Colour Removal from Sugar Cane Juice

COLOUR REMOVAL FROM SUGAR CANE JUICE

Danny M. T. Nguyen

Submitted in fulfilment of the requirements for the degree of Doctor of Philosophy

School of Chemistry, Physics and Mechanical Engineering Science and Engineering Faculty Queensland University of Technology, Brisbane, Australia

June 2013

Supervisor: Professor William O. S. Doherty Sugar Research and Innovation Centre for Tropical Crops and Biocommodities Queensland University of Technology, Brisbane, Australia

Associate Supervisor: Adjunct Associate Professor John P. Bartley School of Chemistry, Physics and Mechanical Engineering Science and Engineering Faculty Queensland University of Technology, Brisbane, Australia

The research was carried out within the Centre for Tropical Crops and Biocommodities at the Queensland University of Technology.

IMPORTANT NOTICE

The information in this thesis is confidential and should not be disclosed for any reason or relied on for a particular use or application. Any invention or other intellectual property described in this document remains the property of the Queensland University of Technology.

iii

DECLARATION OF AUTHORSHIP

The work contained in this thesis has not been submitted for assessment for any other award. Wherever contributions of others are involved, every effort is made to indicate this clearly with proper reference to the literature and acknowledgement of collaborative research and discussions. Some parts of the research work in this thesis have been published and a list of publications arising from this research has been provided.

QUT Verified Signature

.. ...

Danny M. T. Nguyen, BSc (Hons)

Date: ......

Abstract

One of the most important parameters in raw sugar quality is colour. Australian raw sugars are considered to be of high quality with respect to this parameter. However, some raw sugars produced in both Australia and overseas are relatively difficult to decolourise by sugar refiners, and tend to develop colour during storage. A new approach that has the potential to efficiently and cost-effectively decolourise sugar process streams is through the use of the Fenton oxidation and related processes. The Fenton oxidation process involves the catalytic production of hydroxyl radicals from the decomposition of hydrogen peroxide (H2O2) using iron(II), which has the potential to effectively degrade colour and colour precursors present in aqueous systems.

As a first step towards developing this technology, this study determined the colour content and the composition of colour precursors (i.e., phenolic acids), present in sugar cane juices processed by Australian sugar factories. The results showed that caffeic, p–coumaric and ferulic acids (classed as hydroxycinnamic acids) are the main phenolic acids present in sugar cane juice. The study was able to identify flavonoids (e.g., chrysin, morin, quercetin and rutin) because of modifications of the methods used in the evaluation of colourants in sugar cane juice.

The results also show that juice expressed partly or solely from whole crop harvested cane, has significantly higher colour (11,400–20,000 IU) than juices expressed from burnt harvested cane (10,400–12,700 IU). However, the concentrations of phenolic acids in burnt cane were twice as much as those obtained in whole crop cane. This is probably due to the thermal decomposition of HMW phenolics (viz., lignin, polyphenols) during cane burning.

The Fenton oxidation process was used to study the degradation of these hydroxycinnamic acids (i.e., caffeic, p–coumaric and ferulic) in water and sucrose solutions. Central composite design and response surface methodologies were used to evaluate and optimise the interactive effects of the process parameters. Quadratic polynomial models were developed for the degradation of each of the individual

acids, and the total hydroxycinnamic acid mixtures. The optimum degradation efficiency (77%) in an aqueous solution containing the hydroxycinnamic acids (200 mg/L) was optimum at pH 4.7 and at 25 °C. The efficiency dropped in the presence of sucrose to 57% at pH 5.4 and at 36 °C.

In a mixture of these hydroxycinnamic acids, the degradation behaviour of caffeic acid differed from those of p–coumaric and ferulic acids, because unlike the other acids, it forms a complex with iron(III). Iron(III) is produced in situ during the oxidation process.

Analysis of the Fenton degradation products showed the presence of low molecular weight phenolics, aliphatic carboxylic acids as well as several oligomer products. The tentative mechanisms of formation of these compounds have been proposed.

To improve the effectiveness of the Fenton process, aluminium chloride was added to act as a pro-oxidant. This process was evaluated on a synthetic juice solution consisting of sucrose (15% (w/w)), the hydroxycinnamic acids (200 mg/L) and a synthetic glucose-glycine melanoidin (2,000 mg/L). The modified Fenton process degraded the melanoidin and the hydroxycinnamic acid mixture by approximately 69% and 53% respectively. In the absence of aluminium chloride, the Fenton process on its own resulted in 63% and 47% degradation, respectively but only achieved 24% decolourisation. However, the addition of aluminium chloride played a significant role in the removal of colour with up to 43% decolourisation achieved.

The modified Fenton process was then evaluated for the decolourisation of authentic factory juices. There were increases in colour measured at pH 4.0 (≤ 45%) and pH 7.0 (≤ 21%). However, there was decrease for the colour measured at pH 9.0 (≤ 42%). Colour is usually measured at pH 7.0 but additional information about the nature of colourants is obtained at pH 4.0 and pH 9.0. Colour measured at pH 4.0 suggests the presence the presence of high molecular weight colourants, while colour measured at pH 9.0 is due to the presence of natural colourants such as flavonoids and phenolics. The colour at pH 9.0 is more likely to be transferred to the crystal, so there may well be colour reduction if the treated juice is further processed to raw sugar.

The key contribution contained in this thesis is an understanding of the degradation of colour precursors in sugar solutions. A new direction of research for the removal of colour and colour precursors in sugar process streams has been identified.

vii

Keywords

Colour Colourants Colour precursors Sugar Sugar cane juice Sugar quality Sucrose Decolourisation Degradation Fenton Advanced oxidation process Hydroxycinnamic acids Caffeic acid p–Coumaric acid Ferulic acid Response surface methodology Experimental design UV/Visible spectroscopy High-performance liquid chromatography Reaction pathways Clarification Aluminium chloride Melanoidin Reducing sugars

viii

Acknowledgements

A PhD candidature, by its very own nature, is a very unique endeavour. There is no right way to undertake a PhD project. However, there are many wrong ways that one could take throughout their candidature. I am for one, a very glad person, who has taken the best path possible in order to complete my candidature and hopefully graduate with a doctoral degree. I could have not taken this path without the consistent guidance and advice given from the very kind people that I have met throughout my entire candidature to whom I give thanks to.

First and foremost, I would like to sincerely thank my primary supervisor, Prof. William (Bill) Doherty, for his constant patience, guidance, encouragement and commitment to this work. Bill, you have been a great mentor. I have learnt and gained so much from you. Despite our differences and heated discussions on several aspects of this thesis, you have always seen the best in me. Towards the end of writing this thesis, I was asked by many for an inspirational and memorable quote from you. In response to that, that would definitely be, “Danny, could you please come to my office? I need to see you.” I am very glad because every time I walked into your office with the heater running on a warm Brisbane day, I would learn something new, no matter how irrelevant it is to my own work. Thank you.

I would also like to thank my associate supervisor, Adj. A/Prof. John Bartley. You have always been prompt whenever I needed you most and you have been a great mentor. I appreciate all the times, especially at the very early stages of my candidature, assisting me with certain aspects of organic chemistry. I always gained something useful each time we met. Up to today, I still have a strong passion for organic chemistry and to me, drawing chemical structures for reaction mechanisms is genuinely a form of art.

This project would have not happened, if it was not for the financial support from my scholarship sponsors. To the three main sponsors, the Queensland University of Technology (QUT), the Sugar Research and Development Corporation and Sugar Research Limited, a very big thank you for your generosity. My exposure

to the Australian sugar industry has been very worthwhile. This was a very rare opportunity and I am grateful that each sponsor accepted me to undertake this project. In addition, I would also like to thank the production staff of Condong Sugar Mill, Tully Sugar Mill and Isis Central Sugar Mill who gave me access to their facilities during the crushing seasons.

Many thanks must go to all the academic and technical staff who have contributed to this project throughout my candidature. Prof. Robert (Bob) Gilbert (University of Queensland, UQ) for his expertise on food polymers; Dr. Peter Sopade (UQ) and A/Prof. Geoff Kent (QUT) for introducing me into the world of multivariate statistics; Mr. Hakan Bakir (QUT) for his assistance at the mills during the factory trials; Mr. Tony Raftery (QUT) for his assistance on XRD analysis; Dr. Chris Carvalho and Mrs. Leonora Newby for analytical instrument training; and Ms. Wanda Stolz (QUT) for her endless hospitality in the lab.

To my fellow colleagues who work closely with me, thank you for your ongoing support. Chris East, you have been a great mate throughout my candidature and thanks for changing my life that day (you know what happened). William Gilfillan, thanks for keeping a lookout for Bill every time he approaches into our office. Travelling with you to the conferences has been great. You always take the best photos! Josh Howard, thanks for your feedback during our research meetings. All the best with your PhD mate. Darryn Rackemann, massive thanks for your hospitality in general as well as your advice on various aspects of the sugar industry. It has been great to work alongside you. Caroline Thai for her patience and generosity throughout our university lives since the days back at RMIT University. I am sorry, if I ever convinced you to do a PhD but in the end we know it was worth it. Thank you for the seven years of memories. A final message to the whole group, I am very glad to have met all of you and I wish you all the best throughout your careers.

There are far too many people to list all of them individually, but I am indebted to all of them at one time or another, for their support and giving me the motivation to complete my candidature. These people are all the staff and students from the Centre of Tropical Crops and Biocommodities (QUT), the School of Chemistry, Physics and Mechanical Engineering (QUT) and the administrative and

HDR support staff (QUT). Also, to the SEF HDR Student Society at QUT, thank you for giving me the opportunity to be the founding chairman of the society. It has been a pleasure during the inaugural year and I wish all the best for the team in the future.

There is one more group of people that I am very fond and close to that I need to recognise for their long-distance support and love, that is my family. Leaving home and family for a long period of time (once again… sorry mum!) was not so easy. I cannot remember how long since I left home but the words “Không có văn bằng học là big trouble! Okay?” (Vietnamese: No (PhD) degree means big trouble (for you)! Okay?) are still ringing in my ear. Maybe that has been a driving force for me to finish my candidature. To my family back in Melbourne, thank you and I will always make the both of you, mum and dad, proud! To my brother, Steven, good luck with Year 12 exams. Considering taking chemistry in university next year? You should!

I dedicate this thesis to my family and friends for nursing me with affections and love and for their dedication for success in my life.

“The surest way not to fail is to determine to succeed.”

Rt. Hon. Richard Brinsley Sheridan

xii

Publications and Awards

Refereed Journal Papers

Nguyen, D. M. T., & Doherty, W. O. S. (2013). Optimisation of process parameters for the removal of hydroxycinnamic acids in sugar solutions. International Sugar Journal, accepted for publication.

Nguyen, D. M. T., & Doherty, W. O. S. (2012). Optimisation of process parameters for the degradation of caffeic acid in sugar solutions. International Journal of Food Science and Technology, 47(12), 2477-2486.

Nguyen, D. M. T., & Doherty, W. O. S. (2012). Phenolics in sugar cane juice: Potential degradation by hydrogen peroxide and Fenton's reagent. International Sugar Journal, 114(1361), 309-315.

Conference Proceedings

Nguyen, D. M. T., & Doherty, W. O. S. (2012) Process optimisation for the degradation of phenolic compounds in water and sugar solutions. Proceedings of the Second International Conference on Advanced Oxidation Processes, 72- 73.

Nguyen, D. M. T., & Doherty, W. O. S. (2012) Optimisation of process parameters for the removal of hydroxycinnamic acids in sugar solutions. Proceedings of the Australian Society of Sugar Cane Technologists, 34, (electronic format).

Nguyen, D. M. T., & Doherty, W. O. S. (2011) Phenolics in sugar cane juice: Potential degradation by hydrogen peroxide and Fenton’s reagent. Proceedings of the Australian Society of Sugar Cane Technologists, 33, (electronic format).

xiii

Conference Posters

Nguyen, D. M. T., & Doherty, W. O. S. (2012) Combined Fenton oxidation and chemical coagulation for the treatment of melanoidin/phenolic acid mixtures. Presented at the Second International Conference on Advanced Oxidation Processes, Kottayam, Kerala, India.

Conference Lectures

Nguyen, D. M. T., & Doherty, W. O. S. (2012) Process optimisation for the degradation of phenolic compounds in water and sugar solutions. Presented at the Second International Conference on Advanced Oxidation Processes, Kottayam, Kerala, India.

Nguyen, D. M. T., & Doherty, W. O. S. (2012) Optimisation of process parameters for the removal of hydroxycinnamic acids in sugar solutions. Presented at the 34th Australian Society of Sugar Cane Technologists, Palm Cove, Queensland, Australia.

Nguyen, D. M. T., & Doherty, W. O. S. (2011) Phenolics in sugar cane juice: Potential degradation by hydrogen peroxide and Fenton’s reagent. Presented at the 33rd Australian Society of Sugar Cane Technologists, Mackay, Queensland, Australia.

Awards

Presenting Science Award (2013) for the best presentation presented at the Sugar Research and Development Corporation Scholarship Forum, Townsville, Queensland, Australia.

Young Investigators Award (2012) for the best paper presented at the Second International Conference on Advanced Oxidation Processes, Kottayam, Kerala, India.

Denis Foster Chemistry/Chemical Engineering Award (2012) for the best paper presented by a chemistry/chemical engineering tertiary student at the 34th Australian Society of Sugar Cane Technologists Conference, Palm Cove, Queensland, Australia.

xiv

Table of Contents

Abstract...... v Keywords...... viii Acknowledgements...... ix Publications and Awards...... xiii List of Figures...... xx List of Tables...... xxvii List of Abbreviations and Nomenclature...... xxxi 1. GENERAL INTRODUCTION...... 1 1.1 Background and Motivation...... 2 1.2 Research Problem...... 3 1.3 Aims and Objectives...... 4 1.4 Scope of this Thesis...... 5 2. LITERATURE REVIEW...... 9 2.1 Introduction...... 10 2.2 Colourants in Sugar Process Streams...... 10 2.2.1 Naturally Occurring Colourants...... 12 2.2.2 Factory Produced Colourants...... 16 2.3 Reactivity of Colourants during Sugar Manufacturing...... 18 2.3.1 Enzymatic Browning...... 18 2.3.2 Non-enzymatic Oxidation...... 20 2.3.3 Maillard Reaction...... 21 2.3.4 Caramelisation...... 24 2.3.5 Hexose Alkaline Degradation...... 27 2.3.6 Conversion of Anthocyanins to Chalcones...... 29 2.3.7 Biochemical Precursors of Flavonoids...... 30 2.4 Colour in Sugar Process Streams...... 31 2.4.1 Effects of Temperature on Colour Formation...... 35 2.5 Sugar Decolourisation Technologies...... 37 2.5.1 Current Technologies...... 37

2.5.2 Decolourisation using Chemical Additives...... 38 2.5.3 Novel and Potential Technologies...... 42 3. DETERMINATION OF PHENOLIC COMPOUNDS IN FACTORY SUGAR CANE JUICES...... 57 3.1 Introduction...... 58 3.2 Materials and Methods...... 58 3.2.1 Reagents and Solvents...... 58 3.2.2 Specification of Samples...... 59 3.2.3 Sample Preparation...... 60 3.2.4 Instrumental Procedures and Analyses...... 60 3.2.5 Colour, Refractive Index and Total Soluble Solids Measurements...... 62 3.3 Results and Discussion...... 62 3.3.1 Colour Analyses of Juices...... 62 3.3.2 Phenolic Content in Juices...... 63 3.4 Summary...... 70 4. DEGRADATION OF HYDROXYCINNAMIC ACIDS...... 73 4.1 Introduction...... 74 4.2 Materials and Methods...... 75 4.2.1 Reagents and Solvents...... 75 4.2.2 Catalytic and Non-catalytic Oxidation of Caffeic Acid...... 75 4.2.3 Fenton Oxidation Reactions for Caffeic Acid Degradation...... 76 4.2.4 Fenton Oxidation Reactions for the Degradation of Hydroxycinnamic Acid Mixtures...... 78 4.2.5 Instrumental Procedures and Analyses...... 78 4.2.6 Performance Assessment of the Fenton Oxidation Process...... 79 4.2.7 Design of Experiments...... 80 4.2.8 Statistical Analysis...... 82 4.2.9 Evaluation of the Interactions between Fe(II) and Hydroxycinnamic Acids...... 82 4.3 Results and Discussion...... 83 4.3.1 Catalytic and Non-catalytic Oxidation of Caffeic Acid in Aqueous Systems...... 83

xvi

4.3.2 Optimisation of Process Parameters for the Degradation of Caffeic Acid in Sugar Solutions...... 87 4.3.3 Degradation of Hydroxycinnamic Acid Mixtures...... 100 4.4 Summary...... 130 5. SEPARATION AND IDENTIFICATION OF FENTON OXIDATION PRODUCTS DERIVED FROM HYDROXYCINNAMIC ACIDS...... 137 5.1 Introduction...... 138 5.2 Materials and Methods...... 138 5.2.1 Reagents and Solvents...... 138 5.2.2 Fenton Oxidation Reactions for the Degradation of Hydroxycinnamic Acid Mixtures...... 138 5.2.3 Sample Preparation...... 139 5.2.4 Instrumental Procedures and Analyses...... 140 5.2.5 Fenton Oxidation Reactions for the Degradation of Sucrose Mixtures...... 142 5.2.6 Computational Methods...... 142 5.3 Results and Discussion...... 143 5.3.1 Identification of Oxidation Products...... 143 5.3.2 Proposed Degradation Pathways of Selected Hydroxycinnamic Acids...... 153 5.3.3 Oligomerisation of Hydroxycinnamic Acids...... 166 5.4 Summary...... 171 6. DEGRADATION OF MELANOIDIN AND HYDROXYCINNAMIC ACID MIXTURES...... 177 6.1 Introduction...... 178 6.2 Materials and Methods...... 178 6.2.1 Reagents and Solvents...... 178 6.2.2 Preparation of Synthetic Melanoidin...... 179 6.2.3 Modified Fenton Oxidation Process...... 179 6.2.4 Instrumental Procedures and Analyses...... 179 6.2.5 Performance Assessment of the Modified Fenton Oxidation Process...... 180 6.2.6 Design of Experiments...... 181

xvii

6.2.7 Statistical Analysis...... 182 6.3 Results and Discussion...... 182 6.3.1 Monitoring Melanoidin and Hydroxycinnamic Acid Degradation...... 182 6.3.2 Transformation of Data, Regression Modelling and Statistical Analysis...... 184 6.3.3 Oxidation Performance of Melanoidins...... 190 6.3.4 Oxidation Performance of Hydroxycinnamic Acids...... 194 6.3.5 Response Surface Analyses for the Decolourisation of Mixtures. 198 6.3.6 Process Optimisation and Validation...... 200 6.4 Summary...... 203 7. EVALUATION OF FENTON AND FENTON-LIKE PROCESSES FOR THE REMOVAL OF COLOUR FROM FACTORY SUGAR CANE JUICE...... 205 7.1 Introduction...... 206 7.2 Materials and Methods...... 206 7.2.1 Reagents and Solvents...... 206 7.2.2 Specification of Samples...... 207 7.2.3 Decolourisation Procedure...... 207 7.2.4 Preparation of Flocculants...... 207 7.2.5 Preparation of Lime Saccharate...... 208 7.2.6 Clarification Procedure...... 208 7.2.7 Turbidity Measurements...... 209 7.2.8 Sucrose, Dry Substance and Purity Measurements...... 210 7.2.9 Reducing Sugars Composition Analyses...... 210 7.2.10 Colour, Refractive Index and Total Soluble Solids Measurements...... 210 7.2.11 Inorganic Ion Composition Analyses...... 211 7.3 Results and Discussion...... 211 7.3.1 First Decolourisation Trials...... 211 7.3.2 Second Decolourisation Trials...... 215 7.3.3 Economic Considerations...... 222 7.4 Summary...... 222

xviii

8. CONCLUSIONS AND FUTURE ASPECTS...... 227 8.1 Findings of the Thesis...... 228 8.2 Recommendations for Future Work...... 231 Appendices...... 235

xix

List of Figures

Figure 2.1 Schematic flowchart of the sugar manufacturing process in Australia………………………………………………………… 11 Figure 2.2 Flavonoid structures found in sugar process streams and products. Examples are given under the general chemical structures (Ververidis et al., 2007)...... 13 Figure 2.3 Structures of phenolics found in sugar process streams and products. Examples are given under the general chemical structures (Harborne, 1989; Vermerris and Nicholson, 2006)..... 14 Figure 2.4 Polymerisation of monomeric gallic acid to polyphenols ellagic acid and ellagitannin (Ross et al., 2007)...... 15 Figure 2.5 Redox chemistry of phenolics under copper and iron to produce colour forming products as proposed by Danilewicz et al. (2008)...... 20 Figure 2.6 An example of a basic melanoidin structure formed from 3–deoxyhexosuloses (Cämmerer et al., 2002)...... 26 Figure 2.7 An example of a melanoidin polymer formed from 3–deoxyhexosuloses and amino acids proposed by Cämmerer and Kroh (1995)...... 27 Figure 2.8 Condensation product formed from the reaction of HMF and a ketone; followed by an additional condensation reaction with a second equivalent of HMF (Chheda and Dumesic, 2007)…...... 27 Figure 2.9 Formation of colour among three clarification processes; mixed juice (MJ), heated juice (HJ), incubated juice (IJ), limed juice (LJ), flocculated heated limed juice (FHLJ), evaporator supply juice (ESJ), final evaporator syrup (FES) and raw sugar (RS) (Eggleston et al., 2003)...... 34

Figure 3.1 Separation of a typical mixture of compounds in the FEJ extract of burnt harvested cane by HPLC-DAD (Method A, UV/Vis detection at 280 nm). 1 = gallic acid (tentative), 2 = HMF, 3 = 4–hydroxybenzoic acid, 4 = chlorogenic acid, 5 = vanillic acid, 6 = caffeic acid, 7 = 2,3–dihydroxybenozic acid, 8 = protocatechuic acid (tentative), 9 = p–coumaric acid, 10 = ferulic acid...... 64 Figure 3.2 Separation of a standard mixture of compounds by HPLC-DAD (Method B, UV/Vis detection at 280 nm). 1 = gallic acid, 2 = HMF, 3 = protocatechuic acid, 4 = furfural, 5 = 4–hydroxybenzoic acid, 6 = (±)–catechin, 7 = vanillic acid, 8 = caffeic acid, 9 = chlorogenic acid, 10 = vanillin, 11 = p–coumaric acid, 12 = syringaldehyde, 13 = ferulic acid, 14 = sinapinic acid, 15 = coumarin, 16 = o–coumaric acid, 17 = 3,4,5–trimethoxybenzoic acid, 18 = rutin, 19 = diosmin, 20 = chrysin, 21 = morin, 22 = quercetin...... 67 Figure 3.3 Separation of a typical mixture of compounds in the PJ extract of burnt harvested cane by HPLC-DAD (Method B, UV/Vis detection at 280 nm). 1 = gallic acid, 2 = HMF, 3 = protocatechuic acid, 4 = furfural, 5 = 4–hydroxybenzoic acid, 6 = vanillic acid, 7 = caffeic acid, 8 = p–coumaric acid, 9 = syringaldehyde, 10 = ferulic acid, 11 = sinapinic acid, 12 = coumarin, 13 = rutin, 14 = diosmin, 15 = chrysin, 16 = morin, 17 = quercetin...... 68 Figure 4.1 Schematic representation of heating block used for the Fenton oxidation process...... 77

Figure 4.2 Absorption spectra of CaA after the addition of 2.94 mM H2O2 at pH 3.0 at 25 °C...... 84 Figure 4.3 Degradation of CaA (measured at 320 nm) using Fenton’s reagent at different initial pH at 25 °C. Concentrations of

H2O2: (a) 11.8 mM and (b) 2.94 mM...... 86 Figure 4.4 Plot of predicted and experimental (actual) values for the degradation (%) of CaA...... 91

xxi

Figure 4.5 Normal probability plot of residuals for fitted model using CaA degradation data...... 91 Figure 4.6 Three-dimensional surface plots of CaA degradation (%) as a

function of (a) CaA and Fe(II); (b) sucrose and H2O2; (c) sucrose and temperature; and (d) pH and Fe(II). Variables: CaA (1.11 mM); sucrose (0% (w/w)); pH (5.0); Fe(II) (0.45

mM); H2O2 (6.62 mM); temperature (35 °C) and time (120 s).... 93 Figure 4.7 Three-dimensional surface plots of CaA degradation (%) as a

function of (a) pH and H2O2; (b) Fe(II) and H2O2; (c) H2O2 and

temperature; and (d) H2O2 and time. Variables: CaA (1.11 mM); sucrose (0% (w/w)); pH (5.0); Fe(II) (0.45 mM);

H2O2 (6.62 mM); temperature (35 °C) and time (120 s)...... 95 Figure 4.8 Normal probability plot of residuals for fitted model using CaA degradation data before power transformation...... 101 Figure 4.9 Box-Cox plots of (a) CaA and (b) pCoA degradation data for the determination of the optimised power transformed response surface models...... 103 Figure 4.10 Box-Cox plots of (a) FeA and (b) total HCA degradation data for the determination of the optimised power transformed response surface models...... 104 Figure 4.11 Normal probability plots of residuals for fitted model using (a) CaA and (b) pCoA degradation data after power transformation...... 105 Figure 4.12 Normal probability plots of residuals for fitted model using (a) FeA and (b) total HCA degradation data after power transformation...... 106 Figure 4.13 Plots of predicted response and experimental (actual) values for the degradation (%) of (a) CaA and (b) pCoA...... 114 Figure 4.14 Plots of predicted response and experimental (actual) values for the degradation (%) of (a) FeA and (b) total HCA...... 115

xxii

Figure 4.15 Perturbation plots for the degradation (%) of (a) CaA; (b) pCoA and (c) FeA. Coded values are shown for each factor: total HCA (A); sucrose (B); pH (C) and temperature (D); and refer to actual values listed in Table 4.3...... 117 Figure 4.16 Effect of pH (pH 4.0–6.0) on the absorption spectra of CaA (0.055 mM) at 25 °C: (a) in the absence and (b) in the presence of Fe(II) (0.04 mM)...... 119 Figure 4.17 Normalised ATR-FTIR spectra of CaA solutions at 25 °C after subtraction of acetate buffer (pH 5.5): (a) in the absence and (b) in the presence of Fe(II)...... 122 Figure 4.18 Normalised ATR-FTIR spectra of CaA solutions containing sucrose at 25 °C after subtraction of acetate buffer (pH 5.5): (a) in the absence and (b) in the presence of Fe(II)...... 123 Figure 4.19 Three-dimensional surface plots of total HCA degradation (%) as a function of (a) total HCA and sucrose; (b) sucrose and pH; and (c) pH and temperature. Variables: total HCA (155 mg/L); sucrose (7.5% (w/w)); pH (5.0) and temperature (35 °C)...... 124 Figure 5.1 High-performance LC-DAD chromatograms (UV/Vis detection at 280 nm) of (a) CaA; (b) pCoA and (c) FeA; subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C)...... 145 Figure 5.2 Total ion chromatograms (negative ion mode ESI-MS) of (a) CaA; (b) pCoA and (c) FeA; subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C)...... 148 Figure 5.3 Gas chromatograms of SPE extracts of (a) CaA; (b) pCoA and (c) FeA solutions; subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C)...... 152 Figure 5.4 Electrostatic potential maps and equilibrium geometries of (a) CaA; (b) pCoA and (c) FeA as derived from B3LYP/ 6-31+G* calculations...... 154 Figure 5.5 Proposed structure of a tetramer of caffeic acid (m/z 715) by Agha et al., (2009)...... 171

xxiii

Figure 6.1 Typical HPLC-FLD chromatogram (fluorescence detection at

λex = 350 nm and λem = 445 nm) of the melanoidin/HCA mixture in sucrose solution (15% (w/w)) before and after modified Fenton oxidation (t = 2 min) at pH 5.6 and 35 °C...... 183 Figure 6.2 Typical HPLC-DAD chromatogram (UV/Vis detection at 280 nm) of the melanoidin/phenolic acid mixture in sucrose solution (15% (w/w)) before and after modified Fenton oxidation (t = 2 min) at pH 5.6 and 35 °C. 1 = CaA, 2 = pCoA, 3 = FeA. 184 Figure 6.3 Perturbation plot for (%) melanoidin degradation. Coded values are shown for each factor: melanoidin (A); total HCA

(B); pH (C); FeSO4·7H2O dose (D) and AlCl3·6H2O dose (E); and refer to actual values listed in Table 6.1...... 191 Figure 6.4 Contour plots of melanoidin degradation (%) as a function of

(a) melanoidin and AlCl3·6H2O dosage; (b) pH and

FeSO4·7H2O dosage. Variables: melanoidin (1,500 mg/L);

total HCA (150 mg/L); pH (5.25); FeSO4·7H2O dosage (389

mg/L) and AlCl3·6H2O dosage (200 mg/L)...... 192 Figure 6.5 Contour plots of melanoidin degradation (%) as a function of

(a) pH and AlCl3·6H2O dosage; (b) FeSO4·7H2O dosage and

AlCl3·6H2O dosage. Variables: melanoidin (1,500 mg/L); total

HCA (150 mg/L); pH (5.25); FeSO4·7H2O dosage (389 mg/L)

and AlCl3·6H2O dosage (200 mg/L)...... 193 Figure 6.6 Perturbation plot for (%) total HCA degradation. Coded values are shown for each factor: melanoidin (A); total HCA (B); pH

(C); FeSO4·7H2O dose (D) and AlCl3·6H2O dose (E); and refer to actual values listed in Table 6.1...... 195 Figure 6.7 Contour plots of total HCA degradation (%) as a function of

(a) melanoidin and pH; (b) melanoidin and FeSO4·7H2O dosage. Variables: melanoidin (1,500 mg/L); total HCA

(150 mg/L); pH (5.25); FeSO4·7H2O dosage (389 mg/L) and

AlCl3·6H2O dosage (200 mg/L)...... 196

xxiv

Figure 6.8 Contour plots of total HCA degradation (%) as a function of

(a) total HCA and FeSO4·7H2O dosage; (b) total HCA and

AlCl3·6H2O dosage. Variables: melanoidin (1,500 mg/L); total

HCA (150 mg/L); pH (5.25); FeSO4·7H2O dosage (389 mg/L)

and AlCl3·6H2O dosage (200 mg/L)...... 197 Figure 6.9 Three-dimensional surface plots of decolourisation (%) as a

function of (a) melanoidin and AlCl3·6H2O dosage; (b) total

HCA and FeSO4·7H2O; (c) total HCA and AlCl3·6H2O dosage;

and (d) FeSO4·7H2O and AlCl3·6H2O. Variables: melanoidin

(1,500 mg/L); total HCA (150 mg/L); pH (5.25); FeSO4·7H2O

dosage (389 mg/L) and AlCl3·6H2O dosage (200 mg/L)...... 199 Figure 7.1 Sugar Research Institute designed batch settling kit...... 209 Figure A2.1 High-performance LC-DAD chromatograms (UV/Vis detection at 280 nm) of the HCA mixture subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C)…………...…………………………… 259 Figure A2.2 Total ion chromatogram (negative ion mode ESI-MS) of the HCA mixture subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C).….……………....………………...…………… 260 Figure A2.3 Gas chromatogram of a SPE extract of the HCA mixture subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C).……...... 260 Figure A2.4 Negative ion mode ESI-MS full-scan spectrum relevant to the dimer product arising from the Fenton oxidation of FeA, [M]– = 385.1 Da …...... …...……………….………………… 261 Figure A2.5 Negative ion mode ESI-MS full-scan spectrum relevant to the tetramer product arising from the Fenton oxidation of CaA, [M]– = 715.2 Da ……………………….……..………………… 261 Figure A3.1 Normal probability plots of residuals for fitted model using (a) melanoidin and (b) total HCA degradation data after power transformation.….………………………………….…………… 266 Figure A3.2 Box-Cox plots of (a) melanoidin and (b) total HCA degradation data for the determination of the optimised power transformed response surface models.………….….………………………… 267

xxv

Figure A3.3 Plots of predicted response and experimental (actual) values for the degradation (%) of (a) melanoidin and (b) total HCA.……... 268 Figure A3.4 Plot of predicted response and experimental (actual) values for the decolourisation (%).…………………………………….…... 269

xxvi

List of Tables

Table 2.1 Properties of colourants in cane juice (Davis, 2001a)...... 32 Table 2.2 Colour analyses of milled juice at pH 7.0 (Smith et al., 1981)...... 33 Table 2.3 Comparison of colour at pH 7.0 and 9.0 from process streams of a typical sugar mill (Smith et al., 1981)...... 34 Table 2.4 Decolourisation processes on colourants types existing in juice as adapted from Davis (2001b)...... 37 Table 3.1 Colour of factory cane juices recorded at pH 7.0...... 63 Table 3.2 Phenolic acids and HMF (mM on dry content) by HPLC-DAD of sugar cane juices using Method A...... 65 Table 3.3 Phenolic acids (mM on juice) and HMF by HPLC-DAD of PJs using Method A...... 66 Table 3.4 Phenolic acids and HMF (mM on dry content) by HPLC-DAD of sugar cane juices using Method B...... 69 Table 4.1 Volumes of reagents (mM) used for the degradation of CaA...... 76 Table 4.2 Coded and actual values of the experimental design for Design 1. 81 Table 4.3 Coded and actual values of the experimental design for Design 2. 82 Table 4.4 Analysis of variance (ANOVA) results for response surface quadratic model terms for CaA degradation...... 89 Table 4.5 Regression diagnostics for the response surface quadratic model for CaA degradation...... 90 Table 4.6 Optimised conditions under specified constraints for the degradation of CaA and model verification...... 97 Table 4.7 Model verification of optimised conditions under randomly specified constraints for CaA degradation...... 98 Table 4.8 Model verification of optimised conditions in synthetic sugar solutions under specified constraints of selected various sugar process streams for CaA degradation...... 99 Table 4.9 Results of ANOVA for model terms of the response surface reduced quadratic model for CaA degradation...... 109

xxvii

Table 4.10 Results of ANOVA for model terms of the response surface reduced quadratic model for pCoA degradation...... 110 Table 4.11 Results of ANOVA for model terms of the response surface reduced quadratic model for FeA degradation...... 111 Table 4.12 Results of ANOVA for model terms of the response surface reduced quadratic model for total HCA degradation...... 112 Table 4.13 Wavenumbers (cm–1) of selected bands from ATR-FTIR spectra of CaA solution and CaA mixtures containing Fe(II) and/or sucrose at pH 5.5 and 25 °C...... 121 Table 4.14 X-ray diffraction data of the precipitate formed between CaA and Fe(II) at pH 5.5 and 25 °C...... 126 Table 4.15 Optimised conditions under specified constraints for the degradation of total HCA (200 mg/L) and model verification...... 128 Table 5.1 Reaction products formed from the Fenton oxidation of HCAs detected by LC/MS...... 146 Table 5.2 Contents of organic acids (mM) by HPIEC of individual and combined HCA mixtures...... 149 Table 5.3 Reaction products formed from the Fenton oxidation of HCAs detected by GC/MS...... 151 Table 5.4 Electron density distribution of carbon atoms in HCA molecules. 155 Table 6.1 Coded and actual values of the experimental design...... 182 Table 6.2 Results of ANOVA for model terms of the response surface reduced two-factor interaction model for melanoidin degradation. 187 Table 6.3 Results of ANOVA for model terms of the response surface reduced two-factor interaction model for total HCA degradation.. 188 Table 6.4 Results of ANOVA for model terms of the response surface reduced quadratic model for decolourisation...... 189 Table 6.5 Optimised conditions under specified constraints for the degradation of melanoidin (2,000 mg/L) and total HCA (200 mg/L) in sucrose solution (15% (w/w)) at 35 °C; and model verification...... 201

xxviii

Table 6.6 Optimised conditions under specified constraints for the decolourisation of synthetic juice mixtures containing melanoidin (2,000 mg/L), HCA (200 mg/L) and sucrose (15% (w/w)) at 35 °C; and model verification...... 202 Table 7.1 Operating parameters for ICP-OES analyses...... 211 Table 7.2 Clarification performance results on clarified No. 2 mill juices from the Tully Sugar Mill trials...... 212 Table 7.3 Inorganic ion composition results on clarified No. 2 mill juices from the Tully Sugar Mill trials...... 213 Table 7.4 Colour results on clarified No. 2 mill juices from the Tully Sugar Mill trials...... 214 Table 7.5 Clarification performance results on clarified factory juices from the Isis Central Sugar Mill trials...... 216 Table 7.6 Inorganic ion composition results clarified factory juices from the Isis Central Sugar Mill trials...... 217 Table 7.7 Purity and reducing sugar results on clarified factory juices from the Isis Central Sugar Mill trials...... 219 Table 7.8 Colour results on clarified factory juices from the Isis Central Sugar Mill trials...... 220 Table 7.9 Prices of additives in bulk quantities used for the modified Fenton process...... 222 Table A1.1 Experimental design and results for % CaA degradation (i.e., Design 1)……………………………………………………. 236 Table A1.2 Sucrose and reducing sugar results on selected tests at t = 2 min (i.e., Design 1)……………………………………………………. 241 Table A1.3 Experimental design and results for % CaA, % pCoA, % FeA and % total HCA degradation (i.e., Design 2)…………………… 243 Table A1.4 Sucrose and reducing sugar results at t = 2 min (i.e., Design 2)…. 245 Table A2.1 Geometry optimisation, charges and bond order computational calculations of CaA………………………………………………. 247 Table A2.2 Geometry optimisation, charges and bond order computational calculations of pCoA……………………………………………... 251

xxix

Table A2.3 Geometry optimisation, charges and bond order computational calculations of pCoA……………………………………………... 255 Table A2.4 Sucrose and reducing sugar results of Fenton-mediated reactions of sucrose at t = 2 min……………………………………………. 259 Table A3.1 Experimental design for % total HCA, % melanoidin degradation and decolourisation…….……………………………………….... 262 Table A3.2 Results for % total HCA, % melanoidin degradation and decolourisation………………………………………………….... 264

xxx

List of Abbreviations and Nomenclature

Acronyms

BSES Bureau of Sugar Experiment Stations CA California, USA CO Colorado, USA CT Connecticut, USA ICDD International Centre for Diffraction Data IL Illinois, USA MA Massachusetts, USA MD Maryland, USA MN Minnesota, USA MO Missouri, USA NSW New South Wales, Australia QLD Queensland, Australia QSL Queensland Sugar Limited QUT Queensland University of Technology SRI Sugar Research Institute UK United Kingdom UQ University of Queensland USA United States of America VIC Victoria, Australia WI Wisconsin, USA

Scientific Acronyms

2EJ Second Expressed Juice 2FI Two-Factor Interaction 3D Three-Dimensional 3EJ Third Expressed Juice 4EJ Fourth Expressed Juice

xxxi

ALS Automatic Liquid Sampler ANOVA Analysis of Variance AR Analytical Research AOP Advanced Oxidation Process ATR Attenuated Total Reflectance B3 Becke’s Three Parameters (Functional) CaA Caffeic Acid CCD Central Composite Design CV Coefficient of Variance DAD Diode-Array Detection DFT Density Functional Theory DH Degree of Hydrolysis DOE Design of Experiments DS Dry Substance EI Electron Impact ESI Electrospray Ionisation ESJ Evaporator Supply Juice FeA Ferulic Acid FEJ First Expressed Juice FES Final Evaporator Syrup FHLJ Flocculated Heated Limed Juice FLD Fluorescence Detection FTIR Fourier Transform Infrared GC Gas Chromatography HADP Hexose Alkaline Degradation Product HCA Hydroxycinnamic Acid HF Hartree-Fock HJ Heated Juice HMF Hydroxymethylfurfural HMW High Molecular Weight HPAEC High-Performance Anion Exchange Chromatography HPIEC High-Performance Ion Exclusion Chromatography HPLC High-Performance Liquid Chromatography

xxxii

ICUMSA International Commission for Uniform Methods of Sugar Analysis IJ Incubated Juice IV Indicator Value LC Liquid Chromatography LJ Limed Juice LMW Low Molecular Weight LYP Lee-Yang-Parr (Functional) MJ Mixed Juice MRP Maillard Reaction Product MRSM Multi-Response Surface Methodology MS Mass Spectrometry MS/MS Tandem Mass Spectrometry MW Molecular Weight PAD Pulse Amperometric Detection pCoA p–Coumaric Acid PJ Primary Juice PPO Polyphenol Oxidase PRESS Predicted Residual Sum of Squares Q Quadrupole RI Refractive Index RS Raw Sugar RSD Relative Standard Deviation RSM Response Surface Methodology SPE Solid Phase Extraction SS Sum of Squares TIC Total Ion Chromatography TOF Time-of-Flight TSS Total Soluble Solids UV Ultraviolet UV/Vis Ultraviolet/Visible VWD Variable Wavelength Detector XRD X-ray Powder Diffraction

xxxiii

Nomenclature

2° secondary 3° tertiary A absorbance df degrees of freedom

E1 early-time waveform

E2 intermediate-time waveform

E3 late-time waveform e exponential ε random error term i.d. internal diameter k number of factors L luminescence ln natural logarithm λ wavelength

λem emission wavelength

λex excitation wavelength m– prefix (meta–) for substituents on 1,3-positions of aromatic compounds m/z mass-to-charge ratio n number of experiments o– prefix (ortho–) for substituents on 1,2-positions of aromatic compounds p p-value for the probability of obtaining a test statistic p– prefix (para–) for substituents on 1,4-positions of aromatic compounds π– prefix (pi–) for representing covalent chemical bonds pKa acid dissociation constant R2 coefficient of determination T temperature t time tR retention time

xxxiv

Units

Å Ångström AU absorbance units °Bx degree Brix °C degree Celsius cc cubic centimetre cm centimetre Da dalton eV electron volt g gram h hour hp horsepower IU ICUMSA units K Kelvin kg kilogram kPa kilopascals kV kilovoltage kW kilowatt LU luminescence units M molarity mA milliampere mAU milliabsorbance units mg milligram min minute mL millilitre mM millimolarity mm millimetre MΩ megaohm μA microampere μL microlitre μm micrometre nm nanometre ppm parts per million

xxxv

psig pound-force per square inch gauge rpm revolutions per minute s second t tonne TU turbidity units V voltage % (v/v) volume/volume percent % (w/w) weight/weight percent

xxxvi

CHAPTER 1

General Introduction

1.1 Background and Motivation...... 2 1.2 Research Problem...... 3 1.3 Aims and Objectives...... 4 1.4 Scope of this Thesis...... 5

1.1 Background and Motivation

Sugar is an important commodity in world agricultural trade. It is mainly sucrose, a disaccharide made up of glucose and fructose, and is sourced from either sugar cane (Saccharum officinarum L.) or sugar beet (Beta vulgaris L.). Commonly referred to as table or granulated sugar, its use in the food and beverage industries is widespread. Sugar obtained from sugar cane contributes to 70% of the world’s sugar production. However, unlike sugar beet, which is primarily sold as white sugar, a plethora of sugar products are produced from sugar cane (e.g., raw sugar, syrup and molasses).

Australia produces approximately 4.5 million tonnes of raw sugar per year, of which 85% is exported (Canegrowers, 2012). This contributes around $A1.7 billion to Australia’s export earnings (QSL, 2011). The ongoing fluctuating international value of sugar continues to stress the viability of Australian sugar mills and threaten the reliance of regional communities on the industry. A consequence of the increased competition amongst sugar producers in world trade is an increased focus on the delivery of high quality sugar. In Australia, high quality raw sugar attracts a premium value of about $A7 per tonne of sugar (QSL, 2011).

One of the most important parameters in sugar quality is colour. Australian raw sugars are considered to be of high quality with respect to this parameter, but there is room for improvement. However, some raw sugars produced in both Australia and overseas are relatively difficult to decolourise by sugar refiners and tend to develop colour during storage.

The costs of sugar refining increase with the amount of colouring matter in the raw sugar feedstock. Therefore, given a choice, sugar refineries select low colour raw sugar from the markets at a price premium. Sugar refineries are not wanting to affine low quality raw sugars, as affination is expensive due to the high use of fossil fuels for energy to operate the fugals. A reduction of colour in raw sugar or a cheap and effective method of removal in processing would lead to lower refining costs.

1.2 Research Problem

Numerous technologies have been developed over the decades to achieve efficient and effective decolourisation of sugar cane process streams at a reasonable cost, in order to produce low coloured raw sugar. Apart from the crystallisation process, there are very few technologies and modifications to current processes that can significantly and economically reduce colour, except for the use of sulfur dioxide

(SO2) (Paton, 1992; Bento, 1999; Godshall, 1999). The use of SO2 via the sulfitation process for the production of plantation white sugar is discouraged in many countries because of the health risks surrounding the consumption of contaminated sugar containing residual sulfur (6–30 mg/kg) (Steindl and Doherty, 2005). The options that are in current use in Australian sugar factories for colour removal in raw sugar include double purging (i.e., washing) of sugar crystals and modification of crystallisation boiling schemes. These treatment options are not so effective with high molecular weight (HMW) colourants and require refining in order to obtain white sugar (Lindeman and O'Shea, 2001).

Different methods that have been trialled to treat sugar cane process streams for colour removal include clarification techniques (Eggleston et al., 2003), air flotation (Echeverri and Rein, 2006), membrane filtration (Farmani et al., 2008), chemical precipitation (Doherty et al., 2003), ion exchange resins (Broadhurst and Rein, 2003) and activated carbon adsorption (Mudoga et al., 2008). A major disadvantage that sugar manufacturers face is that most of the aforementioned processes are colour selective and are not effective in removing certain types of colourants. To overcome this, combinations of two or more processes are usually required to produce the best low colour raw sugar (Olivério et al., 2010). However, the option of using combinations of technologies is expensive and not viable for raw sugar manufacturers. In addition, sugar manufacturers have also reduced the amount of extraneous matter entering into the factory in an attempt to reduce the amount of colour in the final raw sugar product.

In the last decade or so, there has been an increasing trend towards the evaluation of chemical additives as alternatives to reduce or inhibit colour formation during sugar processing. These include the use of sulfurated and chlorinated compounds (Saska et al., 2010), ozone (Moodley et al., 1999), hydrogen peroxide

(H2O2) (Mane et al., 2000) and ferric ion (Fe(III)) in conjunction with endogenous proteins (Madsen and Day, 2010). Advanced oxidation processes (AOPs) are gaining focus as alternatives to conventional methods for the treatment of organic dyes (Koprivanac et al., 2005) and industrial wastewaters (Pera-Titus et al., 2004). However, these technologies have received limited attention to not only decolourise sugar process streams, but also remove impurities that may affect processing. An example of this is the activation of H2O2 using ferrous iron (Fe(II)), typically referred to as the Fenton oxidation process. It is an attractive process for its low capital costs, low toxicity of reagents and ease of application.

In this context, this present study builds on this line of research by examining the potential decolourisation and oxidative degradation of colourants and colour precursor compounds in water, synthetic juice solutions and sugar cane factory juices using the Fenton oxidation and related processes.

1.3 Aims and Objectives

The overarching aim of the project was to develop a cost-effective decolourisation process for effective removal of colourants and colour precursors from sugar cane process streams using the Fenton oxidation process.

The specific objectives of the project were to:

 Determine the colour and composition of phenolic acid compounds (i.e., colour precursors) present in different juice types.  Optimise process parameters and develop models for the removal of colour precursors (e.g., hydroxycinnamic acids) and synthetic colourants using the Fenton process.  Propose mechanisms for the degradation of hydroxycinnamic acids by the Fenton process.  Evaluate the decolourisation efficiency of the Fenton process on factory juice.

A preliminary cost benefit analysis was also conducted to assess the benefits of the developed Fenton technologies to remove colour during raw sugar manufacture.

1.4 Scope of this Thesis

There are limited reports in the literature with respect to the use of the Fenton process to treat sugar process streams. There are also gaps in the literature with regards to the degradation mechanisms of colour precursor compounds using the Fenton process. Therefore, this thesis examined the use of the Fenton oxidation process and variants of this process as potential technologies for the removal of colourants and colour precursors in aqueous and sugar solutions. This thesis is thus arranged in the following manner.

Chapter 1 provides the background and motivation for the work, the research problem and the specific objectives of the project.

Chapter 2 covers a comprehensive literature review on the types, origins and reactions of colourants in the processing of sugar cane; the formation of colour during processing of sugar cane to produce raw sugar; and discusses known and potential sugar decolourisation technologies.

Chapter 3 provides a study on the colour and phenolic acid composition of sugar cane juices processed in Australian sugar factories. Colour content does not only differ from region to region but also because of differences in cane variety, soil type, climate and harvesting methods. The standard method used for the determination of colour precursors is compared to a modified method developed in the project.

Chapter 4 presents an exhaustive and comprehensive analysis on the effects of the Fenton oxidation process on a selected group of colour precursor compounds (viz., caffeic, p–coumaric and ferulic acids). The use of experimental design coupled with regression modelling through multivariate statistics were used for the optimisation of the operating parameters.

The Fenton oxidation process is capable of mineralising organic compounds

(i.e., decomposition to carbon dioxide (CO2) and water (H2O)) through reactions involving free radicals. However, depending on the operating conditions, this may not imply complete mineralisation. Chapter 5 evaluates the oxidation products obtained from the Fenton process using several chromatographic and spectroscopic

techniques. Attempts were made to determine the degradation pathways of selected colour precursors.

Complex synthetic juice solutions involving more than one type of colourant group have been investigated. Mixtures containing a synthetically-made factory produced colourant (i.e., melanoidin) and hydroxycinnamic acids were degraded and decolourised using the Fenton and modified Fenton processes. The results from this work are presented in Chapter 6.

On the basis of the results obtained from Chapters 3 to 6, the developed technologies and their optimised constraints were then trialled on factory sugar cane process streams, as shown in Chapter 7. A selected number of juice streams from Australian sugar factories were tested under laboratory scale conditions. A preliminary financial analysis based on the indicated benefits and costs of additives on factory process streams was conducted and discussed in this chapter.

Chapter 8 summarises the overall findings of the works carried out throughout this project and provides recommendations for future work.

References

Bento, L. S. M. (1999). Study of colour formation during carbonation in cane sugar refining using GPC with a ELS detector. Proceedings of the AVH Association (pp. 23-27). Reims, France.

Broadhurst, H. A., & Rein, P. W. (2003). Modeling adsorption of cane-sugar solution colorant in packed-bed ion exchangers. AIChE Journal, 49(10), 2519-2532.

Canegrowers (2012). Canegrowers Australia Annual Report 2011/2012. Tingalpa, QLD, Australia: Harding Colour.

Doherty, W. O. S., Fellows, C. M., Gorijan, S., Senogles, E., & Cheung, W. H. (2003). Flocculation and sedimentation of cane sugar juice particles with cationic homo- and copolymers. Journal of Applied Polymer Science, 90(1), 316-325.

Echeverri, L. F., & Rein, P. W. (2006). Numerical study of the flow in air flotation syrup clarifiers. Proceedings of the South African Sugar Technologists' Association, 80, 378-390.

Eggleston, G., Monge, A., & Ogier, B. E. (2003). Sugarcane factory performance of cold, intermediate, and hot lime clarification processes. Journal of Food Processing and Preservation, 26, 433-454.

Farmani, B., Haddadekhodaparast, M. H., Hesari, J., & Aharizad, S. (2008). Determining optimum conditions for sugarcane juice refinement by pilot plant dead-end ceramic micro-filtration. Journal of Agriculture, Science and Technology, 10, 351-357.

Godshall, M. A. (1999). Removal of colorants and polysaccharides and the quality of white sugar. Proceedings of the AVH Association (pp. 28-35). Reims, France.

Koprivanac, N., Kušić, H., Vujević, D., Peternel, I., & Locke, B. R. (2005). Influence of iron on degradation of organic dyes in corona. Journal of Hazardous Materials, B117, 113-119.

Lindeman, P. F., & O'Shea, M. G. (2001). High molecualr weight (HMW) colorants and their impact on the refinability of raw sugar. A study of Australian and overseas raw sugars. Proceedings of the Australian Society of Sugar Cane Technologists, 23, 322-329.

Madsen, L. R., II, & Day, D. F. (2010). Iron mediated clarification and decolourisation of sugarcane juice. Proceedings of the International Society of Sugar Cane Technologists, 27, 1-13.

Mane, J. D., Phadnis, S. P., Jambhale, D. B., & Yewale, A. V. (2000). Mill scale evaluation of hydrogen peroxide as a processing aid: quality improvement in plantation white sugar. International Sugar Journal, 102(1222), 530-533.

