drug particles. The physical properties of the particle that are often manipulated are the size and the shape.

The size of a particle influences nearly all of the particle performance – which includes drug degradation, clearance, flow properties and uptake mechanisms [4]. For example, by simply changing the size of a particle, the rate of dissolution inside the body can be changed. For example, the rate of dissolution of ibuprofen, a drug that is commonly used to treat pain and arthritis, can be increased significantly when it is micronized, as shown in Figure 1.2. For drugs that provide pain relief, rapid drug dissolution is an important factor, as it can determine the rapidity of the onset of action.

25

20

15

Micronized ibuprofen 10 Raw ibuprofen

5 Concentration Concentration (mg/100mL)

0 0 50 100 150 Time (mins)

Figure 1.2: Dissolution rate of micronized (median size 2µm) vs. non-micronized ibuprofen in phosphate buffer solution at pH 6.3. Reproduced from [5]

The manipulation of the size of a particle can affect the route of delivery. For example, in order for a drug to be delivered into the lungs, the maximum particle size needs to be 5µm [6]. The restrictions on the particle size are related to the relative tortuosity of the lung chambers, whereas if a particle is too large the drug will not be able to traverse to the lower lung absorption sites [7]. Conversely, if the particle is too small, it may be exhaled [8]. Thus stringent control over the particle size needs to be achieved. 4

The shape of a pharmaceutical affects the transport properties of a drug, and thus can affect where the drug is delivered into the human body [4]. The shape of a particle is particularly important when a drug is to be inhaled, to be delivered into the lungs [9]. While it is true that the size of a particle will limit the delivery of a drug through the lungs, the ability of the particle to “fly” will determine whether the particle will get to the lungs. Thus control of both the particle size and the particle shape can be crucial to effective drug delivery.

1.2. MICRONIZATION TECHNIQUES

Particle size and shape are typically controlled by micronization techniques. There are a wide range of industrial and experimental techniques that can be employed to micronize compounds. These techniques can be grouped into two categories: top-down and bottom-up methods.

1.2.1. ‘TOP-DOWN’ METHODS

Top-down methods are techniques that comminute larger particles into smaller particles. Conventional top-down size reducing technologies include extrusion, fluidized bed granulators [10], and traditional and jet mills [11]. Often, these methods combine agglomeration and comminution in one step. Many of these methods are capable of producing micron-sized particles with narrow particle size distributions.

The pharmaceutical industry employs both ‘wet’ and ‘dry’ methods to control particle size and mix different ingredients together[12]. Both methods typically have more than 5 unit operations, including mixers, dryers, lubricators, sieves, granulators and mills, as shown in Figure 1.3. These operations combine to reduce large particles into small particles, and mix particles together to create the final pharmaceutical product.

One of the main limitations of top-down methods is that they often generate high shear environments, and are exposed to air [13-14]. Under such conditions there are concerns over the modification of the surface properties of active particles [15]. The surface of the drug is what first comes into contact with the human body and thus surface degradation of a pharmaceutical can hinder the onset of solubility, which can lead to a slower onset of therapeutic action.

5

Figure 1.3: Unit operations involved in synthesizing a pharmaceutical, including primary and secondary processes

6

1.2.1. ‘BOTTOM-UP’ METHODS

‘Bottom-up’ methods typically refer to those methods which produce particles, either crystalline or amorphous, via precipitation of a solute from a solution. These methods are commonly used to produce nanosuspensions and/or nano-scale powders [16]. Many bottom-up methods have the ability to tune particle size by altering process parameters [14]. Supercritical fluid antisolvent techniques are a prime example of a system with a high degree of tunability: where the temperature, pressure, solvent type, nozzle size and a variety of other process variables can be altered to produce distinctly different particle morphologies [13].

There are a number of benefits that bottom-up methods have over top-down methods. One is that the micronization process is more efficient (in terms of unit operations). Figure 1.4 shows the number of steps involved to achieve a micron sized product from both a conventional technique and a supercritical fluid technique. It is possible for bottom-up processes to require only a single step, where conventional methods require at least the five steps to attain a micron- sized product [17]. Secondly, bottom-up methods eliminate the problem of thermal degradation due to high shear forces [18]. The precipitation of particles from the molecular level up to the nano/micro/macro level avoids the high shear forces often involved in reducing granules to smaller size scales.

In order to carry out a rapid precipitation, typically a compound needs to be dissolved in a medium – which can be water, an organic solvent, an acid or a supercritical fluid[14]. A change in temperature, pH or pressure is then imposed to reduce the affinity of the solute for the solvent, which in turn causes a precipitation [19]. The rate at which the particle is formed is determined by the degree of supersaturation brought about by the change in operating conditions, which in turn determines the resulting morphology of the particle [19-20].

Supercritical fluid micronization methods (in which SC-CO2 is used) show great promise in the pharmaceutical processing industry. Supercritical fluid methods have been shown to have a high degree of control over pharmaceutical particle morphology [19]. Furthermore, developments over the last 10 years have demonstrated that supercritical fluid methods can process nearly any pharmaceutical compound currently on the market [18, 21-23]. The ability of SC-CO2 technology to manipulate the morphology of particles in the nano/micro scale has resulted in the scale-up in some of the technologies for particle engineering. Semi-batch pilot- scale plants are operable that can produce 9g of active pharmaceutical compound per run [24].

7

Figure 1.4: Comparison of steps involved in a conventional micronization process and a supercritical fluid micronization process. Image reproduced from [17]

1.2.2. SUPERCRITICAL FLUID MICRONIZATION

Supercritical fluid precipitation techniques are widely lauded as one of the best choices for precipitating pharmaceutical compounds [15]. The low critical temperature of supercritical carbon dioxide (SC-CO2), as well as its benign nature, lends the fluid to be an appropriate solvent for active pharmaceutical ingredients (APIs)[23]. Furthermore, the very fast nucleating times (microseconds and below) can lead to small drug particles in the nano-scale [18].

Supercritical precipitation methods can be categorized as either solvent or antisolvent techniques. The former uses supercritical carbon dioxide solely (SC-CO2) to dissolve or melt a

8 drug or . To precipitate the compound, the solution of API and SC-CO2 is injected through a nozzle into a chamber at low pressure which, due to the rapid change in density and diffusivity of the solvent, leads to a rapid precipitation of the solute. Particle morphology can vary greatly with only slight changes in injection conditions [25]. Rapid precipitation techniques that dissolve a drug in SC-CO2 are referred to as rapid expansion of supercritical solution (RESS) processes. Despite the high degree of tunability and environmentally friendly nature of the RESS process, it is often limited in its applicability to pharmaceuticals, as CO2 is a poor solvent for the majority of the pharmaceuticals available on the market [19].

The supercritical antisolvent (SAS) technique makes use of the poor solvent quality of SC-CO2 for pharmaceutical applications. Typically a drug is dissolved in an organic solvent, such as acetone, and then brought into contact with SC-CO2. The SC-CO2 expands and extracts the solvent from the drug-solvent solution, which induces a rapid precipitation of drug particles [26]. The precipitation is typically controlled by the temperature and pressure of the supercritical fluid, the concentration of the solute in the conventional solvent, the choice of solvent, the size of the injection nozzle, and the flowrate of the injection of CO2/solution into the precipitation chamber [18]. The ability to tune all of these process parameters has led to a variety of different particle morphologies[27]. The best quality of supercritical antisolvent techniques is that they can be used to process most pharmaceutical compounds, as many different organic solvents with a range of polarities can be employed[15].

1.2.3. LIMITATIONS OF RAPID PRECIPITATION TECHNIQUES

Most rapid precipitation techniques rely on organic solvents to dissolve organic compounds. Ideally, rapid precipitation methods should use non-toxic conventional solvents, like water or ethanol to dissolve drug particles. Ethanol and water at room temperature are both highly polar compounds. The polar nature of the fluids limits the applicability to dissolving polar compounds only. Considering that the majority of drugs available to the public are hydrophobic[28], polar solvents have only limited applications for the pharmaceutical processing industry.

Non-polar organic solvents like dichloromethane, and hexane are thus used dissolve APIs and excipients [29-31]. The major disadvantage of using of toxic organic solvents like dichloromethane and methanol for rapid precipitation is that a solvent removal step is required to extract the organic solvent which is absorbed into the drug particles. Solvent removal, when using supercritical fluids, typically requires high volumes of CO2 to flush out the remaining

9 organic solvent. Thus solvent removal is expensive, particularly given the necessarily low levels of trace solvent allowable in the final formulation.

1.2.4. SUBCRITICAL WATER AS AN ALTERNATIVE SOLVENT FOR MICRONIZATION

Subcritical water (SBCW) is typically defined as water that has been heated to between 100°C and 300°C with sufficient pressure applied to retain a liquid state. SBCW has been used extensively as a solvent for the extraction of a variety of foodstuffs and nutraceuticals from natural sources[32]. The reason that SBCW is chosen over other solvents is that the polarity of water can be changed by changing the temperature of the water. The ability to change the polarity of water with temperature means that water may be referred to as a tunable solvent. The tunability of the polarity of water means that water can be used to selectively extract compounds of different polarity at different temperatures [33-34]. The implications of many of the studies carried out on selective extraction are that SBCW may be used to dissolve hydrophobic and hydrophilic compounds by manipulating temperature.

A method in which SBCW is used to micronize hydrophobic compounds is shown schematically in Figure 1.5. By mixing a SBCW solution with water at room temperature, a rapid temperature quench of the solution and a reduction in the polarity of the solvent are obtained. The change in polarity can induce supersaturation of the solution, which can lead to rapid precipitation of the solute. The yield of the precipitation process will be determined by the difference in solubility of the solute in SBCW and at room conditions.

The changing solvent power of SBCW itself is examined in Chapter 2 of this thesis, as well as the applications the fluid has had in extraction technologies. The work described in Chapter 3 provides a quantification of the changing solvent power of SCBW for a number of model APIs, as well as an evaluation of the chemical stability of the APIs at temperatures up to 200°C. In Chapter 4, solubility models are developed from the solubility data obtained in the work described in Chapter 3, as well as the solubility data summarized in Chapter 2. A method by which SBCW can be used as a fluid for precipitation is described in Chapter 5 of this thesis, and a demonstration of how the tailorable dielectric constant of SBCW can be used to rapidly precipitate and micronize APIs is presented.

10

Figure 1.5: Schematic of micronization process using subcritical water as a solvent for the compound to be micronized and water at ambient temperature as the antisolvent.

1.3. SUMMARY

‘Bottom-up’ micronization methods can be effective tools for improving the bioavailability of pharmaceutical compounds. Most bottom-up methods, however, use toxic organic solvents, which can be difficult to remove from the final pharmaceutical formulation. SBCW precipitation may be used as an alternative to antisolvent precipitations, as the tailorable polarity of water lends itself well to dissolving a range of different organic compounds. A comprehensive review of SBCW is presented in Chapter 2. A detailed description of dissolution and precipitation of organic materials from SBCW solutions is presented in Chapter 5.

11

1.4. REFERENCES

1. Hardman, J.G. and Limbird, L.E., Goodman & Gilman’s The Pharmacological Basis of Therapeutics. 10th Edition, McGraw Hill, United States of America, 2001.

2. Pratt, W.B., Taylor, P., and Goldstein, A., Principles of Drug Action: the Basis of Pharmacology. 1990: Churchill Livingstone.

3. Khalafalla, N., Elgholmy, Z.A. and Khalil, S.A., Influence of High Fat Diet on GI Absorption of Griseofulvin Tablets in Man. Die Pharmazie, 1981. 36(10): pp. 692.

4. Champion, J.A., Katare, Y.K. and Mitragotri, S., Particle shape: A New Design Parameter for Micro- and Nanoscale Drug Delivery Carriers. Journal of Controlled Release, 2007. 121(1- 2): pp. 3-9.

5. Charoenchaitrakool, M., Dehghani, F., Foster, N.R. and Chan, H.K. Micronization by Rapid Expansion of Supercritical Solutions to Enhance the Dissolution Rates of Poorly Water- Soluble Pharmaceuticals. Industrial & Engineering Chemistry Research, 2000. 39(12): pp. 4794-4802.

6. Huang, Y.-Y. and Wang, C.-H., Pulmonary Delivery of Insulin by Liposomal Carriers. Journal of Controlled Release, 2006. 113(1): pp. 9-14.

7. Vozone, C.M. and Marques, H.M.C., Complexation of Budesonide in Cyclodextrins and Particle Aerodynamic Characterization of the Complex Form for Dry Powder Inhalation. Journal of Inclusion Phenomena and Macrocyclic Chemistry, 2002. 44(1): pp. 111-116.

8. Fu, J., Fiegel, J., Krauland, E. and Hanes, J., New Polymeric Carriers For Controlled Drug Delivery For Controlled Drug Delivery Followin Inhalation or Injection. Biomaterials, 2002. 23(22): pp. 4425-4433.

9. Maghsoodi, M., Taghizadeh, O., Martin, G.P., and Nokhodchi, A., Particle Design of Naproxen-Disintegrant Agglomerates for Direct Compression by a Crystallo-Co- Agglomeration Technique. International Journal of Pharmaceutics, 2008. 351(1-2): pp. 45-54.

10. Vervaet, C. and Remon, J.P., Continuous Granulation in the Pharmaceutical Industry. Chemical Engineering Science, 2005. 60(14): pp. 3949-3957.

12

11. Nykamp, G., Carstensen, U. and Müller, B.W., Jet Milling--A New Technique for Microparticle Preparation. International Journal of Pharmaceutics, 2002. 242(1-2): pp. 79-86.

12. Bill Bennet, G.C., Pharmaceutical Production; an Engineering Guide, ed. IChemE. 2003: Institution of Chemical Engineers (Great Britain).

13. Tom, J.W. and Debenedetti, P.G., Particle Formation with Supercritical Fluids--A Review. Journal of Aerosol Science, 1991. 22(5): pp. 555-584.

14. Date, A.A. and Patravale, V.B., Current Strategies for Engineering Drug Nanoparticles. Current Opinion in & Interface Science, 2004. 9(3-4): pp. 222-235.

15. Ginty, P.J., Whitaker, M.J., Shakesheff, K.M. and Howdle, S.M., Drug Delivery goes Supercritical. Materials Today, 2005. 8(8, Supplement 1): pp. 42-48.

16. Rabinow, B.E., Nanosuspensions in Drug Delivery. Nat Rev Drug Discov, 2004. 3(9): pp. 785-796.

17. York, P., Strategies for Particle Design using Supercritical Fluid Technologies. Pharmaceutical Science and Technology Today, 1999. 2(11): pp. 430-440.

18. Subramaniam, B., Rajewski, R.A. and Snavely, K., Pharmaceutical Processing with Supercritical Carbon Dioxide. Journal of Pharmaceutical Sciences, 1997. 86(8): pp. 885- 890.

19. Jung, J. and Perrut, M., Particle Design using Supercritical Fluids: Literature and Patent Survey. The Journal of Supercritical Fluids, 2001. 20(3): pp. 179-219.

20. Garside, J., Mersmann, A. and Nývlt, J., Measurement of Crystal Growth and Nucleation Rates. Working Party on Crystallization, ed. E.C.o.C. Engineering. 2002: Institution of Chemical Engineers (IChemE), Davis Building, 165-189 Railway Terrace, Rugby, CV21 3HQ, United Kingdom.

21. Alessi, P., Cortesi, A., Kikic, I., Foster, N.R., Macnaughton, S.J. and Colombo, I., Particle Production of Steroid Drugs using Supercritical Fluid Processing. Ind Eng Chem Res, 1996. 35(12): pp. 4718-4726.

22. Duarte, A.R.C., Gordillo, M.D., Cardoso, M.M., Simplicio, A.L. and Duarte, C.M.M., Preparation of Ethyl Cellulose/methyl Cellulose Blends by Supercritical Antisolvent Precipitation. International Journal of Pharmaceutics, 2006. 311(1-2): pp. 50-54. 13

23. Foster, N., Mammucari, R., Dehghani, F., Barrett, A., K. Bezanehtak, Coen, E., Combes, G., Meure, L., Ng, A. and Regtop, H.L., Processing Pharmaceutical Compounds Using Dense Gas Technology. Industrial and Engineering Chemistry Research, 2003. 42(25): pp. 6476- 6493.

24. Reverchon, E., Marco, I.D., Caputo, G. and Porta, G.D., Pilot Scale Micronization of Amoxillin by Supercritical Anti-solvent Precipitation. Journal of Supercritical Fluids, 2003. 26: pp. 1-7.

25. Debenedetti, P.G., Tom, J.W., Kwauk, X. and Yeo, S.D., Rapid Expansion of Supercritical Solutions (ress ): Fundamentals and Applications. Fluid Phase Equilibria, 1993. 82: pp. 311-321.

26. Yeo, S.D. and Kiran, E., Formation of Polymer Particles with Supercritical Fluids: A Review. Journal of Supercritical Fluids, 2005. 34(3): pp. 287-308.

27. Perrut, M. and Jung, J., Particle Design using Supercritical Fluids; Literature and Patent Survey. Journal of Supercritical Fluids, 2001. 20: pp. 179-219.

28. Lipinski, C., Poor Aqueous Solubility—An Industry wide Problem in Drug Discovery. American Pharmaceutical Review, 2002. 5(3).

29. Martin, T., Bandi, N., Shulz, R., Roberts, C. and Kompella, U., Preparation of Budesonide and Budesonide-PLA Microparticles using Supercritical Fluid Precipitation Technology. AAPS PharmSciTech, 2002. 3(3): pp. 16-26.

30. Rasenack, N., Steckel, H. and Müller, B.W., Micronization of Anti-inflammatory Drugs for Pulmonary Delivery by a Controlled Crystallization Process. Journal of Pharmaceutical Sciences, 2003. 92(1): pp. 35-44.

31. Steckel, H., Thies, J. and Muller, B.W., Micronizing of Steroids for Pulmonary Delivery by Supercritical Carbon Dioxide. International Journal of Pharmaceutics, 1997. 152(1): pp. 99-110.

32. Teo, C.C., Tan, S.N., Yong, J.W.H., Hew, C.S. and Ong, E.S., Pressurized Hot Water Extraction (PHWE). Journal of Chromatography A, 2010. In Press, Corrected Proof. DOI: 10.1016/j.chroma.2009.12.050

14

33. Ibanez, E., Kubatova, A., Senorans, F.J., Cavero, S., Reglero, G. and Hawthorne, S.B., Subcritical Water Extraction of Antioxidant Compounds from Rosemary Plants. Journal of Agricultural and Food Chemistry, 2003. 51(2):pp. 375-382.

34. Ibanez, E., Oca, A., de Murga, G., López-Sebastián, S., Tabera, J. and Reglero, G., Supercritical Fluid Extraction and Fractionation of Different Preprocessed Rosemary Plants. J. Agric. Food Chem, 1999. 47(4): pp. 1400-1404.

15

2. WATER AS A TUNABLE SOLVENT FOR HYDROPHOBIC

ORGANIC COMPOUNDS

The popularity of using subcritical water as a solvent to extract a variety of organic compounds has grown over the last 10 years. A number of review articles have identified subcritical water as an effective solvent, catalyst and reactant for hydrolytic conversions and extractions [1]. Few articles have looked at water solely as a solvent for hydrophobic organic compounds (HOCs) at elevated temperatures. The aim of this chapter is to provide a review of the literature that describes subcritical water purely as a solvent for solubilizing of HOCs. The literature reviewed includes publications on subcritical water extraction, publications on the measurement of the solubility of organic compounds in subcritical water, the use of the solubility data in solubility modelling and a recently developed particle formation technique.

16

2.1. INTRODUCTION

Subcritical water (SBCW, defined as water between 100°C and 300°C [2]), is an effective solvent for both polar and non-polar compounds. The versatility of SBCW as a solvent is related to the tunable polarity of water. The polarity of water is directly dependent upon the temperature. As the temperature of water is increased, the polarity of water decreases. As a result, the solubility of non-polar organics increases, and the solubility of polar organics decreases [3]. A generalization of these effects is shown in Table 2.1.

Table 2.1: Generalized effects of temperature change in water in relation to the polarity of water (as measured by the dielectric constant) and solubility of organic compounds in water

Temperature Polarity Polar Compound Non-polar Solubility Compound Solubility

↓ ↑ ↑ ↓ ↑ ↓ ↓ ↑

The polarity of SBCW is measured by the dielectric constant value [4]. As the temperature of water is raised above 100°C, the dielectric constant of water becomes similar to that of organic solvents, like DMSO, at ambient conditions (Figure 2.1).

The ability to tune the dielectric constant of water to mimic the dissolving power of organic solvents for non-polar compounds has been exploited to selectively extract a large number of hydrophobic organic compounds (HOCs) from plants, soils and foods [5-7]. By achieving low polarities at elevated temperatures, SBCW extraction (SBCWE) technology has produced high extraction yields and fast extraction times for a number of HOCs [8]. SBCWE yields are comparable to techniques that use organic solvents.

17

100

90

80

70 DMSO 60

50 Acetonitrile Methanol 40 Acetone Dielectric Constant Dielectric 30

20 Chloroform

10

0 0 100 200 300 400 500 600 700 800 Temperature (°C)

Figure 2.1: Comparison of the temperature-sensitive dielectric constant of water with the dielectric constants of different solvents at room temperature at saturated liquid pressure. Data was taken from the public dielectric constant release from the International Association for the Properties of Water and Steam (IAPWS)[9]

There is benefit to using SBCW as a solvent for extractions over conventional extraction solvents. Conventional extraction processes require non- or semi-polar solvents to achieve high extraction yields. These solvents are often toxic. Furthermore, rigorous organic solvent removal is often required where the extract is to be ingested as a food or pharmaceutical. Solvent removal is expensive and time-consuming. Water, on the other hand, is ubiquitous, non-toxic and has low disposal costs [2]. SBCW is thus also an ideal candidate for use as a solvent for the engineering pharmaceutical particles for drug delivery.

The properties affecting solute solubility in SBCW are discussed before extraction and particle engineering using SBCW technology is reviewed. Solubility data provides valuable information relating to the mechanisms that govern extraction yields and extraction times [10]. Furthermore, solubility data are needed whenever scale-up of extraction technologies is to be considered [11].

18

2.2. THE SOLUBILITY OF ORGANIC COMPOUNDS IN SUBCRITICAL WATER

The number of publications containing solubility data for HOCs in SBCW has risen over the last decade. The rise in the number of publications is related to the growing popularity of SBCWE and reverse phase high performance liquid chromatography (RP-HPLC) separations. There have been particularly large increases in publications on separations of compounds that are extracted to be used for fragrances and flavors, or as materials to be removed from a matrix due to their toxicity. Solubility data of HOCs in SBCW are summarized in Table 2.2.

The solubility of an organic compound in subcritical water is influenced by the properties of water, the structure of the solid, and the complex interactions between the water and the solid. These factors are reviewed individually in light of recently published solubility data.

19

Table 2.2: Compounds that have quantified solubilities in subcritical water (between 25°C and 280°C)

Temperature Temperature Pressure Notes Reference Compound range (°C) increment (°C) (bar)

Benzo[A]pyrene Propazine First method devised [10] Chlorothalonil 25-250 50 70 for SBCW solubility studies Endosulfan II

Anthracene Pyrene 3rd Order solubility model Chrysene 60 and 400 developed Perylene 25-250 50 [12] Carbazole

d-Limonene Carvone Solubility depression between Eugenol 50°C and 150°C 1,8-Cineole 25-250 50 60 exhibited by 1,8 Cineole [13] Nerol

Atrazine Added ethanol as a co-solvent Cyanazine 100-250 50 50 which increased solubility [2] Simazine of the extracts Caprylic acid Pressure made no difference to Capric acid the solubility in SBCW. Lauric acid solubility of fatty acids Myristic acid 50-250 25 50 and 150 increased dramatically [14] Palmytic acid beyond 150°C Stearic acid

20

Compound Temperature Temperature Pressure Notes Reference range (°C) increment (°C) (bar)

Benzene Toluene m-Xylene 3rd order model from p-Cymene 25-250 50 70 [12] developed with higher [15] Octane accuracy 2,2,4-trimethylpentane Tetrachloroethylene Tetraethyl Tin 1,2 Dichlorobenzene Fluorine Solubility model with pure Fluoranthene 30-125 20 50 component properties [16] developed with higher accuracy than [12, 15] Anthracene 1,2 Benzanthracene 30-260 10,20 50 [17] Triphenylene p-Terphenyl Adamantane New 3rd order Diamantane 30-280 20 50 solubility model developed [18]

Xanthene Anthrone Solubility study carried out on Xanthone compounds with different [19] Thioxanthone 30-250 20 64 oxygen arrangements 9,10 anthraquinone 9,10 phenanthraquinone

21

2.2.1. THE INFLUENCE OF SOLVENT CONDITIONS

The thermodynamic properties of water (at both ambient and subcritical conditions) are typically described in terms of hydrogen bonding strength and hydrogen bonding structure [4]. Hydrogen bonds in water are self associating, in that the strength of one is governed by the presence of other hydrogen bonds grouped around it [20]. Thus a small change to one hydrogen bond affects the entire water volume. Changes in hydrogen bonding strength are reflected in the dielectric constant and heat of values [3]. At lower temperatures, the dielectric constant value is higher; thus the hydrogen bonding strength is stronger [21]. As the temperature of the water is raised, the increased thermal agitation reduces the strength of each hydrogen bond and leads to an amplified reduction in dielectric constant value [22]. The reduction in hydrogen bonding strength within the water molecules and the reduction in water polarity generally leads to an increase in HOC solubility in water.

The decreasing rate of increase in solubility of non-polar organic compounds is not always directly proportional to the dielectric constant value, as shown in Figure 2.2. When a solute is added to the water system, more complex molecular interactions can take place. The complex molecular interactions can cause different solubility trends over a broad temperature range. For most compounds, smooth solubility increases are experienced up to 150oC, followed by larger solubility increases beyond 150oC. However, for compounds such as benzene, toluene, and m-xylene; solubilities decrease between 25°C and 60°C, followed by a large solubility rise beyond 60°C(referred to here as solubility ‘minima)[23]. Larger oxygenated extracts such as 1,8 Cineole demonstrated similar behavior around 100°C, as shown in Figure 2.2 [13].

It has been shown that two dominating processes (positive heat and negative heat) act on the solubility of an organic solute in SBCW. These two forces act in opposition to one another[23]. The positive heat (or driving force) of solubility is provided by the heat of cavity formation. The negative heat in solubility is produced by the formation of hydrogen bonding ‘cages’, or icebergs around the solute molecule [24]. In situations where the formation of icebergs dominate, as the temperature increases the solubility decreases. Similarly, when the solubility increases as the temperature is raised the heat of cavity formation exceeds the heat of iceberg formation.

The temperature range in which solubility minima occur is different for each compound. For example, benzene exhibits a solubility minimum around 60°C and 1, 8 cineole exhibits a solubility minimum around 100°C. With few examples of solubility minima reported in the literature, it is difficult to predict where solubility minima will occur based on molecular shape, size or structure,

22 nor is it possible to predict where a solubility depression will occur based on water properties alone.

600000 100

90 500000

9 80

70 400000 water 60

300000 50

40 constant of 200000 30 Dielectric Solubility (mole fraction) x10 20 100000 10

0 0 0 50 100 150 200 250 Temperature (°C)

Figure 2.2: Solubility of anthracene (◊)[17], p-terphenyl (□)[17] and 1,8-Cineole (Δ)[13] in subcritical water. Trend lines are added as a guide to the eye. The dielectric constant(-) was constructed from literature data[9]

It has been shown that at 160°C and above, the solubility of HOCs begins to increase significantly with temperature [24]. The large increases in solubility have been documented in a number of publications [10, 12, 17]. The rapid increase in solubility demonstrates that the heat of cavity formation has a far greater influence on solubility than the heat of iceberg formation, to the point that the heat of cavity formation is negligible above 160°C[25]. Therefore it may be considered that icebergs no longer form around solute molecules at temperatures of 160°C and above [14].

It should be noted that, while water does form hydrogen bonded clusters in the pure liquid state [26], the icebergs referred to in this work are a phenomenon that occurs only when a solute is present in solution. Furthermore, the disappearance of icebergs in liquid water does not mean that hydrogen bonds are no longer present between water molecules. It has been shown that the hydrogen bonding between water molecules exists up to and beyond the critical point of water [21, 25].

Pressure has a negligible effect on the solubilities of HOCs in SBCW. Typically, as pressure increases the solubility of the organic compound slightly decreases [12]. The magnitude of pressure required to produce a noticeable effect on solubility differs between each compound. For example, the

23 solubility of carboxylic acids decreases by one order of magnitude over a pressure range of 350 bar [14], whereas the solubility of anthracene in SBCW decreases by one order of magnitude over a pressure range of 2800bar [2]. Typically subcritical water extractions and other SBCW processes are conducted at pressures between 20bar and 100bar, where the solubility differences due to pressure changes are negligible [10].

2.2.2. THE INFLUENCE OF THE SOLUTE

The solubility behavior of HOCs in SBCW is dependent upon the degree of conjugation of the aromatic rings, the position of hydrogen bonding side-groups around the HOC, and the presence of different side-groups around the HOC. Solubility data for both the number of solutes studied and the type (such as oxygenated/nitrogenated/chlorinated hydrocarbons, etc.) of solutes studied is limited, though some generalizations on the effect of solute structure on solubility can be made.

2.2.2.1. Aromatic Hydrocarbons

Hydrophobic organic compound solubility typically increases as the solute size decreases. For example, anthracene has a higher solubility than 1,2 benzanthracene and chrysene [12]. A comparison of the solubilities of low molecular weight HOCs with high molecular weight HOCs is shown in Table 2.3.

A higher degree of conjugation in a HOC leads to a higher solubility. For example, p-terphynyl has a solubility that is up to 10 times higher than chrysene, despite the similar molecular weight of both compounds. Furthermore, clustered aromatic hydrocarbons (such as pyrene, see Table 2.4) have a higher solubility than linear aromatic hydrocarbons (such as 1,2 benzanthracene in Table 2.3). The clustered aromatic rings give rise to a cluster of delocalized * electrons, which gives rise to a greater interaction between SBCW and the clustered aromatic hydrocarbon [23].

2.2.2.2. Oxygen and Other Side Groups

The solubility of various oxygenated HOCs is shown in Table 2.6. The presence of oxygen in a solute generally increases the solubility of the solute in SBCW. For example, the solubility of anthrone is 10 times higher than anthracene between 150°C-160°C, because of the extra oxygen on anthrone.

24

The type of oxygen attached to the aromatic chain affects the solubility of a HOC. For example xanthene, which has an ether group, has a lower solubility at 90°C than anthrone, which has a carbonyl group. The presence of both oxygen types leads to a higher solubility than if only one carbonyl or one ether group were present. For example, xanthone contains both the ether and carbonyl oxygen types, and thus the solubility of xanthone is higher than both xanthene and anthrone (Table 2.6).

The position of oxygen side-groups has an effect on the solubility of a HOC. For example, 9, 10 anthraquinone has a lower solubility than anthrone (Table 2.6). Further evidence of the effect of oxygen position can be seen by comparing 9, 10 anthraquinone to 9, 10 phenanthraquinone; compounds with the same molecular weight and molecular formula. The solubility of 9, 10 anthraquinone, which has carbonyl side-groups in an ortho-orientation, is 10 times higher than 9, 10 phenanthraquinone, which has carbonyl side-groups in a para-orientation.

The effect of aliphatic chain length was investigated by examining the solubilities of different fatty acids in SBCW[14]. As the aliphatic chain length increases, the solubility of the decreases. At higher temperatures, it was found that longer chain carboxylic acids experience higher degree of solubility increases per degree of heat increase[14]. Thus the slope of the solubility curve with respect to temperature is steeper for long chain oxygenated aliphatic compounds than short chain oxygenated aliphatic compounds.

With the limited solubility data available relating to side-groups other than oxygen (the data is shown in Table 2.5), only limited observations of the effect of selected side-groups on solubility of HOCs in SBCW can be made. Compounds containing nitrogen tend to give rise to a higher solubility than non-nitrogen containing compounds. For example, carbazole (Table 2.5) has a solubility 7.6 times higher than anthracene (Table 2.3 ) at 200°C, despite having the same molecular weight. The higher solubility of the nitrogen containing compound over the non-nitrogen containing compound is not surprising, as water is a compound that readily forms hydrogen bonds with nitrogen in nitrogen containing compounds (e.g. amines) [27]. The strength of the hydrogen bond is not as strong for nitrogen-containing compounds as it is for oxygenated compounds, and this is reflected in the higher solubility of oxygenated compounds (Table 2.6).

