Assessment of End-of-Life

Opportunities for Reverse

Osmosis Membranes

Will Lawler

A thesis in fulfilment of the requirements for the degree of

Doctor of Philosophy

School of Chemical Engineering Faculty of Engineering

March 2015 i. Abstract

Reverse osmosis (RO) membranes are now core to modern desalination processes and are widely used around the world. Based on the increasing number of desalination plants, and the finite lifespan of the membranes, the resulting number of used RO modules to be discarded is becoming a critical challenge. The overall aim of this study is to identify, develop and assess alternative end-of-life options for used RO elements and investigate the associated technical readiness, environmental impact, financial considerations and legislative challenges. The assessed end-of-life alternatives include, direct reuse of the old membranes within lower throughput systems; chemical conversion into porous, ultrafiltration (UF) like filters; direct reuse or recycling of the various module components; various energy recovery techniques, and landfill disposal. The results show that direct reuse is a promising application that can be utilised with minimal additional treatment or infrastructure; however, module storage techniques are a critical consideration, particularly as membrane drying has a significant and irreversible impact on membrane performance due to pore collapse in the polysulfone support layer. The method for chemical conversion with controlled exposure to NaOCl has been optimised, resulting in promising organic and virus removal properties, comparable to commercially available 10 – 30 kDa molecular weight cut off UF membranes; however, there was significant variation in hydraulic performance, ranging from 8 – 400 L.m-2.h-1.bar-1. A detailed life cycle assessment was completed and demonstrated that module fabrication contributed less than 1% of the CO2-e emissions for the production of potable water from seawater desalination, and that direct reuse over one year is more environmentally favourable than landfill disposal, regardless of the transportation distance required.

However, in terms of direct reduction of waste to landfill, incineration provided the greatest benefit, at the expense of increased greenhouse gas emissions. Applying the knowledge generated within this study, an interactive online educational tool has been developed using a dynamic multi criteria decision analysis system, providing information

1 on end-of-life options to membrane users. Overall, this study provides detailed quantitative information for membrane users and manufacturers to enhance their decision making process when it comes to end-of-life membrane options.

2 ii. List of Publications

Journal Papers

Lawler, W., Bradford-Hartke, Z., Cran, M.J., Duke, M., Leslie, G., Ladewig, B.P., Le-

Clech, P., 2012. Towards New Opportunities for Reuse, Recycling and Disposal

of Used Reverse Osmosis Membranes. Desalination 299, 103–112.

Lawler, W., Antony, A., Cran, M., Duke, M., Leslie, G., Le-Clech, P., 2013. Production

and Characterisation of UF Membranes by Chemical Conversion of Used RO

Membranes. Journal of Membrane Science 447, 203–211.

Lawler, W., Alvarez-Gaitan, J., Leslie, G., Le-Clech, P., 2014. Comparative Life Cycle

Assessment of End-of-life Options for Reverse Osmosis Membranes.

Desalination 357, 45–54.

Lawler, W., Leslie, G., Le-Clech, P. Assessment of Membrane Drying and Subsequent

Rewetting Techniques. To be submitted to the Journal of Membrane Science.

Peer Reviewed Conference Papers

Lawler, W., Wijaya, T., Antony, A., Leslie, G., Le-Clech, P., 2011. Reuse of Reverse

Osmosis Desalination Membranes. Paper and oral presentation. IDA World

Congress. Perth.

Lawler, W., Leslie, G., Le-Clech, P., 2015. Decision Making Tool for End-of-Life Reverse

Osmosis Membrane Users. IDA World Congress. San Diego. Under Review.

Conference Proceedings

Lawler, W., 2011. Reuse, Recycling and Disposal of Used Reverse Osmosis Membranes

and it’s application to steelmaking. Oral presentation at Sustainability Symposium

- Future Pathways for Reducing Greenhouse Gas Impacts of Materials. Sydney,

Australia.

3 Lawler, W., Reusing Old Reverse Osmosis Membranes in Humanitarian Projects.

Oral presentation at Early Career Symposium Adelaide, November 2011.

Lawler, W., Antony, A., Leslie, G., Duke, M., Le-Clech, P. Fate of Aged Reverse Osmosis

Membranes. Poster presentation at IWA Leading Edge Technologies

conference. Amsterdam 2011.

Lawler, W., Rodricks, J., Le-Clech, P. Converted RO membranes for decentralised

gravity driven water treatment. Poster presentation at IWA Toronto 2013.

Lawler, W., Le-Clech, P. The end of it: Alternative options for reuse, recycling and

disposal of old RO membranes. Oral conference presentation at IMSTEC 2013.

Lawler, P. Le-Clech, P. Developing new applications for old reverse osmosis

membranes. Oral conference presentation at IMSTEC 2013.

4 iii. Acknowledgements

First and foremost I would like to thank my supervisor, Dr. Pierre Le-Clech, without whom this thesis would not have been possible. Thank you for taking me on board and teaching me about the world of membranes, research and academic life. I thoroughly enjoyed the journey, and I hope that you did too.

Secondly, I would like to thank my partner, Dr. Gemma Reynolds, for her infinite support and patience. The journey through post graduate life was significantly more bearable with you to share the experience with.

I would also like to thank everyone at UNESCO Centre for Membrane Science and

Technology at UNSW. This includes my co-supervisor Dr. Greg Leslie, whose expertise in the membrane industry was invaluable to my work. Also a big thanks to Dr. Alice

Anthony, for all of your input into my work and your help. It was a pleasure to get to know you, your husband Subbu, and your wonderful kids over the last few years. Thank you to Dr. Yun Ye for all the support in the lab over the years, and also Dr. Deyan Guang in the early days of my work. Thank you to my fellow students and friends in the membrane centre, it was quite a ride and I wish you all the best in your futures. Also thanks to my honours students, Joel, Philip and Sarah, who taught me a lot about what it is to be a supervisor and teacher. Thank you to Juan and Zenah for your help with my life cycle assessment work. Special thank you to Dr. Shane Cox, my honours thesis supervisor, friend, advisor, mentor, and employer. You started me off in the world of water treatment, and I continue to learn from you every day.

Thank you to my closest friends Joe, David and Kosta, with whom I shared the PhD adventure. Also big thanks to Alex, for providing the soundtrack for my research and all the good times.

5 I would also like to give a big thank you to my parents and siblings. You may not have understood what I was doing for the last few years, but you put up with me and supported me anyway.

I would also like to acknowledge the financial support of the National Centre of

Excellence in Desalination Australia, which is funded by the Australian Government through the Water for the Future initiative. Collaborative partners from Monash

University, Sydney water, Water Corporation, Dow and the SkyJuice Foundation are also gratefully acknowledged. Special thank you to Victoria University and particularly Dr.

Marlene Cran for your close participation on this project. Finally, thank you to Dr. Ludovic

Dumee, who brought me into the world of SAXS and helped me greatly with the analysis and modelling. I learnt a lot from you and our journey to the synchrotron.

6 iv. Table of Contents i. Abstract ...... 1 ii. List of Publications ...... 3 iii. Acknowledgements ...... 5 iv. Table of Contents ...... 7 v. List of Figures ...... 11 vi. List of Tables ...... 15 vii. Nomenclature ...... 17 Introduction ...... 19 Justification ...... 19 Aims and objectives ...... 21 Chapter Descriptions ...... 23 Literature Review ...... 25 Introduction ...... 26 Reverse Osmosis Membranes: Process and Materials ...... 27 Membrane Desalination Processes ...... 27 Membrane Structure ...... 29 Membrane Operating Conditions and Lifespan ...... 31 Waste Management ...... 33 Product Stewardship ...... 34 Early Policy Attempts for E-waste ...... 35 National Television and Computer Product Stewardship Scheme ...... 35 The Australian Packaging Covenant ...... 37 Towards Implementation of Product Stewardship for RO Membranes ... 38 End-of-life Options ...... 39 Membrane Reuse ...... 39 Multi-membrane Vessel Design ...... 42 Material Reuse ...... 43 Material Recycling ...... 45 Energy Recovery ...... 46 Energy Recovery in Electric Arc Furnace ...... 50 Chemical Conversion of RO Membranes ...... 51 Exposure of RO membrane to Oxidative Chemicals...... 52 Previous Attempts at Membrane Conversion ...... 54 Effect of Chlorine Exposure on Polysulfone Membranes ...... 55 7 Decentralised Water Treatment Technologies ...... 57 Decentralised Treatment Options ...... 57 Membrane Treatment Options ...... 58 Criteria for Membrane Use in Developing Rural Areas ...... 63 Membrane Storage and Drying ...... 64 Membrane Pore Collapse ...... 65 Previous Work Investigating Membrane Drying ...... 66 Manufacturer Recommended Rewetting Strategies ...... 67 Life Cycle Assessment ...... 68 Waste Management LCA ...... 68 LCA Studies on Water Treatment and Membrane Technology ...... 69 Conclusions ...... 70 Assessment of Membrane Drying and Subsequent Rewetting Techniques ...... 71 Introduction ...... 72 Methodology ...... 73 Membranes ...... 73 Membrane Performance Characterisation ...... 74 Membrane Drying ...... 74 Membrane Rewetting ...... 75 Scanning Electron Microscopy ...... 76 Thermo Gravimetric Analysis ...... 76 Atomic Force Microscopy...... 77 Rejection Characterisation ...... 77 Small Angle X-ray Scattering ...... 78 Results and Discussion ...... 81 Membrane Drying Behaviour ...... 81 Effect of Drying on Membrane Layers ...... 89 Assessment of Pore Structure Change Using SAXS Analysis ...... 92 Effect of Drying on Membrane Rejection ...... 99 Assessment and Application of Rewetting Strategies ...... 100 Change in Membrane Layer Resistance ...... 108 Conclusion ...... 111 Production and Characterisation of UF Membranes by Chemical Conversion of Used RO Membranes ...... 112 Introduction ...... 113 Materials and Methods ...... 113 8 Membranes ...... 113 Membrane Performance Characterisation ...... 114 Degrading Agents ...... 115 Membrane Characterisation by FTIR ...... 115 Rejection Characterisation ...... 115 Silver Nanoparticle Challenge Testing ...... 116 Atomic Force Microscopy...... 117 Fouling Characterisation ...... 117 Module Conversion and Gravity Filtration ...... 117 Results and Discussion ...... 120 Efficiency of Degrading Agents ...... 120 Active Layer Removal ...... 124 Impact of Membrane Type and Condition on Converted Performance . 127 Rejection Properties ...... 130 Fouling Propensity ...... 137 Gravity Fed Water Treatment using Converted RO Membranes ...... 144 Conclusions ...... 149 Comparative Life Cycle Assessment of End-of-life Options for RO Membranes 150 Introduction ...... 151 Methodology ...... 152 Goal and Scope Definition ...... 152 Functional Unit ...... 155 Life Cycle Inventory and Impact Assessment...... 155 Uncertainty Analysis ...... 157 Model Description ...... 159 Membrane Manufacturing ...... 159 Disposal Options ...... 162 Transportation ...... 166 Results and Discussion ...... 168 Membrane Manufacturing ...... 168 Impact of Membrane Module Size on Production Emissions ...... 170 Contribution of RO Manufacturing to Desalination Operation Emissions 171 Comparison of End-of-life Scenarios ...... 173 Effect of Reuse Membrane Lifetime on Reuse Viability ...... 177

9 Effect of Transportation Distance and Reuse Membrane Lifespan on End- of-life Scenario Viability ...... 179 Effect of End-of-life Scenarios on Landfill Loading ...... 183 Conclusions ...... 185 Decision Making Tool for End-of-life Membrane Users ...... 187 Introduction ...... 187 MemEoL Tool Operation ...... 188 Multi-Criteria Decision Analysis ...... 190 Assessment Criteria ...... 193 Environmental ...... 194 Economic ...... 195 Social ...... 197 Alternative Decision Matrix ...... 198 Assessment of Membrane Reuse Quality ...... 200 Conclusion ...... 205 Conclusion and Future Work...... 206 Conclusion ...... 206 Mechanisms for Membrane Performance Decline upon Drying ...... 206 Feasibility of RO Reuse Following Chemical Conversion ...... 209 Assessment and Comparison of End-of-life Options ...... 211 Future Work ...... 213 References ...... 216 Appendix A...... 234 A.1. Membrane Delamination ...... 234 A.2. Additional SEM images ...... 237 A.3. Supplementary Information for SAXS analysis ...... 239 A.4. Supplementary Information for Chapter 4 ...... 242 Appendix B. The Use of RO Membrane Components as a Coke Substitute in EAF Steelmaking...... 243 B.1. Introduction ...... 243 B.2. Material and Methods ...... 243 B.3. Results and Discussion ...... 245 B.3.1. Behaviour of Pure Metallurgic Coke ...... 246 B.3.2. Behaviour of Membrane Components ...... 246 B.3.3. Implications for the EAF process ...... 247 B.4. Conclusion ...... 248

10 Appendix C. Supporting information for Chapter 6 – Comparative Life Cycle Assessment of End-of-life Options for RO Membranes ...... 249 C.1. Scenario Process Flow Diagrams ...... 249 C.2. Input output tables ...... 253 C.3. Additional Results ...... 264 C.4. Uncertainty Calculations ...... 267 C.5. Pair Wise Monte Carlo Comparisons ...... 269 C.6. Status of end-of-life options in Australia ...... 271 v. List of Figures

Figure 1-1: Estimation of the annual mass of RO modules to be discarded worldwide...... 20 Figure 2-1: Structure of RO membrane module (US EPA, 2005)...... 28 Figure 2-2: Cross section structure of a TFC RO membrane...... 29 Figure 2-3: Cross linked fully aromatic PA active layer structure (Lee et al., 2011). .... 30 Figure 2-4. Waste management hierarchy from most to least preferred options...... 33 Figure 2-5: Sources of WaterSurplus reusable membranes...... 42 Figure 2-6: Old membrane movement along the pressure vessel in hybrid system. .... 43 Figure 2-7: TGA of membrane components heated under nitrogen atmosphere at 20°C.min-1 (Prince et al., 2011)...... 49 Figure 2-8: Nitrogen and ring chlorination of PA (Kang et al., 2007)...... 53 Figure 3-1: Flow diagram of membrane testing sequence...... 74 Figure 3-2: Example chromatograph of Humic and BSA feed solution...... 78 Figure 3-3: Membrane drying behaviour in TGA at 25°C in nitrogen atmosphere...... 81 Figure 3-4: XLE membrane drying behaviour in TGA at various temperatures...... 82 Figure 3-5: XLE and 10 kDa membrane mass loss from drying in balance-desiccator apparatus at room temperature...... 83 Figure 3-6: Performance of XLE membrane after varying levels of drying. Tested at 15 bar in high pressure apparatus...... 85 Figure 3-7: Performance of 10 kDa UF membrane after varying levels of drying. Tested at 1.5 bar in low pressure apparatus...... 85 Figure 3-8: SEM images of the backside of the PSf layer of delaminated XLE RO membranes. Left: Virgin, Right: Dried...... 90 Figure 3-9: AFM height map surface scans (active skin layer side) of 10 kDa UF membrane. Left: Virgin, Right: 100% Dry...... 91

11 Figure 3-10: Change root mean square (Rq) roughness for 10 and 30 kDa Virgin and 100% Dry UF membranes...... 91 Figure 3-11: Reduced data Guinier plots for the series of evaporation tests of 10 kDa UF membranes at 25, 40 and 60°C over time, using a 0.6 m camera length...... 94

Figure 3-12: Change in RG of 10 kDa UF membranes with evaporation time and temperature, using 0.6 m camera length...... 95 Figure 3-13: Mean pore size of 10 kDa UF membranes as a function of evaporation time and temperature, using 0.6 m camera length...... 97 Figure 3-14: Organic carbon rejection of virgin and 100% dry 100 kDa UF membranes, assessed with LC-OCD...... 99 Figure 3-15: Effect of ethanol solution concentration (1, 10, 25 and 50% w/w) and exposure time on 100% dry XLE membranes...... 101 Figure 3-16: Effect of alcohol solution (50% w/w for 15 min) rewetting 95 and 100% dry XLE membranes...... 102 Figure 3-17: Effect of prolonged exposure rewetting methods on 100% dry XLE membranes...... 103 Figure 3-18: Swelling volume and permeability of various membranes at various stages; Virgin, and after rewetting (following 100% drying in desiccator) with water (1 h soak), ethanol (50% w/w for 15 min), and with SLS (50 h)...... 105 Figure 3-19: Performances recovery of dried membranes after soaking in 50% ethanol for 15 min and SLS for 50 h...... 106 Figure 3-20: Comparison of (V) measured virgin, (M) measured post rewetting, and (T) calculated theoretical values for RO membrane resistance...... 110 Figure 4-1: Gravity membrane filtration Setup ...... 119 Figure 4-2: Effect of various degrading solutions on the PWP of BW30 membranes. Tested at 10 bar in high pressure stirred dead end cell...... 120 Figure 4-3: Effect of various degrading solutions on salt (NaCl) rejection of BW30 membranes. Tested with 2000 ppm NaCl at 10 bar in high pressure stirred dead end cell...... 121 Figure 4-4: Impact of conversion at 300,000 ppm.h on virgin DOW RO membranes. Tested at 2 bar in stirred low pressure dead end cell...... 123 Figure 4-5: Infrared spectra of BW30FR membrane, virgin and exposed to 300,000 ppm.h of NaOCl...... 124 Figure 4-6: SEM images of membrane surface at 40,000x magnification: A) Virgin BW30FR, B) BW30FR treated with 60,000 ppm.h NaOCl, C) BW30FR treated with 300,000 ppm.h NaOCl, D) Virgin 10 kDa UF...... 126

12 Figure 4-7: Effect of membrane condition on converted performance. Permeability tested at 2 bar...... 128 Figure 4-8: Dextran rejection results for UF and converted RO membranes. Tested at 2 bar and analysed using TOC detection...... 131 Figure 4-9: Organic carbon substance rejection of (A) Virgin UF, (B) converted virgin RO membranes, and (C) converted industrially used RO membranes, assessed with LC- OCD...... 134 Figure 4-10: AFM height map surface scans (active layer side) of; A) 10 kDa UF, B) 30 kDa UF, and C) converted BW30FR...... 138

Figure 4-11: Root mean squared roughness (Rq) for dry virgin UF and converted BW30FR RO membranes...... 139 Figure 4-12: Depiction of fouling effect for 60 min filtration and 10 min cleaning cycles at 30 L.m-2.h-1...... 141 Figure 4-13: TMP behaviour during cyclical fouling and cleaning...... 142 Figure 4-14: Fouling and cleaning performance of converted membrane samples from DOW-M1 and CSM-M1 modules in gravity filtration set up...... 145 Figure 4-15: Fouling on membrane sheet from gravity fed filtration...... 146 Figure 4-16: Permeability decline of converted DOW-M2 membranes used in gravity driven membrane treatment...... 147 Figure 5-1: Membrane Life Cycle...... 154 Figure 5-2: Simplified process flow diagram for membrane manufacturing and end-of-life scenarios including offsets...... 160 Figure 5-3: Relative impact from different components during the manufacturing of one RO module. Values above bars are the total emissions for the impact category. .. 168 Figure 5-4: Uncertainty propagation analysis of the membrane manufacturing model using Monte Carlo assessment. 2000 runs, 95% confidence interval...... 170 Figure 5-5: Greenhouse gas emissions and resource depletion for the disposal of one RO membrane element. Results are displayed in terms of relative offset of membrane production...... 174 Figure 5-6: Effect of secondary use phase lifespan on viability of reuse scenarios relative to landfill and recycling...... 178

Figure 5-7: Contribution of transportation and process to the climate change (CO2-e) emissions of the different scenarios...... 179 Figure 5-8: Effect of transportation distance on the viability of reuse scenarios relative to membrane reuse lifetime. (RO: direct reuse, UF: membrane conversion then reuse)...... 181

13 Figure 5-9: Benefit of direct reuse of RO membranes in terms of reduction in CO2-e emissions depending on transportation method used...... 182 Figure 5-10: Mass of waste material requiring landfill disposal for each of end-of-life scenarios for one RO membrane ...... 183 Figure 6-1: Structure of MemEoL tool...... 188 Figure 6-2: Capture of MemEoL input screen...... 189 Figure 6-3: Steps for solving a discrete MCDA problem. Adapted from (Zarghami and Szidarovszky, 2011)...... 191 Figure 6-4: Assessment criteria tree for decision tool...... 193 Figure 6-5: Relative assessment scores of end-of-life scenarios across assessment categories. (0) least favourable to (1) most favourable...... 199 Figure 6-6: Performance of new RO membranes from five of the major manufacturers...... 201 Figure 6-7: NaCl rejection and permeability performance degradation for end-of-life membranes (origin of the arrows indicates the initial performances, while the end of the arrow shows the end-of-life performance)...... 202 Figure A-1: SEM images of the backside of PSf layer of XLE RO membrane after the PET support layer has been removed...... 234 Figure A-2: SEM images of the PET support layer of XLE RO membrane after it has been removed from the PSf layer. The viewed side was directly in contact with the PSf layer...... 235 Figure A-3: SEM images of the active membrane surface for; Top) 30 kDa UF, and Bottom) 100 kDa UF...... 237 Figure A-4: Cross section SEM images of XLE RO membrane; Top) virgin, and Bottom) after use and drying. Images show the physical delamination of the PET and PSf layers in the dry sample...... 238 Figure A-5: Scattering patterns for the 25°C evaporation tests on the delaminated UF membranes for image numbers 1 and 100 (7 and 700 s)...... 239 Figure A-6: Scattering patterns for the 40oC evaporation tests on delaminated UF membranes for images 1, 100, 200 and 300 (7, 700, 1400, 3100 s)...... 240 Figure A-7: Patterns for the 60°C evaporation tests on delaminated UF membranes for images 1, 20, 25, 30 and 40 (7, 140, 175, 210, 280 s)...... 240 Figure A-8: Porod regressions for 10 kda UF membranes at maximum times for 25, 40 and 60°C series, using 0.6m camera length. The Porod law is express in its standard form for a flat material...... 241 Figure A-9: Effect of NaOCl degradation intensity on the PWP of DOW XLE membranes. Tested at 2 bar in low pressure stirred dead end cell...... 242 14 Figure B-1: A schematic representation of the experimental sessile-drop arrangement (Zaharia et al., 2009a)...... 244 Figure B-2: Images of slag droplets in contact with raw metallurgical coke at 1550°C as a function of time...... 245 Figure B-3: Carbon/slag interactions for metallurgical coke/slag and material blends at 1550°C...... 246 Figure C-1: Manufacturing process for RO TFC Membranes ...... 249 Figure C-2: Flow diagram and boundaries for landfill scenario ...... 249 Figure C-3: Flow Diagram and boundaries for incineration scenario ...... 250 Figure C-4: Flow diagram and boundaries for gasification scenario ...... 250 Figure C-5: Flow diagram and boundaries for electric arc furnace scenario ...... 251 Figure C-6: Flow diagram and boundaries for material recycling ...... 251 Figure C-7: Flow diagram and boundaries for conversion to UF scenario ...... 252 Figure C-8: Flow diagram and boundaries for direct membrane reuse ...... 252

vi. List of Tables

Table 1-1: Details on Australian major desalination plants (Water Desalination Report, 2013) ...... 21 Table 2-1: Comparison of proposed RO membrane and e-waste management schemes (data given for Australian market)...... 39 Table 2-2: Carbon content of membrane components (Prince et al., 2011)...... 48 Table 2-3: Summary of available decentralised membrane systems...... 63 Table 3-1: Key properties of studied membranes ...... 73 Table 3-2: Description of rewetting techniques...... 76 Table 3-3: Mass and performance change during membrane wetting and drying in the balance-desiccator at room temperature...... 88 Table 4-1: Condition and key properties of studied membranes...... 114 Table 4-2: Initial specifications of used RO modules ...... 118 Table 4-3: ICP results from Ag model virus particle removal...... 136 Table 4-4: Cleaning efficiency and changes in TMP...... 142 Table 5-1: Composition of a dry 8” and 16” SWRO membrane elements...... 155 Table 5-2: Selected ReCiPe impact categories and units ...... 157 Table 5-3: Summary of scenario transportation distances and sources ...... 167

Table 5-4: Summary of available literature data on SW desalination CO2-e emissions ...... 172

15 Table 5-5: Monte-Carlo simulation results for CO2-e emissions and Oil-e Consumption produced from pairwise comparison of end-of-life membrane options. Values represent the percentage of runs where the column variable (A) has higher emission/consumption compared to its corresponding row variable (B). A result can be considered statistically significant if the direction of results was consistent for over 90% of runs (Goedkoop et al., 2010)...... 177 Table 6-1: Qualitative criteria assessment matrix...... 194 Table 6-2: Normalised decision matrix...... 198 Table 6-3: Performance categories for membrane reuse ...... 204 Table 7-1: Summary of results regarding membrane drying and pore collapse...... 208 Table C-1: Life cycle inventory and uncertainty values for membrane manufacturing 253 Table C-2: Life cycle inventory and uncertainty values for end-of-life scenarios ...... 257 Table C-3: Overall impacts for membrane manufacturing, including breakdown into sub components...... 264 Table C-4: Overall impacts of disposal scenarios, including membrane manufacturing and distribution ...... 265 Table C-5: Material composition breakdown of different sizes of RO Membranes ..... 266 Table C-6: Basic uncertainty factors for inputs, outputs and elementary flows. C: Combustion emissions, P: Process emissions, A: agricultural emissions (Frischknecht et al., 2007) ...... 267 Table C-7: Pedigree matrix for uncertainty calculation (Frischknecht et al., 2007). ... 268 Table C-8: Monte-Carlo simulation results for emissions and consumptions produced from pairwise comparison of end-of-life membrane options. Values represent the percentage of runs where the column variable (A) has higher emission/consumption compared to its corresponding row variable (B). A result can be considered statistically significant if the direction of results was consistent for over 90% of runs (Goedkoop et al., 2010)...... 269 Table C-9: Summary of end-of-life options for RO membranes...... 271 Table C-10: Summary of survey of Australian landfill and recycling costs ...... 272

16 vii. Nomenclature

ABS Acrylonitrile butadiene styrene AFM Atomic force microscopy APC Australian Packaging Covenant BB Building Blocks BP Biopolymers BSA Bovine serum albumin BW Brackish water CA Cellulose acetate CE Cleaning efficiency DMF Dimethylformamide DOC Dissolved organic carbon DTF Drop tube furnace EAF Electric arc furnace FTIR Fourier transform infrared spectroscopy GDMD Gravity driven membrane disinfection HA Humic Acid ICP Inductively coupled plasma LCA Life cycle assessment LCI Life cycle inventory LCIA Life cycle impact assessment LC-OCD Liquid chromatography - organic carbon detection LRV Log removal value MAUT Multi-attribute utility theory MC Metallurgical coke MCDA Multi-criteria decision analysis MED Multi-effect distillation MF Microfiltration MPD m-phenylenediamine MWCO Molecular weight cut off NF Nanofiltration NPC National Packaging Covenant PA Polyamide PEG Polyethylene glycol PES Polyethersulfone PET Polyethylene terphthalate PP Polypropylene 17 PSf Polysulfone PVP Polyvinyl Chloride PWP Pure water permeability RO Reverse osmosis SAW Simple additive weighting SAXS Small angle x-ray scattering SEM Scanning electron microscopy SLS Sodium lauryl sulfate SMBS Sodium metabisulfite SW Sea water TDS Total dissolved solids TFC Thin film composite TGA Thermo gravimetric analysis TMC Trimesoyl chloride TMP Trans membrane pressure UF Ultrafiltration UV Ultraviolet WPC Wood plastic composites

18 Introduction

1. Introduction

1.1. Justification

Over the last two decades, there has been a 4 fold increase in the number of reverse osmosis (RO) desalination plants, and a 10 fold increase in capacity, with over 10,000 installations of various sizes currently online around the world. In Australia alone, the total RO capacity has risen from less than 100 ML.d-1 in 1990 to nearly 3000 ML.d-1 in

2012 (Global Water Intelligence, 2012). The size of these RO plants has also increased significantly, now exceeding 500,000 m3.d-1 in many parts of the world. In Australia again, six large municipal sea water desalination plants plus over a hundred commercial plants of various sizes are accounted for (Water Desalination Report, 2013). Plant expansion is largely attributable to population growth, while decreased process energy costs and improved membrane performance remain critical contributing factors (Peñate and

García-rodríguez, 2012; Schrotter et al., 2010).

Notwithstanding the constantly growing need to secure water production, the desalination industry still faces the challenge of improving its environmental sustainability

(Mezher et al., 2011; Roberts et al., 2010). Although the large amount of energy needed to pressurise the feed water during RO desalination is the primary environmental concern, the disposal of old RO modules is emerging as a critical issue to be addressed.

So far, used RO modules are considered as waste and are generally discarded to municipal landfills, with only limited disposal alternatives proposed to RO users.

Given the range of pre-treated water qualities and their associated operating conditions, it is estimated that an average of 100 of the 8” modules are generally needed to produce each mega litre per day of water. Based on an average of 13.5 kg per 8” module, the use of a basic single pass system, and a mean membrane life of 6 years, the inventory of the current plants worldwide allows the estimation of the total mass of modules to be disposed annually (Figure 1-1). The steady increase in the mass of disposed 19 membranes, reaching 12,000 tonnes.yr-1 globally by 2015, clearly indicates the magnitude of this disposal problem. In Australia, the current growth in the industry has resulted in an estimated 800 tonnes.yr-1 of used membranes to be disposed by 2015.

15

10

5

Annual mass of membranesmassofAnnual disposed globaly ( x1000 Tonnes) globaly x1000 ( disposed 0 1985 1990 1995 2000 2005 2010 2015

Figure 1-1: Estimation of the annual mass of RO modules to be discarded worldwide.

These calculations assume a constant rate of desalination plant construction and expansion; however, that is proving to not be the case. In Australia, the large plants, which can produce up to 440,000 m3.d-1, were all planned and constructed during a time of extreme drought. However, due to recent rainfall, political and budget concerns, a number of the plants have recently entered standby modes of varying degrees (Water

Desalination Report, 2013). A summary of the major desalination plants in Australia, and their current status, can be found in Table 1-1. While this may decrease the number of modules to be continually replaced and disposed of, it does pose an interesting challenge for the membranes stored at the plant. While temporary membrane storage is possible, little investigation into long term mothballing has been completed, and it is generally assumed that membrane performance and quality will decrease during extended

20 storage. If an option besides landfill disposal can be proposed, then it will provide an

alternative use for these membranes.

Table 1-1: Details on Australian major desalination plants (Water Desalination Report,

2013)

Project Capacity (m3.d-1) Commissioned Status Operating at 100% Perth 1 – Kwinana 140,000 Feb 2007 capacity Phase 1 operating at Perth 2 - 100% capacity. Phase 2 100,000 2011 – Sep 2013 Binningup due to be online by end of 2013 Gold Coast – Hot standby since Dec 133,000 Jan 2009 Tugun 2010 Mothballed after 2 years Sydney – Kurnell 250,000 Apr 2010 operation Western Corridor Recycled Water 232,000 (over 2007-2008 Mothballed Scheme – three plants) Queensland Operating at 100% but Adelaide 270,000 2012 will be reassessed after 2 year proving period Victoria – Standby mode after 440,000 2012 Wonthaggi commissioning

1.2. Aims and objectives

The overall aim of this thesis is to build on the needs of the desalination industry by

exploring and assessing the alternative end-of-life options available for used RO

membrane elements following the waste management hierarchy. This work focuses on

converting a potential environmental and economic liability, into an opportunity for value

recovery. Outcomes for membrane users and manufacturers include an interactive,

adaptive and informative tool which can assist in identifying the optimum end-of-life

avenue. Some aspects of this study, particularly the life cycle assessment (LCA), is

primarily focused on the Australian desalination industry and the related available

technologies, but does also explore end-of-life option in a global context. 21 The major objectives of this thesis therefore are to:

 Identify and assess the technical viability of available end-of-life options for RO membranes.  Investigate and propose mechanisms for performance loss during improper storage of ultrafiltration (UF) and RO membranes, and determine an optimum rewetting methodology for dried membranes.  Develop an optimum method for converting used RO modules into UF microporus membranes for subsequent reuse, and characterisation of the resulting membranes.  Assess the environmental impact of membrane manufacturing and available end-of-life options.  Develop a practical, educational web-based decision making tool to assist membrane users in identifying optimum end-of-life strategies for their specific situation.

22 1.3. Chapter Descriptions

Chapter 2 discusses the use and lifespan of RO membranes in seawater desalination, identifies potential end-of-life options for used membranes based on the waste management hierarchy, outlines the legislation for product stewardship in Australia, and reviews the available literature on end-of-life scenarios and LCA.

Chapter 3 investigates the cause of membrane performance degradation from drying due to improper storage, and develops an optimum strategy for permeability recovery through rewetting.

Chapter 4 builds on previous research by optimising the protocol to remove the active layer from used RO membranes, resulting in a low cost UF-like membrane. Following this conversion, the membranes are characterised using a wide range of techniques and potential applications are explored.

Chapter 5 uses LCA to investigate the RO membrane manufacturing process, identifies areas of potential improvement, and puts its environmental impact into the context of the complete desalination process. Additionally, the LCA study compares the identified end- of-life options for used membranes based on environmental and landfill impact, and assesses the effect of transportation distance on reuse viability.

Chapter 6 outlines the development of an online educational tool aimed at providing information about end-of-life options to membrane users. The chapter discusses the dynamic multi criteria decision analysis system employed and details the inputs used in the tool.

Chapter 7 summarises the major outcomes of this work, and provides recommendations for future research activities.

Appendix A provides supplementary information for Chapters 3 and 4, including additional experimental data and images from SAXS and SEM analysis.

23 Appendix B reports on the methodology and findings of a preliminary investigation into the application of membrane components as a carbon source for electric arc furnace steel making.

Appendix C provides supplementary information for Chapter 5. The complete life cycle inventory for both membrane manufacturing, and the alternative end-of-life options is provided. Additionally, detailed process diagrams show the boundaries and inputs for each of the end-of-life options, and complete statistical results from Monte Carlo simulations are provided.

24 Chapter 2:

Literature Review

This chapter is an expanded version of the following peer-reviewed journal article:

Will Lawler, Zenah Bradford-Hartke, Marlene J. Cran, Mikel Duke, Greg Leslie,

Bradley P. Ladewig, Pierre Le-Clech, 2012. Towards New Opportunities for

Reuse, Recycling and Disposal of Used Reverse Osmosis Membranes.

Desalination, 299, pp.103–112

Literature Review

25 2.1. Introduction

As identified in Chapter 1, the accumulation of used membranes is becoming a serious challenge for the desalination industry. As the first step in addressing this problem, this literature review studies membrane desalination technology to assess opportunities related to the recycling, reuse and disposal of old RO membranes. It aims to investigate waste management techniques, including previous attempts at product stewardship, to understand how similar problems have been faced by other industries. A wide range of possible end-of-life options for RO membranes are identified and the current state of their development is discussed. Specific focus is given to the chemical treatment of membranes, in order to recondition them for a secondary reuse. A promising application of using the membranes for decentralised water treatment in developing areas is discussed, and the main criteria that are required for the products to be successful are identified. The significant challenges for membrane reuse of storage and transportation are identified and assessed. Finally, the methodology of LCA is discussed, and the current state of its application in terms of desalination and membrane technologies is reviewed.

26 2.2. Reverse Osmosis Membranes: Process and Materials

As the global population rises, and due to water scarcity issues like drought, there has been a growing need for water sources independent of rainfall. The most common process to fill this demand is seawater (SW) desalination. Desalination, which can also be used to purify water for industry applications, involves the removal of salts and other solutes from the water, thus producing potable water.

2.2.1. Membrane Desalination Processes

There are a number of different technologies available for desalination, including; Multi- stage Flash, multi-effect distillation (MED), vapour compression distillation, reverse osmosis (RO), nanofiltration (NF), and electrodialysis. Out of these available technologies, MED and RO are the most commonly used, with 40 and 44% of desalinated water being generated from these techniques respectively (Greenlee et al.,

2009). The MED process is favoured where low cost fossil fuel based thermal energy is available, such as in the Middle-East (Fritzmann et al., 2007), but RO has the lowest electrical energy consumption of all desalination process.

RO uses a semi-permeable membrane and a pressurised feed to extract and purify water. The process is commonly used with SW (15,000 – 50,000 ppm total dissolved solids (TDS)) and brackish water (BW) (1,500 – 15,000 ppm TDS) (Malaeb and Ayoub,

2011). The basis of the technology revolves around the semi-permeable polymer membranes, which allows the diffusion of water molecules, while rejecting other solutes.

An applied pressure greater than the osmotic pressure of the feed solution is required to drive the process. A pressure of 50-70 bar is required for SW and 10-15 bar for BW, depending on the solute concentration. Modern membranes can reject 98 – 99.5 % of monovalent salts, determined by the product used and operating conditions.

The membrane is a critical component of the overall RO process, which also involves water intake, pre-treatment, multiple membrane passes to increase product quality, and

27 various post treatment processes to bring the water quality in-line with drinking water standards. The membranes used for this process are of flat sheet construction and are in a spiral wound configuration to maximise surface area and reduce the footprint of the desalination plant. The structure of the membrane module, which can be seen in Figure

2-1, involves two membrane sheets with the active side facing out, sealed over a permeate spacer. The sheets are then wound up around a central tube, with a feed spacer in between each sheet. The water flows through the feed spacer channel and permeates through either of the adjacent membrane sheets, spirals down the permeate spacer and is collected in the central tube.

Figure 2-1: Structure of RO membrane module (US EPA, 2005).

The current standard size for RO membranes is 8” (203 mm) in diameter and 40” (1016 mm) long; however, 2.5” and 4” configurations are also used for smaller systems. Larger elements, for example 16”, have been developed and have seen limited adoption (Peery et al., 2004). Up to eight membranes are placed in series into one pressure vessel, sharing a common feed stream. A number of pressure vessels are organised into trains with independent pumping and energy recovery systems.

28 2.2.2. Membrane Structure

RO membranes where originally developed in the 1960s and were constructed of asymmetric cellulose acetate (CA). These membranes had the drawback of slowly degrading in water due to hydrolysis, and suffered from severe polymer compaction over time, reducing the performance. These weaknesses lead to the development of thin-film- composite (TFC) membranes using an active layer of dense non-porous polymer

(Cadotte, 1979).

Figure 2-2: Cross section structure of a TFC RO membrane.

As seen in Figure 2-2, TFC membranes typically consists of a dense polyamide (PA) active layer, a microporous polysulfone (PSf) supporting layer, and a considerably thicker non-woven polyester (PET) base (Lee et al., 2011). The membranes are generally manufactured by interfacial polymerisation which physically anchors the active and support layers together through interlocking of the pore structures (Soice et al., 2004).

The most common active layer structure used today is constructed of cross linked fully 29 aromatic PA. This structure, which can be seen in Figure 2-3, is commonly referred to as FT-30, and is based on the reaction of m-phenylenediamine (MPD) and trimesoyl chloride (TMC) (Lee et al., 2011). These membranes showed higher chemical stability and resistances, and are not affected by compaction to the same degree as CA membranes.

Figure 2-3: Cross linked fully aromatic PA active layer structure (Lee et al., 2011).

30 2.2.3. Membrane Operating Conditions and Lifespan

It is important to understand the mechanisms of membrane ageing and to assess its consequences, which ultimately results in the decline of RO performance to a point that it no longer meets required specifications. Several factors contribute to RO damage, which can be functionally determined by a decrease (or increase) in permeate throughput and/or a decline in salt rejection. Another common indicator for the decline in system performance is an increase in pressure drop along the pressure vessel.

The loss of RO performance can result from irreversible organic and/or inorganic fouling

(Kogutid and Kunst, 2002; Tang et al., 2011), chemical degradation of the active membrane layer (Antony et al., 2010), and the slow creep of the polymer due to long term application of pressure. Microbiological fouling, generally defined as the consequence of irreversible attachment and growth of bacterial cells on the membrane, is also a common reason for discarding old membranes (Matin et al., 2011). A combination of cleaning agents can be used to remove a variety of foulants, including acids for iron and metal oxides (Schrotter et al., 2010); commercial anti-scalants (Koo et al., 2001; Malaeb and Ayoub, 2011); alkalines, biocides, detergents and enzymes for bio-fouling (Ang et al., 2011a, 2011b; Matin et al., 2011). The repetitive and incidental exposure of these solutions can adversely affect the membranes, generally through the decrease in their rejection efficiencies (Gao et al., 2011; Porcelli and Judd, 2010).

All of these factors lead to the expected lifetime of the membranes to be widely variable.

It can depend on the feed quality, operating conditions, cleaning routine, and acute incidents such as dechlorination failure or abrupt water quality change. Manufacturers generally provide a three year limited warranty for purchased membranes, as long as the provided operating conditions are met. It is commonly stated that the membranes have a lifespan of 5-10 years; however, there is limited reliable public information to confirm this (Baker, 2004; US EPA, 2005). The longevity of membranes can be measured in two different ways; first, as absolute membrane lifespan in years, and 31 secondly in replacement rates. For example, if a plant has a replacement rate of 5%, it is expected that 5% of membranes will be swapped each year of operation.

There have been a number of reports of membrane lifespan using these metrics. One study, citing a number of private industry sources, reports a RO membrane lifespan of

5-8 years, depending on the varying salinity of the Mediterranean sea (Raluy et al.,

2005a). Similar numbers were also reported by another team in Spain, with some of the plants discussed using CA membranes (Fernandez-Álvarez et al., 2010). Another study, assessing the lifespan of membranes from a variety of manufacturers, and a number of plants in Greece, reported that the membrane lasted 2-5 years before requiring replacement (Avlonitis et al., 2003). A general review of desalination technology reports a replacement rate of 20% (2-5 year membrane lifespan) for SW application and 5% (5-

7 year lifespan) for BW applications (Greenlee et al., 2009). Studies focusing on the economic performance of desalination systems have to take into account membrane replacement rates, some using retrospective data and others using assumptions for cost forecasting. One economic model from Hydranautics assumes a 20% replacement rate due to the high salinity of the Mediterranean (Wilf and Klinko, 2001), while another study uses a manufacturer supplied rate of 13% to model the cost-effectiveness of small scale systems (Gilau and Small, 2008), and a study published by GE membranes reported that using their UF pre-treatment system would reduce the replacement rate from 14% to

10% per year (Wolf et al., 2005). A more recent study, reporting the trends in the industry and major parameters influencing desalination cost, mentions a 14-20% membrane replacement rate (Ghaffour et al., 2013).

Ultimately, membrane lifespan varies considerably and no one value can be universally applied. Nevertheless, at some point, the membranes will require replacing, leading the disposal of the original modules. Currently, landfill disposal is the most commonly fate for the used modules, leaving an opportunity for improvement.

32 2.3. Waste Management

Concepts inherent in the waste management hierarchy, shown in Figure 2-4, can help to assess and prioritise the optimum strategies for the management of old RO modules

(EPA, 1996). It is generally accepted that disposal options for waste and end-of-life products follow in order of preference from the top of the inverted pyramid: reduction, reuse, recycling, energy recovery, treatments, and then disposal.

Figure 2-4. Waste management hierarchy from most to least preferred options.

The first option involves reducing the amount of end-of-life membranes requiring disposal. This could be done through increasing the lifespan of the membranes, reducing the number of membranes required per m3 of water treated, or using an alternate technology. These broad categories are being actively pursued by membrane researchers around the world, and while their primary goals are to make membrane technologies more efficient and economically viable, they have the additional benefit of reducing membrane waste. While the act of avoiding and reducing the use of RO modules is not within the primary focus of this study, it will be considered for comparison when assessing new end-of-life options.

Reuse is the second most favourable option to consider for end-of-life products. The membrane elements as a whole, or in part, could be used in lower specification

33 applications. This could be as simple as using sea water RO membranes in BW applications at lower pressures, or components of the element could be used in a secondary process. Recycling involves physically transforming the membrane or its components into new materials that can be used to make new products. This process could use traditional recycling techniques, or novel ones developed specifically for membranes. The category of energy recovery covers any technique that converts the waste material into a usable form of energy, be it heat or electricity. These include techniques like pyrolysis, gasification and incineration. Similar to energy recovery is the treatment category. This involves converting the waste material into a form that has a lower impact on the environment. In the case of membrane modules, which do not produce emissions during landfill, this option would focus on reducing the amount of waste to be sent to landfill. The final and least preferred option is disposal, which refers to traditional landfilling and is currently the major available option for end-of-life membranes, especially in Australia.