Moodley, M., Davis, S. B., & Adendorff, M. W. (1999). Full scale decolourisation trials with ozone. International Sugar Journal, 101, 165-171.

Mudoga, H. L., Yucel, H., & Kincal, N. S. (2008). Decolorization of sugar syrups using commercial and sugar beet pulp based activated carbons. Bioresource Technology, 99, 3528-3533.

Olivério, J. L., Boscariol, F. C., Mantelatto, P. E., Ciambelli, J. R., Gabardo, H., & Oliveira, A. A. (2010). DRD–Dedini Refinado Direto (Dedini Direct Refined) improvements in refined and crystal white sugar production. Proceedings of the International Society of Sugar Cane Technologists, 27, 1-13.

Paton, N. H. (1992). The origin of colour in raw sugar. Proceedings of the Australian Society of Sugar Cane Technologists, 14, 8-17.

Pera-Titus, M., García-Molina, V., Baños, M. A., Giménez, J., & Esplugas, S. (2004). Degradation of chlorophenols by means of advanced oxidation processes: a general review. Applied Catalysis, B: Environmental, 47, 219-256.

QSL (2011). Key Facts | Queensland Sugar. Retrieved March 14, 2013, from http://www.qsl.com.au/about-qsl/key-facts

Saska, M., Zossi, S., & Liu, H. (2010). Colour behaviour in cane juice clarification by defecation, sulfitation and carbonation. Proceedings of the International Society of Sugar Cane Technologists, 27, 1-14.

Steindl, R. J., & Doherty, W. O. S. (2005). Syrup clarification for plantation white sugar to meet new quality standards. International Sugar Journal, 107(1282), 581-589.

CHAPTER 2

Literature Review

2.1 Introduction...... 10 2.2 Colourants in Sugar Process Streams...... 10 2.2.1 Naturally Occurring Colourants...... 12 2.2.2 Factory Produced Colourants...... 16 2.3 Reactivity of Colourants during Sugar Manufacturing...... 18 2.3.1 Enzymatic Browning...... 18 2.3.2 Non-enzymatic Oxidation...... 20 2.3.3 Maillard Reaction...... 21 2.3.4 Caramelisation...... 24 2.3.5 Hexose Alkaline Degradation...... 27 2.3.6 Conversion of Anthocyanins to Chalcones...... 27 2.3.7 Biochemical Precursors of Flavonoids...... 30 2.4 Colour in Sugar Process Streams...... 31 2.4.1 Effects of Temperature on Colour Formation...... 35 2.5 Sugar Decolourisation Technologies...... 37 2.5.1 Current Technologies...... 37 2.5.2 Decolourisation using Chemical Additives...... 38 2.5.3 Novel and Potential Technologies...... 42

2.1 Introduction

The development of colour during sugar processing is a common problem for the sugar manufacturing industry. Juices and syrups produced as a result of processing contain compounds that end up in the sugar crystal. This chapter presents an overview of the previous work on sugar colour and provides the essential background for the current research. A review of the literature on the properties of colourants, their behaviour during processing and evaluation of decolourisation technologies is described in this chapter. The review provides one understanding of the fundamental mechanisms of colour formation in sugar cane processing.

2.2 Colourants in Sugar Process Streams

A representation of the typical sugar manufacturing process in Australia is shown in Figure 2.1. Sugar cane is harvested and cut on a seasonal basis. Harvested sugar cane is transported in large containers or bins to the sugar mill. The cane is then shredded and crushed (i.e., milled) to extract the juice. The juice is incubated and limed to remove impurities (e.g., starch) that affect subsequent processes and minimise sucrose inversion. The limed juice is then boiled (≥ 100 °C) and flashed before a flocculant is added to enhance the bridging of impurity aggregates. The treated juice is then clarified to separate and remove flocculated impurities, fibre and soil. Clarified juice then passes to the evaporation stage, where water is removed to form syrup. In the crystallisation process, the syrup is seeded and the crystals grow in vacuum pans, followed by separation of crystals by centrifugation. The separated crystals are washed and then dried to produce raw sugar ready for export or transferred to a sugar refinery for the production of white sugar.

Cane Sugar Cane Milling Incubation Harvesting

Raw Sugar Crystallisation Evaporation Clarification

Exported to Consumers

Transferred to Refinery

Figure 2.1 Schematic flowchart of the sugar manufacturing process in Australia.

Colour in sugar process streams consist of a complex mixture of compounds. They are introduced naturally from the cane plant or produced during processing in the factory. The compounds formed have different molecular weights, chemical structures and properties as a result of degradation and polymerisation reactions caused by changes in process parameters such as pH and temperature. The colourants that are difficult to remove are mainly hydrophobic in nature and persist throughout the sugar manufacturing process occluding within the sugar crystals. Moreover, their behaviour and reactivity at various stages of the sugar manufacturing process is extremely complex. Therefore, it is important to understand the process parameters that contribute to the formation of colour in order to develop technologies suitable for the subsequent removal of colour during processing.

2.2.1 Naturally Occurring Colourants

Chlorophylls and Carotenoids

Sugar cane pigments are predominantly made up of chlorophylls, carotenoids (carotenes and xanthophylls) and flavonoids. These colourants are present in expressed juices after the milling of cane. Extraneous matter such as the tops and leaves of the sugar cane plant contribute significantly to colour in juice (Mersad et al., 2003). Colloidal in nature, chlorophylls and carotenoids are insoluble in water. Therefore, they do not contribute to the colouring of the final product as they are easily removed during clarification.

Flavonoids

Flavonoids are soluble and weakly acidic in nature and persist throughout the milling and refining processes. These compounds are essential for the growth of the sugar cane plant. However, their presence in processing significantly impacts on the colouring of raw sugar. Flavonoids contribute up to a third of the colouring in raw sugar according to Smith and Paton (1985). This amount can considerably rise with juices expressed from whole green cane crop that contain tops and leaves. The colouring of raw sugar from flavonoids is attributable to the occlusion of flavonoid glycosides in the sugar crystals during crystallisation. These naturally occurring compounds are divided into various subgroups such as flavones, flavanols and anthocyanins and only differ in the numbering and positioning of hydroxyl groups on the C6–C3–C6 flavonoid backbone structure. A summary of these structures is presented in Figure 2.2 (Ververidis et al., 2007).

Anthocyanidins (Flavylium ion) Flavones (2–phenylchromen–4–one) Flavonols (2–phenylchromen–4–one) 3' 3' 3' 4' 4' 4' 2' 2' 2' 8 8 8 O O O 7 5' 7 5' 7 5' 6' 6' 6' 6 6 6 3 3 3 OH 5 5 5 O O

3,5,7,3’,5’–OH: Cyanidin 5,7,4’–OH: Apigenin 5,7,4’–OH: Kaempferol 3,5,7,4’–OH, 3’–OCH3: Peonidin 5,7,3’,4’–OH: Luteolin 5,7,3’,4’–OH: Quercetin 3,5,7,4’–OH, 3’,5’–OCH3: Malvidin 5,7,4’–OH, 3’,5’–OCH3: Tricin 5,7,3’,4’,5’–OH: Myricetin

3–O–sugar: Anthocyanidin Glycosides 5,7 or 6,8–sugar: Flavones Glycosides 3,7–O–sugar: Flavonol Glycosides

Flavanols or Flavan-3-ols Isoflavones (3–phenylchromen–4–one) Flavanone (2,3–dihydro–2–phenylchromen–4–one) (2–phenyl–3,4–dihydro–2H–chromen–3–ol) 3' 8 3' O 4' 7 4' 2' 2' 8 2' 8 O O 7 5' 6 3' 7 5' 3 6' 5 6' O 4' 6 6' 6 OH 3 3 5 5' 5 O 7,4’–OH: Daidzein 5,7,4’,5’–OH: Catechin 5,7,4’–OH: Narnigenin 5,7,4’–OH: Genistein 5,7,3’,4’,5’–OH: Gallocatechin 5,7,3’–OH, 4’–OCH3: Hesperetin 7,4’–OH, 6–OCH3: Glycitein 3,5,7,4’,5’–OH: Taxifolin (also a flavanonol) 7–O–sugar: Isoflavone Glycosides 6,8–sugar: Flavanol Glycosides 7–O–sugar: Flavanone Glycosides

Figure 2.2 Flavonoid structures found in sugar process streams and products. Examples are given under the general chemical structures (Ververidis et al., 2007).

Phenolic Compounds

The term phenolic comprises a wide range of compounds which possess an aromatic ring with one or more hydroxyl groups. Their presence is widespread throughout the plant kingdom. Phenolics in nature can exist in their free and bound forms, as esters or glycosides (e.g., flavonoids). Mainly colourless, phenolics are endogenous to the cane and are introduced into the sugar process streams after the crushing of cane. Subsequently, these phenolics may participate in enzymatic, complexation or polymerisation reactions yielding coloured compounds which survive throughout the milling process. A summary of the phenolic compounds is described in Figure 2.3 (Harborne, 1989; Vermerris and Nicholson, 2006).

Basic Phenolics (C6) Hydroxybenzoic Acids (C6–C1) Phenylacetic Acids (C6–C2) Hydroxycinnamic Acids (C6–C3)

OH O OH O OH OH O 2 6 2 6 2 6 2 6 3 5 3 5 4 3 5 4 3 5 4 4 2–OH: Catechol 2–OH: Salicylic acid 4–OH: 4–Hydroxyphenylacetic acid 4–OH: p–Coumaric acid 3–OH: Resorcinol 3,4–OH: Protocatechuic acid 3,4–OH: 3,4–Dihydroxyphenylacetic acid 3,4–OH: Caffeic acid 3,5’–OH, 3’–OCH3: Phloroglucinol 3,4,5–OH: Gallic acid 4–OH, 3–OCH3: Homovanillic acid 4–OH, 3–OCH3: Ferulic acid

3' 8 Coumarins (C –C ) 2' 4' 6 3 Stillbenoids (C –C –C ) HO 6 6 2 6 6–OH: Umbelliferone 5 3,5–OH: Pinosylvin 6,7–OH: Aesculetin 5' 3,5,4’–OH: Resveratrol HO O O 6–OCH : Herniarin 3 6' 3,3’,4’–OH, 2–glc: Astringin 5 4 2

O 3 2 O Naphthoquinones (C6–C4) 4 5 Anthraquinones (C –C –C ) 3 6 2 6 5–OH: Juglone 3 6 1,8–OH: 1,8–Dihydroxyanthraquinone 4 2 7 1,3,8–OH, 6–OCH3: Emodin 5 O 1 8 O 3 4 5 4 O Xanthonoids (C6–C1–C6) 2 Chalconoids (C6–C3–C6) 3 6 3' 1–OH, 7–glu: Euxanthin 4' 5 3–OH, 5’–CH3: Methyl hydroxychalcone 1,3,6,7–OH, 2–glc: Mangiferin 2 7 6 3,6,8–OH, 2–OCH , 1,7–CH CH(CH ) : Mangostin 1 8 3 2 3 2 O 5' 6' O

Figure 2.3 Structures of phenolics found in sugar process streams and products. Examples are given under the general chemical structures (Harborne, 1989; Vermerris and Nicholson, 2006).

Polyphenolic Compounds

Polymers consisting of multiple phenolic units are termed polyphenols. The number of repeating phenolic units varies; hence each polymer has a different molecular weight and structure. The disambiguation of polyphenols is shown in Figure 2.4. The simplest polyphenols are dimers of the monomeric phenolic units such as ellagic acid (i.e., gallic acid dimer). Intermediate polyphenols consist of two or more dimers of monomeric phenolic units (e.g., ellagitannin (Ross et al., 2007)). The molecular weights and structures of simple and intermediate polyphenols can be determined. However, this is not possible for complex polyphenols (e.g., lignin) which consist of repetitive monomeric phenolic units resulting in a macromolecule with an extremely HMW. In most cases, the molecular structures of these products are undefined and only approximations can be given.

O HO O HO O HO OH HO OH O OH HO O

Ellagic acid Gallic acid Dimer containing 2 gallic acid units Phenolic acid monomer

HO OH

HO O OH HO HO OH O O OH HO O HO O O O O HO OH HO O O O O O O O HO OH O O HO OH HO HO OH OH Ellagitannin HO OH HO OH Polymer containing 8 gallic acid units

Figure 2.4 Polymerisation of monomeric gallic acid to polyphenols ellagic acid and ellagitannin (Ross et al., 2007).

Sugar cane polyphenols include lignins and tannins. Lignin is a complex macromolecule present in the cell wall of plants. The rigidity of plant stems is attributable to the presence of lignin with cellulose. Lignin comprises of three different phenolic units (viz., p–hydroxyphenyl, guaiacyl and syringyl); the proportions vary according to the type of cane plant and the extraction conditions (Alves et al., 2012). Tannins are polymeric products of phenolic compounds. They have the ability to form strongly coloured complexes with proteins to form stable, hydrophobic co-polymers.

Nitrogenous Compounds

The main group of nitrogen-containing compounds present in cane juice are amino acids and proteins. Proteins are complex HMW compounds made up of amino acids. The amount of proteins in juice is dependent on the cane variety, soil type and harvesting conditions. Moreover, the levels of proteins in juice are relatively lower in sugar cane than in sugar beet. Proteins are of different isoelectric points some of which are removed during clarification while the remainder persist in the later stages of processing. Proteins denature and degrade to individual amino acid units as a result of heat and changes in pH. Amino acids produced from protein denaturation coupled with those endogenous to the cane plant are not removed during clarification and can react with reducing sugars via the Maillard reaction to form HMW dark coloured compounds.

2.2.2 Factory Produced Colourants

Melanins

Polyphenolic products formed by the enzymatic oxidation of phenolic compounds during processing are called melanins. A typical structure is shown in Section 2.3.1. The enzymatic browning is catalysed by the polyphenol oxidase (PPO) enzyme responsible for the conversion of phenols into quinones (Bucheli and Robinson, 1994). The quinones can then bind to proteins to form coloured polymers or undergo condensation to form dark colourants.

Melanoidins

Melanoidins by definition are the coloured end products of the Maillard reaction between an amine (e.g., amino acid) and a carbonyl compound (e.g., reducing sugar). Also known as the non-enzymatic browning reaction, the reaction mechanisms are complex, consisting of repetitive condensation, dehydration and polymerisation reactions resulting in dark brown coloured substances (Rizzi, 1997). The coloured Maillard reaction products (MRPs) formed are of varying molecular

weights, which are dependent on temperature and reaction time. A description on how a melanoidin is formed is presented in Section 2.3.3.

Aroma Compounds

Aroma compounds are reaction intermediates formed as a result of sucrose degradation, sucrose fragmentation and amino acid degradation. Some of these products are similar to those obtained from Maillard and caramelisation reactions. Intermediate products are capable of further reacting amongst each other to yield volatile products such as pyrazines, imidazoles and thiophenes. These products act as precursors of melanoidins since they either possess amino nitrogen or carbonyl groups, initiating the Maillard reaction.

Caramels

Caramels are produced by the polymerisation of thermally degraded products of sucrose at high temperatures (Baunsgaard et al., 2001). The products contain mixtures of oligosaccharides, polysaccharides and coloured matter (Lindeman and O'Shea, 2001). These colloidal compounds formed have a tendency to remain on the outer surface of the sugar crystals, which affect the quality of the final raw sugar product. A description on the formation of caramels is shown in Section 2.3.4.

Hexose Alkaline Degradation Products (HADPs)

Alkaline degradation products of hexose sugars are coloured products formed as a result of the thermal decomposition of reducing sugars. The end products mainly consist of carboxylic acids, carbonyl compounds and lower molecular weight (LMW) polymers, which can lead to inversion of sucrose and further colour formation. The degradation rate and composition are heavily dependent on temperature, juice pH and the presence of divalent cations (Coca et al., 2004). The alkaline degradation rate of hexose sugars is much faster than under acidic conditions. Typical structures of HADPs are presented in Section 2.3.5.

2.3 Reactivity of Colourants during Sugar Manufacturing

2.3.1 Enzymatic Browning

Enzymatic browning is a colour forming reaction involving a phenolic and a nitrogenous compound, occurring prior to the heating of sugar cane juice to form melanins. The reaction is likely to take place after crushing and milling of sugar cane when the juice makes contact with atmospheric air. In general, the reaction involves an enzyme that acts as a catalyst to oxidise o–diphenolic substrates to o–benzoquinones (Li et al., 2008). The o–benzoquinone can further react with a phenolic compound or an amino acid to produce a highly dark coloured condensation product (i.e., melanins) (Kort, 1979; Riffer, 2000).

The presence of PPO catalyses two reactions: the production of a diphenol (Singleton, 1987) and the oxidation of the diphenol to an o–benzoquinone (Li et al., 2008). The first reaction is described in Scheme 2.1, where the monophenol is oxidised (1) to a diphenol (2). The following reaction (Scheme 2.2) involves the oxidation of the diphenol (2) to yield o–benzoquinone (3) and water.

OH OH +O OH

Phenol Catechol (Monophenol) (Diphenol)

(1) (2)

Scheme 2.1

OH O

OH O2 O

2 2 + 2H2O

Catechol oBenzoquinone (Diphenol) (Quinone)

(2) (3)

Scheme 2.2

In a separate example, tyrosine, an amino acid present in cane juice, readily participates in this reaction using the PPO catalyst described in Scheme 2.3 (Wiggins, 1953; Cleary, 1988). Tyrosine (4) undergoes oxidation to dihydroxyphenylalanine (5). Subsequent catalytic oxidation yields dopaquinone (6). Dopaquinone is converted to a leuco compound (7) and then oxidised to give dopachrome (8). Decarboxylation of (8) yields 5,6–dihydroxyindole (9). Further oxidation of (9) yields indole–5,6–quinone (10) and slower oxidation over time will eventually produce a melanin (11).

O O O +O HO +O O OH OH OH slow fast NH2 NH2 NH2 HO HO O

Tyrosine Dihydroxyphenylalanine Dopaquinone (4) (5) (6)

HO O O +O HO O + CO2 slow fast HO N O N OH HO N OH H H H

5,6-Dihydroxyindole Dopachrome (Leuco Compound) (9) (8) (7)

+O fast

CH O

H O N +O O slow O

CH N H

Indole-5,6-quinone (Melanin) (10) (11)

Scheme 2.3

2.3.2 Non-Enzymatic Oxidation

Non-enzymatic (i.e., chemical) oxidation can occur throughout the sugar manufacturing process via several reaction pathways with phenolic compounds. Phenolics that have two or more hydroxyl functional groups on the aromatic ring such as caffeic acid and its esters (hydroxycinnamic compounds), gallic (hydroxybenzoic compounds), catechin (flavanols) and malvidin (anthocyanidins) are considered to be vulnerable to oxidation and produce colour during the manufacturing process of sugar (Fernandez-Zurbano et al., 1995; Fernández-Zurbano et al., 1998; Kilmartin et al., 2001). The primary reaction pathway is through the oxidation of phenolics by reactive oxygen species, catalytically produced from O2 under the presence of transistion metal ions (viz., copper (Cu) and iron). As shown in Figure 2.5, the oxidised products are semiquinone radicals and benzoquinones, while the reduction product of O2, mediated by the redox cycle of Fe(II)/Fe(III) and Cu(I)/Cu(II), is H2O2 (Danilewicz et al., 2008). The quinones formed as a result of oxidation are unstable and due to their highly electrophilic nature, they can spontaneously react with other phenolics and amine compounds present in juice to produce coloured polymeric substances (Oliveira et al., 2011).

OH O

Fe(III) Cu(I) O2 R OH R O (Diphenol) (Benzoquinone) H2O2 O O

Fe(II) Cu(II) HO2 R OH R OH (Semiquinone Radical) (Semiquinone Radical) OH O

R OH R O (Diphenol) (Benzoquinone)

Figure 2.5 Redox chemistry of phenolics under copper and iron to produce colour forming products as proposed by Danilewicz et al. (2008)

Other reaction pathways that may occur during the manufacturing process of raw sugar include condensation reactions of phenolics with aldehydes (viz., acetaldehyde) and organic compounds with an aldehyde functional group (viz., glyoxylic acid). These condensation reactions are common in the wine industry and are mainly responsible for the colouring and flavouring of wines (Ferreira et al., 1997; Silva Ferreira et al., 2003). These reactions involve the protonation of an aldehyde to give a carbocation under acidic conditions, followed by the nucleophilic addition of the C8 position of the C6-C3 moiety of a flavonoid compound (cf. Figure 2.2) (Li et al., 2008). The intermediate produced is then protonated and can react with another phenolic compound of any type producing coloured polymers (Fulcrand et al., 2006).

2.3.3 Maillard Reaction

The Maillard reaction is a non-enzymatic browning process that involves the reaction of an amino compound with a reducing sugar to produce to a melanoidin. A common example of the Maillard reaction is described by Cleary (1988) between glucose and glycine. In this reaction, the formation of a Schiff’s base (or commonly known as an imine) occurs followed by an Amadori rearrangement to yield an enol.

Scheme 2.4 shows that initial nucleophilic addition occurs where the active lone pair of electrons on the amine nitrogen atom of glycine (13) attacks the electrophilic carbonyl carbon of glucose (12) to form a zwitterion (14) (Carey and Sundberg, 2007). The zwitterion is then converted to an unstable carbinolamine (15).

OH OH O O OH OH O HO H + OH HO H NH OH OH NH2 OH OH O

Glucose Glycine (Zwitterion) (12) (13) (14)

OH OH OH HO H N H OH OH O

(Unstable Carbinolamine) (15) Scheme 2.4

Nucleophilic addition with a base to (15) and removal of water yields an imine (16) (Scheme 2.5). The imine (16) undergoes an Amadori rearrangement where the hydrogen atom bonded to the carbon atom adjacent to the carbon-nitrogen double bond (C=N) relocates to bond with the nitrogen atom forming an enol (17), as shown in Scheme 2.6. The Amadori product could also participate in a keto-enol tautomerism rearrangement to its keto-form (18) (Scheme 2.7). It is also possible for these products to take part in further colour forming reactions (Belitz et al., 2009).

OH OH OH OH OH H + H2O HO H HO N N H - H2O OH OH OH OH O O

OH OH B (Unstable Carbinolamine) (Imine/Schiff's Base) (15) (16)

Scheme 2.5

OH OH H OH OH H H HO HO N NH OH OH O OH OH O

OH OH

(Imine/Schiff's Base) (Enol/Amadori Product) (16) (17)

Scheme 2.6

OH OH H OH OH HO HO NH NH OH OH O OH O O

OH OH

(Enol/Amadori Product) (17) (18)

Scheme 2.7

2.3.4 Caramelisation

Under the harsh operating conditions in sugar factories (e.g., elevated temperatures, acidic pH), caramelisation takes place. The chemistry of caramelisation is poorly understood due to the complexity of the reactions taken place. Various general mechanisms have been proposed in the literature and are shown in Scheme 2.8 (Riffer, 1988; Suárez-Pereira et al., 2010).

Heating sucrose syrup at elevated temperatures can form levoglucosan or decompose to glucose and fructose. These simple sugars form hydroxyl- methylfurfural (HMF) which is cleaved to yield one equivalent of formic acid and levulinic acid or react with other volatile compounds to yield melanoidins or polymeric colourants (Figures 2.6–2.8). Under increasing acidic conditions and subsequent losses of water (i.e., evaporation) formation of polymers with difructose dianhydride units can also occur (Madsen, 2006; Suárez-Pereira et al., 2010). Oxygen does not influence or contribute to further colouring of the caramel formed but could possibly affect the solubility of the caramel in water or acidic solution (Belitz et al., 2009).

Scheme 2.8a is an older proposed mechanism of the caramelisation reaction by Riffer (1988). Thermal degradation of sucrose yields levoglucosan. Dehydration of levoglucosan yields levoglucoseone. The precursor of HMF, 3,4–dideoxyglucosulose–3–ene can be obtained from levoglucosenone. (Daniels and Lohneis, 1997). The latter product can undergo cleavage to yield one equivalent of levulinic acid and formic acid or react with other volatiles to form melanoidins, coloured polymers and condensation products (Scheme 2.8d).

O OH O CH3 HO OH HO O H OH O H H O H H HO OH H Levulinic acid Formic acid O H OH

OH H H OH Sucrose Other O Volatiles (a) Heat O Melanoidins (cf., Figures 2.6-2.7) Heat OH O + HO H H /H O Condensation Products (cf., Figure 2.8) O (b) 2 HMF OH OH (d) - H2O Levoglucosan OH - H2O HO OH O H H O H OH H + H HO O HO OH H OH O H OH OH H Glucose Fructose O

Levoglucosenone (c) (e) (f) H HO OH OH Glucooligosaccharides H O + H2O O H H HO H OH H H HO OH

OH OH H OH H O Fructosyl Oxocarbenium Cation Fructopyranose

OH O

3,4-Dideoxyglucosulose-3-ene

- H2O CH3 O HO HO OH O O HO OH O Levulinic acid HO H Fructosyl Oxacarbenium Cation

HMF O

Other H OH Volatiles Formic acid O O

Melanoidins (cf., Figures 2.6-2.7) HO O Condensation Products (cf., Figure 2.8) HO

Fructose Disaccharides

O O

O O Difructose Dianhydride Compounds

Scheme 2.8

A more recent proposed mechanism of the caramelisation reaction is shown in Scheme 2.8b (Suárez-Pereira et al., 2010). Sucrose is hydrolysed and thermally degraded to glucose and fructose. Glucose is in equilibrium with the glucooligosaccharides (Scheme 2.8c). On the other hand, fructose may take part in several different reaction pathways.

In Scheme 2.8d, dehydration of fructose yields HMF which can then further participate in colour-forming reactions. In both Schemes 8e and 8f, fructose can form 5-membered or 6-membered fructosyl oxacarbenium cations and subsequently to fructose disaccharides and other difructose dianhydride compounds.

O glc HO O

H3C OR R RO N OH OR O glc glc O OH OH OH O

R = H, glc or (glc) n CH3

Figure 2.6 An example of a basic melanoidin structure formed from 3–deoxyhexosuloses (Cämmerer et al., 2002).

CH2OH

CH2OH CHOH COO CHOH CHOR

H OH H CHR1 CHOR H H CH2 1 1 1 + 1 R = H or sugars C C C N C C C C C N C C N R1 = Amino acid side chain CHOH H CH2 OH H CR1H CHOH CHOR COOH

CH2OH CHOH (e.g. , amide, ester)

CH2OH

Figure 2.7 An example of a melanoidin polymer formed from 3–deoxyhexosuloses and amino acids proposed by Cämmerer and Kroh (1995).

OH O O OH O Aldol Condensation O Aldol Condensation O O H O HO H H HO H3C H

Figure 2.8 Condensation product formed from the reaction of HMF and a ketone; followed by an additional condensation reaction with a second equivalent of HMF (Chheda and Dumesic, 2007).

2.3.5 Hexose Alkaline Degradation

Degradation of fructose and glucose under hot alkaline conditions show both reducing sugars undergo similar reaction pathways and form similar end products (Yang and Montgomery, 1996; Knill and Kennedy, 2003; Sinnott, 2007). In one study, over 50 products (including lactic acid, oxalic acid, saccharinic acids, short and long-chain carboxylic acids) were identified from the degradation of fructose and glucose in calcium hydroxide solutions (Yang and Montgomery, 1996).

The reaction mechanisms are often referred to as the Nef-Isbell-Richards mechanisms for the degradation of reducing sugars (Sinnott, 2007). There are at least six reaction steps in this mechanism:

Step 1: Keto-enol tautomerism of a reducing sugar (e.g., glucose, fructose).

Step 2: Enediol deprotonation.

Step 3: Anion isomerisation.

Step 4: Elimination of β–hydroxycarbonyl group.

Step 5: Keto-enol tautomerism.

Step 6: Benzilic acid rearrangement.

Scheme 2.9 illustrates the initial enolisation of fructose (or glucose) (Reaction Steps 1 and 2) to form a 1,2–enediolate. Elimination of the hydroxyl group (Reaction Step 4) yields a 1–aldehydo–3–deoxy–2,3–enol. The product undergoes further tautomerisation (Reaction Step 5) to yield an α–ketoaldehyde followed by addition of a hydroxide anion. The product then undergoes the final step, a benzilic acid rearrangement (Reaction Step 6) to yield metasaccharinic acid.

Alternatively, subsequent anion isomerisation of the 1,2–enediolate may also occur (Reaction Step 3). The two other isomers then participate in the same reactions (Reaction Steps 4–6) forming isosaccharinic and saccharinic acids respectively.

There are other numerous reactions that may take place under alkaline conditions such as the Lobry de Bruyn-van Ekenstein transformation, which is the interconversion between an aldose sugar and a ketose sugar (Hajek et al., 2013). The reaction steps involved in these reactions are similar to the aforementioned pathways.

- - O - OH O O O O H OH O O OH (Step 1, 2) (Step 4) OH (Step 5) O + OH O (Step 6) OH HO HO OH OH OH OH OH OH OH OH OH OH OH OH OH OH OH OH OH OH

Fructose Metsaccharinic (Step 3) Acid

OH O OH - OH OH O - - O (Step 4) O (Step 5) O + OH O (Step 6) OH O OH OH O O OH OH

OH OH OH OH OH OH OH OH OH OH

Isosaccharinic (Step 3) Acid

OH - CH2 CH3 OH O OH (Step 5) OH - (Step 6) (Step 4) OH O + H3C O O - O O O O H3C OH OH OH OH OH CH3 OH OH OH OH OH OH OH OH OH OH

Saccharinic Acid

Scheme 2.9

2.3.6 Conversion of Anthocyanins to Chalcones

When sugar cane juice is heated at pH 7.0, anthocyanins are decomposed to yield one equivalent of a chalconoid and a coumain-glucoside (Smith and Paton, 1985). An example of this is depicted in Scheme 2.10 as the glycoside of malvidin (19) is degraded by heat to its corresponding chalcone (20) and a coumarin-glucoside (21) is formed.

OCH3 OCH3 + OH O HO O HO HO O O OH Heat OH + O O O OCH3 OCH3 OH Sugar OH Sugar OH Sugar

Malvidin Chalcone Coumarin-Glucoside (19) (20) (21)

Scheme 2.10

2.3.7 Biochemical Precursors of Flavonoids

There is a relation between phenolic compounds and flavonoids. Cinnamic acid derivatives are biochemical precursors of flavonoids (Smith and Paton, 1985). For example, in Scheme 2.11, a tricin aglycone molecule (22) is decomposed to yield two products, phloroglucinol (23) and sinapic acid (24). Further oxidation of (24) yields syringic acid (25). Other phenolic acid derivatives such as caffeic and p–hydroxybenzoic acids are related to luteolin and apigenin aglycones respectively as they undergo similar reaction mechanisms.

OCH3

OCH3 HO O HO OH OH + HO OH OCH3

OH O OH O OCH3

Tricin Aglycone Phloroglucinol Sinapinic Acid (22) (23) (24)

OCH3 HO OH O

OCH3

Syringic Acid (25) Scheme 2.11

2.4 Colour in Sugar Process Streams

As reported in the previous section, the formation of colourants produced during factory processing is mainly due to sugar degradation reactions. Reducing sugars, such as glucose and fructose, formed by the inversion of sucrose, play an important role in the formation of colour. These sugars degrade due to changes in operating conditions such as pH and temperature to form highly reactive intermediates, which undergo condensation and polymerisation reactions to form highly coloured polymers. Colour precursors are of interest as they are not removed during juice clarification and polymerise to HMW coloured polymers and subsequently contribute to the colour in raw sugar (Lindeman and O'Shea, 2001).

A wide range of cane pigments and natural colourants are introduced into the manufacturing process as a result of milling and crushing of harvested cane. The cane plant primarily consists of LMW compounds that contribute approximately 30% of the colouring in raw sugar (Paton, 1992). The remaining 70% is attributable to colourants produced in the factory, mostly polymeric and of HMW with different chemical structures and properties. Lindeman and O’Shea (2001) reported that 50–60% of colourants by weight were of HMW and its contribution of these to the total colour in the final product, based on a standard spectrophotometric procedure at 420 nm, was approximately twice that of LMW colourants.

Generally in Australia and most other parts of the world, colour is measured at pH 7.0, however the colour at pH 7.0 is the least stable. Moreover, additional information about the nature of the colourants present can be obtained by taking measurements at pH 4.0 and pH 9.0. The classes of compounds attributable to colour in various process streams exhibit different colour sensitivity according to the pH of the aqueous media. For example, HMW colourants (e.g., caramels, melanoidins) are pH insensitive; therefore their colour does not change across pH 4.0–9.0. On the other hand, flavonoids and phenolic compounds (i.e., colour precursors) are highly pH sensitive. The colours of these compounds are lightly coloured at pH 4.0 but darken greatly at pH 9.0 (Smith et al., 1981; Paton, 1992). This is because at pH 9.0, the ionisation of these compounds is almost complete. Hence, these compounds are more highly coloured in their anionic form than in their neutral form. As the pH significantly affects the molecular structure and association-dissociation equilibria of

colourant types in sugar process streams, it is possible to determine the different mechanisms of colour formation taken place during processing by measuring the indicator value (IV) (Eggleston, 1998). The IV is the ratio of colour at pH 9.0 to that at pH 4.0 and reflects on the pH sensitivity of the colourants present in sugar process streams (Godhsall et al., 1991). For example, a decrease in IV value shows a higher presence of HMW colourants and may be attributable to the Maillard and/or caramelisation reactions taking place. It is also important to note that lower pH sensitive compounds (i.e., HMW compounds) will appear to be visually darker than higher pH sensitive compounds (i.e., LMW compounds). This is due to the higher absorption of the lower pH sensitive compounds over most of the visible region and can be avoided if colour is only measured spectrophotometrically at 420 nm (Riffer, 1988). The properties of colourants present in sugar cane juice are summarised in Table 2.1 (Davis, 2001a).

Table 2.1 Properties of colourants in cane juice (Davis, 2001a).

Monomeric Intermediate Polymeric Colourant type Flavonoids HADPs Caramels, Melanoidins MW (Da) Less than 1,000 1,000–2,500 Greater than 2,500 Ion Neutral at low Cationic at Cationic at pH pH 1.0–5.0 pH 1.0–5.0 Anionic at Anionic at pH 6.0–14 pH 6.0–14 Indicator value 5–40 3–4 1–2 pH sensitivity Sensitive Intermediate Insensitive Polarity Weak Intermediate High

Smith et al., (1981) further investigated the colour of the juice during the milling process. In this process, pre-cut sugar cane, also known as billets (ca. 30 cm), enters a sequence of mills where the sugar cane is crushed and the juice is extracted. Each mill has a certain number of rollers in which are processed in tandem. The juice that is extracted as of the crushing and milling of cane from the pinch of the first two

rollers (i.e., No. 1 mill) is termed first expressed juice (FEJ). Hence, the remaining juices extracted from the pinches of the subsequent pairs of rollers are named according to the order of expressed juice: second expressed juice (2EJ), third expressed juice (3EJ) and fourth expressed juice (4EJ). Table 2.2 shows an increase in colour was observed across the milling train. Lower brix content, higher temperatures and extensive decomposition of fibres in the final mills, increases the colour of the expressed juice (Curtin and Paton, 1980).

Table 2.2 Colour analyses of milled juice at pH 7.0 (Smith et al., 1981).

Expressed juice Colour at pH 7 (IU) FEJ 11,100 2EJ 33,000 3EJ 57,100 4EJ 90,800

Eggleston et al., (2003) investigated colour formation across the various stages of the sugar manufacturing process using three different liming techniques (viz., cold, intermediate and hot). The results are presented in Figure 2.9. Colour increases after the liming process due to the reactions of alkaline degradation of reducing sugars. Paton (1992) and Eggleston et al., (2003) agree that colour decreases during clarification due to the removal of colourants by the calcium phosphate flocs.

As the brix content increases during the evaporation stage, the juice colour increases. Several factors such as reaction time, juice pH and sugar concentration contribute to the increase in colour. Colour formation from the Maillard reaction is dominant at the earlier stages of evaporation followed by alkaline degradation reactions (Eggleston, 1998). Based on this colour profile across the sugar manufacturing stage (Figure 2.9), to produce low coloured sugar, colour removal strategies should be targeted at mixed juice (MJ) (i.e., combined juice from No. 1 and No. 2 mills) and/or juice during the evaporating stage. The colourants are partitioned between the sugar crystals and liquor in the crystallisation step; hence the significant decrease in colour of the liquor will result in low colour sugar. Table 2.3 shows the

extent of colour formed in various sugar process streams in a typical sugar mill and the partition of colour between liquor and sugar is 6:1 (Smith et al., 1981).

12 ) )

3 10 10 x 8

Colour (IU) ( Colour(IU) 4 Hot Intermediate 2 Cold 0 MJ HJ IJ LJ FHLJ ESJ FES RS Processing Stage of Sample

Figure 2.9 Formation of colour among three clarification processes; mixed juice (MJ), heated juice (HJ), incubated juice (IJ), limed juice (LJ), flocculated heated limed juice (FHLJ), evaporator supply juice (ESJ), final evaporator syrup (FES) and raw sugar (RS) (Eggleston et al., 2003).

Table 2.3 Comparison of colour at pH 7.0 and 9.0 from process streams of a typical sugar mill (Smith et al., 1981).

Processing stage Colour at pH 7.0 (IU) Colour at pH 9.0 (IU) FEJ 10,700 23,300 Mixed juice 19,200 37,100 Liquor 15,400 44,500 Magma 27,800 57,400 Massecuite (A-grade) 25,500 63,900 Raw sugar (A-grade) 2,600 5,700

2.4.1 Effects of Temperature on Colour Formation

Paton and McCowage (1987) investigated colour formation in factory and synthetic evaporator supply juice (ESJ) when heated (up to 100 °C) for 5 h. In one of their studies, the authors compared samples of factory ESJ and deaminated factory ESJ. The colour in the factory ESJ sample at pH 7.0 increased by about 20% of the original colour. The colour of the deaminated factory ESJ at pH 7.0 was similar to the aminated sample despite the latter having 25% less colour than mill ESJ prior to heating. The authors further investigated this result using a synthetic ESJ consisting of reducing sugars (equimolar quantities of glucose and fructose), four amino acids (alanine, aspartic acid, leucine and valine), sucrose and other organic materials (e.g., dimethyl formamide). The use of synthetic ESJ allowed better understanding on the mechanisms of colour formation. A summary of their study is as follows:

 The formation of colour in the model ESJ after 5 h was of the same order as the mill ESJ.  The model compounds appeared to have an induction period (0–1.5 h) and a rapid increase in colour formation was observed over time.  The contribution of caramel colourants was small with or without an amino acid.  Consistency in the behaviour of amino acid (i.e., did not affect colour greatly) in both synthetic and factory ESJ samples.  Lower levels of reducing sugars showed slower colour formation.  Lower pH retarded colour formation, however higher pH levels rapidly produced more colour.  Colour formation was prominent at 100 °C and negligible at 65 °C.  Higher brix content in synthetic ESJ samples resulted in an increase of melanoidins and a decrease in HADP.

Paton and McCowage (1987) concluded that temperature was the largest factor that contributed to the formation of colour in ESJ and the formation of colour in synthetic ESJ (with or without amino acids) was primarily due to HADPs. The rate of colour formation in synthetic ESJ was slower compared to factory ESJ and this may be due to the absence of intermediate products of HADPs, MRPs and other impurities (e.g., phenolic compounds, iron and copper) in the factory ESJ.

However, the authors only investigated colour formation for temperatures of 100 ºC and below. The range of interest for temperatures during clarification and evaporation in a typical sugar mill is considered to be between 65 ºC and 125 ºC. De Ambrosis (1964) studied the effect of clarified juices at temperatures above 100 °C on clarified juices. The juices were held in a closed stainless steel vessel (with a mild steel cap) and heated to the required temperatures.

Comparing the juice colour formation data from De Ambrosis’s (1964) work and those obtained from Paton and McCowage’s (1987), it can be seen that Paton and McCowage’s (1987) work showed reduced colour formation rates, while the opposite was observed in De Ambrosis’s (1964) work. The high colour formation rates in De Ambrosis’s (1964) work may be due to the vessel’s material of construction (e.g., iron) having a catalytic effect. Further laboratory analyses on the formation of colour from three different sugar mills were conducted by Wright and Jegaraj (1992). The analyses were designed to obtain additional experimental data on colour formation in juice at higher temperatures (80–125 °C) to complement De Ambrosis’s (1964) work. The authors examined the colouration rate as functions of temperature, time, juice (sucrose concentration) and pH, which may impact on the colouring of raw sugar during the clarification and evaporation stages. An Arrhenius expression for the rate of colour formation in sugar manufacture was proposed as described in Equation 2.1.

8502 (2.1)  ye0.8930 109 T   where y is the colour formation rate (% of the initial colour per min)

T is the absolute temperature (K).

Wright and Jegaraj (1992) concluded that the extent of colour in raw sugar can be minimised by reducing the residence time of process streams at high temperatures (i.e., lower residence times during clarification, evaporation and crystallisation processes), lowering the ESJ pH and reducing the content of nitrogenous compounds (i.e., proteins and amino acids) present in sugar cane plants.

2.5 Sugar Decolourisation Technologies

2.5.1 Current Technologies

Methods used to treat sugar process streams to reduce impurity and colour loadings prior to crystallisation include modified clarification techniques (Eggleston et al., 2003; Lindeman and O'Shea, 2004); dissolved air floatation (Smith et al., 2000; Echeverri and Rein, 2006); membrane filtration (Hamachi et al., 2003; Farmani et al., 2008); chemical precipitation (Moodley, 1993; Doherty et al., 2003); ion exchange (Broadhurst and Rein, 2003; Bento, 2004); activated carbon adsorption (Mudoga et al., 2008; Simaratanamongkol and Thiravetyan, 2010) and chemical oxidation via ozonolysis (Moodley et al., 1999). A summary of the effectiveness of a decolourisation technique mentioned according to the corresponding types of colourants, adapted from Davis (2001b) is shown in Table 2.4. The tick symbols () represent effective removal of the colourant and cross symbols () for poor colour removal. Some cells have been left blank due to insufficient data in the literature.

Table 2.4 Decolourisation processes on colourants types existing in juice as adapted from Davis (2001b).

acids

henolics

Flavonoids P Amino Melanoidins Caramels HADPs Defecation      Carbonatation      Sulfitation     Phosphatation      Filtration    Precipitation    Oxidation    Activated carbon      Ion exchange     

Conventional processes such as defecation, carbonatation, sulfitation and phosphatation performed during clarification are colour selective and are not so effective in removing certain types of colourants. To overcome this, combinations of two or more processes are usually required to produce the best low coloured raw sugar. Amongst these techniques to produce white or low coloured raw sugar, activated carbons or ion exchange resins are used. However, there are common problems with the use of adsorbents and resins such as fouling and exhaustion. These adsorbents and resins can be regenerated to minimise costs, but there are also problems associated with the management of waste produced from regeneration.

Therefore to further minimise costs, many sugar manufacturers use SO2 as a pretreatment step, thereby reducing the amount of adsorbent or resin used for decolourisation (Olivério et al., 2007). In some cases, SO2 is further used in syrup clarification to produce plantation white sugar (Kulkarni, 2010).

2.5.2 Decolourisation using Chemical Additives

Chemical additives in the form of oxidants, precipitants, coagulants and inhibitors have been used to assist in the colour removal of sugar process streams.

Arguably, SO2 is one of the best performing decolourising agents. The use of SO2 as a bleaching gas, during sulfitation for plantation white sugar production, is known to produce very low coloured sugar with a lustre appearance (Saska et al., 2010). It is widely used in less developed countries but discouraged in developed countries because of the residual sulfur contamination that is hazardous to human health. In addition, the low colour in these treated sugars are only temporary, with residual iron compounds, not removed during the sulfitation process, oxidise and colourise sugar crystals within a few months of storage (Kulkarni, 2010).

Organic polymers, such as polyacrylamides and polyamines, are commonly used for coagulation, flocculation and sedimentation processes during the clarification of juice (Thai et al., 2012). However, the amounts of polymeric material added are limited due to cost and the possible presence of toxic residual monomers at higher dosages (Bae et al., 2007); hence these are usually dosed at lower concentrations with an additional process for optimum colour removal (Moodley, 1993).

There has been an increasing trend towards the evaluation of oxidative chemical additives as an alternative to reduce or inhibit colour formation during sugar processing. These include the use of chlorinated compounds (Riffer, 1980), ozone

(Moodley et al., 1999), H2O2 (Mane et al., 2000) and Fe(III) in conjunction with endogenous proteins (Madsen and Day, 2010). The use of chlorinated compounds is not recommended because of toxicological concerns surrounding the production of unwanted by-products and residual sulfite/sulfate or chloride/chlorite present in the final raw sugar product (Davis, 2001a). On the other hand, oxidative decolourants such as ozone and H2O2 are non-toxic and have shown good decolourisation on sugar factory process streams.

The reason for the difference between the action of oxidative chemicals and other colour removal technologies lies in the decolourisation mechanism. Oxidants destroy colourants rapidly by cleaving unsaturated bonds (i.e., conjugated species), converting them to non-reactive intermediates which are unable to form colour (Shore et al., 1984; Riffer, 2000). Examples of these are shown in Schemes 12 and 13 for ozone and wet peroxide oxidation, respectively (Davis et al., 1998; Neyens and Baeyens, 2003).

In Scheme 2.12, electrophilic ozone reacts with a nucleophilic alkene (26) to form an unstable 5-membered ring (i.e., ozonide) (27). The unstable ozonide decomposes to a carbonyl compound and a zwitterion. Cycloaddition of the two decomposition products form a stable ozonoide intermediate (28). The ozonoide (28) is then oxidised in the presence of H2O2 to yield one equivalent of a carboxylic acid (29) and an aldehyde (30) or ketone.

O H O O O R O O O O O R H R H O R H O H R R O H R H R H (Alkene) (Unstable Ozonide) (Stable Ozonide)

(26) (27) (28)

H2O2

O O + R OH R H

(Carboxylic Acid) (Aldehyde) (29) (30)

Scheme 2.12

Highly reactive •OH radicals produced from the decomposition of H2O2 are mainly responsible for the oxidation of colourants. Under mild alkaline conditions, – H2O2 dissociates to give water and the perhydroxyl anion (HOO ), a strong – nucleophile (Equation 2.2). The HOO anion can decompose H2O2 to give water, oxygen and the •OH radical (Equation 2.3).

– – H2O2 + OH  H2O + HOO (2.2) – HOO + H2O2  H2O + O2 + •OH (2.3)

In Scheme 2.13, hydroxyl radicals generated from the decomposition of H2O2 attach to the aromatic ring of benzene (31) causing the ring to open and yield muconic acid (32). The product (32) can possibly undergo further oxidative degradation to produce harmless reaction products such as CO2 and water.

O H OH H OH OH OH OH OH OH OH

O Benzene Muconoic acid

(31) (32) Scheme 2.13

Studies showing the destruction of flavonoids and colour precursors in sugar cane juice using H2O2 as the oxidant have been conducted previously (Mane et al., 1992; Mane et al., 1998; Mane et al., 2000; Saska, 2007). Saska (2007) reported up to 30% reduction in colour by using H2O2 on Colombian plantation white sugar at dosages of 100–500 ppm, which is also in agreement with the findings from Mane et al., (2000). Also, in their most recent study, Mane et al., (2000) reported more than

20% reduction in both colour precursor compounds and SO2 content in Indian plantation white sugars. They observed a decrease in phenolic content (40–50%) in

H2O2 treated raw syrup, therefore minimising the chance for these compounds to take part in further colour forming reactions. There were also reductions in raw sugar colour (12–35%), amino acids (15–25%) and starch (12–13%). Furthermore, the authors reported lower colour development upon stored raw sugars treated directly with H2O2.

Ozone, on the other hand, is a stronger oxidant than H2O2. A decrease by about one third of the initial syrup colour was achieved with 250 ppm ozone (Davis et al., 1998). However, unlike H2O2, ozone is very expensive to produce and is not cost effective for juice or syrup decolourisation (Moodley et al., 1999). Therefore, a technology based on H2O2 has the potential to produce low coloured raw sugar at a reasonable cost.

2.5.3 Novel and Potential Technologies

Advanced Oxidation Processes (AOPs)

In recent years, AOPs have become increasingly attractive to treat a wide range of azo dyes (Joseph et al., 2000), contaminated soils (Kong et al., 1998), polluted oceans and streams (Trovó et al., 2009) and industrial wastewaters (Lucas and Peres, 2009). These processes involve the in situ generation of highly reactive •OH radicals by chemical (e.g., acids, inorganic salts), photocatalytic (e.g., solar, ultraviolet (UV) light), electrochemical (e.g., cathode electrodes), radiolytic (γ–radiolysis) and physical (e.g., ultrasound) methods. The oxidation potential of

•OH radicals is stronger (2.80 V) than ozone (2.07 V) and H2O2 (1.80 V) and can completely degrade and mineralise organic compounds and impurities.

Fenton Oxidation Process

An example of an established and commercialised AOP is the catalytic activation of H2O2 using Fe(II), typically referred to as the Fenton oxidation process. The conventional homogenous Fenton oxidation process is already in use for the treatment of industrial wastewaters (Guedes et al., 2003; Cañizares et al., 2007; Lu et al., 2009). It is an attractive process for its low capital costs, low toxicity of reagents and ease of application. The Fenton process involves the production of •OH radicals through the homogenous catalytic decomposition of H2O2 using Fe(II). The generally accepted free radical chain mechanism for the oxidation of organic compounds (RH) via the Fenton process is shown in Equations 2.4–2.10 (Walling, 1975; Pignatello, 1992; Kang and Hwang, 2000; Sun et al., 2007).

2+ 3+ – Fe + H2O2  Fe + OH + •OH (2.4)

RH + •OH  R• + H2O (2.5)

H2O2 + •OH  H2O + •O2H (2.6) Fe2+ + •OH  Fe3+ + OH– (2.7)

•OH + •OH  H2O2 (2.8) 3+ 2+ + H2O2 + Fe  Fe + H + •O2H (2.9) 3+ 2+ + Fe + HO2•  Fe + H O2 (2.10)

As depicted in Equation 2.5, a radical chain oxidation reaction is initiated through the formation of organic radicals (R•) by hydrogen atom abstraction, electron transfer or electrophilic addition (Neyens and Baeyens, 2003; Pignatello et al., 2006). These organic radicals are highly reactive, which form peroxyl radicals (Equation 2.11) (Lipczynska-Kochany et al., 1995), and further oxidation through the addition of •OH or molecular oxygen, in turn would oxidise and mineralise to harmless products such as CO2 and H2O (Equation 2.12) (Sun et al., 2009; Oturan et al., 2011).

R• + O2  ROO• (2.11)

ROO• + •OH/O2  CO2 + H2O (2.12)

The two important factors to consider in the Fenton process are the dosage levels of H2O2 and Fe(II) (Chamarro et al., 2001). The H2O2 dose improves the decolourisation whilst the reaction kinetics is dependent on the amount of Fe(II) added.

In the last few years, much attention has been paid to the variations and development of advanced Fenton processes to improve the oxidation performance and alleviate one of the major drawbacks of the Fenton process, which is the production of iron sludge. These include photo-Fenton (e.g., solar and UV light) (Kuo et al., 2012; Lucas et al., 2012), electro-Fenton (Wang et al., 2012), sono-Fenton (Babuponnusami and Muthukumar, 2011), Fenton-like (e.g., Fe(III), chelated iron) (Li et al., 2007; Nichela et al., 2010) and heterogeneous Fenton (e.g., Fe-pillared clays, zero valent iron) (Catrinescu et al., 2012; Segura et al., 2012). However, most of these technologies have not yet been commercialised. Therefore, the conventional Fenton process, which is simple and requires no specialised equipment, is still the only cost effective process to treat a wide range of compounds and convert them into less

harmful compounds that are easier to be removed through other purification techniques (i.e., filtration, coagulation, ion exchange) (Üstün et al., 2007; Arsene et al., 2011; Elías-Maxil et al., 2011).

In previous studies, Fenton oxidation has been employed to target individual model phenolic acids in synthetic industrial process streams. Rivas et al. (2001) evaluated the degradation of p–hydroxybenzoic acid (10 mM) and found under optimum conditions of 5.0 mM Fe(II), 2.7 M H2O2 and pH 3.2; 95% of the phenolic acid was degraded after 30 min at 20 °C. In a later study, Rivas et al. (2005) reported 98% removal of protocatechuic acid (0.65 mM), under similar operating conditions to those of p–hydroxybenzoic acid. Benitez et al. (2005) reported 79% degradation of –2 gallic acid (0.59 mM) after 40 min at 25 °C using 2.5 × 10 mM Fe(II), 2.5 mM H2O2 and at pH 3.0.