The presence of chlorine lowers the solubility of organic compounds. The solubility of chlorothalonil and endosulfan II are significantly lower than non-chlorinated aromatic hydrocarbons (Table 2.5). It is unknown how the presence of sulphur in a compound affects its solubility in SBCW as there is insufficient data available.

25

Table 2.3: Solubility of linear aromatic compounds in SBCW at a pressures between 45 and 50 bar

Solubility§ (109) Linear Aromatics T (deg C) Naphthalene Anthracene 1,2 benzanthracene p-terphenyl Chrysene 25 8.1 0.63 40 6920 11.9 3.37 45 8600 50 11400 1 55 15500 60 20700 42.2 8.46 0.849 65 26400 70 34800 75 43500 80 137 29.5 3.85 100 457 113 21.9 13 120 1540 418 90 140 1600 372 150 2960 600 160 15900 1540 170 180 48400 6260 190 195 200 130000 24100 15800 210 220000 39300 225 75800

Chemical Structure # aromatic bonds 5 7 9 9 9 #aliphatic chains 0 0 0 0 2 #aromatic rings 2 3 4 4 3 other side-groups 0 0 0 0 0 Molecular weight 128.16 178.22 228.29 230.31 228.29 Reference [17] [12, 17] [17] [17] [12] §Solubility in mole fraction

26

Table 2.4: Solubility of clustered aromatics in SBCW at a pressures between 45 and 50 bar

Solubility§, (109) Clustered Aromatics T (deg C) Triphenylene Pyrene Perylene 25 10.7 0.29 40 1.82 45 50 38 2.1 55 60 6.07 65 70 75 80 23 100 89.9 900 120 120 353 140 1260 150 5000 160 3770 170 180 12300 190 23000 195 28300 200 210 225

Chemical Structure # aromatic bonds 9 8 10 #aliphatic chains 0 0 0 #aromatic rings 4 4 5 other side-groups 0 0 0 Molecular weight 228.29 202.25 252.32 Reference [17] [12] [12] §Solubility in mole fraction

27

Table 2.5: Solubility of other aromatics in SBCW at pressures between 45 and 55 bar

Solubility§, (109) Other Aromatics T (deg C) Carbazole Propazine Chlorothalonil endosulfan II 25 110 6.3 0.18 0.27 40 45 50 450 13.7 0.8 1.1 55 60 65 70 75 80 100 9900 106 28 30 120 140 150 162000 2560 950 720 160 170 180 190 195 200 1900000 26800 23400 4500 210 225

Chemical Structure # aromatic bonds 6 0 3 1 #aliphatic chains 1 4 0 3 #aromatic rings 2 0 1 0 other side-groups N 6N, Cl 2N, 4Cl S,6Cl, 3O Molecular weight 167.08 207.62 265.92 406.92 Reference [10] [10] [10] [10] §Solubility in mole fraction

28

Table 2.6: Solubility and structure of some oxygenated aromatic compounds at pressures between 50 and 55 bar [19]

Solubility§ (109) Compound T (oC) Xanthene Anthrone Xanthone Thioxanthone 9,10 anthraquinone 9,10 phenanthraquinone 40 252 345 709 118 72.5 550 50 471 60 844 1060 1930 313 205 1630 70 1450 80 2570 3180 5440 879 592 4180 90 4270 100 9490 12900 2220 1560 10600 110 120 27900 37000 6440 4100 39400 130 140 98900 91300 19800 10600 84600 150 126000 160 271000 55000 29600 255000 170 180 135000 842000 190 200 1830000

Chemical Structure

Molecular weight 182.22 194.23 196.20 214.43 208.22 208.22 §Solubility in mole fraction

29

2.3. MODELING THE SOLUBILITY OF ORGANIC COMPOUNDS IN SUBCRITICAL WATER

The solubility of a substance is a fundamental property, which can be used to calculate extraction yields. Solubility models should accurately predict how soluble a compound is in a solvent, in this case SBCW, and facilitate the determination of whether a process is viable or not before experiments are conducted. It is also imperative to have solubility models if the process is to be scaled up to an industrial level [11].

SBCW solubility models have been developed since 1998, when limited solubility data were publicly available [12]. Miller proposed that as more data became available, models based on solute and solvent characteristics would become possible [12]. As more solubility data was published, more models were proposed, with varying degrees of accuracy. Many models attempted to model solubility by taking into account pure water properties alone (in terms of hydrogen bond influences), which neglected solute structure. These models tended to be accurate at either high temperature or at low temperature, but not at both. The first model to apply both solvent and solute properties was the modified universal functional activity coefficient (M-UNIFAC) model published in 2008 [28]. The result of the application of solute properties resulted in a model that was accurate at both high and low temperatures.

A brief review of the major solubility models for organic compounds in SBCW developed over the last 10 years is presented in this section. The models are divided into four subgroups (Empirical, Solubility Parameter, Pure Component Property and M-UNIFAC models) and are presented in the chronological order of their publication.

2.3.1. EMPIRICAL MODELS

The work in this section will focus on models that have used solubility data to construct a generalized first, second or third order equation based, and applied to predict the solubility of a compound in SBCW. Three empirical models have been used for the prediction of the solubility of polycyclic aromatic hydrocarbons (PAHs) [12, 15, 29] in SBCW. The basic form of the empirical models is shown in Equation 2.1, where x2(T) is the solubility in mole fraction at temperature T, and x2(To) is the known solubility of the compound at ambient temperature.

Equation 2.1

30

Equation 2.2

Equation 2.3

Miller et al. [12] altered Equation 2.1 to fit a cubic equation to describe the solubility data for anthracene, pyrene, chrysene, perylene and carbazole (Equation 2.2). Equation 2.2 was then used to predict the solubility of benzo[a]pyrene, propazine, chlorothalonil, and endosulfan II. A table of errors is shown in Table 2.7. The error is calculated using Equation 2.2 (x is the solubility and E is the error in %).

Table 2.7: Errors in modeling the solubility of HOCs in SBCW [12]. Errors were calculated using Equation 2.3

Benzo[a]Pyrene Propazine Chlorothalonil Endosulfan II

500C 0% -43% 7% -2%

1000C 0% -39% 67% 55%

1500C -26% -46% 70% 75%

2000C -338% -5% 70% 7%

The Miller model predicted the solubility at lower temperatures well, but failed to predict as accurately at elevated temperatures. It is proposed that the large errors in the model exist for two reasons. The first is that the compounds being modeled were dissimilar (the chemical structures are shown in Table 2.5) to the compounds for which the model was fitted (shown in Table 2.4). It has already been demonstrated that the placement of various side groups on HOCs can have a dramatic effect on the solubility of HOCs. The effect of sidegroup orientation in the HOC on solubility model calculations is reflected by looking at the different magnitudes of error between each compound in Table 2.7.

The other source of error was due to the changing hydrogen bonding properties in water. It can be seen from Table 2.7 that if the model is accurate above 150°C, then it is inaccurate below 150°C. The aforementioned disappearance of hydrogen-bonded icebergs at 160°C leads to different solute-solvent interactions above160°C than below 160°C[24] . These results demonstrate that models that take into account hydrogen bonding interactions between water and the solute are required. SBCW-solute interaction models have been recently constructed which, to varying extents, account for hydrogen bonding between the solute and the solvent.

31

2.3.2. HANSEN SOLUBILITY PARAMETERS

Hansen Solubility Parameters (HSP) have been used to establish the onset of solubility for both polar and non-polar compounds in a variety of solvents. While the method does not predict solubility, it can establish whether a solvent is ‘good’ or ‘poor’ (establishment of whether a solvent is nominally good or poor is done quantitatively) [30].

The HSP method takes into account partial contributions of dispersion, polar and hydrogen bonding forces involved in the dissolution of an organic compound in a solvent [30]. Dispersion forces are calculated from a series of group contribution tables for organic molecules [30-31]. The polar forces are calculated using dipole moment data. The hydrogen bonding forces are accounted for using the heat of vaporization data. The contribution of these three forces places the solute and the solvent in a 3-dimensional space. The relative distance of the solute to the solvent determines how soluble the solute is in the solvent.

The HSP has been used to calculate solubility parameters for SBCW. Srinivas, King and Hansen [31] used the HSP to predict the ability of SBCW to dissolve a number of complex organic compounds, such as catechin and glucose. Detailed information is not available in the literature regarding any hydrogen bond contribution corrections done by Srinivas et al. to the original HSP. It is possible that the model incorporated correction terms to the hydrogen bonding parameter for water specifically, using a correction applied by Hansen and Andersen [32]. The method demonstrated the ability to determine qualitatively whether water was a good solvent with both binary and ternary systems. The inability of the method to calculate an actual solubility value limits the applicability of the HSP model.

2.3.3. SOLUBILITY MODEL BASED ON PURE COMPONENT PROPERTIES

In 2006, a semi-empirical model was developed to calculate the solubility of PAHs in SBCW[16]. The intent of the model was to take into account properties of high temperature water that were indicators of hydrogen bonding interactions to predict solubility. Modelling of the hydrogen bonding interactions was done by using equations that modeled the cohesive energy density (a function of internal energy) and the dielectric constant as functions of temperature. The model did not take into account interaction between the solvent and the HOC.

32

The model was used to calculate solubilities of 17 polycyclic aromatic hydrocarbons such as anthracene and naphthalene from 25°C to 250°C, using only pure water properties and pure solute properties. The model exhibited an (absolute) average error of 44%. Considering that the model predicted solutes that had solubilities which differed by up to 7 orders of magnitude from each other, the correlation was reasonably accurate. Furthermore the model was consistent (in terms of error) from room temperature to 250°C for all of the tested compounds, implying that the errors in calculations, due to poor representation of hydrogen bonding interactions, were no longer present. The model was therefore an improvement over previous empirical models in that it took into account hydrogen bonding at both high and low temperatures, which improved the predictive power.

2.3.4. MODIFIED UNIVERSAL FUNCTIONAL ACTIVITY COEFFICIENT (M- UNIFAC) MODEL

The M-UNIFAC model calculates activity coefficients of materials in solution by taking into account interactions between organic compound side-groups and solvent side groups [33]. The activity coefficients are converted into mole fraction solubilities by using a thermodynamic relationship between the melting point properties of the solute and the solute/solvent fugacity ratio [16]. Recently, Fornari et al. fitted a M-UNIFAC model for a number of HOCs (specifically PAHs) in SBCW[28]. By the time Fornari et al. had constructed the corrected M-UNIFAC model, the solubility of over 17 PAHs had been published (most are shown in Table 2.3 and Table 2.4), which allowed for the correction of the aromatic carbon-water interaction parameters.

The M-UNIFAC model (corrected by Fornari) demonstrated an absolute average standard deviation (AASD) from all published aromatic hydrocarbon solubilities of 4.7%[28]. The improvement in the solubility of the M-UNIFAC model for the PAH anthracene is shown in Figure 2.3.The accuracy of the M-UNIFAC model between room temperature and 200°C is greatly improved over the original UNIFAC model for all PAHs.

33

Figure 2.3: SBCW solubility of anthracene, experimental results from Miller et al.[12] using the MF-UNIFAC model, modified by Fornari et al.[28]. Model constructed in Excel 2007 using literature data[28, 34-35].

Another UNIFAC model was constructed by Fornari that used hydrogen bonding correction (associating) terms (the A-UNIFAC model)[28]. The association term took into account the dielectric constant of water, and has been used when strong hydrogen bonding occurs between a solute and a solvent[36]. The A-UNIFAC model is not as accurate as the M-UNIFAC model.

2.3.5. DISCUSSION ON MODELLING SBCW SOLUBILITY DATA

Over the past decade models have become more accurate with the increasing database of HOC solubilities. As more data becomes available, models will become applicable to a wider range of organic compounds. The M-UNIFAC model, in its current stage of development, can accurately model the solubilities of PAHs by taking into account both hydrogen bonding behavior, and solute-solvent interactions. Corrections to the solute-solvent interaction parameters are relatively simple to make. Thus, as more solubility data becomes available for compounds with similar structures and a variety of side-groups, the M-UNIFAC model may be further improved.

As discussed earlier, some HOCs exhibit solubility plateaus. It is unlikely that generalized solubility models would be able to account for this behavior. Karásek has published a number of

34 models that may be able to account for non-ideal behavior, though the models require a full set of solubility data for a specific compound to be fitted [18-19].

2.4. SUBCRITICAL WATER EXTRACTIONS OF THERAPEUTIC COMPOUNDS

SBCW is a suitable solvent for HOC extraction for a number of reasons, which are elaborated upon in this section. Comparisons between SBCWE and conventional extraction techniques are presented. Comparisons are based on extraction times and extraction yields, which are measures of the success of a technique at obtaining an extract. A discussion is presented on the experimental conditions which affect these yields and extractions times. A thorough review of extraction and chromatographic separations using subcritical water (or pressurized hot water) has recently been published[37]. In the following section seeks to provide an overview specifically related to therapeutic organic compounds is provided.

2.4.1. EXTRACTION

Over 300 papers that describe the use of SBCW as a solvent to extract HOCs have been published. SBCW extractions in earlier years focused on extracting polycyclic aromatic hydrocarbons (PAHs) and metal complexes from soils and other matrices [5, 38-39]. More recently, SBCW has been used to extract therapeutic organic compounds, such as antioxidants, proteins and anti-inflammatory agents from plants and foods [6, 40-43]. The diverse range of compounds extracted using SBCW is listed in Table 2.8.

Thermal degradation is not an issue for extracts like PAHs, which typically are toxins that need to be removed, rather than separated for further use. However, degradation is an issue for therapeutic HOCs, such as the antioxidants found in oregano and coriander. In these cases, activity tests are required to be able to evaluate whether the extraction was a success. The method for determining the activity of an extract varies depending on the extract. For oregano and a host of other herbs and plants, the extracted antioxidants retained their antioxidant activity [7, 41].

The successful extraction of therapeutic compounds from plants and other materials without inducing chemical degradation has attracted focus on the processing of therapeutic compounds for medicinal purposes. Pharmaceuticals have also been dissolved in SBCW at temperatures up

35 to 200°C without degradation to the chemical structures, as tested by H-NMR[44]. SBCW can thus be used as a solvent for structure-sensitive therapeutic compounds that require preservation of their chemical activity up to 200°C.

In the cases studied, SBCW extraction was more rapid than the best conventional method of extraction for that compound [45-47]. The greatest improvement in extraction time was found from the extraction of anti-oxidants from coriander seeds [41]. Conventional methods required at least 3 hours to collect the desired amount of extract, whereas SBCW took 30 minutes. The speed of extraction is related to the higher temperatures used for SBCWE [46] than for the conventional methods.

SBCWE, in general, had higher extraction yields than conventional alternatives, such as soxhlet, hydrodistillation and supercritical fluid extraction methods [45-49]. Of the papers reviewed, only one conventional method had a higher optimum yield than SBCWE; hydrodistillation extraction of essential oils from coriander seeds[41].

36

Table 2.8: Therapeutic substances extracted using SBCW

Optimum Conventional Optimum SBCW conventional optimum yield extraction time Extract Application Method extraction time (w/w, %) Tmax(°C) (mins) Ref.

Damnacanthal Anti-cancer Continuous hot - - 220 45 [40] water flow past noni-roots

Lignans, Natural food Continuous pH- - - 220 100 [43] proteins, component altered hot flow carbohydrates through packing

Essential oils Therapeutic Continuous hot 60mins (CO2) 175 30 [45] from Thymbra supplement in tea water flow past spicata Thymbra spicata 3hrs - 69% leaves (hydrodistillation)

24 hrs (Soxhlet)

Anti-oxidant Nutraceutical Batch hot water 200 30 [7] extract from (therapeutic) extraction across oregano oregano leaves

37

Optimum SBCW Extract Application Method Optimum Conventional Tmax(°C) extraction time Ref. conventional optimum yield (mins) extraction time (w/w, %)

Essential oils Indigestion Continuous hot 3hrs 21.7% from Coriander therapeutic water flow past (hydrodistillation) (hydrodistillation) seed C. sativum seeds 175 120 [41] 24hrs (soxhlet) 19.4% (Soxhlet)

Sweetener from Natural sweetener Batch hot water 24hrs (soxhlet) 5.1% (soxhlet)- 170 5 [46] Siraitia and analgesic extraction past grosvenorii powdered 90mins (sc-CO2) 70% (sc-CO2) substrate

Extracts from Veterinary Batch hot water Ultrasonic animal feeds: antibiotics extractions past extraction: animal feed 120 5 [48] oxytetracycline purchased in 65.4% (OTC) China - tetracycline 52% (TC) chloramphenicol 64.4 (CAP)

38

Optimum SBCW Extract Application Method Optimum Conventional Tmax(°C) extraction time Ref. conventional optimum yield (mins) extraction time (w/w, %)

Saponins, Nutraceuticals Continuous hot - - 160 70 [42] cyclopeptides water flow past cow cockle seed

Anti-oxidants Nutraceuticals, Continuous 200 30 [6, 50] and flavonoids natural flavours extraction flow from rosemary past dried leaves

Lignans from Pectin removal Semi-continuous 3hrs (hot acid 160 1 [47] flavedo of citrus from industrial flow past pectin extraction) fruits waste; jams and extract jellies

39

2.4.2. OPTIMUM EXTRACTION CONDITIONS

The majority of the materials that have been extracted using SBCW do not have published solubility data. Thus it is difficult to determine what effect the solute sidegroups had on extraction yield or extraction time (except for whether the compound was oxygenated or not). Determination of the optimum extraction conditions for SBCWE can only be based on water properties, rather than on water properties and solute chemical structure. Thus, only the influences of extraction temperature, flowrate, pH and pressure are discussed.

2.4.2.1. Temperature

The temperature of the extraction affects the polarity of water, which defines the type of compound that will be extracted. It is possible to selectively tune SBCW to extract a desired compound. Ibanez et al. showed that this was possible for various extracts of rosemary [6]. High concentrations of polar rosemary extracts were present in SBCW between the temperatures of 50°C and 100°C. Between 150°C and 200°C the concentrations of the polar extracts were much lower, and non-polar extracts were present in high concentrations. Thus the optimum extraction temperature, in terms of recovery of extract, will vary depending on the polarity of the extract.

Some compounds will degrade at elevated temperatures. However, degradation of a HOC is not only dependent upon the temperature, but also the exposure time of that compound at that temperature. For example; the extraction of damnacanthal had a higher yield at 170°C (100% extract collected) than at 200°C (27% extract collected)[40]. After 45 minutes at 200°C, the non- extracted compound was completely degraded. Evaluation of the optimum extraction temperature thus needs to take into account the temperature at which hydrolysis occurs [51], as well as quantification of the amount of time the extract may be exposed to this temperature until hydrolysis occurs.

Higher temperatures in SBCWE typically lead to rapid extractions. Studies on retention factor thermodynamics (for SBCW RP-HPLC) have shown a link between the temperature, the extent of hydrogen bonding and the speed of extraction. Coym et al. [52] compared the experimental retention factor behavior of SBCW between 150°C and 250°C to water at room temperature with organic co-solvents. It was shown, in both cases, that a comparable reduction in retention factor was experienced. The reduction in retention factor was directly related to the reduction in hydrogen bonding strength, caused by the high separation temperature or presence of co- solvent [53]. The decreased hydrogen bonding cohesion not only leads to the reduced dielectric

40 constant, but acts to decrease . The reduced surface tension improves solute- solvent contact [54], thereby improving the speed of extraction.

2.4.2.2. Flowrate

The flowrate affects the rate at which a compound is extracted. Most articles have reported that higher flowrates at constant temperature lead to faster extraction times [40, 45].

In terms of efficient water use, higher flowrates may not be optimal if the SBCW-solute product is highly diluted. Guclu-Ustundag et al.[42] reported that low flowrates were more concentrated than high flowrates, whereas Anekpankul et al.[40] showed that the extraction of damnacanthal from noni-roots was equally concentrated in SBCW at all tested flowrates. Similar flowrates and temperatures were used by both groups of researchers, implying that the ideal flowrate will differ for each compound that is extracted. Thus the determination of optimum extraction efficiency must take into account individual compound interactions at different water flowrates.

2.4.2.3. pH

Adjustment of water pH, by temperature change or with a buffer, is important where the extract is soluble at selected pH levels. Some separations are also more rapid when the pH is altered. Teutenberg carried out separations on model anti-cancer drugs using SBCW with added phosphate buffer[55]. The pH was modified to 3.5 and 11.5 using phosphoric acid and potassium , respectively. Optimum separations were achieved at 150°C at a pH of 3.5. Improvement in extraction times by altering the pH has not yet been fully investigated for SBCW extractions.

2.4.2.4. Pressure

Pressure has a negligible effect on the extraction of compounds using SBCW. The dielectric constant is only mildly affected by changes in pressure above 1000 bar[3]. Pressures above 1000 bar are rarely used for extractions and separations with SBCW technology. Ozel et al.[45] confirmed that pressure changes in the extraction of essential oils made no difference to extraction time or yield.

41

2.5. SUBCRITICAL WATER AS A SOLVENT FOR RAPID PRECIPITATION OF ACTIVE PHARMACEUTICAL INGREDIENTS

Recently, SBCW has been used as a solvent for rapid precipitation of active pharmaceutical ingredients (APIs)[56]. The solubility of selected APIs in SUBCW may increase by up to 7 orders of magnitude when the temperature is increased from 25°C to 200°C. The wide solubility gap ensures that high supersaturation ratios can be generated when a SBCW solution is rapidly quenched to room temperature [12, 19]. A visual interpretation of this concept is shown in Figure 2.4.

Figure 2.4: The solubilization and precipitation of griseofulvin from subcritical water solutions[56]

The primary advantage of using SBCW as a solvent for rapid precipitation techniques is to overcome the requirements of using toxic organic solvents to dissolve hydrophobic APIs, which are commonly used in the majority of modern rapid precipitation techniques [57-58]. The significance of such an advance stems from the fact that the majority of the APIs on the market are hydrophobic [59], which means that solvents with dielectric constants equivalent to or

42 higher than ethanol cannot be used to dissolve them. The use of SBCW eliminates the use of any organic solvent, while being able to process a range of hydrophobic APIs[44, 60].

The processing of some APIs with SBCW has produced small crystals with narrow particle size distributions. For example, bi-pyramidal crystals of griseofulvin have been precipitated with an average particle length of 1µm [56], where previously the smallest crystal size by a supercritical antisolvent method was 100 µm [61].

Some observations have been made on the effect of processing conditions on the resulting morphology of the precipitates. Temperature does not have a large effect on the particle morphology[56]. For example, the size and shape of the griseofulvin crystals precipitated were largely unaltered when the precipitations were carried out at 140°C and 170°C. Some experiments have been conducted with potential pharmaceutical excipients, as modifiers of the precipitate morphology. In some cases, the presence of excipient was able to produce spherical plate-agglomerates of naproxen crystals[62]. Such particles have beneficial aerodynamic properties, which are effective for the delivery of APIs to the lungs[63].

While the technology has the potential to micronize pharmaceuticals with narrow particle size distributions, there are a number of limitations. Firstly, the product post-processing is a suspension of drug particles in water. The drying of water from a suspension can be energy intensive, whether it is filtered out or removed by heat. If the latter is the method of choice, the slow evaporation of water may change the resulting morphology of the precipitated API. Secondly, while the technology has been shown to be inert with a number of APIs, degradation can still be an issue, particularly if proteins are intended to be dissolved in SBCW, which are typically more sensitive to heat than small molecular weight materials like naproxen or griseofulvin.

Despite the limitations, SBCW has the potential to eliminate organic solvent usage for micronization of some APIs, and should be investigated as an alternative to other rapid precipitation techniques.

2.6. CONCLUSIONS

Recent breakthroughs in the prediction of HOC solubility have improved solubility predictions for HOCs in SBCW at both high and low temperatures. Model applicability is limited, as no models have been able to account for solubility trends of compounds other than PAHs.

43

Establishment of a more comprehensive databank is still necessary to complete and supplement the current models reported in the literature. An improved understanding of the interactions between SBCW and solute will lead to more accurate solubility models.

SBCW has been demonstrated to be an effective, environmentally friendly solvent for the extraction of HOCs. SBCWE can reduce extraction times by up to 50% of conventional method extraction time. As the temperature of SBCW is increased, the yield increases and extraction time decreases. Extracting therapeutics, such as antioxidants, above 200°C can lead to hydrolysis. However, many extractions have been carried out below 180°C without hydrolysis of the chemical structure of the API, and with higher yields than conventional methods. Thus subcritical water, even at moderate temperatures, is an effective tunable solvent for therapeutic HOCs, and a viable replacement for toxic organic solvents such as acetone.

The replacement of organic solvents has a direct advantage over precipitation techniques that employ organic solvents to dissolve hydrophobic organic compounds. Water acts in a similar way to hydrophobic organic solvents, and as such can be used as a substitute solvent for HOCs. If cold water is used as the antisolvent, a rapid precipitation technique can be used which could eliminate the use of toxic organic solvents altogether. It has been shown that the SBCW- precipitation process can produce particles with narrow particle size distributions, and with crystalline morphologies. The precipitation technique has a future in any industry where small suspensions of hydrophobic materials are needed in water. Thus the technique warrants further research.

44

2.7. REFERENCES

1. Brunner, G., Near Critical and Supercritical Water. Part I. Hydrolytic and Hydrothermal Processes. The Journal of Supercritical Fluids, 2009. 47(3): pp. 373-381.

2. Curren, M.S.S. and King, J.W., Solubility of Triazine Pesticides in Pure and Modified Subcritical Water. Analytical Chemistry, 2001. 73(4): pp. 740-745.

3. Fernández-Prini, R.J., Corti, H.R. and Japas, M.L., High-Temperature Aqueous Solutions: Thermodynamic Properties. 1991: CRC Press.

4. Franks, F., Water. Revised 1st edition ed. 1983: The Royal Society of Chemistry.

5. Jiménez-Carmona, M.M., Manclús, J.J., Montoya, A. and de Castro, M.D.L., Sub-and Supercritical Fluid Extraction of Trichloropyridinol from Soil Prior to Immunoassay. Journal of Chromatography A, 1997. 785(1-2): pp. 329-336.

6. Ibanez, E., Kubatova, A., Senorans, F.J., Cavero, S., Reglero, G. and Hawthorne, S.B., Subcritical Water Extraction of Antioxidant Compounds from Rosemary Plants. Journal of Agricultural and Food Chemistry, 2003. 51(2): pp. 375-382.

7. Rodriguez-Meizoso, I., Marin, F.R., Herrero, M., Senorans, F.J, Reglero, G., Cifuentes, A. and Ibanez, E., Subcritical Water Extraction of Nutraceuticals with Antioxidant Activity from Oregano. Chemical and Functional Characterization. Journal of Pharmaceutical and Biomedical Analysis, 2006. 41(5): pp. 1560-1565.

8. Herrero, M., Cifuentes, A. and Ibanez, E., Sub- and Supercritical Fluid Extraction of Functional Ingredients from Different Natural Sources: Plants, Food-by-products, Algae and Microalgae: A Review. Food Chemistry, 2006. 98(1): pp. 136-148.

9. IAPWS. Release on the Static Dielectric Constant of Ordinary Water Substance for Temperatures from 238K to 873K and Pressures up to 1000MPa. 1997; Available from: http://www.iapws.org/.

10. Miller, D.J. and Hawthorne, S.B., Method for Determining the Solubilities of Hydrophobic Organics in Subcritical Water. Analytical Chemistry, 1998. 70(8): pp. 1618-1621.

11. McHugh, M.A. and Krukonis, V.J., Supercritical Fluid Extraction. Vol. 2. 1994: Elsevier.

45

12. Miller, D.J., Hawthorne, S.B., Gizir, A.M. and Clifford, A.A., Solubility of Polycyclic Aromatic Hydrocarbons in Subcritical Water from 298 K to 498 K. Journal of Chemical & Engineering Data, 1998. 43(6): pp. 1043-1047.

13. Miller, D.J. and Hawthorne, S.B., Solubility of Liquid Organic Flavor and Fragrance Compounds in Subcritical (Hot/Liquid) Water from 298 K to 473 K. Journal of Chemical & Engineering Data, 2000. 45(2): pp. 315-318.

14. Khuwijitjaru, P., Adachi, S. and Matsuno, R., Solubility of Saturated Fatty Acids in Water at Elevated Temperatures. Biosci. Biotechnol. Biochem, 2002. 66(8): pp. 1723-1726.

15. Mathis, J., Gizir, A.M. and Yang, Y., Solubility of Alkylbenzenes and a Model for Predicting the Solubility of Liquid Organics in High-Temperature Water. Journal of Chemical & Engineering Data, 2004. 49(5): pp. 1269-1272.

16. Karásek, P., Planeta, J. and Roth, M., Solubility of Solid Polycyclic Aromatic Hydrocarbons in Pressurized Hot Water: Correlation with Pure Component Properties. Industrial & Engineering Chemistry Research, 2006. 45(12): pp. 4454-4460.

17. Karasek, P., Planeta, J. and Roth, M., Solubility of Solid Polycyclic Aromatic Hydrocarbons in Pressurized Hot Water at Temperatures from 313 K to the Melting Point. Journal of Chemical & Engineering Data, 2006. 51(2): pp. 616-622.

18. Karásek, P., Planeta, J. and Roth, M., Solubilities of Adamantane and Diamantane in Pressurized Hot Water. Journal of Chemical & Engineering Data, 2008. 53(3): pp. 816- 819.

19. Karásek, P., Planeta, J. and Roth, M., Solubilities of Oxygenated Aromatic in Pressurized Hot Water†. Journal of Chemical & Engineering Data, 2009: pp. 294-301.

20. Nezbeda, I. and Pavlicek, J., Application of Primitive Models of Association: A Simple Theoretical of Water. Fluid Phase Equilibria, 1996. 116(1-2): pp. 530- 536.

21. Nakahara, M., Matubayasi, N., Wakai, C. and Tsujino, Y., Structure and Dynamics of Water: From Ambient to Supercritical. Journal of Molecular , 2001. 90(1-3): pp. 75-83.

22. Caffarena, E.R. and Grigera, J.R., On the Hydrogen Bond Structure of Water at Different Densities. Physica A: Statistical Mechanics and its Applications, 2004. 342(1-2): pp. 34- 39.

46

23. Bohon, R.L. and Claussen, W.F., The Solubility of Aromatic Hydrocarbons in Water1. Journal of the American Chemical Society, 1951. 73(4): pp. 1571-1578.

24. Shinoda, K., "Iceberg" Formation and Solubility. The Journal of Physical Chemistry, 1977. 81(13): pp. 1300-1302.

25. Kalinichev, A.G. and Bass, J.D., Hydrogen Bonding in Supercritical Water. 2. Computer Simulations. Journal of Physical Chemistry A, 1997. 101(50): pp. 9720-9727.

26. Frank, H.S., Covalency in the Hydrogen Bond and the Properties of Water and Ice. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences, 1958. 247(1251): pp. 481-492.

27. Marten, B., Kim, K., Cortis, C., Friesner, R., Murphy, R., Ringnalda, M., Sitkoff, D. and Honig, B., New Model for Calculation of Solvation Free Energies: Correction of Self- Consistent Reaction Field Continuum Dielectric Theory for Short-Range Hydrogen-Bonding Effects. J. Phys. Chem, 1996. 100(28): pp. 11775-11788.

28. Fornari, T., Stateva, R.P., Señorans, F.J., Reglero, G. and E. Ibañez, Applying UNIFAC-based Models to Predict the Solubility of Solids in Subcritical Water. The Journal of Supercritical Fluids, 2008. 46(3): pp. 245-251.

29. Pongnaravane, B., Goto, M., Sasaki, M., Anekpankul, T., Pavasant, P. and Shotipruk, A., Extraction of Anthraquinones from Roots of Morinda Citrifolia by Pressurized Hot Water: Antioxidant Activity of Extracts. The Journal of Supercritical Fluids, 2006. 37(3): pp. 390- 396.

