Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An ion exchange loading correlation for the simulation of base metal resin-in-pulp circuits McKevitt, Bethan Ruth 2012

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2012_spring_mckevitt_bethan.pdf [ 5.57MB ]
Metadata
JSON: 24-1.0072620.json
JSON-LD: 24-1.0072620-ld.json
RDF/XML (Pretty): 24-1.0072620-rdf.xml
RDF/JSON: 24-1.0072620-rdf.json
Turtle: 24-1.0072620-turtle.txt
N-Triples: 24-1.0072620-rdf-ntriples.txt
Original Record: 24-1.0072620-source.json
Full Text
24-1.0072620-fulltext.txt
Citation
24-1.0072620.ris

Full Text

    AN ION EXCHANGE LOADING CORRELATION FOR THE SIMULATION OF BASE METAL RESIN-IN-PULP CIRCUITS  by BETHAN RUTH MCKEVITT  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Materials Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  March 2012  © Bethan Ruth McKevitt, 2012 ii  ABSTRACT This thesis investigates the loading rate of nickel onto an iminodiacetic ion exchange resin, with the goal of gaining insight into operation of a base metals resin-in-pulp (RIP) circuit.  A modified pH-stat method was used to generate loading data under infinite solution volume conditions.  A comparison of several commercially available iminodiacetic resins was performed, using this method. The standard linear approach to equilibrium was found to fit well to the portion of loading under film diffusion control, but none of the standard engineering approximations could adequately describe the data from the intraparticle diffusion / exchange rate limited regime.  A hybrid correlation was developed and was found to adequately describe single element loading of nickel, copper, and cobalt from synthetic solution.  In order to know which of the two models to use at a given point in time, a modified Helfferich number was derived.  This dimensionless number can be used to track when the resin bead switches from film diffusion control to intraparticle diffusion control. The  experimental  fit  parameters  of  the  film  diffusion  model  and  the  hybrid  correlation  were used to successfully predict the results of batch experiments with varying solution concentration.  The effectiveness of these equations were also assessed and verified through the operation of a five stage RIP miniplant, using synthetic solutions. Using these circuit models, various RIP circuit operating configurations were simulated. Results from these simulations suggest that all operating configurations and strategies have both advantages and disadvantages.  A table summarizing the process sensitivity of a cascade circuit to various operating conditions has been generated.  iii  PREFACE Much of the work presented in this thesis has previously been presented at conferences, or is under review for journal publication.  The overall supervision of this research program was provided by Dr. Dreisinger. Modifications to the pH-stat procedure, as described in Chapter 3, were developed by the candidate.  The batch tests reported in Chapter 4 and in Figure 8.3 were performed by Ms. Abbasi; the remainder of the batch tests were performed by the candidate. The candidate worked with the Materials Engineering Shop to design and set up the RIP miniplant.  For operation of the RIP miniplant, the candidate developed the test procedures, provided training to the dayshift operators, and operated the RIP miniplant on nightshift. The data analysis of all test results, the development of the hybrid correlation, the derivation of the modified Helfferich number, the construction of the circuit models, and the circuit sensitivity analysis of Chapter 12 was done by the candidate.  The following is a list of publications and articles submitted to a journal for consideration, that have had material reprinted as part of this thesis.  The following manuscripts were prepared by the candidate and edited by Dr. Dreisinger, prior to submission for publication.  1)   McKevitt, B. & Dreisinger, D.  (2010)  Lessons from the past: A literature review of the resin-in-pulp process for the uranium industry.  Proceedings of the third international conference on uranium held August 15-18, 2010.  Uranium 2010: The Future is U.  Canadian Institute of Mining, Metallurgy and Petroleum: vol. II (ISBN 1-894475-98-4), pp. 3-14.  This paper was based on information gained by the candidate during the literature review portion of this project.  Portions of this paper appear in Sections 2.2 and 2.3.   iv  2)   McKevitt, B., Abbasi, P. & Dreisinger, D.  (2011)  A Comparison of Large Bead Ion Exchange Resins for the Recovery of Base Metals in a Resin-in-Pulp (RIP) Circuit. Proceedings of the 6th Southern African Base Metals Conference.  SAIMM: pp. 337-352.  This paper was based on laboratory tests conducted by Ms. Abbasi, using the experimental procedure developed as part of this thesis.  Ms. Abbasi prepared the resin samples, conducted the experiments in the laboratory, and performed the AA assays of the samples taken.  Portions of this paper appear in Chapter 4.  3)   McKevitt, B. & Dreisinger, D.  (under consideration).  Development of an engineering model for nickel loading onto an iminodiacetic resin for resin-in-pulp applications: Part I – Method development and discussion of rate limiting factors.  Parts of this manuscript appear in Section 2.2.1, Chapter 3, Section 4.1, Chapter 5, Chapter 6, Section 8.1.4 and Section 8.1.5.  4)   McKevitt, B. & Dreisinger, D.  (under consideration).  Development of an engineering model for nickel loading onto an iminodiacetic resin for resin-in-pulp applications: Part II – Evaluation of existing models and their extension into a hybrid correlation.  Parts of this manuscript appear in Section 2.1.2, Chapter 7, and Chapter 8.  5)   McKevitt, B. & Dreisinger, D.  (under consideration).  Development of an engineering model for nickel loading onto an iminodiacetic resin for resin-in-pulp applications: Part III – Verification of hybrid correlation.  Parts of this manuscript appear in Chapter 9 and Chapter 10.   v  TABLE OF CONTENTS Abstract.................................................................................................................................... ii Preface .................................................................................................................................... iii Table of Contents ..................................................................................................................... v List of Tables ......................................................................................................................... xii List of Figures ....................................................................................................................... xiv List of Abbreviations and Acronyms .......................................................................................xx List of Symbols ..................................................................................................................... xxi Acknowledgements ..............................................................................................................xxiv 1.0  Introduction ....................................................................................................................... 1 2.0  Literature Review .............................................................................................................. 2 2.1  Key Concepts from Ion Exchange Fundamentals ................................................... 2 2.1.1  Equilibrium Capacity ............................................................................... 2 2.1.2  Rate of Ion Exchange .............................................................................. 5 2.1.2.1  Fundamental Models for Ion Exchange Loading Rate ............... 5 2.1.2.2  Engineering Models for Ion Exchange Loading Rate ................. 7 2.1.2.2.1  Linear Approach to Equilibrium ................................. 7 2.1.2.2.2  Vermeulen’s Quadratic Approach to Equilibrium ....... 8 2.1.2.2.3  Shrinking Core Model ................................................ 9 2.1.2.2.4  Second Order Rate Model ..........................................10 2.1.3  The Iminodiacetic Chelating Resin .........................................................11 2.2  Development of the Commercial RIP Process: The Uranium Industry ...................12 vi  2.2.1  Definition of Terms Relevant to RIP.......................................................13 2.2.2  United States of America Basket RIP Process .........................................15 2.2.3  South Africa: Floating Resin Processes ...................................................17 2.2.4  United States of America: Continuous RIP .............................................18 2.2.5  Davy-McKee Contactor ..........................................................................20 2.2.6  Union of the Soviet Socialist Republics Technologies ............................21 2.3  Modern RIP Installations (Commercial Scale).......................................................22 2.3.1  Gold RIP ................................................................................................22 2.3.2  Uranium RIP ..........................................................................................23 2.3.3  Base Metals RIP .....................................................................................24 2.3.4  Modern RIP Systems Marketed in the Literature ....................................25 2.3.4.1  Anglo American Corporation (AAC) / Kemix Pump Cells .......25 2.3.4.2  Mintek / Bateman MetRIX .......................................................26 2.3.4.3  Clean-TeQ Clean iX .................................................................26 2.4  Research Studies for Base Metals RIP ...................................................................27 2.4.1  Mintek Studies .......................................................................................27 2.4.2  Clean-TeQ Study ....................................................................................28 2.4.3  Tshwane University Study ......................................................................28 2.4.4  Vale Study ..............................................................................................29 2.4.5  UBC Study .............................................................................................29 2.4.6  Murdoch Study .......................................................................................29 2.5  Challenges for Base Metal RIP Circuits ................................................................30 2.5.1  Resin Losses ...........................................................................................31 2.5.2  Resin Distribution Across RIP Circuit ....................................................32 2.6  Objectives for Current Work .................................................................................33 3.0  Method Development for Batch Tests ...............................................................................34 3.1  Method Development ............................................................................................35 3.2  Detailed Description of Experimental Procedure ...................................................40 vii  4.0  Comparison of Large Bead Iminodiacetic Resins ..............................................................42 4.1  Size Distributions of Resins ..................................................................................42 4.2  Loading Rate and Capacities .................................................................................46 4.2.1  Loading Rate at 500 ppm Nickel.............................................................46 4.2.2 Loading Rate at 2500 ppm Nickel .........................................................50 4.3  Stripping Comparison ...........................................................................................54 4.4  Conclusions from Resin Comparison ....................................................................59 4.5  Selection of Resin for Further Work ......................................................................59 5.0  Qualitative Investigation into Rate Determining Step(s) ...................................................60 5.1  Effect of Resin Bead Diameter on Loading Rate ...................................................62 5.2  Effect of Bulk Solution Concentration on Loading Rate ........................................65 5.3  Interruption Test ...................................................................................................65 5.4  Effect of Complex Formation Rate ........................................................................66 5.5  Conclusions from Qualitative Tests .......................................................................68 6.0  Quantitative Model for Film Diffusion Component...........................................................69 6.1  Film Diffusion Model Fit ......................................................................................69 6.2  Theoretical Estimate of Film Diffusion Coefficient ...............................................71 7.0  Development of Hybrid Correlation for Intraparticle Diffusion .........................................73 7.1  Evaluation of Existing Engineering Loading Models .............................................73 7.1.1  Linear Approach to Equilibrium .............................................................74 7.1.2  Vermeulen’s Quadratic Approach to Equilibrium ...................................76 7.1.3  Shrinking Core Model ............................................................................79 viii  7.1.4  Second Order Rate Model .......................................................................81 7.2  Development of Hybrid Correlation ......................................................................84 8.0  Application of Hybrid Correlation to Other Batch Systems ...............................................88 8.1  Fitting the Hybrid Correlation to Additional Data Sets ..........................................88 8.1.1  Nickel Loading onto TP207 ....................................................................89 8.1.2  Nickel Loading onto TP207XL (Lot CHC 5034 –Resin Comparison) .....90 8.1.3  Nickel Loading onto TP207XL (Lot CHE 50019 –Miniplant).................91 8.1.4  Cobalt Loading onto TP207MP ..............................................................92 8.1.5  Copper Loading onto TP207MP .............................................................93 8.1.6  Copper Loading onto TP207XL (Lot CHE 50019 –Miniplant) ...............94 8.2  Physical Significance of Hybrid Correlation Fit Parameters ..................................95 8.2.1  Apparent Diffusivity ...............................................................................95 8.2.2  Alpha .....................................................................................................96 8.2.3  Qualitative Evaluation of Effect of pH on Alpha .....................................97 8.3  Effect of Bead Size on Hybrid Correlation .......................................................... 100 8.4  Limitations of Hybrid Correlation ....................................................................... 101 9.0  Verification of Hybrid Correlation (Batch Tests) ............................................................ 103 9.1  Modified Helfferich Number ............................................................................... 103 9.2  Batch Tests with Changing Solution Concentrations ........................................... 105 9.2.1 Dual Concentration Test ........................................................................ 105 9.2.2 Test with Increasing Solution Composition ............................................ 107 9.2.3 Tests with Decreasing Solution Composition ......................................... 111 9.2.4  Summary of Batch Test Results and Assessment of Mixed Control Regime........................................................................................................... 115 10.0  Development of RIP Circuit Models ............................................................................. 116 ix  10.1  Cascade Circuit Model ...................................................................................... 116 10.2 Carousel Circuit Model ...................................................................................... 121 11.0  Verification of Circuit Models (RIP Miniplant) ............................................................ 124 11.1  Nickel Cascade Runs ........................................................................................ 127 11.2  Copper Cascade Run ......................................................................................... 130 11.3  Nickel Carousel Run ......................................................................................... 132 11.4  Copper Carousel Run ........................................................................................ 136 11.5  Validity of Constant Volume Assumption ......................................................... 139 11.6  Conclusions from Miniplant Testwork .............................................................. 140 12.0  RIP Circuit Simulation and Results ............................................................................... 141 12.1  Effect of Resin Volume in Tanks and Solution Backflow on Cascade Circuit .... 141 12.1.1  Resin Volumes & Solution Back-Mixing: Nickel Cascade .................. 142 12.1.2  Resin Volumes & Solution Back-Mixing: Copper Cascade ................. 148 12.2  Effect of Size and Number of Tanks on Cascade Circuit Performance ............... 150 12.2.1  Nickel Cascade Circuit ....................................................................... 151 12.2.2  Copper Cascade Circuit ...................................................................... 152 12.3  Process Sensitivity of a Cascade Circuit ............................................................ 153 12.3.1 Effect of Uneven Resin Distribution .................................................... 154 12.3.2  Effect of Losing Resin Volume in Loading Circuit ............................. 157 12.3.3  Effect of Swings in Feed Concentration .............................................. 158 12.3.4  Summary: Cascade Circuit Process Sensitivity ................................... 159 12.4  Resin Utilization: Cascade vs. Carousel Operation ............................................ 159 12.5  Effect of Contactor Size, Cycle Time, and Number of Contactors -Carousel ..... 162 12.5.1  Effect of Contactor Size ...................................................................... 162 x  12.5.2  Effect of Cycle Time .......................................................................... 163 12.5.3  Effect of the Number of Contactors .................................................... 165 12.6  Effect of Resin Properties on Circuit Performance ............................................ 166 12.6.1  Effect of Resin Capacity ..................................................................... 167 12.6.2  Effect of Variation in Dapp on Nickel Loading ..................................... 170 12.6.3  Effect of Resin Bead Diameter on Nickel Loading .............................. 172 12.6.4  Effect of Recirculating Load of Nickel on Circuit Performance .......... 174 12.6.5  Summary of Effect of Resin Properties on Circuit Performance .......... 176 12.7  Resin Comparison: Lewatit TP207 vs Lewatit MonoPlus TP207XL .................. 177 12.8  Key Findings from Circuit Simulations ............................................................. 180 13.0  Conclusion ................................................................................................................... 182 13.1  Conclusions from Research ............................................................................... 182 13.1.1  Adequately Describe Loading Under Infinite Solution Volume Conditions ...................................................................................................... 182 13.1.2  Verification of Engineering Models .................................................... 184 13.1.3  Quantify Effect of Various Operating Conditions on Circuit Performance  ...................................................................................................................... 185 13.2  Recommendations for Further Work ................................................................. 187 13.2.1  Fundamental Loading Model for Iminodiacetic Resins in Hydrogen Form  ...................................................................................................................... 187 13.2.2  Application of Hybrid Correlation to Actual Slurries .......................... 188 13.2.3  Circuit Modeling With Variations in Resin Diameters ........................ 189 References ............................................................................................................................ 190 Appendix A:  Blank Batch Test Check Sheet ......................................................................... 203 Appendix B: Manufacturer Provided Resin Properties ........................................................... 204 xi  Appendix C: Loading Isotherms ............................................................................................ 205 Appendix D: Qualitative Tests on Effect of pH (Copper Alpha Value) .................................. 209 Appendix E: Derivation of Modified Helfferich Number ....................................................... 210 E.1  Derivation of Standard Helfferich Number ......................................................... 210 E.2  Derivation of Modified Helfferich Number ......................................................... 211 Appendix F: Carousel Finite Difference Model ..................................................................... 213 Appendix G: RIP Miniplant Operating Procedures –Cascade ................................................ 217 Appendix H: Estimate of Airlift Total Flowrates ................................................................... 223 Appendix I: RIP Miniplant Cascade Circuit Model Output .................................................... 224 Appendix J: RIP Miniplant Operating Procedures: Carousel .................................................. 241 Appendix K: Copper Resin Volume Optimization ................................................................. 249  xii  LIST OF TABLES Table 2.1: Iminodiacetic Functional Group at Different pH Values –from (1) ..........................11 Table 2.2: Gold Recoveries at the Penjom Gold Mine 1998-2002 ............................................23 Table 3.1:  Results of Blanks with Different Neutralizing Agents ............................................37 Table 4.1: Effective Resin Diameters for Modeling Purposes ..................................................46 Table 4.2: Mass Balances and Capacities at 500 ppm Nickel ...................................................50 Table 4.3: Mass Balances and Capacities at 2500 ppm Nickel .................................................54 Table 5.1: Impact of Various Parameters Under Different Rate Limiting Regimes...................61 Table 7.1: Development of Hybrid Correlation Equation .........................................................86 Table 8.1: Isotherm Fit Parameters ..........................................................................................88 Table 8.2: Summary of Hybrid Fit Parameters .........................................................................95 Table 8.3:  Effect of Target pH on Test Conditions ..................................................................98 Table 8.4: Hybrid Correlation Fit Parameters for Copper Loading at pH 4, 3, 2 .......................99 Table 11.1: Summary of Model Inputs and Sampling Times for Nickel Cascade Tests .......... 128 Table 11.2: Modified Helfferich Numbers for Nickel Cascade Tests ..................................... 130 Table 11.3: Summary of Model Inputs and Sampling Times for Copper Cascade Test .......... 131 Table 11.4: Modified Helfferich Numbers for Copper Cascade Test ...................................... 132 xiii  Table 11.5: Summary of Model Inputs and Sample Times for Nickel Carousel Test .............. 134 Table 11.6: Summary of Model Inputs and Sample Times for Copper Carousel Test ............. 137 Table 11.7: Estimate of Solution Volume Errors.................................................................... 140 Table 12.1: Key Parameters for Nickel Loading onto TP207 and MonoPlus TP207XL.......... 177 Table 12.2:  Sensitivity of Cascade Circuits........................................................................... 181 Table 13.1:  Sensitivity of Cascade Circuits (Repeat of Table 12.2) ....................................... 186 Table B.1: Resin Properties from Manufacturer Data Sheets ................................................. 204  xiv  LIST OF FIGURES Figure 2.1: Favourable vs Unfavourable Equilibrium Isotherms ............................................... 3 Figure 2.2: Iminodiacetic Resin Decomplexing pH for Divalent Ions –Based on (32) ..............12 Figure 2.3: Cascade Operating Strategy ...................................................................................14 Figure 2.4: Carousel Operating Strategy ..................................................................................15 Figure 3.1:  Batch Loading Equipment ....................................................................................35 Figure 3.2:  Nickel Concentration in Beaker for Duration of Test ............................................36 Figure 3.3:  Repeatability at 250 ppm Nickel ...........................................................................38 Figure 3.4: Effect of Increasing Agitation Speed at 250 ppm Nickel ........................................39 Figure 4.1: Resin Volume Change as Converted from Hydrogen Form to Sodium Form..........43 Figure 4.2: Cumulative Fraction Passing –Hydrogen Form ......................................................44 Figure 4.3: Comparison of Wet-Screened Results with Data Provided by Lanxess ..................45 Figure 4.4: Loading of 500-600 Micron Beads at 500 ppm Nickel ...........................................47 Figure 4.5: Loading of 710-850 Micron Beads at 500 ppm Nickel ...........................................48 Figure 4.6: Loading of 1000-1180 Micron Beads at 500 ppm Nickel .......................................49 Figure 4.7: Loading of 500-600 Micron Beads at 2500 ppm Nickel .........................................51 Figure 4.8: Loading of 710-850 Micron Beads at 2500 ppm Nickel .........................................52 xv  Figure 4.9: Loading of 1000-1180 Micron Beads at 2500 ppm Nickel .....................................53 Figure 4.10: Cumulative Nickel Stripped from 500 – 600 Micron Beads .................................55 Figure 4.11: Effect of Stripping Flowrate on IRC78 and TP207 (500-600 Micron) ..................56 Figure 4.12: Cumulative Nickel Stripped from 710 – 850 Micron Beads .................................57 Figure 4.13: Cumulative Nickel Stripped from 1000 – 1180 Micron Beads .............................58 Figure 5.1: Effect of Resin Bead Diameter on Loading Rate of TP207 ....................................62 Figure 5.2:  Effect of Resin Bead Diameter on Loading Rate ...................................................64 Figure 5.3: Effect of Solution Concentration on Nickel Loading Rate ......................................65 Figure 5.4: Interruption Test -68 Hour Interruption after one hour of loading ..........................66 Figure 5.5: Effect of Solution Concentration on Copper Loading Rate.....................................67 Figure 6.1: Loading Rate of Nickel and Copper onto TP207MP -First 30 Minutes ..................69 Figure 6.2: Linear Approach to Equilibrium Fit to First 12 Minutes of Data ............................71 Figure 7.1: Equilibrium Isotherm for Nickel Loading onto TP207MP......................................74 Figure 7.2:  Linear Approach to Equilibrium Fit to Individual Loading Curves .......................75 Figure 7.3: Comparison of Vermeulen’s Approximation and Boyd’s Exact Solution ...............77 Figure 7.4: Vermeulen Model Fit to Individual Loading Curves ..............................................78 Figure 7.5: Vermeulen Model Fit to All Curves Simultaneously ..............................................79 Figure 7.6:  Shrinking Core Model Fit to Individual Loading Curves.......................................80 xvi  Figure 7.7: Shrinking Core Model Fit to All Curves Simultaneously .......................................81 Figure 7.8  Thomas Model Fit to Individual Loading Curves ...................................................82 Figure 7.9: Thomas Model Fit to All Curves Simultaneously ..................................................83 Figure 7.10: Using Vermeulen Model to Approximate Shrinking Core Model .........................85 Figure 7.11: Hybrid Correlation Fit to All Curves Simultaneously ...........................................87 Figure 8.1: Hybrid Correlation Fit to Nickel Loading onto TP207 ...........................................89 Figure 8.2: Hybrid Correlation Fit to Nickel Loading onto TP207XL, Lot CHC 5034 .............90 Figure 8.3: Hybrid Correlation Fit to Nickel Loading onto TP207XL, Lot CHE 50019............91 Figure 8.4: Hybrid Correlation Fit to Cobalt Loading onto TP207MP ......................................92 Figure 8.5: Hybrid Correlation Fit to Copper Loading onto TP207MP.....................................93 Figure 8.6: Hybrid Correlation Fit to Copper Loading onto TP207XL, Lot CHE 50019 ..........94 Figure 8.7: Photographs of Resin Beads Loaded to 1 eq / L .....................................................97 Figure 8.8: Effect of Solution pH on Copper Loading at 5412, 541, 54 ppm Copper ................99 Figure 8.9: TP207MP/TP207XL Combined Hybrid Correlation Fit to TP207MP Data .......... 100 Figure 8.10: TP207MP/TP207XL Combined Hybrid Correlation Fit to TP207XL Data ........ 101 Figure 9.1: Dual Concentration Test ...................................................................................... 106 Figure 9.2: Loading While Nickel Concentration Increases with Time .................................. 109 Figure 9.3: Modified Helfferich Number for Increasing Nickel Concentration Test ............... 110 xvii  Figure 9.4: Loading From Solution with Initial Concentration of 500 ppm Nickel ................. 111 Figure 9.5: Modified Helfferich Number –No Nickel Stock Solution, 500 ppm Ni Initial ...... 112 Figure 9.6: Loading From Solution with Initial Concentration of 250 ppm Nickel ................. 113 Figure 9.7: Modified Helfferich Number –No Nickel Stock Solution, 250 ppm Ni Initial ...... 114 Figure 10.1: Block Diagram for a Single Stage in a Cascade Circuit ...................................... 118 Figure 10.2: Cascade Circuit Model Inputs & Outputs ........................................................... 119 Figure 10.3: Flow and Mass Balance for Cascade Circuit Model ........................................... 120 Figure 11.1: RIP Miniplant Flowsheet (Cascade Operation) .................................................. 125 Figure 11.2: Photographs of RIP Miniplant (Cascade Operation) ........................................... 126 Figure 11.3: Nickel Cascade Miniplant Results –Model vs Actual ......................................... 129 Figure 11.4: Copper Cascade Miniplant Results –Model vs Actual ........................................ 131 Figure 11.5: Physical Circuit Configuration During Sampling -Carousel ............................... 133 Figure 11.6: Nickel Carousel Miniplant Results -Solution ..................................................... 135 Figure 11.7: Nickel Carousel Miniplant Results –Resin ......................................................... 136 Figure 11.8: Copper Carousel Miniplant Results -Solution .................................................... 137 Figure 11.9: Copper Carousel Miniplant Results –Resin ........................................................ 138 Figure 11.10: Measured Wet-Settled Resin Volumes During Carousel Tests ......................... 139 Figure 12.1: Solution Profile Across Circuit –Nickel Cascade, No Back-Mixing ................... 143 xviii  Figure 12.2: Resin Profile Across Circuit –Nickel Cascade, No Back-Mixing ....................... 143 Figure 12.3: Solution Profile Across Circuit –Nickel Cascade, With Back-Mixing ................ 144 Figure 12.4: Resin Profile Across Circuit –Nickel Cascade, With Back-Mixing .................... 144 Figure 12.5: Final Resin Loading vs. Percent Resin in Tank –Nickel Cascade ....................... 145 Figure 12.6: Total Resin Inventory vs. Percent Resin in Tank –Nickel Cascade ..................... 146 Figure 12.7: Effect of Resin Residence Time on Final Resin Loading ................................... 147 Figure 12.8: Total Resin Inventory vs. Resin Loading –Nickel Cascade ................................ 148 Figure 12.9: Final Resin Loading vs. Percent Resin in Tank –Copper Cascade ...................... 149 Figure 12.10: Total Resin Inventory vs. Final Resin Loading –Copper Cascade .................... 150 Figure 12.11: Effect of Number of Stages on Total Resin Inventories –Nickel Cascade ......... 152 Figure 12.12:  Effect of Number of Stages on Total Resin Inventories ................................... 153 Figure 12.13: Resin Profile Across RIP Circuit –Uneven Resin Distribution Scenarios ......... 154 Figure 12.14: Solution Profile Across Circuit –Uneven Resin Distribution ............................ 155 Figure 12.15: Circuit Recovery for Different Resin Distribution Profiles ............................... 156 Figure 12.16: Effect of Losing Resin Volume in Loading Section of RIP Circuit ................... 157 Figure 12.17: Effect of Increased Feed Concentration on Circuit Recoveries ......................... 158 Figure 12.18: Carousel Solution Profile at 25 Volume Percent Resin ..................................... 161 Figure 12.19: Carousel Resin Profile at 25 Volume Percent Resin ......................................... 161 xix  Figure 12.20: Effect of Contactor Volume on Carousel Cycle Time ...................................... 162 Figure 12.21: Resin Required per Stage vs. Carousel Cycle Time .......................................... 163 Figure 12.22: Effect of Carousel Cycle Time on Resin Loadings ........................................... 164 Figure 12.23: Effect of Number of Contactors on Carousel Cycle Time ................................ 165 Figure 12.24: Effect of Number of Stages on Carousel Resin Loadings ................................. 166 Figure 12.25: Effect of Resin Capacity on Total Resin Residence Time................................. 167 Figure 12.26: Effect of Total Resin Capacity on Total Resin Inventory ................................. 168 Figure 12.27: Effect of Capacity on Fractional Nickel Loading ............................................. 169 Figure 12.28: Effect of Apparent Diffusivity on Total Resin Inventory .................................. 171 Figure 12.29: Effect of Apparent Diffusivity on Final Resin Loadings................................... 172 Figure 12.30: Effect of Resin Bead Diameter on Total Resin Inventory ................................. 173 Figure 12.31: Effect of Resin Bead Diameter on Final Resin Loading ................................... 174 Figure 12.32  Effect of Resin Recirculating Load on Final Resin Loadings ........................... 175 Figure 12.33:  Effect of Incomplete Stripping on Resin Inventory Requirements ................... 176 Figure 12.34: Final Resin Loadings –Nickel Cascade Comparison of TP207 vs TP207XL .... 178 Figure 12.35: Total Resin Inventory –Nickel Cascade  Comparison of TP207 vs TP207XL .. 179   xx  LIST OF ABBREVIATIONS AND ACRONYMS AA  Atomic Absorption AAC  Anglo American Corporation ACS   American Chemical Society BV   Bed Volumes CCD  Counter-Current Decantation CIL   Carbon In Leach CIP   Carbon In Pulp DI   Deionized EQ   Equivalents ICP   Inductively Coupled Plasma ISV   Infinite Solution Volume IX   Ion Exchange PPM  Parts Per Million RIL   Resin In Leach RIP   Resin In Pulp RPM   Revolutions Per Minute SX   Solvent Extraction USBM  United States Bureau of Mines USSR  Union of Soviet Socialist Republics  xxi  LIST OF SYMBOLS a1 empirical fit parameter for a linear isotherm [-] a2 empirical fit parameter for a Freundlich isotherm [-] [A] concentration of ion A in bulk solution [mol A / L] [As] concentration of ion A in solution at interface with resin surface [mol A / L] [Aഥ] concentration of ion A loaded on ion exchange resin [mol A / L resin] [B] concentration of ion B in solution [mol B / L] [Bs] concentration of ion B in solution at interface with resin surface [mol B / L] [Bഥ] concentration of ion B loaded on ion exchange resin [mol B / L resin] dc diameter of unreacted core [m] dp resin bead diameter [m] Dapp  apparent diffusion coefficient through resin bead in hybrid correlation [m2/s] DA isotopic intraparticle diffusion coefficient of A through resin gel [m2/s] DB isotopic intraparticle diffusion coefficient of B through resin gel [m2/s]   Ds diffusion coefficient through reacted layer in shrinking core model [m2/s] Df diffusion coefficient through film [m2/s] Dp diffusion coefficient through resin bead [m2/s] E residence time distribution function [-] f empirically fit Freundlich isotherm exponent [-] F  fractional attainment of equilibrium [-] ۃFۄ mean fractional attainment of equilibrium for resin in a tank [-] xxii  j summation counter [-] kc constant, shrinking core diffusion control [1/s] kf constant, film diffusion control [1/s] kfwd kinetic rate constant, forward reaction in pseudo-second order model [1/mol s] kg constant, Gluckauf’s linear intraparticle diffusion control [1/s] kh constant, diffusion control using hybrid correlation [1/s] kl constant, linear approach to equilibrium [1/s] krev kinetic rate constant, forward reaction in pseudo-second order model [1/mol s] kv constant, Vermeulen’s approximation to intraparticle diffusion control [1/s] K selectivity coefficient [-] L empirical fit parameter to Langmuir isotherm [-] [M] concentration of divalent ion M in solution [mol M / L] [M]eqb concentration of divalent ion M in solution in equilibrium with the resin [mol M / L] [M]feed concentration of M in the feed solution to an RIP contactor [mol M / L Resin] [Mഥ ]  concentration of divalent ion M, loaded onto ion exchanger [mol M / L Resin] [Mഥ ]eqb equilibrium concentration of M loaded onto ion exchanger [mol M / L Resin] [Mഥ ]max capacity of ion exchanger for divalent ion M [mol M / L Resin] Q flowrate through an RIP circuit [L/s] SBA separation factor for ion exchange resin for ion A over ion B [-] t time [s] tr time required under current conditions for resin to achieve the initial loading value [s] t1 2Τ  time to reach 50% loading of an ion exchange bead [s] xxiii  't finite differences model time step [s] VR volume of resin in an RIP contactor [L] VS volume of solution in an RIP contactor [L] zA charge of ion A [electron charge units] zB charge of ion B [electron charge units]  Greek Letters: Ƚ exponent for hybrid correlation [-] Ɂ film thickness [m] Ș Gluckauf’s shape factor [-] Wr average resin residence time in tank [s]    xxiv  ACKNOWLEDGEMENTS There are many people whose contributions to this research need to be acknowledged.  First of all, I would like to thank my supervisor, Dr. Dreisinger, for his guidance, support and encouragement.  I would also like to thank Dr. Wassink for his guidance in the lab. The assistance of Parisa Abbasi, who spent two summers working with me in the lab on this project is gratefully acknowledged.  Parisa helped with converting the resin samples from sodium form to hydrogen form, screening the various resin samples, assaying many nickel solution samples, running batch experiments for many of the resins in the resin comparison, and performing the batch experiments to determine the nickel kinetic parameters for the TP207XL sample used in the miniplant.   Additionally,  Parisa was the day-shift  operator for some of the RIP miniplant  tests.   Additionally,  the  assistance  of  Fernando Parada as day-shift operator of the RIP miniplant when Parisa was unavailable is acknowledged. The  staff  of  the  Materials  Engineering  Shop did  an  excellent  job  with  the  construction  of  the RIP miniplant.  Thanks to Ross McLeod for  helping  with  the  design  and  for  overseeing  the project;  thanks  to  Carl  Ng for the many hours spent constructing the miniplant; thanks to Wonsang Kim for wiring all the pH controllers; and thanks to Dave Torok for manufacturing the impeller blades. Finally, I would like to thank the resin manufacturers: Lanxess, Purolite, and Dow for supplying the resin samples for this research.  Additionally, the assistance of Dr. Fries and Dr. Halle of Lanxess for providing the detailed size analysis of their resin beads in the hydrogen form is gratefully acknowledged.   1  1.0  INTRODUCTION Ion exchange (IX) is a solution processing technique where small spherical resin beads are used to extract ions from solution, while replacing them with another ion in solution.  In the mining industry, ion exchange is used for recovery of pay metals, removal of impurities from process streams, and for treatment of waste streams prior to disposal.  Typically, ion exchange systems consist of a series of packed beds and filtration of the feed solution is required.  In order to be able to clean-up residual dissolved metals in mine waste streams containing large amounts of solids, or to be able to economically recover dissolved metals in cases where the processed ore is difficult to filter, a resin-in-pulp (RIP) process can be used.  A resin-in-pulp process contacts the resin beads directly with the solution and solids, typically in a series of stirred tanks.  The solids and resin beads then need to be separated, typically by screening the larger resin beads (~ 0.5 – 1.2 mm in diameter) from the slurry. Commercial resin-in-pulp processes have been implemented successfully for gold and uranium, but no successful full-scale base metals resin-in-pulp plants have yet been built.  The purpose of this research is to study the loading rate of base metals onto a chelating resin, under conditions that would be representative of a resin-in-pulp circuit.  This work focuses primarily on the loading rate of nickel from synthetic solutions, although a few tests were also conducted for copper and cobalt removal from synthetic solutions. A mathematical model of the system was developed to provide a means of predicting how quickly the resin will extract metals under various conditions.  Once verified, the loading rate model  was  incorporated  into  a  circuit  model  to  simulate  the  performance  of  different  circuit designs and operating configurations to better understand the potential behaviour of a full-scale base metal RIP circuit. 2  2.0  LITERATURE REVIEW This chapter presents a review of the key literature that is relevant to the current research project.  This discussion assumes a basic familiarity with ion exchangers; a detailed introduction to the subject is available in the literature (1).  The review initially focuses on key concepts from ion exchange that pertain to this thesis, and then focuses specifically on the development of the RIP process.  2.1  KEY CONCEPTS FROM ION EXCHANGE FUNDAMENTALS For base metals RIP, the most common ion exchange reaction is the loading of a divalent cation, M2+, onto an iminodiacetic chelating resin in the hydrogen form (to be described in Section 2.1.3).  This can be described by Equation 2.1, where R represents the resin phase:  2R-H + M2+ Æ R2-M + 2H+         [2.1]  In this work, concentrations of ions in the solution are depicted in square brackets, e.g. [M], while concentrations of ions in the resin phase use square brackets and an overbar, e.g. [Mഥ ]. Most engineering models describe loading rates in terms of fractional approach to equilibrium. Therefore, knowledge of both the total amount of metal that the resin can load under the current operating conditions and the rate at which this loading occurs is required.  2.1.1  EQUILIBRIUM CAPACITY The total capacity of a resin for ion M is expressed in this work as [Mഥ ]max, and uses the units mol M / L resin.  Under most operating conditions, the resin reaches an equilibrium loading value, [Mഥ ]eqb,  which  is  lower  than  the  total  capacity.   The  equilibrium  curve  is  typically described using an adsorption isotherm.  While ion exchange is a separate phenomenon to adsorption, many engineering textbooks treat it as a special case of adsorption since many of the design techniques are similar.  Isotherms are typically plotted with the solution 3  concentration on the horizontal axis and the resin concentration on the vertical axis.  Thus, an isotherm that is concave downwards is a favourable isotherm, while an isotherm that is concave upwards is unfavourable.  This is depicted in Figure 2.1.     FIGURE 2.1: FAVOURABLE VS UNFAVOURABLE EQUILIBRIUM ISOTHERMS  The simplest isotherm is the linear isotherm.  Strictly speaking, if a linear isotherm were used to describe the full range of exchange of two equal valence ions (ion A and ion B) in a binary mixture (i.e. from the point at which there is no ion A on the exchanger to the point at which there is no ion B on the exchanger), a linear isotherm would describe a process with no selectivity  for  one  ion  over  the  other.   An  example  of  an  ion  exchange  process  with  no selectivity is isotopic exchange (e.g. exchange of 58Ni2+ for 63Ni2+).  However, in systems where the resin is selective for one ion over another, the first portion of the loading curve can often be approximated by a linear isotherm.  When using such an equation, it is important to be aware that it can be prone to large errors if extrapolated beyond the region for which it was developed. The equation for a linear isotherm is described in Equation 2.2.  Linear:  [Mഥ ]eqb=a1[M]eqb       [2.2]  The Freundlich isotherm is an empirical equation that has been observed to fit adsorption and ion exchange data fairly well.  The Freundlich isotherm is able to model both favourable and unfavourable isotherms; however it has been criticized for poor predictions at both extremes of the isotherm.  It overpredicts extraction at high solution concentrations, since it does not place a limit  on  adsorption  capacity  (2),  and  it  also  tends  to  overpredict  extraction  at  low  solution Resin Loading Solution Concentration Favourable Unfavourable 4  concentrations yielding insufficient adsorbent/resin requirements for the final stages of a plant (3).  It is described by Equation 2.3, where a2 and f are both empirically fitted constants: Freundlich: [Mഥ ]eqb=a2[M]eqbf        [2.3]  The Langmuir adsorption isotherm was first presented in a lengthy article on crystal structures and their impact on heterogeneous reactions, where adsorption rate was postulated as the rate of condensation  of  gas  on  a  crystal  surface  minus  the  rate  of  evaporation  of  gas  from  the  same surface (4).  Inherent in the derivation of the Langmuir isotherm is that only a single layer of molecules adsorb onto an active site.  The Langmuir isotherm also implies the exchange of two ions of equal valency, and becomes essentially an empirical relationship when applied to a more complex system.  The Langmuir isotherm can be described by Equation 2.4, where L is a constant fitted to the dataset: Langmuir:    [Mഥ ]ୣ୯ୠ = [୑]౛౧ౘ୐[୑ഥ ]ౣ౗౮ଵା୐[୑]౛౧ౘ       [2.4]  When solutions contain relatively few species, ion exchange is often expressed using the Mass Action Law.  This approach involves describing ion exchange as a reversible reaction, and determining the equilibrium constant for it.  One problem with this method lies in the difficulty of obtaining the activity coefficients for the various species loaded on the resin.  As pointed out by Reynolds, “The best solid-phase activity model is still being debated, and may not be the same  for  all  systems.”   In  this  same  paper,  it  is  pointed  out  that  the  activity  coefficient parameters for various models are determined empirically in binary systems and that these parameters  may  not  necessarily  be  extendable  to  systems  with  multiple  components  (5).   To avoid this difficulty, the convention is to use concentrations, rather than activities in the Mass Action  Law  equations.   This  results  in  the  development  of  a  constant,  K,  known  as  the selectivity coefficient, which will vary based on the composition of the feed solution.  Thus, K can not be calculated from fundamental parameters, and is determined by fitting to experimental data.  