Even though, the Fenton process or AOPs in general are non-selective processes, the degradation efficiency of phenolic compounds via •OH radical attack differs from one type to another, as it depends on many factors including number of substituents (e.g., hydroxyl and methoxyl groups) attached on the aromatic ring; •OH radical positioning and bonding sites on the aromatic ring; and preference for •OH radical attack on additional functional groups (e.g., vinyl groups) than the aromatic ring (Rice-Evans et al., 1996; Sroka and Cisowski, 2003). Also, the degradation of phenolic compounds in mixtures is expected to be different due to competing reactions between the phenolics, the •OH radicals and the intermediates formed during the course of the oxidation reaction.

The degradation of mixtures of phenolic compounds has been studied using Fenton oxidation (Heredia et al., 2001), Fenton-like oxidation (Du et al., 2006), ozone (Amat et al., 2003) and other AOPs, mainly photocatalysis processes (Gernjak et al., 2003; Kusvuran et al., 2005; Azabou et al., 2007; Monteagudo et al., 2011). Heredia et al. (2001) developed a kinetic model for the oxidation of phenolic compounds (viz., caffeic, p–coumaric and ferulic acids) by the Fenton process. These compounds are the main colour precursors present in sugar cane juice and are known to participate in reactions producing colour that results in the final raw sugar product. The rate constants for the degradation of the individual phenolic acids in a mixture of acids, were deduced from the developed model and it was found that the degradation

process at a constant Fe(II) concentration at 30 °C proceeded in the following order; ferulic acid > p–coumaric acid > caffeic acid. No reason was given for the differences in the rate of degradation among these phenolic acid mixtures. The ozonisation of solutions containing a mixture of cinnamic, caffeic, p–coumaric and ferulic acids was studied by Amat et al. (2003). These workers found that the behaviour of caffeic acid though followed a similar mechanism as that of cinnamic acid had a different reaction rate due to a greater number of intermediates. None of these studies optimised the degradation process of the individual acids within a mixture of phenolic acids by the Fenton process, nor examined the interactive effects of various operating parameters on the degradation of each acid. Also, none of the aforementioned studies were conducted in sucrose solutions, and the reaction times were generally an order of magnitude higher than that required in the various stages of the sugar manufacturing process. The role of sucrose (apart from its free radical scavenging ability) in the degradation process of these acids in a mixture has not been reported (Morelli et al., 2003). As such, the focus in this present study was to provide further insight into the degradation of these phenolic acids by the Fenton process.

The decolourisation of a baker’s yeast waste product, which primarily consists of cane molasses, using Fenton oxidation was investigated by Pala and Erden (2005). The colourants present in molasses include caramels, melanoidins, colour precursors, iron-phenol complexes and some cane pigments. Neyens and Baeyens (2003) stated that acidic pH levels (about pH 3.0) are usually optimum for Fenton oxidation. However, Pala and Erden (2005) established the optimum pH for colour removal via the Fenton process was at pH 4.0. The best colour removal efficiency of 99% at

25 °C and pH 4.0 was achieved with dosages of 22 mM Fe(II) and 24 mM H2O2 for 20 min. Under the optimum pH and treatment times, dosages of 11 mM Fe(II) and

18 mM H2O2 were enough to remove colour with an efficiency of 97%.

Madsen and Day (2010), though not using the Fenton oxidation process, demonstrated the removal of phenolic and other colourants from raw juice using endogenous proteins as well as Fe(III) as an oxidative catalyst. The treatment produced clarified juice with up to 70% lower colour in cold liming clarification (i.e., addition of lime before juice incubation) than juice produced by hot liming clarification. However, clarification via cold liming results in less precipitation of calcium phosphate precipitates and impurities due to the higher solubility of calcium

ions at lower temperatures (35–40 °C) (Doherty et al., 2002). This would result in more turbid clarified juices and higher sucrose losses. Hence, the presence of high calcium levels in clarified juice will heavily impact on the downstream processes, particularly during evaporation where the formation of scale takes place in the evaporators.

On the basis of the information obtained from the literature, the aim of this present project was to develop, optimise and evaluate the Fenton oxidation process for the degradation and decolourisation of selected colour precursors and colourants present in sugar process streams.

References

Alves, E. F., Bose, S. K., Francis, R. C., & El Moussaouiti, M. (2012). Trapping of p- coumaryl and coniferyl alcohol during soda–anthraquinone treatment: A means of estimating uncondensed β-O-4 structures in native lignin. Journal of Agricultural and Food Chemistry, 60(36), 9202-9210.

Amat, A. M., Arques, A., Beneyto, H., García, A., Miranda, M. A., & Seguí, S. (2003). Ozonisation coupled with biological degradation for treatment of phenolic pollutants: A mechanistically based study. Chemosphere, 53, 79-86.

Arsene, D., Musteret, C. P., Catrinescu, C., Apopei, P., Barjoveanu, G., & Teodosiu, C. (2011). Combined oxidation and ultrafiltration processes for the removal of priority organic pollutants from wastewaters. Environmental Engineering and Management Journal, 10(12), 1967-1976.

Azabou, S., Najjar, W., Gargoubi, A., Ghorbel, A., & Sayadi, S. (2007). Catalytic wet peroxide photo-oxidation of phenolic olive oil mill wastewater contaminants Part II. Degradation and detoxification of low-molecular mass phenolic compounds in model and real effluent. Applied Catalysis, B: Environmental, 77, 166-174.

Babuponnusami, A., & Muthukumar, K. (2011). Degradation of phenol in aqueous solution by Fenton, sono-Fenton and sono-photo-Fenton methods. Clean: Soil, Air, Water, 39(2), 142-147.

Bae, Y.-H., Kim, H.-J., Lee, E.-J., Sung, N.-C., Lee, S.-S., & Kim, Y.-H. (2007). Potable water treatment by polyacrylamide base flocculants, coupled with an inorganic coagulant. Environmental Engineering Research, 12(1), 21-29.

Baunsgaard, D., Nørgaard, L., & Godshall, M. A. (2001). Specific screening for color precursors and colorants in beet and cane sugar liquors in relation to model colorants using spectrofluorometry evaluated by HPLC and multiway data analysis. Journal of Agricultural and Food Chemistry, 49, 1687-1694.

Belitz, H. D., Grosch, W., & Schieberle, P. (2009). Food Chemistry. Berlin: Springer- Verlag.

Benitez, F. J., Real, F. J., Acero, J. L., Leal, A. I., & Garcia, C. (2005). Gallic acid degradation in aqueous solutions by UV/H2O2 treatment, Fenton’s reagent and the photo-Fenton system. Journal of Hazardous Materials, B126, 31-39.

Bento, L. S. M. (2004). Decolorization of sugar solutions with oxidants and ion exchange resins. Proceedings of the Sugar Industry Technologists, 63, 265- 269.

Broadhurst, H. A., & Rein, P. W. (2003). Modeling adsorption of cane-sugar solution colorant in packed-bed ion exchangers. AIChE Journal, 49(10), 2519-2532.

Bucheli, C. S., & Robinson, S. P. (1994). Contribution of enzymic browning to color in sugarcane juice. Journal of Agricultural and Food Chemistry, 42(2), 257- 261.

Cämmerer, B., Jalyschkov, V., & Kroh, L. W. (2002). Carbohydrate structures as part of the melanoidin skeleton. International Congress Series, 1245, 269-273.

Cämmerer, B., & Kroh, L. W. (1995). Investigation of the influence of reaction conditions on the elementary composition of melanoidins. Food Chemistry, 53, 55-59.

Cañizares, P., Lobato, J., Paz, R., Rodrigo, M. A., & Sáez, C. (2007). Advanced oxidation processes for the treatment of olive-oil mills wastewater. Chemosphere, 67(4), 832-838.

Carey, F. A., & Sundberg, R. J. (2007). Advanced Organic Chemistry - Part A: Structure and Mechanisms (5th ed.). New York: Springer.

Catrinescu, C., Arsene, D., Apopei, P., & Teodosiu, C. (2012). Degradation of 4- chlorophenol from wastewater through heterogeneous Fenton and photo- Fenton process, catalyzed by Al–Fe PILC. Applied Clay Science, 58(0), 96- 101.

Chamarro, E., Marco, A., & Esplugas, S. (2001). Use of Fenton reagent to improve organic chemical biodegradability. Water Research, 35(4), 1047-1051.

Chheda, J. N., & Dumesic, J. A. (2007). An overview of dehydration, aldol- condensation and hydrogenation processes for production of liquid alkanes from biomass-derived carbohydrates. Catalysis Today, 123, 59-70.

Cleary, M. (1988). Carbonatation Process in Beet Sugar Manufacture. In M. A. Clarke & M. A. Godshall (Eds.), Chemistry and Processing of Sugarbeet and Sugarcane (pp. 20-34). Amsterdam: Elsevier.

Coca, M., Teresa García, M., González, G., Peña, M., & García, J. A. (2004). Study of coloured components formed in sugar beet processing. Food Chemistry, 86, 421-433.

Curtin, J. H., & Paton, N. H. (1980). The quantitative analysis of phenolic acids from sugar liquors by high performance liquid chromatography. Proceedings of the International Society of Sugar Cane Technologists, 17, 2361-2371.

Daniels, V., & Lohneis, G. (1997). Deterioration of sugar artifacts. Studies in Conservation, 42(1), 17-26.

Danilewicz, J. C., Seccombe, J. T., & Whelan, J. (2008). Mechanism of interaction of polyphenols, oxygen, and sulfur dioxide in model wine and wine. American Journal of Enology and Viticulture, 59(2), 128-136.

Davis, S. B. (2001a). The chemistry of colour removal: a processing perspective. Proceedings of the South African Sugar Technologists' Association, 75, 328- 336.

Davis, S. B. (2001b). A review of colour removal mechanisms in sugar refining processes. Communications from the Sugar Millling Research Institute, 169, 1-25.

Davis, S. B., Moodley, M., Singh, I., & Adendorff, M. W. (1998). The use of ozone for colour removal at the Malelane refinery. Proceedings of the South African Sugar Technologists' Association, 72, 255-260.

De Ambrosis, J. L. (1964). The effect of temperature and time on colour formation in clarified juice, Unpublished Sugar Research Institute Internal Report.

Doherty, W. O. S., Greenwood, J., Pilaski, D., & Wright, P. G. (2002). The effect of liming conditions in juice clarification. Proceedings of the Australian Society of Sugar Cane Technologists, 24, 443-451.

Du, Y., Zhou, M., & Lei, L. (2006). Role of the intermediates in the degradation of phenolic compounds by Fenton-like process. Journal of Hazardous Materials, 136(3), 859-865.

Echeverri, L. F., & Rein, P. W. (2006). Numerical study of the flow in air flotation syrup clarifiers. Proceedings of the South African Sugar Technologists' Association, 80, 378-390.

Eggleston, G. (1998). Changes in chemistry across evaporators. International Sugar Journal, 100, 49-95.

Eggleston, G., Monge, A., & Ogier, B. E. (2003). Sugarcane factory performance of cold, intermediate, and hot lime clarification processes. Journal of Food Processing and Preservation, 26, 433-454.

Elías-Maxil, J. A., Rigas, F., Orta de Velásquez, M. T., & Ramírez-Zamora, R.-M. (2011). Optimization of Fenton's reagent coupled to Dissolved Air Flotation to remove cyanobacterial odorous metabolites and suspended solids from raw surface water. Water Science and Technology, 64(8), 1668-1674.

Fernández-Zurbano, P., Ferreira, V., Escudero, A., & Cacho, J. (1998). Role of hydroxycinnamic acids and flavanols in the oxidation and browning of white wines. Journal of Agricultural and Food Chemistry, 46(12), 4937-4944.

Fernandez-Zurbano, P., Ferreira, V., Pena, C., Escudero, A., Serrano, F., & Cacho, J. (1995). Prediction of oxidative browning in white wines as a function of their

chemical composition. Journal of Agricultural and Food Chemistry, 43(11), 2813-2817.

Ferreira, V., Escudero, A., Fernández, P., & Cacho, J. F. (1997). Changes in the profile of volatile compounds in wines stored under oxygen and their relationship with the browning process. Zeitschrift für Lebensmitteluntersuchung und -Forschung A, 205(5), 392-396.

Fulcrand, H., Dueñas, M., Salas, E., & Cheynier, V. (2006). Phenolic reactions during winemaking and aging. American Journal of Enology and Viticulture, 57(3), 289-297.

Gernjak, W., Krutzler, T., Glaser, A., Malato, S., Caceres, J., Bauer, R., & Fernández- Alba, A. R. (2003). Photo-Fenton treatment of water containing natural phenolic pollutants. Chemosphere, 50, 71-78.

Godhsall, M. A., Clarke, M. A., Dooley, C. D., & Blanco, R. S. (1991). Progress in Beet Sugar Colorant Research. Journal of Sugar Beet Research, 28(3/4), 155- 165.

Guedes, A. M. F. M., Madeira, L. M., Boaventura, R. A. R., & Costa, C. A. V. (2003). Fenton oxidation of cork cooking wastewater—overall kinetic analysis. Water Research, 37, 3061-3069.

Hajek, J., Murzin, D. Y., Salmi, T., & Mikkola, J.-P. (2013). Interconversion of lactose to lactulose in alkaline environment: Comparison of different catalysis concepts. Topics in Catalysis, 56(9-10), 839-845.

Hamachi, M., Gupta, B. B., & Ben Aim, R. (2003). Ultrafiltration: a means for decolorization of cane sugar solution. Separation and Purification Technology, 30, 229-239.

Harborne, J. B. (1989). Methods in Plant Biochemistry, 1: Plant Phenolics. London: Academic Press.

Heredia, J. B. D., Torregrosa, J., Dominguez, J. R., & Peres, J. A. (2001). Kinetic model for phenolic compound oxidation by Fenton's reagent. Chemosphere, 45, 85-90.

Joseph, J. M., Destaillats, H., Hung, H.-M., & Hoffman, M. R. (2000). The sonochemical degradation of azobenzene and related azo dyes: rate enhancements via Fenton's reactions. Journal of Physical Chemistry A, 104(2), 301-307.

Kang, Y. W., & Hwang, K.-Y. (2000). Effects of reaction conditions on the oxidation efficiency in the Fenton process. Water Research, 34(10), 2786-2790.

Kilmartin, P. A., Zou, H., & Waterhouse, A. L. (2001). A cyclic voltammetry method suitable for characterizing antioxidant properties of wine and wine phenolics. Journal of Agricultural and Food Chemistry, 49(4), 1957-1965.

Knill, C. J., & Kennedy, J. F. (2003). Degradation of cellulose under alkaline conditions. Carbohydrate Polymers, 51, 281-300.

Kong, S.-H., Watts, R. J., & Choi, J.-H. (1998). Treatment of petroleum-contaminated soils using iron mineral catalyzed hydrogen peroxide. Chemosphere, 37(8), 1473-1482.

Kort, M. J. (1979). Colour in the Sugar Industry. In G. G. Birch & K. J. Parker (Eds.), Sugar: Science and Technology (pp. 97-130). London: Applied Science Publishers Ltd.

Kulkarni, V. M. (2010). Modified single sulfitation process for producing better quality plantation white sugar. Proceedings of the International Society of Sugar Cane Technologists, 27, 1-8.

Kuo, C. Y., Pai, C. Y., Wu, C. H., & Jian, M. Y. (2012). Effects of oxidant concentration and temperature on decolorization of azo dye: comparisons of UV/Fenton and UV/Fenton-like systems. Water Science and Technology, 65(11), 1970-1974.

Kusvuran, E., Samil, A., Atanur, O. M., & Erbatur, O. (2005). Photocatalytic degradation kinetics of di- and tri-substituted phenolic compounds in aqueous solution by TiO₂/UV. Applied Catalysis, B: Environmental, 58, 211-216.

Li, H., Guo, A., & Wang, H. (2008). Mechanisms of oxidative browning of wine. Food Chemistry, 108(1), 1-13.

Li, Y., Bachas, L. G., & Bhattacharyya, D. (2007). Selected chloro-organic detoxifications by polychelate (poly(acrylic acid)) and citrate-based Fenton reaction at neutral pH environment. Industrial & Engineering Chemistry Research, 46(24), 7984-7992.

Lindeman, P. F., & O'Shea, M. G. (2004). Colorant removal during clarification and decolourisation processes. Proceedings of the Australian Society of Sugar Cane Technologists, 26, 51-64.

Lipczynska-Kochany, E., Sprah, G., & Harms, S. (1995). Influence of some groundwater and surface waters constituents on the degradation of 4- chlorophenol by the Fenton reaction. Chemosphere, 30(1), 9-20.

Lu, L.-l., Li, Y.-h., & Lu, X.-y. (2009). Kinetic study of the complexation of gallic acid with Fe(II). Spectrochimica Acta, Part A: Molecular and Biomolecular Spectroscopy, 74, 829-834.

Lucas, M. S., & Peres, J. A. (2009). Treatment of olive mill wastewater by a combined process: Fenton’s reagent and chemical coagulation. Journal of Environmental Science and Health Part A, 44, 198-205.

Lucas, M. S., Peres, J. A., Amor, C., Prieto-Rodríguez, L., Maldonado, M. I., & Malato, S. (2012). Tertiary treatment of pulp mill wastewater by solar photo- Fenton. Journal of Hazardous Materials, 225–226(0), 173-181.

Madsen, L. R., II (2006). Simulated Multiple Effect Evaporation of Clarified and Decolorized Sugarcane Juices and Accelerated Storage Tests of the Syrups: Evolution of Color and the Behavior of Carbohydrates and Related Compounds. Paper presented at the SPRI Conference on Sugar Processing: Sugar Processing Research Institute.

Madsen, L. R., II, & Day, D. F. (2010). Iron mediated clarification and decolourisation of sugarcane juice. Proceedings of the International Society of Sugar Cane Technologists, 27, 1-13.

Mane, J. D., Pachpute, S. P., & Phadnis, S. P. (1998). Effects of hydrogen peroxide treatment on cane syrup. International Sugar Journal, 100(1193), 210-212.

Mane, J. D., Phadnis, S. P., & Jadhav, S. J. (1992). Effects of hydrogen peroxide on cane juice constituents. International Sugar Journal, 94(1128), 322-324.

Mersad, A., Lewandowski, R., Heyd, B., & Decloux, M. (2003). Colorants in the sugar industry: laboratory preparation and spectrometric analysis. International Sugar Journal, 105(1254), 269-281.

Monteagudo, J. M., Durán, A., Aguirre, M., & San Martín, I. (2011). Optimization of the mineralization of a mixture of phenolic pollutants under a ferrioxalate- induced solar photo-Fenton process. Journal of Hazardous Materials, 185, 131-139.

Moodley, M. (1993). The application of cationic flocculants as decolourising agents in the sugar industry. Proceedings of the South African Sugar Technologists' Association, 67, 155-159.

Moodley, M., Davis, S. B., & Adendorff, M. W. (1999). Full scale decolourisation trials with ozone. International Sugar Journal, 101, 165-171.

Morelli, R., Russo-Volpe, S., Bruno, N., & Lo Scalzo, R. (2003). Fenton-dependent damage to carbohydrates: free radical scavenging activity of some simple sugars. Journal of Agricultural and Food Chemistry, 51(25), 7418-7425.

Mudoga, H. L., Yucel, H., & Kincal, N. S. (2008). Decolorization of sugar syrups using commercial and sugar beet pulp based activated carbons. Bioresource Technology, 99, 3528-3533.

Neyens, E., & Baeyens, J. (2003). A review of classic Fenton’s peroxidation as an advanced oxidation technique. Journal of Hazardous Materials, B98, 33-50.

Nichela, D., Haddou, M., Benoit-Marquié, F., Maurette, M.-T., Oliveros, E., & García Einschlag, F. S. (2010). Degradation kinetics of hydroxy and hydroxynitro derivatives of benzoic acid by fenton-like and photo-fenton techniques: A comparative study. Applied Catalysis, B: Environmental, 98, 171-179.

Oliveira, C. M., Ferreira, A. C. S., De Freitas, V., & Silva, A. M. S. (2011). Oxidation mechanisms occurring in wines. Food Research International, 44, 1115-1126.

Olivério, J. L., Boscariol, F. C., & Gabardo Filho, H. (2007). DRD – Dedini Refinado Direto (Dedini Direct Refined): Refined sugar without refinery. Proceedings of the International Society of Sugar Cane Technologists, 26, 1612-1621.

Oturan, M. A., Oturan, N., Edelahi, M. C., Podvorica, F. I., & El Kacemi, K. (2011). Oxidative degradation of herbicide diuron in aqueous medium by Fenton’s reaction based advanced oxidation processes. Chemical Engineering Journal, 171, 127-135.

Pala, A., & Erden, G. (2005). Decolorization of a baker's yeast industry effluent by Fenton oxidation. Journal of Hazardous Materials, B127, 141-148.

Paton, N. H. (1992). The origin of colour in raw sugar. Proceedings of the Australian Society of Sugar Cane Technologists, 14, 8-17.

Paton, N. H., & McCowage, R. J. (1987). Colour forming mechanisms during mill processing. Proceedings of the Australian Society of Sugar Cane Technologists, 11-20.

Pignatello, J. J. (1992). Dark and photoassisted Fe3+-catalyzed degradation of chlorophenoxy herbicides by hydrogen peroxide. Environmental Science and Technology, 26(5), 944-951.

Pignatello, J. J., Oliveros, E., & MacKay, A. (2006). Advanced oxidation processes for organic contaminant destruction based on the Fenton reaction and related chemistry. Critical Reviews in Environmental Science and Technology, 36, 1- 84.

Rice-Evans, C. A., Miller, N. J., & Paganga, G. (1996). Structure-antioxidant activity relationships of flavonoids and phenolic acids. Free Radical Biology and Medicine, 20(7), 933-956.

Riffer, R. (1980). A study of chemical additives for use in sugar refining. Proceedings of the Technical Session on Cane Sugar Refining Research, 84-102.

Riffer, R. (2000). Nonsugars and Sugar Refining. In C. C. Chou (Ed.), Handbook of Sugar Refining: A Manual for the Design and Operation of Sugar Refining Facilities (pp. 627-660). New York: John Wiley & Sons.

Riffer, R. (Ed.). (1988). The Nature of Colorants in Sugarcane and Cane Sugar Manufacture. Amsterdam: Elsevier.

Rivas, F. J., Beltrán, F. J., Frades, J., & Buxeda, P. (2001). Oxidation of p- hydroxybenzoic acid by Fenton's reagent. Water Research, 35(2), 387-396.

Rivas, F. J., Frades, J., Alonso, M. A., Montoya, C., & Monteagudo, J. M. (2005). Fenton's oxidation of food processing wastewater components. Kinetic modeling of protocatechuic acid degradation. Journal of Agricultural and Food Chemistry, 53, 10097-10104.

Rizzi, G. P. (1997). Chemical structure of colored Maillard reaction products. Food Reviews International, 13(1), 1-28.

Ross, H. A., McDougall, G. J., & Stewart, D. (2007). Antiproliferative activity is predominantly associated with ellagitannins in raspberry extracts. Phytochemistry, 68, 218-228.

Saska, M. (2007). Reducing sugar color by washing and bleaching with dilute hydrogen peroxide. International Sugar Journal, 109(1306), 616-622.

Segura, Y., Martínez, F., Melero, J. A., Molina, R., Chand, R., & Bremner, D. H. 0 (2012). Enhancement of the advanced Fenton process (Fe /H2O2) by ultrasound for the mineralization of phenol. Applied Catalysis, B: Environmental, 113–114(0), 100-106.

Shore, M., Broughton, N. W., Dutton, J. V., & Sissons, A. (1984). Factors affecting white sugar colour. Sugar Technology Reviews, 12, 1-99.

Silva Ferreira, A. C., Oliveira, C., Hogg, T., & Guedes de Pinho, P. (2003). Relationship between potentiometric measurements, sensorial analysis, and some substances responsible for aroma degradation of white wines. Journal of Agricultural and Food Chemistry, 51(16), 4668-4672.

Simaratanamongkol, A., & Thiravetyan, P. (2010). Decolorization of melanoidin by activated carbon obtained from bagasse bottom ash. Journal of Food Engineering, 96, 14-17.

Singleton, V. L. (1987). Oxygen with phenols and related reactions in musts, wines, and model systems: Observations and practical implications. American Journal of Enology and Viticulture, 38(1), 69-77.

Sinnott, M. L. (2007). Carbohydrate Chemistry and Biochemistry. Cambridge: RSC Publishing.

Smith, I. A., Schumann, G. T., & Walthew, D. C. (2000). Some developments in flotation clarification. Proceedings of the South African Sugar Technologists' Association, 74, 263-266.

Smith, P., & Paton, N. H. (1985). Sugarcane Flavonoids. In R. A. McGinnis & E. G. Muller (Eds.), Sugar Technology Reviews (Vol. 12, pp. 117-141). Amsterdam: Elsevier.

Smith, P., Paton, N. H., Delaney, H. R., & Ames, R. (1981). Colour studies in milling. Proceedings of the Australian Society of Sugar Cane Technologists, 4, 71-80.

Sroka, Z., & Cisowski, W. (2003). Hydrogen peroxide scavenging, antioxidant and anti-radical activity of some phenolic acids. Food and Chemical Toxicology, 41, 753-758.

Suárez-Pereira, E., Rubio, E. M., Pilard, S., Mellet, C. O., & García Fernández, J. M. (2010). Di-D-fructose dianhydride-enriched products by acid ion-exchange resin-promoted caramelization of D-fructose: chemical analyses. Journal of Agricultural and Food Chemistry, 58(3), 1777-1787.

Sun, J.-H., Shi, S.-H., Lee, Y.-F., & Sun, S.-P. (2009). Fenton oxidative decolorization of the azo dye Direct Blue 15 in aqueous solution. Chemical Engineering Journal, 155, 680-683.

Sun, J.-H., Sun, S.-P., Fan, M.-H., Guo, H.-Q., Qiao, L.-P., & Sun, R.-X. (2007). A kinetic study on the degradation of p-nitroaniline by Fenton oxidation process. Journal of Hazardous Materials, 148, 172-177.

Trovó, A. G., Nogueira, R. F. P., Agüera, A., Fernandez-Alba, A. R., Sirtori, C., & Malato, S. (2009). Degradation of sulfamethoxazole in water by solar photo- Fenton. Chemical and toxicological evaluation. Water Research, 43, 3922- 3931.

Üstün, G. E., Solmaz, S. K. A., & Birgül, A. (2007). Regeneration of industrial district wastewater using a combination of Fenton process and ion exchange— A case study. Resources, Conservation and Recycling, 52(2), 425-440.

Vermerris, W., & Nicholson, R. (2006). Phenolic Compound Biochemistry. Doetinchem, Netherlands: Springer.

Ververidis, F., Trantas, E., Douglas, C., Vollmer, G., Kretzschmar, G., & Panopoulos, N. (2007). Biotechnology of flavonoids and other phenylpropanoid-derived natural products. Part I: Chemical diversity, impacts on plant biology and human health. Biotechnology Journal, 2(10), 1214-1234.

Walling, C. (1975). Fenton's reagent revisited. Accounts of Chemical Research, 8, 125-131.

Wang, Y., Li, X., Zhen, L., Zhang, H., Zhang, Y., & Wang, C. (2012). Electro-Fenton treatment of concentrates generated in nanofiltration of biologically pretreated landfill leachate. Journal of Hazardous Materials, 229–230, 115-121.

Wiggins, L. F. (1953). The Nitrogen-Containing Nonsugars (The Amino Acids and Proteins). In P. Honig (Ed.), Principles of Sugar Technology. Amsterdam: Elsevier.

Wright, P. G., & Jegaraj, T. J. (1992). A study of colour formation in the processing of cane juice, Sugar Research and Development Corporation.

Yang, B. Y., & Montgomery, R. (1996). Alkaline degradation of fructofuranosides. Carbohydrate Research, 280, 47-57.

CHAPTER 3

Determination of Phenolic Compounds in Factory Sugar Cane Juices

3.1 Introduction...... 58 3.2 Materials and Methods...... 58 3.2.1 Reagents and Solvents...... 58 3.2.2 Specification of Samples...... 59 3.2.3 Sample Preparation...... 60 3.2.4 Instrumental Procedures and Analysis...... 60 3.2.5 Colour, Refractive Index and Total Soluble Solids Measurements...... 62 3.3 Results and Discussion...... 62 3.3.1 Colour Analyses of Juices...... 62 3.3.2 Phenolic Content in Juices...... 63 3.4 Summary...... 70

3.1 Introduction

A new approach that has the potential for efficient and cost-effective decolourisation of sugar process streams during the manufacture of raw sugar is through the use of the Fenton oxidation process. As a first step towards developing this technology, the colour and composition of phenolic acids (i.e., colour precursors) present in sugar cane juices obtained from three different harvesting methods were determined. These methods include burnt cane harvesting with all trash (i.e., tops and leaves) extracted; green cane harvesting with a proportion of the trash extracted and whole crop cane harvesting with no trash extracted.

The colour content of each juice sample at pH 7.0 was measured spectrophotometrically at a wavelength of 420 nm according to the international (ICUMSA) method. The phenolic content in the juices was determined using reversed-phase high-performance liquid chromatography (HPLC). Juice samples were hydrolysed based in the standard method and extracted based on the procedures reported by Paton (1978) and Schieber et al. (2001) respectively, prior to HPLC analysis. This procedure was modified by changing the sample preparation method and HPLC operating conditions in order to improve the response of the phenolic compounds for accurate quantification. The most concentrated phenolic compounds were selected for oxidative degradation by the Fenton process in subsequent chapters.

3.2 Materials and Methods

3.2.1 Reagents and Solvents

All chemicals purchased were of analytical reagent (AR) grade and used as supplied without further purification. Solvents for chromatographic analyses were of super gradient HPLC grade from Scharlau (Sentmenat, Spain). Solutions were prepared using ultrapure (Milli-Q) water from a Millipore system (Bedford, MA, USA) with a resistivity of 18.2 MΩ.cm.

Caffeic acid, (±)–catechin, chlorogenic acid, chrysin, m–coumaric acid, o–coumaric acid, p–coumaric acid, coumarin, 2,3–dihydroxybenzoic acid, 5,7–dihydroxy–4–methoxyisoflavone, diosmin, ferulic acid, gallic acid, hesperidin,

hesperetin, HMF, homogentisic acid, 4–hydroxybenzoic acid, kojic acid, morin, quercetin, α–resorcylic acid, β–resorcylic acid, rutin, syringaldehyde, 3,4,5–trimethoxybenzoic acid and vanillic acid were purchased from Sigma-Aldrich (St. Louis, MO, USA). Ammonium chloride, ammonium hydroxide, glacial acetic acid, lead acetate and sodium hydroxide were obtained from Ajax Finechem (Seven Hills, NSW, Australia). Methyl orange indicator and vanillin were supplied from Merck (Darmstadt, Germany) and furfural was from Fluka (Buchs, Switzerland). Protocatechuic acid and sinapinic acid were purchased from Acros Organics (Geel, Belgium). Celite 577 (diatomaceous earth) was obtained from World Minerals (Santa Barbara, CA, USA).

3.2.2 Specification of Samples

First expressed juice from burnt harvested sugar cane was obtained from the processing lines at Condong Sugar Mill (Condong, NSW, Australia). Whole crop harvested cane FEJ was obtained by harvesting sugar cane located around Condong Sugar Mill in the field and expressing the juice with a laboratory hammer mill designed by the Sugar Research Institute (SRI) (Brisbane, QLD, Australia). The specification of the mill was as follows: 430 × 220 mm roll; 12.8 mm groove pitch; 12.0 mm groove depth; 4 rpm operating speed; and a 10 hp powered motor running at 7.5 kW. The juice was collected by pressing through a 1 mm mesh sieve.

Both FEJs were obtained during the crushing season in 2009. Samples of primary juice (PJ) (i.e., incubated MJ prior to lime addition) from burnt cane and green cane were obtained at ca. 76 °C and pH 5.15 from Condong Sugar Mill during the crushing season in 2010. All juices were stored at –22 °C. In total, four juices (2 × FEJs and 2 × PJs) were analysed. The following analyses of the four juice samples are unrelated and not comparable. The results obtained provide an insight on the levels of colour and phenolic content present in each juice type.

3.2.3 Sample Preparation

The mill juices were analysed as phenolic extracts after alkaline hydrolysis. Hydrolysis was carried out under ambient temperature using 2.0 M NaOH on centrifuged juice (50% (v/v)) for 30 min at ambient temperature with magnetic stirring (280 rpm).

Two different liquid-liquid extraction procedures were carried out for the determination of phenolic acid content in the juices collected as follows:

Method A. The hydrolysed juice was neutralised by adjusting the pH to 3.0 with 6.0 M HCl and extracted three times with diethyl ether (20 mL). The combined extracts were dried over anhydrous sodium sulfate followed by evaporation to dryness in vacuo to constant weight. The individual residues were weighed, dissolved in water (10 mL) and membrane filtered (0.45 µm) prior to analysis by HPLC. The procedure is that developed by Paton (1978).

Method B. Hydrolysed juices were treated in the same manner as described in Method A but instead were extracted three times with ethyl acetate (50 mL). The dried individual residues were dissolved in HPLC grade methanol (10 mL) prior to membrane filtration followed by HPLC analysis.

3.2.4 Instrumental Procedures and Analyses

The organic extracts were analysed using reversed-phase HPLC with UV/Visible (UV/Vis) diode-array detection (DAD). The method was adapted from a previously reported method for the determination of phenolic acids in apple and pear juices (Schieber et al., 2001). Analyses were performed on a Hewlett Packard HP/Agilent 1100 Series HPLC system (G1379A micro-degasser, Japan; G1311A quaternary pump, Germany; G1313A automatic liquid sampler (ALS), Germany; G1315B diode-array detector, Germany) using a Waters Symmetry C18 column (150 × 3.9 mm i.d.) with a Waters Guard-Pak guard holder containing a Waters Guard-Pak Resolve C18 guard insert (10 µm) (Milford, MA, USA). The mobile phase consisted of 2.0% (v/v) glacial acetic acid in water (as eluent A) and methanol

(as eluent B). The gradient programs for extracts produced from each method were as follows:

Method A. 10% B to 17% B (18 min), 17% B to 23% B (12 min), isocratic (10 min), 23% B to 31% B (13 min), 31% B to 46% B (12 min), 46% B to 55% B (5 min), 55% B to 100% B (5 min), isocratic (8 min), 100% B to 10% B (2 min) and isocratic (5 min).

Method B. 2% B to 5% B (10 min), 5% B to 20% B (50 min), 20% B to 50% B (20 min), isocratic (5 min) and 50% B to 2% B (5 min).

Simultaneous detection at specific wavelengths (280 nm, 320 nm, 370 nm and 420 nm) was subtracted against a reference wavelength (600 nm). The wavelengths were chosen for identification and quantification of the various types of phenolic compounds. Data on hydroxybenzoic acids can be collected at 280 nm; hydroxycinnamic acids at 320 nm; flavanols and chalcones at 370 nm; and other flavonoid derivatives at 420 nm (Cai et al., 2004; Stalikas, 2007).

Aliquots of samples were membrane filtered (0.45 μm) prior to injection into the HPLC system. Injection volumes for all samples were 10 μL and 8.0 μL for extracts produced from Methods A and B respectively. Column temperature was 25 °C; flow rate was 1.0 mL and run time was 90 min. After each run, the chromatographic system was equilibrated for 10 min. Data acquisition was performed using the Agilent ChemStation (Rev. A.09.03) software package. Analyses of samples were carried out in triplicate.

Identification of peaks was based on the conformance of UV/Vis spectra and retention times with the corresponding authentic standards. Calibration curves for 18 compounds were constructed using five different standard concentrations over the concentration ranges expected in sugar process streams (Curtin and Paton, 1980; Payet et al., 2006). The calibration curves were linear (R2 = 1.00). The peak heights of the target compounds were within the linear range of the calibration curve. Analyses of standards were carried out in triplicate.

3.2.5 Colour, Refractive Index and Total Soluble Solids Measurements

Celite 577 (7.5 g) was suspended in 50 mL of juice and stirred magnetically for 10 min at ambient temperature. The adsorbed fine particles present in the suspension were removed by vacuum filtration. The filtrate was diluted to an appropriate absorbance range and membrane filtered (0.45 μm) before adjusting the pH to 4.0 using 0.01 M HCl, and pH 7.0 and pH 9.0 using 0.01 M NaOH.

Absorbance measurements were conducted spectrophotometrically at 420 nm (A420) on a GBC Scientific Cintra 40 double-beam UV/Vis spectrophotometer (Braeside, VIC, Australia) using cells of 1.0 cm path length. Data acquisition was performed using the GBC Spectral 1.50 software package. The resulting colour of each sample was calculated as:

1000 A (3.1) Colour (IU)  420 Cell Length Sucrose Concentration

The total soluble solids (TSS) of the juice and refractive index (RI) of the filtrate were measured at ambient temperature using a Bellingham and Stanley RFM 342 refractometer (Tunbridge Wells, UK) accurate to ± 0.01 °Bx and ± 0.00001 RI units respectively. The RI values were used to determine the corresponding concentration of sucrose in solution (g/mL) based on Table XII in the Bureau of Sugar Experiment Stations (BSES) Laboratory Manual for Australian Sugar Mills (BSES, 2001).

3.3 Results and Discussion

3.3.1 Colour Analyses of Juices

Colour is conventionally measured at pH 7.0. Flavonoids and phenolic compounds are pH sensitive and their colour profile increases greatly from minimal colour in untreated MJ and FEJ (at pH 4.0–5.0) up to near-maximum colour at pH 9.0 (Paton, 1992). Therefore, colour measured at pH of 7.0 or higher would provide satisfactory measurement of the presence of flavonoids and phenolic compounds. The colour of Condong Sugar Mill juices is presented in Table 3.1. High colour and impurity was recorded with juices expressed from whole crop harvested cane. This is

primarily due to green cane harvesting where green tops and brown leafy trash are processed (Eggleston et al., 2010).

Table 3.1 Colour of factory sugar cane juices recorded at pH 7.

PJ FEJ Green cane Burnt cane Whole crop Burnt cane TSS (°Bx) 13.19 14.97 16.39 19.43 RI 1.3537 1.3601 1.3579 1.3690 Colour (IU) 20,000 12,700 11,400 10,400 *Mean values (n = 3). % Relative standard deviation (RSD) was ≤ 0.7%.

3.3.2 Phenolic Content in Juices

The phenolic compounds separated from the cane juice extracts using Method A are shown in Figure 3.1. Baseline separation was achieved for all identified components. The m– and o–isomers of coumaric acid were not detected in any of the extracts analysed. The elution order of the phenolic compounds was consistent with previous studies under different chromatographic conditions with the exception of 2,3–dihydroxybenzoic acid and chlorogenic acid (Curtin and Paton, 1980).

50 1 40 2 30

20 8 9 5 10 3 6 7 10 Absorbnace(mAU) 4 0

-10 0 5 10 15 20 25 30 35 Retention Time (min)

The concentrations of each compound varied with the juice type. These are tabulated in terms of mM on dry content as shown in Table 3.2. The concentrations of phenolic compounds in whole crop and green harvested cane juices are substantially lower than burnt harvested cane juices. This is probably due to the valorisation of lignin, a component of bagasse, during cane burning.

Table 3.2 shows that higher amounts of HMF were identified in both FEJ and PJ extracts of burnt cane compared to the extracts of green cane and whole crop. This is due to the dehydration of sugars, particularly reducing sugars, to HMF as a result of high temperatures generated during the burning of cane prior to harvesting (Huber et al., 2006). Prior to this work, the quantification of HMF in Australian FEJ and PJ extracts using this method has not been reported in the literature. Also, the concentration of caffeic acid (Table 3.2) was relatively lower than other phenolic acids in comparison to previous work on Australian factory cane juice (Curtin and Paton, 1980).

Table 3.2 Phenolic acids and HMF (mM on dry content) by HPLC-DAD of sugar cane juices using Method A.*

PJ FEJ Green cane Burnt cane Whole crop Burnt cane Caffeic 10 20 0.68 6.9 Chlorogenic 2.2 19 0.20 8.6 p–Coumaric 14 87 1.2 19 2,3–Dihydroxybenzoic 12 24 0.80 7.7 Ferulic 6.0 11 0.48 5.0 4–Hydroxybenzoic 8.3 20 0.45 5.0 Vanillic 14 25 0.68 7.3 HMF 0.43 7.8 0.40 1.4 *Mean values (n = 3). % RSD was < 5.0%.

Higher concentrations of phenolic compounds are present in PJs compared to FEJs (Table 3.2). This is probably due to the decomposition of flavonoids followed by oxidation of the intermediate products and further degradation of lignin products at the relatively higher processing temperatures of PJ.

Table 3.3 shows a comparison of the phenolic acid and HMF composition based on the PJs from Table 3.2 in terms of mM on juice, to those reported by Curtin and Paton (1980). The total amount of phenolic compounds are considerably higher than those previously reported by Curtin and Paton (1980). The differences between the two sets of data may be related to differences in the cane varieties or the differences in the analytical procedures used for phenolic composition analysis.

Evident from Table 3.1, the juices expressed from green cane and whole crop contain higher colour but lower amounts of phenolic compounds than the juices expressed from burnt cane (Table 3.2). It is therefore deduced that the juices expressed from green cane and whole crop cane harvesting contain a higher proportion of cane pigments (e.g., flavonoids).

Table 3.3 Phenolic acids (mM on juice) and HMF by HPLC-DAD of PJs using Method A.*

Green cane Burnt cane Burnt cane† Caffeic 0.044 0.055 0.083 Chlorogenic 0.009 0.051 0.000 p–Coumaric 0.060 0.240 0.002 2,3–Dihydroxybenzoic 0.053 0.065 0.001 Ferulic 0.026 0.028 0.002 4–Hydroxybenzoic 0.036 0.055 0.001 Vanillic 0.058 0.067 0.002 *Mean values (n = 3). % RSD was < 5.0%. †Cane juice data based from Curtin and Paton (1980).

The reversed-phase HPLC method (Method B) was then optimised for separation and identification of phenolic compounds. Different solvents for the liquid-liquid extraction of phenolics and dissolution of the dried residues as well as modifications of the gradient program were conducted to obtain chromatograms with good resolution of peaks with an acceptable analysis time. Changes in the injection volume and gradient program significantly affected the resolution and separation of peaks.

Ethyl acetate was chosen as the solvent for liquid-liquid extraction in place of diethyl ether as it has a better extraction efficiency and higher recovery yields for phenolic acids but not phenolic aldehydes (Simón et al., 1990). The improved solubility of the dried residues in methanol resulted in better detector response (Robbins, 2003). The separation of a standard mixture of 20 phenolic compounds (viz., hydroxybenzoic and hydroxycinnamic acids), HMF and furfural monitored at 280 nm using Method B is shown in Figure 3.2.

60 4 50 2 40

30 1 20 11 3 5 8 10 16 15 10 7 12 13 22 Absorbance(mAU) 21 6 9 14 17 18 19 20 0

-10 0 10 20 30 40 50 60 70 80 90 Retention Time (min)

Figure 3.2 Separation of a standard mixture of compounds by HPLC-DAD (Method B, UV/Vis detection at 280 nm). 1 = gallic acid, 2 = HMF, 3 = protocatechuic acid, 4 = furfural, 5 = 4–hydroxybenzoic acid, 6 = (±)–catechin, 7 = vanillic acid, 8 = caffeic acid, 9 = chlorogenic acid, 10 = vanillin, 11 = p–coumaric acid, 12 = syringaldehyde, 13 = ferulic acid, 14 = sinapinic acid, 15 = coumarin, 16 = o–coumaric acid, 17 = 3,4,5–trimethoxybenzoic acid, 18 = rutin, 19 = diosmin, 20 = chrysin, 21 = morin, 22 = quercetin.

As shown in Figure 3.2 the baseline separation was achieved for virtually all components evenly across the whole chromatogram except for chlorogenic acid and sinapinic acid which are overlapped by caffeic acid and coumarin respectively. For these four compounds, chlorogenic acid was not quantified as it was superimposed by caffeic acid.

The HPLC chromatogram of burnt cane PJ using Method B is shown in Figure 3.3. The concentration of phenolic components of the cane juice extracts are shown in Table 3.4. The phenolic acid content detected in the sugar cane extracts in this study were similar to the values obtained by Payet et al., (2006) for various sugar process streams and products after juice clarification. The concentrations of the phenolic compounds detected are consistent to those found in orange juice (Rapisarda

et al., 1998), mandarin juice (Kelebek and Selli, 2011) and commercial fruit juices (viz., apple, grape strawberry) (Díaz-García et al., 2013).

As shown in Figure 3.3 and Table 3.4, the juice extracts mainly consisted of hydroxybenzoic and hydroxycinnamic acids, with the latter having the highest total concentration. Caffeic, p–coumaric and ferulic acids are the most concentrated hydroxycinnamic acids, while 4–hydroxybenzoic and vanillic acids are the most concentrated hydroxybenzoic acids.

60 8 50

30 10 3 2 4 15 16 20 7 14 17 13 6 10 1 5 Absorbance(mAU) 9 11 12 0

-10 0 10 20 30 40 50 60 70 80 90 Retention Time (min)

Figure 3.3 Separation of a typical mixture of compounds in the PJ extract of burnt harvested cane by HPLC-DAD (Method B, UV/Vis detection at 280 nm). 1 = gallic acid, 2 = HMF, 3 = protocatechuic acid, 4 = furfural, 5 = 4–hydroxybenzoic acid, 6 = vanillic acid, 7 = caffeic acid, 8 = p–coumaric acid, 9 = syringaldehyde, 10 = ferulic acid, 11 = sinapinic acid, 12 = coumarin, 13 = rutin, 14 = diosmin, 15 = chrysin, 16 = morin, 17 = quercetin.

Table 3.4 Phenolic acids and HMF (mM on dry content) by HPLC-DAD of sugar cane juices using Method B.*

PJ FEJ Green cane Burnt cane Whole crop Burnt cane Hydroxybenzoic acids 2,3–Dihydroxybenzoic 20 17 13 25 Gallic 22 8.1 0.80 29 4–Hydroxybenzoic 33 25 24 43 Protocatechuic 12 6.0 6.9 11 Vanillic 54 34 34 82 Hydroxycinnamic acids Caffeic 53 36 6.8 190 p–Coumaric 140 120 66 160 Ferulic 180 100 62 190 Sinapinic 6.8 4.7 4.0 15 Flavonoids Chrysin 12 1.9 2.2 – Morin 8.4 7.7 7.5 15 Quercetin 16 7.5 10 16 Rutin 5.1 2.7 2.5 5 Other phenolic compounds Coumarin 7.7 11 6.5 27 Syringaldehyde 18 15 11 24 Non-phenolic compounds Furfural 6.0 3.4 – 7.7 HMF 0.83 0.12 – 5.7 * Mean values (n = 3). % RSD was ≤ 13.4%.

The flavonoid compounds, chrysin, diosmin, morin, quercetin and rutin were also detected using Method B. These compounds were eluted towards the end of the chromatogram (> 60 min) as shown in Figure 3.3. It is assumed that the unidentified peaks within the 60–90 min timeframe of each chromatogram are attributable to

flavonoid compounds. The m– and o– isomers of coumaric acid were not found. The components (±)–catechin, 5,7–dihydroxy–4–methoxyisoflavone, hesperidin, hesperetin, homogentisic acid, kojic acid, α–resorcylic acid, β–resorcylic acid and 3,4,5–trimethoxybenzoic acid were also not identified in any of the four juice extracts. These compounds are typically found in commercial products such as honeys (Gómez-Caravaca et al., 2006) and fruit juices (Díaz-García et al., 2013).

3.4 Summary

Fifteen phenolic compounds, HMF and furfural were quantified in juice extracts of FEJ and PJ process streams expressed from burnt, green and whole crop harvested cane. The results show that juice expressed from whole crop cane has significantly higher colour but lower concentrations of phenolic acids than juices expressed from burnt cane. It was deduced that the juices expressed from green cane and whole crop cane harvesting contain a higher proportion of cane pigments.

Changes to the extraction procedure, sample preparation and chromatographic conditions as outlined in the modified method (Method B), gave more definitive peak separation and showed an overall improved response to phenolic acids and revealed the presence of flavonoid compounds. Interestingly, the concentrations of phenolic acids separated using Method A showed a higher proportion of hydroxybenzoic acids than hydroxycinnamic acids, possibly due to the solubility effect of the dried extracts. However, the opposite was observed when the extracts were separated using Method B. Using the modified method, the HPLC results reveal that caffeic, p–coumaric and ferulic acids were the three main phenolic acids present in FEJ and PJ extracts sourced from burnt cane, green cane and/or whole crop harvested cane.

References

BSES (2001). Table XII, Laboratory Manual for Australian Sugar Mills (Vol. 2, pp. 1-7). Indooroopilly, QLD, Australia: Bureau of Sugar Experiment Stations.

Cai, Y., Luo, Q., Sun, M., & Corke, H. (2004). Antioxidant activity and phenolic compounds of 112 traditional Chinese medicinal plants associated with anticancer. Life Sciences, 74(17), 2157-2184.

Díaz-García, M. C., Obón, J. M., Castellar, M. R., Collado, J., & Alacid, M. (2013). Quantification by UHPLC of total individual polyphenols in fruit juices. Food Chemistry, 138(2–3), 938-949.

Eggleston, G., Grisham, M., & Antoine, A. (2010). Clarification properties of trash and stalk tissues from sugar cane. Journal of Agricultural and Food Chemistry, 58(1), 366-373.

Gómez-Caravaca, A. M., Gómez-Romero, M., Arráez-Román, D., Segura-Carretero, A., & Fernández-Gutiérrez, A. (2006). Advances in the analysis of phenolic compounds in products derived from bees. Journal of Pharmaceutical and Biomedical Analysis, 41(4), 1220-1234.

Huber, G. W., Iborra, S., & Corma, A. (2006). Synthesis of transportation fuels from biomass: chemistry, catalysts and engineering. Chemical Reviews, 106(9), 4044-4098.

Kelebek, H., & Selli, S. (2011). Identification of phenolic compositions and the antioxidant capacity of mandarin juices and wines. Journal of Food Science and Technology, 1-8.

Paton, N. H. (1978). A method for the separation and identification of phenolic acids in sugar products. Proceedings of the International Society of Sugar Cane Technologists, 16, 2633-2643.

Paton, N. H. (1992). The origin of colour in raw sugar. Proceedings of the Australian Society of Sugar Cane Technologists, 14, 8-17.

Payet, B., Shum Cheong Sing, A., & Smadja, J. (2006). Comparison of the concentrations of phenolic constituents in cane sugar manufacturing products with their antioxidant activities. Journal of Agricultural and Food Chemistry, 54, 7270-7276.

Rapisarda, P., Carollo, G., Fallico, B., Tomaselli, F., & Maccarone, E. (1998). Hydroxycinnamic acids as markers of Italian blood orange juices. Journal of Agricultural and Food Chemistry, 46(2), 464-470.

Robbins, R. J. (2003). Phenolic acids in foods: an overview of analytical methodology. Journal of Agricultural and Food Chemistry, 51(10), 2866- 2887.

Schieber, A., Keller, P., & Carle, R. (2001). Determination of phenolic acids and flavonoids of apple and pear by high-performance liquid chromatography. Journal of Chromatography A, 910, 265-273.

Simón, B. F., Pérez-Ilzarbe, J., Hernández, T., Gómez-Cordovés, C., & Estrella, I. (1990). HPLC study of the efficiency of extraction of phenolic compounds. Chromatographia, 30(1-2), 35-37.

Stalikas, C. D. (2007). Extraction, separation, and detection methods for phenolic acids and flavonoids. Journal of Separation Science, 30, 3268-3295.