30. Hansen, C.M., Hansen Solubility Parameters, A Users Handbook. 2000: CRC Press.

31. Srinivas, K., King, J.W. and Hansen, C.M., Prediction and Modeling of Solubility Phenomena in Subcritical Fluids Using An Extended Solubility Parameter Approach. ACS-AIChE National Meeting, Spring April 6-10, 2008, New Orleans, Louisiana 2008.

32. Hansen, C.M. and Andersen, B.H., The Affinities of Organic Solvents in Biological Systems. American Industrial Hygiene Association Journal, 1988. 49(6): pp. 301-308.

33. Aage Fredenslund, Jones, R.L. and Prausnitz, J.M., Group-contribution Estimation of Activity Coefficients in Nonideal Liquid Mixtures. AIChE Journal, 1975. 21(6): pp. 1086- 1099.

47

34. Gmehling, J., Li, J. and Schiller, M., A Modified UNIFAC Model. 2. Present Parameter Matrix and Results for Different Thermodynamic Properties. Industrial & Engineering Chemistry Research, 1993. 32(1): pp. 178-193.

35. Magnussen, T., Rasmussen, P. and Fredenslund, Aa., UNIFAC Parameter Table for Prediction of Liquid-liquid Equilibriums. Industrial & Engineering Chemistry Process Design and Development, 1981. 20(2): pp. 331-339.

36. Ferreira, O., Macedo, E.A., and Bottini, S.B., Extension of the A-UNIFAC Model to Mixtures of Cross- and Self-associating Compounds. Fluid Phase Equilibria, 2005. 227(2): pp. 165- 176.

37. Teo, C.C., Tan, S.N., Yong, J.W.H., Hew, C.S. and Ong, E.S., Pressurized Hot Water Extraction (PHWE). Journal of Chromatography A, 2010. In Press, Corrected Proof., DOI: 10.1016/j.chroma.2009.12.050

38. Priego-López, E. and Luque de Castro, M.D., Demetalisation of Soils by Continuous Acidified Subcritical Water Extraction. Talanta, 2002. 58(2): pp. 377-385.

39. Fernández-Pérez, V., Jiménez-Carmona, M.M. and Luque de Castro, M.D., Continuous Liquid–liquid Extraction using Modified Subcritical Water for the Demetalisation of used Industrial Oils. Analytica Chimica Acta, 2001. 433(1): pp. 47-52.

40. Anekpankul, T., Goto, M., Sasaki, M., Pavasant, P. and Shotipruk, A., Extraction of Anti- cancer Damnacanthal from Roots of Morinda Citrifolia by Subcritical Water. Separation and Purification Technology, 2007. 55(3).

41. Eikani, M.H., Golmohammad, F., and Rowshanzamir, S., Subcritical Water Extraction of Essential Oils from Coriander Seeds (Coriandrum Sativum L.). Journal of Food Engineering, 2007. 80(2): pp. 735-740.

42. Guclu-Ustundag, O. and Mazza, G., Extraction of Saponins and Cyclopeptides from Cow Cockle Seed with Pressurized Low Polarity Water. LWT - Food Science and Technology, 2007. 80 (2): pp. 1028.

43. Ho, C.H.L., Cacace, J.E. and Mazza, G., Extraction of Lignans, Proteins and Carbohydrates from Flaxseed Meal with Pressurized Low Polarity Water. LWT - Food Science and Technology, 2007. 40 (9).

48

44. Smith, R.M., Cheinthavorn, O., Wilson, I.D., Wright, B. and Taylor, S.D., Superheated Heavy Water as the Eluent for HPLC-NMR and HPLC-NMR-MS of Model Drugs. Analytical Chemistry, 1999. 71: pp. 4493-4497.

45. Ozel, M.Z., Gogus, F., and Lewis, A.C., Subcritical Water Extraction of Essential Oils from Thymbra Spicata. Food Chemistry, 2003. 82(3): pp. 381-386.

46. Xia, Y., Rivero-Huguet, M.E., Hughes, B.H. and Marshall, W.D., Isolation of the Sweet Components from Siraitia Grosvenorii. Food Chemistry, 2008. 107(3): pp. 1022-1028.

47. Ueno, H., Tanaka, M., Hosino, M., Sasaki, M. and Goto, M., Extraction of Valuable Compounds from the Flavedo of Citrus Junos using Subcritical Water. Separation and Purification Technology, 2008. 63(3).

48. Wang, L., Yang, H., Zhang, C., Mo, Y. and Lu, X., Determination of Oxytetracycline, Tetracycline and Chloramphenicol Antibiotics in Animal Feeds using Subcritical Water Extraction and High Performance Liquid Chromatography. Analytica Chimica Acta, 2008. 619.

49. Ogunsola, O.M. and Berkowitz, N., Extraction of Oil Shales with Sub- and Near-critical Water. Fuel Processing Technology, 1995. 45(2): pp. 95-107.

50. Ibanez, E., Oca, A., de Murga, G., López-Sebastián, S., Tabera, J. and Reglero, G., Supercritical Fluid Extraction and Fractionation of Different Preprocessed Rosemary Plants. J. Agric. Food Chem, 1999. 47(4): pp. 1400-1404.

51. Hardman, J.G. and Limbird, L.E., Goodman & Gilman’s The Pharmacological Basis of Therapeutics. 2001.

52. Coym, J.W. and Dorsey, J.G., Reversed-phase Retention Thermodynamics of Pure-water Mobile Phases at Ambient and Elevated Temperature. Journal of Chromatography A, 2004. 1035(1): pp. 23-29.

53. Coym, J.W. and Dorsey, J.G., Superheated Water Chromatography: A Brief Review of an Emerging Technique. Analytical Letters, 2004. 37(5): pp. 1013-1023.

54. Ramos, L., Kristenson, E.M., and Brinkman, U.A.T., Current use of Pressurised Liquid Extraction and Subcritical Water Extraction in Environmental Analysis. Journal of Chromatography A, 2002. 975(1): pp. 3-29.

49

55. Teutenberg, T., Lerch, O., Gotze, H.-J. and Zinn, P., Separation of Selected Anticancer Drugs Using Superheated Water as the Mobile Phase. Analytical Chemistry, 2001. 73: pp. 3896- 3899.

56. Carr, A., Mammucari, R. and Foster, N., The Solubility and Micronization of Griseofulvin using Subcritical Water. Industrial & Engineering Chemistry Research, 2010.

57. King, J.W., Critical Fluid Technology for the Processing of Lipid-related Natural Products. C.R.Chimie, 2004. 7: pp. 647-659.

58. Yeo, S.D. and Kiran, E., Formation of Polymer Particles with Supercritical Fluids: A Review. Journal of Supercritical Fluids, 2005. 34(3): pp. 287-308.

59. Date, A.A. and Patravale, V.B., Current Strategies for Engineering Drug Nanoparticles. Current Opinion in Colloid & Interface Science, 2004. 9(3-4): pp. 222-235.

60. Carr, A., Mammucari, R. and Foster, N.. Controlled Precipitation of Hydrophobic Pharmaceuticals in Subcritical Water. in ISSF 2009. 2009. Arcachon, France.

61. De Gioannis, B., Jestin, P. and Subra, P., Morphology and Growth Control of Griseofulvin Recrystallized by Compressed Carbon Dioxide as Antisolvent. Journal of Crystal Growth, 2004. 262(1-4): pp. 519-526.

62. Carr, A., Mammucari, R. and Foster, N., The Solubility, Solubility Modeling and Precipitation of Naproxen from a Subcritical Water Solution. 2010. Industrial and Engineering Chemistry Research (Corrections received - awaiting resubmission)

63. Maghsoodi, M., Taghizadeh, O., Martin, G.P. and Nokhodchi, A., Particle Design of Naproxen-Disintegrant Agglomerates for Direct Compression by a Crystallo-Co- Agglomeration Technique. International Journal of Pharmaceutics, 2008. 351(1-2): pp. 45-54.

50

3. SOLUBILITY STUDIES

It has been shown that subcritical water may be viable as an organic solvent for precipitation processes. The solvation power of water has been manipulated to dissolve and process organic foodstuffs, salts and contaminants for many years, however there is very little solubility data available in the published literature at temperatures above 100°C. In this section of the thesis, accurate and reproducible solubility data are presented for:

Anthracene Griseofulvin Naproxen Pyrimethamine Budesonide

The solubility of budesonide was established in SBCW with the addition of ethanol as a modifier. The organic solvent was added to reduce the dielectric constant of water. Budesonide was selected because the solubility in SBCW without organic solvents was low. Ethanol was selected as an organic-solvent modifier because of its effect on the dielectric constant of the SBCW, its low toxicity and acceptance for processing pharmaceuticals.

51

3.1. INTRODUCTION

As discussed in Chapter 2, current solubility models are not capable of predicting accurate and reliable solubility predictions for organic compounds in SBCW. Furthermore, the current database of compounds for which solubility has been measured and modelled is limited to polycyclic aromatic hydrocarbons. Collection of solubility data in this work allows for updates of the current state-of-the-art solubility model (M-UNIFAC model) parameters, which are developed in Chapter 4, as well as providing fundamental information for the particle formation experiments to be described in Chapter 5.

A number of studies of the determination of solubilities in SBCW using dynamic (steady state) systems have been reported [1-4]. In this study, a new batch method was devised to determine the solubility of hydrophobic compounds in SBCW. The development of this method allowed for a quick conversion of the solubility method into a batch precipitation apparatus.

The new method developed in this work was validated based on literature solubility data for anthracene [1], to ensure that the data being produced by our method was accurate and reliable. The selection of the other pharmaceutical compounds was based on known low water solubilities at room temperature, and in some cases, the potential for improvement in bioavailability upon micronization [5]. All of the drugs selected were treated as model compounds, as the aim of this thesis was to develop a new particle formation technology, rather than develop a new product to be used to treat ailments.

3.2. EXPERIMENTAL METHOD

A schematic of the experimental apparatus is shown in Figure 3.1. The fittings and tubing were of (type 316). Stainless steel has been used in the past for subcritical water experiments up to 250°C [6-7], and it has been proven suitable for operation below 250°C in the absence of salts or oxygen [8-9]. A Druck pressure transducer and indicator was fitted and a Shimadzu GC-8A chromatography oven was used as the heating unit.

52

Figure 3.1: Schematic of the solubility apparatus

53

The solubility vessel (SV) had an internal volume of 6.4mL. For each run, the SV was loaded with an excess of drug (confirmed by inspecting the vessel post processing for trace drug powder). The vessel was filled with water from the syringe pump P1. The “line end” was left open during the filling period, and once water dripped out, it was sealed off with a stainless steel cap. The water overflow ensured that air was purged from the system.

The operating pressure was set to 70 bar via the syringe pump, thus ensuring that water was in the liquid state throughout the experiment[10]. The system was allowed to equilibrate for 5 minutes with all valves closed prior to heating. The system was then brought to the selected temperature using the GC oven with pressure generated from thermal expansion relieved through V4.

Once the set temperature was reached (which took 15-20 minutes depending on the final temperature) the system was left to equilibrate for 10 minutes whilst being internally stirred by an oscillating magnetic bar. The internal magnet was guided by an external iron ring magnet. Both the internal magnet and the iron ring magnet were purchased from AMF Magnetic. A schematic of the magnetic stirrer setup is shown in Figure 3.3. The ring magnet was attached to stainless steel rod, which was guided out of a small hole in the oven. The rod was driven by an electric motor outside of the oven. The external ring magnet oscillated at 36 downstrokes per minute.

After 10 minutes mixing, the magnetic stirrer was stopped and the nitrogen supply - preset to 72 bar- was allowed to contact the solution via the opening of V3. High pressure nitrogen enabled constant vessel pressure to be maintained, thereby preventing the vaporization of SBCW inside the apparatus during product collection. Valve V4 was opened slightly to permit the slow flow of SBCW solution into a capped vial. The purpose of the cap was to minimize the amount of water lost through evaporation out of the collection line to atmosphere. Once nitrogen started to flow into the collection vial, V4 was shut and the oven was turned off.

In order to prevent the flow of undissolved solute through SV: two stainless steel filters were added to the fittings at either end of SV. A diagram of the filter setup is shown in Figure 3.2.

54

Figure 3.2: Filter setup at the ends of SV

The presence of undissolved drug in SV was observed after each experiment. The collection vial with the solution was then removed and weighed. Valve V3 was shut to isolate the nitrogen supply and the system was depressurized through V4. After the system was cooled, the collection line was removed and subjected to a flow of 20mL reagent grade acetone to collect deposited griseofulvin. Acetone solutions were collected in separate vials. The vial filled with acetone and the vial filled with water were dried for 24 hours in air and in an oven at 50°C, respectively. Both vials were then re-weighed. In order to confirm that solvents were quantitatively removed thermogravimetric analysis (TGA) was conducted (showing negligible solvent content), as shown in Section 3.3.2. Each experiment was repeated a minimum of 4 times to ensure reproducibility.

Solute quantification was determined gravimetrically for samples above 10 mg. Smaller samples were analysed by UV spectrometry. A comparison between the accuracy of the mass balance method and the UV method is discussed in Section 3.3.3. The system was validated against the solubility of anthracene from the published literature [1], discussed in Section 3.5.1.

55

Crankshaft

a) c)

b)

Vessel internals – Disc magnet

Figure 3.3: Design of the solubility apparatus a) stirrer crankshaft, b) internal disc magnet and c) overall vessel design showing the crankshaft with the ring magnet held outside SV

56

3.2.1. ADDITION OF ORGANIC SOLVENTS AS MODIFIERS TO SBCW

The effect of organic solvent modifiers on the solubility of organic compounds in SBCW has been investigated previously [11]. In the presented work, ethanol was used to modify the SBCW to reduce the polarity of the solvent. Ethanol was added to de-ionized water at 5% and 20% v (ethanol)/v (water) fractions. The solubility was measured using the same method outlined in Section 3.2 of this chapter.

3.3. EXPERIMENTAL CONSIDERATIONS

3.3.1. EQUILIBRATION TIME Equilibration time was set based on two considerations:

1. Avoidance of hydrothermal reactions between the solute and water 2. Achievement of solute saturation in the SBCW-API solution

The first consideration was dependent upon both the temperature at which experiments were conducted and the exposure time of the solute to SBCW [12]. It has been shown that, for many hydrolysis reactions, temperatures of 250°C and exposure times of at least 15 minutes are required before hydrolysis begins to occur [12-13]. However, for some compounds, hydrolysis can occur at 180°C within 15 minutes [14]. To test the stability of model compounds upon processing, infra-red spectroscopy was performed on each compound at the highest temperature used in the solubility studies. The method and results of these tests are presented in Sections 3.4.1 and 3.5.1 respectively.

The time required to reach saturation was evaluated by comparing the levels of dissolved solute observed with 10 minute and 30 minute equilibration times. The experiments were conducted at the lowest temperature within the tested range on the material that had the lowest solubility in room temperature and subcritical water: budesonide. The experiments were conducted in triplicate and the results are shown in Table 3.1. There was no statistically relevant difference between the solubility results at 10 minutes and 30 minutes, as the mean of each sample was well within the margin of error of the other sample. A two-tailed t-test was conducted on the two points of data in question. It was found that the mean value of each data point was within the 95% confidence interval of being equal to the other data point (thus a success in proving the null hypothesis that x10 minutes = x30 minutes). Thus it was determined that 10 minutes was sufficient time to reach saturated conditions.

57

Table 3.1: Solubility of budesonide in SBCW at 100°C at 10 and 30 minutes exposure times

Exposure time Temperature Solubility Standard Deviation (minutes) (°C) (mole fraction x106) (x106) 10.0 100 5.03 2.15

30.0 100 6.09 1.47

3.3.2. DRYING METHOD

The effectiveness of the drying technique was evaluated by determining the amount of residual solvent (water) in the sample post processing. Thermogravimetric analysis (TGA) was perfomed on the most hygroscopic material processed: naproxen. The TGA was operated using a standard temperature ramp program. The temperature was raised by 15°C/minute in an air environment up to 1000°C. The TGA curve is shown in Figure 3.4. There was a 1.1% mass difference between 19.8°C and 150.2°C, and therefore the percentage of water in the drug was negligible. Thus the drying method was effective.

2.5

2

1.5

1

0.5 Mass (g) Mass

0

-0.5

-1 0 50 100 150 200 250 Temperature (°C)

Figure 3.4: TGA curve of naproxen samples produced at 160°C and subjected to drying in the oven

58

3.3.3. SOLUTE QUANTIFICATION

Solute determination by gravimetry and UV analysis were used to calculate solubility data for budesonide in SBCW at different temperatures. The minimum mass of solute that could be measured by gravimetry was determined by comparing the mass reading of the balance with the mass measured by UV spectrometry for the experimental points of low solubility. Results are shown in Figure 3.5. The solubility readings are equivalent for solubility values no less than a mole fraction of 3.7x10-5, which corresponded to a mass reading of 6mg (±0.68mg). For this reason, solutes extracted during solubility experiments that had a mass reading of less than 10mg by mass difference were measured by UV. Samples requiring UV analysis were budesonide and anthracene at 150°C and below (though for budesonide, UV spectrometry was carried out at all temperatures and all tested co-solvent conditions).

1.20E-04

1.00E-04

8.00E-05

6.00E-05

4.00E-05

Mole Fraction Fraction Mole Solubility 2.00E-05

0.00E+00 80 100 120 140 160 180 200 220 Temperature /°C

Figure 3.5: Comparison of the solubility data produced from measuring the mass of budesonide and water using gravimetry (■) and UV analysis (♦)

3.3.4. OVEN TEMPERATURE FLUCTUATIONS

To ensure that the oven was heating to the preset temperature (controlled by a thermocouple), two independent thermometers were used during experimental runs. Due to the age of the oven, it was necessary to check that throughout the time of the experimental runs (30-40 minutes depending on the required temperature) the oven was maintaining the required set temperature. Both a mercury bulb and a temperature sensor (JENCO 7000CH microcomputer)

59 were placed at different points around the oven. Both thermometers showed that at all points, and at all times, the oven maintained a constant temperature.

3.4. ANALYTICAL TECHNIQUES

3.4.1. INFRARED SPECTROSCOPY

Fourier Transform Infrared Spectroscopy (FTIR) is the primary method to assess whether a chemical has undergone any chemical degradation [15]. The stability of each of the model compounds in SBCW was tested by running a solubility experiment at 200 °C with an exposure time of 20 minutes. Samples were analysed by FTIR on the dried sample of both the raw drug and the drug processed in SBCW at 200°C using the KBr disc method. The results are shown in Section 3.5.1.

3.4.2. DIFFERENTIAL SCANNING CALORIMETRY

A TA Instruments differential scanning calorimeter (DSC) 2010 was used. Dried powder samples (5 to 10mg) were loaded into an aluminum pan and sealed. The samples were cooled to -50oC and the temperature was ramped at 10oC/min to 300oC in a nitrogen atmosphere.

3.5. RESULTS

3.5.1. CHEMICAL STABILITY OF THE SOLUTES IN SUBCRITICAL WATER

The FTIR spectra of each of the model compounds were unchanged upon SBCW treatment, as shown in Figure 3.6. As the compounds were exposed to SBCW at the highest of the tested temperatures for a time that was double the experimental exposure time during solubility measurements, it was concluded that all the selected model compounds were stable in the range of temperatures and times tested in this study.

60

Figure 3.6: Comparative FTIR spectra of a) griseofulvin, b) naproxen, c) pyrimethamine and d) budesonide in both raw and processed states

61

3.5.2. SOLUBILITY

The solubility profile of anthracene in SBCW measured using the method described in Section 3.2 is shown in Figure 3.7. The comparison of our results with the results of Miller et al. [1] is also shown in Figure 3.7. The solubility of anthracene found from our work was typically higher than the solubility of anthracene reported in the published literature. The highest deviation from the published literature is at 200°C, where our results had a 10% difference to the literature values. The deviation is calculated according to equation (1). However, discrepancies were well within the experimental error, demonstrating a good agreement between the two sets of data.

(1)

The solubilities of each drug in SBCW have been calculated as mole fractions. The individual model compound solubility data are shown in Table 3.2-Table 3.5. In all cases, solubilities increased steeply with temperature beyond 160°C. The steep increase in solubility beyond 160°C is consistent with the solubility behavior of other HOCs in SBCW [1, 16-17]. The change in solubility behavior may be related to a termination in the hydrogen-bonds that form cages around solute molecules at 160°C, which hinders the solubility of materials in SBCW below 160°C [18].

0.00035

0.0003

0.00025

0.0002

0.00015

0.0001 Solubility (mole fraction) (mole Solubility 0.00005

0 50 70 90 110 130 150 170 190 210 Temperature (°C)

Figure 3.7: Solubility of anthracene in SBCW from our results (◊) and Miller et al. (▪)[1]. The lines serve as guides to the eye.

62

Table 3.2: Solubility of griseofulvin from 25°C to 170°C

T x2 (106) SD (106) Ref (°C) (mole fraction) (mole fraction) 25 0.52 0.12 [19-20] 140 160 6.30 150 226 4.20 155 295 28.0 160 378 150 170 528 25.0

Table 3.3: Solubility of naproxen from 25°C to 170°C

T x2 (106) SD (106) Ref (°C) (mole fraction) (mole fraction) 25 5.27 0.27 [21-23] 110 302 53.0 120 678 57.0 130 1270 44.0 140 1860 260 150 2040 280

Table 3.4: Solubility of pyrimethamine from 25°C to 180°C

T x2 (106) SD (106) Ref (°C) (mole fraction) (mole fraction) 25 2.53 0.51 [24-25] 140 121 15.0 150 174 20.0 160 372 21.0 170 675 90.0 180 792 110

Table 3.5: Solubility of budesonide from 25°C to 170°C

T x2 (106) SD (106) Ref (K) (mole fraction) (mole fraction) 25 0.75 0.13 [26-27] 100 6.10 1.47 130 15.4 0.10 140 20.6 4.20 150 28.3 2.40 180 67.3 7.50 200 98.1 8.00

63

A comparison between the solubility, molecular structure and molecular weight of the APIs at each of the measured temperatures is shown in Table 3.6. Larger molecules, such as budesonide, tend to be less soluble than smaller molecules, such as naproxen. More highly oxygenated compounds, such as naproxen, tend to be more soluble than compounds containing nitrogen, such as pyrimethamine. Compounds that have chlorine side-groups, such as griseofulvin and pyrimethamine tend to have lower solubilities than compounds that do not contain chlorine, such as naproxen. The solubility behaviour of the APIs follows similar trends to other organic hydrocarbons, as described in Chapter 2.

64

Table 3.6: Solubilities and structures of APIs and anthracene in subcritical water, from 25°C to 200°C

Solubility* Temperature (°C) Anthracene Budesonide Griseofulvin Naproxen Pyrimethamine 25 1.20E-08 8.40E-07 6.10E-07 5.50E-06 2.20E-06 100 2.40E-07 6.10E-06

110 3.02E-04

120 6.80E-04

130 1.50E-05 1.20E-03

140 2.00E-05 1.60E-04 1.90E-03 1.20E-04

150 9.00E-06 2.80E-05 2.30E-04 1.90E-03 1.70E-04 155 3.00E-04

160 3.00E-05 3.80E-04 3.70E-04

170 5.30E-04 6.80E-04

180 1.30E-04 6.70E-05 7.90E-04

200 2.30E-04 9.80E-05

Structure

Molecular Weight 178.23 430.53 352.76 230.26 248.71

*Solubility in mole fraction

65

3.5.3. THE INFLUENCE OF DIELECTRIC CONSTANT ON SOLUBILITY

Budesonide exhibited low solubility in SBCW, even at 200°C. Low solubility in a large scale micronization process will lead to low process efficiency (in terms of water usage). It has been documented that the solubility of organic compounds may be increased in SBCW when organic solvent modifiers are added to the SBCW-organic compound solution [11]. The effect of the organic modifiers is to reduce the dielectric constant of the solvent, thereby increasing the affinity of the solvent for the solute. Thus it is possible to increase the solubility of budesonide in the subcritical water system by adding organic solvent modifiers to SBCW.

The organic solvent modifier selected for budesonide was ethanol. While other solvents may increase the solubility of budesonide in SBCW more than ethanol, ethanol is relatively non-toxic. The use of toxic organic solvents as polarity-modifiers would mitigate one of the primary benefits of using SBCW: namely the environmentally friendly and biologically benign nature of using only water.

The solubility results for budesonide in SBCW and SBCW-ethanol modified solutions are shown in Figure 3.8. The addition of ethanol increased the solubility of budesonide in SBCW by up to 10 times the solubility in SBCW in the absence of ethanol. Similar increases in solubility with the addition of ethanol have been observed previously [11].

66

0.00045

0.0004 0% ethanol 0.00035 5% ethanol 0.0003 20% ethanol 0.00025

0.0002

0.00015 Solubility (mol/mol) 0.0001

0.00005

0 0 25 50 75 100 125 150 175 200 225 Temperature (°C)

Figure 3.8: Solubility of budesonide in SBCW with different volume fractions of ethanol. Room temperature solubility data was from [28]

-4

-6

-8

) y = -0.2031x + 1.5745 2 R² = 0.9539 -10 ln (x ln

-12

-14

-16 80 75 70 65 60 55 50 45 40 Dielectric constant (ɛ)

Figure 3.9: Solubility of budesonide in pure and ethanol-modified SBCW

67

Table 3.7: Dielectric constant values for temperatures between 0°C and 200°C and 0% and 20% (v/v) ethanol in water solutions

Ethanol % (v/v) T (°C) 0 5 20 0 80.4 77.4 68.7 50 72.5 69.8 61.7 100 65.5 63.0 55.4 110 64.2 61.7 54.3 120 62.8 60.4 53.1 130 61.6 59.2 52.0 140 60.3 58.0 50.9 150 59.1 56.8 49.8 160 57.9 55.6 48.8 170 56.7 54.5 47.7 180 55.6 53.3 46.7 190 54.4 52.3 45.7 200 53.3 51.2 44.8

It is possible to estimate the dielectric constant of water and organic solvent mixtures at elevated temperatures using a relationship described by Akerlof [29]. The Akerlof relationship was used to estimate the dielectric constant of water and ethanol mixtures at 0%, 5% and 20% concentrations for temperatures between 25°C and 200°C. The results are shown in Table 3.7. The Akerlof correlation predictions are accurate to 3 significant figures [29], thus the dielectric constant predictions are displayed to 2 significant figures.

A plot of the solubility of natural logarithm of the solubility against the dielectric constant of water is shown in Figure 3.9. The conformation of all of the data points to the linear relationship shown in Figure 3.9 indicates that the solubility of behaviour of budesonide is directly dependent upon the dielectric constant of the SBCW/solvent mixture.

The line that best fits the data in Figure 3.9 can be used as a predictive solubility model for any solvent in any fraction that is to be added to the SBCW-budesonide solution. The solubility model, referred to throughout the rest of this thesis as the Dielectric Constant model is presented and evaluated in Chapter 4. The model is also validated against the solubility of budesonide in SBCW with methanol as a modifier in Chapter 4.

68

3.6. CONCLUSIONS

The influence of molecular structure on the solubility of APIs in SBCW mirrors the behaviour observed for other organic compounds, as identified in Chapter 2. All organic compounds studied tend to exhibit a dramatic increase in solubility at and above 160°C, which is believed to be a response to the disappearance of hydrogen bonded cages around solute molecules.

The solubility of budesonide in SBCW was increased by adding different fractions of ethanol into the SBCW-budesonide mixture. The addition of ethanol acted to decrease the overall dielectric constant of the SBCW-organic solvent mixture. A relationship between the solubility of organic solutes in subcritical water and the dielectric constant was found by comparing the solubility of budesonide in SBCW with different fractions of ethanol added. The log of the solubility increases linearly as the dielectric constant of water decreased. A model based on this trend, and is presented in Chapter 4.

69

3.7. REFERENCES

1. Miller, D.J., Hawthorne, S.B., Gizir, A.M. and Clifford, A.A., Solubility of Polycyclic Aromatic Hydrocarbons in Subcritical Water from 298 K to 498 K. Journal of Chemical & Engineering Data, 1998. 43(6): pp. 1043-1047.

2. Karásek, P., Planeta, J. and Roth, M., Solubility of Solid Polycyclic Aromatic Hydrocarbons in Pressurized Hot Water: Correlation with Pure Component Properties. Industrial & Engineering Chemistry Research, 2006. 45(12): pp. 4454-4460.

3. Karásek, P., Planeta, J. and Roth, M., Solubilities of Adamantane and Diamantane in Pressurized Hot Water. Journal of Chemical & Engineering Data, 2008. 53(3): pp. 816- 819.

4. Karásek, P., Planeta, J. and Roth, M., Solubilities of Oxygenated Aromatic Solids in Pressurized Hot Water. Journal of Chemical & Engineering Data, 2009: pp. 294-301.

5. Date, A.A. and Patravale, V.B., Current Strategies for Engineering Drug Nanoparticles. Current Opinion in Colloid & Interface Science, 2004. 9(3-4): pp. 222-235.

6. Smith, R.M., Superheated Water: The Ultimate Green Solvent for Separation Science. Analytical and Bioanalytical Chemistry, 2006. 385(3): pp. 419-421.

7. Smith, R.M. and Burgess, R.J., Superheated Water as an Eluent for Reversed-Phase High- Performance Liquid Cromatography. Journal of Chromatography A, 1997. 785(1-2): pp. 49-55.

8. Kritzer, P., Corrosion in High-Temperature and Supercritical Water and Aqueous Solutions: A Review. The Journal of Supercritical Fluids, 2004. 29(1-2): pp. 1-29.

9. Ozel, M.Z., Bartle, K.D., Clifford, A.A. and Burford, M.D., Extraction, Solubility and Stability of Metal Complexes using Stainless Steel Supercritical Fluid Extraction System. Analytica Chimica Acta, 2000. 417: pp. 177-184.

10. Smith, J.M., Van Ness, H.C. and Abbot, M.M., Chemical Engineering Thermodynamics. 6 ed. 2001: McGraw Hill.

11. Curren, M.S.S. and King, J.W., Solubility of Triazine Pesticides in Pure and Modified Subcritical Water. Analytical Chemistry, 2001. 73(4): pp. 740-745.

70

12. Siskin, M. and Katritzky, A.R., A Review of the Reactivity of Organic Compounds with Oxygen-Containing Functionality in Superheated Water. Journal of Analytical and Applied , 2000. 54(1-2): pp. 193-214.

13. Holliday, R.L., King, J.W. and List, G.R., Hydrolysis of Vegetable Oils in Sub- and Supercritical Water. Industrial & Engineering Chemistry Research, 1997. 36(3): pp. 932- 935.

14. Smith, R.M., Cheinthavorn, O., Wilson, I.D., Wright, B. and Taylor, S.D., Superheated Heavy Water as the Eluent for HPLC-NMR and HPLC-NMR-MS of Model Drugs. Analytical Chemistry, 1999. 71: pp. 4493-4497.

15. Council of Europe, European Pharmacopoeia 5.0., Council of Europe, 2005. p. 1691.

16. Miller, D.J. and Hawthorne, S.B., Method for Determining the Solubilities of Hydrophobic Organics in Subcritical Water. Analytical Chemistry, 1998. 70(8): pp. 1618-1621.

17. Karasek, P., Planeta, J. and Roth, M., Solubility of Solid Polycyclic Aromatic Hydrocarbons in Pressurized Hot Water at Temperatures from 313 K to the Melting Point. Journal of Chemical & Engineering Data, 2006. 51(2): pp. 616-622.

18. Shinoda, K., "Iceberg" Formation and Solubility. The Journal of Physical Chemistry, 1977. 81(13): pp. 1300-1302.

19. Trotta, M., Gallarate, M., Carlotti, M.E. and Morel, S., Preparation of Griseofulvin Nanoparticles from Water-Dilutable Microemulsions. International Journal of Pharmaceutics, 2003. 254(2):p p. 235-242.

20. Elamin, A.A., Ahlneck, C., Alderborn, G. and Nyström, C., Increased Metastable Solubility of Milled Griseofulvin, Depending on the Formation of a Disordered Surface Structure. International Journal of Pharmaceutics, 1994. 111(2): pp. 159-170.

21. Liu, R., Water-Insoluble Drug Formulation. 2000: CRC Press, Taylor and Francis Group, Florida, USA. p.409.

22. Abraham, M.H. and Le, J., The Correlation and Prediction of the Solubility of Compounds in Water using an Amended Solvation Energy Relationship. Journal of Pharmaceutical Sciences, 1999. 88(9): pp. 868-880.