2.4. Product Stewardship

The disposal of used RO membranes must be considered in the context of its relative environmental impact, compared with other types of waste. In Australia alone, over 21 million tonnes of waste is disposed into landfill each year, including more than 80,000 tonnes of PET water bottles, and in excess of 300,000 tonnes of electronic waste (e- waste) (DEWPC, 2010). These large numbers have caused governments around the world to consider novel strategies to deal with domestic and industrial wastes, including high-end products that have reached the end of their lifetime. This section briefly describes practices recently implemented for e-waste in Australia as an illustration of strategies to consider for the management of old RO membranes.

34 2.4.1. Early Policy Attempts for E-waste

Because of their large environmental impact (both quantitative and qualitative), recent efforts have focused on the fate of e-waste, such as television and computer equipment, and a number of recycling programs have been implemented around Australia. For example, the Australian Capital Territory government has already banned landfill disposal of e-waste, while the State of Victoria has setup a computer collection and recycling scheme (Environment Protection and Heritage Council, 2009a). In addition, a number industrial suppliers (e.g. Dell and Apple) propose various recycling schemes for their own products (Environment Protection and Heritage Council, 2009a). However, these industry programs have not significantly expanded in practice since product recycling is rarely financially viable without government support. As a result, the reported recycling rate was as low as 10% by weight during 2007/2008 (Environment Protection and Heritage Council, 2009b). While suppliers and importers usually demonstrate great interest in participating in e-waste recycling schemes, they could suffer from potential financial disadvantage if some of their competitors do not also join the program. As a result, the setup of a compulsory recycling scheme appears critical in this highly competitive market (DEWHA, 2010).

2.4.2. National Television and Computer Product Stewardship Scheme

In Australia, the “National Television and Computer Product Stewardship Scheme” is the first instalment of the Product Stewardship Scheme, developed under the new “National

Waste Policy”. The Product Stewardship Bill was passed by the Australian parliament in

June 2011, and provides legislation for nationwide practice for evaluating the entire product life-cycle, particularly end-of-life options, minimising their environmental impact and sharing of responsibility among manufacturers, importers, governments and consumers (The Parliament of the Commonwealth of Australia, 2011).

35 Specifically, this document provides a framework to impose obligations on key stakeholders regarding the avoidance, reduction and management of waste from products. Under the scheme, liable parties will meet their obligations by becoming a member of an approved product stewardship arrangement, available in three configurations:

1. Voluntary product stewardship arrangements are intended to encourage product

responsibility without regulation, and to provide community assurance through an

accreditation system. Once accredited, the party may use the accreditation logos

to promote the environmental and social responsibility of the party (Cain, 2010).

However, the legislation does not require all businesses within an industry to

participate in this type of arrangement.

2. Co-regulatory product stewardship features more formalised arrangements, for

which the government is expected to set minimum outcomes and requirements,

providing the industry with the flexibility to control how these outcomes will be

met.

3. Mandatory product stewardship arrangements are controlled by government,

responsible to initiate regulations, and are backed up by criminal and/or civil

penalties if the requirements are not met.

In Australia, this scheme has been used to target television and computer equipment under a co-regulatory framework (i.e. “e-waste management scheme” (The Parliament of the Commonwealth of Australia, 2011)), and is expected to quickly expand to tyres, mercury containing lights and packaging (Department of the Environment Water

Heritage and the Arts, 2010). Under the e-waste regulations, the key commitments for television and computer producers includes covering the cost of implementing the scheme. This will include collection infrastructure, recycling, awareness and education programs, governance activities, and development and provision of information used to identify relevant products to be covered by the scheme. A threshold will be set to exclude 36 corporations that import or manufacture less than 5,000 television and computer products annually.

One of the significant challenges facing the electronics industry is the number of product locations, with every household being a potential source of end-of-life products. The collection of such used products is a major obstacle to the effective implementation of this scheme. Therefore, collection points, including council pick-up and designated drop- off locations have been organised.

2.4.3. The Australian Packaging Covenant

Another example of an existing framework for reducing the environmental impacts of consumer and industrial wastes is the Australian Packaging Covenant (APC), formerly known as the National Packaging Covenant (NPC) (Australian Packaging Covenant,

2010). Established in 1999, the NPC was developed to manage the environmental impacts of consumer packaging and paper wastes. The key areas targeted by NPC include: designing packaging that is more resource efficient and more recyclable; increasing the recovery and recycling of used packaging from households and other sources; and taking action to reduce the incidence and impacts of packaging litter.

The early stages of the NPC development involved seeking cooperation between stakeholders who include all sectors of the packaging chain and included both industry, local, state and commonwealth governments. Stakeholders could sign the voluntary

Covenant with the provision that they submit action plans and annual reports based on their activities specific to their industry. The original form of the NPC proposed a term of five years which was extended for the same period in 2005. From July 1st, 2010, the

APC emerged in an open-ended revised form that included committing signatories to national targets and key performance indicators. Concurrently, the APC released the

“Sustainable Packaging Guidelines” which signatories can use to assist them to analyse and document their packaging sustainability strategy (Australian Packaging Covenant,

37 2011). At present, a total of 713 Australian and international stakeholders are signatories of the APC indicating a strong commitment to the principals of the Covenant.

2.4.4. Towards Implementation of Product Stewardship for RO Membranes

As discussed earlier, many challenges still limit the prompt implementation of a systematic approach to the management of RO modules from cradle to grave. However, important lessons could be learned from the current implementation of the Australian e- waste management scheme and the APC towards the stewardship of RO modules. For example, it is important to note that extensive consultation of all stakeholders was undertaken during the drafting of the Stewardship Bill and the NPC/ACP, and it is expected that the desalination industry at large will be keen to work towards the implementation of a similar scheme for RO membranes. It also remains crucial to clearly define the main difference between post-consumer wastes and old RO modules, and the resulting challenges and opportunities for implementing a similar scheme to RO products. When considering the significantly smaller volume of solid waste disposed of by the desalination industry, government intervention is not expected to be a national priority, although state and/or federal involvement could be necessary in order to implement this type of recycling scheme. At this early stage, the Product Stewardship

Act already allows the desalination industry to seek voluntary accreditation for RO modules. However, no membrane suppliers or users are known to have joined this scheme in Australia to date.

As summarised in Table 2-1, the potential RO membrane regeneration scheme offers advantages over the e-waste program, such as a significantly lower number of collection points, and therefore a much higher volume of product to collect in each location. This characteristic is expected to greatly affect the transportation costs associated with this initiative as used RO membranes are located in a relatively small number of desalination plants in known locations. Although the complex network of collection points used for the e-waste scheme will not be required, membranes will still need to be transported from 38 the original plant to the facility responsible for reuse/recycling treatment and then to another plant for reuse in the potential second life application.

Table 2-1: Comparison of proposed RO membrane and e-waste management

schemes (data given for Australian market).

E-waste RO membrane Voluntary (so far) Co-regulatory Industry-run Scheme Type Industry-run No mandatory participation Government controlled Accredited promotion Waste Generation Households Membrane plants Amount of waste 300,000+ 800+ (tonnes)

Waste Locations 5 million + 180 + Direct material recycling Direct and indirect reuse, End-of-Life options after disassembly, limited material recycling after reuse disassembly, waste reduction.

2.5. End-of-life Options

The first step towards successfully minimising membrane waste and its impact on landfill is the identification of alternative end-of-life options. Using the concepts of the waste management hierarchy and the lessons learnt from the application of product stewardship for other products, a number of potential alternative end-of-life options can be generated. This section discusses the background and previous studies investigating the available options.

2.5.1. Membrane Reuse

Direct membrane reuse involves taking membranes that have been deemed unsuitable for their primary application from one plant and transporting them to a secondary plant, potentially for use in harsh conditions where high replacement rates makes low cost second hand modules economically attractive (Frick et al., 2014). Published only recently, a limited number of reports have mentioned the potential of direct RO reuse.

39 The two identified studies were based on the detailed characterisation of used RO filters after SW filtration (Ould Mohamedou et al., 2010; Prince et al., 2011). One study from

France (Ould Mohamedou et al., 2010), used autopsy techniques including hydraulic permeability, salt rejection, morphological and topographical parameters with field emission scanning electron and atomic force microscopes. The used RO element showed performances similar to those usually obtained by NF membranes including an increase in permeability from 1.0 to 2.1 L.m-2.h-1.bar-1 and a decrease in NaCl rejection from >90% to 35-50% (Ould Mohamedou et al., 2010). Given their average molecular weight cut off (MWCO) ranging between 100 and 1000 Da, and their capacity to remove small organics and divalent ions, NF membranes are generally used in the food processing, water and wastewater industries (industrial and domestic). Similar findings were obtained when the performances of used membrane elements were compared to those reported for new membranes during another study in Australia (Prince et al., 2011).

It was concluded that although the sampled membranes were no longer in accordance with the manufacturer’s performance criteria of 99.5% salt rejection, all tested membranes showed more than 96% rejection. The authors proposed that this level of salt-rejection would allow repurposing of the membranes for applications including brackish-water treatment (Prince et al., 2011). Other direct reuse applications such as

SW pre-treatment and selective demineralisation of BW can be considered as suitable reuse strategies for used RO membranes (Anne et al., 2001; Van Der Bruggen and

Vandecasteele, 2003; Wang et al., 2009).

Although direct reuse of used membranes without any additional treatment is obviously preferred, appropriate assessment of their current performances, further validation and potential chemical cleaning will most likely be required. Information on the used membrane performance from the plant of origin, as well as a number of additional monitoring and characterisation methods, could enable a tailored cleaning system to be developed (Pontié et al., 2005). Based on the type of feed water used and location of the

40 RO membranes within the filtration train, standard chemical cleaning protocols could be established in order to efficiently remove most of the fouling/scaling that may have occurred during the filtration. A critical step in this reuse process is the validation of the cleaned membranes, as integrity, permeability and rejection must be demonstrated before the membranes can be successfully reused. This could be obtained through detailed reporting of the RO performance at the end of their initial life time.

To date, one US-based company, WaterSurplus, has recognised the potential of directly reusing membranes. The company offers a wide range of surplus new, as well as used, cleaned and repackaged, RO, UF and NF Membranes and associated equipment. A survey and review of products available from WaterSurplus was completed between

2012 and 2014 (WaterSurplus, 2014). The unit cost for 8” RO elements ranges from $75 to $400 USD, with batches of up to 400 elements available to purchase. The membranes available come from a variety of previous applications, as seen in Figure 2-5, and have performances ranging from 90-99.5% salt rejection and permeability between 40-157% of the manufacturers virgin performance. This information highlights the variety of conditions of the used membranes after their primary use. Membranes with permeabilities lower than their virgin performance may have undergone significant fouling and require intensive cleaning, while membranes with higher permeabilities may have physical or chemical damage to their surfaces.

41 Well Water 17% Not Specified 38% Municipal Drinking Water 17%

Tertiary City Water Waste 14% Water 14%

Figure 2-5: Sources of WaterSurplus reusable membranes.

Another potential reuse option involves chemically removing the active PA layer of the

RO membrane. Theoretically, this would leave the PSf support layer exposed and act as a typical UF membrane (Ghosh and Hoek, 2009). If possible, these membranes could be reused in a number of UF applications, providing a low cost alternative to new membrane modules, and extending the life of the original RO product. This technique will be discussed in detail in Section 2.6.

2.5.2. Multi-membrane Vessel Design

The concept of using RO elements with different performances within the same pressure vessel in order to optimise the overall process efficiency has been recently introduced

(Peñate and García-rodríguez, 2012). This novel hybrid system generally proposes the use of high rejection, low productivity membranes in the upstream section of the filtration train, followed by high productivity, low energy membranes downstream. This design benefits from either a reduction in power consumption as a result of lower pressure requirements, or an increase in productivity. With fewer modules and pressure vessels required for a given application, this concept could also result in significant decrease in capital costs (Peñate and García-Rodríguez, 2011). 42 Fresh Old membranes Oldest membrane 1 moves down 2 3 6 7 membrane exits enters

Figure 2-6: Old membrane movement along the pressure vessel in hybrid system.

It is proposed to adapt this original concept, by internally reusing older RO membranes within the same pressure vessel. The sequential, strategic movement of membranes to higher productivity positions can coincide with a change in performance, resulting from membrane ageing and degradation. When a fresh high rejection membrane is inserted into the lead position of the module, the removal of the oldest module at the end of the vessel will be conducted simultaneously, allowing a gradual movement to older membranes downstream. This movement is demonstrated in Figure 2-6. A more rigorous cleaning protocol could be applied just prior to this changeover, thus sparing the fresh module from the potentially damaging process. In order to assess the viability of this concept, the membrane performance will first need to be validated and monitored, as well as an economic impact of the extra labour required.

2.5.3. Material Reuse

Material reuse involves the repurposing of the component materials for a secondary application, thus extending the products lifecycle. These applications include use as aggregate in concrete, geotextiles, and in wood plastic composite materials.

It has been shown that a number of plastic types (including ones used in the construction of RO membranes) can be ground into an aggregate that can be used in concrete (Saikia and de Brito, 2012). However, the resulting concrete has a decreased bulk density, with reduced compressive and splitting tensile strength (Siddique et al., 2008). This decrease in performance is directly proportional to the substitution ratio of plastic aggregate, and is attributed to the low binding strength between the plastic particles and cement paste.

Currently, there is no known Australian companies using plastic aggregate substitute in

43 their concrete mix and therefore this options in not currently available for end-of-life membrane modules.

One interesting application of membrane components involves their use as geotextiles.

Geotextiles are generally permeable synthetic fabrics used to separate, protect or retain, soils or gravel in landscaping or construction applications. There has been some previous work showing that the membrane sheets and spacers can be used in a variety of these applications (Ould Mohamedou et al., 2010). Additionally, potential agricultural applications have been explored for the membrane spacers, including bird netting, wind- breakers or nets for lawn protection (Prince et al., 2011). However, while membrane sheets and spacers hold similar properties to commonly used geotextiles, an unrolled membrane sheet (and accompanying spacer) is no more than 2 m in length per leaf.

When compared to the lengths generally required in geotextile applications, this disadvantage limits membrane component use to small scale applications in the home garden.

One emerging application for waste plastics is their use in wood plastic composites

(WPC) (Kazemi Najafi, 2013). WPC are produced by mixing virgin or recycled plastics

(including all types used in membrane module construction) with ground wood particles.

The mixture can be extruded or moulded into a desired shape and can be used in a number of applications including decking, roof tiles, cladding, landscaping timbers etc.

WPC have a number of advantages including increased resistance against biological deterioration, when compared to untreated timber products for outdoor applications, high environmental sustainability and low cost (Wechsler and Hiziroglu, 2007). As there are a growing number of companies manufacturing WPC products in Australia and around the world, this is a possible application for waste plastics like used RO membranes.

44 2.5.4. Material Recycling

Plastic solid waste treatment and recycling can be separated into four major categories, primary (re-extrusion), secondary (mechanical), tertiary (chemical) and quaternary

(energy recovery). Primary recycling is generally conducted within the manufacturing plant by reintroduction of clean scrap back into the extrusion cycle. Generally, this process cannot be applied to dirty waste products, such as used or even cleaned RO modules, as the recycling materials are not expected to meet the required quality (Al-

Salem et al., 2009).

During the process of mechanical recycling, plastics are physically ground into suitable size, separated from contaminants, washed, then used as feed stock for the production of new products (Welle, 2011). This process remains difficult to apply to mixed polymers or contaminated materials, as immiscible or incompatible polymers can cause mechanical property deterioration during the process. Thus, in order to make mechanical recycling economically viable, it is important to have a large volume waste stream of single-polymer plastic that is clean and homogenous (Brandrup and Wiesbaden, 2005;

Goodship, 2007; Siddique et al., 2008).

With regards to membrane modules, each component must be considered individually to determine their potential suitability for mechanical recycling, assuming they can be successfully and economically separated. For example, the polypropylene (PP) feed spacer (Prince et al., 2011), has the ability to be directly recycled using this method

(Howell, 1992). Indeed, PP is commonly recycled into containers and packaging due to the strength, thermal and chemical resistances it can maintain, even after being recycled

(Meran et al., 2008; Recoup, 2005). Depending on the type of polyester used in components such as the permeate spacer, these also have the capacity to be mechanically recycled. Because of their copolymer nature, acrylonitrile butadiene styrene (ABS) materials such as the end caps and permeate tubes can suffer a deterioration of physical properties when recycled through this method (Arostegui et al., 45 2006). Finally, the membrane sheets, which make up a large proportion of the module, are constructed from a number of different polymers and additives and would therefore be inherently difficult to accurately and efficiently separate. In addition, the membranes sheets can be contaminated by a wide variety of substances after extensive use, further complicating the process. Due to the nature of the process and the aforementioned reasons, direct mechanical recycling the module as a whole may prove to be labor and cost prohibitive.

Chemical (or feedstock) recycling is a process which chemically breaks down the polymer, to be used as raw materials for petrochemicals processes, by using the reverse of the method used to create the polymer chains, such as depolymerisation and degradation (Al-Salem et al., 2009; Ram, 1997). Polyester materials (such as in the permeate spacer and components of the membrane sheet) are suitable for chemical recycling processes, and hydrolysis is used to reverse the polycondensation reaction used to make the polymer, with the addition of water to cause decomposition (Scheirs,

1998). Chemical recycling cannot typically be used with contaminated materials

(Goodship, 2007), and while more expensive and complex than mechanical and primary recycling, its main advantage is that heterogeneous polymers with limited use of pre- treatment can be processed (Al-Salem et al., 2009). Chemical recycling processes are tailored for individual materials and further classification will be required to determine if this avenue is suitable for the recycling of RO membrane modules.

2.5.5. Energy Recovery

The next category considered by the waste management hierarchy is energy recovery and thermal processing, which can be defined as the conversion of solid wastes to a secondary product with a release of heat energy (Goodship, 2007). The major categories of thermal processing commonly used in industry include: incineration; pyrolysis or thermal processing in the absence of oxygen; gasification, which is the partial combustion with limited air to produce syngas; and catalytic conversion to fuel oil (Arena, 46 2012; Keane, 2009; Wu and Williams, 2010). Environmentally, gasification and pyrolysis offer advantages over simple incineration, as they produce fewer emissions, reduce waste residues, and increase energy recovery (Zaman, 2010). Most importantly, these processes can be applied to mixed plastic wastes, such as the combination of materials used in the manufacturing of RO modules (Al-Salem et al., 2009; Brandrup and

Wiesbaden, 2005; Goodship, 2007). Incineration of plastic solid waste can reduce the volume by 90-99%, greatly reducing the strain on landfill. In addition, heat energy can be recovered and used for electricity generation or other heat related processes (Yassin et al., 2005).

The thermal decomposition of the polymeric components of RO membranes has been reported recently (Prince et al., 2011). With the exception of the fiberglass outer casing, the membrane components are comprised of synthetic polymers. As shown in Table 2-2, the carbon content of the polymers is between 62.2 and 88.3% by mass for the major membrane components. With a total mass of 13.5 kg, a typical 8” RO membrane element is therefore expected to contain approximately 9.1 kg of carbon.

47 Table 2-2: Carbon content of membrane components (Prince et al., 2011).

Membrane element Approximate carbon Composition Component content (%)

Outer casing Fiberglass 30-50

Feed spacer Polypropylene 85.7

Permeate spacer Polyester 62.5

Aromatic polyamide 71.6 Membrane sheet (thin Polysulfone 73.7 film composite) Polyester support 62.5

Permeate tube/end caps Acrylonitrile butadiene styrene 88.3

Glues Epoxy resin 62.2

Rubber o-rings EPDM rubber 83.6

Figure 2-7 shows the results of thermo gravimetric analysis (TGA) of the membrane components performed under nitrogen atmosphere. Most of the materials are comprised of multi-component systems including the fiberglass casing which showed considerable residual content (i.e. inorganic). The components comprised of ABS or PP are the least thermally stable, showing almost complete degradation at 400°C. The other parts are comprised of components that are more thermally stable, with the membrane sheet completely degraded at 600°C, followed by the permeate spacer and glued parts at

900°C. It is therefore possible to thermally degrade the polymer components to carbon using thermal treatments (Dollimore and Heal, 1967).

48 1.0

0.8

Fiberglass 0.6 Glued parts Polysulfone 0.4 Membrane sheet

Normalised Mass Normalised Permeate spacer 0.2 Permeate tube Feed spacer 0.0 0 200 400 600 800 Temperature (°C)

Figure 2-7: TGA of membrane components heated under nitrogen atmosphere at

20°C.min-1 (Prince et al., 2011).

Waste incinerators can generally operate from 760 to 1100°C (Johmke, 1992) and would therefore be capable of removing all combustible material, with the exception of the residual inorganic filler of the fibreglass casing. For a sample of the 8” RO membrane, only 7% of the original mass remained after combustion at 900°C. Pyrolysis of the same membrane under nitrogen at a temperature less than 450°C results in a higher residual of 26%. In either case, the overall volume of waste sent to landfill is significantly reduced.

The process of waste incineration, if not properly controlled, can emit greenhouse gases

(Australian Government Productivity Commission, 2006) as well as other harmful products (Al-Salem et al., 2009). As a result, additional emission capture technologies are expected to be considered and implemented, in accordance to the current legislation.

The recovery of energy from the process can also offset the greenhouse gas emissions from traditional energy generation, further increasing the sustainability of the incineration process (Assamoi and Lawryshyn, 2012).

49 2.5.6. Energy Recovery in Electric Arc Furnace

The use of polymeric wastes as a coke substitute in electric arc furnace (EAF) employed steel making is a relatively new concept, although the process has already been implemented on large scale. Currently, old rubber tyres and other waste plastics are used as a partial coke substitute in EAFs in a number of plants (Ayed et al., 2007;

O’Kane, 2011). This section briefly describes the primary concepts underlying the potential use of old RO modules as a carbon and energy source in EAFs.

Although the most common method used to produce steel products remains the blast furnace, the use of EAFs, which accounts for 35% of global production (Ghosh and

Chatterjee, 2008), usually results in higher quality steel and offers the opportunity to use plastic waste as a partial energy source. The EAF productivity can be further improved by the formation of foaming slag (Hara et al., 1990). In this process, carbon, commonly found in the form of metallurgical coke, is injected into the reactor, where it combines with oxygen in the molten steel and generates CO and CO2. These gasses bubble through the molten metal, producing the foaming slag which improves the electrical energy transfer from the electrodes to the molten bath, while also protecting the electrodes and furnace walls. The reaction between the carbon and oxygen is exothermic and produces thermal energy, which lowers the operation costs (Stadler et al., 2007).

The use of waste plastics in EAFs presents the potential to provide an extra energy source and to promote slag foaming (Sahajwalla et al., 2010). The thermal decomposition of the polymeric substances into CO and H2 make them ideal for metallurgical processes, as these gasses help the reduction of oxide ores (Ogaki et al.,

2003). Using waste polymers in EAFs provides extensive environmental advantages, such as a reduction of the amount of both the specialty cokes commonly required and the solid plastic wastes sent to landfill. Increased furnace efficiency, decreased power usage and decreased carbon consumption are also environmental and economic advantages. As various carbon sources are already used in EAF steelmaking, no 50 significant modifications to the plant design are required to implement these new processes.

Throughout the literature, a standard methodology has been utilised to assess the feasibility of using polymeric compounds as a coke substitute in EAFs (Gupta et al.,

2006). After undergoing pyrolysis in a drop tube furnace (DTF), a process which replicates the injection of coke into the EAF, the remaining char of the tested samples are collected and placed in a horizontal drop tube furnace at 1200°C, which replicates the EAF operating conditions. Previous tests have been successfully conducted on blends of metallurgic cokes and different proportions of polypropylene and high-density polyethylene (Sahajwalla et al., 2010). These studies show that up to 30% plastic/coke blend can be used without adversely affecting the combustion efficiency of coals. Indeed, similar or slightly higher combustion efficiencies, faster gas generation and improvements in slag foaming volume and behaviour were obtained when compared to operation with standard coke (Ogaki et al., 2003). This increased performance was attributed to the amount of hydrogen available in the polymer materials (Sahajwalla et al., 2011). Given that the RO feed spacers are generally made of PP, this fraction of the old modules could potentially be used in EAF. While the chemical composition and structure of the waste plastics are known to influence the coke reactivity and subsequent combustion performance, the same tests would need to be applied to the other plastics comprising the modules in order to fully assess the feasibility of using some or all of the parts of the RO membrane in an EAF.

2.6. Chemical Conversion of RO Membranes

Given the nature of the composite membrane construction used in RO production, and especially the structure of the supporting layer, relatively simple conversion of the dense

RO into a porous material may be possible through the degradation of the PA layer

(Baker, 2004). Indeed, the combination of the PSf and PET support layers within RO membranes is markedly similar to UF materials (Ghosh and Hoek, 2009). The relative 51 vulnerability of the dense PA layer towards conventional oxidative agents could be used as a conversion method (Avlonitis et al., 1992). The resulting converted RO membrane could then be expected to feature hydraulic and removal performances that are comparable to commercially available UF products. Potential applications for the converted RO membranes include use in pre-treatment filtration in desalination plant, waste water treatment or for low cost water treatment in developing areas. It is expected that converted RO modules could remove all suspended solids, large organics species, and a significant fraction of pathogens from feed streams. This technique of oxidative chemical treatment to remove the PA active layer of the RO membrane resulting in an

UF membrane is referred to as membrane conversion. In terms of end-of-life options, membrane conversion has the benefit over direct reuse of being possible when the performance is already significantly below the original specifications and/or when there has been fouling on the membrane surface.

2.6.1. Exposure of RO membrane to Oxidative Chemicals.

The effect of oxidative chemicals on RO membranes has been studied extensively due to the detrimental effect that they have, and because they are used in desalination process to disinfect feed water before RO treatment. This is generally studied in the context of membrane ageing, investigating the effect of incidental and extended exposure to low concentrations of oxidative chemicals. Specifically, chlorinated oxidants have seen research focus due to their common use for water disinfection and fouling removal at various stages of the water treatment process (Gao et al., 2011; Porcelli and

Judd, 2010). In water reuse applications, monochloramine or sodium hypochlorite are dosed continuously to maintain residual chlorine levels. However, in desalination process, feed water is chlorinated to minimise biogrowth and then the feed is dechlorinated to minimise the effects of membrane exposure. Despite this, membranes are exposed to low concentrations of oxidants over the years of operation.

52 Once in contact with chlorinated oxidants, the active PA layer is degraded through a number of suggested reaction pathways, including N-chlorination of amine groups followed by aromatic ring chlorination (as seen in Figure 2-8). This can be summarised as the weakening of PA intermolecular hydrogen bond interactions with the incorporation of chlorine in a process called Orton Rearrangement (Cran et al., 2011; Kang et al.,

2007). The presence of ferrous iron (Fe(II)) has been shown to have a catalytic affect during this process (Gabelich et al., 2005). Additionally, it has been suggested that hydrolysis of the amidic group is a competing mechanism, as it changes the hydrophobicity of the surface (Do et al., 2012).

Figure 2-8: Nitrogen and ring chlorination of PA (Kang et al., 2007).

A number of studies have looked at the effect of hypochlorite on the membrane surface using a range of exposure intensities from 10 to 26000 ppm.h, and using doses of 10 to

4000 ppm of NaOCl (see (Donose et al., 2012) for an up-to-date review of available studies). Overall, it has been shown that when hypochlorite ions dominate the solution

(high pH), membranes exhibits increased water and salt permeability. However, when hypochlorous acid dominates the solution (low pH), there is a slight decrease in membrane permeability. While most studies do not use the extreme exposure intensities required to completely degrade the active layer, it is expected that as the PA chains

53 break down, the structure will weaken. As the PA degrades, the physical anchoring to the PSf support will become weaker, and eventually the remaining PA layer will come away from the support.

2.6.2. Previous Attempts at Membrane Conversion

Initial attempts to convert used RO into porous membranes have been conducted a decade ago in Spain (Rodriguez-Gonzalez et al., 2002). In this early work, sodium hypochlorite (NaOCl) and other strongly oxidative chemicals, including hydrogen peroxide (H2O2), sodium dodecyl sulphate and potassium permanganate (KMnO4), were tested under different operating conditions (active recirculation versus passive immersion) to remove the active layer from the membrane. From these initial tests,

KMnO4 was found to be the most effective agent to convert old RO membranes, with an optimal dose of approximately 1000 mg.L-1 for 1 to 2 h (Rodriguez-Gonzalez et al., 2002).

In addition, it was concluded that an active recirculation contact method was more successful than a passive immersion method. The use of the newly converted membranes within the tertiary treatment step of municipal wastewater was then assessed in a following study (Veza and Rodriguez-Gonzalez, 2003). The converted membranes demonstrated a potential to remove up to 96% of the suspended solids before further RO treatment. Although a high level of fouling was recorded during the filtration, the deposition was easily reversible by hourly flushing and alkaline chemical cleaning every 3 to 4 days. A nearly complete recovery of the converted membrane permeability was possible for the few weeks of the test (Veza and Rodriguez-Gonzalez,

2003).

A recent study assessed the viability of conversion using KMnO4 treatment, and displayed that it is possible to obtain stable performance resulting in high permeate flux at the expense of lower salt rejection (Ambrosi and Tessaro, 2013). A cleaning step with citric acid was investigated to address a decrease in permeability due to the formation of a manganese oxide layer, caused by the oxidative treatment. This study has shown that 54 even the most effective KMnO4 treatment resulted in only a 2 fold increase in permeability with salt rejection reduced to 85%. Moreover, it has not been clearly demonstrated that this techniques completely removes the active layer of the membrane and therefore presents limitations for RO conversion into UF.

A similar methodology has been used in a recent study to convert SWRO membranes with NaOCl into BWRO, and to briefly consider the formation of UF membranes through prolonged exposure (Raval et al., 2012). This examination of NaOCl treatment only utilised membranes manufactured in-house for the prolonged exposure tests, with limited discussion of the potential applications of the converted membranes from major manufacturers. Additionally, discussion of the effect of used membrane condition on subsequent converted performance has not yet be presented and converted RO membranes are still to be benchmarked against commercially available UF membranes.

2.6.3. Effect of Chlorine Exposure on Polysulfone Membranes

As oxidative chemicals are used to remove the active layer of RO membranes, the resulting PSf layer will also be exposed. Therefore, it is important to consider the effect that this oxidative chemical exposure has on the freshly exposed PSf surface.

In various water treatment processes, UF membranes are fouled with organic matter and are traditionally cleaned with hypochlorite, leading to numerous studies on the effect this exposure has on the membrane structure. One issue with PSf is its high natural hydrophobicity, which can lead to serious fouling by the adsorption or deposition of molecules on the membrane surface and pores. Commonly, during the manufacturing of

UF membranes, a hydrophilic polymeric additive is blended with the membrane polymer to create a hydrophilic membrane (such as polyvinyl chloride (PVP) or polyethylene glycol (PEG)) (Rahimpour and Madaeni, 2010). However, this is not included to the PSf layer during RO manufacturing, as the hydrophobicity is generally dictated by the PA active layer.

55 One study investigated the impact of hypochlorite exposure (up to 120,000 ppm.h) on polyethersulfone (PES) membranes (Arkhangelsky et al., 2007). It was shown that the chemical cleaning of the membranes with hypochlorite appeared to dislodge the PVP from the membrane matrix, causing an increase in flux and pore size; these results were determined using streaming potential, AFM and virus rejection. As a result, this led to an increase in fouling and significant flux decline during filtration. In addition to these performance decreases, the study also addressed the change in mechanical strength, and showed that possible chain scission of ethersulfone resulted in the deterioration in strength and an acceleration in membrane integrity loss (Kuzmenko et al., 2005; Levitsky et al., 2011).

A number of studies have corroborated these results, stating that hydrophilicity decreases with bleach contact due to PVP consumption (Gaudichetmaurin and

Thominette, 2006; Yadav and Morison, 2010), and decreased in tensile strength

(Causserand et al., 2006; Rouaix et al., 2006). One study notes that tensile strength was mainly governed by the backing layer of the membrane (Yadav et al., 2009), which was shown to be unaffected by hypochlorite treatment; when the backing layer was removed, tensile testing showed a considerable decrease in strength with ageing. This study also reported surface pitting and cracking after hypochlorite exposure, which increases the pore size causing increased clean water flux and a reduction in protein rejection. Lower pH NaClO exposure and higher concentrations had a greater negative effect on the surface and a documented increase in susceptibility to fouling was attributed to an increase in roughness and interaction between PES and the proteins filtered. The effect of oxidative chemicals on the PSf layer of the RO membranes was also studied using x- ray photoelection spectroscopy (XPS) and fourier transform infrared spectroscopy

(FTIR), and it was shown that there was an absence of chemical or elemental modification to the PSf layer (Ettori et al., 2011).

56 2.7. Decentralised Water Treatment Technologies

In order to properly assess the ability to reuse RO membranes, the actual applications need to be considered and reviewed. One of the main reuse options that has been considered is the conversion of used RO membranes to UF membranes by removing the active PA layer via oxidation. This could lead to the supply of low cost UF membranes whose performance may be consistent with current industry standards. Currently, one potential application for these membranes is their use in decentralised, small scale water treatment solutions for developing areas of the world. In order to determine the suitability of these membranes to this application, a review of the currently available membranes treatment options for this application has been completed. Additionally, these application could be considered for end-of-life UF membranes, however this is generally beyond the scope of this study.

One of the major factors to consider is the location of the improved water source and the time taken to obtain safe drinking water every day. In several African nations, more than a quarter of the population take longer than 30 minutes to make one water collection round trip, that will supply enough water to their family for the day, which significantly affects quality of life (WHO and UNICEF, 2010). The 2010 UN report states that focus needs to be shifted away from urban areas to the rural and decentralised communities, posing new challenges.

2.7.1. Decentralised Treatment Options

Large scale water treatment facilities and distribution systems can be quite cost prohibitive for a small developing community. Thus decentralised systems need to be low cost and have a small footprint. As these systems will be installed away from major centres, they also need to be simple to operate and maintain. Here is a list of currently available methods for water disinfection for decentralised communities (Peter-varbanets et al., 2009b).

57  Heat and ultraviolet (UV) light, including boiling with fuel, solar radiation and UV

lamps.

 Chemical Treatment methods, including, coagulation, flocculation, precipitations,

Adsorption, Ion exchange, and chemical disinfection.

 Physical removal processes including, sedimentation or settling, membranes,

and fibre filters, granular media filters, including sand filters, and aeration

When centralised systems are not available, the source of the water as well as the treatment method needs to be considered. Common feed water sources include groundwater wells and rainwater harvesting, and the quality of these sources can vary greatly. The most effective method of treatment will depend of the quality of feed water as well as the volumes required to be treated. While many of the listed methods are effective in removing the required pathogens and can be suitable for many applications, they also have some significant disadvantages. Heat and UV treatment techniques require a significant amount of time, produce small volumes and require excessive energy. Chemical treatment methods, by definition, require chemicals to treat the water, and the transport and storage of chemicals is generally not suitable for decentralised systems. Many physical removal processes require large footprints as well as expert knowledge to run and maintain the systems. While membrane systems share some of these disadvantages they also have the potential to be simple low cost solutions to water quality issues (Peter-Varbanets, 2010; Peter-varbanets et al., 2009b).

2.7.2. Membrane Treatment Options

In order to understand how a membrane process could be used effectively in decentralised drinking water treatment, current options and techniques need to be assessed. While there are a wide variety of options, they all have different intended applications, as well as a variety of currently installed capacities. This review will focus on widely implemented, small to medium scale systems with a focus on gravity filtration 58 with low maintenance requirements. Please refer to the review by Peter-varbanets et al, for a complete study of decentralised treatment systems (Peter-varbanets et al., 2009b).

A summary of the advantages and disadvantages for each system can be seen in Table

2-3.

2.7.2.1. Lifestraw

The Lifestraw system is a well-known and commercially available point-of-use UF systems. It consists of a UF module with a 20 nm pore size, inline pre-filter and a chlorine chamber. The system is contained in a small unit which is gravity fed from a small elevated container connected by a flexible hose, providing 0.1-0.2 bar of pressure. It is a simple system with basic manual backwash which can remove bacteria and viruses from up to 18,000 L of water in its operational life time, with a flow rate of 6-8 L.h-1

(Vestergaard Frandsen, 2011). This system presents many advantages, including its simplicity and small footprint, the only disadvantage is its extremely small scale which means that every family requires their own unit.

2.7.2.2. Homespring

The Homespring system by GE is one of the bestselling household units in developed countries for additional water treatment at the point of use; however, it is also rated to be used with various feed water qualities. The system uses a submerged hollow fibre UF configuration which runs off tap or hydrostatic pressure. It includes a activated carbon filter and requires annual maintenance (GE, 2011). The activated carbon filter and maintenance increases the cost of this system and its operation, making it less suitable for its use in developing areas. While technically it can be used as a decentralised water treatment solution, its design and marketing is aimed towards domestic household use.

2.7.2.3. Aquapot

Since 1996, AQUAPOT international projects have been using fully autonomous UF water treatment systems (Arnal et al., 2004) . The system uses a sand filter and 25-50 59 µm microfilter for pre-treatment. Following this pre-treatment, they have tested UF membranes in both a hollow fibre and spiral wound configuration, to produce permeate flow of 1000 L.h-1 with a pressure of 2.5-5 bar (Arnal et al., 2008). The pressure is provided by petrol driven pumps, however it has been stated that it is possible to run with hydrostatic pressure. After running the system for three years, the results show that UF technology is suitable for drinking water treatment in decentralised areas across a variety of feed water qualities. The membranes show a high rejection of total coliforms and thermotolerant coliforms and guarantee continuous drinking water supply (Arnal et al.,

2002). Part of the study included the use of 10, 30, 50 and 100 kDa MWCO membranes to determine which membranes remove all of the required pathogens. The membranes all showed significant removal of bacteria, however the 100kDa had the greatest flux and thus was selected for further testing (Arnal et al., 2004). Nearly 50% of the system costs can be attributed to the UF membrane system, including the membrane module cost.

On top of this membrane cost, it was also identified that another major disadvantage of this option is the local technical knowledge required to maintain and run the system, as well as the fuel source required.

2.7.2.4. SkyHydrant and Skybox

The SkyHydrant produced by SkyJuice is designed for community water supply in decentralised areas as well as disaster relief applications. The system uses a hollow fibre (microfiltration) MF membrane with a 0.04 µm pore size and chlorine disinfection. It can be operated in a number of configurations, including the use of 0.3 bar of hydrostatic pressure. The system is modular, meaning that multiple units can be combined to increase the output as required, with each module producing up to 700 L.h-1 (SkyJuice

Foundation, 2014a). The membrane is cleaned with manual back flushing and physical agitation as well are regular hypochlorite chemical cleaning, requiring a semi-skilled operator. The PVDF MF membrane module has a cost between US$1250 and $1400 which represents 50% of the manufacturing cost of the system (Butler, 2011). Due to the

60 relatively large pore size of the MF membrane, the chlorine disinfection is required to kill the remaining pathogens in the water. Another disadvantage of this system includes the relatively high cost, coupled with the maintenance required for operation, however this product has been widely implemented and has proven to be extremely successful.

Another product developed by Skyjuice is the SkyBox, which uses the same type of hollow fibre MF membrane as the SkyHydrant, but has significantly less membrane area.

The membrane is mounted at the bottom of a plastic box, which the feed water is poured into by the user. The water is then filtered using hydrostatic pressure. The SkyBox has many advantages over the SkyHydrant including lower cost and simpler operation; however, it also produces significantly less water at 25 L.h-1 (SkyJuice Foundation,

2014b).

2.7.2.5. PAUL

The Portable Aqua Unit for Life Saving (PAUL) is a portable gravity fed membrane filtration system designed for use in emergency and disaster situations. The UF membranes in the system have a pore size between 20 and 100 nm (150 kDa MWCO), and is driven with 1.15 m of hydrostatic pressure (Frechen, 2014). Like the Skybox, the user simply pours feed water into the top of the unit, and potable water will then flow out of the permeate hose. The system produces up to 1200 L.d-1 of water, and can be carried using the included backpack straps if required. One issue that has been reported is that the units eventually become non-operational due to the membrane drying out when not used (Frechen, 2014).

61 2.7.2.6. EAWAG

The research institute EAWAG has been recently working on gravity driven membrane disinfection (GDMD) using 100 kDa PES UF membranes, using pressures of 0.04 - 0.15 bar. The system has a dead end filtration configuration without any pre-treatment and minimum maintenance. After challenging a number of UF and MF membranes with various bacteria and MS2 phage, it was concluded that UF membranes with a maximum pore size of 20nm was the most suitable, with a 150 kDa membrane showing a 4 log removal of MS2 phage (Peter-varbanets et al., 2009a). It was observed that after one week of operation the membrane flux stabilised around 4 - 10 l.m-1.h-2, depending on what feed water was used. After this stabilisation, constant flux was seen for several months of operation. It was determined that this flux stabilisation occurred due to the formation of cavities and channels that where observed in the fouling layer (Peter-

Varbanets et al., 2010). This process occurs in the biofouling layer, which led to an increase of flux in equilibrium with the deposition of retained materials which decreased the flux. This phenomenon was observed for a wide variety of feed waters including a variety of natural organic matter, all feed waters where shown to have the same stabilisation effect (Peter-Varbanets et al., 2011). Due to this flux stabilisation, dead end filtration was deemed suitable for decentralised systems, with no back flushing and cleaning required for months of continuous operation. These findings and the new method of operation enable extremely low cost, simple systems to be created and if a recycled UF membrane can be applied to this method, the cost will be further reduced.

While the flux values are low compared to conventional systems, having cheap membrane available will allow greater surface areas to be used allowing a high volume of water to be treated for minimum cost and maintenance.

62 Table 2-3: Summary of available decentralised membrane systems.

System Membrane Used Advantages Disadvantages Simple operation, Small scale, chemical free, Lifestraw 20 nm UF multiple units gravity driven, needed small footprint Expensive, short operational High quality Homespring Hollow fibre UF lifespan, requires product regular maintenance Expensive, spiral wound UF complicated Aquapot High productivity 100 kDa system, high operational costs Expensive, some High Productivity, technical Hollow fibre MF SkyHydrant Gravity driven, knowledge 40 nm manual cleaning required, chemicals required Hollow fibre MF Gravity driven, SkyBox Low productivity 40 nm simple operation. Higher cost than Gravity driven, PAUL 150 kDa other gravity simple operation. options Gravity driven, Flat sheet UF Eawag GDMD simple operation, Low productivity 100 kDa minimal cleaning.

2.7.3. Criteria for Membrane Use in Developing Rural Areas

After assessing the membrane systems that are currently available and the previous work conducted on decentralised operations, an outline of the minimum requirements can be created. It has been widely concluded that UF membranes with a pore size of 20

- 40 nm or around 100kDa MWCO is sufficient to have significant removal of viruses, bacteria and organic matter. UF also has many advantages in comparison with conventional disinfecting processes, including production of water of invariable quality, and smaller quantity of required chemical additives. Gravity filtration is generally favoured because the cost of pump equipment and running costs make up the largest proportion of the total system cost. After pumping costs, the second most expensive part

63 of a traditional system is the membrane modules themselves. Therefore, recycled membranes can further reduce the cost of the process, essentially creating a water treatment system constructed out of entirely recycled materials. Low cost and low maintenance is a critical requirement for the application to decentralised areas, where there are no trained personal, and at the very least manual back flushing without chemicals is desired. It is also a requirement that the system is able to deal with feed water of various quality, while still producing safe drinking water. Finally, it is desirable that the system features a modular design so that it can be adapted to different sized applications, as well as having a small footprint.

An assessment of the current options available to treat a variety of feed waters to produce drinking water in decentralised and rural areas has shown that there is an application for low cost UF membranes. Provided that the recycled RO membranes, that will theoretically show UF properties, can be shown to operate effectively under these conditions they will be suitable for creating ultra-low cost gravity driven membrane systems for decentralised applications.

2.8. Membrane Storage and Drying

For RO applications, the membrane elements are contained within a pressure vessel and are constantly used, or maintained in an aqueous environment. However, in certain situations, the membranes may be removed from this environment for storage, transportation or equipment maintenance. The primary focus for storage protocols is to prevent biological growth, to which used membranes are extremely susceptible. The standard protocol for this type of membrane storage is through the use of a buffered 1 wt% solution of food-grade sodium metabisulfite (SMBS), at a pH >3 (DOW, 2012). The solution is circulated through the membranes, which are then individually packaged in oxygen-barrier plastic bags and vacuum sealed. If the plant is temporarily shut down, this preserving technique can also be used in situ; in this case, the solution is circulated

64 through the membrane pressure vessels and it monitored for pH and concentration over the storage duration (Holgate, 2014).