CHAPTER 4

Degradation of

Hydroxycinnamic Acids

4.1 Introduction...... 74 4.2 Materials and Methods...... 75 4.2.1 Reagents and Solvents...... 75 4.2.2 Catalytic and Non-catalytic Oxidation of Caffeic Acid...... 75 4.2.3 Fenton Oxidation Reactions for Caffeic Acid Degradation 76 4.2.4 Fenton Oxidation Reactions for the Degradation of Hydroxycinnamic Acid Mixtures...... 78 4.2.5 Instrumental Procedures and Analyses...... 78 4.2.6 Performance Assessment of the Fenton Oxidation Process 79 4.2.7 Design of Experiments...... 80 4.2.8 Statistical Analysis...... 82 4.2.9 Evaluation of the Interactions between Fe(II) and Hydroxycinnamic Acids...... 82 4.3 Results and Discussion...... 83 4.3.1 Catalytic and Non-catalytic Oxidation of Caffeic Acid in Aqueous Systems...... 83 4.3.2 Optimisation of Process Parameters for the Degradation of Caffeic Acid in Sugar Solutions...... 87 4.3.3 Degradation of Hydroxycinnamic Acid Mixtures...... 100 4.4 Summary...... 130

4.1 Introduction

As reported in Chapter 3, the main phenolic acids present in sugar cane juice are caffeic acid (CaA), p–coumaric acid (pCoA) and ferulic acid (FeA), which are classed as hydroxycinnamic acids (HCAs). Thus, the aim of this chapter was to determine the optimum conditions and develop models for the rapid degradation of these colour precursors by the Fenton oxidation process.

In the first section, Section 4.3.1, a preliminary investigation compared the performance of the Fenton process on the degradation of caffeic acid in aqueous solution to that of H2O2 alone.

The outputs from the study were then used to identify the necessary process parameters and their numeric constraints for the development of a mathematical model (Section 4.3.2). Response surface methodology (RSM) and central composite experimental design were used to determine the optimum conditions for the degradation of CaA. Also, the model was used to predict the optimum conditions for the degradation of caffeic acid at particular stages of the sugar manufacturing process.

In Section 4.3.3, the study builds on the results and observations from the previous sections by examining the degradation of a mixture of three HCAs; CaA, pCoA and FeA, using the Fenton oxidation process in the presence and absence of sucrose. Multi-response surface methodology (MRSM) was used for modelling and optimisation of process parameters for the degradation process by examining individual and interactive influences of the parameters. The rigorous optimisation process undertaken in this study was to accurately determine the exact amounts of phenolic acids degraded under the chosen Fenton process conditions, as any excess iron would result in an increase in the amount of iron sludge and colour formed during processing.

4.2 Materials and Methods

4.2.1 Reagents and Solvents

All chemicals purchased were of AR grade and used as supplied without further purification. Solvents for chromatographic analyses were of super gradient HPLC grade from Scharlau (Sentmenat, Spain). Solutions were prepared using Milli-Q water from a Millipore system (Bedford, MA, USA) with a resistivity of 18.2 MΩ.cm.

The phenolic acids (CaA, pCoA and FeA), fructose, glucose, lactose and sucrose were purchased from Sigma-Aldrich (St. Louis, MO, USA). Ferrous sulfate heptahydrate (FeSO4·7H2O), glacial acetic acid, H2O2 (30% (w/v)), potassium permanganate, sodium acetate, sodium hydroxide, sodium oxalate and sulfuric acid were obtained from Ajax Finechem (Seven Hills, NSW, Australia). Ethanol (absolute) was supplied from Merck (Darmstadt, Germany). Stock solutions of HCAs (i.e., CaA, pCoA and FeA) were prepared individually by dissolution in degassed ethanol solution (50% (v/v)) and stored at 4.0 °C, unless otherwise stated.

4.2.2 Catalytic and Non-catalytic Oxidation of Caffeic Acid

Caffeic acid solution (55.5 mM) was prepared by dissolving CaA in degassed absolute ethanol solution (50% (v/v)). Aqueous Fe(II) solution (179 mM) was prepared by dissolving solid FeSO4·7H2O in Milli-Q water. Dilute H2O2 solution

(147 mM) was prepared from stock H2O2 with Milli-Q water and standardised iodometrically. The materials were used to prepare a series of solutions with a final concentration of CaA (1.11 mM), Fe(II) (0 or 0.72 mM) and H2O2 (2.94 or 11.8 mM) according to the sample matrix given in Table 4.1.

Table 4.1 Volumes of reagents (mM) used for the degradation of CaA.

Sample Water CaA Fe(II) H2O2 Total Final [H2O2]

(μL) (μL) (μL) (μL) (μL) (mM) Non-catalytic oxidation Control 49,000 1,000 0 0 50,000 0 Test 1 48,000 1,000 0 1,000 50,000 2.94 Test 2 45,000 1,000 0 4,000 50,000 11.8 Catalytic oxidation Control 48,880 1,000 200 0 50,000 0 Test 3 47,800 1,000 200 1,000 50,000 2.94 Test 4 44,800 1,000 200 4,000 50,000 11.8

Reactions were carried out in 50 mL Erlenmeyer flasks at ambient temperature. The procedure for the catalytic oxidation can be described as follows:

(i) adjusting the pH to 3.0, 4.0 or 5.0 of the CaA solution using 0.01 M H2SO4 or

0.1 M NaOH; (ii) addition of Fe(II); (iii) addition of H2O2; and (iv) the reaction allowed to run for up to 30 min with continuous magnetic stirring (280 rpm). The procedure for the non-catalytic oxidation was identical with the exceptions that no Fe(II) was added and that the reaction was allowed to run for up to 60 min. The pH was measured using a Radiometer Analytical MeterLab PHM 220 pH meter (Lyon, France). Aliquots (1 mL) were taken at 5 min intervals, diluted 10-fold and analysed spectrophotometrically. Spectrophotometric measurements were conducted at wavelengths ranging between 190 nm and 800 nm on a GBC Cintra 40 double beam UV/Vis spectrophotometer using cells of 1.0 cm path length. Data acquisition was performed using the GBC Spectral 1.50 software package.

4.2.3 Fenton Oxidation Reactions for Caffeic Acid Degradation

Reactions were carried out in 10 mL glass scintillated reaction vessels housed in an 18971 Pierce Reacti-Therm heating/stirring module (Rockford, IL, USA) with continuous magnetic stirring (280 rpm) (Figure 4.1). In each run, a predetermined amount of Milli-Q water, sucrose and CaA were added to the reaction vessel and the

whole adjusted to the desired pH value with 0.01 M H2SO4 or 0.1 M NaOH. Known amounts of FeSO4·7H2O and H2O2 solutions were added to achieve a final volume of

10 mL. The reaction was initiated as soon as H2O2 was added. For pH measurements, a Hach H160 portable pH meter (Loveland, CO, USA) with a Eutech Instruments glass pH electrode (Singapore) was used. Temperature was monitored using a Comark C9001 thermometer probe (Sheffield, UK). At the required time of sampling, 1.0 mL of the solution was taken, diluted 10-fold to quench the reaction and measured immediately spectrophotometrically at 320 nm on a GBC Cintra 40 double beam UV/Vis spectrophotometer (Braeside, VIC, Australia) using cells of 1.0 cm path length. Data acquisition was performed using the GBC Spectral 1.50 software package.

Figure 4.1 Schematic representation of heating block used for the Fenton oxidation process.

4.2.4 Fenton Oxidation Reactions for the Degradation of Hydroxycinnamic Acid Mixtures

The procedure for the Fenton oxidative degradation of HCA mixtures is similar to that described in Section 4.2.3. In each run, a predetermined amount of Milli-Q water, sucrose and each HCA (equivalent mg/L concentration) were added to the reaction vessel. Known amounts of FeSO4·7H2O (50 mM, 0.498 mL) and H2O2 (500 mM, 0.150 mL) solutions were added to achieve a final volume of 10 mL and a final concentration of 2.49 mM and 7.50 mM, respectively. The working molar ratio of 1:15 (Fe(II)/H2O2) for the Fenton reaction of HCA mixtures was chosen based on the optimum molar ratio of 1:13 for CaA solutions (cf. Section 4.3.2). The reaction was initiated as soon as H2O2 was added. At 2 min, 3 mL of the solution was taken, diluted 10-fold to quench the reaction and kept frozen. Samples were defrosted and prepared for instrumental analysis.

4.2.5 Instrumental Procedures and Analyses

HPLC-DAD. The proportion of each HCA degraded was monitored by reversed-phase HPLC-DAD. The analysis was performed on a Hewlett Packard HP/Agilent 1100 Series HPLC system (G1379A micro-degasser, Japan; G1311A quaternary pump, Germany; G1313A ALS, Germany; G1315B DAD, Germany) using a Waters Symmetry C18 column (150 × 3.9 mm i.d.) with a Waters Guard-Pak guard holder containing a Waters Guard-Pak Resolve C18 guard insert (10 μm) (Milford, MA, USA). The mobile phase consisted of 1.0% (v/v) acetic acid in water (as eluent A) and methanol (as eluent B). The gradient program was as follows: 20% B to 25% B (5 min), 25% B to 50% B (15 min) and 50% B to 20% B (5 min). Simultaneous detection at specific wavelengths (280 nm and 320 nm) subtracted against a reference wavelength (620 nm). Aliquots of samples were membrane filtered (0.45 μm) prior to injection into the HPLC system. Injection volume for all samples was 50 μL; column temperature was ambient; flow rate was 1.0 mL/min and run time was 25 min. After each run, the chromatographic system was equilibrated for 5 min. Data acquisition was performed using the Agilent ChemStation (Rev. A.09.03) software package. Identification of peaks was based on the conformance of UV/Vis spectra and retention times with the corresponding authentic standards.

HPAEC-PAD. Sucrose and reducing sugar contents in the reaction mixtures were monitored by high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD). The analysis was performed on a Waters HPLC system (Milford, MA, USA) equipped with a 626 pump, a 600S controller, a 717plus autosampler and a 2465 electrochemical detector (fitted with a solid-state Ag/AgCl reference electrode and a gold working electrode). The waveforms:

E1 = + 0.08 V for 0.4 s; E2 = +0.73 V for 0.4 s and E3 = –0.57 V for 0.2 s were employed with a PAD intensity of 10 μA. Aliquots of samples were diluted 100-fold, membrane filtered (0.45 µm) and injected (20 µL) on a Dionex CarboPac PA-1 guard column (50 × 4 mm i.d.) attached to a Dionex CarboPac PA-1 anion exchange column (250 × 4 mm i.d.) (Waltham, MA, USA). The columns were equilibrated at 27 °C. The sugars were eluted isocratically with 150 mM NaOH (sparged with helium at 30 mL/min) at a flow rate of 1.0 mL/min. Data acquisition was performed using the Waters Empower 2 (Build 2154) software package. Quantification of sugars was carried out by external calibration using standard solutions of sucrose, glucose and fructose in combination with lactose (as an internal standard).

4.2.6 Performance Assessment of the Fenton Oxidation Process

The efficiency of the Fenton process on the degradation of CaA, pCoA and FeA was determined based on the change in absorbance of the corresponding HPLC chromatographic peak using Equation 4.1:

AA (4.1) % CaA, p CoA or FeA degradation = 0 t  100 A0

where, A0 initial absorbance of HCA in mAU (at t = 0 min)

At absorbance of HCA in mAU at time of aliquot taken (at t = 2 min)

4.2.7 Design of Experiments

Design of experiments (DOE), mathematical modelling and optimisation of process parameters were evaluated using the Stat-Ease, Inc. Design-Expert 7.0.0 software package (Minneapolis, MN, USA). Two experimental designs were developed for the two separate batch Fenton oxidation reaction studies. The first DOE was developed for the Fenton oxidation of CaA (cf. Section 4.2.3). Meanwhile, the second DOE was developed for the Fenton oxidation of HCA mixtures (cf. Section 4.2.4).

Fenton Oxidation of Caffeic Acid (Design 1)

A face-centred central composite design (CCD) was used to evaluate the main effect for each condition and the possible interaction effects on the residual stresses between two factors. The factors (independent variables) used in this study were CaA concentration (x1), sucrose concentration (x2), initial solution pH (x3), Fe(II) dosage

(x4), H2O2 dosage (x5), reaction temperature (x6) and reaction time (x7). The selected response factor (dependent variable) for optimisation was % CaA degradation (y). The coded and actual values of each factor and their levels for this experimental design used in this study are shown in Table 4.2. The ranges for each parameter were determined by preliminary experiments based on previous works published in the literature (Pala and Erden, 2005; Nguyen and Doherty, 2012). The reaction time was kept to 2 min in order to minimise sucrose degradation in order to allow treatment of sugar cane process streams, where the main objective is to preserve the sucrose content.

The design consisted of a 2k factorial augmented by 2k axial points and a centre point, where k is the number of factors investigated (k = 7). For this study, a total of 152 experiments were conducted in random order with 128 factorial points, 14 axial points and 1 centre point (duplicated 9 times for experimental error calculation).

Table 4.2 Coded and actual values of the experimental design for Design 1.

Coded levels of parameters Notation Factor Unit –1 0 +1

A (x1) CaA mM 0.28 0.70 1.11

B (x2) Sucrose % (w/w) 0 17 34

C (x3) Solution pH 3.5 5.0 6.5

D (x4) Fe(II) dosage mM 0.18 0.45 0.72

E (x5) H2O2 dosage mM 2.21 6.62 11.03

F (x6) Temperature °C 35 65 95

G (x7) Time s 10 65 120

Fenton Oxidation of Hydroxycinnamic Acid Mixtures (Design 2)

A rotatable circumscribed CCD was used to evaluate the main effect for each condition and the possible interactive effects on the residual stresses between two variables. The process parameters (independent variables) used in this study were the initial total HCA concentration (x1), the initial sucrose concentration (x2), the solution pH (x3) and the reaction temperature (x4). The selected response factors (dependent variables) for optimisation were % CaA degradation (y1), % pCoA degradation (y2),

% FeA degradation (y3) and % total HCA degradation (y4). The coded and actual values of each variable and their levels for the experimental design used in the study are shown in Table 4.3. The ranges for each parameter were determined by preliminary experiments based on the previous experimental design (i.e., Design 1) and were selected to closely mimic operating parameters during the processing of sugar cane juice for raw sugar manufacture. Concentrations of HCAs vary depending on season, region and type of cane and the method of harvesting (e.g., burnt cane, green cane, whole crop cane), hence, an initial total HCA concentration range of 20–200 mg/L was chosen for this study in order to account for other HCAs and phenolic compounds present in sugar cane juice.

Table 4.3 Coded and actual values of the experimental design for Design 2.

Coded levels of parameters Notation Factor Unit –2 –1 0 +1 +2

A (x1) Total HCA mg/L 20 65 110 155 200

B (x2) Sucrose % (w/w) 0 3.75 7.50 11.25 15.00

C (x3) pH 4.50 3.75 5.00 5.25 5.50

D (x4) Temperature °C 25.00 31.25 37.50 43.75 50.00

The design consisted of a 2k factorial augmented by 2k axial points and a centre point, where k is the number of factors investigated (k = 4). For this study, a total of 54 experiments were conducted in random order with 16 factorial points (in duplicate), 8 axial points (in duplicate) and 1 centre point (duplicated 5 times). Duplicate runs were required for experimental error calculation.

4.2.8 Statistical Analysis

Analysis of variance (ANOVA) was used for model adequacy and analysis of the experimental data. The quality of the fit polynomial model was expressed by the regression coefficient, R2 and its statistical significance was checked using Fisher’s F-test. Model terms were determined based on the significance of each term at a confidence level of 95%.

4.2.9 Evaluation of the Interactions between Fe(II) and Hydroxycinnamic Acids

Studies were conducted to investigate the interaction between Fe(II) and each of the HCAs in the presence and absence of sucrose using UV/Vis and attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy. Sodium acetate (100 mM) and acetic acid (100 mM) solutions were used to make buffer solutions having pH values of 4.0 to 6.0. For each analysis, a predetermined amount of buffer, sucrose and FeSO4·7H2O were added to achieve a final HCA concentration of 5.5 mM. Samples were diluted to the desired concentration and immediately membrane filtered (0.45 μm) for analysis. The pH of each solution was checked

before and after dilution. The UV/Vis spectra were recorded on a Perkin Elmer Lambda 35 double-beam UV/Vis spectrophotometer (Shelton, CT, USA) using cells of 1.0 cm path length and at a wavelength range of 190–450 nm in 1.0 nm increments. Data acquisition was performed using the Perkin Elmer UV WinLab (Ver. 2.85.04) software package. Infrared absorbance spectra were obtained using a Thermo Electron Nicolet Smart Endurance horizontal single bounce, diamond ATR accessory on a Thermo Electron Nicolet Nexus 870 FTIR instrument fitted with a deuterated triglycine sulfate detector (Madison, WI, USA). Spectra were recorded over the 4000–650 cm–1 range at 4 cm–1 resolution for 64 scans with an optical path difference velocity of 0.6329 cm/s. Data acquisition and processing was performed using the OMNIC 7.3 software package. The FTIR peaks were normalised with respect to the main peak at 1045 cm–1.

A light brown precipitate was formed at pH ≥ 5.0 for all the acids with iron. This precipitate was filtered using a polyvinyl chloride membrane filter (5 μm). It was analysed by X-ray powder diffraction (XRD). Sample analysis was performed on a PANalytical X’Pert PRO multi-purpose X-ray diffractometer (Almelo, Netherlands) using Cu Kα radiation (λ = 1.5406 Å) at 40 kV and 40 mA. Patterns were recorded in the 2θ range from 3.5° to 75° with a scan step size of 0.017° and a count time per step of 50 s. Data was acquired and processed using the X’Pert Data Collector 2.2 and MDI Jade 9.0 software packages respectively.

4.3 Results and Discussion

4.3.1 Catalytic and Non-catalytic Oxidation of Caffeic Acid in Aqueous Systems

The average concentration of CaA is approximately 20 mg/L on PJ obtained from burnt cane, although an initial CaA concentration of 200 mg/L (i.e., 1.11 mM) was chosen for the degradation studies in order to account for other phenolics and colour precursors present in cane juice. The degradation of CaA in aqueous solutions

(at pH 3.0, 4.0 and 5.0) at 25 °C was studied with 2.94 or 11.8 mM H2O2.

The absorption spectra obtained for the degradation of CaA is shown in Figure 4.2. Two maxima at 320 nm and 292 nm are attributable to the CaA molecule (1) and the deprotonated caffeate anion (2) respectively as shown in Scheme 4.1 (Cornard et al., 2006).

0.80

0.70 t = 0 min t = 30 min 0.60 t = 60 min 0.50

0.40

0.30

Absorbance(AU) 0.20

0.10

0.00 200 250 300 350 400 Wavelength (nm)

Figure 4.2 Absorption spectra of CaA after the addition of 2.94 mM H2O2 at pH 3.0 at 25 °C.

O O

HO H HO O O + H2O

- HO HO HO

Caffeic Acid Caffeate Anion (1) (2) Scheme 4.1

These two maxima were also present in reaction mixtures containing CaA and Fenton’s reagent at pH 3.0, 4.0 and 5.0. After 60 min, 17% CaA was degraded at pH

3.0 with 2.94 mM H2O2. Working with solutions at pH 4.0 and 5.0, there was no observable CaA degradation. At the higher H2O2 dosage of 11.8 mM, no further reduction in absorbance was noticeable at either maximum even at pH 3.0. However, there appears to be some degradation occurring at lower wavelengths, but this was not conclusive. It is speculated that with the addition of 11.8 mM of H2O2 after the initial reactions between the •OH radicals and CaA, there were subsequent recombination reactions.

The degradation of CaA with Fenton’s reagent monitored at 320 nm is shown in Figure 4.3. The reaction was virtually complete within 5 min. Within 30 min, 86% of CaA was destroyed upon addition of 0.72 mM Fe(II) and 11.8 mM H2O2 at pH 5.0 (Figure 4.3a). At pH 3.0 and 4.0, the degradation of CaA was 62% and 66% respectively. The degradation of the deprotonated caffeate anion was also observed. The degradation ratio of the neutral and anionic forms was approximately 1:1; hence the Fenton’s reagent is capable of attacking both forms of CaA.

The degradation trends were similar for both 2.94 mM and 11.8 mM H2O2 dosages with the latter having a larger decrease in absorbance (Figure 4.3b). At the lower H2O2 concentration, approximately 70% degradation occurred after 30 min. A faster degradation rate was observed at pH 5.0 for both H2O2 dosages despite a higher initial absorbance. The higher initial absorbance is attributable to the chelating ability of Fe(II)/Fe(III) on CaA to produce coloured complexes (Smith, 1983).

No other prominent peaks were observed across the spectral wavelength range during the course of both experiments. This suggests that the degradation products formed from the use of Fenton’s reagent are relatively LMW compounds with weak chromophores or compounds without a chromophore.

In summary, the Fenton process is significantly more effective than H2O2 to degrade CaA in aqueous systems. The Fenton process at 25 °C was optimum at pH 5.0 to degrade CaA in water, which better reflects the pH of sugar cane juices anyway.

(a) 1.00

pH 3.0 0.75 pH 4.0 pH 5.0

0.50

Absorbance(AU) 0.25

0.00 -5 0 5 10 15 20 25 30 35 Time (min)

(b) 1.00

pH 3.0

0.75 pH 4.0 pH 5.0

0.50 Absorbance(AU) 0.25

0.00 -5 0 5 10 15 20 25 30 35 Time (min)

Figure 4.3 Degradation of CaA (measured at 320 nm) using Fenton’s reagent

at different initial pH at 25 °C. Concentrations of H2O2: (a) 11.8 mM and (b) 2.94 mM.

4.3.2 Optimisation of Process Parameters for the Degradation of Caffeic Acid in Sugar Solutions

Section 4.3.1 has shown that CaA can be degraded by the Fenton process. Therefore, the aim of this work was to determine the optimal conditions and develop a model for the degradation of CaA, both in aqueous and sucrose solutions using the Fenton process. Response surface methodology, a powerful statistical tool, was used for the experimental design and development of the model.

Regression Modelling and Statistical Analysis

Central composite design and RSM were used to evaluate the relationships between the response (i.e., % CaA degradation) and the process parameters (i.e., H2O2 dosage, temperature and sucrose concentration). To achieve this, the experimental data obtained from the experimental design using the constraints from Table 4.2, were modelled by the system described through an empirical second-order polynomial function (Montgomery, 2008):

kk (4.2) 2 y0   xi    ii x i     ij x i x ij   i1 i  1 i  1 i  j  1 where, y predicted response (i.e., % CaA degradation)

β0 constant coefficient

βi linear coefficient

βii quadratic coefficient (for the independent factor i)

βij interaction effect coefficient (between independent factors i and j)

xij independent factors (i.e., process parameters shown in Table 4.2)

k number of process parameters investigated

ε random error

The ANOVA results are presented in Table 4.4. The analysis indicated that all independent variables and some of their interactions were significant and contributed to the degradation of CaA by Fenton oxidation. The model F-value of 22.28 implies that the model is significant. There is only a 0.01% chance that a model F-value this large could occur due to noise.

The model for % CaA degradation was improved after the exclusion of insignificant coefficients (Table 4.4) is shown as follows:

CaA degradation (%)

y = 39.84 – 8.47A – 14.13B + 1.75C + 5.83D + 5.15E – 2.90F (4.3) + 5.66G + 5.41AB + 3.20AD + 5.01AE – 2.35BD – 2.62BE – 4.02BF – 1.81BG + 4.09CD – 2.35CF – 1.96DG – 2.55FG + 16.65B2

Based on the coefficients in Equation 4.3, it is evident that % CaA degradation increases with solution pH (C), Fe(II) dosage (D), H2O2 dosage (E) and reaction time (G) but decreases with initial CaA concentration (A), sucrose concentration (B) and reaction temperature (F). Amongst the variables, key interaction effects between initial CaA and sucrose concentrations (AB), CaA and Fe(II) (AD), CaA and H2O2

(AE), sucrose and Fe(II) (BD), sucrose and H2O2 (BE), sucrose and temperature (BF), sucrose and time (BG), pH and Fe(II) (CD), pH and temperature (CF), Fe(II) and time (DG) and temperature and time (FG) are also observed.

The response surface quadratic model diagnostics for % CaA degradation is summarised in Table 4.5. A satisfactory R2 coefficient of 0.87 meant that the model explains 87% of the variability in the data. The predicted R2 of 0.75 is in reasonable agreement with the adjusted R2 of 0.83, and a plot of predicted values of % CaA degradation against the observed values was degenerated as shown in Figure 4.4. As a reasonable linear relationship was obtained, Equation 4.3 (i.e., the quadratic model) is suitable for predicting the % degradation of CaA.

Table 4.4 Analysis of variance (ANOVA) results for response surface quadratic model terms for CaA degradation.*

Source SS df Mean sq. F-value p-value Remarks Model 67,843.69 35 1,938.39 22.28 < 0.0001 Significant A 9,319.02 1 9,319.02 107.11 < 0.0001 Significant B 25,955.48 1 25,955.48 298.33 < 0.0001 Significant C 399.74 1 399.74 4.59 0.0342 Significant D 4,414.70 1 4,414.70 50.74 < 0.0001 Significant E 3,446.42 1 3,446.42 39.61 < 0.0001 Significant F 1,094.64 1 1,094.64 12.58 0.0006 Significant G 4,162.00 1 4,162.00 47.84 < 0.0001 Significant AB 3,642.45 1 3,642.45 43.01 < 0.0001 Significant AC 85.34 1 85.34 0.98 0.3240 AD 1,310.89 1 1,310.89 15.07 0.0002 Significant AE 3,213.26 1 3,213.26 36.93 < 0.0001 Significant AF 153.53 1 153.53 1.76 0.1867 BC 87.00 1 87.00 1.00 0.3194 BD 72.75 1 72.75 0.84 0.3624 BE 707.20 1 707.20 8.13 0.0052 Significant BF 879.30 1 879.30 10.11 0.0019 Significant BG 2,070.87 1 2,070.87 23.80 < 0.0001 Significant BH 419.35 1 419.35 4.82 0.0301 Significant CD 2,145.37 1 2,145.37 24.66 < 0.0001 Significant CE 91.71 1 91.71 1.05 0.3067 CF 705.64 1 705.64 8.11 0.0052 Significant CG 84.64 1 84.64 0.97 0.3260 DE 10.29 1 10.29 0.12 0.7315 DF 15.19 1 15.19 0.17 0.6768 DG 489.94 1 489.94 5.63 0.0193 Significant EF 289.07 1 289.07 3.32 0.0709 EG 50.24 1 50.24 0.58 0.4489 FG 829.72 1 829.72 9.54 0.0025 Significant A2 88.84 1 88.84 1.02 0.3143

B2 644.82 1 644.82 7.41 0.0075 Significant C2 264.65 1 264.65 3.04 0.0838 D2 64.11 1 64.11 0.74 0.3924 E2 298.99 1 298.99 3.44 0.0663 F2 158.59 1 158.59 1.82 0.1796 G2 0.25 1 0.25 0.0029 0.9571 Residual 10,092.40 116 87.00 Lack of fit 10,092.37 107 94.32 26,870 < 0.0001 Significant Pure error 0.032 9 0.0035 Corr. total 77936.09 151 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

Table 4.5 Regression diagnostics for the response surface quadratic model for CaA degradation.

Criteria Standard deviation 9.33 Mean 43.40 Coefficient of variance (CV) (%) 21.49 Predicted residual sum of squares (PRESS) 19,391.88 R2 0.87 Adjusted R2 0.83 Predicted R2 0.75 Adequate precision 21.00

Figure 4.4 Plot of predicted and experimental (actual) values for the degradation (%) of CaA.

Figure 4.5 Normal probability plot of residuals for fitted model using CaA degradation data.

The residuals from the least squares of fit are important for judging model adequacy. Through constructing the plot of studentised residuals versus the normal percentage of probability as shown in Figure 4.5, a check was made for the normality assumption, which was found to be satisfied for the % CaA degradation as the residual plots approximated a straight line.

Interaction Effects between Process Parameters

For the graphical interpretation of the interactions between % CaA degradation and the process parameters, three-dimensional (3D) surface plots of the regression model (Equation 4.3) were used. These plots are shown in Figures 4.6 and 4.7 and some of the interactions are significant as the curvature of the 3D surfaces was obvious.

Influence of Initial CaA Concentration

It is observed that, at a given time, a higher initial CaA concentration results in lower degradation (Figure 4.6a). However, in relation to the reaction rate, a higher initial CaA concentration will result in a higher degradation rate of CaA. In other words, increasing the concentration of CaA involves higher uptake of •OH radicals produced from decomposed H2O2. It is presumed that the degradation efficiency of CaA in a mixture of other phenolic acids (and other juice components) would decrease because of competing reactions between •OH radicals and the other phenolic compounds. This will be further investigated in the next section of this chapter (cf. Section 4.3.3).

Influence of Sucrose Concentration

The influence of sucrose concentration of the oxidation of CaA was investigated as shown in Figure 4b and 4c. The addition of sucrose clearly inhibited the oxidation of CaA. The results show that up to 61% of CaA was degraded at 13% (w/w) sucrose, the concentration typically encountered in factory cane MJs. The

H2O2 concentration has a greater negative influence on the amount of CaA degraded relative to temperature effect. Morelli et al., (2003) investigated whether the •OH radicals generated by the Fenton process were scavenged by simple carbohydrates. Their results not only show the scavenging ability of simple sugars but show that disaccharides such as maltose and sucrose were more effective than monosaccharides in removing •OH radicals. So, the reduced effectiveness of the Fenton process for CaA degradation in the presence of sucrose is related to the reduction of available •OH radicals.

Design-Expert® Software Design-Expert® Software % CA Degradation % CA Degradation 96.8358 96.8358 0.146016 (a) 0.146016 (b) X1 = A: CA X1 = B: Sucrose X2 = D: Fe(II) X2 = E: Peroxide

Actual Factors Actual Factors B: Sucrose = 0.00 110 A: CA = 1.11 85 C: pH = 5.00 C: pH = 5.00 E: Peroxide = 6.62 D: Fe(II) = 0.45

F: Temp. = 35.00 F: Temp. = 35.00 G: Time = 120.00 95 G: Time = 120.00 70

80 55

65 40

50 25

% CA Degradation %Degradation CA %Degradation CA

% CaA Degradation % CaA Degradation % CaA

0.72 1.11 11.03 34 0.58 0.90 8.82 26 0.45 0.70 6.62 17 Design-Expert® Software 0.31 Design-Expert® 0.49 Software 4.41 9 D: Fe(II) A: A: CaA CA E:E: Peroxide H2O2 B: Sucrose % CA Degradation 0.18 0.28 % CA Degradation 2.21 0 96.8358 96.8358

0.146016 0.146016

X1 = B: Sucrose (c) X1 = C: pH (d) X2 = F: Temp. X2 = D: Fe(II) Actual Factors 95 Actual Factors A: CA = 1.11 A: CA = 1.11 85 C: pH = 5.00 B: Sucrose = 0 D: Fe(II) = 0.45 E: Peroxide = 6.62 E: Peroxide = 6.62 F: Temp. = 35 G: Time = 120.00 80 G: Time = 120.00 75

65 65

50 55

35 45

% CaA Degradation % CaA Degradation % CaA

% CA Degradation %Degradation CA

95 34 0.72 6.50 80 26 0.58 5.75 65 17 0.45 5.00 F: Temp. 50 9 B: Sucrose D: Fe(II) 0.31 4.25 C: pH 35 0 0.18 3.50 Figure 4.6 Three-dimensional surface plots of CaA degradation (%) as a

function of (a) CaA and Fe(II); (b) sucrose and H2O2; (c) sucrose and temperature; and (d) pH and Fe(II). Variables: CaA

(1.11 mM); sucrose (0% (w/w)); pH (5.0); Fe(II) (0.45 mM); H2O2 (6.62 mM); temperature (35 °C) and time (120 s).

Influence of Solution pH

Experiments were conducted at initial pH of 3.5, 5.0 and 6.5. Maximum CaA degradation was observed at pH 4.5–5.5 (Figure 4.6d and 4.7a). This is in line with the results of Tang and Huang (1996) and Deng (2007). At higher pH values, Fe(III) produced from Fe(II) oxidation precipitates as oxyhydroxides instead of being regenerated back to Fe(II). Hence, the total amount of Fe(II) required to catalyse the decomposition of H2O2 to produce the reactive •OH radicals is reduced (Cortez et al., 2011). As such, the lower degradation of CaA at pH 6.5 is mainly attributable to the generation of reduced amounts of •OH radicals in comparison with •OH radicals generated at pH 3.5–5.5. Also, H2O2 is unstable under alkaline conditions and itself may rapidly decompose to water and oxygen (Chang et al., 2010).

Influence of Fe(II) and H2O2

The interactive effects of both Fe(II) and H2O2 are shown in the surface plot of

Figure 4.7b. A greater proportion of CaA is degraded with increasing Fe(II) and H2O2 concentrations. The availability of increasing amounts of H2O2 will result in an increase in the proportion of •OH radicals formed as Fe(II) can readily be generated by Fe(III). Also, increasing the concentration of Fe(II) will result in an increase in the amount of H2O2 formed. However, there is an optimum molar ratio of Fe(II) to H2O2 required for the generation of •OH radicals. In this study, the optimum molar ratio of

Fe(II) to H2O2 for the degradation of CaA is 1:13. The value mentioned in the literature varied from 1:1 to 1:400 as different feed compositions and operating conditions were examined (Tang and Huang, 1997; Kitis et al., 1999; de Souza et al., 2006). According to the stoichiometric equation (Equation 4.4), a molar ratio of 1:18 is required for the complete mineralisation of CaA by H2O2.

C9H8O4 + 18H2O2  9CO2 + 22H2O (4.4)

So, the present result confirms the catalytic influence of Fe(II) on CaA degradation.

Design-Expert® Software Design-Expert® Software % CA Degradation 96.8358 % CA Degradation 96.8358 0.146016 (a) 0.146016 (b) X1 = C: pH X2 = E: Peroxide X1 = D: Fe(II) X2 = E: Peroxide Actual Factors 90 A: CA = 1.11 Actual Factors 90 B: Sucrose = 0 A: CA = 1.11 D: Fe(II) = 0.45 B: Sucrose = 0 C: pH = 5.00

F: Temp. = 35 G: Time = 120.00 75 F: Temp. = 35 G: Time = 120.00 75

60 60

45 45

30 30

% CA Degradation %Degradation CA

% CaA Degradation % CaA Degradation % CaA

11.03 6.50 11.03 0.72 8.82 5.75 8.82 0.58 6.62 5.00 6.62 0.45 Design-Expert® Software 4.41 4.25 E:E: Peroxide H2O2 Design-Expert® SoftwareC: pH E:E: Peroxide H2O2 4.41 0.31 D: Fe(II) % CA Degradation 2.21 3.50 2.21 0.18 96.8358 % CA Degradation 96.8358 0.146016 0.146016 X1 = E: Peroxide (c) (d) X2 = F: Temp. X1 = E: Peroxide X2 = G: Time Actual Factors Actual Factors A: CA = 1.11 90 A: CA = 1.11 85 B: Sucrose = 0 B: Sucrose = 0 C: pH = 5.00 C: pH = 5.00 D: Fe(II) = 0.45 D: Fe(II) = 0.45 G: Time = 120.00 75 F: Temp. = 35 70

60 55

45 40

30 25

% CaA Degradation % CaA Degradation % CaA

% CA Degradation %Degradation CA

95 11.03 120 11.03 80 8.82 93 8.82 65 6.62 65 6.62 50 4.41 38 4.41 F: Temp. E:E: Peroxide H2O2 G: Time E:E: Peroxide H2O2 35 2.21 10 2.21 Figure 4.7 Three-dimensional surface plots of CaA degradation (%) as

function of (a) pH and H2O2; (b) Fe(II) and H2O2; (c) H2O2 and

temperature; and (d) H2O2 and time. Variables: CaA (1.11 mM);

sucrose (0% (w/w)); pH (5.0); Fe(II) (0.45 mM); H2O2 (6.62 mM); temperature (35 °C) and time (120 s).

Influence of Temperature

The effect of temperature on CaA degradation was studied at 35 °C, 65 °C and 95 °C (Figures 4.6c and 4.7c). Degradation of CaA occurred at a faster rate with increasing temperature. This is because raising the temperature increased the

decomposition rate of H2O2 and hence the formation of reactive •OH radicals (Sun et

al., 2009). However, the decomposition of H2O2 is not directly linked to the amount of CaA degraded, because in addition to the formation of •OH radicals, at higher

temperatures non-reactive species such as H2O and O2 are formed (Rodrigues et al., 2009a). These counteractive effects are clearly illustrated in Figure 4.6c where maximum degradation of CaA is obtained at 35 °C. However, the degradation of CaA at 95 °C is more effective than at 65 °C because of the contributing effect resulting from the thermal degradation of CaA. It has been reported by Kulik et al. (2011) that the decarboxylation of CaA and other HCAs (e.g., pCoA and FeA) occurs at temperatures > 70 °C.

Influence of Time

Figure 4.7d shows that reaction time has a positive effect on the degradation of CaA. Maximum CaA degradation is achieved within 120 s, as there was no increase thereafter. The short degradation time obtained in this study implies that the Fenton oxidation process will be suitably applied in a sugar factory for the degradation of CaA and other phenolic compounds.

Sugars Analysis

During raw sugar manufacture, sucrose loss through inversion to glucose and fructose, and degradation to organic acids are minimised to maintain sugar yield by working at selected pH and temperatures. Sucrose degradation by Fenton oxidation was evaluated by HPAEC-PAD (cf. Appendices, Table A1.2). The results showed minimal losses of sucrose (< 0.01%) were present in reactions carried out at 35 °C after 10 min. Conversion of sucrose to glucose and fructose was observed in reactions carried out at 65 °C and 95 °C (< 1.0%), the latter showing higher amounts of reducing sugars. This means that the Fenton process may only find applications in the sugar manufacturing process at far lower temperatures.

Model Validation and Optimisation

Numeric optimisation was used to determine the optimum process parameters for CaA degradation. The optimum and worst conditions for CaA degradation were obtained on the basis of the model (Equation 4.3) and the desirability function. The desirability function is expressed as a numeric value and denotes the degree of importance in obtaining the desired target response value. To validate the accuracy and robustness of the predicted model and the reliability of the obtained conditions, additional experiments were carried out under those conditions, as well as randomly selected conditions within the ranges investigated. As shown in Table 4.6, the experimental values of the optimum and worst conditions agree well with the predicted values.

Table 4.6 Optimised conditions under specified constraints for the degradation of CaA and model verification.

Experiments* 1 2 3 4 CaA (mM) 1.11 1.11 1.11 1.11 Sucrose (% (w/w)) 0 0 14 34 pH 5.1 5.5 5.1 3.5 Fe(II) (mM) 0.68 0.72 0.64 0.18

H2O2 (mM) 8.88 9.44 8.47 2.21 Temperature (°C) 95 35 95 64 Time (s) 120 120 120 120 Observed degradation (%) 91 80 59 10 Predicted degradation (%) 92 85 62 11 Error 1.00 5.00 3.00 1.00 Standard deviation 0.71 3.53 2.12 0.71 Desirability function 0.95 0.87 0.64 0.89 *Experiments: (1) Optimum; (2) Optimum without thermal degradation; (3) Optimum with 14% (w/w) sucrose; and (4) worst case.

The experimental values of randomly selected conditions are shown in Table 4.7. The low error in the experimental and predicted values indicates good agreement of the results. The desirability functions obtained with solutions containing sucrose were comparatively less than the values obtained in the absence of sucrose (Table 4.7).

Table 4.7 Model verification of optimised conditions under randomly specified constraints for CaA degradation.

Experiments 5 6 7 CaA (mM) 1.11 1.11 1.11 Sucrose (% (w/w)) 0 14 34 pH 5.0 5.0 5.0 Fe(II) (mM) 0.72 0.72 0.36

H2O2 (mM) 4.41 2.21 2.21 Temperature (°C) 35 35 76 Time (s) 120 120 120 Observed degradation (%) 67 31 31 Predicted degradation (%) 69 33 29 Error 2.00 2.00 2.00 Standard deviation 1.41 1.41 1.41 Desirability function 0.84 0.57 0.55

The applicability of the proposed model was also investigated using the raw sugar processing constraints of a typical Australian sugar cane factory. On the basis of the colour profile across the sugar manufacturing stage as shown in Figure 2.8 (Eggleston et al., 2003), to reduce colour in raw sugar, colour removal strategies should be targeted at MJ (i.e., juice prior to incubation), PJ (i.e., prior to liming) and/or on juices during the evaporation stage. From the information obtained from the model, sugar process streams operating at temperatures > 95 °C (because of

sucrose degradation) and/or at sucrose concentrations > 34% (w/w) may not be suitable to be treated with the Fenton oxidation process.

Table 4.8 shows results obtained for synthetic juice solutions, under processing conditions similar to that of MJ, PJ and juice from the third effect of a quintuple evaporator set. It shows that the best result is obtained with MJ followed by PJ. A higher error in the experimental and predicted values for the optimised conditions for the optimised conditions for the third effect juice, compared with the conditions of other juice process streams, was observed. It is probable that solution pH may have contributed to the inaccuracy of the prediction as it is outside the range used to develop the proposed model (Equation 4.3).

Table 4.8 Model verification of optimised conditions in synthetic juice solutions under specified constraints of selected sugar process streams for CaA degradation.

Experiments Mixed Juice Primary Juice Third Effect CaA (mM) 1.11 1.11 1.11 Sucrose (% (w/w)) 13 17 30 pH 5.4 5.4 6.8 Fe(II) (mM) 0.68 0.66 0.64

H2O2 (mM) 8.67 8.62 8.59 Temperature (°C) 35 76 94 Time (s) 120 120 120 Observed degradation (%) 61 49 27 Predicted degradation (%) 62 52 41 Error 1.00 3.00 14.0 Standard deviation 0.71 2.12 9.90 Desirability function 0.64 0.54 0.42

The aforementioned results have shown that the Fenton process is reasonably effective in degrading CaA in sucrose solutions. The reduced effectiveness of the Fenton process for the degradation of CaA for these systems is related to the reduction of available •OH radicals by the scavenging action of sucrose (Morelli et al., 2003). Despite the free radical scavenging ability of sucrose, minimal losses of sucrose (< 0.01%) were obtained after 2 min of treatment.

4.3.3 Degradation of Hydroxycinnamic Acid Mixtures

The work described in this section builds on the results previously discussed in Sections 4.3.1 and 4.3.2 by examining the degradation of a mixture of the three main HCAs present in Australian sugar cane juice (viz., CaA, pCoA and FeA) using the Fenton process in the presence and absence of sucrose. The results obtained were used to develop a model for the degradation of each individual HCA within a mixture as well as a model for total HCA degradation. No previous study has reported on the optimisation of the degradation process of individual acids within a mixture of other phenolic acids by the Fenton process, nor examined the interactive effects of various operating parameters on the degradation of each acid.

Optimal Data Transformation and Test for Normality

Rotatable CCD and RSM were used to investigate the relationships between the response factors (dependent variables) and the process parameters (independent variables). In order to achieve this, an empirical second-order polynomial function identical to Equation 4.2, for each response factor was used to fit the experimental results obtained.

The assumption used to estimate the response based on the model given in Equation 4.2 is that the random error terms (ε) for all levels of the independent factors are distributed normally and independently with a mean zero and a common variance (Tunali and Batmaz, 2000). Graphical residual analysis was used to verify the adequacy of different aspects of the model. The residuals from the least squares fit are important for judging model adequacy. A normal probability plot of residuals

100

based on the experimental data obtained for CaA degradation (Figure 4.8) indicates a non-linear pattern in the middle of the trend line, and short tails with the first and last few points showing increasing departure from the trend line.

Figure 4.8 Normal probability plot of residuals for fitted model using CaA degradation data before power transformation.

To address the non-linearity of these plots, the Box-Cox power transformation was used to improve linearity. The power transformation on the predicted response can be described as follows (Box and Cox, 1964):

 y 1 (4.5)     0 y     ln y   0 where λ indicates the power to which all data should be raised. The initial value of λ in the standard quadratic function (i.e., Equation 4.2) is λ = 1.00.

101

To determine the λ value for each response, a Box-Cox plot was used as a guide for the selection of the optimised λ value for the power transformation of the experimental data. Figures 4.9 and 4.10 show the Box-Cox plots for each response investigated. From the Box-Cox plots for the degradation of pCoA (Figure 4.9b) and FeA (Figure 4.10a), the recommended λ values ranged from 0.70 to 2.40 and 0.59 to 2.23, respectively at a 95% confidence interval. On the other hand, the λ value range within the 95% confidence interval were not shown for the degradation data of CaA (Figure 4.9a) and total HCA (Figure 4.10b), due to the values being outside the λ = ± 3.00 limits. Hence, the optimum λ values used to transform the CaA and total HCA degradation were both maximised at λ = 3.00. For pCoA and FeA degradation, the optimum λ values were determined by observing the minimum of the curve, which was 1.56 and 1.43 respectively.

Using the optimised λ values, the normal probability plot for each response surface model shown in Figures 4.11 and 4.12 indicate improved linearity of data points. There are only a minimal number of data points deviating from the line of fit. The data for all fitted response surface models show good correspondence to a normal distribution and validated the normality assumption.

102

(a)

(b)

Figure 4.9 Box-Cox plots of (a) CaA and (b) pCoA degradation data for the determination of the optimised power transformed response surface models.

103

(a)

(b)

Figure 4.10 Box-Cox plots of (a) FeA and (b) total HCA degradation data for the determination of the optimised power transformed response surface models.

104

(a)

(b)

Figure 4.11 Normal probability plots of residuals for fitted model using (a) CaA and (b) pCoA degradation data after power transformation.

105

(a)

(b)

Figure 4.12 Normal probability plots of residuals for fitted model using (a) FeA and (b) total HCA degradation data after power transformation.

106

Regression Modelling and Statistical Analysis

On the basis of the sequential model sum of squares (Type I), the power transformed response surface models for CaA (y1), pCoA (y2), FeA (y3) and total

HCA (y4) degradation were selected based on the highest order polynomial, where the additional model terms were significant and the models were not aliased. The data obtained for all four responses fit a quadratic polynomial function.

In each model, there are some unimportant model terms that should be removed to improve the accuracy of fitting. In this study, the significant coefficients of the models were identified using ANOVA statistics and stepwise regression. Stepwise regression involves the selection of the most appropriate independent variables for a regression model. In this case, a subset of variables from the full set is determined. The stepwise regression method is a combination of forward selection and backward elimination (regression) statistical methods. From the full set of available variables, the stepwise procedure builds or depletes the regression model, one variable at a time. Sequentially, variables are added (i.e., forward selection) at an alpha-to-enter significance level of 0.1 and removed (i.e., backward elimination) at an alpha-to-exit significance level of 0.1 until an added variable does not yield a Student t–test probability (p–value) of ≤ 0.1. The chosen stepwise alpha range applied to all four response surface models should result in final models with significant model terms included at the approximate 95% confidence level.

The ANOVA results for the partial sum of squares (Type III) for the four response surface reduced quadratic models after stepwise regression are shown in Tables 4.9–4.12. The analysis indicates that most independent variables and some of the interactions are significant and contribute to the degradation of the HCAs. The model F-values of 43.30, 44.14, 88.37 and 19.03 for CaA, pCoA, FeA and total HCA degradation respectively, imply that the models are significant. There is only a 0.01% chance that a model F-value this large could occur due to noise. The lack-of-fit F-values of 1.40 and 1.38 for the pCoA and FeA models in that order imply that the lack of fit is not significant relative to the pure error. There is a 21% and 23% chance respectively that the lack-of-fit F-values this large would occur due to noise. Non- significant lack-of-fit is good as it confirms the predictability of the model. On the other hand, the lack-of-fit F-values of 34.50 and 2.65 for CaA and total HCA models

107

respectively, imply that the lack-of-fit is significant. A significant lack-of-fit is undesirable as the proposed models do not fit the data well. This may be because the experimental data of CaA degradation showed little variation (p < 0.0001) under the constraints of the experimental design compared to pCoA and FeA degradation data. Therefore, it also affected the lack-of-fit of the total HCA degradation model (p = 0.0138). Despite this, further ANOVA statistics (discussed later) demonstrate that the data is suitable for the modelling and prediction of CaA and total HCA degradation.

Model terms with a p–value < 0.0500 indicate model terms are significant at the 95% confidence level. Values > 0.1000 indicate the model terms are insignificant at the 90% confidence level and are removed from the proposed models via stepwise regression, with the exception of the first-order temperature model term for all models. Temperature was regarded as statistically insignificant but was added to all models to make each model hierarchical. In other words, parent (i.e., first-order) model terms are added to the model to complete the family of any significant higher- order (i.e., second-order) model terms.

The independent variables in the models were initial total HCA concentration, initial sucrose concentration, solution pH and reaction temperature; and were coded A, B, C and D respectively. The final empirical quadratic equations in terms of coded factors for each response are as follows:

CaA degradation (%)

3 5 5 (y1) = 7.459 × 10 – 22685.04A + 87649.64B –1.893 × 10 C (4.6) – 2787.88D + 38875.43BC + 25613.66BD – 48866.47B2 – 55229.82C2 + 21771.66D2 pCoA degradation (%)

1.56 (y2) = 452.03 – 25.39A – 112.96B + 56.46C – 9.29D + 25.34AB (4.7) + 24.11CD + 51.51B2 – 13.11C2

108

Table 4.9 Results of ANOVA for model terms of the response surface reduced quadratic model for CaA degradation.*

Source SS df Mean sq. F-value p-value Remarks Model 2.49 × 1012 9 2.77 × 1011 43.30 < 0.0001 Significant A 2.47 × 1010 1 2.47 × 1010 3.87 0.0556 B 3.69 × 1011 1 3.69 × 1011 57.74 < 0.0001 Significant C 1.72 × 1012 1 1.72 × 1012 269.21 < 0.0001 Significant D 3.73 × 1008 1 3.73 × 1008 0.058 0.8101 Insignificant BC 4.84 × 1010 1 4.84 × 1010 7.57 0.0086 Significant BD 2.10 × 1010 1 2.10 × 1010 3.29 0.0766 B2 1.15 × 1011 1 1.15 × 1011 17.95 0.0001 Significant C2 1.46 × 1011 1 1.46 × 1011 22.93 < 0.0001 Significant D2 2.28 × 1010 1 2.28 × 1010 3.56 0.0657 Residual 2.81 × 1011 44 87.00 Lack of fit 2.66 × 1011 15 94.32 34.50 < 0.0001 Significant Pure error 1.49 × 1010 29 0.0035 Corr. total 2.77 × 1012 53 Criteria Standard deviation 8.00 × 1004 Mean 6.73 × 1005 CV (%) 11.88 PRESS 4.54 × 1011 R2 0.90 Adjusted R2 0.88 Predicted R2 0.84 Adequate precision 23.28 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

109

Table 4.10 Results of ANOVA for model terms of the response surface reduced quadratic model for pCoA degradation.*

Source SS df Mean sq. F-value p-value Remarks Model 1.00 × 106 8 1.25 × 105 44.14 < 0.0001 Significant A 3.09 × 104 1 3.09 × 104 10.90 0.0019 Significant B 5.49 × 105 1 5.49 × 105 193.36 < 0.0001 Significant C 1.53 × 105 1 1.53 × 105 53.93 < 0.0001 Significant D 4.14 × 103 1 4.14 × 103 1.46 0.2333 Insignificant AB 2.05 × 104 1 2.05 × 104 7.24 0.0100 Significant CD 1.81× 103 1 1.81× 103 6.56 0.0139 Significant B2 1.10 × 105 1 1.10 × 105 38.77 0.0001 Significant C2 8.78 × 103 1 8.78 × 103 3.09 0.0855 Residual 1.25 × 105 44 87.00 Lack of fit 5.55 × 104 16 94.32 1.40 0.2121 Insignificant Pure error 6.94 × 104 28 2.48 × 103 Corr. total 1.13E+06 52 Criteria Standard deviation 53.27 Mean 487.18 CV (%) 10.93 PRESS 1.82 × 105 R2 0.89 Adjusted R2 0.87 Predicted R2 0.84 Adequate precision 27.21 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

110

Table 4.11 Results of ANOVA for model terms of the response surface reduced quadratic model for FeA degradation.*

Source SS df Mean sq. F-value p-value Remarks Model 3.30 × 105 9 3.67 × 104 88.37 < 0.0001 Significant A 1.03 × 104 1 1.03 × 104 24.74 < 0.0001 Significant B 2.02 × 105 1 2.02 × 105 485.42 < 0.0001 Significant C 3.16 × 104 1 3.16 × 104 76.07 < 0.0001 Significant D 528.86 1 528.86 1.27 0.2656 Insignificant AB 1.92 × 104 1 1.92 × 104 46.22 < 0.0001 Significant BD 2.70 × 103 1 2.70 × 103 6.49 0.0146 Significant CD 1.24 × 103 1 1.24 × 103 2.98 0.0915 B2 2.89 × 104 1 2.89 × 104 69.46 < 0.0001 Significant D2 4.57 × 103 1 4.57 × 103 11.01 0.0019 Significant Residual 1.74 × 104 42 415.39 Lack of fit 7.57 × 103 15 504.72 1.38 0.2262 Insignificant Pure error 9.88 × 103 27 365.76 Corr. total 3.48 × 105 51 Criteria Standard deviation 20.38 Mean 308.78 CV (%) 6.60 PRESS 2.81 × 104 R2 0.95 Adjusted R2 0.94 Predicted R2 0.92 Adequate precision 37.35 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

111

Table 4.12 Results of ANOVA for model terms of the response surface reduced quadratic model for total HCA degradation.*

Source SS df Mean Sq. F-value p-value Remarks Model 2.17 × 1011 10 2.17 × 1010 19.03 < 0.0001 Significant A 2.46 × 1010 1 2.46 × 1010 21.57 < 0.0001 Significant B 1.16 × 1011 1 1.16 × 1011 102.20 < 0.0001 Significant C 5.41 × 1009 1 5.41 × 1009 4.75 0.0349 Significant D 8.13 × 1008 1 8.13 × 1008 0.71 0.4028 Insignificant AB 1.11 × 1010 1 1.11 × 1010 9.79 0.0032 Significant BC 9.27 × 1009 1 9.27 × 1009 8.14 0.0067 Significant CD 6.14 × 1009 1 6.14 × 1009 5.39 0.0252 Significant B2 4.20 × 1009 1 4.20 × 1009 3.68 0.0617 C2 1.89 × 1010 1 1.89 × 1010 16.62 0.0002 Significant D2 6.11 × 1009 1 6.11 × 1009 5.37 0.0254 Significant Residual 4.78 × 1010 42 415.39 Lack of Fit 2.72 × 1010 14 504.72 2.65 0.2262 Significant Pure Error 2.06 × 1010 28 365.76 Corr. total 2.64 × 1011 52 Criteria Standard Deviation 3.37 × 1004 Mean 2.68 × 1005 CV (%) 12.58 PRESS 8.04 × 1010 R2 0.82 Adjusted R2 0.78 Predicted R2 0.70 Adequate Precision 15.61 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

112

FeA degradation (%)

1.43 (y3) = 274.79 – 14.82A – 69.60B + 25.99C – 3.36D + 24.97AB (4.8) – 9.36BD + 6.34 CD + 26.42B2 + 9.46D2

Total HCA degradation (%)

= 2.670 × 105 – 22911.69A – 49869.11B – 10752.76C (4.9) 3 (y4) – 4168.93D + 19018.79AB + 17344.68BC +14113.81CD + 9351.73B2 – 19861.54C2 + 11289.82D2

The predicted R2 values of all response surface models are in reasonable agreement with the adjusted R2 values, which show that the fitted models are adequate. The accuracy of the models is shown in Figures 4.13 and 4.14, which compares the predicted responses against the experimental data. As reasonable linear relationships were obtained, Equations 4.6–4.9 are suitable for predicting the degradation of CaA, pCoA, FeA and total HCA, respectively.