71

23. Ran, Y. and Yalkowsky, S.H., Prediction of Drug Solubility by the General Solubility Equation (GSE). Journal of Chemical Information and Computer Sciences, 2001. 41(2): pp. 354-357.

24. de Araujo, M.V.G., Macedo, O.F.L., Nascimento, C.d.C., Conegero, L.S., Barreto, L.S., Almeida, L.E., de Costa Jr, N.B. and Gimenez, I.F., Characterization, Phase Solubility and Molecular Modeling of [Alpha]-Cyclodextrin/pyrimethamine Inclusion Complex. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2009. 72(1): pp. 165-170.

25. Lázaro, G.S., Meneses Jr, A.L., de Macedo, O.F.L., Gimenez, I.d.F., da Costa Jr, N.B., Barreto, L.S. and Almeida, L.E., Interaction of Pyrimethamine and Sulfadiazine with Ionic and Neutral Micelles: Electronic Absorption and Fluorescence Studies. and Surfaces A: Physicochemical and Engineering Aspects, 2008. 324(1-3): pp. 98-104.

26. Bandi, N., Wei, W., Roberts, C.B., Kotra, L.P. and Kompella, U.B., Preparation of Budesonide- and Indomethacin-Hydroxypropyl-[Beta]-Cyclodextrin (HPBCD) Complexes using a Single-step, Organic-solvent-free Supercritical Fluid Process. European Journal of Pharmaceutical Sciences, 2004. 23(2): pp. 159-168.

27. Lin, H., Yoo, J.W., Roh, H.J., Lee, M.K., Chung, S.J., Shim, C.K. and Kim, D.D., Transport of anti-allergic drugs across the passage cultured human nasal epithelial cell monolayer. European Journal of Pharmaceutical Sciences, 2005. 26(2): pp. 203-210.

28. Ali, H.S.M., York, P., Blagden, N., Soltanpour, S., Acree, W.E., and Jouyban, A., Solubility of Budesonide, Hydrocortisone, and Prednisolone in Ethanol + Water Mixtures at 298.2 K. Journal of Chemical & Engineering Data, 2009. 55 (1): pp.578-582.

29. Akerlof, G., Dielectric Constants of some Organic Solvent-Water Mixtures at Various Temperatures. Journal of the American Chemical Society, 2002. 54(11): pp. 4125-4139.

72

4. SOLUBILITY MODELLING

The Modified UNIversal Functional Activity Coefficient (M-UNIFAC) model is applied to predict the solubility of anthracene, budesonide, griseofulvin, naproxen and pyrimethamine in subcritical water. Using the experimental solubility data reported in Chapter 3 and literature solubility data, model outputs are optimized by changing interaction parameters of the M-UNIFAC model. Correction to the carboxylic acid-water and chlorine-water interaction parameters is shown to be required to fit the M-UNIFAC model to the reported solubility data. A new solubility model is also presented that correlates the dielectric constant of a water/organic solvent solution with the solubility of a test API in SBCW/organic solvent solutions. The model has an average error of 3.4%.

73

4.1. INTRODUCTION

In Chapter 2, the peculiarities of modelling the solubility of organic compounds in SBCW using both empirical and semi-empirical models were discussed. The primary limitation of the current models is that they cannot accurately predict the solubility of an organic compound in water over broad temperature ranges. It was also shown that solubility models using water properties alone (as represented by either the dielectric constant or heat of vaporization values) cannot accurately predict the solubility of a HOC in SBCW. The interaction between organic compound side-groups (e.g. hydroxide side-groups, or carboxylic side-groups on an aromatic ring) and water molecules as a function of temperature needs to be taken into account directly to accurately model organic compound solubility in SBCW.

The M-UNIFAC model takes into account the interactions between water and organic compound side-groups as a function of temperature [1]. Recently, the M-UNIFAC model has been used to predict the solubility of a number of polycyclic aromatic hydrocarbons (PAHs) in SBCW [2]. To reduce the error of M-UNIFAC model outputs, the interaction parameters between water and aromatic carbon (ArC) and aromatic carbon-hydrogen (ArCH) side-groups were optimized [2]. The optimization reduced the average model error from over 50% to 4.7% for the 13 organic compounds tested. It was also shown that model errors were consistent over a broad temperature range, which implied that hydrogen bonding interactions between water and the side-groups was either not an issue, or appropriately corrected for, during the model fitting exercise [3].

The main limitation of the updated M-UNIFAC model is that only two of over 45 interaction parameters have been corrected[4]. Furthermore the ArC side-groups and water do not undergo hydrogen bonding with water. Side-groups that contain oxygen, such as ether (CHO) and ethyl (CH3O) side-groups, which undergo hydrogen bonding with water, would be more affected by changes in SBCW temperature than aromatic hydrocarbons in which oxygen is not present. The impact of oxygenated side-groups on the solubility of aromatic hydrocarbons has not yet been thoroughly investigated through rigorous solubility modelling. The work described in this chapter seeks to extend the M-UNIFAC model to be able to account for oxygenated aromatic hydrocarbons by fitting M-UNIFAC parameters to recently acquired solubility data from the published literature, as well as the data reported in Chapter 3.

74

Solubility relationships that attempt to take into account association between solutes and solvents (which includes hydrogen bonding) have been incorporated into the UNIFAC model [5]. Bonding/association interactions between the solvent and the solute are modelled using pure solvent properties such as the dielectric constant and heat of vaporization values. Solute interactions are taken into account by using melting point temperature and heat of fusion data [5]. Though the UNIFAC model can be inaccurate over broad temperature ranges [6], with some corrections to empirical interaction parameters, the accuracy of the model between 25°C and 300°C can be improved [2]. The UNIFAC model (and some state-of-the art variations on the UNIFAC model) is used in this work to evaluate their applicability to SBCW systems.

The results of the various UNIFAC models are analyzed based on the ability of each model to predict the solubility of various HOCs in SBCW between 25°C and 200°C. Model outputs are compared to the measured solubilities of the APIs investigated in Chapter 3. Model parameters were obtained from the published literature. In the absence of literature data, thermal methods such as DSC were employed to estimate important model inputs (particularly the heat of fusion and melting point temperature).

A new empirical model is also presented which correlates the dielectric constant of a SBCW system with the solubility of a model compound (measured in Chapter 3). Though many authors have commented on the dependency of HOC solubility in water to the dielectric constant of water [7-9], no authors have attempted to correlate the solubility of a HOC in SBCW with the dielectric constant numerically. The model was constructed by plotting the solubility data to the dielectric constant of the water system. The dielectric constant was calculated using a relationship developed by Akerlof [10].

4.2. THERMODYNAMIC FRAMEWORK

The UNIFAC method predicts activity coefficients based on functional group interactions. The functional group interactions are determined empirically, based on pure component properties from phase equilibrium data. The activity coefficient may be used to calculate multi-phase equilibrium data, or solubility of multiple materials within a solvent, or combination of solvents. The benefit of using a method such as this is that the modelling of complex organic materials may be simplified, by subdividing the compound structure into a number of simple, common subgroups. The interaction of each subgroup with the solvent is then estimated, and the

75 interaction values are summed to produce a net interaction value, which can be used to calculate solubility.

At equilibrium, the fundamental fugacity condition for the solid solute (s) to dissolve in water (l) is:

Equation 4.1

Where f2s is the fugacity of the solute in the solid phase and f2l is the fugacity of the solute in the water phase. It is assumed that water is not absorbed in the solid phase [11]. The pure solute fugacity is defined as f20 [2]. The relationship of this property with the fugacity of the solute in the liquid phase is:

Equation 4.2

Where γ2 is the activity coefficient of the solute in the water phase and x2 is the mole fraction of the solute in the water phase (the solubility). By rearranging Equation 4.2 and using the fundamental fugacity condition (Equation 4.1), the solubility relationship shown in Equation 4.3 is reached.

Equation 4.3

Where the γ2 is estimated using the UNIFAC or M-UNIFAC models. Karásek [13] presented a

simple estimation of the fugacity relationship, shown in Equation 4.3 (i.e. ). This is a standard derivation that can be found commonly throughout literature and thermodynamic texts [12]. The derivation of this relationship is shown below.

The standard state fugacity ( ) is arbitrary. Typically, the standard state for solid-liquid equilibrium (SLE) is taken as the fugacity of a pure subcooled liquid of compound 2 at the temperature of the solution. The subcooled state is hypothetical, and can be determined by extrapolating the vapour-liquid equilibrium line below the melting point. However, it is often tedious and inaccurate to extrapolate this line far away from the of the solute. A more accurate method of estimating has been developed [12]. A relationship exists that relates the fugacity ratio of the solid and the pure subcooled liquid state to the Gibbs energy change for component 2 transforming from a solid (subscript S) to subcooled liquid (subscript L), as shown in Equation 4.4.

Equation 4.4

76

And:

Equation 4.5

Where H and S refer to the enthalpy and entropy change associated with compound 2 changing from a solid to a pure subcooled liquid state. For enthalpy:

Equation 4.6

Where the superscript f refers to the heat of fusion. Similarly for entropy:

Equation 4.7

And at the triple point:

Equation 4.8

By substituting Equations 4.8, 4.7, 4.6 into Equation 4.5, and equating Equation 4.5 to Equation 4.4, Equation 4.9 is reached:

Eq. 4.9

Commonly, the triple point temperature can be taken as the melting point temperature (Tm), as

the values are the same for most solids. Furthermore, the second two terms (incorporating ) often cancel out, with negligible error [12]. Thus the expression for solubility reduces to equation 4.10.

Equation 4.10

The heat of fusion (ΔHm2) and Tm2 (Where the subscript 2 refers to the solute) are normally available in the published literature, or, can be determined by differential scanning calorimetry

(DSC). The ΔHm and Tm data for the compounds used in our work are shown Table 4.1. Equation 4.10 also assumes that pressure has a negligible effect, which is reasonable at the pressures at which solubility data is typically collected for our experiments.

Table 4.1: Heat of fusion and melting temperature for the studied APIs

Anthracene Griseofulvin Naproxen Pyrimethamine Budesonide ΔHm (J/mol) 29000 45860 32930 34040 12150 Tm (K) 490.15 492.15 429.15 516.86 499.65 Reference [14] [15] [16] [17] [18]

77

To construct the UNIFAC model, the activity coefficient is divided into combinatorial (γC) and residual (γR) parts. The combinatorial part takes into account the surface properties of the solute molecules, including size and shape based on the molecular structure and its side-groups. The residual part calculates the molecular interactions between solute and solvent, the solute sidegroups on each other sidegroup in the solute, and the solvent sidegroups on each other sidegroup in the solvent.

Mathematically:

Equation 4.11

Equations 4.12 and 4.13 describe the combinatorial and residual activity coefficients. Equations 4.14-4.15 describe the parameters of Equations 4.6 and 4.7.

Equation 4.12

Equation 4.13

Equation 4.14

Equation 4.15

Equation 4.16

The subscripts j and l are the indices of the subdivided side-groups (e.g. OH in budesonide, see

Table 4.2). The values ri and qi are the volume and surface area parameters respectively, and θl and τli are a function of the interaction parameters within the solute and solvents, running across all side-groups l and I (See Appendix A for a sample calculation). Further functions and in-depth detail can be found in Smith and Van Ness 2006, Appendix H [15].

The alterations of the UNIFAC model to make the M-UNIFAC and A-UNIFAC models can be found in detail in the published literature [3-4]. For the M-UNIFAC model, τli is a function of temperature, as shown in Equation 4.17.

Equation 4.17

where am,n, bm,n and cm,n are empirically derived interaction parameters between a one subgroup (n) and another subgroup (m).

78

The A-UNIFAC model adds a correction term for hydrogen bonding interactions [4-5]. The correction term for the A-UNIFAC model is shown in Equation 4.18, [4].

Equation 4.18

Where lnXinf is the association correction to the activity coefficient calculated at infinite dilution using the parameters shown in Equations 4.19 to 4.23.

Equation 4.19

Equation 4.20

Equation 4.21

Equation 4.22

Equation 4.23

where and are the hydrogen bonding correction terms for density and hydrogen bond strength respectively between group 1 (solute side-group) and group 2 (solvent OH group). Fornari altered the association parameter in Equation 4.17 to make ε/k a function of temperature, as shown in Equation 4.24 [2]. The constants were fitted by correlating ε/k to the dielectric constant of water, where the ε/k was 3125 at 25°C.

Equation 4.24

Fornari found that the accuracy of the A-UNIFAC model solubility prediction using the above correction to the association energy was improved. In this work, the ε/k correction was applied to the A-UNIFAC model and compared to the outputs of the Fornari modified M-UNIFAC model (MF-UNIFAC), original M-UNIFAC model and original A-UNIFAC model.

79

4.2.1. ESTIMATION OF ACTIVITY COEFFICIENTS

The solubilities of the APIs processed in the SBCW apparatus were modelled using Equation 4.4. The activity coefficients were calculated using:

1. The original UNIFAC model[1, 20] 2. The modified UNIFAC model (M-UNIFAC or Dortmund method)[4-5] , both with and without optimized aromatic parameters by Fornari for the M-UNIFAC model (MF-UNIFAC)[2] 3. The Associating UNIFAC model (A-UNIFAC)[5] with variable temperature parameters, and the A-UNIFAC model modified by Fornari (AF-UNIFAC)[2]

The original UNIFAC model construction, as well as the modified (Dortmund method) UNIFAC and associating UNIFAC methods, including revisions with updated parameter values, can readily be found in the published literature [1-2, 4-5, 20-22]. Several parameter values were not available [22]. In these cases the interaction parameters (amn, bm,n and cm,n) were assumed to be zero. Although errors would be present in this assumption, there is no basis for interpolation of an amn value for a particular compound, as the interaction parameter values follow no identifiable trend. A full list of parametric data is included in Appendix A. Only the subdivision of side-groups, corrected parameters and solubility model predictions are presented in this chapter.

4.2.2. DIVISION OF FUNCTIONAL GROUPS FOR THE SOLUTES

Division of functional groups and subgroups was relatively simple for anthracene, naproxen and budesonide. Griseofulvin and pyrimethamine were more difficult, as some side groups were not accounted for in the model, such as the furan ring in griseofulvin and the pyrimidine ring in pyrimethamine. Due to the lack of published data, the interaction parameter values for these subgroups had to be redefined by substituting the subgroups with similar sidegroups for which interaction parameters were available. The redefinition of compound side-groups is identified as the compounds are presented. The subgroup designations and numbers of subgroups (vk) are shown in Table 4.2.

80

Table 4.2: Subgroup designation for the modelled solutes

Anthracene Naproxen Griseofulvin Pyrimethamine Budesonide Water

Subgroup vk vk vk vk vk vk

CH3 0 1 1 1 3 0

CH2 0 0 1 0 7 0

C=C 0 0 1 0 3 0

Ar1C 4 3 1 1 3 0

Ar1CH 10 6 5 4 0 0

Ar1CCH2 0 0 0 1 1 0

Ar1CCH 0 1 0 0 3 0

H2O 0 0 0 0 0 1

CHO 0 0 2 0 0 0

COOH 0 1 0 0 2 0

CH3O 0 1 4 0 0 0

Ar1CCl 0 0 1 1 4 0

Ar1CNH2 0 0 0 2 0 0

Pyridine 0 0 0 1 0 0

1 Ar denotes an aromatic ring

81

4.2.2.1. Anthracene

Anthracene has a three-benzene ring structure, and has been modelled before using the UNIFAC model, and all successive modifications to the UNIFAC model [2]. Original optimization of the MF-UNIFAC and AF-UNIFAC models for the ArC and ArCH parameters has been carried out, and is applied in this work, as shown in Table 4.2.

Figure 4.1: Chemical structure of anthracene

4.2.2.2. Naproxen

Naproxen contains 2 conjugated benzene rings with additional CH3O, CH3 and COOH sidegroups, as shown in Table 4.2.

Figure 4.2: Chemical structure of naproxen

4.2.2.3. Griseofulvin

Griseofulvin has a furan ring; for which there is no interaction parameter available in the published literature. Though corrections for furan rings are available for the A-UNIFAC model, these parameters cannot be added into the M-UNIFAC model [6]. Parameters from the A- UNIFAC model could not be applied to the M-UNIFAC model because they were not updated in the same way as the M-UNIFAC model. In order to be able to assign the functional group for M- UNIFAC and original UNIFAC models; the bond between the furan oxygen and the adjoined carbon on the right of that oxygen (shown below) was removed.

82

Figure 4.3: Chemical structure of griseofulvin

4.2.2.4. Pyrimethamine

No interaction parameter for the pyrimidine ring in pyrimethamine was available in the published literature. Pyridine, however, is defined and was used in place of pyrimidine in developing the M-UNIFAC model for pyrimethamine. In so doing, one nitrogen molecule was eliminated from the molecule. In place of the nitrogen molecule (based on subgroup designations), a hydrogen and carbon atom would have been present. The elimination of one nitrogen atom should have little effect on the volume and surface area parameters (based on the similar atomic size of carbon and nitrogen).

Figure 4.4: Chemical structure of pyrimethamine

4.2.2.5. Budesonide

While budesonide is a large molecule (relative to the other molecules modeled in this chapter), the side-group allocation was simple. The cycloalkanes were treated as aliphatic hydrocarbon chains. The allocation of side-groups is shown in Table 4.2.

83

Figure 4.5: Chemical structure of budesonide

4.3. UNIFAC MODEL RESULTS AND DISCUSSION

The results presented show the solubility as a natural logarithm of mole fraction as a function of temperature for up to four different model outputs, including the solubility data. Errors were evaluated using Equation 4.25 at each solubility data point, where x is the mole fraction of the solute.

Equation 4.25

4.3.1. ANTHRACENE

The UNIFAC, M-UNIFAC and A-UNIFAC model outputs for anthracene are shown in Figure 4.6. A similar comparison of model and experimental outputs can be seen in the article published by Fornari [2]. While the unaltered M-UNIFAC model outputs has errors below 7% up to 150°C, the model outputs had errors of 20% above 150°C, as shown in Figure 4.6. The MF-UNIFAC improved on the correlation, reducing the overall average error to 13% (Table 4.3).

Despite the improvement in accuracy of the A-UNIFAC model outputs when the Fornari modifications were applied, the model followed a trend that did not accurately depict the solubility behavior of anthracene below 170°C, as shown in Figure 4.6. Further corrections need to be implemented to improve model outputs at both high and low temperatures, despite the equivalent numerical value at 180°C. Clearly, the MF-UNIFAC is the most accurate of the unoptimized UNIFAC models.

84

Table 4.3: Average errors of the solubility models for anthracene in SBCW

Model Average error 100°C - 200°C UNIFAC 39% M-UNIFAC 16% MF-UNIFAC 13% A-UNIFAC 61% AF-UNIFAC 10%

4.3.2. NAPROXEN

The UNIFAC model outputs for naproxen, with the subgroup assignments applied in Table 4.2, are shown in Figure 4.7. The MF-UNIFAC model was more accurate than the M-UNIFAC model. In addition to the improvement of the general solubility-temperature trend, the average model error was reduced from 59% to 28% (Table 4.4). However, the resulting error is still large and it is clear that other interaction parameters need to be optimized to improve the fit of the model to the experimental data.

The AF-UNIFAC model had the lowest error of all of the models across the temperature range tested. However, the prediction of the model was not as accurate at room temperature, and based on the trend, would also be inaccurate at temperatures above 160°C (shown in Figure 4.6). The MF-UNIFAC model was therefore selected for optimization rather than the AF-UNIFAC model.

Table 4.4: Average errors of the solubility models for naproxen in SBCW

Model Average Error 130°C – 170°C UNIFAC 55% M-UNIFAC 59% MF-UNIFAC 28% A-UNIFAC 23% AF-UNIFAC 24%

85

3 Experimental

Poly.M-UNIFAC (M-UNIFAC) -2

Poly.MF- UNIFAC(M-Uni (Fornari correction)) -7 Poly.AF-UNIFAC (A-UNIFAC (Fornari ) 2 correction)) ln (x ln

-12

-17

-22 100 120 140 160 180 200 Temperature (°C)

Figure 4.6: Anthracene solubility and solubility prediction using the M-UNIFAC the MF-UNIFAC and the AF-UNIFAC models

0 Experimental Data

Poly.MF- UNIFAC(MF-UNIFAC) -5 Poly.M-UNIFAC (M-UNIFAC)

-10 Poly.AF-UNIFAC (AF-UNIFAC) ) 2 ln(x -15

-20

-25 0 20 40 60 80 100 120 140 160 180

Temperature (°C) Figure 4.7: Naproxen solubility and solubility predictions using the M-UNIFAC, MF-UNIFAC and AF- UNIFAC models

86

4.3.3. GRISEOFULVIN

The UNIFAC models were less capable of reproducing the solubility behavior of griseofulvin in SBCW than all other compounds, as shown in Figure 4.8. The average errors of each of the UNIFAC model outputs are shown in Table 4.5. The Fornari correction to the M-UNIFAC model reduced the error of the M-UNIFAC model, but not the A-UNIFAC model.

Table 4.5: Average errors of the solubility models for griseofulvin in SBCW

Model Average Error 140°C-170°C UNIFAC 166% M-UNIFAC 280% MF-UNIFAC 64% A-UNIFAC 128% AF-UNIFAC 162%

The large errors in the model may result from outdated interaction parameters between either the oxygen groups and water and/or chlorine and water [2]. The error in the solubility model could be from either of these two groups because the MF-UNIFAC model carbon-hydrogen and carbon-carbon interactions in the presence of SBCW have already been optimized.

4.3.4. PYRIMETHAMINE

The UNIFAC model predictions were more accurate for pyrimethamine than for griseofulvin. Model output errors were reduced when the MF-UNIFAC model and the AF-UNIFAC model were used (Table 4.6). The MF-UNIFAC model outputs are the most accurate predictions of all the tested models for pyrimethamine.

Table 4.6: Average errors of the solubility models for pyrimethamine in SBCW

Model Average Error 140°C-170°C UNIFAC 160% M-UNIFAC 194% MF-UNIFAC 27% AF-UNIFAC 76% A-UNIFAC 103% .

87

-2 ExperimentalExperimental -4 UNIFACUNIFAC -6 MMF-Uni-UNIFAC (Fornari -8 correction) AA-UNIFAC-UNIFAC (Fornari -10 correction) ) 2 -12 ln (x ln -14

-16

-18

-20

-22 100 120 140 160 180 200 Temperature (°C)

Figure 4.8: Solubility and solubility predictions for griseofulvin using the UNIFAC, MF-UNIFAC and AF-UNIFAC models -2 ExperimentalExperimental -4 Poly.M-UNIFAC (M-UNIFAC) -6 Poly.MF -(MUNIFAC-Uni (Fornari correction)) -8 Poly.AF- UNIFAC(A-UNIFAC (Fornari correction)) -10 ) 2 -12 ln (x ln -14

-16

-18

-20

-22 100 125 150 175 200 Temperature (°C) Figure 4.9: Solubility and solubility predictions for pyrimethamine using the M-UNIFAC, MF- UNIFAC and AF-UNIFAC models

88

4.3.5. BUDESONIDE

The errors of the 4 best model outputs for budesonide are shown in Figure 4.10. The MF- UNIFAC models reproduced the trend in solubility well, though the lowest (average) model error was 27%. A table of the average error of each of the UNIFAC model outputs used is shown in Table 4.7. The AF-UNIFAC model outputs were more accurate than the original A-UNIFAC model.

Table 4.7: Average errors of the solubility models for budesonide in SBCW

Model Average Error 100°C-200°C M-UNIFAC 35% MF-UNIFAC 27% A-UNIFAC 50% AF-UNIFAC 46%

0 ExperimentalExperimental Data

Poly.MF (MF-UNIFAC-UNIFAC) -5 Poly.M- UNIFAC(M-UNIFAC)

Poly.AF -(AFUNIFAC-UNIFAC) -10 2 ln x ln -15

-20

-25 100 120 140 160 180 200

o Temperature ( C)

Figure 4.10: Solubility and solubility predictions of budesonide using the UNIFAC and MF-UNIFAC models

89

4.4. UNIFAC MODEL PARAMETER OPTIMIZATION

While the MF-UNIFAC model is an improvement on the original M-UNIFAC model, water-oxygen interaction parameters need to be revised. The AF-UNIFAC model takes into account water- oxygen effects on solubility by adding a correction term that uses the dielectric constant of water (as shown in Equation 4.19). While the dielectric constant is a variable that can indicate hydrogen bonding strength, the AF-UNIFAC model was not accurate for any of the model compounds in this work. Thus, the MF-UNIFAC model was chosen to be optimized.

Recently, solubility data for oxygenated compounds in SBCW have been published. The data are summarized in Chapter 2 [23]. The oxygenated compounds in the published literature had either CHO or CH3O side-groups. The position of the oxygen in these compounds was similar to the position of the oxygen in the APIs used in our studies. The literature data was modelled to determine if the error in solubility prediction of the MF-UNIFAC models was due to any of the oxygen side groups present in the compounds published in literature. The results of the model output errors are reported in Table 4.8. Errors were calculated according to Equation 4.25.

The model could reproduce the solubility of the oxygenated compounds anthone, xanthone, anthrone and 9, 10 anthraquinone to within 6% of experimental values (Table 4.8). While it may be seen that the error for all compounds increases with temperature, it should also be noted that the solubility data published by Karásek et al. is typically lower than other literature data at higher temperatures (such as the published data by Miller et al., reported in Chapter 3 of this thesis) [7, 14]. If the solubility data were higher than what were reported in the published literature, then the error at elevated temperatures would be lowered. The deviations at elevated temperatures of the MF-UNIFAC models in Karáseks work may be due to experimental error rather than model error.

To correct the solubility model output of each API, solubility data at room temperature were incorporated into the optimization of the MF-UNIFAC model. The addition of the room temperature data into the fit allowed for a more accurate correlation over a larger temperature range, which has already been identified to be a problem with most SBCW solubility models (see Chapter 2).

90

Table 4.8: MF-UNIFAC model of xanthenes, anthrone, xanthene and 9, 10 anthraquinone

Xanthene Anthrone Xanthone 9,10 anthraquinone n (Ghmeling) Subgroup vk vk vk vk

2 CH2 1 1 0 0 3 ArC 4 4 4 4 3 ArCH 8 8 8 8 10 CHO 1 0 1 0

13 CH3O 0 1 1 2

# subgroups 14 14 14 14

dHmelt 19200 26800 26120 32570

Tmelt 373.7 428.15 434.1 558 Reference [23]

Errors T(°C) Xanthene Anthrone Xanthone 9,10 anthraquinone 40 6% 2% 1% 1% 60 7% 0% 3% 4% 100 - 3% 8% 10% 140 - 4% 10% 15% 160 - - 9% 16%

Average SD 6%

4.4.1. NAPROXEN

Naproxen contains CH3O, CHO and COOH sidegroups. It has been shown that the CH3O and CHO interaction parameters do not require alteration based on the already accurate MF-UNIFAC modelling of the solubility of compounds published by Karásek [23] over a broad temperature range. It was proposed that the carboxylic acid (COOH) –water interaction parameters be modified to fit the data. It should be noted that naproxen melts at 152°C, as reported in Chapter 3. At and above the melting point of naproxen, the infinite dilution assumption is not valid. Thus, when the model was fitted to the data, only the solubility data at 150°C and below were used to optimize the M-UNIFAC model. The original and modified parameters are shown in Table 4.9. The parameters were optimized by minimizing the error (calculated using Equation 4.25) using the goal seek function in Microsoft Excel by modifying parameter c. The parameters are displayed to the same number of significant figures as were found in the published literature [23].

91

Table 4.9: Original and fitted water-carboxylic acid (side-group 11,18) and carboxylic acid-water (side-group 18,11) interaction parameters. The letters a, b and c represent M-UNIFAC interaction parameters

Original M-UNIFAC Fitted M-UNIFAC

(18,11) COOH-H2O (11,18) H 2O-COOH (18,11) COOH-H2O (11,18) H 2O-COOH a 624.97 -1795.2 624.97 -1795.2 b -4.6878 12.7080 -4.6878 12.7080 c 5.24E-03 -1.55E-02 5.15-03 -1.80E-02

The results of the adjusted M-UNIFAC model parameters are shown in Figure 4.11. The average error (from room temperature to 170°C) was lowered from 28% to 5%.

0 Experimental Data

-5 Poly.MF-UNIFAC (MF-UNIFAC)

-10 ) 2 ln(x -15

-20

-25 0 20 40 60 80 100 120 140 160 180

Temperature (°C)

Figure 4.11: Solubility of naproxen (Δ) and the MF-UNIFAC model (-) with optimized carboxylic interaction parameters

The correction to the MF-UNIFAC carboxylic acid-water interaction parameter was tested by predicting the solubility other carboxylic-acid-containing HOCs in SBCW. There is some data available in the published literature for fatty acid solubility in SBCW from 60°C to 230°C [24]. As the M-UNIFAC model was designed to predict activity coefficients for organic compounds independent of the phase of the solute/solvent, it is entirely possible to compare the original and corrected interaction parameter M-UNIFAC model to the literature data.

92

The activity coefficients for the liquid-liquid solubility were predicted using the M-UNIFAC model in the same way as the activity coefficient was predicted for the solid-liquid solubility. Conversion of the activity coefficients into mutual solubilities (of the naproxen melt into the water phase and the water into the naproxen melt phase) was calculated using a relationship derived from binary phase liquid-liquid equilibrium (LLE) thermodynamics, assuming that the solutes were very dilute in each phase (i.e. that naproxen was dilute in the water phase, and vice-versa, a similar assumption used to predict the SLE for hydrophobic APIs in the work presented in this thesis). The development of the mutual solubility relationship in binary liquid- liquid equilibrium systems is common in the published literature, and presented in Appendix B. The relationships used to predict the mutual solubilities of water and naproxen melt in the water and melted naproxen phases are shown in Equations 4.26-4.27, where the activity coefficients of each compound ( and ) were calculated using the M-UNIFAC model.

Equation 4.26

Equation 4.27

Predicted solubilities using the original and corrected MF-UNIFAC parameters are shown in Figures 4.13 and 4.14, respectively. The correction made to the activity coefficient from the naproxen data, when applied to the LLE of the fatty acids resulted in predictions that were less accurate than when the correction was not applied. The average error using the original M- UNIFAC parameters was 3%, whereas the average error using the corrected MF-UNIFAC parameters was 19%. While the correction clearly increased the average error, the curve fit was specific to naproxen. Future work should evaluate the parameter with a larger database of phase equilibrium data, including the work done here with naproxen.

93

0 Caprylic acid data

Capric acid data

-5 Lauric acid data

Myristic acid data ) 2 Palmytic acid data -10 Stearic acid data

Poly.Caprylic (Caprylicacid predictionacid MF-UNIFAC) -15 Poly.Capric (Capric acid acidprediction MF-UNIFAC) Log solubility solubility Log (ln x Poly.Lauric (Lauric acid acidprediction MF-UNIFAC)

-20 Poly.Myristic (Myristicacid predictionacid MF-UNIFAC) Poly.Palmytic (Palmyticacid predictionacid MF-UNIFAC)

Poly.Stearic (Stearic acid acidprediction MF-UNIFAC) -25 50 100 150 200 250 Temperature (°C)

Figure 4.12: Solubility of fatty acids in SBCW as published by Khwujitjaru et al. [24] and predicted by the MF-UNIFAC model with original COOH interaction parameters

0 Caprylic acid data -2 Capric acid data Lauric acid data -4 Myristic acid data -6 Palmytic acid Stearic acid -8 Poly.Caprylic (Caprylicacid prediction Acid MF-UNIFAC)

-10 Poly.Capric (Capric acid prediction acid MF-UNIFAC) Log solubility (ln x2) (ln solubility Log Poly.Lauric (Lauric acid prediction acid MF-UNIFAC) -12 Poly.Myristic (Myristicacid prediction acid MF-UNIFAC) -14 Poly.Palmytic (Palmyticacid prediction acid MF-UNIFAC) Poly. (Stearic acid MF-UNIFAC) -16 Stearic acid prediction 50 100 150 200 250

Temperature (°C)

Figure 4.13: Solubility of fatty acids in SBCW as published by Khwujitjaru et al. [24] and predicted by the MF-UNIFAC model with updated COOH interaction parameters

94

4.4.2. GRISEOFULVIN AND PYRIMETHAMINE

Griseofulvin contained both CH3O and CHO sidegroups, which do not require optimization (as demonstrated in Section 4.4). Thus, the only interaction parameter that was not properly accounted for was the aromatic chlorine side-group. Pyrimethamine also had a chlorine side- group. However, the solubility of pyrimethamine is affected by both the nitrogen as well as the chlorine. The nitrogen sidegroups have not yet been optimized, which means that an optimization minimizing the error on the chlorine group included in pyrimethamine would not be accurate. Thus, the minimization on the sum of square errors of the MF-UNIFAC solubility output due to the chlorine side group was conducted for griseofulvin only, as the parameters for all side-groups apart from chlorine have been optimized. The optimized chlorine-water interaction parameters were then applied to predict the solubility of pyrimethamine in SBCW. The original and fitted MF-UNIFAC parameters are shown in Table 4.10.