2.8.1. Membrane Pore Collapse

If these preservation methods are not undertaken or hydration is not maintained, the membranes will dry out, resulting in significantly reduced permeate flow. However, the rejection characteristics are expected to remain similar, do to minimal changes in the PA layer (CSM, 2006). This permeability loss cannot be recovered through normal operation.

This type of performance change is shared by all RO and tight PES/PSf UF membranes, regardless of construction material (Subrahmanyan, 2003).

The cause of this extreme decrease in permeability has been attributed to the collapse of the pore structure of the PES/PSf membrane layer, which involves the shrinkage and volume reduction of the pores (Subrahmanyan, 2003). As the water in the membrane pores evaporates, the pores collapse and form a denser structure (Pinnau and Freeman,

2000). This effect can be described by the Young-Laplace relationship:

2훾 ∆푃 = cos 휃 Equation 2-1 푟

Where P is the force pulling the pores shut (Pa), γ is the surface tension (N.m-1), r is the pore radius (m), and θ is the contact angle (°) between the liquid and the membrane material. The capillary pressure is directly proportional to the surface tension of the wetting liquid, and inversely proportional to the pore radius. If the swollen membrane modulus is lower than the exerted forces of the drying pores, then the pores will collapse and form a denser structure (Beerlage, 1994). As the membrane pores get smaller, the forces increases, and thus the pressure required to rewet the membrane pores increases

(Noble and Stern, 1995). This effect is closely related to the bubble point of the membrane, and pore collapse is effectively increasing the bubble point (Perry and Green,

65 2008). As asymmetric membrane skin layer pores are significantly smaller than pores in the rest of the membrane, pore collapse is believed to be most predominant in the surface region (Subrahmanyan, 2003). As TFC RO membranes are, in their simplest form, an asymmetric UF membrane coated with a PA layer, the drying mechanisms are reported to be similar (DOW, 2012).

2.8.2. Previous Work Investigating Membrane Drying

Very few studies have directly investigated the effect of membrane drying or rewetting methods. One study used the systematic drying of PA UF membranes to determine its structural properties and briefly investigate the effect of surfactants on rewetting (Staude and Passlack, 1986). The study used a custom balance suspended inside a desiccator to selectively dry the membranes, and the effect of rewetting was measured by the mass of water uptake. The results showed that submerging the dried membranes in a surfactant could increase the water uptake of the dried membranes, and that the uptake could be further increased by applying pressure. The study showed that the permeability decrease caused by drying could not be fully recovered by these rewetting techniques.

This early investigation is the first indication that the smallest pores near the surface of the membrane are being closed due to the effect of drying. Two additional separate studies briefly investigated drying mechanisms for in-house manufactured polyimide and

PSf asymmetric UF membranes (Beerlage, 1994; Subrahmanyan, 2003). Both studies concluded that pore collapse played a significant role in hydraulic performance loss; however, they failed to properly quantify the changes through the methods employed.

Challenges related to drying has been recognised since the beginning of membrane development and there is evidence of early attempts to produce membranes that are impervious to the phenomenon. These include a dry manufacturing technique for integrally-skinned asymmetric cellulose acetate RO membranes that results in a larger void (pore) size in the support layer. This larger pore size means that the capillary forces are not sufficient to collapse the pores, resulting in a membrane that can be wetted and 66 dried repetitively (Ann and W, 1972; Kesting, 1975). Additionally, a number of methods have been developed to dry the membrane while retaining its original performance

(Macdonald and Pan, 1974). However, these processes were primarily developed for cellulose acetate membranes, and required considerable foresight to prepare the membranes for proper storage.

2.8.3. Manufacturer Recommended Rewetting Strategies

Membrane manufacturers are also aware of the negative effects of element drying, and note that it results in irreversible flux loss due to pore structure change in the PSf layer

(DOW, 2012). Therefore, they emphasise that all membranes should be kept wet during transportation, storage and shutdowns (CSM, 2006; Toray Industries, 2005). Some of the manufactures also have recommendations for rewetting dry, or partially dry membranes (CSM, 2006; DOW, 2012). These methods include, soaking the membrane in 50/50% w/w ethanol/water or propanol/water solutions for 15 min to 2 h; pressurising the element at 10 bar with a closed permeate port for 30 min; or soaking the element in

1% HCl or 4% HNO3 for 1-100 h. However, there has yet to be a published systematic study of the effectiveness of these techniques.

It is clear that maintaining membrane hydration throughout transport or maintenance is a priority, otherwise irreversible performance loss is expected to occur. If membranes are not properly preserved during long term plant shutdown, or during transportation to a secondary location during membrane reuse, then the membranes will require disposal.

Therefore, it is critical to identify the factors leading to pore collapse, and optimise a process for rewetting dried membranes. From this, membranes that have dried have the potential to be reused, thus increasing their lifecycle and decreasing the overall environmental impact.

67 2.9. Life Cycle Assessment

Increasing awareness of the environmental impact of products and processes has led to the development of environmental management tools to better understand and manage these effects. LCA is a systematic tool for assessing potential environmental consequences and has been increasingly applied to the water (Lassaux et al., 2007;

Ribera et al., 2013; Vince et al., 2008), wastewater (Lassaux et al., 2007; Pasqualino et al., 2009) and membrane industries (Raluy et al., 2006; Tangsubkul et al., 2006; Zhou et al., 2011a).

2.9.1. Waste Management LCA

LCA can be used to evaluate the impact of a product or service on the ecosystem and human health by compiling and interpreting the relevant inputs and outputs. The assessment can include all parts of the products lifecycle, including the extraction of raw materials through to the end-of-life disposal. The associated environmental impacts of the inputs and outputs are used to interpret the processes inventory and impacts in relation to the objectives of the study. The general categories of environmental impacts commonly considered in LCA studies include resource use, human health, and ecological consequences (Zuin et al., 2013).

An LCA performed on waste management options generally compares a conventional disposal process, such as landfilling, where the mixed waste stream is treated, with alternatives, whereby the waste is separated into different fractions and treated with the aim of reducing environmental impact through recycling or reuse options (Ekvall et al.,

2007; Heijungs et al., 2011). The compared treatment options may range from cleaning a used product to make it fit for reuse, to recycling of materials or energy recovery. An important consideration for this type of study is that any environmental benefit derived from recycling must overcome the increased environmental impact from transportation and processing. Generally, direct comparison of the impact from the various reuse and

68 recycling methods is not sufficient, as some of the options produce useful products, such as heat, electricity or products which create and open recycling loop. Waste management LCAs deal with this through system expansion, or by taking into account the offset of additional production of the new product. For example, if one used RO membrane was cleaned and reused, one less membrane needs to be produced.

2.9.2. LCA Studies on Water Treatment and Membrane Technology

A number of LCA studies have been conducted on various water treatment processes including desalination with RO (Vince et al., 2008). The majority of these studies focus on the operation phase of the process, including chemical and energy requirements, as they have been shown to have an overwhelming majority of the contribution to environmental impact (Tarnacki et al., 2011; Zhou et al., 2011b). However, a number of studies have briefly explored the impact of RO membrane production and renewal, showing that it generally contributes less than 5% to the overall impact of the process. A small number of studies include RO manufacturing, however detailed information of the construction process is not provided (Hancock et al., 2012; Raluy et al., 2005a). Two separate studies calculated that the emissions from membrane transportation, production and replacement account for less than 1% of the total CO2-e emissions from the life time of a desalination plant (Biswas, 2009; Raluy et al., 2005b). A study on a MF process reported that membranes have a low contribution to the overall process, with only 0-11% of CO2-e contribution coming from manufacturing and up to 5% and 90% from process chemical and energy use respectively (Tangsubkul et al., 2006). However, these studies have not assessed the impact of membrane manufacturing in detail, or explored possible end-of-life disposal options for used membranes.

69 2.10. Conclusions

The development of alternative end-of-life avenues for a complex product such as an

RO membrane is an involved process. This has been seen before, with other industries requiring a concerted effort by all stakeholders to make it viable. The first major step of this process is to identify and assess all possible end-of-life options for the product for technical feasibility, financial viability and environmental sustainability.

For RO membrane reuse, the effect of membrane drying during storage and transportation has been identified as a significant challenge and knowledge gap requiring further investigation. This will be investigated in Chapter 3. Additionally, the concept of membrane conversion through chemical treatment requires significant refinement before it can be considered a feasible option, and therefore will be further studied in Chapter 4.

Finally, a number of applications, including the use of membrane material as a coke substitute in EAF steelmaking, need to be tested for the specific materials found in RO membranes. The results of this study are presented in Appendix B.

Once the end-of-life options have been identified and tested, they need to be compared based on a number of different parameters. As the aim of this study is to increase the environmental sustainability of the RO water treatment industry, the primary focus is on comparing the impact of each end-of-life option. To this end, the tool of LCA will be utilised in Chapter 5 to systematically investigate the impact of membrane waste disposal and compare the alternative methods. Finally, when all the methods are tested and assessed for environmental impact, this information can be combined with qualitative values like ease of implementation and expected financial viability into a decision making tool. This tool, which is described in Chapter 6, will adapt to the decision maker’s individual situation and inform them of the optimum avenue for their end-of-life membranes.

70 Chapter 3:

Assessment of Membrane Drying

and Subsequent Rewetting

Techniques

Assessment of Membrane Drying and Subsequent Rewetting Techniques

71 3.1. Introduction

An extensive literature review in Chapter 2 has determined that direct membrane reuse could be one of the simplest and most beneficial end-of-life options for used RO membranes. However, the performance of reused membranes can be directly affected by membrane transportation and storage. If the membranes are not preserved correctly, premature drying results, leading to significantly reduced permeate flow. This effect is also noted with low MWCO UF membranes of PES and PSf construction (Perry and

Green, 2008). The aim of this chapter is to firstly, quantify the performance change of

RO and UF membranes due to unwanted drying, and secondly to assess and optimise various rewetting strategies.

Not only is the structural changes due to drying a critical consideration for membrane reuse, it also has implications for research and development. When imaging techniques such as scanning electron microscopy (SEM), atomic force microscopy (AFM) and contact angle are used, sample preparation often includes drying. The subsequent change in structure, performance and surface interaction means that the sample does not accurately represent the industrial membrane (Abdullah et al., 2014; Patterson et al.,

2009). High temperature surface modification methods, such as plasma treatment, have also been found to permanently alter membrane structure (Razmjou et al., 2011;

Subrahmanyan, 2003).

The current study tests the primary hypothesis that performance degradation is caused by pore collapse in the PES/PSf layers of UF and RO membranes, due to capillary forces generated during solvent evaporation (as discussed in Section 2.8). The UF membranes used in this study are of PES construction, which are more commonly used in industrial applications than PSf, but have similar pore structure and operating parameters.

Additionally, PES membranes are readily available in varying MWCO, allowing for the assessment of pore size on performance loss due to drying.

72 3.2. Methodology

3.2.1. Membranes

The characteristics of the membranes used in this study are summarised in Table 3-1.

The information includes the pure water permeability (PWP) as tested using the equipment described in Section 0. All RO membranes were PA thin-film composite construction with a PSf support layer, while the UF membranes have a PES active layer.

RO membranes were purchased in 8” modules, and after the modules were cut open, the membranes were extracted, sealed in oxygen barrier plastic bags and refrigerated.

UF membranes were purchased in sheets and were stored in the same method as the

RO membranes. For testing, membrane coupons with a diameter of 5.9 cm were cut with an aluminium template. These coupons are compatible with both membrane testing systems used, meaning that all tests used samples of the same size.

Table 3-1: Key properties of studied membranes

Label Membrane Type PWP (L.m-2.h-1.bar-1)

DOW Filmtec extra low XLE BWRO 7.51 ± 0.81 energy BW30 DOW Filmtec BW30 BWRO 3.74 ± 0.28

Dow Filmtec high SW30 SWRO 0.84 ± 0.20 rejection SW30 10 kDa flat sheet PAL 10 kDa UF UF 210 ± 34 Omega PES UF 30 kDa flat sheet PAL 30 kDa UF UF 430 ± 30 Omega PES UF 100 kDa flat sheet PAL 100 kDa UF UF 1520 ± 115 Omega PES UF

73 3.2.2. Membrane Performance Characterisation

A flow diagram describing the testing sequence can be seen in Figure 3-1. Performance testing for RO membranes was completed in a Sterlitech HP4750 high pressure dead end cell using a nitrogen gas cylinder to supply pressure. All membranes were initially compacted at 15 bar for RO and 1.5 bar for UF with Milli-Q™ water for up to 3 h to ensure complete wetting, prior to any drying or further testing. RO membrane PWP tests were performed with Milli-Q™ at 15 bar in the same dead end cell. Salt rejection tests were conducted under the same conditions but with 2000 ppm NaCl as the feed. Performance testing for UF membranes was completed at 1.5 bar using a custom built plastic dead end cell with an active area of 17 cm2, connected to a 3L feed tank. Pressure was applied using a centralised nitrogen supply line.

Controls

Compaction Controlled Virgin Rewetting Permeability at 15 bar for drying in membrane technique testing 3 hours desiccator

Dry sample testing

Figure 3-1: Flow diagram of membrane testing sequence.

3.2.3. Membrane Drying

This study required the precise measurement of the degree of membrane drying that has occurred. Typically, a desiccator is used when complete drying of a sample is required; however, timing mass loss using this technique is unreliable. To enhance the accuracy and reproducibility, an analytical balance (KERN AES 220-4) was used and adapted.

The balance was made airtight, and silica gel beads were placed inside the compartment, thus replicating a desiccator in function. This enabled the controlled drying

74 of the membrane samples. After membrane compaction, surface water was removed to decrease variation in drying time and weight measurements by briefly compressing the samples between two KIMTECH KimwipesTM. The error of this technique was calculated by completing cycles of weighing the sample, then soaking the membrane for 10 min in

Milli-Q™, and then repeating the process. Overall this technique was extremely consistent, with a relative standard error of 0.45%. To increase consistency, all membranes were dried with the active layer facedown, therefore the slight curve in the membranes resulted in contact with the balance only at the edges.

The degree of membrane drying (measured in %) is calculated by Equation 3-1.

푀푤 − 푀푓 % 퐷푟푦 = 푥 100 Equation 3-1 푀푤 − 푀푑푎푣푔

Where Mw is the wet mass of the membranes, Mf is the final mass, and Mdavg is the average dry mass of the sample type. This value represents the amount of solvent evaporated from the membranes.

3.2.4. Membrane Rewetting

A number of membrane rewetting techniques were tested during this study. These include various rewetting solutions and exposure times, as seen in Table 3-2.

Membranes were submerged in the rewetting solution in a glass petri dish for the required time. After the predetermined rewetting exposure time, membranes were rinsed and soaked in Milli-QTM water for at least 1 h prior to testing. A minimum of 3 separate membrane samples repeats were tested for each set of experiments. Thus for all graphs below, the data point represents the average of the pooled data across repeats, and the error bars represent the standard deviation across the pooled data.

75 Table 3-2: Description of rewetting techniques.

Solution surface Rewetting Exposure tension (mN.m-1) Concentration solution time @ 20°C (Vhquez et al., 1995) Water Milli-QTM 1-100 h 72.8

Ethanol 1 - 50% w/w 15 min 28.5

Isopropanol 50% w/w 15 min 24.8

Hydrochloric 1% w/v 1-100 h 73.0 acid (HCl) Nitric acid - 4% w/v 1-100 h (HNO3) Sodium Lauryl - 0.25 g.L-1 SLS 1-100 h Sulfate (SLS)

3.2.5. Scanning Electron Microscopy

Membrane samples were imaged using an FEI Nova NanoSEM 230 SEM. Samples were dried and coated in chromium using an Emitech K575x sputter coater. Imaging was done at a variety of magnifications using an ETD lens with a spot size of 3.0 at 5 kV, and a working distance of 5 mm.

3.2.6. Thermo Gravimetric Analysis

TGA was completed using a TA Instruments Q5000 machine. Wetted samples were cut and weighed on an analytical balance before being loaded into a platinum high temperature pad. The sample purge was air flowing at 25 ml.min-1, while the balance purge was Nitrogen flowing at 15 ml.min-1. The furnace was bought to the test temperature at a rate of 25°C.min-1 and held for 20 min.

76 3.2.7. Atomic Force Microscopy

Surface morphology and roughness of membranes was assessed using AFM. Samples were scanned in air on a Bruker MultiMode 8 in ScanAsystTM (tapping) mode using

ScanAsyst-air probes (nominal tip radius of 2 nm; nominal spring constant of 0.4 N.m-1).

Three samples of each membrane type were imaged at three randomly chosen locations using a 400 nm2 scan size. All scans were taken with 512 points per line with 512 scan lines at a scan rate of 1.6 µm.s-1. This scan size was selected to maximise the resolution of the small pores on 10 kDa UF membranes. Samples roughness was assessed using the root mean square roughness (Rq) as calculated by Gwyddion SPM analysis software

(version 2.35). Surface topography images were created using Nanoscope Analysis software (Bruker, version 1.4). Average pore size was calculated by measuring a large number of pores from three representative images using Photoshop CS6 measurement recorder tool.

3.2.8. Rejection Characterisation

To assess the effect of membrane drying of rejection performance, a solution of 10 mg.L-

1 humic acid (HA) and 10 mg.L-1 bovine serum albumin (BSA) (Sigma Aldrich, Australia) was passed through samples and analysed using duel column liquid chromatography- organic carbon detection (LC-OCD), system model 8, based on the Gräntzel thin film reactor developed by DOC Labor, Germany (Huber et al., 2010). Chromatographs were interpreted using DOC Labor ChromCALC 2013 software, and rejection was calculated using Equation 3-2 for each major component.

푃푒푟푚푒푎푡푒 푐표푛푐푒푛푡푟푎푡푖표푛 Equation 3-2 푅푒푗푒푐푡푖표푛 (%) = (1 − ) 푥 100 퐹푒푒푑 푐표푛푐푒푛푡푟푎푡푖표푛

77

Figure 3-2: Example chromatograph of Humic and BSA feed solution.

Figure 3-2 displays an example chromatograph of the feed solution, showing its major components. The prominent biopolymers (PB) peak includes substances with a molecular weight above 10 kDa, including the proteins contained in the BSA. Humic substances (HS) includes material roughly between 10 and 1 kDa, while materials less than 1 kDa in size make up the building blocks (BB), low molecular weight acids and neutrals (Huber et al., 2010). The total dissolved organic carbon (Total DOC) is measured by the column bypass within the LC-OCD system.

3.2.9. Small Angle X-ray Scattering

3.2.9.1. Sample Preparation and SAXS Operation

The small angle x ray scattering (SAXS) beam-line at the Australian Synchrotron was used with a 0.6 and 7 m camera lengths to investigate the scattering patterns of the UF and RO membranes corresponding to a 0.015 to 0.09 Å−1 scattering vector, q range. The end-station uses a 1 M Pilatus detector for that provides excellent dynamic range, single photon per pixel sensitivity, low noise and fast time resolution (30 and 150 frames per second, respectively). An in-vacuum undulator source was used (22 mm period, 3 m

78 length maximum, Kmax 1.56), and energy beam values were maintained constant during all the measurements at 11 keV (resolution of 10−4 from a cryo-cooled Si(1 1 1) double crystal monochromators). The beam size at the sample was 250 horizontal ×150 vertical

μm2. A AgBe reference was used for calibration of the beam (Australian Synchrotron,

2014).

Membrane samples were prepared as described in Section 0 and stored in Milli-QTM water for transportation and storage. To reduce interference, the PET support layer of the membranes was removed by carefully pealing it off the PES/PSf layer. The samples were heated within the SAXS chamber with a Linkam Scientific HFSX350 heating stage fitted with a THMS600 heating block. Tests were performed at 25, 40 and 60°C; and the heating rate was fixed at 50°C.min-1. A thermocouple was placed on the heating stage and the accuracy of the temperature measurement was estimated to ± 0.1 °C. Due to the long acquisition time required for the 25°C test for 10 kDa UF membrane, the sample was removed from the beamline after 25°C was reached (after ~10 min), was maintained at 25°C for 100 min, and then retested. Scattering patterns were acquired upon heating and at the plateau temperature for after stabilization of the patterns for at least 1 min.

3.2.9.2. SAXS Analysis

First, background normalizations were performed with an air shot across the Linkam stage including Kapton windows with Scatterbrain v2.08. Scattering patterns were plotted with normalized intensities (range of pixel colours between 2 and 1500) and the weak peaks were expanded by consistently adjusting contrast between the images

(range of pixel colours between 2 to 615).

The analysis of the dynamic evaporation tests was performed in Origin Pro 8.0. Local inflexion points, corresponding to Guinier knees were identified and used for modelling pore size distribution and dynamics of pore collapse as a function of time and heating temperature.

79 The Guinier knee radius, RG, of the different knees were calculated following previous work (Dumée et al., 2012) and Equation 3-3. The q values of the knees were determined by locally fitting the peaks with Gaussian distributions.

(퐼푞/퐼0) 푅퐺 = √ Equation 3-3 1 2 (3 푞 )

With Iq the intensity of the knee at q, and I0 the intensity of the reference spectra corresponding to the sample shot prior to starting the heating experiment.

At long evaporation times, and upon complete disappearance of the knees, the I=f(q) curves were fit with power laws to confirm the presence of a Porod region upon homogenisation of the structure. Linear regressions in log domains were calculated to confirm the presence of a Porod region, typically corresponding to a power law in q of -

4. The surface area of the pores S was calculated from Equation 3-4.

−4 퐼푞 = 푆푞 Equation 3-4

The pore size distributions were then calculated with Irena, installed into Igor Pro 6.31 and by using the Unified Level Model within the Modelling II option. The scatterer distributions were assumed to by spheroidal and scatterers to be dispersed into a dilute system (Grigoriew and Chmielewki, 1998). Gaussian distributions were used to evaluate the distributions and the mean diameter of the scatterers was recorded and compared over time and across the series of heating temperatures. The average scatterer size and volume were calculated.

80 3.3. Results and Discussion

3.3.1. Membrane Drying Behaviour

3.3.1.1. Impact of Temperature and Membrane Type on Drying Rate

The first step was to assess the rate of membrane drying, and determine how much mass a wet membrane could lose during the drying process. Small samples of wet compacted membranes were analysed using TGA at 25°C in a nitrogen atmosphere.

Figure 3-3 shows the change in the normalised rate of mass loss with time (Δmg.min-

1.°C-1) during the experiment.

x10-3

18

) 1

- SW30

C

°

. 1 - 15 BW30 XLE

mg.min 12 10 kDa UF Δ 30 kDa UF 9 100 kDa UF

6

3

0 Normalised mass loss ( rateNormalised 0 5 10 15

Drying time (min)

Figure 3-3: Membrane drying behaviour in TGA at 25°C in nitrogen atmosphere.

The time at which the mass loss rate reaches zero, indicates a point where all moisture has evaporated from the sample and the membrane is dry. The UF membranes took longer (13 - 15 min) to completely dry when compared to the RO samples (8 – 11 min).

81 This is due to the thicker PET and PES layers, which contain a larger amount of water per membrane area, when compared to the RO. Interestingly, as the MWCO of the UF membranes increases, so does the drying time. It is likely that the evaporation of the water trapped in the large voids of the PET layer takes place first during the consistent mass loss rate period. The water that is bound to the polymer and in the membrane pores evaporates later, as it requires a higher amount of energy (Stokke et al., 2013).

This effect has been studied and characterised in detail for in-house manufactured UF membranes (Subrahmanyan, 2003).

x10-3

25 50°C

20 35°C

) 1

- 25°C

C

°

. 1

- 15

10

mg.min

Δ (

5 Normalised mass loss rateNormalised

0 0 5 10 Drying time (min)

Figure 3-4: XLE membrane drying behaviour in TGA at various temperatures.

To investigate how temperature affects the evaporation of water from the membranes, samples of XLE were dried using TGA at 25, 35 and 50°C, as seen in Figure 3-4. The results show that the peak evaporation rate and drying time is linearly proportional to the

TGA temperature. The dry nitrogen atmosphere in the TGA further accelerates evaporation, thus it is expected that drying in humid environments or in a desiccator would be considerably slower.

82 3.3.1.2. Assessment of Performance Loss Due to Drying

Initial testing indicated that all six membranes in this study showed drastic reduction in

PWP after overnight desiccation, up to 100% flux loss in most cases. Two membranes,

XLE to represent RO and 10 kDa for UF, were selected for further assessment of gradual drying. Compacted/wet samples were dried to varying degrees in the balance-desiccator apparatus, as described in Section 3.2.3. Figure 3-5 shows the mass remaining (from starting compacted/wet mass) over time in the balance-desiccator. Figure 3-5 shows the average of over 20 samples per membrane type which were selectively dried and tested for PWP (Figure 3-6 and Figure 3-7), with the error bars showing one standard deviation.

100

90 XLE 10 kDa

80

70

60 Percent of wet (%) massofremainingwet Percent 50 0 20 40 60 80 Drying time (min)

Figure 3-5: XLE and 10 kDa membrane mass loss from drying in balance-desiccator

apparatus at room temperature.

Due to the thinner PSf and PET layers of the RO membrane, it loses less of its total wet mass when dried, compared to the UF membrane. The XLE membrane is completely dry after 75 min, while the thicker UF membrane which takes up to 85 min. These results

83 are similar to the TGA in terms of relative drying time. However, due to the static atmosphere, drying takes significantly longer.

While the main metric for this study is the change in hydraulic performance due to drying, the physical integrity of the membrane also needs to be considered. All six membrane types physically curled towards the active layer when dried, even if they initially curled the other way (while virgin before initial wetting). This indicates that the PES and PA/PSf layer of the membrane has contracted more than the PET support layer. This curling effect is substantial, with a relatively flat sample completely curling up so that opposite edges of the sample touched. The majority of this curling effect happens towards the end of the drying process, within the last 5 min before there is no further mass loss.

Interestingly, this is also the period of lowest mass loss rate, i.e. lowest evaporation rate.

This further supports the idea that the bulk of the PES or PA/PSf drying occurs after the

PET layer has dried. This curling, which has been previously reported (Kesting, 1973), results in an increase in stress between the PSf/PES and PET support layers (Siddique et al., 2014), leading to layer separation, which can increase the chance of membrane delamination in the case of sudden backpressure or change in trans-membrane pressure

(TMP). An example of this separation can be seen in Figure A-4 of Appendix A, and the effect of this separation on rewetting water uptake is further discussed in Section 3.3.5.2.

To test the impact of drying on PWP, XLE and 10 kDa UF were subjected to varying degrees of drying (as determined by Equation 3-1), in the balance-desiccator. Following this treatment, they were tested for PWP in the respective high pressure RO and low pressure UF apparatuses. The results of these experiments can be seen in Figure 3-6 and Figure 3-7, for XLE and 10 kDa UF respectively. Three samples were tested for each level of drying, and the standard deviation is shown as error bars on the graph.

84

8

)

1 -

6

.bar

1

-

.h 2 - 4

2 PWP (L.m PWP

0 0% 20% 40% 60% 80% 100% Degree of drying

Figure 3-6: Performance of XLE membrane after varying levels of drying. Tested at 15

bar in high pressure apparatus.

250

) 1 - 200

.bar Surface water removed

1 -

.h 150

2 -

100 PWP (L.m PWP 50

0 0% 20% 40% 60% 80% 100% Degree of drying

Figure 3-7: Performance of 10 kDa UF membrane after varying levels of drying. Tested

at 1.5 bar in low pressure apparatus.

85 The results for both membranes follow a similar pattern, with some notable exceptions.

The XLE samples showed no significant deviation in hydraulic performance throughout the drying process until a level of 85% dry was reached. At this point, the PWP drastically decreased with increased drying until the minimum PWP was obtained at 100% dry.

Figure 3-7 shows a significant drop of PWP at the start of the drying process for the 10 kDa UF membranes. This occurs when the surface water is removed with the technique described in Section 3.2.3. At this point, the membrane has spent no time in the desiccator, but still has consistent performance loss due to the surface water removal with Kimwipes. Further tests showed that if surface water is only removed from the PET support side, no change in PWP occurs; however, if surface water is removed only from the active side, the reported performance drop occurs. This indicates that when surface water is removed using this technique, moisture is also drawn from the membrane structure, which is in turn reducing permeability. Interestingly, the RO membrane is not susceptible to the same phenomenon. If performance degradation from drying occurs with the same mechanism for both membrane types, the dense nonporous PA layer may be protecting the PSf layer of the RO membrane during this phase. Using the rewetting methods described in Section 3.3.5, this permeability lost is completely recovered in all cases.

Following this initial drop in PWP, the effect of drying for the 10 kDa UF membrane follows a similar trend to the XLE samples. Performance is consistent until drying reaches 90%, at which point the PWP drops rapidly to its minimum value. At 100% dry, all membranes show extremely low PWP. This rapid performance decline is in line with the observed curling effect described earlier and occurs during the period of lowest evaporation rate.

86 The other performance metric that needs to be assessed is membrane rejection. It has been previously reported that RO membrane salt rejection does not change during drying

(CSM, 2006; DOW, 2012). After PWP tests, samples was tested for salt rejection using the method described in Section 0. The results showed no statistical variation in rejection due drying, indicating no change to the PA layer. Additionally, membranes that had not been subjected to the initial wetting stage (i.e. dry as purchased) were also desiccated and then tested for PWP, however no difference in permeability or rejection was observed when compared to the control samples.

Table 3-3 reports the mass change of all six membranes studied at different stages of testing. All membranes, except SW30 have a statistically significant mass loss over the wetting-drying cycle. This leads to the conclusion that some material is flushed away during the pure water compaction phase. This material could be glycerol, preserving agent, or unreacted monomer. This is confirmed by analysing PWP permeate in a total organic carbon (TOC) detector, which results in increased organic carbon detection when compared to pure water. This amount of organic carbon is significantly reduced after 3 h of flushing, indicating that it has been removed from the membrane structure.

Table 3-3 also shows the hydraulic performance of the membranes after complete drying in the desiccator. All of the membranes, apart from 100 kDa, showed almost complete

PWP loss after 100% drying. Regardless of filtration time, no measureable permeability was observed for SW30 and 10 kDa UF membranes. However, for the membranes that did exhibit reliable dried permeability (XLE and 30 kDa), flow was only observed after

20-60 min of filtration. The comparison between PWP loss for 10, 30 and 100 kDa membranes indicates that pore size is a determining factor.

87 Table 3-3: Mass and performance change during membrane wetting and drying in the balance-desiccator at room temperature.

Percent of wet PWP after Membrane Virgin dry sample Wet sample Dried Sample Mass loss from mass remaining 100% drying type mass (mg) mass (mg) Mass (mg) virgin to dry (mg) after drying (%) (L.m-2.h-1.bar-1) XLE 213.1 ± 2.1 310.8 ± 4.9 208.3 ± 1.8 4.8 ± 1.2 67.0 0.11 ± 0.09

BW30 210.2 ± 0.2 311.1 ± 2.3 205.1 ± 1.6 5.1 ± 1.6 65.9 < 0.1

SW30 199.8 ± 0.3 288.0 ± 0.6 199.8 ± 1.3 0 ± 1.3 69.4 0

10 kDa UF 241.3 ± 5.0 352.5 ± 1.5 194.6 ± 7.7 46.6 ± 2.9 55.2 0

30 kDa UF 245.5 ± 5.9 384.1 ± 7.4 195.2 ± 12.4 50.3 ± 13.7 50.8 0.16 ± 0.22

48.1 100 kDa UF 238.5 ± 3.4 405.2 ± 2.1 195.0 ± 7.3 43.5 ± 8.1 757 ± 15

88 Overall, it can be seen that both RO and UF membranes can lose a significant amount of their wet mass before any performance degradation occurs. Visually, a 90% dry membrane appears and feels dry. While this does mean that unprotected membranes can be stored for longer than expected, the dramatic drop in performance makes predicting the impact of uncontrolled drying difficult. With membrane drying rate, the impact of performance of drying, and associated mass loss have been investigated in detail, the next step is to characterise the impact of drying of the membrane layers.

3.3.2. Effect of Drying on Membrane Layers

The PES membranes, and the support layers of RO membranes assessed in this study are asymmetric, meaning that the pore structure on the active side will be tighter than on the backside of the membrane The backside and active layer side of the membrane are assessable using SEM and AFM imaging techniques; however, to probe the inner structure of the membranes, SAXS is required. To assess the change due to drying of the membrane backside, RO and UF samples were delaminate using a process involving the removal of the PET base layer to expose the backside of the PSf/PES layer. SEM images of the PSf layer backside of both virgin and dried XLE RO membranes are shown in Figure 3-8. Further information and images using this technique can be found in

Appendix A.

89

Figure 3-8: SEM images of the backside of the PSf layer of delaminated XLE RO

membranes. Left: Virgin, Right: Dried.

The SEM images clearly show pore structures with a sizes between 0.2 and 1.9 µm. Due to the asymmetric nature of these membrane layers, these pore are significantly larger than the expected active layer pore size. No significant change in pore size or geometry can be seen between the virgin and dried samples, and this was observed for all membrane types tested. Therefore, it can be concluded that the large pore size of the membrane backside are unaffected by drying, and therefore will play no role in the decrease in hydraulic performance.

The expected active layer pore size of the PES and PSf layers of the studied membranes is between 2.5 and 10 nm (Pall Corporation, 2012), which is beyond the resolving power of available SEM equipment. Therefore, to assess the effect of membrane drying on the active layer, the surfaces were imaged using AFM. Only UF membranes were used, as the PA layer hides the PSf surface pore structure of RO samples. Figure 3-9 shows representative surface scans for the different membrane types, and Figure 3-10 displays the root mean squared roughness values.

90

Figure 3-9: AFM height map surface scans (active skin layer side) of 10 kDa UF

membrane. Left: Virgin, Right: 100% Dry.

6

Virgin Dry

4 Rq Rq (nm) 2

0 10 kDa UF 30 kDa UF

Figure 3-10: Change root mean square (Rq) roughness for 10 and 30 kDa Virgin and

100% Dry UF membranes.

A significant surface morphology change can be seen in the surface AFM scans of 10 kDa UF membranes shown in Figure 3-9, which was consistent across all samples.

Firstly, distinct pore structures can be seen in the virgin sample, with an average pore size of 12.1 ± 6.5 nm. The surface structure of the dry sample is distinctly different, not

91 only with less defined features and pores, but significantly larger variation is topography.

This change can be attributed to the contraction of the layer during drying, resulting in a wavy surface structure. The Rq roughness of virgin samples was calculated to be 1.2 ±

0.3 and 2.8 ± 1.4 nm for 10 and 30 kDa UF membranes respectively. These values are in line with previously reported measurements using similar dry cell AFM techniques and similar membranes (ElHadidy et al., 2013; Rahimpour and Madaeni, 2010). The 10 kDa membranes showed dried roughness of 1.8 ± 0.4 nm, resulting in a statistically significant increase. For 30 kDa samples, large surface nodules (as seen in Figure A-3 Appendix

A) dominated the quantitative roughness measurements; however, a visual morphology change, similar to the 10 kDa samples, was seen.

3.3.3. Assessment of Pore Structure Change Using SAXS Analysis

To further define the mechanisms responsible for the observed performance decline, analysis of the internal membrane structure was completed using SAXS, while being dynamically dried at varying temperatures.

The scattering patterns (Figure A-5 to A-7 Appendix A) of 10 kDa UF membrane samples show a main scattering ring, visible around q ~ [0.06; 0.1], composed of a series of broad and low intensity Guinier knees. These scattering rings were found consistently across the series of samples. The intensity of the scattering patterns were radially integrated and reduced (Figure 3-11). At fixed evaporation temperature, the curves are found to shift towards higher intensities over time and the knees to progressively smoothen until complete disappearance. The knees were observed to smoothen faster with increasing evaporation temperature, indicating that evaporation energy plays a key role in the disappearance of the knees.

As discussed in Section 3.2.9, the 25°C sample was removed from the beamline after the first 10 min and maintained at 25°C until being reinserted and tested at 111 min. As

92 a result, complete knee reduction was not seen. This may be due to the change in beamline location on the sample, the variable humid atmosphere of the testing chamber, or the test not being conducted long enough. However, the overall trend was consistent with other samples.

-1 Using this data, the gyration radius (RG) for the two major Guinier knees (0.06393 Å and 0.09493 Å-1) were calculated for the three series of samples as a function of time, as seen in Figure 3-12.

93 10000 Time (min) 111 1000 10.5 Increasing time 7 4.66 100 1.16 0

10

Scattering Intensity Scattering Intensity (a.u.) 25°C 1 0.1 -1 0.2 0.05 q (Å ) 0.3

10000 Time (min) 16.33 1000 10.5 Increasing time 7 3.5 100 0

10

Scattering Intensity Scattering Intensity (a.u.) 40°C 1 0.1 -1 0.2 0.05 q (Å ) 0.3

10000 Time (min) 16.33 1000 Increasing time 8.17 4.1 3.5 100 2.95 2.33

10

Scattering Intensity Scattering Intensity (a.u.) 60°C 1 0.1 -1 0.2 0.05 q (Å ) 0.3

Figure 3-11: Reduced data Guinier plots for the series of evaporation tests of 10 kDa

UF membranes at 25, 40 and 60°C over time, using a 0.6 m camera length.

94 30

25

20

G 15 R 10 0.06393 Å-1 5 0.09493 Å-1 25°C 0 0 50 100 150 Time (min)

60

50

40

G 30 R 20

-1 10 0.06393 Å 40°C 0.09493 Å-1 0 0 5 10 15 20 Time (min)

30

25

20

G 15 R 10 0.06393 Å-1 5 0.09493 Å-1 60°C 0 0 5 10 15 20 Time (min)

Figure 3-12: Change in RG of 10 kDa UF membranes with evaporation time and

temperature, using 0.6 m camera length.

95 The radius of gyration (RG) is a direct representation of the morphology of a polydisperse system and may be translated into the effective size of the scattering feature. This feature may be a macro-molecule, part of a semi-crystalline domain across a macro-molecule, a micelle in suspension, a pore or a single domain in a multiphase system (Dumée et al.,

2012). In the present case, RG leads to an estimate of overall size of the membrane pores. The RG for the two main knees, visible in Figure 3-12, increased for every series prior to plateauing at a fixed value, suggesting that the system reached a steady state independently of the evaporation conditions (i.e. Sample was completely dry). This trend directly correlates with the kinetic of evaporation; i.e. the plateau was reached quicker at higher temperatures.

Upon reaching the plateau, no knee can be visibly seen on the scattering patterns, which are found to then follow a Porod law, suggesting a shift from an isotropic to an evenly distributed matrix (Dumée et al., 2012). The initial knees can be attributed to the presence of the sub-100 nm pores across the structure of the PES, which are reducing in size during the evaporation of water from the membrane structure. In addition, the surface area of the scattering material is found to increase as a function of water evaporation rate from ~0.033 at 25oC and to plateau at 60oC around 0.035 (+/- 0.01)Å-2.

This suggests that although the reduction in scattering from both volume and size is tangible, slight variations of final geometries might be achieved depending on the water evaporation rate.

To quantitatively assess the effect of drying on pore geometry, the size distributions of the knees was evaluated (using methods described in Section 3.2.9.2), and the average pore size is plotted as a function of time for the series of evaporation tests (Figure 3-13).

96 12

10

8

6 25°C 4

Mean pore size size pore(nm)Mean 40°C 2 60°C 0 0 20 40 60 80 100 120 Time (min)

Figure 3-13: Mean pore size of 10 kDa UF membranes as a function of evaporation

time and temperature, using 0.6 m camera length.

The results show a consistent decrease in pore size with increasing drying time, and an increase in the rate of change with temperature. Again, the value measured for 25°C after 111 min is clearly not in line with expected results, due to the impact of being removed and reinserted into the beamline. However, a trend comparable to 40 and 60°C samples is observed. A previous study using cellulose UF membranes reported that there was distinct features (similar to the Guinier knees reported here) that differed between wet and dry samples (Grigoriew and Chmielewki, 1998). The authors hypothesised that these were due to water filled nano scale pores, which were not present in dry samples; however, no direct investigation into the effect of drying was conducted. Nevertheless, the similarity in results supports the model for change in pore structure due to drying.

Results obtained when imaging with longer (i.e. 7m) camera length and using RO membranes show similar results, with no new knees visible, and a similar shift to higher intensities and knee smoothening. Unfortunately, due to calibration issues with this long

97 camera length and limitations in experimental time, the results are incomplete, and are thus not discussed at length here. However, camera length mainly impacts the size of the features detectable, with 0.6 m sizes ranging from 0.7 – 40 nm and 7m between 7 –

400 nm. Therefore, the shorter camera length is optimum for imaging small membrane pores.

It should be noted that while the membranes were tested using a transmission technique, and therefore all pore structures within the asymmetric PES/PSf layers were within the beam-line, structures which are bigger than the spatial resolution limit of the setup scatter at inaccessibly small angles and thus are not detected. Therefore, data represents only the skin and sub-skin layers (down to a few µm) of the top UF surface (Dumée et al.,

2013).

Overall, in the context of the assumptions made in the modelling methodology, the results indicate a significant change in the internal membrane structure during the evaporation of water, with high consistency between sample series. This, along with AFM results and the behaviour of samples during selective drying, provides evidence for the theory of pore collapse caused by capillary forces (as discussed in section 2.8), which are driven by the evaporation of water from within the pores of the PES layers. The force at the air-liquid interface is large enough that the walls of the capillary sized pores are pulled together and the structure collapses on to itself, restricting flow. There is also an indication that the final pore structure has a dependence of the evaporation rate, and that this collapse occurs only at, or close to, the membrane skin layer. Preliminary SAXS results and the similarities in drying behaviour between UF and RO membranes, indicates that the same effect occurs in the PSf layer of RO membranes during drying, and is the dominant factor in the reduction of hydraulic permeability.

98 3.3.4. Effect of Drying on Membrane Rejection

A hypothesised effect of capillary pore collapse in UF membranes is that rejection will increase as a function of drying, due to a decrease in mean pore size. To assess this effect, a solution of proteins and humic substances was filtered through virgin and dry

100 kDa UF samples, and analysed by LC-OCD (Figure 3-14). The 100 kDa UF membrane was selected for these tests as it is the only membrane with a low rejection of the test solution, therefore a significant change can be detected. Additionally, as shown in Table 3-3, drying causes a change in permeability from 1520 to 760 L.m-2.h-

1.bar-1, while the other UF membranes experienced complete permeability loss, making filtration tests challenging.

100 Virgin 100 kDa UF

80 Dry 100 kDa UF

60

40 Rejection(%)

20

0 Total DOC Biopolymers Humic substances Building blocks

Figure 3-14: Organic carbon rejection of virgin and 100% dry 100 kDa UF membranes,

assessed with LC-OCD.

The results show an expected general trend of decreasing rejection with decreasing component size, with the highest rejection of the large biopolymers. When compared to the virgin control samples, the dry membranes show a statistically significant increase in rejection of total DOC and across all three size categories. This phenomenon can be explained by the effect of partial pore collapse and has been previously seen in PES 99 membranes during heat treatment (Gholami et al., 2003; Razmjou et al., 2011). As drying occurs, and the pores collapse due to capillary pressure, the mean pore size decreases, effectively increasing the MWCO of the membrane.

3.3.5. Assessment and Application of Rewetting Strategies

3.3.5.1. Initial Assessment of Rewetting Strategies

A number of techniques designed to recover performance loss due to drying were identified and selected for testing in this study. Further information about these techniques is summarized can be found in Table 3-2 and in Section 2.8.3 of the literature review.

The rewetting technique most commonly proposed by manufacturers and by previous research involves soaking the membrane in alcohol solutions. To test this method, solutions of various alcohols from 1 – 50% w/w were prepared and membranes were exposed up to 2 h. These initial assessments where conducted with XLE membranes, as the high initial PWP allows for more accurate differentiation between resulting permeabilities. To initially validate the methodology of rewetting solution exposure, membranes were tested after being completely submerged in a petri dish (allowing contact with both sides of the membrane) and solution recirculation in a membrane cell

(restricting contact to the active layer). No statistically significant difference in recovered

PWP was observed; as a result, petri dish submersion was selected to enable greater control of exposure and thus consistency between samples. The effect of ethanol concentration and time was investigated in detail, the results of which can be seen in

Figure 3-15.

100

3

)

1 - 1%

.bar 2

1 -

.h 10%

2 - 25%

1 50% PWP (L.m PWP

0 0.25 1 2 Rewetting time (h)

Figure 3-15: Effect of ethanol solution concentration (1, 10, 25 and 50% w/w) and

exposure time on 100% dry XLE membranes.