On the basis of the coefficients of the first-order model terms in Equations 4.6–4.9, it is evident that the degradation efficiency of all HCAs decreases with initial total HCA concentration (A). Sucrose concentration (B) is the most influential parameter with the highest coefficient in all equations and shows a negative influence in pCoA and FeA degradation but a positive influence for CaA degradation. Also, the degradation efficiency of pCoA and FeA increases with solution pH (C) but the opposite is observed for CaA. Temperature (D) has a negative effect on all responses but its minuscule coefficient has little effect on the respective response. Hence, this model term is statistically insignificant and was only included in all of the equations to make the models hierarchical.

For the degradation of mixtures (i.e., the acids combined), there are strong interactions between total HCA concentration and sucrose (AB); sucrose and pH (BC); and pH and temperature (CD).

113

(a)

(b)

Figure 4.13 Plots of predicted response and experimental (actual) values for the degradation (%) of (a) CaA and (b) pCoA.

114

(a)

(b)

Figure 4.14 Plots of predicted response and experimental (actual) values for the degradation (%) of (a) FeA and (b) total HCA.

115

Perturbation Analysis

Perturbation plots were analysed in order to further identify the most influential variables on the degradation of each HCAs investigated in this study (Figure 4.15). Sucrose concentration and solution pH appeared to be the most influential parameters. Temperature showed an insignificant effect as expected, whilst the initial total HCA concentration exhibited a consistent effect for the degradation of each HCA.

As shown in Figure 4.15, the higher the initial total HCA concentration, the lower the amount of each HCA degraded. The reason behind the decrease in degradation efficiency is simply due to a higher uptake of •OH radicals by the increased amounts of HCA molecules.

The presence of sucrose significantly affected the degradation efficiency of the HCAs. The fate of sucrose during the degradation process was evaluated by HPAEC- PAD (cf. Appendices, Table A1.3). The results showed up to 0.01% sucrose loss due to complete mineralisation, as no glucose and/or fructose are detected. This is related to the effective scavenging ability of sucrose in removing •OH radicals (Morelli et al., 2003), and accounts for the decrease in degradation efficiency with increasing in sucrose concentration for pCoA and FeA (Figures 4.15b and 4.15c), but not for CaA (Figure 4.15a). The reason for the increased degradation efficiency of CaA with increasing sucrose concentration, may be related as will be shown in the next subsection, to a strong association between CaA and sucrose which increased with increasing sucrose concentration.

Degradation of CaA decreases with increasing pH whereas the opposite was observed for pCoA and FeA degradation. The reason for the results obtained with pCoA and FeA is not known but may be related to the various species that exist in the acid-base equilibria that influences the logarithmic acid dissociation constants (pKa’s) of these acids.

116

(a)

(b)

(c)

117

Complex Formation

In order to obtain insights into the apparent differences in the behaviour among the three HCAs, the UV/Vis spectra of the individual acids, mixtures of each acid with Fe(II) and mixtures of each acid with Fe(II) and sucrose at pH 4.0 to 6.0 were obtained. The UV/Vis spectra obtained with mixtures of FeA or mixtures of pCoA were not dissimilar to that of their corresponding acids. However, as Figure 4.16 shows, there is a significant difference between the spectra of CaA with Fe(II) and those spectra without Fe(II). In these acidic conditions, Fe(II) and Fe(III) will be present in equilibrium (Morgan and Lahav, 2007). The change in the profile of the spectra is likely due to complexation between Fe(III) and CaA, as shown in the spectra; similar to that obtained for aluminium-caffeic acid in aqueous acidic solutions by Cornard and co-workers (2006). In fact, Hynes and O’Coinceanainn (2004) have reported the formation of 1:1 complex between Fe(III) and CaA at pH between 1.0 and 2.5 (Scheme 4.2). Moreover, previous studies have shown the accelerated decomposition of H2O2 to •OH radicals by Fe(III) complexes of analogous phenolic acids (Rivas et al., 2002).

O O

HO 2+ O OH [Fe(H2O)5(OH)] OH Fe + H HO O Scheme 4.2

There is a shape drop in peak intensities at pH ≥ 5.0 for CaA and Fe(III) mixtures (Figure 4.16b), likely to be associated with increased complex formation due to increasing amounts of caffeate ions with pH rise. As the pKa1 of CaA is 4.38, there is an increasing amount of deprotonation with increasing pH (Adams et al., 2002). The drop in intensity may also be due to the removal of CaA by adsorbing onto the iron precipitate formed under these pH conditions. The spectra of Figure 4.16b also show that there was no change in the shape of the curves with increasing pH, so it is probable that only one type of complex is formed between Fe(III) and CaA under the conditions investigated.

118

(a)

) 25 3

10 pH 4.0

x x ) ( )

1 pH 4.5

– 20

cm pH 5.0 1 – pH 5.5 15 pH 6.0

Molar Absorptivity (L mol (L Molar Absorptivity 0 200 250 300 350 400 Wavelength (nm)

(b)

) 3

10 pH 4.0 x x

) ( ) pH 4.5 1

– 20

pH 5.0

cm 1 – pH 5.5 15 pH 6.0

Molar Absorptivity (L mol (L MolarAbsorptivity 0 200 250 300 350 400 Wavelength (nm)

Figure 4.16 Effect of pH (pH 4.0–6.0) on the absorption spectra of CaA (0.055 mM) at 25 °C: (a) in the absence and (b) in the presence of Fe(II) (0.04 mM).

119

The CaA mixtures were further characterised using ATR-FTIR spectroscopy. From the FTIR data, a number of bands were used to monitor changes in CaA as a result of the presence of Fe(II), and the presence of Fe(II) and sucrose. The spectral bands of CaA and sucrose solutions, and CaA mixtures containing Fe(II) or Fe(II) and sucrose are given in Table 4.13. Spectral bands were assigned based on literature data for CaA (Sánchéz-Cortés and García-Ramos, 1999; Dürüst et al., 2001; Machado et al., 2009; Świsłocka, 2013), similar phenolic acids (Dobson and McQuillan, 2000; Hanna and Quilès, 2011; Kalinowska et al., 2011; Świsłocka et al., 2012) and sucrose (Vasko et al., 1971; Huvenne et al., 1981; Kodad et al., 1994; Kačuráková and Mathlouthi, 1996; Max and Chapados, 2001). Bands attributable to aromatic ring vibrations are numbered using the Wilson notation adapted by Varsányi (1974). The main differences between the spectrum of CaA and that of Fe(II)–CaA are shown in –1 Figure 4.17. The ν(CC)ar aromatic bands (i.e., 8a and 19a) that occur at 1554 cm and 1483 cm–1 (Świsłocka, 2013) are of increased intensity in the Fe(II)–CaA mixture than that of CaA (Figure 4.17). The peak at ~1386 cm–1 associated with

ν(CC) + β(OH)ar (i.e., 14) (Świsłocka, 2013) is also of higher intensity in the spectrum containing both Fe(II) and CaA. These increases in intensity may be attributed to complex formation between the aromatic –OH group in CaA and Fe (III) (Hanna and Quilès, 2011). The peak at 1275 cm–1 attributable to ν(C–OH) for CaA (Yost et al., 1990; Machado et al., 2009) has shifted to a lower wavelength of 1265 cm–1 with increase in intensity. This is a further confirmation of a strong association between Fe(III) and CaA and that the complex formed is between Fe(III) and the phenolic hydroxyl group (Rivas et al., 2002; Hynes and O'Coinceanainn, 2004). There was no change in the band at 1672 cm–1 associated with ν(C=O) implying no evidence of Fe(III) bonding to the carboxylic acid group of CaA. Previous works have shown that with other phenolic acids, linkages are formed with their carboxylic acid groups (Hanna and Quilès, 2011; Kalinowska et al., 2011; Świsłocka et al., 2012).

120

Table 4.13 Wavenumbers (cm–1) of selected bands from ATR-FTIR spectra of CaA solution and CaA mixtures containing Fe(II) and/or sucrose at pH 5.5 and 25 °C.

CaA mixtures CaA Fe(II) Fe(II)/Sucrose Sucrose Band assignments* 3401 3495 ν(OH)

3274 3247 ν(OH)ar 3182 3113 ν(OH)

2981 2981 2980 2980 ν(CH)C=C + ν(CH) 2921 2933 2933 ν(CH) 20a 2900 2900 ν(CH) 2854 2852 ν(OH) 1672 1672 1669 1674 ν(C=O)

1618 1608 1611 1619 ν(CC)C=C

1554 1550 1567 1578 ν(CC)ar 8a

1524 1524 ν(CC)ar 8b

1483 1483 ν(CC)ar 19a

1454 1454 1454 1454 ν(CC)ar 19b 1426 1426 β(COH)

1386 1388 1377 1377 ν(CC) + β(OH)ar 14

1328 1329 1332 1332 β(CH)C=C 1275 1265 1274 1266 ν(C–OH) 1210 1210 β(CH) 1160 1160 β(CH) 18a 1118 1118 β(CH) 18b 1085 1085 β(OH)

1045 1045 1045 1045 γ(CH)C=C + γ(CH) 17b 1018 1018 ν(C–O) 998 998 β(COH) 927 927 ν(CC) 877 877 876 876 830 830 β(CCH)

121

1 1

– –

–

– –

cm cm

1483 1386

1554

1265 1672 1672

(a) Absorbance (Arbitrary Units) Absorbance(Arbitrary

(b)

1800 1600 1400 1200 1000 800 Wavenumber (cm–1)

Figure 4.17 Normalised ATR-FTIR spectra of CaA solutions at 25 °C after subtraction of acetate buffer (pH 5.5): (a) in the absence and (b) in the presence of Fe(II).

The spectrum for CaA, Fe (II) and sucrose (Figure 4.18) show that the broad band that occurs at 3495 cm–1 ν(OH) (Max and Chapados, 2001) which is associated with sucrose has shifted by 94 cm–1 to a lower wavenumber of 3401 cm–1. This implies hydrogen-bonding interactions between CaA, Fe(III) and sucrose and could well explain why CaA degradation increases with increasing sucrose concentration (Gilfillan et al., 2012). These interactions provide supporting evidence of the differences in the degradation behaviour of CaA and the other two HCAs (viz., pCoA and FeA).

122

–

3401 3401 3495 3495

(a)

Δν = 94 cm–1

(b) Absorbance (Arbitrary Units) Absorbance(Arbitrary

4000 3600 3200 2800 2400 Wavenumber (cm-1)

Figure 4.18 Normalised ATR-FTIR spectra of CaA solutions containing sucrose at 25 °C after subtraction of acetate buffer (pH 5.5): (a) in the absence and (b) in the presence of Fe(II).

Response Surface Analysis

Graphical representations of the regression model in the form of 3D surface plots were used to provide a pictorial view of the interactions between the independent variables on total HCA degradation. These plots are shown in Figure 4.19, where two independent variables were varied within the experimental ranges investigated while the remaining variables were kept constant. The interactions are significant as the curvature of the surfaces is obvious.

123 Design-Expert® Software Original Scale (% Total HCA Degradation)^3 76.7742

49.6923 X1 = A: Total HCA (a) X2 = B: Sucrose

Actual Factors C: pH = 5.00 72 D: Temperature = 35.00 69

% Total HCA Degradation Degradation HCA % Total 4 65 6 88 Design-Expert® Software Original Scale 8 110 (% Total HCA Degradation)^3 9 133 76.7742 B: Sucrose A: Total HCA 11 155 49.6923

X1 = B: Sucrose (b) X2 = C: pH

Actual Factors A: Total HCA = 155 67 D: Temperature = 35.00 65

% Total HCA Degradation Degradation HCA % Total 4 5.25 6 5.13 Design-Expert® Software Original Scale 8 5.00 (% Total HCA Degradation)^3 9 4.88 76.7742 B: Sucrose C: pH 11 4.75 49.6923

X1 = C: pH (c) X2 = D: Temp.

Actual Factors A: Total HCA = 155.00 65 B: Sucrose = 7.50 63

% Total HCA Degradation Degradation HCA % Total 4.75 44 4.88 41 5.00 38 C: pH 5.13 34 D: Temp. 5.25 31

Figure 4.19 Three-dimensional surface plots of total HCA degradation (%) as a function of (a) total HCA and sucrose; (b) sucrose and pH; and (c) pH and temperature. Variables: total HCA (155 mg/L); sucrose (7.5% (w/w)); pH (5.0) and temperature (35 °C).

124

The variables of sucrose concentration and initial total HCA concentration were varied as shown in Figure 4.19a, whilst the other variables, namely pH and temperature were kept constant at 5.0 and 35 °C respectively. These fixed values were chosen as they were similar to that typical of process sugar cane juice (Nguyen and Doherty, 2012). The total HCA degradation efficiency decreases with increasing sucrose concentration and the initial total HCA concentration. Increasing the initial total HCA concentration did not significantly decrease the degradation efficiency of the HCAs. This can be seen by both the coefficient of the first- and second-order model term (Equation 4.9) for total HCA concentration (i.e., A) and in Figure 4.19a where there was only a 5.9% discrepancy between 65 mg/L and 155 mg/L of initial total HCA at 3.75% (w/w) sucrose. This discrepancy is not noticeable at higher sucrose concentrations. It can be said from these observations, that the optimal Fenton dosage is capable of degrading higher concentrations of HCAs and other components (similar to that of HCAs) than at the highest concentration studied (i.e., > 200 mg/L).

Sucrose concentration showed a significant effect on the degradation of the HCAs (Figure 4.19b). Degradation increases smoothly with an increase in pH from 4.75 to 5.0 but decreases gradually when the pH exceeds 5.0, at any given concentration of sucrose. The negative effect on total HCA degradation at lower pH than the optimal may be attributed to the scavenging effect of H+ or •OH radicals which can inhibit the reduction of Fe(III) to Fe(II) and prevent the further generation of •OH radicals (Rivas et al., 2005; Deng, 2007). On the other hand, the negative effect at pH above the optimal may be attributable to the deactivation of the Fe(II) catalyst with the formation of Fe(III) oxyhydroxide in lieu of being regenerated back to Fe(II) (Bigda, 1995). The formation of Fe(III) oxyhydroxides in the present study was confirmed by analysing the precipitates obtained at pH 5.5 and 25 °C, by XRD (Table 4.14). The d-spacing values 6.21 Å, 3.28 Å, 2.46 Å and 2.36 Å correspond to lepidocrocite (i.e., Fe(III) oxide hydroxide), FeO(OH), while the peaks at 5.20 Å and 2.04 Å is associated with CaA (Dong et al., 2012). The formation of oxyhydroxide is derived from the following reaction equation (Equation 4.10):

2+ – Fe + ¼O2 + 2OH  FeOOH + ½H2O (4.10)

125

Table 4.14 X-ray diffraction data of the precipitate formed between CaA and Fe(II) at pH 5.5 and 25 °C.

d-spacing (Å) Intensity (counts) Precipitate FeO(OH)* 126.31 6.2164 6.2580 80.550 5.2003 204.48 3.2821 3.2933 161.52 2.4644 2.4737 138.89 2.3609 2.3635 543.59 2.0409 78.600 1.9320 1.9365 36.960 1.7282 1.7350 28.340 1.5252 1.5360 112.57 1.2941 1.2990 *Based on a FeO(OH) reference pattern (ICDD PDF card 04-010-4300).

The reduction in degradation effectiveness at pH above the optimal, in addition could be because H2O2 is relatively unstable and may rapidly decompose to

H2O and O2 (Kuo, 1992; Chang et al., 2010). Thus at pH below or above the optimal, the amounts of Fe(II) and/or H2O2 required to catalyse the Fenton oxidation process is reduced.

Figure 4.19c shows the interaction effects of pH and temperature on HCA degradation. The non-significance of the temperature variable is evident by the narrow range on the response axis (i.e., 57–64%). Despite this, the degradation trend on the HCAs in terms of temperature is still observable. Increasing temperature leads to less degradation of the HCAs. The decomposition of H2O2 by Fe(II) is not directly linked to the amount of HCA degraded. In addition to the formation of •OH radicals by the Fenton process, non-reactive species such as H2O and O2 are also formed at higher temperatures (> 40 °C) (Rodrigues et al., 2009b). The Fenton process was the only contributor to the degradation of HCAs as there was no thermal decomposition of any of the HCAs within the temperature range studied (25–50 °C).

126

Process Optimisation and Validation

Numerical optimisation was performed on the basis of the desirability function to determine the optimum process parameters for the degradation of the HCAs. The desirability function is expressed numerically from a scale of 0 to 1 (lowest to highest desirability) and denotes the degree of importance in obtaining the desired response value (Harrington, 1965). A desirability function value can be constructed by using five different goal optimisation constraints: none, maximum, minimum, target and within range. On the basis of the fitted quadratic models, an optimised response value can then be predicted by using the chosen goal optimisation criteria that maximises the desirability function. In order to simultaneously optimise numerous responses (i.e., multi-response optimisation), the desirability function values for each response (i.e., CaA, pCoA and FeA) are combined into an overall desirability function by computing their geometric mean of different desirability values, as shown in Equation 4.11 (Derringer and Suich, 1980).

1 (4.11) 1 n n n  D( d1  d 2  d 3  ...  dn )   d 1 i1 where, D overall desirability function

di desirability of the response

n number of responses investigated

In order to confirm the accuracy and robustness of the predicted models and assess its reliability to predict the (%) degradation of HCAs, additional experiments were carried out under those conditions, as well as selected conditions of process streams close to that of a typical Australian sugar mill.

For this study, the desirability functions for the three individual HCA degradation models were combined into one value and compared to the desirability function of the total HCA model (Table 4.15). The combined desirability function values of the three individual HCA models for the experiments were relatively close to the desirability values produced for the single total HCA degradation model. This

127

indicates that there is little variation between the simultaneously predicted values of each HCA degraded and the predicted value for the total HCA degraded.

Table 4.15 Optimised conditions under specified constraints for the degradation of total HCA (200 mg/L) and model verification.*

Experiments Water Worst Synthetic Synthetic case juice 1 juice 2 Sucrose (% (w/w)) 0 14.00 13.00 21.00 pH 4.7 4.5 5.4 4.9 Temperature (°C) 25 40 36 30 CaA degradation (%) 92 (90) 87 (90) 73 (72) 78 (68) pCoA degradation (%) 69 (68) 33 (37) 48 (52) 52 (75) FeA degradation (%) 70 (64) 40 (46) 51 (56) 54 (84) Total HCA degradation (%) 77 (73) 53 (49) 57 (58) 61 (67) Desirability Combined models 0.720 0.542 0.383 0.655 Total HCA degradation model 0.743 0.632 0.332 0.621 *Values in parentheses indicate model predicted % degradation for each individual/total HCA model. Measurements were conducted in triplicate. RSD was < 5.0%.

As shown in Table 4.14 the experimental and predicted values (in parentheses) for the degradation of each and the total of the HCAs, under specified constraints. The optimum conditions for maximum degradation of HCAs (200 mg/L) using the

Fenton process (0.5 mM Fe(II) and 7.5 mM H2O2) are 0% (w/w) sucrose (i.e., aqueous), pH 4.7 and 25 °C. Under these conditions 92% CaA, 69% pCoA and 70% FeA was degraded (i.e., total HCA degradation of 77%). The experimental values of the optimum conditions agree well with the predicted values deduced from each of the four models. The low error in the experimental and predicted values indicates good agreement of the results. The experimental values obtained for the

128

worst conditions were also in good agreement with the predicted values. The good agreement between values is attributable to the high combined desirability function value. It is worth mentioning that the sum of the predicted degradation values of the individual HCA degradation models (i.e., CaA, pCoA and FeA) is not equal to the predicted total HCA degradation values. Hence, the individual degradation models should be only used as a guide to predict the degradation of the total HCAs present in a mixture.

The best results for the synthetic juices are obtained with solutions having similar sucrose content and operating temperature as factory MJ (synthetic juice 1) followed by factory No. 1 mill juice (i.e., juice expressed from the first mill of a quintuple set of mills) (synthetic juice 2). Despite a low desirability function value predicted under the synthetic juice 1 conditions, the experimental results were in close agreement with the predicted values for all four models. The lower desirability may be due to some of the constraints that were not close to any of the design points of the CCD. On the other hand, a higher error was observed for synthetic juice 2, despite a reasonable desirability value. In addition, the experimental values obtained for pCoA and FeA degradation were significantly lower than the predicted values. It is highly probable therefore, that the presence of sucrose may have contributed to the inaccuracy of this prediction as its concentration was outside the range used to develop the proposed models. Therefore, it is not recommended to use constraints outside the ranges studied for multi-response optimisation, as the responses are all dependent on each other.

From these results, the Fenton process can successfully be used to degrade HCAs (i.e., colour precursor compounds) under the operating conditions in a raw sugar manufacturing factory.

129

4.4 Summary

In this chapter, the experimental procedures and statistical treatment of experimental results for the degradation of three phenolic acids (CaA, pCoA, and FeA) is described in great detail. For each selected acid there are up to seven independent variables considered, with conditions covering the normal ranges experienced in sugar cane juice processing and about 28 combinations of these are examined for significance in a quadratic regression to determine an expression for the % degradation. The statistical manipulations to obtain a meaningful expression for degradation are given in detail. The effect of variations in individual parameters was sometimes rationalised by appealing to their effect on the availability of free •OH radicals for degrading the particular phenolic acid being studied.

The degradation of HCA mixtures by the Fenton process has been studied in water and sucrose solutions. From the information obtained from the initial model for the degradation of CaA; four additional quadratic models were developed and showed the working relationship between the degradation efficiency of each HCA (i.e., CaA, pCoA and FeA) with four independent variables (i.e., initial HCA mixture concentration, sucrose concentration, solution pH and reaction temperature). Under the optimised conditions for a 200 mg/L initial HCA mixture concentration, the degradation efficiencies of the mixture in water and sugar solutions (i.e., 13% (w/w)) were 77% and 57% respectively.

The behaviour of CaA degradation in the composite system is different from that of pCoA and FeA possibly because of its ability to form complexes with Fe(III), as its aromatic ring is highly activated with the presence of two hydroxyl groups. In addition, CaA has been shown to hydrogen-bond with sucrose – a free radical scavenger.

The Fenton process has been shown to degrade HCAs in sucrose solutions with minimum sucrose breakdown. This means that the process may find use in the raw sugar manufacturing process for the removal of these and other colour precursors that are significantly prevalent when the juice expressed from the whole sugar cane biomass (instead of the stalk) is processed. As the sugar cane industry around the world is looking towards diversification by value-adding with the excess biomass produced from whole crop processing, the use of the Fenton or similar processes will

130

allow juice expressed from the whole sugarcane plant to be cost-effectively processed. The advantages of the use of the Fenton process in the sugar manufacturing process include its simplicity, its non-specific oxidation property and the use of inexpensive equipment. Also, the sludge that is produced has the potential to remove colourants and other impurities (including proteins and polysaccharides) improving the quality of the juice feedstock.

131

References

Adams, M. L., O'Sullivan, B., Downard, A. J., & Powell, K. J. (2002). Stability constants for aluminum(III) complexes with the 1,2-dihydroxyaryl ligands caffeic acid, chlorogenic acid, DHB, and DASA in aqueous solution. Journal of Chemical and Engineering Data, 47(2), 289-296.

Bigda, R. J. (1995). Consider Fenton's chemistry for Fenton wastewater treatment. Chemical Engineering Progress, 91(12), 62-66.

Box, G. P. E., & Cox, D. R. (1964). An analysis of transformations. Journal of the Royal Statistical Society, B26, 211-243.

Chang, M.-W., Chung, C.-C., Chern, J.-M., & Chen, T.-S. (2010). Dye decomposition kinetics by UV/H2O2: Initial rate analysis by effective kinetic modelling methodology. Chemical Engineering Science, 65, 135-140.

Cornard, J.-P., Caudron, A., & Merlin, J.-C. (2006). UV–visible and synchronous fluorescence spectroscopic investigations of the complexation of Al(III) with caffeic acid, in aqueous low acidic medium. Polyhedron, 25, 2215-2222.

Cortez, S., Teixeira, P., Oliveira, R., & Mota, M. (2011). Evaluation of Fenton and ozone-based advanced oxidation processes as mature landfill leachate pre- treatments. Journal of Environmental Management, 92, 749-755. de Souza, D. R., Duarte, E. T. F. M., de Souza Girardi, G., Velani, V., da Hora Machado, A. E., Sattler, C., de Oliveira, L., & de Miranda, J. A. (2006). Study of kinetic parameters related to the degradation of an industrial effluent using Fenton-like reactions. Journal of Photochemistry and Photobiology, A: Chemistry, 179, 269-275.

Deng, Y. (2007). Physical and oxidative removal of organics during Fenton treatment of mature municipal landfill leachate. Journal of Hazardous Materials, 146, 334-340.

Derringer, G., & Suich, R. (1980). Simultaneous optimization of several response variables. Journal of Quality Technology, 12, 214-219.

Dobson, K. D., & McQuillan, A. J. (2000). In situ infrared spectroscopic analysis of the adsorption of aromatic carboxylic acids to TiO2, ZrO2, Al2O3, and Ta2O5 from aqueous solutions. Spectrochimica Acta, Part A: Molecular and Biomolecular Spectroscopy, 56(3), 557-565.

Dong, W., Xu, Y., Ren, J., Chu, H., Li, J., Liu, X., & Chen, M. (2012). Synthesis and characterization of biodegradable polymers composed of 3,4- dihydroxycinnamic acid and poly(ethylene glycol). Journal of Applied Polymer Science, 125(3), 1657-1662.

Dürüst, N., Özden, S., Umur, E., Dürüst, Y., & Küçükislamoğlu, M. (2001). The isolation of carboxylic acids from the flowers of Delphinium formosum. Turkish Journal of Chemistry, 25(1), 93-97.

132

Eggleston, G., Monge, A., & Ogier, B. E. (2003). Sugarcane factory performance of cold, intermediate, and hot lime clarification processes. Journal of Food Processing and Preservation, 26, 433-454.

Gilfillan, W. N., Nguyen, D. M. T., Sopade, P. A., & Doherty, W. O. S. (2012). Preparation and characterisation of composites from starch and sugar cane fibre. Industrial Crops and Products, 40(0), 45-54.

Hanna, K., & Quilès, F. (2011). Surface complexation of 2,5-dihydroxybenzoic acid (gentisic acid) at the nanosized hematite−water interface: An ATR-FTIR study and modeling approach. Langmuir, 27(6), 2492-2500.

Harrington, E. C. (1965). The desirability function. Industrial Quality Control, 21, 494-498.

Huvenne, J. P., Vergoten, G., Fleury, G., & Legrand, P. (1981). Champ de forces de symetrie locale de molecules cycliques: III. Interprétation des spectres de vibration du glucose cristallisé sous les deux formes α et β. Journal of Molecular Structure, 74(2), 169-180.

Hynes, M. J., & O'Coinceanainn, M. (2004). The kinetics and mechanisms of reactions of iron(III) with caffeic acid, chlorogenic acid, sinapic acid, ferulic acid and naringin. Journal of Inorganic Biochemistry, 98, 1457-1464.

Kačuráková, M., & Mathlouthi, M. (1996). FTIR and laser-Raman spectra of oligosaccharides in water: characterization of the glycosidic bond. Carbohydrate Research, 284(2), 145-157.

Kalinowska, M., Świsłocka, R., & Lewandowski, W. (2011). Zn(II), Cd(II) and Hg(I) complexes of cinnamic acid: FT-IR, FT-Raman, 1H and 13C NMR studies. Journal of Molecular Structure, 993(1–3), 404-409.

Kitis, M., Adams, C. D., & Daigger, G. T. (1999). The effects of Fenton's reagent pretreatment on the biodegradability of nonionic surfactants. Water Research, 33(11), 2561-2568.

Kodad, H., Mokhlisse, R., Davin, E., & Mille, G. (1994). Étude IRTF par réflexion totale atténuée (ATR) de sucres en solution aqueuse. Canadian Journal of Applied Spectroscopy, 39(4), 107-112.

Kulik, T. V., Barvinchenko, V. N., Palyanytsya, B. B., Lipkovska, N. A., & Dudik, O. O. (2011). Thermal transformations of biologically active derivatives of cinnamic acid by TPD MS investigation. Journal of Analytical and Applied Pyrolysis, 90, 219-223.

Kuo, W. G. (1992). Decolorizing dye wastewater with Fenton's reagent. Water Research, 26(7), 881-886.

Machado, N. F. L., Calheiros, R., Gaspar, A., Garrido, J., Borges, F., & Marques, M. P. M. (2009). Antioxidant phenolic esters with potential anticancer activity: Solution equilibria studied by Raman spectroscopy. Journal of Raman Spectroscopy, 40(1), 80-85.

133

Max, J.-J., & Chapados, C. (2001). Sucrose hydrates in aqueous solution by IR spectroscopy. Journal of Physical Chemistry A, 105(47), 10681-10688.

Morgan, B., & Lahav, O. (2007). The effect of pH on the kinetics of spontaneous Fe(II) oxidation by O2 in aqueous solution – basic principles and a simple heuristic description. Chemosphere, 68(11), 2080-2084.

Nguyen, D. M. T., & Doherty, W. O. S. (2012). Phenolics in sugar cane juice: Potential degradation by hydrogen peroxide and Fenton's reagent. International Sugar Journal, 114(1361), 309-315.

Pala, A., & Erden, G. (2005). Decolorization of a baker's yeast industry effluent by Fenton oxidation. Journal of Hazardous Materials, B127, 141-148.

Rivas, F. J., Beltrán, F. J., Garcia-araya, J. F., Navarrete, V., & Gimeno, O. (2002). Co-oxidation of p-hydroxybenzoic acid and atrazine by the Fenton’s like system Fe(III)/H2O2. Journal of Hazardous Materials, B91, 143-157.

Rodrigues, C. S. D., Madeira, L. M., & Boaventura, R. A. R. (2009a). Optimization of the azo dye Procion Red H-EXL degradation by Fenton’s reagent using experimental design. Journal of Hazardous Materials, 164, 987-994.

Rodrigues, C. S. D., Madeira, L. M., & Boaventura, R. A. R. (2009b). Treatment of textile effluent by chemical (Fenton’s Reagent) and biological (sequencing batch reactor) oxidation. Journal of Hazardous Materials, 172, 1551-1559.

Sánchez-Cortés, S., & García-Ramos, J. V. (1999). Photoinduced coupling and adsorption of caffeic acid on silver surface studied by surface-enhanced Raman spectroscopy. Spectrochimica Acta, Part A: Molecular and Biomolecular Spectroscopy, 55(14), 2935-2941.

Smith, P. (1983). A Review of Raw Sugar Colorant Chemistry and its Relevance to Storage and Refining Quality. Paper presented at the Discussions on Colour at the Meeting in Sydney. Sydney, Australia: Sugar Division, CSR Ltd.

Sun, J.-H., Shi, S.-H., Lee, Y.-F., & Sun, S.-P. (2009). Fenton oxidative decolorization of the azo dye Direct Blue 15 in aqueous solution. Chemical Engineering Journal, 155, 680-683.

Świsłocka, R. (2013). Spectroscopic (FT-IR, FT-Raman, UV absorption, 1H and 13C NMR) and theoretical (in B3LYP/6-311++G** level) studies on alkali metal salts of caffeic acid. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 100(0), 21-30.

134

Świsłocka, R., Kowczyk-Sadowy, M., Kalinowska, M., & Lewandowski, W. (2012). Spectroscopic (FT-IR, FT-Raman, 1H and 13C NMR) and theoretical studies of p-coumaric acid and alkali metal p-coumarates. Spectroscopy: An International Journal, 27(1), 35-48.

Tang, W. Z., & Huang, C. P. (1996). 2,4-Dichlorophenol oxidation kinetics by Fenton's reagent. Environmental Technology, 17(12), 1371-1378.

Tang, W. Z., & Huang, C. P. (1997). Stoichiometry of Fenton's reagent in the oxidation of chlorinated aliphatic organic pollutants. Environmental Technology, 18, 13-23.

Tunali, S., & Batmaz, I. (2000). Dealing with the least squares regression assumptions in simulation metamodeling. Computers & Industrial Engineering, 38, 307- 320.

Varsányi, G. (1974). Assignments for Vibrational Spectra of Seven Hundred Benzene Derivatives (Vol. 1). London: Adam Hilger.

Vasko, P. D., Blackwell, J., & Koenig, J. L. (1971). Infrared and raman spectroscopy of carbohydrates: Part I: Identification of O–H and C–H-related vibrational modes for D-glucose, maltose, cellobiose, and dextran by deuterium- substitution methods. Carbohydrate Research, 19(3), 297-310.

Yost, E. C., Tejedor-Tejedor, M. I., & Anderson, M. A. (1990). In situ CIR-FTIR characterization of salicylate complexes at the goethite/aqueous solution interface. Environmental Science and Technology, 24(6), 822-828.

135

136

CHAPTER 5

Separation and Identification of Fenton Oxidation Products Derived from Hydroxycinnamic Acids

5.1 Introduction...... 137 5.2 Materials and Methods...... 138 5.2.1 Reagents and Solvents...... 138 5.2.2 Fenton Oxidation Reactions for the Degradation of Hydroxycinnamic Acid Mixtures...... 138 5.2.3 Sample Preparation...... 139 5.2.4 Instrumental Procedures and Analyses...... 140 5.2.5 Fenton Oxidation Reactions for the Degradation of Sucrose Mixtures...... 142 5.2.6 Computational Methods...... 142 5.3 Results and Discussion...... 143 5.3.1 Identification of Oxidation Products...... 143 5.3.2 Proposed Degradation Pathways of Selected Hydroxycinnamic Acids...... 153 5.3.3 Oligomerisation of Hydroxycinnamic Acids...... 166 5.4 Summary...... 171

137

5.1 Introduction

In this chapter, attempts were made to determine the oxidation products and the degradation pathways of the HCAs studied (viz., CaA, pCoA and FeA). To date, inadequate information exists in relation to the oxidation of these compounds using the Fenton process and other established AOPs. Not all of the oxidation products of these HCAs have been previously isolated or identified. The identification of these products will assist in proposing possible mechanistic pathways of the oxidation process. Therefore, the investigations in this chapter are further attempts to identify these products and propose probable mechanisms for the degradation of these phenolic acids.

5.2 Materials and Methods

5.2.1 Reagents and Solvents

All chemicals, solvents and reagents were obtained in their purest form from the suppliers as described in the previous chapters or as otherwise stated. Acetic acid, cis–aconitic acid, trans–aconitic acid, butyric acid, citric acid, formic acid, fumaric acid, glycolic acid, glyoxylic acid, isobutyric acid, lactic acid, malic acid, oxalic acid and succinic acid were purchased from Sigma-Aldrich (St. Louis, MO, USA). Potassium chloride was obtained from Ajax Finechem (Seven Hills, NSW, Australia). Stock solutions of HCAs (i.e., CaA, pCoA and FeA) were prepared by dissolution in degassed ethanol solution (50% (v/v)) and stored at 4.0 °C.

5.2.2 Fenton Oxidation Reactions for the Degradation of Hydroxycinnamic Acid Mixtures

The procedure for the Fenton oxidative degradation of HCA mixtures is similar to that described in Section 4.2.4. Four HCA solutions were investigated (viz., CaA, pCoA, FeA and their mixture) under the optimised operating conditions in water, pH 4.7 and 25 °C (cf. Section 4.3.3).

138

In each run, a predetermined amount of Milli-Q water and each HCA were added to the reaction vessel. To improve the analytical detector response of the reaction products, the starting materials for the combined HCA mixture were carried out at two orders of magnitude higher than those studied in Section 4.3.3. The final concentrations for each HCA component were CaA (37 mM), pCoA (41 mM) and FeA (34 mM). The total HCA concentration is approximately equivalent to 20,000 mg/L (i.e., 100× more than 200 mg/L used for total HCA concentration in Chapter 4). For the three individual HCA mixtures, an initial HCA concentration of 100 mM (approximately equivalent to 20,000 mg/L) was added to the reaction vessel.

Known amounts of FeSO4·7H2O (0.5 M, 2.49 mL) and H2O2 (5.0 M, 0.75 mL) solutions were added to achieve a final volume of 50 mL and a final concentration of

24.9 mM and 75 mM, respectively. The working molar ratio (Fe(II)/H2O2) for the

Fenton reaction was 1:15. The reaction was initiated as soon as H2O2 was added. At 2 min, 3.0 mL of the solution was taken, diluted 10-fold to quench the reaction and stored at 4.0 °C. The remainder of the mixtures were immediately snap-frozen in liquid nitrogen and stored at –80 °C. Samples were defrosted and prepared for instrumental analysis.

5.2.3 Sample Preparation

Undiluted samples were pre-concentrated using solid phase extraction (SPE) prior to gas chromatography/electron impact-mass spectrometry (GC/EI-MS) analysis. Undiluted samples for high-performance ion exclusion chromatography (HPIEC) analysis did not require any sample preparation. Diluted samples for the analyses on all other analytical techniques required no further sample preparation.

Samples for GC/EI-MS analyses did not require any further adjustment prior to SPE. Waters Sep-Pak tC18 vacuum cartridges (3 cc, 500 mg, 37-55 µm) (Wexford, Ireland) were placed in a Waters Sep-Pak 24-port vacuum manifold (Milford, MA, USA) and first conditioned with 2 × 2.5 mL HPLC grade methanol followed by 2 × 2.5 mL of Milli-Q water. After the conditioning step, 2 × 1.0 mL aliquots of the reaction sample were loaded at a flow rate lesser than 2.0 mL/min by adjusting the vacuum to ca. 15 kPa. The column was washed with 2.5 mL of Milli-Q

139

water. Finally, elution was performed with 2 × 1.0 mL HPLC grade methanol at a flow rate ≤ 1.0 mL/min by adjusting the vacuum to ca. 10 kPa. The eluates obtained were concentrated by solvent evaporation under a gentle stream of N2 and recomposed to a final volume of 1 mL in HPLC grade methanol. The extracts were membrane filtered (0.45 μm) prior to GC/EI-MS analysis.

5.2.4 Instrumental Procedures and Analyses

HPLC-DAD/ESI-Q-TOF-MS/MS. Identification of organic reaction products was evaluated using reversed-phase HPLC coupled with UV/Vis DAD and electrospray ionisation quadrupole time-of-flight tandem mass spectrometry (ESI-Q- TOF-MS/MS). Analyses were performed on an Agilent 1290 Infinity LC system (G4220A binary pump, Germany; G4226A ALS, Germany; G1330B ALS thermostat, Germany; G1316C thermostatted column compartment, Germany; G1314E variable wavelength detector (VWD), Germany) coupled with an Agilent Accurate-Mass Q- TOF mass spectrometer (G6520B, USA). The chromatographic conditions were identical as described in Section 4.2.5 for all samples with the exception of single wavelength UV/Vis detection (280 nm) and injection volume (20 μL). The column effluent from the VWD was then introduced into the dual ESI source of the Q-TOF mass spectrometer without post-column splitting. Mass spectra were acquired in negative ion mode and the conditions were set as follows: gas temperature: 350 °C; drying gas flow (N2): 12 L/h; nebuliser (N2): 35 psig; capillary voltage: 3.5 kV; fragmentor: 170 V; skimmer: 60 V; OCT1 RF Vpp: 250 V. Data acquisition was performed using the Agilent Masshunter Data Acquisition TOP/Q-TOF B.02.00 software package, scanning from a mass-to-charge ratio (m/z) 100 to 1500 in profile (continuum) mode with a scan cycle time of 2.242 s and an acquisition time of 714.1 ms/spectrum. Two reference masses (m/z 121.050873 and m/z 922.009798) were used. Tandem MS product ions were produced by collision-induced dissociation of selected precursor ions in the collision cell of the Q-TOF mass spectrometer at fixed collision energy voltage of 50 V. The Agilent Masshunter Qualitative Analysis (B.02.00) software package was used for data analysis.

140

HPIEC. Identification of carboxylic acids was evaluated using high- performance ion exclusion chromatography (HPIEC). Analyses were performed on a Waters HPLC system (Milford, MA, USA) equipped with a 626 pump, a 600S controller, a 717plus autosampler, a 2487 dual λ absorbance detector and a 410 differential refractometer. The separation was carried out on two Bio-Rad Aminex HPX-87H ion exclusion columns (300 × 7.8 mm i.d.) (Heracles, CA, USA) connected in series and protected by a Bio-Rad Cation H+ micro-guard cartridge (30 × 4.6 mm i.d.). Reaction products were eluted isocratically with 8 mM H2SO4 (sparged with helium at 10 mL/min). Simultaneous UV/Vis detection at specific wavelengths (190 nm and 210 nm) without reference wavelength subtraction. Refractive index detection was set to negative polarity and run at an attenuator setting of 20×. Aliquots of samples were membrane filtered (0.45 μm) prior to injection into the HPLC system. Injection volume for all samples was 20 μL; first and second column temperatures were equilibrated at 35 °C and 85 °C respectively; flow rate was 0.4 mL/min and run time was 90 min. Data acquisition was performed using the Waters Empower 2 (Build 2154) software package. Identification of peaks was based on the conformance of UV/Vis spectra and retention times with the corresponding authentic standards.

GC/EI-MS. Identification of volatile reaction products was evaluated using GC/EI-MS. Analyses were performed on a Hewlett Packard HP/Agilent 6890, 7683 and 5973 Series GC/MS system (G1530A (6890A) gas chromatograph system, USA; G2614A (7683) ALS Tray, China; G2613A (7683) ALS Injector, China; G1926A (7683) Bar Code Reader, China; G2589A (5973N) Mass Selective Detector; USA) using a Phenomenex Zebron ZB-1 GC capillary column (30 m, 0.25 mm i.d. and 0.25 μm thick film) (Torrance, CA, USA). Helium was used as carrier gas at a constant flow rate of 0.5 mL/min. Sample aliquots (3.0 μL) were injected in splitless mode at an injector temperature of 250 °C. The oven temperature program was 4 min at 40 °C; 8 °C/min to 180 °C (2 min); and 8 °C/min to 280 °C (9 min). The mass range scanned was 35 m/z to 500 m/z using EI ionisation at 70 eV. Data acquisition and analysis was performed using the Agilent MSD ChemStation (G1701EA E.01.00.237) software package.

141

5.2.5 Fenton Oxidation Reactions for the Degradation of Sucrose Mixtures

Four 50 mL mixtures containing only sucrose at various concentrations (3.75%, 7.50%, 11.25% and 15.0% (w/w)) were subjected to Fenton oxidation under the optimum operating conditions described in Section 4.2.4 (i.e., 2.49 mM

FeSO4·7H2O, 7.50 mM H2O2, pH 5.4 and 36 °C). These experiments were carried out to determine the presence of organic acids and reducing sugars formed from the Fenton oxidation of sucrose. At 2 min, the mixtures were immediately snap-frozen in liquid nitrogen and stored at –80 °C. Samples were defrosted and prepared for HPIEC and HPAEC-PAD analyses for the determination of organic acids and reducing sugars, respectively.

Prior to HPIEC analysis, samples containing sugars were adjusted to pH 8.5 using 0.1 M NaOH prior to SPE to facilitate the ion exchange in the packed cartridge. Waters AccellPlus QMA vacuum cartridges (3 cc, 500 mg, 37-55 µm) (Wexford, Ireland) were placed in a Waters Sep-Pak 24-port vacuum manifold and first conditioned with 2 × 2.5 mL of 0.5 M potassium chloride solution followed by 2 × 2.5 mL of Milli-Q water. After the conditioning step, 2 × 2.0 mL aliquots of the reaction sample were loaded at a flow rate ≤ 1.0 mL/min by adjusting the vacuum to ca. 5 kPa. The column was washed with 8 × 2.5 mL of Milli-Q water. Finally, elution was performed with 2 × 2.0 mL of 0.1 M sulfuric acid at a flow rate ≤ 0.5 mL/min by adjusting the vacuum to ≤ 5 kPa. The extracts were membrane filtered (0.45 μm) prior to HPIEC analysis. The operating procedure for the chromatographic system is identical to that described in Section 5.2.3.

Sucrose and reducing sugar contents in the reaction mixtures were monitored by HPAEC-PAD. Sample preparation and the operating procedure for the chromatographic system are identical to that described in Section 4.2.5.

5.2.6 Computational Methods

Geometry optimisations of HCA molecular systems in their ground state were performed using the density functional theory (DFT) methods implemented using the Wavefunction, Inc. Spartan ′10 (1.1.0) software package (Irvine, CA, USA). Density functional theory was chosen as the method for computation as it provides a

142

reasonable description of the electronic correlation of a molecule in a quantum system within minimal computational time and cost (Fifen et al., 2009). Also, the precision of the DFT is typically better than that of other methods (e.g., Hartree-Fock (HF), semi-empirical) where the electron spin is not considered (Nsangou et al., 2008). Density functional approximations were calculated based on the B3LYP hybrid functional, which consists of the Becke’s three parameters exact exchange functional (B3) (Becke, 1988) combined with the non-local gradient corrected correlation functional of Lee-Yang-Parr (LYP) (Lee et al., 1988). The standard split valence double-zeta Gaussian basis set 6-31G augmented by a set of d polarisation functions (Frisch et al., 1984) on heavy atoms was chosen. Solvent effects of water are computed in the framework of a restricted HF-DFT self-consistent field SM8 model using the Pulay direct inversion iterative subspace approach (Pulay, 1980) with geometric direct minimisation (Van Voorhis and Head-Gordon, 2002). The outputs produced from the theoretical calculations are presented in Appendices, Tables A2.1–A2.3.

5.3 Results and Discussion

5.3.1 Identification of Oxidation Products

In comparison to other AOPs (e.g., ozonation, UV/H2O2 oxidation), the mechanism of the Fenton process between Fe(II) and H2O2 is already intricate and is further complicated when an organic compound is involved in the reaction. So, mechanistic pathways for the degradation of organic compounds using the Fenton process have only been proposed for simpler compounds. Structural elucidation and proposal of reaction schemes is complex in the case of HCAs, since these phenolic derivatives are molecularly larger and have more available sites for free radical attack. In turn, they may produce complex intermediates or produce several smaller products at lower concentrations that are difficult to detect. Identification of these products will assist in proposing possible mechanistic pathways for the oxidative degradation of the HCAs investigated.

143

Liquid Chromatography Techniques

The stoichiometry for the complete mineralisation of CaA (C9H8O4), pCoA

(C9H8O3) and FeA (C10H10O4) by H2O2 in the Fenton process is as follows:

C9H8O4 + 18H2O2  9CO2 + 22H2O (5.1)

C9H8O3 + 19H2O2  9CO2 + 23H2O (5.2)

C10H10O4 + 21H2O2  10CO2 + 26H2O (5.3)

It is expected that the optimised working Fenton molar ratio (Fe(II)/H2O2) of 1:15 is insufficient for complete degradation of all acids, individually or combined. The incomplete depletion of HCA peaks and the presence of new peaks detected as depicted in the HPLC-DAD profiles of the reaction mixtures (Figure 5.1) suggest that there are reaction products remaining in solution. The identification of these compounds is important in order to assess and predict their role in downstream processes of the sugar manufacturing process. It is presumed that after 2 min, under the optimum operating conditions in water or sucrose solutions, the reaction had reached equilibrium as there were no changes in the response of the chromatographic peaks of the starting compounds after 2 min of the reaction initiated by H2O2.

Figure 5.1 shows numerous chromatographic peaks corresponding to the starting materials and the reaction products formed at 2 min of Fenton oxidation. The numbers directly labelled on the peaks of the HPLC-DAD chromatograms (Figure 5.1) are associated with the identified products listed in Table 5.1. Products were identified based on the comparison of retention time data of the available pure standards and/or accurate mass measurements. Proposal of chemical structures were evaluated based on the predicted oxidation mechanism of the Fenton process and mass spectral fragmentation patterns of similar compounds suggested in the literature (Fulcrand et al., 1994; Antolovich et al., 2004). Additional information was obtained from the isotopic distribution in the mass spectra for certain molecules. Table 5.1 also shows data related to the experimental and calculated masses of the deprotonated ions and proposed empirical formulae related to the identified compounds. The resulting accurate masses were found with an error lower than 0.04 Da.

144

(a) 150

110

2 70

1 3

Absorbance(mAU) 30

-10 0 5 10 15 20 25 Retention Time (min) (b) 150 4 5

110

Absorbance(mAU) 30

-10 0 5 10 15 20 25 Retention Time (min) (c) 150 7

110

70 6

8 Absorbance(mAU) 30

-10 0 5 10 15 20 25 Retention Time (min)

Figure 5.1 High-performance LC-DAD chromatograms (UV/Vis detection at 280 nm) of (a) CaA; (b) pCoA and (c) FeA; subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C). Numbers correspond to compound numbers in Table 5.1.

145

Table 5.1 Reaction products formed from the Fenton oxidation of HCAs detected by LC/MS.

Peak Compound tR Molecular ion Formula Error (min) (m/z) (Da) Caffeic acid – 1 protocatechuic aldehyde 4.24 137.048 C7H5O3 –0.02 – 2 caffeic acid 7.02 179.061 C9H7O4 –0.03 3 caffeic acid tetramer 12.75 715.183 p–Coumaric acid – 4 4–hydroxybenzaldehyde 6.83 121.050 C7H5O2 –0.02 – 5 p–coumaric acid 10.64 163.065 C9H7O3 –0.03 Ferulic acid – 6 vanillin 7.12 151.065 C8H7O3 –0.03 – 7 ferulic acid 11.60 193.079 C10H9O4 –0.03 – 8 ferulic acid dimer 18.87 385.133 C20H17O8 –0.04

The strong intense peak at the retention time (tR) range of 0.83–0.98 min in the HPLC-DAD and is attributable to the solvent. Five products were identified by means of reversed-phase HPLC with UV/Vis DAD and negative ion mode ESI-MS detection and had retention times of less than 20 min. The chromatogram obtained for the combined HCA mixture (cf. Appendices, Figure A2.1) revealed no new peaks than those that already appear in the individual HCA mixtures in Figure 5.1. This may suggest that there were no side reactions taken place among the HCAs.