The average error of griseofulvin was reduced from 64% to 2% and the average error of pyrimethamine was reduced from 46% to 29%. A graph of the fitted curves is shown in Figure 4.14. While the accuracy of the MF-UNIFAC model outputs for pyrimethamine was improved, the solubility output error was still large. Further refinements of the interaction parameters for pyrimethamine are needed. For griseofulvin the modified MF-UNIFAC model outputs described the solubility data well between 25°C and 200°C.

Table 4.10: Original and fitted water-chlorine acid (side-group 11,19) and chlorine-water (side- group 19,11) interaction parameters.

Original M-UNIFAC Fitted M-UNIFAC

19,11 Cl-H2O 11,19 H 2O Cl 19,11 Cl-H2O 11,19 H2O-Cl a 591.6 -1895 591.6 -1895 b -3.08 9.33 -3.08 9.33 c 0 0 -0.0016 -0.0052

95

0 GriseofulvinGriseofulvin ExperimentalExperimental -2 PyrimethaminePyrimethamine Experimental -4 Experimental Griseofulvin Mod. MF-UNIFAC -6 Poly. (Griseo fitted MF-UNI)

-8 Poly.Pyrimethamine (Pyri fitted MF Mod.-UNI) MF-UNIFAC ) 2 -10 ln(x -12

-14

-16

-18

-20 0 20 40 60 80 100 120 140 160 180 Temperature (°C)

Figure 4.14: Solubility of griseofulvin and pyrimethamine and the MF-UNIFAC model with optimized chlorine-water interaction parameters

Correction for the nitrogen groups in pyrimethamine could not be conducted because the pyrimidine interaction parameter does not exist yet in the published literature. Creation of a pyrimidine side-group requires a large number of -liquid equilibrium data to construct the coefficients [1], and to do so is outside the scope of this thesis.

4.4.3. BUDESONIDE AND 9-ANTHRACENEMETHANOL

The OH-water interaction parameters were optimized to the values shown in Table 4.11. The result of the changed interaction parameters is shown in Figure 4.15. The average error of the MF-UNIFAC model outputs was reduced from to 30% to 5%. While the fit was sufficient between 100°C and 170°C, the trend produced an error of 12% at room temperature. Though the error at room temperature is large, it is an improvement over the original MF-UNIFAC model.

96

Table 4.11: Original and fitted water-OH (side-group 11,10) and OH-water (side-group 10,11) interaction parameters.

Original M-UNIFAC Fitted M-UNIFAC

(10,11) OH-H2O (11,10) H2O-OH (10,11) OH-H2O (11,10) H2O-OH a -801.9 1460 -801.9 1460 b 3.824 -8.673 3.824 -8.673 c -7.51E-03 1.64E-02 -8.27E-03 1.62E-02

0 BudesonideBudesonide Data Experimental -2 Poly.Budesonide (BudesonideMod Opt MF -UNIFAC -4 UNIFAC)

-6

) -8 2

ln(x -10

-12

-14

-16

-18 0 50 100 150 200 250 Temperature (°C)

Figure 4.15: Solubility of budesonide and the MF-UNIFAC model with optimized OH-water interaction parameters

There was no solubility data available in the literature for which the corrected interaction parameters were applicable. In order to determine the applicability of the optimized OH-water interaction parameter, the solubility of 9-anthracenemethanol was established using the solubility method described in Chapter 3. The chemical structure of 9-anthracenemethanol is shown in Figure 4.16. The sidegroup subdivision was the same for 9-anthracenemethanol as it was for anthracene, except with an additional OH sidegroup and an additional CH3 sidegroup. The mass of the solubilized 9-anthracenemethanol was determined using UV spectrometry with DMSO as the solvent. The solubility results are shown in Table 4.12, as well as the error of the original MF-UNIFAC model and the MF-UNIFAC model with altered OH-water interaction parameters.

97

Figure 4.16: Chemical structure of 9-anthracenemethanol

Table 4.12: 9-anthracenemethanol solubility from 25°C to 170°C and MF-UNIFAC and budesonide- optimized MF-UNIFAC (MFB-UNIFAC) model errors

x2 (106) SD (106) Error Error T (mole (mole MF- MFB- Ref (°C) fraction) fraction) UNIFAC UNIFAC 25 7.74 - 22% 5% [25] 100 2.25 0.003 16% 27% 120 15.5 0.80 11% 25% 140 37.5 2.40 14% 31% 150 621 14.0 11% 13% 160 984 95.0 11% 17%

The average error of the MF-UNFAC model outputs was increased when the budesonide- corrected OH-water interaction parameters were applied, though the data at room temperature had a higher accuracy in the modified MF-UNIFAC model output. It would be beneficial to establish solubility data of more organic compounds with OH sidegroups in SBCW to further optimize the OH -water interaction parameter.

4.5. ALTERNATIVE MODEL BASED ON THE DIELECTRIC CONSTANT OF WATER

One of the outcomes of the dependency of solubility on the dielectric constant, reported in Chapter 3, is that a generalized solubility model can be made of the form shown in Equation 4.26, where D may be calculated from the relationship correlated by Akerlof [10], and A and B are the linear constants fitted from a linear regression preformed on the solubility and dielectric constant data.

98

Equation 4.28

For budesonide:

The A and B values were used to calculate the solubility of budesonide from 25°C to 200°C in 0%, 5% and 20% ethanol/water solution. The number of significant figures used for the constants was determined by level of accuracy of the solubility output (i.e. to 3 significant figures). The average error (calculated using Equation 4.25) of the Dielectric Constant model over the entire range tested is 0.14%. A comparison of the model outputs to experimental data for budesonide in pure SBCW is shown in Figure 4.17.

0 ExperimentalSolubility Data

FittedFitted MF MF-UNIFAC-UNIFAC -5 Poly.MF -(MFUNIFAC-UNIFAC) Poly.Original (OriginalM-UNIFAC UNIFAC) -10 Poly.Dielectric (Carr Model) Constant Model ) 2 ln(x -15

-20

-25 0 50 100 150 200 250

o Temperature ( C)

Figure 4.17: M-UNIFAC, MF-UNIFAC and Dielectric Constant models of budesonide

The accuracy of the predicted solubility values of the Dielectric Constant model was tested against the solubility of budesonide in a 10% methanol/SBCW solution. The solubility of budesonide in a 10% (v/v) methanol-water solution was predicted using Equation 4.28. The solubility of budesonide in a 10% (v/v) methanol-SBCW was measured between 100°C and

99

160°C, according to the method described in Chapter 3 (Section 3.2.1). The model outputs and the solubility data are shown in Figure 4.18. The predicted solubility of budesonide in the methanol/SBCW solution had an average error of 3%.

1.2E-04

Model 1.0E-04 Experimental

8.0E-05

6.0E-05

4.0E-05 Solubility(mol/mol))

2.0E-05

0.0E+00 0 50 100 150 200 Temperature (°C)

Figure 4.18: Experimental and predicted solubility data for budesonide. Solubility predictions were calculated from Equation 4.20 with the parameters A and B fitted from 0% to 20% ethanol/SBCW solubility data. The line for the solubility model was added as a guide to the eye

4.5.1. DIELECTRIC CONSTANT MODEL FITS FOR OTHER APIS

A Dielectric Constant Model constructed for each of the APIs and test compounds dissolved in SBCW in this work. The fitted parameters, R2 value, correlation coefficient and average error (calculated using equation 4.25) for each compound is shown in Table 4.13. The average error over all temperatures for every compound is 3.20%.

Table 4.13: Dielectric Model parameters and errors

Compound A B R2 Average Error (%) Griseofulvin -0.346 12.1 0.999 0.3 Naproxen -0.330 13.1 0.971 5.5 Pyrimethamine -0.276 7.97 0.982 3.0 Budesonide -0.233 3.27 0.998 0.3 Anthracene -0.440 14.7 0.966 4.7 9-anthracenemethanol -0.817 40.2 0.924 5.1 Average model error - - - 3.2

100

A comparison of Dielectric Constant model outputs to experimental data for each compound over the temperature range tested is shown in Figure 4.19. The model is accurate from 25°C to 200°C. While the Dielectric Constant model is more accurate than the M-UNIFAC model, the dielectric constant model is fitted to experimental data and as such cannot predict of any compound in SBCW without solubility data points.

Figure 4.19: Solubility data and Dielectric Constant model of solubility for a) naproxen, b) griseofulvin, c) pyrimethamine, d) anthracene and e) 9-anthracenemethanol

101

Although the Dielectric Constant model requires SBCW solubility data to produce a solubility curve, it is still useful for the prediction of the solubility at a range of temperatures, and with different levels of co-solvents. The implications of this are that a limited number of solubility studies may be done for any HOC in SBCW, and the solubility at another condition may be accurately predicted using the Dielectric Constant model. Future work on further solidifying the applicability of the Dielectric Constant model to a variety of organic solvents and HOCs is discussed in Chapter 6.

4.6. MODELLING CONCLUSIONS

The solubility of 6 compounds in SBCW was modelled using a variety of different M-UNIFAC models and a new empirical model based on the dielectric constant of water and modified water-mixtures. Accuracy of the M-UNIFAC model has been significantly improved for all compounds from the original and modified M-UNIFAC models, to within 10% of the solubility results obtained from experiments. The errors commonly encountered over a large temperature range have been minimized. It was found that some oxygen-water interaction parameters do not require further optimization based on the recent solubility data presented in the literature.

A new model based on the dielectric constant of water and water modified with organic solvents was constructed, and had an average error of 3.2% for all compounds at all temperatures tested in this work. The work presented in this chapter suggests that the application of the Dielectric Constant model to a host of different organic compounds with organic modifiers in the SBCW solution is possible.

102

4.7. REFERENCES

1. Fredenslund, Aa., Jones, R.L. and Prausnitz, J.M., Group-contribution Estimation of Activity Coefficients in Nonideal Liquid Mixtures. AIChE Journal, 1975. 21(6): pp. 1086-1099.

2. Fornari, T., Stateva, R.P., Señorans, F.J., Reglero, G. and Ibañez, E., Applying UNIFAC-based Models to Predict the Solubility of Solids in Subcritical Water. The Journal of Supercritical Fluids, 2008. 46(3): pp. 245-251.

3. Miller, D.J., Hawthorne, S.B., Gizir, A.M. and Clifford, A.A., Solubility of Polycyclic Aromatic Hydrocarbons in Subcritical Water from 298 K to 498 K. Journal of Chemical & Engineering Data, 1998. 43(6): pp. 1043-1047.

4. Gmehling, J., Li, J. and Schiller, M., A Modified UNIFAC Model. 2. Present Parameter Matrix and Results for Different Thermodynamic Properties. Industrial & Engineering Chemistry Research, 1993. 32(1): pp. 178-193.

5. Ferreira, O., Macedo, E.A., and Bottini, S.B., Extension of the A-UNIFAC Model to Mixtures of Cross- and Self-associating Compounds. Fluid Phase Equilibria, 2005. 227(2): pp. 165- 176.

6. Ferreira, O., Brignole, E.A., and Macedo, E.A., Phase Equilibria in Sugar Solutions using the A-UNIFAC Model. Industrial & Engineering Chemistry Research, 2003. 42(24): pp. 6212- 6222.

7. Miller, D.J. and Hawthorne, S.B., Method for Determining the Solubilities of Hydrophobic Organics in Subcritical Water. Analytical Chemistry, 1998. 70(8): pp. 1618-1621.

8. Smith, R.M., Superheated Water: The Ultimate Green Solvent for Separation Science. Analytical and Bioanalytical Chemistry, 2006. 385(3): pp. 419-421.

9. Smith, R.M. and Burgess, R.J., Superheated Water: A Clean Eluent for Reversed-Phase High- Performance Liquid Chromatography. Analytical Communications (Print), 1996. 33(9): pp. 327-329.

10. Akerlof, G., Dielectric Constants of some Organic Solvent-Water Mixtures at Various Temperatures. Journal of the American Chemical Society, 2002. 54(11): pp. 4125-4139.

11. Hansen, C.M., Hansen Solubility Parameters, A Users Handbook. , Taylor and Francis Group, Boca Raton, Florida, USA, 2000: CRC Press.

103

12. Prausnitz, J.M., Lichtenthaler, R.N., Azevedo, E.G. de, Molecular Thermodynamics of Fluid Phase Equilibria, Second Edition, 1986, Prentice Hall, New Jersey, United States .

13. Karásek, P., Planeta, J. and Roth, M., Solubility of Solid Polycyclic Aromatic Hydrocarbons in Pressurized Hot Water: Correlation with Pure Component Properties. Industrial & Engineering Chemistry Research, 2006. 45(12): pp. 4454-4460.

14. Karásek, P., Planeta, J. and Roth, M., Solubility of Solid Polycyclic Aromatic Hydrocarbons in Pressurized Hot Water at Temperatures from 313 K to the Melting Point. Journal of Chemical & Engineering Data, 2006. 51(2): pp. 616-622.

15. Ahmed, H., Buckton, G. and Rawlins, D.A., Crystallisation of Partially Amorphous Griseofulvin in Water Vapour: Determination of Kinetic Parameters using Isothermal Heat Conduction Microcalorimetry. International Journal of Pharmaceutics, 1998. 167(1-2): pp. 139-145.

16. Bettinetti, G., Mura, P., Faucci, M.T., Sorrenti, M. and Setti, M., Interaction of Naproxen with Noncrystalline Acetyl [beta]- and Acetyl [gamma]-Cyclodextrins in the Solid and Liquid State. European Journal of Pharmaceutical Sciences, 2002. 15(1): pp. 21-29.

17. de Araujo, M.V.G., Macedo, O.F.L., Nascimento, C.d.C., Conegero, L.S., Barreto, L.S., Almeida, L.E., de Costa Jr, N.B. and Gimenez, I.F., Characterization, Phase Solubility and Molecular Modeling of [Alpha]-Cyclodextrin/pyrimethamine Inclusion Complex. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2009. 72(1): pp. 165-170.

18. Su, C.S. and Chen, Y.P., Correlation for the Solubilities of Pharmaceutical Compounds in Supercritical Carbon Dioxide. Fluid Phase Equilibria, 2007. 254(1-2): pp. 167-173.

19. Smith, J.M., Van Ness, H.C. and Abbott, M.M., Chemical Engineering Thermodynamics. 6 ed. 2001: McGraw Hill.

20. Magnussen, T., Rasmussen, P. and Fredenslund, Aa., UNIFAC Parameter Table for Prediction of Liquid-liquid Equilibriums. Industrial & Engineering Chemistry Process Design and Development, 1981. 20(2): pp. 331-339.

21. Hafasoui, S.L. and Mahmoud, R., Solid-liquid Equilibria of Binary Systems Containing n - tetracosane with Naphthalene or Dibenzofuran: Prediction with UNIFAC Model Journal of Thermal Analysis and Calorimetry, 2007. 88(2): pp. 565-570.

104

22. Gmehling, J., Wittig, R., Lohmann, J. and Joh, R., A Modified UNIFAC (Dortmund) Model. 4. Revision and Extension. Industrial & Engineering Chemistry Research, 2002. 41(6):pp. 1678-1688.

23. Karásek, P., Planeta, J. and Roth, M., Solubilities of Oxygenated Aromatic Solids in Pressurized Hot Water†. Journal of Chemical & Engineering Data, 2009, 54 (5): pp.1475- 1461.

24. Khuwijitjaru, P., Adachi, S. and Matsuno, R., Solubility of Saturated Fatty Acids in Water at Elevated Temperatures. Biosci. Biotechnol. Biochem, 2002. 66(8): pp. 1723-1726.

25. Southworth, G.R. and Keller, J.L., Hydrophobic Sorption of Polar Organics by Low Organic Carbon Soils. Water, Air, & Soil Pollution, 1986. 28(3): pp. 239-248.

105

5. PARTICLE FORMATION

In this chapter, a proof of concept for the application of SBCW to particle engineering with an emphasis on processing APIs is presented. Model APIs budesonide, griseofulvin and naproxen have been successfully micronized via the newly developed technique. Comparisons of the results with similar morphology particles from other rapid precipitation technologies are presented, as well as suggestions for future work based on the sensitivity of the process to certain variables.

The potential of combining the subcritical water micronization technology with a common spray- drying process to produce pharmaceutical particles suitable for inhalation drug delivery has been evaluated. A model powder of budesonide and lactose was produced by the SBCW micronization process, and then spray dried. The aerodynamic performance of the dried product was tested. Results demonstrated that the SBCW micronization technique was suitable for the production of inhalable particles.

106

5.1. PARTICLE ENGINEERING USING SBCW AS A SOLVENT

The benefits of tuning the size and shape of a particle to enhance the bioavailability of an API were discussed in Chapter 1. Two methods of controlling particle size were presented: the comminution/milling method and the rapid precipitation method. Rapid precipitation techniques have an advantage over milling techniques in that they do not require high shear forces to comminute particles, which can degrade the surface properties of a drug [1]. In a typical rapid precipitation method, a material is dissolved in a solvent and then crystallized from the molecular level up to a nano- or micro- level. Rapid precipitation techniques include spray drying (SD) [2], controlled crystallization (CCP) [3] and supercritical fluid precipitation (SCF) techniques [4-6]. Many of these methods have achieved particle sizes within a desirable size range (which differs depending on the application of the drug). Moreover, the precipitation methods are able to be tailored to produce amorphous [2] or crystalline[6] particles.

Most rapid precipitation techniques rely on organic solvents to dissolve organic compounds. Ideally, rapid precipitation methods should use non-toxic organic solvents, such as water or ethanol, to dissolve drug particles. Ethanol and water at room temperature are both highly polar compounds. The polar nature of the fluids limits these solvents to dissolving polar compounds only. Considering that the majority of commercial drugs are hydrophobic [7], polar solvents have only limited applications for the pharmaceutical processing industry.

Non-polar organic solvents like dichloromethane, methanol and hexane are often used dissolve APIs and excipients [3, 5-6]. The major disadvantage of using toxic organic solvents like dichloromethane and methanol for rapid precipitation is the difficulty of solvent removal. Solvent removal is expensive, particularly given the necessarily low levels of trace solvent allowable in the final formulation [8]. Replacement of these organic solvents with subcritical water is potentially a viable alternative, particularly based on the fact that a number of APIs are stable in SBCW up to 200°C, as reported in Chapter 3.

The advantages of using subcritical water technology for particle formation have been discussed in other sections of this thesis. In this chapter, the methodologies of particle formation are presented in more depth. Three methods are developed:

1. Particle formation by rapid temperature quench of a SBCW-API solution 2. Particle formation by rapid temperature quench of a SBCW-API solution in excipient- laden water

107

3. Particle formation by rapid expansion of a SBCW solution into a vacuum chamber

The first two methods induce precipitation by temperature quench to give fine particles in a suspension. The last method induces precipitation by rapidly removing SBCW to give a dry powder product.

A method using SBCW to micronize hydrophobic compounds was shown schematically in Figure 1.5. By mixing a SBCW solution with water at room temperature, a rapid temperature quench of the solution and a reduction in the polarity of the solvent results. The change in polarity can induce a supersaturation of the solution, which can lead to rapid precipitation of the solute. The yield of the precipitation process will be determined, in part, by the difference in solubility of the solute in SBCW and at quench conditions.

5.2. METHODS

5.2.1. MATERIALS

Griseofulvin (purum, 95%), budesonide (>99% purity) and polyethylene glycol (average molecular weight 400amu, PEG400) were purchased from Sigma Aldrich. Naproxen (98% purity) and anhydrous lactose were purchased from Fluka. HPLC grade methanol, HPLC grade ethanol and reagent grade acetone was purchased from UNIVAR. De-ionized water was used for all experiments.

5.2.2. INJECTION INTO COLD WATER PARTICLE FORMATION METHOD

The design of the particle formation apparatus was similar to the solubility apparatus (shown in Section 3.2). A schematic of the equipment used is shown in Figure 5.1. A tee permitted the flow of SBCW solution from the SV to a collection vessel where precipitation occurred. The line to the precipitation vessel was 1/16” OD stainless steel tubing. The line was located inside the oven to minimize temperature changes prior to PC. A nozzle with an ID of 1mm was used to deliver the SBCW solution into PC. The flow from the SV to the precipitation vessel was controlled by V3 (Figure 5.1). The pressure of the system was controlled by the opening/closing of V4.

108

The collection vessel (PC) was filled with 60 ml of water at room temperature and a backpressure of 20 bar was applied to ensure that the water remained in the liquid state throughout the experiment.

Figure 5.1: Subcritical water particle formation apparatus

0433588107 A weighed amount of API was loaded into the SV. The system was filled with water and pressure and temperature were equilibrated as described for solubility measurements. The system was then stirred for 10 minutes and subsequently contacted with nitrogen at 72 bar. The valve V3 was then cracked open ¼ turn to allow the solution to flow to the PC vessel. The flow was stopped once the first nitrogen bubble was seen. The pressure reading of the sight gauge was continually monitored to ensure a constant backpressure was maintained.

The content of the PC vessel was collected using a plastic syringe and transferred to a glass collection vial. The API was separated from the water suspension by vacuum filtration through a 0.45μm HV hydrophilic Millipore membrane filter. The vacuum was provided by an Adixen Pascal 2000SD vacuum pump.

109

5.2.3. PRECIPITATION IN THE PRESENCE OF EXCIPIENTS

One of the benefits of the SBCW micronization technique is that there is potential to process a number of APIs and pharmaceutical excipients simultaneously. By doing this, it may be possible to produce a pharmaceutical formulation in a single processing step. The mass ratio of API to the excipient can be controlled as the solubility of each API in SBCW is known (see Chapter 3). The work presented in this section was an investigation of whether the presence of excipients during the precipitation of the API in the PC has an effect on particle morphology.

The effect of two commonly used pharmaceutical excipients, lactose and polyethylene glycol 400 (PEG400), on the morphology of the precipitated model API particles was investigated. The materials were selected based on the common usage of both materials as pharmaceutical excipients. Both Lactose and PEG400 were selected because they are commonly used carriers for naproxen and budesonide [9-13]. The excipients were added to PC in proportions similar to those found in common API formulations. These ratios are outlined in Section 5.2.3.1 and Section 5.2.3.2.

5.2.3.1. Precipitation in the Presence of Lactose

Lactose was dissolved in the PC chamber water to investigate the effect a common pharmaceutical excipient had on the morphology of the precipitate. The effect of injection temperature of the SBCW solution and of the presence of lactose in the precipitation vessel on the morphology of naproxen and budesonide was also investigated.

The SBCW-naproxen solutions were prepared at 130°C, 160°C and 170°C. Lactose/water solutions were prepared by dissolving 1%w/v lactose in de-ionized water, which gives a similar proportion of naproxen:lactose ratio found in the published literature [9]. To investigate the effect of lactose on product morphology, 60mL of the solutions were placed inside the PC vessel.

Experiments were also conducted on budesonide. Lactose is a commonly used carrier/excipient for the delivery of budesonide into the lungs [10]. The therapeutic ratios have been described in the literature as 1:15 and 1:30w/w ratios of budesonide API to lactose[11]. Thus, for each temperature, a mass of lactose was added to the 60mL water in PC, where the mass of lactose added was 15 or 30 times the mass of budesonide injected into PC at each temperature (based on the solubility data reported in Chapter 3).

110

5.2.3.2. Precipitation in the Presence of PEG

PEG has been used as an excipient for the pulmonary delivery of budesonide into the lungs [10, 12]. The effect of PEG as a modifier for particle formation in the PC was investigated on the morphology of budesonide only. PEG is added to budesonide at 4:1 and 8:1w/w ratios (PEG:budesonide), as they are common ratios used in budesonide formulations [13]. A mass of PEG was added to the water in PC in the aforementioned ratios, based on the solubility data of budesonide in SBCW (determined in Chapter 3).

5.2.4. SBCW-SOLUTE SPRAY PARTICLE FORMATION METHOD

A spray apparatus to produce dry particles from SBCW-solutions was devised and is shown schematically in Figure 5.2. A 150mL stainless steel capture vessel (SB) was fitted at the outlet of V2 inside the oven. The outlet of SB was connected to 1/8” stainless steel tubing fitted with a 0.5µm filter stone. The tubing was connected to a venturi tube attached to a water supply. The venturi tube provided a maximum vacuum of 17.5mmHg. The purpose of the vacuum was to assist in the vaporization of the injected solution.

The precipitate was collected from SB after the system had cooled to room temperature. In some cases water was not completely removed from the precipitate and the powder was dried further by a flow of warm air (30°C) for 5 minutes.

111

Figure 5.2: Spray apparatus for the formation of dry particles from SBCW solutions

5.2.5. SEM AND LS METHODS

A Malvern Mastersizer S was used to determine the particle size of naproxen. Particle size and size distribution (PSD) analyses were carried out on suspensions taken directly from the particle formation vessel. Samples produced via the spray technique were captured by washing the internals of SB with distilled water; the resulting suspension was collected and analyzed.

Results are reported as a number median diameter (X50) of a distribution. While the volume mean diameter is typically used to size particles, the number mean diameter gives a more accurate representation of the actual sizes of the particles produced from this work. The number mean is also comparable to the sizes of the particles found from SEM images.

A Brookhaven DLS was used to determine the particle size distributions of budesonide in water. Suspensions were collected directly from the PC and injected into the cuvettes for analysis.

Results are reported as the X10, X50, and X90 number diameter (corresponding to the diameter of particles at the 10th percentile, 50th percentile and 90th percentile respectively).

A Hitachi S900 SEM was used to image the product. The API powders were dispersed onto double sided carbon tape and then placed on sample holders. Samples were chromium coated.

112

5.2.6. XRD METHOD

The XRD machine used was a Philips multi-purpose X-ray diffraction system (MPD). The API powder was placed on a polished iron sample holder. The beam angle was varied from (2θ = ) 6° to 60° with a 0.0206 step size. The X-ray generator was set at 45kV and 40mA. The diffraction patterns of the processed materials were analyzed and compared to the diffraction patterns of the raw material to evaluate changes in the crystal structure.

5.2.7. DSC METHOD

A TA Instruments differential scanning calorimeter (DSC) 2010 was used. Dried powder samples (5 to 10mg) were loaded into an aluminum pan and sealed using a DSC-disc press. The samples were cooled to-50oC and the temperature was ramped at 10°C/min to 300°C in a nitrogen atmosphere.

5.2.8. SPRAY DRYING METHOD

A spray dryer was used to dry lactose solutions and budesonide suspensions in lactose solutions. A Buchi 290 spray dryer was used. The spray nozzle had an ID of 1/16”. The inlet temperature was set to 170°C. The compressed air flow was set to 40 on the instrument gauge. The flowrate of the liquid solutions/suspensions was delivered at 15% pump rate (gauge reading). The aspirator was set to 100% (gauge reading). Before injecting the test solutions to the drying chamber, water was injected for 25 minutes, which was the time required to stabilize the outlet temperature of the drying chamber. The outlet temperature for all experiments was recorded for all spray dried samples, which differed for each solution.

Experimental conditions were kept constant between samples, as the purpose of the spray dryer was to produce a powder post-SBCW processing and not to optimize the spray drying process.

113

5.2.9. AERODYNAMIC PARTICLE SIZING METHOD

A model 8301 Anderson Cascade Impactor was used to measure the aerodynamic particle sizes of spray dried budesonide and lactose suspensions/solutions. A basic schematic of the Cascade Impactor is shown in Figure 5.3. The experiments were run with 8 stages (stages 0 – 7) at 28.3L/min [11]. Each stage corresponded to an aerodynamic particle size, and a corresponding stage in the human respiratory system. In order to properly simulate a standard breath (4L of air per breath), a flowtime of 8.5s was used. All of the powders tested in the Cascade Impactor were sealed within a gelatin capsule (capsule size 3GS, purchased from Capsugel). A standard inhaler was used to break the pill and subject the contents to a flow of air. A Copley vacuum pump and TPK solenoid control valve were used to control the vacuum flowrate (measured with a Copley flowmeter) and inhalation pressure for all experiments [14].

Figure 5.3: Standard setup of the Andersen Cascade Impactor

114

In order to establish the mass of the powders at each stage, the plate at each stage was weighed pre- and post-experiment. The mass of material at each stage was converted into a mass fraction. When budesonide was deposited on the plates, the masses were also measured using UV spectrometry. Each plate was put in a glass Petri dish. 4mL of methanol was added to the dish to dissolve the budesonide. The plates were washed in the methanol and then removed. The methanol-budesonide solution was then transferred by pipette into a 10mL grade-A volumetric flask. The volume was made up to 10mL in the volumetric flask and the concentration of budesonide in the API-methanol solution was determined through UV spectrometry using the Beer-Lambert law.

In order to measure the concentrations of lactose and budesonide in a spray dried formulation, both the mass difference method and the UV method were used. Methanol is a poor solvent for lactose, and a good solvent for budesonide [15]. Thus UV spectrometry could be carried out accurately for the methanol-budesonide solutions, as lactose was not present in the methanol. The proportion by mass of lactose in the formulation was determined by subtracting the difference of the weighed plates with the deposited budesonide and lactose on them and the UV results for budesonide.

In order to compare the micronized budesonide/spray dried lactose product to an unprocessed product, a blend of the raw budesonide and lactose was prepared. Budesonide and lactose were blended in a 1:4 budesonide:lactose mass ratio. The budesonide and lactose were combined using a mortar and pestle.

115

5.3. RESULTS

5.3.1. PRECIPITATION INTO A COLD WATER-FILLED VESSEL

5.3.1.1. Griseofulvin

Griseofulvin is an antifungal drug that is typically delivered in a tablet form[16]. Commonly, it is used to treat tineal infections, such as ringworm, and athlete’s foot. The oral bioavailability of griseofulvin is typically 40%[17]. Upon micronization, the oral bioavailability of griseofulvin can increase to 60%[18]. The improvement in overall bioavailability of micronized and ultra- micronized griseofulvin has caused the drug to be subject to micronization and formulation via many different rapid precipitation techniques, particularly supercritical fluid methods [18-23].

The limitations, in terms of organic solvent usage, of micronization via rapid precipitation have already been identified in Chapter 2. Typically, methylene chloride is used as a solvent for rapid precipitation process to micronize griseofulvin [21]. Replacing methylene chloride with SBCW has obvious environmental and economic benefits. The stability of griseofulvin in subcritical water up to 200°C in SBCW has already been reported in Chapter 3. The effects of injection temperature and concentration on the morphology of the precipitated griseofulvin particles are presented here. Experimental conditions and results are summarized in Table 5.1.

Table 5.1: Precipitation experiment variables and results

Temperature Concentration Product Size Experiment of SBCW ×104 F* morphology (m) solution (oC) (mole fraction) G1 140 1.6 1 Bi-pyramidal 20** G2 160 2.9 1 Plate-agglomerates 35** G3 170 5.3 1 Bi-pyramidal 15** G4 170 2.7 0.5 Bi-pyramidal 10-20 G5 170 1.8 0.25 Bi-pyramidal 15

* F  Cexperiment /Csaturated ** X50 number average measured by light scattering

The particle size distributions (by number) of griseofulvin crystals from experiments G1, G2, and G3 are shown in Figure 5.4. The particle size distributions were unimodal in all cases.

116

Figure 5.4: Particle size distribution of griseofulvin produced by SBCW micronization from saturated SBCW solutions at different temperatures.