The results show that the highest PWP recovery was obtained using 50% w/w ethanol, with no significant effect of exposure time observed over the tested range. It should be noted that the virgin PWP of the XLE membranes is 7.5 ± 0.8, leading to the most effective ethanol strategy resulting in only 34% permeability recovery. This increase in wetting ability with concentration is in line with previous work showing that the lower surface tension of the fluid increases its ability to rewet the membrane pores (Mulder,

2003). This can be expected as a direct result of Equation 2-1, where a lower surface tension overcomes the forces required for pore collapse. It can be concluded from these results that 15 min is a reasonable time for alcohol exposure to obtain optimum results.

Further experiments were conducted to explore the impact of wetting fluid surface tension, particularly through the use of different alcohols, and the degree of membrane dryness on rewetting. As discussed in Section 3.3.1.2, XLE membranes that are 95% dry show a significantly reduced PWP of 2.9 ± 0.6 L.m-2.h-1.bar-1, while samples that are dried to 100% have almost no measureable permeability. Figure 3-16 shows this effect, along with the resulting PWP after exposure to ethanol and isopropanol at a concentration of 50% for 15 min. 101 10 No rewetting

8

) 1

- Ethanol

.bar 1

- 6 Isopropanol

.h

2 -

4

PWP (L.m PWP 2

0 Virgin Control 95% 100%

Degree of drying

Figure 3-16: Effect of alcohol solution (50% w/w for 15 min) rewetting 95 and 100% dry

XLE membranes.

The results show that samples which are not completely dry have a higher PWP recovery from alcohol exposure. Additionally, at 95% dry, ethanol and isopropanol have the same rewetting effect; while at 100% dry, ethanol has a superior impact with a recovery of 2.6

L.m-2.h-1.bar-1, compared to the 1.6 L.m-2.h-1.bar-1 from isopropanol. From these results, it can be concluded that as degree of drying increases, flux recovery from rewetting methods is expected to decrease.

The final stage in the preliminary investigation into rewetting strategies is the assessment of long term exposure techniques proposed by membrane manufacturers, including

HNO3 and HCl, as discussed in Section 3.2.4 . These solutions also include long term exposure to 1% w/w ethanol and sodium lauryl sulfate (SLS). SLS is a common alkaline solution used for membrane cleaning of inorganic fouling (Ang and Elimelech, 2009) and was selected for testing due to its availability, low cost, and low surface tension (DOW,

2012; Greenlee et al., 2009).

102 8 Virgin membrane performance 7

HCl

) 1 - 6

HNO3

.bar 1

- 5 .h

2 Ethanol 1% - 4 SLS

3 50% w/w ethanol soak for 15 min PWP (L.m PWP

2

1

0 1 50 100 Rewetting time (h)

Figure 3-17: Effect of prolonged exposure rewetting methods on 100% dry XLE

membranes.

The results in Figure 3-17 show that none of the tested long term exposure methods performed better than the recovered PWP of 2.6 L.m-2.h-1.bar-1 from 50% w/w ethanol exposure. Notably, wetting by HCl and HNO3, which are recommended by manufacturers, showed only minor improvement over completely dry samples. However, the commonly used surfactant SLS showed a permeability recovery of 1.5 L.m-2.h-1.bar-

1 after 50 h of exposure, with no improvement after an additional 50 h. Long term soaking in Milli-QTM water was also assessed, but no duration of exposure (up to one week) showed any improvement.

103 3.3.5.2. Application of Rewetting Strategies

In the previous section, it was demonstrated that soaking dried membranes in ethanol

(50% w/w for 15 min) was the optimum short exposure duration alcohol treatment, and that soaking in SLS for 50 h was a promising long duration method. To further assess these methods, they were applied to six commercially available RO and UF membranes which had been previously compacted, tested for PWP and then completely dried in a desiccator. As seen in Figure 3-18, the permeability and swelling volume was measured for each membrane in its virgin state, after drying, and after rewetting with both ethanol and SLS. Swelling volume is defined as the volume of water uptake within the membrane structure per cm2 of membrane area, and was measured gravimetrically as described in

Section 3.2.3. This swelling volume, which is also known as macroscopic swelling volume, gives an indication of the state of the pores and water uptake affinity (Izak et al.,

2007).

All the membrane types tested had significant performance decline after drying, with all but the 100 kDa samples showing almost zero PWP. Due to the significantly larger pores of the 100 kDa UF membrane, according to Equation 2-1, the forces generated during the drying process are not sufficient to overcome the modulus of the polymer. The 10 and 30 kDa UF membranes were also assessed for bubble point, with results of 9-10 and 3-4 bar respectively.

During the drying process, the closed membrane pores absorbed less water, leading to a uniform reduction in swelling volume across the entire membrane. It should be noted that an extended swelling study was completed on each dry membrane type, with no significant increase in water uptake being observed up to 72 h.

104

Figure 3-18: Swelling volume and permeability of various membranes at various stages; Virgin, and after rewetting (following 100% drying in desiccator) with water (1 h

soak), ethanol (50% w/w for 15 min), and with SLS (50 h).

Out of the tested membranes, only SW30 showed complete flux recovery from ethanol soaking after drying, as seen in Figure 3-19. In fact, SW30 membranes showed 115 ±

9% flux recovery after ethanol rewetting, a phenomenon that has been previously

105 reported and is attributed to increased swelling in the PA layer (Louie et al., 2011). It is possible that this occurs in all the RO membranes tested here, however its minor effects are potentially masked by the overall reduction in hydraulic permeability due to pore collapse in the PSf layer. The 10 and 30 kDa UF membranes showed an extremely low permeability recovery of between 1-5% for both methods. All other membranes show significant PWP recovery after both ethanol and SLS rewetting techniques have been applied. It should also be noted that NaCl rejection was tested for all three RO membranes after ethanol and SLS rewetting, and no statistically significant difference from virgin performance was observed.

120 15 min 50% ethanol 100 50 hr SLS

80

60

40 PWP Recovery (%) Recovery PWP 20

0 SW30 BW30 XLE UF 10kDa UF 30kDa UF 100kDa

Figure 3-19: Performances recovery of dried membranes after soaking in 50% ethanol

for 15 min and SLS for 50 h.

Interestingly, there was no correlation between recovered swelling volume and recovered PWP. In fact, there was an observed trend of increasing swelling volume after ethanol and SLS rewetting. This is likely caused by either increased swelling in the bulk polymer of the membranes, or due to a slight delamination of the PES/PSf and PET layers throughout the drying process, thus increasing the available uptake volume. As the membrane dries, the PES/PSf layer contracts more than that of the PET layer,

106 leading to a curling effect (as previously discussed in Section 3.3.1.2). This curling adds extra strain on the membrane layers and physically delaminates them from the PET support. Therefore, while the usable pore volume shrinks slightly, the overall water uptake volume is observed to increase.

A previous study investigating the effect of various surface coating methods on SWRO membranes also briefly reported the effect of drying and rewetting in ethanol (Louie et al., 2011). The study reported a significant increase in permeability of virgin RO membranes when soaking them in ethanol, and shows a significant decrease upon drying in an oven at 60°C overnight. It was concluded that this decrease in flux is due to increased interchain hydrogen bonding in the PA layer during water evaporation. While this may play a role in membrane flux decline, evidence from the current study, and previous literature, indicates that the primary cause is due to pore collapse in the PSf layer.

While the current study did not directly investigate the effect of relative humidity on drying rate and resulting performance, the effect of air drying was briefly assessed. XLE RO membrane samples were dried in uncontrolled laboratory air for 3 days. When dried, the membranes showed a PWP of 0.14 ± 0.17 L.m-2.h-1.bar-1, and had a PWP of 2.46 ± 0.13

L.m-2.h-1.bar-1 after ethanol rewetting. These results are similar to the desiccated samples, indicating that pore collapse still occurs in atmospheric conditions.

To further assess practical applications of membrane rewetting, the developed ethanol rewetting technique was applied to a used XLE-2520 membrane module, which had been stored in variable atmospheric conditions for more than five years. The module was initially tested in a RO testing apparatus with RO permeate quality water as feed at 6 bar for over 2 h; during this time no flux was observed. The module was then submerged in

50% w/w ethanol for 15 min before being flushed with pure water and again tested for

PWP. After rewetting, the module showed a PWP of 1.01 L.m-2.h-1.bar-1 and a NaCl salt rejection of 94.8%. While lower than the previous results reported in this study, this 107 module had an unknown history and visible inorganic fouling was observed during autopsy. These results show that this rewetting of used and completely dry RO membranes does partially recover lost flux, despite the membrane being in a spiral wound module format.

3.3.6. Change in Membrane Layer Resistance

The initial hypothesis that the reduction in membrane performance is due to changes in the PES/PSf membrane layers is supported by the similarities in drying behaviour, results from AFM and SAXS analysis, and previous literature (Section 2.8). While the RO membranes have significantly higher PWP recovery using rewetting techniques, when compared to UF, their virgin PWP is several orders of magnitude lower as well. To more accurately contrast UF and RO performance variation during drying and wetting, change in membrane layer resistance can be assessed. The relationship between membrane flux and resistance to flow is given by Equation 3-5.

푇푀푃 Equation 3-5 퐹푙푢푥 = 휇푅

Where µ is the viscosity of the feed water, TMP is the trans-membrane pressure and R is the total resistance. When pure water is being filtered, and there is no fouling layer to consider, R is simply the total membrane resistance Rm. This Rm value can be calculated as the sum of the resistances from the individual membrane layers, as per Equation 3-6 for RO membranes.

푅푅푂 = 푅푃퐴 + 푅푃푆푓 + 푅푃퐸푇 Equation 3-6

Where RPA is the resistance from the PA layer, RPSf from the PSf layer and RPET is from the PET support layer. For UF membranes, only the PES and PET layers are to be considered.

108 As the resistance of the PSf layer in RO membranes cannot be measured without dramatically changing the membrane structure, the resistance of the studied UF membranes can be used as a surrogate. By substituting the change in resistance of a

UF membrane, a theoretical dried membrane permeability can be calculated and compared to the measured values.

To test the hypothesis of PSf pore collapse, it is assumed that there is no resistance change in the PA layer. Additionally, the PET layers delaminated from UF and RO membranes were tested for PWP, and they had a resistance two orders of magnitude lower than the UF membranes, and did not change with drying; therefore, RPET is negligible. Finally, resistance values for membranes rewetted for 50 h in SLS are used, as the PWP of 100% dry samples are too low to be accurately measured.

The first step is to calculate the resistance of the PA layer of the RO membrane, by subtracting the virgin total UF resistance (RUF) from the total RO resistance. Using this method, RUF for UF is equal to RPSf for RO and therefore, RPA = RRO – RUF. From this, the theoretical dried RO membrane resistance can be calculated using the dry 10 UF resistances and Equation 3-6. With these calculated resistances, using the 10 kDa UF membrane as a PSf layer surrogate, can be seen in Figure 3-20, compared to the virgin and experimentally measured values. The error bars show one standard deviation, which is calculated for the theoretical values from the variance in measured layer permeability.

109 x1014

6 )

1 PA layer resistance - 5 PSf layer resistance 4

3

2

1 Membrane Resistance (m ResistanceMembrane 0 V M T V M T V M T SW30 BW30 XLE

Figure 3-20: Comparison of (V) measured virgin, (M) measured post rewetting, and (T)

calculated theoretical values for RO membrane resistance.

The calculated theoretical results from this layer substitution technique closely follow the measured values, further justifying that the same change is occurring in the PSf layer of the RO membranes as in the PES layer of the UF membranes.

For a virgin RO membrane, the PA layer has over two orders of magnitude higher resistance than the PSf layer (as seen on virgin samples in Figure 3-20), meaning that it dominates the hydraulic performance. When the RO membrane is dry, the combination of the dense PA and the collapsed pores of the PSf, results in extremely low PWP (The resistance of a dried membrane with zero PWP is effectively infinite). When a rewetting technique is applied, the relatively minor recovery in the PSf layer permeability means the PA layer resistances dominates again, leading to significant PWP recovery. This is most pronounced in the SW30 samples, where the PA layer resistance is highest. The

XLE samples, which have the lowest PA layer resistance, show the lowest relative permeability recovery, as seen in Figure 3-19.

110 These results supports the hypothesis that both the RO and UF membranes are being effected by the same mechanism. While RO membranes have decent permeability recovery using ethanol and SLS rewetting techniques, the technique only has a minor effect on the PSf layer. This means that while the SW30 membranes have almost complete permeability recovery, there is still significant pore collapse present in the PSf layer.

3.4. Conclusion

The work presented in this chapter investigated the impact of drying on the performance of commercially available RO and UF membranes, and assessed various rewetting strategies. The results provide significant evidence to support the hypothesis of major structural changes in the PSf/PES layer due to drying. Further discussion of the key outcomes will be made in Chapter 7.

The outcomes of this study has implications for sample preparation, membrane storage, direct membrane reuse, conversion to UF and for temporary plant shutdowns. Therefore, identifying the mechanisms of drying and importance of proper storage is the first step for assessing end-of-life options. As a result, it has been identified that proper storage protocols should be included in membrane reuse scenarios when modelled in the LCA in Chapter 5.

111 Chapter 4:

Production and Characterisation of

UF Membranes by Chemical

Conversion of Used RO Membranes

This chapter is an expanded version of the following peer-reviewed journal article:

Lawler, W., Antony, A., Cran, M., Duke, M., Leslie, G., Le-Clech, P., 2013.

Production and Characterisation of UF Membranes by Chemical Conversion of

Used RO Membranes. Journal of Membrane Science, 447, pp.203–211.

And the extended abstract:

Lawler, W., Wijaya, T., Antony, A., Leslie, G., Le-Clech, P., 2011. Reuse of

Reverse Osmosis Desalination Membranes. IDA World Congress. Perth.

Production and Characterisation of UF Membranes by Chemical Conversion of Used RO Membranes

112 4.1. Introduction

The previous chapter investigated critical considerations for facilitating direct membrane reuse, which, despite being one of the simplest reuse options, requires high quality used membranes. Another possible option is to reuse RO membranes as low-cost UF membranes after the dense PA layer has been chemically removed. As identified in the literature review (Section 2.6), this concept has been previously investigated, but requires further practical assessment before it can be considered as a feasible option.

The aim of this chapter is to build on previously published work and systematically investigate opportunities for RO membrane reuse by; (a) determining the optimum protocol for converting RO membranes, including the effect of storage conditions; (b) examining the effect of active layer removal on flux, rejection and surface morphology on a range of virgin and industrially used membranes; (c) benchmarking the efficacy and fouling of converted RO membranes against UF membranes; and (d) investigating potential applications and discussing limitations of the use of the converted RO membranes.

4.2. Materials and Methods

4.2.1. Membranes

The characteristics of the membranes used in this study, including UF, SWRO, and

BWRO are summarised in Table 4-1. All RO membranes tested were PA thin-film composite construction with a PSf support layer, and have a manufacturer stated chlorine tolerance rating of less than 1,000 ppm.h or a continuous free chlorine exposure of less than 0.1 mg.L-1. Initial PWP was determined using the methods described in

Section 4.2.2.

113 Table 4-1: Condition and key properties of studied membranes.

Initial PWP Label Membrane Type Condition (L.m-2.h-1.bar-1) Virgin long term BW30 DOW BW30 BWRO 3.43 ± 0.50 uncontrolled storage BW30FR DOW BW30FR BWRO Virgin 3.73 ± 0.29

XLE DOW XLE BWRO Virgin 7.51 ± 0.81

SW30 DOW SW30-HR SWRO Virgin 0.84 ± 0.20

Koch1 Koch 8822HR BWRO Used/wet 2.93 ± 0.41

CSM CSM RE8040-FE BWRO Used/moist 2.42 ± 0.63

Hydranautics Hydra BWRO Used/moist 1.01 ± 0.62 CPA5-LD

Toray Toray TML820 SWRO Used/moist 0.39 ± 0.19

Koch2 Koch TFC-SW SWRO Virgin 1.11 ± 0.30 10 kDa flat sheet 10 kDa UF PAL Omega PES UF Virgin 210 ± 34 UF 30 kDa flat sheet 30 kDa UF PAL Omega PES UF Virgin 430 ± 30 UF 100 kDa Flat sheet 100 kDa UF PAL Omega PES UF Virgin 1520 ± 115 UF

4.2.2. Membrane Performance Characterisation

Performance testing for RO membranes was completed using a Sterlitech HP4760 high pressure dead end cell with a magnetic stirrer, using a nitrogen gas cylinder to supply pressure. Unless otherwise stated, RO membranes were initially compacted at 10-15 bar with Milli-QTM water for 1-2 h, prior to conversion or further testing. RO membrane PWP tests were performed with Milli-Q™ at 10-15 bar in the same dead end cell. Performance testing for UF and converted RO membranes was completed at using a custom built acrylic dead end cell with an active area of 17 cm2 with a magnetic stirrer, connected to a 3L feed tank. Pressure of 1-2 bar was applied using a centralised nitrogen supply line.

114 Membranes samples were cut from either autopsied modules leafs or flat sheets, depending on source, and were stored at 4°C in Milli-QTM water or dry, depending on the processing stage.

4.2.3. Degrading Agents

For this study, the conversion method was as follows: membrane samples were cut to size from sheets of spiral wound elements and pre-conversion wetting was applied.

Following the wetting step, samples were exposed to the degrading solution, which was protected from UV light to minimise deterioration. The degrading solutions tested were sodium hydroxide (NaOH), potassium permanganate (KMnO4) and sodium hypochlorite

(NaOCl), technical grade from Ajax Finechem Pty Ltd. The concentration and exposure time of the degrading agent was varied to achieve exposure intensities ranging between

28,000 and 500,000 ppm.h. The concentration of NaOCl was determined using a Cary

100 Bio UV-Visible spectrophotometer at a wavelength of 292.5 nm using a molar extinction coefficient of 360 M-1.cm-1 (Hussain et al., 1970), and the pH of the solution was maintained at 12.0 ± 0.2. Following conversion, samples were thoroughly rinsed and then stored in Milli-Q™ water at 4°C prior to characterisation.

4.2.4. Membrane Characterisation by FTIR

Membrane samples were characterized by FTIR spectroscopy using a Shimadzu

IRAffinity-1 FTIR spectrophotometer. The membranes were analysed by attenuated total reflectance (ATR) using a Pike Technologies VeeMAXTM II variable angle specular reflectance accessory using a 45 degree ZnSe ATR crystal. The spectra were recorded at a resolution of 4.0 cm-1 in the range 4000-600 cm-1 with an average of 56 scans per membrane.

4.2.5. Rejection Characterisation

Salt rejection characterisation was conducted using 2000 ppm of NaCl (Sigma Aldrich,

Australia) at 15 bar for RO and 1-2 bar for converted RO and UF in the aforementioned 115 dead end cells. For advanced rejection characterisation, a solution of 10 mg.L-1 HA and

10 mg.L-1 BSA (Sigma Aldrich, Australia) was passed through membrane samples and analysed using duel column liquid chromatography-organic carbon detection (LC-OCD), system model 8, based on the Gräntzel thin film reactor developed by DOC Labor,

Germany (Huber et al., 2010). Chromatographs are interpreted using DOC Labor

ChromCALC 2013 software. Dextrans of various molecular weights were obtained from

Pharmacosmos and prepared in 2 g.L-1 solutions in Milli-Q™ water. Filtration was conducted in a stirred dead end cell at 2 bar and permeate quality was analysed using

TOC detection, Shimadzu TOC-VCSH.

4.2.6. Silver Nanoparticle Challenge Testing

Pathogen removal was characterised using citrate stabilised silver nanoparticles. The solution was prepared by dissolving 9 mg of silver nitrate in 50 mL of water and bringing to boil. To the boiling solution, 1 mL of 1% sodium citrate solution was added dropwise under vigorous stirring. After complete addition, the mixture was further boiled for 1 h and allowed to cool at room temperature. The mixture was centrifuged at 25000 rpm for

30 min to separate the nanoparticles from the free silver ions. The nanoparticles were then redispersed in water by sonication and stored for testing. All the reagents used to prepare the nanoparticles were of analytical grade and Milli-Q™ water was used throughout experiments. The particles were filtered through the membrane samples using the setup described in Section 4.2.2, feed and permeate samples were collected after 15 and 30 min of operation and analysed for Ag content by Perkin Elmer optima

7300 inductively coupled plasma (ICP) optical emission spectrometer. Further information about the preparation process and validation of these nanoparticles can be found in (Antony et al., 2013).

116 4.2.7. Atomic Force Microscopy

Surface morphology and roughness of membranes was assessed using AFM. Samples were scanned in air on a Bruker MultiMode 8 in ScanAsystTM (tapping) mode using

ScanAsyst-air probes (nominal tip radius of 2 nm; nominal spring constant of 0.4 N.m-1).

Three samples of each membrane type were assessed by conducting three 400 nm2 scans at randomly chosen locations. All scans were taken with 512 points per line with

512 scan lines at a scan rate of 1.6 µm.s-1. Samples roughness was assessed using the root mean square roughness (Rq) as calculated by Gwyddion SPM analysis software

(version 2.35). Surface topography images were created using Nanoscope Analysis software (Bruker, version 1.4).

4.2.8. Fouling Characterisation

Fouling and cleaning experiments were performed using a laboratory-scale cross flow unit, which allows for long-term automated cycles. The tests used 0.5% w/v NaOCl cleaning solution and a solution of 10 mg.L-1 BSA, 10 mg.L-1 HA, 2 mg.L-1 alginic acid,

10 mg.L-1 colloidal silica and 10 mg.L-1 calcium carbonate (sourced from Sigma Aldrich,

Australia) in Milli-QTM water as a synthetic feed. Before filtration, the flat sheet module and all membranes were rinsed with Milli-Q™ water at 30 L.m-2.h-1 for 30 min prior to the synthetic feed testing. Cole-Parmer Masterflex peristaltic pumps provided up to 2 bar of

TMP, recorded by pressure transducers, located across a custom built plate and frame membrane cell featuring an active surface area of 0.00275 m2. In addition, contact angle measurements were made using the sessile drop method (Akin and Temelli, 2011).

4.2.9. Module Conversion and Gravity Filtration

To validate the techniques developed, and explore potential applications, the conversion process was applied to a number of small scale 2.5” used RO modules.

117 4.2.9.1. Module Characterisation

The membranes were compacted in a 2.5” RO filtration apparatus with RO quality feed

water for 2 h at 10 bar prior to any further testing. Salt rejection characterisation was

conducted using 2000 ppm of NaCl (Sigma Aldrich, Australia) at 10 bar. The membrane

details and initial performance can be seen in Table 4-2.

Table 4-2: Initial specifications of used RO modules

Number Membrane Salt Initial PWP Label of Previous use module rejection (L.m-2.h-1 .bar-1) modules DOW Commercial (5 DOW-M1 1 86% years in 1.4 BW30 - 2521 storage) CSM Demonstrations CSM-M1 1 93% (3 years in 2.4 RE2521 -TL storage) Commercial DOW 88.8 - DOW-M2 3 (Months in wet 2.0 - 2.6 98.1% BW30 - 2540 storage)

4.2.9.2. Module Conversion

Conversion was completed by circulating 12,500 ppm NaOCl for 24 h at a flow rate of 30

L.h-1. Following conversion, the modules were flushed with recirculating RO water for at

least 2 h. Pressure decay tests were conducted before and after conversion by

connecting the pressure vessel feed to a centralised air supply line and pressurising the

module to 25 - 28 kPa, with the retentate and permeate values closed. The feed line was

then closed and the pressure loss was monitored for 10 min, via a pressure transducer

connected to the feed line.

4.2.9.3. Gravity Driven Filtration

After post conversion testing, the modules were autopsied and samples of the membrane

sheets were mounted on custom build plate and frame membrane holders. Each frame

holder has a membrane surface area of 0.12 m2 and up to three frames can be used in 118 the gravity filtration tank at a time. As seen in Figure 4-1, these frames were placed at the bottom of a 60 cm high tank, with a permeate outlet 80 cm below the tank, producing a total hydraulic head of 0.14 bar. Fouling and rejection tests were conducted with a

-1 -1 ground water analogue model solution containing, 12 mg.L BSA, 12 mg.L HA, 340

-1 -1 -1 -1 mg.L Bentonite, 100 mg.L Sodium Azide, 400 mg.L NaCl, and 80 mg.L Magnesium

Sulfate. This model solution was developed to induce accelerated fouling, with its composition based on the finding of a number of studies of rural African river and surface water sources (Aremu, 2011; Davies, 1996; Omo-irabor et al., 2008). Turbidity was measured using an Entech TN-100 turbidity meter and organics removal was measured using a TOC detector, Shimadzu TOC-VCSH.

Figure 4-1: Gravity membrane filtration Setup

119 4.3. Results and Discussion

4.3.1. Efficiency of Degrading Agents

The first step in developing an optimum conversion protocol for used RO membranes is to assess the various oxidative agents available. To select the optimum degradation agent, BW30 was exposed to chemicals known to damage the PA layer; including

NaOCl, KMnO4, and NaOH. The membrane samples, which had previously been compacted for 1-2 h with Milli-QTM water, were exposed to the agents for varying times.

Figure 4-2 shows the resulting PWP of the exposed membranes and Figure 4-3 shows the salt rejection performance of the same samples.

150

NaOCl

) 1

- 120 KMnO4

.bar 1

- 90 NaOH

.h

2

- .

60

PWP (L.m PWP 30

0 0 200000 400000 600000 Degradation Intensity (ppm.h)

Figure 4-2: Effect of various degrading solutions on the PWP of BW30 membranes.

Tested at 10 bar in high pressure stirred dead end cell.

120 100

NaOCl 80

KMnO4 60 NaOH

40 Salt Rejection(%) Salt 20

0 0 200000 400000 600000 Degradation Intensity (ppm.h)

Figure 4-3: Effect of various degrading solutions on salt (NaCl) rejection of BW30

membranes. Tested with 2000 ppm NaCl at 10 bar in high pressure stirred dead end

cell.

The results show that exposure to NaOCl resulted in the most rapid increase in permeability with increasing exposure intensity. Membranes exposed to degradation intensities greater than 300,000 ppm.h of NaOCl produced a permeability of 144 ± 6 L.m-

2.h-1.bar-1, all rejecting less than 6% of the NaCl in the test feedwater. This value appears to be the end point, beyond which no further exposure to NaOCl increased the permeability. While many studies have investigated the effect of NaOCl on RO membranes (as discussed in Section 2.6), only one has exposed RO membranes to these extremely high intensities (Raval et al., 2012). The study reported a converted flux of 124 L.m-2.h-1.bar-1 when applying a similar conversion protocol (300,000 ppm.h of

NaOCl exposure) to an in-house manufactured PA/PSf TFC SWRO membrane, following a wetting phase.

Both KMnO4 and NaOH treatment resulted in a significant decrease in salt rejection, indicating discernible damage to the membrane surface. However, flux increase

121 following treatment was significantly lower than after NaOCl treatment, suggesting that complete removal of the active layer was not achieved. These results differ from the study conducted by Rodriguez et al. (2002), who reported that KMnO4 most effectively degraded the RO membranes when compared to NaOCl and NaOH. However, their methodology used significantly lower degradation intensities (up to 12,000 ppm.h) and focused on variation in oxidant dose rather than intensity (ppm.h). From this previous study, little information was provided regarding the hydraulic performances (i.e. permeability of the degraded membranes), thus direct comparisons cannot be made. A separate study using low concentrations (up to 24,000 ppm.h) of KMnO4, reported similar results to the ones here, with a maximum of 2 fold increase in hydraulic performance exposure (Ambrosi and Tessaro, 2013). However, this low permeability was reportedly caused by the formation of manganese oxide on the membrane surface restricting flow, and the implementation of a citric acid cleaning step realised further improvements up to

10 L.m-2.h-1.bar-1.

It has been concluded that, in terms of both permeability and salt rejection, NaOCl presents the most promising performance as a conversion solution. Following this preliminary investigation, the study was expanded and a number of new virgin RO membranes (XLE, BW30FR, SW30) were acquired. Additionally, the testing protocol was modified to better assess the membranes under conditions of UF membranes (i.e. low pressure). Testing at 1-2 bar in the low pressure apparatus, these virgin membranes were converted using 300,000 ppm.h of NaOCl (see Figure 4-4), and over the complete range of exposure intensities (see Appendix A.4). From this, it was determined that exposure intensities above 300,000 ppm.h showed no additional performance change, indicating complete removal of the PA layer. Therefore, 300,000 ppm.h of NaOCl was selected as the optimum conversion protocol and has been used for all following work.

Interestingly, the converted permeability of these membranes, which are from the same manufacturer, range from 70 – 400 L.m-2.h-1.bar-1. This indicates a difference in PSf

122 support layer construction or thickness among the different membranes. Overall, it is evident that all converted RO membranes will feature varying hydraulic performances, which will need to be considered when identifying reuse applications. Additionally, these membranes had hydraulic performances similar to those of commercially available 10 and 30 kDa UF membranes (as seen in Table 4-1), and therefore further comparative characterisation was conducted.

) 1

- 400

.bar

1

-

.h

2 -

200 PWP (L.m. PWP

0 XLE BW30FR SWHR

Figure 4-4: Impact of conversion at 300,000 ppm.h on virgin DOW RO membranes.

Tested at 2 bar in stirred low pressure dead end cell.

123 4.3.2. Active Layer Removal

During preliminary investigation discussed in 4.3.1, the primary method to assess active layer removal was the change in membrane permeability and NaCl rejection. To further support the evidence of conversion, chemical changes in the membrane were investigated by FTIR spectroscopy, and visually assessed with SEM. In Figure 4-5, a select range of FTIR spectra for virgin and converted samples of BW30FR membrane are shown. To investigate the effect the extreme NaOCl exposure of the PSf layer, converted and virgin samples that had the PET support layer removed (as described in

Section 3.3.2) were tested using FTIR on the backside of the PSf layer.

Amide I Amide II Active layer virgin

Active layer Converted

PSf layer

virgin Absorbance (a.u.)Absorbance PSf layer converted 1750 1650 1550 1450 1350 Wavenumber (cm-1)

Figure 4-5: Infrared spectra of BW30FR membrane, virgin and exposed to 300,000

ppm.h of NaOCl.

-1 The spectra from the untreated membrane shows peaks at 1667 and 1542 cm , corresponding to amide I and amide II bands respectively, and are associated with predominately C=O stretching and N-H plane bending (Ettori et al., 2011). The peak at

1610 cm−1 is representative of the C=C stretching vibrations from the aromatic amide bonds (Donose et al., 2012; Ettori et al., 2011). The converted membrane exhibits suppression and possibly elimination of the characteristic aromatic PA peaks which has

124 been previously shown to be an indication of chlorine damage via the mechanism of

Orton rearrangement (Antony et al., 2010). When comparing these results to previous studies on free chlorine damage to PA membranes, the relative peak suppression is extreme and nearly complete (Cran et al., 2011; Kang et al., 2007), a difference which can be explained by the significantly higher oxidant exposures used. This suppression corresponds to breaking of various bonding groups and the subsequent deterioration of the chemical structure.

Analysis of the PSf layer before and after NaOCl exposure shows no significant change across the entire spectra. The peaks at 1586 and 1487 cm-1, which corresponds to a characteristic peaks of PSf (CH2 stretch), do not show any significant suppression. This lack of suppression indicates that the PSf layer has not been negatively affected in this band by the conversion process, and is therefore expected to retain is structure and strength (Cran et al., 2011).

To investigate how the membrane surface morphology changes during the conversion process and to visually confirm the removal of the PA layer, samples exposed to various intensities of NaOCl were imaged with SEM. These images, along with an image of a typical UF membrane for comparison, can be seen in Figure 3-6. Additional SEM images of studied membranes, including 30 and 100 kDa UF, can be seen in Appendix A.

125

Figure 4-6: SEM images of membrane surface at 40,000x magnification: A) Virgin

BW30FR, B) BW30FR treated with 60,000 ppm.h NaOCl, C) BW30FR treated with

300,000 ppm.h NaOCl, D) Virgin 10 kDa UF.

Figure 4-6A shows the typical surface expected with PA based RO membranes. In Figure

4-6B, the effect of chlorine exposure can be observed, with the 60,000 ppm.h exposure resulting in the polymer nodule structure loosening and starting to separate. When the membrane has been exposed to 300,000 ppm.h, the PA structure has been completely removed, as seen in Figure 4-6C. Additionally, polymer nodules from were the PA layer was physically interlocked with the PSf layer can be seen protruding from the PSf surface. This morphology is extremely consistent accros all samples images and closelly

126 resembles the visual structure of the commercially available 10 kDa UF membrane

(Figure 4-6D).

Based on these qualitative assessments, the PA active layer has undergone significant changes to its physical and chemical structure. Along with the changes in salt rejection and hydraulic performance, it is suggested that the conversion process results in substantial and possibly complete removal of the PA layer from the underlying support layers, leaving the PSf layer exposed and closely resembling a typical UF membrane.

4.3.3. Impact of Membrane Type and Condition on Converted Performance

To assess the variation in performance of converted membranes, the conversion process was applied to a range of virgin and used BWRO and SWRO membranes in varying conditions. The effect of membrane pre-compaction was assessed by applying the conversion process to samples left as stored, and with additional compaction.

Compaction was conducted at 15 bar with a pure water feed and conversion was completed with 300,000 ppm.h exposure of NaOCl. After conversion, the samples were tested for permeability at 2 bar in the low pressure apparatus.

127 120

As stored

) 1

- Compacted

.bar 1

- 80

.h

2 -

40 PWP (L.m. PWP

0 BW30FR Koch1 CSM Hydra Toray Koch2

Figure 4-7: Effect of membrane condition on converted performance. Permeability

tested at 2 bar.

As shown in Figure 4-7, the pre-wetted membranes demonstrated up to an 11 fold increase in permeability when compared to the un-wetted membranes; however, there was great variation in performance among the different membrane types. The virgin

BW30FR showed significantly lower permeability than when wetting was applied; while

Koch1, which was stored wet, achieved only 10% increase with additional wetting.

Additionally, the BW30FR and Koch1 membranes showed extremely poor converted performance when completely desiccated, even with attempted rewetting, with permeabilities of 1 and 5 L.m-2.h-1.bar-1 respectively. The salt rejection for all converted membranes was tested, showing less than 5% removal. This indicates that the PA layer is still significantly compromised on converted membranes with low permeability, included completely desiccated samples.

The two SWRO membranes, Toray and Koch2 showed extremely low converted hydraulic performance, with and without pre-wetting. Toray was stored moist, and Koch2 was a virgin sample that had not been previously wetted. While converted performance

128 was low compared to the other membrane types, the compacted samples still showed up to a 20 fold increase over initial permeability, reaching 7.8 and 11.2 L.m-2.h-1.bar-1 for

Toray and Koch2 respectively. It cannot be concluded that this is a result of their SWRO construction as the SW30 samples shown in Figure 4-4 had the highest permeability of all tested membrane types. Conversely, two of the BWRO membranes (CSM and Hydra), which were extremely fouled/damaged during coal seam gas water treatment applications, showed a high permeability of 123 and 110 L.m-2.h-1.bar-1 respectively, when pre-wetting was applied. Interestingly, the Koch1, CSM and Hydra membranes were visibly fouled with organic material prior to conversion. This fouling layer was not removed during the passive conversion process, but visual evidence was immediately washed away during the rinsing phase of the conversion protocol.

When comparing the converted performance of as stored, dried and wetted membranes, the results show that the level of moisture in the membrane structure affects the success of the conversion process. As discussed in Chapter 3, improper storage can cause pore collapse in the PSf membrane layer due to drying. However, as the stored used membranes still demonstrated reasonable hydraulic performance prior to conversion

(Table 4-1), and it was previously shown that the bulk of permeability loss only occurs in the final stages of drying, it is unlikely that pore collapse is the primary factor affecting the conversion process.

When visually inspected, the converted membranes were more translucent than non- converted membranes. However, the rim created by the pressure cell o-ring remained opaque on the converted membranes, and the same effect can be seen with the un- wetted samples. This shows the effect that compaction and wetting has on the conversion process, as the area outside of the o-ring does not experience high pressure during compaction. One proposed mechanism for the performance difference between the wet and dry samples is that increased inter-chain hydrogen bonding occurs with low moisture samples, and subsequently affects the process of ring chlorination (Louie et al.,

129 2011). For the samples that have been completely desiccated, the combination of this effect and pore collapse lead to extremely low membrane flux.

The different performances between the converted SWRO, BWRO and damaged membranes, indicate that converted performance is affected, not only by storage conditions, but also by the membrane condition and construction. Supporting this concept, it has been shown that different PSf construction methods and additives can affect the hydraulic performance of the support layer (Ghosh and Hoek, 2009).

Additionally, it is possible that further degradation of un-wetted samples would result in hydraulic performances comparable to the wetted samples, however this was not explored in the current study.

4.3.4. Rejection Properties

It has been shown that converted RO membranes have the hydraulic performance and surface morphology expected from commercial UF membranes. To characterise the rejection properties of the converted membranes, the membrane were challenge tested using a number of methods, including linear polymers, mixed organics and model virus solutions, and compared to 10, 30 and 100 kDa UF reference membranes.

4.3.4.1. Dextran Rejection

Dextrans are commonly used for UF membrane characterisation because only simple analytical techniques are required and because solute, solvent and membrane interactions are avoided (Baker, 2004). For this study, prepared monodisperse dextran solutions were consecutively filtered through the sample membranes in a stirred dead end cell at 2 bar. After 5 min of filtration a permeate sample was collected and measured using TOC detection. The MWCO of a membrane generally refers to the lowest molecular weight solute for which 90% is retained.

130 100

80 10kDa UF 60 Koch1 40

BW30FR Rejection(%) 20 100kDa UF

0 0 100 200 300 400 500 Molecular weight of dextran (kDa)

Figure 4-8: Dextran rejection results for UF and converted RO membranes. Tested at 2

bar and analysed using TOC detection.

As seen in Figure 4-8, the MWCO using this technique was approximately 250 kDa for the 10 kDa UF and Koch1 converted membranes, and was outside the measured range for the converted BW30FR and 100 kDa UF membranes. While these MWCO values are extremely high, given the 10 and 100 kDa MWCO of the reference UF membranes, these manufacturer stated values are determined using proteins instead of dextrans (Pall

Corporation, 2012). Comparison of protein and dextran properties is required to enable assessment of membrane properties using dextran solutions.

In contrast to proteins, which are generally globular and charged, dextrans are neutral, linear, water soluble molecules. It follows that rejection properties vary between the two techniques. Furthermore, polysaccharides, such as dextrans, have a loose structure and high degree of deformability, rendering them more permeable across a membrane than globular proteins of equivalent molecular weight (Perry and Green, 2008). This effect has been seen in numerous studies (Baker, 2004; Peeva et al., 2011; Venturoli and Rippe,

2005). However, this does not prohibit the direct comparison between membranes, if a constant feed solution is used and the unknown membranes are compared to known

131 reference samples. Based on this assessment, it can be concluded that the converted

Koch1 and BW30FR show a dextran rejection profile similar to the commercially available 10 kDa UF membrane.

An additional hypothesis for the low observed dextran rejection is the relatively high pressure used during filtration. It has been proposed that dextran rejection decreases dramatically with increasing applied pressure in a stirred dead end cell. Baker et al. showed that dextran filtration at 2 bar can result in up to 40% less rejection than the same dextran’s filtered at 0.5 bar, and that this effect is significant lower when filtering proteins (Baker and Strathmann, 1970). Furthermore, a separate study showed that TMP of around 1.5 bar displayed the lowest rejection of dextrans through UF membranes

(Schock et al., 1989). A more recent study reported the rejection of linear polymers, including PEG and dextran, by PES UF membranes (Levitsky et al., 2011). The reported results are similar to the ones presented here, with a 30 kDa membrane showing less than 90% rejection of PEGs up to 600 kDa. It was reported that the rejection of PEG and dextran is effected significantly by the hydrophilicity of the surface, as well as the linear nature of the molecules.

Therefore, while the rejection of dextrans does not provide an accurate representation of the MWCO of the converted membranes, it does provide a relative performance when compared to the commercially available UF membranes.

4.3.4.2. Protein and Humic Substance Rejection

To more accurately assess the performance of the converted membranes, they were challenge tested with a solutions of BSA and HA as described in Section 4.2.5, and the concentrations of solutes were determined by LC-OCD analysis. For further information about the LC-OCD methodology used see Section 3.2.8 and 3.3.4.

Figure 4-9 shows the total DOC, BP, HA and BB rejection for the three reference UF membranes (10, 30 and 100 kDa MWCO), three converted virgin RO membranes

132 (SW30, BW30FR and XLE), and three converted industrially used RO membranes

(Koch1, CSM and Hydra). As this model feed contains no particulate organic carbon, the total DOC is similar to TOC for this application, and is measured by the column bypass within the LC-OCD.

100 A

80

60 10 kDa UF 30 kDa UF 40

100 kDa Rejection(%) 20 UF

0 Total DOC BP HS BB

100 B

80

60 SW30 BW30FR

40 XLE Rejection(%) 20

0 Total DOC BP HS BB

133 100 C

80

60 Koch1 CSM 40 Hydra

Rejection(%) 20

0 Total DOC BP HS BB

Figure 4-9: Organic carbon substance rejection of (A) Virgin UF, (B) converted virgin

RO membranes, and (C) converted industrially used RO membranes, assessed with

LC-OCD.

The results show an expected general trend of decreasing rejection with decreasing component size, with BP showing the highest rejection and BB the lowest. The 100 kDa

UF membrane only has a BP rejection of 46% and a total DOC rejection of 18%. This result is expected, due to the large pore size, and the 67 kDa average molecular weight of the BSA. While the 10 and 30 kDa membranes have a similar PB rejection of greater than 99%, the main difference can be observed in the rejection of HS, with retention of

97 and 87% respectively. All of the converted RO membranes showed promising BP rejection of above 99%, except for the SW30 and CSM samples with an average rejection of 97 and 98% respectively. Additionally, the converted membranes showed average HS rejections between the 10 and 30 kDa membranes, except for the industrially used Koch and CSM samples; however, significant sample variation was seen. While the converted membranes perform favourably when compared to the 10 and 30 kDa membranes in terms of BP and HS, their rejection is lower for BB and when the total

DOC is considered. This is likely attributed to a wide pore size distribution due to a lack of defined skin layer on the PSf surface of the converted membranes (discussed further in Section 4.3.5). Additionally, while the samples were extensively flushed after 134 conversion, it is possible that residual chlorine trapped within the membrane structure degraded the organic feed into lower molecular weight components, or the filtration process flushed amide degradation products left over from conversion.

Overall, the converted membranes showed significantly higher rejection performance than the 100 kDa UF reference membranes across all categories and performed similar to 10 and 30 kDa membranes in the critical rejection categories of BP and HS.

Additionally, no significant difference was observed between the industrially used and virgin membranes. However, when compared to the 10 and 30 kDa membranes, the converted membranes showed wide sample to sample variation. Nevertheless, the all of the converted membranes show rejection performances associated with high MWCO UF membranes.

4.3.4.3. Pathogen Removal

To confirm the converted membrane rejection properties, investigation into pathogen removal by challenge testing was conducted to determine the suitability of converted membranes in possible water treatment applications. Citrate stabilised silver nanoparticles were used as a virus analogue and were selected based on consistency, ease of manufacture and testing (Antony et al., 2013). Particles were filtered in a cross flow configuration, followed by analysis of the permeate using ICP. The membrane removal efficacy was calculated as log removal value (LRV) using Equation 4-1.

Equation 4-1

Where Cin and Cout are the average particle concentrations in the feed and permeate respectively. The results and calculated LRVs are displayed in Table 4-3.

135 Table 4-3: ICP results from Ag model virus particle removal.

Feed Permeate Rejection Membrane concentration concentration LRV (%) mg Ag.L-1 mg Ag.L-1 Converted 9.19 0.035 ± 0.009 99.62 2.42 BW30FR

10 kDa UF 8.05 0.024 ± 0.003 99.75 2.57

The 10 kDa UF and the converted BW30FR membranes displayed rejections above

99.5%, equivalent to LRVs of 2.57 and 2.42 respectively for the tested particles. Previous work conducted by our research group has revealed that the particles closely resemble the physical characteristics of some water-borne viruses (Antony et al., 2013).

Specifically, the particles were 62 ± 10 nm in size, had a strong negative charge of -25

± 4 mV and were nearly spherical in shape (Antony et al., 2012).