It is observed that at 2 min, the Fenton oxidation of CaA produced two main products observable at a wavelength of 280 nm. Protocatechuic aldehyde

(3,4–dihydroxybenzaldehyde) was assigned to the eluted peak at tR = 4.24 min. This product is the initial breakdown product of CaA as a result of •OH radical attack to the vinyl functional group of the phenolic acid. The later oxidation product had a m/z – ion of 715 which was tentatively assigned to the tetramer of CaA (i.e., [M4–H] ). It is assumed that the tetramer could have possibly been formed by the oxidative coupling – of two dimers of CaA (MW of 358) (i.e., [M2–H] ) (Pati et al., 2006). Unlike previous reports, the dimers formed by the Fenton oxidation of CaA were not

146

observed in this study (Cilliers and Singleton, 1991; Tazaki et al., 2001; Antolovich et al., 2004).

The main product observed from the Fenton oxidation of pCoA is

4–hydroxybenzaldehyde (tR = 6.83 min). There were several other unidentified peaks with intensities lower than 4–hydroxybenzaldehyde. Similar to CaA, only two main oxidation products were produced from the oxidation of FeA, they are vanillin

(4–hydroxyl–3–methoxybenzaldehyde) at tR = 7.12 min and a dimer of FeA – (i.e., [M2–H] ) at tR = 18.87 min. The dimeric products detected are in consistency with those found in similar oxidation studies (Antolovich et al., 2004; Šmejkalová et al., 2006).

In addition to the HPLC-DAD chromatograms shown in Figure 5.1 for the individual HCA solutions, the total ion chromatograms (TICs) recorded in negative ion mode ESI-MS for each phenolic acid mixture is shown in Figure 5.2. The mass spectral fragmentation pattern data extracted from peaks obtained in the TICs were similar to those extracted from the HPLC-DAD chromatographic peaks. Hence, the results shown in the TICs are consistent with the results obtained with DAD.

The peak at tR = 6.15 min shown in Figures 5.1b and 5.2b has a m/z of 121 and the peak at tR = 6.83 min also has the same m/z. The peak at tR = 6.83 min has been reported previously to be attributable to 4–hydroxybenzaldehyde. Therefore, the peak at tR = 6.15 min may be an isomer formed as a result of hydroxylation during

(Poerschmann et al., 2010). Three unidentified peaks at tR = 17.7, 19.7 and 21.3 min observed in each of the TICs shown in Figure 5.2 are possibly due to impurities present in the chromatographic system and are not associated to the starting materials and its reaction products.

147

(a) 25

) 6

10 15 x

10 3

Intensity ( Intensity 2 1 5

0 0 5 10 15 20 25 Retention Time (min) (b) 25

20 5

) 6

10 15 x

10 Intensity ( Intensity

0 0 5 10 15 20 25 Retention Time (min) (c) 25

) 6

10 15 x x

6 10 7

Intensity ( Intensity 8 5

0 0 5 10 15 20 25 Retention Time (min)

Figure 5.2 Total ion chromatograms (negative ion mode ESI-MS) of (a) CaA; (b) pCoA and (c) FeA; subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C). Numbers correspond to compound numbers in Table 5.1.

148

Ion Chromatography Techniques

Despite oligomer formation, the presences of the phenolic aldehydes produced from the Fenton oxidation of the HCA mixtures show that the Fenton process is decomposing the HCAs into smaller products. Based on the oxidative degradation mechanisms of aromatic compounds by the Fenton process, proposed by Neyens and Baeyens (2003), it is expected that the phenolic aldehydes would undergo further oxidation via hydroxylation causing the aromatic rings to open and form LMW aliphatic carboxylic acids. Hence, HPIEC was used in this project to determine the presence of carboxylic acids.

Between the two HPIEC detection methods used (viz., UV/Vis and RI), improved baseline resolution and peak separation was achieved with RI detection. Problems associated with UV/Vis detection at 190 nm and 210 nm wavelengths include broadening and overlapping of chromatographic peaks that may be attributable to other reaction intermediates and products that strongly or partially absorb at the chosen wavelengths. In the present study, butyric, cis–aconitic, formic, acetic, glyoxylic, isobutyric, lactic, and oxalic acids were detected in each of the reaction mixtures (Table 5.2).

Table 5.2 Contents of organic acids (mM) by HPIEC of individual and combined HCA mixtures.*

CaA pCoA FeA Mixture cis–Aconitic 0.016 0.015 0.016 0.015 Butyric 0.065 0.054 0.044 0.061 Formic 0.16 0.39 0.45 0.36 Glyoxylic 0.21 0.19 0.19 0.19 Isobutyric 5.0 2.6 3.9 3.6 Lactic 0.0038 0.0076 0.0026 0.003 Oxalic 5.2 4.8 4.7 4.6 *Mean values (n = 3). % RSD was < 5.0%.

149

However, these compounds except for oxalic and isobutyric acids were found at low concentrations (≤ 0.45 mM). This may indicate that progressive oxidative degradation from HCA is minimal under the operating conditions; or that these carboxylic acids may have decomposed to CO2 and H2O within 2 min of the reaction. Oxalic acid and isobuytic acid concentrations in each of the three individual acids ranged from 4.7–5.2 mM and 2.6–5.0 mM, respectively. The highest concentrations of oxalic, isobutyric and glyoxylic acids were obtained with CaA. However, higher amounts of butyric and formic acids were produced from pCoA and FeA degradation. Interestingly, these organic acids are typically organic acids found in sugar cane juice (Thai and Doherty, 2011).

Mixtures only containing Fenton’s reagent and sucrose at varying concentrations (3.75%, 7.50%, 11.25% and 15.0% (w/w)) were examined by HPIEC and HPAEC-PAD for the determination of carboxylic acids and reducing sugars, respectively. With the HPIEC method, no peaks were observed in both UV/Vis and RI chromatograms indicating that no carboxylic acids were produced from sucrose degradation at 2 min of the reaction. However, the HPAEC-PAD analyses in fact showed sucrose degradation (≤ 0.01%) and the presences of glucose and fructose (≤ 0.02%) in the 3.75% and 7.50% (w/w) sucrose mixtures (cf. Appendices, Table A2.4). It should be noted that the amount of sucrose degraded was not significant.

Gas Chromatography Techniques

The products produced from the Fenton oxidation of the individual acids and their mixture were analysed by GC/EI-MS analysis. Gas chromatographic studies for the monitoring of HCA degradation products have not been previously reported. This may be due to the low volatilities of the products.

Figure 5.3 shows the gas chromatograms obtained for SPE extracts of each HCA solution at 2 min of oxidation. The relatively smaller intensities and fewer peaks on the GC chromatogram of degraded CaA (Figure 5.3a) show that CaA has fewer volatile compounds than the other HCAs. The numbers directly labelled on the peaks of the GC chromatograms shown in Figure 5.3 correspond to the identified products listed in Table 5.3. These products have been identified based on their

150

molecular ion and mass fragmentation patterns. A compound identification program of the National Institute of Standards and Technology library (Gaithersburg, MD, USA) was also used to confirm these compounds with a fit value of ≥ 90% in all cases.

The identification program matched several compounds with fitting values of 80–90% to peaks found in the extracts of pCoA and FeA reaction mixtures as shown in Figures 5.3b and 5.3c, respectively. However, the structures of these matching compounds were not strongly associated with any of the products with fit values of ≥ 90% and products detected by other techniques. Therefore, these compounds were not considered for the proposal of mechanistic pathways for the degradation of HCAs.

Table 5.3 Reaction products formed from the Fenton oxidation of HCAs detected by GC/MS.

Peak Compound tR Formula EI/MS Spectrum Ions (min) (m/z) Caffeic acid

8 p–vinylguaiacol 18.09 C9H10O2 150, 135, 107, 77 p–Coumaric acid

9 chavicol 18.52 C9H10O 134, 107, 91, 77

10 4–hydroxybenzaldehyde 18.66 C7H6O2 121, 93, 65, 39

11 4–hydroxybenzoic acid 20.92 C7H6O3 138, 121, 93, 65, 39

12 p–coumaric acid methyl ester 25.06 C10H10O3 178, 147, 119, 91, 65

13 p–coumaric acid 25.68 C9H8O3 164, 147, 119, 107, 91 Ferulic acid

14 p–vinylguaiacol 18.14 C9H10O2 150, 135, 107, 77

15 vanillin 19.26 C8H8O3 152, 123, 109, 81

16 trans–isoeugenol 20.36 C10H12O2 164, 149, 131, 103, 91, 77

17 ferulic acid 27.08 C10H10O4 194, 179, 133

151

(a) 0.05

0.04

)

6 10 × 0.03 8

0.02 Intensity ( Intensity

0.01

0.00 15 20 25 30 35 Retention Time (min) (b) 2.00 10

1.60 ) 6 12 10 1.20 × 11 13 0.80

Intensity ( Intensity 9 0.40

0.00 15 20 25 30 35 Retention Time (min) (c) 2.00

14 17

1.60

) 6

× 10 1.20 15

0.80

16 Intensity ( Intensity

0.40

0.00 15 20 25 30 35 Retention Time (min)

Figure 5.3 Gas chromatograms of SPE extracts of (a) CaA; (b) pCoA and (c) FeA solutions; subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C). Numbers correspond to compound numbers in Table 5.1.

152

5.3.2 Proposed Degradation Pathways of Selected Hydroxycinnamic Acids

On the basis of the results obtained from LC/MS, HPIEC and GC/MS analyses of the various products obtained from oxidation of HCAs using the Fenton process, possible reaction pathways are tentatively proposed in this section with support from the literature.

Hydroxyl radicals are mainly responsible for the degradation of HCAs. Free radicals, in general, are unstable and highly reactive due to their unpaired electron. Hence, it is desirable for an •OH radical to regain a lost electron, misplaced from the catalytic decomposition of H2O2, to become stable. There are three typical fates of these radicals as their main purpose is to become stable in the presence of other molecules: (i) addition to a π-bond; (ii) atom transfer; and (iii) radical combination. However, when a free radical reacts with another compound, it removes an electron from that compound and in turn, that compound becomes a free radical. Hence, this leads to a sequence of reactions until the reaction is terminated when two radicals react with each other to give a non-radical species.

In this context, it is probable that the oxidative degradation of HCAs begins with the electrophilic attack of the •OH radical. Therefore, it was suggested that the position with the highest electron density is the most probable site for the HCAs to be attacked by •OH (Marusawa et al., 2002). Figure 5.4 shows the electrostatic potential maps and the equilibrium geometries of each of the three HCAs investigated, where red indicates a negative charge and blue indicates a positive charge.

Observing the carbon atoms of each HCA molecule, there are various areas coloured in orange and yellow which show a slightly higher negative charge than those areas coloured in blue or green (i.e., positive charge). To determine the intensity of these charges, the electron density distribution was calculated for each atom of each HCA molecule. Table 5.4 shows the natural electron density distribution of the carbon atoms of each HCA molecule. It is obvious from both Figure 5.4 and Table 5.4, that the C8 atom has the highest electron density of –0.371, –0.374 and –0.370 for CaA, pCoA and FeA, respectively. Hence, this is the most potential site for the electrophilic attack of the •OH radical.

153

(a) (b)

O2 H10 O2 H10

H9 H8 C9 C8 C9 C8 H1 H1 O1 O1 C7 H11 C1 C7 O3 C1 C2 C6 H9 C2 C6 H8 H2 C3 H3 O4 C3 C5 O4 C5 C4 C4 H3

H4 H6 H6 H4 H10 (c) O2 C9 C8 H9 O1

H1 C7 C1 O3 H8 C2 H7 C10

C6 H3 H2 C3 C4 H5 O4 C5 H4

Figure 5.4 Electrostatic potential maps and equilibrium geometries of (a) CaA; (b) pCoA and (c) FeA as derived from B3LYP/6-31+G* calculations.

So, the initial degradation pathway is an attack at the C8 atom by the •OH radical. The mechanism proposed by Krimmel et al. (2010), as depicted in Scheme 5.1, shows the formation of a new bond involving the •OH radical and one electron from the π-bond (of the vinyl functional group) of the HCA (1). The other electron from the same π-bond is transferred to the more stable carbon atom (2) (i.e., a secondary (2°) radical). The 2° radical (2) is oxidised in air to form a peroxyl radical (3).

154

Table 5.4 Electron density distribution of carbon atoms in HCA molecules.

Natural atomic charges Caffeic acid p–Coumaric acid Ferulic acid C1 –0.248 –0.179 –0.230 C2 –0.112 –0.130 –0.116 C3 –0.203 –0.178 –0.204 C4 –0.302 –0.314 –0.300 C5 +0.291 +0.347 +0.286 C6 +0.277 –0.288 +0.268 C7 –0.120 –0.119 –0.120 C8 –0.371 –0.374 –0.370 C9 +0.785 +0.785 +0.785 C10 – – –0.313

O H H O H O H O O 1 7 9 H H R 6 2 R R OH OH O2 OH 8 H H 5 H OH OH HO 3 H HO H HO H 4 OH H H H (1) (2) (3)

R = H (pCoumaric acid) R = OH (Caffeic acid)

R = OCH3 (Ferulic acid)

Scheme 5.1

155

The peroxyl radicals (3) can undergo numerous fragmentation and rearrangements reactions. However, based on the works presented by von Sonntag and Schuchmann (1991), it is most probable that a bimolecular reaction occurs between two equivalents of peroxyl radicals (3), as shown in Scheme 5.2. The subsequent losses of oxygen atoms from each of the two peroxyl radicals (3) form oxyl radicals (4). This is then followed by the molecular rearrangement and subsequent fragmentation of the oxyl radicals (4) to produce two equivalents of an aldehyde (5) and another 2° radical (6).

O H O O H O O H H R R -O OH 2 OH 2 H 2 H OH OH HO H HO H H H (3) (4)

H O O R H 2 2 + OH H HO H OH H (5) (6)

R = H (4Hydroxybenzaldehyde) R = OH (Protocatechuic aldehyde)

R = OCH3 (Vanillin)

Scheme 5.2

156

A unimolecular reaction between the newly formed 2° radical (6) and O2 gives an α–hydroxyperoxyl radical (7) as shown in Scheme 5.3. Elimination of the perhydroxyl radical (HO2•) then occurs by the simultaneous dissociation of the C–O bond and the intramolecular transfer of the hydrogen atom from one oxygen atom to another (8–9), which then gives glyoxylic acid (10) (Denisov and Denisova, 2006).

Hence, based on the presences of the phenolic aldehydes produced from the corresponding HCAs as well as the presence of glyoxylic acid detected by HPIEC (cf. Table 5.2), it is suggested that the formation of phenolic aldehydes from HCAs via the Fenton process is likely to occur by the reaction pathways described in Schemes 5.1–5.3.

O O O O O +O2 O O O OH OH O OH O OH H H H H OH OH H O H O

(6) (7) (8) (9)

O O HO2 + OH H (10)

Glyoxylic acid

Scheme 5.3

To date, there has been no work reported on the direct oxidative degradation of phenolic aldehydes in aqueous systems. However, a majority of work published in the literature on the degradation of phenolic compounds has been on phenolic acids, particularly hydroxybenzoic acids (Beltran-Heredia et al., 2001; Heredia et al., 2001; Peres et al., 2004). Aldehydes can easily be oxidised in air to yield carboxylic acids because the hydrogen atom from the –CHO functional group can be abstracted during oxidation (Larkin, 1990). Under the operating conditions of both aqueous and sucrose

157

systems, there are numerous reactions that can take place for the conversion of aldehydes to carboxylic acids such as oxidation with O2 (Larkin, 1990), radical formation (McElroy and Waygood, 1991), Dakin oxidation and the Cannizzaro reaction. However, both the Dakin oxidation and Cannizzaro reactions are optimum under basic conditions, with the latter only applicable to aldehydes without α– hydrogen atoms. It is still possible that these two reactions can take place under mild acidic conditions (e.g., pH 4.0–6.0). However, as an aromatic alcohol was not detected in the present study, the Cannizzaro reaction may not have taken place. Moreover, the Dakin oxidation reaction may not have taken place, as no benezediols or dihydroxybenzenes were detected in the present study.

A more possible oxidation pathway for the conversion of aldehydes to carboxylic acids is via a combination of hydration, radical formation (i.e., hydrogen abstraction) and oxidation reactions (McElroy and Waygood, 1991; Chudasama et al., 2010). The first reaction is a reversible reaction that is in equilibrium between the aldehyde and the aldehyde hydrate (Scheme 5.4). In this case, the phenolic aldehyde (11) rapidly undergoes hydration to form an aldehyde hydrate (12).

O O OH R R R H H H O O H H H OH HO HO H HO

(5) (11) (12)

R = H (4Hydroxybenzaldehyde) R = H, OH, OCH3 R = H, OH, OCH3 R = OH (Protocatechuic aldehyde) Aldehyde Hydrate

R = OCH3 (Vanillin)

Scheme 5.4

As shown in Scheme 5.5, the •OH radical abstracts a hydrogen atom from the aldehyde hydrate (12) forming water and a tertiary (3°) radical (13). A peroxyl radical (14) is formed by the oxygenation of the 3° radical. Dissociation of the C–O bond and intramolecular transfer of the hydrogen atom between oxygen atoms

(15–16), eliminates HO2• and give a phenolic acid (i.e., hydroxybenzoic acid) (17).

158

H OH H OH H OH R R R O O2 H O OH H2O + OH OH OH HO H HO H HO H H H H (12) (13) (14)

R = H, OH, OCH3 R = H, OH, OCH3 R = H, OH, OCH3 Aldehyde Hydrate

H O H OH H OH R R O R O OH O O + HO2 O O H H HO H HO H HO H H H H

(17) (16) (15)

R = H (4Hydroxybenzoic acid) R = H, OH, OCH3 R = H, OH, OCH3 R = OH (Protocatechuic acid) R = OCH3 (Vanillic acid)

Scheme 5.5

The oxidative pathway for the degradation of hydroxybenzoic acids via the Fenton process have been established previously in numerous reports (Rivas et al., 2002; Rivas et al., 2005; Duesterberg and Waite, 2007). Some of the intermediate products from these oxidation reactions such as phenols, quinones and hydroquinones were not detected in the present study. However, several carboxylic acids detected in the present study (viz., oxalic, glyoxylic, formic and acetic) were reported to be produced from the degradation of these intermediates. Hence, it is presumed that the progressive degradation of the HCAs is through the following degradation reaction pathways (Scheme 5.6).

On the basis of the kinetic models developed by Duesterberg and Waite (2007), electrophilic attack by •OH on the hydroxybenzoic acid (17), particularly for 4–hydroxybenzoic and protocatechuic acids, leads to the formation of isomeric hydroxycyclohexadienyl radicals (18) (Scheme 5.6). The isomers (18) are then oxygenated to form two different peroxyl radicals, an 1,3–cyclohexadienyl (19) and an 1,4–cyclohexadienyl radical (20) (Fang et al., 1995; Krimmel et al., 2010).

Elimination of HO2• from the 1,3–cyclohexaidenyl radical (19) produces a

159

hydroxylated phenolic acid (21), meanwhile the 1,4–cyclohexadienyl radical (20) is subjected to further oxidation which may lead to ring-opened products.

H O H O H O R R O R OH OH 2 OH H HO H HO H HO O H OH H OH H OH O (17) (18) (19)

R = H (4Hydroxybenzoic acid) R = H, OH R = H, OH R = OH (Protocatechuic acid)

O2 -HO2

O O H O H O R Ring-Opened Products R OH OH End Products (e.g., Carboxylic Acids) HO H HO H H OH OH (20) (21)

R = H, OH R = H (Protocatehuic acid) R = OH (Gallic acid)

Scheme 5.6

However, for vanillic acid (22), an oxidation product of FeA, the methoxyl group of (23) undergoes oxidative demethoxylation by •OH to produce a phenoxyl radical (24) and methanol, as shown in Scheme 5.7 (O'Neill et al., 1977). The phenoxyl radical (24) reacts with HO2• to form protocatechuic acid (25) and O2. Subsequently, the newly formed protocatechuic acid can be subjected to electrophilic addition by the •OH radical and react in the same manner as described in Scheme 5.6.

160

HO H O H O H O HO H CO 3 H CO O OH 3 OH OH + CH3OH HO H HO H HO H H H (22) (23) (24)

Vanillic acid

+HO2

H O HO OH + O2

HO H H (25) Protocatehuic acid

Scheme 5.7

Based on the reactions discussed so far, oxidation by radicals or oxygenation will not influence any further degradation of the hydroxybenzoic acids. Moreover, dehydration of the hydroxycyclohexadienyl radical (29), initially produced from the electrophilic attack on the hydroxybenzoic acid, would give a phenoxyl radical (Anderson et al., 1987). The phenoxyl radical can rapidly undergo oxidative coupling to form oligomers; or undergo a cycle by reacting with non-reacted hydroxybenzoic acids, quinones or Fe(II)/Fe(III) to produce hydroxylated products (e.g., 4–hydroxybenzoic and protocatechuic acids) (Anderson et al., 1987; Lind et al., 1990; Chen and Pignatello, 1997).

Therefore, it is only possible that the degradation of hydroxylated (or non- hydroxylated) hydroxybenzoic acids can occur through chelating with Fe(II)/Fe(III) as described in Scheme 5.8. The process not only regenerates or converts Fe(II)/Fe(III) vice versa, but also oxidises these acids to give quinones through electron-transfer reactions. The formation of quinones can readily be attacked by •OH radicals and consequently, opens the aromatic ring forming carboxylic acids. In this case, the quinone produced from the reactions between Fe(II)/Fe(III) and protocatechuic acid, is subject to •OH radical attack and decomposes to give ring- opened products such as carboxylic acids (Duesterberg and Waite, 2007).

161

H O H O H O H O

R Fe(III) R -Fe(II) R 1. Fe(III) R OH OH OH OH 2. -Fe(II) HO H HO H HO H O H OH Fe OH OH O (17) (26) (27) (28)

R = H (Protocatehuic acid) R = H, OH R = H, OH R = H, OH R = OH (Gallic acid) OH

Ring-Opened Products End Products (e.g., Carboxylic Acids)

Scheme 5.8

An alternate and more probable pathway for the opening of the aromatic ring than the pathway proposed by Duesterberg and Waite (2007), previously shown in Scheme 5.8 is through the initial decarboxylation of the hydroxybenzoic acid to give a phenol. Subsequent •OH radical attack on phenol would result in ring-opening of the aromatic ring and give a carboxylic acid.

Scheme 5.13 shows the postulated oxidation pathways of protocatechuic acid and gallic acid leading to the formation of carboxylic acids based on the kinetic models developed by Rivas et al., (2005) and Chen and Pignatello (1997). Hydroxyl radical attack on the hydroxybenzoic acid (17) followed by radical abstraction by Fe(III) gives a stable decarboxylated phenol (30). The phenol (30) can then undergo two reaction pathways. The first pathway is with Fe(III) to form a quinone (33) which will then lead to formation of a C6 dicarboxylic acid (34). The other pathway is through hydrogen abstraction, followed by the elimination of the HO2• radical to give an intermediate (32). The intermediate (32) is then subjected to subsequent hydrogen abstraction and HO2• elimination to give a C6 dicarboxylic acid with a hydroxyl group at the C3 position (35). Alternatively, the intermediate (32) can react with Fe(III) to produce a phenoxyl radical intermediate (36).

162

H O H

H OH R H OH + H2O + CO2 HO H HO H OH OH (17) (29)

R = H (Protocatehuic acid) R = H, OH R = OH (Gallic acid) 1. Fe(III) 2. H

H H R H R H 1. Fe(III) Fe(II) + + Fe(II) 2. H HO H HO H OH O (30) (31)

R = H, OH R = H, OH

1. OH 1. Fe(III)

2. O2 2. H

H H H R H R H R H 1. Fe(III) O OH Fe(II) + + HO2 + Fe(II) 2. H HO HO O H OH OH O (36) (32) (33)

R = H, OH R = H, OH R = H, OH

1. OH OH 2. O2

(cf., Scheme 5.10) O R O OH OH HO HO OH O R O

(35) (34) R = H, OH R = H, OH

Scheme 5.9

163

Further oxidation of the phenoxyl radical intermediate (36) gives two equivalents of malonic acid or tartronic acid (38) from protocatechuic acid and gallic acid, respectively (Scheme 5.14). Oxidation of these carboxylic acids (38) could give rise to acetic and glycolic acid (39), which in turn decomposes to formic acid (40) and mineralises to CO2 and H2O (Cui et al., 2012).

H H R H R H 1. Fe(III) O O + Fe(II) 2. H HO HO OH OH (36) (37)

R = H, OH R = H, OH

O O O 2 OH + H2O + CO2 HO OH HO R R

(38) (39)

R = H (Malonic acid) R = H (Acetic acid) R = OH (Tartronic acid) R = OH (Glycolic acid)

HO H

(40)

(Formic acid)

H2O + CO2

Scheme 5.10

164

The C6 dicarboxylic acids produced from the ring cleavage of (32) and (33) by the •OH radical can oxidise and break down to smaller products. For example, as shown in Scheme 5.11, the cleavage of one of the carbon-carbon unsaturated bonds of muconic acid (34), produced from protocatechuic acid, gives rise to two intermediates

(41) and (42). In the presence of O2, •OH radical attack on the intermediate (41) gives oxalic acid (43). Oxalic acid (43) under these conditions readily decomposes to formic acid (40) and mineralises to CO2 and H2O (Rivas et al., 2005; Cui et al., 2012). Moreover, the aldehyde functional group in glyoxylic acid (10), produced from the cleavage of the vinyl functional group of the HCAs (Scheme 5.3), can oxidise in the same manner as described in Scheme 5.4 and 5.5 to give oxalic acid (43).

HO O O O OH OH OH H HO H + HO O O O OH (34) (41) (42)

(Muconic acid)

O O 2 OH OH CO HO + H2O + 2 HO H O (43) (40)

(Oxalic acid) (Formic acid)

H2O + CO2

Scheme 5.11

165

The remaining carboxylic acids detected in the HCA mixtures; cis–aconitic acid, butyric acid, isobutyric acid and lactic acid have not been reported in any of the proposed mechanisms published in the literature. It is probable that these acids may not have originated from the aromatic moiety of the HCAs. It is hypothesised that the formation of butyric, isobutyric and lactic acids may have originated from the aliphatic substituent of the HCAs after bond cleavage of the vinyl functional group. On the other hand, the presence of cis–aconitic acid may have also originated from the aliphatic substituent. Additional carboxylic acid groups on cis–aconitic acid may have originated from other reaction intermediates and products as a result of radical combination.

5.3.3 Oligomerisation of Selected Hydroxycinnamic Acids

From the results obtained from the LC/MS analyses, oligomeric products were formed within 2 min of oxidation via the Fenton process. Oxidative coupling reactions can lead to the formation of dimer products consisting of two equivalents of HCAs (Antolovich et al., 2004). Moreover, reactions with organic radicals produced from HCAs, are susceptible to further polymerisation to form higher oligomers such as trimers and tetramers (Arakawa et al., 2004). Initial oligomerisation of HCAs can possibly lead to the formation of two types of products, cyclodimers and dehydrodimers (Ford and Hartley, 1990; Dobberstein and Bunzel, 2010). However, cyclodimerisation reactions are photochemical and not dependent on reactions involving free radicals (i.e., radical coupling) (Ford and Hartley, 1989).

In Scheme 5.1, the electrophilic addition of the •OH radical on a HCA was shown. However, radicals in general not only partake in addition reactions but can also go through abstraction and radical combination reactions as well. In Scheme 5.12, hydrogen atom abstraction by the •OH radical from the phenolic functional group can occur, giving rise to the formation of a cinnamoyl radical (44).

166

O O R R OH OH H2O + H O O OH (1) (44)

R = H (pCoumaric acid) R = H, OH, OCH3 R = OH (Caffeic acid)

R = OCH3 (Ferulic acid)

Scheme 5.12

Possible resonance structures of the cinnamoyl radical are shown in Scheme 5.13, where (45) and (48) are the least and most stablilised forms of the radical by resonance, respectively. Due to electron delocalisation of this radical, the coupling of two cinnamoyl radicals (i.e., radical combination) can theoretically give rise to the formation of dimers through 5–5-, 4–O–5-, 8–5-, 8–8-, and 8–O–4-coupling (Bunzel, 2010).

O OH O OH O OH O OH 9

7 8 1 2 6 3 R 5 R R R 4 O O O O

(45) (46) (47) (48)

R = H, OH, OCH 3 R = H, OH, OCH3 R = H, OH, OCH3 R = H, OH, OCH3

Scheme 5.13

167

The structure of the FeA dimer was determined on the basis of the MS fragments obtained from the LC/MS data with confirmation from literature data

(Antolovich et al., 2004). The MS spectrum for the peak at tR = 18.87 min (cf. Appendices, Figure A2.4), tentatively assigned to the FeA dimer shows fragments at m/z 385, 341, 297 and 155. Losses of two CO2 molecules suggest that the formation of the FeA dimer, under the Fenton conditions investigated was due to the 5–5-coupling of feruloyl radicals (Scheme 5.14).

+ + H H O

OCH3 OCH3 HO H2C

OH OH CO2 HO HO

OH OH H3CO H3CO O O

m/z 385 m/z 341

CO2

+ H OCH3 H2C

H O O H

CH2 H3CO

m/z 297 Scheme 5.14

168

The structure for the tetramer of CaA was not determined due to insufficient information obtained from the MS data. The MS spectrum of the peak at tR = 12.75 min tentatively assigned to the tetramer of CaA had fragment ions of m/z 715, 471, 357, 269 and 145 (cf. Appendices, Figure A2.5).

Caffeic acid, [M]– has a m/z of 179 and its deprotonated dimer from any of the five suggested radical coupling reaction pathways would give a m/z of 357 (Arakawa et al., 2004). Hence, further deprotonation between the coupling of two equivalents of CaA dimers or the subsequent coupling of two caffeoyl radicals in succession to the CaA dimer would give rise to a tetramer that would have a m/z ratio of 713 (Pati et al., 2006), not a m/z ratio of 715. Representative structures of other tetramers of CaA that exist naturally in plants have a m/z ratio of 713 (Bunzel, 2010). So, it is probable that the m/z 715 ion is an adduct of the CaA dimer.

The fragmentation of m/z 357 proposed by Pati et al. (2006) is given in

Scheme 5.15. Loss of two CO2 molecules from the 5–5-coupled dimer (m/z 357) gives a structure for the m/z 269 ion. Ring opening of one of the aromatic moieties of the dimer and subsequent rearrangement reactions lead to the formation of a quinone fragment ion (m/z 159) and a neutral fragment (M = 110). The m/z ion of 393 is – possibly attributable to the CaA dimer and two H2O molecules [M–2H2O] .

The m/z ions 471 and 715 could not be elucidated on the basis of the current information ascertained. Agha et al. (2009) proposed a structure for a tetramer of CaA with a m/z ratio of 715, showing the coupling of a 8–8-coupled dimer and a 5–5- coupled dimer, both at the O–4 positions (Figure 5.5). However, it is not possible for the oxygen atoms to be coupled under these circumstances due to the instability of the O–O single bond. In addition, it is highly unlikely for a phenolic group in only one of the CaA moieties to exist in a semiquinone form within the tetramer. From the foregoing, the structure of the presumed CaA tetramer (or CaA dimer adduct) assigned to the m/z ion of 715 is not known.

169

+ + O H H

OH OH HO H2C

2CO OH 2 H O HO O H

OH CH2 HO HO

m/z 357 m/z 269

+ + H H

CH2 CH2 O O H

O O O O

CH2 H O CH2 HO

m/z 269 m/z 269

+ H H2C CH2 O + O

CH2 O O

m/z 159 M = 110

Scheme 5.15

170

+ H O

OH HO O

OH OH HO O

O O OH HO OH O HO O

Figure 5.5 Proposed structure of a tetramer of caffeic acid (m/z 715) by Agha et al., (2009).

5.4 Summary

This chapter outlines an intense treatise in organic chemistry in an attempt to determine whether the degradation products of three selected phenolic acids when subjected to Fenton oxidation were liable to be colour precursors. Attempts were made to identify and propose the tentative reaction pathways of oxidation products of CaA, pCoA and FeA via the Fenton process. Eleven aromatic products and eight aliphatic products were identified. Cleavage of the conjugated vinyl substituent of the HCAs by electrophilic addition of •OH; and hydrogen abstraction from the phenol group of the HCAs were the two major mechanisms initiating the degradation of HCA via the Fenton oxidation process. The initial products undergo a series of successive oxidation steps which lead to the formation of carboxylic acids.

171

References

Agha, A. E., Abbeddou, S., Makris, D. P., & Kefalas, P. (2009). Biocatalytic properties of a peroxidase-active cell-free extract from onion solid wastes: caffeic acid oxidation. Biodegradation, 20(2), 143-153.

Anderson, R. F., Patel, K. B., & Stratford, M. R. L. (1987). Radical spectra and product distribution following electrophilic attack by the OH[radical dot] radical on 4-hydroxybenzoic acid and subsequent oxidation. Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases, 83(10), 3177-3187.

Antolovich, M., Bedgood, D. R., Jr., Bishop, A. G., Jardine, D., Prenzler, P. D., & Robards, K. (2004). LC-MS investigation of oxidation products of phenolic antioxidants. Journal of Agricultural and Food Chemistry, 52(4), 962-971.

Arakawa, R., Yamaguchi, M., Hotta, H., Osakai, T., & Kimoto, T. (2004). Product analysis of caffeic acid oxidation by on-line electrochemistry/electrospray ionization mass spectrometry. Journal of the American Society for Mass Spectrometry, 15, 1228-1236.

Becke, A. D. (1988). Density-functional exchange-energy approximation with correct asymptotic behavior. Physical Review A, 38(6), 3098-3100.

Beltran-Heredia, J., Torregrosa, J., Dominguez, J. R., & Peres, J. A. (2001). Comparison of the degradation of p-hydroxybenzoic acid in aqueous solution by several oxidation processes. Chemosphere, 42, 351-359.

Bunzel, M. (2010). Chemistry and occurrence of hydroxycinnamate oligomers. Phytochemistry Reviews, 9, 47-64.

Chen, R., & Pignatello, J. J. (1997). Role of quinone intermediates as electron shuttles in Fenton and photoassisted Fenton oxidations of aromatic compounds. Environmental Science and Technology, 31, 2399-2406.

Chudasama, V., Fitzmaurice, R. J., & Caddick, S. (2010). Hydroacylation of α,β- unsaturated esters via aerobic C–H activation. Nature Chemistry, 2(7), 592- 596.

Cilliers, J. J. L., & Singleton, V. L. (1991). Characterization of the products of nonenzymic autoxidative phenoilc reactions in a caffeic acid model system. Journal of Agricultural and Food Chemistry, 39(7), 1298-1303.

Cui, Y.-H., Feng, Y.-J., Liu, J., & Ren, N. (2012). Comparison of various organic compounds destruction on rare earths doped Ti/Sb-SnO2 electrodes. Journal of Hazardous Materials, 239–240(0), 225-232.

Denisov, E. T., & Denisova, T. G. (2006). Kinetics and thermodynamics of formation and degradation of α-hydroxyperoxyl radicals. Petroleum Chemistry, 46(6), 373-383.

172

Dobberstein, D., & Bunzel, M. (2010). Identification of ferulate oligomers from corn stover. Journal of the Science of Food and Agriculture, 90(11), 1802-1810.

Duesterberg, C. K., & Waite, T. D. (2007). Kinetic modeling of the oxidation of p- hydroxybenzoic acid by Fenton's reagent: Implications of the role of quinones in the redox cycling of iron. Environmental Science & Technology, 41(11), 4103-4110.

Fang, X., Pan, X., Rahmann, A., Schuchmann, H.-P., & von Sonntag, C. (1995). Reversibility in the reaction of cyclohexadienyl radicals with oxygen in aqueous solution. Chemistry – A European Journal, 1(7), 423-429.

Fifen, J. J., Nsangou, M., Dhaouadi, Z., Motapon, O., & Lahmar, S. (2009). Single or double hydrogen atom transfer in the reaction of metal – associated phenolic acids with ˙OH radical: DFT study. Journal of Molecular Structure: THEOCHEM, 901, 49-55.

Ford, C. W., & Hartley, R. D. (1989). GC/MS characterisation of cyclodimers from p- coumaric and ferulic acids by photodimerisation—a possible factor influencing cell wall biodegradability. Journal of the Science of Food and Agriculture, 46(3), 301-310.

Ford, C. W., & Hartley, R. D. (1990). Cyclodimers of p-coumaric and ferulic acids in the cell walls of tropical grasses. Journal of the Science of Food and Agriculture, 50(1), 29-43.

Frisch, M. J., Pople, J. A., & Binkley, J. S. (1984). Self-consistent molecular orbital methods 25. Supplementary functions for Gaussian basis sets. The Journal of Chemical Physics, 80(7), 3265-3269.

Fulcrand, H., Cheminat, A., Brouillard, R., & Cheynier, V. (1994). Characterization of compounds obtained by chemical oxidation of caffeic acid in acidic conditions. Phytochemistry, 35(2), 499-505.

Heredia, J. B. D., Torregrosa, J., Dominguez, J. R., & Peres, J. A. (2001). Kinetic model for phenolic compound oxidation by Fenton's reagent. Chemosphere, 45, 85-90.

Krimmel, B., Swoboda, F., Solar, S., & Reznicek, G. (2010). OH-radical induced degradation of hydroxybenzoic- and hydroxycinnamic acids and formation of aromatic products—A gamma radiolysis study. Radiation Physics and Chemistry, 79, 1247-1254.

Larkin, D. R. (1990). The role of catalysts in the air oxidation of aliphatic aldehydes. The Journal of Organic Chemistry, 55(5), 1563-1568.

Lee, C., Yang, W., & Parr, R. G. (1988). Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. Physical Review B, 37(2), 785-789.

173

Lind, J., Shen, X., Eriksen, T. E., & Merenyi, G. (1990). The one-electron reduction potential of 4-substituted phenoxyl radicals in water. Journal of the American Chemical Society, 112(2), 479-482.

Marusawa, H., Ichikawa, K., Narita, N., Murakami, H., Ito, K., & Tezuka, T. (2002). Hydroxyl radical as a strong electrophilic species. Bioorganic & Medicinal Chemistry, 10(7), 2283-2290.

McElroy, W. J., & Waygood, S. J. (1991). Oxidation of formaldehyde by the hydroxyl radical in aqueous solution. Journal of the Chemical Society, Faraday Transactions, 87(10), 1513-1521.

Neyens, E., & Baeyens, J. (2003). A review of classic Fenton’s peroxidation as an advanced oxidation technique. Journal of Hazardous Materials, B98, 33-50.

Nsangou, M., Fifen, J. J., Dhaouadi, Z., & Lahmar, S. (2008). Hydrogen atom transfer in the reaction of hydroxycinnamic acids with •OH and •HO2 radicals: DFT study. Journal of Molecular Structure: THEOCHEM, 862, 53-59.

O'Neill, P., Schulte-Frohlinde, D., & Steenken, S. (1977). Formation of radical cations and zwitterions versus demethoxylation in the reaction of OH with a series of methoxylated benzenes and benzoic acids. An example of the electrophilic nature of the OH radical. Faraday Discussions of the Chemical Society, 63(0), 141-148.

Pati, S., Losito, I., Palmisano, F., & Zambonin, P. G. (2006). Characterization of caffeic acid enzymatic oxidation by-products by liquid chromatography coupled to electrospray ionization tandem mass spectrometry. Journal of Chromatography A, 1102(1–2), 184-192.

Peres, J. A. S., Melo de Carvalho, L. H., Boaventura, R. A. R., & Costa, C. A. V. (2004). Characterisitics of p-hydroxybenzoic acid oxidation using Fenton's reagent. Journal of Environmental Science and Health, Part A: Toxic/Hazardous Substances & Environmental Engineering, A39(11-12), 2897-2913.

Poerschmann, J., Trommler, U., & Górecki, T. (2010). Aromatic intermediate formation during oxidative degradation of Bisphenol A by homogeneous sub- stoichiometric Fenton reaction. Chemosphere, 79, 975-986.

Pulay, P. (1980). Convergence acceleration of iterative sequences. the case of scf iteration. Chemical Physics Letters, 73(2), 393-398.

174

Šmejkalová, D., Piccolo, A., & Spiteller, M. (2006). Oligomerization of humic phenolic monomers by oxidative coupling under biomimetic catalysis. Environmental Science & Technology, 40(22), 6955-6962.

Tazaki, H., Taguchi, D., Hayashida, T., & Nabeta, K. (2001). Stable isotope-labeling studies on the oxidative coupling of caffeic acid via o-quinone. Bioscience, Biotechnology, and Biochemistry, 65(12), 2613-2621.

Thai, C. C. D., & Doherty, W. O. S. (2011). The composition of sugarcane juices derived from burnt cane and whole green cane crop. Proceedings of the Australian Society of Sugar Cane Technologists, 33

Van Voorhis, T., & Head-Gordon, M. (2002). A geometric approach to direct minimization. Molecular Physics, 100(11), 1713-1721. von Sonntag, C., & Schuchmann, H.-P. (1991). The elucidation of peroxyl radical reactions in aqueous solution with the help of radiation-chemical methods. Angewandte Chemie International Edition in English, 30(10), 1229-1253.

175

176

CHAPTER 6

Degradation of Melanoidin and

Hydroxycinnamic Acid Mixtures

6.1 Introduction...... 177 6.2 Materials and Methods...... 178 6.2.1 Reagents and Solvents...... 178 6.2.2 Preparation of Synthetic Melanoidin...... 179 6.2.3 Modified Fenton Oxidation Process...... 179 6.2.4 Instrumental Procedures and Analyses...... 179 6.2.5 Performance Assessment of the Modified Fenton Oxidation Process...... 180 6.2.6 Design of Experiments...... 181 6.2.7 Statistical Analysis...... 182 6.3 Results and Discussion...... 182 6.3.1 Monitoring Melanoidin and Hydroxycinnamic Acid Degradation...... 182 6.3.2 Transformation of Data, Regression Modelling and Statistical Analysis...... 184 6.3.3 Oxidation Performance of Melanoidins...... 190 6.3.4 Oxidation Performance of Hydroxycinnamic Acids...... 194 6.3.5 Response Surface Analyses for the Decolourisation of Mixtures...... 198 6.3.6 Process Optimisation and Validation...... 200 6.4 Summary...... 203

177

6.1 Introduction

The Fenton process has been shown to oxidise and degrade model colour precursor compounds in water and sucrose solutions. However, the numerous oxidation products formed and the presence of starting materials (viz., CaA, pCoA and FeA) indicate that the Fenton process on its own, would not effectively decolourise factory sugar cane juice.

Dwyer et al., (2009) studied the removal of factory and synthetic melanoidins (factory produced colourants during sugar manufacture), using hydrated aluminium sulfate (Al2(SO4)3.xH2O). The results from the study showed that an Al2(SO4)3 dose of 30 mg/L as aluminium ion (Al(III)) was sufficient to remove 75% of colour from factory effluents contaminated with melanoidins. Aluminium is known for its significant pro-oxidant activity, and so can enhance the Fenton process by reducing Fe(III) to Fe(II) under the presence of superoxide (Exley, 2004; Ruipérez et al.,

2012). This is because Fe(II) is a more effective catalyst for H2O2 decomposition than Fe(III).

This study builds on these investigations by examining the degradation and decolourisation of a complex mixture of a synthetic melanoidin and HCAs using a modified Fenton process consisting of Fe(II), Al(III) and H2O2.

6.2 Materials and Methods

6.2.1 Reagents and Solvents

All chemicals, solvents and reagents were of AR grade and obtained from the suppliers as described in the previous chapters or as otherwise stated. Aluminium chloride hexahydrate (AlCl3·6H2O) and glycine were supplied from Merck (Darmstadt, Germany). Stock solutions of hydroxycinnamic acids, HCAs (i.e., CaA, pCoA and FeA) were prepared individually by dissolution in degassed ethanol solution (50% (v/v)) and stored at 4.0 °C.

178

6.2.2 Preparation of Synthetic Melanoidin

The synthesis procedure used for the preparation of a synthetic melanoidin was adapted from Shore et al., (1984). Glucose (72 g) and glycine (30 g) were dissolved in water (60 mL) and incubated at 50 °C for 72 h. The resulting co-polymer was then stored at 4.0 °C.

6.2.3 Modified Fenton Oxidation Process

The procedure for the degradation of the mixtures is similar to that described in Sections 4.2.4 and 5.2.2. In each run, a predetermined amount of Milli-Q water, melanoidin, sucrose and each HCA (equivalent mg/L concentration) were added to the reaction vessel. Known amounts of FeSO4·7H2O, AlCl3·6H2O and H2O2 solutions were added to achieve a final volume of 50 mL, while maintaining the working

Fenton molar ratio (Fe(II)/H2O2) at 1:15. The final sucrose concentration in each mixture was 15% (w/w). At 2 min, the reaction was immediately quenched by neutralising the mixture to a pH of 7.0 with 2.0 M NaOH and stored at –80 °C. Samples were defrosted and prepared for instrumental analysis.

6.2.4 Instrumental Procedures and Analyses

HPLC-DAD/FLD. The proportion of each HCA and melanoidin degraded were monitored by reversed-phase HPLC-DAD and fluorescence detection (FLD). The analysis was performed on a Hewlett Packard HP/Agilent 1100 Series HPLC system (G1379A micro-degasser, Japan; G1311A quaternary pump, Germany; G1313A Automatic Liquid Sampler (ALS), Germany; G1315B DAD, Germany; G1321A FLD, Germany) using a Waters Symmetry C18 column (150 × 3.9 mm i.d.) with a Waters Guard-Pak guard holder containing a Waters Guard-Pak Resolve C18 guard insert (10 μm) (Milford, MA, USA). The mobile phase consisted of 1.0% (v/v) acetic acid in water (as eluent A) and methanol (as eluent B). The gradient program was as follows: 0% B to 10% B (1 min), 10% B to 20% B (1 min), 20% B to 25% B (3 min), 25% B to 50% B (15 min) and 50% B to 20% B (5 min). Simultaneous UV/Vis detection at specific wavelengths (280 nm and 320 nm) subtracted against a

179

reference wavelength (620 nm). Fluorescence detection was performed at an excitation wavelength (λex) of 350 nm and an emission wavelength (λem) of 445 nm. Aliquots of samples were membrane filtered (0.45 μm) prior to injection into the HPLC system. Injection volume for all samples was 50 μL; column temperature was ambient; flow rate was 1.0 mL/min and run time was 25 min. After each run, the chromatographic system was equilibrated for 5 min. Data acquisition was performed using the Agilent ChemStation (Rev. A.09.03) software package. Identification of peaks was based on the conformance of UV/Vis spectra and retention times with the corresponding authentic standards.

HPAEC-PAD. Sucrose and reducing contents in the reactions were monitored by HPAEC-PAD. Sample preparation and chromatographic conditions are described in Section 4.2.5.

6.2.5 Performance Assessment of the Modified Fenton Oxidation Process

The efficiency of the modified Fenton process on the degradation of the CaA, pCoA and FeA was determined individually based on the change in absorbance of the corresponding HPLC chromatographic peak using Equation 4.1. The degradation efficiency of the melanoidin was determined on the basis of the changes in luminescence of the corresponding HPLC chromatographic peak using Equation 6.1:

LL (6.1) % Melanoidin degradation = 0 t  100 L0

where, L0 initial luminescence of melanoidin in LU (at t = 0 min)

Lt luminescence of melanoidin in LU at time of aliquot taken (at t = 2 min)

The decolourisation efficiency of the synthetic mixtures was determined based on the change in colour of the mixtures prior to oxidation and at 2 min, using Equation 6.2. Procedures for the measurements of colour, RI and TSS are described in Section 3.2.6.

180

Colour Colour (6.2) % Decolourisation = 0 t  100 Colour0

where, Colour0 initial colour of the mixture in IU (at t = 0 min)

Colourt colour of the mixture in IU at time of aliquot taken (at t = 2 min)

6.2.6 Design of Experiments

Design of experiments, mathematical modelling and optimisation of process parameters were evaluated using the Stat-Ease, Inc. Design-Expert 7.0.0 software package (Minneapolis, MN, USA).

A rotatable circumscribed CCD with a half-fractional factorial was used to evaluate the main effect for each condition and the possible interactive effects on the residual stresses between two variables. The process parameters (independent variables) used in this study were the melanoidin concentration (x1), the initial total

HCA concentration (x2), the solution pH (x3), FeSO4·7H2O dosage (x4) and the

AlCl3·6H2O (x5). The selected response factors (dependent variable) for optimisation were % melanoidin degradation (y1), % total HCA degradation (y2) and

% decolourisation (y3). Sucrose concentration and temperature parameters required no further optimisation and remained constant at 15% (w/w) and 35 °C respectively, to closely mimic conditions of MJ during the sugar manufacturing process. The coded and actual values of each variable and their levels for the experimental design used in the study are shown in Table 6.1.

181

Table 6.1 Coded and actual values of the experimental design.

Coded Levels of Parameters Notation Factor Unit –2 –1 0 +1 +2

A (x1) Melanoidin mg/L 0 500 1000 1500 2000

B (x2) Total HCA mg/L 0 50 100 150 200

C (x3) pH 4.50 4.88 5.25 5.63 6.00

D (x4) FeSO4·7H2O mg/L 86.0 238 389 541 692

E (x5) AlCl3·6H2O mg/L 0 100 200 300 400

The design consisted of a 2k factorial augmented by 2k axial points and a centre point, where k is the number of factors investigated (k = 5). For this study, when the one-half fraction is used in the factorial portion of the CCD, a total of 32 experiments were conducted in random order with 16 factorial points, 10 axial points and 1 centre point (duplicated 5 times). Duplicate runs were required for experimental error calculation.

6.2.7 Statistical Analysis

Analysis of variance was used for model adequacy and analysis of the experimental data. The quality of the fit polynomial model was expressed by the regression coefficient, R2 and its statistical significance was checked using Fisher’s F-test. Model terms were determined based on the significance of each term at a confidence level of 95%.

6.3 Results and Discussion

6.3.1 Monitoring Melanoidin and Hydroxycinnamic Acid Degradation

The reversed-phase HPLC analyses show that the melanoidin components were eluted between tR = 0.85–5.0 min (Figure 6.1), while the HCAs were eluted between tR = 8.5–13.5 min (Figure 6.2). The FLD method is a far more superior detection method than the DAD method for the monitoring of melanoidins and humic-

182

like substance, as evident from the chromatogram of Figures 6.1 and 6.2 (Westerhoff et al., 2001). Meanwhile, the latter is more sensitive for the detection of HCAs.

Figure 6.1 shows that the synthetic glucose-glycine melanoidin consists of several products which are relatively polar in comparison to the non-polar HCAs. A distinctive large peak prior to oxidation (t = 0 min) at tR = 2.11 min was chosen to monitor the degradation of the melanoidin. At 2 min, the peaks at tR = 1.79 and 2.11 min at tR = 1.79 min are reduced in size. However, the peak at tR = 1.18 min increased in size. This probably suggests that the components associated with tR = 1.79 and 2.11 min are oxidised to polar compounds at 2 min. On the other hand, a large response was observed for the peak at tR = 1.18 min after oxidation. This suggests that the two components of the melanoidin present at t = 0 are being oxidised to form polar compounds.

t = 0 min

Absorbance (Arbitrary Units) Absorbance(Arbitrary t = 2 min

0 2 4 6 8 10 Retention Time (min)

Figure 6.1 Typical HPLC-FLD chromatogram (fluorescence detection at

λex = 350 nm and λem = 445 nm) of the melanoidin/HCA mixture in sucrose solution (15% (w/w)) before and after modified Fenton oxidation (t = 2 min) at pH 5.6 and 35 °C.