Therefore, there is currently minimal value gained to a design engineer in employing the Mass Action Law in lieu of one of the above mentioned isotherms as it is subject 5  to the same constraint: namely that it is only valid for the feed solution in which the tests were performed. The Mass Action Law for the binary system described in Equation 2.1, is described by Equation 2.5:  Mass Action Law   K = ሾୌ]మ [୑ഥ ] [୑][ୌഥ]మ       [2.5]  2.1.2  RATE OF ION EXCHANGE In theory, there could be five possible rate limiting steps for ion exchange: diffusion of the ion to be loaded through a stagnant film around the bead, diffusion of the ion to be loaded through the ion exchange particle, the ion exchange reaction, the diffusion of the displaced ion through the ion exchange particle, and diffusion of the displaced ion through the stagnant film to the bulk solution.  However,  in order to maintain charge neutrality,  the flux of the charge carried into  the  exchanger  by  the  ions  to  be  loaded  onto  the  resin  is  equal  to  the  flux  of  the  charge carried by the displaced ions out of the exchanger.  Therefore, there are only three possible rate limiting steps: film diffusion, intraparticle diffusion, or the ion exchange process itself. Modeling of ion exchange loading rates is a complex process and it is important to note the difference between an engineering model and a fundamental model.  The goal of a fundamental model is to describe the ion exchange loading as scientifically accurately as possible.  The problem with such models is that they typically result in complex sets of differential equations with no analytical solution, and consequently are quite cumbersome when incorporated into system design equations.  Typically, engineering models are much simpler, are often empirical or semi-empirical, and their goal is to adequately describe system performance.  Both types of models shall be discussed in turn.  2.1.2.1  Fundamental Models for Ion Exchange Loading Rate In terms of fundamental models, there is much debate in the literature as to the best approach to the problem.  For example, even selecting the governing equations for the diffusion process 6  continues  to  be  under  debate.   A  recent  paper  on  competitive  loading  used  Fick’s  laws  and modeled  the  diffusion  process  based  on  diffusion  gradients  alone  (6).   However,  most researchers prefer the Nernst-Planck equations, which also take into account the electric potentials.  More complex than the Nernst-Planck equations are the Stefan-Maxwell equations which  are  based  on  the  kinetic  theory  of  diffusion  in  ideal  gases.   In  fact,  the  Nernst-Planck equation  can  be  taken  as  a  limiting  case  of  the  Stefan-Maxwell  equations  (7),  if  one  ignores frictional interactions between diffusing ions (8).  Note that in a thorough fundamental study, each ion’s intraparticle diffusion coefficient should be determined experimentally, and this is typically achieved by measuring isotopic exchange with radioactive tracers (9).  Using the Nernst-Planck equation to describe the exchange of ions A and B, the intraparticle diffusion coefficient through the gel phase of an exchanger can be described by Equation 2.6, where z represents the valence of the ions (10):  Dp= DഥADഥB൫zA 2 [Aഥ]+zB2 [Bഥ]൯ zA 2 [Aഥ]DഥA+zB2 [Bഥ]DഥB         [2.6] For the case of chelating macroporous resins, such as those being studied in this work, a parallel diffusion mechanism based on the Nernst-Planck equations has been proposed (11).  In this model, both the rate of diffusion in the macropores and in the microgel contribute significantly to the overall rate of loading.  The model is demonstrated for predicting the loading  rate  of  copper  onto  an  iminodiacetic  resin  (9).   It  is  also  used  by  another  set  of researchers in a fundamental study of the sodium/nickel system for a macroporous aminophosphonic resin, where all the individual model parameters were determined experimentally (12).  However, it should be noted that in the aminophosphonic resin study, the parallel diffusion equations were so complex that the authors used an alternative, approximate method that is based on the exchange of equal valence ions (13).  This approximation significantly overpredicted experimental loading values (e.g. 85% final loading at 1000s vs 75% final loading observed), but was much closer than the predictions of the classic shrinking core model (to be discussed in section 2.1.2.2.3), which predicted only 10% final loading at 1000s. One drawback with the fundamental models discussed above is that they all require numerical solutions to differential equations.  While fundamental models often perform better than empirical models when extrapolated to conditions beyond those tested, their use in ion 7  exchange  system  design  tends  to  be  limited  due  to  their  complexity.   Helfferich  and  Hwang acknowledge this limitation by stating, “In the design of ion exchange equipment –e.g. columns- there are usually so many other factors introducing uncertainties or calling for approximations that the simplest rate law seems the best choice” (14).  Michael Streat also acknowledges this limitation by stating, “exact solutions of the diffusion equations exist in the literature  but  they  are  complex  and  ill-suited  to  process  design  calculations”  (15).   Since  the objective  of  this  work  is  to  understand  RIP  system  performance,  the  use  of  complex fundamental models would be questionable.  Therefore, the decision was made to focus on the engineering models available in the literature to describe system performance.  2.1.2.2  Engineering Models for Ion Exchange Loading Rate Studies to develop an engineering model to predict loading rates onto resin have been published for gold recovery from cyanide solutions (16) and for recovery of uranium (17).  No extensive study to try to capture the loading rates over a wide range of test conditions has yet been published for the loading of nickel onto an iminodiacetic ion exchange resin.  Such a model could prove useful in determining the bounds for which the linear approach to equilibrium assumption is valid and may help extend the modeling of RIP circuits in situations where this assumption does not hold. Four engineering models that are commonly applied to ion exchange systems are: the linear approach to equilibrium, the quadratic approach to equilibrium, the shrinking core model, and the pseudo-second order reaction.  These models will each be reviewed in turn.  2.1.2.2.1  Linear Approach to Equilibrium The linear approach to equilibrium is the simplest model, and it is frequently selected over more accurate models due to ease of application.  Some authors have presented their results for nickel loading onto an iminodiacetic acid resin assuming a linear approach to equilibrium based on a co-current loading test (18,19), or by assuming that the fraction of metal recovered to the resin in each tank is constant (20), which implies a linear approach to equilibrium. 8  Film diffusion is typically described using the linear approach to equilibrium, and Gluckauf also proposed approximating intraparticle diffusion by a linear approach to equilibrium (21). With this approach, the situation where both film diffusion and intraparticle diffusion are significant can also easily be accounted for since both film diffusion and intraparticle diffusion are described by the same equation set. The linear approach to equilibrium is described by Equation 2.7, with the solution in terms of fractional approach to equilibrium, F, given in Equation 2.8.  The general model constant, kl, is described by kf, in the case of film diffusion, and by kg, in Gluckauf’s model, both described in Equation 2.9.  Note that the shape factor, Ș, in Gluckauf’s model is < 0.5 and that kf is a function of solution concentration, while kg is not.  d dt [Mഥ ]=kl([Mഥ ]eqb-[Mഥ ])          [2.7]  F= [Mഥ ][Mഥ ]eqb =1- exp(-klt)          [2.8]  kf = 6Df[M]dpį[Mഥ ]max ;     kg= DpȘdp2          [2.9]  2.1.2.2.2  Vermeulen’s Quadratic Approach to Equilibrium Particularly in the case of gel-type exchangers, it is possible that the rate of diffusion within the actual resin bead is the rate-limiting step.  This intraparticle diffusion limitation was modeled by Boyd, Adamson, and Myers and the exact solution to the differential equations for the case of isotopic exchange under infinite solution volume and constant diffusivities is the infinite series of Equation 2.10 (22):   F=1- 6 ʌ2 σ 1 j2 exp ൜ -4Dpʌ2j 2t dp 2 ൠ f j=1       [2.10] In order to simplify the application of this result, Vermeulen developed an approximate fit to Equation 2.10, commonly referred to as Vermeulen’s quadratic approach to equilibrium. Assuming no initial loading, it is written in differential form as per Equation 2.11, with the 9  solution as given in Equations 2.12-2.14 (23).  Note that kv is independent of solution concentration.  d dt [Mഥ ]= 2kv൫[Mഥ ]eqb2 -[Mഥ ]2൯[Mഥ ]        [2.11]   t= -1 4kv ln ൬ [Mഥ ]eqb2 -[Mഥ ]2[Mഥ ]eqb2 ൰       [2.12]   F= ට1-exp൛-4kvtൟ        [2.13]   kv= ʌ2Dp dp 2          [2.14]  2.1.2.2.3  Shrinking Core Model The concept of an outer layer of loaded resin surrounding an unreacted core was first proposed for ion exchange systems accompanied by reactions by Helfferich, based on theoretical considerations.  The equation for this model, and its’ solution, assuming infinite solution volume and constant diffusivities are described in Equations 2.15-2.18 (24).  Note that kc is a function of solution concentration.  dF dt = kcdc 2൫dp-dc൯          [2.15]   t = 1+2൫1-F൯-3(1-F)2 3ൗ kc         [2.16]   F=1- ቄ1 2 + sin ቀarcsinൣ1-2kct൧ 3 ቁቅ 3       [2.17]   kc= 24Ds[M] dp 2[Mഥ ]max         [2.18] Experimental observations confirming the presence of a sharp boundary between the reacted and unreacted zones have been reported for the case of acid elution of a cupric amine complex from a strong acid resin (25), acid elution of a weak acid resin (26), and for the simultaneous 10  advancement of a copper chelate front and a cobalt chelate front into an iminodiacetic resin in the sodium form (6).  2.1.2.2.4  Second Order Rate Model Early attempts to model ion exchange loading often involved representing the system as a chemical reaction.  A list of several of these early models is available in the literature (10). Perhaps the most famous of these was the Thomas model, in which the system is modeled as a second-order reversible reaction (27).  In an early paper on the subject, it is pointed out that this model is not meant to be reflective of the mechanism, but is simply a design tool suggesting that perhaps the diffusional equations may be a similar type of differential equation as these kinetic equations (28). While the second order reaction rate model has minimal theoretical basis, it is still employed on occasion today.  In fact, The Handbook of Separation Processes states, “Rate laws that describe bimolecular second-order chemical reactions can be treated as an analogue to the ion exchange process” (29).  Perry’s Chemical Engineers Handbook echoes this by stating, “When diffusion steps control the process, it is still possible to describe the system by its apparent second order kinetic behavior, since it usually provides a good approximation…” (30). For the case of a divalent metal loading onto an iminodiacetic resin, as described in Equation 2.1, the corresponding second order differential equation is described by Equation 2.19, which can be further simplified to Equation 2.20 in a binary system.  As can be seen from the form of Equation 2.19, the Thomas model is analogous to a second order reaction and not an elementary reaction.  Also note that the forward (kfwd) and reverse (krev) rate constants are determined empirically, and are not necessarily connected via the selectivity coefficient, K (10).  2R-H + M2+ Æ R2-M + 2H+         [2.1]    d dt [Mഥ ]=kfwd[M][Hഥ]-krev[H][Mഥ ]      [2.19]    d dt [Mഥ ]={kfwd[M]+2krev[H]}[Mഥ ]-kfwd[M][Mഥ ]max   [2.20]  11  2.1.3  THE IMINODIACETIC CHELATING RESIN As previously mentioned, the iminodiacetic chelating resin is the most commonly recommended resin for potential base metals RIP applications.  The iminodiacetic group loses protons as the pH increases, as illustrated in Table 2.1 (1).  The loss of a proton from the iminodiacetic acid resin at pH 4 has been confirmed by titration (31). TABLE 2.1: IMINODIACETIC FUNCTIONAL GROUP AT DIFFERENT PH VALUES –FROM (1) pH 1.2 pH 4.0 pH 7.4 pH 12.3                 CH2COOH               /             R-N+H               \                CH2COOH                  CH2COO-               /             R-N+H               \                CH2COOH                  CH2COO-               /             R-N+H               \                CH2COO-                  CH2COO-               /              R-N+               \                CH2COO-   The selectivity of chelating resins can be difficult to define, as the apparent selectivity can vary with solution pH, temperature, solution concentration, and complexing agents.  One option for quantifying the stability of the metal-iminodiacetate complexes is the pH at which desorption of metal from the resin begins.  This “decomplexing pH” can provide insight into the selectivity of the resin for a given ion over the selectivity of the resin for hydrogen ions.  Therefore, an ion with a low decomplexing pH will remain bound to the resin under more acidic conditions than an ion with a high decomplexing pH.   One of the resin manufacturers has published decomplexing pH data for divalent ions (32).  This information is displayed graphically in Figure 2.2.  12   FIGURE 2.2: IMINODIACETIC RESIN DECOMPLEXING PH FOR DIVALENT IONS –BASED ON (32)  2.2  DEVELOPMENT OF THE COMMERCIAL RIP PROCESS: THE URANIUM INDUSTRY The RIP process was first developed in the early 1950s to allow for the processing of uranium without having to first clarify leach solutions.  Several processes were piloted and several full- scale  RIP  plants  were  constructed.   One  of  the  major  objectives  of  the  RIP  process  was  to generate significant capital cost savings in the construction of a uranium plant through the elimination of the filtration step.  In the Union of Soviet Socialist Republics (USSR), RIP had become the industry standard by 1983 and was credited with “reduced energy costs, enhanced productivity, and saved millions of square meters of filtering fabrics and hundreds of tons of acids, alkali, and other chemicals, while the production capacity increased severalfold” (33).  In the Western World, solvent extraction replaced RIP as the process of choice for uranium processing.  Critics claim that one of the reasons for the demise of the RIP process in the Western World was that the high cost of specialized contactors and high operating costs for the 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 D ec om pl ex in g pH 13  first generation of RIP plants generally negated the cost savings from the elimination of the thickening / filtration step (34).  2.2.1  DEFINITION OF TERMS RELEVANT TO RIP Like many terms, the definition of RIP has evolved and changed since it was first coined.  In the early days, RIP simply referred to any process that would allow for ion exchange resin to be used  on  a  stream  that  had  not  been  completely  clarified.   Today,  it  is  more  common  to categorize ion exchange systems into four categories: fixed bed, moving packed bed, fluidized bed, and resin-in-pulp.  This can lead to confusion when reading historical documentation on RIP processes.  For example, the ion exchange installation at St. Pierre du Cantal in France was documented as a successful implementation of “RIP” technology (35); however, this installation used fluidized bed columns based on the United States Bureau of Mines (USBM) design, which had to be retrofitted to operate using air pulsation in 1981 due to operational problems caused by the build-up of solids on the interstage platforms (36). For the purposes of this review, processes that involve moving packed beds (resin that moves continuously through the system in a plug-flow manner) and processes involving fluidized beds (resin that has been fluidized due to the upflow of the feed stream) will not be considered as RIP technologies.  For a review of the development, commercial implementation, and current status of moving packed bed and fluidized bed technology, the reader is referred to publications available in the literature (37,38,39). Generally speaking, an RIP system would be expected to be able to handle feed slurry with a higher solids content than a fluidized bed process.  Note that care must be taken when comparing solids loadings from technologies of different eras, since advances in areas such as resin size and interstage screening can also contribute significantly to the maximum solid handling capabilities of a system. RIP systems usually operate in counter-current fashion and are generally designed based on one of two operating strategies.  The first operating strategy is the continuous or “cascade” operating strategy.  In this type of system, the order in which the feed passes through the tanks remains fixed, and resin is transferred from tank to tank.  Resin exiting the lead tank is directed 14  to a regeneration step, while the regenerated resin is returned to the final lag tank.  Resin transfer between tanks can be periodic (e.g. once every 4 hours) or continuous.  Depending on the method chosen for resin transfer, the degree of back-mixing of pulp can vary significantly. A three stage cascade circuit is depicted in Figure 2.3.   FIGURE 2.3: CASCADE OPERATING STRATEGY  The second operating strategy is the semi-continuous or “carousel” operating strategy.  This type of system has a continuous flow of slurry and batch processing of resin.  In these systems, the order in which the slurry passes through the tanks rotates, in the same way as the order in which the feed to a multi-column fixed bed ion exchange rotates.  For example, with 3 tanks in the loading circuit, if the tanks start out in the sequence 1-2-3, under normal operation the sequence of tanks would rotate through the 2-3-1, and the 3-1-2, sequences before returning to the 1-2-3 sequence.  Each time the sequence changes, the resin from the lead tank would be transferred to elution and refilled with eluted resin prior to coming back into the final lag position.  Alternatively, elution can be done in the same vessel, and all tanks would then cycle through the various positions in the loading and elution circuits.  An example of such a circuit is depicted in Figure 2.4. 3 2 1 R Feed Slurry Tails Slurry Resin Movement 15                 FIGURE 2.4: CAROUSEL OPERATING STRATEGY  2.2.2  UNITED STATES OF AMERICA BASKET RIP PROCESS The first RIP process developed in the United States of America was a semi-continuous loading operation referred to as the “Basket RIP” process.  In this process, the pulp was passed through a series of tanks in which square baskets containing specially manufactured large resin beads (0.81-1.65 mm diameter) were reciprocated up and down in a “tea-bag-like movement”. Elution was done batch-wise in the same tanks used for loading (40).     R 2 1 3 3 1 R 2 1 3 2 R 2 R 3 1 16  A pilot plant for this process was built at Grand Junction, Colorado, in 1953 and operated on fine pulp (-325 mesh) containing approximately 8% solids.  Results indicated that optimum performance would be achieved with a pulp to resin ratio of 6:1, using 6-9 banks for loading, and 6 banks for elution.  Resin loadings of 3 lbs U3O8 per cubic foot of resin (~48 g U3O8 / L resin) and a concentration factor of 225 g/L U3O8 on resin / g/L U3O8 in solution were reported (40). Six mills were constructed in the United States using this technology, and were located at: Monticello, Utah; Moab, Utah; Grants, New Mexico; Tuba City, Arizona; Jeffery City, Wyoming; and Edgemont, South Dakota (41). The full-scale plant at the Uranium Reduction Company’s mill in Moab, Utah, consisted of two RIP circuits that operated in parallel.  Each circuit contained 14 stages: 9-10 stages were used for loading, and only 4 stages for elution.  Each stage contained four baskets.  The resin was reported to have loaded 3.5 lbs U3O8 per cubic foot of resin (~56 g U3O8 / L resin) , and an eluate grade of 9 g/L of U3O8 present as UO2(SO4)3-4 was produced (42).  By 1967, this mill had been taken over by Atlas Minerals and the RIP circuit was being used on alkaline leached ores in lieu of acid leached ores.  Throughput through the alkaline leach circuit was ~850-950 tons per day, and resin losses were reported at 0.0014 cubic feet per ton ore processed.  Given the resin inventory was reported at 152 cubic feet per bank (43), this calculates to approximately 10% resin inventory replacement per annum.  The two RIP lines were replaced by a solvent extraction circuit in 1976 (44). Anaconda’s Bluewater Mill near Grants, New Mexico, started up their basket RIP circuit in 1955 (45), and it appears to have been operating on acid leached slurries with minimal design changes as late as 1974 (46).  The mill had a capacity of 3000-3300 tons per day, although in 1967 was operating at a “reduced capacity” (actual rate not specified).  Bluewater operated two RIP circuits in parallel, each circuit containing 14 sections and a resin inventory of 140 cubic feet per section.  Resin losses were reported at 100 cubic feet per month (43), which calculates to 30% resin replacement per annum.  While this value appears higher than what was reported at the plant in Moab, this plant was likely operating at a higher throughput value.  If the plant had been operating at capacity, the resin losses would calculate to approximately 0.0010 cubic feet per ton of ore processed, which is less than the value reported at Moab; so it is likely that 17  the resin losses were comparable.  A photograph of the RIP section of this mill is available in the literature (46). A number of challenges have been reported with the Basket process.  The Mill at Tuba City (Rare  Metals  Corporation  /  El  Paso  Natural  Gas)  processed  ore  with  a  high  bentonite  clay content and reported problems with resin beads cementing together.  Root cause of this problem was not determined; however, it appears to have cleared up around the time pH adjustment was switched from ammonia to lime (43). The Mill at Moab reported severe corrosion of the baskets when the process was converted from an acid leach process to an alkaline leach process; however, the problem appears to have been addressed by utilizing polyethylene screens in lieu of the stainless steel screens.  The polyethylene screens were also found to have the following additional benefits over the original stainless steel screens: lower cost, easier to repair small tears, and improved drainage of pulp from the baskets (47) Himsley criticized the Basket design by stating, “high maintenance costs of the mechanical equipment along with excessive resin losses in the region of 30% (due mainly to mechanical failures) and the high initial costs resulted in that method becoming obsolete” (48).  2.2.3  SOUTH AFRICA: FLOATING RESIN PROCESSES In order to avoid using screens to separate the resin from the pulp, research was performed in South Africa to develop a process to take advantage of the difference in density between resin and the pulp.  It had been observed that when agitation was stopped, resin tended to float on top of  the  pulp  surface,  and  several  attempts  were  made  to  develop  a  process  to  move  the  resin between stages by taking advantage of this phenomenon. Results of the first two attempts were published in 1959 and 1960.  The first attempt involved a circuit consisting of air agitated Pachuca tanks, each coupled with a settler, in which the separation of resin from pulp would occur (49).  The second attempt involved the design and construction of a pilot plant that utilized one long rectangular tank, subdivided into fourteen small agitated stages, much like a bank of flotation cells.  In these tests, resin was transferred counter-current to the movement of pulp by a set of mechanical scoops attached to an 18  oscillating line shaft.  Within each stage, the pulp density was maintained at an SG of 1.4-1.5 so that the resin would float on top of the cell.  Difficulty in maintaining resin inventories in the various stages was reported (50). The third attempt was known as the Relix process.  This was developed and piloted in the early 1970s.  The principle of the Relix process was to create an “inverted fluidized bed” by having the resin float on a pulp containing 61% solids and using a Pachuca tank to circulate the pulp to a level above the resin.  The inverted fluidized bed would then be created as the pulp percolated down through the resin layer.  The resin-slurry separation between stages could not be achieved at the West Driefontein Pilot Plant, which ran from 1971-1973.  During piloting, airlifts were used to move resin and significant back-mixing of pulp occurred.  After the installation of interstage screening, better metallurgical performance was obtained; however, computer simulation of results after the piloting was complete demonstrated significant breakthrough of resin due to screen apertures being too large on the pulp discharge of the tanks and significant recirculating loops of resin within the tanks (51). No full scale plants based on a floating resin technique were built.  A major design flaw with transferring floating resin is that there is a significant difference in density in resin beads when they are loaded with uranium compared to when they have been regenerated.  In the pulsed bed at Saint-Pierre du Cantal in France, eluted resin was reported as having an SG of 1.12 and resin loaded with 60 g/L uranium as having an SG of 1.34 (36).  Consequently, the resin that is most buoyant, and hence most likely to float and be moved to the previous stage, is small beads with minimal loadings, while the resin most likely to remain in the current tank is large beads with significant resin loadings.  2.2.4  UNITED STATES OF AMERICA: CONTINUOUS RIP The continuous counter-current RIP process was developed and implemented by the same organization as the Basket RIP process.  A pilot plant was run from August 12-23, 1957.  The RIP  adsorption  train  consisted  of  3  conical  bottomed  tanks,  each  with  a  residence  time  of  3 hours.  Both the pulp and the resin, in a ratio of 50:1, flowed by gravity from each contactor to a 50 mesh Sweco screen.  The resin (screen overflow) then advanced to the previous tank by 19  gravity, while the pulp (screen underflow) was pumped to the next tank.  Resin losses were of the order of 300 mL per ton of ore processed (52). Test results indicated that uranium could be adsorbed in significantly fewer contactors than were required for the basket RIP process (3-4 stages vs. 8-10 stages for basket).  Elution was developed utilizing continuous countercurrent elution in columns or tanks.  Due to the success of these tests, the first continuous countercurrent RIP circuit was built at Maybell Mill in Colorado (41).  The Maybell Mill operated from 1957-1964 and utilized 6 stages for adsorption and 10 stages for elution and was able to treat slurry containing 20-30% solids (43).  This plant used Infilco CST Exchangers to try and perform a separation of resin and slurry based on the difference in specific gravity.  In direct contrast to the South African floating resin techniques, the Infilco CST Exchanger was designed based on the pulp floating and the resin sinking to the bottom of the settling chamber.  In these exchangers, a slurry of incoming liquid and resin was fed  down  through  a  feed  pipe  to  a  distributor  near  the  base  of  the  reactor.   Air  was  sparged through porous fabric covering a perforated plate on the bottom of the reactor.  The slurry and resin mixture would rise up through a contacting compartment, and then overflow over internal baffling to a settling compartment of the reactor.  Piloting trials using this system on ore from Charles A. Steen’s Mi Vida Mine near Moab, Utah, showed promising results and a commercial installation was built and scheduled to come on-line in October 1957 (53). Unfortunately, settling of the resin in the full scale units could not provide a satisfactory separation of resin from the slurry in the adsorption circuit and vibrating screens needed to be added to perform resin and slurry separation (43). Three commercial installations were operating in 1967 using continuous counter-current RIP in the Gas Hills area of Wyoming.  The Federal-American Partners Mill processed acid leached material through seven mechanically agitated loading tanks, and the regenerated resin passed through a 13 stage elution circuit (final stage of elution circuit a water wash).  The Union Carbide Mill (Umetco) used air agitation in their tanks, allowing for the operation of up to 30% solids in the feed (the mechanical agitation used at the other two plants limited operation to a maximum of 11-12% solids).  Umetco used six stages for loading and 10 stages in the elution circuit: the first seven elution tanks were for stripping the resin and the final three elution tanks were used to convert the resin from nitric acid form to sulphuric acid form.  Western Nuclear’s Split Rock Mill (Jeffery City) was upgraded from a Basket RIP circuit to continuous counter- 20  current RIP in October 1966.  Their RIP circuit consisted of eight mechanically agitated tanks for loading and fourteen for elution (43). These  plants  all  used  interstage  transfer  of  resin  and  pulp  by  airlifting  to  screens.   After separation on the screen deck, the pulp would proceed to the downstream tank, while the resin would be directed to the upstream tank.  This resulted in resin content of less than 5 vol% in the adsorption tanks.  At the Union Carbide plant, it was desired to increase the resin content in the tanks to 5-10 vol%, so a portion of the resin was recycled to the tank it came from (54).  The Federal American Partners Mill reported that screen decks required replacement every 50-60 days and resin inventory losses of 0.076% per day (43), which would equate to approximately 25-28% per year.  It should be noted when comparing this number to the losses incurred in Basket RIP that the overall resin inventory in a continuous plant is typically less than in a Basket plant. The continuous countercurrent RIP system has been criticized for unpredictable metallurgical performance due to difficulty in maintaining the appropriate resin inventories in the various tanks (48).  2.2.5  DAVY-MCKEE CONTACTOR The Davy-McKee Contactor was designed to try and develop a more compact RIP circuit.   It consisted of a long rectangular cell, subdivided into a series of agitated compartments, much like  a  bank  of  flotation  cells.   Between each  compartment  was  a  common wall  containing  an air-swept screen that would retain resin in the compartment, but allow pulp to pass through. The pilot RIP cell was tested on an unspecified leached uranium ore for 48 hours.  The slurry tested  was  40%  solids  and  contained  25  vol%  resin.   The  pilot  plant  was  also  run  for  a  CIP application (55). Two demonstration plants were operated at Rand Mines’ ERPM Boksburg gold plant.  The first demonstration plant was run with both resin and carbon to test screen throughput at a flowrate of 60 m3/h.  The second demonstration plant was to evaluate the rate of carbon adsorption. Based on the success of the demonstration runs, Davy McKee was contracted to fabricate and 21  supply two 3000 tpd CIP contactors to the Ashanti gold plant in Ghana for commissioning in 1987 (56).  No follow up information on these full scale units could be found in the literature.  2.2.6  UNION OF THE SOVIET SOCIALIST REPUBLICS TECHNOLOGIES In the USSR, a “nonfiltrational sorptive-extractive method … of dense and viscous pulp” was developed in 1953, and had become their industry standard by 1983.  In a discussion about various forms of industrial ion exchange equipment, “standard reactors with mechanical mixing mounted as cascades of 3-6 plants” and “units with pneumatic mixing of a volume from 0.3 to 500m3 for processing of dense pulp up to solid-to-liquid ratio 1:1” were listed in addition to various forms of ion exchange columns (33).  Both pneumatic mixing and mechanical mixing systems were built on an industrial scale.  Pneumatic mixing was used for ores containing at least 95% size fraction less than 150 microns, in a slurry with a specific gravity of no more than 1.5.  Resin was continuously transferred countercurrent to the pulp, by use of special airlifts in these systems.  Mechanical mixing systems consisted of contact tanks with impeller mixers, which were described as follows: “Contact tanks at each sorption stage are combined into cascades of 3-6 apparatuses with a total of 3-5 stages” (57).  Details of the resin transfer mechanism for the mechanical mixing systems were not reported. RIP was also used extensively for gold recovery in the USSR and is believed to have been a carry-over of the uranium technology.  The first resin in pulp process for gold in the USSR was built in the early 1970s.  This plant was located in Western Uzbekistan, and processed ore from the Muruntau deposit.  The RIP circuit utilized Pachuca tanks and gold was loaded onto the resin in a countercurrent fashion to the pulp flow.  This technology was considered a success and rapidly spread through the Soviet Union.  For example, the Ararat Mill for the Zod deposit was originally planned to use conventional processing, but the design switched to resin in pulp due to the success of the process at Muruntau (58). The separation of resin and pulp has been reported as “using flat external screens”, which were “manually  cleaned  once  per  day  with  the  help  of  compressed  air”.   Each  tank  was  equipped with 8-12 screens with an area of one square meter and with apertures of 0.5 mm.  One screen would collect resin for interstage transfer, while all other screens would direct the resin back to the  tank  it  came  from.   At  the  time  of  the  review,  50%  of  the  gold  produced  in  the  former 22  Soviet Union was reported as coming from RIP processes.  Mills were typically grinding to a d95 of 74 microns and the resin was added to the last RIP tank at a concentration of 2.5-3 vol%. One of the most significant drawbacks to the traditional Soviet RIP process for gold was the elution method used resulted in large resin inventories and gold lockup (59). It  is  clear,  and  well  known,  that  several  resin-in-pulp  processes  were  used  extensively  in  the USSR for uranium and gold extraction; however, it is difficult to find published primary data on these plants, or to find a published comparison of the performance of the various methods employed.  2.3  MODERN RIP INSTALLATIONS (COMMERCIAL SCALE) There are a few commercial scale RIP and resin-in-leach (RIL) plants that have been built and operated in more recent years.  These will be reviewed by industry.  2.3.1  GOLD RIP In the western world, RIP/RIL for gold has been used for highly preg-robbing ores (i.e. ores containing carbonaceous material that adsorbs the gold cyanide complex).  There have been three commercial scale RIP/RIL plants that have been operated: Golden Jubilee, Penjom Gold Mine, and Barbrook. The Golden Jubilee Mine converted its’ process from carbon-in-leach (CIL) to RIP in February 1988.  It operated as a semi-continuous operation with pulp transfer between tanks occurring by air-lifting  the  resin-pulp  mixture  to  a  screen.   Pulp  would  advance  to  the  next  tank  in  series, while  the  resin  would  return  to  the  tank  it  originated  from.   Once  per  day,  a  portion  of  resin would be collected and manually transferred to the previous tank (60,61,62).  Golden Jubilee was closed due to depleted gold reserves; however, it is important to note that the RIP circuit operated successfully for six years (63). The  Penjom Gold  Mine  converted  its’  process  from a  CIL process  to  a  batch  RIL process  in 1998, and then upgraded to a continuous RIL process in 2000, once an appropriate gold- selective resin was available.  Dramatic increases in gold recoveries were obtained (64) as 23  illustrated in Table 2.2; however, RIL can not take all the credit for these increased recoveries, since there was also a significant redesign of the gravity circuit of the mill (65,66). TABLE 2.2: GOLD RECOVERIES AT THE PENJOM GOLD MINE 1998-2002 Year Ending Process Recovery (67,68,69,70) Process Recovery Including Tailings Retreatment (71) Comments Prior to April 1997 > 90%   March 31, 1998 86% Started processing carbonaceous ore March 31, 1999 62% 63% Hit low of < 50% recovery at some point March 31, 2000 70% (66% Apr-Sep; 74% Oct-Mar) 71% Continuous gold RIL circuit implemented March 31, 2001 78% 84% First Full year with RIL circuit March 31, 2002  87%    The RIL circuit  at  the Barbrook Mill  was commissioned in mid-December of 2002.  The RIL circuit was a semi-continuous circuit, consisting of six Anglo-American Corporation (AAC) Pump cells, provided by Kemix (72).  This site was placed on care and maintenance due to industrial action caused by labour unrest in the fourth quarter of 2006 (73).  2.3.2  URANIUM RIP The RIP process has once again become quite topical in the uranium industry as it is perceived to hold promise for the economic extraction of high throughputs of low grade ores (74,75). Many uranium projects are evaluating RIP as a potential processing option.  For example, pilot plant results for RIP feasibility studies at SGS Lakefield have recently been published (76,77); an RIP circuit has been selected in the development plan by Khan Resources for their Dornod 24  Project in Mongolia (78); Mantra Resources has reported that RIP would be preferable to CCD/SX  for  their  Mkuju  River  Project  in  Southern  Tanzania  (79),  AREVA  has  conducted testwork for applying RIP to Cominak’s mill site in Niger, as well as for the Kiggavik project in Nunavut (80); and a demonstration plant (scale 1:150) has been operated for uranium recovery at Harmony gold mine (81,82).  Perhaps most significant of all is that a full scale RIP circuit for uranium recovery has been built by Paladin Resources for their Kayelekera Project in Northern Malawi (83). The Kayelekera Mine utilizes a sulphuric acid leach process.  Since there is some tetravalent uranium in the ore, it is first oxidized to hexavalent uranium using ferric as an oxidant.  Ferrous is then oxidized back to ferric using hydrogen peroxide (84).  The RIP loading circuit consists of ten Kemix Pumpcells, which will be discussed in detail in Section 2.3.4.1.  The operational design was to have eight pumpcells on-line at any given time, with the other two pumpcells off- line and isolated from the circuit.  One of the offline contactors would be transferring resin to the elution circuit, while the other offline contactor would be in the process of being refilled with stripped resin.  Reduced foot-print, elimination of the SX fire hazard, elimination of the requirement for solution clarity, increased recoveries and reduction in concerns for health, safety, and the environment were listed in the reasons for the choice of RIP over a conventional CCD/SX process (85).  2.3.3  BASE METALS RIP To date, the only full scale RIP circuit that has been constructed for base metals RIP was at Compass  Resources’  Browns  Oxide  Plant  in  Australia.   At  the  Brown’s  Oxide  operation, leached slurry was directed to a counter current decantation (CCD) circuit.  The CCD overflow reported to copper solvent extraction and electrowinning, while the CCD underflow was to report first to an RIP circuit for copper recovery and then to a second RIP circuit for nickel and cobalt recovery (86).  The first copper cathode from this plant was stripped in September, 2008. At this time, the RIP circuit was reported as having been constructed, and commissioning was scheduled for October 2008 (87).  Mining stopped December 15th, 2008 (88), and Compass Resources went into administration in January 2009, citing the fall in metals prices and the 25  delays encountered to full  metals production at  the Browns Oxide Plant as the cause (89).   It has since been reported that no nickel or cobalt was produced at this site (90).  2.3.4  MODERN RIP SYSTEMS MARKETED IN THE LITERATURE There  are  a  number  of  companies  that  have  published  papers  in  recent  years  on  RIP  system design, and three such systems are reviewed here.  2.3.4.1  Anglo American Corporation (AAC) / Kemix Pump Cells The AAC pumpcell was developed as an alternative configuration for carbon-in-pulp circuits. The AAC pumpcell process uses a cascade operating philosophy to allow for the elimination of interstage transfer of carbon, thus reducing the problems of maintaining adsorbent inventory in the various loading stages and eliminating the inefficiencies associated with pulp back-mixing. The AAC pumpcell was developed and tested from 1987-1990, was used commercially for scavenger duties starting in 1991, and has been used for gold recovery from cyanide leach liquors since 1998.  Residence times for pulp in the AAC pumpcell are typically < 10 minutes per stage and the gold loading on the carbon is typically double that of conventional CIP circuits (91). At  the  heart  of  the  pumpcell  process  is  Kemix’s  Mineral  Processing  Separating  Pumping (MPS(P)) Interstage Screen.  In addition to the agitator blades, the central shaft in a pumpcell also provides the rotation for screen cleaning and turns the pumping impeller to discharge the screened slurry from the cell (92).  The interested reader is directed to the referenced paper for a sketch of this mechanism. There have been two commercial RIP/RIL installations of the AAC / Kemix Pump cell to date. The first is at the Barbrook Gold Mine for a resin in leach gold recovery circuit.  The second installation is the uranium RIP circuit at Paladin Resources’ Kayelekera uranium project in Northern Malawi.  26  2.3.4.2  Mintek / Bateman MetRIX A continuous RIP configuration, marketed under the name of MetRIX, was developed by Mintek and Bateman.  Pulp flows continuously through submerged interstage screens to the next contactor, while resin is transferred hydraulically to the previous contactor and the interested reader is directed to the reference paper for a sketch of this process (93).  Since the mechanisms of resin and pulp transfer are completely independent of each other, this allows for easily operating the circuit with significantly different residence times for the resin compared to the pulp.  In order to address the question of the most durable resin for uranium RIP applications, Mintek has published a paper detailing their development of accelerated resin degradation test procedures for uranium ion exchange resins; however, results do not include the names of the resins tested (94). A cost comparison performed by Mintek and Bateman suggested that a MetRIX circuit combined with solvent extraction of the eluate would be advantageous over a conventional CCD circuit followed by solvent extraction for uranium feed grades of < 900 ppm U3O8, where Bateman  Pulsed  Columns  were  used  for  the  cost  evaluation  of  the  solvent  extraction  circuits (95,96).  No mention of an industrial MetRIX plant could be found in the literature.  2.3.4.3  Clean-TeQ Clean iX Clean-TeQ also markets an RIP system for slurries consisting of 10-60% solids: “Clean-iX cRIP”.  This system is a continuous counter-current system in a series of either air-agitated (“Clean-iX  Air”)  or  mechanically  agitated  (“Clean-iX  Mech”)  tanks.   Resin  and  pulp  are “continuously separated and transported counter-currently” by airlifting to flexible woven wire static screens located above the contactors (97). The Clean-TeQ technology is likely based on the RIP technology developed in the Soviet Union.  As per their 2009 annual report: “Clean-TeQ has been granted the exclusive rights, outside the former USSR, to the Base Technology developed by the All Russian Research Institute of Chemical Technology in Moscow”.  Clean-TeQ has granted exclusive rights to their cRIP system to BHP Billiton and Vale for nickel and cobalt extraction, while still retaining the 27  rights for all other metals (98).  No reference to a full-scale Clean-iX plant could be found in the literature.  2.4  RESEARCH STUDIES FOR BASE METALS RIP While there has not yet been a successful implementation of base metals RIP, a significant amount of research has been published by several authors on this subject.  2.4.1  MINTEK STUDIES Mintek began work on developing a base metals RIP process in the late 1990s.  Early work identified that an iminodiacetic chelating resin was preferable over a weak acid selective resin, and that further work with the iminodiacetic resin was warranted (99).  An early study was sponsored by Billiton to investigate the potential of recovering copper from CCD underflows. Preliminary batch tests suggested that the circuit pH would need to be greater than 2.5 to load the resin to near capacity; however, ferric precipitation tests at this low pH suggested that some iron (6 g Fe / L Resin) would co-load with the copper (47 g Cu / L Resin).  A carousel miniplant was run and preliminary plant economics provided for copper feed concentrations of 4, 1, and 0.5 g/L (100). Anglovaal Mining hired Mintek to evaluate RIP for recovery of copper, nickel, and cobalt from a thickener underflow stream.  Three processing options were considered, all of which contained a ferric precipitation step upstream of the RIP circuit to avoid loading ferric onto the unnamed resin.  The various options were evaluated through miniplant campaigns using a carousel arrangement for the RIP step.  Based on results and preliminary costing, the authors concluded that a copper RIP plant would not be economical for this particular stream, but that a combined copper-nickel-cobalt RIP plant could be (101). Mintek also worked with Zincor to evaluate RIP as an option for recovering soluble zinc from their lead-silver leach residue and their iron precipitation residue.  In this study, the TP207 resin was used in an on-site four stage pilot carousel circuit (150L / stage), held at pH 4.5 with the addition of lime slurry.  For the lead-silver residue, soluble zinc was reduced from a 28  concentration of 2000-6800 ppm in the feed to a barren concentration of 12-80 ppm.  For the iron residue, the soluble zinc was reduced from a feed concentration of 800 – 4400 ppm to a barren concentration of 15 – 105 ppm.  Of note is that zinc adsorbed on the iron residue was observed to have loaded onto the resin, which had not been taken into account in the sizing of the pilot circuit (102). A recent study on loading of copper and cobalt from a tailings dump material suggested that optimum total recovery could be achieved if the pH set point was increased across the RIP circuit.  The rationale is to have a lower pH in the first tank where copper precipitation is likely, and then to increase the pH in later tanks to maximize the recovery of cobalt.  The objective is to use sufficient resin to minimize copper displacing cobalt from the resin.  Scale-up from preliminary laboratory tests to a four stage carousel miniplant was done using a McCabe-Thiele diagram, and assuming stage efficiencies of 60%, 70%, 85%, 100%.  These stage efficiencies were estimated from the fractional attainment of equilibrium observed in batch loading tests (103).  2.4.2  CLEAN-TEQ STUDY Clean-TeQ has published preliminary laboratory data and circuit sizing information to evaluate the use of an RIP circuit to replace a CCD circuit as well as to evaluate the benefits of an RIP circuit installed after a CCD circuit.  In both cases, iron was precipitated at pH 4 ahead of the RIP process.  This resulted in nickel units lost to the solid phase due to precipitation, but these nickel units were subsequently recovered during the RIP process due to a “sorption leach” phenomenon.  The approach taken by Clean-TeQ in the lab was to use repeat contact tests.  In these tests a single pulp sample was contacted with samples of fresh resin until all the nickel had been recovered.  Based on the lab results, a ten stage full scale plant was proposed as an alternative to a CCD circuit (104).  2.4.3  TSHWANE UNIVERSITY STUDY A master’s thesis was conducted at the Tshwane University of Technology to evaluate the effect  of  different  operating  strategies  on  the  metallurgical  performance  of  different  resin 29  transfer  strategies  for  a  zinc  RIP  circuit.   Miniplant  tests  were  conducted  under  carousel operation,  discreet  movement  of  resin  built  up  on  transfer  screens  every  30  minutes,  and continuous counter-current operation.  Results indicated that resin loadings under all three resin transfer strategies were similar for all three cases.  