Griseofulvin crystals have been produced by compressed fluid anti-solvent (CFA) processes [20- 21]. CFA processes use a conventional solvent to dissolve the solute. Solute precipitation is triggered by contacting the solution with a compressed fluid antisolvent which expands and extracts the conventional solvent [24]. Griseofulvin has been crystallized from acetone [20] and chloroform [21] by using carbon dioxide as the antisolvent. Precipitation is controlled by the diffusion of the supercritical fluid into the organic solvent (which decreases solute solubility), and the transfer of the organic solvent into the compressed fluid phase (which increases solute concentration). Thus the fluid dynamics of the CFA process have a significant impact on the morphology of the product. While our work may be governed by different mechanisms than CFA processes (such as the rate of heat transfer), the fluid dynamics of the injected solution is likely to strongly affect the morphology of the precipitate. Thus the CFA process and the SBCW process are compared based on the fluid dynamics of each method.

117

Figure 5.5: Griseofulvin precipitated in water at 20bar from saturated SBCW solutions at: a) 140°C b) 160°C and c) 170°C.

Griseofulvin crystals are generally needle shaped. However bi-pyramidal crystals were produced in experiments G1, G3, G4, and G5. Both morphologies correspond to a tetragonal crystal system and are the same polymorph [20]. Bi-pyramidal crystals form when an even growth rate along all surfaces occurs, which in turn results from efficient mixing in the particle formation environment.

Similar observations were made from compressed fluid antisolvent processes. A comparison between products from this work and from the compressed fluid antisolvent processes by De Gioannis et al. [20] and Jarmer et al. [21] are shown in Table 5.2. De Gioannis et al. investigated the effect of mixing procedure whilst Jarmer et al. elaborated on the effect of the additive poly sebacic anhydride (PSA) on griseofulvin morphology.

Table 5.2: Precipitation methods for crystalline griseofulvin micro-particles

Crystal Conditions Dominant Crystal Method length Reference Morphology (μm)† Slow stirring –slow addition of CFA Needle 1000 [20] antisolvent Fast stirring–fast addition of CFA Bi-pyramidal 300 [20] antisolvent CFA 35:1 griseofulvin:PSA Bi-pyramidal 50 [21] SBCWM 140°C Bi-pyramidal 10 This work Plate SBCWM 160°C 20 This work agglomerates SBCWM 170°C Bi-Pyramidal 5 This work Note: CFA: compressed fluid antisolvent process; PSA: poly sebacic anhydride; SBCWM: subcritical water micronization †: Crystal length is compared using the size obtained from SEM images. This is to keep the sizing method consistent between this article and Jarmer et al.[21]

118

De Gioannis et al. found that fast stirring and high fluxes of CO2 in the precipitation chamber resulted in bi-pyramidal crystals whereas low stirring rates and slow introduction of CO2 resulted in needle-like crystals. High stirring rates slowed down crystal growth in the preferred direction, resulting in an even growth rate on all crystal faces. High introduction rates of CO2 can generate faster supersaturation of the solution, thereby contributing to the formation of small particles. Thus small bipyramidal crystals may be precipitated when homogeneous heat and mass transfer are achieved [20].

Jarmer et al. found that a growth inhibitor was required to produce bi-pyramidal crystals of griseofulvin. It was suggested that the PSA selectively adhered to the fastest growing crystalline face of griseofulvin, thus equilibrating crystal growth in all directions. It was thus shown that PSA acted as a crystal growth inhibitor, which prevented the precipitation of needle crystals forming.

The bi-pyramidal crystals produced by SBCW micronization may reflect a high degree of homogeneity throughout the particle formation stage. In particular, it may indicate efficient mass and heat transfer rates during the mixing of the SBCW solution stream with the water in the particle formation vessel. The particle size range produced is an indication of the high super-saturation achieved. In the SBCW micronization process, the level of supersaturation is determined by the difference in solubility at subcritical conditions and at room conditions. Since the solubility of griseofulvin in subcritical water at 170°C is almost double than at 140°C; crystals from experiment G3 were produced with a degree of supersaturation double than crystals from experiment G1. Figure 5.4 shows that particles produced from experiment G3 had number average particle size of almost half that of the particles produced from experiment G1. The effect of supersaturation level on particle size may have been enhanced by the larger temperature quench of experiment 3 (170 to 21°C) compared to experiment 1 (140 to 21 °C). The corresponding faster heat removal may have triggered a more rapid precipitation.

Particles formed at 160°C (experiment G2) were about 10 microns in length with an agglomerated, plate-like morphology (Figure 5.5b). The peculiar morphology obtained from experiment G2 may be related to a discontinuity in water properties at 160°C, where peculiar solute-water interactions take place. It has been documented that at 160°C hydrogen bonding cages that form around solute molecules in a water solution disappear[25].

In experiments G3, G4, and G5 the effect of supersaturation level at constant temperature was investigated. The micronization process was conducted at 170°C at three different

119 concentration levels: 100%, 50 % and 25% of the saturation concentration at 170°C. The results obtained are shown in Figure 5.6.

Figure 5.6: Griseofulvin precipitated from SBCW solutions at 170°C and different concentrations: a) 5.3 × 10-4 mol/mol, b) 2.6 × 10-4 mol/mol and c) 1.3 × 10-4 mol/mol.

Experiments G4 and G5 generated crystals with a prolonged dimension, which is an indication of a limited heat and mass transfer rate. It is possible that the lower saturation levels used in experiments G4 and G5 corresponded to the growth of a reduced number of crystallization nuclei. The crystallization process is generally slower for low supersaturation levels, and thus mass and heat transfer are more likely to limit the process [26].

Particles from experiment G5 were hollow (Figure 5.6c). The lack of griseofulvin deposition in the inner part of the crystal may indicate insufficient mass and heat transfer in the narrow channels within the structure of the forming crystals [27].

120

Figure 5.7: X-ray diffraction of: a) bi-pyramidal crystals from SBCW processing, b) needle-like crystals from SBCW processing and c) raw griseofulvin.

X-ray diffraction and DSC were performed on SBCW produced griseofulvin crystals and on the raw material results are reported in Figure 5.7 and Figure 5.8, respectively. Results indicate that all the morphologies produced by SBCW processing and the raw material were the same polymorphic form of the API.

The XRD of griseofulvin samples from experiment G2 (needle-like morphology) exhibited peculiarities. The peaks at 13o and 26o were more intense than in the other samples, whilst the peak at 14o was less intense (Figure 5.7). Irregularities in the crystal structure may be the reason for the peculiarities in the XRD profile, as the changes appear reasonably small. The irregularities did not affect the thermal profile of any of the products, as shown in Figure 5.8, thus confirming that the irregularities have negligible effect on the crystal lattice bond strength.

121

Figure 5.8: DSC diagrams for griseofulvin crystals: a) bi-pyramidal crystals from SBCW processing b) needle-like

5.3.1.2. Naproxen

The experimental conditions for the precipitation experiments of saturated SBCW-naproxen solutions injected into cold water are summarized in Table 5.3. Each experiment was carried out in duplicate or triplicate. In experiments N1 and N2 the particle morphology was unimodal, and had a broad particle size distribution (between 0.5µm and 100µm). Flake-plate crystals were precipitated (shown in Figure 5.9a). Similar morphologies of naproxen have been precipitated in supercritical antisolvent (SAS) processes using ethanol or acetone as solvents and carbon dioxide as the antisolvent [28]. The particle sizes produced in this work were similar to those produced in the supercritical antisolvent process.

122

Table 5.3: Naproxen experimental conditions and results summary

Experiment Injection conditions Results T(°C) Concentration Max. length Max. length X50 Morphology (mole fraction) (µm, SEM) Aggregates (µm, LS) (µm, SEM) N1 140 186.3 15 15 20 Flake crystals Flake crystals and N2 160* 179.1 12 15 30 some wrapped spheres N3 170* 556.3 2 15 17 Wrapped- Flake crystal spheres * The concentration of the injected mixtures at 160°C and 170°C is presented in Appendix C

The light scattering results were typically larger than the average particle size established from the SEM results, as shown in Table 5.3. In some cases the SEM images showed that particles tended to aggregate into clumps. An average aggregate particle size was reported based on aggregated particles observed in the SEM images (Table 5.3).

Experiments N1 and N2 were conducted at conditions where the concentrations were similar and temperatures differed by 20°C (Table 5.3). The effect of temperature on the product morphology was negligible. The concentration of naproxen in the SBCW solution affected the size of the precipitates. The size of the naproxen crystals produced for saturated solutions at 170°C (experiment 5) were 10 times smaller than size of the particles precipitated from saturated solutions at 140°C. The same trend has been observed from previous precipitation studies of griseofulvin from SBCW solutions [29].

Experimental condition N3 also resulted in the formation of flaky crystals wrapped into spherical agglomerates, as shown in Figure 5.9b. The DSC curve of the crystals precipitated in plain water at 140°C and 170°C were identical.

Crystals that agglomerated into spheres have been precipitated from crystallo-co-agglomeration techniques (CCA) [9]. The CCA technique was designed to precipitate crystals around droplets of a wetting agent. For spheres to form, the droplets of the wetting agent needed to be much larger in diameter than the precipitated crystals [30]. Otherwise, agglomerates would form around the wetting agent in non-ordered clumps.

It is unlikely that the spheres formed in this work were the result of aggregate formation around droplets. The injected jet of the SBCW-naproxen solution does not form droplets around which the solute can precipitate. Absence of droplet formation was observed in the product collection chamber through the view window. Furthermore the requirement for droplets to form are that both the injected and capture liquids need to be immiscible. Hot and cold water are miscible,

123 therefore it is unlikely that droplets would form. With no droplets formed during the injection of the jet, an alternative mechanism for spherical particle agglomeration should be considered.

It should be noted that the lack of visible droplet formation also means that spherical agglomerate formation is not the product of melted naproxen being extruded through the filter frit. It would be expected that, if the precipitated particles were the result of the melt being extruded through the filter (and thus through the nozzle) the particles, would have a size in the same order of magnitude as (or larger than) the nozzle ID. The nozzle ID was 1.5mm, which is 15 times larger than the precipitates being formed. Thus it is unlikely that the spherical agglomerates were the result of melted naproxen being pushed through the nozzle.

Figure 5.9: Precipitates of naproxen at experimental conditions a) N1 and b) N3

Spherical crystal agglomerates of API particles have also been formed by spray drying suspensions of ciprofloxacin particles [31]. It was shown that as the distance between the crystallized particles was reduced, small spherical agglomerates were formed.

It is possible that the spherical naproxen agglomerates formed as a consequence of an increased density of particles close to the nozzle end. At 170°C the supersaturation ratio of the injected SBCW-naproxen jet was higher than at 140°C. The higher supersaturation ratio corresponded to a closer proximity of newly formed particles. The result of closely spaced precipitates may have led to an agglomeration of naproxen particles into spheres.

124

5.3.2. PRECIPITATION WITH CO-SOLVENTS

5.3.2.1. Budesonide Dissolved in SBCW-alcohol mixtures

Budesonide is a glucocorticoid steroid with a high topical to systemic activity ratio that is typically used to treat asthma and rhinitis [15]. For the treatment of asthma, the drug is delivered into the lungs via inhalation. Often, budesonide is delivered in combination with formoterol fumarate; a bronchodilator [32]. Other compounds can be added as excipients to drug formulations to facilitate the APIs absorption into the body. Pharmaceutical excipients can include compounds such as lactose [33], polyethylene glycol (PEG) [12], dextrose and potassium sorbate [13].

In order for a drug to be delivered efficiently into the lungs, the maximum particle size needs to be 5µm [34]. The restrictions on the particle size are related to the relative tortuosity of the lung chambers – where, if a particle is too large the particle will not be able to penetrate to the lower lung absorption sites[35]. Conversely, if the particle is too small, the particle will be expelled by other areas of the lungs[36]. Thus, stringent control over the particle size needs to be achieved.

Micronization of budesonide has been achieved using a number of methods, which include traditional milling and precipitation techniques [33]. Micronization of budesonide via supercritical antisolvent (precipitation) techniques require the use of methylene chloride or acetone as a solvent (with CO2 as the antisolvent) [5, 37]. These solvents are toxic and can be difficult to remove from the drug matrix post-micronization [38]. The use of SBCW as a solvent to micronize budesonide has an advantage over these methods, as SBCW (even with small amounts of alcohol modifiers) is non-toxic and does not require rigorous solvent removal, even if present in non-trace levels in the drug matrix.

The solubility of budesonide in SBCW and SBCW modified with organic solvents has been established. Both ethanol (see Chapter 3) and methanol (see Chapter 4) were used to decrease the dielectric constant of water. The purpose of using organic solvents is to increase the solubility of budesonide in SBCW, and thus increase the yield of the precipitation process.

Experimental conditions used to micronize budesonide are shown in Table 5.4. The budesonide precipitated from subcritical water solutions, in the absence of organic solvents, crystallized into spherical particles with a number average diameter of 300nm, as shown in Figure 5.10b. The particles precipitated from SBCW-ethanol solutions tended to have larger morphologies than the budesonide precipitated from pure SBCW, as shown in Figure 5.10c and d. The

125 particles precipitated from ethanol modified SBCW solutions tended to crystallize as plate crystals. When methanol was used as a co-solvent, small spherical particles with a diameter of 50nm were precipitated.

Similar spherical particles of budesonide have been formed from a Solution Enhanced Dispersion of Supercritical fluids (SEDS) process [39]. The particle size distribution of the spheres produced in the subcritical water system was comparable to the distributions of the SEDS process.

A change in SBCW-budesonide injection temperature did not affect the budesonide precipitate morphology over the tested temperature range when no excipients were present. While the X50 number diameter of budesonide precipitated from a 130°C with 20% ethanol-SBCW solution was 630nm and from budesonide precipitated from a 150°C with 20% ethanol-SBCW solution was 933nm, the particle size distribution of each condition spanned the same size range. The unchanged particle morphology is reflected by observing the similar average particle size obtained from SEM imaging (as shown in Table 5.4) Thus, while it appeared that the lower temperature injection led to smaller particles, the particle size distribution was unchanged.

126

Table 5.4: Budesonide particle formation conditions and results summary

Experiment Temperature Ethanol Methanol Solubility Average Number X10 X50 X90 Morphology number fraction fraction Particle Diameter (°C) (%, v/v) (%, v/v) mg/mL (SEM, nm) (nm) (nm) (nm) Shape

B1 200 0 0 2.35 200 60 100 200 Spheres 130 20 0 2.22 730 300 630 1290 Spherical particles and B2 plates 150 20 0 6.72 778 530 930 1630 Spherical particles and B3 plates B4 160 0 10 2.15 50.1 24.0 50 100.0 Spheres

127

Figure 5.10: SEM of raw budesonide (a), budesonide processed using conditions B1 (b), B2 (c), B3 (d) and B4 (e)

128

Methanol-modified subcritical water solutions yielded particles smaller than the other experimental conditions, as shown in Figure 5.11. It has been shown that the presence of organic solvents can affect the morphology of a particle during a rapid precipitation process [40]. The inclusion of organic solvent in an antisolvent process results in a change in of the injected solution. It has been shown that the higher the viscosity of the injected solution, the larger the particle size [41]. In supercritical fluid systems, a change in viscosity translates to a change in droplet size, which hinders or facilitates mass transfer rates between the droplet and the antisolvent [42]. In the case of subcritical water precipitation processes: water, SBCW, methanol and ethanol are miscible with each other. Thus, it is unlikely that droplets are forming when the SBCW-API solution is injected into PC.

It is possible that, despite the lack of droplets, the solution viscosity may have affected the intimacy of the contact between the solution jet and the water antisolvent. At higher , the area of contact between the solution jet and the water antisolvent may have decreased, leading to hindered heat transfer between the hot and cold water solutions. It is not possible to analyze the results based on viscosity data of the solvent system, as there is no data available in the published literature on the systems studied in this work. In order to confirm whether viscosity plays a role in determining the morphology of a precipitated particle from a SBCW- organic solvent system, high temperature viscosity studies on the systems used in this work need to be done, as discussed in Chapter 6.

105 B1 95 B2 85

75 B3

65 B4

55 Number Number Count 45

35

25 1 10 100 1000 10000

Hydrodynamic Diameter (nm)

Figure 5.11: Particle size distributions of budesonide processed using experimental conditions B1, B2, B3 and B4 according to Table 5.4

129

The DSC and XRD results, shown in Figure 5.12 and Figure 5.13 respectively, indicate that the budesonide was crystalline when precipitated from ethanol-modified solutions. The processed budesonide DSC curve indicated a peak melting temperature that was 1.3°C less than the peak melting temperature of the raw material. The onset of melting occurred at 250°C for both samples. Considering that both the onset of melting and the XRD patterns are identical for the raw and SBCW-ethanol processed budesonide samples, it can be concluded that the precipitated budesonide has the same crystal structure as the raw material.

0.5

0

-0.5

-1

-1.5

-2 Raw Material

Heat Flow, exo (W/mg)exo up Flow, Heat 130B2 -2.5 150B3

-3 200 220 240 260 280 300 Temperature (°C)

Figure 5.12: DSC curves of raw and processed budesonide

130

Figure 5.13: XRD patterns of raw budesonide and budesonide processed in 20% ethanol/SBCW solutions

5.3.3. PRECIPITATION INTO A SOLUTION CONTAINING EXCIPIENTS

5.3.3.1. Naproxen-SBCW solutions injected into a water-lactose solution

A summary of the experimental conditions for the micronized naproxen from SBCW-solutions injected into a lactose-water solution is shown in Table 5.5.

Table 5.5: Experimental condition and results of SBCW-naproxen solutions injected into a lactose- water environment

Experiment Injection conditions Results Concentration Max. length Max. length X50 T(°C) (mole fraction) (µm, SEM) Aggregates (µm) Morphology (µm, SEM) N4 140 180 3 15 20 Plate crystals N5 160 180 2 12 - Plate crystals N6 170 550 2 20 29 Spheres

131

The addition of lactose into the collection chamber (for experiments N4-N6) resulted in the precipitation of smaller particles. The crystals precipitated in the presence of lactose were visibly different to the crystals produced in experiments N1 and N2, as shown by the SEM images in Figure 5.14a.

Figure 5.14: Precipitates of naproxen at experimental conditions a) N4 and b) N6

The thermal properties of the crystals precipitated in the presence of lactose differed from the crystals precipitated without lactose (Figure 5.15). In particular, the melting peak of the naproxen precipitated in the presence of lactose was sharper than the corresponding signal for naproxen precipitated without lactose. The melting peak of the naproxen precipitated with lactose was 3.6°C lower than for naproxen precipitated without lactose, however, the onset of melting occurred at the same temperature for all samples.

132

2

0

-2

-4

-6

Heat Flow, Flow, exoHeat up(W/mg) -8

-10 50 70 90 110 130 150 170 190 Temperaure (°C)

Figure 5.15: DSC of naproxen precipitated from SBCW without lactose (- - -) and with lactose (-)

The presence of lactose did not substantially affect the XRD patterns of the micronized naproxen crystals, as shown in Figure 5.16. Thus, the crystal phase of naproxen precipitated in the presence of lactose was the same as the raw naproxen and the naproxen precipitated into pure water.

The narrowing of the melting curve indicated that the crystals precipitated in the presence of lactose had a more ordered structure than the crystals precipitated without lactose [43]. It is likely that lactose acted as a crystal growth promoter for naproxen. The hydrophilic nature of the lactose present in the precipitation chamber may have drawn water away from the hydrophobic bonding sites during precipitation, allowing for a more ordered naproxen crystal to form [44].

The presence of lactose during the precipitation from 170°C did not hinder the formation of spherical crystal agglomerates, as shown in Figure 5.14b.

133

Figure 5.16: XRD patterns of naproxen produced at different processing conditions and unprocessed naproxen.

Lactose was effective in reducing the particle size of the naproxen crystals, indicating that there is scope to widen the study of the effect of different excipients and their concentration levels on product morphology.

134

5.3.4. PRECIPITATION INTO A SOLUTION WITH EXCIPIENTS AND CO-SOLVENTS

5.3.4.1. Budesonide-SBCW-alcohol solutions injected into a water- lactose solution

Masses of lactose were added to the collection chamber to be in a 15:1 and a 30:1 ratio with the budesonide. The 15:1 ratio was chosen as it is a commonly used proportion for budesonide preparations [11]. The 30:1 ratio was used to examine whether a change in lactose concentration would affect the precipitate morphology. The experimental conditions and results are summarized in Table 5.6

Precipitation into a lactose environment did not change the shape of the budesonide crystal, which can be seen by comparing the SEM images in Figure 5.17 and Figure 5.10. The particle morphology was the same as that which was found when lactose was not present. The budesonide particles produced from solutions of SBCW modified with ethanol were larger than particles produced when SBCW or SBCW modified with methanol were used (Figure 5.17).

The presence of lactose in the PC tended to result in slightly larger precipitates than when lactose was not present. A higher proportion of lactose placed in the precipitation chamber had no effect on the product morphology, as shown in Figure 5.17 and Figure 5.18.

It appeared that the selection of solvent affected the morphology and the size of the precipitate more than the concentration of the solution jet injected into the chamber or concentration of the lactose dissolved in the precipitation vessel. The solubility of budesonide in the solvent in all experiments, except at 150°C, was equivalent.

135

Figure 5.17: Budesonide precipitated according to experimental conditions a) B3A, b) B3B, c) B2A, d) B2B , e) B1A and f) B4A described in Table 5.6

136

An experiment was conducted (experiment B4A1) whereby the volume of water-lactose solution in the PC was halved. The particle size distribution of the budesonide particles did not change, as shown in Figure 5.18, which indicates that, under the 160°C injection conditions, only half the volume of water may be used to precipitate budesonide micro-particles. The unchanged particle size when the capture water volume was lowered implies that the volume of capture water may be reduced to lower process costs for industrial scale-up.

106

96

86

76 20%B3A eth 150 1:15 lactose 20%B3B eth 150 1:30 lactose 66 20%B2A ethanol 130 1:15 Count 0%B1A eth 200 1:15 lactose 56 10%B4A1 meth 160 1:15 lactose 10%B4A meth half vol 1:15 lact 46 20%ethB2B 130 003

36

26 10 100 1000 10000 Hydrodynamic Diameter (nm)

Figure 5.18: Budesonide precipitated according to the experimental conditions summarized in Table 5.6

137

Table 5.6: Particle formation conditions and summary of budesonide precipitated into lactose and PEG400 environments using SBCW and co-solvents

Temperature Ethanol Methanol Solubility Fraction Fraction Average Particle X10 X50 X90 Morphology fraction fraction Budesonide/ Budesonide/ Diameter Experiment lactose PEG number (mass/mass) (mass/mass) (°C) (%, v/v) (%, v/v) mg/mL - - (SEM, nm) (nm) (nm) (nm) Shape B1A 200 0 0 2.35 0.02 - 70 30 60 110 Spheres B2A 130 20 0 2.22 0.02 - 400 70 210 670 Spherical particles and plates B2B 130 20 0 2.22 0.03 - 420 110 290 720 Spherical particles and plates B3A 150 20 0 6.72 0.06 - 1050 430 1150 3070 Spherical particles and plates B3B 150 20 0 6.72 0.03 - 1370 450 1170 3030 Spherical particles and plates B4A 160 0 10 2.15 0.02 - 210 100 280 830 Spheres B4A1 160 0 10 2.15 0.04 - 250 140 540 980 Spherical particles and plates B1C 200 0 0 2.35 - 0.25 85 40 70 130 Spheres B1D 200 0 0 2.35 0.125 95 50 90 140 Spheres B2C 130 20 0 2.22 - 0.25 1100 730 1670 2770 Plates B2D 130 20 0 2.22 0.125 2140 770 1770 4030 Plates B3C 150 20 0 6.72 - 0.25 830 380 750 1500 Plates B3D 150 20 0 6.72 0.125 750 440 800 1460 Plates B4C 160 0 10 2.15 - 0.25 85 15 30 80 Spheres B4D 160 0 10 2.15 - 0.125 65 25 45 80 Spheres

138

5.3.4.1. Budesonide-SBCW-alcohol solutions injected into a water-PEG solution

Budesonide was precipitated into solutions containing PEG with a 1:4 and 1:8 budesonide:PEG ratio under various injection conditions, as outlined in Table 5.6.These ratios were chosen as they are commonly found in budesonide pharmaceutical formulations [12].

Figure 5.19: Budesonide precipitated according to conditions a) B2C, b) B3D, c) B4C and d) B1D as described in Table 5.6

139

A change in the ratios of PEG added to the capture chamber had no effect on the particle size distribution of the budesonide particles, as shown in Figure 5.20. The particles precipitated from higher temperatures tended to be smaller than the materials precipitated at lower temperatures.

110

100

90

B2C 80 B2D 70 B3C B3D

Count 60 B1C B1D 50 B4C B4D 40

30

20 1 10 100 1000 10000 Hydrodynamic Diameter (nm)

Figure 5.20: Budesonide precipitated according to the experimental conditions summarized in Table 5.6

Aside from the particles precipitated from ethanol modified solutions at 150°C, the particles precipitated into PEG environments were larger than the particles precipitated in the presence of lactose, as shown in Figure 5.20.The particles precipitated into the PEG environments tended to have a uniform morphology. The formation of plate crystals was found in samples from methanol-modified SBCW. Small spheres were the dominating morphology in particles precipitated into PEG where either pure SBCW or methanol was used as a modifier. The precipitation of uniform-morphology particles was a change from the particles precipitated without using PEG. In those cases (shown in Section 5.3.2.1), mixtures of smaller spheres and larger plates were present. The precipitation of particles with a uniform morphology indicates that PEG acted to stabilize particle growth as either spheres or plates.

140

The precipitation of spheres or plates was dependent on the solvent. When ethanol was used as a co-solvent, the dominant particle morphologies of the budesonide precipitates were plates, as shown in Figure 5.19a and b. Similarly, when pure SBCW was used, the dominant particle morphology was small spheres (Figure 5.19c and d).

It is possible that the addition of PEG acted to hinder the efficiency of the heat transfer between the injected solution jet and the cold water. It was suggested in Section 5.3.2 that presence of different solvents acted on the viscosity of the solution jet, which affected the contact between the SBCW-API-organic solvent solutions and the water in PC. If the viscosity of the injected solution has a role in the precipitation of spheres or plates, the change in viscosity of the water in PC would also have a role in the injected jet hydrodynamics. Rigorous studies on the effect of viscosity on the crystallization of organic compounds from SBCW would be required to draw any conclusions as to the mechanisms determining precipitated particle morphology.

It would be beneficial to evaluate the effect of viscosity on the particle morphology by changing the concentrations of ethanol in the solution jet at a constant temperature. Such as study is outside the scope of this work, which is limited to exploring the application of the newly developed precipitation technique to various APIs. Nevertheless, it has been shown that the particle morphology of APIs precipitated from SBCW solutions can be controlled by adding different levels of organic solvents adds a degree of tunability to the process.

5.4. THE DRYING OF THERAPEUTIC FORMULATIONS

5.4.1. PRECIPITATION BY SPRAYING A SBCW-SOLUTION INTO A VACUUM CHAMBER

The experimental conditions and results for the injection of naproxen solutions in SBCW into a heated vacuum chamber are shown in Table 5.7. The spray process (experiments N1 and N2) produced flake-plate crystals, as shown in Figure 5.21. There was no difference between the particles precipitated from SBCW solutions at 140°C and 170°C. The particle size distribution is shown in Figure 5.22. The crystals that were precipitated using the SBCW spray process were similar to the crystals that have been precipitated from supercritical fluid anti-solvent techniques, in which acetone was used as the solvent, and supercritical CO2 as the antisolvent [28].

141

Table 5.7: Experimental conditions and results of naproxen micronization by injection of SBCW solutions into a heated vacuum chamber

Experiment Injection conditions Results

T(°C) Concentration Diameter Number X50 Morphology (mole fraction) (µm, SEM) (µm) NS1 140 179.1 10-500 20 Flake crystals NS2 170 556.3 50-500 25 Flake crystals

Figure 5.21: Precipitate collected from SBCW-naproxen solution spray into a vacuum vessel

The particle size distribution of the precipitates collected from the spray process was bimodal, as shown in Figure 5.22. In some cases, water remained in the product. The slow cooling of the remaining water inside the bomb may have precipitated larger naproxen crystals.

142

Figure 5.22: Particle size distribution of naproxen precipitated by the SBCW-spray technique at 160°C

Modifications of the system to improve water removal have been considered. More efficient water removal may be achieved by contacting the injected solution with a countercurrent flow of hot inert gas, such as nitrogen. Future development of this technology is discussed in Chapter 6 of this thesis.

5.4.2. SPRAY-DRYING OF BUDESONIDE FORMULATIONS

Of all of the conditions tested, the particles produced from pure SBCW-API solutions were the most ideally suited to inhalable drug delivery (experimental conditions B1, B1A, B1C and B1D). While the low solubility of budesonide in pure SBCW may lead to low process yields, the particle morphology is uniform and below the 5µm particle size cut-off for aerosol delivery. Thus samples from experimental condition B1A were spray-dried to produce a drug powder for drug delivery.

In order to collect a high enough mass of budesonide to analyze using the Andersen Cascade Impactor from the SBCW process, multiple injections of budesonide-SBCW were conducted into

143 the same precipitation vessel. The multiple injections were performed by loading the solubility vessel with 3 times the regular mass of budesonide used in the solubility experiments. The precipitation was carried out as per the method in Section 5.2.2, and then the solubility system was refilled with water and allowed to equilibrate for another 10 minutes. After the 10 minute equilibration time, another injection was performed. The refill step was performed twice to allow for a total of 3 injections into the precipitation vessel.

In order to establish whether the product budesonide degraded under the increased exposure time to SBCW at 200°C, FTIR was performed on a dry filtered sample and compared to the raw material. Some of the sample was analyzed by laser scattering to observe whether the multiple- injections had any significant effect on product morphology. The results of the FTIR and DLS analysis on the multiple-injection samples and the raw/originally processed materials are shown in Figure 5.23 and Figure 5.24 respectively. Based on the identical FTIR spectra of the SBCW-processed budesonide and the raw material, as well as the identical particle size distribution of the multiple injection particles to the single injection particles, it was established that the multiple injection step had no effect on the budesonide product chemical stability or particle morphology.

Figure 5.23: FTIR spectra of a) budesonide processed at 200°C with multiple injections, b) budesonide processed at 200°C with a single injection and c) raw budesonide

144

110

100

90 Multiple Injections 80 Single Injection 70

Count 60

50

40

30

20 10 100 1000

Hydronamic diameter (nm)

Figure 5.24: Particle size distribution of budesonide particles from a budesonide-SBCW solution injected into the precipitation chamber in single and triple injections

The spray drying experimental method is detailed in Section 5.2.9. A summary of the experimental conditions and results are shown in Table 5.8 and Table 5.9 respectively.

Table 5.8: Spray drier experimental conditions

Suspension/ solution Inlet T Mass Outlet T Mass flowrate Aspirator Air flowrate (°C) Solutes (g) (°C) (g/min) (gauge) (gauge) lactose 170 3.15 90 6.0 100% 40 budesonide 170 0.08 85 5.9 100% 40 B1A 170 3.85 74 6.0 100% 40

Table 5.9: Spray dryer results

Suspension/ Mass solution Runtime collected Efficiency* (mins) (g) (%) Notes lactose 26 0.98 30 budesonide 23 0.001 1 B1A 33 0.05 1 Some * Efficiency calculated as Mass collected/Mass Solutes

145

5.4.3. SPRAY DRYING RESULTS

The particles captured after the spray drying process were analysed by SEM. The lactose produced by spray drying could not be analysed by SEM as the white particles turned clear on the carbon tape. Thus the morphology of the budesonide particles alone is shown in this section. The aerodynamic particle sizes of lactose and budesonide were analysed using the Andersen Cascade Impactor, as the objective was to test the ability of the SBCW micronization process to produce particles suitable for inhalable delivery.

A SEM of the spray-dried budesonide is shown in Figure 5.25. The particle size and morphology did not change upon spray drying of the suspension of budesonide in water. The similarity of both results can be seen by comparing Figure 5.25 to Figure 5.10b.

Figure 5.25: Budesonide precipitated from a 200°C injection temperature into cold water and the dried using the Buchi 290 spray-drier.

146

The aerodynamic particle sizes of the lactose and budesonide raw materials and precipitates are shown in Figure 5.26, Figure 5.27 and Figure 5.29. The raw and spray dried lactose tend to have the same aerodynamic particle size distributions. On average, 80% of the lactose by mass was either caught in the neckpiece of the Cascade Impactor or trapped in the filter after stage 7.