183

2 1 3

t = 0 min

t = 2 min Absorbance (Arbitrary Units) Absorbance(Arbitrary

4 6 8 10 12 14 Retention Time (min)

Figure 6.2 Typical HPLC-DAD chromatogram (UV/Vis detection at 280 nm) of the melanoidin/phenolic acid mixture in sucrose solution (15% (w/w)) before and after modified Fenton oxidation (t = 2 min) at pH 5.6 and 35 °C. 1 = CaA, 2 = pCoA, 3 = FeA.

6.3.2 Transformation of Data, Regression Modelling and Statistical Analysis

Non-linearity of normal probability plots of residuals for the fitted models of melanoidin and total HCA degradation were resolved via Box-Cox power transformation (cf. Appendices, Figure A3.1). The optimum λ values determined by the minimum of the curve of the Box-Cox plots for the degradation of the melanoidin and total HCA were –3.00 and –0.35 respectively (cf. Appendices, Figure A3.2). The fitted model for decolourisation did not require any data transformation (λ = 1.00) as the internally studentised residual points resembled a linear curve. The data for all fitted response surface models show good correspondence to a normal distribution and validated the normality assumption.

On the basis of the sequential model sum of squares (Type I), the power transformed response surface models for melanoidin degradation (y1) and total HCA degradation (y2); as well as the response surface model for decolourisation (y3) were selected based on the highest order polynomial, where the additional model terms were significant and the models were not aliased. The degradation data obtained for the melanoidin and total HCA responses fit a two-factor interaction (2FI) function,

184

while the data for the decolourisation of the mixtures fits a quadratic polynomial function.

Selection of significant coefficients and removal of unimportant model terms for each model were identified on the basis of ANOVA statistics and stepwise regression at an alpha-to enter and alpha-to-exit significance level of 0.1. The chosen stepwise alpha range applied to all three response surface models should result in final models with significant model terms included at the approximate 95% confidence level.

The ANOVA results for the partial sum of squares (Type III) for the three response surface reduced quadratic or 2FI models after stepwise regression are shown in Tables 6.2–6.4. The analyses indicate that most of the independent variables and some of the interactions are significant and contribute to the degradation of the melanoidin and the HCAs, as well as the decolourisation of the mixtures. The model F-values of 8.09, 8.96 and 18.27 for melanoidin degradation, total HCA degradation and decolourisation respectively, imply that the models are significant. There is only a 0.01% chance that a model F-value this large could occur due to noise. The lack of fit F-values of 0.47, 0.72 and 1.91 for the melanoidin degradation, total HCA degradation and decolourisation models in that order imply that the lack of fit is not significant relative to pure error. There is a 87.52%, 70.95% and 27.86% chance respectively that the lack-of-fit F-values this large would occur due to noise. Non- significant lack of fit is good as it confirms the predictability of the model.

The independent variables in the models were initial melanoidin concentration, initial total HCA concentration, solution pH, FeSO4·7H2O dosage and

AlCl3·6H2O dosage; and were coded A, B, C, D and E respectively. The final empirical quadratic equations in terms of coded factors for each response after the exclusion of the insignificant model terms (p > 0.1000), unless retained to make the models hierarchical, are as follows:

185

Melanoidin degradation (%)

–3 –6 –7 –7 (y1) = 3.447 × 10 – 1.614 × 10 A + 1.286 × 10 B (6.3) +1.764 × 10–7C – 8.041 × 10–8D – 1.853 × 10–8E + 1.938 × 10–7AE + 2.606 × 10–7CD + 1.362 × 10–7CE – 4.267 × 10–7DE

Total HCA degradation (%)

–3 –4 –3 (y2)–0.35 = 0.26 + 1.094 × 10 A – 1.207 × 10 B + 1.219 × 10 C (6.4) – 2.202 × 10–3D + 5.081 × 10–4E + 1.737 × 10–3AC – 4.064 × 10–3AD + 1.926 × 10–3BD – 2.491 × 10–3BE Decolourisation (%)

(y3) = 13.78 + 12.06A – 1.49B + 7.11C – 9.66D – 3.34E (6.5) + 14.83AE + 23.60BC – 16.89BD – 12.99BE + 9.12CD – 13.28DE – 10.78A2 + 3.89E2

The predicted R2 values of all response surface models are in reasonable agreement with the adjusted R2 values, which show that the fitted models are adequate. Plots of the predicted response against the experimental values for the three models show good linearity, indicating that the developed mathematical models are suitable for predicting melanoidin degradation, total HCA degradation and decolourisation (cf. Appendices, Figures A3.2–A3.3).

186

Table 6.2 Results of ANOVA for model terms of the response surface reduced two-factor interaction model for melanoidin degradation.*

Source SS df Mean Sq. F-value p-value Remarks Model 5.22 × 10–12 9 5.80 × 10–12 8.09 < 0.0001 Significant A 4.61 × 10–13 1 4.61 × 10–13 6.43 0.0207 Significant B 3.65 × 10–13 1 3.65 × 10–13 5.09 0.0368 Significant C 5.68 × 10–13 1 5.68 × 10–13 7.92 0.0115 Significant D 1.15 × 10–13 1 1.15 × 10–13 1.60 0.2224 Insignificant E 7.58 × 10–15 1 7.58 × 10–15 0.11 0.7488 Insignificant AE 5.31 × 10–13 1 5.31 × 10–13 7.41 0.0140 Significant CD 9.60 × 10–13 1 9.60 × 10–13 13.39 0.0018 Significant CE 2.62 × 10–13 1 2.62 × 10–13 3.66 0.0719 DE 2.57 × 10–12 1 2.57 × 10–12 35.90 < 0.0001 Significant Residual 1.29 × 10–12 18 7.17 × 10–12 Lack of Fit 7.08 × 10–13 13 5.44 × 10–13 0.47 0.8752 Insignificant Pure Error 5.83 × 10–13 5 1.17 × 10–13 Corr. Total 6.51 × 10–12 27 Criteria Standard Deviation 2.68 × 10–07 Mean 3.41 × 10–06 CV (%) 7.85 PRESS 3.14 × 10–12 R2 0.80 Adjusted R2 0.70 Predicted R2 0.52 Adequate Precision 11.45 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

187

Table 6.3 Results of ANOVA for model terms of the response surface reduced two-factor interaction model for total HCA degradation.*

Source SS df Mean Sq. F-value p-value Remarks Model 6.20 × 10–04 9 6.88 × 10–05 8.96 < 0.0001 Significant A 2.27 × 10–05 1 2.27 × 10–05 2.95 0.1038 Insignificant B 2.19 × 10–07 1 2.19 × 10–07 0.028 0.8679 Insignificant C 3.16 × 10–05 1 3.16 × 10–05 4.11 0.0585 Insignificant D 8.18 × 10–05 1 8.18 × 10–05 10.64 0.0046 Significant E 5.49 × 10–06 1 5.49 × 10–06 0.71 0.4098 Insignificant AC 3.46 × 10–05 1 3.46 × 10–05 4.50 0.0489 Significant AD 1.89 × 10–04 1 1.89 × 10–04 24.63 0.0001 Significant BD 4.25 × 10–05 1 4.25 × 10–05 5.53 0.0310 Significant BE 7.11 × 10–05 1 7.11 × 10–05 9.25 0.0074 Significant Residual 1.31 × 10–04 17 7.69 × 10–06 Lack of Fit 9.15 × 10–05 13 7.04 × 10–06 0.72 0.7095 Insignificant Pure Error 3.92 × 10–05 4 9.80 × 10–06 Corr. Total 7.50 × 10–04 26 Criteria Standard Deviation 2.77 × 10–03 Mean 0.26 CV (%) 1.08 PRESS 3.10 × 10–04 R2 0.83 Adjusted R2 0.73 Predicted R2 0.59 Adequate Precision 13.55 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

188

Table 6.4 Results of ANOVA for model terms of the response surface reduced quadratic model for decolourisation.*

Source SS df Mean Sq. F-value p-value Remarks Model 26,633.27 13 2,048.71 18.27 < 0.0001 Significant A 1,691.99 1 1,691.99 15.09 0.0019 Significant B 42.23 1 42.23 0.38 0.5500 Insignificant C 762.90 1 762.90 6.80 0.0217 Significant D 1,792.63 1 1,792.63 15.99 0.0015 Significant E 214.13 1 214.13 1.91 0.1903 Insignificant AE 2,557.93 1 2,557.93 22.81 0.0004 Significant BC 6,452.02 1 6,452.02 57.54 < 0.0001 Significant BD 3,321.20 1 3,321.20 29.62 0.0001 Significant BE 1,962.16 1 1,962.16 17.50 0.0011 Significant CD 967.59 1 967.59 8.63 0.0115 Significant DE 2,043.80 1 2,043.80 18.23 0.0009 Significant A2 1,536.86 1 1,536.86 13.71 0.0027 Significant E2 419.92 1 419.92 3.74 0.0750 Residual 1,457.76 13 112.14 Lack of Fit 1,182.55 9 131.39 1.91 0.2786 Insignificant Pure Error 275.20 4 68.80 Corr. Total 28,091.03 26 Criteria Standard Deviation 10.59 Mean 14.17 CV (%) 74.74 PRESS 6,824.48 R2 0.95 Adjusted R2 0.90 Predicted R2 0.76 Adequate Precision 22.46 *SS (Sum of Squares), df (degrees of freedom), Mean sq. (Mean Square), Corr. (Corrected)

189

6.3.3 Oxidation Performance of Melanoidins

On the basis of the coefficients of the first-order model terms in the cubed inverse model for melanoidin degradation (Equation 6.3), it is obvious that the degradation efficiency of the melanoidin increases with increasing initial melanoidin concentration (A). However, it is evident that the degradation efficiency decreases with increasing initial HCA concentration (B) and at higher initial pH (C). There were no significant positive effects on the degradation of melanoidins at varying dosages of FeSO4·7H2O (D) and AlCl3·6H2O (E). This suggests that the ranges of the working dosages for the Fenton and Al(III) additives used in this study require no further optimisation and can degrade melanoidins up to 2,000 mg/L in a sugar solution containing 15% (w/w) sucrose. Hence, it is possible to use lower dosages of both Fenton and Al(III) additives to effectively degrade melanoidins. However, based on the perturbation plot for melanoidin degradation (Figure 6.3), the Fenton reagents are more effective than Al(III) in the degradation of melanoidins. The plot confirms that the presence of HCAs has a large influence on the degradation of melanoidins as the •OH radicals produced during the modified Fenton process have a stronger preference to oxidise HCAs than melanoidins.

The significant two-factor interactive parameters for the degradation of melanoidins in the mixture via the modified Fenton oxidation process are melanoidin concentration and AlCl3·6H2O dosage (AE); pH and FeSO4·7H2O dosage (CD); pH and AlCl3·6H2O dosage (CE); and FeSO4·7H2O dosage and AlCl3·6H2O dosage (DE). Contour plots (Figures 6.4 and 6.5) were used to investigate the relationships between the pairs of interactive parameters of the developed model (Equation 6.3). Amongst the pairs, CD (p = 0.0018) and DE (p < 0.0001) were the most statistically significant interactions followed by AE (p = 0.0140) then CE (p = 0.0719). Hence, the corresponding plots for AE and CE shown in Figures 6.4a and 6.5a only show slight changes in colour associated to the amount of melanoidin degraded (64–67%). This indicates that Al(III) contributes to the degradation of the melanoidin, only to a small extent, with changing initial melanoidin concentration or solution pH. Therefore, it is not necessary to use higher dosages of Al(III) with the Fenton process to degrade melanoidins. However, there are stronger relationships between CD (Figure 6.4b) and DE (Figure 6.5b), both showing wider ranges in the amounts of melanoidin degraded with 65–69% and 63–69% degradation respectively.

190

Figure 6.3 Perturbation plot for (%) melanoidin degradation. Coded values are shown for each factor: melanoidin (A); total HCA (B); pH (C);

FeSO4·7H2O dose (D) and AlCl3·6H2O dose (E); and refer to actual values listed in Table 6.1.

Optimal degradation of melanoidin is achieved at increasing FeSO4·7H2O dosage at lower initial pH (Figure 6.4b). The optimal degradation performance can be maintained at 69% but gradually decreases with decreasing FeSO4·7H2O dosage and increasing pH. The oxidative performance is reduced at increasing pH and

FeSO4·7H2O dosage, as it is expected that the deactivation of Fe(II) would occur by the precipitation of Fe(III) oxyhydroxides in solution (Cortez et al., 2011).

191

(a)

(b)

Figure 6.4 Contour plots of melanoidin degradation (%) as a function of

(a) melanoidin and AlCl3·6H2O dosage; (b) pH and FeSO4·7H2O dosage. Variables: melanoidin (1,500 mg/L); total HCA

(150 mg/L); pH (5.25); FeSO4·7H2O dosage (389 mg/L) and

AlCl3·6H2O dosage (200 mg/L).

192

(a)

(b)

Figure 6.5 Contour plots of melanoidin degradation (%) as a function of

(a) pH and AlCl3·6H2O dosage; (b) FeSO4·7H2O dosage and

AlCl3·6H2O dosage. Variables: melanoidin (1,500 mg/L); total

HCA (150 mg/L); pH (5.25); FeSO4·7H2O dosage (389 mg/L) and

AlCl3·6H2O dosage (200 mg/L).

193

Interestingly, the modified Fenton oxidation performance depends on both the dosages of both FeSO4·7H2O and AlCl3·6H2O. As shown in Figure 6.5b, lower or higher amounts of both reagents used together would result in improved melanoidin degradation (68–69%). However, the degradation efficiency is reduced slightly to 66%, if Fe(II) and Al(III) were dosed at any other given concentration than the extremes (e.g., the median values of both dosages, 389 mg/L FeSO4·7H2O and

200 mg/L AlCl3·6H2O). High dosages of FeSO4·7H2O/AlCl3·6H2O and low dosages of the other vice versa would result in poor oxidation performance on melanoidin degradation.

6.3.4 Oxidation Performance of Hydroxycinnamic Acids

Similar to Equation 6.3, the model for the degradation of HCAs in a synthetic mixture containing sucrose and melanoidin via the modified Fenton process is also a negative exponent function of the independent variables (i.e., xn, where n = 1–5). On the basis of the coefficients from Equation 6.4, FeSO4·7H2O dosage (D) is the most influential parameter for the degradation of HCAs, where increasing the dosage of

FeSO4·7H2O for the Fenton oxidation process enhances the degradation of HCAs within the mixture. This is also noticeable in the perturbation plot shown in Figure

6.6. However, increasing AlCl3·6H2O (E) does not assist in the degradation of HCAs unlike for the melanoidin component within the same mixture. Hence, it can be concluded that the removal of HCAs (or other similar phenolic compounds) is primarily attributable to the Fenton process only.

There were no changes in total HCA degradation at any given HCA degradation (B) tested. This shows us that the under the various modified Fenton conditions, a consistent amount of HCA will be degraded (ca. 48%), hence, lower dosages of the Fe(II) and Al(III) can be reduced in order to minimise costs. A similar trend in the degradation behaviour of the HCAs to melanoidin degradation for initial solution pH (C) was also observed. Increasing pH would result in deactivation of the radicals and ions required to regenerate and maintain the oxidation process.

194

Figure 6.6 Perturbation plot for (%) total HCA degradation. Coded values are shown for each factor: melanoidin (A); total HCA (B); pH (C);

FeSO4·7H2O dose (D) and AlCl3·6H2O dose (E); and refer to actual values listed in Table 6.1.

Figure 6.7 and 6.8 show the contour plots for the statistically significant two- factor interactions of the developed model for total HCA degradation (Equation 6.4) via the modified Fenton process: melanoidin concentration and solution pH (AC); melanoidin concentration and FeSO4·7H2O dosage (AD); total HCA concentration and FeSO4·7H2O dosage (BD); and total HCA concentration and AlCl3·6H2O dosage (BE). Increasing the initial pH and melanoidin concentration resulted in lower HCA degradation (46%). However, the amount degraded increases up to 48% when both the melanoidin concentration and initial pH decrease (Figure 6.7a). Decreasing the melanoidin concentration and increasing FeSO4·7H2O dosage vice versa reduces the extent of the modified Fenton process on the degradation of HCAs in solution (Figure

6.7b). However, more HCA is degraded (50%) when lower FeSO4·7H2O dosages are used with lower concentrations of melanoidins.

195

(a)

(b)

Figure 6.7 Contour plots of total HCA degradation (%) as a function of

(a) melanoidin and pH; (b) melanoidin and FeSO4·7H2O dosage. Variables: melanoidin (1,500 mg/L); total HCA (150 mg/L); pH

(5.25); FeSO4·7H2O dosage (389 mg/L) and AlCl3·6H2O dosage (200 mg/L).

196

(a)

(b)

Figure 6.8 Contour plots of total HCA degradation (%) as a function of

(a) total HCA and FeSO4·7H2O dosage; (b) total HCA and

AlCl3·6H2O dosage. Variables: melanoidin (1,500 mg/L); total

HCA (150 mg/L); pH (5.25); FeSO4·7H2O dosage (389 mg/L) and

AlCl3·6H2O dosage (200 mg/L).

197

A strong relationship between the total HCA concentration and FeSO4·7H2O dosage is shown in Figure 6.8a. Increasing FeSO4·7H2O dosage with lower amounts of HCAs would result in more degradation because of less uptake of OH radical by the HCAs. However, unlike Fe(II), adding more Al(III) does not provide any benefit in the degradation of HCAs. At an initial HCA concentration of 50 mg/L, an

AlCl3·6H2O dosage of 100 mg/L is enough to degrade nearly half of the HCAs initially present in solution. The degradation extent of HCAs reduces with increasing initial HCA concentration at 50 mg/L AlCl3·6H2O. However, increasing AlCl3·6H2O dosage with increasing initial HCA concentration would maintain optimal degradation of HCA.

6.3.5 Response Surface Analyses for Decolourisation of Mixtures

Graphical representations of the regression model (Equation 6.5) in the form of 3D surface plots are shown in Figure 6.9. The interactions are significant as the curvature of the surfaces is obvious.

As shown in Figure 6.9a, the initial melanoidin and Al(III) dosage concentrations were varied, whilst the other variables, namely pH and temperature were kept constant. At 100 mg/L of AlCl3·6H2O, the decolourisation of the mixture reduces at initial melanoidin concentrations of ≥ 1000 mg/L. As shown in Figure

6.9a, additional colouring is obtained at higher dosages of AlCl3·6H2O with an initial melanoidin concentration of 500 mg/L. This may indicate that residual Al(III) may be forming complexes with other components in the system (Cornard et al., 2006;

Lapouge and Cornard, 2007). However, with a 300 mg/L dosage of AlCl3·6H2O, the decolourisation efficiency increases smoothly with an increase in melanoidin concentration, suggesting that Al(III) is being consumed and contributing to the removal of melanoidins in the mixture. However, excess dosages of AlCl3·6H2O can also give the reverse effect where colour is added into the system (e.g., 300 mg/L

AlCl3·6H2O and 500 mg/L melanoidin) (Figure 6.9a).

198 Design-Expert® Software Design-Expert® Software

% Decolourisation % Decolourisation 51.049 51.049

-120.544 -120.544 X1 = A: Mel (a) X1 = B: Total PA (b) X2 = E: AlCl3.6H2O X2 = D: FeSO4.7H2O Actual Factors Actual Factors 39 B: Total PA = 100.00 31 A: Glc/Gly Melanoidin = 1000 C: Initial pH = 5.25 C: Initial pH = 5.25 D: FeSO4.7H2O = 389.00 E: AlCl3.6H2O = 200 17 26

4 12

-10 -2

-24 -15

% Decolourisation % Decolourisation

300 1500 238 50 250 1250 313 75 200 1000 389 100 Design-Expert® Software Design-Expert® Software 465 125 E: AlClAlCl3.6H2O3·6H2O 150 750 A: Melanoidin A: Mel D: D: FeSO FeSO4.7H2O4·7H2O B: B: Total Total HCA PA % Decolourisation 100 500 % Decolourisation 541 150 51.049 51.049 -120.544 -120.544 X1 = B: Total PA (c) X1 = D: FeSO4.7H2O (d) X2 = E: AlCl3.6H2O X2 = E: AlCl3.6H2O

Actual Factors Actual Factors A: Glc/Gly Melanoidin = 1000 33 A: Glc/Gly Melanoidin = 1000 38 C: Initial pH = 5.25 B: Total PA = 100 D: FeSO4.7H2O = 389 C: Initial pH = 5.25 25 26

16 15

8 3

-1 -9

% Decolourisation % Decolourisation % Decolourisation

100 50 100 238 150 75 150 313 200 100 200 389 250 125 250 465 E:E: AlCl3.6H2OAlCl3·6H2O B: B: Total Total HCA PA E: AlCl3.6H2OAlCl3·6H2O D:D: FeSO4.7H2O FeSO4·7H2O 300 150 300 541

Figure 6.9 Three-dimensional surface plots of decolourisation (%) as a

function of (a) melanoidin and AlCl3·6H2O dosage; (b) total HCA

and FeSO4·7H2O; (c) total HCA and AlCl3·6H2O dosage; and (d)

FeSO4·7H2O and AlCl3·6H2O. Variables: melanoidin

(1,500 mg/L); total HCA (150 mg/L); pH (5.25); FeSO4·7H2O

dosage (389 mg/L) and AlCl3·6H2O dosage (200 mg/L).

Excess Fe(II)/Fe(III) also affects the decolourisation performance of the modified Fenton process despite degradation of the melanoidin and the HCAs. Increasing Fe(II) dosage may improve the oxidative degradation of the colour precursor and colourants in the system. However, excess Fe(II)/Fe(III) can react with non-reacted HCAs forming coloured complexes (Figure 6.9b). For optimal decolourisation performance, lower dosages of Fe(II) should be used even if the degradation efficiencies of the melanoidins and HCAs are reduced. However, as shown in Figure 6.9c, the effects of Al(III) on the modified Fenton process with initial

199

HCA concentration indicate that Al(III) decolourises the system. Increasing dosages of Al(III) significantly improves the decolourisation extent up to 45% at an initial concentration of 50 mg/L total HCA and up to ca. 15% at an initial HCA concentration of 150 mg/L. This suggests that there is a combined effect between

Fe(II)/H2O2 and Al(III). The Fenton process alone rapidly oxidises and degrades the components within the system while the Al(III) acts as an adsorbent/decolourising agent by removing the coloured products within the system (Dwyer et al., 2009).

To investigate the combined effects of Fe(II) and Al(III), the two key variables were compared against each other as depicted in Figure 6.9d. It is evident that for effective decolourisation of the melanoidin/HCA system, lower Fe(II) dosages and higher Al(III) should be used.

6.3.6 Process Optimisation and Validation

Numerical optimisation was performed on the basis of the desirability function to determine the optimum process parameters for the models developed for melanoidin degradation, HCA degradation and decolourisation. Multi-response optimisation was only used for the responses of melanoidin and HCA degradation because both models are of the same function, 2FI. Meanwhile, the quadratic model for decolourisation was optimised and validated separately.

In order to confirm the accuracy and robustness of the predicted models and assess its reliability to predict the degradation of the melanoidin and the HCAs as well as decolourisation, additional experiments were carried out under those conditions. For this study, the desirability functions for the two degradation models were combined into one value (Table 6.5). The experimental values of the additional experiments agree well with the predicted values (in parentheses) deduced from each of the four models. The low error in the experimental and predicted values indicates good agreement of the results.

200

Table 6.5 Optimised conditions under specified constraints for the degradation of melanoidin (2,000 mg/L) and total HCA (200 mg/L) in sucrose solution (15% (w/w)) at 35 °C; and model verification.*

Experiments Optimum Fenton Only Worst Case pH 5.1 5.1 6.0

FeSO4·7H2O (mg/L) 626 626 404

AlCl3·6H2O (mg/L) 265 0 151 Melanoidin degradation (%) 69 (71) 63 (65) 62 (66) Total HCA degradation (%) 53 (56) 47 (49) 40 (42) Desirability 0.890 0.425 0.765 *Values in parentheses indicate model predicted % degradation for each individual/total HCA model. Measurements were conducted in triplicate. RSD was < 5.0%.

It is worth mentioning that the dosages of the Fenton reagents and Al(III) are dependent on the initial pH (Table 6.5). Under the optimum conditions, higher dosages of FeSO4·7H2O and AlCl3·6H2O are required in order to degrade the melanoidin and HCAs. However, there were little differences between the optimum and worst case experiments in terms of melanoidin degradation with only an extra

7.0% degradation achieved when dosing an additional 222 mg/L FeSO4·7H2O and

114 mg/L AlCl3·6H2O to the system. Under the same operating conditions without

AlCl3·6H2O (i.e., Fenton process), the melanoidin and total HCA degradation efficiencies were slightly lower, at 63% and 47% respectively.

A predicted HCA degradation of 42% under the worst case experiment at a higher pH (i.e., pH 6.0) shows that the modified Fenton process is heavily dependent on pH. Interestingly, lower dosages of the reagents were only required under the worst case conditions due to deactivation of reagents at higher pH levels (pH ≥ 5.50).

201

Degradation of compounds does not necessarily imply that a mixture is decolourised. Parameter optimisation and model verification results for the decolourisation model on melanoidin/HCA mixtures in sugar solutions are shown in Table 6.6. The experimental values are in consistency with the predicted values based on the reduced quadratic model. The low error in the experimental and predicted values and reasonably high desirability values (≥ 0.825) indicate good agreement of the results.

Experiments Optimum Fenton Only Worst Case pH 5.3 5.3 4.5

FeSO4·7H2O (mg/L) 289 289 400

AlCl3·6H2O (mg/L) 322 0 350 Decolourisation (%) 43 (42) 24 (25) –109 (–113) Desirability 0.825 0.851 0.936 *Values in parentheses indicate model predicted % degradation for each individual/total HCA model. Measurements were conducted in triplicate. RSD was < 5.0%.

It is obvious that in the worst case experiment, higher dosages of FeSO4·7H2O and AlCl3·6H2O would result in colour formation. At the optimum working pH of

5.3, using 289 mg/L FeSO4·7H2O and 322 mg/L AlCl3·6H2O, 43% decolourisation was achieved. The predicted melanoidin and total HCA degradation under the optimum decolourisation experiments were 62% and 47%, respectively. A significant decrease in decolourisation performance was observed under the same conditions in the absence of AlCl3·6H2O with only 24% decolourisation achieved. Therefore, the modified Fenton process shows promise as the Fenton process is essential for the

202

breakdown of colour and colour precursor compounds, while the presence of Al(III) aids in colour removal.

6.4 Summary

The works presented in this chapter extends the investigations outlined in the previous chapters to evaluate the effect of the Fenton process on the degradation of melanoidins, including HCA degradation and colour removal, by modifying the Fenton process with the addition of Al(III). Changing the independent variables to be more closely aligned to sugar cane factory processing conditions reduced the complexity of the statistical analyses required.

A modified Fenton oxidation process where Al(III) is used to promote the oxidation process is effective in the degradation and decolourisation of synthetic sugar solutions containing a synthetic melanoidin and HCAs (viz., CaA, pCoA and FeA). Ferrous iron does not remove colour but it is essential for the breakdown of the melanoidin and HCAs. Also, Al(III) aids in the removal of the oxidation products and colour. Decolourisation is best achieved with an increased dosage of Al(III). Despite degradation of HCA with Fe(II), higher dosages would result in increased colour. Lower dosages of Fe(II) combined with higher dosages of Al(III) are suitable for the effective reduction of colour and the degradation of melanoidins and HCAs. Higher dosages of Fe(II) and Al(III) are much to be avoided as they actually increase the solution colour. Such addition to factory juices must be tightly controlled otherwise the process would be counterproductive.

203

References

Dwyer, J., Griffiths, P., & Lant, P. (2009). Simultaneous colour and DON removal from sewage treatment plant effluent: Alum coagulation of melanoidin. Water Research, 43, 553-561.

Exley, C. (2004). The pro-oxidant activity of aluminum. Free Radical Biology and Medicine, 36(3), 380-387.

Lapouge, C., & Cornard, J.-P. (2007). Reaction pathways involved in the mechanism of AlIII chelation with caffeic acid: Catechol and carboxylic functions competition. ChemPhysChem, 8(3), 473-479.

Ruipérez, F., Mujika, J. I., Ugalde, J. M., Exley, C., & Lopez, X. (2012). Pro-oxidant activity of aluminum: Promoting the Fenton reaction by reducing Fe(III) to Fe(II). Journal of Inorganic Biochemistry, 117(0), 118-123.

Shore, M., Broughton, N. W., Dutton, J. V., & Sissons, A. (1984). Factors affecting white sugar colour. Sugar Technology Reviews, 12, 1-99.

Westerhoff, P., Chen, W., & Esparza, M. (2001). Fluorescence analysis of a standard fulvic acid and tertiary treated wastewater. Journal of Environmental Quality, 30(6), 2037-2046.

204

CHAPTER 7

Evaluation of the Fenton and Fenton-like Processes for the Removal of Colour from Factory Sugar Cane Juice

7.1 Introduction...... 205 7.2 Materials and Methods...... 206 7.2.1 Reagents and Solvents...... 206 7.2.2 Specification of Samples...... 206 7.2.3 Decolourisation Procedure...... 207 7.2.4 Preparation of Flocculants...... 207 7.2.5 Preparation of Lime Saccharate...... 208 7.2.6 Clarification Procedure...... 208 7.2.7 Turbidity Measurements...... 209 7.2.8 Sucrose, Dry Substance and Purity Measurements...... 210 7.2.9 Reducing Sugars Composition Analyses...... 210 7.2.10 Colour, Refractive Index and Total Soluble Solids Measurements...... 210 7.2.11 Inorganic Ion Composition Analyses...... 211 7.3 Results and Discussion...... 211 7.3.1 First Decolourisation Trials...... 211 7.3.2 Second Decolourisation Trials...... 215 7.3.3 Economic Considerations...... 222 7.4 Summary...... 222

205

7.1 Introduction

The works presented in Chapters 4 and 6 showed that the Fenton and modified Fenton oxidation processes are capable of effectively decolourising and degrading colourant and colour precursor compounds in aqueous and dilute sucrose solutions (≤ 15% (w/w)). In this present study, the effects of the Fenton and modified Fenton processes on the decolourisation of factory sugar cane juice were investigated.

On-site clarification trials were undertaken to investigate the effectiveness of the Fenton oxidation process and variants of this process to remove colour from sugar cane juice. Sugar cane juice colour is usually measured at pH 7.0 but additional information about the nature of the colourants present may be obtained at pH 4.0 and 9.0. As previously stated in Section 2.4, colour measured at pH 4.0 suggests the presence of HMW colourants, while colour at pH 9.0 is essentially due to the presence of colour precursor, phenolic and flavonoid compounds. The results obtained from these tests are reported in this chapter.

7.2 Materials and Methods

7.2.1 Reagents and Solvents

All chemicals, solvents and reagents were obtained in their purest form from the suppliers as described in the previous chapters or as otherwise stated. Ferric chloride (anhydrous) was purchased from Sigma-Aldrich (St. Louis, MO, USA). Magnafloc LT27 flocculant (degree of hydrolysis (DH) of 27%; MW of 18 × 106 Da) was obtained from Chemiplas Australia (Robina, QLD, Australia). Magnafloc LT340 flocculant (DH of 40%; MW of 18 × 106 Da) was obtained from TD Chemicals (East Melbourne, VIC, Australia).

206

7.2.2 Specification of Samples

Sugar cane juice from the No. 2 mill was obtained from the processing lines at Tully Sugar Mill (Tully, QLD, Australia) and Isis Central Sugar Mill (Childers, QLD, Australia). Juices from the MJ and PJ process streams were also obtained from Isis Central Sugar Mill. All juices were obtained as composites during the crushing season in 2012. In total, four juices (2 × No. 2 mill juices, 1 × MJ and 1 × PJ) were treated. The following analyses of the four juice samples are unrelated and not comparable. However, the results obtained provide an insight on the levels of colour present in each juice type before and after treatment with the Fenton and modified Fenton processes.

7.2.3 Preparation of Flocculants

Solutions of flocculants (0.5% (w/v)) were best prepared by dispersing and dissolving the flocculant powder in Milli-Q water (adjusted to pH 8.0–8.5 using 0.1 M NaOH) under gentle stirring at a low shear rate (50 rpm) for 3 h. The powders were added at a rate which allowed good dispersion to ensure each flocculant particle is hydrated to prevent agglomeration. Flocculant solutions were stored at 4.0 °C. These solutions were diluted further to 0.1% (w/v) before being added to hot limed sugar cane juice.

7.2.4 Preparation of Lime Saccharate

Lime saccharate used for juice clarification was obtained directly from the factory of trial (viz., Tully Sugar Mill or Isis Central Sugar Mill). The mixture of lime saccharate typically consists of 100 g of 20% (w/w) calcium oxide solution and 100 g of 68 °Bx factory syrup.

207

7.2.5 Decolourisation Procedure

The procedure for the Fenton oxidation process with Al(III) on sugar cane juices is similar to that described in Section 6.2.3. In each run, a predetermined amount of FeSO4·7H2O or FeCl3, AlCl3·6H2O and H2O2 solutions were added to the reaction vessel containing factory juice to achieve a final volume of ca. 1.0 L, while maintaining the working Fenton molar ratio (Fe(II)/H2O2) at 1:15. The concentrations of the reagents were chosen based on the previous research reported in Chapter 6. After 2 min of oxidative treatment, the treated juice was immediately subject to clarification.

7.2.6 Clarification Procedure

Clarification experiments were conducted in a heated and illuminated clarification test kit designed by SRI (Brisbane, QLD, Australia) as shown in Figure 7.1. Each tube is of 1.0 L capacity with dimension of 460 × 55 mm i.d. The method of clarification was simple defecation and typically involved liming to pH 7.8 at 76 °C, followed by boiling and settling in the settling tubes. A flocculant dosage equivalent of 3 mg/kg of juice was applied prior to settling in the tubes. The flocculant used for the Tully Sugar Mill trials was Magnafloc LT27, which is the flocculant used at Tully Sugar Mill. Magnafloc LT340 was used for the trials at Isis Central Sugar Mill, because of the experiences at Tully Sugar Mill. The initial settling rate (cm/min) was obtained from the graphical analysis of the initial linear slope and can be calculated as:

æ 40 ö æ 40 ö (7.1) ç Initial Juice Level ´ ÷ - ç Mud Level at 0.5min ´ ÷ è 100 ø è 100 ø Settling Rate = 0.5min

As the floc aggregates were unstirred, the mud heights were not indicative of values obtained in commercial clarifier.

208

7.2.7 Turbidity Measurements

Absorbance measurements were conducted spectrophotometrically at 900 nm

(A900) on a GBC Scientific Cintra 40 double-beam UV/Vis spectrophotometer using cells of 1.0 cm path length. The resulting turbidity of each sample was calculated as:

(7.2) Turbidity (TU) = 100 ´ A900

Figure 7.1 Sugar Research Institute designed batch settling kit.

209

7.2.8 Sucrose, Dry Substance and Purity Measurements

The apparent sucrose content in juice (i.e., pol) was calculated by double polarisation measurements performed on a Schmidt Haensch Polartronic Universal digital polarimeter (Berlin, Germany) according to a standard procedure adapted from BSES (2001a). Juice samples were clarified with lead acetate, followed by pol measurements of the filtered clarified solutions (before and after sucrose inversion with HCl. The pol expressed as % (w/w) is calculated from the change in polarisation between the plain and inverted sugar solutions.

The dry substance procedure used to determine water and/or total solids in juice is also adapted from a procedure by the BSES (2001b). Coiled strips of Whatman No. 4 chromatography paper (600 × 50 mm) were saturated in juice and dried in vacuo at < 7 kPa for 12 h in an oven at 65 °C. The loss of sample after drying indicates the amount of water in the juice sample. The purity of juice samples is expressed as a percentage of pol on DS as shown in Equation 7.3.

pol (7.3) Purity (%) =  100 DS

7.2.9 Reducing Sugars Composition Analyses

Reducing sugar contents in the reaction mixtures were monitored by HPAEC- PAD. Sample preparation and the operating procedure for the chromatographic system are identical to that described in Section 4.2.5.

7.2.10 Colour, Refractive Index and Total Soluble Solids Measurements

Sample preparation and the operating procedures for the determination of colour, RI and TSS in juice samples are identical to those described in Section 3.2.6.

210

7.2.11 Inorganic Ion Composition Analyses

Inorganic ion composition analyses were performed on a Varian Vista-MPX simultaneous inductively coupled plasma-optical emission spectrometer (ICP-OES) with megapixel charge coupled device detection (Mulgrave, VIC, Australia). To reduce the interference of the organic sugar matrix, samples were diluted to a sucrose concentration of 2.0% (w/w). The measurements were conducted in duplicate. The operating parameters listed in Table 7.1 were applied for all ICP-OES measurements.

Table 7.1 Operating parameters for ICP-OES analyses.

RF generator 40 MHz Power 1.25 kW Plasma flow 13.5 L/min Auxiliary 0.75 L/min Nebuliser flow 0.75 L/min Viewing height 5 mm Emission lines (nm) Na 589.592 Mg 279.553 Al 396.152 Si 251.611 P 213.618 S 181.972 K 769.897 Ca 317.933 Fe 238.204

7.3 Results and Discussion

7.3.1 First Decolourisation Trials

Clarification Results

The clarification of juice was assessed based on three clarification performance parameters as shown in Table 7.2; turbidity, settling rate and mud height. Juice turbidity values for all tests were higher (9.7–18 TU) than the control (9.2 TU). The settling rate of the flocs for all tests was extremely slow. This may be due to the presence of high starch levels as no incubation was carried out for the naturally occurring α–amylase enzymes to break down the starch (Bruijn and Jennings, 1968). The settling rates of the flocs formed in the juices treated via the Fenton oxidation process were slightly lower (0.8–6.4 cm/min) than the value

211

obtained with the control (7.2 cm/min). Higher Fe(II) content reduced the settling rate of the floc particles. In addition, higher dosages of the Fenton’s reagent (Tests 2 and 4) increased mud height levels by up to 58%. This is possibly due to the lower density of mud particles.

Table 7.2 Clarification performance results on clarified No. 2 mill juices from the Tully Sugar Mill trials.*

Dosage (mM)

† Test Fe(II) H2O2 Al(III) Turbidity Settling rate MH (TU) (cm/min) (%) Untreated 0 0 0 – – – Control 0 0 0 9.2 7.2 12 Test 1 0.28 4.22 0 12 4.0 11 Test 2 0.50 7.50 0 18 2.4 14 Test 3 0.28 4.22 0.093 18 6.4 13 Test 4 0.50 7.50 0.093 9.7 0.8 19 *% RSD was < 5.0%. †MH (Mud Height)

Inorganic Ion Composition Results

Inorganic ion analysis was conducted on both untreated and treated juices (Table 7.3). Higher dosages of Fenton’s reagent carried out in Tests 2 and 4 increased residual Fe levels in clarified juice. However, the addition of Al(III) to the higher Fenton dosage (Test 4) reduced the level of Fe by 20%. Despite the same Al(III) dosage applied to both tests, the higher Fenton dosage (Test 4) showed a lower Al level by 50% than the lower Fenton dosage (Test 3).

The addition of Fe(II) as FeSO4·7H2O contributed to the increase in S levels in all tests conducted. Both FeSO4·7H2O and AlCl3·6H2O are acidic in nature and increased the amount of lime saccharate used to reach the pH set point for clarification. This resulted in increases with residual soluble Ca levels in the clarified juices. More soluble Ca implies higher fouling rates in the evaporator vessels, which

212

is highly undesirable. More interestingly though is the reduction in P levels in Test 4. This may be due to Fe(II)/Fe(III) and Al(III) ions reacting with free phosphate ions in juice to form a precipitate. This is unlikely and may simply be due to the formation of calcium phosphate precipitates as these precipitates have a lower solubility than Fe(II), Fe(III) and Al(III) phosphates.

Table 7.3 Inorganic ion composition results on clarified No. 2 mill juices from the Tully Sugar Mill trials.*

Concentration (mg/kg on TSS) Test Na Mg Al Si P S K Ca Fe Untreated 8 1,040 290 710 1,940 839 7,100 957 108 Control 29 792 19 333 458 875 7,190 1,980 19 Test 1 32 859 23 359 500 1,020 7,830 2,280 62 Test 2 29 888 22 326 494 1,690 7,190 3,030 258 Test 3 33 898 69 352 523 1,020 7,840 2,840 66 Test 4 38 1,040 35 359 346 1,920 8,460 3,970 205 *% RSD was < 5.0%

Colour Results

Table 7.4 shows the colour results obtained from the Tully Sugar Mill trials on No. 2 mill juice. The results show that in the normal clarification process (i.e., the control) the juice colour at pH 7.0 reduced by 24%, but little effects were observed at pH 4.0 (–7.8%) and pH 9.0 (5.3%). The use of the Fenton and modified Fenton processes reduced juice colour at pH 7.0 to a similar extent as the control. There was a significant drop in juice colour at pH 9.0 using both Fenton and modified Fenton processes with Al(III). The higher Fenton dosage (Test 2) achieved a reduction of 37% at pH 9.0 compared to the control. The addition of Al(III) (Test 4) also significantly decreased the colour content at pH 9.0 by up to 42% and also slightly decreased colour at pH 4 by ≤ 1.0%. The drop in colour at pH 9.0 is attributable to the Fenton process degrading phenolic and flavonoid compounds.

213

The IVs of both untreated and treated juices as shown in Table 7.4 were between 6.0 and 12 which are attributable to monomeric colourants such as natural cane pigments (viz., flavonoids) (cf. Table 2.1). During normal clarification, a small decrease in the IV of the control relative to the untreated No. 2 mill juice was observed. This shows that a small percentage of factory produced colourants such as polymers from HADPs and browning reactions involving phenolic compounds were formed, hence an increase in colour at pH 4.0 (Paton, 1992). However, clarification also removed some of the LMW colourants associated with pH 9.0. The IVs of the treated juices were lower than the control (IV 6.0–8.4) indicating a lower presence of LMW colourants. This confirms that the Fenton, Fenton-like and modified Fenton processes are removing colourants and related compounds associated with pH 9.0.

Table 7.4 Colour results on clarified No. 2 mill juices from the Tully Sugar Mill trials.*

Colour (IU) Test pH 4.0 pH 7.0 pH 9.0 IV Untreated 5,640 21,400 66,500 12 Control 6,080 16,400 63,000 10 Test 1 6,440 17,500 53,800 8.4 Test 2 6,950 17,400 41,700 6.0 Test 3 6,910 16,600 53,300 8.0 Test 4 5,560 16,600 38,500 7.0 *% RSD was < 5.0%

In summary, the Fenton process with Al(III) showed the best result with respect to colour reduction. However, the observations for the poor settling rate of the particles (apart from the likely presence of high starch levels) is probably due to the flocculant (i.e., Magnafloc LT27) being incapable of bridging the flocs closely together to increase the floc density and subsequently enhance sedimentation. The lack of light bonds between the floc particles is possibly due to the composition of the treated juices. The presence of residual Fe(II), Fe(III) and Al(III) cations in the

214

treated juice may have reduced the effectiveness of the flocculant to form tightly bound floc structures. It was hypothesised that an alternative flocculant with higher anionicity may improve the clarification performance of the treated juices.

7.3.2 Second Decolourisation Trials

Clarification Results

As previous results in Table 7.2 indicated that treating juices with iron and aluminium salts would impact on the settling rate of the flocs, hence a flocculant with high anionicity and high molecular weight was selected (Madsen and Day, 2010). The flocculant Magnafloc LT340 was chosen for these trials in place of Magnafloc LT27.

Table 7.5 shows the clarification performance results of No. 2 mill, MJ and PJ obtained from the trials conducted at Isis Central Sugar Mill. In addition to the modified Fenton process (Fe(II)/Al(III)/H2O2), Fe(III) was trialled in place of Fe(II) as it has been previously reported that Fe(III) can be readily used as a chelant and oxidant to create flocs for effective juice clarification (Madsen and Day, 2010).

Excellent clarified juice turbidity values (3–9 TU) were obtained with primary juice treated with both modified Fenton and Fenton-like processes, though the turbidity of the control was more pronouced. For this type of juice, the modified Fenton process with Fe(II) gave a lower turbidity (4.6 TU) than the modified Fenton- like process containing Fe(III) (9.4 TU). The reverse trend was observed for the turbidity values of the clarified juices obtained from No. 2 mill juice and MJ. The turbidity values obtained with Fe(II) (Tests 5 and 7) in place of Fe(III) were unacceptably high. The reasons for these observations are not known, although Fe(III) is known to be a more effective coagulant than Fe(II) (Rivas et al., 2002).

The sizes of the flocs formed by visual assessment with both modified Fenton and Fenton-like processes were smaller than those formed by the normal clarification process and the resulting settling rates were extremely slow (≤ 3.2 cm/min). Reasonable mud heights (except Test 9) were obtained for the different types of juices with the different Fenton treatments (Table 7.5).

215

Table 7.5 Clarification performance results on clarified factory juices from the Isis Central Sugar Mill trials.*

Dosage (mM)

† Test Fe(II) Fe(III) H2O2 Al(III) Turbidity Settling rate MH (TU) (cm/min) (%) No. 2 Mill juice Untreated 0 0 0 0 – – – Control 0 0 0 0 9.0 33 19 Test 5 0.50 0 7.5 0.093 40 0.8 9.0 Test 6 0 0.92 7.5 0.093 15 8.0 17 Mixed juice Untreated 0 0 0 0 – – – Control 0 0 0 0 3.8 18 29 Test 7 0.50 0 7.5 0.093 21 3.2 27 Test 8 0 0.92 7.5 0.093 3.4 20 28 Primary juice Untreated 0 0 0 0 – – – Control 0 0 0 0 3.1 42 19 Test 9 0.50 0 7.5 0.093 4.6 0.8 25 Test 10 0 0.92 7.5 0.093 9.4 3.2 21 *% RSD was < 5.0%. †MH (Mud Height)

Inorganic Ion Composition Results

Table 7.6 shows the inorganic ion composition of the clarified juices. The trends in the proportions of the inorganic ion concentrations were similar for both modified Fenton and Fenton-like processes on each type of factory juice tested. Interestingly, the modified Fenton-like process with Fe(III) (Tests 6 and 8) significantly produced less residual Al and Fe than the modified Fenton process with Fe(II) (Tests 5 and 7). However, the treatment of PJ with Fe(III) (Test 10) compared to Fe(II) (Test 9) resulted in improved Fe removal but increased Al content.

216

This may simply be due to the lower turbidity (except Test 10 for Al) obtained with Fe(III) treated juices, and as a consequence of the type and nature of the calcium phosphate flocs formed. This is reflected in the lower levels of P (except for Test 10) and Ca obtained for these treated juices. Therefore, it is presumed that the modified Fenton-like process with Fe(III) does not interfere in the precipitation of calcium phosphate to the same extent as the Fe(II) treated juices. Additions of Fe(III) (Tests 6, 8 and 10) appear not to boost the S content, are better with Ca content and have less residual Fe content than that of Fe(II) additions (Tests 5, 7 and 9).

Table 7.6 Inorganic ion composition results clarified factory juices from the Isis Central Sugar Mill trials.*

Concentration (mg/kg on TSS) Test Na Mg Al Si P S K Ca Fe No.2 Mill juice Untreated 44 694 38 188 706 1,050 8,000 341 45 Control 75 533 5 130 272 1,090 7,720 663 9 Test 5 92 640 49 128 302 1,740 8,840 1,510 267 Test 6 61 459 36 82 247 929 7,770 1,000 68 Mixed juice Untreated 18 719 11 79 676 1,370 5,970 568 9 Control 26 676 2 169 115 1,760 6,350 878 2 Test 7 25 659 11 94 152 2,030 6,300 1,230 67 Test 8 43 809 7 125 132 1,990 7,350 1,180 18 Primary juice Untreated 108 758 8 133 600 1,330 6,920 917 7 Control 133 741 3 170 133 1,560 7,110 1,260 4 Test 9 136 728 8 136 112 2,080 7,120 1,680 70 Test 10 129 694 19 145 194 1,450 7,180 1,530 33 *% RSD was < 5.0%

217

The addition of Fe(II) as FeSO4·7H2O contributed to the increase in S levels in juices treated with Fe(II) (Table 7.6). The salts FeSO4·7H2O, FeCl3 and AlCl3·6H2O are acidic in nature and their addition to juice reduced the pH. This resulted in an increase in the amount of lime saccharate added to reach the pH set point for clarification. The effect of this is an increase in the soluble Ca content (Table 7.6) of the juices treated with these reagents compared to the control experiments where these reagents were not used. The modified Fenton and Fenton-like processes show reductions in Si content in all juice types compared to the control (Table 7.6). This is due to the formation of insoluble aluminium-silicate compounds during clarification (Thai et al., 2012).

Purity and Reducing Sugars Results

During raw sugar manufacture, sucrose loss through inversion to glucose and fructose, and degradation to organic acids are minimised by working within the desired pH ranges. The values in Table 7.7 indicate that no significant changes to the purity levels occurred in the clarified juice due to the treatment using both the Fenton and modified Fenton processes when compared with the control. However, there were increases in the levels of glucose (≤ 14%) and fructose (≤ 9.0%) indicating some sucrose degradation.

218

Table 7.7 Purity and reducing sugar results on clarified factory juices from the Isis Central Sugar Mill trials.*

Reducing Sugar Content % (w/w) Test Glucose Fructose Purity (%) No. 2 Mill juice Untreated 0.09 0.07 83.8 Control 0.04 0.04 90.5 Test 5 0.27 0.25 89.5 Test 6 0.26 0.23 89.9 Mixed juice Untreated 0.21 0.35 82.3 Control 0.06 0.05 92.8 Test 7 0.26 0.23 90.9 Test 8 0.27 0.24 92.8 Primary juice Untreated 0.09 0.09 86.9 Control 0.07 0.09 92.3 Test 9 0.21 0.18 91.5 Test 10 0.18 0.16 91.9 *% RSD was < 5.0%

Colour Results

Table 7.8 shows, significant colour present in the clarified juice of No. 2 mill juice obtained via the normal clarification process compared to the juices clarified from MJ and PJ (Curtin and Paton, 1980). There are increases in the colour measured at pH 4.0 (37–45%) and pH 7.0 (11–21%) for No. 2 mill juice treated by the modified Fenton and Fenton-like processes relative to the control (Table 7.8). The colours in clarified MJs and PJs measured at pH 4.0 and pH 7.0 respectively follow similar trends as the clarified No. 2 mill juice. However, the colour levels measured at pH 9.0 for the clarified No. 2 mill juice decreased significantly for the modified Fenton process with Fe(II) (Test 5) and with Fe(III) (Test 6) by 42% and 38% respectively

219

relative to the control. There were also decreases for the colour measured at pH 9.0 for clarified PJs (≤ 26%) using the modified Fenton and Fenton-like processes, however the magnitude of decolourisation was reduced. There was only a marginal decrease in colour at pH 9.0 with the treated MJs (Tests 7 and 8). It is presumed that some non-sucrose impurities present in higher proportions in MJ may have interfered with the oxidation processes thereby preventing the degradation of colourants. The IVs of all the treated juices were lower (4.6–5.9) when compared with the respective control sample (6.5–12). This indicates and confirms the significant decreases in LMW colourants in the treated juices using the modified Fenton and Fenton-like processes.

Table 7.8 Colour results on clarified factory juices from the Isis Central Sugar Mill trials.*

Colour (IU) Test pH 4 pH 7 pH 9 IV No. 2 Mill Juice Untreated 9,120 18,400 72,400 8.0 Control 5,250 14,200 63,300 12 Test 5 7,200 15,800 36,600 5.1 Test 6 7,600 17,200 39,100 5.1 Mixed Juice Untreated 3,560 8,830 22,300 6.3 Control 3,990 9,020 25,900 6.5 Test 7 4,230 10,800 23,800 5.6 Test 8 5,310 11,600 24,200 4.6 Primary Juice Untreated 4,220 9,660 31,700 7.5 Control 4,000 8,620 33,700 8.4 Test 9 4,710 9,050 24,900 5.3 Test 10 4,490 10,300 26,400 5.9 *% RSD was < 5.0%

220

The modified Fenton-like process with Fe(III) (Tests 6, 8 and 10) show slightly higher colour than the corresponding process with Fe(II) (Table 7.8). This is attributable to the slower rate of H2O2 decomposition to the active •OH radical necessary for the degradation of colourants, when the more stable Fe(III) is used in place of Fe(II) (Sedlak and Andren, 1991; Pignatello, 1992; Arnold et al., 1995). Also, as the total amount of iron in juice is approximately 10–20 ppm on juice (van der Poel et al., 1998) and is present as Fe(III), the optimum working molar ratio of

Fe(III) and H2O2 was not used in these studies.