Furthermore, it was reported that increasing the average resin residence time through the circuit resulted in higher resin loadings, while changes to pulp residence time had minimal effect on resin loadings (105).  2.4.4  VALE STUDY A study on nickel RIP from a laterite process was published by the Companhia Vale do Rio Doce (now Vale), using the iminodiacetic resin Amberlite 748.  In this work, an RIP process was  simulated  in  two  sets  of  experiments,  both  conducted  at  pH  4.   In  the  first  set  of experiments, a single 500 mL pulp sample was contacted five times with 30g fresh resin in the sodium  form.   In  the  second  set  of  experiments,  a  single  100g  resin  sample  was  contacted  8 times with fresh 500 mL pulp samples.  In both cases, evidence of nickel extracted from the solid phase was observed (106).  2.4.5  UBC STUDY An  RIP  study  on  the  tailings  from  the  Moa  Bay  laterite  operation  was  carried  out  at  the University of British Columbia, using Lewatit TP207 and Lewatit Monoplus TP207 iminodiacetic resins.  In this work, soluble iron was oxidized and precipitated upstream of the RIP circuit.  Much of the copper present in the slurry was found to precipitate at the optimum circuit pH of 4, but a substantial amount was found to redissolve into solution during the RIP trial.  Of note, is that no recovery of nickel and cobalt from the solid phase was reported in this work (107).  2.4.6  MURDOCH STUDY An  extensive  study  on  recovery  of  nickel  from  laterite  tails  was  carried  out  at  Murdoch University.  Laboratory tests, miniplant tests consisting of five 2L contactors, and pilot testing 30  using five 40L contactors were all performed.  Testing was done using the iminodiacetic resin Amberlite IRC 748 and slurry supplied by Anaconda Nickel, and partial recovery of nickel from the solids phase of the tails was observed (18). Laboratory tests were first performed with synthetic solutions (no solids) to determine the equilibrium loading isotherms for four species: nickel, cobalt, manganese, and magnesium. Each species was studied in isolation at pH 3, 4, and 5.  Automatic addition of NaOH was done to neutralize the hydrogen ions released from the resin as the metals loaded.  Results showed that the selectivity of the resin could be described as Ni > Co >> Mn, Mg.  Both the Freundlich isotherm and the Langmuir isotherm were fit  to these curves;  the better fit  typically being the Langmuir isotherm.  Loading rate could not be determined from the rate of addition of NaOH since a significant amount of sodium co-loaded onto the resin during these tests (108). Laboratory tests were also performed to compare the loading rates of several different iminodiacetic resins during a four hour co-current loading test.  Results indicated the following order of resin loading rates: Lewatit TP207MP > Amberlite IRC 748 > Purolite S930 > Lewatit TP207 > Lewatit TP208.  Once again loading rate was modeled using a linear approach to equilibrium (19). Amberlite IRC 748 was selected as the optimum resin, and it was used for testing in the miniplant and pilot scale evaluations (note: TP207MP was not available in the early stages of Zainol’s testwork).   The loading kinetics from a co-current laboratory test  “… fitted well  to a simple rate equation based on a first-order approach to equilibrium as a result of rate- determining mass transport in the pulp phase” (18).  Process simulation was performed using these laboratory results and software developed for the carbon in pulp process.  A graph showing good correlation between simulator predictions and a miniplant run was included in this work.  The loading rate model described in the thesis appears to be based on the CIP model published for a linear approach to a linear isotherm (109,110).  2.5  CHALLENGES FOR BASE METAL RIP CIRCUITS Although RIP has been implemented successfully in the gold and uranium industries, many still consider it an unproven technology for base metals.  Base metal circuits differ from gold RIP 31  circuits since the upgrading ratio (concentration of metal on resin / concentration of metal in feed) is orders of magnitude smaller in a base metals application (101), resulting in shorter resin loading cycles and a requirement for higher resin inventories (93).  These differences suggest that the optimum design configuration / design methodology for a base metals circuit could well differ from the optimum design configuration / methodology for a gold RIP circuit.  A few of the challenges facing base metals RIP circuits will be reviewed.  2.5.1  RESIN LOSSES One  very  important  factor  for  RIP  technology  is  the  question  of  resin  life.   This  question  is particularly pertinent for the case of base metals RIP since resin inventories are much higher than in gold RIP circuits. Resin losses are due to several factors including: osmotic shock as the resin expands and contracts due to changes in solution chemistry, mechanical attrition from contact with pulp, agitators, pumps, and screens, as well as from resin spillage when the resin is moved between equipment,  and  resin  lost  through tears  in  retention  screens.   Predicting  actual  expected  resin losses in a full scale plant from laboratory tests is extremely difficult. Mintek is working on developing a set of standard procedures to compare the relative expected life of different resins in RIP applications.  In their paper, various lab resin durability tests are conducted for various uranium RIP resins, and compared to the results obtained by cycling the resin through their demonstration-scale MetRIX plant.  No single test procedure was shown to be conclusive for predicting the resin losses observed in the demonstration plant (94). Quantifying resin losses in small scale miniplants is also very difficult.  When working with a relatively  small  resin  volume  (a  few  litres)  for  short  periods  of  time  (a  few  days),  any  small spill during the manual transfer of resin would result in large, erroneous resin loss predictions. Thus, extended pilot scale / demonstration scale testing would be required to even attempt predicting resin losses. In terms of actual resin losses observed in operating RIP plants, there is some data published in the literature.  For gold RIP/RIL plants, a resin consumption rate of 10 g / t ore processed has 32  been reported for the strong base resin used at Golden Jubilee and a resin consumption rate of < 5 g / t ore processed has been reported for the Minix resin at the Penjom Gold Mine (111). For the uranium plants operating in the former Soviet Union, a resin consumption rate of 360- 560 mL / t ore processed has been reported (112). For the historical uranium plants that operated in the United States, resin losses can be estimated from numbers published in Merritt’s 1967 survey of uranium mills (113,114).  The Moab Mill basket plant reported 0.0014 ft3/ton ore processed (44 mL / tonne), and the Federal American Partners mechanically agitated continuous RIP plant reported a throughput of 950 ton per day, 0.076% resin consumption per day, and a resin inventory of 825 cubic feet (converts to 21  mL  /  tonne).   While  these  numbers  are  significantly  lower  than  those  reported  for  plants operating in the former Soviet Union, it is important to remember that the percent solids of the slurries being handled has changed significantly.  Typical percent solids in a basket plant were reported as 7-10%, 11-12% for a mechanically agitated continuous RIP circuit (43), while the plants operating in the Soviet Union reported significantly higher percent solids of 40-50% (33,115).  2.5.2  RESIN DISTRIBUTION ACROSS RIP CIRCUIT Another question raised for RIP is how robust the process will be for fluctuations in resin inventories.  As mentioned in section 2.2.4, this was one of the criticisms of the first generation of continuous uranium RIP plants, and is one of the reasons carousel plants are selected over continuous cascade plants. The question of carousel vs cascade operation for base metal RIP circuits was investigated experimentally in the Tshwane University study (105); however, this study did not address the question of how a continuous cascade circuit would respond to operational fluctuations in resin inventory control across the circuit.  The results of an error in resin inventory control in one of the tests reported by Nicol and Zainol suggest that a base metals RIP process should have some process forgiveness (18); however, no model to quantify the effect of varying resin inventory has yet been published. 33  The question of whether transferring slurry to upstream contactors with the resin would have a deleterious effect on circuit performance was investigated experimentally in the Tshwane University master’s thesis (105).  Results indicated that the circuit could operate adequately without removing the slurry from the resin transfer streams, but no attempt to quantify the effect was made.  Furthermore, it is reasonable to expect that the effect of such back-mixing may be dependent on circuit operating conditions or on the base metal being loaded.  The ability to quantify such effects with a model could be beneficial for circuit design.  2.6  OBJECTIVES FOR CURRENT WORK Based on the findings from the literature review, the following three objectives for the current research program were set: x Based on batch laboratory data obtained under infinite solution volume conditions, use an engineering model to adequately describe the loading rate for the range of 50 – 5000 ppm nickel in solution.  x Verify the applicability of the engineering model to non infinite solution volume conditions through batch experiments with variable solution concentrations and through the operation of an RIP miniplant.  x Once  a  useful  RIP  circuit  model  is  available,  use  this  model  to  quantify  the  effect  of various different operating configurations.  In particular, compare cascade operation to carousel operation, quantify the effect of solution back-mixing during resin transfer, and quantify the process sensitivity of a cascade circuit to fluctuations in resin inventory.   34  3.0  METHOD DEVELOPMENT FOR BATCH TESTS For the initial tests, the hydrogen form of Lanxess Lewatit Monoplus TP207 (TP207MP) was used to evaluate the loading of nickel.  The capacity for this resin is rated at 2.0 eq/L.  Prior to use,  the  resin  was  wet-screened  to  remove  the  small  amount  of  fines  (<500  micron)  and oversize  (>600 micron).   Note  that  all  resin  volumes  reported  in  this  thesis  correspond to  the bulk volume of the resin in the hydrogen form.  Bulk volumes were obtained by wet-settling the resin in a graduated cylinder and tapping the cylinder until the settled resin volume remained unchanged. To try and best simulate an RIP environment, batch tests for nickel recovery were carried out in beakers.  Overhead mixers were operated at 300 RPM to ensure good suspension of the resin in the test beaker.  The pH-stat method developed by Babjak in his studies of nickel and cobalt loading onto a bispicolylamine resin (116) was initially considered for this work. In the pH-stat method, a pH electrode and controller are used to maintain a constant pH in the test solution.  By tracking the rate at which base needs to be added to the solution to maintain constant pH, one can estimate the rate at which a resin in the hydrogen form is loading cations. One significant advantage of the pH-stat method over other infinite solution volume methods is that it produces a nearly continuous data-set, as opposed to a series of discreet data points corresponding to samples taken during the test. Preliminary runs with the pH-stat method for loading an iminodiacetic acid resin from synthetic solutions of nickel sulphate, maintained at a pH of four by the addition of sodium hydroxide were performed.  Unfortunately, even though the resin was equilibrated to pH 4 ahead of time, significant mass balance errors were observed between the estimated nickel level based on the sodium hydroxide added and the nickel stripped from the resin beads.  This suggests that the resin must have been co-loading another cation.  The only other cation present in the solution was the sodium from the sodium hydroxide added to maintain constant pH.  The loading of some sodium is not surprising since the iminodiacetic group loses protons as the pH increases, as was discussed in Section 2.1.3. In order to maintain the ability to obtain a near-continuous set of data, work was done to modify the pH-stat method and develop a test procedure to minimize the amount of co-loading. 35  This chapter discusses the modifications made to the procedure, and then provides a detailed description of the experimental method.  3.1  METHOD DEVELOPMENT The first modification made to the pH-stat method was to hold the nickel concentration, in addition to the pH, constant for the duration of the test.  This was achieved by adding a second pump head to the motor being used to drive the pump that was adding the sodium hydroxide. The second pump was used to add a stock solution of nickel, made up by calculating the amount of nickel sulphate needed to maintain constant nickel concentration in the test beaker. A photograph of the equipment used for the loading tests can be found in Figure 3.1.  Note that this photograph depicts two sets of equipment that are being operated in parallel.   FIGURE 3.1:  BATCH LOADING EQUIPMENT  To confirm constant nickel concentration in the beaker, samples were typically taken at the beginning, after 1 hour, after 4 hours, and at the end of the test (approximately 20 hours) and assayed by atomic absorption.  Results for tests run at 50, 500, and 5000 ppm nickel are displayed in Figure 3.2 and illustrate that nearly constant nickel concentrations were maintained for the duration of the tests. 36   FIGURE 3.2:  NICKEL CONCENTRATION IN BEAKER FOR DURATION OF TEST TP207MP Resin, pH 4, 30qC  The second parameter investigated to minimize co-loading onto the resin, was the effect of different alkali hydroxides.  It has been reported that alkali loading onto iminodiacetic acid polyurethane foams at pH 4 showed a preference for Na > Li > K at pH 4, based on batch tests with 2 ppm alkali in the feed (117).  Consequently a series of blank tests were run to determine the effect of changing the alkali hydroxide used.  Additionally, a baseline was run in which the resin was placed into a beaker of deionized (DI) water and the effect of using ammonia as a neutralizing agent was also investigated. For a blank test, 10 mL of resin was placed in a small beaker containing 15 mL of deionized water, and drops of the appropriate base were added until the water and resin were equilibrated at  a  pH  of  4.   The  conditioned  resin  was  then  placed  into  a  beaker  containing  500  mL  of  a 0.0325 M alkali  sulphate  solution,  and  stirred  with  an  overhead  mixer.   A pH controller  was used to add the appropriate amount of a 0.5 M alkali hydroxide solution to maintain a pH of 4.0 10 100 1000 10000 0 5 10 15 20 25 N ic ke l i n Te st  B ea ke r [p pm ] Time in Test [h] 50 ppm 500 ppm, Run #1 500 ppm, Run #2 5000 ppm 37  +/- 0.1 in the test beaker.  These tests were then repeated using 5 mL of resin, 0.0187 M alkali sulphate, and 0.25 M alkali hydroxide solution.  Note that 0.0187 M alkali corresponds to the expected final value after fully loading 5 mL of a 2.2 eq/L resin and adding 88 mL of stock solution  to  the  test  beaker,  and  thus  over-represents  the  worst-case  scenario.   The  0.0325  M alkali sulphate solution does not represent this same worst-case scenario, but was based on observations of the amount of sodium hydroxide added during an early, preliminary run in a nickel sulphate solution.  Equilibrium was observed to be attained within 30 minutes, and early runs showed minimal change between 1 and 8 hours;  therefore,  most tests were shut-off after one hour.  The loaded resins were stripped using 5 bed volumes (BV) of 100 g/L sulphuric acid and select samples were submitted to a commercial laboratory for ICP assays.  Results from the blank tests are summarized in Table 3.1. TABLE 3.1:  RESULTS OF BLANKS WITH DIFFERENT NEUTRALIZING AGENTS  Cation  mL Resin  eq/L loaded Fraction capacity of a 2.0 eq/L resin Estimated ppm in Strip Solution ICP Assay of Strip Solution [ppm] DI H2O 10 0.002 0.1 % *After 2 hours, using NaOH  10 0.004 0.2% *After 8 hours, using NaOH Ammonia 10 0.144 7.2 %  10 0.139 7.0 % Lithium 10 0.104 5.2 %  10 0.118 5.9 %  5 0.058 2.9 % Sodium 10 0.114 5.7 % 105  10 0.100 5.0 % 92 109  5 0.070 3.5 % 32 34 Potassium 10 0.118 5.9% 184  10 0.124 6.2% 194 230  5 0.079 4.0 % 62 65   38  These results show that reducing the concentration of resin in the test beaker (and therefore the final total concentration of alkali in the test solution) has a much greater impact than changing the  alkali  used.   Although  it  appears  that  the  amount  of  alkali  loading  appears  to  follow  the order K > Na > Li, the repeat tests for lithium and sodium at 10 mL resin suggest that the differences between these two elements are within experimental error.  Consequently, it was decided that sodium hydroxide would be used for the experiments and that 5 mL of resin would be used in future tests. Repeatability of the method was examined by running three tests at 250 ppm nickel concentration.  Results are shown in Figure 3.3, and indicate good repeatability with an experimental error of approximately +/- 2 g Ni / L Resin.  FIGURE 3.3:  REPEATABILITY AT 250 PPM NICKEL TP207MP Resin, pH 4, 30qC  0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time in Test [h] Run #1 Run #2 Run #3 39  To determine the amount of sodium that co-loaded when the concentration of nickel was kept constant,  a  sample  of  the  strip  solution  from  one  of  the  tests  run  at  250  ppm  nickel  was submitted for ICP analysis.  Results showed that, under these conditions, only 0.004 eq/L of sodium was stripped from the resin.  Therefore, it is believed that the procedures developed allow for near-continuous measurement of the rate of nickel loading, with minimal interference from the sodium introduced to maintain constant pH. In order to ensure that the agitation speed was sufficient to not limit the loading rate, a test was performed at 250 ppm nickel and the agitation rate increased from 300 RPM to 450 RPM. Results  from  the  increased  RPM  test  are  displayed  in  Figure  3.4,  together  with  the  three baseline  tests.   Results  clearly  indicate  that  at  300  RPM,  agitation  speed  is  not  limiting  the loading rate at 250 ppm nickel.  FIGURE 3.4: EFFECT OF INCREASING AGITATION SPEED AT 250 PPM NICKEL TP207MP Resin, pH 4, 30qC  0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] Baseline 1 Baseline 2 Baseline 3 450 RPM 40  3.2  DETAILED DESCRIPTION OF EXPERIMENTAL PROCEDURE A detailed description of the finalized experimental procedure is described in this section. Check sheets were used to ensure that all tests were carried out using the same procedure.  A copy of a blank test sheet can be found in Appendix A. Batch tests were conducted using synthetic solutions, made up from American Chemical Society (ACS) grade reagents.  The test beaker was initially filled with 500 mL of test solution containing the desired concentration of nickel.  This beaker was then placed in a hot water bath at 30qC, and overhead stirring applied using a propeller-type impeller rotated at 300 RPM.  A pH probe with temperature compensation was calibrated using pH buffers 7 and 4.  The pH probe was then used to equilibrate 5 mL of the resin being tested with 15 mL deionized water at pH 4.  As the pH of the deionized water dropped, sodium hydroxide was added until the resin and water were in equilibrium at a pH of 3.9 - 4.1.  The pH probe was then rinsed and placed into the beaker of nickel test solution, which was adjusted to pH 3.9 - 4.1, through the addition of a few drops of dilute sulphuric acid.  Baffling in the test beaker was provided by the pH probe itself, and by placing an additional length of ½” (~13mm) polyethelene rod into the beaker. A stock solution of 0.25 M sodium hydroxide and a stock solution of > 0.125 M nickel sulphate were made up and their specific gravities were measured by weighing 10 mL of each solution. The actual concentration of the nickel stock solution was slightly higher than 0.125 M and was calculated such that the solution concentration in the test beaker would remain constant for the duration of the test.  The target stock nickel concentration, therefore, varied from test to test, depending on the solution concentration in the test beaker and on the slight differences in pumping rates, depending on which pump heads and motor was being used at the time of the test. The two flasks containing the stock solutions were weighed individually and then placed on a Sartorius TE4101 scale, which was connected to a computer through an RS-232 serial port. Windmill  software  was  used  to  communicate  with  the  balance  and  collect  the  weight  on  the scale every 15 seconds. To  start  the  test,  the  5  mL  of  equilibrated  resin  was  transferred  to  the  test  beaker  using graduated 10 mL pipettes in an “upside-down” fashion (i.e. pipette bulb placed over the tip). 41  The resin in the pipettes was left to settle before being added to the test beaker to minimize the amount of water that was added to the test beaker with the resin.  Time of transfer of the resin was typically less than 30 seconds. A Jenco 3676 controller was used to maintain the pH and nickel concentration in the test beaker.   When  the  pH  dropped  below  3.9,  the  controller  would  turn  on  a  single  motor  drive with two Masterflex L/S 13 pumpheads mounted on it.  One pumphead would deliver the stock solution of sodium hydroxide, while the other pumphead would deliver the stock solution of nickel sulphate.  Once the pH reached 4.1, the motor would turn off.  Due to the on/off nature of the controller, the speed of the motor was manually adjusted during the first hour or two of the test to minimize overshooting and undershooting, and the actual pH in the beaker was maintained at 4.0 +/- 0.2.  To ensure constant nickel concentration over the course of the test, samples were taken after approximately 1, 4, and 20 hours.  The test was ended after approximately 20 hours, and the final weight of the two stock solution beakers was obtained. The resin was then stripped downflow at 10 BV/h in a 25 mL burette.  Resin stripping consisted of 5 BV deionized water wash, 5 BV strip with 100 g/L sulphuric acid, a 20 hour acid hold, followed by a 10 BV strip wash with deionized water.  The strip and strip wash solutions were collected in a 100 mL volumetric flask.  Typically, three different dilutions of the final strip sample were run by AA in order to check instrument reliability, and the average value taken for mass balance calculations.   42  4.0  COMPARISON OF LARGE BEAD IMINODIACETIC RESINS As mentioned in Chapter 2, the study at Murdoch University included a comparison of various commercially available resins, and these results have recently been published (3,19).  However, since this work was done, several resin manufacturers have developed large bead products, specifically for use in base metals RIP applications.  The focus of this chapter is to evaluate these new large bead resins (Ambersep XE818, Lewatit MonoPlus TP207XL, Purolite S930+/4888) and compare their performance to several of the traditional resins evaluated in the study of Zainol and Nicol (Amberlite 748, Lewatit TP207, Lewatit MonoPlus TP207, Purolite S930).  Additionally, the traditional size of the new Purolite S930+ product was included in this comparison.  Resins were compared based on size distribution, loading rate, loading capacity, and stripping rate.  Appendix B contains a summary of resin properties reported by the various manufacturers in their resin data sheets.  4.1  SIZE DISTRIBUTIONS OF RESINS Resin bead size is an important factor for RIP circuits.  It is well known that loading rates decrease as the resin bead sizes increases (as illustrated in Equations 2.9, 2.10, 2.14, 2.18). However, for RIP, the resin beads need to be separated from the solids present in the slurry, and the larger the beads, the easier the separation of the two phases.  Consequently, in RIP applications, large bead products are generally preferred. The resins were converted to the hydrogen form in a column, by passing 5 BV of 100 g/L sulphuric acid, and 10 BV deionized water wash over the resin.  The converted resin was wet- screened to determine the cumulative fraction passing curves.  The bulk volume change for each resin as it was converted from sodium form to hydrogen form was recorded and these values are plotted in Figure 4.1.  Note that three samples of the TP207 resin and two samples of the TP207MP resin were converted from sodium form to hydrogen form to get a sense of the error in these measurements. 43   FIGURE 4.1: RESIN VOLUME CHANGE AS CONVERTED FROM HYDROGEN FORM TO SODIUM FORM  Results from Figure 4.1 indicate that, with the exception of the Purolite S930, there is typically a bulk volume change of 15-30% associated with the conversion from sodium form to hydrogen form.  For the three new large bead products, the resin with the least volume change is the Amberlite XE818, followed by the Purolite S930+/4888, and then the Lewatit TP207XL. The  results  from  wet-screening  the  resin  beads  are  plotted  in  Figure  4.2.   Recall  that  all volumes in this work correspond to the bulk volume of resin in the hydrogen form, the resin beads are spherical, and all resin bead diameters reported correspond to the diameter of the resin beads in the hydrogen form. 0% 5% 10% 15% 20% 25% 30% 35% Vo lu m e Ch an ge  fr om  S od iu m  F or m  to  H yd ro ge n Fo rm 44   FIGURE 4.2: CUMULATIVE FRACTION PASSING –HYDROGEN FORM  Of interest from Figure 4.2 is that the resin with the largest beads is the Ambersep XE818, followed by the Purolite S930+/4888.  The size distribution of the new Purolite S930+ is very similar to the Lewatit TP207.  Also note how the Lewatit MonoPlus TP207 was mostly retained on a single screen size, and that this was also observed, to a slightly lesser degree for the Lewatit MonoPlus TP207XL.  The manufacturer (Lanxess) was contacted regarding how best to  estimate  the  typical  bead  diameter  when  their  resin  is  used  in  the  hydrogen  form.   In response, the manufacturer provided typical size distributions for the TP207, TP207MP, and the TP207XL in the hydrogen form.  The cumulative fraction passing curves from the data provided by Lanxess are displayed in Figure 4.3, along with the wet-sieved data.   0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 500 600 700 800 900 1000 1100 1200 1300 1400 Cu m ul at iv e Fr ac ti on  P as si ng Screen Size [um] TP207 TP207 MP TP207 XL S930 S930+ S930+/4888 IRC748 XE818 45   FIGURE 4.3: COMPARISON OF WET-SCREENED RESULTS WITH DATA PROVIDED BY LANXESS  Reasonably good agreement between the data provided and the measured wet-sieved data was observed.  From the detailed supplier size data, an “effective resin bead diameter” for each screen size was calculated for modeling purposes.  The effective diameter was based on the size and number of particles present within each size fraction.  These values are displayed in Table 4.1.  0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 400 500 600 700 800 900 1000 1100 1200 Cu m ul at iv e Fr ac ti on  P as si ng Screen Opening [um] TP207 (Lanxess Data) TP207MP (Lanxess Data) TP207XL (Lanxess Data) TP207 (Wet Screened) TP207 (Wet Screened) -Repeat TP207MP (Wet Screened) TP207XL (Wet Screened) 46   TABLE 4.1: EFFECTIVE RESIN DIAMETERS FOR MODELING PURPOSES Screen Size (microns) TP207 TP207 MP TP207XL +500  -600 555 552 +600  -710 653 +710   -850 770  736 +850  -1000 901 +1000 -1180 1050   4.2  LOADING RATE AND CAPACITIES Loading rate and capacity of various size fractions of each resin were determined using synthetic solutions of nickel sulphate at pH 4, and the procedure described in Section 3.2. Capacity of the resins can be inferred from the final loading value obtained at the end of the 20 hour loading rate tests. Each resin was evaluated at both 500 ppm nickel and 2500 ppm nickel.  Loading rate curves were generated at three size fractions: 500 – 600 micron, 710 – 850 micron, and 1000 – 1180 micron.  Note that due to the size distributions of the resins, many resins could only be tested at one size fraction, and no resin was tested at all size fractions.  4.2.1  LOADING RATE AT 500 PPM NICKEL Results for the 500 – 600 micron size fraction are displayed in Figure 4.4, which contains an inset with the first four hours of loading. 47   FIGURE 4.4: LOADING OF 500-600 MICRON BEADS AT 500 PPM NICKEL Iminodiacetic Chelating Resins, pH 4, 30qC  The results in Figure 4.4 clearly illustrate the large variation in resin capacities (final loading values).  It also shows that the new Purolite S930+ product has a much larger capacity than the old Purolite S930 and loads at a similar rate as the Lewatit TP207.  Note that the error bars in the graph are based on the evaluation of experimental repeatability, reported in Figure 3.3. Most of the resins in Figure 4.4 were previously tested by Zainol and Nicol in a four hour batch loading test with a sample of CCD tails obtained from industry.  In their work, it was found that the rate of resin loading followed the following order: TP207MP >> IRC748 > S930 > TP207 (19).  The results presented in Figure 4.4 confirm that the rate of loading of the TP207MP is noticeably faster than the other resins tested.  This makes sense in terms of data published by this resin manufacturer, which states that the MonoPlus TP207 has a larger internal surface area and more porosity than the traditional TP207 product (118). 0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa de d  [ g N i /  L  R es in ] Time [h] IRC748 S930 S930+ TP207 TP207MP 0 10 20 30 40 50 60 0 1 2 3 4 First Four Hours: 48  However, these results do not indicate that the rate of loading of TP207 is slower than the other resins.  This is likely reflective of the fact that these tests were done under constant nickel composition, in the absence of impurities that would co-load onto the resin.  Of note from the assays in the work of Zainol and Nicol, was that the manganese content of the TP207 was significantly higher than for the IRC748 and S930 resins.  This suggests that in the presence of high  levels  of  impurities  (as  one  would  expect  to  see  in  an  actual  RIP  slurry),  the  impurities initially load onto the resin and are then displaced by nickel.  It is reasonable to expect that the rates associated with this displacement process could be different from the rates of nickel loading directly onto a resin in the hydrogen form.  This highlights the importance of testing ion exchange resins under conditions that are reflective of their actual anticipated operating conditions. Results for loading onto resins in the 710 – 850 micron size fraction are displayed in Figure 4.5.  FIGURE 4.5: LOADING OF 710-850 MICRON BEADS AT 500 PPM NICKEL Iminodiacetic Chelating Resins, pH 4, 30qC  0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa de d  [ g N i /  L  R es in ] Time [h] S930 S930+ S930+/4888 TP207 TP207XL 0 10 20 30 40 50 60 70 0 1 2 3 4 First Four Hours: 49  In  Figure  4.5,  the  new TP207XL is  shown to  have  a  faster  loading  rate  than  the  other  resins tested.  This is not surprising as this resin is considered a larger version of the TP207MP.  In this test, the new S930+ performs almost identically as the TP207, and the large RIP product S930+/4888 is slightly lower than these two curves; however, the results from all three of these curves are within experimental error. Results for loading onto resins in the 1000 – 1180 micron size fraction are displayed in Figure 4.6.  FIGURE 4.6: LOADING OF 1000-1180 MICRON BEADS AT 500 PPM NICKEL Iminodiacetic Chelating Resins, pH 4, 30qC  In Figure 4.6, the new S930+/4888 resin is shown to load to a higher capacity than the new XE818 resin. After loading, all resins were stripped using 100 g/L sulphuric acid and the strip solution was assayed to provide a mass balance check on the resin capacity (final loading value).  Early tests 0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time [h] XE818 S930+4888 50  showed incomplete stripping of resins, so a 20-hour acid hold step was incorporated into the stripping procedure.  Strip assays were obtained as either the average of three separate dilutions assayed by AA (analysis error of 5%), or are the average of two indirect EDTA titrations using zinc sulphate.  One sample was assayed using both methods and assay results between the two methods were within 2%.  The mass balance results are displayed in Table 4.2. TABLE 4.2: MASS BALANCES AND CAPACITIES AT 500 PPM NICKEL Resin +500 -600 Pm +710 -850 Pm +1000 -1180 Pm Load Strip Load Strip Load Strip IRC 748 41.2 40.3 - - - - XE 818 - - - - 37.6 38.0 S930 19.6 20.8 20.8 21.1 - - S930+ 57.8 55.8 58.2   52.7* - - S930+4888 - - 55.6 55.1 51.9 49.4 TP207 61.7 61.0 58.1 57.6 - - TP207MP 55.5 54.8 - - - - TP207XL - - 55.6   50.0* - - * indicates samples without the 20-hour acid hold step All values are reported as g Ni / L Resin.  AA Assay error +/- 5%   4.2.2 LOADING RATE AT 2500 PPM NICKEL Loading rate tests were repeated, using fresh resin samples, at a feed concentration of 2500 ppm nickel.  Figure 4.7 displays the results for the 500 – 600 micron size fraction. 51   FIGURE 4.7: LOADING OF 500-600 MICRON BEADS AT 2500 PPM NICKEL Iminodiacetic Chelating Resins, pH 4, 30qC  Results from Figure 4.7 are quite interesting in that, under these conditions, the TP207MP does not appear to be loading nickel any faster than the TP207 or the S930+.  This was unexpected, and so the test for the TP207MP was repeated and gave virtually the same loading curve.  This suggests that for this size of resin bead in a feed solution of 2500 ppm nickel, the resin porosity is no longer contributing significantly to the overall resin loading rate.  Further work would be required to determine what factors are truly limiting the nickel loading rate under these conditions. Results for loading onto resins in the 710 – 850 micron size fraction are displayed in Figure 4.8. 0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa de d  [ g N i /  L  R es in ] Time [h] IRC748 S930 S930+ TP207 TP207MP 0 10 20 30 40 50 60 70 0 1 2 3 4 First Four Hours: 52   FIGURE 4.8: LOADING OF 710-850 MICRON BEADS AT 2500 PPM NICKEL Iminodiacetic Chelating Resins, pH 4, 30qC  Results in Figure 4.8 are more in-line with the observations made at 500 ppm nickel.  The TP207XL is once again loading at a significantly higher rate than the other resins tested. However, since loading rate is known to decrease as resin diameter increases, these results in no way invalidate the observations made with the smaller resin beads (Figure 4.7). Results for the 1000 – 1180 micron size fraction are displayed in Figure 4.9.  Note that repeat loading tests for each resin were performed, and both results are displayed in this graph to demonstrate the repeatability of the method used.  0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time [h] S930 S930+ S930+4888 TP207 TP207XL 0 10 20 30 40 50 60 70 0 1 2 3 4  First Four Hours: 53   FIGURE 4.9: LOADING OF 1000-1180 MICRON BEADS AT 2500 PPM NICKEL Iminodiacetic Chelating Resins, pH 4, 30qC  Mass balances from the resins loaded at 2500 ppm nickel are displayed in Table 4.3.  Note that the strip values come from the cumulative stripping curves (to be discussed in Section 4.3).  0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time [h] XE818 S930+4888 XE818 (Repeat) S930+4888 (Repeat) 54   TABLE 4.3: MASS BALANCES AND CAPACITIES AT 2500 PPM NICKEL Resin +500 -600 Pm +710 -850 Pm +1000 -1180 Pm Load Strip Load Strip Load Strip IRC 748 41.6 43.3 - - - - XE 818 - - - - 40.7, 40.9 39.1, 41.2 S930 19.5 22.0 20.8 22.2 - - S930+ 60.9 60.1 61.1 60.3 - - S930+4888 - - 59.6 56.0 58.3, 56.9 56.6, 54.0 TP207 65.6 65.9 60.5 64.0 - - TP207MP 56.1, 55.3 55.1, 58.6 - - - - TP207XL - - 57.7 55.8 - - All values are reported as g Ni / L Resin.  AA Assay error +/- 5%   4.3  STRIPPING COMPARISON The stripping comparison of the resins was conducted on the resins loaded at 2500 ppm nickel. Five bed volumes (BV) of wash water were passed over the resin at 10 BV/hr, followed by 5 BV of 100 g/L sulphuric acid.  Cumulative nickel stripped curves were obtained by collecting strip solutions after 1.0, 2.0, 2.6, 3.2, 3.8, 4.4, and 5.0 BV.  A 20 hour acid hold was initiated and then deionized water was used for a 10 BV strip wash.  In the graphs that follow, the strip wash is plotted as the 6th BV of acid to pass over the resin.   All  assays for this portion of the work are based on a single sample dilution analyzed by AA.  55  Results from the 500 – 600 micron beads are plotted in Figure 4.10:  FIGURE 4.10: CUMULATIVE NICKEL STRIPPED FROM 500 – 600 MICRON BEADS Iminodiacetic Chelating Resins, pH 4, 30qC  The results in Figure 4.10 demonstrate that, under the conditions tested, the rate of nickel stripped follows the order: TP207 > S930+ ~ TP207MP ~ IRC748 >> S930.  This is in direct contrast to the results reported by Zainol and Nicol: IRC748 > TP207 ~ S930 (3).  Both sets of tests were loaded with synthetic solutions containing only nickel cations, and used similar strengths of sulphuric acid to strip the resins.  The only known significant difference between the two sets of tests is that the work of Zainol & Nicol was done upflow at 2 BV / h, while this work was done downflow at 10 BV/hr, due to the small resin volumes used in the loading tests (5 mL). To determine if the difference in flow rates was the source of the discrepancy, 20 mL of the IRC748 and 20 mL of the TP207 were loaded at 2500 ppm nickel.  The resins were then 0 10 20 30 40 50 60 70 0 1 2 3 4 5 6 7 Cu m ul at iv e N ic ke l S tr ip pe d [g  N i /  L  R es in ] Bed Volumes IRC748 S930 S930+ TP207 TP207MP 20 HR SOAK 56  stripped at 2 BV/hr, downflow.  Results are displayed in Figure 4.11, along with the stripping curves for these resins generated at 10 BV/h.  FIGURE 4.11: EFFECT OF STRIPPING FLOWRATE ON IRC78 AND TP207 (500-600 MICRON) Iminodiacetic Chelating Resins, pH 4, 30qC  As one might expect, Figure 4.11 shows that at 2 BV/h both resins stripped with less acid than at 10 BV/h.  The rate of stripping for the two resins appeared to be approximately the same at 2 BV/h.  Of note, is that for the TP207, the amount of resin stripped before the acid hold step is the same at 2 BV/h and 10 BV/h.  This suggests that if this resin is used in an RIP application, it will either be returned to the loading tanks after incomplete stripping or a long acid hold step to fully  strip  the  resin  will  be  required.   The  implication  of  this  on  circuit  performance  will  be explored in Section 12.6.4. Perhaps part of the reason Zainol & Nicol noticed fastest stripping kinetics in the IRC748 was due to the size of the resin beads.  In their work, resin beads were tested in the +600 -1000 Pm 0 10 20 30 40 50 60 70 0 1 2 3 4 5 6 7 Cu m ul at iv e N ic ke l S tr ip pe d [g  N i /  L  R es in ] Bed Volumes IRC 748, 10 BV/ h IRC 748, 2 BV / h TP207, 10 BV / h TP207, 2 BV / h 20 HR SOAK 57  size fraction, based on the larger sodium form of the resins.  Based on the size analysis presented in Table 1 of their work, 94% of the IRC748 resin beads tested would have been in the smaller portion of this size fraction (+600 - 710 Pm).  However, for the S930 only 49% were present in the smaller portion of the size fraction, and for the TP207 only 46% were present in the smaller portion of the size fraction.  Since approximately half of the resin beads tested for the S930 and TP207 were in the larger (+710 - 1000 Pm) portion of this size fraction, it is not surprising that the smaller IRC748 beads exhibited faster stripping. Cumulative stripping curves for resins in the 710 – 850 micron size range are displayed in Figure 4.12.  FIGURE 4.12: CUMULATIVE NICKEL STRIPPED FROM 710 – 850 MICRON BEADS Iminodiacetic Chelating Resins, pH 4, 30qC  0 10 20 30 40 50 60 70 0 1 2 3 4 5 6 7 Cu m ul at iv e N ic ke l S tr ip pe d [g  N i /  L  R es in ] Bed Volumes S930 S930+ S930+4888 TP207 TP207XL 20 HR SOAK 58  Results in Figure 4.12 are generally in agreement with those previously reported in Figure 4.10. Once again a significant portion of nickel was observed to have been difficult to strip and required an acid hold step. Cumulative stripping curves for the 1000 – 1180 micron size fraction are displayed in Figure 4.13.  Note that there is a duplicate strip curve for the S930+/4888 resin, illustrating repeatability of this method.  There is no duplicate strip curve for the XE818 resin, as the strip of the first load test was collected in a single flask and assayed using the procedures for the resins loaded at 500 ppm nickel to ensure both methods were yielding similar results.  FIGURE 4.13: CUMULATIVE NICKEL STRIPPED FROM 1000 – 1180 MICRON BEADS Iminodiacetic Chelating Resins, pH 4, 30qC     0 10 20 30 40 50 60 70 0 1 2 3 4 5 6 7 Cu m ul at iv e N ic ke l S tr ip pe d [g  N i /  L  R es in ] Bed Volumes XE 818 S930+4888 S930+4888 Repeat 20 HR SOAK 59  4.4  CONCLUSIONS FROM RESIN COMPARISON Loading rate and stripping rates of several large bead ion exchange resins designed for base metals RIP circuits have been compared.  Each of the three new RIP grade resins excels in at least one area. In terms of size, the large bead resin with the largest size distribution is the Ambersep XE818. The Purolite S930+/4888 resin is slightly smaller, while the Lewatit MonoPlus TP207XL is a significantly smaller product.  The Ambersep XE818 also exhibits the smallest change in bulk volume during conversion from sodium to hydrogen form. The large bead resin with the largest capacity is the Purolite S930+/4888.  The Lewatit MonoPlus TP207XL resin has a slightly lower capacity, while the Ambersep XE818 has a significantly lower capacity. The large bead resin with the fastest loading rate is the Lewatit MonoPlus TP207XL.  This resin also has the tightest size distribution of resin bead diameters.  4.5  SELECTION OF RESIN FOR FURTHER WORK The Lanxess Lewatit MonoPlus resins were selected for use in the remainder of this thesis.  The rationale for this choice is that the MonoPlus resins have a narrow size distribution.  Since resin bead diameter is known to have a significant impact on loading rates, using a resin with a narrow  size  distribution  will  allow  for  a  better  evaluation  of  other  parameters  impacting  the system.  60  5.0  QUALITATIVE INVESTIGATION INTO RATE DETERMINING STEP(S) In order to develop an engineering model for RIP circuit design, it is important to have an understanding of the rate limiting mechanisms present in the system.  As mentioned in Section 2.1.2, the three possible rate-limiting steps for ion exchange loading are: film diffusion, intraparticle  diffusion,  or  the  ion  exchange  process  itself.   Futhermore,  intraparticle  diffusion can be subdivided into two different processes.  In the first case, a diffusion gradient exists across the resin bead, and intraparticle diffusion can be described using Vermeulen’s Approximation to the infinite series solution of the Nernst-Planck equation.  In the second case, a sharp boundary exists within the resin bead where an unreacted core is surrounded by a shell of fully loaded resin. Being aware of the rate-limiting step (or steps if under some form of mixed control) will help in the selection of the appropriate engineering model.  However, it should be noted that the rate- limiting step may change as the resin bead loads.  For example, all ion exchange processes start out film diffusion limited, although sometimes for only a very short (and hence negligible) period of time. There are a number of experiments that can be conducted to qualitatively identify rate limiting steps.   First  of  all,  from the  form of  Equations  2.9,  2.14,  and  2.18,  it  is  clear  that  resin  bead diameter plays an important role in loading rates when loading is diffusion controlled.  For the case of film diffusion, loading rate is proportional to the resin bead diameter, whereas for both types of intraparticle diffusion, loading rate is proportional to the square of the resin bead diameter. The concentration of the ion to be loaded in the bulk solution also has an impact on loading rate.  This is true for all cases, except for the intraparticle diffusion described by Vermeulen’s Approximation. Loading rate can also be affected by the rate of agitation.  When the loading rate is film diffusion controlled, increasing the agitation rate can increase the rate of loading.  Note that this only  holds  for  a  certain  range,  as  there  is  an  upper  limit  to  the  effect  of  agitation  in  the  film diffusion limited regime.  Increasing the agitation rate has no effect when the resin is being controlled by intraparticle diffusion or the exchange rate. 61  The interruption test is helpful for identifying processes that are intraparticle diffusion controlled.  In an interruption test, the resin beads are separated from the solution mid-way through a test, left apart for a length of time, and then reintroduced to the solution.  If a discontinuity is observed in the loading rate curve, then intraparticle diffusion is playing a significant role in the system. The rate of complex formation / rate of ligand exchange could also be rate limiting.  How the various parameters affect loading rate under the different rate-controlling regimes are summarized in Table 5.1. TABLE 5.1: IMPACT OF VARIOUS PARAMETERS UNDER DIFFERENT RATE LIMITING REGIMES Increasing Parameter Film Diffusion Intraparticle Diffusion (Vermeulen) Shrinking Core Intraparticle Diffusion Exchange Process Resin Bead Diameter: dp Rate v dp  Rate v dp2 Rate v dp2 No Effect Bulk Solution Concentration [M] Increases No Effect Increases Increases Agitation Rate Increases to a point No Effect No Effect No Effect Interruption Test No Effect Discontinuity Discontinuity No Effect Rate of Complex Formation No Effect No Effect No Effect Increases  It has already been shown in Figure 3.4 that increasing the agitation rate from 300 to 450 RPM had  no  effect  at  a  feed  concentration  of  250  ppm  nickel.   The  effect  of  resin  bead  diameter, solution concentration, an interruption test, and the rate of complex formation on loading rate will each be discussed in turn.   62  5.1  EFFECT OF RESIN BEAD DIAMETER ON LOADING RATE To determine the effect of resin bead diameter on loading rate, the Lewatit TP207 resin was used, since the MonoPlus products have such a narrow size distribution.  Loading tests were conducted at 250 ppm nickel for the +500/-600, +600/-710, +710/-850, +850/-1000, and +1000/-1180 micron size fractions.  Results showing the rate of loading are displayed in Figure 5.1.  FIGURE 5.1: EFFECT OF RESIN BEAD DIAMETER ON LOADING RATE OF TP207 250 ppm Nickel, pH 4, 30qC  As expected, the larger resin beads load more slowly than the smaller ones.  However, Figure 5.1 also demonstrates that the final loading value for the large resin beads is significantly less than for the smaller resin beads.  To ensure that this was not simply due to the fact that the larger resin beads did not have enough time to reach equilibrium, the test on the +710 -850 Pm size fraction was run for 36 hours.  During this extended run, the loading curve was observed to 0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time [h] +500 -600 um +600 -710 um +710 -850 um +850 -1000 um +1000 -1180 um +1000 -1180 repeat 63  level off at a value of 55 g Ni / L Resin –a value significantly lower than the observed loading of 60 g Ni / L Resin for the +500 – 600 Pm size fraction at 20 hours. The fact that the final capacity varies with resin bead diameter suggests that the large resin beads do not contain as high a concentration of functional groups as the small beads. Furthermore, it implies that the distribution of functional groups within a resin bead is not uniform, and that the outer surface of the bead may contain a higher concentration of functional groups than the centre of the bead.  This is likely caused by the fact that the functionalization reagents need to diffuse into the bead during the manufacturing process.  If such a diffusion process is intraparticle diffusion limited, it makes sense that the smaller beads contain a higher concentration of functional groups. This raises a few practical issues of interest for RIP applications.  Since RIP requires large beads to allow for the separation of pulp and slurry, these results highlight the fact that tests need  to  be  conducted  on  only  the  portion  of  the  beads  that  will  be  used.   If  scoping  tests  are conducted  using  the  full  size  distribution  of  the  resin,  results  could  well  overpredict  the operating capacity of the resin being tested.  