These rejection values are slightly low when compared to other direct integrity testing methods including the commonly used MS2 bacteriophage. For example, this indicator has been shown to demonstrate an LRV of 3-4 for hollow fibre UF membranes of similar

MWCO values (Antony et al., 2012). However, as operating conditions significantly affect

LRVs, evaluating different methods can be difficult and further testing is required to adequately compare the various techniques. Ultimately however, the results indicate that the conversion process has not compromised the integrity of the porous UF supporting layer of the original RO membrane or its ability to remove virus-sized compounds from the feed water.

136 4.3.5. Fouling Propensity

Fouling remains a major cause of membrane productivity loss, and often dominates the design and operation of treatment systems. The susceptibility of a membrane to fouling is dependent on many factors, including surface morphology, hydrophobicity, surface charge, operating conditions and the nature of the feed water. While the converted RO membranes have been shown to have promising removal and hydraulic performances when compared to commercial UF membranes, their susceptibility to fouling is an important consideration. To investigate this, samples of converted BW30FR and UF membranes were imaged using AFM. Figure 4-10 shows representative surface scans for the different membrane types, and Figure 4-11 displays the root mean squared roughness values.

The AFM surface scans show a distinct difference between the PES UF and the converted RO membrane surfaces on the nanometre scale, with the converted RO samples having significantly larger surface features with a globular structure. It should be noted that the height scales are different for each membrane type; this was required to properly show the larger features of the converted RO membrane.

137

Figure 4-10: AFM height map surface scans (active layer side) of; A) 10 kDa UF, B) 30

kDa UF, and C) converted BW30FR.

138 8

6

(nm) 4

q q R

2

0 10 kDa UF 30 kDa UF Converted RO

Figure 4-11: Root mean squared roughness (Rq) for dry virgin UF and converted

BW30FR RO membranes.

The quantitative results displayed in Figure 4-11 show a statistically significant difference between 10, 30 kDa UF and converted RO membranes. As discussed in Section 3.3.2, the values for UF membranes are in line with previously reported results. The converted

RO membrane samples show an average roughness of 5.7 ± 2.1 nm, which is a significant decrease when compared to virgin RO membranes, which has been reported to be between 30 and 100 nm (Rq), depending on membrane type and imaging methodology (Ghosh and Hoek, 2009; Kwon and Leckie, 2006; Liu et al., 2008). This can also be clearly seen in the SEM images (Figure 4-6); while the PA layer of virgin RO is extremely thin, the surface features are comparably large. While the converted sample roughness is lower than virgin RO membranes, it is still higher than commercially available 10 and 30 kDa UF membranes.

It has been consistently reported that there is a direct correlation between membrane roughness and fouling affinity (Johnson and Hilal, 2015; Wong et al., 2009). High roughness results in a higher total surface area, and this, along with the resulting ridge and valley structure of the surface, leads to favourable conditions for the accumulation

139 of foulants (Johnson and Hilal, 2015). This effect has been specifically observed with the use of PES membranes and humic substances (Lalia et al., 2013; Peeva et al., 2011).

Due to the high roughness of the converted membranes and the extreme nature of chemically removing the physically interlocked PA layer from the PSf support, it is likely that they do not have a defined skin layer like the commercial UF membranes. This would indicate that they have a tighter average pore structure, to compensate for the lack of skin layer, while having comparable rejection and hydraulic properties. This significant difference between membrane morphology, suggests that the converted membranes may be more susceptible to fouling (Baker, 2004).

To investigate the fouling potential of the converted membranes, accelerated long term filtration experiments were performed using both converted BW30FR and 10 kDa UF membranes for comparison. A solution modelling river water with high organic content

(Described in Section 4.2.8) was used to accelerate fouling and cross flow cleaning was conducted using NaOCl.

Figure 4-12 shows a typical series of fouling and cleaning cycles from a converted RO membrane. The plot illustrates the change in TMP expressed as fouling rate

(TMP/tcycle) for each filtration cycle, as well as the overall TMP rate increase during the entire filtration period (TMP/toverall). The results demonstrate a consistent fouling gradient for each cycle, as well as an overall increase in TMP. Following each fouling cycle, the cleaning cycle was shown to partially recover the TMP increase caused by fouling build-up.

140 140 TMP/t TMP/t 120 cycle overall 100 80 60

TMP (kPa) TMP 40 20 Filtration Cleaning 0 0 2 4 6 8 Time (h)

Figure 4-12: Depiction of fouling effect for 60 min filtration and 10 min cleaning cycles

at 30 L.m-2.h-1.

Figure 4-13 shows the TMP behaviour over 32 cycles of 70 min each for the two membrane types tested. To assess the effectiveness of the cleaning cycles, cleaning efficiency (CE) was calculated using Equation 4-2.

Equation 4-2

Where TMPcycle is the TMP increase during the each fouling cycle and TMPresidual is the unrecovered TMP increase after the cleaning cycle. Table 4-4 summarises the TMP and cleaning efficiencies during consecutive cycles.

141 160

120

80 TMP (kPa) TMP 40 Converted BW30FR 10 kDa UF 0 0 5 10 15 20 25 30 35 40 Time (h)

Figure 4-13: TMP behaviour during cyclical fouling and cleaning.

Table 4-4: Cleaning efficiency and changes in TMP.

TMP/t cycle TMP/t overall Average cleaning Membrane type average (kPa.h-1) (kPa.h-1) efficiency (%) Converted 18.6 1.81 86.4 BW30 FR

10 kDa UF 22.6 1.01 91.7

The membrane resistance of the UF and converted RO was initially calculated as 0.30 and 0.58 × 1013 m-1 respectively. As fouling occurs during each cycle, a TMP increase was observed and this was partially mitigated during each subsequent cleaning cycle as the cleaning protocol removed the reversible fouling build-up. Over the course of the experiments, the overall TMP increase was consistent with the effects of irreversible fouling build-up, which was not recovered during the hypochlorite cleaning cycles.

However, as the silica and calcium carbonate components of the feed do not respond to the hypochlorite cleaning, further TMP restoration could be realised with an acidic clean.

142 The overall TMP increase was 1.8 and 1.0 kPa.h-1 for the converted and UF membranes respectively, showing a greater build-up of irreversible fouling on the converted membranes over the same time period. However, the commercially available UF membrane featured an average TMP increase per cycle of 22.6 kPa.h-1 compared to the converted membrane rate of 18.6 kPa.h-1, suggesting that the UF membranes fouled at a slightly faster rate during each filtration cycle. Results by Hajibabania et al obtained using a similar protocol, model solution and hollow fibre PVDF membranes, showed a comparable average fouling rate of 31.2 kPa.h-1 (Hajibabania et al., 2012). The difference in cleaning efficiency explains how the converted membranes fouled at a slower rate per cycle but faster overall. On average, the cleaning process was 92% effective for the UF membranes, but only 86% effective for the converted membranes, suggesting that while the UF membranes fouled faster, they were cleaned more effectively leading to a lower increase in TMP/toverall.

Contact angle measurements determined that the 10 kDa UF membrane was more hydrophobic at 83 ± 1° than the converted RO membranes at 70 ± 3°. This difference in hydrophobicity, along with the potential difference between the protein affinity of the PES construction of the UF membrane and the PSf construction of the converted RO membrane, could also make a significant difference in fouling performance (Baek et al.,

2012).

An important consideration, not explicitly addressed in these experiments, is the effect of the feed spacers on membrane fouling. Feed spacers can limit the deposition of gel particles and reduce concentration polarization (Baker, 2004; Schwinge et al., 2004), while promoting biofouling growth (Vrouwenvelder et al., 2009). Further consideration of the effect of the thinner RO feed spacers on UF applications is required as these may be susceptible to clogging or increased fouling, especially with feed of high suspended solids (Frick et al., 2014).

143 Under the relatively extreme fouling conditions imposed during this study, the converted membranes did not suffer from the catastrophic fouling rates initially hypothesised.

However, they are still quite susceptible to resistance increases and when compared to a commercially available UF membrane, did not respond as favourably to a basic cleaning modality. The differences between the performance of converted RO and manufactured UF will be an important consideration when seeking potential applications for the converted membranes.

4.3.6. Gravity Fed Water Treatment using Converted RO Membranes

The results of this study show that converted membranes have promising technical performance, making them potentially suitable as an alternative to UF membranes for a wide range of applications. As discussed in Chapter 2, one potential application is their use in gravity driven membrane water treatment, for use in rural communities and developing countries, and another is to use them as pre-treatment for further membrane processes (Ould Mohamedou et al., 2010). To validate this, a number of used membrane modules, were converted, characterised, and used in a prototype gravity driven filtration apparatus. A model solution, as described in Section 4.2.2.3, was used and cleaning methods include washing with tap water, and gently scrubbing with a sponge. These methods were aimed at replicating the cleaning methods available to non-technical operators in decentralised communities.

4.3.6.1. Preliminary Tests

Preliminary tests applied the 300,000 ppm.h NaOCl conversion protocol to DOW-M1 and

CSM-M1, 2.5” membrane modules. The conversion was applied as described in Section

4.2.2.2, and the membranes sheets where mounted on the gravity filtration frames. The model solution was filtered for up to 300 h and intermittent cleaning was applied.

144

20 1

- DOW-M1 with cleaning

.bar 1

- 15 DOW-M1 without cleaning

.h 2 - CSM-M1 with cleaning

10 CSM-M1 without cleaning

5 Permeability L.m Permeability 0 0 50 100 150 200 250 Time (h)

Figure 4-14: Fouling and cleaning performance of converted membrane samples from

DOW-M1 and CSM-M1 modules in gravity filtration set up.

As seen in Figure 4-14, initial tests showed an expected flux decline due to fouling, but also promising recovery using basic cleaning. However, based on previous experience in membrane conversion outlined in this chapter, the permeability was lower than

-2 -1 -1 expected, with initial PWP of 4 and 13 L.m .h .bar for the CSM-M1 and DOW-M1 membranes respectively. Upon removing the frames from the gravity fed tank, it was observed that fouling was only visible on portions of the membrane sheet. Figure 4-15 shows a representative example of this effect, which recurred though cleaning and fouling cycles, and was visible to various degrees on all membrane frames.

145

Figure 4-15: Fouling on membrane sheet from gravity fed filtration.

It was hypothesised that filtration was only occurring though the membrane sections that displayed fouling. This was confirmed by testing fouled and unfouled sections of the membrane sheets using dead end cell filtration tests. The results showed that the unfouled sections showed no permeability at 2 bar, while the fouled sections exhibited a

-2 -1 -1 PWP of 35.2 ± 0.6 and 52 ± 1.5 L.m .h .bar for CSM-M1 and DOW-M1 membranes respectively. As described in Chapter 3, it is likely that sections of the membranes dried during the years in storage, and therefore the PSf support layer could have suffered from pore collapse. Through basic image analysis using contrast thresholding, it was estimated that only 11 and 24% (averaged over 6 membrane sheets each) of the membrane area was being actively used for filtration in the gravity experiments, for the

CSM-M1 and DOW-M1 membranes respectively.

4.3.6.2. Optimisation and Further Testing

To build on the lessons learnt in the first tests, a number of BW30-2540 modules, which had been used for commercial water treatment but stored correctly, were obtained. The membranes featured PWP of 2.03 - 2.58 L.m-2.h-1.bar-1, and salt rejection of 88.8 - 98.1%.

146 Pressure decay tests before and after conversion indicated that the integrity of the modules had not been compromised, with an average pre-conversion decay rate of 0.34

± 0.02 kPa.min-1, and an average post conversion decay rate of 0.44 ± 0.14 kPa.min-1.

The membranes were then autopsied, and samples from different areas of the module were tested in the dead end cell. Over the 12 samples taken, the average PWP was 134

± 13 L.m-2.h-1.bar-1. Samples were also mounted on the gravity filtration frames and tested with a ground water analogue feed solution, with the aim of testing the same basic cleaning method (wash and rinse). Due to the high permeability of these membranes compared to the ones in the previous section (DOW-M1 and CSM-M1), long term fouling cycles were not possible using the current apparatus.

160

)

1 -

120

.bar

1

-

.h

2 -

80

Without Cleaning 40

Permeability (L.m Permeability With Cleaning

0 0 10 20 30 Time (h)

Figure 4-16: Permeability decline of converted DOW-M2 membranes used in gravity

driven membrane treatment.

Figure 4-16 shows the permeability loss experienced by the system over a period of 36 h, as well as the effect of simple cleaning methods. Without cleaning, the permeability continually declined over the testing period, starting at 129 L.m-2.h-1.bar-1 and decreasing to 75 L.m-2.h-1.bar-1 after 36 h. For each cleaning cycle, the membranes were removed 147 and wiped with a sponge until visible build-up was removed, resulting in almost complete permeability recovery. This simple cleaning method, which is manageable by an untrained operator, allows for the long term operation of the system and the effective restoration of flux.

Unlike the first batch of membranes in Section 4.3.6.1, the second batch showed fouling distributed across the entirety of the mounted sheets, and dead end cell tests confirming uniform performance. However, when the gravity tank was allowed to empty, the membrane dried out, resulting in no permeability upon refilling the tank. This again highlights susceptibility to drying, and is a complication for simple operation in decentralised area.

In a 24 h period, the system produced 19.4 L of water which is considered enough to supply water to a family of four in developing communities (Peter-varbanets et al.,

2009b). The membranes reduced the turbidity by greater than 99.9%, as well as removing 51 ± 9% of organic species, with no drop in rejection performance after cleaning cycles.

The difference in performance between the first and second batch of tested membranes highlights the importance of proper membrane storage after their primary use. Not all membranes are suitable for conversion and reuse, and each batch will have to be tested before mass conversion can take place. This testing would require a number of modules from each batch to be converted using the protocol outlined here, then autopsied and characterised to obtain a detailed picture of the possible membrane performances achievable.

As discussed in Section 2.7.3, the criteria for membrane systems in decentralised developing areas include compatibility with basic cleaning methods, high production with low energy requirements, low cost, and rejection similar to a 100 kDa MWCO UF membranes. The use of converted RO membranes passes all of these criteria. However,

148 the additional conversion steps, and the variability of converted performance of used membranes lead to the system being sub-optimum when compared to other available products with high membrane area such as the Skyhydrant and SkyBox (SkyJuice

Foundation, 2014b). However, this method does provide a source of low cost membranes that could be used with a flat sheet gravity fed membrane system such as the one developed by EAWAG (Peter-Varbanets, 2010). However, as the model solution used in the current study did not contain any natural organic matter content and included sodium azide, the effects and benefits of flux stabilisation in gravity systems that have been previously identified could not be studied (Peter-Varbanets et al., 2011).

4.4. Conclusions

In this chapter, NaOCl was demonstrated to be the most efficient agent in removing the active PA layer from virgin and used RO membranes. A number of characterisation techniques were applied to the resulting PSf based membranes to determine their suitability for reuse in UF applications. A wide range of permeabilities for the converted membranes were observed, rejection performance was comparable to 10 and 30 kDa

UF membranes, and they had limited susceptibility to fouling. As initially studied in

Chapter 3, membrane storage conditions were further assessed as a critical factor in membrane performance.

From this work, it can be concluded that membrane conversion is a valid end-of-life option. In the following chapter, this scenario, along with the other alternatives identified in chapter 2, will be assessed and compared for environmental impact using the tool of

LCA.

149 Chapter 5

Comparative Life Cycle Assessment

of End-of-life Options for RO

Membranes

This chapter is an expanded version of the following peer-reviewed journal article:

Lawler, W., Alvarez-Gaitan, J., Leslie, G., Le-Clech, P., 2014. Comparative Life

Cycle Assessment of End-of-life Options for Reverse Osmosis Membranes.

Desalination 357, 45–54.

Comparative Life Cycle Assessment of End-of-life Options for RO Membranes

150 5.1. Introduction

One of the core considerations when addressing an end-of-life waste management issue is the environmental impact of each option. The literature review in Chapter 2 highlighted limitations of previous LCA studies applied to membrane technologies used in water applications, and identified a knowledge gap regarding the detailed assessment of the environmental impact of membrane manufacturing and end-of-life disposal or reuse options.

This chapter aims to provide a novel perspective on the specific disposal challenges of the RO industry and to increase its environmental sustainability. The first detailed and process based life cycle model of the production and transportation of RO membranes will be provided, including all stages of membrane casting and module assembly. The model will be used to assess the impact of membrane manufacturing and transportation in the context SW desalination, and to compare the impact of 8” or 16” membrane modules. The chapter will also assess a variety of end-of-life disposal options, addressing the specific challenges in Australia, including transportation distances and industry regulations. The LCA will include end-of-life scenarios identified in Chapter 2, including the novel application of converting used RO modules to UF membranes, which has been explored in detail in Chapter 4. The effect of variation in membrane reuse lifetime and required transportation distance will be explored through a sensitivity analysis and the ultimate mass sent to landfill will be calculated for each end-of-life option.

151 5.2. Methodology

This LCA study follows the ISO 14040-44 guidelines (ISO, 2006a, 2006b), and comprises of four major steps. 1) Goal and scope definition, which identifies the purpose and objectives of the study, including the objects and processes to be studied and their system boundaries; 2) Life cycle inventory (LCI), which involves the systematic collection of all relevant inputs and outputs of all processes included within the system boundaries;

3) Life cycle impact assessment (LCIA), where collected data is grouped and assigned to specific impact categories and characterized using a suitable LCIA model; and 4) life cycle interpretation, where the LCIA model is used to make conclusions and recommendations in the context of the original study goal. Scenario models were developed for membrane manufacturing and available end-of-life options. These models were generated and assessed using Simapro version 8 software and the Ecoinvent 3 and AusLCI databases.

5.2.1. Goal and Scope Definition

This LCA study was undertaken to assist in the decision making process to determine the most preferable disposal options, driving membrane users and manufacturers to more sustainable practices. The intended audience of this study is a combination of policy makers, membrane experts, manufacturers and users. The study aims to answer the question:

“Which end-of-life waste treatment option is best for used reverse osmosis membranes from Australian municipal seawater desalination plants, from an environmental, and resource consumption point of view?”

152 In addition to the general goal, special consideration is given to:

- The impact and significance of membrane manufacturing,

- The sensitivity of reuse viability to transportation distance and secondary life

span,

- Diversion of waste from landfill.

Figure 5-1 shows a representation of a standard membrane lifecycle from extraction of raw materials to various end-of-life options. This study is focused on the disposal options, highlighted in green, and the impact of the different options is assessed on a comparative basis. To truly understand the benefits of the various options and to put the waste issue into context, a detailed model of membrane production (highlighted in orange), has also been completed. This model of membrane production includes the extraction of raw materials and manufacturing, packaging and distribution of the module. Therefore, the only component not considered in this study is the “use” phase of the membrane. This is contrary to nearly every previous LCA study on membrane water treatment, as they have primarily focused on the use phase. The use phase is not considered in this study as it is assumed that all membranes are equivalent after reaching the end of their life and that the energy and material consumption during their life span does not affect the impacts from the disposal options.

153 Extraction of Disposal raw materials Recycling

Reuse and Design and recycling Production

Reuse Production

Use Packaging Use and and Maintenance End-of-life Distribution

Figure 5-1: Membrane Life Cycle.

The LCA boundaries define what is included within the model for all processes considered. The membrane systems under discussion for this study are being used in desalination plants located in Australia’s major cities, i.e. Sydney, Melbourne, Adelaide and Perth. The membranes used are assumed to be manufactured in the United States of America, transported to Australia, then used, recycled and disposed of locally. As the membrane construction and use phases occur in different geographical locations, respective local data has been used. For example, local US electricity and material sources are used for membrane production, while Australian energy and material sources are used for end-of-life options. The boundaries of these models include all inputs and emissions associated with the contained processes (processing, manufacturing, transportation etc), and with the infrastructure required (machinery, buildings, vehicles etc). Due to the geographically spare nature of the various end-of-life options, a thorough transportation model is also included.

154 5.2.2. Functional Unit

The functional unit that is used for this study is one 8” standard TFC RO membrane module, specifically a FT-30 type with a PA active layer, PSf support layer, and PET base. The exact material breakdown of the element modelled in this study is summarised in Table 5-1. The 8” module is approximately one metre in length and 20cm in diameter, with an average dry weight of 13.5 kg and a drained weight of 15 kg. This module was selected due to its state-of-the-art technology and current wide spread use (Lee et al.,

2011). To compare the standard element size to the larger 16” alternative, the functional unit of membrane surface area (m2) is used.

Table 5-1: Composition of a dry 8” and 16” SWRO membrane elements.

16” element 8” element mass Component Mass of material Type of material of material (kg) (kg) Outer Fiberglass with 1.83 6 casing polyester resin PET base with PSf Membrane 4.63 19.7 supporting layer and sheet PA active layer Feed spacer 1.45 6.16 Polypropylene (PP) Permeate 1.81 7.6 Polyester (PET) spacer Acrylonitrile Tube and 2.38 7 butadiene styrene end caps (ABS) Glued parts 1.37 5.2 Polyurethane glue

5.2.3. Life Cycle Inventory and Impact Assessment

As this study includes a wide number of waste treatment options, consistency and quality of data is a primary challenge. Where possible, primary data has been collected from known industrial processes, or from experimental data generated in this thesis. This data includes, membrane material composition, material thermal properties, membrane

155 manufacturing techniques, and membrane cleaning, storage and chemical conversion methods. All known direct and indirect emissions, for processes within the system boundary, have been included in the scenario models. Where direct data collection was not possible, information from personal communication with membrane manufacturers and companies utilising the waste scenario has been used. Finally, where no other data source has been identified, literature data has been utilised. To complement the collected information, this study utilises the Ecoinvent 3 database and the AusCI database to provide locally relevant data.

To help with the interpretation of the LCA model and thus the comparison of the different disposal scenarios, the ReCiPe midpoint hierachist life cycle impact assessment (LCIA) method has been used (Goedkoop et al., 2008). This method has been selected due to its relevance to the study type, recent development, and compatibility with data sets used. It is a problem-oriented midpoint approach, meaning that the impact category indicators are directly derived from the inventory results. The impact categories for this study were selected based on the requirements and aims identified in the goal and scope definition phase of the LCA, and can be seen in Table 5-2. While 9 impact categories from the ReCiPe method were chosen and assessed in this study, climate change potential and fossil fuel depletion were selected for further discussion, as they are easily interpretable, have low data uncertainty and are globally relevant (Goedkoop et al., 2008;

Shonfield, 2008). Detailed input and output tables can be found in the Table C-1 and C-

2 of Appendix C.

156 Table 5-2: Selected ReCiPe impact categories and units

Characterisation Area of Concern Impact Category Units

Climate Change Global warming potential kg CO2 eq

Resource Depletion Fossil fuel depletion kg oil eq Human Health and Human toxicity Potential Kg 1,4-DB eq Toxicity

Ecotoxicity Potential kg 1,4-DB eq

Marine eutrophication kg P eq

Freshwater eutrophication kg N eq

Air Pollution Ozone layer depletion potential CFC-11 eq

Photochemical oxidant formation kg NMVOC eq

Terrestrial acidification potential kg SO2 eq

5.2.4. Uncertainty Analysis

The quality and uncertainty of data used in a model remains a critical aspect of LCA studies. All data which is used in LCA models has varying degrees of uncertainty, which can come in a number of forms, from uncertainty about the correctness (accuracy) of the included data, to uncertainties caused by incomplete models (excluded data).

During this study, uncertainty was calculated through a number of different methods. For processes adapted from the LCI databases in Simapro, the uncertainty in the unit processes was used (Ecoinvent Centre, 2013). For data generated specifically for this study, either experimentally or through literature surveys, either an estimated geometric standard deviation or Max/Min approach was used. Where a range of possible values was obtained, either through experimentation or literature reporting, those values were used for the uncertainty calculation using triangular distribution. An example of this is the

157 inputs and outputs of the gasification scenario, which are based on a detailed assessment of available technologies and provide a range of possible emission values.

To assess the uncertainty of values that did not have an existing uncertainty, an estimated geometric standard deviation was calculated using a Pedigree matrix

(Goedkoop et al., 2010). In this method, each data point is rated against six criteria, from reliability to sample size, and a basic uncertainty factor. Further information about this technique can be found in Section C.4 in Appendix C, including all uncertainty values calculated for the data used in the various models. Using a combination of these discussed techniques, an uncertainty value was calculated for each data point used in this model.

Using these values, Monte Carlo simulations were used to calculate the uncertainties in

LCIA results. This technique used function built into the Simapro software to randomly vary each value within the giving uncertainty range that was specified and calculate the

LCIA results.

When comparing two alternative options using LCA analysis, care needs to be taken to account for any shared uncertainties in the models (Grant, 2013). For example, as all of the end-of-life options for used RO membranes studied include the uncertainties for the membrane manufacturing, a raw comparison of uncertainties from Monte Carlo simulations would be interpreted as not statistically significant (Clavreul et al., 2012).

This is due to the many shared processes within the models, such as materials, electricity, fuel and transport.

The Simapro software accounts for this problem by assessing two systems together and by changing the common variables at the same time, thus directly comparing the impact of the non-common variables. Monte Carlo simulations were completed and the software calculated the proportion of times scenario A has higher emissions (or material consumption, depending on category) than scenario B. If 90% of the Monte Carlo runs

158 result in higher emission (or lower) then the results are considered statistically significant

(Goedkoop et al., 2010).

A limitation of this process is that only two scenarios can be compared at once; therefore,

21 pairwise comparisons were conducted to completely assess the uncertainty of each option relative to another. This method for multiple pairwise comparisons within complex assessments has been previously used and validated (Guo and Murphy, 2012).

5.3. Model Description

5.3.1. Membrane Manufacturing

In order for this study to determine the significance of various end-of-life options for used

RO membranes, particularly the reuse options, a detailed model of the membrane manufacturing is required. This manufacturing model can also provide additional insight into the impact of used different membrane types and configurations, and their contribution to the entire water treatment process. Figure 5-2 shows a simplified process flow diagram for membrane manufacturing and the end-of-life scenarios considered in this study; detailed schematics of these models, including process boundaries can be found in Appendix C.

159 Manufacturing and use phase End-of-life processing Process offsets

Membrane Used Landfill Transport manufacturing Membrane

Electricity Plastic bag and Storage Transport Shredding Incineration Packaging cardboard box solution production

Filament Resin and Syngass Electricity Fibre glass Disassembly Shredding Gasification winding glass strand combustion production case mouling extrusion

Coke Extrusion Granular Transport Disassembly Crushing Washing EAF End caps moulding plastic production

Shredding Offset Material Membrane Membrane Support layer Transport Disassembly and Sorting material Recycling sheets layer casting manufacturing cleaning production

Packaging Extrusion Granular Reuse as Membrane Spacers Compaction Conversion and Spunbonding plastic UF module manufacturing Transport storage Packaging Extrusion Granular Membrane Permeate tube Cleaning and Direct Reuse moulding plastic manufacturing storage

Figure 5-2: Simplified process flow diagram for membrane manufacturing and end-of-life scenarios including offsets.

160 The manufacturing and use phase of Figure 5-2 shows how the construction of the module is broken up into a number of components. While the exact composition of the membranes can vary from type and manufacturer, average values are generated from numerous membrane autopsies and manufacturer information.

The manufacturing of the various plastic components including the permeate tube, spacers and end caps, uses common moulding techniques. Each part was modelled individually with respect to their specific moulding style and the materials involved. The membrane sheets comprise of 37m2 of a three-layer membrane constructed of varying polymers. By weight, the most significant layer is the spun-bonded PET support layer, which provides the structural foundation for the remaining two layers. Information from a variety of manufactures has been adapted for this support layer (Dahiya et al., 2004;

Edwards and Fry, 2006; Migliavacca, 2010). The second layer of the RO membrane is constructed of PSf via phase inversion. This model was constructed from the Ecoinvent data using stoichiometric calculations of the base chemical components in addition to estimated energy and water inputs (Zhang et al., 2011). A number of different solvents can be used for the casting phase, however dimethylformamide (DMF) is used in this model as it is the most widely reported (Baker, 2004; Bonton et al., 2012). The process consumes 4kg of DMF for every 1kg of membrane produced, assuming a 20wt% polymer mix and no solvent recovery (Bonton et al., 2012). The final membrane layer is constructed of dense PA and is laid on the PSf via interfacial polymerisation. RO membranes of FT-30 construction are aromatic with a highly crosslinked structure, and are based on the reaction of m-phenylenediamine (MPD) and trimesoyl chloride (TMC).

This layer is around 0.02 µm thick, and thus only 30g of the polymer is found in each membrane module. As no specific data on the chemical components of the PA that is used for RO manufacturing, and due to the relatively small amount in use, inputs for

Nylon 6.6 (an aliphatic PA) have been used in its place. The total membrane sheet production process includes two casting and curing phases (Caddotte et al., 1981).

161 Water, which is used in both casting stages for precipitation and washing, is estimated based on available literature and existing knowledge on membrane casting to be recirculated at a rate of 20L per module (Baker, 2004; Rangarajan et al., 2011). Energy use from the heating and machine operation has been estimated based on Ecoinvent data on similar material manufacturing.

The membrane module is assembled by gluing of the membrane sheets to the permeate tube, with layering of the feed and permeate spacers, and then rolled up and glued into place. The endcaps are fitted onto the permeate tube and the fibre glass case is cast around the membrane module. Energy consumption and production time of this process has been based of personal communications with membrane manufacturers and module rolling machines. The final component of the membrane element is the O-ring that is fitted around the feed side of the module, it is made out of ethylene propylene rubber and weighs 30g.

5.3.2. Disposal Options

5.3.2.1. Landfill

Landfill involves the disposal of waste by burial and is the most common method utilised in Australia (Australian Bureau of Statistics, 2012), and is therefore the baseline scenario for comparison. Due to their mostly polymeric composition, membranes are considered inert municipal solid waste in the case of landfill disposal. The inert components are not toxic and thus the main problem associated with their existence in the landfill is due to land occupation and transportation (Nielsen and Hauschild, 1998).

To model the transportation distance, a survey of all municipal landfills that process this waste type, near the six major desalination plants in Australia was completed and the average transportation distance was calculated. It was assumed that the end-of-life modules are sent to one of the near municipal landfills for disposal and that it would be transported by an Australian fleet average rigid truck. The landfill transportation

162 distances can vary widely from membrane plant location. For example, it is 30km from the Perth Southern Seawater Desalination Plant to Dardanup, but 129 km from the plant in Tugun Queensland to Veolia’s Ti-tree landfill site in Willowbank QLD. Where possible, primary data of the specific landfill disposal site used by the plant was collected via personal communication. Over 50 landfill sites were assessed and the average distance was calculated as 46.7 km, with minimum and maximum distances of 2 and 129 km respectively.

While it is estimated that 13% of Australia’s landfills have some form of gas recovery system ranging from 300 kilowatts to 13 megawatts (Harris, 2012), the nature of inert plastic waste makes the membranes impact on these systems negligible, and thus it is not to be considered in this study.

Due to the inert nature of the plastic construction of the membrane module and relatively short transportation distances, the initial hypothesis states that the landfill scenario does not comprise of a large volume of emission. However, this does not mean that landfill is by default the best option, as reuse and recycling generally gain emission offset due to avoided production of new products.

5.3.2.2. Incineration

Incineration is a thermal waste treatment method that involves the combustion of materials to produce ash, gas emissions and heat, while reducing the volume of the waste material. Incinerators generally operate at above 850 °C and combustion occurs in an excess of oxygen (Department for Environment Food and Rural Affairs, 2013).

Plastic waste is considered a good fuel source because they have a heating value almost equivalent to that of the coal. The thermal energy produced from the waste incineration is used to generate electrical energy, which is this scenarios primary offset. Additionally, the recovered energy can reduce pressure on existing generation facilities. Incineration also greatly reduces the volume of waste by about 90-95% (Siddique et al., 2008). While

163 incineration is commonly used in countries with strict land use requirements (Tan and

Khoo, 2006), and in some cases has been shown to be environmentally favourable to landfill (Assamoi and Lawryshyn, 2012; Finnveden et al., 2005; Morao et al., 2008), current Australian regulation states that incineration cannot be used for municipal solid wastes (Australian Bureau of Statistics, 2010). Therefore, due to lack of local information, this model uses international literature data, coupled with bench scale experimental work

(Prince et al., 2011).

5.3.2.3. Syngas Production

Gasification is the partial oxidation of carbon-based feedstock to generate syngas, which is directly combusted onsite in an internal combustion engine generator to produce electricity (Al-Salem et al., 2009). Oxygen is adjusted to maintain a reducing atmosphere, but the quantity is maintained lower than the stoichiometric ratio for complete combustion. The major input and emissions data for this scenario has been adapted from a detailed survey gasification facilities using plastic feed components in North America

(American Chemistry Council, 2012; US EPA, 2012), with information on membrane collection, disassembly and sorting added to complete the model.

5.3.2.4. Energy Recovery in EAF Steelmaking

Another possible option is the use of the polymeric membrane components as a substitute carbon source in an EAF steel making process. The use of waste plastics and rubbers as a substitute or metallurgical coke has been extensively tested in recent years and has seen commercial use (O’Kane, 2011; Sahajwalla et al., 2011). This method has been specifically tested with membrane components and the results show a similar benefit of their use, when compared to other waste materials (see Appendix B for complete study). A partial waste polymeric material substitute actually improves the process though increased energy retainment and promotion of foamy slag. A number of detailed LCA studies on polymeric substitution have shown an insignificant change in

164 flue gas emissions, and highlighted that the primary benefit from using this technique is the diversion of polymeric materials from landfill, and the reduction in coke consumption

(Clauzad, 2005; Rocheta et al., 2011). The process requires the removal of the membrane case, as its silica composition is unsuitable for EAF applications and rigorous plastic washing to remove contaminates is also required. While substitution rates between 1-1.7 kg of polymer to kg of coke have been reported (Ayed et al., 2007; Gorez et al., 2003; Zaharia et al., 2009b), a 1 kg/kg rate using a mixture ratio of 30:70 polymer to coke has been selected for this study, as this was the composition specifically assessed during bench scale testing (Appendix B).

5.3.2.5. Material Recycling

The primary recycling method for end-of-life plastic components is mechanical recycling

(A’Vard and O’Farrell, 2013; Lee et al., 2005), which involves the shredded plastic flakes being melted and reformed via melt-extrusion, to produce uniformly-sized pellets which can be used as a raw material for new products (Brulliard et al., 2012). This process required the module to be disassembled and sorted prior to the plastic washing and grinding stages. Due to the requirements of mechanical recycling, only the ABS components, including the tube and end caps, and the spacers are suitable for this method, with all other components being sent to landfill (Brandrup and Wiesbaden, 2005;

Department of Environment Climate Change and Water, 2010a; Goodship, 2007).

Following the sorting and shredding phase (were 5% material loss is assumed), the material goes through the process of melt-extrusion, and it is assumed that there is a

10% loss of material and product quality (Department of Environment Climate Change and Water, 2010b). The benefit of this process is realised through the production offset of virgin plastics.

165 5.3.2.6. Direct RO Reuse and Conversion to UF Membranes

Direct membrane reuse involves taking membranes that have been deemed unsuitable for their primary application from one plant and transporting them to a secondary plant.

The most viable secondary application involves harsh feed water conditions that require regular membrane replacement. As the functional unit of this study is one RO membrane, it is assumed that the module is suitable for direct reuse. To estimate the potential lifespan of the membrane during its second use, a survey of membrane conditions from companies facilitating membrane reuse has been used (see section 2.2.3 and 2.5.1).

Using this data, manufacturer specifications and expected membrane lifetime, it was estimated that a membrane reused in harsh BWRO conditions would have a virgin membrane production offset of 33%, followed by disposal in landfill. While the membrane reuse will offset production, it also requires an additional cleaning, packaging and preservation step prior to transport. These have been modelled on industrial best practice and incorporated into this scenario (DOW, 2012; Toray Industries, 2005).

If the membranes are not suitable for direct reuse as RO, they can be converted with chemical treatment to UF. This technique exploits the relative vulnerability of the PA layer to oxidative chemicals to remove it, leaving the PSf layer intact, which can act as a UF membrane. The details of the process, which this model is based on, has been extensively studied discussed in Chapter 4. This model used the application of UF pre- treatment for RO systems and assumes the lifetime of the reused membrane is 2 years.

5.3.3. Transportation

A number of studies have shown that the exclusion of a detailed transportation model in an LCA study can result in a severe underestimation of the environmental impacts of disposal scenarios (Brambilla Pisoni et al., 2009; Merrild et al., 2012). This is particularly important in this study as each scenario has a different corresponding transportation distance and method. Additionally, due to Australia’s size and sparsity, transportation

166 becomes a significant factor when considering economic or environmental impacts.

Table 5-3 summaries the average, minimum and maximum transportation

distances required for each scenario and outlines the method used. The distances to

landfill and processing sites were calculated from extensive surveys of available

locations. The incineration and syngas scenarios share the same transportation distance

as landfill, as examples of these processes are currently limited in Australia, and due to

their common placement near landfill sites.

Table 5-3: Summary of scenario transportation distances and sources

Average Min Max Transport Scenario transportation distance distance Calculated from method distance (km) (km) (km) Dow manufacturing Rail freight location in Membrane to port then Minneapolis 16,274 14,846 17,700 manufacturing international Minnesota to shipping Australian RO plants Rigid truck Desalination plant Landfill 47 2 129 3.5-16t to local landfill sites Rigid truck Desalination plant Incineration 47 2 129 3.5-16t to local landfill sites Rigid truck Desalination plant Syngas 47 2 129 3.5-16t to local landfill sites Desalination plant Articulated EAF 1770 68 4074 to OneSteel plant in truck >20t Sydney Desalination plant Material Rigid truck to local recycling 85 6 209 recycling 3.5-16t facilities that can process waste type One major Articulated Reuse as RO 2480 76 4556 desalination plant truck >20t to another One major Conversion to Articulated 2480 76 4556 desalination plant UF truck >20t to another

167 5.4. Results and Discussion

5.4.1. Membrane Manufacturing

A detailed model of the impacts from the manufacturing processes of RO membranes is vital for understanding the effectiveness of the studied reuse, recycling and disposal options. In the situation that only landfill, recycling and incineration are considered, a model including only the material components may prove suitable. However, as this study includes a number of scenarios that rely on the offset of membrane production itself, a detailed model with the energy required for production and the consumables like solvents and water is vital for getting a clear picture of the viability of the various scenarios. Figure 5-3 shows the contribution of the different components to the total environmental impact of the membrane manufacturing process. This type of representation is useful for determining which parts of the manufacturing process have the most significant impact, and thus identifying area of possible improvement.

-6 87.7 kg 5.2x10 Kg 6.5 kg 0.369 kg 0.45 kg 0.02 kg 0.14 kg 0.15 kg 38.5 kg CO -e CFC-11-e 2 1,4-DB-e NMVOC SO2-e P-e N-e 1,4-DB-e oil-e 100

Spacers

80 Polyester support layer Polysulfone and 60 polyamide layers Fibre glass case

40 Structural components Assembly

Relative impact (%) impact Relative 20 Transportation

0

Figure 5-3: Relative impact from different components during the manufacturing of one

RO module. Values above bars are the total emissions for the impact category.

168 The relative environmental burden for 9 difference impact categories relevant to this study are displayed, and can be used in conjunction with the absolute values to provide detailed context for the emissions. In terms of climate change potential for the module manufacturing, the results show that, overall, the production and transportation of an 8”

RO element contributes 87 kg CO2-e emissions to the atmosphere. The largest contributor for this impact is from the membrane sheet manufacturing; with this being the heaviest component within the module (Table 5-1). With 4.63 kg of membrane sheets required for each module, they make up 45% of the total CO2-e emissions, with 25% originating from the support layer manufacturing and 12% coming from the solvents used in the PSf and PA layers. Interestingly, despite the vast distances associated with the transportation of the membrane from its manufacturing location in the United States to the point of use in Australia, it only contributes to 8% of the total climate change impact.

Interestingly, the marine eutrophication emissions is dominated by the production of the

PSf and PA layer with the solvents required (particularly the DMF used for the PSf production) having 90% of the contribution, due to the volumes of nitrogen that their production releases into the environment. Finally, the results show that the production of a 13.5 kg membrane element requires the consumptions of 38.5 kg oil-e.

Figure 5-4 shows the results of the uncertainty analysis using Monte Carlo simulations for the manufacturing of 8” RO membrane modules. The uncertainty of the impact categories varies, with the highest values obtained for marine eutrophication and ecotoxicity. The high uncertainty for these categories is primarily attributed to the general uncertainty of current scientific knowledge on the specific environmental impact of the various compounds (Grant, 2013), and the compounding uncertainties of the numerous layers in the model. The results show that impact categories with the lowest data uncertainty for this model are climate change, and fossil fuel depletion, with an overall variation between 10% and 15%. This uncertainty is acceptably low, considering that the model comprises of over 10,000 values each with an associated uncertainty value.

169 200

150 % 100

50

Figure 5-4: Uncertainty propagation analysis of the membrane manufacturing model

using Monte Carlo assessment. 2000 runs, 95% confidence interval.

5.4.2. Impact of Membrane Module Size on Production Emissions

There is increasing interest in the desalination industry in the larger standard format of

16” elements. These modules are constructed out of similar materials to the 8” elements but can be over 60 kg in weight and have 4 times the membrane surface area. To compare the environmental impact of the two formats, the 8”manufacturing model was adapted for the 16” elements, including changes in associated production inputs and transportation requirements. The results show that the production of the larger module is associated with 348 kg of CO2-e, while the equivalent four 8” modules results in 351 kg of CO2-e. While there is some variation in the other impact categories, Monte Carlo simulations show that these results do not show a significant difference between the two.

170 This suggests that there is no benefit to be gained from manufacturing a smaller number of larger modules. However, this analysis does not take into account the potential reduction of plant equipment and maintenance which are part of the advertised benefits of the larger format. Therefore, this LCA model cannot conclusively show that there is benefit from using the larger elements and further information about the variations in membrane manufacturing is required to make a more detailed and conclusive assessment.

5.4.3. Contribution of RO Manufacturing to Desalination Operation

Emissions

While any reduction in process waste can help the industry become more environmentally friendly, it is important to put the impact of membrane manufacturing and replacement into context. For a consumable component, such as a RO membrane, it is logical to compare its impact relative to the overall process. A number of LCA studies of varying depths and qualities have been completed on SW desalination, with the majority focusing on the energy consumption, which is considered to be relatively high for RO processes. This focus on energy generally means that infrastructure and consumables are either ignored, or simplified. Several studies have stated that the inclusion of membrane manufacturing contributes less than 1% to emissions and can therefore can be excluded (Raluy et al., 2005a; Vince et al., 2008), while others have assumed the membranes to be comprised entirely of one material, such as PA (Raluy et al., 2005b; Tarnacki et al., 2012). Interestingly, one study, which has a reasonably detailed assessment of the material component impacts of NF membranes, reported that up to 7% of the CO2-e emissions came from the manufacturing process (Bonton et al.,

2012). However, this higher contribution may be attributed to the significantly lower energy requirement of the NF process (0.55 kWh.m-3).

171 Table 5-4: Summary of available literature data on SW desalination CO2-e emissions

Inclusion Energy of Conditions consumption Kg CO -e.m-3 Reference membrane 2 (kwh.m-3) production European context. 46,000 m3/day. 8 Trains of 82 pressure vessels, 7 Implied but (Raluy et al., 4 1.78 membranes each. not specified 2005a) Model SU-820. Assumed membrane life of 5 years BWRO and sea water RO. Not included United States, 2 for BW (assumed to BW – 1.15 (Zhou et al., Singapore and Spain 4.9 for SW be similar for SW – 2.75 2011b) context (different all scenarios) energy mixes) Comparison of RO Yes but treatment to a thermal entire direct contact (Tarnacki et 2 - 4 membrane 1.77 membrane distillation al., 2012) assumed to process. Spanish be PA context. LCA of Perth SW 3.3 for water Said to be desalination plant. treatment, 3.82 included, but (Poussade et 4.2 125,000 m3/day. including no specific al., 2011) Australian context. distribution details. 4.79 (Calculated using AusLCI database value Literature review of (Plappally for average energy requirements Average of 4.6 n/a and Lienhard Australian low for water production. V, 2012) voltage electricity production)

In order to put membrane production impacts into context with the entire RO process, without performing a detailed LCA of RO water treatment, the obtained values can be compared to literature data. To do this, an extensive literature search of available LCA and carbon footprint studies related to RO treatment was completed (Summarised in

Table 5-4). Surprisingly, out of all the studies reviewed, only four included absolute emission values (primarily kg CO2-e). A fifth study by Plappally et al, provided a review of energy consumptions for RO processes around the globe and was thus consider

172 significant enough to include by using Simapro and the AusLCI database to estimate the emissions.