0.9

0.8

0.7

0.6

0.5

0.4 Mass fraction Mass 0.3

0.2

0.1

0 0 2 4 6 8 10 Aerodynamic Particle Size (µm)

Figure 5.26: Raw (◊) and spray-dried (▲) lactose aerodynamic particle sizes.

On average 85% by mass of the raw budesonide was either caught in the neckpiece of the Cascade Impactor or caught in the filter. Of the material that deposited on the Cascade Impactor plates, the majority of the powder deposited between stages 2 and 4, which corresponded to an aerodynamic particle size between 3µm and 6µm, as shown in Figure 5.27.

It was not possible to compare the raw budesonide aerodynamic particle size distribution to the spray dried sample, as not enough budesonide was recovered from the spray dryer to run the Cascade Impactor. The majority of the budesonide particles were either caught on the walls of the cyclone separator, or expelled through the aspirator of the Buchi spray drier. In order to attempt to compare the particle sizes of the raw budesonide to the micronized budesonide, a 60mL sample of the micronized budesonide suspension (taken directly from the particle formation apparatus) was freeze-dried for 48 hours. The aerodynamic particle size of the

147 freeze-dried, micronized sample is shown in Figure 5.27. The aerodynamic particle size of the freeze-dried micronized budesonide and the raw budesonide sample are identical.

The budesonide spray-dried from the budesonide-lactose-water mixtures tended to have similar proportions by mass of material distributed at every stage of the Cascade Impactor, as shown in Figure 5.28. However, unprocessed budesonide had a higher deposition on plate 4 and a lower proportion on plate 7 than the SBCW-micronized budesonide. Plate deposition may be related to the different particle sizes of the raw and SBCW-micronized budesonide. The difference in size can be observed by comparing Figure 5.10a, and Figure 5.10b.

It is possible that the raw budesonide was already micronized. An SEM of the raw budesonide is shown in Figure 5.10. The particle size of the non-SBCW processed budesonide was up to 10um in length. Furthermore the different morphology of the budesonide particle in the raw material may have allowed the particle to fly more effectively than the SBCW-micronized budesonide. Elongated particle morphologies typically fly better than spherical particles [45]. While the efficiency and aerodynamic performance of the budesonide and lactose powder produced by the SBCW-micronization process are not optimal, it can be seen that the technology developed in this thesis is capable of producing particles that can be delivered into the lungs.

0.3

0.25

0.2

0.15

Mass fraction Mass 0.1

0.05

0 0 1 2 3 4 5 6 7 8 9 10 Aerodynamic Diameter (µm)

Figure 5.27: Aerodynamic particle diameter of raw (◊) and processed (⧠) budesonide

148

0.45

0.4 Processed Budesonide

0.35 Raw Budesonide 0.3

0.25

0.2

Mass Fraction Mass 0.15

0.1

0.05

0 0 2 4 6 8 10

Aerodynamic Size (µm) Figure 5.28: Aerodynamic particle sizes of the raw and spray-dried budesonide from the budesonide-lactose mixture

5.4.4. OVERALL PROCESS EFFICIENCY

The efficiency of the overall process is shown in Table 5.10. The largest loss in mass was through spray drying the liquid suspensions. It is possible to optimize spray drier conditions to increase the efficiency of solids recovery [46-48]. If the process is to be scaled up using the spray dryer to dry the suspensions, optimization on the recovery of solids is needed. It may also be possible to dry the suspensions/solutions using the drying step outlined in section 5.4.1. However, the drying conditions need to be optimized to completely vaporize the solution before the aerodynamic properties of the particles can be tested using the Cascade Impactor.

As the aim of this thesis is the development of a technology that utilizes the dielectric constant of water to dissolve and precipitate model pharmaceuticals, optimization of the spray-drying process lies outside of the scope of this work, and was thus not conducted.

149

Table 5.10: Process efficiency in terms of mass loss at each process step

Material → Lactose Budesonide Lactose and budesonide Process ↓ Lactose Budesonide Lactose Budesonide Lactose+Budesonide Mass in (mg) 3150 70 1050 70 1120 SBCW Micronization Mass out (mg) 3150 70 1050 70 1120 Recovery (%) 100% 100% 100% 100% 100% Mass in (mg) 3150 70 1050 70 1120 Spray Drying Mass out (mg) 980 1 nd nd 50 collect % 31% 0.14% nd nd 4%

Process efficiency 31% 0.14% 4% 5% 4%

Mass in (mg) 96.1 8.1 ∾ nd nd 36.4 Cascade Impactor Mass out (mg) 20.3 1.94∾ 10.9 0.79 11.69 collect % 21% 24% 32%

Overall efficiency 1% nd 1% 1% 1% nd Not determined ∾ Mass of material was from a freeze-dried sample of micronized budesonide suspension

150

5.5. CONCLUSIONS

It has been demonstrated that the morphology of API particles can be manipulated by changing injection conditions of SBCW-API solutions into a chamber containing cold water. Particles with narrow size distributions have been precipitated from a variety of SBCW solutions. Particle morphology can be controlled by changing the injected solution temperature, adding of co- solvents, and/or adding excipients. At higher injection temperatures, particle sizes are generally reduced. The choice of organic solvent alters the interaction between the SBCW solution and the cold water, where injection with ethanol tends to produce larger particles. The presence of the excipients lactose and PEG tends to be able to limit the formation of either sphere of plate morphologies of particles, although a change in concentration of the excipients (in the ratios tested) does not change particle size.

Dry powders of budesonide and budesonide/lactose have been formulated using a combination of SBCW processing and spray-drying. While the spray drying process used was inefficient, the SBCW-micronized powders tended to allow budesonide to penetrate lower sections of the Cascade Impactor.

The SBCW micronization process is a tailorable system that can be controlled by a number of variables. Further research into producing therapeutic combinations through the methods described here, or precipitating two APIs simultaneously is warranted based on the success of the technology at precipitating a single API with excipients.

151

5.6. REFERENCES

1. Ginty, P.J., Whitaker, M.J., Shakesheff, K.M. and Howdle, S.M., Drug Delivery goes Supercritical. Materials Today, 2005. 8(8, Supplement 1): pp. 42-48.

2. Tajber, L., Corrigan, D.O., Corrigan, O.I. and Healy, A.M., Spray Drying of Budesonide, Formoterol Fumarate and their Composites--I. Physicochemical Characterisation. International Journal of Pharmaceutics, 2009. 367(1-2): pp. 79-85.

3. Rasenack, N., Steckel, H. and Müller, B.W., Micronization of Anti-inflammatory Drugs for Pulmonary Delivery by a Controlled Crystallization Process. Journal of Pharmaceutical Sciences, 2003. 92(1): pp. 35-44.

4. Bandi, N., Wei, W., Roberts, C.B., Kotra, L.P. and Kompella, U.B., Preparation of Budesonide- and Indomethacin-Hydroxypropyl-[Beta]-Cyclodextrin (HPBCD) Complexes using a Single-step, Organic-solvent-free Supercritical Fluid Process. European Journal of Pharmaceutical Sciences, 2004. 23(2): pp. 159-168.

5. Martin, T., Bandi, N., Shulz, R., Roberts, C. and Kompella, U., Preparation of Budesonide and Budesonide-PLA Microparticles using Supercritical Fluid Precipitation Technology. AAPS PharmSciTech, 2002. 3(3): pp. 16-26.

6. Steckel, H., Thies, J. and Muller, B.W., Micronizing of Steroids for Pulmonary Delivery by Supercritical Carbon Dioxide. International Journal of Pharmaceutics, 1997. 152(1): pp. 99-110.

7. Lipinski, C., Poor Aqueous Solubility—An Industry wide Problem in Drug Discovery. American Pharmaceutical Review, 2002. 5(3).

8. Rogers, T.L., Nelsen, A.C., Sarkari, M., Young, T.J., Johnston, K.P. and Williams, R.O., Enhanced Aqueous Dissolution of a Poorly Water Soluble Drug by Novel Particle Engineering Technology: Spray-Freezing into Liquid with Atmospheric Freeze-Drying. Pharmaceutical Research, 2003. 20(3): pp. 485-493.

9. Nokhodchi, A., Maghsoodi, M., Hassan-Zadeh, D. and Barzegar-Jalali, M., Preparation of Agglomerated Crystals for Improving Flowability and Compactibility of Poorly Flowable and Compactible Drugs and Excipients. Powder Technology, 2007. 175(2): pp. 73-81.

152

10. Naikwade, S. and Bajaj, A., Preparation and In Vitro Evaluation of Budesonide Spray Dried Microparticles for Pulmonary Delivery. Scientia Pharmaceutica, 2009. 77: pp. 419-441.

11. Kinnarinen, T., Jarho, P., Järvinen, K. and Järvinen, T., Pulmonary Deposition of a Budesonide/[gamma]-Cyclodextrin Complex in Vitro. Journal of Controlled Release, 2003. 90(2): pp. 197-205.

12. Steckel, H. and Wehle, S., A Novel Normulation Technique for Metered Dose Inhaler (MDI) Suspensions. International Journal of Pharmaceutics, 2004. 284(1-2): pp. 75-82.

13. Nilsson, H. and Santesson, G., Low Dose Budesonide Formulations and Uses Thereof in http://www.freepatentsonline.com/6291445.html. 2001, Astra Aktiebolag: United States. p. 4.

14. Scientific, C., Metred Dose Inhalers, Dry Powder Inhalers, Nebulizers and Nazal Sprays: Quality Solutions for Inhaler Testing, C.S. Limited, Editor. 2006 Edition, Copley Scientific, United Kingdom.

15. Budavari, S., The MERCK Index, 12th edn. Merck & Co. Inc., Whitehouse Station, USA, 1996. p. 1512.

16. Khalafalla, N., Elgholmy, Z.A. and Khalil, S.A., Bioavailability of Different Brands of Griseofulvin Tablets and its Correlation to Dissolution Data. Die Pharmazie, 1980. 35(8): pp. 482.

17. Aoyagi, N., Ogata, H., Kaniwa, N., Koibuchi, M., Shibazaki, T. and Ejima, A., Bioavailability of Griseofulvin from Tablets in Humans and the Correlation with its Dissolution Rate. Journal of Pharmaceutical Sciences, 1982. 71(10): pp. 1165-1169.

18. Reverchon, E., Porta, G.D., Spada, A. and Antonacci, A., Griseofulvin Micronization and Dissolution Rate Improvement by Supercritical Assisted Atomization. Journal of Pharmacy and Pharmacology, 2004. 56: pp. 1379-1387.

19. Chattopadhyay, P. and Gupta, R.B., Production of Griseofulvin Nanoparticles using Supercritical CO2 Antisolvent with Enhanced Mass Transfer. International Journal of Pharmaceutics, 2001. 228(1-2): pp. 19-31.

20. De Gioannis, B., Jestin, P. and Subra, P., Morphology and Growth Control of Griseofulvin Recrystallized by Compressed Carbon Dioxide as Antisolvent. Journal of Crystal Growth, 2004. 262(1-4): pp. 519-526.

153

21. Jarmer, D.J., Lengsfeld, C.S., Anseth, K.S., Randolph, T.W., Supercritical Fluid Crystallization of Griseofulvin: Crystal Habit Modification with a Selective Growth Inhibitor. Journal of Pharmaceutical Sciences, 2005. 94(12): pp. 2688-2702.

22. Türk, M., Manufacture of Submicron Drug particles with Enhanced Dissolution Behaviour by Rapid Expansion Processes. The Journal of Supercritical Fluids, 2009. 47(3): pp. 537- 545.

23. Zhiyi, L., Jingzhi, J., Xuewu, L., Shunxuan, Z., Yuanjing, X. and Jian, W., Preparation of Griseofulvin Microparticles by Supercritical Fluid Expansion Depressurization Process. Powder Technology, 2007. 182(3): pp. 459-465

24. Yeo, S.D. and Kiran, E., Formation of Polymer Particles with Supercritical Fluids: A Review. Journal of Supercritical Fluids, 2005. 34(3): pp. 287-308.

25. Shinoda, K., "Iceberg" Formation and Solubility. The Journal of Physical Chemistry, 1977. 81(13): pp. 1300-1302.

26. Garside, J., Mersmann, A., and Nývlt, J., Measurement of Crystal Growth and Nucleation Rates. Working Party on Crystallization, ed. E.C.o.C. Engineering. 2002: Institution of Chemical Engineers (IChemE), Davis Building, 165-189 Railway Terrace, Rugby, CV21 3HQ, United Kingdom.

27. Carr, A., Mammucari, R. and Foster, N.R., The Solubility and Micronization of Griseofulvin using Subcritical Water. Industrial & Engineering Chemistry Research 2010, 49 (7): pp. 3403-3410.

28. Muntó, M., Ventosa, N., Sala, S. and Veciana, J., Solubility Behaviors of Ibuprofen and Naproxen Drugs in Liquid “CO2–Organic Solvent” Mixtures. The Journal of Supercritical Fluids, 2008. 47(2): pp. 147-153.

29. Carr, A., Mammucari, R., and Foster, N., Controlled Precipitation of Hydrophobic Pharmaceuticals in Subcritical Water. in International Symposium of Supercritical Fluids. 2009, Arcachon, France.

30. Amaro-Gonzalez, D. and Biscans, B., Spherical Agglomeration during Crystallization of an Active Pharmaceutical Ingredient. Powder Technology, 2002. 128(2-3): pp. 188-194.

31. Zhao, H., Le, Y., Liu, H., Hu, T., Shen, Z., Yun, J. and Chen, J.F., Preparation of Microsized Spherical Aggregates of Ultrafine Ciprofloxacin Particles for Dry Powder Inhalation (DPI). Powder Technology, 2009. 194(1-2): pp. 81-86.

154

32. Banerjee; P.S., Akapo, S.O., Chaudry, I.A., Bronchodilating Compositions and Methods., United States Patent Number 6667344, 2001, Dey, L.P. (Napa, CA): United States.

33. Dudognon, E., Willart, J.F., Caron, V., Capet, F., Larsson, T., and Descamps, M., Formation of Budesonide/[alpha]-Lactose Glass Solutions by Ball-Milling. Solid State Communications, 2006. 138(2): pp. 68-71.

34. Huang, Y.Y. and Wang, C.H., Pulmonary Delivery of Insulin by Liposomal Carriers. Journal of Controlled Release, 2006. 113(1): pp. 9-14.

35. Vozone, C.M. and Marques, H.M.C., Complexation of Budesonide in Cyclodextrins and Particle Aerodynamic Characterization of the Complex Solid Form for Dry Powder Inhalation. Journal of Inclusion Phenomena and Macrocyclic Chemistry, 2002. 44(1): pp. 111-116.

36. Fu, J., Fiegel, J., Krauland, E., and Hanes, J., New Polymeric Carriers For Controlled Drug Delivery For Controlled Drug Delivery Followin Inhalation or Injection. Biomaterials, 2002. 23(22): pp. 4425-4433.

37. Lobo, J., Schiavone, H., Palakodaty, S., York, P., Clark, A. and Tzannis, S., SCF-Engineered Powders for Delivery of Budesonide from Passive DPI Devices. Journal of Pharmaceutical Sciences, 2005. 94(10): pp. 2276-2288

38. Lakatos, M., Measurement of residual solvents in a drug substance by a purge-and-trap method. Journal of Pharmaceutical and Biomedical Analysis, 2008. 47(4-5): pp. 954-957.

39. Schiavone, H., Palakodaty, S., Clark, A., York, P. and Tzannis, S.T., Evaluation of SCF- engineered particle-based lactose blends in passive dry powder inhalers. International Journal of Pharmaceutics, 2004. 281(1-2): pp. 55-66.

40. Jung, J. and Perrut, M., Particle Design using Supercritical Fluids: Literature and Patent Survey. The Journal of Supercritical Fluids, 2001. 20(3): pp. 179-219.

41. Magnan, C., Badens, E., Commenges, N. and Charbit, G., Soy Lecithin Micronization by Precipitation with a Compressed Fluid Antisolvent--Influence of Process Parameters. The Journal of Supercritical Fluids, 2000. 19(1): pp. 69-77.

42. Carretier, E., Badens, E., Guichardon, P., Boutin, O., and Charbit, G., Hydrodynamics of Supercritical Antisolvent Precipitation: Characterization and Influence on Particle Morphology. Industrial & Engineering Chemistry Research, 2002. 42(2): pp. 331-338.

155

43. Hobokin, N.J., Characterization of Materials. 2008: John Wiley & Sons, New York

44. Myerson, A.S., The Handbook of Industrial Crystallization. 2002: Butterworth- Heinemann, Boston

45. Zeng, X.M., Martin, G.P., Marriott, C. and Pritchard, J., The influence of Carrier Morphology on Drug Delivery by Dry Powder Inhalers. International Journal of Pharmaceutics, 2000. 200(1): pp. 93-106.

46. Das, D., Husni, H.A. and Langrish, T.A.G., The Effects of Operating Conditions on Lactose Crystallization in a Pilot-scale Spray Dryer. Journal of Food Engineering. In Press, Accepted Manuscript. DOI: 10.1016/j.jfoodeng.2010.05.005

47. Gilani, K., Rouholamini Najafabadi, A., Barghi, M. and Rafiee-Tehrani, M., Aerosolisation of Beclomethasone Dipropionate using Spray Dried Lactose/polyethylene Glycol Carriers. European Journal of Pharmaceutics and Biopharmaceutics, 2004. 58(3): pp. 595-606.

48. Islam, M.I.U. and Langrish, T.A.G., An Investigation into Lactose Crystallization under High Temperature Conditions During Spray Drying. Food Research International, 2010. 43(1): pp. 46-56.

156

6. CONCLUSIONS AND RECOMMENDATIONS

SBCW is a highly tunable fluid that can be used to dissolve a host of organic compounds. By manipulating the dissolution temperature, the solubility of organic compounds can be altered. The hot solutions of organic compounds can be subjected to a rapid temperature quench, which rapidly precipitates the hydrophobic organic compounds dissolved in SBCW. The precipitates from the quench are typically crystalline microparticles with narrow particle size distributions. Control over the morphology of the particles can be achieved by manipulating the process temperature, by adding organic solvents as modifiers, and by precipitating SBCW-API solutions into an excipient- containing environment. Precipitated particle sizes were within a size range of 10nm to 100µm.

The use of subcritical water as an alternative solvent for particle engineering overcomes a number of limitations present in both conventional and supercritical fluid techniques. The rapid precipitation involved in SBCW quenching eliminates the high shear environments that commonly degrade pharmaceutical compounds during comminution. The use of only water, or combinations of water with a small volume of ethanol, eliminates the use of toxic organic solvents for rapid precipitation methods. Future development of this technology should evaluate the costs involved in the scale-up of the process, and compare these costs to conventional methods.

157

6.1. CONCLUSIONS

6.1.1. DEVELOPMENT OF A NOVEL, ENVIRONMENTALLY FRIENDLY PARTICLE ENGINEERING TECHNIQUE

A new environmentally friendly particle engineering technique has been developed that uses SBCW to dissolve organic compounds. By quenching the SBCW-organic compound solutions in cold water, it is possible to rapidly precipitate compounds with narrow particle size distributions. In this thesis, APIs were used as model compounds to explore the potential of SBCW as a solvent for particle engineering. APIs were selected as model compounds, as tuning the size and shape of drug particles can have pharmacokinetic benefits (outlined in Chapter 1). It was found that the product morphology could be tuned by altering the supersaturation ratio of the SBCW-API solution, by adding pharmaceutical excipients to the precipitation chamber, and by adding organic solvent modifiers to the SBCW-API solution.

In most cases, increasing the supersaturation ratio of the SBCW-API solution decreased the average particle diameter. Supersaturation levels can be controlled by changing the temperature and/or the concentration of the API in the SBCW solution. Changes in particle morphology due to variations in injected solution temperature were not as pronounced when other SBCW-API injection parameters were altered. In part, this may have been due to the small temperature range of operation, which may not have been large enough to have any noticeable effect on precipitate morphology.

For naproxen, the agglomeration of plate crystals into spheres was observed. The precise mechanism for the formation of these spheres is not known, however it is proposed that the high supersaturation ratio at certain experimental conditions (170°C SBCW-API saturated solution injected into a cold water-filled vessel) led to crystals precipitating in close proximity to one another, which led to the formation of spherical crystal-agglomerates.

The addition of excipients had an effect on the particle morphology for both naproxen and budesonide. A change in excipient type affected the particle morphology of precipitated budesonide when the excipient was changed from lactose to PEG, resulting in higher proportions of spheres or plate crystals precipitated, respectively. However, a change in excipient concentration did not change the particle morphology. Furthermore, while the shape of the precipitated particles sometimes changed when excipients were added, the addition of API did not affect the crystal structure of the precipitates.

158

The addition of organic solvents to modify the dielectric constant of SBCW had a pronounced effect on particle morphology. In the case of ethanol, precipitates tended to be larger and had a different morphology to particles precipitated without ethanol. In the case of methanol- modified SBCW-solutions, particles tended to be smaller than particles precipitated both without methanol, and the particles precipitated with ethanol. Furthermore, particles precipitated with methanol tended to have the same shape as the particles precipitated from pure SBCW-API solutions. The change of solvent type, while also affecting the dielectric constant and thus the supersaturation ratio, led to a change in interaction between the SBCW-API and cold water during precipitation, which affected the precipitate morphology.

The ability to alter the morphology of an API by changing the operating conditions, such as the supersaturation ratios, the presence of excipients into PC, and the addition of organic solvent modifier to the SBCW-API solution demonstrates that SBCW precipitation technology is highly tunable. The spray drying of budesonide-lactose formulations demonstrates that it is possible to combine SBCW processing with a conventional drying technique to produce dry powder products for drug delivery via inhalation.

6.1.2. ESTABLISHMENT OF CHEMICAL STABILITY, SOLUBILITY DATA AND SOLUBILITY MODELS FOR APIS IN SBCW

The chemical stability of each API was tested to ensure that no chemical degradation had occurred in SBCW during processing. The stability of each chemical in SBCW was determined by FTIR. It was found that no degradation of the chemical structure occurred for any of the tested compounds. Thus the APIs tested were chemically stable up to 200°C. The work serves as further evidence that SBCW can be used as a solvent for therapeutic organic compounds.

The solubilities of model APIs in SBCW were measured up to 200°C. The solubilities of the APIs griseofulvin, naproxen, pyrimethamine and budesonide in SBCW were measured. The solubilities of anthracene and 9-anthracenemethanol in SBCW were also measured to both calibrate the solubility apparatus to published literature data and to compare M-UNIFAC solubility model outputs to data, respectively. Solubility was established for all organic compounds to within 10% standard deviation, which was comparable to published literature data. Thus, reliable solubility data was obtained to which a model could be fitted, and for which supersaturation levels in the particle formation technique could be known.

159

The addition of co-solvents to the SBCW-API solutions resulted in increased solubilities of the model compounds. A direct relationship between solute solubility data and the dielectric constant at elevated temperatures was found. The dependence of the solubility on the dielectric constant of the solvent mixtures led to the construction of a solubility model, which used the Akerlof relationship to correlate the decrease in dielectric constant to the percentage increase of organic solvent modifier to water. The model was able to accurately predict the solubility of organic compounds in SBCW between 25°C and 200°C with and without co-solvents; to within 5% of measured solubility values. The outcome of the new model is that solubility can be predicted in SBCW up to 200°C with up to 20% of co-solvent.

The solubilities of all of the organic compounds dissolved in SBCW were also modelled using the UNIFAC, A-UNIFAC, M-UNIFAC, MF-UNIFAC, AF-UNIFAC models. Corrections were made to the MF-UNIFAC model that resulted in the prediction of the solubility of all APIs in SBCW between 25°C and 200°C to within 5% of experimental solubility values. However, it was shown that the correction to the water-hydroxide interaction parameter for budesonide was not applicable to the 9-anthracenemethanol-water system, as the model output error was increased.

6.2. RECOMMENDATIONS

6.2.1. FUNDAMENTAL STUDIES

There are a number of investigations that could be carried out in the future to further the understanding of solubility and particle formation mechanisms using SBCW as a solvent. To gain further understanding of the mechanisms governing the solubility of hydrophobic organic compounds in SBCW; subcritical water solubility studies on additional aromatic hydrocarbons with a variety of different sidegroups should be conducted. Such sidegroups should include hydroxide (alcohol), carboxylic acid and other sidegroups containing oxygen, chlorine and other halogens in different positions around the solute molecule. Modelling the solubilities of such compounds can give greater insight into the effects that different sidegroups have on the solubility behavior of organic compounds in SBCW. Fitting the UNIFAC parameters to the solubility data will increase the accuracy of predicting solubilities using the UNIFAC models.

Further work on the dielectric constant model developed in this thesis is recommended. The model has the potential to be a powerful tool for the prediction of solubilities of a range of organic compounds in SBCW and SBCW-co-solvent systems. Testing of the model in other

160 solute-solvent systems would be of interest, which includes systems applicable to other SBCW technologies that were beyond the scope of this work (i.e. extraction, reaction engineering and reversed phase chromatographic separations). Initially, it would be beneficial to investigate alcohols, such as 1-propanol, to determine whether the Dielectric Constant model can be extended to higher molecular weight organic solvent modifiers. An outcome of this work would be that the dielectric constant solubility model could be tested more rigorously. In time, it may be possible to predict the dielectric constant model parameters based on solute/solvent structure. The eventual development of a dielectric constant model based on chemical structure interactions between water and the solute would be useful.

There are a number of investigations that can be carried out on the SBCW particle formation system. Identification of primary particle formation mechanisms and clarification on how organic solvents influence product morphology when added into the SBCW-API solution would be useful. Further investigations into the variation in different process parameters also need to be conducted, particularly for scale-up to pilot scale.

It would be useful to conduct both computerized fluid modelling and spectral/visual experiments to examine the point of nucleation of the crystals, the rates of heat transfer at each point within the injected jet, and the identification phase boundaries (if they exist). Such an investigation would give greater insight into the hydrodynamics of the injected SBCW-API solution jet, which has been shown to potentially affect the precipitate morphology. Technologies have been developed for supercritical fluid particle formation studies that could be applied to work with SBCW. The spectral technologies make use of multiple view cells with video and spectral software, which can characterize injection hydrodynamics.

6.2.2. PARTICLE FORMATION APPLICATION STUDIES

Given the successful achievement of this study in assessing the feasibility of using SBCW as a solvent for particle engineering, the next stage of investigations should focus on process optimization and scalability. Optimization of the injection nozzle diameter, the precipitation vessel temperature and the quench water volume in the precipitation vessel has not yet been carried out. These variables can affect the heat transfer between hot and cold water solutions, which is the primary factor influencing particle morphology. These variables need to be investigated before the system can be scaled up to produce larger quantities of micronized APIs.

161

Finally, while there was a preliminary investigation into precipitating dry particles via SBCW- evaporation, there is room for process optimization, particularly for the production of particles with narrow particle size distributions. Further recommendations for improving the technology to produce dry powder products could involve combining the particle formation process with the drying of the water-API suspension/SBCW-API solutions in a single step. This may be achieved by either drying the SBCW solution with a controllable flash vaporization technique (i.e. with a vacuum that can be controlled), or coupling the SBCW precipitation process with a spray-drier. The implementation of some of these process modifications may result in a narrower particle size distribution than that which was encountered in the vacuum-drying system described in this thesis.

The use of a SBCW precipitation system coupled with a spray drier may be more effective than a direct spraying process, as the dielectric constant quench is still used to produce small particles. It has been shown in this thesis that the spray drying of SBCW micronized budesonide suspensions did not affect the particle size distribution of SBCW-precipitated budesonide particles. The result of this technology is that an API may be produced in powder form that is unaffected by the drying process. There is scope for the preparation of other drug formulations either in suspension or as a dry powder, using SBCW technology. Further investigations may include the testing of the performance of a SBCW-produced pharmaceutical formulation (e.g. dissolution tests for quantification of bioavailability, aerodynamic size determination for inhalable formulations, and/or anti-bacterial activity tests).

162

7. APPENDIX A

M-UNIFAC sample calculation

The M-UNIFAC model calculations are shown in this Appendix, using all of the equations presented in Chapter 4. The sample calculation is demonstrated using griseofulvin as the model compound at 160°C. The calculations are shown in a similar way to how they appear in the excel calculation file. The similarity is so that the sample calculation can be followed on the spreadsheet itself, and then applied to other organic compounds. The numerical constants and results are displayed to 4 decimal places, as the data from the published literature are put into the model with 4 decimal places1.

1 Gmehling, J., Li, J. and Schiller, M., A Modified UNIFAC Model. 2. Present Parameter Matrix and Results for Different Thermodynamic Properties. Industrial & Engineering Chemistry Research, 1993. 32(1): p. 178- 193.

163

The model input data incorporating the infinite dilution assumption are shown in Table 7.1.

Table 7.1: Inputs for the calculation of solubility using the UNIFAC-type models

INPUTS Temperature 160 deg C R 8.314 Infinite dilution assumption x1 0 solute T 433.15 x2 1 solvent

Table 7.2 shows the subdivision of the sidegroups, parameter values Qk and Rk for griseofulvin and water, and the calculation of ideal solubility (required for the calculation of actual solubility), where the ideal solubility is calculated according to the relationship:

r and q were calculated according to the relationships below. The values of r and q for griseofulvin and water (the solvent) are shown in Table 7.2.

r = Σk(vk * Rk)

q = Σk(vk * Qk)

Table 7.3 shows the list of interaction parameters and the calculation of τlj. In some cases interaction parameter data were not available. The unavailable values are listed in the table.