Furthermore, it also suggests that variations in the size distribution during the resin manufacturing process could result in lot to lot variations in the operating capacity of the large bead fraction of the resin that will be used for RIP. As was mentioned in Table 5.1, the loading rate is proportional to the resin bead diameter in film diffusion limited processes; it is proportional to the square of the resin bead diameter in intraparticle limited processes; and it is independent of the resin bead diameter in ion exchange limited processes.  In order to determine the dependence of loading rate on particle diameter, the time required for each size fraction to reach a loading of 10, 15, 20, 25, 30, 35, 40, and 45 g Ni / L Resin was determined.  To normalize the data, the time required to reach the target loading for each size fraction was divided by the time required for the 555 Pm size fraction to reach the same target loading.  This normalized loading time was plotted against the effective bead diameter (from Table 4.1) in Figure 5.2.  64   FIGURE 5.2:  EFFECT OF RESIN BEAD DIAMETER ON LOADING RATE TP207 Resin, 250 ppm Nickel, pH 4, 30qC  The power law fit to the data in Figure 5.2 suggests that the dependence of loading rate on particle diameter is best fit by raising dp to the 1.8 power.  Note that there is a significant amount of spread in the results for the different final target loadings at the large bead diameters. To get a sense on the bounds, a power law was fit to the data at 10 g Ni / L Resin and at 45 g Ni / L Resin.  Results from these fits yielded a fit of dp to the 1.5 power at 10 g Ni/ L Resin, and a fit of dp to the 2.6 power at 10 g Ni/ L Resin.  Although there is a significant spread in the data (likely caused by the fact that the final capacity is also a function of resin bead diameter), these results suggest that intraparticle diffusion is the dominant process in the system being studied.    y = (9.5x10-06)dp1.8 0 1 2 3 4 5 6 500 600 700 800 900 1,000 1,100 Ti m e Re la ti ve  to  T im e at  5 55  u m Effective Resin Bead Diameter (Hydrogen Form) [um] 10 g Ni / L R 15 g Ni / L R 20 g Ni / L R 25 g Ni / L R 30 g Ni / L R 35 g Ni / L R 40 g Ni / L R 45 g Ni / L R 65  5.2  EFFECT OF BULK SOLUTION CONCENTRATION ON LOADING RATE To determine if solution concentration affects loading rate, a series of loading tests on the TP207MP resin were carried out at 50, 100, 250, 500, 1000, 2500, and 5000 ppm nickel.  In order to better be able to see the various curves, results were filtered and only every 25th data point  was  plotted  in  Figure  5.3.   Results  clearly  show  that  solution  concentration  has  a significant effect on the overall loading rate.  Thus, the loading rate is dependent on more than just the intraparticle diffusion process described by Vermeulen’s Approximation.  FIGURE 5.3: EFFECT OF SOLUTION CONCENTRATION ON NICKEL LOADING RATE TP207MP Resin, pH 4, 30qC  5.3  INTERRUPTION TEST The interruption test started as a typical test at 250 ppm nickel; however, after one hour the resin  was  separated  from  the  test  solution  using  a  425  micron  screen.   The  test  solution  was placed  in  a  refrigerator  for  68  hours,  while  the  resin  was  kept  in  a  sample  vial  at  room 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time [h] 50 100 250 500 1000 2500 5000 66  temperature.  To resume the test, the solution was first placed in the hot water bath, heated to 30qC, and then the resin was reintroduced to the solution.  Results are displayed in Figure 5.4, along with the baseline for 250 ppm.  This figure clearly shows a discontinuity at 1 hour, confirming that intraparticle diffusion plays a significant role in this system.  FIGURE 5.4: INTERRUPTION TEST -68 HOUR INTERRUPTION AFTER ONE HOUR OF LOADING TP207MP Resin, 250 ppm Nickel, pH 4, 30qC   5.4  EFFECT OF COMPLEX FORMATION RATE For this particular system, both the selectivity and the rate of ligand exchange are higher for copper than for nickel.  Selectivity of iminodiacetic resins was discussed in Section 2.1.3, using the concept of decomplexing pH.  The rate of ligand exchange for octahedral nickel complexes is  typically  slower  than  for  the  corresponding  copper  complexes.   A  study  of  the  rate  of exchange of water ligands from various transition metal ions showed that the exchange half-life 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time [h] Baseline 250 ppm Ni 68 hour Interruption Test 67  for  nickel  was  on  the  order  of  10-4 – 10-5 seconds; cobalt was on the order of 10-5 – 10-6 seconds; whereas copper was on the order of 10-8 – 10-9 seconds (119).  Thus, running a series of tests for copper loading onto TP207MP should illustrate whether or not complex formation is affecting the overall rate of loading. The copper loading tests were conducted at feed concentrations that would correspond to the equivalent molar concentration of the nickel feed concentrations used in Section 5.2.  Results for the copper loading tests are displayed in Figure 5.5, and show that copper loaded much faster than nickel (Figure 5.3).  Thus, the rate of complex formation appears to affect the overall loading rate.  FIGURE 5.5: EFFECT OF SOLUTION CONCENTRATION ON COPPER LOADING RATE TP207MP Resin, pH 4, 30qC    0 10 20 30 40 50 60 70 0 5 10 15 20 Co pp er  L oa de d [g  C u / L R es in ] Time [h] 54 108 271 541 1082 2706 5412 68  5.5  CONCLUSIONS FROM QUALITATIVE TESTS Results from the qualitative tests presented in this chapter, suggest that although the process must initially be film-diffusion limited, it is predominantly limited by some combination of intraparticle diffusion and exchange rate.  For a fundamental loading rate model, the effect of selectivity and rate of complex formation would need to be quantified and separated from the intraparticle diffusion limitation.  To separate out these effects, experiments on very small resin beads would be required to determine the fundamental kinetics of both the forward and reverse reactions at various pH values.  Once the fundamental kinetics were determined, the detailed nature of the intraparticle diffusion mechanism would need to be ascertained.  This would likely require isotopic tracer tests such as those previously presented for this purpose (11). As mentioned in the literature review, such fundamental models typically require numerical solutions to partial differential equations.  Such models tend to be cumbersome and overly detailed for most engineering design applications.  Since the objective of this study is to develop an engineering model to quantify the effect of various parameters on an RIP circuit, such experiments are beyond the scope of the current work.  To achieve the goal of developing an adequate loading rate model, the film diffusion limited regime will be defined separately from the intraparticle diffusion / exchange rate limited regime.  An outline of the work necessary for developing a fundamental model can be found in the recommendations for future work.  69  6.0  QUANTITATIVE MODEL FOR FILM DIFFUSION COMPONENT As previously mentioned, the rate limiting step often changes as a resin bead loads, and all exchange processes start out film diffusion limited (although sometimes for a negligible amount of time).  The purpose of this chapter is to quantify the film diffusion component for the system being studied.  6.1  FILM DIFFUSION MODEL FIT To quantify the effect of film diffusion, results from the copper and nickel tests are compared. Since the rate of diffusion across the film should be similar for the two metals, quantifying the film diffusion component of loading can be done for the region where the loading curves for the two metals overlap.  The loading rates at 0.00085M (50 ppm nickel) and 0.0017M (100 ppm nickel) feed concentrations are plotted as fractional loading of resin in Figure 6.1.  FIGURE 6.1: LOADING RATE OF NICKEL AND COPPER ONTO TP207MP -FIRST 30 MINUTES pH 4, 30qC, constant metal concentration 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0 0.1 0.2 0.3 0.4 0.5 Fr ac ti on al  L oa di ng  o f R es in Time [h] Copper (Observed) Nickel (Observed) 0.0017 M Feed (100 ppm Ni) 0.00085 M Feed (50 ppm Ni) 70  Figure 6.1 shows that the two elements load at a similar rate early on, and then diverge.  The engineering model most commonly applied to film diffusion limitation is the linear approach to equilibrium.  This model is based on the analytical solution to isotopic exchange under infinite solution volumes and was described by Equations 2.7-2.9, repeated below:   d dt [Mഥ ] = kl([Mഥ ]eqb-[Mഥ ])          [2.7]  F = 1- exp(-klt)           [2.8]  kf = 6Df[M]dpį[Mഥ ]max           [2.9]  The linear approach to equilibrium was fitted to the first 0.2 hours of loading by adjusting the ratio of the film diffusion coefficient to the thickness of the boundary layer (Df į) as a fitting parameter.  The model fit (Df į = 2.7 x 10-5 m/s) is plotted in Figure 6.2 and shows good correlation at both 0.00085 M and 0.0017 M feed concentrations.  Note how the rate of nickel loading deviates from the film diffusion model earlier than the rate of copper loading, which is in-line  with  the  previous  observations  that  the  overall  rate  of  nickel  is  much  slower  than  the overall rate of copper loading.  71   FIGURE 6.2: LINEAR APPROACH TO EQUILIBRIUM FIT TO FIRST 12 MINUTES OF DATA pH 4, 30qC, constant metal concentration  6.2  THEORETICAL ESTIMATE OF FILM DIFFUSION COEFFICIENT To  confirm  the  validity  of  the  fitting  parameter,  a  rough  set  of  diffusion  calculations  were performed to obtain an estimate of the film diffusion coefficient.  These calculations involved estimating the Sherwood Number from the Frössling correlation for spheres, shown in Equation 6.1 (120).  The various dimensionless numbers are defined in Equation 6.2, and the velocity of the particles was estimated using a large particle settling correlation (> 100 micron), shown in Equation 6.3 (121).  Harriott has pointed out that using the terminal velocity is suitable for an order of magnitude estimate of the mass transfer coefficient (122).  The diffusion coefficient was estimated at 8 x 10-10 m2/s, based on interpolation from the literature (123), and adjusted for temperature using the common correlation shown in Equation 6.4 (120).  0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0 0.05 0.1 0.15 0.2 Fr ac ti on al  L oa di ng  o f R es in Time [h] Model Copper Nickel 0.0017 M Feed (100 ppm Ni) 0.00085 M Feed (50 ppm Ni) 72   Sh = 2 + 0.6 Re1 2ൗ  Sc1 3 ൗ                  [6.1]   Sh = kmdpDf = dpɁ  ;  Re = dpupUP  ; Sc = PUDf      [6.2]   up= ቈdp1.6gቀUp-Uቁ14 U0.4P0.6 ቉1 1.4ൗ          [6.3]   DT2=DT1 PT1PT2 T2T1          [6.4]  Results from these calculations estimate the Sherwood Number to be approximately 23, giving an estimated film thickness of 2 x 10-3 cm.  This is in good agreement with the observed fit parameter, which yields a film thickness of 3 x 10-3 cm for a diffusion coefficient of 8 x 10-10 m2/s.  These values are of the expected order of magnitude, reported as 10-2 to  10-3 cm (10). Therefore, the fit parameter was accepted as adequately representing the film diffusion component of the process.   73  7.0  DEVELOPMENT OF HYBRID CORRELATION FOR INTRAPARTICLE DIFFUSION As demonstrated during the qualitative tests, the rate of loading is predominantly controlled by some combination of intraparticle diffusion and exchange reaction.  To determine whether it is possible to obtain an adequate fit to the system, four engineering models that are commonly applied to ion exchange systems were fitted to the data.  7.1  EVALUATION OF EXISTING ENGINEERING LOADING MODELS The four models evaluated were the linear approach to equilibrium, the quadratic approach to equilibrium, the shrinking core model, and the pseudo-second order reaction.  Each of these models are fitted to the data for the loading of nickel onto TP207MP, presented in Figure 5.3, using the estimated resin bead diameter reported in Table 4.1.  To obtain the fit parameters, the sum  of  the  absolute  values  of  the  difference  between  the  observed  values  and  the  model predictions was minimized. Since many of the engineering models are written in terms of fractional approach to equilibrium, the nickel loading isotherm must first be defined.  To achieve this, the final loading values obtained in each of the nickel loading tests were used.  To describe the isotherm, the resin capacity and the selectivity coefficient, K, were taken as fitting parameters for the mass  action  law.   Figure  7.1  displays  the  isotherm fit  (K =  9.8  x  10-5, capacity = 1.97 eq/L), drawn on a semi-log plot. 74   FIGURE 7.1: EQUILIBRIUM ISOTHERM FOR NICKEL LOADING ONTO TP207MP pH 4, 30qC  7.1.1  LINEAR APPROACH TO EQUILIBRIUM The linear approach to equilibrium is known to give a poor fit to intraparticle diffusion, but it is frequently selected over other models due to ease of application.  Gluckauf and Coates first proposed applying the linear approach to equilibrium to intraparticle diffusion (21).  Their model was described in Equations 2.7 – 2.9, which are repeated here.  Note that the shape factor, Ș, is < 0.5.  ddt [Mഥ ] = k([Mഥ ]eqb-[Mഥ ])          [2.7]  F = 1- exp(-klt)           [2.8]  kl = kg = DpɄdp2           [2.9] 0 10 20 30 40 50 60 10 100 1,000 10,000 Eq ui lib ri um  N ic ke l L oa de d [g  N i /  L  R es in ] Solution Nickel Concentration [ppm] TP207 MP Mass Action Law Fit 75  The linear approach to equilibrium was fitted to three individual tests: 50, 500, and 5000 ppm nickel.  The equilibrium loading values ([Mഥ ]eqb)  at  each  concentration  were  based  on  the isotherm displayed in Figure 7.1, and the kl value adjusted to minimize the absolute value of the error between predicted and observed values.  The individual model fits of the linear approach to equilibrium to the observed data are displayed in Figure 7.2, and the best fit parameters of the kl parameter are tabulated as an inset to this graph.  FIGURE 7.2:  LINEAR APPROACH TO EQUILIBRIUM FIT TO INDIVIDUAL LOADING CURVES Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  Based on this figure, the linear approach to equilibrium provides a good fit at 50 ppm nickel, but the fit of the curve shape becomes poorer as the solution concentration of nickel increases. Furthermore, the inset table shows that the fitting parameter is not constant, but varies with the solution concentration.  This suggests that the linear approach to equilibrium can not adequately model the full dataset. 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 500 5000 Individual Run Model 5000 ppm 500 ppm 50 ppm ppm Ni kl [1/s] 50 0.00013 500 0.00023 5000 0.00049 76  7.1.2  VERMEULEN’S QUADRATIC APPROACH TO EQUILIBRIUM As  mentioned  in  the  literature  review,  the  exact  solution  to  the  differential  equations  for  the case of infinite solution volume and constant diffusivities is the infinite series of Equation 2.10 (22), which can be approximated by Vermeulen’s Approximation, described in Equations 2.11 – 2.14 (23).   F=1- 6 ʌ2 σ 1 j2 exp ൜ -4Dpʌ2j 2t dp 2 ൠ f j=1       [2.10]  d dt [Mഥ ]= 2kv൫[Mഥ ]eqb2 -[Mഥ ]2൯[Mഥ ]        [2.11]   t= -1 4kv ln ൬ [Mഥ ]eqb2 -[Mഥ ]2[Mഥ ]eqb2 ൰       [2.12]   F= ට1-exp൛-4kvtൟ        [2.13]   kv= ʌ2Dp dp 2          [2.14]  In order to determine whether or not Vermeulen’s approximation to the infinite series was adequate for modeling this system, both models were fitted to the curve generated at 250 ppm nickel, using the same fitted diffusion coefficient for both models.  Figure 7.3 displays these results and demonstrates that the simpler Vermeulen approximation is adequate for describing this type of intraparticle diffusion. 77   FIGURE 7.3: COMPARISON OF VERMEULEN’S APPROXIMATION AND BOYD’S EXACT SOLUTION Nickel Loading onto TP207MP at pH 4, 30qC, 250 ppm Ni  Figure 7.4 displays the individual curve fits for Vermeulen’s quadratic approach to equilibrium for tests run at 50, 500, and 5000 ppm nickel.  As before, the equilibrium resin loadings were taken from the isotherm generated in Figure 7.1, and the effective diffusion coefficient in the resin bead, Dp was adjusted as a fitting parameter to each individual curve. 0 10 20 30 40 50 60 0 2 4 6 8 10 12 14 16 18 20 N ic ke l L oa de d [g  N i /  L  R es in ] Time [h] Data -250ppm, 500um Converged Series Vermeulen Approximation Fit parameter: Dp = 8.9 x 10-13 m2/s 78   FIGURE 7.4: VERMEULEN MODEL FIT TO INDIVIDUAL LOADING CURVES Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  Once again the fitted effective intraparticle diffusion coefficient, Dp, varies from test to test, suggesting that the Vermeulen model is not adequately describing the effect of solution concentration on nickel loading rate.  However, the profile of the loading curves do match the shape of the Vermeulen model, suggesting that the Vermeuelen model should make an excellent starting point for modeling nickel loading in this system.  Figure 7.5 shows the best Vermeulen fit to the full set of test results, using only a single effective diffusion coefficient for the entire dataset.  From this figure, it is clear that the Vermeulen model underpredicts the effect of solution concentration when using a single effective diffusivity coefficient (Dp = 1.0 x 10-12 m2/s). 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 500 5000 Individual Run Model 5000 ppm 500 ppm 50 ppm ppm Ni Dp [m2/s] 50 5.8 x 10-13 500 9.6 x 10-13 5000 2.1 x 10-12  79   FIGURE 7.5: VERMEULEN MODEL FIT TO ALL CURVES SIMULTANEOUSLY Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  7.1.3  SHRINKING CORE MODEL The shrinking core model was described in Chapter 2 using Equations 2.15-2.18:  dF dt = kcdc 2൫dp-dc൯          [2.15]   t= 1+2൫1-F൯-3(1-F)2 3ൗ kc         [2.16]   F=1- ቄ1 2 + sin ቀarcsinൣ1-2kct൧ 3 ቁቅ 3       [2.17]   kc= 24Ds[M] dp 2[Mഥ ]max         [2.18] 0 10 20 30 40 50 60 0 2 4 6 8 10 12 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 80  Figure 7.6 displays the best fit of the shrinking core model to the individual loading curves at 50, 500, and 5000 ppm.  Once again, the equilibrium resin loading values are based on the isotherm generated in Figure 7.1, and the effective diffusion coefficient through the reacted layer, Ds was adjusted as a fit parameter for each individual curve.  From this figure, it can be seen that the shape of the shrinking core model predicts the data very well for the first portion of loading, but overpredicts the data as the test nears equilibrium.  FIGURE 7.6:  SHRINKING CORE MODEL FIT TO INDIVIDUAL LOADING CURVES Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  Once again, the best fit effective diffusion coefficient through the reacted layer, Ds is  not constant.  Figure 7.7 shows the best simultaneous fit to all the loading curves, using only a single diffusivity coefficient.  From this figure, it is clear that the shrinking core model overpredicts the effect of solution concentration when using a single diffusivity coefficient (Ds = 2.0 x 10-10 m2/s). 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 500 5000 Individual Run Model 5000 ppm 500 ppm 50 ppm ppm Ni Ds [m2/s] 50 4.0 x 10-13 500 6.8 x 10-13 5000 1.5 x 10-12  81   FIGURE 7.7: SHRINKING CORE MODEL FIT TO ALL CURVES SIMULTANEOUSLY Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  7.1.4  SECOND ORDER RATE MODEL As mentioned in the literature review, a second order chemical reaction is often used as an analogy for ion exchange.  The Thomas model, described by Equations 2.1, 2.19, and 2.20, was used to fit the data presented in this work.  2R-H + M2+ Æ R2-M + 2H+         [2.1]    d dt [Mഥ ]=kfwd[M][Hഥ]-krev[H][Mഥ ]      [2.19]    d dt [Mഥ ]={kfwd[M]+2krev[H]}[Mഥ ]-kfwd[M][Mഥ ]max   [2.20]  0 10 20 30 40 50 60 0 2 4 6 8 10 12 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 82  Figure 7.8 shows the second order model fitted to the individual test data at 50, 500, and 5000 ppm nickel.  The final equilibrium resin loadings were based on the isotherm presented in Figure 7.1 and the two empirical rate constants, kfwd and krev, were determined by trial and error for each curve.  The differential equation 2.20 was solved numerically, using a fourth-order Runge-Kutta algorithm.  FIGURE 7.8  THOMAS MODEL FIT TO INDIVIDUAL LOADING CURVES Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  The shape of the Thomas model fits each individual dataset fairly well; however, when a single set of fitting parameters is used to model the various infinite solution volume tests run at different concentrations, the model overpredicts the system’s dependence on solution concentration.  This is illustrated in Figure 7.9, where the best fitting parameters to the entire dataset are plotted. 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 500 5000 Individual Run Model 5000 ppm 500 ppm 50 ppm ppm Ni kfwd [L / mol s] krev [L / mol s] 50 0.063 0.21 500 0.012 0.17 5000 0.002 0.13  83   FIGURE 7.9: THOMAS MODEL FIT TO ALL CURVES SIMULTANEOUSLY Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  While it is possible to debate ad infinitum the best analogous chemical reaction model for ion exchange systems, it is doubtful that any model with integer coefficients would be able to fit the range of solution concentrations in this work.  As was seen by the Vermeulen model, if the rate of loading is independent of nickel concentration (i.e. rate proportional to [Ni]0), the family of curves will be too close together to represent the dataset for tests run at 50 – 5000 ppm nickel. On the other hand, the Shrinking Core model demonstrated that if the rate of loading is linearly dependent on nickel concentration (i.e. rate proportional to [Ni]1), the family of curves will be spread too far apart to represent the dataset for tests run at 50 – 5000 ppm nickel.    0 10 20 30 40 50 60 0 2 4 6 8 10 12 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 84  7.2  DEVELOPMENT OF HYBRID CORRELATION Since none of the standard engineering models used for ion exchange were able to adequately describe the system for the full range of concentration, a new correlation was developed.  The approach taken to address the fact that the dependence of this system on solution concentration lies somewhere between [Ni]0 and [Ni]1 was to develop a hybrid correlation that combines the Vermeulen ([Ni]0) and Shrinking Core ([Ni]1)  models.   These  models  were  chosen  as  the starting point for the hybrid correlation, in lieu of the second order rate analogy, because they both have relatively simple analytical solutions.  Note that the physical significance of the various fit parameters in the hybrid correlation will be explored in Section 8.2. Mathematically,  the  two  models  appear  to  be  quite  different,  but  both  appear  to  fit  the  data quite  well  for  the  first  portion  of  the  loading  curve.   Therefore,  an  attempt  was  made  to  see whether or not a correlation between the two constants, kv and kc, could be made that minimizes the error between the two models.  It was found that the error is minimized when kv = 0.667 kc. Given that the Vermeulen model better fit the observed nickel loading data, it was decided that it would be better to approximate the Shrinking Core Model by a Vermeulen expression, rather than the other way around. Figure 7.10 displays the shrinking core model as open circles for kc values of 0.05, 0.01, 0.005, 0.002, 0.001, 0.0005, and 0.0001.  Lines corresponding to the best approximation of the Vermeulen model to these points (i.e. kv = 0.667 kc) are also drawn in.  This graph shows good agreement between the two models up to a fractional conversion of 0.8, at which point the shrinking core model predicts higher values than the Vermeulen model.  Since the Vermeulen approximation is underpredicting the shrinking core model, the approximation should result in conservative predictions if it were ever applied to a situation which followed true shrinking core behaviour. 85   FIGURE 7.10: USING VERMEULEN MODEL TO APPROXIMATE SHRINKING CORE MODEL  To combine the Vermeulen model and the Vermeulen approximation to the Shrinking Core model into a single hybrid correlation, a second fit parameter, D, was introduced.  The various constants in the expression were combined in such a way that as D o 0,  the  correlation becomes the Vermeulen model, and as Do 1, the correlation becomes the Vermeulen approximation to the shrinking core model. The hybrid correlation results in a new constant, kh (Equation 7.1), which can then be used in the standard Vermeulen model to create the hybrid correlation.   Note that kh,  like  kf and  kc, varies with solution concentration.  kh= Ɏ2Dappdp2 ቀ 16[M]Ɏ2[Mഥ]maxቁȽ          [7.1] 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 50 100 150 200 250 300 Fr ac ti on  C on ve rs io n Time [s] Vermeulen Approximation Shrinking Core Model kc = 0.05 kc = 0.0001 kc = 0.0005 kc = 0.001 kc = 0.002 kc = 0.005 kc = 0.01 86  The analytical solution to the Vermeulen model, using kh in lieu of kv is displayed in Equation 7.2.  The steps used to derive the new constant are tabulated in Table 7.1.   F = ට1-exp൛-4khtൟ         [7.2]  TABLE 7.1: DEVELOPMENT OF HYBRID CORRELATION EQUATION Vermeulen Model:  F =ට1-exp൫-4kvt൯ Model Standard Vermeulen Vermeulen Approx to Shrinking Core Constant kv = ʌ2Dpdp2  kvൎ23 kc;   kc =  24Ds[M]dp2[Mഥ ]max F = ඩ1-exp ൭- 4Ɏ2Dp tdp2 ൱ ඩ1-expቌ- 4* 23 *24Ds[M]t[Mഥ ]max  dp2 ቍ Collecting constants and pulling out 4ʌ2  ඪ1-exp൮- 4*ɎଶɎଶ *16(Ds[M]) t[Mഥ ]max dp2 ൲ Hybrid Correlation ۣ ളള ളള ളള ളള ളള ളለ 1-exp ۉ ۈ ۈ ۇ- 4Ɏ2*Dapp ൭16Ɏଶ *[M] [Mഥ ]max൘ ൱ D  t dp2 ی ۋ ۋ ۊ  Simplifying F = ඥ1-exp(-4kht) ,  kh=Ɏ2Dappdp2 ቆ 16[M]Ɏ2[Mഥ ]maxቇȽ  The hybrid correlation was fitted to the set of curves in Figure 7.11.  In this figure, the final resin loadings are all based on the isotherm presented in Figure 7.1, and both Dapp and D were adjusted as fitting parameters. 87   FIGURE 7.11: HYBRID CORRELATION FIT TO ALL CURVES SIMULTANEOUSLY Nickel Loading onto TP207MP at pH 4, 30qC, and constant nickel concentration  Figure 7.11 suggests that the hybrid correlation is able to adequately describe the set of curves, using the following fit parameters: Dapp = 3.9 x 10-12 m2/s, D = 0.28.  The following chapter will investigate whether the hybrid correlation can be applied to the loading of nickel onto another iminodiacetic resin or to the loading of metals other than nickel.  Once this has been explored, the significance of the two fit parameters can be discussed.   0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 88  8.0  APPLICATION OF HYBRID CORRELATION TO OTHER BATCH SYSTEMS In order to ensure that the hybrid correlation is flexible enough to be able to accommodate more than one set of conditions, it was fitted to several additional datasets.  This also generates sufficient data to begin to understand the significance of the various fitting parameters.  8.1  FITTING THE HYBRID CORRELATION TO ADDITIONAL DATA SETS The hybrid correlation was fitted to nickel loading onto TP207 and two lots of TP207XL (lot CHC 5034, which was used in the resin comparison and lot CHE 50019, which was used in the RIP miniplant –see Chapter 10).  It is also fitted to cobalt and copper loading onto TP207MP, and to copper loading onto TP207XL (lot CHE 50019). For each of these datasets, the equilibrium isotherm was first modeled using the mass action law, and then the predicted equilibrium values were used with the hybrid correlation.  The full set of loading isotherms can be found in Appendix C. Since TP207XL is considered a large bead version of TP207MP, a single selectivity coefficient, K, was used for the set of isotherms, but the resin capacity was left as a fitting parameter for each dataset.  For loading of the various metals onto TP207MP, it was found that a single resin capacity value was insufficient to generate a good fit to the data.  Therefore, the resin capacity was left as a fitting parameter for each individual dataset.  It is interesting to note that the slight variation in resin capacity follows the resin selectivity: Cu > Ni > Co.  Table 8.1 displays the isotherm model parameters from the various datasets. TABLE 8.1: ISOTHERM FIT PARAMETERS Resin TP207 TP207XL CHC5034 TP207XL CHE50019 TP207MP TP207MP TP207MP TP207XL CHE50019 Metal Ni Ni Ni Ni Co Cu Cu K 3.9 x 10-5 9.8 x 10-5 5.0 x 10-5 7.5 x 10-4 Capacity [eq/L] 2.12 2.01 2.36 1.97 1.91 2.11 2.50  89  8.1.1  NICKEL LOADING ONTO TP207 A series of nickel loading tests were carried out on the +710 – 850 micron size fraction of the TP207 resin.  The best fit parameters for this data set were D= 0.28, and Dapp = 2.7 x 10-12 m2/s. Test results and the fit of the hybrid correlation can be found in Figure 8.1.  FIGURE 8.1: HYBRID CORRELATION FIT TO NICKEL LOADING ONTO TP207 pH 4, 30qC, constant nickel concentration  The hybrid correlation appears to fit reasonably well to this dataset.  The fitting parameter Dapp is lower for TP207 than was found for the TP207MP resin.  This makes sense since the TP207MP resin has been reported to have a porosity of 44%, while the TP207 resin’s porosity is reported to be 36% (118).  0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 90  8.1.2  NICKEL LOADING ONTO TP207XL (LOT CHC 5034 –RESIN COMPARISON) The lot of TP207XL used in the resin comparison work reported in Chapter 4 (lot CHC 5034) was evaluated at the +710 -850 micron size fraction for the full range of nickel feed concentrations.  The fitting parameters for this data set were D= 0.32, and Dapp = 4.5 x 10-12 m2/s.  Test results and the fit of the hybrid correlation can be found in Figure 8.2.  FIGURE 8.2: HYBRID CORRELATION FIT TO NICKEL LOADING ONTO TP207XL, LOT CHC 5034 pH 4, 30qC, constant nickel concentration  Once again, the hybrid correlation was found to fit the data set quite well.   0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 91  8.1.3  NICKEL LOADING ONTO TP207XL (LOT CHE 50019 –MINIPLANT) A second lot of TP207XL resin (lot CHE 50019) was also evaluated for the full range of nickel feed concentrations at the +710 -850 micron size fraction.  This lot was ordered for use in the RIP miniplant (see Chapter 10).  The fitting parameters for this data set were D = 0.36, and Dapp = 4.4 x 10-12 m2/s.  Test results and the fit of the hybrid correlation can be found in Figure 8.3.  FIGURE 8.3: HYBRID CORRELATION FIT TO NICKEL LOADING ONTO TP207XL, LOT CHE 50019 pH 4, 30qC, constant nickel concentration  The hybrid correlation was shown to provide an adequate fit to this data set.  It is interesting to note that the best fit parameters for the two lots of TP207XL were different, particularly in the case  of  the  reported  alpha  values.   The  principal  difference  in  these  two  lots  is  their  total capacities (2.36 for CHE 50019 vs. 2.11 for CHC 5034). 0 10 20 30 40 50 60 70 0 5 10 15 20 N ic ke l L oa di ng  o n R es in  [g  N i /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 Model 500 1000 ppm Model 1000 2500 Model 2500 92  8.1.4  COBALT LOADING ONTO TP207MP A series of cobalt loading tests were done on the +500 – 600 micron size fraction of the TP207MP resin.  The best fit parameters for this data set were alpha = 0.25, and Dapp = 4.2 x 10-12 m2/s.  Test results and the fit of the hybrid correlation can be found in Figure 8.4.  FIGURE 8.4: HYBRID CORRELATION FIT TO COBALT LOADING ONTO TP207MP pH 4, 30qC, constant cobalt concentration  Once again, the hybrid correlation was shown to fit the data set quite well.  The fit parameters for nickel and cobalt for this resin are quite similar; although, the cobalt alpha value (0.25) is slightly lower than the nickel alpha value (0.28).   0 10 20 30 40 50 60 0 2 4 6 8 10 12 Co ba lt  L oa di ng  [ g Co  /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 93  8.1.5  COPPER LOADING ONTO TP207MP The +500 – 600 micron size fraction of the TP207MP resin was also evaluated for copper loading.  Loading was much faster with copper and the best fit parameters for this data set were D = 0.62, and Dapp = 3.7 x 10-11 m2/s.  Test results and the fit of the hybrid correlation for the first 12 hours of loading can be found in Figure 8.5.  FIGURE 8.5: HYBRID CORRELATION FIT TO COPPER LOADING ONTO TP207MP pH 4, 30qC, constant copper concentration  The hybrid correlation was found to fit copper loading quite well.  It is very interesting to note that the fit parameters for copper loading differ significantly different from those for nickel and cobalt loading.  The apparent diffusivity, Dapp, is an order of magnitude larger, and the alpha fit parameter is approximately double what was reported for nickel and cobalt.  0 10 20 30 40 50 60 70 0 2 4 6 8 10 12 Co pp er  L oa di ng  [ g Cu  /  L  R es in ] Time [h] 50 ppm Model 50 100 ppm Model 100 250 ppm Model 250 500 ppm Model 500 1000 ppm Model 1000 2500 ppm Model 2500 5000 ppm Model 5000 94  8.1.6  COPPER LOADING ONTO TP207XL (LOT CHE 50019 –MINIPLANT) A series of copper loading tests were done on the +710 – 850 micron size fraction of the TP207XL resin, lot CHE 50019 (RIP miniplant).  Note that due to the high capacity of this resin,  only 4.4 mL of resin were used per test,  in lieu of the typical 5 mL resin per test.   The best fit parameters for this data set were D = 0.62, and Dapp = 2.8 x 10-11 m2/s.  Test results and the fit of the hybrid correlation can be found in Figure 8.6.  FIGURE 8.6: HYBRID CORRELATION FIT TO COPPER LOADING ONTO TP207XL, LOT CHE 50019 pH 4, 30qC, constant copper concentration   The hybrid correlation was shown to fit the data set reasonably well.  The fitting parameters for copper loading onto TP207XL are similar to the fitting parameters reported for copper loading onto TP207MP. 0 10 20 30 40 50 60 70 80 90 0 5 10 15 20 Co pp er  L oa di ng  o n R es in  [g  C u / L R es in ] Time [h] 54 ppm Model 54 108 ppm Model 108 271 ppm Model 271 541 ppm Model 541 1082 ppm Model 1082 2706 ppm Model 2706 5412 ppm Model 5412 95  8.2  PHYSICAL SIGNIFICANCE OF HYBRID CORRELATION FIT PARAMETERS Ideally, even in the case of empirical correlations, the various fit parameters should trend with experimental observations, in order to allow one to estimate coefficients for similar systems. Table 8.2 displays the empirical fit parameters for the various systems studied. TABLE 8.2: SUMMARY OF HYBRID FIT PARAMETERS Resin Metal Dapp [m2/s] Alpha TP207MP Cobalt 4.2 x 10-12 0.25 TP207 (+710 -850um) Nickel 2.7 x 10-12 0.28 TP207MP Nickel 3.9 x 10-12 0.28 TP207XL (CHC 5034) Nickel 4.5 x 10-12 0.32 TP207XL (CHE 50019) Nickel 4.4 x 10-12 0.36 TP207MP Copper 37 x 10-12 0.62 TP207XL (CHE 50019) Copper 28 x 10-12 0.62   8.2.1  APPARENT DIFFUSIVITY The first fit parameter, Dapp, trends well with the rate of loading.  From the experimental observations, the rate of loading onto TP207MP can be described as Cu > Ni ~ Co; which matches the magnitudes of the Dapp values.  Furthermore, the resin comparison work for nickel in the +710 -850 micron size range showed that TP207XL > TP207, which is also reflected in the magnitude of the Dapp values.  Comparing the nickel Dapp for  the  two  lots  of  TP207XL suggests that for a given resin, Dapp is only accurate to one significant figure (i.e. Dapp = 4 +/- 1 m2/s). To check whether the apparent diffusivities are of the correct order of magnitude, they were compared to resin diffusivities published in the literature.  As a rule of thumb, diffusivities in non-selective ion exchangers, compared to diffusivities in water, are typically 1/10 for monovalent ions, 1/100 for divalent ions, and 1/1000 for trivalent ions (30).  Diffusion coefficients for copper and nickel in water are reported as being of the order of 10-10 m2/s (123), suggesting intraparticle diffusivities of the order of 10-12 m2/s.  In the case of copper and cobalt 96  loading onto the old Dowex A1 iminodiacetic chelating resin, self diffusion coefficients of 9.78 x 10-12 m2/s for copper, and 3.93 x 10-12 m2/s for cobalt are reported (9).  While a self-diffusion coefficient for nickel loading onto an iminodiacetic resin was not found in the literature, nickel loading onto a macroporous aminophosphonic chelating resin was reported as 1.92 x 10-13 m2/s in the gel phase, and 6.35 x 10-11 m2/s in the internal pores, resulting in an overall effective diffusivity of 1.63 x 10-12 m2/s (12).  Consequently, the empirical apparent diffusivity values (10-11 – 10-12 m2/s), obtained with this correlation appear to at least be of the right order of magnitude.  8.2.2  ALPHA The second empirical fit parameter, D, can be considered a measure of how sharp a boundary exists between the loaded and unloaded portion of the ion exchanger.  As D approaches 1, the resin should be loading in a shrinking core fashion, whereas, when D approaches 0, the resin should be loading in a Vermeulen intraparticle diffusion fashion.  This means that it is reasonable to expect that a much sharper boundary would exist in systems with a high D value. To test this hypothesis, TP207MP beads were loaded to approximately 1 eq/L (50% capacity) with each metal from 0.0085M solutions (500 ppm nickel, 502 ppm cobalt, and 541 ppm copper).  The resin was removed from the test solution, mounted using a rapid cold cure acrylic mount (Lecoset 100), polished to approximately 27 microns, and observed under an optical microscope  to  view  whether  or  not  a  sharp  boundary  was  present.   Note  that  the  time  from removing the resin from solution to the time the resin was observed under the microscope was on the order of an hour. Figure 8.7 shows photographs of the three resin beads loaded to 1 eq/L.  Note that no boundary or gradient is visible for nickel or cobalt (under the microscope the beads simply had a greenish or pinkish hue), but that a clear ring can be seen for copper (strong blue ring could even be seen without any magnification).  This is in agreement to what one would expect based on the values of the fit parameter D.   It  is  interesting  to  note  that  the  decomplexing  pH for  copper  (1.0)  is significantly lower than for nickel (2.1) or cobalt (2.6), as previously mentioned in Chapter 2 (32).  Further work would be required to determine if a relationship exists between the decomplexing pH and the fit parameter D. 97    FIGURE 8.7: PHOTOGRAPHS OF RESIN BEADS LOADED TO 1 EQ / L  8.2.3  QUALITATIVE EVALUATION OF EFFECT OF PH ON ALPHA To determine if the operating pH affects the fitting parameter D, copper was loaded at pH 2 and pH 3 under semi-constant solution concentrations of 5412 ppm, 541 ppm, and 54 ppm.  It should be noted that the pH controller used was factory set to turn on at 0.1 pH units below the set point and to turn off at 0.1 pH units above the set point.  Because the copper stock solution is being added with the sodium hydroxide used to maintain constant pH, tests at lower pH values had larger swings in beaker copper concentration.  At pH 3, the Jenco controller was used to maintain pH in the test beaker; however, at pH 2, the feed pump was manually turned on at a pH read-out of 1.99 and turned off at a pH read-out of 2.01.  Additionally, the pH unit’s accuracy is only +/- 0.02 pH units, so additional noise in mass balances is expected at lower pH values.  Therefore, these tests at lower pH are reported merely for qualitative purposes, not quantitative purposes.  A summary of how test parameters were affected based on the target pH is presented in Table 8.3; data on observed solution concentrations and mass balances can be found in Appendix D.  Co Ni Cu  98  TABLE 8.3:  EFFECT OF TARGET PH ON TEST CONDITIONS Test pH 2 3 4 Controller on/off  1.99 / 2.01 2.90 / 3.10 3.90 / 4.10 Corresponding variation in [Cu] +/- 7 ppm +/- 8 ppm +/- 1 ppm pH unit accuracy +/- 0.02 +/- 0.02 +/- 0.02 Corresponding error in resin loading [g Cu / L Resin] +/- 2  +/- 0.2  +/- 0.02  To see if lowering the solution pH had a dramatic effect on the alpha parameter, the hybrid model was fit to the three datasets at pH 2 and 3.  Since tests were only run at three solution concentrations, insufficient data was available to generate isotherms, and so the observed final resin loading was taken as the equilibrium loading.  The one exception to this was the 54 ppm test at pH 2, where the equilibrium loading was allowed to vary as a fitting parameter since the resin had not reached equilibrium when the test was stopped. The observed loading rates and best fit hybrid correlation curves are plotted in Figure 8.8.  The top plot corresponds to tests at pH 4, the plot on the bottom left corresponds to tests at pH 3, and the plot on the bottom right corresponds to tests run at pH 2.  In these curves, note how the final copper loading values drop significantly as the pH decreases.  99              FIGURE 8.8: EFFECT OF SOLUTION PH ON COPPER LOADING AT 5412, 541, 54 PPM COPPER 30qC, constant copper concentration  The best fit parameters for the hybrid correlation at the various pHs can be found in Table 8.4. TABLE 8.4: HYBRID CORRELATION FIT PARAMETERS FOR COPPER LOADING AT PH 4, 3, 2 pH 4 3 2 Dapp [m2/s] 2.8 x 10-11 2.8 x 10-11 2.2 x 10-11 Alpha 0.62 0.59 0.53  Results suggest that decreasing the pH from 4 to 2 resulted in a significant drop in the final resin loadings, and a slight drop in the alpha parameter.  Understanding how the equilibrium resin loadings, the alpha values, and the decomplexing pH correlate is beyond the scope of this thesis, but would make an interesting study for future work. 0 10 20 30 40 50 60 70 80 90 0 2 4 6 8 10 12 Co pp er  L oa de d on  R es in  [ g Cu  /  L  R es in ] Time [h] Data MODEL 5412 ppm 541 ppm 54 ppm 0 10 20 30 40 50 60 70 80 90 0 2 4 6 8 10 12 Co pp er  L oa de d on  R es in  [g  C u / L R es in ] Time [h] Data MODEL 5412 ppm 541 ppm 54 ppm 0 10 20 30 40 50 60 70 80 90 0 2 4 6 8 10 12 Co pp er  Lo ad ed  o n Re si n [g  C u / L Re si n] Time [h] Data MODEL 5412 ppm 541 ppm 54 ppm 100  8.3  EFFECT OF BEAD SIZE ON HYBRID CORRELATION The Lewatit  MonoPlus  TP207XL (dp = 736 Pm) is  a  larger  version  of  the  Lewatit  MonoPlus TP207 (dp = 552 Pm), so it would be reasonable to expect that the kinetic parameters for these two resins should be similar. Given that the observed capacity for TP207XL, Lot CHC 5034 (2.01 eq/L) is quite close to the observed capacity of the TP207MP lot tested (1.97 eq/L), these two resins were modeled as a single system to see if the hybrid correlation adequately captures the effect of resin bead diameter during nickel loading at pH 4.  The best fit parameters to the combined system were  D = 0.30 and Dapp = 4.3 x 10-12 m2/s. Figure 8.9 plots the hybrid correlation fit to the TP207MP data at 50, 500, and 5000 ppm nickel, and Figure 8.10 plots the hybrid correlation fit to the TP207XL data at the same concentrations.  In these graphs, the hybrid correlation is plotted using the best fit parameters from the TP207MP dataset in blue (D = 0.28 and Dapp = 3.9 x 10-12 m2/s), the best fit parameters from the combined dataset in black, and the best fit parameters from the TP207XL dataset in red (D = 0.33 and Dapp = 5.0 x 10-12 m2/s).  FIGURE 8.9: TP207MP/TP207XL COMBINED HYBRID CORRELATION FIT TO TP207MP DATA pH 4, 30qC, constant nickel concentration 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm 500 ppm 5000 ppm MODEL MP MODEL XL Combined Model MP 101   FIGURE 8.10: TP207MP/TP207XL COMBINED HYBRID CORRELATION FIT TO TP207XL DATA pH 4, 30qC, constant nickel concentration  Results show that there is very little difference between the three sets of best fit parameters, and both the TP207MP (dp = 552 Pm), and the TP207XL (dp = 736 Pm) are adequately described. Therefore, it would appear that the hybrid correlation is adequately capturing the effect of resin bead diameter.  8.4  LIMITATIONS OF HYBRID CORRELATION The hybrid correlation has been shown to be adaptable to model single element loading of nickel, copper, and cobalt from pure synthetic solutions onto iminodiacetic resins in the hydrogen form.  Note that all experiments presented thus far to develop this correlation have been conducted under infinite solution volume conditions. 0 10 20 30 40 50 60 0 5 10 15 20 N ic ke l L oa di ng  [g  N i /  L  R es in ] Time [h] 50 ppm 500 ppm 5000 ppm MODEL MP MODEL XL Combined Model XL 102  The next three chapters of this thesis explore whether the hybrid correlation can be extended to non-infinite solution volume conditions.  This is achieved through running batch tests with variable solution concentration and through the operation of an RIP miniplant. Application of the hybrid correlation to a system containing impurities is beyond the scope of this thesis.  However, it is believed that the hybrid correlation should be able to provide bounds for loading a desired element (such as nickel) in the presence of impurities (such as manganese and calcium).  The upper bound would be determined by applying the hybrid correlation to the loading of nickel onto the resin in the hydrogen form.  To determine the lower bound, the resin would first need to be fully loaded with the impurities present in the feed stream.  The impurity- laden resin could then be used in a series of nickel loading experiments.  The hybrid correlation could be fit to the data from these experiments to determine the lower bound on loading.  This should be explored in future work. The effect of solids on the loading rate is also beyond the scope of this thesis.  It is important to note that several researchers have observed the redissolution of various metals from the solid phase to the solution phase during the operation of RIP circuits (18,102,104,5,106), and this would need to be taken into account in a circuit model.  103  9.0  VERIFICATION OF HYBRID CORRELATION (BATCH TESTS) Based on the results presented thus far, base metal loading onto an iminodiacetic resin at pH 4 is initially film diffusion limited, but then becomes limited by some combination of intraparticle diffusion and/or the exchange process.  In Chapter 6, a quantitative model was developed to describe the film diffusion limited regime, and in Chapter 7, a new correlation was developed to describe the intraparticle diffusion / exchange process limited regime.  Since the film diffusion model is based on a linear approach to equilibrium and the intraparticle diffusion model is based on a non-linear approach to equilibrium, combining the two rate limiting regimes into a single equation with a combined constant would be highly complex. In lieu of a combined model, a modified Helfferich number was developed to determine which process was rate-limiting at any given point in time.  Furthermore, to determine whether or not an  additional  correlation  was  required  to  describe  mixed  control  (i.e.  both  film  diffusion  and intraparticle diffusion / exchange process significant), a series of batch tests with changing solution composition were conducted and modeled.  9.1  MODIFIED HELFFERICH NUMBER In order to qualitatively determine whether an ion exchange process was predominantly film diffusion limited or intraparticle diffusion limited, Helfferich developed a criterion, now referred to as the Helfferich number.  This criterion was developed by taking the ratio of the predicted time to reach 50% loading under film diffusion conditions to the predicted time to reach 50% loading under intraparticle diffusion control. The Helfferich number was defined for the case of complete conversion of an ion exchange resin  initially  fully  loaded  with  ion  A  to  being  fully  loaded  with  ion  B.   Furthermore,  it  was defined assuming that the two ions were of equal mobility and equal valence, and that the reaction was carried out under infinite solution volume conditions.  Helfferich’s criterion for finding the overall rate determining step is described in Equations 9.1 and 9.2 (10).  Note that the separation factor, SBA , is only equal to the selectivity coefficient, K, for the exchange of ions of unit valence.  104   For film diffusion:    2[Aഥ]maxDpɁ[A]Dfdp ൫5+2SBA൯ب1       [9.1]  For intraparticle diffusion:   2[Aഥ]maxDpɁ[A]Dfdp ൫5+2SBA൯ا1      [9.2]   where:  SBA = [Aഥ][B][A][Bഥ]          [9.3] As proposed by Helfferich, this criterion only gives the rate limiting step at the moment that the resin is loaded to 50% capacity.  For the ion exchange process being studied, the resin initially starts out film-diffusion controlled, goes through a zone of mixed control, and then ends up controlled by intraparticle diffusion and/or the exchange reaction.  For development of an RIP circuit model, there would be value in modifying the Helfferich criterion so that it is able to track the rate limiting step as the resin loads.  To achieve this goal, the ratio of the time it would take to load a resin to any given fraction of loading, F, under film diffusion control (tf) over the time to achieve the same loading using the hybrid correlation (th) was taken.  This is illustrated in Equations 9.4 – 9.6; a full derivation of the modified Helfferich Number can be found in Appendix E.  