Assuming that one membrane produces on average 10 m3.day-1, and an average lifespan of 7.5 years, 27400 m3 of water are produced in its lifespan. Additionally, the production, transportation and subsequent disposal (in landfill) of one membrane module is associated with the emission of 88.4 kg of CO2-e, meaning that the membrane

-3 -3 contributes 3.23x10 kg CO2-e.m of water produced. Using these values, the calculated total process contribution of the membrane modules is between 0.18 and

0.07% for the lowest and highest process emission cases respectively. This contribution increases with decreasing membrane lifespan, but even with a 3 year replacement frequency, the contribution only reaches 0.45%. As a result of this comparison, which included the most detailed public LCA of membrane manufacturing to date, it is safe to assume that membrane manufacturing impacts are negligible when considering overall

SWRO process emissions. However, these numbers are more of an indication of the extreme energy requirements of the RO process, rather than the insignificance of membrane manufacturing impacts.

5.4.4. Comparison of End-of-life Scenarios

The primary goal of this study was to determine which end-of-life options for used RO membranes originating at Australian municipal sea water desalination plants, is the most environmentally favourable. While nine environmental impact categories were investigated, climate change (CO2-e) and non-renewable resource depletion (Oil-e) will be the focus of the discussion, as they are considered to be global environmental issues, easy to interpret and have the lowest data uncertainty (Goedkoop et al., 2008; Shonfield,

2008). Figure 5-5 shows the impact of each disposal scenario for climate change and fossil fuel deletion relative to membrane manufacturing. Absolute values and other impact categories can be found in Table C-4 of Appendix C.

173 15

10 CO2-e emissions

5 Oil-e consumption

0

-5

-10

-15 Relative Impact (%) Impact Relative -20

-25

-30 Landfill Incineration Gasification EAF Recycling Conversion Direct RO to UF reuse

Figure 5-5: Greenhouse gas emissions and resource depletion for the disposal of one

RO membrane element. Results are displayed in terms of relative offset of membrane

production.

The impact of landfill is relatively small when compared to membrane manufacturing, which is expected, due to the inert materials and short transportation distances. Even though the polymeric membrane components are mostly inert, there is still a degradation of 1 – 5% of the material during the 100-year-surveyable time period (Arena et al., 2003;

Paul P. Rumps, 1991). While small, this effect and the transportation of the membranes from the plant to the landfill location, contributes to the CO2-e emissions and resource use of landfill disposal.

The results show that the reuse scenarios are highly environmentally favourable across the studied impact categories. Direct RO reuse has both the greatest reduction in CO2- e emissions and fossil fuel depletion of all the scenarios, with conversion to UF scenario is only slightly behind, due to the extra chemical treatment steps involved.

174 While the reuse scenarios gain the benefit from avoided production of virgin membranes, the recycling scenario gains environmental credit from the offset of virgin plastic production. The results also show that after the reuse scenarios, recycling has the greatest environmental benefit. The recycling of the PET permeate spacer and PP feed spacer generate a CO2-e emissions saving of 0.93 and 1.25 kg CO2-e per kg recycled respectively. This can be compared to the average benefits reported by the Australian recycling sector, which are stated to generate a saving of 1.07 and 1.64 kg CO2-e per kg recycled for PET and PP respectively (Brulliard et al., 2012). The recycling of ABS provided a considerably higher offset of 2.5 kg CO2-e per kg due to the higher energy requirements for the manufacturing of virgin ABS (Ecoinvent Centre, 2013). Close agreement with those previously reported values adds to the validity of these comparisons and the viability of the recycling scenario as an alternative disposal option for end-of-life membranes.

The scenarios of incineration, syngas production and EAF are under the category of energy recovery. While the incineration scenario generates significant electricity production, which offsets fossil resource consumption, the combustion and subsequent release of flue gas generates considerable CO2-e emissions. The gasification process provides a modest environmental benefit across all impact categories, which is gained from electricity production through the combustion of the generated syngas. While the benefits are small when compared to the reuse and recycling scenarios, this model suggests that gasification is favourable over incineration.

The third energy recovery scenario is the use of the membrane material as polymeric carbon source in EAF for steelmaking. Unlike the incineration and gasification scenarios, this process does not gain its benefits from electricity generation, but rather from negating the use of metallurgical coke. This offset includes the required mining, processing and transportation of the coke to steel mill. While the EAF scenario does

175 provide a positive emissions offset of 3.5 kg CO2-e, the main benefit comes from the reduction in non-renewable resource use with an offset of 6.2 kg oil-e.

When comparing the various end-of-life scenarios, it is important to determine if the differences between them are significant, given the various uncertainties in the model.

The results of uncertainty analysis using pairwise Monte Carlo assessment can be seen in Table 5-5. Only the outputs for CO2-e emission and fossil fuel consumption are shown, as they are the primary indicators used in this study and the results for the seven remaining impact categories can be found in Table C-8 Appendix C. As an example of interpreting the results, the comparison between direct reuse (column) and landfill (row) shows that, out of the 1000 Monte Carlo runs that were completed, none of them produced a result in which the direct membrane reuse option had higher CO2-e emissions than the landfill one.

The majority of comparisons show a significant difference, with the important exceptions being comparisons with either EAF, Conversion to UF or direct reuse scenarios. The uncertainty in comparisons involving these scenarios is due to their highly variable transportation distances and membrane reuse lifetime, which can range from 1770-4556 km and 0.5-5 years. Additionally, the gasification scenario has high uncertainty in CO2-e emissions due to range of technologies used in the model, and incineration in Oil-e consumption due to uncertainty in the amount of energy production offset.

Overall, all end-of-life options studied apart from incineration show a reduced environmental impact over landfill. The results show that reuse after UF conversion and direct reuse as second hand RO membranes have statistically significantly lower CO2-e emissions and Oil-e consumption, when compared to the majority of other end-of-life scenarios. However, it is evident that membrane lifetime and transportation distances have a major impact on the results. Therefore, these parameters will be investigated in detail in the following sections through sensitivity analysis.

176 Table 5-5: Monte-Carlo simulation results for CO2-e emissions and Oil-e Consumption produced from pairwise comparison of end-of-life membrane options. Values represent

the percentage of runs where the column variable (A) has higher emission/consumption compared to its corresponding row variable (B). A result can be considered statistically significant if the direction of results was consistent for over 90%

of runs (Goedkoop et al., 2010).

CO2-e Emissions A B Incineration Gasification EAF Recycling UF Conversion Direct Reuse Landfill 100% 0% 9% 0% 2% 0% Incineration 0% 0% 0% 0% 0% Gasification 80% 20% 15% 6% EAF 1% 6% 0% Recycling 24% 7% Conversion 71% Oil-e Consumption Landfill 0% 0% 0% 0% 0% 0% Incineration 100% 19% 6% 10% 11% Gasification 1% 0% 4% 4% EAF 35% 21% 23% Recycling 28% 21% Conversion 58% A has significantly lower emissions than B A has significantly higher emissions than B Difference between A and B not significant

5.4.5. Effect of Reuse Membrane Lifetime on Reuse Viability

Comparison of the various end-of-life scenarios has shown that direct reuse of the RO membranes is a heavily favourable options with many environmental benefits. However, uncertainty analysis shows that the assumptions for the membranes initial and reuse lifespans have high variance, as they depend heavily on the application, system design, feed water quality and cleaning regimes. The estimates used in this study have been based on surveys of measured membrane lifespan and reported literature values as 177 discussed in detail in Section 2.2.3. To understand how these changes will affect the environmental impacts of the reuse scenario, a breakeven analysis was competed.

Figure 5-6 shows the trend of CO2-e emissions for direct reuse and conversion to UF scenarios with increasing secondary use lifespan, with landfill and recycling scenarios shown for comparison.

110

90

70

eemissions -

2 Landfill

50 Recycling kg CO kg Conversion to UF RO Reuse 30 0 1 2 3 4 Reused membrane lifetime (yr)

Figure 5-6: Effect of secondary use phase lifespan on viability of reuse scenarios

relative to landfill and recycling.

As expected the two reuse scenarios (direct membrane reuse and reuse after conversion to UF) display a decreasing CO2-e emission as reuse membrane lifetime increases, offsetting the production and consumption of virgin module. As reuse lifespan increases, the breakeven point at which the reuse scenario CO2-e emissions are equal to the emissions from the landfill, is the minimum reuse lifespan in which it has an environmental benefit over traditional disposal. Modules that have been converted to UF- like membranes need to be used for over 1 year to be more environmentally favourable to landfill, while membranes directly reused as RO only require 8 months of operation to break even. Similarly, if the membranes are reused as RO for less than 17 months (or

178 23 months in the case of conversion to UF) then recycling the membrane would produce less CO2-e emissions.

5.4.6. Effect of Transportation Distance and Reuse Membrane Lifespan on

End-of-life Scenario Viability

Due to Australia’s size and coastal population distribution, the various municipal SW desalination plants are located towards the coast and are a significant distance apart.

For the end-of-life scenarios where the used membranes require substantial relocation, transportation emissions have the potential to play a significant role in the scenarios environmental sustainability. Figure 5-7 shows the contribution of each scenarios transportation and how it effects the benefits of the process, and also show the difference between the transportation scenarios. The bars represent the results from the average transportation distance with the error bars representing the results from the maximum and minimum distances, as determined by surveys individually described in Section

5.3.3.

10

0

-10

emoduleemissions per - 2 Transport

-20 CO

kg kg Process

-30 Landfill Incineration Gasification EAF Recycling Conversion Direct reuse to UF

Figure 5-7: Contribution of transportation and process to the climate change (CO2-e)

emissions of the different scenarios. 179 While the reuse methods have the highest environmental benefit, they also feature the most strenuous transportation scenarios. Using the average transportation distance of

2480 km, the reuse transport scenarios contribute 3.73 kg CO2-e to the atmosphere with between 0.11 to 6.85 kg being emitted for the low and high cases respectively. Energy recovery through EAF has a similar transportation impact with an average of 2.65 kg

CO2-e being emitted with a high and low of 6.11 and 0.1 kg respectively. The scenarios only requiring local transportation of 47 km, such as landfill, incineration, and gasification have a low impact of 0.232 kg CO2-e emitted, with the recycling scenario having a slightly higher emission of 0.42 kg CO2-e.

As the transportation distance required for the reuse scenarios is highly variable, it is important to understand how this affects emissions and to calculate the maximum distance that they can be transported, while the options is still environmentally favourable. Figure 5-8 shows the effect of increasing transportation distance on the CO2- e emissions associated with the relevant reuse scenarios, contrasted to landfill disposal and recycling. Additionally, as it has already been identified that the reuse lifespan of the membrane has a significant impact on reuse sustainability, a number of membrane lifespans have been included.

180 95 Landfill

Recycling

UF - 0.5 yr 85 UF - 1 yr

UF - 2 yr emoduleemissions per

- RO - 0.5 yr 2 75

RO - 1 yr kg CO kg RO - 2 yr

65 0 1000 2000 3000 4000 5000 Reused membrane transportation distance (km)

Figure 5-8: Effect of transportation distance on the viability of reuse scenarios relative to membrane reuse lifetime. (RO: direct reuse, UF: membrane conversion then reuse).

As expected, CO2-e emissions rise with increasing transportation distance and decreasing membrane lifetime. The results show that membranes which have been converted to UF, and only last half a year, produce higher emissions than landfill at all transportation distances. If the converted membrane is reused for one year, it can be transported up to 1537 km before disposal in landfill is more environmentally sustainable.

Similarly, a directly reused membrane lasting only half a year will generate higher emissions than landfill if transported more than 1173 km. It has been identified that 4550 km is the maximum reasonable transport distance required for a membrane to be reused within Australia. Therefore, membranes directly reused which last at least 11 months and

UF converted membranes lasting at least 1.4 years, will be more environmentally beneficial than landfill at all possible transportation distances. The greater emission offsets that the recycling scenario offers means that substantially shorter distances or longer lifespans are required to make reuse comparably more beneficial.

181 Alternative transportation modes were also considered for the membrane reuse scenarios including domestic shipping and transportation by rail. Figure 5-9 displays the

CO2-e emissions offset for the direct reuse scenario using the transportation methods of road, rail and sea freight. The bars show the results for the average distance values, while the error bars show the results for the maximum and minimum distances possible.

Road Sea Rail 0

-10

eemissions

- 2

-20 kg CO kg

-30

Figure 5-9: Benefit of direct reuse of RO membranes in terms of reduction in CO2-e

emissions depending on transportation method used.

The average distance of transportation by rail for reused membranes was calculated as

2335 km with a max and min of 5035 and 730 km respectively, and the average shipping distance between ports of cities containing desalination plants was calculated to be

2400km with a max and min of 4860 and 950 km respectively (Bendall and Brooks, 2010;

Farnel Capital inc, 2013). In addition, it was assumed that the membranes require transportation of 20 km to and from the nearest train yard (or port) and the base reuse lifespan of 2.5 years was used. The results show a marginal benefit for both sea and rail transportation of the reused membranes. However, coastal shipping has significant challenges including increased transportation times and the requirement of longer distances to become economically viable (Bendall and Brooks, 2010). Overall, all three

182 transportation methods have relatively similar impacts, with sea and rail being slightly favourable over road, while road is more practical in the case of used RO membranes.

5.4.7. Effect of End-of-life Scenarios on Landfill Loading

One of the primary goals of this study is to reduce the burden that membrane disposal has on landfill, and therefore it is important to consider the waste streams of each scenario. To do this, and to avoid the difficult and inherently subjective nature of estimating land use (Baumann and Tillman, 2004), the total mass of the final waste to be disposed is used (shown in Figure 5-10). This includes components that are not suitable for the given scenario (e.g. the fibreglass case in EAF) and the residue from the processes themselves (e.g. slag from the gasification and incineration processes).

16

14

12

10

8 (kg) 6

4

2 Mass going to landfill per modulelandfillto pergoing Mass 0 Landfill Incineration Gasification EAF Recycling Direct Conversion reuse to UF

Figure 5-10: Mass of waste material requiring landfill disposal for each of end-of-life

scenarios for one RO membrane

183 As expected, the landfill scenario involves the highest mass of material to be disposed of, with the entire 13.5 kg (dry mass) module going to landfill. While the reuse scenarios do not directly require any mass to be sent to landfill, the membranes still need disposed of after the second lifespan. Therefore, the mass is calculated from the avoided membrane disposal that comes from extending the membrane useful lifetime, and as a result, the reuse scenarios have the second highest impact on landfill. This is an interesting contrast to the LCA results which demonstrated that the reuse methods are heavily environmentally favourable and highlights one of the major disadvantages of membrane reuse.

The recycling scenario still requires the disposal of 8 kg of material as 40% of the membrane module is unsuitable for this application. Similarly, the contribution from the

EAF process is mainly comprised of the disposal of the fibreglass casing, which cannot go into the furnace. The gasification scenario produces substantial residue due to the incomplete combustion taking place and the incineration process produces up to 1 kg of slag requiring landfill. The majority of the waste from the gasification and incineration processes comprise of residual silica from the fibre glass casing.

While the reuse and thus lifecycle extension of the membrane provides many environmental benefits, it does not reduce the amount of material requiring eventual disposal to the same extent as the other scenarios. If the absolute priority is the aversion of waste from landfill, over all other impacts, then incineration is the most beneficial option. This is the main reason that incineration is commonly used where land space is at a high premium, such as Singapore and Japan (Tan and Khoo, 2006).

184 5.5. Conclusions

This chapter utilised the tool of LCA to assess and compare the environmental impact of the technically feasible end-of-life scenarios for used RO membranes identified in

Chapter 2. Overall, it shows the potential environmental benefits that could be realised through an organised reuse system and encourages the development of a membrane disassembly and component recycling system for final disposal. In the following chapter, this information will be used as key inputs for a decision making tool aimed at providing membrane users with a dynamic assessment of end-of-life options.

185 Chapter 6:

Decision Making Tool for End-of-life

Membrane Users

This chapter is an expanded version of the following submitted peer-reviewed conference paper:

Lawler, W., Leslie, G., Le-Clech, P., 2015. Decision Making Tool for End-of-Life

Reverse Osmosis Membrane Users. IDA World Congress. San Diego.

186 6. Decision Making Tool for End-of-life Membrane Users

6.1. Introduction

In previous chapters, the technical viability and environmental sustainability of end-of-life options for used RO membranes has been assessed in detail, with the aim of promoting better practices in the desalination industry. To apply these findings, an online tool has been developed to help RO membrane users identify and select the optimum end-of-life option for their specific situation and environment. The tool has been named MemEoL, which stands for Membrane End-of-Life. The tool has been designed with an Australian focus and geographic environment, but the educational components are applicable globally. This tool forms part of the National Centre for Excellence in Desalination

Australia’s (NCEDA) Desal Wiki project, which is aimed at providing operators and the general public a platform to learn, share and communicate knowledge and data regarding desalination technologies.

The aim of this chapter is to describe how the tool has been developed and the algorithms used to provide recommendations to users.

Note on website:

At the time of printing the MemEoL tool can be found at the following URL: http://www.desalwiki.che.unsw.edu.au/w/index.php/Membrane_end-of- life_(MemEOL)_Tool

Note that MemEoL is being constantly updated and improved, and therefore may not exactly match the original tool described in this work. Also, as the tool becomes more developed it may move to a dedicated website. If the above URL is not valid, please attempt a web search for MemEoL.

187 6.2. MemEoL Tool Operation

The MemEoL tool, as seen in Figure 6-1, is comprised of a number of user interaction steps, and background calculations based of these inputs. The tool algorithm uses these inputs to rank the end-of-life options in order of preference.

Membrane performance data from user

Physical Performance still Reuse membrane Assess reuse membrane within usable for secondary NO YES quality damage? range? application

YES

User for Convert and NO converted UF YES reuse membrane available? as UF

Compute User ranks optimum solution Assessment NO based on user Criteria weightings

Use alternative end-of-life option

Local landfill

Figure 6-1: Structure of MemEoL tool.

When using the tool, the user is prompted to provide information about condition of the membranes. These questions can be seen in the Figure 6-2, which is a capture of the input screen of the MemEoL tool. The tool also provides links to educational information on the measurement of these parameters.

188

Figure 6-2: Capture of MemEoL input screen.

189 The user is also prompted to rank five assessment criteria, which will be discussed in

Section 6.4, in order of importance to their specific end-of-life program. The tool then uses the data on performance and damage to assess if the membranes are within a suitable range for direct reuse, or chemical conversion and reuse. If the membrane is deemed unsuitable for reuse given the parameters provided, the option is removed from the decision making process. Following these considerations, the tool provides a number of end-of-life options in order of favourability and provides information on these options, along with contact details for providers. If the membranes are deemed suitable for reuse, then further information about their potential applications is provided.

6.3. Multi-Criteria Decision Analysis

The tool is based off a discrete multi-criteria decision analysis (MCDA) system, which is a powerful tool for modelling and solving such problems (Munier, 2011). Figure 6-3 shows the main steps that have been followed to solve this MCDA problem.

190 Identify the decision problem and the goals

Gather information and Identify criteria

Identify alternatives

Evaluate each alternative with respect to the criteria

Weight criteria

Identify prefer alternative and make decision

Figure 6-3: Steps for solving a discrete MCDA problem. Adapted from (Zarghami and

Szidarovszky, 2011).

The strength of MCDA is that criteria measured in different units, and both qualitative and quantitative information, can be used (Soltani et al., 2014). Specifically, the quantitative results such as environmental impact from an LCA and the qualitative information such as public acceptance and project complexity can be combined in a single analysis. The specific MCDA method used is simple additive weighting (SAW), which is also known as the weighted sum method (WSM), and is a subset of multi- attribute utility theory (MAUT) (Zarghami and Szidarovszky, 2011). In this method, each criteria is given a weight, and the sum of all the weights must equal one. Each alternative i is assessed against the selected criteria j, and the overall performance score P of the alternative i is calculated using Equation 6-1.

191 푛

푃푖 = ∑ 푤푗푎푖푗 Equation 6-1 푗=1

Where w is the weight for criteria j, and aij is the normalised assessment score for criteria j and alternative i. The assessment scores are scaled with a min-max approach, using

Equation 6-2.

performance − min peformance of group normalised assessment score = 1 − max performance of group − min performance of group

Equation 6-2

All the performance values used in this tool are negative, meaning that a higher value is

less favourable in the decision making process. For example, higher CO2 emissions or cost is less desirable over lower values. This scaling process converts the range of performances of the end-of-life options relative to the assessment criteria to a score between 0 and 1, which can be used by the MCDA tool. For this tool, a value of 0 is the least favourable outcome, and a value of 1 is the most favourable outcome.

The final preparation step in the MCDA process is assigning weights to the assessment criteria. Traditionally, the weights of different criteria are determined through iterative collaboration from the project stakeholders. The MemEoL tool, however, uses a novel dynamic weighting method, to cater for users with different needs. This method relies on the internet based nature of the tool, allowing the user to ranks the criteria in order of importance to their specific end-of-life program. The ranks are converted to weights using a rank order centroid method, using Equation 6-3, where n is the number of criteria used in the tool and m represents the rank of the given criteria j.

푛 1 1 푤 = 푥 ∑ Equation 6-3 푗 푛 푚 푚

192 6.4. Assessment Criteria

The MemEoL tool uses simple, non-overlapping, criteria to assess the performance of the end-of-life alternatives for used membranes. The criteria have been selected to reflect parameters significant to the user, and that facilitate comprehensive and meaningful assessment of the end-of-life options, based on the goals of the tool. The assessment criteria tree, which can be seen in Figure 6-4, shows the criteria used in the categories of environmental, economic and social impacts.

Categories Primary Critera

Environmental Impact Environmental

Landfill Impact

Financial Impact Economic Project complexity

Social Public Perception

Figure 6-4: Assessment criteria tree for decision tool.

The following sections outline how a literature review, and data generated in the previous chapters are used within the MemEoL tool. The assessment scores for the quantitative criteria (environmental and landfill impact) are generated using data from previous chapters in this thesis, and weighted with the min/max method discussed in Section 6.3.

The assessment scores for the qualitative criteria (financial impact, project complexity and public perception) are determined using the assessment matrix (Table 6-1), and the information outlined in the following sections. The numerical scales of this assessment matrix are positive, with 1 being the most favourable and 0 being the least favourable.

193 Together, these assessment scores are displayed in a normalised decision matrix (Table

6-2), and used directly as assessment scores aij in Equation 6-1.

Table 6-1: Qualitative criteria assessment matrix.

Medium Medium Scale Low Medium High low high name Numerical 0 0.25 0.5 0.75 1 Scale Partial or Expensive No additional Low return on complete Financial process or cost compared Small net cost membrane return on investment, to landfill Impact cost membrane large net cost. disposal purchase High amount Significant Process Small amount of effort time available but of process Process is required. New commitment to development development Project ready to research to be set up, new required to required. Low receive Complexity done or large collaboration make it man hour membranes. team to be and process suitable for commitment organised adaption membranes for user Highly Negative Indifferent Positive public Highly positive negative public public opinion, public opinion, opinion. public opinion Public opinion process seen no public Process seen towards the towards option Perception as dirty or relations as socially option and low with high polluting. benefit. responsible. social impact social impact

6.4.1. Environmental

The environmental results for the MemEoL tool are derived from the LCA detailed in

Chapter 5. To further distil the findings of the LCA, the results have been aggregated using the ReCiPe endpoint hierarchist method, with a world perspective normalisation and an average weighting set (Goedkoop et al., 2008). An endpoint LCA method combines all midpoint impact categories and attempts to quantify environmental concerns, such as impact of human health, extinction of species and rise in seawater level in a single comparative score. These endpoint scores are then normalised using the min-max approach using Equation 6-2. The landfill impact criteria has been directly adapted from the Section 5.4.7, and are also normalised using the min-max approach.

194 6.4.2. Economic

The financial criteria are developed from a justified qualitative assessment, based on the cost or return for the primary membrane user. As the MemEoL tool is generic in both geolocation and technology, a complete cost/benefit analysis is not currently viable. The cost of landfill, the current standard disposal practice, is used as a reference point for direct comparison. Project complexity is a qualitative assessment of the amount of organisational time and resources required to set up an end-of-life program with the given scenario. While these are subjective and qualitative assessments, the goal of the

MemEoL tool is to provide rapid initial information about the available end-of-life options.

If a large number of membranes are to be disposed of, or a permanent program is to be set up, a detailed cost/benefit analysis with specific information can be added to the tool.

A survey of commercial landfill disposal costs around Australia revealed an average cost of $200 AUD/tonne with range of $75 – 329 AUD/tonne (roughly $3 dollar per membrane), depending on location, size and technology of the landfill site (See Table C-

10 Appendix C for more information). Additionally, the landfill scenario has the lowest transportation costs, as local facilities can be used. When compared to landfill, it has been shown that incineration can cost up twice as much for the user, despite the energy recovery potential (Beukering et al., 2014; Ecocycle, 2011). The abundance of compatible waste tyres for the EAF energy recovery processes means that there is minimal market value for membrane components (Zaharia et al., 2009b), and the necessity of disassembly to remove the fibreglass outer casing and relatively large transportation distances increases the cost of this process. Landfill has the lowest project complexity of any of the scenarios, as it is the current standard practice and requires no additional effort to establish. Incineration has extremely high complexity, as there is currently very few incineration facilities in Australia, and legislation prohibits their use for general solid waste. EAF energy recovery has moderate complexity, as this

195 process is currently being used with other materials, but individual collaborations with the involved companies will have to be established.

It has been reported that for gasification energy recovery to be commercially viable, the feedstock plastic waste needs to be obtained for free (Ricardo-Aea, 2013). Another report states that disposal of plastic waste costs 50 USD/ton using gasification, which is comparable to US landfill at 30-75 USD/ton, however this value does not include the sale of product energy (American Chemistry Council, 2012). Therefore, it is unlikely that there will be any market value for used membrane in this application. Gasification has a moderate project complexity, as there are an increasing number of companies using this technology in Australia (Prince et al., 2011).

The economics behind material recycling is dynamic, with high sensitivity to the local conditions, cost of processing, and transportation (Beukering et al., 2014). The value of the recycled material is also a large contributor, which, for plastic recycling, also fluctuates with the price of oil (Foster, 2008; Hopewell et al., 2009). Therefore, it is challenging to determine net financial cost of a recycling program without assessing a specific case. However, from the waste disposers’ perspective, a survey of waste processing facilities in Australia, showed that they generally accept recyclable plastic material for no charge, or at a significantly reduced price compared to landfill (See Table

C-10 of Appendix C for more information). However, these materials have to be properly separated, creating a significant challenge for membrane users, as disassembly and sorting is required, increasing the cost. This disassembly step not only increases the cost, but also the project complexity, as it will have to be organised by the membrane users or a third party.

As previously discussed in Section 2.5.1, direct membrane reuse has already seen application in the United States (WaterSurplus, 2014). The used membranes are still considered valuable products and are valued at an average of 34% of the new membrane price, depending on condition and previous application. Membranes 196 converted to UF membranes will have considerably less value, and will require additional treatment steps, decreasing the return on original purchase. Another factor to consider is the transportation distances involved in these scenarios, which have been shown to be significant in Australia, and would greatly increase the cost. The process of reusing

RO membranes is quite simple, and can be further streamlined by an organised reuse program. On the other hand, the additional treatment steps required to reuse membranes converted to UF makes this end-of-life options significantly more complex.

6.4.3. Social

Public perception is a qualitative measure of the general acceptance of the various waste management options. This includes the general public view of the social and environmental impacts, as well as the impact on local communities and job creation. This is an important consideration for membrane users as a positive public perception can simplify the implementation of large scale projects and increase the company reputation

(McDougal et al., 2008).

Public opinion on the construction of new landfill sites has generally been negative, with a “not in my backyard” attitude, and with the diversion of waste from landfill seen as a positive (Eltridge, 2005). However, as sending waste to landfill is the current practice and the most common national waste disposal technique, little public resistance or pressure is expected. Therefore, the perception of social responsibility can be gained from using an end-of-life option with a more positive public perception.

Generally recycling has a highly positive public perception in Australia, with some minor concerns about the amount of waste being exported for processing and the details of what the process involves (Ha and Santucci, 2006). Setting up a membrane processing centre for the disassembly and sorting of membrane components also has the additional social benefit of creating jobs. Reuse techniques are expected to have a positive public

197 perception, with potential concerns about quality of water produced with reused membranes, and the processing requirements of chemically converted membranes.

Incineration and other waste to energy technologies generally have a negative public perception due to early facility operations that were not subject to rigorous environmental controls, and their use has remained controversial (Fanning, 2013). The public perception of waste recovery technologies like gasification varies, based on previous exposure to incineration technologies and public engagement, but is for favourable over incineration (GBB, 2013).

6.5. Alternative Decision Matrix

Following the steps for solving a discrete MCDA problem as shown in Figure 6-3, the criteria have been developed, and the alternatives have been identified and characterised. The next step is to evaluate each alternative with respect to the criteria.

Using the information outlined in the previous sections and quantitative data developed in other chapters, each assessment score has been assigned and normalized. The results of this process can be seen in the normalised decision matrix (Table 6-2). This information which is directly used within the MemEoL tool, can be better visualised for comparison in Figure 6-5.

Table 6-2: Normalised decision matrix.

Inciner Gasifica Material Conversion Reuse Criteria Landfill EAF ation tion Recycling to UF as RO Financial 0.5 0.25 0.25 0.25 0.25 0.5 0.75 Impact Project 1 0 0.5 0.25 0.5 0.25 0.75 Complexity Environmental 0.07 0 0.32 0.39 0.53 0.85 1 Impact

Landfill Impact 0 1 0.75 0.93 0.43 0.32 0.32 Public 0.25 0 0.5 0.5 1 0.5 0.75 Perception

198 Financial Impact 1 Landfill 0.75 Incineration 0.5 Public Project Gasification Perception 0.25 Complexity EAF 0

Recycling

Conversion to UF Landfill Environmental Direct RO Impact Impact Reuse

Figure 6-5: Relative assessment scores of end-of-life scenarios across assessment

categories. (0) least favourable to (1) most favourable.

Figure 6-5 allows for the comparison of the various end-of-life options for used RO membranes. A number of observations can be made, which will provide insight into the outcomes provided by the decision making tool. Direct RO reuse is highly favourable in many of the considered categories, indicating that it will be preferred for a number of combinations of user input criteria rankings. Additionally, there are no assessment criteria in which the conversion to UF is preferable over direct RO reuse, therefore it will only be recommended when direct reuse is not feasible, or presented as a secondary option. However, if the user indicates that physical damage is present, or the membranes are not within a usable performance range (as discussed in the following sections), the tool will assess alternative options. As previously discussed in Chapter 6, the incineration, gasification, and EAF scenarios will be favoured if reduction in landfill

199 waste is the users priority. Sending the membranes to landfill is the option with the lowest project complexity, but is extremely unfavourable in all other categories.

If the tool determines that direct membrane reuse is the preferred end-of-life option, it will then use available information to classify the membrane condition and recommend potential applications. To do this however, the impact of membrane performance on reuse suitability needs to be assessed.

6.6. Assessment of Membrane Reuse Quality

The primary criteria for determining the suitability of direct reuse is membrane performance. Unless the modules are surplus or have been removed due to plant shutdown or mothballing, their performance will have deviated from the original specifications. RO membranes come in many shapes and sizes, and are designed for varying applications, resulting in a wide range of available performances. While performance will change during the membranes lifespan, it could still be within a usable range for another application. Additionally, with a supply of lower cost membranes, operators will be able to run them in harsh operating conditions with less concern about cleaning or lifespan. The first step in assessing used membrane suitability is to benchmark the current performances of RO membranes. Figure 6-6 shows the results from a survey of nearly 100 RO membranes currently available from five of the leading manufacturers, DOW, Toray, Hydranautics, CSM and GE.

200 100.0 SWRO 99.8 BWRO

99.6

99.4

99.2 NaCl rejection (%) rejection NaCl 99.0

98.8 0 2 4 6 8 -2 -1 -1 Permeability (L.m .h .bar )

Figure 6-6: Performance of new RO membranes from five of the major manufacturers.

The membranes have been assessed with a standard testing method, involving a feed solution of 32,000 ppm NaCl and pressure of 55 bar for SWRO, and 2,000 ppm at 15.5 bar for BWRO. SWRO membranes have performances ranging between 99.6 - 99.9%

NaCl rejection and a permeability of 0.45 – 0.94 L.m-2.h-1.bar-1. While BWRO membranes have a performance that ranges from 98.5 – 99.8% NaCl rejection and a permeability of

1.3 – 7.4 L.m-2.h-1.bar-1. Overall, there is a trade-off between salt rejection and permeability, with an increase in productivity resulting in a decrease in product quality.

Additionally, RO membranes can be modified for a variety of requirements, including high productivity, high rejections, low energy, low fouling and high boron rejection. For comparison, NF membranes by the same manufactures have a NaCl rejection between

55 – 90% with a permeability of between 6.3 and 8.6 L.m-2.h-1.bar-1.

Over the course of this project, a wide number of used RO membranes have either been tested directly, or assessed from literature. By excluding the membranes that have a

201 physical reason that they cannot be reused, such as damage or improper storage, a catalogue of end-of-life membranes has been made. The NaCl rejection and permeability for each of these membranes has been calculated and compared to the performance of a virgin sample. Figure 6-7 shows the salt rejection plotted against the permeability of both the virgin and used samples. The arrow for each membrane leads from the virgin to the used sample, and shows the performance variation over its lifespan.

Superimposed on the figure is the approximate performance range of new membranes, as determined from a survey of commercially available membranes from the major manufactures shown in Figure 6-6.

100 BWRO 98 SWRO

96

94

92

90

88 NaCl rejection NaCl (%) 86 NF 84

82 0 2 4 6 8 -2 -1 -1 Permeability (L.m .h .bar )

Figure 6-7: NaCl rejection and permeability performance degradation for end-of-life membranes (origin of the arrows indicates the initial performances, while the end of the

arrow shows the end-of-life performance).

The results show that membranes studied have been affected in many different ways.

The most common trend for performance change is a decrease in both permeability and salt rejection, most likely due to a combination of fouling, chemical damage and

202 compaction (see Section 2.2.3 for further information). Membranes that have experienced an increase in permeability and a decrease in rejection have likely undergone significant chemical damage. It should be noted that some of these membranes have undergone fouling during their primary lifetime, and their performance could be potentially improved using basic cleaning techniques. From the performances of both the new and used membranes, classification criteria for membrane quality can be created. Table 6-3 outlines the five membrane quality categories developed for the

MemEoL tool and outlines the predicted applications and estimated secondary lifespan.

These performance categories can be dynamically adjusted when more information about the requirements for reused membranes is obtained through the use of the

MemEoL tool.

Interestingly, it would be ideal for used RO membranes to exhibit increased permeability and decrease salt rejection, as it would enable them to be simply reused as NF membranes in a secondary applications. However, as seen in Figure 6-7, this trend has not been observed. This is possibly due to a number of factors, including irreversible fouling, membrane compaction, and/or partial drying during storage and transportation.

203 Table 6-3: Performance categories for membrane reuse

Estimated NaCl Permeability reuse rejection range Designation Action -2 -1 -1 lifespan range (%) (L.m .h .bar ) (yr) Direct reuse as High quality 99.9 - 99.6 > 0.45 SWRO possible in 2 –5 SWRO normal applications Direct reuse as High quality 99.7 - 99.2 > 1.6 BWRO possible in 2 – 3 BWRO normal applications

Direct reuse as Medium 99.2 – 98 > 1.6 BWRO in standard 1 – 2 quality BWRO applications possible

Direct reuse as BWRO in harsh Low quality 98 – 96 > 1 applications where 1 BWRO regular replacement is required Medium Direct reuse as NF 96 – 80 > 5 - quality NF membrane possible

Unsuitable for Membrane suitable < 96 < 5 - RO or NF for UF conversion.

204 6.7. Conclusion

The MemEOL tool has been developed to promote better practices in the desalination industry by helping users to identify and select the optimum end-of-life option for their used RO membranes. The interactive online tool, which is part of a larger Desal Wiki providing a wide range of information for desalination researchers and operators, is based on a dynamic MCDA system. Users are able to provide the performance of their membranes and indicate the relative importance of a number of key criteria to their recycling/disposal program. Based on this generic assessment, the tool provides preliminary recommendations and further educative information about the available end- of-life options.

This internet based tool is primarily an educational resource that uses qualitative and quantitative data to rank the end-of-life options. This chapter has outlined the MCDA methodology, identified relevant criteria and evaluated each end-of-life alternative with respect to these criteria. It has also allowed for the information developed in this study to be summarised and communicated to relevant parties. Additionally, the reuse quality and potential applications of RO membranes has been categorised based on testing of end-of-life modules, which the tool uses to make reuse recommendations.

205 7. Conclu sion an d Future W ork 7. Conclusion and Future Work

7.1. Conclusion

The aim of this thesis was to identify and assess the alternative end-of-life options for used RO membrane elements following the waste management hierarchy, with the goal of increasing the sustainability of the desalination industry. A number of viable end-of- life options were identified including, direct reuse of the old membranes within lower throughput systems; chemical conversion into porous, UF like filters; direct reuse or recycling of the various module components; various energy recovery techniques, and landfill disposal. Following precedent from other industries, it was concluded that a consolidated effort by relevant stakeholders, including membrane users and manufactures is required for a successful large scale end-of-life program. The Product

Stewardship Bill Act developed by the Australian Government allows for the desalination industry to seek voluntary accreditation for a reuse and recycling program for used RO membranes.

Overall, this work represents a significant step in providing membrane users with alternative end-of-life options to landfill disposal. Out of this work, three major outcomes were realised and they will be further discussed and summarised in the following sections.

7.1.1. Mechanisms for Membrane Performance Decline upon Drying

The first major challenge identified for the direct reuse of RO membranes, one of the most promising end-of-life scenarios, was the requirement for proper preservation and storage. The work presented in Chapter 3 investigated the impact of drying on performance for commercially available RO and UF membranes, and optimised various rewetting strategies to facilitate the reuse of membranes that have been improperly stored. While it is generally recognised that precaution should be taken to prevent

206 membranes from drying, the precise mechanisms for the performance loss had seen limited investigation.

This effect was systematically investigated using selective drying, synchrotron SAXS analysis, SEM, AFM and rejection characterisation. As a result of this study, it is proposed that the PWP decline in dried membranes is caused by pore collapse in the skin and subskin layers of UF and RO membranes made of PSf and PES. As water evaporates from the membrane, the resulting capillary forces acting on the pores pull the walls together, thus decreasing the pore size and restricting flow. Key evidence for the mechanism proposed in this thesis is summarised in Table 7-1. This evidence is particularly robust for commercially available UF membranes of PES constructions.

However, with the similarities in drying and rewetting behaviour, and with evidence from the reviewed literature including manufacturers, it follows that the mechanisms are similar for RO PSf layers. These findings bring new light to an important practical challenge in membrane research and operation, and is a critical factor in the successful reuse of end-of-life RO membranes.

For RO membranes, drying had no impact on salt rejection; however, for UF it caused an increase in protein and humic substance rejection. This indicates a significant decrease in mean pore size due to pore collapse. It was also observed that PWP loss for both RO and UF membranes only occurs after 90-95% of the water has evaporated.

This means that only extreme drying of stored membranes will lead to irreversible damage. It also suggests that the pore bound water is the last to evaporate, due to the higher energies required. As the 10 and 30 kDa UF membranes lost up to 100% of initial

PWP upon drying, while the 100 kDa membrane lost only 51%, it can be concluded that pore collapse is dependent of initial mean pore size.

207 Table 7-1: Summary of results regarding membrane drying and pore collapse.

Investigated Observed in RO Observed in UF Method Effect Membranes membranes 0 – 1% of virgin PWP after complete drying, except for 100 kDa UF due to large pore size. Membrane PWP Impact of drying Controlled loss only occurs after 90-95% of water evaporated, and on membrane drying in took up to 85 min to completely dry in a desiccator. performance desiccator Membranes curl towards active layer causing delamination. Significant morphology changes, Pore collapse on N/A due to PA layer increase in roughness for UF AFM active layer side covering PSf layer. membranes and closure of pores. Reduction in pore size and Impact of drying Preliminary results Dynamic volume due to drying. Strong on internal suggest similar drying in evidence for structural change of membrane mechanism to PES SAXS pores with 10 kDa UF structure UF membranes. membranes. 100 kDa UF show increase in Rejection Impact of drying No change in salt protein and humic substance testing after on rejection rejection. rejection after drying due to drying changes in pore structure. Rewetting in 50% ethanol for 15 min and 0.25 g.L-1 SLS for 50 h showed the highest performance recovery out Varied of the tested methods. Membrane swelling volume Performance exposure decreased after drying, but increased after subsequent recovery using methods rewetting due to delamination. RO membranes with various rewetting and lower initial PWP displayed higher PWP recovery upon strategies durations rewetting due to the resistance dominance of the PA layer. Ethanol rewetting strategies proved effective on used membrane modules.

208 It was shown that soaking in a 50% w/w ethanol solution for 15 min is an optimum method for rewetting dried membranes; however, comparable performance recovery was also achieved by soaking in the common surfactant SLS for a minimum of 50 h. The manufacturer recommended rewetting methods of soaking in 1% w/v HCl and 4% w/v

HNO3 did not result in adequate performance recovery compared to the aforementioned methods. Due to the resistance dominance of the PA layer, RO membranes show significant PWP recovery, while low MWCO UF membranes show only minor recovery, relative to virgin performance. However, it is theorised that the pores of the PSf layer of rewetted RO membranes remain restricted, potentially leading to more rapid subsequent performance loss. Additionally, the drying process seriously compromises the structural integrity of the membrane construction through contraction of the PSf/PES layers, causing delamination.

7.1.2. Feasibility of RO Reuse Following Chemical Conversion

A variety of different oxidative agents were assessed for their ability to remove the PA layer from RO membranes. It was demonstrated that NaOCl was the most successful conversion agent, and a number of characterisation techniques were applied to the resulting PSf based membranes to determine their suitability for reuse in UF applications.

Large variation in PWP from the converted membranes was recorded, from 8 - 400 L.m-

2.h-1.bar-1, but all displayed similar rejection performances, comparable to 10 and 30 kDa

UF membranes.

Tests were conducted to determine how different storage and pre-treatment conditions affected the extent of the conversion; invariably, all membrane types tested benefited from pre-wetting, indicating that membrane moisture plays an important role in the degradation process. From this it has been concluded that dried membranes are not suitable for conversion, meaning that this end-of-life option is not suitable for improperly stored membranes. Tests show that the conversion process can be successfully applied 209 to membrane modules in a spiral wound format with consistent performance across the membrane surface after autopsy. However, modules that had been stored incorrectly showed poor converted performance. Preliminary results also suggest that the PSf layer is not significantly damaged by the extreme chlorine exposure.

A possible application of this treatment method is to use controlled exposure with NaOCl to partially remove the active layer, trading increased PWP for decreased salt rejection, resulting in a NF or high flux BWRO membrane. Based on work shown here and previously published (Raval et al., 2012), it has been concluded that this controlled removal of the active layer is possible but challenging. This is due to the varied response of different membrane types to the conversion process and the rapid removal of the active layer as exposure time increases. Implementation of this technique would require the testing of optimum exposure for each membrane type and is unlikely to lead to consistent flux and rejection performance.

Despite the promising results obtained in this study, inconsistent performance across different membrane types, requirement for high pressure pre-wetting and high chemical doses may limit the feasibility of large scale implementation. Each batch of membranes to be reused requires individual characterisation and assignment to a suitable application, and the non-sterile nature of the converted membranes excludes their use in pharmaceutical and food processing applications. However, the supply of converted

RO membranes as inexpensive UF membrane replacements for use in low cost humanitarian water treatment projects, low cost RO pre-treatment, or wastewater treatment operations is a possible benefit to the industry.

210 7.1.3. Assessment and Comparison of End-of-life Options

Following the identification and technical assessment of available end-of-life options, this thesis aimed at comparing them based on environmental, economic and social impact.

The results of the comparative LCA completed in Chapter 5 closely follow the outcomes predicted by the waste management hierarchy, with direct membrane reuse being the most favourable and landfill the least favourable option, in terms of CO2-e emissions and oil-e consumption. In terms of aversion of waste from landfill, incineration results in the least mass requiring disposal, while the membrane reuse scenarios feature the greatest impact. Although the most environmentally favourable option, due to its virgin membrane production offset, it is clearly highlighted that membrane reuse does not target the actual disposal of the membrane module, and is highly dependent of reuse lifespan. However, membranes that are reused for at least one year can be transported to any plant in

Australia, while still remaining environmentally favourable when compared to landfill.