164

Table 7.2: Griseofulvin and water-specific constants used for the calculation of solubility using the M-UNIFAC model n (nominal) n (Ghemling) Subgroup vk griseofulvin vk water Rk Qk 1 1 CH3 1 0 0.6325 1.0608 2 1 CH2 1 0 0.6325 0.7081 3 1 CH 1 0 0.6325 0.3554 4 1 C 0 0 0.6325 0 5 2 C=C 0 0 1.2832 0.4582 6 3 ArC 5 0 0.3673 0.2113 7 3 ArCH 2 0 0.3673 0.4321 8 4 ArCCH2 0 0 0.91 0.7962 9 4 ACCH 0 0 0.91 0.3769 10 5 OH (p) 0 0 1.2302 0.8927 11 7 H2O 0 1 1.7334 2.4561 12 10 CHO 2 0 0.7173 0.771 13 13 CH3O 4 0 1.1434 1.6022 14 15 CNH (CHnH) 0 0 1.368 0.7278 15 16 C3N 0 0 1.0746 1.176 16 17 ARCNH2 0 0 1.1849 0.8067 17 18 pyridine 0 0 2.5 2.1477 18 20 COOH 0 0 0.8 0.9215 19 25 ACCl 1 0 0.5365 0.3177 # subgroups 17 1

dHmelt 45859.58 -

Tmelt 492.15 - Ideal Solubility -1.526638661 - r 10.3808 1.7334 q 11.9581 2.4561

165

Table 7.3: Interaction parameter data

Parameter data n,m availability 1,1 0.0000 0.0000 0.0000 1.0000 Available 1,2 0.0000 0.0000 0.0000 1.0000 Available 1,3 0.0000 0.0000 0.0000 1.0000 Available 1,4 0.0000 0.0000 0.0000 1.0000 Available 1,5 189.6600 -0.2723 0.0000 0.8474 Available 1,6 114.2000 0.0933 0.0000 0.6998 Available 1,7 114.2000 0.0933 0.0000 0.6998 Available 1,8 7.3390 -0.4538 0.0000 1.5478 Available 1,9 7.3390 -0.4538 0.0000 1.5478 Available 1,10 2777.0000 -4.6740 0.0016 0.0899 Available 1,11 1391.3000 -3.6156 0.0011 0.9121 Available 1,12 875.8500 0.0000 0.0000 0.1324 Available 1,13 233.1000 -0.3155 0.0000 0.8004 Available 1,14 350.5800 0.0667 0.0000 0.4164 Available 1,15 -175.7000 1.8570 0.0000 0.2342 Available 1,16 958.7400 -0.1484 0.0000 0.1268 Available 1,17 -9.2805 1.9682 -0.0014 0.2623 Available 1,18 1182.2000 -3.2647 0.0092 0.0318 Available 1,19 -1385.0000 15.8900 -0.0483 3763.0427 Available 2,1 0.0000 0.0000 0.0000 1.0000 Available 2,2 0.0000 0.0000 0.0000 1.0000 Available 2,3 0.0000 0.0000 0.0000 1.0000 Available 2,4 0.0000 0.0000 0.0000 1.0000 Available 2,5 189.6600 -0.2723 0.0000 0.8474 Available 2,6 114.2000 0.0933 0.0000 0.6998 Available 2,7 114.2000 0.0933 0.0000 0.6998 Available 2,8 7.3390 -0.4538 0.0000 1.5478 Available 2,9 7.3390 -0.4538 0.0000 1.5478 Available 2,10 2777.0000 -4.6740 0.0016 0.0899 Available 2,11 1391.3000 -3.6156 0.0011 0.9121 Available 2,12 875.8500 0.0000 0.0000 0.1324 Available 2,13 233.1000 -0.3155 0.0000 0.8004 Available 2,14 350.5800 0.0667 0.0000 0.4164 Available 2,15 -175.7000 1.8570 0.0000 0.2342 Available 2,16 958.7400 -0.1484 0.0000 0.1268 Available 2,17 -9.2805 1.9682 -0.0014 0.2623 Available 2,18 1182.2000 -3.2647 0.0092 0.0318 Available 2,19 -1385.0000 15.8900 -0.0483 3763.0427 Available 3,1 0.0000 0.0000 0.0000 1.0000 Available 3,2 0.0000 0.0000 0.0000 1.0000 Available 3,3 0.0000 0.0000 0.0000 1.0000 Available

166

3,4 0.0000 0.0000 0.0000 1.0000 Available 3,5 189.6600 -0.2723 0.0000 0.8474 Available 3,6 114.2000 0.0933 0.0000 0.6998 Available 3,7 114.2000 0.0933 0.0000 0.6998 Available 3,8 7.3390 -0.4538 0.0000 1.5478 Available 3,9 7.3390 -0.4538 0.0000 1.5478 Available 3,10 2777.0000 -4.6740 0.0016 0.0899 Available 3,11 1391.3000 -3.6156 0.0011 0.9121 Available 3,12 875.8500 0.0000 0.0000 0.1324 Available 3,13 233.1000 -0.3155 0.0000 0.8004 Available 3,14 350.5800 0.0667 0.0000 0.4164 Available 3,15 -175.7000 1.8570 0.0000 0.2342 Available 3,16 958.7400 -0.1484 0.0000 0.1268 Available 3,17 -9.2805 1.9682 -0.0014 0.2623 Available 3,18 1182.2000 -3.2647 0.0092 0.0318 Available 3,19 -1385.0000 15.8900 -0.0483 3763.0427 Available 4,1 0.0000 0.0000 0.0000 1.0000 Available 4,2 0.0000 0.0000 0.0000 1.0000 Available 4,3 0.0000 0.0000 0.0000 1.0000 Available 4,4 0.0000 0.0000 0.0000 1.0000 Available 4,5 189.6600 -0.2723 0.0000 0.8474 Available 4,6 114.2000 0.0933 0.0000 0.6998 Available 4,7 114.2000 0.0933 0.0000 0.6998 Available 4,8 7.3390 -0.4538 0.0000 1.5478 Available 4,9 7.3390 -0.4538 0.0000 1.5478 Available 4,10 2777.0000 -4.6740 0.0016 0.0899 Available 4,11 1391.3000 -3.6156 0.0011 0.9121 Available 4,12 875.8500 0.0000 0.0000 0.1324 Available 4,13 233.1000 -0.3155 0.0000 0.8004 Available 4,14 350.5800 0.0667 0.0000 0.4164 Available 4,15 -175.7000 1.8570 0.0000 0.2342 Available 4,16 958.7400 -0.1484 0.0000 0.1268 Available 4,17 -9.2805 1.9682 -0.0014 0.2623 Available 4,18 1182.2000 -3.2647 0.0092 0.0318 Available 4,19 -1385.0000 15.8900 -0.0483 3763.0427 Available 5,1 -95.4180 0.0617 0.0000 1.1718 Available 5,2 -95.4180 0.0617 0.0000 1.1718 Available 5,3 -95.4180 0.0617 0.0000 1.1718 Available 5,4 -95.4180 0.0617 0.0000 1.1718 Available 5,5 0.0000 0.0000 0.0000 1.0000 Available 5,6 174.1000 -0.5886 0.0000 1.2052 Available 5,7 174.1000 -0.5886 0.0000 1.2052 Available 5,8 117.3000 -0.8552 0.0000 1.7939 Available 5,9 117.3000 -0.8552 0.0000 1.7939 Available

167

5,10 2649.0000 -6.5080 0.0048 0.1834 Available 5,11 778.3000 0.1482 0.0000 0.1430 Available 5,12 476.2500 0.0000 0.0000 0.3330 Available 5,13 733.3000 -2.5090 0.0000 2.2616 Available 5,14 224.0000 0.0000 0.0000 0.5962 Available 5,15 165.3000 0.0000 0.0000 0.6828 Available 5,16 2800.0000 -10.7200 0.0134 0.2135 Available 5,17 13.5020 0.0000 0.0000 0.9693 Available 5,18 -2026.1000 8.1549 0.0000 0.0309 Available 5,19 -56.6900 9.8050 0.0000 0.0001 Available 6,1 16.0700 -0.2998 0.0000 1.3004 Available 6,2 16.0700 -0.2998 0.0000 1.3004 Available 6,3 16.0700 -0.2998 0.0000 1.3004 Available 6,4 16.0700 -0.2998 0.0000 1.3004 Available 6,5 -157.2000 0.6166 0.0000 0.7759 Available 6,6 0.0000 0.0000 0.0000 1.0000 Available 6,7 0.0000 0.0000 0.0000 1.0000 Available 6,8 139.2000 -0.6500 0.0000 1.3891 Available 6,9 139.2000 -0.6500 0.0000 1.3891 Available 6,10 3927.0000 -13.1600 0.0121 0.3203 Available 6,11 792.0000 -2.7560 0.0008 1.7879 Available 6,12 -365.5000 1.8740 0.0000 0.3569 Available 6,13 -87.0800 -0.1859 0.0000 1.4725 Available 6,14 139.6700 0.0377 0.0000 0.6976 Available 6,15 -71.4000 0.7078 0.0000 0.5810 Available 6,16 1044.7000 -1.7112 0.0000 0.4963 Available 6,17 1047.0000 -5.5620 0.0088 0.5198 Available 6,18 69.5610 1.8881 0.0000 0.1289 Available 6,19 595.2000 3.3090 -0.0284 2070.7501 Available 7,1 16.0700 -0.2998 0.0000 1.3004 Available 7,2 16.0700 -0.2998 0.0000 1.3004 Available 7,3 16.0700 -0.2998 0.0000 1.3004 Available 7,4 16.0700 -0.2998 0.0000 1.3004 Available 7,5 -157.2000 0.6166 0.0000 0.7759 Available 7,6 0.0000 0.0000 0.0000 1.0000 Available 7,7 0.0000 0.0000 0.0000 1.0000 Available 7,8 139.2000 -0.6500 0.0000 1.3891 Available 7,9 139.2000 -0.6500 0.0000 1.3891 Available 7,10 3927.0000 -13.1600 0.0121 0.3203 Available 7,11 792.0000 -2.7560 0.0008 1.7879 Available 7,12 -365.5000 1.8740 0.0000 0.3569 Available 7,13 -87.0800 -0.1859 0.0000 1.4725 Available 7,14 139.6700 0.0377 0.0000 0.6976 Available 7,15 -71.4000 0.7078 0.0000 0.5810 Available

168

7,16 1044.7000 -1.7112 0.0000 0.4963 Available 7,17 1047.0000 -5.5620 0.0088 0.5198 Available 7,18 69.5610 1.8881 0.0000 0.1289 Available 7,19 595.2000 3.3090 -0.0284 2070.7501 Available 8,1 47.2000 0.3575 0.0000 0.6272 Available 8,2 47.2000 0.3575 0.0000 0.6272 Available 8,3 47.2000 0.3575 0.0000 0.6272 Available 8,4 47.2000 0.3575 0.0000 0.6272 Available 8,5 -113.1000 1.1720 0.0000 0.4022 Available 8,6 -45.3300 0.4223 0.0000 0.7279 Available 8,7 -45.3300 0.4223 0.0000 0.7279 Available 8,8 0.0000 0.0000 0.0000 1.0000 Available 8,9 0.0000 0.0000 0.0000 1.0000 Available 8,10 3989.0000 -14.0900 0.0153 0.1744 Available 8,11 1050.2000 -1.9939 0.0000 0.6501 Available 8,12 683.6000 -1.0200 0.0009 0.3927 Available 8,13 -595.1000 2.9780 0.0000 0.2011 Available 8,14 1250.0000 0.0000 0.0000 0.0558 Available 8,15 -2631.0000 13.5600 -0.0070 0.0118 Available 8,16 4000.0000 -16.6800 0.0211 0.1822 Available 8,17 -189.3000 1.8600 0.2410 Available 8,18 1352.5000 0.0000 0.0000 0.0440 Available 8,19 -113.6000 19.7200 0.0000 0.0000 Available 9,1 47.2000 0.3575 0.0000 0.6272 Available 9,2 47.2000 0.3575 0.0000 0.6272 Available 9,3 47.2000 0.3575 0.0000 0.6272 Available 9,4 47.2000 0.3575 0.0000 0.6272 Available 9,5 -113.1000 1.1720 0.0000 0.4022 Available 9,6 -45.3300 0.4223 0.0000 0.7279 Available 9,7 -45.3300 0.4223 0.0000 0.7279 Available 9,8 0.0000 0.0000 0.0000 1.0000 Available 9,9 0.0000 0.0000 0.0000 1.0000 Available 9,10 3989.0000 -14.0900 0.0153 0.1744 Available 9,11 1050.2000 -1.9939 0.0000 0.6501 Available 9,12 683.6000 -1.0200 0.0009 0.3927 Available 9,13 -595.1000 2.9780 0.0000 0.2011 Available 9,14 1250.0000 0.0000 0.0000 0.0558 Available 9,15 -2631.0000 13.5600 -0.0070 0.0118 Available 9,16 4000.0000 -16.6800 0.0211 0.1822 Available 9,17 -189.3000 1.8600 0.2410 Available 9,18 1352.5000 0.0000 0.0000 0.0440 Available 9,19 -113.6000 19.7200 0.0000 0.0000 Available 10,1 1606.0000 -4.7460 0.0009 1.8976 Available 10,2 1606.0000 -4.7460 0.0009 1.8976 Available

169

10,3 1606.0000 -4.7460 0.0009 1.8976 Available 10,4 1606.0000 -4.7460 0.0009 1.8976 Available 10,5 1566.0000 -5.8090 0.0052 0.9442 Available 10,6 3049.0000 -12.7700 0.0144 0.6158 Available 10,7 3049.0000 -12.7700 0.0144 0.6158 Available 10,8 2673.0000 -5.7650 -0.0003 0.7693 Available 10,9 2673.0000 -5.7650 -0.0003 0.7693 Available 10,10 0.0000 0.0000 0.0000 1.0000 Available 10,11 -801.9000 3.8240 -0.0075 3.6039 Available 10,12 -281.4000 2.3790 -0.0067 3.1864 Available 10,13 1102.0000 -7.1760 0.0097 1.5390 Available 10,14 -355.1000 0.5800 0.0000 1.2710 Available 10,15 104.6000 -5.0140 0.0089 2.5533 Available 10,16 1114.0000 5.9160 -0.0071 0.0045 Available 10,17 3979.0000 -19.7900 0.0269 0.3491 Available 10,18 -1295.0000 4.3634 0.0000 0.2532 Available 10,19 1862.0000 32.0700 -0.0094 0.0000 Available 11,1 -17.2530 0.8389 0.0009 0.3043 Available 11,2 -17.2530 0.8389 0.0009 0.3043 Available 11,3 -17.2530 0.8389 0.0009 0.3043 Available 11,4 -17.2530 0.8389 0.0009 0.3043 Available 11,5 -1301.0000 4.0720 0.0000 0.3436 Available 11,6 332.3000 1.1580 0.0007 0.1077 Available 11,7 332.3000 1.1580 0.0007 0.1077 Available 11,8 24.1440 1.6504 0.0000 0.1816 Available 11,9 24.1440 1.6504 0.0000 0.1816 Available 11,10 1460.0000 -8.6730 0.0164 0.1644 Available 11,11 0.0000 0.0000 0.0000 1.0000 Available 11,12 -1545.0000 6.5120 0.0000 0.0526 Available 11,13 -197.5000 0.1766 0.0000 1.3223 Available 11,14 1524.0000 -2.5310 0.0000 0.3725 Available 11,15 274.5000 -0.5905 0.0022 0.3685 Available 11,16 158.4000 0.5246 0.0000 0.4105 Available 11,17 732.2000 -0.6607 0.0020 0.1489 Available 11,18 -1795.2000 12.7080 -0.0155 0.1546 Available 11,19 -1895.0000 9.3300 0.0000 0.0070 Available 12,1 256.2100 0.0000 0.0000 0.5535 Available 12,2 256.2100 0.0000 0.0000 0.5535 Available 12,3 256.2100 0.0000 0.0000 0.5535 Available 12,4 256.2100 0.0000 0.0000 0.5535 Available 12,5 202.4900 0.0000 0.0000 0.6266 Available 12,6 1011.0000 -2.1670 0.0000 0.8461 Available 12,7 1011.0000 -2.1670 0.0000 0.8461 Available 12,8 1963.0000 2.6560 -0.0136 0.2675 Available

170

12,9 1963.0000 2.6560 -0.0136 0.2675 Available 12,10 1590.0000 -24.5700 0.0621 0.0025 Available 12,11 512.0000 -2.1450 0.0000 2.6194 Available 12,12 0.0000 0.0000 0.0000 1.0000 Available 12,13 209.0000 -0.6241 0.0000 1.1521 Available 12,14 0.0000 0.0000 0.0000 1.0000 Not available 12,15 0.0000 0.0000 0.0000 1.0000 Not available 12,16 0.0000 0.0000 0.0000 1.0000 Not available 12,17 0.0000 0.0000 0.0000 1.0000 Not available 12,18 435.6400 0.0000 0.0000 0.3658 Available 12,19 0.0000 0.0000 0.0000 1.0000 Not available 13,1 -9.6540 -0.0324 0.0000 1.0562 Available 13,2 -9.6540 -0.0324 0.0000 1.0562 Available 13,3 -9.6540 -0.0324 0.0000 1.0562 Available 13,4 -9.6540 -0.0324 0.0000 1.0562 Available 13,5 -844.3000 2.9450 0.0000 0.3694 Available 13,6 179.0000 0.0562 0.0000 0.6254 Available 13,7 179.0000 0.0562 0.0000 0.6254 Available 13,8 375.0000 -1.5700 0.0000 2.0223 Available 13,9 375.0000 -1.5700 0.0000 2.0223 Available 13,10 1631.0000 -7.3620 0.0118 0.2237 Available 13,11 140.7000 0.0568 0.0000 0.6828 Available 13,12 235.7000 0.1314 0.0000 0.5089 Available 13,13 0.0000 0.0000 0.0000 1.0000 Available 13,14 0.0000 0.0000 0.0000 1.0000 Not available 13,15 0.0000 0.0000 0.0000 1.0000 Not available 13,16 0.0000 0.0000 0.0000 1.0000 Not available 13,17 957.8000 -5.7730 0.0107 0.3361 Available 13,18 521.4800 0.0000 0.0000 0.3000 Available 13,19 974.0000 -1.3680 -0.0198 2227.6909 Available 14,1 207.2600 -1.0916 0.0000 1.8462 Available 14,2 207.2600 -1.0916 0.0000 1.8462 Available 14,3 207.2600 -1.0916 0.0000 1.8462 Available 14,4 207.2600 -1.0916 0.0000 1.8462 Available 14,5 -124.3000 0.0000 0.0000 1.3324 Available 14,6 105.6300 -0.6067 0.0000 1.4374 Available 14,7 105.6300 -0.6067 0.0000 1.4374 Available 14,8 -316.2200 0.0000 0.0000 2.0752 Available 14,9 -316.2200 0.0000 0.0000 2.0752 Available 14,10 -660.2000 1.7430 0.0000 0.8035 Available 14,11 -851.0000 1.0340 0.0000 2.5362 Available 14,12 0.0000 0.0000 0.0000 1.0000 Not available 14,13 0.0000 0.0000 0.0000 1.0000 Not available 14,14 0.0000 0.0000 0.0000 1.0000 Available

171

14,15 402.6000 -1.6140 0.0000 1.9828 Available 14,16 0.0000 0.0000 0.0000 1.0000 Not available 14,17 0.0000 0.0000 0.0000 1.0000 Not available 14,18 0.0000 0.0000 0.0000 1.0000 Not available 14,19 3888.0000 -16.2600 0.0000 1456.6502 Available 15,1 205.6500 -1.4436 0.0000 2.6348 Available 15,2 205.6500 -1.4436 0.0000 2.6348 Available 15,3 205.6500 -1.4436 0.0000 2.6348 Available 15,4 205.6500 -1.4436 0.0000 2.6348 Available 15,5 -131.5000 0.0000 0.0000 1.3547 Available 15,6 16.2900 -0.6022 0.0000 1.7587 Available 15,7 16.2900 -0.6022 0.0000 1.7587 Available 15,8 978.3000 -6.4810 0.0071 3.1653 Available 15,9 978.3000 -6.4810 0.0071 3.1653 Available 15,10 1876.0000 11.5000 0.0900 0.0000 Available 15,11 -446.0000 -0.7738 0.0026 1.9397 Available 15,12 0.0000 0.0000 0.0000 1.0000 Not available 15,13 0.0000 0.0000 0.0000 1.0000 Not available 15,14 -639.9000 2.5610 0.0000 0.3383 Available 15,15 0.0000 0.0000 0.0000 1.0000 Available 15,16 0.0000 0.0000 0.0000 1.0000 Not available 15,17 0.0000 0.0000 0.0000 1.0000 Not available 15,18 0.0000 0.0000 0.0000 1.0000 Not available 15,19 1622.0000 -4.8120 -0.0186 9014.6183 Available 16,1 2257.3000 -0.5668 0.0000 0.0096 Available 16,2 2257.3000 -0.5668 0.0000 0.0096 Available 16,3 2257.3000 -0.5668 0.0000 0.0096 Available 16,4 2257.3000 -0.5668 0.0000 0.0096 Available 16,5 3982.0000 -19.7200 0.0278 0.2170 Available 16,6 154.3900 1.2458 0.0000 0.2014 Available 16,7 154.3900 1.2458 0.0000 0.2014 Available 16,8 3969.0000 8.4970 -0.0059 0.0000 Available 16,9 3969.0000 8.4970 -0.0059 0.0000 Available 16,10 1325.0000 -6.2630 0.0076 0.9222 Available 16,11 -131.0000 0.7957 0.0000 0.6106 Available 16,12 0.0000 0.0000 0.0000 1.0000 Not available 16,13 0.0000 0.0000 0.0000 1.0000 Not available 16,14 0.0000 0.0000 0.0000 1.0000 Not available 16,15 0.0000 0.0000 0.0000 1.0000 Not available 16,16 0.0000 0.0000 0.0000 1.0000 Available 16,17 1489.0000 -13.4800 0.0299 0.0545 Available 16,18 0.0000 0.0000 0.0000 1.0000 Not available 16,19 0.0000 0.0000 0.0000 1.0000 Not available 17,1 258.5700 -2.1156 0.0016 2.3091 Available

172

17,2 258.5700 -2.1156 0.0016 2.3091 Available 17,3 258.5700 -2.1156 0.0016 2.3091 Available 17,4 258.5700 -2.1156 0.0016 2.3091 Available 17,5 -13.3170 0.0000 0.0000 1.0312 Available 17,6 -590.0000 2.9160 -0.0049 1.7927 Available 17,7 -590.0000 2.9160 -0.0049 1.7927 Available 17,8 214.2000 -1.2790 0.0000 2.1913 Available 17,9 214.2000 -1.2790 0.0000 2.1913 Available 17,10 -1496.0000 9.3530 -0.0141 1.2314 Available 17,11 -619.3000 1.9300 -0.0034 2.6285 Available 17,12 0.0000 0.0000 0.0000 1.0000 Not available 17,13 460.3000 -5.6870 0.0098 1.4768 Available 17,14 0.0000 0.0000 0.0000 1.0000 Not available 17,15 0.0000 0.0000 0.0000 1.0000 Not available 17,16 245.8000 -0.1692 -0.0007 0.9089 Available 17,17 0.0000 0.0000 0.0000 1.0000 Available 17,18 -502.2100 1.0583 0.0000 1.1064 Available 17,19 0.0000 0.0000 0.0000 1.0000 Not available 18,1 2017.7000 -9.0933 0.0012 49.3034 Available 18,2 2017.7000 -9.0933 0.0012 49.3034 Available 18,3 2017.7000 -9.0933 0.0012 49.3034 Available 18,4 2017.7000 -9.0933 0.0012 49.3034 Available 18,5 -347.5000 1.2160 0.0000 0.6612 Available 18,6 613.3200 -1.5950 0.0000 1.1961 Available 18,7 613.3200 -1.5950 0.0000 1.1961 Available 18,8 29.7470 0.0000 0.0000 0.9336 Available 18,9 29.7470 0.0000 0.0000 0.9336 Available 18,10 1525.8000 -4.9155 0.0000 4.0266 Available 18,11 624.9700 -4.6878 0.0052 2.6553 Available 18,12 -188.0000 0.0000 0.0000 1.5435 Available 18,13 -310.8200 0.0000 0.0000 2.0495 Available 18,14 0.0000 0.0000 0.0000 1.0000 Not available 18,15 0.0000 0.0000 0.0000 1.0000 Not available 18,16 0.0000 0.0000 0.0000 1.0000 Not available 18,17 -504.2500 0.4034 0.0000 2.1399 Available 18,18 0.0000 0.0000 0.0000 1.0000 Available 18,19 -1398.7000 0.0000 0.0000 25.2578 Available 19,1 3264.0000 -20.8400 0.0332 0.3454 Available 19,2 3264.0000 -20.8400 0.0332 0.3454 Available 19,3 3264.0000 -20.8400 0.0332 0.3454 Available 19,4 3264.0000 -20.8400 0.0332 0.3454 Available 19,5 215.5000 -1.5190 0.0000 2.7773 Available 19,6 1885.0000 -10.9800 0.0166 0.5675 Available 19,7 1885.0000 -10.9800 0.0166 0.5675 Available

173

19,8 -69.2300 -0.7359 0.0000 2.4491 Available 19,9 -69.2300 -0.7359 0.0000 2.4491 Available 19,10 3664.0000 34.1300 0.0030 0.0000 Available 19,11 591.6000 -3.0800 0.0000 5.5522 Available 19,12 0.0000 0.0000 0.0000 1.0000 Not available 19,13 381.1000 -5.6820 0.0168 0.0860 Available 19,14 -868.8000 2.9480 0.0000 0.3898 Available 19,15 -94.8700 -9.6120 0.0372 0.0019 Available 19,16 0.0000 0.0000 0.0000 1.0000 Not available 19,17 0.0000 0.0000 0.0000 1.0000 Not available 19,18 1000.0000 0.0000 0.0000 0.0994 Available 19,19 0.0000 0.0000 0.0000 1.0000 Available

Where τlj was calculated using the equation below:

For example, to calculate the τ value for CH3 and C=C, or τ1,5:

189.66 0.2723 433.15 0 433.152 τ1,5 = exp = 0.8474 433.15

Calculation of the residual activity coefficient was done in a similar way to Smith and Van Ness2.

was calculated using the relationship shown below. The results of the calculation are shown in Table 7.4

For example, , or the value for CH3 is calculated as: = = 0.0887

2 Smith, J.M., H.C. Van Ness, and M.M. Abbot, Chemical Engineering Thermodynamics. 6 ed. 2001: McGraw Hill, Appendix H

174

Table 7.4: Calculated values of e(k) for griseofulvin and water, where the numbers are the nominal interaction parameters as described by Table 7.2

Griseofulvin Water 1 0.0887 0 2 0.0592 0 3 0.0297 0 4 0.0000 0 5 0.0000 0 6 0.0883 0 7 0.0723 0 8 0.0000 0 9 0.0000 0 10 0.0000 0 11 0.0000 1 12 0.1290 0 13 0.5359 0 14 0.0000 0 15 0.0000 0 16 0.0000 0 17 0.0000 0 18 0.0000 0 19 0.0266 0

θk was calculated using the relationship shown below, where 1 is the solute and 2 is water. The results of the equation are shown in Table 7.5.

175

Table 7.5: Theta values calculated for griseofulvin and water

v(k) (griseofulvin) (water) 1 0.0000 0.0000 2 0.0000 0.0000 3 0.0000 0.0000 4 0.0000 0.0000 5 0.0000 0.0000 6 0.0000 0.0000 7 0.0000 0.0000 8 0.0000 0.0000 9 0.0000 0.0000 10 0.0000 0.0000 11 1.0000 1.0000 12 0.0000 0.0000 13 0.0000 0.0000 14 0.0000 0.0000 15 0.0000 0.0000 16 0.0000 0.0000 17 0.0000 0.0000 18 0.0000 0.0000 19 0.0000 0.0000

The values of θ andτ were used to calculate s, shown in the equation below. The results of these values are shown in Table 7.6.

For example, s1 for griseofulvin is calculated as:

s1 = = 0*1 0*1 0*1 … 1*0.304278 … 0*0.345400 = 0.304278

Beta was calculated according to the equation below. The results are shown in

Table 7.7.

For example, the value for griseofulvin is calculated in a similar way to s1:

= 0.0887*1 0.0592*1 0*1 … 0*0.3042 … 0.0265*1 = 1.0331

176

Table 7.6: s-values calculated for griseofulvin

k v(k) k (griseofulvin) (water) 1 0.3043 0.3043 2 0.3043 0.3043 3 0.3042 0.3043 4 0.3042 0.3043 5 0.3435 0.3435 6 0.1077 0.1077 7 0.1077 0.1077 8 0.1815 0.1815 9 0.1815 0.1815 10 0.1643 0.1643 11 1.0000 1.0000 12 0.0525 0.0525 13 1.3222 1.3222 14 0.3725 0.3725 15 0.3685 0.3684 16 0.4105 0.4105 17 0.14894 0.1489 18 0.1546 0.1545 19 0.0070 0.0070

Table 7.7: β values calculated for griseofulvin

v(k) βk (griseofulvin) βk (water) 1 1.0331 0.3042 2 1.0331 0.3042 3 1.0331 0.3042 4 1.0331 0.3042 5 0.6277 0.3435 6 0.7442 0.1077 7 0.7442 0.1077 8 1.6814 0.1815 9 1.6814 0.1815 10 0.1876 0.1643 11 1.3003 1.0000 12 0.5090 0.0525 13 1.0654 1.3222 14 0.8612 0.3725 15 0.7998 0.368 16 0.7936 0.4105 17 0.4657 0.1489 18 0.2369 0.1545 19 2195.1485 0.0070

177

Rek (“Residual” in the spreadsheet file) was calculated according to the equation below. The sum of the Re values is shown in Table 7.8.

=

For example, R1 is calculated below:

= = = -0.108446

The sum of the Re(k) values is then used to calculate the residual activity coefficient.

Table 7.8: R values calculated for griseofulvin

v(k) Re(k) 1 -0.1084 2 -0.0723 3 -0.0363 4 0.0000 5 0.0000 6 -0.1707 7 -0.1397 8 0.0000 9 0.0000 10 0.0000 11 1.3003 12 -0.2927 13 0.1157 14 0.0000 15 0.0000 16 0.0000 17 0.0000 18 0.0000 19 -0.3361

ΣRk 0.2597

The residual activity coefficient was calculated using the equation below:

178

The combinatorial activity coefficient was calculated as shown in Chapter 4. The results for the calculations are shown in Table 7.9.

Table 7.9: Calculation of activity coefficients for griseofulvin at 160°C

Ji 5.9887

Li 4.8687

Ji' 3.8282

γC -0.1111

γR 8.8524 γ 8.7412

The calculation of the A-UNIFAC and AF-UNIFAC model is outlined in Chapter 4. Sample calculations are not presented in this section. As the relationships and values calculated in this sample calculation simply need to be substituted into equations 4.18 – 4.23, with equation 4.24 being used to calculate the AF-UNIFAC model activity coefficients.

179

8. APPENDIX B

Liquid-Liquid Equilibrium Mutual Solubility Relationship Development

The following section presents the development of the thermodynamic relationship between activity coefficients and the mutual solubilities (or miscibilities) of a binary liquid mixture. The presented derivation is similar to that of Prausnitz et al. [1]. This section is to be used as a reference to equation 4.20 presented in Chapter 4.

180

8.1. THERMODYNAMIC DEVELOPMENT

At equilibrium, the fugacity condition for a two phases of liquid for component 1 is:

Where the subscript 1 designates component 1, and a and b designates phase a and b respectively. Similarly, for component 2:

Non-ideal phase behaviour under these two conditions is thus written as:

The standard state fugacity for component 1 in both phases is generally taken at the same condition in both phases, thus the equation simplifies to:

With a similar result for component 2:

A mass balance performed on the above two equations leads to the formation of the following expression for the equilibrium relationship for component 2:

In the case where species 1 and 2 are very dilute in phase a and b respectively (as in the case in Chapter 4 for liquid fatty acid solubility in water, where the mole fraction in the order of 10-4), the four activity coefficients reduce to:

Thus the expressions for phase equilibrium for very dilute species become:

By solving these equations simultaneously by estimating the activity coefficient at infinite dilution of species 1 and 2, the mole fraction of solute in each phase may be solved.

181

9. APPENDIX C

Naproxen data above 150°C

9.1. INTRODUCTION

Naproxen solubility in SBCW and particle formation from SBCW solutions was originally studied between 130°C and 170°C. At 152°C, naproxen melts. An experiment was performed to determine whether the naproxen melt, at 160°C and above, was extruded through the filter in the solubility vessel (SV), shown in Chapter 3 and 5. It was found that naproxen was extruded through the filter. Thus, the solubility data collected above 150°C was rendered void, as the data no longer accurately characterized solid-liquid or liquid-liquid equilibrium (SLE or LLE).

Despite the loss of accurate solubility data, the collected information was useful. The mass of extruded naproxen and dissolved naproxen was still known, albeit in two phases. The injection of the SBCW-naproxen mixture into cold water resulted in unique spherical crystal agglomerates forming. As discussed in Chapter 5, the formation of agglomerates was likely the result of extruded naproxen spheres acting as nucleation points for crystal growth. As the temperature of the solution decreased, the dissolved naproxen precipitated as regular flake crystals (shown in Chapter 5) around the nucleation points. This is a valuable result, as one of the major challenges in the pharmaceutical industry is to produce rough-surfaced spheres for inhalable drug delivery (as discussed in Chapter 5). Thus the particle formation experiments were not made void, despite the failure of the apparatus to be capable of producing accurate solubility data, as valuable results were recorded.

The solubility of naproxen was initially recoded at 160°C and 170°C is presented in this appendix, as well as the method of determining whether naproxen extruded through the filter frits at 170°C.

9.2. NAPROXEN SOLUBILITY DATA Table 9.1: Naproxen mixture mole fractions extracted from SV above 150°C

T x2 (106) SD (106) (°C) (mole fraction) (mole fraction) 160 1790 8.65 170 5563 357

182

9.3. EXTRUSION EXPERIMENT

An experiment was performed on the naproxen-SBCW system at 170°C. The aim of the experiment was to extrude melted naproxen solid through the filter frit to determine whether the naproxen melt remained in the vessel (in which case the apparatus was sufficient in measuring LLE near the melting point of naproxen), or was extruded into the collection vial, thus affecting solubility results.

For the extrusion study, the same system as presented in the SBCW solubility studies (Chapter 3 of this thesis) was used with an excess of naproxen (0.5g) in the solubility equilibrium vessel (SV). The solubility method was then reproduced at 170°C, without adding water into SV. After a 30 minute equilibration time, the vessel contents were then purged into a vial. The vessel and vial contents post-experiment were then observed to see whether naproxen melt was extruded through the filter frit.

9.3.1. RESULT Small amounts of naproxen were extruded through the filter and into the collection beaker. The valve blocked midway through the collection. This result implies that naproxen beyond the melting point was extruded through the filter.

183