Film Diffusion Control: tf = -ln൫1-F൯kf         [9.4]  Hybrid Correlation:  th = -ln൫1-F2൯4kh         [9.5]  Modified Helfferich No: 4khln൫1-F൯kfln൫1-F2൯         [9.6]  The same criteria for determining rate limiting step can be used as are used with the original Helfferich criterion (i.e. film diffusion if >>1, intraparticle diffusion if << 1, mixed control if | 1).  For the overall loading rate model, film diffusion control will be used for modified Helfferich values > 1, and the hybrid correlation will be used for modified Helfferich values < 1.  Therefore, by examining test results where the modified Helfferich number is | 1, it will be possible to determine whether an additional model is required to describe mixed control. 105  9.2  BATCH TESTS WITH CHANGING SOLUTION CONCENTRATIONS In Chapter 7, the hybrid correlation was developed under conditions where the concentration of nickel is held constant over the course of the test.  For the ion exchange process being studied, each resin bead will be exposed to more than one solution concentration during loading. Therefore, a series of tests to see how the model performs under varying nickel solution concentrations were conducted.  In this section, the hybrid correlation fit to the combined TP207MP/TP207XL dataset developed in Section 8.3 is being used for modeling.  9.2.1 DUAL CONCENTRATION TEST To see if the overall loading model is expected to perform under conditions reflective of an RIP circuit, a dual concentration batch test was conducted.  In this test, TP207XL resin (lot CHC 5034) was placed in a beaker held at constant nickel concentration of 50 ppm for 2.6 hours. The resin was then separated from the test solution using a screen and introduced into a second beaker held at a constant nickel concentration of 2500 ppm for the remainder of the test (total duration  21  hours).   The  objective  of  this  test  was  to  determine  whether  the  overall  loading model could account for the loading rate when resin from one tank is moved upstream to a higher concentration nickel tank, as would be expected in a counter-current RIP circuit. The first portion of the model uses either the lower of the two values calculated from the film diffusion model and the hybrid correlation for this resin at 50 ppm for the first 2.6 hours.  The final predicted loading from the hybrid correlation is 27.9 g Ni / L Resin.  To model the second portion of the curve, one must take into account that the time required to achieve a given resin loading at 2500 ppm is less than the time required to reach the same resin loading at 50 ppm.  In this case, it takes 2.59 hours to load the resin to 27.9 g Ni / L Resin at 50 ppm, but it would have only taken 0.53 hours to reach this same loading if the feed composition had been 2500 ppm.  This equivalent loading time was calculated using Equation 9.7, which is a slight variation on Equation 2.12, where the Vermeulen constant kv is replaced by the hybrid correlation constant, kh.  tr= -14kh ln{ͳ െ F2}          [9.7] 106  Therefore, the time used in the model to calculate the resin loading during the 2500 ppm portion of the curve has to be adjusted by subtracting a value of 2.06 hours.  Results from the dual concentration test are plotted in Figure 9.1.  FIGURE 9.1: DUAL CONCENTRATION TEST TP207XL Resin (Lot CHC5034), pH 4, 30qC  Figure 9.1 shows good correlation between the predicted and observed values.  In particular, there does not seem to be any discrepancy at the point when the resin is transferred from the beaker at 50 ppm to the beaker at 2500 ppm, suggesting that the model should be appropriate for modeling a counter-current RIP circuit.  0 10 20 30 40 50 60 70 0 4 8 12 16 20 N ic ke l L oa di ng  [g  N i /  L  r es in ] Time  [h] Part I: 50 ppm Part II: 2500 ppm Strip Value Model Prediction 2500 ppm50 ppm 107  9.2.2 TEST WITH INCREASING SOLUTION COMPOSITION A test was run with TP207MP (kinetic parameters used: Dfį = 2.7 x 10-5 m/s, Dapp = 4.3 x 10-12 m2/s, D = 0.30, dp = 5.52 x 10-4 m), where the solution nickel concentration was increased over the course of the test.  This was achieved by adding an overly-strong nickel stock solution, so that as the resin loaded, the concentration of nickel in the beaker increased.  The test started at 50 ppm nickel, and reached nearly 500 ppm nickel when the resin was fully loaded. To model the changing solution composition, the same approach was taken as with the dual concentration test.  To model the constantly changing solution environment, a finite difference approach was taken where resin loadings were calculated as if the resin was moved to a new solution every 30 seconds (i.e. 't = 30 s).  To ensure that this time step was sufficiently small, a single run with a time step of 10 seconds was also conducted and found to yield the same results.  Thus the mathematical solution to the loading rate problem becomes an initial value problem, based on the initial solution concentration (50 ppm), the initial resin loading (0 g Ni / L Resin), and the initial volume of solution present in the beaker (495 mL). At any given time t, the solution concentration [M], and resin loading [Mഥ ], are known from the previous time step calculation.  The model then calculates the equilibrium resin loading [Mഥ ]eqb, using the equilibrium isotherm presented in Section 7.1 (K = 9.8 x 10-5, capacity = 1.97 eq/L). The fractional attainment of equilibrium F, the constant for the film diffusion model kf, and the constant for the hybrid correlation kh,  are  calculated  and  used  to  determine  the  modified Helfferich  number.   The  equivalent  time  required  to  reach  the  current  resin  loading  tr is calculated using Equation 9.8 for modified Helfferich numbers > 1, or Equation 9.7 for modified Helfferich numbers d 1.  For film diffusion:  tr= -14kf ln{ͳ െ F}      [9.8]  For hybrid correlation: tr= -14kh ln{ͳ െ F2}      [9.7] The predicted resin loading at time t + 't, can now be calculated.  This loading should be the same as that which would have been achieved under infinite solution volume conditions after a time tr + 't.  The resin loading is calculated using either the film diffusion model (Equation 9.9; 108  a slight variation on Equation 2.8) for modified Helfferich numbers less than one, or the hybrid correlation (Equation 9.10; a slight variation on Equation 7.2) for modified Helfferich numbers greater than or equal to one.  For film diffusion: [Mഥ ]୲ାο୲= [Mഥ ]ୣ୯ୠ(ͳ െ exp{െk୤(t୰ ൅ οt)})   [9.9]  Hybrid correlation: [Mഥ ]୲ାο୲= [Mഥ ]ୣ୯ୠඥͳ െ exp {െ4k୦(t୰ ൅ οt)}  [9.10] In order to determine the predicted solution concentration at time t + 't, both a volume balance and a nickel mass balance are required.  The general balance equation can be described in words as: ACCUMULATION = INPUTS – OUTPUTS + GENERATION. For the volume balance, the inputs correspond to the amount of stock solution added.  Since the volume of resin used in the test is known (5 mL) and each mole of nickel that loads onto the resin should displace 2 moles of hydrogen ions (see Equation 2.1), the number of moles of sodium hydroxide required to neutralize this displaced acid can be calculated.  This value can then be converted into a volume of 0.25M sodium hydroxide solution that would be added.  For the particular set of pumpheads used in this particular test, the two pumpheads delivered solution to the test beaker at essentially the same rate, so the volume of nickel stock solution added was equal to the volume of sodium hydroxide solution added. The outputs in the volume balance correspond to water lost due to evaporation, and volume lost due to solution sampling.  To estimate the rate of evaporation, a test was run at 500 ppm nickel and no ion exchange resin was added to the beaker.  Samples were taken over the course of the test and assayed.  Based on the assay results, an approximate solution evaporation rate of 2.8 mL  /  hr  was  obtained.   In  terms  of  sampling,  a  5  mL  sample  of  the  feed  solution  was  taken before the start of the test (i.e. an initial volume of 495 mL), and subsequent samples taken were 4 mL in volume. It  was  assumed  that  the  rate  of  resin  swelling  was  negligible  in  this  test  and  that  the  rate  of volume generation from the loading reaction could be ignored.  This assumption is investigated in detail in Section 11.5.  Therefore, the expected solution volume in the beaker at time t + 't can be calculated as the volume at time t plus the increase due to stock addition less the volume lost to samples and evaporation. 109  For the nickel mass balance, the inputs correspond to the nickel added in the stock solution. The stock solution was made up to target 0.225 M, and this value was confirmed by AA assay. The nickel outputs correspond to the nickel lost in the solution samples.  The generation term in the nickel mass balance corresponds to nickel that is produced as a result of a reaction.  Since the nickel is loading from the solution onto the resin, this is a negative value and it corresponds to the amount of nickel that is predicted to load onto the resin between time t and t + 't.  Thus the  total  amount  of  nickel  present  in  the  beaker  at  time t  +  't can be taken as the amount of nickel present at time t plus the nickel added in stock less the nickel lost to samples less the nickel loaded onto the resin.  Once both the predicted volume and predicted total amount of nickel are calculated, the predicted solution concentration at time t + 't can be calculated. A comparison of the model predictions to the actual observed test data are plotted in Figure 9.2.  FIGURE 9.2: LOADING WHILE NICKEL CONCENTRATION INCREASES WITH TIME TP207MP Resin, pH 4, 30qC, 50 ppm Ni at t=0, 0.225M Ni stock solution  0 0.1 0.2 0.3 0.4 0.5 0.6 0 10 20 30 40 50 60 0 5 10 15 20 So lu ti on  C on ce nt ra ti on  [ g/ L N i] N ic ke l L oa de d on  R es in  [g  N i /  L  R es in ] Time [h] Observed Loading Predicted Loading Final Strip Assay Predicted Soln Assays Observed Soln Assays 110  The fit of the model to the observed data indicates that the model captures the rate of loading adequately.  Figure 9.3 plots the modified Helfferich number against time to illustrate how the rate limiting step changes over the course of the test.  For this test, the loading rate is initially film-diffusion limited, as expected.  The mixed control portion of the curve is around the region where the modified Helfferich number is | 1, which occurs around 8 minutes, and it appears that minimal error is introduced by not having a separate model to describe mixed control. Most of the test occurs under intraparticle diffusion control (modified Helfferich number << 1).  FIGURE 9.3: MODIFIED HELFFERICH NUMBER FOR INCREASING NICKEL CONCENTRATION TEST TP207MP Resin, pH 4, 30qC, 50 ppm Ni at t=0, 0.225M Ni stock solution   0.000 0.100 0.200 0.300 0.400 0.500 0.600 0.01 0.1 1 10 100 0 5 10 15 20 So lu ti on  C on ce nt ra ti on  [ g/ L N i] M od ifi ed  H el ff er ic h N um be r Time [h] Modified Helfferich Number Observed Solution Assays 111  9.2.3 TESTS WITH DECREASING SOLUTION COMPOSITION Two tests were run with TP207MP using sodium hydroxide to maintain the solution pH, but no nickel stock solution was added.  The first test was run with an initial solution concentration of 500 ppm nickel; the second test was run with an initial solution concentration of 250 ppm nickel.  The finite difference model presented in Section 9.2.2 was used to predict solution concentration and resin loadings over the course of the tests. Results for the first test, run at 500 ppm nickel are displayed in Figure 9.4.  FIGURE 9.4: LOADING FROM SOLUTION WITH INITIAL CONCENTRATION OF 500 PPM NICKEL TP207MP Resin, pH 4, 30qC, 500 ppm Ni at t=0, No nickel stock solution  The model was found to predict the observed values quite well.  Of note is that near the end of the test, the observed resin loading (based on the amount of sodium hydroxide added to the beaker) is significantly higher than the model prediction.  This is likely due to the fact that at such a low nickel concentration in the beaker (final assay 16.5 ppm), sodium co-loads onto the 0 0.1 0.2 0.3 0.4 0.5 0.6 0 10 20 30 40 50 60 0 5 10 15 20 So lu ti on  C on ce nt ra ti on  [ g/ L N i] N ic ke l L oa de d on  R es in  [g  N i /  L  R es in ] Time [h] Observed Loading Predicted Loading Final Strip Assay Predicted Soln Assays Observed Soln Assays 112  resin, as discussed in Section 2.1.3.  To confirm this hypothesis, the final resin loading value from the strip assay is also plotted on this graph.  Since the strip value lines up better with the model values than the values calculated from sodium hydroxide addition, it is reasonable to conclude that the discrepancy is due to sodium co-loading. To get a sense of rate-limiting step over the course of the test, the modified Helfferich number is  plotted  in  Figure  9.5,  along  with  the  solution  assays.   The  test  began  with  a  modified Helfferich number of 0.97, suggesting mixed control.  As the resin bead loads, the modified Helfferich number decreased to a value of 0.26.  However, since the solution concentration in the beaker was depleting over the course of the test, the modified Helfferich number increased back to a value of 0.5.  These values suggest that particle diffusion / exchange reaction was the dominant mechanism in this test.  FIGURE 9.5: MODIFIED HELFFERICH NUMBER –NO NICKEL STOCK SOLUTION, 500 PPM NI INITIAL TP207MP Resin, pH 4, 30qC, 500 ppm Ni at t=0, No nickel stock solution  0 0.1 0.2 0.3 0.4 0.5 0.6 0.01 0.1 1 10 100 0 5 10 15 20 So lu ti on  C on ce nt ra ti on  [ g/ L N i] M od ifi ed  H el ff er ic h N um be r Time [h] Modified Helfferich Number Observed Solution Assays 113  Results for the second test are displayed in Figure 9.6.  FIGURE 9.6: LOADING FROM SOLUTION WITH INITIAL CONCENTRATION OF 250 PPM NICKEL TP207MP Resin, pH 4, 30qC, 250 ppm Ni at t=0, No nickel stock solution  Results from the second test once again show reasonable correlation between model predictions and experimental observations.  It is interesting to note that the observed loading is once again significantly higher than the predicted value, or the final strip assay.  This discrepancy was larger in this test than in the previous test.  Since the final nickel assay in the previous test was 16.5 ppm nickel, and the final nickel concentration in this test was 0.5 ppm, it makes sense that the effect of sodium co-loading would be more prominent in this test.  0 0.1 0.2 0.3 0.4 0.5 0.6 0 10 20 30 40 50 60 0 5 10 15 20 So lu ti on  C on ce nt ra ti on  [ g/ L N i] N ic ke l L oa de d on  R es in  [g  N i /  L  R es in ] Time [h] Observed Loading Predicted Loading Final Strip Assay Predicted Soln Assays Observed Soln Assays 114  The plot of the modified Helfferich number is plotted with the solution concentrations for the first hour of the test in Figure 9.7.  FIGURE 9.7: MODIFIED HELFFERICH NUMBER –NO NICKEL STOCK SOLUTION, 250 PPM NI INITIAL TP207MP Resin, pH 4, 30qC, 250 ppm Ni at t=0, No nickel stock solution  This plot shows that the test begins in the film diffusion limited regime, with a modified Helfferich number of 2.9.  However, the modified Helfferich number drops to a value of 0.8, before increasing back to a value of 3.8.  Most of the loading occurs in the time period where the modified Helfferich number is close to one, suggesting that mixed control is predominant in this  test.   Although  the  assay  taken  at  0.5  hours  is  slightly  higher  than  that  predicted  by  the model, the overall fit of the model to the test data appears to be adequate.   0 0.1 0.2 0.3 0.4 0.5 0.6 0.01 0.1 1 10 100 0.0 0.5 1.0 1.5 2.0 So lu ti on  C on ce nt ra ti on  [ g/ L N i] M od ifi ed  H el ff er ic h N um be r Time [h] Modified Helfferich Number Predicted Solution Assay Observed Solution Assays 115  9.2.4  SUMMARY OF BATCH TEST RESULTS AND ASSESSMENT OF MIXED CONTROL REGIME The results from the batch tests carried out with changing nickel concentration indicate that the overall loading rate model developed in Chapters 6 and 7 should be adequate for modeling a counter-current RIP circuit.  In particular, the test run with an initial solution concentration of 250 ppm without any stock solution being added suggests that the overall loading rate model is adequate, even when most of the test is carried out under mixed control conditions.  Therefore, no further work was done to develop a more rigorous model for mixed control, and the research focus shifted to model verification with a continuous counter-current RIP miniplant.  116  10.0  DEVELOPMENT OF RIP CIRCUIT MODELS In order to verify the hybrid correlation with an RIP miniplant, a five stage countercurrent circuit model was developed.  This chapter provides a step-by-step description of the methodology and assumptions used in the development of the five stage cascade circuit model and the five stage carousel circuit model.  Recall that the differences in cascade and carousel operation were discussed in Section 2.2.1.  10.1  CASCADE CIRCUIT MODEL The cascade circuit model developed calculates the steady-state solution profile and the steady- state resin profile across the circuit.  In order to achieve this, the following inputs are required: x Working volume of each contactor x Flowrate and metal concentration of feed stream x Resin volume in each contactor x Resin transfer rate between contactors (constant for circuit) x Factor for estimating solution transferred with resin x Initial metal loading on resin entering circuit x Metal concentration in solution coming into circuit with resin x Resin isotherm parameters: K, capacity x Resin kinetic parameters: dp, Df Ɂ  , D, Dapp The circuit model is calculated by providing an initial guess for the concentration of metal leaving the circuit in the tails stream.  Assuming perfect mixing in the contactor, the concentration of metal in the contactor is the same as the concentration of the stream exiting the contactor.  The average resin loading in the tank, ۃFۄ, is calculated using the Segregated Flow model, and then the concentration of nickel in the stream feeding the tank can be obtained by mass balance.  This procedure is repeated until the overall concentration of metal in the feed is obtained.  This value is compared to the input value, and the guess for the tails value changed until the calculated feed value matches the model’s input feed value.  With Microsoft Excel, this can be automated using the program’s Solver routine. 117  The fractional resin loading function, F(t), for each tank was based on either the film diffusion model of Chapter 6 or the hybrid correlation of Chapter 7, depending on which calculation gave a lower value for ۃFۄ.  Since the concentration of metal in the tank is known, the equilibrium loading value for the resin,  [Mഥ ]eqb , can be calculated from the mass action isotherm. Furthermore, since the resin enters each tank partially loaded, the time it would have taken the resin to achieve this loading, tr, if it had entered the tank fully stripped is required.  This is calculated for the film diffusion model using Equation 10.1, and for the hybrid correlation using Equation 10.2.  Film Diffusion: tr  = -1kf ln ൬1 - [Mഥ ][Mഥ ]eqb൰     [10.1]  Hybrid Correlation: tr  = -14kh ln ൬[Mഥ ]eqb2 -[Mഥ ]2[Mഥ ]eqb2 ൰     [10.2]  The Segregated Flow Model was selected for calculating ۃFۄ, since it is reasonable to assume that there is no significant interaction between resin beads that have been in the RIP contactor for different lengths of time.  It was also assumed that, from the resin bead’s point of view, each contactor  behaved  as  an  ideal  stirred  tank  reactor  with  an  average  resin  residence  time  of  Wr. This can be summarized by Equations 10.3 – 10.4 (120).  Note that the integrals for average resin loading were solved numerically using the trapezoidal rule.  ۃFۄ= ׬ F(t+tr)E(t)dtλ0         [10.3]  E(t)= 1 ɒr exp( -– ɒr)ൗ         [10.4] A simple mass balance is used to solve for the metal concentration of the solution entering each contactor.  Figure 10.1 provides a block diagram of a single stage in the circuit.  118         FIGURE 10.1: BLOCK DIAGRAM FOR A SINGLE STAGE IN A CASCADE CIRCUIT  The first step in the contactor balance is to calculate the volumetric flowrates of the solution in and out of the contactor.  The volumetric flowrates of the resin transfer and the solution transferred  with  the  resin  are  known from the  model  inputs.   The  flowrate  of  the  feed  to  the circuit is also known, and so the flowrate of the solution exiting each contactor can be calculated by a simple flow balance.  The implicit assumption in this calculation is that there is no change in resin volume as the beads load. Once the volumetric flowrates are all known, the mass balance across the contactor can be completed.  At this point in the calculation, the concentration of metal on the resin entering and exiting  the  contactor  is  known,  as  well  as  the  concentration  of  metal  in  solution  exiting  the contactor.  Assuming perfect mixing, the metal concentration within the contactor is equal to that exiting the contactor.  Therefore, it is a matter of simple algebra to calculate the concentration of metal in the solution feeding the contactor. The output for an example run with this model is displayed in Figures 10.2 and 10.3.  All input cells are in grey, all other values are calculated.  Note that the model also includes a Mix Tank (contains no resin), whose purpose is to combine the recycled solution from Stage 1 with the fresh feed.  Figure 10.2 displays the inputs and outputs of the model; Figure 10.3 displays the actual flow and mass balances for each individual stream.   Solution/Slurry from Upstream Contactor   Solution/Slurry to Downstream Contactor Resin with Solution/Slurry from Downstream Contactor Resin with Solution/Slurry to Upstream Contactor  RIP Contactor  119  FIGURE 10.2: CASCADE CIRCUIT MODEL INPUTS & OUTPUTS Solution Side Resin Side TP207XL Kinetics Df/delta 2.73E-05 m/s K 9.78E-05 alpha 0.36 Tot Vol 600 mL per tank Cap 2.36 Dapp 4.43E-12 m2/s Flow Rate 20 mL / min FRESH FEED Resin Rate 2 mL/min bead dp 736 um molar Ni 58.71 g/mol 0 gpL Ni in resin feed solution 0.000736 m Initial Load 3 g Ni / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 250 50 50 50 50 50 120 % Resin 8.3% 8.3% 8.3% 8.3% 8.3% 20.0% Airlift total flow to upstream tank [mL/min]: 36.0 36.0 36.0 36.0 36.0 15.0 Segregated Flow Model, OUTPUTS: input circuit feed: 4.000 model circuit feed: 4.000 0.000 Soln (gpL) 4.000 2.733 2.030 1.286 0.611 0.137 0.012 Mix Tk 1 2 3 4 5 Resin [g Ni / L R] 42.8 35.8 27.1 17.0 6.3 3.0 Mod He # 0.03 0.05 0.10 0.39 3.74 Eqbm Load [g Ni / L R] 67.57 67.13 66.19 62.92 49.96 quadratic "a" 1.35E-05 8.57E-06 4.07E-06 9.14E-07 7.86E-08 quadratic "b" -3.2E-05 -2E-05 -9.6E-06 -2.2E-06 -2E-07 quadratic "c" 1.88E-05 1.19E-05 5.67E-06 1.27E-06 1.09E-07 Film Diffusion Model kf 0.006521 0.004131 0.001962 0.000441 3.79E-05 tr [s] 116          125          151          240          1,634 Hybrid Correlation kh 2.7E-05 2.29E-05 1.75E-05 1.02E-05 4.23E-06 tr [s] 3,047       1,951       971          247          214  120  FIGURE 10.3: FLOW AND MASS BALANCE FOR CASCADE CIRCUIT MODEL Volumetric Flows: mL/min Ni mass flow: gNi/min Assays STREAM total flow soln flow resin flow total soln resin g /L Ni g Ni / L R feed 20.00 20.00 0.00 80.00 80.00 4.00 mix o/f 56.00 56.00 0.00 153.07 153.07 2.73 1 o/f 54.00 54.00 0.00 109.61 109.61 2.03 2 o/f 54.00 54.00 0.00 69.44 69.44 1.29 3 o/f 54.00 54.00 0.00 32.98 32.98 0.61 4 o/f 54.00 54.00 0.00 7.41 7.41 0.14 5 o/f (tails) 33.00 33.00 0.00 0.39 0.39 0.01 soln rec'c 36.00 36.00 0.00 73.08 73.08 2.03 resin exit 2.00 0.00 2.00 85.61 85.61 42.80 airlift to 1 36.00 34.00 2.00 115.22 43.72 71.50 1.29 35.75 airlift to 2 36.00 34.00 2.00 75.05 20.77 54.28 0.61 27.14 airlift to 3 36.00 34.00 2.00 38.59 4.67 33.93 0.14 16.96 airlift to 4 36.00 34.00 2.00 13.02 0.40 12.62 0.01 6.31 airlift to 5 15.00 13.00 2.00 6.00 0.00 6.00 0.00 3.00 Balance by Stage: IN OUT IN OUT Mix Tank 56.00 56.00 153.07 153.07 1 92.00 92.00 268.30 268.30 2 90.00 90.00 184.66 184.66 3 90.00 90.00 108.03 108.03 4 90.00 90.00 46.00 46.00 5 69.00 69.00 13.41 13.41 OVERALL 35.00 35.00 86.00 86.00 Volumetric Flow Nickel Mass Flow CIRCUIT RECOVERY 99.51% [mL/min] [g Ni / min] 121  10.2 CAROUSEL CIRCUIT MODEL The computer model of a five stage carousel circuit was programmed such that feed is directed to the lead contactor, which overflows to the high centre contactor (“HC”), the centre contactor (“C”), the low centre contactor (“LC”), the lag contactor, and finally overflows to tails. The carousel circuit model consists of a pair of coupled ordinary differential equations.  The first differential equation is the mole balance for the metal in the solution phase and the second differential equation is the mole balance for the metal in the resin phase.  In words, the mole balances can be written as: INPUTS – OUTPUTS + GENERATION = ACCUMULATION. For a single contactor, assuming perfect mixing and minimal change in resin and solution volumes, the system balance for the solution phase is expressed by Equation 10.5, where Q is the solution flowrate through the contactor, VS is the volume of solution in the contactor, and VR is the volume of resin in the contactor.  Note that the validity of the constant volume assumption will be explored in Section 11.5.  Solution Phase  Q[M]feed-Q[M]-VR ddt [Mഥ ]= VS ddt [M]   [10.5] The mole balance for the resin phase is trivial, since no resin is being added or removed from the contactor.  Therefore, the rate of loading is described by Equation 2.7 when the resin is loading under film diffusion control (modified Helfferich number greater than one), and a slight variation on Equation 2.11 (kv replaced here with kh) when loading under intraparticle diffusion / exchange reaction limited control (modified Helfferich number less than or equal to one).  Mod He > 1: ddt [Mഥ ]=kf൫[Mഥ ]eqb-[Mഥ ]൯        [2.7] Mod He d 1: ddt [Mഥ ]=2kh ൬[Mഥ ]eqb2 -[Mഥ ]2[Mഥ ] ൰      [2.11] To solve the set of coupled ordinary differential equations, a finite difference approach was taken.  Euler’s method (1st order Runge-Kutta) was selected for ease of application, and is described by Equation 10.6.  Euler’s Method: [M]t+οt=[M]t൅οt ddt [M]     [10.6] 122  The initial concentrations of the metal in both the resin and solution phases for each contactor must be input into the model.  The model uses these initial conditions for t = 0, and works through a time-step sequence, calculating all subsequent resin and solution values at time t + 't, based on the previous values.  The time step, 't, used in this model was 0.24 minutes.  To confirm that this time-step was sufficiently small, a single run of the model was made with a time step of 0.06 minutes and was found to yield the same steady-state results.  Note that for the carousel circuit, the concentrations in each contactor are always a function of time and steady- state occurs when the same solution and resin profile repeat with each operating cycle. The following inputs are required for this circuit model: x Working volume of each contactor x Flowrate and metal concentration of feed stream x Resin volume in each contactor x Carousel cycle time (how often contactor positions change) x Initial solution concentration in each contactor at time t = 0 x Initial metal loading on resin in each contactor at time t = 0 x Metal loading on resin entering the circuit from elution x Resin isotherm parameters: K, capacity x Resin kinetic parameters: dp, Df Ɂ  , D, Dapp The outputs of the circuit model are the solution concentration and metal concentration within each contactor as a function of time.  At the end of each cycle (i.e. when the lead contactor is taken offline and the role of the contactors changes), the model assumes an instantaneous switch of contactor position, and then continues. At any given time t, the current concentration of metal in the contactor solution phase [M], and the current resin loading [Mഥ ]t  are known.  The current solution concentration is used to calculate the equilibrium resin loading value.  From this information, the fractional attainment of equilibrium, the modified Helfferich number, and the time it would have taken the resin to reach the current loading tr, can be calculated. 123  To obtain the resin loading for the next time step [Mഥ ]t+οt , the analytical solution to the film diffusion model (modified He >1) or the analytical solution to the hybrid correlation (modified He d 1), is calculated using t = tr + 't , as per Equation 9.9 and 9.10.  For film diffusion: [Mഥ ]୲ାο୲= [Mഥ ]ୣ୯ୠ(ͳ െ exp{െk୤(t୰ ൅ οt)})   [9.9]  Hybrid correlation: [Mഥ ]୲ାο୲= [Mഥ ]ୣ୯ୠඥͳ െ exp {െ4k୦(t୰ ൅ οt)}  [9.10] To obtain the solution concentration at the next time step [M]t+οt, one must first estimate the rate of resin loading in the “generation” term.  This is achieved with Equation 10.7. ddt [Mഥ ] =  [Mഥ ]t+οt െ [Mഥ ]t         [10.7] Rearranging Equation 10.5, substituting in Equation 10.7, and applying Euler’s method, the concentration of solution in the contactor at time t + 't can be expressed using Equation 10.8.  [M]t+οt=[M]t+ ൛Q οt ൫[M]feed-[M]t൯-VR൫[Mഥ ]t+οt-[Mഥ ]t൯ൟVS     [10.8] For  modeling  the  full  set  of  five  reactors,  it  is  important  to  realize  that  the  concentration  of metal entering each subsequent reactor, [M]feed,  is  the  concentration  of  metal  exiting  the previous reactor.  Thus, the finite differences model was solved for all reactors using the same time step sequence. The output for a single run of the carousel model would be over one thousand pages long. Appendix F displays the output for the first 0.072 hours for the lead, high centre, and centre contactors.   124  11.0  VERIFICATION OF CIRCUIT MODELS (RIP MINIPLANT) A five stage RIP miniplant (working volume: 560 mL / stage) was constructed for the purpose of circuit model verification for both nickel and copper loading onto Lewatit MonoPlus TP207XL, lot CHE 50019.  Five litres of resin were converted to the hydrogen form and wet- screened to the +710 -850 size fraction. The miniplant was designed to operate in a continuous cascade mode, although it was also used in a carousel fashion. During cascade operation, fresh feed was pumped to the Feed Mix Tank and then was fed by gravity through the five RIP loading stages.  Resin was fed continuously from the Resin Feed Tank to the final RIP stage using an airlift pump.  In order to maintain the resin inventory in the Resin Feed Tank, fresh resin was manually added every 15 minutes.  The resin was moved countercurrently through the circuit using airlift pumps.  Note that some solution was also transferred with the airlift pumps, and no attempt was made to separate the solution from the resin feeding the five RIP contactors.  However, resin exiting the first RIP stage was passed over  a  screen  to  remove  the  excess  solution  prior  to  the  resin  being  transferred  to  the  elution columns.  The excess solution was pumped to the Feed Mix Tank, where it was combined with the fresh feed prior to being returned to the first RIP contactor.  The pH in each contactor was held constant at 4.0 +/- 0.1 pH units through the addition of 1M sodium hydroxide, and temperature controllers held each stage at 30 +/ -1 qC. Elution was done in columns at approximately 2 BV/h.  Stripping consisted of 10 BV of 100 g/L sulphuric acid, followed by 5 BV water wash.  At the end of elution, the resin retained a greenish hue, suggesting incomplete stripping of nickel.  For the miniplant, the decision was made to run with a recycle load of nickel on the resin to avoid an acid hold step.  Samples of resin returning to the circuit were taken and stripped with a full 20 hour acid hold step to determine the level of recycle load.  Results indicated that the resin was returning to the circuit with a loading of approximately 3 g Ni / L Resin, and there was no indication of the amount of nickel left on the resin increasing after five cycles.  At the end of the run, the total resin volume in each contactor was recorded and a 5 mL subsample of the resin was taken and stripped to determine the resin loading in each contactor. A flowsheet of the cascade circuit can be found in Figure 11.1; photographs of the circuit can be found in Figure 11.2, and detailed operating procedures can be found in Appendix G.  125               FIGURE 11.1: RIP MINIPLANT FLOWSHEET (CASCADE OPERATION) Mix Tank 1 2 3 4 5 X X X X X X Resin Tank TC AC AIR COMPRESSOR   Feed  Recycle Effluent Water TC AC TC AC TC AC TC AC TC AC  AC Resin Screen  Elution Column (Batch Step) 100 gpLH2SO4 Water 1 0 g/  H2SO4 Eluate TC = Temperature Control (30qC) AC = pH Control (4.0 with 1M NaOH)   X  = Solution Sample Point   Air Flow Solution Flow   Continuous Resin Flow   Periodic Manual Resin Transfer 126    FIGURE 11.2: PHOTOGRAPHS OF RIP MINIPLANT (CASCADE OPERATION) 127  11.1  NICKEL CASCADE RUNS Three continuous cascade runs were performed with nickel.  In the first run (Run A), fresh resin was used, but in all other runs recycled resin (~ 3 g Ni / L resin) was used.  Maintaining constant resin inventories in all the tanks with the airlifts was found to be difficult.  Since the objective of the exercise was model verification, the decision was made that it was more important to reach a steady-state condition rather than to try and hold the total resin inventories constant.   Therefore,  no  adjustments  were  made  to  resin  addition  rates  or  the  airflow  pumps after the first two hours of the test.  Note that the effect of solution back-mixing with the airlifts will be investigated in Section 12.1, and the effect of operating the circuit with a recycle load of nickel on the resin will be investigated in Section 12.6.4. Inputs to the cascade circuit model developed in Chapter 10 include the resin equilibrium and kinetic parameters for TP207XL, lot CHE50019 reported in Chapter 8 and the film diffusion model parameters reported in Chapter 6.  Additional inputs to the model included: feed flowrate (15 mL/min), resin flowrate (1.5 mL/min), initial resin loading (0 or 3 g Ni/L resin), final measured resin volume in each tank, total airlift flowrates, and the measured nickel concentration exiting the Mix Tank.  Note that the assay concentration of solution exiting the Mix Tank was used in lieu of the concentration of the feed solution (5 g/L) since some solution was removed from the circuit with the resin exiting the circuit, and the exact amount was not measured.  For the total airlift flowrate, the total volume was not recorded for all tests and an estimate was required for the first two tests.  Details on how this estimate was calculated can be found in Appendix H. The model inputs for each run are tabulated in Table 11.1, and the model results are compared to the observed values in Figure 11.3.  Note that the two model lines in Figure 11.3 correspond to the model being run once using the highest observed mix tank assay, and again using the lowest observed mix tank assay.  The three charts on the left side of Figure 11.3 correspond to the steady-state solution profile across the circuit, while the figures on the right side of Figure 11.3 display the steady-state resin loading profile across the circuit.  Appendix I contains the detailed model results.  128   TABLE 11.1: SUMMARY OF MODEL INPUTS AND SAMPLING TIMES FOR NICKEL CASCADE TESTS TEST Nickel A Nickel B Nickel C Feed Flowrate [mL/min] 15 15 15 Mix Tank Assay [g/L Ni] 3.5 / 3.7 3.5 / 3.4 / 3.2 4.3 / 4.0 / 3.8 Resin Flowrate [mL/min] 1.5 1.5 1.5 Resin Loading Entering Circuit [g Ni / L Resin] 0 3 3 Final Resin Volumes in Each Tank [mL]  (volume %) Tank 1 40½    (7.2%) 63    (11.2%) 92    (16.3%) Tank 2 58    (10.3%) 63    (11.2%) 83    (14.7%) Tank 3 50      (8.9%) 75    (13.3%) 49      (8.7%) Tank 4 18½   (3.3%) 64    (11.4%) 38      (6.7%) Tank 5 61    (10.8%) 57    (10.1%) 14       (2.5%) Resin Feed Tank 120*  (21.3%) 120*  (21.3%) 120*  (21.3%) Airlift Flowrate Exiting Contactor [mL] Tank 1 31* 20* 12 Tank 2 22* 20* 12 Tank 3 25* 17* 20 Tank 4 69* 20* 36 Tank 5 21* 22* 52 Resin Feed Tank 11* 11* 25 Time Assays Taken [h] 14 / 16 14 / 15 / 16 14 / 15 / 16 Resin Sample Taken [h] 16 16 16  * indicates estimated values Resin Isotherm Parameters:  K = 9.8x10-5, capacity = 2.36 eq/L Resin Kinetic Parameters for Ni:  Dfį = 2.7x10-5m/s, Dapp = 4.4x10-12 m2/s, D = 0.36, dp = 7.36x10-4m   129                       FIGURE 11.3: NICKEL CASCADE MINIPLANT RESULTS –MODEL VS ACTUAL TP207XL Resin (Lot CHE 50019), 30qC, pH 4 0 10 20 30 40 50 60 0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 1 2 3 4 5 0 10 20 30 40 50 60 0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 1 2 3 4 5 0 10 20 30 40 50 60 0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0 1 2 3 4 5 RUN A SOLUTION RUN A RESIN RUN B SOLUTION RUN B RESIN RUN C SOLUTION RUN C RESIN SOLUTION PROFILE ACROSS MINIPLANT  [g/L Ni] RESIN LOADING ACROSS MINIPLANT  [g Ni / L Resin] ASSAYS:        Æ 14 h    X Æ 15 h       Æ 16 h RESIN STRIP ASSAYS:   Lines are model predictions based on the highest and lowest observed Mix Tank Solution Assay  Mix Tank Mix Tank Mix Tank     130  While there is some deviation in the model predictions from observed values, the overall model appears to be adequate, given the amount of uncertainty that is associated with miniplant scale testwork.  In particular, no section of the circuit appears to consistently be overpredicted or underpredicted, suggesting that the deviations do not highlight a major model deficiency. Therefore, the Segregated Flow Model selected in Chapter 10 for modeling resin movement through the circuit appears to be adequate for this application. In  terms  of  the  modified  Helfferich  numbers,  most  of  the  circuit  is  observed  to  be predominantly rate-limited by intraparticle diffusion / exchange reaction, rather than by film diffusion.  The only tank predicted to be under film diffusion control was the final tank in the second nickel run.  The calculated modified Helfferich numbers, based on the observed solution assays and observed resin loadings are displayed in Table 11.2. TABLE 11.2: MODIFIED HELFFERICH NUMBERS FOR NICKEL CASCADE TESTS TEST Nickel A Nickel B Nickel C Tank 1 0.025-0.026 0.027-0.029 0.026-0.027 Tank 2 0.035-0.037 0.041-0.043 0.046-0.048 Tank 3 0.055-0.057 0.131-0.138 0.10-0.11 Tank 4 0.094 0.57-0.94 0.28-0.29 Tank 5 0.22-0.23 9.6-11 0.76-0.84   11.2  COPPER CASCADE RUN A cascade run was conducted for loading copper onto TP207XL using a solution feed flowrate of 18.5 mL /min.  This test was run for 11 hours, and solution samples were taken after 8, 9, 10, and 11 hours.  The recirculating load of copper on the resin appears to have been negligible (resin samples only showed differences of +/- 0.2 g Cu / L Resin when assayed with and without an acid hold step).  The model inputs for this test are summarized in Table 11.3, and the observed values are compared to the model predictions in Figure 11.4.  As was the case for the nickel cascade run, two model lines are generated, correspond to the highest and lowest observed mixed tank assay being used as a model input parameter. 131  TABLE 11.3: SUMMARY OF MODEL INPUTS AND SAMPLING TIMES FOR COPPER CASCADE TEST Feed Flowrate [mL/min] 18.5 Resin Flowrate [mL/min] 1.5 Observed Mix Tank Assays [g/L Cu] 3.9 / 3.7 / 3.8 / 3.9 Resin Loading Entering Circuit [g Cu/L Resin] 0 Solution Assays at [h] 8 / 9 / 10 / 11 Resin Assays at [h] 11 Airlift Flowrates [mL/min] Final Resin Volumes in Tanks [mL] (volume %) Tank 1 28½  Tank 1 43       (7.6%) Tank 2 23 Tank 2 73½  (13.1%) Tank 3 14 Tank 3 39       (6.9%) Tank 4 21 Tank 4 63     (11.2%) Tank 5 21 Tank 5 54       (9.6%) Resin Feed Tank 40 Resin Feed Tank 120*  (21.3%) * indicates estimated values Resin Isotherm Parameters:  K = 7.5x10-4, capacity = 2.5 eq/L Resin Kinetic Parameters for Cu:  Dfį = 2.7x10-5m/s, Dapp = 2.8x10-11 m2/s, D = 0.62, dp = 7.36x10-4m          FIGURE 11.4: COPPER CASCADE MINIPLANT RESULTS –MODEL VS ACTUAL TP207XL Resin (Lot CHE 50019), 30qC, pH 4   0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 0 1 2 3 4 5 Assay 8 hours Assay 9 hours Assay 10 hours Assay 11 hours Model -min assay Model -max assay 0 10 20 30 40 50 60 70 0 1 2 3 4 5 Resin Strip Assay Model -Min Assay Model -Max Assay SOLUTION PROFILE ACROSS MINIPLANT  [g/L Cu] RESIN LOADING ACROSS MINIPLANT  [g Cu / L Resin] 132  In terms of the modified Helfferich numbers, the first three tanks are expected to be intraparticle diffusion limited / exchange reaction limited, and the final tank is expected to be film diffusion limited.  The calculated modified Helfferich numbers, based on the observed solution and resin assays are tabulated in Table 11.4. TABLE 11.4: MODIFIED HELFFERICH NUMBERS FOR COPPER CASCADE TEST Tank 1 Tank 2 Tank 3 Tank 4 Tank 5 0.069-0.071 0.10-0.11 0.22 0.66-0.78 9.1-10.3  It is interesting to note that the tank with a modified Helfferich number closest to one (i.e. the tank most likely to be affected by mixed control) is the fourth tank in this test, and the model underpredicts the resin loading in this tank.  This is in direct contrast to the results from the nickel cascade tests, where the fourth tank in Run B (modified Helfferich value of 0.57-0.94) overpredicts the resin loading.  Furthermore, the observed resin loading is within the model predictions for the fifth tank in Run C (modified Helfferich value of 0.76-0.84).  Since the model overpredicts in one case, underpredicts in another, and predicts the observed value in the third case, it is unlikely that this error is due to the fact that there is no separate model for mixed control.  11.3  NICKEL CAROUSEL RUN The miniplant configuration was slightly modified to run as a five stage carousel circuit, and a 36 hour nickel carousel test was conducted.  In carousel mode, feed was pumped to the lead contactor, and then overflowed to the high centre contactor (“HC”), the centre contactor (“C”), the low centre contactor (“LC”), the lag contactor, and finally overflowed to the tails bucket.  In these tests, the role of the contactors was changed each hour.  To change the role of the contactors, the physical position of the feed line was moved from contactor to contactor.  When necessary, a pump was used after the lowest contactor to direct the overflow from this contactor back to the highest  contactor.   The time required to take the lead contactor off-line and put a new contactor in the lag position could be achieved in less than three minutes.  Therefore, feed was pumped to the circuit for 57 minutes, the roles of the contactors was adjusted, and then the 133  feed was turned back on at 60 minutes, in order to create a constant one hour cycle time. Detailed operating procedures for the carousel runs can be found in Appendix J. To determine the feed flow rate, the total tails volume collected each hour was measured.  It was noticed that the flow rate fluctuated considerably over the course of this test and varied from 13 – 22 mL / min.  Consequently, the test never reached steady-state. Solution samples were taken for assay at 34.75, 35.25, 35.5, 35.75, and 36 hours.  The physical circuit configuration chosen for sampling is shown in Figure 11.5.  This physical configuration was typically the point in time at which the feed flow rate had been most steady, based on the tails volumes collected each cycle.  FIGURE 11.5: PHYSICAL CIRCUIT CONFIGURATION DURING SAMPLING -CAROUSEL  Offline LEAD LAG LC C HC FEED 134  The carousel finite difference model developed in Chapter 10 was run using the resin equilibrium and kinetic parameters listed for TP207XL, lot CHE50019 reported in Chapter 8 and the film diffusion model parameters reported in Chapter 6.  Additional inputs to the model included: the measured hourly feed flowrate (13-22 mL/min), resin volume (40 mL/stage), initial resin loading (3 g Ni/L resin), contactor working volume (560 mL), and the average nickel feed assay (4.14 g/L Ni).  Note that in this model, it is assumed that the circuit operates for 60 minutes in each position and that the contactor change is instantaneous.  These model input values are summarized in Table 11.5.  Note that the sample calculations provided in Appendix F were based on this experimental run. TABLE 11.5: SUMMARY OF MODEL INPUTS AND SAMPLE TIMES FOR NICKEL CAROUSEL TEST Contactor Volume [L] 0.56 Resin per Stage [mL],(%) 0.40 (7.1%) Feed Rate [mL/min] 13-22 Contactor Cycle Time [min] 60 Feed Assay [g/L Ni] 4.14 Eluted Resin [g Ni/L Resin] 3 Solution Samples at [h] 34.75,  35.25,  35.5, 35.75 , 36 Resin Samples at [h] 36 Resin Isotherm Parameters: K = 9.8x10-5, capacity = 2.36 eq/L Resin Kinetic Parameters for Ni: Dfį = 2.7x10-5m/s, Dapp = 4.4x10-12 m2/s, D = 0.36, dp = 7.36x10-4m  Results comparing the circuit model with the actual solution assays indicate that the circuit model is significantly over-predicting circuit performance.  However, the results are still within test error bounds, as the assays are still greater than the predicted values at the lowest feed rate (13 mL/min).  Figure 11.6 displays the circuit model, the circuit model lower error bound, and the actual solution assays.   Note that this graph plots the solution assays for the various tanks against time for the last six hours, since during carousel operation the solution composition in each contactor is continually changing.  135   FIGURE 11.6: NICKEL CAROUSEL MINIPLANT RESULTS -SOLUTION TP207XL Resin (Lot CHE 50019), 30qC, pH 4  The resin in each contactor was sampled and stripped at the end of the run.  These final strip assays are compared to the model values in Figure 11.7.  Once again the model is over- predicting resin loadings, although the final strip assays are again within error bounds.  0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 30 31 32 33 34 35 36 N ic ke l i n So lu tio n [g /L ] Time in Test [h] lead (lower bound) HC (lower bound) C (lower bound) LEAD HC C Lead Assays HC Assays C Assays 136   FIGURE 11.7: NICKEL CAROUSEL MINIPLANT RESULTS –RESIN TP207XL Resin (Lot CHE 50019), 30qC, pH 4  11.4  COPPER CAROUSEL RUN A 30 hour copper carousel test was conducted, using 40 mL fresh resin per stage (no initial loading).  As in the case of nickel, feed was directed to the circuit for 57 minutes.  After shutting off the feed pump, the lead contactor was taken off-line and a new contactor was placed in the lag position.  The circuit was always restarted at the end of three minutes, so that the total cycle time was one hour.  More care was taken as to the placement of the various lines during this test and the feed flow rate was kept between 14 and 17 mL/minute.  The average feed assay for this run was 4.4 g/L copper. Solution samples were taken from each contactor at 24.75, 29.25, 29.5, 29.75, and 30 hours. The  physical  circuit  configuration  at  the  time of  the  sampling  was  the  same as  that  shown in Figure 11.6.  This means that there was a plug flow component (not included in the model) 0 10 20 30 40 50 60 30 31 32 33 34 35 36 N ic ke l L oa de d [g  N i /  L  R es in ] Time in Test [h] lead (lower bound) hc (lower bound) c (lower bound) lc (lower bound) LEAD HC C LC LAG Strip Assays 137  between  the  lead  and  the  high  centre  contactors.   At  the  end  of  the  run,  resin  samples  were taken from each contactor and stripped.  Table 11.6 summarizes the model inputs for this test. Results comparing the solution assays to the model predictions can be found in Figure 11.8. TABLE 11.6: SUMMARY OF MODEL INPUTS AND SAMPLE TIMES FOR COPPER CAROUSEL TEST Contactor Volume [L] 0.56 Resin per Stage [mL],(%) 0.40 (7.1%) Feed Rate [mL/min] 14-17 Contactor Cycle Time [min] 60 Feed Assay [g/L Cu] 4.4 Eluted Resin [g Cu/L Resin] 0 Solution Samples at [h] 24¾,  29¼,  29½, 29¾ , 30 Resin Samples at [h] 30 Resin Isotherm Parameters: K = 7.5x10-4, capacity = 2.5 eq/L Resin Kinetic Parameters for Cu: Dfį = 2.7x10-5m/s, Dapp = 2.8x10-11 m2/s, D = 0.62, dp = 7.36x10-4m   FIGURE 11.8: COPPER CAROUSEL MINIPLANT RESULTS -SOLUTION TP207XL Resin (Lot CHE 50019), 30qC, pH 4 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 24 25 26 27 28 29 30 Co pp er  in  S ol ut io n [g /L ] Time in Test [h] LEAD HC C Lead Assays HC Assays C Assays 138  Figure 11.8 shows a good fit between the observed and predicted values in the lead contactor. In the high centre contactor, the observed values are slightly below the predicted value, which makes sense since the solution from the lead contactor had to be pumped all the way back to the high centre contactor, creating a plug-flow component that the model does not account for. Results comparing the final resin strip assays with predicted model loadings are presented in Figure 11.9.  FIGURE 11.9: COPPER CAROUSEL MINIPLANT RESULTS –RESIN TP207XL Resin (Lot CHE 50019), 30qC, pH 4  Results from the final resin loadings show fairly good correlation with the predicted model values.  Consequently, the results from Figures 11.8 and 11.9 suggest that the overall circuit model adequately describes copper loading under carousel operating conditions.  