Of the end-of-life scenarios that result in module disposal, and are available in Australia, recycling shows the greatest environmental benefit. However, as many complex composite materials are used in membrane construction, only 40% of the components are suitable for recycling, with the remaining parts still requiring landfill disposal. This identifies an opportunity for membrane manufacturers to use more recyclable materials and assembly methods during construction, thus simplifying the disassembly and recycling process. The use of membrane components in EAF is also a relatively favourable options, particularly for the minimisation of waste volume. The sensitivity and uncertainty analysis conducted in this study has identified the role that membrane reuse lifetime and transportation distance have on the difference between end-of-life scenarios, and has led to an increase in the confidence in the LCA findings.

While this study highlights the important issue of end-of-life membrane disposal, it also reinforces the concept that membrane manufacturing and replacement plays an

211 insignificant role when calculating the total emissions from a municipal desalination plant.

However, this only highlights high electricity consumption of the desalination process.

Additionally, no significant benefit in terms of environmental impact was observed when using the larger 16” membrane modules. The membrane manufacturing model developed in this study can be directly applied in future LCA research involving RO desalination plants by replacing common simplified placeholder models for membrane construction and replacement.

While the study focuses on the Australian geographical setting, the principal findings can be easily applied to the international context. In addition to the relevance for RO reuse and recycling, this LCA study is a valid comparison of traditional disposal scenarios for all plastic solid waste. Thus, it can assist in future studies and policy making on general waste management decisions.

The information generated in the LCA study, and a preliminary assessment of economic cost and social impact of the various end-of-life scenarios, was combined in a dynamic online decision making tool. The tool, which is part of the NCEDA’s Desal Wiki project, has been named MemEol. It uses experimental data and user input to provide rapid preliminary assessment of various available end-of-life options for used RO membranes.

Thus helping educate membrane users and promote sustainable practices within the desalination industry.

212 7.2. Future Work

This multidisciplinary project covered a wide variety of topics, from mechanic changes within membrane structures, to collaborative online tools aimed at providing users with practical information. As a result, this work has generated a large number of pathways for future research.

The next step is to collaborate with key stakeholders in the Australia desalination industry, including the NCEDA, to set up a pilot direct membrane reuse program. To begin this, the online MemEOL tool should be expanded.

While this tool is already a valuable source of information for end-of-life membrane users, it can be further enhanced with the feedback from users and through the development of partnerships with operators of end-of-life programs. Through this collaboration, a more detailed cost/benefit analysis, and a thorough assessment of social impacts can also be completed. In future work, the following parameters could be assessed for their impact on the viability of end-of-life options and integrated as additional inputs within MemEoL tool. These parameters include:

 Location of end-of-life membranes  Number of membranes available  Membrane type  Membrane age  Primary membrane application  Membrane autopsy results

The MemEoL tool can also be extended to include a database of membranes that are available for reuse. Users with a supply of reuse suitable membranes will be able to add information about them to the database, and secondary users will be able to search for them based on required performance criteria and the two users will be connected. This will streamline direct RO membrane reuse, making it easier and cheaper to implement.

213 In addition to this consorted industry effort, there are a number of key future scientific research activities that have been identified. This include the expansion for comparative

LCA, which can be updated and expanded when further end-of-life options are identified, or when more specific, local data becomes available for existing options. Additionally, the detailed LCA on membrane manufacturing can be directly used as an input in future studies looking at the sustainability of desalination processes. These values will replace the current simplified placeholders that are commonly used in this type of work, thus increasing their accuracy and applicability.

In terms of other end-of-life scenarios, such as material recycling and energy recovery in EAF, further practical study into the energy requirements and costs of membrane disassembly, crushing, and washing is required. Preliminary results suggest that the use of membrane components as a coke substitute in EAF steelmaking is a viable end-of-life option. However, a study of the effectiveness of a combined blend of membrane materials is required so that component separation can be avoided, and a pilot scale trial is needed for further validation.

To further the investigation into membrane drying, work should focus on the impact of drying on PSf based UF membranes, and membranes of different geometries, such as hollow fibre. Of particular interest are the PSf layers used in the production of RO TFC membranes, which unfortunately were unable to be obtained for the current study. The lessons learnt during the current work about SAXS analysis methods to membrane characterisation can be further applied. An expanded systematic study by dynamically drying different membrane types at low q-ranges, potentially with a variety of different solvents, could provide valuable further insight into the structural changes occurring during drying, and differences between membrane constructions in general. To this end, the utilisation of neutron scattering could provide further contrast between the solvent and membrane materials, and wide angle x-ray scattering can be used to see smaller pore features. As even the most promising rewetting strategies assessed in the current

214 work did not completely restore initial PWP for the majority of membranes. Further development of these techniques, potentially with the use of extremely low surface tension rewetting agents, could prove valuable to the desalination industry.

The next step in the validation of using converted RO membranes in UF applications is the design and operation of a pilot scale plant, potentially as pre-treatment for further RO processes. This would allow for the validation of the conversion process on full sized 8” modules, and assess if the narrow feed channels of RO modules are susceptible to increased clogging or fouling with high suspended solids feeds. A critical factor that was not systematically addressed in the current work is the changes in mechanical properties

(such as tensile strength) for PSf layer following extreme NaClO exposure. It has also been identified that the presence of transition metal ions can effectively catalyse the degradation process (Cran et al., 2011); therefore, future work could apply this principle to RO to UF conversion, potentially decreasing the extreme exposure time or dose requirements.

215 8. Referen ces 8. References

A’Vard, D., O’Farrell, K., 2013. 2011-2012 National Recycling Survey Final Report. Plastics and Chemicals Industries Association, Carlton, VIC.

Abdullah, S.Z., Bérubé, P.R., Horne, D.J., 2014. SEM imaging of membranes: Importance of sample preparation and imaging parameters. Journal of Membrane Science 463, 113–125.

Akin, O., Temelli, F., 2011. Probing the hydrophobicity of commercial reverse osmosis membranes produced by interfacial polymerization using contact angle, XPS, FTIR, FE-SEM and AFM. Desalination 278, 387–396.

Al-Salem, S.M., Lettieri, P., Baeyens, J., 2009. Recycling and recovery routes of plastic solid waste (PSW): A review. Waste Management 29, 2625–2643.

Ambrosi, A., Tessaro, I.C., 2013. Study on Potassium Permanganate Chemical Treatment of Discarded Reverse Osmosis Membranes Aiming their Reuse. Separation Science and Technology 48, 1537–1543.

American Chemistry Council, 2012. Environmental and Economic Analysis of Emerging Plastics Conversion Technologies. RTI International, Washington, DC.

Ang, W.S., Elimelech, M., 2009. Optimization of chemical cleaning of organic-fouled reverse osmosis membranes. US Department of the Interior Bureau of Reclamation.

Ang, W.S., Tiraferri, A., Chen, K.L., Elimelech, M., 2011a. Fouling and cleaning of RO membranes fouled by mixtures of organic foulants simulating wastewater effluent. Journal of Membrane Science 376, 196–206.

Ang, W.S., Yip, N.Y., Tiraferri, A., Elimelech, M., 2011b. Chemical cleaning of RO membranes fouled by wastewater effluent: Achieving higher efficiency with dual- step cleaning. Journal of Membrane Science 382, 100–106.

Ann, J.M., W, M.J., 1972. High flux semipermeable cellulosic derivative membranes with dry handling capabilities and their method of preparation. US Patent 3666508.

Anne, C.O., Trebouet, D., Jaouen, P., Quemeneur, F., 2001. Nanofiltration of seawater : fractionation of mono- and multi-valent cations. Desalination 140, 67– 77.

Antony, A., Blackbeard, J., Angles, M., Leslie, G., 2013. Non microbial indicators for monitoring virus removal by ultrafiltration membranes. Journal of Membrane Science 454, 193–199.

Antony, A., Blackbeard, J., Leslie, G., 2012. Removal Efficiency and Integrity Monitoring Techniques for Virus Removal by Membrane Processes. Critical Reviews in Environmental Science and Technology 42, 891–933.

216 Antony, A., Fudianto, R., Cox, S., Leslie, G., 2010. Assessing the oxidative degradation of polyamide reverse osmosis membrane—Accelerated ageing with hypochlorite exposure. Journal of Membrane Science 347, 159–164.

Aremu, M., 2011. Physicohemical properties of well, borehole and stream waters in Kubwa, Bwari area councial, FTC, Nigeria. Electronic Journal of Environmental, Agricultural and Feed Chemicatry 10, 2296–2304.

Arena, U., 2012. Process and technological aspects of municipal solid waste gasification: A review. Waste management 32, 625–39.

Arena, U., Mastellone, M.L., Perugini, F., 2003. The environmental performance of alternative solid waste management options: a life cycle assessment study. Chemical Engineering Journal 96, 207–222.

Arkhangelsky, E., Kuzmenko, D., Gitis, V., 2007. Impact of chemical cleaning on properties and functioning of polyethersulfone membranes. Journal of Membrane Science 305, 176–184.

Arnal, J.M., Sancho, M., Verdú, G., Lora, J., Marín, J.F., Cháfer, J., 2004. Selection of the most suitable ultrafiltration membrane for water disinfection in developing countries. Water 168, 265–270.

Arnal, J.M.A., Fernández, M.S., Martín, G.V., García, J.L., Zafrilla, J.M.G., Candela, J.I., Peiró, I.F., Martínez, I.T., 2002. Design and construction of a water potabilization membrane facility and its application to the third world countries . Preliminary tests. Desalination 145, 305–308.

Arnal, J.M.Ã., Verdu, G., Lora, J., 2008. Ultrafiltration as an alternative membrane technology to obtain safe drinking water from surface water : 10 years of experience on the scope of the AQUAPOT project. Desalination 251, 34–41.

Arostegui, A., Sarrionandia, M., Aurrekoetxea, J., Urrutibeascoa, I., 2006. Effect of dissolution-based recycling on the degradation and the mechanical properties of acrylonitrile–butadiene–styrene copolymer. Polymer Degradation and Stability 91, 2768–2774.

Assamoi, B., Lawryshyn, Y., 2012. The environmental comparison of landfilling vs. incineration of MSW accounting for waste diversion. Waste management (New York, NY) 32, 1019–30.

Australian Bureau of Statistics, 2010. 4613.0 - Australia’s Environment: Issues and Trends. Available at www.abs.gov.au, Australia’s Environment: Issues and Trends.

Australian Bureau of Statistics, 2012. 1301.0 - Environment - Waste management. Available at www.abs.gov.au.

Australian Government Productivity Commission, 2006. Waste Management Report. Canberra.

Australian Packaging Covenant, 2010. Sustainable Packaging Guidelines. APC Sustainable Packaging Guidelines. 217 Australian Packaging Covenant, 2011. Australian Packaging Covenant: Recycling Data.

Australian Synchrotron, 2014. SAXS Beameline Technical Specifications. Available at http://www.synchrotron.org.au/aussyncbeamlines/saxswaxs/saxs-specifications.

Avlonitis, S., Hanbury, W.T., Hodgkiess, T., 1992. Chlorine degradation of aromatic polyamides. Desalination 85, 321–334.

Avlonitis, S.A., Kouroumbas, K., Vlachakis, N., 2003. Energy consumption and membrane replacement cost for seawater RO desalination plants. Desalination 157, 151–158.

Ayed, P., Clauzade, C., Gros, B., Huber, J.-C., Lebrun, C., Vassart, N., 2007. Charging used tyres in the EAF, as a substitute for carbon: A success story for LME and Industeel Belgium. Revue de Metallurgie Cahiers D’Informations Techniques 104, 128–135.

Baek, Y., Kang, J., Theato, P., Yoon, J., 2012. Measuring hydrophilicity of RO membranes by contact angles via sessile drop and captive bubble method: A comparative study. Desalination 303, 23–28.

Baker, R.W., 2004. Membrane Technology and Applications, Second Edi. ed. John Wiley and Sons Ltd, Chichester, England.

Baker, R.W., Strathmann, H., 1970. Ultrafiltration of macromolecular solutions with high-flux membranes. Journal of Applied Polymer Science 14, 1197–1214.

Baumann, H., Tillman, A., 2004. The Hitchhiker’s Guide to LCA. Studentlitteratur, Lund, Sweden.

Beerlage, M. a. M., 1994. Polyimide ultrafiltration membranes for non-aqueous systems (Thesis).

Bendall, H.B., Brooks, M.R., 2010. Short sea shipping: lessons for or from australia? International association of maritime economists 1–23.

Beukering, P. Van, Kuik, O., Oosterhuis, F., 2014. Handbook of Recylcing: Chapter 31 - The Economics of Recycling. Institute for Environmental Studies, VU University, Amsterdam, The Netherlands.

Biswas, W.K., 2009. Life Cycle Assessment of Seawater Desalinization in Western Australia. World Academy of Science , Engineering and Technology 369–375.

Bonton, A., Bouchard, C., Barbeau, B., Jedrzejak, S., 2012. Comparative life cycle assessment of water treatment plants. Desalination 284, 42–54.

Brambilla Pisoni, E., Raccanelli, R., Dotelli, G., Botta, D., Melià, P., 2009. Accounting for transportation impacts in the environmental assessment of waste management plans. The International Journal of Life Cycle Assessment 14, 248–256.

Brandrup, J., Wiesbaden, 2005. Polymers, Polymer Recycling, and Sustainability, Plastics and the Environment. John Wiley & Sons, Inc. 218 Brulliard, C., Cain, R., Do, D., Dornom, T., Lim, B., Olesson, E., Young, S., 2012. The Australian recycling sector, Department of Sustainability, Environment, Water, Population and Communities.

Butler, R., 2011. MemRegen: A Program to recycle membranes from desalination plants. Portofino, Italy.

Caddotte, J.E., King, R.S., Majerle, R.J., Petersen, R.J., 1981. Interfacial Synthesis in the Preparation of Reverse Osmosis Membranes. Journal of Macromolecular Science: Part A - Chemistry: Pure and Applied 727–755.

Cadotte, J.E., 1979. Interfacially synthesized reverse osmosis membrane. US Patent 4277344.

Cain, R., 2010. Product Stewardship Common Data. Net Balance - Department of Sustainability, Environment, Water, Population and Communities.

Causserand, C., Rouaix, S., Lafaille, J.-P., Aimar, P., 2006. Degradation of polysulfone membranes due to contact with bleaching solution. Desalination 199, 70–72.

Clauzad, C., 2005. End-of-Life tyres in electric arc furnaces: an industrial success story. Aliapur, http://www.aliapur.fr.

Clavreul, J., Guyonnet, D., Christensen, T.H., 2012. Quantifying uncertainty in LCA- modelling of waste management systems. Waste management (New York, NY) 32, 2482–95.

Cran, M.J., Bigger, S.W., Gray, S.R., 2011. Degradation of polyamide reverse osmosis membranes in the presence of chloramine. Desalination 283, 58–63.

CSM, 2006. Technical manual: Reverse osmosis membrane. Available at http://www.csmfilter.com/.

Dahiya, A., Kamath, M.G., Hegde, R.R., 2004. Spunbond Technology, Available from http://www.engr.utk.edu/mse/Textiles/Spunbond Technology.htm.

Davies, T.C., 1996. Chemistry and pollution of natural waters in western Kenya. Journal of African Earth Sciences 23, 547–563.

Department for Environment Food and Rural Affairs, 2013. Incineration of Municipal Solid Waste. Available from www.defra.gov.uk.

Department of Environment Climate Change and Water, 2010a. Environmental benefits of recycling. Sydney, Australia, Available from www.environment.nsw.gov.au.

Department of Environment Climate Change and Water, 2010b. Environmental benefits of recycling Appendix 6 – Plastics. Sydney, Australia, Available from www.environment.nsw.gov.au.

Department of the Environment Water Heritage and the Arts, 2010. National Waste Policy: Implementation Plan, Policy. Canbera.

219 DEWHA, 2010. Television and Computer Scheme E-Bulletin Issue 1-6. National Television and Computer Product Stewardship Scheme.

DEWPC, 2010. National Waste Policy: Less Waste, More Resources. Department of the Environment, Water, Population and Communities. Australian Government.

Do, V.T., Tang, C.Y., Reinhard, M., Leckie, J.O., 2012. Degradation of polyamide nanofiltration and reverse osmosis membranes by hypochlorite. Environmental science & technology 46, 852–9.

Dollimore, D., Heal, G.R., 1967. The degradation of selected polymers to carbons in an inert atmosphere. Carbon 65–72.

Donose, B.C., Sukumar, S., Pidou, M., Poussade, Y., Keller, J., Gernjak, W., 2012. Effect of pH on the ageing of reverse osmosis membranes upon exposure to hypochlorite. Desalination 309, 97–105.

DOW, 2012. FILMTEC TM Reverse Osmosis Membranes Technical Manual. Available from ww.dowwaterandprocess.com.

Dumée, L., Sears, K., Mudie, S., Kirby, N., Skourtis, C., Mcdonnell, J., Lucas, S., Schütz, J., Finn, N., Huynh, C., Hawkins, S., Kong, L., Hodgson, P., Duke, M., Gray, S., 2013. Characterization of carbon nanotube webs and yarns with small angle X-ray scattering: Revealing the yarn twist and inter-nanotube interactions and alignment. Carbon 63, 562–566.

Dumée, L., Thornton, A., Sears, K., Schütz, J., Finn, N., Spoljaric, S., Shanks, R., Skourtis, C., Duke, M., Gray, S., 2012. In situ small angle X-ray scattering investigation of the thermal expansion and related structural information of carbon nanotube composites. Progress in Natural Science: Materials International 22, 673–683.

Ecocycle, 2011. Waste-of-energy: Why incineration is bad for our economy, environment and community. Available from www.ecocycle.org/zerowaste.

Ecoinvent Centre, 2013. Ecoinvent database v3 ecoinvent report. Swiss Centre for Life Cycle Inventories, Dübendorf.

Edwards, C., Fry, J.M., 2006. Life cycle assessment of supermarket carrier bags: a review of bags available in 2006. Environment Agency, Bristol.

Ekvall, T., Assefa, G., Björklund, A., Eriksson, O., Finnveden, G., 2007. What life-cycle assessment does and does not do in assessments of waste management. Waste management (New York, NY) 27, 989–96.

ElHadidy, A.M., Peldszus, S., Van Dyke, M.I., 2013. Development of a pore construction data analysis technique for investigating pore size distribution of ultrafiltration membranes by atomic force microscopy. Journal of Membrane Science 429, 373–383.

Eltridge, F., 2005. Working with the Community. Government of South Australia, Prepared by Harrison Market Research, Eastweood SA.

220 Environment Protection and Heritage Council, 2009a. Decision Regulatory Impact Statement : Televisions and Computers. Hyder Consulting, Canberra.

Environment Protection and Heritage Council, 2009b. Consultation Regulatory Impact Statement: Televisions and Computers. Hyder Consulting, Canberra.

EPA, 1996. Environmental guidelines: solid waste landfills, Water. Sydney NSW.

Ettori, A., Gaudichet-Maurin, E., Schrotter, J.-C., Aimar, P., Causserand, C., 2011. Permeability and chemical analysis of aromatic polyamide based membranes exposed to sodium hypochlorite. Journal of Membrane Science 375, 220–230.

Fanning, M., 2013. Investigation into the Performance (Environmental and Health) of Waste to Energy Technologies Internationally. Department of Environment and Conservation - Western Australia, written by WSP Environmental Ltd.

Farnel Capital inc, 2013. Sea-rates.com: Transit time and distance calculator. Available at http://www.searates.com/reference/portdistance/.

Fernandez-Álvarez, G., Garralón, G., Plaza, F., Garralón, A., Pérez, J., Gómez, M. a., 2010. Autopsy of SWRO membranes from desalination plant in Ceuta after 8 years in operation. Desalination 263, 264–270.

Finnveden, G., Johansson, J., Lind, P., Moberg, Å., 2005. Life cycle assessment of energy from solid waste—part 1: general methodology and results. Journal of Cleaner Production 13, 213–229.

Foster, S., 2008. Domestic Mixed Plastics Packaging Waste Management Options. WRAP - Available at www.wrap.org.uk.

Frechen, F.-B., 2014. Portable Aqua Unit For Life Saving. University of Kassel. Available from http://wasserrucksack.de/.

Frick, J.M., Féris, L. a., Tessaro, I.C., 2014. Evaluation of pretreatments for a blowdown stream to feed a filtration system with discarded reverse osmosis membranes. Desalination 341, 126–134.

Frischknecht, R., Niels, J., Althaus, H., Doka, G., Dones, R., Heck, T., Hellweg, S., Hischier, R., Nemecek, T., Rebitzer, G., Spielmann, M., 2007. Ecoinvent report no 1 - Overview and Methodology.

Fritzmann, C., Lowenberg, J., Wintgens, T., Melin, T., 2007. State-of-the-art of reverse osmosis desalination. Desalination 216, 1–76.

Gabelich, C., Frankin, J., Gerringer, F., Ishida, K., Suffet, I., 2005. Enhanced oxidation of polyamide membranes using monochloramine and ferrous iron. Journal of Membrane Science 258, 64–70.

Gao, W., Liang, H., Ma, J., Han, M., Chen, Z., Han, Z., Li, G., 2011. Membrane fouling control in ultrafiltration technology for drinking water production: A review. Desalination 272, 1–8.

221 Gaudichetmaurin, E., Thominette, F., 2006. Ageing of polysulfone ultrafiltration membranes in contact with bleach solutions. Journal of Membrane Science 282, 198–204.

GBB, 2013. Gasification of Plastics From Municipal Solid Waste In the United States. The American Chemistry Council , writen by Gershman, Brickner & Bratton, Inc.

GE, 2011. Homespring. Available at http://www.homespring.com/.

Ghaffour, N., Missimer, T.M., Amy, G.L., 2013. Technical review and evaluation of the economics of water desalination: Current and future challenges for better water supply sustainability. Desalination 309, 197–207.

Gholami, M., Nasseri, S., Feng, C.Y., Matsuura, T., Khulbe, K.C., 2003. The effect of heat-treatment on the ultrafiltration performance of polyethersulfone (PES) hollow- fiber membranes. Desalination 155, 293–301.

Ghosh, A., Chatterjee, A., 2008. Ironmaking And Steelmaking: Theory And Practice. Prentice-Hall of India Private Limited, New Delhi, India.

Ghosh, A., Hoek, E., 2009. Impacts of support membrane structure and chemistry on polyamide–polysulfone interfacial composite membranes. Journal of Membrane Science 336, 140–148.

Gilau, A.M., Small, M.J., 2008. Designing cost-effective seawater reverse osmosis system under optimal energy options. Renewable Energy 33, 617–630.

Global Water Intelligence, 2012. DesalData.com online database. Global Water Intelligence. Available from www.desaldata.com.

Goedkoop, M., Schryver, A. De, Oele, M., Durksz, S., Roest, D. de, 2010. SimaPro 7 - Introduction into LCA. PRe Consultants.

Goedkoop, M.., Heijungs, R., Huijbregts, M., De Schryver, A., Struijs, J., Van Zelm, R., 2008. A life cycle impact assessment method which comprises harmonised category indicators at the midpoint and the endpoint level.

Goodship, V., 2007. Introduction to Plastic Recycling. Smithers Rapra Technology Limited, Shawbury UK.

Gorez, J., Groz, B., Birat, J., Grisvard, C., Huber, J., Coq, X. Le, 2003. Charging tires in the EAF as a substitute to carbon. In: 7th European Electric Steelmaking Conference. Venice, Italy, pp. 17–23.

Grant, T., 2013. Inclusion of Uncertainty in LCA. ALCAS conference, Sydney, Australia. Available from www.conference.alcas.asn.au.

Grant, T., James, K., 2005. Life Cycle Impact Data for Resource Recovery from Commercial and Industrial and Construction and Demolition Waste in Victoria. RMIT University, Victoria.

222 Greenlee, L.F., Lawler, D.F., Freeman, B.D., Marrot, B., Moulin, P., 2009. Reverse osmosis desalination: water sources, technology, and today’s challenges. Water research 43, 2317–48.

Grigoriew, H., Chmielewki, A.G., 1998. X-ray study of the influence of water on the cellulose membrane structure. Journal of Material Science Letters 16, 1945–1947.

Guo, M., Murphy, R.J., 2012. LCA data quality: Sensitivity and uncertainty analysis. Science of The Total Environment 435-436, 230–243.

Gupta, S., Sahajwalla, V., Wood, J., 2006. Simultaneous combustion of waste plastics with coal for pulverized coal injection application. Energy & Fuels 20, 2557–2563.

Ha, T., Santucci, D., 2006. Recycling Myths: Sorting the Facts from the Fiction. Planet Ark, Sydney, Australia.

Hajibabania, S., Antony, A., Leslie, G., Le-Clech, P., 2012. Relative impact of fouling and cleaning on PVDF membrane hydraulic performances. Separation and Purification Technology 90, 204–212.

Hancock, N.T., Black, N.D., Cath, T.Y., 2012. A comparative life cycle assessment of hybrid osmotic dilution desalination and established seawater desalination and wastewater reclamation processes. Water research 46, 1145–54.

Hara, S., Kitamura, M., Ogino, K., 1990. The surface biscosities and the foamines of molten oxides. ISIJ International 30, 714–721.

Harris, J., 2012. Focus issue: Landfill gas recovery, Environment Business Media. Available from http://www.wme.com.au/categories/waste_managemt/Dec6_2012.php.

Heijungs, R., Huppes, G., Zamagni, A., Masoni, P., 2011. Life Cycle Assessment : Past, Present, and Future 45, 90–96.

Holgate, P., 2014. Change in RO membrane characteristics during storage and mothballing. Thesis, The University of new South Wales.

Hopewell, J., Dvorak, R., Kosior, E., 2009. Plastics recycling: challenges and opportunities. Philosophical transactions of the Royal Society of London Series B, Biological sciences 364, 2115–26.

Howell, S.G., 1992. A ten year review of plastics recycling. Journal of Hazardous Materials 29, 143–164.

Huber, S.A., Balz, A., Abert, M., Pronk, W., 2010. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography - organic carbon detection - organic nitrogen detection (LC-OCD-OND). Water Research 879–885.

Hussain, A., Trudell, P., Repta, A.J., 1970. Quantitative Spectrophotometric Methods for Determination of Sodium Hypochlorite in Aqueous Solutions. Journal of Pharmaceutical Sciences.

223 ISO, 2006a. ISO 14040, Environmental Management - Life cycle assessment - Principles and framework.

ISO, 2006b. ISO 14044, Environmental Management - Life cycle assessment - Requirements and guidelines.

Izak, P., Hovorka, S., Bartovsky, T., Bartovska, L., Crespo, J., 2007. Swelling of polymeric membranes in room temperature ionic liquids. Journal of Membrane Science 296, 131–138.

Johmke, B., 1992. Waste incineration - An important element of the integrated waste management system in germany. Waste Management & Research 10, 303–315.

Johnson, D., Hilal, N., 2015. Characterisation and quantification of membrane surface properties using atomic force microscopy: A comprehensive review. Desalination 356, 149–164.

Kang, G.D., Gao, C.J., Chen, W.D., Jie, X.M., Cao, Y.M., Yuan, Q., 2007. Study on hypochlorite degradation of aromatic polyamide reverse osmosis membrane. Journal of Membrane Science 300, 165–171.

Kazemi Najafi, S., 2013. Use of recycled plastics in wood plastic composites - A review. Waste management (New York, NY) 33, 1898–1905.

Keane, M.A., 2009. Catalytic Transformation of Waste Polymers to Fuel Oil. ChemSusChem 207–214.

Kesting, R.E., 1973. Concerning the Microstructure of dry-RO Membranes. Journal of Applied Polymer Science 17, 1771–1784.

Kesting, R.E., 1975. Preparation of reverse osmosis membranes by complete evaporation of the solvent system. US Patent 3884801.

Khanna, R., Rahman, M., Leow, R., Sahajwalla, V., 2007. Novel Sessile Drop Software for Quantitative Estimation of Slag Foaming in Carbon/Slag Interactions. Metallurgical and Materials Transactions B: Process Metallurgy and Materials Processing Science 38, 719–723.

Kogutid, K., Kunst, B., 2002. RO and NF membrane fouling and cleaning and pore size distribution variations. Desalination 150, 113–120.

Koo, T., Lee, Y.J., Sheikholeslami, R., 2001. Silica fouling and cleaning of reverse osmosis membranes. Desalination 139, 43–56.

Kuzmenko, D., Arkhangelsky, E., Belfer, S., Freger, V., Gitis, V., 2005. Chemical cleaning of UF membranes fouled by BSA. Desalination 179, 323–333.

Kwon, Y.-N., Leckie, J.O., 2006. Hypochlorite degradation of crosslinked polyamide membranes. Journal of Membrane Science 283, 21–26.

Lalia, B.S., Kochkodan, V., Hashaikeh, R., Hilal, N., 2013. A review on membrane fabrication: Structure, properties and performance relationship. Desalination 326, 77–95. 224 Lassaux, S., Renzoni, R., Germain, A., 2007. Life Cycle Assessment of Water: From the pumping station to the wastewater treatment plant. The International Journal of Life Cycle Assessment 118–126.

Lee, K.P., Arnot, T.C., Mattia, D., 2011. A review of reverse osmosis membrane materials for desalination—Development to date and future potential. Journal of Membrane Science 370, 1–22.

Lee, S.G., Lye, S.W., Khoo, M.K., 2005. A Multi-Objective Methodology for Evaluating Product End-of-Life Options and Disassembly. Advanced Manufacturing Technology 148–156.

Levitsky, I., Duek, A., Arkhangelsky, E., Pinchev, D., Kadoshian, T., Shetrit, H., Naim, R., Gitis, V., 2011. Understanding the oxidative cleaning of UF membranes. Journal of Membrane Science 377, 206–213.

Liu, M., Yu, S., Tao, J., Gao, C., 2008. Preparation, structure characteristics and separation properties of thin-film composite polyamide-urethane seawater reverse osmosis membrane. Journal of Membrane Science 325, 947–956.

Louie, J.S., Pinnau, I., Reinhard, M., 2011. Effects of surface coating process conditions on the water permeation and salt rejection properties of composite polyamide reverse osmosis membranes. Journal of Membrane Science 367, 249– 255.

Macdonald, W., Pan, C., 1974. Method for drying water-wet membranes ( No. US3842515A). US Patent 3842515A.

Malaeb, L., Ayoub, G.M., 2011. Reverse osmosis technology for water treatment: State of the art review. Desalination 267, 1–8.

Matin, A., Khan, Z., Zaidi, S.M.J., Boyce, M.C., 2011. Biofouling in reverse osmosis membranes for seawater desalination: Phenomena and prevention. Desalination 281, 1–16.

McDougal, F., White, P., Franke, M., Hindle, P., 2008. Integrated Solid Waste Management: a Life Cycle Inventory, 2nd ed. John Wiley & Sons.

Meran, C., Ozturk, O., Yuksel, M., 2008. Examination of the possibility of recycling and utilizing recycled polyethylene and polypropylene. Materials & Design 29, 701– 705.

Merrild, H., Larsen, A.W., Christensen, T.H., 2012. Assessing recycling versus incineration of key materials in municipal waste: The importance of efficient energy recovery and transport distances. Waste Management 32, 1009–18.

Mezher, T., Fath, H., Abbas, Z., Khaled, A., 2011. Techno-economic assessment and environmental impacts of desalination technologies. Desalination 266, 263–273.

Migliavacca, M., 2010. Environmental product declaration - Spunbon reinforcements for bituminous membranes made of recycled polyester. Freudenberg Politex Group. Available from www.freudenbergpolitex.com.

225 Morao, A., Pessoadeamorim, M., Lopes, A., Escobar, I., Queiroz, J., 2008. Characterisation of ultrafiltration and nanofiltration membranes from rejections of neutral reference solutes using a model of asymmetric pores. Journal of Membrane Science 319, 64–75.

Mulder, M., 2003. Basic Principles of membrane Technology, Second. ed. Kluwer Academic, Dordrecht, The Netherlands.

Munier, N., 2011. A Strategy for Using Multicriteria Analysis in Decision-Making - A Guide for Simple and Complex Environmental Projects. Springer.

Nagasaka, T., Hino, M., Ban-ya, S., 2000. Interfacial Kinetics of Hydrogen with Liquid Slag Containing Iron Oxide. Metallurgical and Meterial Transactions B 31.

Nielsen, H., Hauschild, M., 1998. Product Specific Emissions from Municipal Solid Waste Landfills Part I : Landfill Model 3, 158–168.

Noble, R.D., Stern, S.A., 1995. Membrane Separations Technology, Volume 2: Principles and Applications. Elsevier Science.

O’Kane, P., 2011. Recycling Rubber Tyres in EAF Steelmaking. In: SMaRT Sustainability Symposium. Sydney, Australia.

Ogaki, Y., Tomikoa, K., Watanabe, A., Arita, K., Kuriyama, I., Sugayoshi, T., 2003. Recycling of Waste Plastic Packaging in a Blast Furnace System, NKK Technical Review. JFE Steel Corporation, Tokyo, Japan.

Omo-irabor, O., Olobaniyi, S.B., Oduyemi, K., Akunna, J., 2008. Surface and groundwater water quality assessment using multivariate analytical methods: A case study of the Western Niger Delta, Nigera. Physics and chemistry of the Earth 33, 666–673.

Ould Mohamedou, E., Penate Suarez, D.B., Vince, F., Jaouen, P., Pontie, M., 2010. New lives for old reverse osmosis (RO) membranes. Desalination 253, 62–70.

Pall Corporation, 2012. Omega Ultrafiltration Membrane Disc Filters - Datasheet.

Pasqualino, J., Meneses, M., Abella, M., Castells, F., 2009. LCA as a Decision Support Tool for the Environmental Improvement of the Operation of a Municipal Wastewater Treatment Plant. Environmental Science & Technology 43, 330.

Patterson, D. a., Havill, A., Costello, S., See-Toh, Y.H., Livingston, A.G., Turner, A., 2009. Membrane characterisation by SEM, TEM and ESEM: The implications of dry and wetted microstructure on mass transfer through integrally skinned polyimide nanofiltration membranes. Separation and Purification Technology 66, 90–97.

Paul P. Rumps, 1991. Plastic wastes: Management, control, recycling and disposal - pollution technology review. U.S. Environmental Protection Agency, Park Ridge, NJ.

226 Peery, M., Hallan, M., Bartels, C., Shelby, I., Metcalfe, P., Knappe, P., 2004. Industry Consortium Analysis of Large Reverse Osmosis / Nanofiltration Element Diameters.

Peeva, P.D., Palupi, A.E., Ulbricht, M., 2011. Ultrafiltration of humic acid solutions through unmodified and surface functionalized low-fouling polyethersulfone membranes – Effects of feed properties, molecular weight cut-off and membrane chemistry on fouling behavior and cleanability. Separation and Purification Technology 81, 124–133.

Peñate, B., García-Rodríguez, L., 2011. Reverse osmosis hybrid membrane inter-stage design: A comparative performance assessment. Desalination 281, 354–363.

Peñate, B., García-rodríguez, L., 2012. Current trends and future prospects in the design of seawater reverse osmosis desalination technology. Desalination 284, 1– 8.

Perry, R.H., Green, D.W., 2008. Perry’s Chemical Engineers' Handbook. Section 22: Alternative Separation Processes, 8th ed. McGraw-Hill.

Peter-Varbanets, M., 2010. Ultra-low pressure ultrafiltration for decentralized drinking water treatment. PhD Thesis.

Peter-Varbanets, M., Hammes, F., Vital, M., Pronk, W., 2010. Stabilization of flux during dead-end ultra-low pressure ultrafiltration. Water Research 44, 3607–16.

Peter-varbanets, M., Koetzsch, S., Müller, S., Pronk, W., 2009a. Gravity-driven membrane disinfection (GDMD): choice of a suitable membrane. In: IWA Specialist Conference on Membrane Technology. Aachen Germany.

Peter-Varbanets, M., Margot, J., Traber, J., Pronk, W., 2011. Mechanisms of membrane fouling during ultra-low pressure ultrafiltration. Journal of Membrane Science 377, 42–53.

Peter-varbanets, M., Zurbru, C., Swartz, C., Pronk, W., 2009b. Decentralized systems for potable water and the potential of membrane technology. Water Research 43, 245–265.

Pinnau, I., Freeman, B.D., 2000. Formation and Modification of Polymeric Membranes : Overview. ACS Symposium Series.

Plappally, A.K., Lienhard V, J.H., 2012. Energy requirements for water production, treatment, end use, reclamation, and disposal. Renewable and Sustainable Energy Reviews 16, 4818–4848.

Pontié, M., Rapenne, S., Thekkedath, A., Duchesne, J., Jacquemet, V., Leparc, J., 2005. Tools for membrane autopsies and antifouling strategies in seawater feeds: a review. Desalination 181, 75–90.

Porcelli, N., Judd, S., 2010. Chemical cleaning of potable water membranes: A review. Separation and Purification Technology 71, 137–143.

227 Poussade, Y., Vince, F., Robillot, C., 2011. Energy consumption and greenhouse gases emissions from the use of alternative water sources in South East Queensland. Water Science & Technology: Water Supply 11, 281.

Prince, C., Cran, M., Le-Clech, P., Uwe-Hoehn, K., Duke, M., 2011. Reuse and recycling of used desalination membranes. OzWater.

Rahimpour, A., Madaeni, S.S., 2010. Improvement of performance and surface properties of nano-porous polyethersulfone (PES) membrane using hydrophilic monomers as additives in the casting solution. Journal of Membrane Science 360, 371–379.

Raluy, G., Serra, L., Uche, J., 2006. Life cycle assessment of MSF, MED and RO desalination technologies. Energy 31, 2361–2372.

Raluy, R.G., Serra, L., Uche, J., 2005a. Life Cycle Assessment of Water Production Technologies Part 1 : Life Cycle Assessment of Different Commercial Desalination Technologies (MSF , MED , RO). International Journal of Life Cycle Assessment 10, 285–293.

Raluy, R.G., Serra, L., Uche, J., Valero, A., 2005b. Life Cycle Assessment of Water Production Technologies. Part 2 : Reverse Osmosis Desalination versus the Ebro River Water Transfer. International Journal of Life Cycle Assessment 10, 346– 354.

Ram, A., 1997. Fundamentals of Polymer Engineering. Springer - Verlag.

Rangarajan, R., Desai, N.V., Daga, S.L., Joshi, S.V., Prakash Rao, a., Shah, V.J., Trivedi, J.J., Devmurari, C.V., Singh, K., Bapat, P.S., Raval, H.D., Jewrajka, S.K., Saha, N.K., Bhattacharya, a., Singh, P.S., Ray, P., Trivedi, G.S., Pathak, N., Reddy, a. V.R., 2011. Thin film composite reverse osmosis membrane development and scale up at CSMCRI, Bhavnagar. Desalination 282, 68–77.

Raval, H.D., Chauhan, V.R., Raval, A.H., Mishra, S., 2012. Rejuvenation of discarded RO membrane for new applications. Desalination and Water Treatment 48, 349– 359.

Razmjou, A., Mansouri, J., Chen, V., Lim, M., Amal, R., 2011. Titania nanocomposite polyethersulfone ultrafiltration membranes fabricated using a low temperature hydrothermal coating process. Journal of Membrane Science 380, 98–113.

Recoup, 2005. The Recycling of Polypropylene Containers. Recoup Limited, Peterborough.

Ribera, G., Clarens, F., Martínez-Lladó, X., Jubany, I., V Martí, Rovira, M., 2013. Life cycle and human health risk assessments as tools for decision making in the design and implementation of nanofiltration in drinking water treatment plants. The Science of the total environment 466-467C, 377–386.

Ricardo-Aea, 2013. Case Study 3: Cynar Plastics to Diesel. Available from http://www.zerowaste.sa.gov.au/industry/case-studies.

228 Roberts, D., Johnston, E., Knott, N., 2010. Impacts of desalination plant discharges on the marine environment: A critical review of published studies. Water research 44, 5117–28.

Rocheta, E., Schulz, M., Short, M., 2011. Electric Arc Furnace steelmaking – a life cycle assessment of two options., ALCAS conference. ALCAS conference, Melbourne, Australia. Available from www.conference.alcas.asn.au.

Rodriguez-Gonzalez, J., Jimenez, V., Trujillo, O., Veza, J., 2002. Reuse of reverse osmosis membranes in advanced wastewater treatment. Desalination 150, 219– 225.

Rouaix, S., Causserand, C., Aimar, P., 2006. Experimental study of the effects of hypochlorite on polysulfone membrane properties. Journal of Membrane Science 277, 137–147.

Sahajwalla, V., Rahman, M., Khanna, R., Saha-chaudhury, N., Kane, P.O., Skidmore, C., Knights, D., 2008. Recycling waste plastics in EAF steelmaking : carbon / slag interactions of HDPE-coke blends. Process Metallurgy 535–543.

Sahajwalla, V., Zaharia, M., Kongkarat, S., Khanna, R., Rahman, M., Saha-Chaudhury, N., O’Kane, P., Dicker, J., Skidmore, C., Knights, D., 2012. Recycling end-of-life polymers in an electric arc furnace steelmaking process: fundamentals of polymer reactions with slag and metal. Energy & Fuels 26, 58–66.

Sahajwalla, V., Zaharia, M., Kongkarat, S., Khanna, R., Saha-Chaudhury, N., O’Kane, P., 2010. Recycling plastics as a resource for electric arc furnace (EAF) steelmaking: combustion and structural transformations of metallurgical coke and plastic blends. Energy & Fuels 24, 379–391.

Sahajwalla, V., Zaharia, M., Rahman, M., Khanna, R., Saha-Chaudhury, N., O’Kane, P., Dicker, J., Skidmore, C., Knights, D., 2011. Recycling rubber tyres and waste plastics in EAF steelmaking. Steel Research International 82, 566–572.

Saikia, N., de Brito, J., 2012. Use of plastic waste as aggregate in cement mortar and concrete preparation: A review. Construction and Building Materials 34, 385–401.

Scheirs, J., 1998. Polymer Recycling: Science, Technology and Application, first ed. ed. Wiley-Blackwell.

Schock, G., Miquel, A., Birkenberger, R., 1989. Characterization of ultrafiltration membranes: cut-off determination by gel permeation chromatography. Journal of Membrane Science 41, 55–67.

Schrotter, J., Rapenne, S., Leparc, J., Remize, P., Casas, S., 2010. Current and Emerging Developments in Desalination with Reverse Osmosis Membrane Systems. Water Research Center of Veolia.

Schwinge, J., Neal, P.R., Wiley, D.E., Fletcher, D.F., Fane, A.G., 2004. Spiral wound modules and spacers review and analysis. Journal of Membrane Science 242, 129–153.

229 Shonfield, P., 2008. LCA of Management Options for Mixed Waste Plastics. WRAP - Material change for a better environment, Banbury, UK.

Siddique, H., Bhole, Y., Peeva, L.G., Livingston, a. G., 2014. Pore preserving crosslinkers for polyimide OSN membranes. Journal of Membrane Science 465, 138–150.

Siddique, R., Khatib, J., Kaur, I., 2008. Use of recycled plastic in concrete: a review. Waste management (New York, NY) 28, 1835–52.

SkyJuice Foundation, 2014a. SkyHydrant Water Filtration Unit. Available from http://www.skyjuice.com.au/skyhydrant.htm.

SkyJuice Foundation, 2014b. SkyBox Water Filtration Unit. Available from http://www.skyjuice.com.au/skybox.htm.

Soice, N.P., Greenberg, A.R., Krantz, W.B., Norman, A.D., 2004. Studies of oxidative degradation in polyamide RO membrane barrier layers using pendant drop mechanical analysis. Journal of Membrane Science 243, 345–355.

Soltani, A., Hewage, K., Reza, B., Sadiq, R., 2014. Multiple stakeholders in multi- criteria decision-making in the context of Municipal Solid Waste Management : A review. Waste Management 35, 318–328.

Stadler, S., Eksteen, J., Aldrich, C., 2007. An experimental investigation of foaming in acidic, high FexO slags. Minerals Engineering 20, 1121–1128.

Staude, E., Passlack, J., 1986. Characterization of ultrafiltration membranes by drying. Journal of Membrane Science 28, 209–223.

Stokke, D.D., Wu, Q., Han, G., 2013. Introduction to Wood and Natural Fibre Composites. 4.5.3.2: Bound Water: Differential Head of Sorption. Introduction to Wood and Natural Fibres, John Wiley & Sons.

Subrahmanyan, S., 2003. An Investigation of Pore Collapse in Asymmetric Polysulfone Membranes (PhD Thesis). Virginia Polytechnic Institute and State University.

Tan, R.B.H., Khoo, H.H., 2006. Impact assessment of waste management options in Singapore. Journal of the Air & Waste Management Association (1995) 56, 244– 54.