The modified Fenton and Fenton-like processes resulted in a decrease in colour at pH 9.0, but this decrease was offset by an increase in colour at pH 4.0 and pH 7.0. As stated previously, a decrease in colour at pH 9.0 indicates a reduction in flavonoids and phenolics, and these colourants have a major influence on raw sugar colour inclusion into the crystals of raw sugar (Smith and Paton, 1985; Clarke et al., 1986; Riffer, 1988; Davis, 2001). However, the major contributors to the impurities and colour in the raw sugar lie in the molasses layer of the surface of the crystals.

The crystallisation of raw sugar, if ideal, rejects all impurities from the crystal structure. In practice, impurities are trapped (layered in) within the crystalline structure, co-crystallised with sucrose into the crystal lattice. As well, some impurities are present in gross molasses inclusions within the crystal, and a good deal are left as the molasses or syrup film around the crystals. Little firm data is available on these aspects, particularly in relation to the relative magnitude of the layering effect and the impurity inclusion effect.

The inherent colour in the modified Fenton and Fenton-like processes on its own has minimal absorbance at 420 nm where colour is measured. However, in the presence of a colourant or colour precursor compound (e.g., CaA), the colours of the clarified juices are inflated (Riffer, 1988). It should therefore be noted that although reduction in colour at pH 7.0 (the usual measurement) of the clarified juice were not obtained with the modified Fenton and Fenton-like processes, significant colour reduction may have been realised if clarified juices were furthered processed to raw sugar, based on the aforementioned observations. Such an investigation should be undertaken in future research work.

221

7.3.3 Economic Considerations

Bulk quantities of FeSO4·7H2O, AlCl3·6H2O and H2O2 reagents required for the modified Fenton oxidation process are all available commercially. The pricing of these additives, exclusive of GST and delivered to the metropolitan area of Brisbane, Australia are listed in Table 7.9.

Table 7.9 Prices of additives in bulk quantities used for the modified Fenton process.

Chemical Company Origin Price (AUD $/t)

FeSO4·7H2O Swancorp Australia $350

AlCl3·6H2O Shanghai Smart Chemicals China $450

H2O2, 50% (w/v) Solvay Interox Australia $1,050

The approximate cost of the best treatment, conducted in this study

(i.e., Test 7), for one tonne of factory cane juice at FeSO4·7H2O (2.49 mM),

AlCl3·6H2O (0.83 mM) and H2O2 (7.5 mM) is $A0.24, $A0.06 and $A0.42, respectively (i.e., total of $A0.72/t of juice). The additional costs for the uses of increased lime saccharate and the flocculant (i.e., Magnafloc LT340) needed for all Fenton-mediated processes used as well as possible sucrose losses have not been taken into account. As approximately eight tonnes of Australian MJ is required to produce one tonne of raw sugar, it would cost ca. $A5.76/t of sugar. The costs of reagents can be further reduced if bulk quantities are sourced.

7.4 Summary

This study was aimed at the decolourisation of factory cane juice using Fenton, Fenton-like and modified Fenton oxidation processes. Results have shown that the modified Fenton oxidation process (i.e., Fe(II)/Al(III)/H2O2) significantly reduced colour measured at pH 9.0 (associated with LMW colourants and colour precursors) for clarified juices of No. 2 Mill and PJs. However, the modified Fenton process did not reduce colour levels measured at pH 4.0 (associated with HMW

222

colourants) and at pH 7.0. The results obtained from the second decolourisation trials conducted at Isis Central Sugar Mill also confirm the results obtained from the initial trials carried out at Tully Sugar Mill. Problems associated with small floc size and slow settling of flocs should be addressed in future studies. Furthermore, treated juices should be used to produce sugar, in order to establish whether the Fenton and modified Fenton processes can produce low colour sugar. If low colour sugar can be produced, it will be necessary in a future project to investigate ways to minimise sucrose degradation using Fenton oxidation technologies.

223

References

Arnold, S. M., Hickey, W. J., & Harris, R. F. (1995). Degradation of atrazine by Fenton's reagent: Condition optimization and product quantification. Environmental Science and Technology, 29(8), 2083-2089.

Bruijn, J., & Jennings, R. P. (1968). Enzymatic hydrolysis of starch in cane juice. Proceedings of the South African Sugar Technologists' Association, 45-52.

BSES (2001a). Method 18. Sucrose – Determination in Mill Products by Double Polarisation, Laboratory Manual for Australian Sugar Mills (Vol. 2, pp. 1-2). Indooroopilly, QLD, Australia: Bureau of Sugar Experiment Stations.

BSES (2001b). Method 19. Total Solids (Dry Substance) – Determination in Mill Products, Laboratory Manual for Australian Sugar Mills (Vol. 2, pp. 1-2). Indooroopilly, QLD, Australia: Bureau of Sugar Experiment Stations.

Clarke, M. A., Blanco, R. S., & Godshall, M. A. (1986). Colorant in raw sugars. Paper presented at the Proceedings of the International Society of Sugar Cane Technologists.

Davis, S. B. (2001). The chemistry of colour removal: a processing perspective. Proceedings of the South African Sugar Technologists' Association, 75, 328- 336.

Madsen, L. R., II, & Day, D. F. (2010). Iron mediated clarification and decolourisation of sugarcane juice. Proceedings of the International Society of Sugar Cane Technologists, 27, 1-13.

Paton, N. H. (1992). The origin of colour in raw sugar. Proceedings of the Australian Society of Sugar Cane Technologists, 14, 8-17.

Pignatello, J. J. (1992). Dark and photoassisted Fe3+-catalyzed degradation of chlorophenoxy herbicides by hydrogen peroxide. Environmental Science and Technology, 26(5), 944-951.

Riffer, R. (Ed.). (1988). The Nature of Colorants in Sugarcane and Cane Sugar Manufacture. Amsterdam: Elsevier.

Sedlak, D. L., & Andren, A. W. (1991). Oxidation of chlorobenzene with Fenton's reagent. Environmental Science and Technology, 25(4), 777-782.

224

Smith, P., & Paton, N. H. (1985). Sugarcane Flavonoids. In R. A. McGinnis & E. G. Muller (Eds.), Sugar Technology Reviews (Vol. 12, pp. 117-141). Amsterdam: Elsevier.

Thai, C. C. D., Bakir, H., & Doherty, W. O. S. (2012). Insights to the clarification of sugar cane juice expressed from sugar cane stalk and trash. Journal of Agricultural and Food Chemistry, 60, 2916-2923. van der Poel, P. W., Schiweek, H., & Schwartz, T. (1998). Sugar Technology: Beet and Cane Manufacture. Berlin: Verlag Dr. Albert Bartens KG.

225

226

CHAPTER 8

Conclusions and Future Aspects

8.1 Findings of the Thesis...... 228 8.2 Recommendations for Future Work...... 231

227

8.1 Findings of the Thesis

The sugar industry is constantly looking at ways to cost effectively remove impurity loadings in sugar process streams as these impurities impact on the colour formed in raw sugar. Besides, the industry is concerned with progressive colour formation of raw sugar during storage due to oxidation of phenolic compounds, present in these impurities. This thesis has presented a detailed study on the degradation of HCAs and also the decolourisation of sugar cane juice.

The phenolic acid and colour composition of factory cane juices processed by Australian sugar factories was investigated. Phenolic compounds are of interest, as they are known to be natural colour precursors. These compounds can react with other organic and inorganic components in juice through enzymatic and non- enzymatic reactions to produce highly coloured polymeric compounds that contribute considerably to raw sugar colour.

Amongst the juice extracts of FEJ and PJ process streams, fifteen phenolic compounds, HMF and furfural were quantified. Changes to the conventional procedure by dissolving the dried extracts in methanol in place of water, showed an overall improved response to phenolic acids and revealed the presence of flavonoids. The chromatographic results reveal that the phenolic acids; CaA, pCoA and FeA were of the highest concentrations, which are classed as HCAs, present in juice extracts from Australian factory FEJ and PJ. Moreover, the concentrations of phenolic acids in burnt cane were twice as much as those obtained in whole crop cane. This is probably due to the thermal decomposition of HMW phenolics (viz., lignin, polyphenols) during cane burning.

The colour analyses showed that juice expressed from whole crop harvested cane has significantly higher colour than juices (11,400–20,000 IU) expressed from burnt harvested cane (10,400–12700 IU) attributable to the higher amounts of impurities and natural colourants entering the manufacturing process.

A detailed investigation on the degradation of CaA was conducted. A quadratic polynomial model was obtained for CaA degradation through the use of CCD and RSM, and indicated that initial sucrose and CaA concentration significantly decreased the amount of CaA degraded. Numeric optimisation based on the

228

desirability function was used to determine the optimum process parameters. It showed that in water at 35 °C, 80% of CaA was degraded at pH 5.5 using 0.72 mM

Fe(II) and 0.44 mM H2O2. However, for a synthetic sugar solution (13% (w/w) sucrose), under processing conditions similar to that of MJ, only 61% of CaA was degraded.

The Fenton process was also used to degrade phenolic compounds in synthetic juice mixtures containing HCAs (viz., CaA, pCoA and FeA), sucrose and water. Numerous models were developed and validated to predict the degradation of HCAs through the use of RSM. The models were not only used to predict the optimum conditions for the degradation of the HCAs but to also understand and probe the effects of each significant parameter and their interaction with one another on the degradation of HCAs. Under the optimised conditions for a 200 mg/L initial HCA mixture concentration, the degradation efficiencies of the mixture in water and sugar solutions (i.e., 13% (w/w) sucrose) were 77% and 57% respectively.

Sucrose was the most influential parameter that significantly lowered the degradation efficiencies of the HCAs in the Fenton process. The behaviour of CaA degradation in the composite system is different from that of pCoA and FeA, possibly due to its ability to form complexes with Fe(III), as its aromatic ring is highly activated with the presence of two hydroxyl groups.

Attempts were made to identify and quantify some of the reaction products from the Fenton oxidation of HCAs by means of LC/MS, HPAEC-PAD, HPIEC and GC/MS. Mechanistic oxidation pathways were proposed with support from previous works in the literature. The presence of phenolic aldehydes and aliphatic carboxylic acids suggest that the Fenton process is oxidising and breaking down the HCAs. However, the formation of oligomeric products from the oxidative coupling of cinnamoyl radicals indicates that the Fenton process is also polymerising some of the oxidised products.

Modifications to the Fenton process were made by either adding AlCl3·6H2O to the mixture prior to oxidation and/or replacing Fe(II) with Fe(III) (i.e., Fenton-like process). The oxidation performance of these additives was evaluated on both complex synthetic juice systems (containing a synthetic melanoidin, HCAs, sucrose and water) and factory sugar cane juice.

229

In a synthetic juice solution consisting of sucrose (15% (w/w)), the HCAs (200 mg/L) and a synthetic glucose-glycine melanoidin (2,000 mg/L), the addition of

AlCl3·6H2O in the modified Fenton process degraded the melanoidin and the HCAs by approximately 69% and 53% respectively. However, AlCl3·6H2O did not play a significant role in degradation because the Fenton process on its own (i.e., without

AlCl3·6H2O), under the same conditions resulted in 63% and 47% degradation, respectively. On the other hand, the addition of AlCl3·6H2O played a significant role in the removal of colour with up to 43% decolourisation at pH 5.3 using 289 mg/L

FeSO4·7H2O, 107 mg/L H2O2 and 322 mg/L AlCl3·6H2O. The Fenton process on its own, under the same conditions only gave 24% decolourisation.

In factory sugar cane juice, the modified Fenton oxidation process (i.e.,

Fe(II)/Al(III)/H2O2) showed a decrease in colour at pH 9.0 (≤ 42%) for various factory juices (No. 2 mill, mixed and primary) with minimal sucrose loss. However, there were increases in colour at pH 4.0 (≤ 45%) and pH 7.0 (≤ 21%) under the same conditions. Moreover, it is noted that colour measured at pH 9.0 is readily transferred to the sugar crystal relative to the colour measured at pH 4.0 and pH 7.0, and so some colour reduction will be realised if these clarified juices were processed to raw sugar.

Overall, the studies conducted throughout this project have shown that the Fenton and modified Fenton processes are capable of degrading and decolourising sugar process streams with minimal losses of sucrose.

A preliminary minimum cost calculation indicated that the modified Fenton processes were found to be reasonably inexpensive for decolourisation of sugar process streams. Under the optimum working conditions of the modified Fenton process (i.e., 0.50 mM Fe(II), 0.093 mM Al(III) and 7.5 mM H2O2), effective decolourisation of factory cane juice at pH 9.0 can be achieved at a cost of $A0.72/t of juice.

230

8.2 Recommendations for Future Work

A number of suggestions are proposed for future work, based on the research findings of this thesis. The advantages of the use of the modified Fenton process in the sugar manufacturing process include its simplicity, its non-specific oxidation property and the use of inexpensive equipment. Also, the sludge that is produced has the potential to remove colourants and other impurities (including proteins and polysaccharides) improving the quality of sugar process streams.

Clarification of Treated Juice

Problems associated with small size and slow settling of flocs need to be addressed to achieve optimum decolourisation and prevent any carryover of colour in downstream processes attributable to the finer particles that are not separable during sedimentation. A suitable coagulating agent such as an anionic polyacrylamide or polydiallyldimethylammonium chloride for the binding and precipitating for these flocs needs to be looked into for the effective clarification of juice using the modified Fenton process.

The sludge produced during Fenton oxidation means that it must be used before clarification to remove that sludge. It cannot be used on evaporator syrup unless the treated syrup then undergoes a flotation-type process in order to completely remove the residual sludge. Otherwise high turbidity will be carried through into the product sugar, and this is not acceptable.

Raw Sugar Production

In order to determine the extent of Fenton and modified Fenton oxidation as viable decolourisation processes in sugar production, treated juices should be used to produce raw sugar. If low colour sugar can be produced, it will be necessary to investigate ways to minimise sucrose degradation using Fenton oxidation technologies.

231

Toxicity Measurements

Apart from the measurement of colour, other measurements such as chemical oxygen demand, total polyphenolic content, total aromaticity and toxicity. Evaluation of toxicity is important for assessing the impact of oxidised compounds produced from the Fenton process of food process streams.

Degradation Products

A thorough investigation in probing the oxidation of HCAs and other phenolic compounds in solution is still required. It is important to understand the degradation mechanism of these compounds via the Fenton process in order to propose detailed mechanistic pathways for the conversion of the starting organic materials to their mineralisation products (i.e., CO2 and H2O). One solution that could be used to better determine and quantify the reaction products is to initially isolate and purify them first via preparative HPLC. The combined use of various spectroscopic techniques including UV/Vis, NMR and FTIR will assist in the characterisation and structure elucidation of these compounds, especially oligomeric products which are not available commercially. Changes to the voltages applied during GC/MS and LC/MS analyses can also be envisioned, to improve fragmentation and assist in the determination of unknown reaction products of phenolic compounds via the Fenton process or any other catalytic oxidation process.

Other Oxidants and Catalysts

It is recommended to investigate the oxidative performance of the Fenton process by using other oxidants in place of H2O2 such as organic hydroperoxides and peroxy acids or by using other iron-based materials as catalysts. Recently, significant attention has been paid to the use of cheap heterogeneous catalysts in place of the typical homogeneous Fe(II) catalysts to overcome the high amounts of iron- containing sludge formed after oxidation. Bulk iron-containing materials (e.g., red mud from alumina processing) and natural iron-containing clay minerals (e.g., goethite, hematite or magnetite) should be used as Fenton catalysts as they require

232

minimal catalyst preparation and activation. Another approach is to incorporate aluminium and/or iron onto activated carbons, clays, polymers and zeolites. These heterogeneous catalysts would not only assist in the degradation of the target compounds but provide synergies in assisting in the clarification and removal of intermediate and by-products.

233

234

APPENDICES

235

Table A1.1 Experimental design and results for % CaA degradation (i.e., Design 1).

Test CaA Sucrose pH [Fe(II)] [H2O2] Temp. Time Degradation (mM) % (w/w) (mM) (mM) (°C) (s) (%) 1 0.28 34.00 6.5 0.72 2.21 95 120 48 2 1.11 17.00 5.0 0.45 6.62 65 65 52 3 1.11 34.00 3.5 0.72 11.03 95 120 21 4 0.28 34.00 3.5 0.72 2.21 95 10 33 5 0.695 17.00 5.0 0.45 6.62 65 65 40 6 0.695 17.00 5.0 0.45 6.62 65 65 40 7 1.11 34.00 6.5 0.72 11.03 35 10 42 8 0.28 0.00 6.5 0.18 11.03 95 120 67 9 1.11 0.00 6.5 0.18 11.03 95 10 23 10 0.28 0.00 3.5 0.18 2.21 35 10 33 11 1.11 34.00 3.5 0.18 2.21 95 120 16 12 0.695 0.00 5.0 0.45 6.62 65 65 77 13 0.28 0.00 3.5 0.18 11.03 35 120 76 14 0.28 34.00 3.5 0.72 2.21 35 10 33 15 1.11 0.00 3.5 0.72 2.21 95 120 32 16 1.11 34.00 6.5 0.18 2.21 95 120 18 17 0.28 0.00 6.5 0.72 11.03 35 120 80 18 0.695 17.00 5.0 0.45 6.62 35 65 45 19 1.11 0.00 6.5 0.18 11.03 35 120 38 20 0.28 34.00 6.5 0.18 11.03 35 10 46 21 1.11 34.00 3.5 0.72 11.03 35 120 45 22 1.11 0.00 6.5 0.72 2.21 35 120 64 23 0.695 17.00 5.0 0.45 6.62 65 65 40 24 0.695 17.00 5.0 0.45 6.62 65 10 40 25 0.28 0.00 6.5 0.72 11.03 95 120 87 26 0.28 34.00 6.5 0.72 11.03 35 120 34 27 1.11 0.00 6.5 0.72 11.03 35 120 90 28 1.11 0.00 3.5 0.18 2.21 35 120 24 29 0.28 34.00 6.5 0.72 11.03 35 10 28

236

30 0.28 0.00 3.5 0.18 2.21 35 120 65 31 0.28 34.00 3.5 0.72 11.03 35 10 26 32 0.28 0.00 3.5 0.18 11.03 95 10 75 33 1.11 0.00 3.5 0.18 2.21 35 10 1 34 1.11 34.00 6.5 0.72 2.21 35 120 26 35 0.28 0.00 6.5 0.18 2.21 35 120 81 36 0.28 34.00 3.5 0.18 2.21 95 10 32 37 0.28 0.00 6.5 0.72 2.21 35 120 81 38 0.28 34.00 3.5 0.72 11.03 95 10 7 39 1.11 0.00 6.5 0.72 2.21 35 10 35 40 0.695 17.00 5.0 0.45 11.03 65 65 27 41 0.28 34.00 6.5 0.72 2.21 95 10 43 42 1.11 34.00 3.5 0.72 2.21 35 10 18 43 0.695 17.00 5.0 0.45 6.62 65 65 40 44 0.695 17.00 5.0 0.45 6.62 65 65 40 45 1.11 0.00 6.5 0.18 11.03 35 10 8 46 1.11 0.00 3.5 0.18 11.03 95 120 77 47 0.695 17.00 3.5 0.45 6.62 65 65 19 48 0.28 34.00 6.5 0.72 11.03 95 120 43 49 0.28 34.00 3.5 0.18 2.21 35 10 27 50 1.11 0.00 3.5 0.72 11.03 35 120 67 51 1.11 34.00 6.5 0.18 2.21 35 10 0 52 1.11 0.00 6.5 0.18 2.21 35 120 23 53 0.28 34.00 6.5 0.18 11.03 95 120 19 54 1.11 34.00 6.5 0.18 2.21 35 120 16 55 1.11 0.00 3.5 0.72 11.03 35 10 50 56 1.11 34.00 6.5 0.18 11.03 95 10 14 57 1.11 34.00 6.5 0.72 2.21 95 120 31 58 0.28 0.00 6.5 0.72 11.03 95 10 83 59 1.11 34.00 6.5 0.72 11.03 95 10 48 60 0.695 17.00 6.5 0.45 6.62 65 65 38 61 1.11 34.00 6.5 0.72 11.03 95 120 46 62 1.11 34.00 3.5 0.72 2.21 95 10 33

237

63 1.11 0.00 3.5 0.18 11.03 35 10 11 64 0.28 0.00 3.5 0.72 2.21 95 120 80 65 0.28 34.00 6.5 0.18 2.21 35 120 56 66 1.11 34.00 6.5 0.18 2.21 95 10 3 67 0.28 0.00 6.5 0.18 11.03 35 120 81 68 0.695 17.00 5.0 0.45 6.62 65 65 40 69 0.28 34.00 6.5 0.18 2.21 35 10 30 70 1.11 0.00 6.5 0.18 2.21 95 10 9 71 0.28 34.00 3.5 0.18 11.03 35 120 35 72 0.28 34.00 6.5 0.18 2.21 95 120 27 73 0.28 34.00 6.5 0.18 2.21 95 10 25 74 0.28 0.00 3.5 0.18 2.21 95 10 72 75 1.11 0.00 3.5 0.18 11.03 35 120 60 76 1.11 0.00 3.5 0.72 2.21 95 10 32 77 0.695 34.00 5.0 0.45 6.62 65 65 35 78 1.11 0.00 6.5 0.18 2.21 35 10 3 79 0.28 34.00 6.5 0.72 11.03 95 10 34 80 0.28 34.00 3.5 0.72 2.21 95 120 33 81 0.28 17.00 5.0 0.45 6.62 65 65 39 82 0.695 17.00 5.0 0.45 6.62 65 120 40 83 0.28 0.00 6.5 0.18 2.21 95 10 72 84 1.11 0.00 3.5 0.72 11.03 95 120 66 85 0.28 0.00 6.5 0.72 11.03 35 10 72 86 0.28 0.00 6.5 0.72 2.21 95 10 81 87 1.11 34.00 3.5 0.18 11.03 35 10 15 88 1.11 34.00 3.5 0.72 2.21 95 120 36 89 0.695 17.00 5.0 0.45 6.62 65 65 40 90 1.11 0.00 3.5 0.72 2.21 35 10 21 91 1.11 0.00 6.5 0.18 2.21 95 120 28 92 0.28 34.00 3.5 0.18 11.03 95 120 29 93 0.28 0.00 6.5 0.18 2.21 95 120 82 94 1.11 0.00 6.5 0.72 11.03 35 10 67 95 1.11 0.00 6.5 0.72 11.03 95 10 77

238

96 1.11 34.00 3.5 0.72 2.21 35 120 23 97 1.11 34.00 3.5 0.72 11.03 35 10 38 98 0.28 34.00 3.5 0.18 2.21 35 120 33 99 0.28 34.00 3.5 0.18 11.03 95 10 28 100 0.695 17.00 5.0 0.45 6.62 65 65 40 101 0.695 17.00 5.0 0.45 6.62 65 65 40 102 0.28 0.00 6.5 0.72 2.21 35 10 67 103 0.28 0.00 3.5 0.72 2.21 95 10 79 104 1.11 0.00 6.5 0.72 2.21 95 120 57 105 0.695 17.00 5.0 0.45 6.62 95 65 50 106 0.695 17.00 5.0 0.45 6.62 65 65 40 107 0.28 0.00 3.5 0.72 2.21 35 120 67 108 1.11 34.00 3.5 0.18 11.03 95 120 41 109 0.28 0.00 3.5 0.72 11.03 35 120 75 110 0.695 17.00 5.0 0.72 6.62 65 65 24 111 1.11 0.00 6.5 0.72 2.21 95 10 57 112 0.28 34.00 6.5 0.72 2.21 35 10 43 113 0.28 0.00 6.5 0.72 2.21 95 120 85 114 1.11 0.00 6.5 0.18 11.03 95 120 53 115 1.11 34.00 3.5 0.18 11.03 95 10 35 116 0.28 0.00 3.5 0.18 2.21 95 120 77 117 0.28 34.00 3.5 0.72 2.21 35 120 35 118 1.11 34.00 6.5 0.18 11.03 95 120 39 119 1.11 34.00 6.5 0.18 11.03 35 120 32 120 0.28 34.00 3.5 0.72 11.03 95 120 28 121 1.11 34.00 3.5 0.18 2.21 35 120 12 122 0.695 17.00 5.0 0.45 2.21 65 65 30 123 0.28 34.00 6.5 0.72 2.21 35 120 45 124 1.11 34.00 3.5 0.72 11.03 95 10 17 125 1.11 34.00 3.5 0.18 2.21 95 10 15 126 0.28 0.00 3.5 0.18 11.03 35 10 52 127 1.11 0.00 3.5 0.18 11.03 95 10 70 128 0.28 0.00 3.5 0.72 11.03 35 10 59

239

129 0.28 34.00 3.5 0.18 2.21 95 120 32 130 0.28 0.00 6.5 0.18 2.21 35 10 35 131 0.28 0.00 3.5 0.72 11.03 95 10 95 132 0.28 34.00 3.5 0.18 11.03 35 10 31 133 0.28 34.00 3.5 0.72 11.03 35 120 32 134 0.28 0.00 3.5 0.18 11.03 95 120 86 135 1.11 34.00 6.5 0.72 2.21 35 10 19 136 1.11 34.00 6.5 0.72 2.21 95 10 30 137 1.11 0.00 3.5 0.18 2.21 95 120 43 138 1.11 0.00 6.5 0.72 11.03 95 120 81 139 0.695 17.00 5.0 0.18 6.62 65 65 44 140 0.28 0.00 3.5 0.72 2.21 35 10 49 141 0.28 0.00 6.5 0.18 11.03 35 10 72 142 0.28 0.00 6.5 0.18 11.03 95 10 48 143 1.11 34.00 6.5 0.72 11.03 35 120 63 144 1.11 34.00 3.5 0.18 2.21 35 10 5 145 1.11 0.00 3.5 0.72 2.21 35 120 30 146 1.11 34.00 3.5 0.18 11.03 35 120 43 147 1.11 0.00 3.5 0.18 2.21 95 10 42 148 0.28 0.00 3.5 0.72 11.03 95 120 97 149 1.11 0.00 3.5 0.72 11.03 95 10 66 150 1.11 34.00 6.5 0.18 11.03 35 10 16 151 0.28 34.00 6.5 0.18 11.03 95 10 21 152 0.28 34.00 6.5 0.18 11.03 35 120 51

240

Table A1.2 Sucrose and reducing sugar results on selected tests at t = 2 min (i.e., Design 1).*

Sugar Content % (w/w) Sugar Content % (w/w) Test Glucose Fructose Sucrose Test Glucose Fructose Sucrose 1 0.01 0.02 33.98 1B 0.00 0.00 33.98 3 0.53 0.25 34.01 3B 0.00 0.00 34.02 11 0.05 0.03 33.99 11B 0.00 0.00 34.00 16 0.00 0.00 34.05 16B 0.00 0.00 34.05 21 0.24 0.21 33.98 21B 0.00 0.00 33.98 26 0.02 0.03 34.00 26B 0.00 0.00 34.00 34 0.01 0.01 34.05 34B 0.00 0.00 34.05 48 0.28 0.24 34.12 48B 0.00 0.00 34.11 53 0.02 0.02 34.07 53B 0.00 0.00 34.08 54 0.00 0.00 34.05 54B 0.00 0.00 34.05 57 0.02 0.03 34.02 57B 0.00 0.00 34.02 61 0.05 0.07 33.96 61B 0.00 0.00 33.96 65 0.02 0.01 33.98 65B 0.00 0.00 33.98 71 0.00 0.01 33.97 71B 0.00 0.00 33.97 72 0.03 0.02 34.02 72B 0.00 0.00 34.02 80 0.02 0.04 34.00 80B 0.00 0.00 34.00 82 0.14 0.12 16.98 82B 0.00 0.00 16.98 88 0.01 0.03 34.02 88B 0.00 0.00 34.02 92 0.10 0.11 34.05 92B 0.00 0.00 34.05 96 0.01 0.00 34.00 96B 0.00 0.00 34.01 98 0.00 0.00 34.02 98B 0.00 0.00 34.02 108 0.03 0.02 34.10 108B 0.00 0.00 34.10 117 0.00 0.00 34.01 117B 0.00 0.00 34.01 118 0.17 0.20 34.02 118B 0.00 0.00 34.02 119 0.00 0.00 34.01 119B 0.00 0.00 34.01 120 0.10 0.11 33.98 120B 0.00 0.00 33.98 121 0.00 0.00 34.00 121B 0.00 0.00 34.00 123 0.05 0.04 34.01 123B 0.00 0.00 34.01 129 0.01 0.02 33.97 129B 0.00 0.00 33.97

241

133 0.01 0.00 33.98 133B 0.00 0.00 33.99 143 0.01 0.01 33.95 143B 0.00 0.00 33.95 146 0.02 0.02 34.01 146B 0.00 0.00 34.01 152 0.00 0.01 34.03 152B 0.00 0.00 34.04 *Tests denoted with B indicate blank tests (i.e., t = 0 min)

242

Table A1.3 Experimental design and results for % CaA, % pCoA, % FeA and % total HCA degradation (i.e., Design 2).

Degradation (%) Test Total HCA Sucrose pH Temp. CaA pCoA FeA Total (mg/L) % (w/w) (°C) HCA 1 200 7.50 5.0 38 85 37 43 55 2 155 11.25 4.8 31 95 42 44 60 3 110 7.50 5.0 38 90 52 53 65 4 65 3.75 5.3 31 71 65 70 69 5 110 15.00 5.0 38 94 46 43 61 6 110 7.50 5.0 25 96 53 55 68 7 155 3.75 5.3 44 74 60 62 65 8 65 11.25 4.8 31 96 45 45 62 9 110 7.50 5.0 38 90 48 50 63 10 155 11.25 4.8 31 95 46 48 63 11 155 11.25 4.8 44 95 40 44 59 12 110 7.50 5.5 38 39 57 56 51 13 65 11.25 4.8 31 96 47 50 64 14 110 0.00 5.0 38 61 77 79 73 15 20 7.50 5.0 38 94 62 57 71 16 110 0.00 5.0 38 60 76 78 72 17 65 3.75 4.8 31 95 66 69 77 18 110 7.50 5.0 38 91 52 50 64 19 155 3.75 4.8 31 93 58 59 70 20 110 15.00 5.0 38 93 – – 52 21 155 11.25 5.3 31 79 48 51 59 22 65 11.25 4.8 44 97 33 37 56 23 155 3.75 4.8 44 94 52 56 67 24 155 11.25 4.8 44 95 41 43 60 25 200 7.50 5.0 38 85 43 47 58 26 65 3.75 5.3 44 70 70 74 71 27 155 3.75 4.8 44 93 56 57 69 28 155 3.75 5.3 31 82 67 66 –

243

29 110 7.50 4.5 38 97 39 43 59 30 65 3.75 4.8 44 96 57 65 72 31 65 3.75 4.8 31 95 58 63 72 32 65 3.75 4.8 44 95 55 64 72 33 155 3.75 4.8 31 93 55 54 67 34 65 11.25 5.3 31 80 49 – 66 35 110 7.50 5.0 25 95 48 53 65 36 65 11.25 4.8 44 96 35 40 57 37 65 11.25 5.3 44 91 46 45 60 38 65 11.25 5.3 44 88 52 49 63 39 155 11.25 5.3 44 87 49 49 62 40 110 7.50 5.0 50 90 51 57 66 41 65 3.75 5.3 31 67 63 68 66 42 110 7.50 5.5 38 45 49 54 50 43 155 3.75 5.3 31 80 60 60 67 44 20 7.50 5.0 38 97 47 53 65 45 65 11.25 5.3 31 81 48 48 59 46 110 7.50 4.5 38 97 34 39 57 47 155 11.25 5.3 31 82 53 57 64 48 110 7.50 5.0 38 90 52 53 65 49 110 7.50 5.0 50 90 50 52 64 50 110 7.50 5.0 38 90 50 50 63 51 155 3.75 5.3 44 72 60 63 65 52 65 3.75 5.3 44 73 72 75 73 53 155 11.25 5.3 44 85 53 52 63 54 110 7.50 5.0 38 90 48 50 63

244

Table A1.4 Sucrose and reducing sugar results at t = 2 min (i.e., Design 2).*

Sugar Content % (w/w) Sugar Content % (w/w) Test Glucose Fructose Sucrose Test Glucose Fructose Sucrose 1 0.00 0.00 7.47 1B 0.00 0.00 7.47 2 0.00 0.00 11.20 2B 0.00 0.00 11.20 3 0.00 0.00 7.53 3B 0.00 0.00 7.53 4 0.00 0.01 3.81 4B 0.00 0.00 3.82 5 0.00 0.00 15.01 5BB 0.00 0.00 15.01 6 0.00 0.00 7.55 6B 0.00 0.00 7.55 7 0.00 0.01 3.75 7B 0.00 0.00 3.76 8 0.00 0.00 11.30 8B 0.00 0.00 11.30 9 0.00 0.01 7.84 9B 0.00 0.00 7.85 10 0.00 0.00 11.27 10B 0.00 0.00 11.27 11 0.00 0.00 11.22 11B 0.00 0.00 11.22 12 0.01 0.00 7.90 12B 0.00 0.00 7.90 13 0.00 0.00 11.28 13B 0.00 0.00 11.28 14 0.00 0.00 0.00 14B 0.00 0.00 0.00 15 0.01 0.00 7.55 15B 0.00 0.00 7.55 16 0.00 0.00 0.00 16B 0.00 0.00 0.00 17 0.01 0.01 3.75 17B 0.00 0.00 3.76 18 0.00 0.00 7.49 18B 0.00 0.00 7.49 19 0.00 0.01 3.80 19B 0.00 0.00 3.81 20 0.00 0.00 14.99 20B 0.00 0.00 14.99 21 0.00 0.00 11.26 21B 0.00 0.00 11.26 22 0.00 0.00 11.20 22B 0.00 0.00 11.20 23 0.00 0.01 3.82 23B 0.00 0.00 3.82 24 0.00 0.00 11.31 24B 0.00 0.00 11.31 25 0.01 0.00 7.60 25B 0.00 0.00 7.60 26 0.01 0.01 3.78 26B 0.00 0.00 3.79 27 0.01 0.00 3.82 27B 0.00 0.00 3.82 28 0.00 0.01 3.72 28B 0.00 0.00 3.72 29 0.00 0.00 7.52 29B 0.00 0.00 7.52 30 0.01 0.00 3.72 30B 0.00 0.00 3.73

245

31 0.00 0.01 3.73 31B 0.00 0.00 3.74 32 0.02 0.01 3.76 32B 0.00 0.00 3.77 33 0.00 0.00 3.79 33B 0.00 0.00 3.79 34 0.00 0.00 11.20 34B 0.00 0.00 11.20 35 0.00 0.00 7.42 35B 0.00 0.00 7.42 36 0.00 0.00 11.24 36B 0.00 0.00 11.24 37 0.01 0.01 11.20 37B 0.00 0.00 11.20 38 0.00 0.00 11.22 38B 0.00 0.00 11.22 39 0.00 0.00 11.27 39B 0.00 0.00 11.27 40 0.00 0.00 7.52 40B 0.00 0.00 7.52 41 0.00 0.00 3.76 41B 0.00 0.00 3.76 42 0.00 0.00 7.50 42B 0.00 0.00 7.51 43 0.01 0.00 3.74 43B 0.00 0.00 3.76 44 0.00 0.00 7.51 44B 0.00 0.00 7.51 45 0.01 0.00 11.24 45B 0.00 0.00 11.24 46 0.00 0.00 7.47 46B 0.00 0.00 7.47 47 0.00 0.00 11.22 47B 0.00 0.00 11.22 48 0.00 0.00 7.41 48B 0.00 0.00 7.41 49 0.00 0.01 7.50 49B 0.00 0.00 7.50 50 0.00 0.00 7.52 50B 0.00 0.00 7.52 51 0.00 0.01 3.73 51B 0.00 0.00 3.74 52 0.00 0.01 3.74 52B 0.00 0.00 3.74 53 0.00 0.00 11.27 53B 0.00 0.00 11.27 54 0.01 0.00 7.60 54B 0.00 0.00 7.61 *Tests denoted with B indicate blank tests (i.e., t = 0 min)

246

Table A2.1 Geometry optimisation, charges and bond order computational calculations of CaA.

SPARTAN '10 MECHANICS PROGRAM: PC/x86 1.1.0 Frequency Calculation Adjusted 1 (out of 63) low frequency modes Reason for exit: Successful completion Mechanics CPU Time : .14 Mechanics Wall Time: .29 SPARTAN '10 Quantum Mechanics Program: (PC/x86) Release 1.1.0v4 WARNING: Parallel not implemented with this method

Job type: Geometry optimization. Method: RB3LYP Basis set: 6-31G(D) Number of shells: 68 Number of basis functions: 211 Multiplicity: 1 SCF model: A restricted hybrid HF-DFT SCF calculation will be performed using Pulay DIIS + Geometric Direct Minimization Solvation: water [SM8] Optimization: Step Energy Max Grad. Max Dist. 1 -648.682524 0.011054 0.031973 2 -648.683422 0.002900 0.011599 3 -648.683483 0.001034 0.002655 4 -648.683490 0.000395 0.001440 5 -648.683490 0.000140 0.000430

247

Reason for exit: Successful completion Quantum Calculation CPU Time : 13:44.31

Atomic Charges: Electrostatic Mulliken Natural 1 H1 : +0.212 +0.146 +0.248 2 C1 : -0.374 -0.258 -0.248 3 C4 : -0.243 -0.170 -0.302 4 C2 : +0.208 +0.164 -0.112 5 C6 : +0.308 +0.261 +0.277 6 C5 : +0.200 +0.392 +0.291 7 C3 : -0.244 -0.229 -0.203 8 H3 : +0.167 +0.165 +0.250 9 H4 : +0.176 +0.195 +0.258 10 C7 : -0.143 -0.163 -0.120 11 H8 : +0.150 +0.163 +0.248 12 C8 : -0.310 -0.188 -0.371 13 H9 : +0.174 +0.175 +0.247 14 C9 : +0.663 +0.461 +0.785 15 O1 : -0.575 -0.497 -0.668 16 O2 : -0.588 -0.607 -0.718 17 H10 : +0.439 +0.445 +0.517 18 O3 : -0.567 -0.680 -0.710 19 H2 : +0.453 +0.453 +0.515 20 O4 : -0.582 -0.699 -0.712 21 H6 : +0.476 +0.470 +0.530

248

Bond Orders Mulliken 1 C1 H1 : 0.907 2 C1 C4 : 0.073 3 C1 C2 : 1.338 4 C1 C6 : 1.472 5 C1 C8 : 0.031 6 C1 O3 : 0.045 7 C4 C5 : 1.394 8 C4 C3 : 1.421 9 C4 H4 : 0.901 10 C4 O4 : 0.037 11 C2 C5 : 0.070 12 C2 C3 : 1.374 13 C2 C7 : 1.104 14 C6 C5 : 1.281 15 C6 C3 : 0.066 16 C6 O3 : 0.987 17 C6 O4 : 0.046 18 C5 C8 : 0.028 19 C5 O4 : 0.889 20 C3 H3 : 0.915 21 C3 C7 : 0.034 22 C3 C8 : 0.050 23 C7 H8 : 0.910 24 C7 C8 : 1.704 25 C7 C9 : 0.045 26 C7 O1 : 0.057 27 C8 H9 : 0.906 28 C8 C9 : 1.043

249

29 C9 O1 : 1.945 30 C9 O2 : 1.015 31 O1 O2 : 0.055 32 O1 H10 : 0.026 33 O2 H10 : 0.737 34 O3 H2 : 0.721 35 O4 H2 : 0.041 36 O4 H6 : 0.727

Reason for exit: Successful completion Properties CPU Time : 1.14 Properties Wall Time: 1.35

250

Table A2.2 Geometry optimisation, charges and bond order computational calculations of pCoA.

SPARTAN '10 MECHANICS PROGRAM: PC/x86 1.1.0 Frequency Calculation Adjusted 1 (out of 60) low frequency modes Reason for exit: Successful completion Mechanics CPU Time : .13 Mechanics Wall Time: .06 SPARTAN '10 Quantum Mechanics Program: (PC/x86) Release 1.1.0v4 WARNING: Parallel not implemented with this method

Job type: Geometry optimization. Method: RB3LYP Basis set: 6-31G(D) Number of shells: 64 Number of basis functions: 196 Multiplicity: 1 SCF model: A restricted hybrid HF-DFT SCF calculation will be performed using Pulay DIIS + Geometric Direct Minimization Solvation: water [SM8] Optimization: Step Energy Max Grad. Max Dist. 1 -573.459367 0.014428 0.082246 2 -573.462031 0.004680 0.029876 3 -573.462253 0.001063 0.003240 4 -573.462261 0.000340 0.001653 5 -573.462262 0.000109 0.000325

251

Reason for exit: Successful completion Quantum Calculation CPU Time : 11:46.03 Quantum Calculation Wall Time: 11:59.19

Atomic Charges: Electrostatic Mulliken Natural 1 H1 : +0.161 +0.168 +0.247 2 C1 : -0.212 -0.178 -0.179 3 C4 : -0.335 -0.171 -0.314 4 C2 : +0.211 +0.148 -0.130 5 C6 : -0.234 -0.190 -0.288 6 C5 : +0.410 +0.352 +0.347 7 C3 : -0.161 -0.219 -0.178 8 H3 : +0.152 +0.170 +0.250 9 H4 : +0.180 +0.189 +0.255 10 C7 : -0.191 -0.159 -0.119 11 H8 : +0.158 +0.159 +0.246 12 C8 : -0.302 -0.186 -0.374 13 H9 : +0.182 +0.178 +0.248 14 C9 : +0.670 +0.453 +0.785 15 O1 : -0.576 -0.496 -0.669 16 O2 : -0.598 -0.607 -0.719 17 H10 : +0.441 +0.445 +0.517 18 O4 : -0.569 -0.670 -0.694 19 H6 : +0.440 +0.458 +0.515 20 H11 : +0.173 +0.157 +0.255

252

Bond Orders Mulliken 1 C1 H1 : 0.907 2 C1 C4 : 0.078 3 C1 C2 : 1.361 4 C1 C6 : 1.489 5 C1 C8 : 0.034 6 C4 C5 : 1.369 7 C4 C3 : 1.448 8 C4 H4 : 0.903 9 C4 O4 : 0.044 10 C2 C5 : 0.071 11 C2 C3 : 1.364 12 C2 C7 : 1.108 13 C6 C5 : 1.367 14 C6 C3 : 0.075 15 C6 O4 : 0.032 16 C6 H11 : 0.917 17 C5 C3 : 0.029 18 C5 C8 : 0.026 19 C5 O4 : 0.965 20 C3 H3 : 0.913 21 C3 C7 : 0.031 22 C3 C8 : 0.046 23 C7 H8 : 0.911 24 C7 C8 : 1.701 25 C7 C9 : 0.047 26 C7 O1 : 0.057 27 C8 H9 : 0.905 28 C8 C9 : 1.044

253

29 C9 O1 : 1.950 30 C9 O2 : 1.013 31 O1 O2 : 0.055 32 O1 H10 : 0.026 33 O2 H10 : 0.738 34 O4 H6 : 0.736

Reason for exit: Successful completion Properties CPU Time : 1.14 Properties Wall Time: 1.19

254

Table A2.3 Geometry optimisation, charges and bond order computational calculations of pCoA.

SPARTAN '10 MECHANICS PROGRAM: PC/x86 1.1.0 Frequency Calculation Adjusted 4 (out of 72) low frequency modes Reason for exit: Successful completion Mechanics CPU Time : .17 Mechanics Wall Time: .07 SPARTAN '10 Quantum Mechanics Program: (PC/x86) Release 1.1.0v4 WARNING: Parallel not implemented with this method

Job type: Geometry optimization. Method: RB3LYP Basis set: 6-31G(D) Number of shells: 76 Number of basis functions: 230 Multiplicity: 1 SCF model: A restricted hybrid HF-DFT SCF calculation will be performed using Pulay DIIS + Geometric Direct Minimization Solvation: water [SM8] Optimization: Step Energy Max Grad. Max Dist. 1 -687.885006 0.782278 0.172563 2 -687.946054 0.309605 0.069009 3 -687.962560 0.152032 0.158684 4 -687.967209 0.046289 0.101182 5 -687.974773 0.023712 0.037747 6 -687.975517 0.011028 0.013128

255

7 -687.975705 0.003954 0.003702 8 -687.975724 0.001689 0.002490 9 -687.975729 0.000477 0.001537 10 -687.975730 0.000136 0.000451

Reason for exit: Successful completion Quantum Calculation CPU Time : 35:13.30 Quantum Calculation Wall Time: 36:02.80

Atomic Charges: Electrostatic Mulliken Natural 1 H1 : +0.217 +0.149 +0.248 2 C1 : -0.560 -0.264 -0.230 3 C4 : -0.299 -0.182 -0.300 4 C2 : +0.384 +0.169 -0.116 5 C6 : +0.405 +0.285 +0.268 6 C5 : +0.241 +0.375 +0.286 7 C3 : -0.304 -0.224 -0.204 8 H3 : +0.184 +0.167 +0.250 9 H4 : +0.194 +0.193 +0.255 10 C7 : -0.251 -0.162 -0.120 11 H8 : +0.165 +0.163 +0.247 12 C8 : -0.219 -0.186 -0.370 13 H9 : +0.170 +0.177 +0.248 14 C9 : +0.601 +0.461 +0.785 15 O1 : -0.555 -0.496 -0.668 16 O2 : -0.588 -0.607 -0.718 17 H10 : +0.447 +0.445 +0.517 18 O3 : -0.283 -0.559 -0.549

256

19 O4 : -0.561 -0.678 -0.700 20 H6 : +0.452 +0.463 +0.520 21 C10 : -0.373 -0.216 -0.313 22 H2 : +0.177 +0.174 +0.217 23 H5 : +0.177 +0.174 +0.217 24 H7 : +0.179 +0.178 +0.227

Bond Orders Mulliken 1 C1 H1 : 0.904 2 C1 C4 : 0.074 3 C1 C2 : 1.358 4 C1 C6 : 1.443 5 C1 C8 : 0.033 6 C1 O3 : 0.037 7 C4 C5 : 1.371 8 C4 C3 : 1.447 9 C4 H4 : 0.900 10 C4 O4 : 0.044 11 C2 C5 : 0.068 12 C2 C3 : 1.359 13 C2 C7 : 1.102 14 C6 C5 : 1.287 15 C6 C3 : 0.067 16 C6 O3 : 0.925 17 C6 O4 : 0.032 18 C5 C8 : 0.027 19 C5 O3 : 0.025 20 C5 O4 : 0.917 21 C3 H3 : 0.913

257

22 C3 C7 : 0.031 23 C3 C8 : 0.046 24 C7 H8 : 0.910 25 C7 C8 : 1.705 26 C7 C9 : 0.046 27 C7 O1 : 0.058 28 C8 H9 : 0.905 29 C8 C9 : 1.042 30 C9 O1 : 1.946 31 C9 O2 : 1.014 32 O1 O2 : 0.055 33 O1 H10 : 0.026 34 O2 H10 : 0.738 35 O3 C10 : 0.880 36 O4 H6 : 0.731 37 C10 H2 : 0.930 38 C10 H5 : 0.930 39 C10 H7 : 0.941

Reason for exit: Successful completion Properties CPU Time : 1.33 Properties Wall Time: 1.46

258

Table A2.4 Sucrose and reducing sugar results of Fenton-mediated reactions of sucrose at t = 2 min.*

Sugar Content % (w/w) Sugar Content % (w/w) Test Glucose Fructose Sucrose Test Glucose Fructose Sucrose 1 0.02 0.01 3.75 1B 0.00 0.00 3.76 2 0.01 0.01 7.50 2B 0.00 0.00 7.51 3 0.00 0.00 11.25 3B 0.00 0.00 11.24 4 0.00 0.00 15.01 4B 0.00 0.00 15.01 *Tests denoted with B indicate blank tests (i.e., t = 0 min)

150

110

Absorbance(mAU) 30

-10 0 5 10 15 20 25

Retention Time (min)

Figure A2.1 High-performance LC-DAD chromatograms (UV/Vis detection at 280 nm) of the HCA mixture subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C).

259

)

6 10 × 15

10 Intensity ( Intensity 5

0 0 5 10 15 20 25 Retention Time (min)

Figure A2.2 Total ion chromatogram (negative ion mode ESI-MS) of the HCA mixture subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C).

2.00

1.60

)

6 10 × 1.20

0.80 Intensity ( Intensity

0.40

0.00 15 20 25 30 35 Retention Time (min)

Figure A2.3 Gas chromatogram of a SPE extract of the HCA mixture subjected to Fenton oxidation at 2 min (pH 4.7, 25 °C).

260

6.0

357.1 )

6 4.0 10 × 471.1 269.1 2.0 Intensity ( Intensity 393.1 715.2 179.1

0.0 100 300 500 700 900 Mass-to-Charge (m/z) Ratio

Figure A2.4 Negative ion mode ESI-MS full-scan spectrum relevant to the dimer product arising from the Fenton oxidation of FeA, [M]– = 385.1 Da.

4.0 385.1

3.0

)

6 10 × 297.1

2.0 Intensity ( Intensity 1.0 341.1 155.0 189.1 249.0 0.0 100 200 300 400 Mass-to-Charge (m/z) Ratio

Figure A2.5 Negative ion mode ESI-MS full-scan spectrum relevant to the tetramer product arising from the Fenton oxidation of CaA, [M]– = 715.2 Da.

261

Table A3.1 Experimental design for % total HCA, % melanoidin degradation and decolourisation.

Test Melanoidin Total HCA pH FeSO4·7H2O AlCl3·6H2O (mg/L) (mg/L) (mM) (mM) 1 1500 150 5.63 0.85 0.41 2 1000 100 5.25 1.40 0.83 3 1000 100 6.00 1.40 0.83 4 500 50 4.88 1.94 0.41 5 1000 0 5.25 1.40 0.83 6 500 150 5.63 1.94 0.41 7 1500 150 4.88 1.94 0.41 8 1500 50 5.63 1.94 0.41 9 500 150 4.88 0.85 0.41 10 0 100 5.25 1.40 0.83 11 1500 50 4.88 0.85 0.41 12 1000 200 5.25 1.40 0.83 13 1000 100 5.25 1.40 0 14 1000 100 4.50 1.40 0.83 15 500 150 4.88 1.94 1.24 16 500 150 5.63 0.85 1.24 17 1500 50 4.88 1.94 1.24 18 1000 100 5.25 1.40 0.83 19 500 50 4.88 0.85 1.24 20 500 50 5.63 0.85 0.41 21 2000 100 5.25 1.40 0.83 22 1000 100 5.25 1.40 1.66 23 1000 100 5.25 1.40 0.83 24 1000 100 5.25 1.40 0.83 25 1500 50 5.63 0.85 1.24 26 1500 150 4.88 0.85 1.24 27 1000 100 5.25 0.31 0.83 28 1000 100 5.25 1.40 0.83 29 1000 100 5.25 1.40 0.83

262

30 500 50 5.63 1.94 1.24 31 1000 100 5.25 2.49 0.83 32 1500 150 5.63 1.94 1.24

263

Table A3.2 Results for % total HCA, % melanoidin degradation and decolourisation.

Degradation (%) Test Melanoidin Total HCA Decolourisation (%) 1 70 43 51 2 65 47 12 3 64 45 23 4 64 51 33 5 67 – 8 6 – 46 47 7 69 50 -42 8 64 51 14 9 63 – 18 10 – 50 – 11 71 – 6 12 66 48 28 13 66 47 34 14 – 49 – 15 74 46 -121 16 62 52 22 17 76 52 45 18 65 51 20 19 65 47 23 20 71 51 – 21 70 48 -3 22 64 46 27 23 66 48 – 24 71 47 8 25 64 40 34 26 63 48 25 27 65 – 42 28 65 49 -3 29 66 – 10

264

30 69 48 – 31 – 49 10 32 64 52 12

265

(a)

(b)

Figure A3.1 Normal probability plots of residuals for fitted model using (a) melanoidin and (b) total HCA degradation data after power transformation.

266

(a)

(b)

Figure A3.2 Box-Cox plots of (a) melanoidin and (b) total HCA degradation data for the determination of the optimised power transformed response surface models.

267

(a)

(b)

Figure A3.3 Plots of predicted response and experimental (actual) values for the degradation (%) of (a) melanoidin and (b) total HCA.

268

Figure A3.4 Plot of predicted response and experimental (actual) values for the decolourisation (%).

269

270