0 10 20 30 40 50 60 70 80 24 25 26 27 28 29 30 Co pp er  L oa de d [g  C u / L Re si n] Time in Test [h] LEAD HC C LC LAG Strip Assays 139  11.5  VALIDITY OF CONSTANT VOLUME ASSUMPTION In the development of both circuit models, the implicit assumption was made that the volumes of  the  resin  and  solution  do  not  change  with  time.   However,  it  is  well  known that  there  is  a significant change in resin volume for iminodiacetic resins when converted from the sodium form to the hydrogen form, as was seen in Figure 4.1 in the comparison of resins.  Therefore, the validity of this assumption to the RIP circuit studied must be scrutinized. In terms of the rate of metal loading onto the resin, the change in resin volume is irrelevant. This is because both the film diffusion model and the hybrid correlation were defined with respect to the original bead diameter, in the hydrogen form.  Thus, as the resin swells, the only error that is incurred in the model is the change in the solution volume present in the contactor (i.e. since the resin takes up more volume, there is slightly less room for solution). To determine how much resin swelling occurred during these tests, the resin volumes exiting the circuit in the carousel tests can be examined.  In these tests, 40 mL of resin were added to the offline contactor, and was removed when the contactor had cycled through all positions and was taken offline.  These results are displayed in Figure 11.10.  FIGURE 11.10: MEASURED WET-SETTLED RESIN VOLUMES DURING CAROUSEL TESTS 37 38 39 40 41 42 43 44 45 0 6 12 18 24 30 36 W et -S et tle d Re si n Vo lu m e [m L] Time Sample Taken [h] Volume Input Nickel Test Out Copper Test Out 140  From Figure 11.10, it is clear that some resin swelling is occurring.  Ignoring the first three measurements, when the resin may not have yet swelled completely, the average resin volume exiting the circuit was 42.6 mL.  This corresponds to a resin volume change of only 6.1% between the loaded form of the resin and the hydrogen form of the resin.  This is significantly less than the observed volume change of 27% for the TP207XL reported in Chapter 4 when converted from the sodium form to the hydrogen form. The percent resin during the carousel tests was 7.1%, so the corresponding error in solution volume estimate is only 0.5% (negligible).  If the circuit were run with a higher percent resin, the expected error would increase, but is still fairly low, as is tabulated in Table 11.7.  Thus, the constant volume assumption appears to be adequate for this application. TABLE 11.7: ESTIMATE OF SOLUTION VOLUME ERRORS % Resin 1 5 7.1 10 15 20 25 % Error  0.07% 0.3% 0.5% 0.7% 1.2% 1.6% 2.2%  11.6  CONCLUSIONS FROM MINIPLANT TESTWORK Results from the miniplant testwork suggest that the film diffusion model from Chapter 6 and the hybrid correlation model from Chapter 7 are adequate for describing loading of a single element RIP circuit.  Both the results for nickel (low alpha) and for copper (high alpha) appear to be adequately described. The segregated flow model appears to be adequate for describing loading under cascade operating conditions.  While the finite differences model for carousel operation did not provide a very good fit to the nickel carousel run, results are within error bounds and the discrepancy between the model and observations is believed to be due to the poor feed rate control experienced during this run.  In the copper carousel run, much tighter control of the feed rate was  achieved  and  there  was  good  agreement  between  observations  and  the  finite  differences model. Therefore, both the cascade and carousel models appear to be adequate at predicting circuit operation, and are, therefore, believed to be suitable for simulating the effect of various parameters on circuit performance. 141  12.0  RIP CIRCUIT SIMULATION AND RESULTS In this chapter,  a number of different RIP operating scenarios are evaluated.  The RIP circuit simulation is based on the circuit models verified in the previous chapter.  For these simulations, the base case consists of five 20 m3 RIP loading tanks fed with 5 g/L solution (either nickel or copper) at 30 m3/hour, and a circuit recovery of 99.5%.  To determine total resin volume, it is assumed that the resin feed tank is the same size as a loading tank, but holds twice the resin volume.  Furthermore, it is assumed that resin spends six hours in the elution circuit (5 BV wash @ 10 BV/hr, 5 BV acid strip @ 2 BV/hr, 10 BV strip wash @ 10 BV/hour + 1.5  hours  for  resin  transfer  and/or  acid  hold  step).   In  terms  of  resin  properties,  the  kinetic parameters for the TP207XL, lot CHE 50019, were used, except that the manufacturer specification for resin capacity (2.0 eq/L) was used in lieu of the observed capacity for the particular lot tested.  12.1  EFFECT OF RESIN VOLUME IN TANKS AND SOLUTION BACKFLOW ON CASCADE CIRCUIT The cascade RIP circuit model was used to determine how the amount of resin in the loading tanks affects overall circuit performance.  In these runs, the total resin volume present in the loading circuit (5 loading tanks + resin feed tank) was modeled at 5 m3, 10 m3, 15 m3, 20 m3, 25 m3, 30 m3, and 35 m3 (corresponds to 3.6%, 7.1%, 10.7%, 14.3%, 17.9%, 21.4%, and 25.0% resin in each loading contactor).  To ensure that the target recovery of 99.5% was achieved, the rate of resin transfer between the tanks was adjusted.  Circuit performance was evaluated in terms of final resin loading exiting the circuit and the total resin inventory required. The set of scenarios were run twice to determine the effect of solution back-mixing during resin transfer.  In the first set, it was assumed that there was no solution back-mixing occurring (i.e. perfect separation of resin and solution when resin is transferred between contactors).  In the second set, it was assumed that all solution in the airlift with the resin was back-mixed (i.e. no interstage screening on airlift discharges). Removal of interstage screening during resin transfer could have significant benefits to an RIP operation.  During the original resin evaluations for continuous uranium RIP circuits in the 1950s, it was the vibrating interstage screens that were expected to produce the largest amount 142  of resin attrition in the circuit (52).  In modern RIP circuits, sieve-bend screens are often used, but can cause operational concerns due to fouling of the screens in some environments.  Note that interstage screening can only be removed on the resin transfer stream; interstage screening (internal or external) on pulp transfer is still required.  12.1.1  RESIN VOLUMES & SOLUTION BACK-MIXING: NICKEL CASCADE All six resin volume scenarios were run for nickel loading with no solution back-mixing, but only four resin volume scenarios were run for nickel loading with solution back-mixing.  This was because at  3.6% and 7.1% resin volumes, so much dilution was occurring that the circuit could not achieve the target 99.5% recovery. The solution profile and resin profile across the circuit have been plotted in Figures 12.1 and 12.2 for the case of no solution back-mixing; and in Figures 12.3 and 12.4 for the case of solution back-mixing.  Work done at Mintek on the optimization of carbon-in-pulp plants in the 1980s was based on a linear approach to a linear isotherm (109,110).  If this assumption were applied to an RIP circuit, it would result in an exponential decrease in solution concentration across a multi-stage circuit with equal resin distribution in each stage.  However, when the hybrid correlation is used to predict loading rates, a very different solution profile across the circuit is observed, as can be seen in Figures 12.1 and 12.3.  Thus, the conclusions from the CIP optimization study may or may not hold true for a base metal RIP circuit.  143   FIGURE 12.1: SOLUTION PROFILE ACROSS CIRCUIT –NICKEL CASCADE, NO BACK-MIXING  FIGURE 12.2: RESIN PROFILE ACROSS CIRCUIT –NICKEL CASCADE, NO BACK-MIXING 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Feed Mix 1 2 3 4 5 N ic ke l i n So lu ti on  [ g/ L] Tank 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% Solution Profile NO BACK-MIXING 0 10 20 30 40 50 60 1 2 3 4 5 Resin Feed N ic ke l L oa de d on  R es in  [g  N i /  L  R es in ] Tank 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% Resin Profile NO BACK-MIXING 144   FIGURE 12.3: SOLUTION PROFILE ACROSS CIRCUIT –NICKEL CASCADE, WITH BACK-MIXING  FIGURE 12.4: RESIN PROFILE ACROSS CIRCUIT –NICKEL CASCADE, WITH BACK-MIXING 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Feed Mix 1 2 3 4 5 N ic ke l i n So lu tio n [g /L ] Tank 10.7% 14.3% 17.9% 21.4% 25.0% Solution Profile WITH BACK-MIXING 0 10 20 30 40 50 60 1 2 3 4 5 Resin Feed N ic ke l L oa de d on  R es in  [g  N i /  L  R es in ] Tank 10.7% 14.3% 17.9% 21.4% 25.0% Resin Profile WITH BACK-MIXING 145  The effect of resin volume per stage on final resin loading is plotted in Figure 12.5.  This graph shows that the final resin loading increases with the amount of resin in each tank, but the incremental benefit of increasing the resin volume diminishes.  In order to reach the target of 99.5% recovery when less resin was present in the tank, the resin had to be cycled through the circuit more quickly.  Thus, these predictions are in agreement with those reported by researchers at the Tshwane University of Technology, who found that the final resin loading increased with the residence time of the resin in the circuit (105).  FIGURE 12.5: FINAL RESIN LOADING VS. PERCENT RESIN IN TANK –NICKEL CASCADE   0 10 20 30 40 50 60 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% Fi na l R es in  L oa di ng  [g  N i /  L  R es in ] Percent  Resin Volume in Each Tank Resin Load with back-mixing Resin Load with no back-mixing 146  The total circuit resin inventory (loading circuit + elution circuit) is plotted against the percent resin in each tank in Figure 12.6.  This graph shows that the minimum total resin inventory is achieved near 10% resin volume for the case of no solution back-mixing and somewhere between 10-15% for the case of complete back-mixing of airlift solution.  Furthermore, this graph suggests that separation of the solution and resin has minimal benefit on the overall resin inventory.  In fact, at the point where the discrepancy between the two models is greatest (10% resin volume), the total resin inventory differs by only 3 m3 (approximately 10% difference in total resin inventory).  FIGURE 12.6: TOTAL RESIN INVENTORY VS. PERCENT RESIN IN TANK –NICKEL CASCADE   0 10 20 30 40 50 60 70 0% 5% 10% 15% 20% 25% 30% To ta l R es in  In ve nt or y [m 3 ] Percent Resin Volume in Each Tank no back-mixing with back-mixing 147  To further understand the effect of solution back-mixing, the final resin loading value was plotted against the total resin residence time in the five RIP loading contactors.  These results are plotted in Figure 12.7 and show that the two curves (solution recycle and no solution recycle) are almost overlapping.  This suggests that very similar circuit performance can be achieved under both operating conditions and that it is simply necessary to increase the resin transfer rate when solution back-mixing is occurring.  FIGURE 12.7: EFFECT OF RESIN RESIDENCE TIME ON FINAL RESIN LOADING  Figure 12.8 compares the two metrics used to evaluate circuit performance against each other. Ideally, an RIP circuit will have a high resin loading and a low total resin inventory.  In an industrial RIP circuit high resin loading results in more concentrated eluate with fewer impurities, while lowering the total resin inventory results in cost savings.  Ultimately, the selection of optimum conditions will vary from system to system, and one will need to take the eluate processing steps into consideration when designing the loading circuit. 0 10 20 30 40 50 60 0 2 4 6 8 10 12 14 Re si n Lo ad in g [g  N i /  L  R es in ] Resin Residence Time in Loading Circuit [h] no back-mixing with back-mixing 148   FIGURE 12.8: TOTAL RESIN INVENTORY VS. RESIN LOADING –NICKEL CASCADE  12.1.2  RESIN VOLUMES & SOLUTION BACK-MIXING: COPPER CASCADE The same six scenarios were run for a copper RIP circuit.  The solution and resin profiles for these simulations can be found in Appendix K.  Note that for copper loading, all six resin volume scenarios were able to be evaluated for the case of solution back-mixing. The final resin loadings under various resin volume fractions are displayed in Figure 12.9. Note that the final resin loading reaches a maximum at 14% resin volume and then decreases with increasing resin volume.  This illustrates that too much resin in a circuit can lead to decreased circuit performance, and in an industrial RIP plant this would translate into increased impurity loadings on the resin exiting the circuit.  0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 To ta l R es in  In ve nt or y [m 3 ] Resin Loading [g Ni / L Resin] no back-mixing with back-mixing 149   FIGURE 12.9: FINAL RESIN LOADING VS. PERCENT RESIN IN TANK –COPPER CASCADE  Since copper has a much higher alpha value than nickel (0.62 vs 0.36) it is reasonable to expect that the effect of solution back-mixing should be more pronounced.  However, from the column graph in Figure 12.9, it only appears that this is the case at low resin concentrations.  Figure 12.10 plots the total resin inventory against final resin loading to further investigate this point.  0 10 20 30 40 50 60 70 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% Fi na l R es in  L oa di ng  [g  C u / L Re si n] Percent  Resin Volume in Each Tank Resin Load -with solution back-mixing Resin Load -no solution back-mixing 150   FIGURE 12.10: TOTAL RESIN INVENTORY VS. FINAL RESIN LOADING –COPPER CASCADE  Results from Figure 12.10 illustrate that at low final resin loadings, the discrepancy in total resin inventory is much greater for copper than for nickel.  However, it is unlikely that a copper RIP circuit would be operated in this regime, since low loadings would result in high levels of impurities co-loading onto the resin.  Thus, as in the case of nickel, it would appear that separation of resin from the pulp on the resin transfer streams may be unnecessary.  12.2  EFFECT OF SIZE AND NUMBER OF TANKS ON CASCADE CIRCUIT PERFORMANCE Since the cascade circuit simulation assumes perfect mixing and no short-circuiting of solution in  the  contactors,  the  only  effect  tank  size  has  on  circuit  performance  is  in  the  amount  of solution transferred with the resin in the airlifts.  Thus, if one assumes no solution back-mixing, the  simulation  results  are  independent  of  tank  size.   Note  that  this  was  also  true  of  the 0 10 20 30 40 50 60 0 10 20 30 40 50 60 70 To ta l R es in  In ve nt or y [m 3 ] Resin Loading [g Cu / L Resin] with back-mixing no back-mixing 151  simulation  results  presented  in  the  work  of  Zainol  and  Nicol,  for  their  simulation  based  on  a linear approach to equilibrium (3). To determine the effect of the number of stages on cascade circuit performance, a series of scenarios were run for both nickel and copper recovery.  In these scenarios 5 g/L feed was modeled through a three tank cascade circuit, a four tank cascade circuit, and a five tank cascade circuit.  For this evaluation, it was assumed that there was no solution back-mixing, and the recovery target of 99.5% was used once again.  12.2.1  NICKEL CASCADE CIRCUIT To determine the effect of the number of contactors for nickel, the amount of resin in each circuit was adjusted to achieve the target loadings of 45, 48, 50, and 52 g Ni / L Resin.  Note that the three stage circuit could not achieve a loading target of 46 g Ni / L Resin or greater, and the four stage circuit could not attain a resin loading of 52 g Ni / L Resin.  In order to achieve the  target  of  99.5%  recovery,  the  rate  at  which  the  resin  was  moved  through  the  circuit  was adjusted. Results  showed  that  to  achieve  a  higher  resin  loading  and  still  maintain  the  same  circuit recovery, one needs to slow down the rate of resin transfer and increase the amount of resin present in the circuit.  Assuming, once again, that 6 hours worth of resin are required outside the loading circuit (elution, resin transfer steps), increasing the total number of stages results in lower total resin inventories.  This is illustrated in Figure 12.11. 152   FIGURE 12.11: EFFECT OF NUMBER OF STAGES ON TOTAL RESIN INVENTORIES –NICKEL CASCADE  12.2.2  COPPER CASCADE CIRCUIT For copper, the target final resin loadings were: 57, 58.5, 60, and 61.5 g Cu / L Resin.  Note that the four tank circuit could not achieve a resin loading of 60 g Cu / L Resin, and the three tank circuit could not achieve resin loadings of 60 or 61.5 g Cu / L Resin. Results for copper showed the same general trend as those previously reported for nickel, namely that more stages resulted in a lower total resin inventory.  However, in the case of copper, higher resin loadings were achieved and the amount of resin inventory required is significantly less than that required for nickel.  This is illustrated in Figure 12.12. 0 10 20 30 40 50 60 70 3 Stages 4 Stages 5 Stages To ta l R es in  In ve nt or y [m 3 ] 52 g Ni / L Resin 50 g Ni / L Resin 48 g Ni / L Resin 45 g Ni / L Resin 153   FIGURE 12.12:  EFFECT OF NUMBER OF STAGES ON TOTAL RESIN INVENTORIES   12.3  PROCESS SENSITIVITY OF A CASCADE CIRCUIT To determine the process sensitivity of a cascade operation, various parameters were adjusted and the resulting effect on circuit recovery was calculated by the model.  Three of the nickel circuit models with no back-mixing presented in Section 12.1.1 were evaluated for each parameter: 5 m3, 20 m3, and 35 m3 resin volume in the loading circuit.   0 10 20 30 40 50 60 70 3 Stages 4 Stages 5 Stages To ta l R es in  In ve nt or y [m 3 ] 61.5 g Cu / L Resin 60 g Cu / L Resin 58.5 g Cu / L Resin 57 g Cu / L Resin 154  12.3.1 EFFECT OF UNEVEN RESIN DISTRIBUTION To determine the effect of an uneven resin profile across the RIP circuit, the distribution of resin  across  the  circuit  was  adjusted,  and  both  the  total  resin  inventory  and  the  rate  of  resin movement through the circuit were held constant.  In this simulation, the case of complete solution back-mixing with the resin transfer was considered. Four resin distribution profiles for each scenario were investigated.  For the base case, an even resin distribution was considered.  Three additional resin profiles were considered: a resin profile decreasing across the RIP circuit, a resin profile increasing across the RIP circuit, and finally a resin profile where it was assumed that the middle tank lost most of its’ resin inventory and the upstream tank had almost twice the normal resin inventory.  Figure 12.13 displays the resin profiles across the circuit for the four cases being evaluated.  FIGURE 12.13: RESIN PROFILE ACROSS RIP CIRCUIT –UNEVEN RESIN DISTRIBUTION SCENARIOS 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 Even Front end Back end Lost in Middle Fr ac ti on  o f T ot al  R es in  in  E ac h Ta nk Tank 1 Tank 2 Tank 3 Tank 4 Tank 5 155  To get a sense of how the overall resin distribution affects the solution profile across the RIP circuit, results for the four cases when the average resin volume per tank is 18% (25 m3 resin in loading circuit) are displayed in Figure 12.14.  FIGURE 12.14: SOLUTION PROFILE ACROSS CIRCUIT –UNEVEN RESIN DISTRIBUTION  This graph suggests that an RIP circuit should have the ability to somewhat compensate for uneven resin distributions.  This has been observed in miniplant testwork on laterite pressure acid leach tailings by Nicol and Zainol when they experienced higher resin concentrations in the final three stages of their circuit (18). However, different resin distribution patterns do affect the overall circuit recovery.  In the case of gold CIP circuits (110), the highest circuit recovery is expected when there is an even resin distribution  across  the  circuit.   Figure  12.15  plots  the  overall  circuit  recovery  for  each  resin 0 1 2 3 4 5 6 -2 -1 0 1 2 3 4 5 So lu ti on  C on ce nt ra ti on  [ g/ L N ic ke l] Tank Even Front end Back end Resin Lost in Tank 3 Mix Tank Feed Solution Mixing RIP Circuit 156  distribution profile for each of the three scenarios investigated: 15 m3, 25 m3, and 35 m3 resin inventory in loading circuit.  FIGURE 12.15: CIRCUIT RECOVERY FOR DIFFERENT RESIN DISTRIBUTION PROFILES  This  graph  shows  that  the  highest  circuit  recoveries  are  typically  achieved  with  an  even distribution of resin across the circuit, as is the case for gold CIP.  The one exception to this is at 15 m3 total resin inventory, where a very slight increase in recovery was observed when more resin  is  placed  in  the  front  end  of  the  circuit  (i.e.  tanks  1  and  2).   To  determine  if  a  higher recovery could be expected in a circuit operated with a very high resin transfer rate with more resin in the front end, a simulation was run with no solution back-mixing and a total resin inventory of 5 m3 in the circuit.  Results indicated a noticeable increase in predicted recovery in this case: 99.8% vs. 99.5%. 80% 85% 90% 95% 100% 15 25 35 Ci rc ui t R ec ov er y [% ] Volume of Resin in Loading Circuit [m3] Even Front end Back end Lost in Middle 157  More importantly, this graph shows the process sensitivity to fluctuations in resin distribution across the circuit.  In the case of low resin volumes in the loading tanks and a high resin feed rate  (15  m3), the process is relatively robust with respect to fluctuations in resin distribution across the circuit.  On the other hand, at high resin volumes and a low resin feed rate (35 m3), the circuit is much more sensitive to fluctuations in resin distribution across the circuit.  12.3.2  EFFECT OF LOSING RESIN VOLUME IN LOADING CIRCUIT To determine the process sensitivity to losing resin volume in the loading portion of the RIP circuit, four scenarios were then run for each base case.  In each scenario, the resin volume in the loading portion of the circuit was decreased, but the rate of resin transfer was held constant. The model was run using resin volumes of 95%, 80%, 67%, and 50% of the base case values. Figure 12.16 plots the overall circuit recovery against the resin volume in the loading section of the circuit.  FIGURE 12.16: EFFECT OF LOSING RESIN VOLUME IN LOADING SECTION OF RIP CIRCUIT Results from Figure 12.16 suggest that a loading circuit with a high resin content and a low resin transfer rate is much more robust to variance in the total amount of resin present in the 75% 80% 85% 90% 95% 100% 0 0.2 0.4 0.6 0.8 1 O ve ra ll Re co ve ry Fraction of Base Case Resin Inventory 5 m3 Resin 20 m3 Resin 35 m3 Resin 158  loading section of the circuit.  It is important to point out that the changes in resin volumes in Figure 12.16 refer to percent variation from the base case (i.e. poorer recoveries are observed when a circuit decreases from 5 m3 of resin to 2.5 m3 resin than when a circuit decreases from 35 m3 of resin to 17.5 m3 resin).   Thus,  the  circuit  with  a  low resin  volume and  a  high  resin transfer rate is much more sensitive to losing a particular volume of resin (say 1 m3 of resin) than a circuit with a high resin content and a low resin transfer rate.  12.3.3  EFFECT OF SWINGS IN FEED CONCENTRATION To determine the process sensitivity to spikes in feed concentration, the model was run with higher concentrations of nickel in the feed.  In addition to the base case of 5000 ppm nickel, the model was run at 5263, 6250, and 7500 ppm nickel in the feed.  As would be expected, if one makes no changes to circuit operation, the process recoveries drop significantly (in a real operation  one  could  increase  the  speed  of  resin  transfer  to  compensate  for  this).   Results  are plotted in Figure 12.17 and show that the scenario with high resin volumes is most sensitive to swings in feed concentration.  FIGURE 12.17: EFFECT OF INCREASED FEED CONCENTRATION ON CIRCUIT RECOVERIES 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 100% 5 6 7 8 9 10 O ve ra ll Re co ve ry Nickel Concentration in Feed [g/L] 5 m3 in loading circuit 20 m3 in loading circuit 35 m3 in loading circuit 159  12.3.4  SUMMARY: CASCADE CIRCUIT PROCESS SENSITIVITY Based on the modeling scenarios presented above, there is no one optimum circuit operating strategy  that  is  robust  under  all  process  variations.   A  circuit  operated  with  a  low  resin concentration in each tank and a high resin flowrate is more sensitive to fluctuations in the total resin volume present in the circuit, the effect of slurry back-mixing with resin transfer, and is expected to have higher levels of impurities reporting to elution.  On the other hand, if a circuit is operated with a high resin concentration in each tank and low resin flowrate, it is predicted to be more sensitive to fluctuations in feed concentration and to variations in the resin distribution profile across the circuit.   12.4  RESIN UTILIZATION: CASCADE VS. CAROUSEL OPERATION Carousel operations have the advantage of eliminating solution back-mixing and allowing for much tighter control of resin inventory.  In order to determine how the two modes of operation compare in terms of resin utilization, the cascade case with 25 volume percent of resin and no solution back-mixing was taken as a base case.  To obtain the same 99.5% recovery with a five stage carousel circuit and 25% resin volume, the time interval between changing the position of the contactors was varied.  Furthermore, it was assumed that when the lead contactor came offline, all the solution in that contactor was recycled back to the feed tank, and that the new contactor coming online in the lag position was full of water. Results from the circuit simulation showed that the carousel mode of operation is more efficient in  terms  of  resin  utilization.   To  achieve  99.5%  recovery  with  the  cascade  circuit,  the  resin needs to be transferred at a rate of 3 m3/hr,  giving  a  total  circuit  resin  residence  time  of  8.3 hours and a final resin loading of 52.9 g Ni / L Resin.  However, to achieve this same recovery with  a  carousel  circuit,  the  contactor  positions  need  to  cycle  every  2.3  hours,  yielding  a  total circuit resin residence time of 11.5 hours and a final resin loading of 57.3 g Ni / L Resin.  This suggests that the optimum resin residence time for a circuit running in cascade mode is not the same as for a circuit running in carousel mode.  This is important, as previous researchers have concluded that either operating strategy will yield the same results, but in their work the same 160  resin residence time was used to experimentally compare circuit performance under the two operating strategies (105). Furthermore, it is interesting to note that very little loading occurs in the lead contactor of the carousel arrangement, suggesting that the carousel circuit may be able to operate with fewer contactors than the cascade circuit.  This will be further explored in Section 12.5.3. The significant disadvantage of the carousel arrangement is the logistics of taking the lead contactor offline, moving the resin through the elution process, and then bringing a fresh contactor back online in the lag position.  For base metals, the required cycle times are quite fast: 2.3 hours in this particular simulation.  Practically speaking, the amount of time required to drain the lead contactor and separate the resin from the slurry could be greater than the optimum cycle time calculated by a process model.  Thus, practical materials handling constraints must be evaluated as part of a carousel circuit evaluation. The simulation results for two cycles of the carousel circuit are displayed in Figures 12.18 (solution) and 12.19 (resin).  Note that the solution and resin profiles for the cascade circuit are flat with respect to time (assuming steady-state) and the values at 25% resin volume were previously reported in Figures 12.1 and 12.2.  161   FIGURE 12.18: CAROUSEL SOLUTION PROFILE AT 25 VOLUME PERCENT RESIN  FIGURE 12.19: CAROUSEL RESIN PROFILE AT 25 VOLUME PERCENT RESIN 0 1 2 3 4 5 0 1 2 3 4 5 N ic ke l i n So lu ti on  [g /L ] Time [h] Mix Tank Lead HC C LC LAG 0 10 20 30 40 50 60 0 1 2 3 4 5 N ic ke l L oa de d on  R es in  [g  N i /  L  R es in ] Time [h] Lead HC C LC LAG 162  12.5  EFFECT OF CONTACTOR SIZE, CYCLE TIME, AND NUMBER OF CONTACTORS -CAROUSEL In terms of carousel circuit design, the cycle time between tanks coming offline, the size of the contactors and the number of contactors all play a significant role.  For this exercise, the target of 99.5% recovery was once again used, and for each scenario the amount of resin in the loading circuit or the carousel cycle time was adjusted to ensure that this target was achieved.  12.5.1  EFFECT OF CONTACTOR SIZE To determine the effect of contactor size, a resin volume of 5 m3 per contactor was used in all scenarios, the size of the contactor was varied, and the cycle time was adjusted to achieve the target of 99.5% recovery.  Results from this simulation showed that increasing the contactor size had negligible impact on the solution and resin profiles across the circuit; although increasing contactor size does result in an increased cycle time, as shown in Figure 12.20.   FIGURE 12.20: EFFECT OF CONTACTOR VOLUME ON CAROUSEL CYCLE TIME 0 0.5 1 1.5 2 2.5 3 0 5 10 15 20 25 30 35 Ca ro us el  C yc le  T im e [h ] Tank Volume [m3] 163  Figure 12.20 suggests that increasing contactor volume results in a linear increase in cycle time. However, by increasing the size of the contactor, the time required to drain the contactor also increases.  If one compares the results from the simulation run at 15 m3 and 30 m3, the cycle time only increases from 2.2 to 2.7 hours, even though the volume of the contactor has doubled. Therefore, large contactors are not necessarily beneficial for reducing the minimum allowable cycle time for practical purposes (i.e. the minimum cycle time that can accommodate the logistics of resin movement).  12.5.2  EFFECT OF CYCLE TIME To evaluate the effect of cycle time, the circuit with five 20 m3 tanks was used.  The cycle time of 1, 1.5, 2, 2.5, and 3 hours was input into the model and the resin volume in the contactors was adjusted to achieve the target of 99.5% recovery.  Results in Figure 12.21 show that for the simulations performed, as the cycle time increased, the amount of resin required in each contactor increased linearly.  FIGURE 12.21: RESIN REQUIRED PER STAGE VS. CAROUSEL CYCLE TIME 0 1 2 3 4 5 6 7 8 0 0.5 1 1.5 2 2.5 3 3.5 R es in  R eq ui re d in  E ac h Co nt ac to r [m 3 ] Contactor Cycle Time [h] 164  The effect of carousel cycle time on resin loading is displayed in Figure 12.22.  Note that the resin loading values plotted in these curves are the final resin loadings at the end of a cycle (i.e. the resin in the lead contactor is about to be transferred to elution).  FIGURE 12.22: EFFECT OF CAROUSEL CYCLE TIME ON RESIN LOADINGS  It  is  interesting  to  note  that  at  a  resin  cycle  time  of  one  hour,  most  of  the  tanks  are  loading approximately the same amount of nickel.  However, as the cycle time increases, most of the loading is taking place in only the last few tanks.  This suggests that perhaps not as many tanks are needed when the amount of resin present in each contactor is high.  0 10 20 30 40 50 60 Lead HC C LC Lag Resin from elution N ic ke l o n R es in  [g  N i /  L  R es in ] Tank 1 h 1.5 h 2 h 2.5 h 3 h 165  12.5.3  EFFECT OF THE NUMBER OF CONTACTORS To investigate the effect of the number of contactors in a carousel RIP circuit, the case of 5 m3 resin in 20 m3 contactors  was  once  again  selected  for  evaluation.   The  resin  cycle  time  was adjusted until the circuit recovery was 99.5%.  Figure 12.23 shows that the number of contactors in series is predicted to have minimal effect on the circuit cycle time.  FIGURE 12.23: EFFECT OF NUMBER OF CONTACTORS ON CAROUSEL CYCLE TIME  The other significant variable being affected by the number of contactors is the final resin loading.  The circuit resin distribution at the end of a cycle is plotted in Figure 12.24.  Note that there is minimal variation in resin loading in the third tank (52.4 – 53.0 g Ni / L Resin), regardless of the number of contactors in use.  For the sake of comparison, the resin loading for the five stage cascade circuit was 52.9 g Ni / L Resin with no back-mixing and 52.5 g Ni / L Resin with back-mixing.  This suggests that, for this particular case study, a three stage carousel circuit would provide similar metallurgical performance as a five stage cascade circuit. 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3 4 5 6 Cy cl e Ti m e [h ] Number of Stages 166   FIGURE 12.24: EFFECT OF NUMBER OF STAGES ON CAROUSEL RESIN LOADINGS   12.6  EFFECT OF RESIN PROPERTIES ON CIRCUIT PERFORMANCE In Chapter 4, a number of different iminodiacetic resins were evaluated.  In terms of the loading curves generated for resins of the same size, the most significant difference between resins was the final loading capacity.  There was also a slight difference in resin loading rate, where the Lanxess MonoPlus products were observed to have slightly faster loading rates than the other resins.  The five stage cascade model with no back-mixing was used to evaluate how these differences in resin properties may affect circuit performance.  Once again, the five cascade model was evaluated at 3.6, 14.3, and 25 vol% resin.  0 10 20 30 40 50 60 70 Lead 2nd 3rd 4th 5th R es in  L oa di ng  [ g N i /  L  R es in ] 5 tank carousel 4 tank carousel 3 tank carousel 167  12.6.1  EFFECT OF RESIN CAPACITY In the resin comparison work, results for the final loading values at 2500 ppm nickel were reported in Table 4.3.  If these values are converted from g Ni / L Resin to eq/L, the capacity of these resins varied from approximately 0.75 – 2.0 eq/L.  The second lot of TP207XL used in the miniplant had an even higher capacity of 2.4 eq/L.  Therefore, for the circuit simulations, capacity values of 0.75, 1, 1.5, 2, and 2.5 eq/L were used. In these simulations, the flowrate of resin through the circuit was adjusted to achieve the target recovery of 99.5%.  The final adjusted tank resin residence times are plotted in Figure 12.25. Note that many of the required resin flowrates at 3.6 volume percent are unrealistically fast –the point plotted at 1.5 eq/L required that the resin residence time per tank be just over 2 minutes; the point at 1 eq/L would require a tank resin residence time of about 12 seconds.  FIGURE 12.25: EFFECT OF RESIN CAPACITY ON TOTAL RESIN RESIDENCE TIME  0.0 0.5 1.0 1.5 2.0 2.5 0 0.5 1 1.5 2 2.5 3 R es in  R es id en ce  T im e pe r Ta nk  [ h] Resin Capacity [eq/L] 5 m3 Resin 20 m3 Resin 35 m3 Resin 168  It is interesting to note that in Figure 12.25, while the three curves appear to be approximately linear, the slopes are quite different, depending on the resin volume present in the circuit.  For all three scenarios, doubling the resin capacity resulted in a reduction of the required residence time by more than a factor of two. Assuming, once again, that 6 hours worth of resin flow are required for elution and resin transfer steps, the resulting total resin inventories are plotted in Figure 12.26:  FIGURE 12.26: EFFECT OF TOTAL RESIN CAPACITY ON TOTAL RESIN INVENTORY  The results in Figure 12.26 show that when a circuit is operated at low resin volume and high transfer rate, the circuit performance is more sensitive to variations in resin capacity.  This is important to be aware of when operating an RIP circuit and replacing resin inventory, as there can be variation in capacity between lots of the same resin (as was experienced in this work with  the  two lots  of  TP207XL),  and  there  can  certainly  be  large  differences  with  resins  from 0 25 50 75 100 125 150 0 0.5 1 1.5 2 2.5 3 To ta l R es in  In ve nt or y  [m 3 ] Resin Capacity [eq/L] 5 m3 Resin 20 m3 Resin 35 m3 Resin 169  different manufacturers.  Furthermore, Figure 12.26 also indicates that the relationship between total resin inventory and resin capacity is non-linear.  Therefore, using a resin with double the capacity does not necessarily mean that only half the resin inventory is required. To get a sense of how different resin capacities could affect the amount of impurities reporting to  elution,  the  final  resin  loading  was  examined.   It  is  reasonable  to  assume that  most  of  the capacity unused by nickel would be occupied by impurities in an operating circuit.  Therefore, the resin with the lowest fractional nickel loading (nickel loaded / nickel loaded at capacity), would be expected to load the highest amount of impurities.  The results for fractional nickel loading are plotted in Figure 12.27.  FIGURE 12.27: EFFECT OF CAPACITY ON FRACTIONAL NICKEL LOADING  Results  from  Figure  12.27  re-iterate  the  fact  that  when  the  circuit  is  operated  with  low  resin volumes and high resin transfer rates greater levels of impurities are expected to report to 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 5 m3 Resin 20 m3 Resin 35 m3 Resin Fr ac ti on  o f C ap ac it y Lo ad ed 0.75 eq/L 1 eq/L 1.5 eq/L 2 eq/L 2.5 eq/L 170  elution.  The general trend is that, assuming no difference in resin selectivities, and that the circuit is optimized to achieve the same recoveries at the different resin capacities, resins with higher capacities should load less impurities. To see how resin capacity affects carousel circuits, a single simulation of a five tank circuit with five cubic meters of resin in each tank was run with a resin capacity of 1.0 eq/L and was compared to the base case simulation presented in Section 12.4 (2.0 eq/L).  Results showed that the carousel cycle time dropped from approximately 2.3 to 1.3 hours and the final resin loading dropped from 98% to 95% of the resin capacity.  This suggests that using a resin with high capacity is beneficial in carousel operation as well.  12.6.2  EFFECT OF VARIATION IN DAPP ON NICKEL LOADING In the resin comparison work, the loading rates of the Lanxess Lewatit MonoPlus TP207 products were noticeably faster than for the other resins evaluated.  In Chapter 8, the hybrid correlation was fitted at the +710 -850 um size fraction to both the regular TP207 product and the TP207XL product (lot CHC 5034).  The fit parameters for these two resins were reported in Table 8.2, and the Dapp value for the MonoPlus TP207XL (5.0 x 10-12 m2/s) was significantly higher  than  the  Dapp value for the TP207 product (2.7 x 10-12 m2/s).  As was previously mentioned,  the  MonoPlus  products  are  reported  to  have  a  more  porous  structure  than  the heterodispersed product (118), and so this increase in apparent diffusivity makes sense. To see how this variation in apparent diffusivity could affect RIP circuit performance, the 5 stage circuit simulation was run with the two Dapp values, assuming a constant D factor of 0.3, and a constant resin capacity of 2.0 eq/L.  Once again, the target of 99.5% recovery was used, and the rate of resin movement was adjusted to ensure this target was reached.  Note that these simulations are not a direct comparison of the TP207 and TP207XL resins since the TP207 resin has a higher capacity than the TP207XL, but the model is holding the capacity constant to better see the effect of the variance in Dapp values. Results show that to compensate for a lower Dapp value, the resin must be moved through the circuit more quickly.  Again, assuming that there are six hours of resin in the elution circuit, 171  Figure 12.28 plots the total resin inventory against the resin volume in each tank for both the high and low Dapp values.  FIGURE 12.28: EFFECT OF APPARENT DIFFUSIVITY ON TOTAL RESIN INVENTORY  Results from this simulation show that increasing the Dapp parameter results in a lower overall resin inventory.  This is particularly pronounced when the circuit is being operated with low volumes of resin in the tanks and a high resin recirculation rate. The effect of the apparent diffusivity on final resin loadings is displayed in Figure 12.29. Results suggest that a higher apparent diffusivity results in higher final resin loading values, and hence lower levels of impurities reporting to the elution circuit. 0 10 20 30 40 50 60 70 80 90 100 0 5 10 15 20 25 30 To ta l R es in  In ve nt or y [m 3 ] Resin Volume in Tank [volume % of a 20m3 tank] Low Dapp High Dapp 172   FIGURE 12.29: EFFECT OF APPARENT DIFFUSIVITY ON FINAL RESIN LOADINGS  To see how the apparent diffusivity affected carousel circuit performance, a single simulation of a five cell circuit, with five cubic meters of resin in each tank was run using the low apparent diffusivity of 2.7 x 10-12 m2/s.  Results were compared to the base case simulation presented in Section 12.4 (Dapp = 4.4 m2/s).  The carousel cycle time decreased slightly from approximately 2.33 to 2.25 hours and the final resin loading dropped from 57.3 to 55.2 g Ni / L Resin.  This suggests that having a high apparent diffusivity is beneficial in carousel circuits as well.  12.6.3  EFFECT OF RESIN BEAD DIAMETER ON NICKEL LOADING The effect of resin bead diameter on nickel loading rate was evaluated using the five stage cascade simulation, assuming no solution back-mixing.  The simulation was run at 3.6, 14.3, 0 10 20 30 40 50 60 3.6 14.3 25 Fi na l R es in  L oa di ng  [ g N i /  L  R es in ] Resin Volume in Tank [volume % of a 20m3 tank] Low Dapp High Dapp 173  and 25 volume percent resin in each tank for resin diameters of 600, 710, 850, 1000, and 1180 Pm. The effect of resin bead diameter on total resin inventory, assuming 6 hours of resin present in the elution circuit, is plotted in Figure 12.30.  Note how the resin inventory required increases with the resin bead diameter, and that this effect is much more pronounced when the circuit is run with low resin concentrations in the tanks.  FIGURE 12.30: EFFECT OF RESIN BEAD DIAMETER ON TOTAL RESIN INVENTORY  The effect of resin bead diameter on final resin loading is plotted in Figure 12.31.  This plot indicates that the final resin loading decreases as the size of the resin bead increases, even though the resin capacity was assumed to be constant in this simulation.  This suggests that the higher loading rate associated with the smaller resin beads should result in fewer impurities 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 1200 1400 To ta l R es in  In ve nt or y  [ m 3 ] Resin Bead Diameter [um] 3.6 vol% 14.3 vol% 25.0 vol% 174  reporting to elution.  However, large resin beads are still benefical in RIP circuits to aid in the separation of the resin beads from the solids present in the feed slurry.  FIGURE 12.31: EFFECT OF RESIN BEAD DIAMETER ON FINAL RESIN LOADING  12.6.4  EFFECT OF RECIRCULATING LOAD OF NICKEL ON CIRCUIT PERFORMANCE During the resin comparison work, it was observed that an acid hold step was required to obtain complete  stripping  of  nickel  from  some  of  the  resin  beads.   During  the  operation  of  the  RIP  miniplant, the decision was made to recycle the resin to the circuit without using an acid hold step, resulting in a recirculating load of approximately 3 g Ni / L Resin.  To quantify how such a recirculating load impacts circuit performance, simulations were run with resin recirculating loads of 0, 0.5, 1, 3, and 5 g Ni / L Resin. In these simulations, the resin flowrate through the circuit was adjusted to ensure the target recovery of 99.5% was achieved in all cases.  In order to attain the target recovery with higher 0 10 20 30 40 50 60 3.6 vol% 14.3 vol% 25 vol% Fi na l R es in  L oa di ng  [ g N i /  L  R es in ] 600 um 710 um 850 um 1000 um 1180 um 175  initial loadings on the resin, the rate of resin transfer through the circuit needed to be increased. Note that 99.5% recovery at a feed rate of 5 g/L nickel corresponds to a tails concentration of 25 ppm nickel.  At 25 ppm nickel, the equilibrium capacity of the resin is still quite high (> 40 g Ni / L Resin) and so incomplete stripping would be expected to have minimal effect on the final attainable equilibrium concentration exiting the circuit.  From Figure 9.6, at a solution concentration of 0.5 ppm nickel, the equilibrium capacity of the resin was 24 g Ni / L Resin. Thus, one would have to be targeting an extremely low concentration of nickel in the tails for these levels of recirculating loads to be equal to or lower than the equilibrium loading of the resin. Figure 12.32 displays the expected final resin loading values for simulations run at 3.6, 14.3 and 25 volume percent.  This graph shows that the presence of a recirculating load on the resin beads results in slightly lower final resin loadings at low resin concentration, but the effect is hardly noticeable in the simulation run at 25 volume percent.  FIGURE 12.32  EFFECT OF RESIN RECIRCULATING LOAD ON FINAL RESIN LOADINGS 0 10 20 30 40 50 60 3.6 vol% 14.3 vol% 25 vol% Fi na l R es in  L oa di ng  [ g N i /  L  R es in ] 0.0 g Ni / L Resin 0.5 g Ni / L Resin 1.0 g Ni / L Resin 3.0 g Ni / L Resin 5.0 g Ni / L Resin 176  The effect of the resin recirculating load on total resin inventory was also investigated.  Results show that at high resin volume concentrations, there is only a slight increase in inventory requirements if resin is returned to the circuit with a recirculating load of nickel.  However, at low resin volume concentrations, returning resin to the circuit without complete stripping results in a significantly higher resin inventory.  This is displayed in Figure 12.33.  FIGURE 12.33:  EFFECT OF INCOMPLETE STRIPPING ON RESIN INVENTORY REQUIREMENTS  12.6.5  SUMMARY OF EFFECT OF RESIN PROPERTIES ON CIRCUIT PERFORMANCE It is not surprising that the results on the effect of resin properties on metallurgical performance can be summarized by saying that the best resin will have both a high capacity and a fast loading rate (high Dapp, and small dp).  The circuit simulation predicts that increasing either the capacity or the loading rate will result in lower total resin inventories and lower impurity levels reporting to elution. 0 10 20 30 40 50 60 70 80 90 0 1 2 3 4 5 To ta l R es in  In ve nt or y [m 3 R es in ] Recirculating Load on Resin [g Ni / L Resin] 5 m3 Resin 20 m3 Resin 35 m3 Resin 177  The simulation also predicts that a circuit operated with low resin volumes in the contactors and at a high resin transfer rate will be particularly sensitive to variation in these parameters. However, it is important to reiterate that other factors not included in this evaluation, such as resin durability, may drive the selection criteria of the resin for a particular application.  12.7  RESIN COMPARISON: LEWATIT TP207 VS LEWATIT MONOPLUS TP207XL Although insufficient tests were conducted to determine the hybrid correlation parameters for all the resins tested in Chapter 4, a full set of loading tests were conducted on the +710 -850 Pm size  fraction  of  the  Lewatit  TP207 and  the  Lewatit  MonoPlus  TP207XL.   These  results  were reported in Chapter 8, and the key resin properties required for circuit simulation are reported in Table 12.2 (note that the results for TP207XL, lot CHC 5034 were used in this comparison). TABLE 12.1: KEY PARAMETERS FOR NICKEL LOADING ONTO TP207 AND MONOPLUS TP207XL Property TP207 MonoPlus TP207XL dp [Pm] 770 736 K [-] 3.9 x 10-5 9.8 x 10-5 Capacity [eq/L] 2.12 2.01 Alpha [-] 0.28 0.33 Dapp [m2/s] 2.7 x 10-12 5.0 x 10-12  From Table 12.2, it can be seen that this size fraction of this lot of TP207 has a greater capacity than this lot of TP207XL.  On the other hand, the TP207XL has a larger apparent diffusivity, and a slightly smaller typical resin bead diameter.  To compare the performance of these two resins, the five stage nickel cascade circuit simulation was used.  The circuit was simulated for each resin at 3.6, 7.1, 10.7, 14.3, 17.9, 21.4, and 25 volume percent resin.  As before, the target recovery of 99.5% was achieved by varying the resin flowrate through the circuit. In order to compare the amount of impurities that would report to elution with the two resins, the final nickel loading was calculated as fractional attainment of resin capacity.  These results are plotted in Figure 12.34, which shows that, for the scenarios simulated, the TP207XL resin should have fewer impurities reporting to elution than the same size fraction of TP207. 178   FIGURE 12.34: FINAL RESIN LOADINGS –NICKEL CASCADE COMPARISON OF TP207 VS TP207XL  Assuming that six hours of resin inventory is present in the elution circuit, the total resin inventory is plotted against the percent resin volume in each tank in Figure 12.35.  This graph shows that  at  low resin  concentrations  in  the  tanks,  the  same recovery  can  be  achieved  using less TP207XL resin than TP207 resin.  At higher resin concentrations in the tanks, the resin inventory for the two resins is approximately the same.  0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% Fi na l R es in  L oa di ng  a s Fr ac ti on  o f C ap ac it y Percent  Resin Volume in a 20m3 Tank TP207 TP207XL 179   FIGURE 12.35: TOTAL RESIN INVENTORY –NICKEL CASCADE  COMPARISON OF TP207 VS TP207XL  This resin comparison suggests that the loading properties of the TP207XL resin should be superior to the loading properties of the TP207 resin.  