Tang, C.Y., Chong, T.H., Fane, A.G., 2011. Colloidal interactions and fouling of NF and RO membranes: a review. Advances in colloid and interface science 164, 126–43.

Tangsubkul, N., Parameshwaran, K., Lundie, S., Fane, A.G., Waite, T.D., 2006. Environmental life cycle assessment of the microfiltration process. Journal of Membrane Science 284, 214–226.

Tarnacki, K., Meneses, M., Melin, T., van Medevoort, J., Jansen, A., 2012. Environmental assessment of desalination processes: Reverse osmosis and Memstill®. Desalination 296, 69–80.

230 Tarnacki, K.M., Melin, T., Jansen, A.E., van Medevoort, J., 2011. Comparison of environmental impact and energy efficiency of desalination processes by LCA. Water Science & Technology: Water Supply 11, 246.

The Parliament of the Commonwealth of Australia, 2011. Product Stewardship Bill 2011. Canberra.

Toray Industries, 2005. Operation, Maintenance and Handling Manual for membrane elements. Available from www.torayindustries.com.

US EPA, 2005. Membrane Filtration Guidance Manual. United States Environmental Protection Agency, Office of Water.

US EPA, 2012. State of practice for Emerging waste conversion technologies. Offie of Research and Development, Research Triangle Park, NC, available at www.epa.gov/ord.

Van Der Bruggen, B., Vandecasteele, C., 2003. Removal of pollutants from surface water and groundwater by nanofiltration : overview of possible applications in the drinking water industry. Environmental Pollution 122, 435–445.

Venturoli, D., Rippe, B., 2005. Ficoll and dextran vs. globular proteins as probes for testing glomerular permselectivity: effects of molecular size, shape, charge, and deformability. American journal of physiology Renal physiology 288, F605–13.

Vestergaard Frandsen, 2011. LifeStraw Family. Available from http://www.vestergaard- frandsen.com/lifestraw.

Veza, J., Rodriguez-Gonzalez, J., 2003. Second use for old reverse osmosis membranes: wastewater treatment. Desalination 157, 65–72.

Vhquez, G., Alvarez, E., Navaza, J.M., 1995. Surface Tension of Alcohol + water from 20 to 50C. Journal of Chemical Engineering Data 611–614.

Vince, F., Aoustin, E., Bréant, P., Marechal, F., 2008. LCA tool for the environmental evaluation of potable water production. Desalination 220, 37–56.

Vrouwenvelder, J.S., Graf von der Schulenburg, D. a, Kruithof, J.C., Johns, M.L., van Loosdrecht, M.C.M., 2009. Biofouling of spiral-wound nanofiltration and reverse osmosis membranes: a feed spacer problem. Water research 43, 583–94.

Wang, X.-L., Shang, W.-J., Wang, D.-X., Wu, L., Tu, C.-H., 2009. Characterization and applications of nanofiltration membranes: State of the art. Desalination 236, 316– 326.

Water Desalination Report, 2013. Ozwater 2013 update.

WaterSurplus, 2014. Surplus Assets - Membrane Elements. Available from http://www.watersurplus.com.

Wechsler, A., Hiziroglu, S., 2007. Some of the properties of wood plastic composites. Building and Environment 42, 2637–2644.

231 Welle, F., 2011. Twenty years of PET bottle to bottle recycling—An overview. Resources, Conservation and Recycling 55, 865–875.

WHO, UNICEF, 2010. Progress on Sanitation and Drinking-Water Update. World Health Organisation.

Wilf, M., Klinko, K., 2001. Optimization of seawater RO systems design. Desalination 138, 299–306.

Wolf, P.H., Siverns, S., Monti, S., 2005. UF membranes for RO desalination pretreatment 182, 293–300.

Wong, P.C.Y., Kwon, Y.-N., Criddle, C.S., 2009. Use of atomic force microscopy and fractal geometry to characterize the roughness of nano-, micro-, and ultrafiltration membranes. Journal of Membrane Science 340, 117–132.

Wu, C., Williams, P.T., 2010. Pyrolysis–gasification of plastics, mixed plastics and real- world plastic waste with and without Ni–Mg–Al catalyst. Fuel 89, 3022–3032.

Yadav, K., Morison, K., Staiger, M.P., 2009. Effects of hypochlorite treatment on the surface morphology and mechanical properties of polyethersulfone ultrafiltration membranes. Polymer Degradation and Stability 94, 1955–1961.

Yadav, K., Morison, K.R., 2010. Effects of hypochlorite exposure on flux through polyethersulphone ultrafiltration membranes. Food and Bioproducts Processing 88, 419–424.

Yassin, L., Lettieri, P., Germana, A., 2005. Energy recovery from thermal processing of waste : a review. ICE Engineering Sustainability 158, 97–103.

Zaharia, M., Sahajwalla, V., Khanna, R., Koshy, P., 2009a. Carbon/Slag Interactions between Coke/Rubber Blends and EAF Slag at 1 550°C. ISIJ International 49, 1513–1521.

Zaharia, M., Sahajwalla, V., Kim, B.-C., Khanna, R., Saha-Chaudhury, N., O’Kane, P., Dicker, J., Skidmore, C., Knights, D., 2009b. Recycling of rubber tires in electric arc furnace steelmaking: simultaneous combustion of metallurgical coke and rubber tyres blends. Energy & Fuels 23, 2467–2474.

Zaman, A.U., 2010. Comparative study of municipal solid waste treatment technologies using life cycle assessment method. International Journal of Environmental Science and Technology 7, 225–234.

Zarghami, M., Szidarovszky, F., 2011. Multicriteria Analysis: Applications to Water and Environment Management. Springer, London, New York.

Zhang, Z., An, Q., Liu, T., Zhou, Y., Qian, J., Gao, C., 2011. Fabrication of polysulfone ultrafiltration membranes of a density gradient cross section with good anti- pressure stability and relatively high water flux. Desalination 269, 239–248.

Zhao, L., Chang, P.C.-Y., Ho, W.S.W., 2012. High-flux reverse osmosis membranes incorporated with hydrophilic additives for brackish water desalination. Desalination 308, 225–232. 232 Zhou, J., Chang, V.W.-C., Fane, A.G., 2011a. Environmental life cycle assessment of reverse osmosis desalination: The influence of different life cycle impact assessment methods on the characterization results. Desalination 283, 227–236.

Zhou, J., Chang, V.W.-C., Fane, A.G., 2011b. Environmental life cycle assessment of brackish water reverse osmosis desalination for different electricity production models. Energy & Environmental Science 4, 2267.

Zuin, S., Scanferla, P., Brunelli, A., Marcomini, A., Wong, J.E., Wennekes, W., Genne, I., 2013. Layer-by-Layer Deposition of Titanium Dioxide Nanoparticles on Polymeric Membranes : A Life Cycle Assessment Study. Industrial & Engineering Chemistry Research.

233 Appendix A.

A.1. Membrane Delamination

A number of experiments in this thesis have required the removal of the non-woven PET support layer from the PSf layer, which are physically interlocked during the casting process. To do this, a method for membrane delamination that minimised damage to the active layer was needed. Slight backpressure was applied in a dead end cell to the membranes to put stress on the membrane interfaces. Following this, the layers were slowly peeled away using tweezers. Once removed from the PET layer, the PSf/PA layer

(or the PES layer for UF membranes) rapidly curled up. The side of the layer facing away from the active layer is referred to as the backside. SEM images of these layers, which were done using the method described in Section 3.2.5, can be seen in Figures A-1 and

A-2.

Figure A-1: SEM images of the backside of PSf layer of XLE RO membrane after the

PET support layer has been removed.

234

Figure A-2: SEM images of the PET support layer of XLE RO membrane after it has been removed from the PSf layer. The viewed side was directly in contact with the PSf

layer.

235 Following delamination a number of tests were conducted to validate the process. The methods used are described in Section 3.2.2. PWP and salt rejection tests of the delaminated PA/PSf RO membranes showed no statistical difference from virgin samples. This indicated that the process was successful in delicately removing the PET layer without adversely affecting the PSf or PA layers. Interestingly, the PET layer comprises 80% of the total membranes weight.

Figure A-1 shows the backside of the PSf layer for a XLE RO membrane, this is the same membrane layer as in Figure 3-8 in Chapter 3, but at significantly reduced magnification.

This image shows a number of interesting features. First, large bubble shaped features can be seen throughout the surface. They are almost perfectly hemispherical in shape, and are likely formed by trapped air during the casting process. Secondly, channels where the strands of non-woven PET (as seen in Figure A-2A) were in contact with the

PSf can be seen. Where the PSf has been physically interlocked with the PET layer can be seen in Figure A-2A. In Figure A-2B, a piece of porous PSf can be seen trapped in the non-woven structure of the PET. During the separation process the PSf has phyiscially torn away from the membrane.

236 A.2. Additional SEM images

Figure A-3: SEM images of the active membrane surface for; Top) 30 kDa UF, and

Bottom) 100 kDa UF.

237

Figure A-4: Cross section SEM images of XLE RO membrane; Top) virgin, and Bottom) after use and drying. Images show the physical delamination of the PET and PSf layers

in the dry sample.

238 A.3. Supplementary Information for SAXS analysis

Figure A-5: Scattering patterns for the 25°C evaporation tests on the delaminated UF

membranes for image numbers 1 and 100 (7 and 700 s).

239

Figure A-6: Scattering patterns for the 40oC evaporation tests on delaminated UF

membranes for images 1, 100, 200 and 300 (7, 700, 1400, 3100 s).

Figure A-7: Patterns for the 60°C evaporation tests on delaminated UF membranes for

images 1, 20, 25, 30 and 40 (7, 140, 175, 210, 280 s).

240

Figure A-8: Porod regressions for 10 kda UF membranes at maximum times for 25, 40

and 60°C series, using 0.6m camera length. The Porod law is express in its standard

form for a flat material.

241 A.4. Supplementary Information for Chapter 4

300

) 1

- 250

.bar

1 -

.h 200

2 -

150

100

50 Permeability (L.m Permeability

0 0 100000 200000 300000 400000 Degradation Intensity (ppm.h)

Figure A-9: Effect of NaOCl degradation intensity on the PWP of DOW XLE

membranes. Tested at 2 bar in low pressure stirred dead end cell.

242 Appendix B. The Use of RO Membrane Components as a Coke Substitute in EAF

Steelmaking.

B.1. Introduction

As previously discussed in Section 2.5.6 of the literature review, the use of the polymeric components of used RO membranes in EAF steelmaking is a promising end-of-life scenario, with many financial and environmental benefits. As many of the major components of RO membrane modules are constructed out of polymeric materials, they can potentially be utilised in EAF using this process. Therefore, the aim of the following study is to use standard EAF testing methodologies to determine the viability of using membrane components in the EAF process and assess any benefit that they provide.

B.2. Material and Methods

A new 8” DOW BW30FR membrane element was autopsied and the sheets and spacers were removed from the permeate tube. The sheets and spacers were cut, ground and sieved to a size of <1 mm and mixed at a ratio of 30:70 with metallurgical coke, which has been determined to be the optimum ratio of increasing slag performance (Sahajwalla et al., 2008). The remaining fibre glass shell and ABS permeate tube and end caps were not suitable for the process due their chemical composition. Mixtures were devolatilised in a DTF in an atmosphere of 20 % O2 and 80 % N2 at a temperature of 1200°C, this process is described in detail in a previous paper (Gupta et al., 2006).

The foaming behaviour of EAF iron oxide rich slag in contact with coke/polymeric blends was studied using the sessile drop method in a horizontal furnace (Sahajwalla et al.,

2011). This technique is used to study carbon/slag interaction by continuously monitoring the changes in droplet volume with time. The sessile drop technique uses a horizontal furnace at a temperature of 1550°C, using argon gas to ensure an inert atmosphere flowing at a rate of 1 L.min-1. The schematic representation of the horizontal tube furnace is presented in Figure B-1.

243

Figure B-1: A schematic representation of the experimental sessile-drop arrangement

(Zaharia et al., 2009a).

The substrate is prepared by hydraulically compacting the carbonaceous residue collected after the DTF. Once the desired furnace temperature was attained, the carbon substrate and slag sample was pushed into the hot zone using a stainless steel rod. The melting of the slag marked the beginning of the experiment and a high resolution camera and data processing software was used to capture the dynamic changes in volume

(Khanna et al., 2007). The dynamic changes in the volume of slag droplet are represented as volume ratio (Vt/V0) as a function of time, where Vt represents the volume of slag droplet at time t and V0 is the initial volume. The experimental methodology followed in his study has been reported in previous papers (Sahajwalla et al., 2008;

Zaharia et al., 2009a).

Slag foaming involves the expansion of molten slag by CO gas bubbles evolving from chemical reactions at the slag/carbonaceous material interface and slag/metal interface.

The primary chemical reaction responsible for carbon/slag interactions is:

FeO(slag) + C(s) = Fe↓(metal) + CO(g)

244 In addition, CO(g) initially produced, is transported through the gas phase and reacts with

FeO in slag to produce CO2. The rate of CO and CO2 gas generation depends on the carbon based material and their composition, including volatile matter and ash content, in addition to slag properties and process conditions such as temperature. As the slag properties and process conditions were constant for this study, variations in foaming behaviour can be attributed to the carbon material properties.

B.3. Results and Discussion

The slag foaming behaviour of several different membrane module components and MC was studied using a horizontal furnace. In order to qualitatively visualize the foaming phenomena studied with this technique, representative dynamic images of the slag droplet with MC are shown in Figure B-2.

Figure B-2: Images of slag droplets in contact with raw metallurgical coke at 1550°C as

a function of time.

The size of the slag droplet in contact with MC showed a slight decrease in volume over time and maintained a reasonably spherical shape, which is beneficial for the process and measurement accuracy. In order to quantify the changes in volumes of the three samples and pure MC, the ratio of current to initial volume (Vt/V0) was plotted against time, seen in Figure B-3.

245 2 Permeate spacer Membrane 1.5 sheet Feed spacer 100%MC

o 1

/V

t V

0.5

0 0 100 200 300 400 500 600 Time (s)

Figure B-3: Carbon/slag interactions for metallurgical coke/slag and material blends at

1550°C.

B.3.1. Behaviour of Pure Metallurgic Coke

As the initial slag melting stage is represented by t=0, the volume ratio starts with a value of 1, following this initial stage, changes are observed corresponding to CO and CO2 generation and entrapment, which is reflected as an increase or decrease in droplet volume. After 480 sec, gas activity is reduced considerably with nearly stable Vt/V0 values, and is the result of complete consumption of carbon.

The results reveal that the volume ratio of the pure MC decreased gradually to a value of 0.5, indicating a low extent of gas entrapment by the slag. After 270 sec, the volume ratio increased to a maximum value of 0.9 before all the carbon was consumed. The small but rapid fluctuation in droplet size, seen in all samples, is associated with the generation and subsequent release of gases.

B.3.2. Behaviour of Membrane Components

The membrane sheet and feed spacer, constructed of primarily polyester and polypropylene respectively, showed significantly different trends with much higher levels 246 of droplet volume when compared to pure MC. Both samples reached a maximum of

1.32 Vt/V0 after 45 sec and gradually declined over the course of the experiments to final ratio between 0.8 and 1. The pure polyester of the permeate spacer showed the most favourable slag foaming behaviour, with a high value of 1.5 reached in the first 45 sec.

The volume ratio fluctuated with gas generation and release but maintained a consistent average of 1.3 until carbon was consumed by 470 sec.

B.3.3. Implications for the EAF process

The total volume of generated gases for raw MC was lower than that produced by the bends involving polymeric membrane components. This increased gas generation can be attributed to the volatiles still trapped in the carbonaceous mixtures, which are available after the initial combustion reaction in the DTF. The larger volume ratio of the polymeric membrane blend components translates to a thicker and more stable foamy slag, preventing excessive heat transfer to the furnace walls and roof, increasing efficiency and preventing damage to vulnerable components.

The membrane sheet, which is primarily made of polyester, did not show comparable performance with the permeate spacer which is constructed entirely of polyester. Thin film composite RO membrane sheets typically consist of three layers, a polyester non- woven structural support, a microporous polysulfone interlayer, and range of thin selective polyamide materials for the upper surface (Zhao et al., 2012). In addition, a number of surface modifiers and coatings are used to enhance membrane performance, including acids, alcohols, poly vinyl alcohol, ammonia or alkyl compounds (Lee et al.,

2011). One possible explanation for this difference in foaming performance is the composition of these membrane additives, which could decrease the levels of available volatile matter or suppress the formation of a foamy slag (Zaharia et al., 2009a).

However, this does not have a significant negative effect on the membrane sheets suitability for this application. When compared to existing literature, all the membrane components show similar performance to MC-recycled rubber tyre and MC-PET plastic 247 blends. However a MC-HDPE plastic blend has been shown to have significantly higher foaming performance than both rubber and membrane component blends (Sahajwalla et al., 2012).

In addition to enhanced slag foaming performance, using polymeric substances can provide a number additional benefits. The calorific content of the plastic can provide an additional source of energy for the steelmaking process, potentially reducing the electrical energy demand (Sahajwalla et al., 2012). It has also been proposed that hydrogen remaining in the volatiles following the DTF process can increase the rate of reduction. This kinetic advantage is attributed to the reduction reaction (FeO + H2(g) =

Fe↓ + H2O(g)), which is up to two magnitudes faster than CO reduction (Nagasaka et al.,

2000). The final benefit of using polymeric membrane components in an EAF is that it is expected to reduce CO2 emissions because of their higher ratio of hydrogen to carbon results in less CO2 as a combustion product (Gupta et al., 2006).

B.4. Conclusion

This preliminary study has shown that blends of specific polymeric membrane components with MC have a possible application in replacing some of the conventional

MC used in EAF steelmaking for its carbon requirements, as a result of enhanced carbon/slag interactions. Higher gas entrapment in the slag was observed when polymeric components partially replaced coke. The size of the slag droplet showed wide fluctuation with time and was associated with trapping and subsequent release of gases.

Additional benefits of using polymeric membrane components as a partial coke substitute, have also been discussed, making this option increasingly beneficial for the environment, membrane users, and steelmakers.

248 Appendix C. Supporting information for Chapter 6 – Comparative Life Cycle

Assessment of End-of-life Options for RO Membranes

C.1. Scenario Process Flow Diagrams

Feed and Permeate Membrane permeate End caps Casing Packaging tube sheets spacer

Integrity testing

Gluing Permeate permeate Membrane End cap Fibre glass Packaging Transportation tube spacer to rolling welding spinning tube

Filament Extrussion winding fibre Cardboard Drilling holes Mouling Gluing moulding glass box moulding

Extrussion Granular Polyamide Granular Resin Plastic bag moulding plastic polymerisation plastic

1% food grade sodium Granular Polkysulfone Glass strand metabisulfite plastic casting extrusion storage solution

Polyester base manufacturing

Figure C-1: Manufacturing process for RO TFC Membranes

Air Emissions Used Transport Landfill Membrane Water Emissions

Figure C-2: Flow diagram and boundaries for landfill scenario

249

Emissions Heat Energy Heat

Used Offset Electricity Transport Shredding Incineration Membrane production

Ash

Landfill

Electricity

Chemicals Natural Gas Natural

Figure C-3: Flow Diagram and boundaries for incineration scenario Air Emissions Air Air Emissions Air

Used Syngas Syngass Offset Electricity Transport Disassembly Shredding Gasification Membrane combustion production

Waste Ash components

Landfill

Water

Electricity

Electricity Electricity Natural Gas Natural

Figure C-4: Flow diagram and boundaries for gasification scenario

250 Chemical waste Chemical

Used Offset Coke Transport Disassembly Crushing Washing EAF Membrane production

Waste

Landfill

Electricity

Electricity

Electricity Chemicals

Figure C-5: Flow diagram and boundaries for electric arc furnace scenario

Material recycling 1

Offset Used Shredding Material Transport Disassembly material Membrane and cleaning recycling 2 production

Material Landfill

recycling 3

Water

Electricity

Chemicals Emissions

Figure C-6: Flow diagram and boundaries for material recycling

251

Electricity Materials

Production offset Membrane manufacturing

Used Reuse as Compaction Conversion Transport Landfill

Membrane UF module

Water Water

Electricity Electricity Chemicals

Chemical waste Chemical

Figure C-7: Flow diagram and boundaries for conversion to UF scenario

Electricity Materials

Production offset Membrane manufacturing

Used Cleaning and Transport Reuse Landfill

Membrane storage

Electricity

Chemicals Water use Water

Chemical waste Chemical

Figure C-8: Flow diagram and boundaries for direct membrane reuse

252 C.2. Input output tables

Table C-1: Life cycle inventory and uncertainty values for membrane manufacturing

Amount Module used Material Unit Notes U1 U2 U3 U4 U5 U6 Ub SD95 Notes on Uncertainty Components per module End caps ABS Kg 0.44 Based on multiple 1 1.02 1 1.02 1 1.05 1.05 1.08 Uncertainty of the membrane autopsies weight of each membrane component Feed spacer Polypropylene kg 1.45 Based on multiple 1 1.02 1 1.02 1 1.05 1.05 1.08 Uncertainty of the membrane autopsies weight of each membrane component Permeate Polyethylene kg 1.81 Based on multiple 1 1.02 1 1.02 1 1.05 1.05 1.08 Uncertainty of the spacer terephthalate membrane autopsies weight of each membrane component Membrane Composite material kg 4.63 Based on multiple 1 1.02 1 1.02 1 1.05 1.05 1.08 Uncertainty of the sheet membrane autopsies weight of each membrane component Spunbonded Support layer kg / 0.75 Measurements and 1.2 1.02 1 1.02 1 1.05 1.05 1.22 production construction, PET kg of calculation of individual MB layer thickness, density sheet and weight Polysulfone Supportive and kg / 0.25 Measurements and 1.2 1.02 1 1.02 1 1.05 1.05 1.22 and active membrane kg of calculation of individual polyamide layers, varied MB layer thickness, density layers material sheet and weight Fibre glass Wound fibre glass in Kg 1.83 Based on multiple 1 1.02 1 1.02 1 1.05 1.05 1.08 Uncertainty of the case polyester resin membrane autopsies weight of each membrane component Tube ABS Kg 1.97 Based on multiple 1 1.02 1 1.02 1 1.05 1.05 1.08 Uncertainty of the membrane autopsies weight of each membrane component

253 Glue Polyurethane Glue Kg 1.37 Based on multiple 1 1.02 1 1.02 1 1.05 1.05 1.08 Uncertainty of the membrane autopsies weight of each membrane component O-ring Synthetic rubber g 30.00 1 1.02 1 1.02 1 1.05 1.05 1.08 Amount used Component Process Units Notes U1 U2 U3 U4 U5 U6 Ub SD95 Notes on Uncertainty per module End caps ABS kg 0.46 Ecoinvent process Built-in uncertainty for adapted to US material Ecoinvent processes and energy mix. Assuming 5% waste Injection moulding kg 0.47 Ecoinvent process Built-in uncertainty for of ABS adapted to US material Ecoinvent processes and energy mix. Feed spacer Polypropylene kg 1.53 Ecoinvent process Built-in uncertainty for adapted to US material Ecoinvent processes and energy mix. Assuming 5% waste Extrusion of plastic kg 1.56 Ecoinvent process Built-in uncertainty for film PP adapted to US material Ecoinvent processes and energy mix. Perm spacer Polyethylene kg 1.90 Ecoinvent process Built-in uncertainty for terephthalate adapted to US material Ecoinvent processes

and energy mix. Assuming 5% waste Extrusion of plastic kg 1.81 Ecoinvent process Built-in uncertainty for film PET adapted to US material Ecoinvent processes and energy mix. Membrane Water L 19.99 Water use for interfacial sheet polymerisation of 1.2 1 1 1 1.2 1.2 1.05 1.38 polyamide layer Drying of the polyamide Light fuel oil kwh 2.83 1.2 1 1 1 1.5 1.2 1.05 1.62 layer

254 Polyamide fibres g 29.98 Ecoinvent process adapted to US material 1.2 1.02 1 1.1 1.2 1 1.05 1.32 and energy mix. Spunbonded Water L 6.55 From spunbonded layer Based on information production manufacturers from public information from a variety of 1.05 1.05 1 1 1.5 1.05 1.05 1.52 spunbonded manufacturers, some of which are for different applications. Polyethylene kg 3.64 Ecoinvent process 1.05 1.1 1 1.01 1 1.05 1.05 1.14 terephthalate adapted to US material and energy mix. Assuming 5% waste Light fuel oil kwh 19.64 From spunbonded layer 1.05 1.05 1 1 1.5 1.05 1.05 1.52 manufacturers Electricity kwh 5.86 US average energy mix. 1.05 1.05 1 1 1.5 1.05 1.05 1.52 From spunbonded layer manufacturers Polysulfone Sulphuric acid g 356.38 Stoichiometric calculation 1.2 1.02 1 1 1.5 1 1.05 1.56 layer Monochlorobenzene g 408.99 Stoichiometric calculation 1.2 1.02 1 1 1.5 1 1.05 1.56 bisphenol A g 400.10 Stoichiometric calculation 1.2 1.02 1 1 1.5 1 1.05 1.56 Sodium carbonate g 385.12 Stoichiometric calculation 1.2 1.02 1 1 1.5 1 1.05 1.56 Light fuel oil kwh 0.26 Membrane drying stages 1.1 1.02 1 1 1.2 1 1.05 1.24 Water l 5.89 Non-solvent 1 1.02 1 1 1.5 1 1.05 1.51 N,N - kg 5.89 Solvent. Ecoinvent 1 1.02 1 1 1.5 1 1.05 1.51 Dimethylformamide process adapted to US material and energy mix. Electricity kwh 0.59 US average energy mix 1.1 1.02 1 1 1.2 1 1.05 1.24 Fibre glass Glass fibre with pipe kg 1.92 Assuming 5% waste 1 1 1 1 1.5 1 1.05 1.50 spinning Electricity kwh 1.83 US average energy mix 1.5 1 1 1 2 1.2 1.05 2.28

255 Tube Acrylonitrile kg 2.06 Ecoinvent process Built-in uncertainty for butadine styrene adapted to US material Ecoinvent processes (ABS) and energy mix. Assuming 5% waste Extrusion of plastic kg 2.07 Ecoinvent process Built-in uncertainty for pipes (ABS) adapted to US material Ecoinvent processes and energy mix. Membrane Polyurethane Glue kg 1.37 Information from Built-in uncertainty for assembly membrane Ecoinvent processes manufacturers Electricity US kwh 1.18 Information from 1.05 1.05 1 1.02 1.2 1.1 1.05 1.25 energy mix membrane manufacturers Packaging Cardboard box kg 1.32 Measured 1 1.05 1 1 1 1.05 1.05 1.09 and preservation polyethylene bag g 70.00 Measured 1 1.05 1 1 1 1.05 1.05 1.09 Sodium g 50.00 Measured 1 1 1 1 1.5 1.02 1.05 1.51 metabisulphate Water L 5.00 Measured 1 1.05 1 1 1 1.05 1.05 1.09 O-ring Synthetic rubber g 31.50 Assuming 5% waste 1 1.05 1 1 1 1.05 1.05 1.09 Outputs Polysulfone Carbon dioxide g 159.91 1 1.02 1 1 1.5 1 1.05 1.51 layer production Waste water g 6089.00 1 1.02 1 1 1.5 1 1.05 1.51 Solved organics g 58.93 1 1.02 1 1 1.5 1 1.05 1.51

256 Table C-2: Life cycle inventory and uncertainty values for end-of-life scenarios

Amount used Min Max Notes on Scenario Process Unit Notes U1 U2 U3 U4 U5 U6 Ub SD95 per value value Uncertainty module Landfill Inputs Rigid truck, 3.5-16t, Transport tkm 0.7 0.03 1.94 Australian fleet average Air and water emissions are dictated by the Outputs material specific processes in the Ecoinvent database Process includes Material to be kg 13.5 landfill construction landfilled and operation Incineration Inputs Rigid truck, 3.5-16t, Transport tkm 0.7 0.03 1.94 Australian fleet average Shredding of module. 0.024 kwh Module kWh 0.32 per kg. (Grant and 1.2 1.2 1 1.1 1.5 1.05 1.05 1.64 shredding James, 2005; Shonfield, 2008)

257 Emissions are included in Ecoinvent processes for incineration. Outputs Specific incineration processes for each material type were used. Electricity kWh 10.18 1.2 1.02 1 1.1 1.5 1.05 1.05 1.58 Heat energy MJ 74.44 1.2 1.02 1 1.1 1.5 1.05 1.05 1.58 (Coal offset) Solid residue kg 1.01 1.2 1.02 1 1.1 1.5 1.05 1.05 1.58 to landfill Gasificatio Inputs n Rigid truck, 3.5-16t, Transport tkm 0.7 0.03 1.94 Australian fleet average Shredding of module. 0.024 kwh Module kWh 0.32 per kg. (Grant and 1.2 1.2 1 1.1 1.5 1.05 1.05 1.64 shredding James, 2005; Shonfield, 2008) Process data based on survey of Uncertainty available technology based on max Electricity kWh 4.79 2.77 6.8 (American Chemistry and min process Council, 2012; US values. EPA, 2012). Water L 56.76 28.36 85.18 natural gas Kg 0.32 0.10 0.55 Outputs Electricity kWh 15.45 12.83 18.06 Residual gas kg 1.35 0 2.69

258 0.056 Sulphur kg 0.069 0.082 6 0.056 Salt kg 0.069 0.082 6 Slag kg 1.41 0.15 2.67 Char kg 0.93 0 1.87 Slag kg 0.24 0 0.47 Solid residue kg 0.46 0.16 0.76 inorganic kg 0.14 0 0.28 sludge Wastewater L 52.51 31.5 73.52 Air emissions 6.29 PM kg 0.0011 0.0022 E-5 3.15E- PM10 kg 0 6.29E-6 06 CO2 bio kg 1.47 0 2.94 CO2 fossil kg 3.47 2.17 6.59 1.26 CH4 kg 0.0063 0.0126 E-6 0.00018 HCL kg 9.44E-5 0 9 SO2 kg 0.00127 0 0.00252 sulphur oxide kg 1.57E-7 0 3.15E-7 6.29 N2O kg 0.00127 0.00252 E-6 0.001 NO2 kg 0.00377 0.00629 26 0.000 CO kg 0.00346 0.00629 66 Hg kg 1.89E-9 0 3.78E-9

259 Cd kg 2.52E-8 0 5.03E-8 Lead kg 3.15E-8 0 6.29E-8 EAF Inputs Articulated truck, 20t, Transport tkm 26.5 1.02 61.11 Australian fleet average Plastic washing Water L 88.3 (Ecoinvent Centre, 1.05 1.1 1 1.02 1.5 1.1 1.05 1.54 2013) Plastic grinding (Grant and James, Electricity kWh 0.278 1.2 1.2 1 1.1 1.5 1.05 1.05 1.64 2005; Shonfield, 2008) Membrane Electricity kWh 0.225 1.1 1.1 1 1.02 2 1.1 1.05 2.04 disassembly Outputs Offset of Includes mining and metallurgical kg 11.6 refining of 1.05 1.1 1 1.02 1.5 1.05 1.05 1.53 coke metallurgical coke Coal Transport from mine transport - tkm 34.8 to onesteel plant, 1.05 1.05 1 1 1.2 1 2 2.05 Rail Sydney Uncertainty of landfill of the weight of fibreglass kg 1.83 1 1.02 1 1.02 1 1.05 1.05 1.08 membrane casing component Transport of fibreglass 0.003 tkm 0.0756 0.209 case to 2 landfill

260 Additional inputs defined by adapted Recycling Inputs Australasian LCI processes for plastic recycling Rigid truck, 3.5-16t, Transport tkm 1.27 0.09 3.14 Australian fleet average Membrane Electricity kWh 0.225 1.1 1.1 1 1.02 2 1.1 1.05 2.04 disassembly Based on Visy Uncertainty in Plastics process. ABS the amount of kg 2.26 Includes 5% loss 1.2 1.02 1 1.02 1 1.05 1.05 1.23 reprocessing recoverable during disassembly material and sorting Based on Visy Uncertainty in Plastics process. PET the amount of kg 1.72 Includes 5% loss 1.2 1.02 1 1.02 1 1.05 1.05 1.23 reprocessing recoverable during disassembly material and sorting Based on Visy Uncertainty in Plastics process. PP the amount of kg 1.38 Includes 5% loss 1.2 1.02 1 1.02 1 1.05 1.05 1.23 reprocessing recoverable during disassembly material and sorting Outputs 10% loss due to ABS offset kg 2.06 reprocessing and 1.2 1.02 1 1.02 1.5 1.1 1.05 1.58 quality loss 10% loss due to PET offset kg 1.55 reprocessing and 1.2 1.02 1 1.02 1.5 1.1 1.05 1.58 quality loss 10% loss due to Polypropylen kg 1.24 reprocessing and 1.2 1.02 1 1.02 1.5 1.1 1.05 1.58 e offset quality loss

261 Remaining components kg 8.11 to be landfilled Direct RO Inputs reuse Articulated truck Transport tkm 37.24 1.137 68.34 >20t, fleet average AU Membrane Polyethylene g 70 1 1.05 1 1 1 1.05 1.05 1.09 repackaging Sodium Membrane metabisulpha g 50 1 1 1 1 1.5 1.02 1.05 1.51 preservation te Membrane Water L 5 1 1.05 1 1.02 1 1.02 1.05 1.08 preservation Citric acid kg 1.2 Membrane cleaning 1.1 1.05 1 1.02 1.5 1.1 1.05 1.54 Sodium kg 0.12 Membrane cleaning 1 1.02 1 1.02 1.2 1.1 1.05 1.24 hydroxide Water L 120 Membrane cleaning 1 1.02 1 1.02 1.2 1.1 1.05 1.24 Ammonium kg 0.012 Membrane cleaning 1 1.02 1 1.02 1.2 1.1 1.05 1.24 hydroxide Electricity kWh 0.248 Membrane cleaning 1.05 1.05 1 1.02 1.2 1.1 1.05 1.25 Average membrane year 7.5 3 10 lifetime Membrane year 2.5 0.5 5 reuse lifetime Membrane .33 .1 .5 offset ratio Outputs Wastewater L 120 1 1.02 1 1.02 1.2 1.1 1.05 1.24 Conversion Inputs to UF

262 Articulated truck Transport tkm 37.24 1.137 68.34 >20t, fleet average AU Membrane Polyethylene g 70 1 1.05 1 1 1 1.05 1.05 1.09 repackaging Sodium Membrane metabisulpha g 50 1 1 1 1 1.2 1.02 1.05 1.21 preservation te Membrane Water L 5 1 1.05 1 1.02 1 1.02 1.05 1.08 preservation Membrane Water L 100 1 1.02 1 1 1.2 1.02 1.05 1.21 compaction Membrane Electricity kWh 3.9 1 1.02 1 1 1.2 1.02 1.05 1.21 compaction Sodium Membrane Hypochlorite kg 3.6 1 1.02 1 1 1.5 1 1.05 1.51 conversion 15% in H2O Membrane Water L 27 1 1.02 1 1 1.2 1.02 1.05 1.21 conversion Membrane Electricity kWh 0.31 1 1.02 1 1 1.2 1.02 1.05 1.21 conversion Average membrane year 7.5 3 10 lifetime Membrane year 2.5 0.5 5 reuse lifetime Membrane .33 .1 .5 offset ratio Outputs Wastwater L 137 1 1.02 1 1 1.2 1.02 1.05 1.21 Sodium hypochorite kg 3.6 1 1.02 1 1 1.5 1 1.05 1.51 in wastewater

263 C.3. Additional Results

Table C-3: Overall impacts for membrane manufacturing, including breakdown into sub components.

Fibre Polysulfone Polyester Structural Impact category Unit Total Transport Assembly glass and polyamide support Spacers components case layers layer kg CO2 Climate change 87.72 7.41 0.89 19.06 9.80 17.96 21.83 10.78 eq mg CFC- Ozone depletion 5.24 0.46 0.03 0.72 0.62 1.86 1.27 0.28 11 eq kg 1,4-DB Human toxicity 6.53 0.35 0.03 0.50 2.49 1.62 1.05 0.49 eq Photochemical kg 0.37 0.08 0.00 0.06 0.06 0.07 0.06 0.04 oxidant formation NMVOC Terrestrial kg SO2 0.45 0.08 0.01 0.07 0.05 0.11 0.09 0.05 acidification eq Freshwater g P eq 20.33 1.46 0.44 2.17 2.62 6.11 5.12 2.41 eutrophication Marine g N eq 138.10 4.76 0.17 4.86 3.22 120.82 2.80 1.47 eutrophication kg 1,4-DB Ecotoxicity 0.16 0.01 0.00 0.02 0.02 0.06 0.04 0.01 eq

Fossil depletion kg oil eq 38.48 2.32 0.23 8.57 3.05 9.05 9.15 6.10

264 Table C-4: Overall impacts of disposal scenarios, including membrane manufacturing and distribution

Electric Arc Material Direct RO Conversion Impact category Unit Landfill Incineration Syngas Furnace Recycling reuse to UF

Climate change kg CO2 eq 88.44 98.65 81.44 84.22 79.02 66.82 72.17

Ozone depletion mg CFC-11 eq 5.27 5.38 5.25 5.36 5.05 3.74 3.77

Human toxicity kg 1,4-DB eq 6.55 6.80 6.62 6.55 6.15 4.63 4.77

Photochemical kg NMVOC 0.37 0.33 0.35 0.38 0.37 0.28 0.30 oxidant formation Terrestrial kg SO2 eq 0.46 0.37 0.40 0.46 0.44 0.34 0.37 acidification Freshwater g P eq 20.4 20.5 20.5 20.4 18.6 14.2 14.5 eutrophication Marine kg N eq 0.25 0.14 0.14 0.14 0.23 0.16 0.16 eutrophication

Ecotoxicity kg 1,4-DB eq 0.16 0.20 0.16 0.16 0.14 0.11 0.11

Fossil depletion kg oil eq 38.61 34.40 36.24 32.69 32.30 29.14 29.85

265

Table C-5: Material composition breakdown of different sizes of RO Membranes

4” Element 8” Element 16” Element Material

wt% of wt% of wt% of Component kg kg kg element element element Membrane PET, PSf, 0.24 19 4.63 34.3 19.7 38 Sheet PA TFC Feed Spacer 0.09 7 1.45 10.8 6.16 12 PP

Permeate 0.1 8 1.81 13.4 7.6 14.7 PET Spacer

Glue 0.11 9 1.37 10.1 5.2 10.1 PU Fiberglass Fibre Glass 0.37 30 1.83 13.6 6 11.6 with PET Case resin Permeate tube/end 0.34 27 2.38 17.6 7 13.5 Abs caps Total 1.25 13.5 51.7

266 C.4. Uncertainty Calculations

Equation C-1: Formula for the geometric standard deviation calculated using the

pedigree matrix.

Table C-6: Basic uncertainty factors for inputs, outputs and elementary flows. C:

Combustion emissions, P: Process emissions, A: agricultural emissions (Frischknecht

et al., 2007)

267 Table C-7: Pedigree matrix for uncertainty calculation (Frischknecht et al., 2007).

268 C.5. Pair Wise Monte Carlo Comparisons

Table C-8: Monte-Carlo simulation results for emissions and consumptions produced from pairwise comparison of end-of-life membrane options. Values represent the percentage of runs where the column variable (A) has higher emission/consumption compared to its corresponding row variable (B). A result can be considered statistically significant if result direction is consistent for over 90% of runs (Goedkoop et al., 2010).

Freshwater ecotoxicity A Incineratio Gasificatio Recyclin UF Direct EAF B n n g Conversion Reuse Landfill 100% 1% 20% 0% 0% 0% Incineration 0% 0% 0% 0% 0% Gasification 96% 0% 0% 0% EAF 0% 0% 0% Recycling 6% 2% Conversion 57% Freshwater eutrophication Landfill 62% 87% 62% 0% 0% 0% Incineration 52% 43% 0% 0% 0% Gasification 36% 0% 0% 0% EAF 0% 0% 0% Recycling 3% 1% Conversion 59% Human Toxicity Landfill 73% 63% 64% 0% 0% 0% Incineration 30% 25% 1% 0% 0% Gasification 45% 0% 0% 0% EAF 0% 0% 0% Recycling 2% 0% Conversion 61% Marine ecotoxicity Landfill 99% 1% 61% 0% 0% 0% Incineration 1% 0% 0% 0% 0% Gasification 86% 0% 0% 0% EAF 0% 0% 0% Recycling 4% 3% Conversion 59% A has significantly lower emissions than B A has significantly higher emissions than B Difference between A and B not significant

269 Marine eutrophication A UF Direct B Incineration Gasification EAF Recycling Conversion Reuse Landfill 0% 0% 0% 0% 2% 5% Incineration 15% 60% 100% 84% 79% Gasification 100% 100% 86% 82% EAF 100% 83% 82% Recycling 5% 6% Conversion 57% Ozone depletion Landfill 56% 3% 89% 0% 0% 0% Incineration 45% 48% 37% 17% 15% Gasification 94% 0% 0% 0% EAF 0% 0% 0% Recycling 0% 0% Conversion 57% Photochemical oxidant formation Landfill 0% 0% 82% 55% 2% 0% Incineration 98% 99% 100% 30% 17% Gasification 100% 100% 15% 6% EAF 20% 1% 0% Recycling 2% 0% Conversion 66% Terrestrial acidification Landfill 0% 0% 77% 1% 2% 0% Incineration 99% 100% 100% 60% 29% Gasification 100% 100% 34% 12% EAF 1% 1% 0% Recycling 7% 0% Conversion 73% Terrestrial ecotoxicity Landfill 55% 0% 77% 0% 6% 12% Incineration 36% 41% 27% 13% 15% Gasification 100% 0% 11% 11% EAF 0% 9% 11% Recycling 13% 16% Conversion 47% A has significantly lower emissions than B A has significantly higher emissions than B Difference between A and B not significant

270 C.6. Status of end-of-life options in Australia

Table C-9: Summary of end-of-life options for RO membranes.

Manual Plastic Companies capable of this type of recycling in Name Fibreglass AUS World disassembly Comments components Australia required Landfill      Current standard practice Many landfills around Australia Technically possible but currently no Incineration      Not currently commercially possible operating plants in Australia Bioplant – www.Bioplant.com – VIC Pacific Pyrolysis Pty Ltd - www.pacificpyrolysis.com Technically possible but currently no Pyrolysis      – NSW large scale plants in Australia New Energy Corp - www.newenergycorp.com.au - WA Technically possible but many Bluescope Steel - http://www.bluescopesteel.com.au Electric Arc Furnace      challenges involved - NSW Material reuse options - Concrete aggregate   # #  Not currently used n/a This application needs to be self-directed by the - Geotextiles   # #  Limited, small scale applications user or additional third party Advanced Plastic Recycling - Possible for spacers and membrane - Wood plastic composite      www.advancedplasticrecycling.com.au – SA sheet. Cosset - www.cosset.com.au - SA SITA – www.sita.com.au – NSW Only mechanical recycling available in Material recycling      AP Recycling - http://www.aprecycling.com – Nation Australia wide Life cycle extension options - Direct Reuse      Technically possible and proven Third party initiative - UF Conversion      Technically possible but unproven Third party initiative  Currently possible  Currently not possible Technically possible but with # limitations

271 Table C-10: Summary of survey of Australian landfill and recycling costs (2015)

Commercial waste Location State Recycling Cost landfill (AUD/tonne) Artarmon NSW 329 Auburn NSW 308 Belrose NSW 329 chullora NSW 308 Rockdale NSW 329 Ryde NSW 308 Seven Hills NSW 308 spring farm NSW 308 wetherill NSW 308 Free for non- Maitland NSW NSW 246 commercial if sorted Free if sorted and Shellharbour NSW 312 clean Summerhill Waste Management NSW 230 Centre - newcastle Kimbriki - terrey NSW 290 hills Bundaberg QLD 86 Free if sorted regional council Gladstone QLD 133 Free if sorted Mackay - Paget waste management QLD 145 Free if sorted facitlity Mackay - Hogan's waste management QLD 115 Free if sorted facitlity Gold coast QLD 89 Brisbane city QLD 116 council Tablelands Free if sorted and QLD 78 regional council clean Free if clean and Logan city council QLD 126 sorted Toowoomba QLD 107 Free if sorted Regional Council City of Armadale VIC 197 West Melbourne VIC 155 Grantville VIC 160 Mornington VIC 215 Fleurieu Peninsula SA 124 Darwin council NT 75 ACT ACT Free if sorted Free if sorted Perth EMRC WA 128

272