The TP207XL product is superior in terms of the amount of impurities reporting to elution and in term of requiring less total resin inventory.  However, note that other properties, such as resin durability were not evaluated in this work.   0 10 20 30 40 50 60 70 80 90 100 0% 5% 10% 15% 20% 25% 30% To ta l R es in  In ve nt or y [m 3 ] Percent Resin Volume in Each Tank TP207XL TP207 180  12.8  KEY FINDINGS FROM CIRCUIT SIMULATIONS Key findings from the circuit simulations are presented in bullet form below, and the cascade circuit sensitivity analysis results are tabulated in Table 12.2. x The  same  recovery  from  an  RIP  circuit  can  be  achieved  in  many  ways.   If  the  resin volume in each tank is reduced, it is possible to compensate for this by increasing the resin flowrate through the circuit.  x There is a minimum resin inventory that can be achieved for a given recovery. However, this minimum resin inventory may not occur at the ideal operating point, in terms of process robustness.  x Optimum results are achieved when the resin is distributed evenly in all loading tanks.  x The total resin inventory required in a cascade circuit decreases with the number of stages used.  Furthermore, the amount of impurities reporting to elution also decreases with the number of stages.  x For the case of both nickel and copper, the deleterious effect of solution back-mixing when transferring resin to the upstream contactor is minimal, if the circuit is operated with high resin volumes and a low resin flowrate.  x A resin with high capacity and fast loading rates results in lower total resin inventories and fewer impurities reporting to elution, especially when operating with low resin volumes and a high resin flowrate.  x The optimum resin residence time for a circuit running in cascade mode is not the same as for a circuit running in carousel mode.  Results from these simulations suggest that if the optimum resin residence time is used in both cases, the carousel arrangement results in higher final resin loadings due to a longer resin residence time.  Alternatively, it may be possible to reduce the number of contactors in a carousel circuit and still achieve a similar final resin loading as occurs in a cascade circuit. 181   x Base metal circuit operation in carousel mode is probably going to be limited by the logistics of taking the lead contactor offline and transferring resin to and from the elution circuit.  In order to increase the carousel cycle time, the circuit should use a high capacity resin and be operated with high resin volumes in each tank.  x Process sensitivity of a cascade operation varies depending on whether the circuit is operated with high resin volumes and a low resin transfer rate or with low resin volumes and a high resin transfer rate.  Process sensitivity is summarized in Table 12.2.   TABLE 12.2:  SENSITIVITY OF CASCADE CIRCUITS Parameter More Sensitive When… High Resin Volume in Tank and Low Resin Transfer Rate Low Resin Volume in Tank and High Transfer Rate Total resin volume variations in loading circuit  X Uneven resin distribution across loading circuit X Solution back-mixing, especially at high D  X Swings in feed concentration X Impurities reporting to elution  X Changes in Resin Loading Capacity  X Changes in Resin Apparent Diffusivity  X Changes in Resin Bead Diameter  X Incomplete Stripping of Resin Beads  X  182  13.0  CONCLUSION The three objectives for this research program have been met.  The conclusions from this work and how they relate to the research objectives will be presented.  This will be followed by recommendations for further work.  13.1  CONCLUSIONS FROM RESEARCH The three objectives for this research program were to adequately describe the rate of nickel loading by examining data obtained under infinite solution volume conditions, verify the developed loading model under non-infinite solution volume conditions, and then use the model to quantify the effect of various RIP operating configurations.  Conclusions from this research program can be grouped according to these three objectives.  13.1.1  ADEQUATELY DESCRIBE LOADING UNDER INFINITE SOLUTION VOLUME CONDITIONS 1)  In order to obtain laboratory data that could be analyzed under infinite solution volume conditions, the pH-stat method was modified to minimize the effect of sodium co-loading onto an iminodiacetic acid resin.  The modified method held the concentration of the pay metal (nickel, cobalt, or copper) approximately constant over the course of the test, in addition to the solution pH.  For the specific case of nickel loading, the method was shown to have good repeatability with a maximum error of +/- 2 g Ni / L Resin. 2)  The modified pH-stat method was used to compare the performance of three new large bead resins that have been manufactured specifically for RIP.  Each of the three new RIP grade resins excels in at least one area: x In terms of size, the large bead resin with the largest size distribution is the Ambersep XE818.  The Purolite S930+/4888 resin is slightly smaller, while the Lewatit MonoPlus TP207XL is a significantly smaller product.  The Ambersep XE818 also has the least bulk volume change when converted from sodium form to hydrogen form.  183  x The large bead resin with the largest capacity is the Purolite S930+/4888.  The Lewatit MonoPlus TP207XL resin has a slightly lower capacity, while the Ambersep XE818 has a significantly lower capacity.  x The large bead resin with the fastest loading rate is the Lewatit MonoPlus TP207XL. This resin also has the tightest size distribution of resin bead diameters.  Since resin bead diameter is known to have a significant impact on loading rates, using a resin with a narrow size distribution will allow for a better evaluation of other parameters affecting the  system.   Therefore,  the  Lewatit  MonoPlus  resins  were  selected  for  use  in  this research program.  3)  Qualitative tests were conducted to determine the predominant rate-limiting step.  Results suggest that the process is initially film-diffusion limited, and then becomes limited by some combination of intraparticle diffusion and ion exchange rate.  The portion of loading that is film diffusion limited was able to be described by a linear approach to equilibrium –the standard engineering model for this scenario.  The portion of loading that is limited by intraparticle diffusion and/or ion exchange rate could not be adequately modeled by a standard engineering model.  In particular, the dependence on solution concentration could not be adequately described by any model. 4)  A hybrid correlation was proposed to describe the loading rate in the intraparticle diffusion / ion exchange limited regime.  The hybrid correlation combines the Vermeulen Approximation to intraparticle diffusion limited loading and an approximation to the shrinking core model.  A new model fit parameter, D, was introduced to describe the dependence of the system on solution concentration.  As D approaches zero, the model predicts standard intraparticle diffusion limitation, as described by the Vermeulen Approximation.  As D approaches one, the hybrid correlation predicts shrinking core type intraparticle diffusion.  Thus, D can be taken as a measure  of  how strong  a  boundary  is  present  between the  reacted  and  unreacted  portion  of  a resin bead during loading.  The hybrid correlation was fit to experimental results for copper, nickel, and cobalt loading.  The hybrid correlation was found to adequately describe the loading rate of nickel over the range of 50 – 5000 ppm, meeting the first objective of this work.  The alpha parameter for copper (0.62) was much higher than that for nickel and cobalt (0.28 & 0.25 184  respectively).  When resin beads loaded to approximately 50% of capacity were placed under the microscope, a distinct shell and core structure was observed for copper, but not for nickel or cobalt.  13.1.2  VERIFICATION OF ENGINEERING MODELS 5)  Three batch loading tests were conducted with non-constant nickel concentrations in solution.  The early portion of loading was modeled using the film diffusion model and the latter portion of loading was modeled using the hybrid correlation.  To determine the rate- limiting step as the resin loaded, a modified Helfferich number was developed.  This modified Helfferich number yields a value greater than one during film diffusion control and a value less than one during intraparticle diffusion / ion exchange reaction limited control.  Results showed that the two models provided an adequate fit to the experimental data, without an additional model being required to describe mixed control.  In particular, a test was carried out without any stock solution being added, and the modified Helfferich number suggests that most of this test was carried out under mixed control conditions.  6)  A five stage RIP miniplant was constructed and operated in both cascade and carousel modes for both nickel loading and copper loading.  Loading rates were predicted using the film diffusion model, the hybrid correlation, and the modified Helfferich number.  A cascade circuit model was assembled, assuming perfect mixing in each reactor and using the segregated flow model for the resin beads.  This model was verified with experimental data for both nickel and copper.  A carousel model was assembled using finite differences to calculate the changes in reactor solution concentrations and metal loadings with time.  This model was verified experimentally for copper; the results from the nickel run were within error bounds; however the error bounds for this test were large due to fluctuations in solution feed rate during circuit operation.  Results from the RIP miniplant suggest that both the hybrid correlation and the circuit models generated are adequately describing observations so that the second objective of model verification was met.  185  13.1.3  QUANTIFY EFFECT OF VARIOUS OPERATING CONDITIONS ON CIRCUIT PERFORMANCE 7)  The circuit models were used to simulate various scenarios for a potential RIP plant.  Key results from these simulations are: x The  same  recovery  from  an  RIP  circuit  can  be  achieved  in  many  ways.   If  the  resin volume in each tank is reduced, it is possible to compensate for this by increasing the resin flowrate through the circuit.  x There is a minimum resin inventory that can be achieved for a given recovery. However, this minimum resin inventory may not occur at the ideal operating point, in terms of process robustness.  x Optimum results are achieved when the resin is distributed evenly in all loading tanks.  x The total resin inventory required in a cascade circuit decreases with the number of stages used.  Furthermore, the amount of impurities reporting to elution also decreases with the number of stages.  x For the case of both nickel and copper, the deleterious effect of solution back-mixing when transferring resin to the upstream contactor is minimal, if the circuit is operated with high resin volumes and a low resin flowrate.  x A resin with high capacity and fast loading rates results in lower total resin inventories and fewer impurities reporting to elution, especially when operating with low resin volumes and a high resin flowrate.  x The optimum resin residence time for a circuit running in cascade mode is not the same as for a circuit running in carousel mode.  Results from these simulations suggest that if the optimum resin residence time is used in both cases, the carousel arrangement results in higher final resin loadings due to a longer resin residence time.  Alternatively, it may be possible to reduce the number of contactors in a carousel circuit and still achieve a similar final resin loading as occurs in a cascade circuit. 186   x Base metal circuit operation in carousel mode will probably be limited by the logistics of taking the lead contactor offline and transferring resin to and from the elution circuit. In order to increase the carousel cycle time, the circuit should be operated with high resin volumes in each tank a resin with a high capacity should be used.  x Process sensitivity of a cascade operation varies depending on whether the circuit is operated with high resin volumes in the contactors and a low resin transfer rate, or operated with low resin volumes in the contactors and a high resin transfer rate.  Table 12.2 is reprinted here to summarize which operating situations are more likely to affect a circuit, depending on which resin transfer strategy is used.  TABLE 13.1:  SENSITIVITY OF CASCADE CIRCUITS (REPEAT OF TABLE 12.2) Parameter More Sensitive When… High Resin Volume in Tank and Low Resin Transfer Rate Low Resin Volume in Tank and High Transfer Rate Total resin volume variations in loading circuit  X Uneven resin distribution across loading circuit X Solution back-mixing, especially at high D  X Swings in feed concentration X Impurities reporting to elution  X Changes in resin loading capacity  X Changes in resin apparent diffusivity  X Changes in resin bead diameter  X Incomplete stripping of resin beads  X  187  13.2  RECOMMENDATIONS FOR FURTHER WORK As with most research projects, several interesting areas and questions arose over the course of this work that were not critical to achieving the research objectives and so were not explored in detail.  Recommendations for further work can be grouped into three research areas, and each will be discussed in turn.  13.2.1  FUNDAMENTAL LOADING MODEL FOR IMINODIACETIC RESINS IN HYDROGEN FORM In the hybrid correlation, both the intraparticle diffusion and ion exchange reaction components of the process are grouped together.  Separating out these effects could provide valuable insight for manufacturers of resin beads.  In order to be able to separate out these effects, fundamental kinetic studies are recommended to model the ion exchange reaction component, and isotopic tracer tests are recommended to model the intraparticle diffusion component.  A good reference in the literature that could help to plan such a fundamental study is the work of Price et al. for the study of nickel loading onto an aminophosphonic resin in the sodium form (12). When developing such a fundamental model, it will be important to remember that the loading capacity of a resin typically varies with manufacturing lot, and the work done with the heterodispersed TP207 product suggests that loading capacity can vary with resin bead diameter within the same lot  of resin.   Therefore,  it  is  recommended that sufficient resin of a single size fraction from the same lot be on hand to run the full set of fundamental tests. To obtain the fundamental kinetics, the resin beads will need to be crushed to eliminate mass transfer effects.  It is recommended that loading rate kinetics be determined at various pH values, feed metal concentrations, and temperatures.  Past work on fundamental models for the iminodiacetic resin have typically been done using resin in the sodium form, and development of fundamental loading kinetics for a resin in the hydrogen form should be interesting. To obtain the intraparticle diffusion coefficients, isotopic tracer tests can be done to obtain the diffusivities of individual components, and these can then be used to estimate a system’s overall diffusivity by use of the Nernst-Planck equation, as other researchers have done for iminodiacetic resins in the sodium form (11,9). 188  13.2.2  APPLICATION OF HYBRID CORRELATION TO ACTUAL SLURRIES The hybrid correlation developed in this work was only applied to the case of a single metal loading onto an iminodiacetic resin in the hydrogen form.  In an actual application, it is expected that the resin will first load predominantly with impurities (calcium, magnesium, manganese, et cetera) and that these impurities will then be displaced by the pay metal.  Thus, running a series of experiments to determine how the hybrid correlation’s fit parameters (Dapp and D) change depending on the initial form of the resin (hydrogen form vs saturated with impurities) would be useful.  Furthermore, determining how the hybrid correlation’s fit parameters vary with temperature and feed pH would also provide insight into the expected behaviour of an actual RIP system. When the linear approach to equilibrium is used to model the loading of a base metal circuit in the intraparticle diffusion limited regime, the implicit assumption is that the intraparticle diffusion follows Gluckauf’s model.  Strictly speaking, Gluckauf’s model was derived for the case of single element loading; however, engineers often use the model in more complex situations.   This  is  also  true  for  the  application  of  all  other  engineering  models:  Vermeulen’s Approximation, the Shrinking Core Model, and even the linear approach to equilibrium for film diffusion control.  Therefore, while the hybrid correlation was developed and proven for the case of single element loading, it may be possible to use it to model the loading of base metals from some slurries. To investigate this possibility, the hybrid correlation should first be fit to loading tests conducted with synthetic solutions containing only one pay metal, and various common impurities.  Next, the hybrid correlation should be fit to tests run with the filtrate from actual slurries, in order to determine if the loading rates can be described without the additional complication  of  solids  redissolving  as  the  resin  loads.   Note  that  if  the  filtrate  contains  both copper and nickel /  cobalt,  then tests should first  be run at  a low pH to only recover only the copper, and then repeated at a higher pH to recover the nickel and cobalt.  Tests could also be run with the complete filtrate at higher pH to determine whether the hybrid correlation fails when trying to model the loading of two pay metals with such different loading behaviour. From  a  practical  stand-point,  it  is  interesting  to  note  that  the  Brown’s  Oxide  RIP  Flowsheet contained two RIP circuits: the first RIP circuit to recover copper and the second RIP circuit to recover nickel and cobalt. 189  13.2.3  CIRCUIT MODELING WITH VARIATIONS IN RESIN DIAMETERS In this thesis, the RIP miniplant was operated with screened Lewatit MonoPlus resin, which has quite a narrow size distribution.  Since resin bead diameter plays a significant effect on loading rate, it would be beneficial to extend the circuit models to be able to account for variations in resin bead diameters.  This work is complicated by the fact that, even within a single manufacturing lot, the capacity of the resin can vary with resin bead diameter. To account for a distribution of resin bead sizes, a statistical approach will likely be required. Such statistical methods have been applied to hydrometallurgical leaching systems and are reviewed in the literature (124).  For CIP circuits, population balance models (125) have been used to simulate the counter-current movement of carbon through multiple tanks (126,127). Since CIP can be described in terms of a linear approach to equilibrium, the population model can be reduced to a set of coupled ordinary differential equations.  However, since the hybrid correlation is based on a non-linear approach to equilibrium, a population balance model will result in a set of coupled partial differential equations.  In this regard, it is interesting to note that a population balance model has been used with a non-linear approach to equilibrium to model the continuous ion exchange system at Rossing Uranium in Namibia (128).  190  REFERENCES 1. Dorfner K. (1991)  Introduction to ion exchange and ion exchangers.  In: Ion Exchangers; Dorfner K, (ed).  Walter de Gruyter, Berlin: pp. 7-188. 2. Proctor A, Toro-Vazquez JF. (1996)  The Freundlich isotherm in studying adsorption in oil processing.  Journal of the American Oil Chemists' Society, 73 (12): pp. 1627-1633. 3. Zainol Z. (2005)  The development of a resin-in-pulp process for the recovery of nickel and cobalt from laterite leach slurries.  PhD Thesis, Murdoch University. 4. Langmuir I. (1916)  The constitution and fundamental properties of solids and liquids. Part I: Solids. Journal of the American Chemical Society, 38 (11): pp. 2221-2295. 5. Reynolds JG. (2009)  A multiresponse method for fitting solid-phase activity coefficient models to ternary ion-exchange data.  Geoderma, 151 (3-4): pp. 2644-4649. 6. Mijangos F, Tikhanov N, Ortueta M, Dautov A. (2002)  Modeling ion exchange kinetics in bimetallic systems. Industrial & Engineering Chemistry Research, 41: pp. 1357-1363. 7. Graham EE, Dranoff JS. (1982)  Application of the Stefan-Maxwell equations to diffusion in ion exchangers. Part I: Theory.  Industrial Engineering Chemistry: Fundamentals, 21 (4): pp. 360-365. 8. Nesbitt A, Abrahams J. (2005)  Computational power and ion-exchange modeling.  In: Computational Analysis in Hydrometallurgy; Dixon DG, Dry MJ, (eds).  CIM Metsoc: pp. 191-203. 9. Yoshida H, Kataoka T, Fujikawa S. (1986)  Kinetics  in  a  chelate  exchanger.  Part  II: Experimental. Chemical Engineering Science, 41 (10): pp. 2525-2530. 10. Helfferich FG. (1962)  Chapter 6: Kinetics. In: Ion Exchange.  McGraw-Hill, New York: pp. 250-322.  191  11. Yoshida H, Kataoka T, Fujikawa S. (1986)  Kinetics in a chelate ion exchanger. Part I: Theoretical Analysis. Chemical Engineering Science, 41 (10): pp. 2517-2524. 12. Price  SG,  Hebditch  DJ,  Streat  M.  (1988)   Diffusion or chemical kinetic control in a chelating ion exchange resin system.  In: Ion Exchange for Industry; Streat M (ed). Ellis Horwood, West Sussex: pp. 275-285. 13. Yoshida H, Kataoka T. (1985)  Intraparticle mass transfer in bidispersed porous ion exchangers. Part II: Mutual ion exchange.  The Canadian Journal of Chemical Engineering, 63 (3): pp. 430-435. 14. Helfferich FG, Hwang YL. (1991)  Ion Exchange Kinetics. In: Ion Exchangers; Dorfner K (ed).  Walter de Gruyter, Berlin: pp. 1277-1309. 15. Streat M. (1991)  General Ion Exchange Technology.  In: Ion Exchangers; Dorfner K (ed). Walter de Gruyter, Berlin: pp. 685-715. 16. Fleming CA, Cromberge G. (1984)  The extraction of gold from cyanide solutions by strong- and weak-base anion exchange resins. Journal of the South African Institute of Mining and Metallurgy, 84 (5):pp. 125-137. 17. Fleming CA, Nicol MJ. (1980)  A comparative study of kinetic models for the extraction of uranium by strong base anion exchange resins.  Journal of the South African Institute of Mining and Metallurgy, 80 (2): pp. 89-99. 18. Nicol MJ, Zainol Z. (2003)  The development of a resin-in-pulp process for the recovery of nickel and cobalt from laterite leach slurries.  International Journal of Mineral Processing, 72 (1-4): pp. 407-415. 19. Zainol Z, Nicol MJ. (2009)  Comparative study of chelating ion exchange resins for the recovery of nickel and cobalt from laterite leach tailings.  Hydrometallurgy, 96 (4): pp. 283-287. 20. Adams MD. (2006)  Towards a virtual metallurgical plant: CCD vs RIP case study. In: ALTA 2006 Proceedings.  ALTA Metallurgical Services. 192  21. Gluckauf E, Coates JI. (1947)  Theory of Chromatography. Part IV: The influence of incomplete equilibrium on the front boundary of chromatograms and on the effectiveness of separation.  Journal of the Chemical Society (Resumed), pp. 1315-1321. 22. Boyd GE, Adamson AW, Myers LS. (1947)  The exchange adsorption of ions from aqueous solutions by organic zeolites. II: Kinetics.  Journal of the American Chemical Society, 69 (11): pp. 2836-2848. 23. Vermeulen T. (1953)  Theory for irreversable and constant pattern solid diffusion. Industrial Engineering Chemistry, 45 (8): pp. 1664-1670. 24. Helfferich F. (1965)  Ion Exchange Kinetics. V: Ion exchange accompanied by reaction. The Journal of Physical Chemistry, 69 (4): pp. 1178-1187. 25. Dana PR, Wheelock TD. (1974)  Kinetics of a moving boundary ion-exchange process. Industrial and Engineering Chemistry: Fundamentals, 13 (1): pp. 20-26. 26. Holl W, Sontheimer H. (1977)  Ion exchange kinetics of the protonation of weak acid ion exchange resins.  Chemical Engineering Science, 32(7): pp. 755-762. 27. Thomas HC. (1944)  Heterogeneous ion exchange in a flowing system.  Journal of the American Chemical Society, 66 (10): pp. 1664-1666. 28. Hiester NK, Vermeulen T. (1952)  Saturation performance of ion exchange and adsorption columns.  Chemical Engineering Progress, 148 (10): pp. 505-516. 29. Streat M, Cloete FLD. (1987)  Chapter 13: Ion Exchange. In: Handbook of Separation Process Technology; Rousseau R (ed).  John Wiley & Sons: pp. 709. 30. LeVan MD, Carta G. (2008)  Section 16: Adsorption and Ion Exchange.. In: Perry’s Chemical Engineer’s Handbook: 8th Edition; Green DW, Perry RH, (eds).  McGraw-Hill: pp. 16.23. 31. Diaz M, Mijangos F. (1987)  Metal recovery from hydrometallurgical wastes.  Journal of Metals, 39 (7): pp. 42-44. 193  32. Bayer.   Lewatit Selective Ion Exchange Resins: Properties and applications of Lewatit TP207.  Technical Information Brochure.. 33. Laskorin  BN,  Mamilov  VA,  Koreisho  YuA,  Skorovarov  DI,  Vodolazov  LI,  Smirnov IP, Kedrouskii OL, Shulika VP, Nevski BV, Mosinets VN. (1983)  Extraction and processing of uranium in the USSR.  Atomic Energy, 54 (4): pp. 301-309. 34. Rosenbaum JB, George DR. (1971)  Cost reductions in ion exchange processing of uranium  ores.  In:  The  Recovery  of  Uranium.   International  Atomic  Energy  Agency.  pp. 297-309. 35. Mussard A. (1979)  Innovative uranium mill makes small deposit pay. Mining Engineering, 31 (10): pp. 1443-1454. 36. Angel M, Teissié J. (1982)  Résines en pulpe. Le procédé en lit fluidisé pulse de l'usine d'uranium de Saint-Pierre (Cantal).  Industrie Minérale -Les Techniques, pp. 563-569. 37. Lermigeaux G, Roques H. (1972)  Échange d'ions continu: État actuel de développement des procédés.  Chimie & Industrie Génie Chimique, 105 (12): pp. 725-748. 38. Slater MJ. (1981)  Recent industrial scale applications of continuous resin ion exchange systems.  Journal of Separation Process Technology, 2 (3): pp. 2-12. 39. Taylor A. (2008)  Review of CIX and RIP systems for uranium extraction. In: ALTA 2008 Proceedings.  ALTA Metallurgical Services. 40. Hollis RF, McArthur CK. (1955)  The resin-in-pulp process for recovery of uranium.  In: Proceedings of the International Conference on the Peaceful Uses of Atomic Energy. International Atomic Energy Agency, vol. VIII, pp. 54-63. 41. Beverly RG, McArthur CK. (1959)  Summary Report: 1954-1959.  United States Atomic Energy Commission Raw Materials Development Laboratory, WIN-115.   194  42. Painter LA, Izzo TF. (1958)  Operation of the resin-in-pulp uranium processing mill at Moab, Utah. In: Proceedings of the Second United Nations International Conference of the Peaceful Uses of Atomic Energy. International Atomic Energy Agency, vol. III, pp. 383- 386. 43. Merritt RC. (1971)  The Extractive Metallurgy of Uranium.  Colorado School of Mines. 44. White L. (1976)  Renovated Atlas Mill will produce uranium, vanadium, and copper. Engineering and Mining Journal, 177 (1): pp. 87-91. 45. Peterson EC, Matthew DC. (1957)  Uranium milling at the Bluewater Plant. The Mining Congress Journal, 43 (1): pp. 58-59, 66. 46. Anon. (1974)  Anaconda's resin-in-pulp process: another route to yellowcake. Mining Engineering, 26 (8): pp. 31-33. 47. Curfman  RL.  (1962)   Plastic  becomes  panacea  for  Uranium  Reduction  Co.   Mining Engineering, 14 (9): pp. 46-47. 48. Himsley A. (1977)  Uranium extraction from turbid liquids by continuous ion exchange processes. CIM Bulletin, 70 (785): pp. 148-154. 49. Read  FO.  (1959)   Development  of  a  resin-in-pulp process for the extraction of uranium from acid-leached pulps.  Journal of the South African Institute of Mining and Metallurgy, 60 (3): pp. 105-125. 50. Carman EHD. (1960)  A resin-in-pulp process using a floating resin technique for extracting uranium from acid leached pulp.  Journal of the South African Institute of Mining and Metallurgy, 60 (12): pp. 647-662. 51. Cloete  FLD.  (1981)   The  Relix  process  for  the  resin-in-pulp  recovery  of  uranium.   The Southern African Institute of Mining and Metallurgy, 81: pp. 66-73.   195  52. Breymann JB, Hollis ET, Lynch JT. (1958)  A continuous countercurrent resin-in-pulp process.  United  States  Atomic  Energy  Commission  Raw  Materials Development Laboratory, WIN-102. 53. Rau EL. (1957)  Uranium recovery by continuous countercurrent ion exchange.  The Mines Magazine, September 1957: pp. 39-41, 59. 54. Streat M, Naden D. (1987)  Ion exchange in uranium extraction. In: Ion Exchange and Sorption Processes in Hydrometallurgy. John Wiley and Sons, Chichester: pp. 1-55. 55. Naden D, Willey G, Bicker E, Lunt D. (1984)  Development of an in-pulp contactor for metals recovery. In: Ion Exchange Technology; Naden D, Streat M (eds).  Ellis Horwood, Chichester: pp. 690-697. 56. Naden D, Bicker E, Willey G, Whittaker CJ. (1986)  Development of a new in-pulp contactor for gold recovery.  In: Gold 100 -Proceedings of the International Conference on Gold; Fivaz CE, King RP (eds).  South African Institute of Mining and Metallurgy: pp. 41- 55. 57. Laskorin BN, Vodolazov LI, Tokarev NN, Vyalkov VI, Goldovina VA. (1977)  A filtration free method of processing uranium ores.  Atomic Energy, 43 (6): pp. 1127-1131. 58. McQuiston FW, Shoemaker RS. (1981)  Soviet Union production by ion exchange. In: Gold and Silver Cyanidation Plant Practice. Society of Mining Engineers, New York: pp. 239-241. 59. Bolinski L, Shirley J. (1996)  Russian resin-in-pulp technology, current status and recent developments. In: Proceedings of the 1996 Randol Gold Forum: pp. 419-423. 60. Fleming CA. (1989)  Recovery of gold by resin-in-pulp at the Golden Jubilee Mine. In: Precious  Metals;  Jha  MC,  Hill  SD (eds).  The  Mining,  Metals  and  Materials  Society.  pp. 105-119. 61. Seymore D, Fleming CA. (1989)  Golden Jubilee resin-in-pulp plant for gold recovery. In: Proceedings of the 1989 Randol Gold Forum, pp. 297-307. 196  62. Fleming CA, Seymore D. (1990)  Golden Jubilee RIP Plant -process economics and recent changes & improvements in plant performance. In: Proceedings of the 1990 Randol Gold Forum, pp. 237-241. 63. Fleming CA. (2003)  Section 3: Gold and Silver Recovery. In: The leaching and recovery of gold from ores and concentrates.  CIM Hydrometallurgy Short Course. 64. Dow Chemical Company. (2003)  Avocet mining strikes gold with Dow's developmental gold-selective resin.  Dowex ion exchange resins case study brochure. 65. Lewis G. (1999)  Increased recovery from preg-robbing gold ore at Penjom Gold Mine. In: Proceedings of the 1999 Randol Gold and Silver Forum. 66. Lewis GV. (2000)  The Penjom Process: An innovative approach to extracting gold from carbonaceous ore.  In: Proceedings of the 2000 Randol Gold and Silver Forum. 67. Avocet Mining PLC. (1998)  Annual Report and Accounts. 68. Avocet Mining PLC. (1999)  Annual Report and Accounts. 69. Avocet Mining PLC. (2000) Annual Report and Accounts. 70. Avocet Mining PLC. (2001)  Annual Report and Accounts. 71. Avocet Mining PLC. (2002)  Annual Report and Accounts. 72. Caledonia Mining. (2002)  Caledonia Mining Annual Report. 73. Pearton  T.  (2007)   Technical  statement  regarding  assets  for  sale  at  the  Barbrook Gold Mine. Venmyn Read. 74. Kotze M. (2009)  The resurrection of hydrometallurgy's own RIP van Winkle: Uranium. Mintek 75; June 4-5, 2009.   197  75. Leung S, Heymann L, King D. (2010)  The recovery of uranium from acid leached ore using resin-in-pulp technology.  In: Uranium 2010 -The future is U; Lam EK, Rowson JW, Ozberk, E (ed).  Canadian Institute of Mining, Metsoc: pp. 15-25. 76. Brown  J,  Goode  JR,  Fleming  C.  (2010)   The  re-emergence  of  resin  in  pulp  with  strong base resins as a low cost, technically viable process for the recovery of uranium.  In: Proceedings of the 42nd Annual Meeting of the Canadian Mineral Processors. 77. Goode JR, Brown JA. (2010)  The  Michelin  Uranium  Project,  Labrador,  Canada: Metallurgical testwork, economic studies and process design. In: Uranium 2010 -The future is U; Lam EK, Rowson JW, Ozberk, E (ed).  Canadian Institute of Mining, Metsoc: pp. 393-413. 78. Khan Resources. (2009)  Khan lays out plan for Dornod Uranium Project in Mongolia. Press Release: September 28, 2009. 79. Mantra Resources. (2009)  Positive results from resin-in-pulp metallurgical testwork on Nyota Prospect.  Press Release: December 1, 2009. 80. Ling Y, Durupt N, Banton N. (2010)  RIP Studies at AREVA: R&D and Applications for Niger and Canada Projects.  In: Uranium 2010 -The  future  is  U;  Lam  EK,  Rowson  JW, Ozberk, E (ed).  Canadian Institute of Mining, Metsoc: pp. 27-38. 81. Auerswald D, Udayar T, Kotze M, Scheepers J. (2011)  Operation of a resin in pulp (RIP) demonstration plant for recovering uranium from a South African gold pulp.  In: Proceedings of ALTA 2011.  ALTA Metallurgical Services. 82. Udayar T, Kotze MH, Yahorava V. (2011)  Recovery of uranium from dense slurries via resin-in-pulp.  In: Proceedings of the 6th Southern African Base Metals Conference. The Southern African Institute of Mining and Metallurgy,  pp. 49-64. 83. Borshoff J. (2009)  Challenges for uranium development in Africa -an industry perspective. In: 53rd International Atomic Energy Agency's General Conference Round Table. 198  84. Hladun D. (2010)  Start-up of Kayelekera Uranium Mine.  In: Uranium 2010 -The future is U; Lam EK, Rowson JW, Ozberk, E (ed).  Canadian Institute of Mining, Metsoc: pp. 415- 426. 85. Marsh D, Hladun D. (2010)  An overview of the Kayelekera Uranium Mine in Malawi. In: Proceedings of ALTA 2010.  ALTA Metallurgical Services. 86. Anon. (2008)  There's more than rum in this jungle. The Australian Mining Review, May 2008: pp. 39-40. 87. Anon. (2008)  First copper for Compass.  Compass News, November 2008. 88. Jones M. (2009)  Circular to creditors (Compass Resources Limited): April 22, 2009. 89. Anon. (2009)  Compass Resources in administration.  The Sydney Morning Herald: January 29, 2009. 90. Jones  M.  (2009)   Minutes  of  a  meeting  of  the  committee  of  inspection  of Compass Resources Limited held at 10 am, June 19, 2009. 91. Whyte RM. (2009)  Necessity the mother of invention: A reflection on the requirements for successful innovation in the metallurgical industries. In: Proceedings of the Fifth Southern African Base Metals Conference, pp. 115-126. 92. Dippenaar A, Proudfoot MR. (2005)  The performance of the AAC pumpcell circuits at the Gold Fields Limited Driefontein and Kloof Operations. In: Proceedings of the 2005 Randol Innovative Metallurgy Forum. 93. Van Hege B, van Tonder D, Bell R, Wyethe J, Kotze M. (2006)  Recovery of base metals using MetRIX.  In:Proceedings of ALTA 2006.  ALTA Metallurgical Services. 94. Yahorava V, Scheepers J, Kotze MH, Auerswald D. (2009)  Evaluation of various durability tests to assess resins for in-pulp applications.  In: Proceedings of the Fifth Southern African Base Metals Conference: pp. 341-358. 199  95. van Tonder D, Kotze M, Van Hege B. (2007)  Uranium recovery from acid leach liquors: IX or SX? In: Proceedings of ALTA 2007.  ALTA Metallurgical Services. 96. van Tonder D, Van Hege B. (2007)  Uranium recovery from acid leach liquors: The optimization of RIP/SX based flowsheets.  In: Proceedings of the Fourth Southern African Base Metals Conference: pp. 241-260. 97. Carr J, Zontov N, Yamin S. (2008)  Meeting the future challenges of the uranium industry. In: Proceedings of ALTA 2008.  ALTA Metallurgical Services. 98. Clean-TeQ. (2009)  Annual Report. 99. Green BR, Kotze MH, Engelbrecht JP. (1998)  Resin-in-pulp -After gold, where next? In: EPD Congress; Mishra, B. (ed). The Mining, Metals, and Materials Society: pp. 119-136. 100. Greager IP, Wyethe JP, Kotze MH, Dew D, Miller D. (2001)  A resin-in-pulp process for the recovery of copper from bioleach CCD underflows.  In: Proceedings of the Conference on Copper, Cobalt, Nickel and Zinc Recovery; held in Victoria Falls, Zimbabwe.  The South African Institute of Mining and Metallurgy. 101. Wyethe JP, Kotze MH, Greager IP, Swarts AC. (2002)  Cobalt nickel and copper recovery with resin-in-pulp. In: Proceedings of ALTA 2002.  ALTA Metallurgical Services. 102. Kotze MH, Wyethe JP, Meyer EHO. (2005)  Resin in pulp for the recovery of zinc from zinc refining residues.  In: Proceedings of Lead and Zinc '05.  The Mining and Materials Processing Institute of Japan. 103. Yahorava V, Kotze MH. (2011)  RIP pilot plant for the recovery of copper and cobalt from tailings. In: Proceedings of the 6th Southern African Base Metals Conference.   The Southern African Institute of Mining and Metallurgy,  pp. 149-168. 104. Zontov N. (2001)  Potential benefits of resin-in-pulp for PAL plants. In:  Proceedings  of ALTA 2001.  ALTA Metallurgical Services.  200  105. Makhubela TR. (2006)  Base metal recovery by resin-in-pulp (RIP) technology. Master's Thesis: Tshwane University of Technology. 106. Mendes FD, Martins AH. (2005)  Recovery of nickel and cobalt from acid leach pulp by ion exchange using chelating resin.  Minerals Engineering, 18: pp. 945-954. 107. Wassink B, Neufeld M, Dreisinger D, Freeman GKW. (2006)  Towards a resin-in-pulp process for recovery of nickel and cobalt from laterite leach tailings: an iminodiacetic acid ion exchange resin as a prospective resin. In: Proceedings of the 67th International Water Conference, paper IWC-06-33. 108. Zainol Z, Nicol MJ. (2009)  Ion-exchange equilibria of Ni2+, Co2+, Mn2+, and Mg2+ with iminodiacetic acid chelating resin Amberlite IRC748.  Hydrometallurgy, 99 (3-4): pp. 175-180. 109. Nicol MJ, Fleming CA, Cromberge G. (1984)  The absorption of gold cyanide onto activated carbon. I: The kinetics of absorption from pulps.  Journal of the South African Institute of Mining and Metallurgy, 84 (2): pp. 50-54. 110. Nicol MJ, Fleming CA, Cromberge G. (1984)  The absorption of gold cyanide onto activated carbon. II: Application of the kinetic model to multistage absorption circuits. Journal of the South African Institute of Mining and Metallurgy, 84 (3): pp. 70-78. 111. Kotze M, Green B, Mackenzie J, Virnig M. (2005)  Resin-in-pulp and resin-in-solution. In: Advances in Gold Ore Processing; Adams, MD (ed).  Elsevier: pp. 603-635. 112. Mikhaylenko M. (2010)  Solution Purification III -IX from Solution (from ISR) and RIP. Canadian Institute of Mining Short Course held in conjunction with U2010, August 14-15. 113. Merritt  RC.  (1971)   Descriptions  of  active  uranium milling  plants  in  the  United  States  - 1967.  In: The Extractive Metallurgy of Uranium.  Colorado School of Mines: pp. 379- 525. 114. Merritt RC. (1971)  Inactive and past uranium milling operations in the United States.  In: The Extractive Metallurgy of Uranium.  Colorado School of Mines: pp. 526-545. 201  115. Laskorin BN. (1961)  The sorption extraction of uranium from pulps and solutions. Atomic Energy, 9 (4): pp. 822-832. 116. Babjak J. (1982)  A study on ion exchange kinetics involving chelating resin.  PhD Thesis, The University of Waterloo. 117. El-Shahat  M,  Moawed  E,  Burham  N.  (2008)   Preparation, characterization and applications of novel iminodiacetic polyurethane foam (IDA-PUF) for determination and removal of some alkali metal ions from water.  Journal of Hazardous Materials, 160 (2-3): pp. 629-633. 118. Halle  O,  Rossoni  D.  (2005)   A new (monodisperse)  chelating  resin  designed for mining application -Lewatit MonoPlus TP207. In: Proceedings of ALTA 2005.  ALTA Metallurgical Services. 119. Eigen M. (1963)  Fast elementary steps in chemical reaction mechanisms. Pure and Applied Chemistry, 6 (1): pp. 97-116. 120. Fogler HS. (1992)  Elements of Chemical Reaction Engineering: Second Edition. Prentice- Hall: New Jersey. 121. Dixon DG (2007).  Mixing and Mass Transfer in Hydrometallurgy.  Hydrometallurgy Short Course (presented as part of MTRL 564). The University of British Columbia. 122. Harriott P. (1962)  Mass Transfer to Particles. Part I -Suspended in Agitated Tanks. American Institute of Chemical Engineers Journal, 8 (1): pp. 93-101. 123. Awakura Y, Doi T, Majima H. (1988)  Determination of the Diffusion Coefficients of CuSO4, ZnSO4, and NiSO4 in aqueous solution.  Metallurgical Transactions B, 19B (1): pp. 5-12. 124. Dixon DG. (1996)  The multiple convolution integral: a new method for modeling multistage continuous leaching reactors.  Chemical Engineering Science, 51(21): pp. 4759- 4767. 202  125. Herbst JA, Ashihene SW. (1993)  Modelling and simulation of hydrometallurgical processes using the population balance approach.  In: Modeling, Simulation and Control of Hydrometallurgical  Processes;  Papangelakis  VG,  Demopoulos  GP  (eds).   Canadian Institute of Mining: pp. 3-44. 126. Stange W, King RP. (1987)  A population balance approach to the modeling of the CIP process. In: APCOM 1987: Application of Computers and Mathematics in the Minerals Industries; King RP, Barker JJ (eds).  South African Institute of Mining and Metallurgy: pp. 209-221. 127. Stange W, King RP, Woollacott L. (1990)  Towards more effective simulation of CIP and CIL processes. 2: A population-balance-based simulation approach.  Journal of the South African Institute of Mining and Metallurgy, 90 (11): pp. 307-314. 128. McInnes CA, Stange WW. (1993)  The steady-state modeling of the Rossing Uranium extraction circuit. In: Modelling and simulation of hydrometallurgical processes using the population balance approach.  In: Modeling, Simulation and Control of Hydrometallurgical Processes; Papangelakis  VG,  Demopoulos  GP  (eds).   Canadian Institute of Mining: pp. 45-58.   203  APPENDIX A:  BLANK BATCH TEST CHECK SHEET  Resin: Size: Ni Conc: Date: Prep Test Solution:  500mL in 1L beaker.  Place into hotwater bath to bring up to 30C target g NiSO4: actual g NiSO4 added: Prep stock sol'n: 0.25M NaOH,  NiSO4.  Accuracy very important. NiSO4: g into a 50 mL flask Actual g = NaOH: 25 mL into a 100 mL flask Measure SG to three decimal places by pipetting 10 mL into flask on scale.  Take 4-6 measurements. SG of NaOH: Calc SG SG of NiSO4: Calc SG Prep Equipment: Turn on scales,  calibrate pH probe (stand = pH 7, slope = pH 4) Flush stock sol'n addition lines.  Use solutions of appropriate pH so controller turns pump on and off Weigh stock sol'n:  Weights with no stoppers, tubing connected to flask.  Record min. three decimal places. gross weight NaOH: gross weight NiSO4: Prep resin: 5mL in H+ form, place in beaker and fill to 15 mL with DI water.   Initial resin pH = pH adjust with NaOH to target 3.9-4.1 when stirred.   Actual resin pH= pH adjust test solution in beaker to target 3.9 - 4.1.   Beaker intial pH = final pH = Take sample #1.  (Approx 5 mL) Prep beaker for test: check pH probe, stock addition lines in place, pump is on "auto" Measure rpm on motor. Target is 275-325 rpm; note that some fluctuation is normal.  RPM = Prep Scale:  Zero scale with nothing on it.  Replace flasks and record initial weight Start Windmill Logger.  N.B.  Parsing of data requires that a "+" sign preceeds all data -> else get "error 114" File> Restore logger setup: RIPBatch.  Enter a file name & start logging data: Start test:  Transfer resin to test beaker and record start time as displayed on logger: Sampling:  Take sample after 1 hr, 4hrs, end of test Time of Samples: Ending Test: Shut off controller, mixer, stop data-logging.  Record final wt on recording scale Measure gross weights stock sol'n, no stoppers, tubing: gross NaOH: gross NiSO4: Take sample of NiSO4 stock solution for assay Transfer resin into stripping column and change the DI water in the stripping flask (want NO ACID) Pass at least 5 BV DI water over resin at 10BV/hr  (~0.8 mL/min).  Solution to waste Pass 5 BV 100 gpL sulphuric acid over resin and collect in 100 mL volumetric flask Leave resin in column 20 hours for acid hold step.  Close stopcock at base of burrette Pass 10 BV DI water over resin and collect in the same 100 mL flask Make up volume in flask, ready for assay LOADING RATE TEST RESIN ELUTION 204  APPENDIX B: MANUFACTURER PROVIDED RESIN PROPERTIES The following resin properties are taken from the published data sheets from the various resin manufacturers.  Note that all resins tested were macroporous resins with iminodiacetic acid functional group and were shipped in the sodium form. TABLE B.1: RESIN PROPERTIES FROM MANUFACTURER DATA SHEETS Resin Amberlite IRC748 Ambersep XE818 Lewatit MonoPlus TP207 Lewatit MonoPlus TP207XL Lewatit TP207 Purolite S930 Purolite S930 Plus Purolite S930 Plus /4888 Lot Tested (92) 78544 6605081 CHT 5007 CHC 5034 CHE 50019 CHK 00444 107-92 SR 602078/1 SR 602078/2 Total exchange capacity t 1.35 eq/L (Naform) - t 2.0 eq/L (H form) t 2.0 eq/L (H form) t 2.2 eq/L (H form) - t 2.9 eq/L - Reversible Swelling 30% typical - 25% max 25% max 30% max - 35% max - Size Range [mm] - - - - >90% 0.4-1.25  - 0.425 – 1.000 - Mean Size [mm] 0.50-0.65 - 0.61 r 0.05 0.79 r 0.05 - - - - Effective Size [mm] - - - - 0.55 r 0.05 - - - Fines Content [mm] max 1% < 0.3 - - - - - max 2% < 0.425 - Coarse Beads [mm] max 5% > 1.1 - - - - - - - Uniformity Coefficient d 1.7 - d 1.1 d 1.1 d 1.7 - d 1.5 - Data Source [B1] [B2] [B3] [B4] [B5] [B6] [B7] [B8]  Data Sources [B1]  Rohm and Haas.  Datasheet: Amberlite IRC748, version 2006. [B2]  New Dow product –no datasheet available at time of publication of thesis. [B3]  Lanxess.  Product Information LEWATIT MonoPlus TP207, edition 2011-10-13. [B4]  Lanxess.  Product Information LEWATIT MonoPlus TP207XL, edition 2011-10-13. [B5]  Lanxess.  Product Information LEWATIT TP207, edition 2011-10-28. [B6]  Historical Purolite Product.  Unsure as to appropriate datasheet. [B7]  Purolite.  Product DataSheet Purolite S930 Plus Chelation Resins, edition Feb 29, 2012. [B8]  New Purolite product –no datasheet available at time of publication of thesis. 205  APPENDIX C: LOADING ISOTHERMS The equilibrium isotherms in this appendix were generated from the resin final loading values obtained during the laboratory batch tests.  For completeness, both the loading value and the stripping value are included on the curves (note: strip values for tests run before the 20 hour acid hold step was introduced are not plotted on the graphs).  The mass action law was fit to the final loading values.  The best fit parameters were reported in Table 8.1, and are repeated here:  Resin TP207 TP207XL CHC5034 TP207XL CHE50019 TP207MP TP207MP TP207MP TP207XL CHE50019 Metal Ni Ni Ni Ni Co Cu Cu K 3.9 x 10-5 9.8 x 10-5 5.0 x 10-5 7.5 x 10-4 Capacity [eq/L] 2.12 2.01 2.36 1.97 1.91 2.11 2.50   Nickel loading onto TP207 (+710 / -850 um size fraction):  0 10 20 30 40 50 60 70 0.01 0.10 1.00 10.00 N ic ke l l oa di ng  [g  N i /  L  R es in ] Nickel in solution [g Ni / L] Isotherm for TP207 Loaded Value Stripped Value Mass Action Law Model Fit 206  Nickel loading onto TP207XL, lot CHC 5034 (Resin Comparison Work):   Nickel loading onto TP207XL, lot CHE 50019 (RIP Miniplant):  Note that no acid hold step was conducted to predict recirculating load on resin   0 10 20 30 40 50 60 70 0.01 0.1 1 10 N ic ke l l oa di ng  [g  N i /  L  R es in ] Nickel in solution [g Ni / L] Isotherm for TP207XL (Resin Comparison) Loaded Stripped Mass Action Law Model Fit 0 10 20 30 40 50 60 70 0.01 0.1 1 10 N ic ke l l oa d in g [g  N i /  L  R es in ] Nickel in solution [g Ni / L] Isotherm for TP207XL (RIP Miniplant) Loaded Mass Action Law Model Fit Stripped -no acid hold step 207  Nickel loading onto TP207MP:   Cobalt loading onto TP207MP:    0 10 20 30 40 50 60 70 0.01 0.1 1 10 N ic ke l l oa di ng  [g  N i /  L  R es in ] Nickel in solution [g Ni / L] Isotherm for TP207MP Loaded Stripped Mass Action Law Model Fit 0 10 20 30 40 50 60 70 0.01 0.1 1 10 Co ba lt  [g  C o / L Re si n] Cobalt in solution [g Co / L] Isotherm for TP207MP Loaded Stripped Mass Action Law Model Fit 208  Copper loading onto TP207MP:   Copper loading onto TP207XL, lot CHE 50019 (RIP Miniplant):  Note that no acid hold step was conducted to predict recirculating load on resin   0 10 20 30 40 50 60 70 0.010 0.100 1.000 10.000 Co pp er  lo ad in g [g  C u / L Re si n] Copper in solution [g Cu / L] Isotherm for TP207MP Loaded Stripped Mass Action Law Model Fit 0 10 20 30 40 50 60 70 80 0.010 0.100 1.000 10.000 Co pp er  lo ad in g [g  C u / L Re si n] Copper in solution [g Cu / L] Isotherm for TP207XL (RIP Miniplant) Loaded Mass Action Law Model Fit Stripped -no acid hold step 209  APPENDIX D: QUALITATIVE TESTS ON EFFECT OF PH (COPPER ALPHA VALUE) Qualitative tests were run at 54, 541, and 5412 ppm copper at pH 2, 3, and 4.  Mass balances for these tests are displayed in the table below: Target pH 4 3 2 Solution ppm Cu 5412 541 54 5412 541 54 5412 541 54 Copper Loaded 80.3 78.6 72.5 76.1 69.1 60.9 58.6 46.3 25.9 Copper Stripped 81.2 80.2 72.8 76.3 73.9 62.3 64.3 52.1 24.9  Ideally, the solution composition for these tests was to be held constant over the course of the test.  The following graph plots the measured solution composition (analyzed by AA) over the duration of the test.  Results show that solution composition was held fairly constant for most tests (an exception being the 54 ppm test at pH 3).   10 100 1,000 10,000 0 5 10 15 20 25 N ic ke l i n Te st  B ea ke r [ pp m ] Time in Test [hrs] pH 2 pH 3 pH 4 210  APPENDIX E: DERIVATION OF MODIFIED HELFFERICH NUMBER The Helfferich number, is a dimensionless number used to predict the nature of the rate controlling step in an ion exchange reaction.  It was defined for the case of complete conversion of an ion exchange reaction for two ions of equal mobility and equal valence, under infinite solution volume conditions.  E.1  DERIVATION OF STANDARD HELFFERICH NUMBER The following section summarizes the key steps Helfferich used in developing his criterion for the rate determining step.  The full derivation can be found on page 277 of his text on ion exchange (10).  Given: Ion exchange reaction:    Aഥ+ B ĺ A+ Bഥ   ,  F= [Bഥ] [Bഥ]eqbൗ Fractional Attainment of Equilibrium (equal valence ions, equal mobilities, ISV conditions): Film Diffusion Control: ln൛1-Fൟ+ ቀ1- 1 SB Aቁ F= - 6Df[A] dpį[Aഥ]eqbSBA t Intraparticle Diffusion Control: F=1- 6 ʌ2 σ ൜ 1 n2 exp ൬- 4Dpʌ 2n2 dp 2 t൰ൠ’n=1 Half-time for exchange:  For film diffusion control:  t1 2Τ =൫0.167+0.064 SB A൯ dpį[Aഥ ]eqb 2Df[A]  For intraparticle diffusion control: t1 2Τ =0.030 dp 2 4 Dp   211  Calculate: Helfferich Number= ൫t1 2Τ ൯film ൫t1 2Τ ൯particle ൙ Helfferich Number= ൫0.167+0.064 SB A൯ dpį[Aഥ]eqb 2Df[A] 0.030 dp 2 4 Dp ൚ Helfferich Number= 2[Aഥ]eqbDpį[A]Dfdp ൫5.57+2.13SBA൯  Compare to Helfferich’s Criterion:  For film diffusion:    2 [Aഥ ]eqbDpį[A]Dfdp ൫5+2SBA൯ ب1  For intraparticle diffusion:    2[Aഥ ]eqbDpį[A]Dfdp ൫5+2SBA൯ ا1  E.2  DERIVATION OF MODIFIED HELFFERICH NUMBER The modified Helfferich number used in this thesis was derived using a similar approach.  In lieu of assuming equal valence, equal mobility ions for film diffusion control, the simpler case of isotopic exchange that is typically used in engineering models was considered.  In lieu of the exact solution to the differential equation for intraparticle diffusion, the hybrid correlation presented in this thesis for approximating intraparticle diffusion control was used. Since it is reasonable to expect that the initial loading of the resin will be film diffusion controlled, while most of the loading will be intraparticle diffusion controlled, it was desired to derive the modified Helfferich number to be a function of the current fraction reacted.  Under such a scheme, the Helfferich number will decrease as the reaction progresses, illustrating how 212  the mechanism changes from initial film diffusion control to finally being intraparticle diffusion controlled. Given: Ion exchange reaction:    Aഥ+ B ĺ A+ Bഥ   , F= [Bഥ] [Bഥ]eqbൗ Fractional Attainment of Equilibrium (isotopic exchange, infinite solution volume): Film Diffusion Control:  F = 1- exp൫-kft൯ Intraparticle Diffusion Control: F = ට1-exp൫-4kht൯ Time required to achieve fractional loading of F:  Film Diffusion Control:  tf = -ln൫1-F൯kf  Intraparticle Diffusion Control: tp = -ln൫1-F2൯4kh Definitions of diffusion constants:  Film Diffusion Control:  kf = 6Df[B] dpį[Bഥ]eqb  Intraparticle Diffusion Control: kh= ʌ2Deff൫1.622[B] [Bഥ]eqbൗ ൯Įdp2 Calculate:  Modified Helfferich Number= tf tpൗ  Modified Helfferich Number= 4khln ൫1-F൯ kf ln൫1-F2൯ ൘ Modified Helfferich Number= ʌ2(1.622)Į 3 Deffį Df dp ቊ [B][Bഥ]eqbቋĮ-1  ln(1-F)ln൫1-F2൯  213  APPENDIX F: CAROUSEL FINITE DIFFERENCE MODEL The  output  from  a  single  run  of  the  carousel  finite  difference  model  would  be  over  one thousand pages long.  Consequently, only the output for the first few minutes of the first three contactors from the nickel run have been printed out for inclusion in this appendix. The model starts at t = 0 and assumes that all the contactors contain no metal ions in solution (they were filled with water), and all the resin starts with the recirculated load from elution (i.e. 3 g Ni / L Resin or 0 g Cu / L Resin).  The model then works through a time-step sequence, calculating all subsequent resin and solution values at time t + 't, based on the previous values. To determine the resin loading in each time step, [Mഥ ]t+οt , the equilibrium resin loading value, and the equivalent resin loading times, tr , for both the film diffusion model and hybrid correlation are calculated based on the previous resin loading, [Mഥ ]t, and the previous solution concentration, [M]t. The resin loading value is then calculated at time tr + 't, using both the film diffusion model and the hybrid correlation.  The lower of these two values is then taken for the resin loading [Mഥ ]t+οt . The amount of metal in the contactor at time t + 't can be expressed as the amount of metal present at time t, plus the amount of metal that entered the system in the feed stream, less the amount of metal that overflowed the reactor.  For a feed flowrate of Q, and a reactor containing a volume VR of resin and VS of solution, this is expressed as: VR[Mഥ ]t+οt+VS[M]t+οt= VR[Mഥ ]t+VS[M]t+Qοt[M]feed-Qοt[M]t Rearranging, the solution concentration at time t + 't can then be calculated using the following formula: [M]t+οt= {VS[M]t+Q οt ([M]feed-[M]t)-VR([Mഥ ]t+οt-[Mഥ ]t)}VS  214  Resin Side TP207XL Solution Side Contactors 0.56 L tot K 9.78E-05 Feed 4.140 gpL Ni working vol 0.52 L Cap 2.36 Flow based on tails volume switch 60 mins bead dia 736 um delta t 0.24 min 0.000736 m Initial Nickel Concentrations: Resin Distribution [mL] Kinetics Lead 0.00 gpL Ni 3.00 g Ni / LR Lead 40 alpha 0.36 HC 0.00 gpL Ni 3.00 g Ni / LR HC 40 Dapp 4.43E-12 m2/s C 0.00 gpL Ni 3.00 g Ni / LR C 40 Df/delta 2.73E-05 m/s LC 0.00 gpL Ni 3.00 g Ni / LR LC 40 Ni 58.71 g/mol Lag 0.00 gpL Ni 3.00 g Ni / LR Lag 40 Offline 0.00 gpL Ni 3.00 g Ni / LR t [hr] gpL Ni g Ni/L R eqb Ni kh eqv s to load resin-hybrid kl eqv s to load resin-film gpL Ni tk g Ni/LR 21.5 0.000 0.000 3.00 0.000 3.00 21.5 0.004 0.000 3.00 3.00 0.00E+00 0.0 0.00 0.00E+00 0.0 0.00 0.041 3.00 21.5 0.008 0.041 3.00 58.11 6.62E-06 100.8 3.21 1.32E-04 401.6 3.10 0.074 3.10 21.5 0.012 0.074 3.10 60.76 8.17E-06 80.0 3.37 2.37E-04 221.5 3.30 0.099 3.30 21.5 0.016 0.099 3.30 61.86 9.08E-06 78.5 3.59 3.18E-04 172.5 3.57 0.118 3.57 21.5 0.020 0.118 3.57 62.46 9.69E-06 84.3 3.86 3.81E-04 154.5 3.89 0.136 3.86 21.5 0.024 0.136 3.86 62.89 1.02E-05 92.7 4.15 4.37E-04 145.1 4.23 0.153 4.15 21.5 0.028 0.153 4.15 63.25 1.06E-05 101.3 4.43 4.93E-04 137.6 4.57 0.171 4.43 21.5 0.032 0.171 4.43 63.56 1.11E-05 110.2 4.71 5.50E-04 131.5 4.90 0.189 4.71 21.5 0.036 0.189 4.71 63.82 1.15E-05 119.2 4.99 6.07E-04 126.4 5.23 0.207 4.99 21.5 0.040 0.207 4.99 64.06 1.18E-05 128.4 5.26 6.65E-04 121.9 5.55 0.225 5.26 21.5 0.044 0.225 5.26 64.26 1.22E-05 137.6 5.53 7.23E-04 118.1 5.87 0.243 5.53 21.5 0.048 0.243 5.53 64.45 1.26E-05 147.0 5.79 7.82E-04 114.7 6.19 0.262 5.79 21.5 0.052 0.262 5.79 64.61 1.29E-05 156.4 6.05 8.41E-04 111.6 6.50 0.280 6.05 21.5 0.056 0.280 6.05 64.77 1.32E-05 165.8 6.31 9.01E-04 108.9 6.81 0.299 6.31 21.5 0.060 0.299 6.31 64.90 1.35E-05 175.4 6.56 9.60E-04 106.5 7.11 0.318 6.56 21.5 0.064 0.318 6.56 65.03 1.38E-05 184.9 6.81 1.02E-03 104.2 7.41 0.336 6.81 21.5 0.068 0.336 6.81 65.15 1.41E-05 194.6 7.05 1.08E-03 102.2 7.71 0.355 7.05 21.5 0.072 0.355 7.05 65.25 1.44E-05 204.2 7.30 1.14E-03 100.3 8.00 0.374 7.30 flowrate mL/min LEAD  215  gpL Ni g Ni/L R eqb Ni kh eqv s to loadresin-kh kl eqv s to loadresin-kl gpL Ni tk g Ni/LR 0.0000 3 3 0 0 3 0.0000 3 3 0 0 0 0 0 0 0.000 3.00 0.0004 3 14.09092 1.26E-06 9218.916 3.002288 1.31E-06 182808.4 3.000209 0.001 3.00 0.0011 3.000209 24.95144 1.81E-06 2012.295 3.010847 3.59E-06 35640.09 3.001345 0.002 3.00 0.0020 3.001345 31.84478 2.23E-06 999.8156 3.022785 6.43E-06 15388.64 3.004017 0.003 3.00 0.0030 3.004017 36.34553 2.57E-06 667.8738 3.036117 9.49E-06 9094.415 3.008571 0.004 3.01 0.0039 3.008571 39.49151 2.84E-06 512.0419 3.050458 1.26E-05 6289.435 3.015189 0.005 3.02 0.0049 3.015189 41.8395 3.08E-06 422.8743 3.065961 1.57E-05 4754.766 3.023983 0.006 3.02 0.0059 3.023983 43.67376 3.29E-06 365.6301 3.08281 1.89E-05 3805.186 3.035019 0.007 3.04 0.0068 3.035019 45.15614 3.47E-06 326.0513 3.101161 2.2E-05 3167.35 3.048341 0.008 3.05 0.0078 3.048341 46.38604 3.64E-06 297.2625 3.121138 2.51E-05 2713.045 3.063974 0.009 3.06 0.0088 3.063974 47.42814 3.79E-06 275.5572 3.142841 2.81E-05 2374.961 3.081935 0.010 3.08 0.0097 3.081935 48.32635 3.94E-06 258.7677 3.166347 3.12E-05 2114.692 3.102233 0.011 3.10 0.0106 3.102233 49.11164 4.07E-06 245.5437 3.191716 3.42E-05 1908.871 3.124874 0.012 3.12 0.0116 3.124874 49.80643 4.2E-06 235.0022 3.218997 3.72E-05 1742.536 3.149864 0.013 3.15 0.0125 3.149864 50.4274 4.31E-06 226.5418 3.248229 4.02E-05 1605.694 3.177203 0.013 3.18 0.0134 3.177203 50.98724 4.43E-06 219.7391 3.279447 4.31E-05 1491.438 3.206893 0.014 3.21 0.0143 3.206893 51.49573 4.53E-06 214.2883 3.312676 4.61E-05 1394.848 3.238937 0.015 3.24 0.0153 3.238937 51.96059 4.64E-06 209.9627 3.347941 4.9E-05 1312.333 3.273334 0.016 3.27 HIGH CENTRE  216  gpL Ni g Ni/L R eqb Ni kh eqv s to loadresin-kh kl eqv s to loadresin-kl gpL Ni tk g Ni/LR 0.0000 3 3 0 0 3 0.0000 3 3 0 0 0 0 0 0 0.000 3.00 0.0000 3 0.218954 2.39E-07 0 0.000812 1.29952E-08 0 4.1E-08 0.000 3.00 0.0000 3 0.80396 3.84E-07 0 0.003782 4.85347E-08 0 5.62E-07 0.000 3.00 0.0000 3 1.799828 5.19E-07 0 0.009839 1.11886E-07 0 2.9E-06 0.000 3.00 0.0001 3 3.164204 6.45E-07 888072.7 3.000006 2.04904E-07 14438704 3 0.000 3.00 0.0001 3 4.812368 7.64E-07 161006 3.000104 3.27772E-07 2979370 3.000009 0.000 3.00 0.0001 3.000009 6.633939 8.75E-07 65332.6 3.000303 4.785E-07 1257857 3.000034 0.000 3.00 0.0002 3.000034 8.534228 9.8E-07 33645.81 3.000635 6.54683E-07 661601.3 3.000086 0.000 3.00 0.0003 3.000086 10.43704 1.08E-06 19996.96 3.00112 8.53273E-07 397176.1 3.000178 0.000 3.00 0.0003 3.000178 12.28749 1.17E-06 13133.96 3.001772 1.07085E-06 261411.8 3.000321 0.000 3.00 0.0004 3.000321 14.05013 1.26E-06 9290.458 3.002591 1.30387E-06 184235.2 3.000528 0.000 3.00 0.0005 3.000528 15.70506 1.34E-06 6956.809 3.003575 1.54888E-06 136888.4 3.000812 0.001 3.00 0.0006 3.000812 17.24368 1.41E-06 5447.503 3.004714 1.80269E-06 106057.8 3.001181 0.001 3.00 0.0006 3.001181 18.66506 1.48E-06 4420.549 3.006002 2.06242E-06 84995.27 3.001647 0.001 3.00 0.0007 3.001647 19.97307 1.55E-06 3692.144 3.007428 2.3256E-06 70026.59 3.002215 0.001 3.00 0.0008 3.002215 21.17437 1.61E-06 3157.32 3.008984 2.59015E-06 59031.63 3.002893 0.001 3.00 0.0009 3.002893 22.27699 1.67E-06 2753.015 3.010664 2.85439E-06 50726.18 3.003685 0.001 3.00 0.0010 3.003685 23.28947 1.72E-06 2439.686 3.012462 3.11696E-06 44299.8 3.004595 0.001 3.00 CENTRE 217  APPENDIX G: RIP MINIPLANT OPERATING PROCEDURES –CASCADE EMERGENCY SHUT-DOWN INSTRUCTIONS: ***See next page for photographs of all key shut-offs*** 1. SHUT OFF AIR COMPRESSOR –off switch at unit & unplug at South Wall.  Open tank drain valve to bleed air.  2. SHUT OFF TEMPERATURE CONTROLLERS –unplug YELLOW extension cord on North Wall  3. SHUT OFF MOTORS –unplug orange extension cord on North Wall  4. SHUT OFF PUMPS –use power bar switch on wall just to the left of sink  5. SHUT OFF pH CONTROLLERS –use switch or unplug power bar on East Wall  6. SHUT OFF SCALES & COMPUTER –use switch or unplug power bar on East Wall     5,6  4  1 EAST NORTH SOUTH  1  2  3 218  SOUTH WALL FOR OUTLET: 1) AIR COMPRESSOR:   On/off switch    Bleed Valve  Outlet: Panel#1 Circuit 2 NORTH WALL: 2) TEMPERATURE CONTROLLERS:  3)  MOTORS:            Outlet: Panel 1 Circuit 9 EAST WALL: 4) PUMPS:     5,6)   pH CONTROLLERS and SCALES: O       Fed from Panel 1 Circuit 13    Outlet: Panel 1 Circuit 11       219  SYSTEM START UP: 1. Calibrate the pH probes on each controller using the Temperature Probe.  When calibration is complete, switch for a 0.1% 111.7 ohm resistor (except for the resin feed tank which is calibrated with and keeps any other resistor) 2. Mount contactors to frame and connect tubing between them.  Do not connect feed to first tk. 3. Place heaters with cages, airlifts, pH probes, and temperature sensors into each contactor 4. Measure out and add desired wet-settled resin volumes for each contactor 5. Add water and lower agitators into each contactor.  Start mixing at 300 RPM. 6. Submerge NaOH addition lines into each contactor (must be below solution level!) 7. Turn on and tare all scales with NaOH containers on them 8. Start feed pump with water.  Adjust speed to get desired flowrate for test. 9. Turn off feed pump and connect to feed mix tank 10. Start overflow recycle pump with water.  Adjust speed to desired flowrate. 11. Turn off recycle pump and place into feed mix tank 12. Start flow of water through circuit.  Check for any leaks 13. Check that temperature controllers are in contactors and are not touching contactor walls 14. Plug in yellow extension cord on North wall to provide power to temperature controllers 15. Plug heaters into temperature controllers and turn on individual heater power bars 16. Check that temperature control is set to 30C and then initiate heaters by pressing the hand on the temperature controller. 17. Turn on the water feed pump to the resin feed tank (labeled “carousel recycle”) 18. Check that all airlifts are recycling back to the tank that they sample from 19. Start air compressor and connect to air manifold.  Open manifold valves if necessary. 20. Measure rate of RESIN transfer for each contactor.  Adjust valves to attain a value of 1.5 mL resin per minute through each airlift. 21. Once resin transfer is balanced, adjust airlifts so feeding to previous (higher) contactor. Start 15 minute timer and begin adding resin to Resin Feed Tank every 15 minutes. Start by adding half a resin volume at time zero. 22. Check resin flows via airlifts again, and make adjustments as necessary. 23. Re-tare all scales, turn on computer and start datalogging using Windmill. 24. When ready to start the test, place feed line into nickel feed solution.  RECORD TIME OF STARTING!!!!  Switch Loaded Resin Screen. 25. Monitor test –see circuit performance monitoring checklist                         220  CASCADE CIRCUIT MONITORING: 1. DATALOGGING: x Is windmill recording the data? x Are the scale weights dropping as the NaOH pumps turn on?  2. RESIN FEED TANK: x Is water flowing into and out of the Resin Feed Tank? x Is there still water in the fresh water bucket? x Does the water overflow bucket need changing?  If so, pH adjust overflow water to 6-10 and dispose of down sink.  NOTE: DO NOT REUSE THIS WATER TO FEED THE RESIN FEED TANK AS IT WILL CONTAIN SODIUM FROM THE CAUSTIC ADDED –as sodium levels increase, the sodium will react with the ion exchange resin! x Is the airlift transferring resin to Tank #5 as per usual?  3. CONTACTORS: x Check that there is good mixing of resin within contactor x Check that airlift is moving resin appropriately x Check that the overflow screen is not plugging x Check no resin is breaking through overflow screen (fines okay) x Check temperature in tank is 28-32C x Check that heater is not too close to tank walls and is not smoking x Check that pH is between 3.8-4.2.  (Ideally between 3.9-4.1).  Adjust speed of motor on NaOH pumps as needed to avoid overshooting / undershooting  4. SOLUTION FLOW: x Is there still plenty of fresh feed? x Does the tails bucket need changing?  5. OTHER: x Is the Resin Overflow Screen plugging with resin? x How is the oil level in the air compressor? x Are the pH controllers reading okay? Is the resistor still firmly in place?   221  SYSTEM SHUT-DOWN AT END OF RUN: 1. Take any required samples from contactors before starting shut-down! 2. Turn off heaters using switch at individual power bars.  Unplug yellow extension cord at north wall.  Check lights off at all temperature controllers. 3. Turn off sodium hydroxide addition to all 7 tanks (feed tank, 5 RIP contactors, resin prep tank), by switching off the appropriate Masterflex drive units. 4. Turn off air compressor.  Use switch at unit, unplug from extension cord, bleed air. 5. Turn off feed pumps at appropriate Masterflex drive units: fresh feed, recirculating solution to feed tank, water to resin prep tank 6. Stop windmill logger software on logging computer 7. Turn off individual agitators for each contactor  REMOVAL OF RIP CONTACTOR: x Put on chemical goggles in addition to safety glasses. x Remove feedline into contactor, if necessary (use container to catch spills!) x Unhook airline to contactor’s airlift x Raise & Rinse Agitator x Remove from contactor and rinse the following: o Resin transfer tube from airlift feeding this contactor o Immersion heater & cage o pH probe & holder o airlift & holder o NaOH feedline x Remove outlet hose to next contactor (use container to catch spills!) x Unscrew the three screws holding contactor onto frame x Drain tank.  Use outlet hose for first portion, then tip contactor sideways towards you. Use deionized water to rinse out and collect the resin. x Separate nickel solution and resin.  Nickel solution to go to waste drum, resin to be rinsed with deionized water.  Measure resin volume, keep and label sample. 222   CASCADE OPERATING SHEET TEST START TIME: Add 22.5 mL resin at: (15 mins) Change screen at: (30 mins) Sampling at: target ~45 Tank 1 Tank 2 Tank 3 Tank 4 Tank 5 End of hr Resin Vol Tails Vol 0.5 36 Final Resin Volumes 1 24 Tank 1 1.5 48.5 2 105 Tank 2 2.5 80.5 3 72 Tank 3 3.5 65.5 4 53 Tank 4 4.5 59 5 52 Tank 5 5.5 30 6 56 6.5 52 7 40 7.5 52 8 42 8.5 47 9 51 9.5 48 10 48.5 10.5 52 11 38 11.5 48 12 48.5 12.5 47 13 51 13.5 48 14 48.5 14.5 48 15 48.5 15.5 48 16 48.5 NaOH Weights 223  APPENDIX H: ESTIMATE OF AIRLIFT TOTAL FLOWRATES During the first few cascade runs, the total airlift volumetric flowrate was not recorded. However, the total resin volume in each tank was measured, as well as the rate of resin transfer. Thus, an empirical calculation was needed to estimate the total volumetric flowrate from these two measured values.  The proposed empirical correlation was:  Total Airlift Flowrate = Resin FlowrateVolume Fraction Resin in Tank x Empirical Factor  The  proposed  empirical  correlation  was  fit  to  all  the  measured  total  airlift  flowrates,  and  the empirical factor adjusted to a value of 1.13, in order to get an appropriate fit to the data.  The predicted vs. actual values are plotted in the graph below, and the diagonal line corresponds to a 1:1 correlation between predicted and actual.   0 10 20 30 40 50 60 0 10 20 30 40 50 60 M ea su re d To ta l A ir lif t V ol um e [m L/ m in ] Predicted Total Airlift Volume [mL/min] 224  APPENDIX I: RIP MINIPLANT CASCADE CIRCUIT MODEL OUTPUT  This appendix contains the Excel print outs for the cascade circuit models presented in Chapter 10.  The first page contains all the inputs (outlined cells) and key outputs.  The second page contains the flows and mass balances. On the first page, the solution and resin calculations printed out in the vicinity of the circuit are the outputs from the segregated flow model.  The actual integrals were evaluated using the trapezoidal rule. To provide a check for the segregated flow model, the equivalent loading that would be experienced if all the resin in the tank moved in a plug flow fashion through each contactor with a residence time equal to tmean was calculated.  These calculations are done using both the film diffusion model and the hybrid correlation, and the outputs are lower down on the spreadsheet, under the headings “film diffusion” and “hybrid correlation”.  Since these calculations are sufficiently close to the values obtained with the segregated flow model, the calculations for the segregated flow model are believed to be programmed correctly.      225  Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.0000978 alpha 0.36 Tot Vol 563 mL per tank Cap 2.36 Dapp 4.43E-12 m2/s Flow Rate 15 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Ni 58.71 g/mol 0 nickel in resin feed solution 0.000736 m Initial Load 0 g Ni / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 228 40.5 58 50 18.5 61 120 % Resin 7% 10% 9% 3% 11% 21% Airlift total flow to upstream tank (estimated): 31.3 21.8 25.3 68.5 20.8 10.6 Segregated Flow Model, OUTPUTS: model mix tank exit: 3.547 observed mix tank assay: 3.547 0.000 Soln (gpL) 4.800 3.547 2.947 1.978 1.115 0.841 0.108 Mix Tk 1 2 3 4 5 Resin [g Ni / L R] 46.3 40.9 31.8 21.7 16.1 0.0 Mod He # 0.02 0.03 0.06 0.10 0.47 Eqbm Load [g Ni / L R] 67.85 67.54 66.98 66.63 62.14 quadratic "a" 1.96E-05 1.32E-05 7.43E-06 5.6E-06 7.163E-07 quadratic "b" -4.6E-05 -3.1E-05 -1.75E-05 -1.32E-05 -1.7E-06 quadratic "c" 2.73E-05 1.84E-05 1.03E-05 7.8E-06 9.974E-07 Film Diffusion kf 0.009466 0.006354 0.00358 0.0027 0.0003453 eqv s already loaded 97            100          109           102           - total eqv time in this tank (s) 1,717       2,420       2,109        842           2,440 resin loading moving forward at tmean 67.85 67.54 66.94 59.78 35.38 Hybrid Correlation kh 3.08E-05 2.67E-05 2.17E-05 1.96E-05 9.36E-06 eqv s already loaded 3,657       2,342       1,270        766           - total eqv time in this tank (s) 5,277       4,662       3,270        1,506        2,440 resin loading moving forward at tmean 46.93 42.30 33.31 22.25 18.36 RUN A: Low Feed Assay  226  FLOWS: mL/min Ni mass flow: gNi/min Assays STREAM total flow soln flow resin flow total soln resin g /L Ni g Ni / L R feed 15.00 15.00 0.00 72.00 72.00 4.80 mix o/f 46.28 46.28 0.00 164.16 164.16 3.55 1 o/f 35.34 35.34 0.00 104.14 104.14 2.95 2 o/f 38.84 38.84 0.00 76.81 76.81 1.98 3 o/f 81.97 81.97 0.00 91.36 91.36 1.11 4 o/f 34.27 34.27 0.00 28.80 28.80 0.84 5 o/f (tails) 24.06 24.06 0.00 2.59 2.59 0.11 soln rec'c 31.28 31.28 0.00 92.16 92.16 2.95 resin exit 1.50 0.00 1.50 69.41 69.41 46.27 airlift to 1 21.84 20.34 1.50 101.55 40.23 61.32 1.98 40.88 airlift to 2 25.34 23.84 1.50 74.23 26.56 47.66 1.11 31.77 airlift to 3 68.47 66.97 1.50 88.78 56.30 32.48 0.84 21.65 airlift to 4 20.77 19.27 1.50 26.22 2.07 24.15 0.11 16.10 airlift to 5 10.56 9.06 1.50 0.00 0.00 0.00 0.00 0.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 46.28 46.28 164.16 164.16 1 68.12 68.12 265.71 265.71 2 60.68 60.68 178.36 178.36 3 107.31 107.31 165.59 165.59 4 102.74 102.74 117.58 117.58 5 44.82 44.82 28.80 28.80 OVERALL 25.56 25.56 72.00 72.00 CIRCUIT RECOVERY 96.41%  227  Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.0000978 alpha 0.36 Tot Vol 563 mL per tank Cap 2.36 Dapp 4.43E-12 m2/s Flow Rate 15 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Ni 58.71 g/mol 0 nickel in resin feed solution 0.000736 m Initial Load 0 g Ni / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 228 40.5 58 50 18.5 61 120 % Resin 7% 10% 9% 3% 11% 21% Airlift total flow to upstream tank (estimated): 31.3 21.8 25.3 68.5 20.8 10.6 Segregated Flow Model, OUTPUTS: model mix tank exit: 3.738 observed mix tank assay: 3.738 0.000 Soln (gpL) 4.986 3.738 3.139 2.163 1.276 0.984 0.166 Mix Tk 1 2 3 4 5 Resin [g Ni / L R] 47.2 42.0 33.1 23.1 17.7 0.0 Mod He # 0.02 0.03 0.05 0.09 0.33 Eqbm Load [g Ni / L R] 67.90 67.62 67.12 66.83 63.47 quadratic "a" 2.09E-05 1.44E-05 8.5E-06 6.56E-06 1.106E-06 quadratic "b" -4.9E-05 -3.4E-05 -2.01E-05 -1.55E-05 -2.62E-06 quadratic "c" 2.91E-05 2.01E-05 1.18E-05 9.13E-06 1.54E-06 Film Diffusion kf 0.010084 0.00695 0.004098 0.003161 0.0005333 eqv s already loaded 95            97            103           97             - total eqv time in this tank (s) 1,715       2,417       2,103        837           2,440 resin loading moving forward at tmean 67.90 67.62 67.11 62.09 46.20 Hybrid Correlation kh 3.15E-05 2.76E-05 2.28E-05 2.08E-05 1.094E-05 eqv s already loaded 3,815       2,479       1,383        874           - total eqv time in this tank (s) 5,435       4,799       3,383        1,614        2,440 resin loading moving forward at tmean 47.83 43.35 34.59 23.67 20.20 RUN A: High Feed Assay  228  FLOWS: mL/min Ni mass flow: gNi/min Assays STREAM total flow soln flow resin flow total soln resin g /L Ni g Ni / L R feed 15.00 15.00 0.00 74.78 74.78 4.99 mix o/f 46.28 46.28 0.00 172.97 172.97 3.74 1 o/f 35.34 35.34 0.00 110.94 110.94 3.14 2 o/f 38.84 38.84 0.00 84.01 84.01 2.16 3 o/f 81.97 81.97 0.00 104.56 104.56 1.28 4 o/f 34.27 34.27 0.00 33.72 33.72 0.98 5 o/f (tails) 24.06 24.06 0.00 3.99 3.99 0.17 soln rec'c 31.28 31.28 0.00 98.19 98.19 3.14 resin exit 1.50 0.00 1.50 70.79 70.79 47.19 airlift to 1 21.84 20.34 1.50 106.95 44.00 62.94 2.16 41.96 airlift to 2 25.34 23.84 1.50 80.02 30.40 49.62 1.28 33.08 airlift to 3 68.47 66.97 1.50 100.56 65.90 34.66 0.98 23.11 airlift to 4 20.77 19.27 1.50 29.73 3.20 26.53 0.17 17.68 airlift to 5 10.56 9.06 1.50 0.00 0.00 0.00 0.00 0.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 46.28 46.28 172.97 172.97 1 68.12 68.12 279.91 279.91 2 60.68 60.68 190.96 190.96 3 107.31 107.31 184.58 184.58 4 102.74 102.74 134.28 134.28 5 44.82 44.82 33.72 33.72 OVERALL 25.56 25.56 74.78 74.78 CIRCUIT RECOVERY 94.66%  229 Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.0000978 alpha 0.36 Tot Vol 563 mL per tank Cap 2.36 Dapp 4.43E-12 m2/s Flow Rate 15 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Ni 58.71 g/mol 0 nickel in resin feed solution 0.000736 m Initial Load 3 g Ni / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 322 63 63 75 64 57 120 % Resin 11% 11% 13% 11% 10% 21% Airlift total flow to upstream tank (estimated): 20.1 20.1 16.9 19.8 22.2 10.6 Segregated Flow Model, OUTPUTS: model mix tank exit: 3.208 observed mix tank assay: 3.208 0.000 Soln (gpL) 4.308 3.208 2.388 1.479 0.498 0.038 0.003 Mix Tk 1 2 3 4 5 Resin [g Ni / L R] 46.0 38.1 27.8 13.5 3.8 3.0 Mod He # 0.03 0.04 0.11 1.00 10.45 Eqbm Load [g Ni / L R] 67.70 67.28 65.86 57.73 37.53 quadratic "a" 1.59E-05 9.86E-06 3.32E-06 2.54E-07 2.186E-08 quadratic "b" -3.8E-05 -2.3E-05 -7.84E-06 -6.09E-07 -6.158E-08 quadratic "c" 2.22E-05 1.37E-05 4.62E-06 3.54E-07 3.043E-08 Film Diffusion kf 0.007672 0.004753 0.001599 0.000122 1.054E-05 eqv s already loaded 108          112          144           558           7,907 total eqv time in this tank (s) 2,628       2,632       3,144        3,118        10,187 resin loading moving forward at tmean 67.70 67.28 65.43 18.32 3.82 Hybrid Correlation kh 2.86E-05 2.41E-05 1.63E-05 6.44E-06 2.665E-06 eqv s already loaded 3,332       1,948       663           169           601 total eqv time in this tank (s) 5,852       4,468       3,663        2,729        2,881 resin loading moving forward at tmean 47.28 39.77 30.32 15.05 6.53 RUN B: Low Feed Assay  230  FLOWS: mL/min Ni mass flow: gNi/min Assays STREAM total flow soln flow resin flow total soln resin g /L Ni g Ni / L R feed 15.00 15.00 0.00 64.61 64.61 4.31 mix o/f 35.11 35.11 0.00 112.63 112.63 3.21 1 o/f 33.61 33.61 0.00 80.26 80.26 2.39 2 o/f 30.39 30.39 0.00 44.96 44.96 1.48 3 o/f 33.29 33.29 0.00 16.57 16.57 0.50 4 o/f 35.72 35.72 0.00 1.36 1.36 0.04 5 o/f (tails) 24.06 24.06 0.00 0.08 0.08 0.00 soln rec'c 20.11 20.11 0.00 48.02 48.02 2.39 resin exit 1.50 0.00 1.50 69.04 69.04 46.02 airlift to 1 20.11 18.61 1.50 84.68 27.53 57.15 1.48 38.10 airlift to 2 16.89 15.39 1.50 49.38 7.66 41.72 0.50 27.81 airlift to 3 19.79 18.29 1.50 20.99 0.70 20.29 0.04 13.53 airlift to 4 22.22 20.72 1.50 5.78 0.07 5.71 0.00 3.81 airlift to 5 10.56 9.06 1.50 4.50 0.00 4.50 0.00 3.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 35.11 35.11 112.63 112.63 1 55.21 55.21 197.31 197.31 2 50.50 50.50 129.64 129.64 3 50.18 50.18 65.95 65.95 4 55.52 55.52 22.35 22.35 5 46.28 46.28 5.86 5.86 OVERALL 25.56 25.56 69.11 69.11 CIRCUIT RECOVERY 99.88%  231  Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.0000978 alpha 0.36 Tot Vol 563 mL per tank Cap 2.36 Dapp 4.43E-12 m2/s Flow Rate 15 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Ni 58.71 g/mol 0 nickel in resin feed solution 0.000736 m Initial Load 3 g Ni / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 322 63 63 75 64 57 120 % Resin 11% 11% 13% 11% 10% 21% Airlift total flow to upstream tank (estimated): 20.1 20.1 16.9 19.8 22.2 10.6 Segregated Flow Model, OUTPUTS: model mix tank exit: 3.525 observed mix tank assay: 3.525 0.000 Soln (gpL) 4.573 3.525 2.744 1.844 0.790 0.140 0.010 Mix Tk 1 2 3 4 5 Resin [g Ni / L R] 48.6 41.4 32.1 18.7 6.0 3.0 Mod He # 0.02 0.04 0.08 0.35 4.26 Eqbm Load [g Ni / L R] 67.80 67.48 66.55 62.98 48.41 quadratic "a" 1.83E-05 1.23E-05 5.26E-06 9.33E-07 6.53E-08 quadratic "b" -4.3E-05 -2.9E-05 -1.24E-05 -2.21E-06 -1.641E-07 quadratic "c" 2.55E-05 1.71E-05 7.33E-06 1.3E-06 9.092E-08 Film Diffusion kf 0.008814 0.005924 0.002538 0.00045 3.148E-05 eqv s already loaded 107          109          130           224           2,032 total eqv time in this tank (s) 2,627       2,629       3,130        2,784        4,312 resin loading moving forward at tmean 67.80 67.48 66.53 44.97 6.15 Hybrid Correlation kh 3E-05 2.6E-05 1.92E-05 1.03E-05 3.952E-06 eqv s already loaded 3,892       2,463       1,068        225           243 total eqv time in this tank (s) 6,412       4,983       4,068        2,785        2,523 resin loading moving forward at tmean 49.70 42.94 34.47 20.73 9.57 RUN B: High Feed Assay  232  FLOWS: mL/min Ni mass flow: gNi/min Assays STREAM total flow soln flow resin flow total soln resin g /L Ni g Ni / L R feed 15.00 15.00 0.00 68.59 68.59 4.57 mix o/f 35.11 35.11 0.00 123.76 123.76 3.53 1 o/f 33.61 33.61 0.00 92.21 92.21 2.74 2 o/f 30.39 30.39 0.00 56.05 56.05 1.84 3 o/f 33.29 33.29 0.00 26.30 26.30 0.79 4 o/f 35.72 35.72 0.00 5.00 5.00 0.14 5 o/f (tails) 24.06 24.06 0.00 0.24 0.24 0.01 soln rec'c 20.11 20.11 0.00 55.17 55.17 2.74 resin exit 1.50 0.00 1.50 72.86 72.86 48.57 airlift to 1 20.11 18.61 1.50 96.48 34.32 62.16 1.84 41.44 airlift to 2 16.89 15.39 1.50 60.31 12.16 48.15 0.79 32.10 airlift to 3 19.79 18.29 1.50 30.57 2.56 28.01 0.14 18.67 airlift to 4 22.22 20.72 1.50 9.26 0.20 9.06 0.01 6.04 airlift to 5 10.56 9.06 1.50 4.50 0.00 4.50 0.00 3.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 35.11 35.11 123.76 123.76 1 55.21 55.21 220.24 220.24 2 50.50 50.50 152.52 152.52 3 50.18 50.18 86.61 86.61 4 55.52 55.52 35.57 35.57 5 46.28 46.28 9.50 9.50 OVERALL 25.56 25.56 73.09 73.09 CIRCUIT RECOVERY 99.66%  233  Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.0000978 alpha 0.36 Tot Vol 563 mL per tank Cap 2.36 Dapp 4.43E-12 m2/s Flow Rate 15 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Ni 58.71 g/mol 0 nickel in resin feed solution 0.000736 m Initial Load 3 g Ni / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 276.08 92.16 82.56 48.96 38.4 14 120 % Resin 16% 15% 9% 7% 2% 21% Airlift total flow to upstream tank (measured): 12.0 12.0 20.0 36.0 52.0 25.0 Segregated Flow Model, OUTPUTS: model mix tank exit: 3.757 observed mix tank assay: 3.757 0.000 Soln (gpL) 4.641 3.757 2.650 1.293 0.476 0.123 0.035 Mix Tk 1 2 3 4 5 Resin [g Ni / L R] 48.5 38.1 25.1 15.0 6.3 3.0 Mod He # 0.02 0.05 0.13 0.46 2.13 Eqbm Load [g Ni / L R] 67.78 67.14 65.79 62.57 57.33 quadratic "a" 1.77E-05 8.61E-06 3.17E-06 8.17E-07 2.359E-07 quadratic "b" -4.2E-05 -2E-05 -7.5E-06 -1.94E-06 -5.667E-07 quadratic "c" 2.46E-05 1.2E-05 4.42E-06 1.14E-06 3.284E-07 Film Diffusion kf 0.008515 0.004152 0.00153 0.000394 0.0001137 eqv s already loaded 97            113          169           267           473 total eqv time in this tank (s) 3,783       3,415       2,127        1,803        1,033 resin loading moving forward at tmean 67.78 67.14 63.25 31.82 6.35 Hybrid Correlation kh 2.97E-05 2.29E-05 1.6E-05 9.81E-06 6.275E-06 eqv s already loaded 3,200       1,641       833           256           109 total eqv time in this tank (s) 6,887       4,943       2,791        1,792        669 resin loading moving forward at tmean 50.65 40.52 26.61 16.31 7.40 RUN C: Low Feed Assay  234  FLOWS: mL/min Ni mass flow: gNi/min Assays STREAM total flow soln flow resin flow total soln resin g /L Ni g Ni / L R feed 15.00 15.00 0.00 69.62 69.62 4.64 mix o/f 27.00 27.00 0.00 101.43 101.43 3.76 1 o/f 25.50 25.50 0.00 67.59 67.59 2.65 2 o/f 33.50 33.50 0.00 43.30 43.30 1.29 3 o/f 49.50 49.50 0.00 23.57 23.57 0.48 4 o/f 65.50 65.50 0.00 8.03 8.03 0.12 5 o/f (tails) 38.50 38.50 0.00 1.36 1.36 0.04 soln rec'c 12.00 12.00 0.00 31.81 31.81 2.65 resin exit 1.50 0.00 1.50 72.76 72.76 48.51 airlift to 1 12.00 10.50 1.50 70.72 13.57 57.15 1.29 38.10 airlift to 2 20.00 18.50 1.50 46.44 8.81 37.63 0.48 25.09 airlift to 3 36.00 34.50 1.50 26.71 4.23 22.48 0.12 14.99 airlift to 4 52.00 50.50 1.50 11.17 1.79 9.38 0.04 6.25 airlift to 5 25.00 23.50 1.50 4.50 0.00 4.50 0.00 3.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 27.00 27.00 101.43 101.43 1 39.00 39.00 172.15 172.15 2 45.50 45.50 114.03 114.03 3 69.50 69.50 70.01 70.01 4 101.50 101.50 34.74 34.74 5 90.50 90.50 12.53 12.53 OVERALL 40.00 40.00 74.12 74.12 CIRCUIT RECOVERY 98.04%  235  Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.0000978 alpha 0.36 Tot Vol 563 mL per tank Cap 2.36 Dapp 4.43E-12 m2/s Flow Rate 15 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Ni 58.71 g/mol 0 nickel in resin feed solution 0.000736 m Initial Load 3 g Ni / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 276.08 92.16 82.56 48.96 38.4 14 120 % Resin 16% 15% 9% 7% 2% 21% Airlift total flow to upstream tank (measured): 12.0 12.0 20.0 36.0 52.0 25.0 Segregated Flow Model, OUTPUTS: model mix tank exit: 4.300 observed mix tank assay: 4.300 0.000 Soln (gpL) 5.163 4.300 3.220 1.797 0.832 0.333 0.143 Mix Tk 1 2 3 4 5 Resin [g Ni / L R] 51.0 41.5 29.2 19.1 9.1 3.0 Mod He # 0.02 0.04 0.08 0.20 0.67 Eqbm Load [g Ni / L R] 67.91 67.46 66.62 65.13 63.04 quadratic "a" 2.15E-05 1.2E-05 5.55E-06 2.22E-06 9.502E-07 quadratic "b" -5.1E-05 -2.8E-05 -1.31E-05 -5.25E-06 -2.252E-06 quadratic "c" 2.99E-05 1.67E-05 7.72E-06 3.09E-06 1.323E-06 Film Diffusion kf 0.010345 0.005774 0.002674 0.001071 0.0004581 eqv s already loaded 91            98            127           140           106 total eqv time in this tank (s) 3,778       3,401       2,085        1,676        666 resin loading moving forward at tmean 67.91 67.46 66.37 54.32 16.58 Hybrid Correlation kh 3.18E-05 2.58E-05 1.96E-05 1.41E-05 1.036E-05 eqv s already loaded 3,673       2,013       1,101        350           55 total eqv time in this tank (s) 7,359       5,316       3,059        1,886        615 resin loading moving forward at tmean 52.96 43.83 30.73 20.67 10.00 RUN C: High Feed Assay  236  FLOWS: mL/min Ni mass flow: gNi/min Assays STREAM total flow soln flow resin flow total soln resin g /L Ni g Ni / L R feed 15.00 15.00 0.00 77.45 77.45 5.16 mix o/f 27.00 27.00 0.00 116.10 116.10 4.30 1 o/f 25.50 25.50 0.00 82.12 82.12 3.22 2 o/f 33.50 33.50 0.00 60.21 60.21 1.80 3 o/f 49.50 49.50 0.00 41.20 41.20 0.83 4 o/f 65.50 65.50 0.00 21.84 21.84 0.33 5 o/f (tails) 38.50 38.50 0.00 5.49 5.49 0.14 soln rec'c 12.00 12.00 0.00 38.64 38.64 3.22 resin exit 1.50 0.00 1.50 76.46 76.46 50.97 airlift to 1 12.00 10.50 1.50 81.13 18.87 62.26 1.80 41.50 airlift to 2 20.00 18.50 1.50 59.22 15.40 43.83 0.83 29.22 airlift to 3 36.00 34.50 1.50 40.21 11.50 28.70 0.33 19.13 airlift to 4 52.00 50.50 1.50 20.85 7.20 13.65 0.14 9.10 airlift to 5 25.00 23.50 1.50 4.50 0.00 4.50 0.00 3.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 27.00 27.00 116.10 116.10 1 39.00 39.00 197.23 197.23 2 45.50 45.50 141.34 141.34 3 69.50 69.50 100.42 100.42 4 101.50 101.50 62.05 62.05 5 90.50 90.50 26.34 26.34 OVERALL 40.00 40.00 81.95 81.95 CIRCUIT RECOVERY 92.91%  237  Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.000745 alpha 0.62 Tot Vol 563 mL per tank Cap 2.5 Dapp 2.8E-11 m2/s Flow Rate 18.5 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Cu 63.55 g/mol 0 Cu in resin feed solution 0.000736 m Initial Load 0 g Cu / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 272.5 43 73.5 39 63 54 120 % Resin 8% 13% 7% 11% 10% 21% Airlift total flow to upstream tank (measured): 28.5 23.0 14.0 21.0 21.0 40.0 Segregated Flow Model, OUTPUTS: model mix tank exit: 3.725 observed mix tank assay: 3.725 0.000 Soln (gpL) 5.153 3.725 2.798 1.508 0.387 0.020 0.001 Mix Tk 1 2 3 4 5 Resin [g Cu / L R] 63.5 53.0 27.9 9.5 0.4 0.0 Mod He # 0.07 0.10 0.30 2.33 100.80 Eqbm Load [g Cu / L R] 78.82 78.60 77.79 72.49 55.69 quadratic "a" 0.000131 7.07E-05 1.82E-05 9.54E-07 6.274E-08 quadratic "b" -0.00033 -0.00018 -4.54E-05 -2.4E-06 -1.669E-07 quadratic "c" 0.000205 0.00011 2.84E-05 1.49E-06 9.803E-08 Film Diffusion kf 0.00784 0.004224 0.001085 5.7E-05 3.749E-06 eqv s already loaded 142          104          120           109           - total eqv time in this tank (s) 1,862       3,044       1,680        2,629        2,160 resin loading moving forward at tmean 78.82 78.60 65.22 10.09 0.45 Hybrid Correlation kh 8.65E-05 5.89E-05 2.54E-05 4.09E-06 7.556E-07 eqv s already loaded 1,736       570          148           2               - total eqv time in this tank (s) 3,456       3,510       1,708        2,522        2,160 resin loading moving forward at tmean 65.83 58.97 31.03 14.57 4.49 Copper: Low Feed Assay  238  FLOWS: mL/min Cu mass flow: gCu/min Assays STREAM total flow soln flow resin flow total soln resin g /L Cu g Cu / L R feed 18.50 18.50 0.00 95.32 95.32 5.15 mix o/f 47.00 47.00 0.00 175.08 175.08 3.73 1 o/f 40.00 40.00 0.00 111.93 111.93 2.80 2 o/f 31.00 31.00 0.00 46.74 46.74 1.51 3 o/f 38.00 38.00 0.00 14.71 14.71 0.39 4 o/f 38.00 38.00 0.00 0.77 0.77 0.02 5 o/f (tails) 57.00 57.00 0.00 0.08 0.08 0.00 soln rec'c 28.50 28.50 0.00 79.75 79.75 2.80 resin exit 1.50 0.00 1.50 95.25 95.25 63.50 airlift to 1 23.00 21.50 1.50 111.86 32.41 79.44 1.51 52.96 airlift to 2 14.00 12.50 1.50 46.66 4.84 41.82 0.39 27.88 airlift to 3 21.00 19.50 1.50 14.64 0.40 14.24 0.02 9.49 airlift to 4 21.00 19.50 1.50 0.70 0.03 0.67 0.00 0.45 airlift to 5 40.00 38.50 1.50 0.00 0.00 0.00 0.00 0.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 47.00 47.00 175.08 175.08 1 70.00 70.00 286.93 286.93 2 54.00 54.00 158.60 158.60 3 52.00 52.00 61.38 61.38 4 59.00 59.00 15.41 15.41 5 78.00 78.00 0.77 0.77 OVERALL 58.50 58.50 95.32 95.32 CIRCUIT RECOVERY 99.92%  239  Solution Side Resin Side TP207XL Kinetics Df/delta 0.0000273 m/s K 0.000745 alpha 0.62 Tot Vol 563 mL per tank Cap 2.5 Dapp 2.8E-11 m2/s Flow Rate 18.5 mL / min FRESH FEED Resin Rate 1.5 mL/min bead dia 736 um molar Cu 63.55 g/mol 0 Cu in resin feed solution 0.000736 m Initial Load 0 g Cu / L Resin TOTAL IN CIRCUIT Resin Fd Tk Resin Volume [mL] 272.5 43 73.5 39 63 54 120 % Resin 8% 13% 7% 11% 10% 21% Airlift total flow to upstream tank (measured): 28.5 23.0 14.0 21.0 21.0 40.0 Segregated Flow Model, OUTPUTS: model mix tank exit: 3.910 observed mix tank assay: 3.910 0.000 Soln (gpL) 5.285 3.910 3.017 1.743 0.533 0.029 0.002 Mix Tk 1 2 3 4 5 Resin [g Cu / L R] 65.1 55.4 31.5 13.1 0.6 0.0 Mod He # 0.07 0.09 0.24 1.53 65.62 Eqbm Load [g Cu / L R] 78.84 78.66 78.03 73.58 58.61 quadratic "a" 0.000141 8.18E-05 2.5E-05 1.36E-06 8.591E-08 quadratic "b" -0.00035 -0.0002 -6.25E-05 -3.42E-06 -2.248E-07 quadratic "c" 0.000221 0.000128 3.91E-05 2.13E-06 1.342E-07 Film Diffusion kf 0.008453 0.004884 0.001493 8.14E-05 5.133E-06 eqv s already loaded 143          105          123           108           - total eqv time in this tank (s) 1,863       3,045       1,683        2,628        2,160 resin loading moving forward at tmean 78.84 78.66 71.70 14.17 0.65 Hybrid Correlation kh 9.06E-05 6.45E-05 3.09E-05 5.09E-06 9.182E-07 eqv s already loaded 1,877       678          229           4               - total eqv time in this tank (s) 3,597       3,618       1,789        2,524        2,160 resin loading moving forward at tmean 67.30 61.27 34.78 16.47 5.21 Copper: High Feed Assay  240  FLOWS: mL/min Cu mass flow: gCu/min Assays STREAM total flow soln flow resin flow total soln resin g /L Cu g Cu / L R feed 18.50 18.50 0.00 97.78 97.78 5.29 mix o/f 47.00 47.00 0.00 183.76 183.76 3.91 1 o/f 40.00 40.00 0.00 120.68 120.68 3.02 2 o/f 31.00 31.00 0.00 54.04 54.04 1.74 3 o/f 38.00 38.00 0.00 20.25 20.25 0.53 4 o/f 38.00 38.00 0.00 1.10 1.10 0.03 5 o/f (tails) 57.00 57.00 0.00 0.10 0.10 0.00 soln rec'c 28.50 28.50 0.00 85.99 85.99 3.02 resin exit 1.50 0.00 1.50 97.67 97.67 65.12 airlift to 1 23.00 21.50 1.50 120.58 37.48 83.09 1.74 55.40 airlift to 2 14.00 12.50 1.50 53.94 6.66 47.28 0.53 31.52 airlift to 3 21.00 19.50 1.50 20.15 0.57 19.58 0.03 13.05 airlift to 4 21.00 19.50 1.50 1.00 0.04 0.96 0.00 0.64 airlift to 5 40.00 38.50 1.50 0.00 0.00 0.00 0.00 0.00 Balance by Stage: mass flow mass flow Flow in Flow out in out Mix Tank 47.00 47.00 183.76 183.76 1 70.00 70.00 304.34 304.34 2 54.00 54.00 174.62 174.62 3 52.00 52.00 74.19 74.19 4 59.00 59.00 21.25 21.25 5 78.00 78.00 1.10 1.10 OVERALL 58.50 58.50 97.78 97.78 CIRCUIT RECOVERY 99.89% 241  APPENDIX J: RIP MINIPLANT OPERATING PROCEDURES: CAROUSEL  EMERGENCY SHUT-DOWN INSTRUCTIONS: ***See next page for photographs of all key shut-offs***  1.  SHUT OFF TEMPERATURE CONTROL –unplug yellow extension cord on North Wall 2.  SHUT OFF MOTORS –unplug orange extension cord on North Wall 3.  SHUT OFF PUMPS –use power bar switch on wall just to the left of sink 4.  SHUT OFF pH CONTROLLERS –use switch or unplug power bar on East Wall 5.  SHUT OFF SCALES & COMPUTER –use switch or unplug power bar on East Wall    4,5  3 EAST WALL NORTH  2  1 242   NORTH WALL: 3) TEMPERATURE CONTROLLERS:  2)  MOTORS:            Outlet: Panel 1 Circuit 9 EAST WALL: 3) PUMPS:     4,5)   pH CONTROLLERS and SCALES: O       Fed from Panel 1 Circuit 13          Outlet: Panel 1 Circuit 11     243  SYSTEM START UP: 1. Calibrate the pH probes on each controller using the Temperature Probe.  When calibration is complete, switch for a 0.1% 111.7 ohm resistor (except for the resin feed tank which is calibrated with and keeps any other resistor) 2. Mount contactors to frame and connect tubing between them.  Do not connect feed to first tk. 3. Place heaters with cages, pH probes, and temperature sensors into each contactor 4. Measure out and add desired wet-settled resin volumes for each contactor 5. Add water and lower agitators into each contactor.  Start mixing at 300 RPM. 6. Submerge NaOH addition lines into each contactor (must be below solution level!) 7. Turn on and tare all scales with NaOH containers on them 8. Start feed pump with water.  Adjust speed to get desired flowrate for test. 9. Turn off feed pump and connect to feed mix tank 10. Start flow of water through circuit.  Check for any leaks 11. Check that temperature controllers are in contactors and are not touching contactor walls 12. Plug in yellow extension cord on North wall to provide power to temperature controllers 13. Plug heaters into temperature controllers and turn on individual heater power bars 14. Check that temperature control is set to 30C and then initiate heaters by pressing the hand on the temperature controller. 15. Turn on motors to NaOH pumps so they control on pH 16. Re-tare all scales, turn on computer and start datalogging using Windmill. 17. When ready to start the test, place feed line into nickel feed solution.  RECORD TIME OF STARTING!!!!                   244  CAROUSEL SWITCHING OF CONTACTORS: Lead – High Centre – Centre – Low Centre – Lag   (“Lead – H.C. – C – L.C. –Lag”) x Take samples from all contactors 45 minutes into cycle  Label: Circuit Position – Hour – Physical Position (e.g. H.C. – 2 - F)  x Record NaOH Weights from Logger onto Test Sheet 55 minutes into cycle x Three minutes before hour ends (i.e. 57 minutes into cycle), turn off Feed Pump and Carousel  Recycle  Pumps  at  Power  Bar  by  Sink.   Manually  turn  on/off  Carousel Recycle Pump if necessary.  x Remove Feed Line to Lead Contactor and place into High Centre Contactor. Disconnect lead o/f.  x Place bucket under Lag Contactor and disconnect Tails line.  Attach tail line to offline contactor  x Connect Lag overflow line to offline contactor.  Switch and adjust position of Tails bucket.  x Turn on NaOH pump to offline contactor.  x Add ____ mL resin to each contactor in use.  Turn Pumps back on at Power Bar. Contactors are all now in new circuit positions! ______________________________________________________________________________________________________ x Switch off power bar for heater to new offline contactor (formerly Lead Contactor)  x Switch off NaOH pump to offline contactor.  Direct offline contactor O/F to bucket  x Remove and Rinse: pH probe, heater, temp sensor, agitator, NaOH line at offline contactor  x Remove offline contactor from frame.  Use Caution.  Do not handle contactor by overflow or inlet.  Be careful around screens  x Mount spare contactor with ____ mL resin onto frame.  Have overflow directed to a water rinse bucket.  Lower agitator into contactor.  Place pH probe, temp sensor, NaOH line into contactor.  x Double check position of heater.  Point the sealed edge of the screen towards the contactor wall.  Ensure that there is sufficient space between heater screen and contactor.  Turn powerbar back on  x Clean-Out the offline contactor (formerly Lag Contactor) ***Use Caution.  Do not handle contactor by overflow or inlet.  Be careful around screens*** Rinse out Resin Fines Separately into Fines Container.  x Measure out and Record Final Resin Volume.  Keep Resin Sample.  Record volume of tails. 245  CAROUSEL CIRCUIT MONITORING: 1. DATALOGGING: x Is windmill recording the data? x Are the scale weights dropping as the NaOH pumps turn on? 2. CONTACTORS: x Check that there is good mixing of resin within contactor x Check that the overflow screen is not plugging x Check no resin is breaking through overflow screen (fines okay) x Check temperature in tank is 28-32C x Check that heater is not too close to tank walls and is not smoking x Check that pH is between 3.8-4.2.  (Ideally between 3.9-4.1).  Adjust speed of motor on NaOH pumps as needed.  Check solution levels of NaOH and replenish with 1M solution as needed.  If pH reading stuck ~ pH 7, check resistor. 3. SOLUTION FLOW: x Is there still plenty of fresh feed? x Does the tails bucket need changing?   SYSTEM SHUT-DOWN AT END OF RUN: 1. Take any required samples from contactors before starting shut-down! 2. Turn off heaters using switch at individual power bars.  Unplug yellow extension cord at north wall.  Check lights off at all temperature controllers. 3. Turn off sodium hydroxide addition to all 7 tanks (feed tank, 5 RIP contactors, resin prep tank), by switching off the appropriate Masterflex drive units. 4. Turn off feed pumps at appropriate Masterflex drive units: fresh feed, recirculating solution to F 5. Stop windmill logger software on logging computer 6. Turn off individual agitators for each contactor    246  REMOVAL OF RIP CONTACTOR: x Put  on  chemical  goggles  if  using  safety  glasses  for  PPE.   MUST USE GOGGLES FOR THIS STEP!!! x Remove feedline into contactor, if necessary (use container to catch spills!) x Remove from contactor and rinse the following: o Immersion heater & cage & temperature probe o pH probe & holder o NaOH feedline o Raise & Rinse Agitator x Remove outlet hose to next contactor (use container to catch spills!) x Unscrew the three screws holding contactor onto frame x Drain tank.  Use outlet hose for first portion, then tip contactor sideways towards you.  Use deionized water to rinse out and collect the resin. x Separate feed solution and resin.  Feed solution to go to waste drum, resin to be rinsed with deionized water.  Measure resin volume, keep and label sample.  247    RIP MINIPLANT OPERATING SHEET DATE: Test Start Time: Sample at:_______________ Weights at:_______________ Turn pump off at:___________ Pump on:__________ SAMPLE! Hour Time for Change Tails Volume Resin Volume 1 2 3 4 5 F 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 4-5-F-1-2 5-F-1-2-3 4-5-F-1-2 5-F-1-2-3 F-1-2-3-4 1-2-3-4-5 2-3-4-5-F 3-4-5-F-1 4-5-F-1-2 5-F-1-2-3 F-1-2-3-4 1-2-3-4-5 2-3-4-5-F 3-4-5-F-1 4-5-F-1-2 5-F-1-2-3 F-1-2-3-4 1-2-3-4-5 2-3-4-5-F 3-4-5-F-1 4-5-F-1-2 5-F-1-2-3 F-1-2-3-4 1-2-3-4-5 2-3-4-5-F 3-4-5-F-1 NaOH Scale Weights Tank Order Lead-HC-C-LC-Lag F-1-2-3-4 1-2-3-4-5 2-3-4-5-F 3-4-5-F-1 248    Carousel Test Cont… Hour Time for Change Tails Volume Resin Volume 1 2 3 4 5 F 31 32 33 34 35 36 FINAL TANK RESIN VOLUMES: POSITION During 36th hour: F Sample all contactors at: 1 15 mins 30 mins 45 mins 60 mins (start of 37th hour!) 2 3 4 5 2-3-4-5-F 3-4-5-F-1 4-5-F-1-2 5-F-1-2-3 VOLUME NaOH Scale Weights Tank Order Lead-HC-C-LC-Lag F-1-2-3-4 1-2-3-4-5 249  APPENDIX K: COPPER RESIN VOLUME OPTIMIZATION   0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Feed Mix 1 2 3 4 5 Co p pe r i n So lu ti on  [g pL ] Tank 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% No Back-Mixing 250     0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Feed Mix 1 2 3 4 5 Co pp er  in  S ol ut io n [g pL ] Tank 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% with Back-Mixing 0 10 20 30 40 50 60 70 1 2 3 4 5 Resin Feed Co pp er  Lo ad ed  o n Re si n [g  C u / L Re si n] Tank 3.6% 7.1% 10.7% 14.3% 17.9% 21.4% 25.0% with Back-Mixing 251   0 10 20 30 40 50 60 0% 5% 10% 15% 20% 25% 30% To ta l R es in  In ve nt or y [m 3 ] Percent Resin Volume in Each Tank no back-mixing with back-mixing 0 10 20 30 40 50 60 70 0 2 4 6 8 10 12 14 16 R es in  L oa di n g [g  C u / L R es in ] Resin Residence Time in Loading Circuit [hr] no back-mixing with back-mixing

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0072620/manifest

Comment

Related Items